key: cord-022889-lv6fy6e6
authors: Dávalos, Alberto; Henriques, Rossana; Latasa, María Jesús; Laparra, Moisés; Coca, María
title: Literature review of baseline information on non‐coding RNA (ncRNA) to support the risk assessment of ncRNA‐based genetically modified plants for food and feed
date: 2019-08-07
journal: nan
DOI: 10.2903/sp.efsa.2019.en-1688
sha: 
doc_id: 22889
cord_uid: lv6fy6e6

This report is the outcome of an EFSA procurement (NP/EFSA/GMO/2016/01) reviewing relevant scientific information on ncRNA and on RNA interference(RNAi) that could support the food and feed risk assessment of ncRNA‐based genetically modified (GM) plants. Information was retrieved through key words and key questions covering the stability and degradation of ncRNAs after oral ingestion, the passage of ncRNAs from food and feed to human and animal organs and tissues via the gastrointestinal tract and other barriers, as well as the potential effects on the gastrointestinal tract, the immune system or the entire organism.Full description of the strategy used for the literature search and for studies selectionis provided and the number of retrieved publications is reported. This report is divided into four partsdiscussing the kinetics of exogenous ncRNAs in humans and animals, with focus on ingested ncRNAs (Part 1); the possible effects of ncRNAs on the gastrointestinal tract (Part 2), systemically(Part 3)and on the immune system (Part 4). This report suggests that some plant ncRNAs (e.g miRNAs and siRNAs) show higher stability as compared to other ncRNAs due to peculiar chemical characteristics (2’‐O‐methylation at 3’ end).However, ingested or administered ncRNA must overcome many extracellular and cellular barriers to reach the intended target tissue or functional location in sufficient amount to exert any biological effect. Literature data indicate that chemically unmodified and unformulated ncRNAs exhibit very low stability in the gastrointestinal tract and in biological fluids and, in general, do not elicit major biological effects.This report also provides an overview of the RNA content in plant‐derived foods and diets and discusses the controversies on the presence of dietary exogenous RNAs in the biological fluids of humans and animals and their effects. Finally, gaps in the scientific literature are highlighted and recommendations provided

1. Introduction

This scientific report is the result of a contract awarded by the European Food Safety Authority (EFSA) to the Madrid Institute of Advanced Studies (IMDEA)-Food, under the contract title: "Literature review of baseline information on non-coding RNA (ncRNA) that could support the food/feed risk assessment of ncRNA-based GM plants" (contract number NP/EFSA/GMO/2016/01).

For the literature review of the available scientific information on foreign (exogenous) ncRNAs that could support the risk assessment of ncRNA-based GM plants, the following tasks were defined by EFSA.

Based on available scientific literature (narrative review) to review the kinetics profile of foreign (exogenous) ncRNA in humans and animals (experimental animals, livestock and pets). This to be based on information from research and development of ncRNAs intended to be used as therapeutics and/or from research on biomarkers in humans (pharmaceuticals/medicine area) and focused primarily on the oral route of administration of ncRNA, with detailed information on the absorption, distribution, metabolism and excretion of ncRNA molecules. Other routes of administration, however, not to be excluded.

Based on available scientific literature (narrative review), to review the effects (physiological, paraphysiological, and pathological) of foreign (exogenous) ncRNAs on the GI tract and annex glands (e.g. liver, pancreas, salivary glands) in humans and animals (experimental animals, livestock and pets). This to be based on information from research and development of ncRNA molecules intended to be used as therapeutics and/or from research on biomarkers in humans (pharmaceuticals/medicine area).

Based on available scientific literature (narrative review), to provide information on the possibility of systemic effects of foreign (exogenous) ncRNAs in humans and animals (experimental animals, livestock and pets) with focus on gastrointestinal barrier to absorption, and other barriers as relevant (e.g. placenta).

Based on available scientific literature (narrative review), to assess the plausibility of effects on the immune system of humans and animals (experimental animals, livestock and pets) of foreign (exogenous) ncRNAs.

GM plants based on silencing approaches by RNAi (through ncRNA expression) are developed for food and feed purposes and assessed by EFSA within the EU GMO regulatory framework. 1 The team considered that pursuing the four tasks defined by EFSA to support the risk assessment of GM plants in this regulatory context required gathering preparatory baseline information on ncRNA and refinement of the scope of the tasks. Therefore, further elaboration on the terms of reference provided by EFSA has been conducted and it is described below. It should be emphasized that the available information on ncRNAs which could be possibly relevant to food and feed safety, overlaps with and is barely distinguishable from the broader information on RNAs. Information on RNA in general has therefore been included in the search as warranted; when possible, the relevance of distinguishing peculiar ncRNAs features is described. The term "foreign ncRNA" or "exogenous ncRNA" used in this report refers to ncRNA molecules to which humans/animals can be exposed through the diet or via therapeutic treatment. Similarly, "foreign RNA" or "exogenous RNA" refers to RNA molecules to which humans/animals can be exposed through the diet or via therapeutic treatment.

Part 1: Kinetics of exogenous ncRNA in humans and animals (EFSA Task 1) EFSA Task 1: Based on available scientific literature (narrative review) to review the kinetics profile of foreign (exogenous) ncRNA in humans and animals (experimental animals, livestock and pets) . This to be based on information from research and development of ncRNAs intended to be used as therapeutics and/or from research on biomarkers in humans (pharmaceuticals/medicine area) and to be focused primarily on the oral route of administration of ncRNA, with detailed information on the absorption, distribution, metabolism and excretion of ncRNA molecules. Other routes of administration, however, not to be excluded.

To address EFSA Task 1, preparatory baseline information is provided i) on general features of plant ncRNAs including their function, movement in the plant, biogenesis and degradation (section 3.1.1), ii) on their stability (i.e. for how long ncRNA molecules retain their original structure, and how they resist degradation over time and in various conditions, both inside and outside the plant), and turnover, also in comparison with turnover of endogenous ncRNA in animals (section 3.1.2). If necessary, comparisons between the ncRNAs class and other RNA classes are made. General information on RNAs used/intended for use as therapeutics (including ncRNAs) is provided, expanding on the chemical modifications necessary to avoid degradation and to achieve a target effect after administration (section 3.1.3). Lessons from RNA-based therapeutics as regards the pharmacokinetics of exogenous RNA are discussed, with a focus on non chemically modified (naked)ncRNAs. In addition, aspects of the pharmacodynamics of RNAs as learnt by therapeutics are addressed (section 3.1.4). A description of cellular uptake of RNA (in general and with focus on ncRNAs), and of the many barriers represented by human/animal cells after oral ingestion is addressed insection 3.1.5. Further relevant information on dietary exposure to plant RNAs (i.e. exposure to plant RNAs following dietary consumption) by different diet types, and presentation of the gaps in the literature on dietary plant RNA exposure is provided in section 3.1.6.

EFSA Task 2: Based on available scientific literature (narrative review), to review the effects (physiological, paraphysiological, pathological) of foreign (exogenous) ncRNA on the gastrointestinal tract and annex glands (e.g. liver, pancreas, salivary gland) 

To address EFSA Task 2 preparatory baseline information is provided on i) the GI tract barriers to ingested exogenous RNAs, including ncRNAs (section 3.2.1); ii) experience in RNA-based therapeutics for oral (GI) administration, with focus on delivery of naked (non chemically modified) ncRNAs (section 3.2.2). Biological effects of dietary exogenous ncRNAs (plant and nonplant origin) on the GI tract and its annex glands is presented in section 3.2.3.

www.efsa.europa.eu/publications 8 EFSA Supporting publication 2019: EN-1688 The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

To address EFSA Task 3, preparatory baseline information is provided on the molecular pathways relevant for the uptake of exogenous ncRNAs by human and animal cells and the intracellular trafficking that follows, with description of the effects by exogenous ncRNAs once they reach the specific subcellular location (section 3.3.1). To exert systemic biological effects, in most cases the ingested exogenous RNA must enter the systemic circulation to reach the target tissues. Information is therefore provided on the landscape of exogenous RNAs in biological fluids and tissues from humans and animals, with focus on ncRNA when possible (section 3.3.2). A description of possible systemic biological effects of dietary exogenous ncRNAs is provided in section 3.3.3, with details on tissue barriers (i.e. placenta, brain) encountered. A detailed review is made of the contradictory information on the systemic effects exerted by plant-derived exogenous ncRNAs. Finally, studies on the safety and in general on risk assessment on ncRNAs from GMOs are reviewed (section 3.3.4).

EFSA Task 4: Based on available scientific literature (narrative review), to assess the plausibility of effects on the immune system of humans and animals (experimental animals, livestock and pets) of foreign (exogenous) ncRNA molecules.

To address EFSA Task 4, preparatory baseline information is provided on ncRNA mediated regulation of the immune system in humans and animals. The many pathways of sensing exogenous RNAs, both at the endosomal compartment and in the cellular cytosol, and the mechanism of triggering the immune system are reviewed (section 3.4.1). An overview is done of the plausibility of biological effects of exogenous plant ncRNAs on the regulation and function of the immune system upon uptake by mammalian cells (section 3.4.2). In addition, since the gut microbiota influences the immune system, a review on the possible effects of exogenous ncRNAs as gut microbiota modulators was considered relevant and it is provided (section 3.4.3).

In the final section of this report, concluding remarks inform on the gaps existing in the scientific literature, and highlight areas needing further investigations to better understand and support the food and feed risk assessment of ncRNA-based GM plants. 

An extensive search was done to identify as many studies as possible relevant to the literature review questions. To include all possibly relevant information multidisciplinary databases and information resources were explored. Unpublished research reports ("grey literature") including dissertations, thesis or other scientific reports were also included, which were retrieved using general search engines (i.e. Google). The following information sources, databases or search engines ( Table 2) were used for the literature review: Pubmed (biomedical), World Wide Science (multidisciplinary), Web of Science (multidisciplinary), SpringerLink (multidisciplinary), Scopus (multidisciplinary), SciELO (multidisciplinary) and bioRxiv (preprint biology).

Finally, the publications reporting the outcome of two EFSA procurements aiming respectively at investigating and summarising the state of knowledge on the mode-of-action of dsRNA and miRNA pathways, the potential for non-target gene regulation by dsRNA-derived siRNAs or miRNAs, the determination of siRNA pools in plant tissues and the importance of individual siRNAs for silencing 6 ; and reviewing relevant scientific information on RNA interference that could serve as baseline information for the environmental risk assessment of RNAi-based GM plants ) 7 were also used.

www.efsa.europa.eu/publications 11 EFSA Supporting publication 2019: EN-1688 The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

The methodology used for this literature review process and report preparation followed three basic principles: i) methodological rigour and coherence in study retrieval and selection; ii) transparency; and iii) reproducibility. To ensure that these principles were implemented in this literature review process, the search methodology design, study selection and data and results presentation follow, whenever possible, most of the methods and techniques described in the EFSA guidelines on application of systematic review methodology for food and feed safety assessments to support decision-making (EFSA, 2010) . 7 The following section describes the general key steps followed in the literature search methodology; detailed descriptions of the literature search methodology for each topic are also provided in this section.

To find a comprehensive set of scientific literature related to exogenous ncRNAs and their biological effects on a target organism, either local or systemically, key questions were identified and key words were used as search queries for each task section according to a pre-defined review process.

To provide a fit-for-purpose literature search, the team defined key questions based on the EFSA tender specifications and their interpretation as above described. To support this phase, the team identified key elements of the questions considering the PICO approach of EFSA (2010). 8

The key elements identified include the population, in this case humans or animals (experimental animals, livestock and pets) as defined by the terms of reference of EFSA. The intervention was identified as exogenous RNAs (focusing on ncRNAs). The comparator was identified as the control or reference group (not exposed to exogenous RNAs) in most studies, whenever possible. The comparator could also be identified as the modified (either chemically or biologically) version of the exogenous RNAs when referring to the stability or pharmacokinetic of the exogenous RNA. The outcome is the biological effect (if any) as a response to the intervention. Other specific key questions are determined for every part of the specific tasks defined by EFSA. Evidence is generally scarce for most of the topics covered by this scientific review, and in some cases the information was available from not directly related studies (i.e. pharmaceutical industry studies). The literature review was helpful in identifying both knowledge gaps and unexpectedly relevant studies or evidence not previously known to exist. These knowledge gaps are used to make informed proposals for future research designs (section 4.2).

The key methodological steps in the literature search, selection and review process are shown in Figure  1 . Three phases were defined in this process.

The preparatory phase started with team discussions of topics to be included in the literature review.

During this phase key elements addressing the EFSA tender specifications and their interpretation as above described were identified and key questions were defined. The key questions aimed to to assist the literature search in order to produce a comprehensive overview of the topic that best fits tender specifications.

The bibliography search and selection phase included several steps. The initial search in bibliographic databases and search engines was done using task-specific key words. The key words used for the search were derived from the EFSA tender specifications and their interpretation (Tasks 1-4) and were complemented with other key words identified in the preparatory phase. Identified key words are detailed for each task in section 2. The final phase of review and final selection involved two additional steps, i.e. a detailed review of all relevant previously selected documents and a systematic evaluation of the final selected studies. A detailed review of the selected documents was done for studies with full text availability (either freely available or available for purchase) and, based on this, a final selection was made of full-text studies most relevant to answer the key questions for Tasks (1-4) and apt for review in this report. A systematic evaluation of these final documents was done to develop a comprehensive overview of the topic. Conflicting or contradictory results were reported in a neutral way and supported by references and The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

citations of all available evidence, both from the published and grey literature. Retracted papers were carefully identified and noted. Finally, each section of the narrative review was assessed by one reviewer.

If more specific data was needed because of missing or novel information, additional bibliographic research was performed. Additional key words that could directly address the specific data needed were identified and used to run a new specific search. The search among full-text studies was focused on obtaining only specific missing or required data (e.g. quantitative data on RNAs levels in plants, earlier studies on exogenous RNA exposure to mammalian cells). This new specific search generally represented less than 1% of the literature included in this report and it was performed only once for certain topics.

Finally, to reduce the risk of introducing bias into the literature review process and to assure reproducibility in the review methodology the following general approach was applied:

when the search of the identified "key words" in different databases resulted in a myriad of duplicate documents, manual removal of these duplicates was performed in each section by each team expert;

when relevant studies that might answer the identified key questions were identified after the first search run, these were used to further refine the literature search, for example refining the keywords;

each selected full-text study to be included in the result section was manually examined by the team.

To identify literature relevant to the topic, the methodology described in section 2.2.1.2 ( Figure 1 ) was applied, with adaptations. A multiple-step approach was followed based on the expertise of the team members who carried out research in the area. Following a first search based on identified keywords, a careful selection of retrieved review studies was done, followed by a search of specific publications referred to in these reviews. Additional specific key words were identified, and an additional search was run. This search retrieved a relatively small number of studies; therefore, team members reviewed all these to determine which were relevant to the topic. Duplicates, irrelevant or erroneously assigned publications were removed. Team members also cross-checked each others' findings, sharing discussing the results of their searches. The following areas were investigated.

The purpose of this search was to provide general information on RNAs function, movement in the plant and aspects of biogenesis and degradation, with a focus on ncRNA (either small or long). By comparing animal and plant ncRNA pathways, the aim was to identify similarities between animal and plant ncRNA processing machineries that would indicate the possibility of ingested plant-derived ncRNAs being processed in animals. In addition, this section was organized to avoid details that were outside the scope of this literature review in the topic of plant ncRNA biology. This serves as an introductory chapter allowing easier integration of the knowledge presented in later chapters. The following combination of key words or phrases was used in some of the databases indicated above:

Plant small RNAs; plant AND siRNA; plant AND miRNAs; plant AND lncRNAs; plant and circRNAs; plant RNA AND movement.

www.efsa.europa.eu/publications 14 EFSA Supporting publication 2019: EN-1688 The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

The initial search using the above-mentioned key words or phrases yielded a large number (≈50 000) of documents ( Table 3) , most of which were not relevant to the topic. Particularly, the databse Web of Science retrieved documents not directly related to the topic.

To assist and refine the search, studies were further filtered by their relevance to the key questions identified:

-What type of ncRNAs are described in plants?

-What are the main differences between mammalian and plant ncRNAs?

-What is the function of plant ncRNAs?

-What is known about the movement of plant ncRNAs?

Title and abstracts were evaluated, and studies further refined using the key questions. Detailed study of the most relevant full-text documents answering the key questions, retrieved 173 documents which were finally selected as relevant to give an overview of plant ncRNAs. search was then refined by using other key words including "half-life", "turnover", "fate" or "degradation". The identified studies were then manually evaluated by title and abstract and their number further limited due to their low relevance to the below identified key questions for this specific subchapter:

-Are plant ncRNAs more stable than animal (mammalian) ncRNAs?

-If so, what makes a plant ncRNA more stable than its animal (mammalian) counterpart?

-Why is it relevant to consider ncRNA stability outside the organism of origin (plant)?

-Are certain types of plant ncRNAs more or less stable than others?

-What is known about the half-life of ncRNAs? Documents on the stability of RNAs (not only ncRNAs) were further screened for relevance to the question. Most of the studies in the literature describe the stability of a specific class of intracellular RNAs (i.e. mRNAs). The stability of RNAs other than ncRNAs (e.g. mRNA, tRNA), unless are used as model of general ncRNA stability, is outside the scope of this literature review, since it is very unlikely they would be used as baseline information for food/feed risk assessment of ncRNA-based GM plants. These keywords were specifically selected since they describe the kinetic profile of exogenous ncRNAs in human and animals. The word "pharmacokinetics" was included in the search because it is a general term that retrieves data on absorption, distribution, metabolism or excretion of any molecule, particularly those intended for use as therapeutics (pharmaceutical/medicine areas). The keywords "increased intestinal permeability" or "increased plasma clearance" are intended to retrieve any study indicating changes in these parameters. Although a large number of publications unrelated to ncRNAs were expected, these keywords were chosen to initially explore the general aspects of physiological or pathological conditions associated with modification in these parameters. The search was further refined after this initial exploration. The search of the above-mentioned key words on different databases yielded ≈27.000 entries ( Table 6) . A quick exploration of titles and abstracts suggested that (as expected) a number were unrelated to or were not useful to appropriately address the topic. It is worth noting that the search in the previous section (section 2.2.2.1.) indicated that the chemical modifications generally introduced for RNA-based therapeutics are normally absent in nature. Scientific data that could serve as a baseline to support risk assessment of ncRNA-based GM plants should contain primarily information on ncRNAs resembling those present in the nature or in GM plants. Therefore, studies on naked exogenous RNAs were analysed to prepare this narrative review. To promote comparisons between natural and synthetic RNAs, a very few examples of slightly chemically modified RNAs were included. Heavily chemically modified RNAs were not considered due to their unlikely existence in nature. Only in vivo studies were considered, both in animals (mammals) or humans. Studies were further evaluated for their relevance to answer the key questions identified for this section:

-What are the pharmacokinetic properties of RNA-based therapeutics?

-What routes of administration have been used for RNA-based therapeutics?

-What is it known about the pharmacodynamic properties of RNA-based therapeutics?

-Do disease conditions modify the pharmacokinetic properties of RNAs when administered to mammals? The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

After detailed evaluation of the full-text reports, 119 documents meeting the described criteria were finally selected for review. 

Understanding whether and how exogenous RNAs are taken up by cells and what barriers are encountered is relevant to study the possible absorption and systemic exposure to ncRNAs introduced by food and feed. This topic is reviewed in this dedicated section and excluded from the section on the pharmacokinetics of exogenous RNAs (section 2.2.2.1. above). To search and select the relevant studies concerning the uptake of exogenous RNAs in mammals, the following combination of key words was used in some of the databases listed above:

RNA AND cellular Uptake; ncRNA AND cellular uptake; miRNA AND cellular uptake; lncRNA AND cellular uptake.

These key words were used to screen for studies where "cellular uptake" was mentioned in the title or abstract for any type of RNAs. It was also decided to include specific types of ncRNAs including "miRNA" or "lncRNA" to expand the search to literatures where the more general term "RNA" was not necessarily used. The initial search of scientific documents using the above keywords yielded ≈7000 documents ( Table 7) . These were explored by their relevance (title and abstract), and only studies addressing naked ncRNAs were considered for full text analysis, since highly chemically modified RNAs are not common in nature. To address the reference terms for task-1 of this procurement, and after expert discussion for this section (preparatory phase) key questions were identified and the studies then further filtered for their relevance to the these:

-Are exogenous RNAs taken up by mammalian cells?

-How does exogenous RNAs uptake in mammalian cells occur?

-What are the biological barriers to consumed RNAs?

-What are the biological barriers for the exogenous RNAs after absorption and entering into the mammalian circulatory system?

Only studies where exogenous RNAs were evaluated were included in the literature review. After detailed evaluation of full-text reports and expert judgement on their relevance to EFSA Task 1, 39 documents were finally selected for review. 

To select relevant studies for plant RNA exposure, the following combination of keywords were used: plant AND RNA content; plant AND miRNA content; plant AND siRNA content; plant AND lncRNA content; plant AND dsRNA content; RNA intake AND vegetarians; vegetables/fruit intake AND vegetarians.

These specific keywords were used because the wording "RNA (ncRNA, miRNA, siRNA, dsRNA, lncRNA) content" would yield most of the documents related to quantities of RNAs in a specific substrate. These were combined with "plant" to retrieve only those studies where plants were used to evaluate RNA content. It was also decided to use "RNA intake", "vegetable intake" or "fruit intake" to specifically refer to consumption, and to combine this with "vegetarian" to search only for studies where comparisons are made between vegetarian and other types of diets. Alternative wordings including "RNA amount" or "vegan" were used to search for alternative information. Using the above keywords, ≈3300 studies were retrieved (Table 8) during the initial search. To more appropriately cover this section and to refine the search, the following key questions were identified during the preparatory phase:

-What is the general amount of RNAs naturally occurring in edible plants?

-What percentage of total plant RNAs are ncRNAs, miRNAs, siRNAs, lncRNAs or dsRNAs?

-Does exposure to RNAs change in different diet types?

-Do changes in RNAse activity, due to pathologies or human polymorphisms, influence dietary exposure?

Titles and abstracts were evaluated, and studies further filtered using the key questions above. The search was also refined to focus on those studies where quantitative RNAs data were mentioned. Only in vivo exposure studies were used (either epidemiological or experimental). Common types of diets were only considered (i.e. vegetarian, vegan or omnivorous). For RNA exposure by different diet types, studies where comparative analysis of at least two different diets (one of which was the general population diet type) were mainly considered. In all cases, full-text studies were evaluated to screen for relevance to answer the identified key questions. A detailed study of the full-text documents resulted in the identification of 39 studies most relevant to address this specific topic of Task-1, and which were reviewed for this section of the report. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. To identify literature relevant to the topic, the methodology described in section 2.2.1.2 (Figure 1 ) was applied. During the preparatory phase key questions were identified to assist in the literature search.

The document search was first done using specific key words and then refined according to the key questions. Detailed analysis of the full-text documents was done by team members with expertise in the area. Search and analysis results were discussed collectively among team members. The following areas were investigated:

In this subsection, there is a general description of the biological barriers within the GI tract that exogenous RNAs would need to overcome to exert a biological effect, either locally or systemically. To select relevant studies, the following combination of key words or phrases were used:

"Gastrointestinal tract anatomy"; "gastrointestinal tract physiology"; intestinal transport pathway; physicochemical properties of RNAs.

An exploration of titles and abstract of the yielded entries (Table 9 ) suggested that a large number of documents were unrelated to the topic.

Only one key question was identified in the preparatory phase:

-What are the biological barriers within the gastrointestinal tract to ingested exogenous RNAs?

The search for relevant studies addressing this key question was based mainly on team member judgement. Detailed study of the most relevant full-text documents answering the key question retrieved 21 documents which were finally selected as relevant to this section of the report. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

This search was intended to collect data from RNA-based therapeutics employing oral administration (and other GI routes) which could support the search on the possible biological effects of plant dietary exogenous RNA and ncRNAs on the GI tract and its annex glands. The initial search was done using the following key words on the different databases:

ncRNA oral delivery; RNA oral delivery; siRNA oral delivery; aptamer AND Oral; siRNA gastrointestinal delivery; miRNA gastrointestinal delivery; oral exosome RNA; extracellular vesicle AND gastrointestinal.

These key words were selected to search for any documents related to RNAs and oral administration.

Any information on oral administration was considered relevant. To expand the search, different terms referring to different types of ncRNAs (miRNA, aptamer, siRNA or others), and a synonym for "administration", i.e. "delivery", were included. The initial search using the above key words or phrases yielded a small number (≈800) of documents ( Table 10 ). The different databases generated similar number amounts of publications, most of which were duplicates. Using the key word "lncRNAs" in combination with the other key words did not yield any information relevant to the topic. Indeed, just the search using "lncRNA exosome" identified fewer than 100 documents, but none was relevant to oral delivery or GI administration. The wording "dsRNA AND oral" produced ≈300 documents, but none of them was relevant to mammalian organisms. The documents were selected by relevance, evaluating title and abstract. In this section, only studies in animals (mammals) or humans were considered. When replacing the word "delivery" with "administration", ≈8000 additional documents were identified, but almost 100% were unrelated to the review topic. Since most of the documents in the literature describe minimally modified RNAs for oral use, all chemically modified RNAs were considered in this section. These can provide baseline information to support food/feed risk assessment of ncRNA-based GM plants. To assist and refine the search, studies were further filtered for their relevance to the key question identified:

is the oral route of administration relevant for RNA-based therapeutic development?

-What types of formulation (delivery vehicles) are tested for RNA-based therapeutics for the oral route of administration or other gastrointestinal tract routes of administration? The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

-Do exogenous ncRNAs (naked RNAs) resist the conditions of the gastrointestinal tract environment?

-Could ncRNAs-based therapeutics -delivered by oral administration or other GI tract routes of administration-exert a biological effect in the GI tract or its annex glands?

-Are there any natural vehicles for oral administration of exogenous ncRNAs?

From the initial ≈8000 entries identified using the above key words and after filtering for relevance to the key question with evaluation of titles and abstracts and detailed study of the full-text reports, 110 documents were selected for review in this section.

Twenty (20) of these documents reports selected examples of local effects of nucleic acids-based drugs (mainly exogenous RNAs) within the GI tract, using oral administration.

Since EFSA required a literature review also including livestock species, an additional specific search was run for birds and fish. This search was intended to compile information on administration of exogenous ncRNAs in either fish or birds, regardless of administration method, exerting any biological effects in vivo. The initial search was done using the following key words on the different databases: exogenous RNA AND fish; Exogenous RNA AND bird; RNA oral delivery AND fish; RNA oral delivery AND bird; ncRNA administration AND fish; ncRNA administration AND bird.

To expand the search, different terms referring to different types of ncRNAs were used (siRNA, miRNA, circRNA, or dsRNA). The initial search using the key words or phrases produced a small number (≈500) of documents (Table 11 ) in most of the databases. To assist and refine the search, only one key question was identified during the preparatory phase for this subsection:

-Are exogenous ncRNAs also used to exert RNAi effects in fish or birds?

From the initial ≈500 entries identified using the above key words, and after filtering for relevance to the key question, and evaluating titles and abstracts and detailed study of the full-text reports, 18 documents were selected for the review in this subsection. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. 

This chapter provides baseline information supporting food/feed risk assessment of ncRNAs-based GM plants. In the effort to cover the widest amount of literature containing relevant scientific data on exogenous ncRNAs and their possible biological effects, the following key words or phrases were used in the search:

dietary plant non coding RNAs; food plant non coding RNAs; exogenous plant non coding RNAs; diet plant microRNAs; food plant miRNAs; breast milk exosomes; breast milk miRNA; breast milk ncRNAs; breast milk lncRNAs; breast milk RNA; dietary exogenous RNA.

These key words were selected to cover most, if not all, available studies on the topic ( Table 12) . The search of other common examples of exogenous ncRNAs (non-plant origin) consumed by oral intake was focused on breast milk ncRNAs, due to their relevance to human nutrition (Table 13) .

Almost all documents yielded by the three databases (PubMed, Scopus or Web of Science) were duplicates. In this section, studies reporting local effects (in the GI tract or its annex glands when consuming exogenous ncRNAs) were preferentially (but not exclusively) evaluated. Other databases including the preprint server for biology (bioRxiv), SciELO or search engines (i.e. Google) were also used. The initial search in some databases using the above key words yielded a number of entries (Table 12 and 13) not exceeding ≈2600 documents. Because this specific topic is especially salient to future understanding of the possible food/feed risk assessment of ncRNAs-based GM plants, all the identified studies were further evaluated by title and abstract and their relevance to answering the identified key questions:

does the literature describe the effect of exogenous plant-origin ncRNAs when consumed orally?

-Does the literature describe negative results or contradictory results regarding the possible biological effect of ncRNAs when consumed orally?

-Are ncRNA stability properties under GI tract conditions normally evaluated?

-Are exogenous ncRNAs evaluated quantitatively?

-Are exogenous ncRNAs consumed in an amount sufficient to exert a local biological effect?Are there other common examples of exogenous ncRNAs (other than plant-origin) that could exert a biological effect on the GI tract and its annex glands when ingested orally?

Local effects (in the GI tract or its annex glands) are mainly described in this section. The possible mechanisms of plant exogenous ncRNAs cellular uptake, their intracellular trafficking and systemic biological effects (including plasma levels) are reviewed in another section (Section 3.3.). Of the initial ≈2500 documents retrieved using the above key words, and after filtering for relevance to the key questions by evaluating titles and abstracts, and detailed study of the full-text reports, 64 documents were selected for the review in this subsection. From these, 29 documents were selected for plant-origin ncRNAs and 35 for non-plant origin (breast milk) ncRNAs. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. 

In line with the methodology described in section 2.2.1.2. (Figure 1 ), a multiple-step approach based on the expertise of the team members responsible for this section was applied to identify literature relevant to the possible systemic effects of exogenous ncRNAs. The document search was first done using specific key words in the databases and refined based on the appropriate key questions. Detailed analysis of the full-text documents was done by team members and the results discussed collectively. The below areas were investigated.

The search objective was to obtain information on basic molecular cellular uptake pathways of exogenous ncRNA, including proteins involved in this process (if any). A review was also done on several other aspects of intracellular trafficking, once an exogenous ncRNA is internalized, and how it exerts a biological effect. Tissue barriers to exogenous ncRNA function were also investigated. The content of this section deepens on molecular pathways previously described in section 3.1.5. To assist the search the following key words were used in the different databases: exogenous ncRNA AND mechanism AND absorption; ncRNA AND mechanism AND absorption; SIDT1 (SID 1 transmembrane family member 1) AND ncRNA; SIDT2 (SID 1 transmembrane family member 2) AND ncRNA; ncRNA AND uptake; intracellular trafficking AND ncRNA.

These key words and terms were selected to identify any document related to the mechanisms of uptake and intracellular trafficking of ncRNAs exerting a biological effect. The search was expanded by adding and combining different terms for exogenous ncRNAs (see Table 14 ), including the general term "RNA", www.efsa.europa.eu/publications 25 EFSA Supporting publication 2019: EN-1688 The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

or specific ncRNAs such as "miRNA". The initial search using the above key words or combinations thereof produced a moderate number of documents (Table 12 ). The different databases generated similar amounts of documents, most of which were duplicates. The largest proportion of the references was obtained using the key wording "ncRNA AND uptake" (≈2500). However, most of these had no relation to this specific section. The search was refined by filtering for relevance to answering the identified key questions:

is there any clear transport of exogenous ncRNAs into mammalian cells?

-Are orthologous of environmental RNAs transporters in lower organisms (i.e. SID family of proteins) involved in the mammalian exogenous ncRNA mechanism of uptake?

-What molecular mechanisms are involved in exogenous ncRNAs trafficking inside the cell?

-What is the cellular fate of exogenous ncRNAs once inside the cell?

-What are the specific tissue barriers to exogenous ncRNA function?

Very few documents evaluate the molecular mechanism of exogenous ncRNAs uptake, intracellular trafficking and function. From the initial ≈3500 documents retrieved using the above key words, and after evaluation of their titles and abstracts and filtering for relevance to the key question, less than 50 documents initially seemed relevant to the topic and 33 were finally selected for review after detailed study of the full-text reports. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

This subsection is intended to retrieve available information on the presence of exogenous RNAs in human and animal biological fluids (i.e. blood, plasma, serum or any other biological fluids) and tissues. The search focused on the presence of dietary plant-derived exogenous ncRNAs, due to their relevance to the overall literature review. To cover the largest possible portion of the relevant literature, the

following key words were used in the search:

exogenous RNA AND human plasma; exogenous RNA AND serum; plant ncRNA AND human tissue; plant lncRNA AND human plasma; plant miRNA AND blood; plant miRNA AND tissue; plant RNA AND tissue.

To expand the search, the above key words were used in different combinations for the same search (i.e. blood, serum or plasma for biological fluids) ( Table 15 ). The general wording "RNA" was also incorporated into some searches. Specific types of ncRNAs were also added, including "miRNA" or "lncRNA", to avoid missing documents possibly related to these specific molecule types.

The search using these key words or phrases in different databases yielded more than ≈20.000

publications (Table 15) . A quick exploration of the titles and abstracts suggested that only few studies were related to the review topic. Only one key question was identified as relevant which was related to the presence or absence of evidence of exogenous ncRNAs in biological fluids and tissues. Since these findings are important for the risk assessment of ncRNA-based GM plants, the key question was presented in its positive and negative forms as follow:

are there studies describing the presence of exogenous RNAs in the biological fluids and tissues of humans and/or animals?

-Are there studies contradicting the presence of exogenous RNAs in the biological fluids and tissues of humans and/or animals?

After a detailed assessment of the full-text, 20 documents were finally selected for review. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. 

This search primarily aimed to gather information relevant to understand the possibility of systemic effects of dietary exogenous ncRNAs after intake (oral administration) by humans and/or animals (mammals). Information on the possible passage through specific biological barriers (i.e. placenta or brain) was also searched. The search was guided by the following key words to identify studies providing any data on the possibility of systemic effects of dietary exogenous ncRNAs: plant ncRNA AND diet; plant exogenous RNA AND diet; plant RNA AND diet, plant environmental RNAi AND diet; plant RNA AND placenta; plant RNA AND brain; dietary RNA AND cross kingdom.

These key words were used to guarantee that every single document potentially providinginformation on the possibility of systemic effects of exogenous ncRNAs was found. To widen the search, other wording such as "environmental RNAi", "cross kingdom" or "mobile RNAs" were used. Although the wording "systemic effects" was not used in the key words during the search, the above key words were used to retrieve any documents in which a biological effect (local or systemic) might be reported. Different wording for the term ncRNAs was also used (including lncRNA, circular RNA, miRNA, or dsRNA).

The initial search in some databases using the above key words produced a small number of entries (Table 16) . The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. Approximately ≈1900 documents were initially identified using the above key words (Table 16) . After filtering for their relevance to the key question by detailed assessment of the full-text reports, 48 documents were finally selected for review. Within this specific subsection, the selected documents are mainly grouped as studies that "support the evidence" or studies that "contradict the evidence" of systemic biological effects of dietary exogenous ncRNAs in humans and animals (mammals).

Although not specifically included in the scope of this literature review, search for toxicological effects of dietary exposure to exogenous ncRNAs in humans or animals is relevant to risk assessment. This search was aimed at identifying information on possible toxicological effects of plant-derived exogenous ncRNAs following food/feed consumption. The initial search was conducted using the following key words in the different databases: The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Few documents were initially retrieved and after a quick evaluation of titles and abstracts, it became evident that very little information is available in this area. Therefore, different terms were used to refer to exogenous ncRNAs, including circular RNA, lncRNA and miRNAs. The search using the above key words yielded a small number of documents (Table 17) . Only in vivo studies were considered, both in animals (mammals) and humans (if available), with focus on dietary plant-origin exogenous ncRNAs.

The key questions identified to refine the literature search for this specific topic were:

-Is there evidence of any toxicological effects of plant-derived dietary exogenous ncRNAs?

-Are the toxicological effects related to RNAi-mediated interaction with the human or animal genome?

-Are there possible interactions related to unintended RNAi-mediated gene regulation?

Approximately 700 documents were identified using the above key words. After filtering for relevance to the key questions, and assessing the full-text reports, only 12 documents were finally selected for review. A systematic and comprehensive review and collection of the literature relevant to the topic was performed with the aid of the team members' expertise. An extensive search using key words was initially done on multidisciplinary databases to avoid publication bias, followed by a grey literature search using general search engines. The resulting documents were screened by title and abstract to determine relevance to the topic and refined using the identified key questions. After full-text analysis of the documents, irrelevant documents were eliminated. All stages of the screening process were independently assessed and, in case of uncertainty, discussed with other experts to avoid personal biases. The methodology followed the key steps described in section 2.2.1.2. (Figure 1 ) and a total of 91 papers were identified, investigating the areas below described. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

The final goal of this search was to identify relevant information on ncRNAs-mediated processes associated with innate and adaptive immune responses. This section serves as an introduction to the studies presented later, and it is based on the results of the seach described in the next paragraph. Sixty-four (64) documents were reviewed for this section.

The search was aimed primarily at gathering state-of-the-art knowledge on the possible biological effects of exogenous (plant) ncRNAs on the regulation and function of the immune system.

To cover this specific section, a combination of the following key words was used in the different databases:

Exogenous RNA AND immune AND human/mice/rat; exogenous plant RNA/miRNA/siRNA/lncRNA/dsRNA/transgenic RNA AND immune function/inmmunity AND human/mice/rat; genetically modified plants AND immune function/immunity AND human/mice/rat; exogenous plant RNA/miRNA/siRNA/lncRNA/dsRNA AND primary cells AND lymphocyte/monocyte/macrophage; exogenous RNA/genetically modified plants AND immune function/immunity AND zebrafish.

The key words were designed to identify all publications that use in vivo, including non-mammalian animal models, and in vitro ncRNAs testing to evaluate the influence on tolerance processes as well as immune cell function. The search also included publications on in silico methods. The different key words were entered one by one into the search field 'Topic' of the databases ( Table 2 ). The search was designed to identify documents including methods that could use non-mammalian animal models (i.e. zebrafish embryos). All search terms within one key word were combined by the operator 'OR' and 'AND' copied in a second or third search field. 'Review' and 'meeting' documents were excluded. A search for grey literature was done in general search engines (i.e. Google) using the different key words in the 'Search' function of the respective websites. All the information on in vivo studies, using mammalian and/or non-mammalian organisms, and in silico studies was collected. These key words were chosen because this section aimed to describe the possible effects that exogenous ncRNAs, specifically of plant dietary origin, might exert on the immune system of humans and animals.

The search yielded ≈2255 entries (Table 18) . Publications were excluded if one of the following criteria was met: the biological effect could not be unambiguosly attributed to ncRNAs; the publication addressed immunodeficiency or autoimmunity; the publication reported no immunity (innate or adaptive)-associated endpoint or it is not specific for immune function and/or homeostasis evaluation; evaluation of exposure of and/or effect on cells other than immune cells (e.g. epithelial cells); the study was unable to measure processes related to immune function and/or homeostasis; the study reported data on potentially immunocompetent cells not fully differentiated as mature immune cell; no ncRNAbased plant was tested in the publication; and the publication did not address plant-derived exogenous ncRNA.

The unexpectedly low number of publications retrieved using the key words proved a challenge in the selection process. For example, selection for in vivo studies produced only 53 articles after title/abstract screening.

To expand the number of documents, additional key word searches were included and less conservative criteria were set for the title/abstract screening. Retrieval of grey literature was also based on information containing "preliminary results" or on-going projects that could possibly answer the key questions.

www.efsa.europa.eu/publications 31 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. Studies were further evaluated for their relevance to answering the key questions. The key questions identified using the above key words were:

-Which test methods or approaches are available to investigate the possible role of plant-derived exogenous ncRNAs on the immune system of humans or animals? -What human or animal immune response(s), innate or adaptive, are possibly associated with administration of plant-derived exogenous ncRNAs? -Which mediators in humans and animals have been identified and/or ascribed to immune response(s) possibly mediated by administration of plant-derived ncRNAs? -Which innate immunity mediators (i.e. chemokines, cytokines) have been identified in humans and animals as associated to effects of plant-derived ncRNAs? -Which human or animal cells (i.e. lymphocytes, macrophages, myeloid) are possibly involved in immune responses mediated by plant-derived ncRNAs?

After the selection of the full-text documents, publications were excluded if not providing sufficient information (i.e. endpoint, exposure time), if investigating a mechanism other than immune-related processes, or if describing an operational procedure or guideline rather than a research study. From the initial ≈2255 documents identified using the above key words, and after filtering for the relevance to the key questions and detailed study of the full-text reports, 91 documents were finally selected for review in the three sections of Part 4 (EFSA Task 4). Of these, 22 documents were selected for review in this section.

Gut microbiota is important for the function of the immune system, therefore the purpose of this additional search was to provide general information on the effect of dietary exogenous ncRNAs on the gut microbiota and the possible consequent secondary modulation of human or animal immune systems. This would serve as an introduction to identifying information gaps and experimental needs for future studies relevant to food/feed risk assessment of ncRNA-based plants. Selected full-text studies were analysed for their relevance to answering the following key question:

-Is there evidence for influence of exogenous plant ncRNAs on the gut microbiota related to the immune system?

Five documents were finally selected for review in this section. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Results are presented narratively. When appropriate, tables and figures are included to allow better understanding of the topic covered by the literature review. Some sections of this review provide only limited amount of quantitative analysis due to the nature of the available evidence and the key questions identified for each subchapter. At the end of each subchapter, a short summary presents the general points identified in the reviewed literature. Gaps in the literature and open questions identified in the field are presented in a subchapter of certain sections. These gaps in the literature may serve as suggestions for further research in each topic.

Part 1: Kinetics of exogenous ncRNAs in humans and animals (EFSA Task 1)

ncRNAs are transcripts that are not translated into any functional peptides or proteins. These transcripts include structural (transfer, ribosomal, small nuclear, and small nucleolar RNAs) and regulatory ncRNAs. The latter comprise both small and long ncRNAs that can regulate gene expression by acting on chromatin, transcription, RNA processing, RNA stability and translation. This part of the report introduces general features of regulatory ncRNAs including their function and aspects of their biogenesis and degradation. It also establishes possible comparisons between animal and plant ncRNA pathways. This can be considered background information to clarify the possibility of plant ncRNA biological effect in humans and animals that ingest food/feed from ncRNA-expressing GM plants. In addition, ncRNA movement inside the plant is also described. Specific details on the structural ncRNAs and the broad topic of plant ncRNA biology are outside the scope of this literature review. Following an extensive literature search as described above and based on the team expert decisions on the topic, this section is a review of 173 scientific documents.

Small ncRNAs (sRNAs) sRNAs are common to both plants and animals. Two major classes of sRNAs are found in eukaryotes: microRNAs (miRNAs) and small interfering RNAs (siRNAs). They function as regulators of endogenous genes or as defenders from invasive nucleic acids. They are characterized by the double-stranded nature of their precursors, and differ from the less abundant piwi-interacting RNAs (piRNAs) which derive from fragmentation of single-stranded RNAs (Juliano et al., 2011) , and are primarily found in animals; they have not been described in plants. piRNAs exert their functions in germ and stem cells through interaction with Piwi-proteins (Juliano et al., 2011) . By contrast, both miRNAs and siRNAs are bound to Argonaute (AGO) proteins, where they identify 'target' RNAs by base-pairing interactions, and act in both somatic and germ cells. All sRNAs in plants are modified at the 3' -terminus by 2'-O-methylation, including miRNAs and siRNAs, which lack this modification in animals. 2'-O-methylation is essential to conferring stability and protection from 3' uridylation and degradation (Borges and Martienssen, , 2015) . miRNAs and siRNAs are distinguished by their origin and biogenesis (Figure 2) . miRNAs are derived from processing of single-stranded precursors with a hairpin structure, whereas siRNAs are generated from long, fully complementary double-stranded RNA (dsRNA) precursors (Carthew and Sontheimer, 2009 The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

endonucleases (DCLs) that excise RNA precursors into short double-stranded fragments, 20-24 nucleotide (nt) long, with 2-nt 3' overhangs; and Argonaute proteins (AGOs) that engage these duplexes and support their silencing activies based on target complementarity (Carthew and Sontheimer, 2009 ). Further details on siRNAs and miRNAs are provided below.

siRNAs Several different classes of siRNAs with specialized biological roles have been identified in plants. The most abundant class consists of heterochromatic siRNAs (hetsiRNAs), which are usually 23-24 nt long and originate from the repetitive and intergenic regions in the chromosome and transposable elements. hetsiRNA are crucial in DNA methylation and chromatin modification through a process known as RNAdirected DNA methylation (Daxinger et al., 2009) . hetsiRNAs are processed by DCL3 nuclease and preferentially loaded into AGO4. Mature Ago4-siRNA complex can interact with complementary noncoding nascent RNA polymerase triggering the recruitment of DNA methyltransferase (domains rearranged methyltransferse 2) to silence target loci at the transcriptional level via DNA methylation and repressive chromatin modifications (Vaucheret, 2006; Lee and Carroll, 2018) . These hetsiRNAs may account for approximately half of a plant's total mass of sRNAs. Another class of siRNAs are naturalantisense transcript siRNAs (natsiRNAs) which can be generated from dsRNA precursors through hybridization of independently-transcribed complementary RNA strands (Vaucheret, 2006) .

In addition, secondary siRNAs generated as a "secondary effect" of miRNA-mediated target cleavage are found in plants (Axtell, 2013; Vaucheret, 2006) . The miRNA-mediated cleaved target is occasionally used by RNA-dependent RNA polymerase (RDR) to produce secondary siRNAs, which can either give rise to a phased set of siRNAs (phasiRNAs) or trans-acting siRNAs (tasiRNAs) with the ability to target genes different from their loci of origin. This secondary pool of siRNAs can greatly amplify and sustain a systemic silencing throughout the organism. Recognizable RDR-encoding genes are present in the genome of many RNAi-competent eukaryotes, with the notable exceptions of insect and vertebrate species in which these secondary siRNAs are absent (Carthew and Sontheimer, 2009 ). The lack of these secondary siRNAs might have a positive effect on specificity (as siRNA amplification can lead to the silencing of multiple transcripts, specifically if they share a highly conserved sequence or a common exon) and a negative effect on the amplification of RNA silencing in these species.

Plants have more diversified and specialized siRNA-based pathways than other organisms, which are thought to contribute to plant plasticity . It is generally accepted that these pathways evolved as a cellular defence mechanism against RNA viruses and transposable elements, which were later adapted to regulate the expression of endogenous genes. This is consistent with the fact that most small RNA classes have a recognized role in defence responses, as well as in epigenetic regulation, and that plants have larger and repetitive genomes (Borges and Martienssen, 2015) .

The diversification of sRNA-directed silencing pathways in plants occurred through the expansion of the RDR-polymerases, DCL and AGO proteins. RDR genes are found in RNA viruses, plants, fungi, protists and some animals, but are absent in flies, mice and humans; this is consistent with the fact that the vast majority of sRNAs in humans are miRNAs (Kurzynska-Kokorniak et al., 2015) .

Absence of RDR activity may also justify the lack of intercellular spreading of RNA silencing in vertebrates, whereas systemic silencing is a phenomenon widely reported in plants and nematodes. Dicer or Dicer-like proteins in plants constitute a four-member gene family, whereas vertebrates have only one member (Mukherjee et al., 2013) . The diversification of AGOs has resulted in development of distinct gene-silencing processes based on differential AGO affinities to small-RNA duplexes (Borges and Martienssen, 2015) .

www.efsa.europa.eu/publications 34 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. miRNAs miRNAs are well-studied sRNAs deeply conserved over long evolutionary distances, even between the plant and animal kingdoms. However, there are substantial differences between these two kingdoms with regard to miRNA biogenesis, and mechanism and scope of miRNA-mediated gene regulation (Jones-Rhoades et al., 2006) . Detailed pathway comparisons and species information are described elsewhere . Transcription of miRNAs is typically performed by RNA polymerase II, and transcripts are capped and polyadenylated. In animals, most miRNAs are derived from longer hairpin transcripts (pri-miRNA) by the consecutive processing by the RNase III-like enzymes Drosha and Dicer, whereas in plants only Dicer (particularly DCL1) is responsible for miRNA processing. Thus, the biogenesis of most miRNAs in plants occurs in the nucleus, whereas in animals it requires the sequential cleavage of pri-miRNA in the nucleus and then in the cytoplasm by distinct RNAseIII enzymes (Rogers and Chen, 2013) .

Once processed, one strand of the hairpin duplex is loaded into an AGO protein to form the core of the miRNA-induced silencing complexes (miRISCs). miRISCs silence the expression of target genes predominantly at the post-transcriptional level (PTGS). The targets to be silenced are recognised through base-pairing interactions between the loaded miRNA and mRNA target, which contains a partially or fully complementary sequence (Huntzinger and Izaurralde, 2011) . Plant miRNAs recognize fully or nearly complementary binding sites, which are generally located within the open reading frames (ORFs) of the mRNA target. Of importance is that miRNA nt 9-12 are usually engaged in base pairing, which allows target cleavage by AGO proteins (between nt 10 and 11). By contrast, animal miRNAs recognize partially complementary binding sites, which are generally located in 3' UTRs. In both plants and animals, complementarity to the 5' end of the miRNA (the 'seed' sequence, containing nt 2-7) is a major determinant in target recognition and is sufficient to trigger silencing. For most miRNA-binding sites complementarity is limited to the seed sequence (seed-matched sites) or to the seed sequence plus miRNA nucleotide 8. However, in some rare cases complementarity to the 3' region of the miRNA might contribute to target recognition, particularly when the mRNA has a weak seed match. Even for these sites, however, miRNA nucleotides 9-12 generally bulge out, preventing endonucleolytic cleavage by AGOs. In both animals and plants the miRNA 5' terminal nucleotide is buried in the mid domain of AGOs and is not available for pairing with the target (Huntzinger and Izaurralde, 2011) . Therefore, there is an important difference in complementarity between plant and animal miRNAs and their targets; this is less extensive in animals than in plants. In both animals and plants miRNAs can move from cell-tocell. While in mammalian cells miRNAs and other types of RNAs can be transferred through secretory vesicles (Ruvkun, 2008; Valadi et al., 2007) , in plants they mostly use a different mechanism (Kobayashi and Zambryski, 2007) (see section 3.1.1.4. for details). However, recent evidences also suggest that plants can send small RNAs in extracellular vesicles to pathogens to silence virulence genes (Cai et al., 2018) .

Regarding the silencing mechanism, miRNAs were initially thought to inhibit translation in animals and to predominantly promote target endonucleolytic cleavage in plants. However, recent evidence has changed this view by showing that miRNAs can trigger translational repression and mRNA destabilization in both kingdoms. In both plants and animals, the current evidence suggests that target mRNA degradation provides a major contribution to silencing by miRNAs (Huntzinger and Izaurralde, 2011 The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. 

Next generation sequencing approaches have revealed a new transcriptional landscape in which many novel lncRNAs have been identified. These are transcripts longer than 200 bp without any clear protein coding potential. lncRNAs are transcribed either from intergenic regions (lincRNAs), introns (incRNAs) or the opposite strand to protein-coding genes or other lncRNAs; they are thus natural antisense transcripts (NATs) (Ariel et al., 2015; Chekanova, 2015; Shafiq et al., 2016; Yamada, 2017) . As a confirmation of their widespread nature and possible biological relevance, lncRNAs have been identified in plants, fungi and animals. In fact, lncRNAs are involved in diverse biological processes across the eukaryotes, ranging from regulation of mating type in yeast to the pluripotency of embryonic stem cells in mammals (Zofall et al., 2012; Flynn and Chang, 2014) . In plants, lncRNAs play key roles in flowering time regulation, gene silencing, root organogenesis, seedling photomorphogenesis, and reproduction (Ariel et al., 2015; Chekanova, 2015; Shafiq et al., 2016; Yamada, 2017) . Although plant lncRNA biological characterization is still limited, detailed analysis of over 200 Arabidopsis thaliana transcriptome data sets identified 40,000 putative lncRNAs, including approximately 30,000 NATs and over 6000 lincRNAs. Both in plants an mammalians (Jin et al., 2013; Liu, J et al., 2012) these lncRNAs do not show association with small RNAs and are also expressed at lower levels (30-fold to 60fold less) than mRNAs. Plant NAT-lncRNAs can overlap completely (60%) or have complementary sequences in the 5' or 3' regions of mRNAs. They accumulate in a tissue-specific manner and many are modulated in respond to biotic or abiotic stresses, suggesting fine-tuning regulatory roles Wang et al., 2015; Ben Amor et al., 2009; Xin et al., 2011) . NAT-lncRNAs also accumulate under specific environmental conditions such as light exposure, and their expression overlaps with accumulation of histone acetylation marks, suggesting some transcription regulation effect . In plants there are other types of lncRNAs. Such is the case of the lncRNAs involved in: 1) the RNA-dependent DNA methylation silencing pathway (RdDM); and 2) the lncRNAs generated from PHASloci which serve as precursors of 21-nt and 24-nt secondary phased phasiRNAs in many plant genomes (Zhai et al., 2015; Fei et al., 2013; Zheng et al., 2015) . However, the function of most plant lncRNAs is still unknown (Yamada, 2017; Shafiq et al., 2016; Ariel et al., 2015) . The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Many lncRNAs in mammals and yeast originate from specific genomic locations such as the regions around transcription start sites (TSSs), enhancer regions, intron splicing sites, and transcription termination sites. Most of the studied lncRNAs are expressed around the TSS, including exosomesensitive yeast cryptic unstable transcripts (CUTs) and stable un-annotated transcripts (SUTs) (Xu et al., 2009) , and upstream antisense RNAs (uaRNAs) (Flynn et al., 2011) , among others. Many mammalian non-polyadenylated lncRNAs also correspond to divergently transcribed, exosome-sensitive eRNAs mapped to enhancer regions (Andersson et al., 2014) ; so far plant eRNAs have not been reported.

Although most plant lncRNAs are transcribed by RNA Pol II, there are two plant-specific RNA polymerases, Pol IV and Pol V, that also produce lncRNAs (Wierzbicki et al., 2008; Li et al., 2014; Yamada, 2017) . Like many yeast and mammalian lncRNAs, most plant lncRNAs are polyadenylated, although non-polyadenylated lncRNAs do exist (Heo and Sung, 2011; Shin and Chekanova, 2014; Kim and Sung, 2017; Andersson et al., 2014) . In fact, hundreds of non-polyadenylated lncRNAs induced by specific abiotic stresses were identified in Arabidopsis .

The expression of lncRNA is regulated at transcriptional level and in combination with the pathways involved in their biogenesis, 3' end processing and degradation. The main 3'-5' exoribonuclease complex may regulate lncRNA levels since various groups of polyadenylated ncRNAs were originally identified in Arabidopsis exosome mutants (Chekanova et al., 2007) . RNA exosome is an evolutionary conserved cellular RNA processing/degradation complex (Lange and Gagliardi, 2011). Some of these lncRNAs originate from the TSSs of protein-coding genes, resembling CUTs, or they overlap with the 5' ends of protein-coding transcripts and extend into the first intron (Chekanova et al., 2007) .

Gene expression can be regulated by lncRNAs (either in cis or in trans) by sequence complementarity or homology with other RNAs or DNA, and/or by their structure; lncRNAs can thus form specific scaffolds or platforms for assembly of specific complexes (Chekanova, 2015) . Most lncRNAs regulate transcription. In animals this can be achieved by: 1) modulation of transcription factor DNA-binding activity; 2) control of RNA Pol II pausing; or 3) recruitment of chromatin modellers, which will ultimately affect chromatin topology and nuclear organization (Bonasio and Shiekhattar, 2014) . In plants, there are two lncRNAs (HID1, APOLO) that associate with chromatin, promote loop formation, and modulate transcription ), (Ariel et al., 2015 (Figure 3 ). Some lncRNAs act at the post-transcriptional and translational level (Chekanova, 2015; Jabnoune et al., 2013) . Specifically, lncRNAs can act as "decoys" or mimics by blocking certain RNAs and/or DNAs to access their protein regulators (e.g. IPS1) (Franco-Zorrilla et al., 2007; Huang et al., 2011) . Bioinformatics analyses have also identified other putative miRNA target mimics in animals that act similarly to plant miRNA sponges (Chekanova, 2015) . Another plant lncRNA (ASCO) also acts as a decoy of nuclear speckle RNA binding proteins leading to different alternative splicing events and consequent altered root development (Bardou et al., 2014) . Other lncRNAs with biological function are the enhancerRNAs (eRNAs), described in yeast and mammals, but still unknown in plants. These eRNAs can act in cis as scaffolds to recruit co-activators and thus promote chromosome looping between enhancer and promoter regions; they can interact with other lncRNAs to form specific chromosome structures to control gene expression. eRNAs are regulated by the exosome which can affect either RNA synthesis or degradation in these regions (Pefanis et al. 2015) . The exosome can also protect eRNAs from genomic instability by resolving R loops, which are stable RNA-DNA triplexes naturally formed during transcription, which can persist in regions that have divergent transcription, eventually becoming deleterious (Skourti-Stathaki and Proudfoot, 2014 The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

action of eRNAs and exosome can control gene expression and nuclear organization in these enhancer regions.

Interactions between lncRNA and chromatin modifiers can be either dependent or independent of sRNAs. In animals they are sRNA-independent and occur via the Trithorax complex (H3K4 trimethylation) ; or PRC2 (POLYCOMB REPRESSIVE COMPLEX 2, that regulates H3K27 methylation) (Tsai et al., 2010) . In plants they are -sRNA-dependent, with the RNA-dependent DNA methylation pathway (RdDM) (Matzke and Mosher, 2014 ) (see below).Further details on specific aspects are provided below.

The RdDM pathway seems to be a plant-specific pathway that relies on lncRNAs transcribed from Pol IV that will produce 24 nt siRNAs. In parallel, lncRNAs produced by Pol V act as a scaffold that can then recruit the siRNA-Ago complex by sequence complementarity (Matzke and Mosher, 2014) . However, Pol V-dependent lncRNAs are difficult to identify probably due to their low accumulation rates. Nevertheless, the few identified lncRNA are non-polyadenylated and can be either tri-phosphorylated or capped at their 5' end (Wierzbicki et al., 2008) . In addition, Pol V also seems to collaborate with Pol II to promote RdDM (Zheng et al., 2009) . Another group of regulatory ncRNAs that are Pol IV -RDR2 dependent were identified at intergenic regions overlapping mostly with transposons or sequence repeats. These ncRNAs are also non-polyadenylated and at their 5' end have a monophosphate instead of 5' tri-phosphate or cap (Zheng et al., 2009) . The Arabidopsis exosome is involved in metabolism or processing of these lncRNAs generated by Pol IV, V and also some from Pol II (Chekanova, 2015) . Also worth mentioning is that the exosome is constituted by different subunits that are functionally diverse and can affect metabolism of smRNAs and DNA methylation (Chekanova, 2015) .

lncRNAs and regulation of flowering Some of the best characterized plant lncRNAs are those involved in regulation of flowering upon "vernalisation" (the need of plants to experience a period of cold -winter-to enable flowering in spring). Therefore, transgenic plants expressing any of these lncRNAs may have biotechnological value. To assess the future relevance of these lncRNAs it is important to understand the molecular mechanisms behind their function, since they are quite different from those of sRNAs (siRNAs and miRNAs).The best studied lncRNAs in this process are: 1) COLDAIR (Heo and Sung, 2011) ; 2) COLDWRAP (Kim and Sung, 2017); 3) COOLAIR (Swiezewski et al., 2009) and ASL (Shin and Chekanova, 2014) . All these lncRNAs modulate the FLClocus, which encodes a transcription factor able to repress flowering (Berry and Dean, 2015) . Interestingly, these four lncRNAs originate from different regions within the FLClocus (promoter, first intron, and 3' end antisense) suggesting a function in cis. They all contribute to prevent FLC transcription by recruiting the PRC2 silencing complex and the accumulation of repressive chromatin marks (Crevillén and Dean, 2011; Csorba et al., 2014; Rosa et al., 2016; Sun et al., 2013sza; Swiezewski et al., 2009; Liu, F et al., 2007) . Of interest is that certain aspects of this regulation are similar to that of the mammalian lncRNAs HOTAIR and Xist. It has been proposed that some of these lncRNAs could directly bind to PRC2 components, but findings showing that mammalian PRC2 can bind very strongly to unrelated RNAs (Davidovich et al., 2015) question the need for specific lncRNAs in this recruitment step.

The study of components of the Arabidopsis exosome (RRP6L1 and RRP6L2) revealed their role in COOLAIR and ASL expression or processing, although this seemed to occur independently of the exosome core complex. Both RRP6Ls could process the 3' end of COOLAIR and promote ASL accumulation, similarly to Xist in humans (Shin and Chekanova, 2014; Ciaudo et al., 2006 The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

regulate the levels of different LncRNAs and probably act in different mechanisms to silence FLC (Chekanova, 2015) .

lncRNAs and R-loop formation COOLAIR transcript levels are decreased due to R-loop formation at its promoter (Sun et al., 2013b) , which could promote exosome recruitment through the 3' end processing pathway (Chekanova, 2015) . This function is also seen in mammals where the exosome also resolves/degrades deleterious R loops (Pefanis et al.) . Surprisingly, in mutants affected in R loop formation, COOLAIR may accumulate together with FLC, suggesting that this regulation is still not fully understood (Chekanova, 2015) .

lncRNAs and nuclear architecture Animal lncRNAs are involved in tethering RNA, DNA and proteins, and thus affect nuclear 3D structure (Engreitz et al., 2013; Quinodoz and Guttman, 2014; Hacisuleyman et al., 2014) . In plants, extensive evidence supports a similar role, either within the RdDM pathway (Moissiard et al., 2012) , in the regulation of flowering time (Hepworth and Dean, 2015) or in auxin signalling and root development (Ariel et al., 2014; Ariel et al., 2015) . Plant and animal lncRNAs can be intergenic, intronic or natural antisense transcripts, depending on their location in the genome.

Although only a few lncRNAs have been characterized in detail, they are known to have different modes of action, ranging from chromatin modifications including PRC2 recruitment, to promotion of translation, miRNA target mimicry, hijacking splicing factors and formation of chromatin loops. Adapted from (Ariel et al., 2015) . Not all processes occur simultaneously. Different types of lncRNAs are shown schematically. circRNAs Earlier studies of RNAs structure proved that viroids (existing as uncoated RNA molecules and are known to infect plants) have circular RNA (circRNA) molecules. The nature of covalently closed circular RNA molecules was determined due to: i) the inability to phosphorylate at the 5'-terminus; ii) resistance to metaperiodate oxidation or borohydride reduction of the 3'-terminal ribose; or iii) resistance to venom phosphodiesterase degradation (Sanger et al., 1976) The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

al., 1990), plants (Daros and Flores, 1996) , and mammals (Nigro et al., 1991) including humans Memczak et al., 2013) , and across the eukaryotic spectrum. Circular RNAs' lack of coding potential was identified early on by the inability of circular mRNA (Kozak, 1979) or RNA (Konarska et al., 1981) to assemble/adhere to eukaryotic ribosomes. Circular RNAs are also unable to be translated in plants extracts .

As with other ncRNAs, circRNAs in plants exhibit tissue-specific expression, have a considerable number of isoforms, alternative backsplicing (canonical and noncanonical) and alternative circularization patterns (Lu et al., 2015; Sun et al., 2016; Wang et al., 2016; Darbani et al., 2016) . The biogenesis of circRNAs is a conserved feature in animal and plant cells. Transcribed by RNA polymerase II and backsplicing reactions of pre-messenger RNAs, they occur in the spliceosome and require a repeated sequence and RNA-binding proteins. Spliceosome formation is initiated by sequential assembly of small nuclear ribonucleoproteins onto a specific region of the pre-mRNA downstream of the 5' donor splice site (dinucleotide GT) and upstream of the 3' acceptor site (dnucleotide AG). Thus, exonic or intronic circRNAs are generated depending on the initial small nuclear ribonucleoprotein binding site (Lee et al., 2017; Sun et al., 2016) .

Although their mechanisms of action in plants are still unclear, some studies in rice suggest that they do not imply any significant enrichment (as occurring in humans) for miRNA target sites, and that circular RNA and its linear form may act as a negative regulator of its parental gene (Lu et al., 2015) .

Other studies have shown that fluctuations in circRNAs do not correlate with the levels of their parentalloci encoded linear transcripts (Darbani et al., 2016) . Some circRNAs have also been shown to contain putative miRNA binding sites Lu et al., 2015) and have been identified as miRNA sponges (Zuo et al., 2016) . By binding to miRNAs, and consequently repress their function, circRNAs act as miRNA sponge to regulate the response to stress .

Recent data suggest that circRNAs in plants may affect plant response to abiotic and biotic stress. For example, circRNAs may be involved in chilling injury in tomato (Zuo et al., 2016) , or may respond to imbalances in iron and zinc (Darbani et al., 2016) . In another study, Tan and colleagues (Tan et al., 2017) showed that overexpression of a tomato circRNA derived from phytoene synthase 1 (PSY1), reduced PSY1 mRNA abundance, and lycopene and β-carotene accumulation. This was likely due to the continuous highly expressed circRNA and/or the low abundance of linear RNA from the overexpression vector. Similar results were reported for another circRNA derived from the phytoene desaturase gene (Tan et al., 2017) . Their role in developmental processes has also been established (Cheng et al., 2017) , including their possible role in different aspects of development and senescence in Arabidopsis Cheng et al., 2017) . Other biological processes such as photosynthesis (Dou et al., 2017) or mitochondrion function (Darbani et al., 2016) have also been proposed as involving circRNAs.

circRNAs are a popular topic in animal research because of their potential as post-transcriptional regulators (Memczak et al., 2013; Piwecka et al., 2017; Guarnerio et al., 2016; Hansen et al., 2013; Ashwal-Fluss et al., 2014) , their recognized function as miRNA sponges, as sponges for RNA-binding proteins or their competition with linear splicing; and their role as diagnostic markers (Zhao et al., 2017a; Zhao et al., 2017b) . Indeed, circRNAs have also been found in biological fluids including saliva (Bahn et al., 2015) , seminal fluid (Dong et al., 2016) and plasma . Research in plants is just emerging and functional studies are still lacking. It is unknown if this type of novel ncRNAs could resist gastrointestinal tract conditions, due to their expected high stability, and have a biological effect on an organism. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

ncRNAs can regulate many biological processes through their impact on gene expression by acting on chromatin structure and affecting transcription, RNA processing, RNA stability and translation. Among these, sRNAs are produced by cleavage of dsRNAs intermediates, either from hairpin precursors (miRNAs) or from the synthesis of dsRNAs by RDRs (siRNAs); most lncRNAs regulate gene expression without being processed. Despite size and biogenesis differences, most ncRNAs share sequence-specific inhibitory functions. Important differences are present between plants and other organisms in sRNAdirected silencing pathways, which are highly diversified and specialized in plants through the expansion of RDR, DCL and AGO proteins that mediate sRNAs biogenesis and function. The vast majority of small regulatory RNAs in humans and animals are miRNAs. In contrast to plants, no evidence for systemic spreading of RNA silencing in humans and vertebrates have been found. Although very few plant lncRNAs have been studied in detail, they seem to share with their animals' counterparts the ability to recruit chromatin modifiers and thus regulate gene expression. In the same context, very few studies are available on plant circRNAs. Their functional characterization will bring novel insights into their possible role as a novel class of noncoding regulators.

When considering the potential presence of plant ncRNAs in food or feed it is important to determine the mobility potential of these molecules inside the source plant (Figure 4 ). There are several comprehensive reviews (Chitwood and Timmermans, 2010; Dunoyer et al., 2013) addressing siRNA/miRNA movement in detail.

It has been hypothesiezed that sRNAs may move throughout the plant. Although the initial model would suggest non-cell autonomous (one whose action extends beyong the cell producing the signal) RNA silencing by sRNAs, other intermediates could also account for this signal: dsRNAs produced by RDR activity (plant specific), or fold back RNA acting as silencing trigger (Dunoyer et al., 2007; Dunoyer et al., 2005; Himber et al., 2003 and Smith et al., 2007) . These intermediates could also diffuse from one cell to another and be processed again by DCL4, which has been associated with 21nt siRNA movement (Dunoyer et al., 2005) . In detail, secondary siRNAs processed from dsRNA by DCL4 could move cell-tocell to propagate silencing by signal amplification. It has been shown that DCL4 expression from phloem companion cells is critical to the short-term spread of the silencing effect of a siRNA duplex. This was reported in a study in which the viral suppressor P19 was found to be expressed in phloem companion cells and able to bind to 21nt small dsRNA (Vargason et al., 2003) . DCL4-dependent siRNA movement has also been shown to occur within certain organs, such as leaves, to create gradients (Chitwood et al., 2009; Levine et al., 2007) . Other DCLs-dependent siRNAs are involved in siRNA movement and these can also be sorted into different RISC complexes (Montgomery et al., 2008; Mi et al., 2008) . Other examples of mobile siRNAs could be the ta-siRNAs and DCL4-dependent 21nt siRNAs, also acting through AGO1. DCL3 processing of long dsRNA substrates generated by Pol IV (a plant-specific RNA polymerase) and RDR2 (another plant-specific protein) leads to production of 24nt siRNAs that are incorporated into AGO4 and will transcriptionally silence transposons (Chapman and Carrington, 2007; Mi et al., 2008) . In fact, siRNAs derived from transposons or methylated regions in the genome have been shown to be associated with chromatin silencing. These 24-nt mobile siRNAs could then promote epigenetic changes transmissible to following generations. When transgenes are expressed in plants, the generated dsRNAs can be processed by DCL3 and DCL4. Another DCL, DCL2, processes dsRNAs into 22nt siRNAs which can induce gene silencing of viral origin or from transgene expression (Deleris et al., 2006) . siRNAs may also function as mobile signals able to promote epigenetic modifications.

Grafting experiments with roots from dcl2dcl3dcl4 mutants and shoots expressing GFP siRNAs could The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

indicate that siRNAs 21 to 24 nt long move throughout the plant, but still unclear if in free or proteinbound form (i.e. bound to AGO or other proteins) (Molnar et al., 2010) . Within the plant, siRNAs are known to move in a source-to-sink direction and into growing meristems (Molnar et al., 2010; Palauqui et al., 1997; Schwab et al., 2006) . Their accumulation in reproductive cells could promote epigenetic changes in subsequent generations. In fact, Slotkin et al. (Slotkin et al., 2009) showed that DCL3dependent siRNAs generated by transposon activation in the pollen vegetative nucleus can silence transposons in pollen sperm cells. These cells are then able to transmit genetic material to the following generation. A similar effect may occur if maternally-derived siRNAs in the endosperm moved to the embryo and influence genome integrity or even lead to the creation of new epialleles (Molnar et al., 2010) . Similar findings of siRNAs silencing transposable elements in the animal germ line have also been reported in an animal cell line and Tetrahymena (Malone and Hannon, 2009 ).

Considering that plant development relies more on positional effects than cell lineage, the role of miRNAs in the control of several developmental processes has to be strictly regulated at the mobility level (Chitwood et al., 2009; Levine et al., 2007) . For example, as small RNAs diffuse into regions of low mRNA (targets) expression, it eliminates target molecules therein, but cannot affect regions of high mRNA levels (Levine et al., 2007) . As mentioned previously, DCL1 processes imperfect hairpins into 21nt miRNAs, with a 5' U nucleotide incorporated into AGO1, and cleavage the mRNA targets. AGO10 can also incorporate these miRNAs (Brodersen et al., 2008) . While siRNA movement is better understood, this is not the case for miRNA movement. miRNAs can be isolated from phloem sap (miR395, miR398, miR399) but it is unclear if they can leave the phloem (Buhtz et al., 2008; Pant et al., 2008; Yoo et al., 2004) . Some reports support both endogenous miRNA and artificial miRNA movement (Chitwood et al., 2009; Schwab et al., 2006) . It has been shown that miRNAs can move from below the shoot apical meristem into the meristematic layers (Chitwood et al., 2009) , and, similarly to siRNA movement, from the phloem towards the root meristems (Molnar et al., 2010) .

Although initial evidence suggested siRNA mobility, with only DCL4 responsible for this, it is now accepted that almost every known RNAi pathway in plants has non-cell autonomous activity. Movement has been shown for sRNAs 1) of viral origin; 2) induced by transgene expression; 3) ta-siRNAs; 4) miRNAs and 5) repeat-associated siRNAs (Dunoyer et al., 2005; Carlsbecker et al., 2010; Chitwood et al., 2009; Molnar et al., 2010; Slotkin et al., 2009) . However, it is still unclear if sRNAs can move in a free state or associated with specific proteins, and in a single or double strand form.

Whereas some reports suggest that the sRNA molecule movement happens in a duplex form, the strand bias found in the sRNAs present in dcl2dcl3dcl4 grafted roots suggests that single strand siRNAs can also diffuse (Molnar et al., 2010) . Unlike in animals, in which siRNAs can move from cell-to-cell through secretory vesicles (Ruvkun, 2008; Valadi et al., 2007) , in plants the size of small RNA would allow cellto-cell movement through plasmodesmata channels connecting different plant cells (Kobayashi and Zambryski, 2007) . This movement seems to occur in a source-to-sink direction, and could be regulated in some tissues through the formation of siRNAs-AGO complexes (Molnar et al., 2010; Schwach et al., 2005) .

www.efsa.europa.eu/publications 42 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. Plant small RNAs (sRNAs) can move either cell-to-cell or long distance. Cell-to-cell movement occurs through plasmodesmata (green arrows) that allow spreading of sRNAs from the cell where they are generated to 10-15 neighbouring cells (A). This cellto-cell movement can be extended over the surrounding cells by signal amplification (B). In this case, sRNA targets will be converted into new dsRNAs by the combined action of DCLs and RDRs. Several grafting experiments have demonstrated that 21-24nt siRNAs can move over long distances in the plant (C). Whether they move as ssRNA, dsRNA or associated with proteins is still unclear. Adapted from (Dunoyer, P. et al., 2013) , (Melnyk et al., 2011) and (Molnar et al., 2010) . In terms of miRNA mobility, it is still unclear if all miRNAs, or only a subset, can diffuse. Different mobility could be associated with different tissues/structures, and/or the size and stability of the passenger strand miRNA (a.k.a. miRNA*, star strand). In addition, miRNA movement regulation could involve subcellular compartmentalization (by nuclear export proteins such as HASTY (Park et al., 2005) ) and sequestration via AGO proteins (e.g. miR165/6 can be sequestered by AGO10 and affect shoot apical meristem differentiation (Liu et al., 2009; Tucker et al., 2008) ).

Most of the information is available for sRNAs since they represent the most studied class of ncRNA to date. In addition to the role of sRNA movement when considering ingestion of transgenic plants expressing altered levels of sRNAs, it is also important to address the issue of their biogenesis and turnover to assess their temporal availability when ingested. This is especially important since transgenic plants expressing RNA interference constructs will accumulate the different sRNA intermediary forms (pri-miRNAs, pre-miRNAs, mature miRNAs, dRNAs, siRNAs), which may also be processed outside the plant when ingested. Several reviews published in high-impact journals (Rogers and Rogers and Chen, 2013; or scientific reports The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

view of the processes of miRNA maturation and degradation in plants. It is important to consider the homology between the miRNA biogenesis pathways in plants and animals as well as miRNA stability. This knowledge allows understanding 1) whether any aspect of animal metabolism or other biological processes could be modulated by plant miRNAs; and 2) if miRNAs can survive for long periods after ingestion.

As typical Pol II transcripts, pri-miRNAs are stabilized by addition of a 5' 7-methylguanosine cap (Xie et al., 2005) and a 3' polyadenylated tail (Jones-Rhoades and Bartel, 2004; Zhang et al., 2005) . miRNA genes are also subjected to transcription factor regulation and can be alternatively spliced. The size of these transcripts varies and their processing normally occurs in a base-to-loop direction (Rogers and . In plants, there are multiple DCL endonucleases that possess DExD/H-box RNA helicase, DUF283, PAZ, tandem RNaseIII, and dsRNA-binding domains. DCLs cleave the dsRNA precursor generating a 2-nucleotide 3' overhang (Margis et al., 2006) . Processing of the pri-miRNA requires at least two catalytic cycles to free the miRNA:miRNA* duplex. In animals, this requires sequential cleavage of the pri-miRNA in the nucleus first and then in the cytoplasm by different RNase III enzymes. The biogenesis of most miRNAs in Arabidopsis requires DCL1 (Park et al., 2002; Reinhart et al., 2002) , although other DCLs may be involved, but are not essential (Xie et al., 2004; Gasciolli et al., 2005) . In contrast to animals, all plant miRNA processing steps occur in the nucleus (Papp et al., 2003) . DCL1 activity will result in mature miRNAs of 20-21 nucleotides (nt), although variations in size can occur (Rogers and Chen, 2012) . Confirming its fundamental biological relevance, dcl1 mutants are lethal (Golden et al., 2002) . Other DCL proteins that function in different sRNA biogenesis pathways might also have minor roles in miRNA biogenesis. For instance, several miRNA genes are partially processed into 23-25-nucleotide mature miRNAs species whose accumulation depends on DCL3 (Vazquez et al., 2008) . In the absence of DCL3, mature 22-nucleotide miRNAs can accumulate due to DCL2 activity. In contrast, processing of miR822, miR839, and miR869 depends primarily on DCL4, probably as a consequence of highly complementary fold-backs in these pri-miRNAs (Ben Amor et al., 2009; Rajagopalan et al., 2006) . DICER cleavage of animal pri-miRNAs is facilitated by the action of dsRNAbinding domain (dsRBD) proteins (DRBs) (Jiang et al., 2005; Parker et al., 2006) . In Arabidopsis, there are five DRBs that have been shown to bind dsRNA in vitro (Hiraguri et al., 2005) . The best studied member of this family is HYL1 (HYPONASTIC LEAVES 1) which is involved in pri-miRNA processing, but also in pri-miRNA intron splicing (Laubinger et al., 2008; Szarzynska et al., 2009) . HYL1 seems to increase the accuracy of DCL1 processing of most, but not all, pri-miRNAs, suggesting that other mechanisms may also be involved (Liu, C et al., 2012) . From the same family, DRB2 and DRB4 are also involved in pri-miRNA processing, by associating with DCL4 (Rogers and . Another protein involved in miRNA biogenesis is SERRATE (SE), which can interact with HYL1 and DCL1 (Lobbes et al., 2006; Machida et al., 2011) . Like HYL1, SE also seems to increase DCL1 pri-miRNA cleavage efficiency (Dong et al., 2008) . Of interest is that these two proteins (HYL1 and SE) can be regulated by phosphorylation. Although HYL1 phosphorylation can decrease its activity (Manavella et al., 2012) , the biological relevance of this modification in SE is not known (Rogers and . Another regulator, DAWDLE (DDL), a phosphothreonine-binding forkhead associated (FHA) domain protein, seems to stabilize pri-miRNAs, possibly by direct binding (Yu et al., 2008) . Like SE and HYL1, DDL can also interact with DCL1, but does not seem to be involved in pri-miRNA processing. DDL most likely facilitates DCL1 access or recognition of pri-miRNAs. In addition, DDL can recruit phosphorylated SE or HYL1 to the pri-miRNA. The human homologue of DDL, SNIP1, seems, among other functions, to be involved in miRNA biogenesis and interacts with Drosha (Yu et al., 2008) . DDL could therefore be an evolutionarily www.efsa.europa.eu/publications 44 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

conserved factor in miRNA biogenesis. Other examples of conserved sRNA biogenesis machinery include AGO, DICER, HEN1 and exportin 5 (Yu et al., 2008) .

Two components form the cap structure of RNA Pol II transcripts in plants: CAP-BINDING PROTEINs (CBPs) 20 and 80. There is substantial overlap among cbp20, abh1/cbp80, and se mutants in terms of:

1) the specific pri-miRNAs affected; 2) the specific mRNA splicing defects observed, and 3) a bias for accumulation of first introns (Laubinger et al., 2008) . This connection between pri-miRNA processing and pre-mRNA splicing seems to converge at SE and CBPs. In fact, splicing defects are also observed in cbp20 and abh1/cbp80 in plants. In animals, the cap and its associated cap-binding complex are essential for the correct splicing of the first intron (Lewis et al., 1996) .

The connection between pri-miRNA processing and splicing also converges in TOUGH (TGH), a G-patch domain RNA-binding protein.

The tgh mutant accumulates pri-miRNAs in vivo and has been shown to affect processing of pri-miRNAs into 21-nucleotide mature miRNAs (Ren et al., 2012) . TGH may have either a structural or a regulatory role in pri-miRNA processing. TGH binds both pri-and pre-miRNAs in vivo, but this probably occurs via association with the loop structure since TGH binds ssRNA but not dsRNA in vitro (Ren et al., 2012) . TGH paralogs have been described in metazoans (Calderon-Villalobos et al., 2005) , but their roles in miRNA biogenesis are still unknown. A human TGH paralog is reported in spliceosomal preparations (Jurica et al., 2002) . A conserved connection may exist between TGH and splicing and possibly pri-miRNA processing (Rogers and Chen, 2012).

miRNA processing seems to occur in specific subnuclear regions. HYL1, SE and DCL1 can make pairwise interactions that occur in certain subnuclear particles (Fang and Spector, 2007) . TGH can also interact with DCL1, HYL1, and SE in subnuclear foci (Ren et al., 2012) . These foci seem to be zones of functional pri-miRNA processing (Fang and Spector, 2007; Fujioka et al., 2007; Manavella et al., 2012) . Some of these components (DCL1, HYL1, SE, TGH) also co-localize with components of the splicing machinery, further strengthening the connection between the two processes (Calderon-Villalobos et al., 2005; Fang and Spector, 2007; Fujioka et al., 2007) .

Processed miRNA:miRNA* duplex will leave the nucleus to carry out its function in the cytoplasm. Several components have been implicated in this transport and function and are described below.

Nucleo-cytoplasmic transport HASTY (HST) is a member of the importin-ß family of Arabidopsis that accumulates in the nuclear periphery (Bollman et al., 2003) . HST is a paralog of human exportin-5 involved in pre-miRNA transport. HST can participate in DCL1-dependent miRNA:miRNA* duplex export (Zeng and Cullen, 2004) . But this is probably not the sole mechanism for miRNA nucleocytoplasmic transport, since passive diffusion via the nuclear pore may also occur (Jacob et al., 2007; Park et al., 2005) .

EMA1/SAD1 is another importin-ß family protein involved in miRNA function. EMA1 is an orthologue of human importin-8 which facilitates nuclear import of Ago2 and reduces its association with target mRNAs (Weinmann et al., 2009) . It is still unclear if EMA1 1) sequesters miRNAs; 2) is involved in transport of other RISC loading factors or; 3) affects loading of certain miRNAs into RISC (Rogers and . The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

After leaving the nucleus the miRNA:miRNA* duplex is loaded into AGO, a RISC component. The

Arabidopsis AGO family consists of 10 members that associate with small RNAs possessing specific 5'

nucleotides (Mi et al., 2008; Takeda et al., 2008) . In plants, the slicer activity of AGOs has been shown for AGO1, AGO2, AGO4, AGO7, and AGO10 (Rogers and . AGO1 seems to have a predominant role since it preferably associates with 5'U small RNAs (most common 5' nucleotide in miRNAs) (Mi et al., 2008; Takeda et al., 2008; Vaucheret et al., 2004) . However, other AGOs can also associate with specific miRNAs depending on miRNA length or even their accumulation in different tissues (Ebhardt et al., 2010; Rogers, Kestrel and Chen, Xuemei, 2013) , suggesting a certain flexibility in this association.

In plants, miRNA guide strands generally accumulate in higher levels than miRNA* strands, although the exact mechanism for strand selection is still not fully understood (Rogers and Chen, 2012). HYL1 (DRB) seems to be involved in this process (Eamens et al., 2009; Manavella et al., 2012) . In animals, loading of siRNAs into RISC requires the cytosolic DRB protein Loq, but loading of miRNAs into RISC requires Dicer rather than Loqs (Liu et al., 2003; Liu, X et al., 2007) , suggesting different mechanisms in plants and animals. HYL1 is predominantly nuclear, therefore its association with the miRNA:miRNA* duplex suggests that 1) HYL1 may be transported to the cytoplasm together with the processed miRNA duplex; or 2) loading may also occur in the nucleus. In addition, other RISC loading mechanisms can exist (Eamens et al., 2009 ).

In animals, siRNA passenger strand unwinding requires Ago2 slicer activity; miRNAs duplexes are unwound by a slicing-independent mechanism (Matranga et al., 2005) . In plants, AGO1 slicer activity removes siRNA passenger strand but this function is not required for miRNA passenger strand removal (Carbonell et al., 2012; Iki et al., 2010) . As occurs in animals, plant miRNAs use an alternative slicerindependent unwinding mechanism.

AGO1 loading seems to require HEAT SHOCK PROTEIN 90 (HSP90) too (Iki et al., 2010) . In addition, via HSP90, AGO1 is able to interact with SQUINT (SQN) and the phosphatase PP5 (Iki et al., 2012) . These interactions seem to regulate AGO1 loading and consequently RISC activity. AGO1 phosphorylation has been detected in Arabidopsis (de la Fuente van Bentem et al., 2008) in manner similar to human Ago2. Apparently, phosphorylated Ago2 exhibits reduced sRNA binding (Rüdel et al., 2011) . Phosphoregulation by PP5 or other proteins may modulate Arabidopsis AGO1 loading capacity.

Unlike in animals, in which miRNA biogenesis occurs in the nucleus and cytoplasm, the miRNAs in plants are generated in the nucleus in a mostly DCL-1-dependent manner. Studies confirm that several of the major components are evolutionary conserved, such as the DICER, AGO, HEN1 and exportin 5 paralogs involved in nucleocytoplasmic transport of the miRNA:miRNA* duplex. However, shared pathway components among the different species do not reflect totally identical processes, such as, loading of miRNAs into RISC which occurs differently in plants and animals. One interesting common aspect remains the pri-miRNA processing and pre-mRNA splicing which seems to occur both in plants and animals.

www.efsa.europa.eu/publications 51 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. 

Following a systematic literature search as described in section 2.2.2.1 and following the methodology described in section 2.2.1.2., a total of 46 publications were selected as relevant to the stability and turn-over of non-coding RNAs. Of these, only four analyse ncRNAs stability at a genome-wide scale, and one (Enuka et al., 2016) simultaneously addresses the stability of several circRNAs. Using these studies, the stability of small ncRNAs, lncRNAs and circRNAs can be compared. While earlier studies report the degradation of plant nucleic acids when exposed to hard conditions (i.e. cooking), six recent studies suggest that some ncRNAs (i.e. miRNAs) from plants can resist these conditions.

Turnover of ncRNA molecules in plants and mammals depends both on their stability and degradation rate and it is described that chemical modifications can increase stability of different types of sRNAs (e.g. siRNAs, miRNAs). In mammalian cells, artificially introduced 2-O-methyl groups can stabilize siRNAs without affecting their RNA interference activity (Czauderna et al., 2003) . Considerable work has been done to address the issue of miRNA modifications and how these affect their stability. Among these, 2'-O-methylation has been identified as relevant. Plant miRNAs have a naturally occurring methyl group on the 3' nucleotide ribose ( Figure 5 ). In this case, methylation does not require guide RNAs, since HEN1 (HUA ENHANCER 1) can methylate miRNA:miRNA* duplexes. This HEN1-dependent 2'-Omethylation on the 3' terminal ribose is a Mg2+ dependent methylation mechanism that will ultimately stabilize miRNAs (Abe et al., 2010; Molnar et al., 2007; Yu et al., 2005; Yang et al., 2006) .The HEN1 Mg2+-dependent methylation mechanism relies on its two dsRBDs binding the substrate duplex, and the La motif-containing domain interacting with the 5' end of the substrate strand (Huang et al., 2009 ). This methylation step occurs after DCL processing creates the miRNA:miRNA* duplex but before duplex www.efsa.europa.eu/publications 54 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

unwinding and selection of the miRNA guide strand. However, it is still not clear whether HEN1 can methylate free miRNAs or can bind HYL1, DCL1 and AGO1, although it has been shown to interact with HYL1 and DCL1. In animals, the HEN1 homolog has a different structure since it does not possess the dsRBD and La motif-containing domains. Therefore, the animal HEN1 homolog methylates singlestranded sRNAs associated with AGO or PIWI proteins (Saito et al., 2007; Horwich et al., 2007) . The degradation of ncRNA molecules in plants and mammals is mediated by 3' nucleotidyl transferases and exoribonucleases. The Arabidopsis HEN1 SUPPRESSOR1 (HESO1) belongs to the DNA Polymerase β gene family (Zhao et al., 2012; Ren et al., 2012) . In other organisms, HESO1 putative homologs are ribonucleotidyl transferases that are able to add specific nucleotides to the 3' end of different RNAs (Martin and Keller, 2007) . Arabidopsis HESO1 can also add nontemplated nucleotides to the 3' end of unmethylated miRNAs in vitro and exhibits a preference for uridine (Ren et al., 2012; Zhao et al., 2012) . Analysis of hen1 and heso1 mutants indicates that this 3' oligo uridylation triggers degradation of miRNAs (Ren et al., 2012; Zhao et al., 2012) . The Arabidopsis family of SMALL RNA DEGRADING NUCLEASEs (SDNs) of 3'-5' exonucleases has similarity to the yeast Rex exonucleases. SDN1 has 3'-5' exonuclease activity versus short ssRNAs but is not active against longer or double-stranded substrates (Ramachandran and Chen, 2008) . In plants, 3' truncated miRNAs can be modified by the addition of 3' oligonucleotide tails (see above) (Ibrahim, Fadia et al., 2010; Li, J et al., 2005; Lu et al., 2009 ). In addition, these 3' truncated miRNAs also seem to associate with AGO1. This interaction delays their degradation and allow addition of the 3' oligonucleotide tail leading to their degradation at a later stage (Zhao et al., 2012) .

Whereas 2'-O-methylation may protect against 3' oligouridylation by HESO1 (see above) and consequent targeting to degradation, this modification would not protect against SDN1. It is possible that SDN1 can promote miRNA 3' truncation and then facilitate subsequent 3' polyuridylation by HESO1. However, after 3' polyuridylation SDN1 would be unable to degrade these tailed miRNAs, suggesting the existence of other nucleases (Ramachandran and Chen, 2008) .For instance, the addition of 3' polyuridylated tails in miRNAs seems to attract exosome-mediated degradation in algae (Ibrahim, F. et al., 2010) , but in

Arabidopsis this regulation has not been addressed in detail (Rogers and Rogers, Kestrel and Chen, Xuemei, 2013) .

Another possible mechanism would include loss of AGO1 association and protection of miRNAs with long 3' tails due to HESO1 activity. In addition, 3' polyuridylation may act as a nucleases recruitment platform, thus promoting miRNA degradation. Nevertheless, 3' polyuridylation and nuclease processing seem to be interconnected events that will ultimately result in degradation of the tagged miRNAs.

The overall turn-over of RNAs is mainly described using half-life studies in both plants and animals. Most of the available reports assessing ncRNA half-life focus on the role of 3' truncation, 3' uridylation and consequent degradation (see above) as the main processes regulating miRNA turnover in plants. However, a recent report identifies the enzymes (SDN1 and SDN2 exonucleases) responsible for 3' truncation in vivo and its relationship with 3' tailing (uridylation). Also of importance, this report shows the opposite role of AGO1 and AGO10 in miR165/6 degradation by SDNs, specifically SDN1. AGO10 promotes miR165 degradation by SDN1, an unexpected role for a plant AGO which has been associated with a protecting role in miRNA stability (Yu et al., 2017) . This function of AGO10 (enhancing miR165 degradation) keeps miR165 out of the stem cell niche in the shoot apical meristem (where AGO10 is not expressed). Accumulation of ectopic miR165 in the shoot apical meristem fails to maintain the stem cell population. This kind of mechanism is thus needed to clear miR165 since this particular miRNA can move between cell layers (see miRNA movement section, above). Another report evaluates the activity of nucleotidyl transferases in 3' uridylation of miRNAs and in vitro assays indicated a fast enzymatic process (Tu et al., 2015) . Also, XRN -2 nucleases that affect miR homeostasis in C. elegans do not www.efsa.europa.eu/publications 55 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. seem to process mature plant miRs but rather the by-products of pri-miRNA processing (Ji, Lijuan and Chen, Xuemei, 2012) . In summary, there are no specific studies on miRNA half-life in plants. The reports mentioned here evaluate miRNAs levels at steady state in different mutants, for example, when either the exonucleases and/or nucleotidyl transferases involved in miRNA turnover are deleted.

The half-life of miRNAs in mammals has been addressed in studies of Dicer1 ablation in mouse embryonic fibroblasts showing that miRNAs are highly stable inside the cell. Indeed, turnover assays revealed that the average miRNA half-life is 119h (i.e. ≈5 days) (Figure 6 ). Although some miRNAs have a shorter half-life, these data generally indicate that miRNAs are much more stable (≈10x more) than messenger RNAs. (Gantier et al., 2011) . Several features have been described that might explain this high half-life for miRNAs. For instance, partial pairing of miRNA and the mRNA target site generally produces translational repression, while extensive pairing of miRNA and mRNA produce a mRNA cleavage (Pasquinelli, 2010) . In the former case the miRNA remains stable, but in the latter it degrades. Further studies in mammalian cells have also shown that Argonaute proteins stabilize mature miRNAs in a slicing-independent manner, increasing mature miRNA stability (Winter and Diederichs, 2011). For specific miRNAs, it has been found that certain RNA-binding proteins (i.e. roquin) can enhance mature miRNA-146 stability by reducing its 3' end uridylation (Srivastava et al., 2015) . While intracellular levels of miRNAs are mostly affected by cell division (Ghosh et al., 2015; Gantier et al., 2011) , miRNA activity and turnover can also be controlled by subcellular distribution of microribonucleoproteins; that is, polysome sequestration contributes to an increase of cellular miRNA levels, but without an increase in miRNA activity (Ghosh et al., 2015) . Plant small RNA duplexes can be 2'-O-methylated at their 3' terminal ribose by HEN1, loaded into AGO and protected from uridylation and degradation by 3'-5' exonucleases. AGO1-bound unprotected small RNAs, however, are uridylated by HESO1 and degraded by SDN1. Adapted from (Borges and Martienssen, 2015) and (Ji and Chen, 2012) . The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Widespread variation of miRNA decay in slow-dividing primary macrophages. Bone marrow derived macrophages from When analysing lncRNAs stability in mammals at a genome-wide scale, lncRNAs half-life generally varies over a wide time range, comparable to, although on average less than, that of mRNAs (Clark et al., 2012) . In a study of all transcripts, including mammals, it was found that the range of half-lifes for ncRNAs (all types) was similar to that of the mRNA half-life (Tani et al., 2012) . Indeed, a large number of ncRNAs transcripts have a half-life of less than 4h (Tani et al., 2012) , which agrees with the < 2h half-life for hundreds of unstable lncRNAs (Clark et al., 2012) . However, although there are lncRNAs with extreme stability (half-life > 16h) they are still much less stable than miRNAs (Gantier et al., 2011) (see above), but more stable than protein-coding RNAs (Wang, L et al., 2014) .

CircRNAs have a covalently closed loop structure with neither 5'-3' end nor a polyadenylated tail, which confers high resistance to RNA exonuclease or RNase R treatment when compared to that of their linear sequence counterparts (Memczak et al., 2013) . Indeed, earlier studies of the structure of RNAs from viroids, which were found to be circular, suggested that the nature of covalently closed circular RNA molecules confers certain properties including resistance to metaperiodate oxidation or borohydride reduction of the 3'-terminal ribose; the inability to phosphorylate at the 5'-terminus; or resistance to venom phosphodiesterase degradation (Sanger et al., 1976) .

In mammals, some studies have evaluated the stability of circRNAs compared to that of their linear isoform derived from the same host gene. For example, circRNAs were found to be less abundant and dynamic than their counterparts (Enuka et al., 2016) . Calculating the half-life of 60 circRNAs and their linear counterparts, Enuka et al., 2016 showed that the media half-life of circRNAs of mammary cells (18.8-23.7h) was at least 2.5 times longer than the median half-life of their linear counterparts (4.0-7.4h). This suggests that cirRNAs are generally more stable and static compared to linear species (Enuka et al., 2016) (Figure 7 frame A). In a stability study on some circRNAs it was found that while the associated linear transcripts exhibited a half-life of <20h, the circular RNA isoforms were highly stable, with transcript half-life exceeding 48h (Jeck et al., 2013) .

In plants, covalently circularized exogenous RNA incubated with wheat embryo cell extract in vitro have shown that the circular RNAs exhibited considerable stability compared to the linear version (Makino et The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

al. , 2006) . While linear RNA degraded within the first few minutes, its circular counterpart was stable for more than 80 min in vitro (Figure 7 frame B). By contrast, circularization dramatically reduced the translation capability of the RNA, which was restored to some extent after linearization.

Plant sRNAs are naturally methylated at their 3' nucleotide ribose. This modification depends on HEN1 and protects sRNA from degradation. Unmethylated or truncated sRNAs can be uridylated by HESO1 and thus be targeted for degradation by 3'-5' exonucleases of the SDN family. Although AGO binding to miRNAs was believed to protect them from degradation, recent reports suggest that AGO10 in particular may promote miRNA degradation by SDNs. Differently from the plant situation, animal sRNAs are not methylated.

A) MCF10A cells were metabolically labelled using 4sU, for 1 or 2 h. The RNA was then extracted, biotinylated and purified on streptavidin magnetic beads. Flow-through RNA was also collected. Next, RNA was reverse transcribed and quantified using highthroughput real time PCR (Fluidigm). The half-lives of 60 circRNAs and their corresponding linear counterparts are shown. Halflife values were calculated from two samples, labelled with 4sU for 1 or 2 h and then averaged. All data were corrected for any bias introduced due to low uridine (short length) of RNA species (see 'Materials and Methods' section). The circRNAs (red dots) and their linear counterparts (blue dots) were sorted from high to low according to the difference between their half-lifes. Error bars represent standard errors. Source: From (Enuka et al., 2016) . B) Stability of circular RNAs in the wheat germ cell-free translation system. A denaturing polyacrylamide gel separating the synthesized circular and linear RNAs with a (GAAA) 16 sequence after incubation with the wheat embryo extract for the indicated time periods. Untreated RNA is in lane N. Positions of circular, linear, and ribosomal RNAs are indicated as C, L, and rRNA, respectively, on the right. Source: From . 

Compelling evidence supports the significant contribution of sRNAs to communication between hosts and some eukaryotic pathogens, pests, parasites, or symbiotic microorganisms. (Baulcombe, 2015; (Ghag et al., 2014; Koch et al., 2013; Vega-Arreguín et al., 2014; Helber et al., 2011) . SRNAs or dsRNAs, EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

for example, can be transferred from plant to pests such as insects that eat leaves or nematodes that infect roots. In fact, transgenic plants can be created that express dsRNA homologous to essential genes of insect pests or nematodes and thus control them (Baum et al., 2007; Fairbairn et al., 2007; Mao et al., 2007) . Thus, this silencing transfer mechanism is very relevant for the food and feed risk assessment of ncRNA GMO plants, highlighting the need to evaluate the stability of ncRNAs outside the plant. Earlier studies suggested that food nucleic acid content was hydrolysed when cooked (Colling and Wolfram, 1987) . Most of ncRNAs, including sRNAs and lncRNAs have been discovered in the last 20 years, and thus new data has been added to the body of knowledge. In a study of sRNA stability under cooking conditions, plant miRNAs were detected after cooking although at significantly lower levels when compared to fresh plant tissues (see Table 19 ) (Zhang, Lin et al., 2012) . Differences were observed for miRNAs from different food/feed sources (rice, cabbage, wheat or potato), ranging from nearly undetectable values to almost 80% persistance of miRNAs after cooking. Levels of miRNAs were monitored with a stem-loop quantitative reverse transcription polymerase chain reaction (qRT-PCR) assay using the U6 snRNA for normalization, a commonly used housekeeping gene in miRNAs analyses. The same study found that most of the plant miRNAs and mammalian miRNAs could survive under acidic conditions that mimic the gastric acidic environment. Of note is that the degradation rate of mammalian miRNAs under acidic conditions was similar to that of their synthetic form (without 2'-Omethylated 3' ends), whereas plant miRNAs had a much slower degradation rate compared with their synthetic form (without 2'-O-methylated 3' ends), suggesting that methylation had a protective effect on the stability of plant miRNAs (Table 20) . Another study reported that endogenous miRNAs can be detected by qPCR analysis in rice and soybean plant materials after storage, processing and cooking (Philip et al., 2015) . For processing and cooking, soybeans were soaked in RNase-free water with 0.25% w/v NaHCO 3 overnight at 4 ºC, then separated from the soaking liquid, rinsed in fresh RNase-free water, and then boiled in RNase-free water for 80 www.efsa.europa.eu/publications 59 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. min until they became soft in texture. By constrast, a synthetic miRNA showed a significantly higher susceptibility to simulated food processing conditions as compared to plant miRNAs, as well as high molecular weight RNAs (total RNA). The synthetic miRNA used for comparisons was the cel-lin-4, which is a Caenorhabditis elegans miRNA. The synthetic miRNA was synthesized without 2'-O-methylated 3'

ends (Philip et al., 2015) . These results suggest that the methylation and the small size of plant miRNAs make them more resistant to degradation than synthetic (not chemically modified) miRNAs. In addition, the study showed significant plant miRNA stability in a simulated digestion system for 75 min prior to absorption or transport into the blood stream. However, this study provided most of the quantitative data from q-PCR analysis without normalization, and data should be considered as relative compared to the treated control in the experiments.

Studying the stability of plant sRNA in vitro, Liang and colleagues exposed total RNA (5 µg), isolated from Brassica oleracea, to freshly drawn serum from mice. In serum, a large amount of sRNAs survived after 24h of incubation at 37ºC, but after 36h only about 5% of the RNA could be detected. The authors also exposed the plant RNA molecules to fecal suspensions at 37 ºC. After 4h incubation significant amounts of sRNAs were still detectable, but after 18h RNA molecules were undetectable . Suggesting that sRNAs were more resistant to degradation in the presence of serum suspension than in fecal suspensions. These results were produced by gel electrophoresis of sRNAs, which is known to be less sensitive than qRT-PCR or RNA sequencing methods. This study also reports that, when orally ingested, plant miR172 (chosen because it is the most highly enriched plant miRNA from B. oleracea)

can pass intact through the gastrointestinal tract in mice (a maximum of 4.5% recovered from the stomach of some individuals), and can be detected in the blood, liver, spleen and kidney. Levels of miR172 were monitored by qRT-PCR using a TaqMan probe, which can distinguish differences of one nucleotide. Since only one miRNA was evaluated, the above findings cannot be generalized to other miRNAs.

Other studies have reported miRNAs to be stable in watermelon and mixed fruit juices produced by simple extraction without additives . Plant miRNAs were detected by stem-loop qRT-PCR using TaqMan probes and some were validated by Northern-blot analysis. To reduce the nonspecificity signal of qRT-PCR, no-template controls were used. Yang and colleagues studied artificial in vitro digestion systems that simulate mammalian gastric and intestinal conditions, and found that miR2911 was between 10-100 fold more stable than other miRNAs . This effect was observed for cabbage extracts when compared to either the 2'O-methylated or the non-modified form. Detection was also done by qRT-PCR. The same research group reported recently that several plant miRNAs, including miR172, were degraded during storage at 4 ºC in cabbage extracts obtained by mechanical maceration . Degradation of these miRNAs correlated with that of 26S rRNA, whereas miR2911 accumulation was increased over 600-fold after 24 hours of storage, and then gradually decreased over a span of 7 days. miRNA (miR2911) increasing in macerated tissues (rather than degrading) is uncommon for most miRNAs. Indeed, the authors indicated that, in terms of genesis and amplification, miR2911 is an atypical miRNA, which could explain the disparity in serum detection with respect to canonical plant miRNAs. Thus, miR2911 was shown to be derived from 26S rRNA and processed in a DCL1-independent manner, meaning that when 26S rRNA is degraded miR2911 is produced. This study presents the possibility that certain ncRNAs can have non canonical features for synthesis or stability, and should be evaluated individually. In another recent study, the in vitro stability of Curcuma longa miRNA 14 (Clo-miR14) was assessed. When exposed to foetal bovine serum (FBS)

for up to 48h, the miR14 exhibited high resistance .

In summary, food nucleic acid content are generally hydrolysed during processing, with some studies suggesting that miRNAs from plants may be more resistant to degradation than synthetic or animal www.efsa.europa.eu/publications 60 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. miRNAs due to endogenous modifications (2'-O-methylated 3' ends). However, these findings are not consistent with miRNA stability studies in the animal gastrointestinal tract, and further research is required.

Although there are many studies on plant ncRNAs stability (see 3.1.2.1.), these mainly focus on ncRNA stability within the plant itself. Mammalian ncRNA stability differs from that of plants because lack the natural 2'-O-methylation at the 3' ends occurring in plant ncRNAs, mostly in miRNAs (Yu et al., 2005) . Very few reports were identified in which plant ncRNA stability is investigated in a mammalian organism ) (see sections 3.2. and 3.3.) . Understanding the molecular basis that confers plant ncRNA stability is of vital importance and needs to be further evaluated to better inform on the relevance of plant ncRNA for food and feed risk assessment. For instance, modifications plant miRNAs not occurring in mammals, such as the methyl group located on the 3' nucleotide ribose, could hypothetically have an impact on plant miRNAs stability in mammalian systems due to the lack of the appropriate enzymes for their recognition and degradation, or due to increased stability to mammalian RNases . Should this be the case, plant miRNAs could manifest a much longer than expected half-life in mammalian organisms when compared to mammalian miRNA, which could increase the probability of encountering target molecules. The literature contains descriptions of possible in vitro interaction of plant miRNAs with mammalian miRNA silencing complexes which may trigger target repression Chin et al., 2016) . However, this hypothesis needs to be experimentally validated since there are gaps in the literature. Moreover, it needs to be determined if sufficient levels of ingested plant miRNAs reach a target cell to determine an effect in mammalian cells. For instance, in mammalian studies it has been suggested that the threshold for target gene regulation is between 1000 and 10000 copies of mammalian miRNA per cell . Accumulation of plant exogenous ncRNAs has not yet been reported, but recent next generation sequencing studies have reported a plethora of plant miRNAs and other RNAs in circulating in the human organism (see section 3.3). The biological significance of their presence is still unknown. It remains unknown if under certain pathological conditions (i.e. compromised intestinal permeability or renal function) exposure to these plant ncRNAs could change. All these gaps in the literature would need to be experimentally validated.

To determine the half-life of plant ncRNAs in mammalian cells would require experiments utilizing labelled plant ncRNAs, first transfected into mammalian cells (in vitro studies) and then orally administered to experimental animals (in vivo studies). The markers used for these experiments could be radioactive probes or labelling molecules of very low molecular weight, such as biotin or fluorescein, to prevent distorting the actual size of the plant ncRNA under study. Due to their small size, this is important when studying miRNAs stability.

There is evidence that certain plant ncRNAs can induce silencing in some eukaryotic pathogens, pests, parasites or symbiotic microorganisms as a defence strategy. In return, various pathogens develop mechanisms to evade this defence system, proving the existence of a two-way siRNAs traffic between pathogens and their plant hosts. Plant ncRNAs stability outside the plant and in animal systems may play a role in cross-kingdom effects but has received limited attention. The resistance of plant ncRNA to the gastrointestinal environment or processing has been partly investigated but more research would be needed. EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. 

Following the methodology described in the section 2.2.1.2 and based on a literature search as described in section 2.2.2.1.3, a total of 178 documents were selected as relevant to describing the state of the art in the field of ncRNAs used/intended for use as therapeutics in humans (pharmaceuticals/medicine area). This constitutes valuable preparatory material on the pharmacokinetics (and pharmacodynamics) of ncRNA, as required by Task 1. Of these publications at least 130 indicate that to ensure the biological activity of exogenous RNAs for therapeutic requires numerous chemical modifications and delivery methods.

Although composed mainly of only four basic building blocks, RNAs form simple to very complex structures. Its intrinsic base-pair property, enhanced by the ability of the extra hydroxyl groups in the ribose sugar to form hydrogen bonds, are the foundations of its diverse structure and functions (Lu et al., 2016) . RNA can fold into various secondary structures including stem, loop, bugle, pseudoknots, Gquadruplex and kissing hairpin. It can perform a wide range of biological functions, ranging from regulation of gene expression at various levels to catalysing chemical reactions which closely depend on the ribonucleotide chain's spatial structure. Other RNAs, such as ribozymes (Doudna and Cech, 2002) , can form tertiary structures with catalytic activity. Certain RNAs can also undergo transformation between alternative structures, depending for instance on binding to specific ligands or environmental changes, as is the case of riboswitches (Mironov et al., 2002) . Riboswitches are noncoding RNA structures located within mRNAs that bind endogenous ligands to regulate gene expression or RNA splicing events (Cheah et al., 2007) .

The siRNA mechanism of action involves dicing of endogenous dsRNAs from longer RNAs transcripts by Dicer and loading into the protein complex of the RNA-induced silencing complex (RISC). Argonaute (Ago) proteins are the catalytic core of the RISC complex in plants and animals (Tolia and Joshua-Tor, 2007) . One strand (the passenger strand) is discarded and the guide strand is paired to a complementary mRNA sequence via the RISC complex. Gene silencing can be achieved mainly through post-transcriptional gene silencing (PTGS) or transcriptional gene silencing (TGS). In PTGS mechanisms, the mRNA undergoing translation can be sequence-specific cleaved by the RISC complex and degraded when the target mRNA is perfectly complementary to the siRNA. When the interaction is partial or only limited complementarity exists, translational repression and RNA degradation occurs, which is a mechanism exerted by miRNAs.

The mechanism of action for miRNA differs somewhat. The miRNAs are endogenous encoded singlestranded RNAs of about 22 nt long that are important post-transcriptional regulators of gene expression by pairing through a sequence-specific complementary binding to the 3' UTR of the target mRNA. This is usually mediated by a sequence of 2-8 nucleotides, known as the seed region, at the 5' end of the mature small RNA (Bartel, 2009). However, miRNAs can also bind to other mRNA sites including 5' UTR or protein coding exons (Forman and Coller, 2010) . Depending on the degree of sequence complementarity, the interaction will result in mRNA degradation and/or translational repression (Ameres et al., 2010; Bartel, 2009) , with destabilization of target mRNA being the predominant EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. mechanism to reduce protein output exerted by mammalian microRNAs (Guo et al., 2010) . Most animal miRNAs are transcribed by RNA polymerase II into a long primary transcript (pri-miRNA) containing a stem-loop structure. The pri-miRNA is then processed by a multiprotein complex containing the nuclear RNase III Drosha into a ≈70 nt long hairpin-shape precursor (pre-miRNA) and exported to the cytoplasm via an Exportin-5 and Ran-GTP-dependent mechanism. The cytoplasmic RNase III DICER further processes the pre-miRNA by removing the terminal loop to produce a small RNA duplex, containing the functional microRNA and the passenger (*) strand (miRNA/miRNA*). The duplex miRNA is then separated and the functional miRNA incorporated into Ago proteins within the RNA-induced silencing complex (RISC), guiding it to the target mRNA to induce its translational repression or mRNA destabilization (Krol et al., 2010; Bartel, 2009 ).

In 1978, Zamecnik first reported the use of an exogenous synthetic oligodeoxynucleotide complementary to the Rous sarcoma virus 35S RNA as a potent inhibitor of protein translation (Stephenson and Zamecnik, 1978) and virus replication (Zamecnik and Stephenson, 1978) . Since then, the ability to use exogenous (synthetic) agents to control gene expression has revolutionized many aspects of biological research and catalysed the development of a new promising class of exogenous molecules to treat human diseases. However, despite the obvious promise of this approach, progress has been slow because of the need to overcome myriad technical hurdles, particularly those related to the pharmacokinetics, pharmacodynamics and delivery of antisense molecules. Indeed, as most antisense-based potential therapies have not yet produced significant clinical results, very few RNAbased or DNA-based antisense drugs have been approved for clinical use. In 1998, the U.S. Food and Drug Administration (FDA) approved the first ASO-based therapeutics for the treatment of a human disease. In 1999, Fomivirsen (marketed as Vitravene) was approved for clinical use by the European Medicines Agency (EMA) for the European market for treatment of cytomegalovirus (CMV) retinitis in immunocompromised patients. Fomivirsen is a synthetic 21-mer ASO (DNA) that binds complementary to the sequence of the mRNA encoding the IE2 protein from CMV. The first RNA-based therapeutic approved for clinical use was pegaptanib (marketed as Macugen), which was approved by both the FDA (in 2004) and EMA (in 2006), for treatment of neovascular (wet) age-related macular degeneration. Pegaptanib is also the only approved single strand RNA aptamer drug, and it targets the vascular endothelial growth factor as an antagonist. In both cases, fomivirsen and pegaptanib, the route of administration is intravitreal injection, and the unique properties of this route provide benefits in terms of pharmacokinetic and immunological response. In 2013, the FDA (but not the EMA) approved the second RNA-based therapeutic, mipomersen (marketed as Kynamro) for treatment of homozygous familial hypercholesterolemia. Mipomersen is a DNA-RNA ASO that binds to the mRNA of apolipoprotein B (apoB), thus reducing apoB protein and concomitantly LDL cholesterol levels. Antisense oligos like fomivirsen (DNA) and mipomersen (DNA-RNA) act by binding to a complementary sequence of a mRNA, thus preventing production of its encoded protein. Nusinersen (marketed as Spinraza) approved by the FDA (in 2016) and EMA (in 2017) as an orphan-drug for treatment of spinal muscular atrophy is a RNAbased ASO that increases the production of the full-length survival motor neuron 2 (Smn2) protein.

Eteplirsen ( The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

(Bobbin and Rossi, 2016) and many others, which suggests that exogenous RNAs can be delivered to humans and used to modify gene function, protein accumulation and treat human diseases.

The ability of nucleic acids, both DNA and RNA, to form duplexes by base complementarity, has been used to develop oligonucleotide-based drugs for gene silencing. In principle, an oligonucleotide binds a target RNA through Watson-Crick base pairing and exerts gene silencing through different mechanisms (Kole et al., 2012) . A single strand antisense oligonucleotide (ASO) binds to the mRNA, which is also single-stranded, and is translated into proteins. Different molecular mechanisms to either block gene expression or degrade the RNA duplex formed by Watson-Crick base pairing have been described.

Cleavage of the RNA strand of DNA•RNA hybrids is predominantly mediated by the enzyme RNase H (Walder and Walder, 1988) , which is an abundant enzyme in both the nucleus and cytoplasm of eukaryotic cells (Wu et al., 2004; ten Asbroek et al., 2002) . ASOs can also bind RNA and block ("steric blockers") gene expression rather than facilitating RNA cleavage (Dias et al., 1999) . Also considered to be within the ASO family are peptide nucleic acids (PNAs) or phosphorodiamidate morpholino oligomers (PMOs or "morpholinos"). Their modification usually acts through this mechanism (Michel et al., 2003) , which in some cases exerts high efficacy in vivo (Iversen et al., 2003; Kole et al., 2012) . Exogenous RNA molecules or ASOs can alternatively target and bind a splicing site on the target pre-mRNA and thus modify its exon content (Mayeda et al., 1990; Skordis et al., 2003) , which results in the production of alternative splicing products with potential applications in exon skipping therapy (Goemans et al., 2011) . Because many of these ASOs contain DNA with multiple modifications, they are not discussed here in detail. The present literature review focuses mainly on ncRNA and particularly on exogenous ncRNAs.

The discovery of RNA interference (RNAi) (Fire et al., 1998) increased the interest in using exogenous chemically-synthesised RNAs for silencing genes in mammals, and promoted their potential usage in human therapeutics (Song et al., 2003) . To achieve therapeutic results, RNAi can be administered via delivery to the cell of small exogenous RNA duplexes -including short interfering RNAs (siRNAs) (Zuckerman and Davis, 2015), miRNAs mimics (Bouchie, 2013), short hairpin RNAs and Dicer substrate RNAs (dsiRNAs) (Foster et al., 2012; Kim et al., 2005) -to the cell for further processing into an RNA interfering silencing complex. Although double-stranded RNAs have been widely used for siRNA therapy, single-stranded siRNA (ss-siRNA) is capable of activating the RNAi pathway and exerting RNAi effects (Holen et al., 2003; Lima et al., 2012; Prakash et al., 2015) . Moreover, singlestranded miRNA mimics have also been shown to activate the miRNA pathway (Chorn et al., 2012) . In principle, then, both dsRNAs and ss-siRNAs can trigger gene silencing of complementary messenger RNA sequences (Holen et al., 2003) . Other small RNAs with imperfect matches have also been described to repress mRNA translation (Saxena et al., 2003) .

For miRNA therapeutics, different pharmacological tools have been developed which involved either inhibiting or enhancing miRNA function (van Rooij and Olson, 2012; Janssen et al., 2013) . Approaches to inhibiting miRNA function include small-molecule inhibitors; miRNA masking; miRNA sponges; and ASO, such as anti-miRs, locked-nucleic acids (LNA), or cholesterol-modified antagomiRs (Krutzfeldt et al., 2005) . The latter's complementarity allows them to bind to miRNAs, inducing duplex formation or miRNA degradation. The LNA therapeutic approach successfully led to the first miRNA-based clinical trials for the treatment of hepatitis C (Miravirsen; Santaris Pharma, Denmark) with promising results (Janssen et al., 2013) . There are also approaches that can be utilized to enhance miRNAs function called "mimic". These include small-molecule activators of miRNA expression and miRNA mimics, which are www.efsa.europa.eu/publications 66 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. exogenous miRNAs delivered by several methods aiming to repress the function of their endogenous targets. They are also referred to in the literature as "miRNA replacement therapy". As proof of concept, a synthetic version of miR-34a (MRX34, Mirna Therapeutics), delivered using a liposomal delivery formulation, was the first miRNA to enter a clinical trial for cancer (Beg et al., 2016) .

The literature also describes other types of RNAs that can exert physiological functions similar to that of miRNAs and siRNAs. For example, a 30-nt long guide hairpin RNA (ghRNA), that might not require Ago2 or Dicer, has been shown to exert miRNA or siRNA activity in vivo (Ohno et al., 2016) . This novel RNA interference technology may act as a novel platform for RNA-based therapy. Indeed, systemic or local injection of the ghR-form miRNA-34a (ghR-34a) suppressed tumour growth in a mouse model of RAS-induced lung cancer.

Understanding the manifold mechanisms of RNA function requires detailed knowledge of the RNA tertiary structure (Magnus et al., 2014) , which can lead to therapeutic development (DiGiusto et al., 2010; Strobel et al., 2015) . Indeed, different experimental (Lu et al., 2016; Weinreb et al., 2016) and computational methods (Magnus et al., 2014) have been developed to predict RNAs structure and interactions. The capacity of RNA to fold in various ways can generate unique three dimensional secondary structures capable of specific molecular recognition of their target cognate (Zhou and Rossi, 2016; Tuerk and Gold, 1990) . Aptamers are short single-strand RNA molecules (20-100 nt) with a defined structure that can specifically bind to a molecular target via tertiary structures. Aptamers can be rationally designed using SELEX technology (Tuerk et al., 1992) , and can therefore be incorporated into chemically modified RNAs with high nuclease resistance properties suitable for animal and clinical studies (Sullenger and Nair, 2016). Unlike other ncRNAs therapeutics, aptamers can target soluble extracellular proteins or extracellular domains of cell-surface receptors. The latter characteristic is unique and can be used to deliver siRNAs, miRNAs or other compounds to target tissues, binding them to a cell-specific aptamer (Zhou and Rossi, 2014) . Aptamer therapeutics have rapidly progressed to clinical studies, and some are already in phase 3 clinical trials (Sullenger and Nair, 2016; Zhou and Rossi, 2016) . The unique properties of aptamers led to development of the first therapeutic aptamer and the first RNA-based drug approved for clinical use, pegaptanib. However, aptamer and other ncRNAs therapeutics are not free of adverse effects (Lincoff et al., 2016) .

When compared to siRNA, dsiRNA was found to be comparable in potency and duration of effect. However, dsiRNA was found to be more immunostimulatory when compared to the shorter siRNAs (Foster et al., 2012) . Preclinical studies suggest that when administered in lipid nanoparticles (LNP) dsiRNAs can exert biological effects in different mouse models with tumours of diverse origin (Ganesh et al., 2016) . Although preliminary data in nonhuman primates suggest acceptable tolerance for this specific LNP formulation (Ganesh et al., 2016) , the dsiRNA therapeutic arena still needs to demonstrate its efficacy and safety in humans. Catalytic RNAs, the so-called ribozymes, are another family of RNAs for which therapeutics have been actively developed (Kobayashi et al., 2005) , reaching advanced clinical trials (Burnett and Rossi, 2012; Mitsuyasu et al., 2009 ), but this review will not focus on this type of RNA molecules.

Unmodified nucleic acids have been shown to possess limited stability in biological media and are subject to rapid enzymatic-mediated degradation (Braasch et al., 2004; . There are three major classes of intracellular RNA-degrading enzymes (Houseley and Tollervey, 2009) (a.k.a. ribonucleases or RNases): (i) endonucleases acting within the RNA chain, ii) 5' exonucleases and iii) 3' exonucleases degrading RNA from the 5' or 3' end respectively. Very large amounts of nuclear RNAs EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

are rapidly degraded (Brandhorst and McConkey, 1974) , this being considered either as translation noise (Struhl, 2007) or part of the biological function of ncRNAs. Instead, most cellular RNAs are modified to resist to exonuclease degradation (Ren et al., 2014) . Most genomes encode a plethora of RNases, often with overlapping activities, making redundancy a general feature of RNA degradation systems, presumably with the goal of enhancing the overall efficiency and robustness of these degradation pathways (Houseley and Tollervey, 2009). Normal physiological degradation of endogenous RNA molecules or degradation of endogenous RNA molecules with defects in processing, folding or assembly are not addressed in this literature review. Abundant RNase activity has been described in different tissues, including human body fluids (Blank et al., 1981) (Weickmann and Glitz, 1982) and with differential catalytic properties (Leimoni et al., 2003) . Different tissues contribute to body fluid RNases (Neuwelt et al., 1978) , supporting the cellular defence system against, for instance, viruses (Barrangou et al, 2007 )(Gupta et al., 2013 . The mammalian ribonuclease A family seems to contribute to this host defence by exerting antimicrobial activity (Harder and Schroder, 2002; Dyer and Rosenberg, 2006) . RNase 7, for instance, is a RNase A superfamily member with potent ribonuclease activity that is secreted by epithelial tissues including skin, respiratory tract, genitourinary tract, and the gut (Harder and Schroder, 2002) . The abundant RNase A produced in humans very probably functions to reduce RNA contamination, whether endogenous or exogenous, preventing entry into unwanted RNA-processing pathways (Houseley and Tollervey, 2009).

As above described, unmodified RNAs (naked) are generally unstable in biological systems, due to the large amount of ubiquitously expressed nucleases. Thus, several chemical modifications have been tested for exogenous RNA developed for in vivo therapies to increase resistance to nucleases, enhance binding affinity, facilitate cellular uptake, improve the pharmacokinetic and pharmacodynamics profiles, and reduce immunological response or toxicity (Wan and Seth, 2016; Bennett et al., 2017) .

There are several sites of an exogenous ncRNA molecule susceptible to chemical modification ( Figure  8 ) without interfering with its ability for base-paring or enhancing its drug-like properties. The several sites on RNAs that can be modified include the base, the sugars, and the backbone. This allows them to conjugate with a wide variety of molecules.

Different chemical modifications have been tested in the context of developing RNA-based therapeutics aimed at increasing the pharmacological properties of exogenous RNAs and allowing their successful use in different pathologies/targeted therapies. The following section briefly reviews the types of RNA modifications included in different studies in the available literature.

Nucleobase modifications can affect base-paring properties, binding affinity or specificity for the target mRNA, as well as exert adverse effects due to competition with natural nucleotide pools or interfere with polymerases (Wan and Seth, 2016). Replacement of adenosine in the siRNA guide strand with adenosine analogues has been found to modify immunomodulatory activity. Watson-Crick face-localized N-ethylpiperidine triazole modification has been found to be less reactive than the Hoogsteen edge modification, and the 5' end was more effective at blocking cytokine production than those placed at the 3' end (Valenzuela et al., 2015) . C-5 substitution in pyrimidines in the RNA guide strand increase thermal stability of siRNA duplexes. Indeed, smaller 5-methyl group substitutions resulted in better siRNA activity (Terrazas and Kool, 2009). Major groove modifications have also been found to increase the serum stability of siRNAs (Terrazas and Kool, 2009). 7-position purine modifications, where the major groove edge (i.e. the Hoogsteen face) is modified, have also been tested. Thiazole modifications of guide strand siRNA at position 15 is well tolerated in siRNAs, but 7-ethynyl at this location reduces EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

potency (Ibarra-Soza et al., 2012) . Several other nucleobase modifications have been evaluated to modify thermal stability of the siRNA duplex, hydrogen bonding and sterics, or off-target effects (Peacock et al., 2011) .

Although not exactly a base modification, movement of the nucleobase from C-1 to another position on the ribose (i.e. isonucleosides) in siRNAs has been evaluated (Zhang, J et al., 2012) . While maintaining binding capacity to RNA and stability toward nucleases, passenger strand modifications with isonucleoside at the 3' or 5' terminals can retain the silencing activity and minimize the passenger strand specific off-target effect (Zhang, J et al., 2012) . Chiral inversion of the natural D-forms (Spiegelmers) of RNA has also been tested in aptamers, leading to their clinical evaluation (Ludwig et al., 2017) .

Increase nuclease resistance Modify plasma protein binding Prevent rapid renal excretion Improve pharmacokinetics

Enhance RNA-binding affinity Enhance thermal stability

Increase nuclease resistance Enhance RNA-binding affinity

Enhance cellular uptake Reduce renal filtration Enhace delivery to certain tissues Modulate protein binding

Enhance specific cellular uptake Increase possibility of formulations for specific tissues/mucosal mebranes Ability to traverse biological barriers using extracellular vesicles 5´to 3d irection

Sites susceptible for modification on an RNA molecule. Exogenous RNAs for therapeutic use can be subjected to different chemical modifications without interference on their base-pairing ability and enhancing their drug-like properties. Conjugates at the 5´-end or the 3´-end may have the same objectives. 

Earlier studies showed that the 2'-OH of the RNA ribose sugar is not required for siRNA activity (Chiu and Rana, 2003) and 2'-modifications influence the sugar to adopt a 3'-endo sugar pucker due to the gauche effects between the O 4'-and 2'-substitutes, improving affinity properties (Egli et al., 2005) . Therefore, the furanose ring structure at the 2' position of RNAs has been extensively modified to EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

achieve pharmacological properties including increasing nuclease resistance, stability and efficacy (Egli et al., 2005) .

The 2'-O-methyl (2'-OMe) or 2'-O-methoxyethyl (2'-MOE) modifications increase resistance to nucleases and duplex melting temperature (Lamond and Sproat, 1993; Prakash and Bhat, 2007) , and have been widely used to develop miRNAs (Meister et al., 2004) , siRNAs (Judge et al., 2006; Choung et al., 2006) or aptamer-based (Burmeister et al., 2005) therapies. For example, 2'-OMe in siRNA has been shown to selectively protect the particularly vulnerable 5'-end of the guide strand against exonucleolytic degradation in human blood serum (Hoerter and Walter, 2007) . Moreover, 2'-OMe is a naturally occurring modification of RNA found in tRNA and other small RNAs. 2'-OMe has also been used to modulate RNA splicing ( Toxicity related to LNA modifications has also been described (Swayze et al., 2007) . 

Although sugar-modified oligonucleotides or exogenous RNAs, including siRNAs or antimiRs, are more effective RNA inhibitors than their unmodified counterparts, they are still susceptible to serum In addition, the phosphorothioate linkage promotes plasma protein binding (Srinivasan et al., 1995; Wan and Seth, 2016; Sands et al., 1994) . This increases pharmacokinetic benefits by reducing rapid renal clearance (Sands et al., 1994) and facilitating tissue delivery, binding to cell surface proteins and cellular uptake (Miller et al., 2016; Overhoff and Sczakiel, 2005 The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

binding affinity (Tm) for its target complementary nucleic acids (≈0.4ºC/linkage) (Freier and Altmann, 1997) including mRNAs and miRNAs (Krutzfeldt et al., 2007) . Thus, the use of a mixture of phosphodiester and phosphorothioate bonds may be preferred over all phosphorothioate bonds for in vivo applications (Ghosh et al., 1993; Krutzfeldt et al., 2007) . Also, the phosphorothioathe linkage confers chirality at phosphorus and might be relevant for siRNA activity (Jahns et al., 2015) .

Although initial studies have shown exogenous homologous RNA uptake by cell suspension, there was very rapid degradation after uptake (Shanmugam and Bhargava, 1966) . Duplex RNAs may be more stable than single-stranded RNAs, even in the absence of phosphorothioate modifications (Braasch et al., 2003) . Indeed, 2'-OMe hairpin-designed antimiR oligonucleotides do not seem to require phosphorothioate modification to exhibit in vitro activity (Lennox and Behlke, 2010).

Minimally-modified phosphodiester ASO has been reported to have less (Brigui et al., 2003) , similar or superior activity than its complete phosphorothioate counterparts (Uhlmann et al., 2000) , but sugar modifications in phosphodiester backbones can compensate the loss of activity (Monia et al., 1996) . Some adverse effects related to phosphorothioate modifications, compared to their phosphodiester backbone counterparts, have been described in ASO and siRNAs, including nonspecific protein binding (Brown et al., 1994) , The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

(dsRNA). Alpha-tocopherol conjugated siRNAs can be delivered to liver tissue (Nishina et al., 2008) or the brain when administered locally (Uno et al., 2011) . Other lipid conjugates of RNAs have been evaluated including bile acids, long-chain fatty acids, and medium-chain fatty acids (Murakami et al., 2015; Lorenz et al., 2004; Wolfrum et al., 2007) . Conjugation with specific peptides has also targeted siRNAs to specific tissues (Hsu and Mitragotri, 2011; Yamayoshi et al., 2009) . Another approach includes conjugation with high-molecular-weight polyethylene glycol (PEG) to overcome renal filtration. This strategy was used in RNA aptamers either approved (Macugen) or in advanced clinical trials for ocular disorders (Zhou and Rossi, 2016; Drolet et al., 2016) . The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

RNA delivery to a target tissue for appropriate and specific cellular uptake is currently a main hurdle in RNA therapeutics. Several synthetic non-viral methods for RNAs delivery have been developed for local and systemic RNA therapeutics ( Multiple studies have suggested that exosomes, naturally-released extracellular vesicles, can deliver RNAs (mRNAs, miRNAs, lncRNAs) to recipient cells (Zhang, L et al., 2015; . Due to their ability to traverse biological barriers, the use of exosomes or exosome-mimetic nanovesicles for RNA delivery is an open field for future RNA therapeutics (Lunavat et al., 2016; Zhou et al., 2016) . Indeed, preclinical studies suggest that exosome-mediated delivery of RNAs is feasible in vivo (Didiot et al., 2016; El-Andaloussi et al., 2012) . Different RNAs have been found in milk exosomes from different species, including humans Zhou, Q et al., 2012) , but whether these vesicle-protected exogenous RNAs are bioavailable is still in debate (Zempleni et al., 2017) .

It is important to note that most of the above described chemical modifications are not found in nature and have been developed generally to increase the drug-like properties of RNAs and other nucleotides.

To the best of our knowledge there is no evidence of clinical trials using unconjugated naked or unmodified exogenous RNA for human use with any administration route. EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Early studies suggested that RNAs were poor drug candidates due to their relatively high instability, relatively short half-life in vivo, immunomodulatory effects and the many hurdles to their cellular absorption caused by their negative charge. However, several improvements in stabilization chemistry have been identified in recent decades. Indeed, a number of chemical modifications and conjugation strategies have improved their RNA-binding affinity, in vivo nuclease stability and pharmacokinetic and pharmacodynamic properties. An exogenous RNA can be amenable for chemical modification without interfering with its base-pairing ability. Several parts of the exogenous RNA molecule can be chemically modified including the nucleobase, the backbone or the ribose (sugar). RNAs can also be conjugated with a variety of molecules which can exert different biological properties. In addition to these, chemical and biological approaches can be applied to deliver exogenous RNA to cells or specific tissues. Although many aspects are still to be fully understood in the field of chemical modification of exogenous RNAs for therapeutic use, several exogenous RNAs (siRNAs, aptamers, miRNAs, etc.) have entered advanced clinical trials for human use.

Adams The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

www.efsa.europa.eu/publications 77 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. 

Following a systematic literature search as described in section 2.2.1.2 and 2.2.2.1., a total of 119 papers were reviewed on the pharmacokinetics of foreign exogenous ncRNAs used/intended to be used as therapeutics (hereafter referred as exogenous ncRNAs). This section focuses on the pharmacokinetics of exogenous ncRNAs (naked, i.e. nonchemically modified or minimally chemically modified) other than those consumed orally from plants; these will be covered in chapter 3.3. Moreover, only in vivo studies, in mammals and humans, are reviewed here. Of the 119 documents included here, 15 were used for review of ncRNA pharmacokinetics in disease and other conditions, which may be relevant to risk assessment considerations. Although not specifically requested by the mandate, information on pharmacodynamics aspects of ncRNAs was considered. Pharmacodynamics is the study of a drug's effect on an organism, meaning its biological effects and other aspects of its xenobiotic action, (efficacy, potency, and toxicity) and this is considered relevant for ncRNAs (see 1.3).

Pharmacokinetics is dedicated to determining the fate of substances administered to a living organism. It describes the trajectory of a xenobiotic (in this case exogenous ncRNA) after delivery into an organism, and encompasses from the movement of administration, to its movement through the organism, to its complete elimination; in other words, absorption, distribution, metabolism and excretion. A drug's pharmacokinetics depends on an individual's physiology (patient-related factors, i.e. renal function, sex, age) or pathology (i.e. renal failure, hepatic failure), and drug chemical properties.

Pharmacokinetic relevant parameters include: Dose, the amount of drug administered at a single point in time; Dosing interval, the time between drug dose administration; C max , the peak plasma concentration of a drug post administration; T max , the time to reach C max ; Elimination half-life, the time required for the drug concentration to reach half its original value; Area under the curve (AUC), the integral of the concentration-time curve (after a single dose or in steady state); and Clearance, the volume of plasma from which a substance is completely removed per unit time. Pharmacokinetics modelling is done using noncompartmental or compartmental methods; the former estimate exposure to a drug by estimating the AUC of a concentration-time graph, and the latter estimate the concentration-time graph using different kinetic models.

Considering an organism as a homogeneous compartment (monocompartmental model) implies that a drug is distributed equally throughout the entire body (tissues and fluids). Therefore, blood plasma drug concentrations are a true reflection of drug concentration in other fluids or tissues, and drug elimination is directly proportional to its concentration in the organism. Not all tissues in an organism receive the same blood supply and different drugs have different characteristics (i.e. variation in drug passage through natural barriers, such as the brain blood barrier, due to a drug's biochemical properties), meaning other models are sometimes needed. The two-compartmental model assumes that there is a compartment where distribution is more rapid (central compartment) and another compartment www.efsa.europa.eu/publications 82 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

(peripheral compartment) consisting of organs with a lower relative blood flow, which consequently exhibits a slower drug distribution rate. Saturation of the enzymes responsible for drug metabolisation, the presence of active transmembrane transport mechanisms independent of drug plasma concentration, and many other factors not taken into account by the two-compartmental model, require the use of multi-compartmental models.

Very few studies have focused on the pharmacokinetic profile of exogenous RNAs (naked or unmodified), with most of the knowledge on the absorption, distribution, metabolism and excretion profile of ncRNAs coming from the development of antisense oligonucleotides therapeutics (Dirin and Winkler, 2013) . Scaling dosing from preclinical animal studies to humans (i.e. based on body weight or surface area) is challenging and apparently depends on the mechanism of clearance. For example, the pharmacokinetics of a naked siRNA formulated into polymer-based nanoparticles in mice, rats, monkeys and humans exhibited blood C max shortly after intravenous administration and rapid elimination across all species in correlation with the body weight (Zuckerman et al., 2014) . The pharmacokinetic profile investigated during the development of exogenous RNA therapeutics has shown to be influenced by the route of administration, as described below. A comprehensive summary of available information of in vivo studies and related pharmacokinetics of exogenous RNAs is provided in Table 21 .

Most studies on exogenous ncRNAs have focused on iv administration. Studies in mice have shown that naked unconjugated siRNA (400 µg/kg) injected intravenously disappears from peripheral blood 1 min after administration, restricting the likelihood of accumulation to peripheral organs (Gao et al., 2009 ). Similar results were found following iv administration of naked siRNA ( 123 I labelled), which distributed to the kidney and liver within the first minute, peaking in the kidney, liver and bladder within the first 5 minutes (Braasch et al., 2004) . Although distribution of siRNA ( 125 I labelled) decreased markedly after 24h, it persisted in the kidney and liver up to 72h, and lower concentrations were observed in the lung, spleen and heart (Braasch et al., 2004) . In rats, siRNAs were distributed to the kidney and excreted into the urine one hour after injection (van de Water et al., 2006) .

Exogenous miRNA iv administration has also been tested in animal models and entered human clinical trials. In this case (miRNA mimics), they are either chemically modified dsRNAs formulated in different nanoparticles (Xue et al., 2014) to preferentially target a tissue (Trang et al., 2011), or incorporated into naturally circulating extracellular vesicles (Bala et al., 2015) . For example, miRNAs miR-34 and let-7b formulated in a neutral lipid emulsion have been iv administered to mice and have shown to exert a biological effect (Kasinski et al., 2015) . The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

administration site for less than one day, while in a chitosan-formulated version it persisted for up to 7 days (Ma et al., 2014) . Peptide-modified and chemically modified siRNAs applied topically to the skin have been found to exert a local biological effect, as compared to the siRNA alone (Hsu and Mitragotri, 2011) . miRNAs formulated with transfection agents or nanocomplexes have also been injected intradermally (Srivastava et al., 2017) , or subcutaneously (Urgard et al., 2016) , and demonstrated to exert local biological effects. Intranasal administration of siRNAs in rats for brain targeting has resulted in very low levels of radioactive siRNAs in the brain or plasma when naked siRNAs were administered compared to a formulated version (Perez et al., 2012) . miRNAs administered by intranasal injection using transfection reagents reached the dorsal root ganglia and the olfactory bulb (Cheng et al., 2015) and, using another nanovector delivery system through intranasal administration (as drops), it reached the brain, exerting a local biological effect (Zhuang et al., 2016) .

Due to ncRNAs susceptibility to degradation, very few studies have focused on exogenous ncRNA administration in the GI tract. Non chemically modified naked siRNA administered by oral gavage to female mice (78 µg in total), was found intact and at high levels (analyzed by Northern blot and by quantitative PCR analysis) 1 and 5 hours after dosing in the stomach, small intestine and colon when formulated with chitosan nanoparticles (Ballarin-Gonzalez et al., 2013) . By contrast only low siRNAs levels were observed in the stomach, intestine and colon of mice 1 hour after oral administration of the non formulated naked siRNA. At this timepoint intact siRNA was almost exclusively present in the stomach while partial degradation products were observed in the proximal and distal regions of the small intestines. Further reduction of siRNA levels was observed 5 hours after administration, with traces only found in the colon (Ballarin-Gonzalez et al., 2013) .

Orally administered siRNAs, formulated using specific vehicles including thioketal nanoparticles (20 µg kg -1 d - The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

counterparts were present in all epithelial cells (Martirosyan et al., 2016) . Oral administration of exogenous exosomes containing miRNAs of bovine origin has been shown to attenuate arthritis in mice models. Bovine milk derived extracellular vesicles were administered daily by oral gavage starting at week 5 till week 15 after birth and arthritis was induced by collagen gavage. Animals receiving exosomes ameliorated the clinical condition, although no clear mechanisms for this effect have been described . Enteral (large intestine) siRNA administration for therapeutic gene silencing targeting the liver (via the lymphatic route) has also been described, but using chemically modified and formulated siRNAs (Murakami et al., 2015) . Enzyme-and pH-responsive microencapsulated nanogels are also being developed for oral siRNA delivery, as reported in recent in vitro tests (Knipe et al., 2016) .

Naked siRNA formulated in a water-in-oil microemulsion has been administered via the rectal mucosa to mice along 12 days (300-600 µg siRNA/kg, 8 administrations) and has reached the brain where exerted a biological effect through downregulation of the prion protein expression (Lehmann et al., 2014) . Prion infected mice presented an improvement of some neuropathological conditions after receiving the formulated siRNA following the rectal mucosa route of administration.

Being the eye the target for the first RNA-based therapy, siRNA intraocular administration has been investigated. Intravitreally injected naked siRNA (slightly modified, dTdT) distributed throughout the eye (vitreous, iris, retina, retinal pigment epithelium and sclera) of rabbits when administered at 2 mg/eye, and the pattern of ocular distribution was similar in male and female rabbits ( . Conjugated or formulated siRNAs have also been tested in animal models (Zhang et al., 2014; Janout et al., 2014) . Intravitreal injection of miRNAs has been described in the literature using either naked miRNAs in mice , rats (Qin et al., 2016) or using transfection reagent in rats (McArthur et al., 2011) or exosomes in rats (Mead and Tomarev, 2017) .

Some studies have evaluated ip administration of naked siRNAs (slightly modified, dTdT) in animal models (Wilcox et al., 2006; Shimizu et al., 2010) . The half-life of ip administered naked siRNA (dTdT) was found to be ≈11 times lower than the formulated version (Perepelyuk et al., 2016) . siRNA (dTdT) was also quickly excreted into the urine compared to a formulated version (Shimizu et al., 2010) . In a comparison of a formulated (complexed) and naked siRNA after ip administration in mice it was found that after 30 min the formulated version was detectable in muscle, kidney, liver and tumour tissue, while the naked siRNA was completely absent (Urban-Klein et al., 2005) .

While not a common administration option, chemically modified siRNAs formulated in pH-responsive nanoparticle complexes have been administered in the submandibular gland of mice (by retroductal injection) and found to exert RNAi effects (Arany et al., 2013) . The monitoring of the distribution of the chemically modified siRNA 3 h post-administration indicated that the formulated siRNA was internalized www.efsa.europa.eu/publications 85 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

in the cytoplasm of duct cells via the endosomal pathway. At 6 h post-administration, the signal was concentrated in the acinar compartment as well. After 24 h, ≈40% of the submandibular gland cells were positive for the siRNA. The delivery of the siRNA formulated in nanoparticles was more efficient compared to that of naked siRNA (Arany et al., 2013) . In another study, iv administration of chemically modified siRNA (2.5 mg/kg) was shown to accumulate remarkably in the submandibular gland just 30 min after injection, and had a half-life in this tissue of ≈22.7 h, but did not have a RNAi biological effect . In contrast, direct local injection of liposome-encapsulated siRNA (0.3 mg/kg) into the submandibular gland was shown to have strong gene-silencing effects, while naked siRNA injection (2 mg/kg) did not .

Few studies have focused on this administration route. Senn and colleagues compared central nervous system administration of siRNA vs. ASO (Senn et al., 2005) . Compared to phosphothioate and 2'methoxyethyl protected ASO (up to 50 µg), naked (slightly-modified, dTdT) siRNA (50 µg) administered to rats via icv was not found in any brain region, including the area around the injection site. In contrast, 3 h after ASO injection it was distributed in the thalamus and hypothalamus. The naked siRNA clearly exhibited very poor distribution and uptake in the brain compared to the ASO. Although ASO exhibited a capacity to reach different brain structures, its administration at a total dose of 150 µg over two days produced no RNAi effect. In the same context, icv administration of siRNA (up to 100 µg/rat/day) on two consecutive days did not exert RNAi effects. However, in another study, a siRNA chemically modified to allow for delivery in the absence of transfection reagents (Accell siRNA) was administered via the icv route at 5 µg/rat. The siRNA was incorporated into different types of neurons, although not glia, where it exerted a biological effect in regions like the cortex, the caudate subregion of the striatum and in the pyramidal cells of the hippocampal CA1 subregion (Nakajima et al., 2012) .

In a different study, a miRNAs mimic (chemically modified double-stranded RNAs) administered by icv (3 pmol/g body weight, mixed with cationic lipid DOTAP) was found to increase forty fold the level of this miRNA in the brain (Stary et al., 2015) . Several chemically modified miRNAs (agomirs) were also found to be active when administered via this route in rats (Huang et al., 2015; Ge et al., 2015) . A double-stranded modified pre-miRNAs was also reported as active in rats (Davis et al., 2011), but a naked miRNA (double-stranded 22-bp RNA) injected by icv (100 pmol) in this study was unable to increase its levels or exert a biological effect, suggesting that the naked miRNA had a higher susceptibility to RNase degradation (Davis et al., 2011) .

Lipid-complexed naked siRNA is reported to be absorbed by mouse epithelial and lamina propria cells of the vagina and ectocervix (Palliser et al., 2006) . However, uncomplexed and unmodified siRNAs were immediately degraded (<15 min) when exposed to genital wash fluid (Wu et al., 2009) , suggesting high RNAse activity in the cervicovaginal secretion. Also, slightly modified but still naked siRNA (dTdT) was unable to exert a clear biological effect (Zhang et al., 2006) .

In summary, different administration routes have been tested during ncRNA therapeutics development. The administration route used for exogenous ncRNA delivery clearly influences pharmacokinetic profile of chemically synthesized ncRNA. No matter the administration route, it seems that naked and unformulated RNAs rapidly degrade and exert very limited or no biological activity.

Very few studies have evaluated the distribution profile of administered naked oligonucleotides in organs and tissues or compared it to the distribution profile of their chemically modified counterparts. Using The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

imaging techniques in baboons (positron emission tomography -PET -and [ 18 F]-labelling) it was shown that naked oligonucleotides are poor drug candidates due to limited distribution into organs and tissues and very rapid elimination (Tavitian et al., 1998) . Compared to 2'-O-methyl or phosphorothioatemodified oligonucleotides, naked oligonucleotides were found in the heart, liver and kidney and, to a lower extent, in other tissues 5 min after iv injection. At that timepoint, naked oligonucleotides were already found in the bladder suggesting that naked oligonucleotides are very rapidly eliminated compared to 2'-O-methyl or phosphorothioate ones. Moreover, one hour after injection, almost all naked oligonucleotides were found in the bladder or residually in the kidney, while the phosphorothioate oligonucleotides were found largely in the liver and kidney, and the 2'-O-methyl ones were in an intermediate distribution (Tavitian et al., 1998) . Kinetics studies with radiolabelled oligonucleotides also showed a dramatic reduction of naked oligonucleotides in the plasma within the first 10 min, while 2'-O-methyl and phosphorothioate-modified ones showed a much slower reduction during the first 60 min.

Naked siRNA have been reported to distribute to the kidneys and bladder of mice 5 min after iv administration (Braasch et al., 2004; Naeye et al., 2013) . Intravenous administration of 50 mg Kg -1 partially-modified unconjugated siRNA (partial phosphorothioate backbone and 2'-O-methyl sugar modification) or cholesterol-modified siRNA in mice resulted in a broader tissue distribution (liver, heart, kidney, adipose and lung) of the cholesterol-conjugated siRNA 24 h after injection (Soutschek et al., 2004) . In another study, naked siRNA (free siRNA) iv administered (retro-orbitally) was found in the kidney 30 min after injection (Naeye et al., 2013) . Ip administration of synthetic siRNA formulated with low molecular weight polyethylenimine resulted in its presence in muscle, liver, kidney and tumour tissues 30 min after injection, while the naked counterpart was completely absent (Urban-Klein et al., 2005) . In rats, chemically modified naked siRNAs were distributed in a similar manner to those of 2'-O-Methyl or 2'-F modified siRNAs (Viel et al., 2008) , even though the circulation time of the 2'-F was increased.

In other studies with rats, iv injection of siRNA (chelated to indium-111) resulted mainly in renal distribution 1 hour after exposure (van de Water et al., 2006) . Some liver distribution of siRNAs have been also reported in mice, when using either the two phosphodiester RNA strands duplexes or one phosphodiester strand and one phosphorotioate strand siRNA duplex (Braasch et al., 2004) . Orallyadministered naked siRNAs (78 µg siRNA) were found to be present in the stomach, proximal and distal small intestine and colon tissues 1 hour after administration but at very low levels, further lowering 5 hours after administration (Ballarin-Gonzalez et al., 2013) . The same siRNA formulated with chitosan nanoparticles was found to distribute systemically and reach the liver, spleen and kidney (Ballarin-Gonzalez et al., 2013) . In an experimental model of arthritis in mice, iv administered naked siRNA (150 µg) was found to distribute to arthritic joints immediately after injection, but disappeared within 12 hours, compared to that of specific formulations (Komano et al., 2012) . Overall, these finding suggest that naked siRNA, although distributing at very low levels and with short residence times compared to other formulated counterpart versions, does distribute to the GI tract when administered orally or other tissues when administered systemically.

www.efsa.europa.eu/publications 87 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Formulation type (i.e. cationic lipids, neutral lipids, or others) can drive preferential distribution to specific tissues. As described for certain miRNAs, a neutral lipid emulsions (NLE) formulation has been shown to distribute to the lung after iv injection (20 µg) (Trang et al., 2011) . The lack of cationic lipids in the NLE seems to bypass some of the shortcomings that can be attributed to the charge, including formation of aggregates in biofluids, filtration by the liver, adherence to the endothelium or uptake by scavenging macrophages. The result is excellent distribution to lung tissues (Trang et al., 2011) . Extracellular vesicles (exosomes)-formulated synthetic miR-155 was found to distribute to liver, adipose, lung, muscle and kidney tissues 10 min after iv injection (Bala et al., 2015) . Administration of exosomeloaded miR-155 resulted in its detection by isolated hepatocytes and liver mononuclear cells, suggesting cellular uptake (Bala et al., 2015) .

In summary, naked oligonucleotides, including ncRNAs, rapidly distribute to the liver, lung and kidney after iv injection. Within the first minutes they are found in the bladder, suggesting rapid renal elimination. Chemical modifications or formulations increase their distribution to other tissues or delay clearance. Exosome-loaded RNAs have recently been shown to be an alternative tool for distribution to different tissues. ncRNAs by oral administration seem to distribute to the GI tract or, if formulated, systemically.

There are few studies which have evaluated the metabolism of exogenous RNAs administered to animals or humans, most of which only investigate their degradation. Naked siRNAs seem to be rapidly degraded when administered iv (Viel et al., 2008) , decreasing by ≈50% within the first 3 min, although certain chemical modifications (i.e. 2'F, 2'OME) can slightly increase this time. Stability of naked siRNA was reduced (<5 min) compared to that of other chemically modified or formulated siRNA when administered by iv injection in mice (Gao et al., 2009) . Indeed, incubation of naked siRNA with plasma or foetal bovine serum causes a very rapidly degradation of the genetic material (Jiang et al., 2012; Braasch et al., 2004; Urban-Klein et al., 2005) . Unmodified aptamers in general are also very rapidly degraded (≈10 min) (Zhou and Rossi, 2016) . When naked siRNA is administered as formulated (i.e. nanoparticles) it was shown that the renal filtration barrier can separate the siRNA from its carrier (Naeye et al., 2013; Zuckerman et al., 2012) . In the eye, siRNAs are degraded by endonucleases without preferences for one side of the duplex (as observed also for chemically modified siRNAs) (Beverly et al., 2006) . Early in vitro studies also showed that exogenous RNAs taken up by cells are rapidly degraded (Shanmugam and Bhargava, 1966) , and exogenous RNA injected into the blood is rapidly degraded to nucleosides, ribose phosphate and free bases (Sved, 1965).

The renal filtration barrier owns an effective size cut-off of ≈40 Å (Batsford et al., 1987; Bohrer et al., 1978) , but is also influenced by net molecular charge discrimination (Brenner et al., 1978) . This is supposed to facilitate rapid renal clearance of small molecules and free siRNA (Zuckerman et al., 2012; Naeye et al., 2013) since ASO, siRNAs and aptamers are commonly smaller than this cut-off (Huang et al., 2010) . By comparing naked siRNAs and ribose-modified siRNAs (2'-O-Methyl or 2'F) using PET scans in rats and mice, the kidney was identified as the main organ for siRNA elimination (Viel et al., 2008) . However, it is worth noting that also the liver participates in siRNA elimination. In other studies, siRNA appeared to accumulate in the bladder shortly after administration (Hatanaka et al., 2010) . This was also the case for modified siRNAs (LNA) compared to their versions conjugated with PEG administered in mice, in which their levels peaked in the bladder 5 min post injection and they were found in the urine as soon as 1 min after injection (Iversen et al., 2013) . EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Some studies have described rapid elimination of naked siRNA and renal excretion after iv injection at doses including 50 µg (Jiang et al., 2012) and 400 µg/Kg (Gao et al., 2009) in mice or rats (van de Water et al., 2006) . Other studies using dynamic imaging techniques or radiolabelled siRNAs have confirmed that most of the naked siRNA (free siRNA) administered to mice was found in the bladder after just 1 min up to 10 min, faster than the formulated counterparts (Naeye et al., 2013; Zuckerman et al., 2012) . The literature mostly addresses chemically modified RNAs (Dyke et al., 2006) in clinical trials (Zhou and Rossi, 2016) . The small size of RNA aptamers renders them susceptible to rapid clearance by renal filtration which has driven efforts to increase their size by conjugation for therapeutic use (Healy et al., 2004; Haruta et al., 2017) . Overall this indicates that naked exogenous small RNAs are rapidly cleared from systemic circulation and excreted in the urine.

Iv injection of a cocktail of four different plant-based small RNAs (MIR2911, MIR168a, MIR156a, and MIR161; 5 pmol each) showed them to be rapidly cleared from circulation (<5 min) . The exception was MIR2911, which peaked 5 min after injection compared to the other miRNAs administered at equal dosages, and then dramatically decreased at 30 min. When compared to other small RNAs, anomalous or atypical stability and genesis were proposed for this specific miRNA. EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Although scarcely described, there may be several conditions in which, theoretically, the pharmacokinetic properties of exogenous RNAs could be modified. Increased intestinal permeability has been described in different conditions including consumption of nonsteroidal anti-inflammatory drugs, celiac disease, Crohn's disease or other intestinal inflammatory disease (Oman et al., 1996; Sundqvist et al., 1980; Michielan and D'Inca, 2015) . For instance, absorption may be altered in cases where vasculature permeability of the gut is increased, as it has been observed in murine models of ageassociated inflammation (Jeong et al., 2017) . Increased intestinal permeability ("leaky gut") has been described in alcoholism, contributing to the extra-intestinal tissue damage caused by alcohol (Bjarnason et al., 1984) , including liver damage (Keshavarzian et al., 1999) . Similarly, leaky gut has been linked to liver steatosis in obese patients ( Theoretically, leaky gut may increase the possibilities of passively absorbing larger amounts of exogenous RNAs with consequent release into the circulatory system. However, very few studies are available in which siRNAs, aptamers, miRNAs or other exogenous ncRNAs have been tested under these conditions. Absorption of naked siRNA formulated into nanoparticles following intrarectal administration was tested in mice models of dextran sulphate sodium (DSS) induced colonic inflammation. Uptake of siRNA (observed mainly in epithelial cells, lamina propria lymphocytes, and cells from the mesenteric lymph nodes including dendritic cells and T cells) was increased in mice suffering from acute colitis compared to uptake (in the same cell types) under healthy conditions (Frede et al., 2016) , suggesting that inflammatory conditions may alter ncRNA absorption. Although there are no studies using naked unmodified exogenous ncRNAs, this topic deserves further attention.

Acute kidney injury or chronic kidney disease are important public health issues. The kidneys play a key role in elimination of xenobiotics and metabolism products and thus renal clearance should be considered in drug discovery and drug interaction research (Feng et al., 2010) . Very few studies have focused on the clearance of exogenous ncRNAs in pathological renal conditions. For example, plasma clearance of chemically modified siRNA administered by iv injection (10 mg/kg) was slightly reduced in renally impaired rats than in normal animals (Thompson et al., 2012) . Although siRNA concentration in residual kidney tissue of partially nephrectomised animals was slightly greater than the mean in normal animals, the differences did not result in appreciably greater siRNA distribution into other organs (Thompson et al., 2012) . In general, similar results, but using ASOs, were observed in rats after cisplatin-induced renal injury (Masarjian et al., 2004) . Also in a study of an ASO against miRNAs, it was recently reported that in end stage renal disease patients, only relatively modest increases in ASO plasma levels were observed when compared to control subjects (http://regulusrx.com/wpcontent/uploads/2016/11/2016-AASLD-RG-101-PK-and-Safety-in-ESRD-vs-Normal.pdf). Also using ASOs, distribution within the kidney was found to be altered in a mice model with chronic kidney disease (CKD) compared to controls (Donner et al., 2018) . Moreover, concentration of 2`-MOE ASOs in the liver of mice with CKD was elevated as compared to mice without CKD, suggesting that a reduction of renal function and ASO excretion can potentially influence the systemic delivery of ASOs (Donner et al., 2018).

www.efsa.europa.eu/publications 94 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Liver disease may also lead to alterations in drug pharmacokinetics, and thus the potential effects of hepatic impairment are considered in drug development (Allen et al., 2012) . Although exogenous ncRNAs have not been analysed quantitatively, one study showed that chemically modified and formulated siRNAs in lipid nanoparticles were found to be distributed and similarly accumulated in pathological or normal liver tissue from mice (Yu et al., 2012) .

Other physiological, pathological or lifestyle factors, such as aging (Cohen, 1986) and exercise activity (van Baak, 1990) , may influence the pharmacokinetics of exogenous molecules. However, no information is yet available regarding exogenous RNAs and their behaviour under these conditions.

Pharmacodynamics refers to the relationship between the concentration of a drug at the site of action and the resulting effect, including the time course and intensity of therapeutic and adverse effects. There is scarce data on the pharmacodynamics of naked unmodified exogenous ncRNAs. Some information is available on chemically modified RNAs (or nuclease-stabilized RNAs, the latter being outside the scope of this literature review). For example, in a mouse model of viral hepatitis the administration (hydrodynamic injection) of naked or chemically modified (methyl-modified) siRNAs was associated with clear dose-related effects of various RNAi constructs (Peng et al., 2005) . In this study the naked siRNAs had the highest inhibitory effects, but these rapidly declined (shorter-lasting inhibitory effects); in contrast, the methyl-modified siRNA duplexes were more stable inside cells and exerted their effect over a longer time period (Peng et al., 2005) .

Non-compartmental analysis of RNA therapeutics is used in most preclinical studies when analysing pharmacokinetic and pharmacodynamic parameters. For example, a non-compartmental model was used to estimate the properties of intravitreal injection of slightly chemically modified siRNA (dTdT) (Dejneka et al., 2008) . Also, in a miRNA mimic therapy by iv administration of a liposomal formulation, a non-compartmental model was used to evaluate different parameters (Beg et al., 2016) .

Systemic delivery of naked siRNA has generally failed to produce significant biological effects (gene silencing). No biological effects have been reported when naked siRNA (1 µg/eye) was administered by intravitreous injection in rats . Few studies report gene silencing effects following local administration, although some of the tested RNAs were slightly chemically modified (i.e. dTdT overhangs). Intrathracheal administration of naked (but slightly modified, dTdT) siRNA was effective through gene silencing effects in a mouse model of haemorrhagic shock and sepsis (Perl et al., 2005) . In non-human primates, naked (but slightly modified, dTdT) and unformulated siRNA administered intranasally was effective in severe acute respiratory syndrome SARS coronavirus infection reducing viral load and host symptoms (Li et al., 2005) .

It looks that naked ncRNAs require proper formulation to attain efficacy. Naked siRNA administered intranasally to mice was effective in reducing viral replication when complexed with lipofectamine (Fulton et al., 2009) . Naked siRNA formulated in pH-triggered nanoparticles exerted gene silencing effects when administered by iv injection in mice (3 mg/kg) (Kolli et al., 2013) . A formulation of naked siRNAs using cyclodextrin-containing polycations and transferrin, as a targeting ligand for delivery to transferrin receptor-expressing tumor cells, also showed gene silencing effects when administered by iv injection (2.5 mg/kg) in a mouse model (Hu-Lieskovan et al., 2005) . When administered orally EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

(20 µg/kg) to a mice model of thioglycollate-elicited inflammation, a formulated siRNA reached macrophages in the peritoneum, spleen, liver and lung, exerting a siRNA effect (Aouadi et al., 2009) . A liposome-formulated naked miRNA iv administered twice a week was efficacious in clinical trials (Beg et al., 2016) . A cationic liposome formulated Dicer-substrate siRNA (dsiRNA) (a solid dsiRNA core in an envelope of cationic lipids and cholesterol) was found to be efficacious in mice when administered by iv injection (3 mg/kg) (Ganesh et al., 2016) . Preliminary results within this formulation demonstrated a tolerability profile in non-human primates upon repeated administration (Ganesh et al., 2016) .

When comparing the naked siRNA and to a highly formulated siRNA in cyclodextrin-containing polycations with surface PEG and transferrin as the targeting ligand, only the formulated version (2.5 mg/kg iv) was effective in reducing the expression of the EWS-FLI1 gene and tumour engraftment, while the naked version exerted no biological effect (Hu-Lieskovan et al., 2005) . Similarly, in a mouse model of concanavalin A-induced hepatitis a reduction of hepatic injury by liver-specific induction of RNA interference was observed upon iv administration a of galactose-conjugated liposome nanoparticles (Gal-LipoNP) bearing naked siRNA, while no effects were noted following administration of the unformulated naked siRNA (Jiang et al., 2012) . In a collagen-induced arthritis mouse model and using systemic administration, naked siRNA (dTdT) was used against TNF-α, and was completely unable to repress TNF-α mRNA expression, whereas a formulated version repressed it (Komano et al., 2012) . In a mouse tumour model, intraperitoneal injection of naked siRNA exerted no RNAi effects, while its complex-formulated version did (Urban-Klein et al., 2005) . While developing a model to study siRNA pharmacodynamics, intramuscular administration (10 µg) of naked siRNA (chemically stable) in mice showed no effects without use of electroporation (Stevenson et al., 2013) .

There are several ongoing clinical trials using siRNAs, RNA aptamers, miRNAs and other noncoding RNAs intended for use in humans to treat diseases including cancer, cardiovascular disease, inflammation, ocular disease and others (Zuckerman and Davis, 2015; Li and Rana, 2014; Bobbin and Rossi, 2016; Zhou and Rossi, 2016; Sullenger and Nair, 2016) . Most of these ncRNAs are chemically modified or formulated. Adverse effects were identified in some of these studies. The delivery component of the formulation (i.e. liposomes), rather than the naked siRNA component, was primarily responsible for the adverse effects observed in different preclinical toxicity studies (Zuckerman et al., 2014) . For the miR-34a mimic, adverse effects in humans have been reported including acute renal injury, back pain and fatigue, all likely attributed to the liposomal carrier rather than the miRNA (Beg et al., 2016) . Liposomerelated toxicities are considered due to activation of the complement and have been well characterized with the most frequent symptoms observed being flushing, rash, dyspnoea, chest pain, back pain and subjective distress (Szebeni et al., 2011) . In non-human primates, repeated iv administration of escalating doses of a non chemically modified siRNA led to increased IL-6 and INF-gamma levels, as well as kidney and liver toxicity at the highest dose tested (27 mg/kg) (Heidel et al., 2007) .

In several clinical trials on RNA therapeutics, dexamethasone premedication was required to reduce infusion-related adverse effects or reactions (i.e. hypersensitivity, flushing, oedema, etc.) (Beg et al., 2016) . For some siRNAs used in the clinic, a predosing hydration protocol has been used to protect the kidneys (Zuckerman et al., 2014), thus reducing in humans the toxicities observed in animals.

There are also animal studies in which no toxic effects are reported. For instance, when highly formulated naked siRNAs were administered by iv route at 2.5 mg/kg, no apparent liver or kidney toxicity, immunogenicity-related toxicity or organ damage were detected in a murine model of metastatic

Ewing's sarcoma (Hu-Lieskovan et al., 2005) . No histological toxic effects were seen in liver, kidney, lung, heart and spleen in the same model following repeated treatment with the same formulation The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

(twice weekly for 4 weeks) (Hu-Lieskovan et al., 2005) . The same results were obtained for the naked siRNA administered at the same dose and for the same duration (Hu-Lieskovan et al., 2005) . In addition, when administered daily via oral gavage (200 µg siRNA/kg) for 20 days in mice, siRNA nanoparticles in galactose modified trimethyl-cysteine conjugates were reported to have no histological toxicity in major organs (brain, heart, kidney, lung, liver or spleen) (Han et al., 2014) .

Since there are so few studies in the current literature on the pharmacokinetics of naked/unmodified RNA molecules, this section included studies of other, minimally-modified, exogenous RNAs to compare their pharmacokinetic and pharmacodynamic properties to that of naked RNAs, when possible. The findings suggest that naked or unmodified RNAs are rapidly cleared from circulation and have been reported in the kidney and urine immediately after administration (generally intravenously). In general, no major gene silencing effects have been observed for naked RNAs when compared to chemically modified or formulated ones. This lack of efficacy has probably discouraged studies investigating the toxicity and safety of naked unmodified RNAs. While limited studies have reported minor biological effects of certain naked RNAs in animal models (very much dependent on the route of administration), the lack of studies in humans prevents scientists from understanding the kinetic profile of exogenous RNAs and developing effective therapeutics. Finally, scientific literature on toxicological or safety studies provides no evidence for concern about the systemic toxicity of naked exogenous small RNAs under normal physiological conditions.

Allen The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Following a systematic literature search as described in section 2.2.2.1. and following the methodology described in the section 2.2.1.2, a total of 39 papers were selected as relevant to reviewing the topic of RNAs uptake. Information on naked (non chemically modified) exogenous RNAs is mainly presented in this section. Of these publications, 15 support the uptake of exogenous ncRNAs. One study reports no intestinal absorption of RNA (Baintner and Toth, 1986) , and three others suggest that RNA is degraded before absorption (Loretz et al., 2006; O'Neill et al., 2011; Sonoda and Tatibana, 1978) . Uptake of plant exogenous RNAs, after oral intake is not included here, and will be reviewed in chapter 3.3; and uptake of exogenous RNAs from the pharmaceutical/medicinal area is described in section 3.1.4. Herein, uptake refers to passage of the molecule (in this case a ncRNA) through the cellular membrane as a part of the absorption process. An important feature for the possible biological effect that exogenous RNAs, including ncRNAs, may have within an organism is their cellular uptake. Because of their size and negative charges, exogenous RNAs cannot easily passively cross cell membranes (see below). However, earlier studies suggested that exogenous RNAs, either homologous (Shanmugam and Bhargava, 1966; Galand and Ledoux, 1966; Sirdeshmukh and Bhargava, 1983) or heterologous (i.e. from E. coli, yeast or others) (Natural Academy of Sciences, 1960; Herrera et al., 1970; Galand et al., 1966) , could be taken up by mammalian cells (Bensch and King, 1961) . Intact, RNA may be present inside the cell but is reported to degrade rapidly in some studies (Herrera et al., 1970; Shanmugam and Bhargava, 1966) . Extracellular ribonuclease activity seems to inhibit uptake (Shanmugam and Bhargava, 1966) or function (Niu et al., 1961) of exogenous RNAs. Uptake of exogenous RNA is apparently not due to altered membrane permeability, impaired cell viability, or ribonuclease degradation of the macromolecule during incubation experiments (Herrera et al., 1970) . Radioautographic evidences (radioactive assays) have shown the uptake of macromolecular RNAs by normal and neoplastic mouse cells, but degradation was also shown to occur prior to or after RNA uptake (Schwarz and Rieke, 1962) . Although intracellular degradation may occur, early studies also described biological actions in response to uptake of exogenous RNAs (Niu et al., 1961; Niu et al., 1962; 1960; Esposito, 1964) , suggesting that once taken up by a cell, RNA could eventually exert a biological effect. Studies on other types of cellular uptake of exogenous RNAs, such as transfer RNAs, either homologous or heterologous (depending on donor RNA and recipient cell type) have been conducted (Gallagher et al., 1972; Sakai and Cohen, 1977; Crooke et al., 1971) . However, as noted in RNA therapeutics development (see section 3.1.3.1), the large number of barriers encountered by RNAs from intake to cell uptake needs to be considered when evaluating possible exogenous RNA bioavailability (Figure 10 ).

The GI tract offers a large surface area potentially relevant for ncRNAs cellular uptake and following systemic release, since it contains a large variety of cell types at different maturational levels (e.g. stem cells in the crypts of Lieberkühn) located only a few microns from an extensive capillary network (Lozier et al., 1997) . However several barriers exist that could prevent ncRNAs uptake by mammalian cells (O'Neill et al., 2011) . For example, in a study conducted in neonatal pigs, no intestinal absorption of RNA was observed (Baintner and Toth, 1986), suggesting possible RNA degradation or an inability to cross the intestinal barrier. Gastrointestinal barriers can be distinguished in extracellular and cellular barriers and are presented below. EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Orally ingested exogenous ncRNA would have to pass different barriers in the human GI tract. For clarity, only the general barriers are shown.

GI barriers to the uptake of orally ingested RNAs.

The first environment encountered by any orally ingested RNA is the oral cavity, where lytic enzymes such as amylase or lipase are secreted in the saliva, and the pH ranges from 5.6 to 7.9. The stomach is a more acidic environment (pH 1.5-1.9) (Dressman et al., 1990) , and nucleic acids are known to be denatured and depurinated (hydrolysis of nucleic acids to release purine bases) over time in acidic gastric fluid (Loretz et al., 2006) . The gastroenteric fluid flow and peristaltic activity in the GI tract could reduce the contact time between ncRNAs and the epithelial layer, therefore diminishing the opportunities for cellular uptake, while nuclease enzymes present in the GI lumen could degrade nucleic acids before any cellular uptake (O'Neill et al., 2011) . Indeed, purines and pyrimidines in nucleic acids are believed to be absorbed mainly in the form of nucleosides (Sonoda and Tatibana, 1978) . The gut flora, predominant in the distal ileum and in the large intestine, produces a range of enzymes that could degrade ncRNAs as well.

The GI tract is lined by a viscous sticky layer of mucus entrapping foreign particles before reaching the underlying epithelium (Figure 11) . This "trapping" effect could be mediated by mucus composition, which confers a net negative charge to the mucous layer. Since nucleic acids present a net negative charge as well, this mucus could represent an electrostatic barrier difficult to overcome by RNAs, and particularly ncRNA, if not properly delivered or chemically modified. Furthermore, mucus turnover EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

(secretion, shedding and discard) happens in a relatively short time period (50-270 min) (Lehr, 1991), preventing RNAs from traversing the mucous layer and to reach the underlying epithelium. In conditions such as ulcerative colitis this mucus layer can be reduced or missing in areas of acute inflammation (Pullan et al., 1994) , which would allow nucleic acids to reach the epithelial cell layer. Another barrier is the glycocalyx, a glycoprotein and polysaccharide layer (400-500 nm thick) associated with the apical cellular membrane of enterocytes (Frey et al., 1996) . This layer prevents the access of certain viruses, bacteria and particles (such as RNAs) to the underlying plasma membrane by acting as a size-selective diffusional barrier.

The intestinal mucosa is structured as a three-layer barrier: a single layer of epithelial cells, the lamina propria ( Furthermore, the epithelial cells (enterocytes), which represent approximately 90% of the epithelium, have a short lifetime of 5-7 days, and are continuously shed and replaced (Quastler and Sherman, 1959; Jung et al., 2000) . The enterocyte apical membranes are characterized by a high number of microvilli, structures approximately 1 mm long and 50 nm wide (Johnson, 1987) . Since endocytosis occurs primarily at the base of microvilli (Jung et al., 2000) , particles with a diameter greater than 50 nm could not be efficiently endocytosed (Uduehi et al., 1999) . M-cells (cells dedicated to trans-cellular transport and associated to intestinal lymphoid follicles) are fewer in number than enterocytes, but, due to their role in the transport of foreign material from the lumen to the immune cells of the lamina propria, are characterized by a high endocytic activity. Moreover, M-cells have microfolds in their apical membrane, and are covered by a reduced glycocalyx and mucous layer. This determines a low breakdown capacity (e.g. low alkaline phosphatase activity, and a lower number of lysosomes) so intracellular degradation of ncRNAs could be spared. Conversely the capillary network underlying the M-cells is less dense and less permeable (O'Neill et al., 2011) .

Although the literature has described these cellular barriers when studying several molecules, there is little specific data (O'Neill et al., 2011) on the relevance of these to exogenous RNA molecules. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Assuming that ncRNAs could overcome the GI barriers and reach the circulatory system, they would still encounter other biological barriers before reaching the appropriate machinery to exert any effects. Significant degradation occurs of siRNA duplexes (non chemically modified) in 15 min when they are incubated in foetal calf or human sera (Haupenthal et al., 2006; Urban-Klein et al., 2005) . siRNA in human plasma is rapidly degraded, with nearly 75% degraded within 2 min (Layzer et al., 2004) . Nuclease activity in plasma and tissues is relevant to degrading RNAs, including ncRNAs. The major activity in plasma is a 3' exonuclease, but cleavage of internucleotide bonds can also take place (Juliano et al., 2009) . Encapsulation methods, which protect RNAs from degradation, include exosomes, microvesicles and apoptotic bodies, all of which are extracellular vesicles distinguished by size, biogenesis and cargos (ELA et al., 2013; van Niel et al., 2018) . Exosomes, for instance, contain various species of RNA, conferring them protection against degradation and providing them a vehicle for cellular uptake of RNAs by endocytosis (Cui et al., 2017).

The reticuloendothelial system (RES) is comprised of phagocytic cells, including circulating monocytes and tissue macrophages, whose physiological function is to clear the body of foreign pathogens, remove cellular debris generated by tissue remodelling, and clear cells that have undergone apoptosis. Additionally, the RES plays an important role in uptake and clearance of individual "free" oligonucleotides. Phagocytic cells of the RES, particularly the abundant Kupffer cells in the liver and splenic macrophages, express a number of cell surface receptors, including integrins and scavenger receptors, that are potentially involved in uptake, although the role of scavenger receptors in uptake of free oligonucleotides is somewhat controversial (Juliano, 2016). Following internalization of an oligonucleotide, the phagosome fuses with lysosomal compartments, where the contents are subjected to enzymatic degradation by proteases and hydrolases that operate efficiently in the low-pH lysosomal environment. Tissues macrophages are most abundant in the liver and spleen, and both these tissues receive high blood flow (Juliano et al., 2009) .

The capillary lumen is surrounded by a layer of endothelial cells interlinked by adherence junctions (VEcadherin containing) and by tight junctions (occluding and claudin-containing) and tightly adherent to the underlying extracellular matrix. This structure forms a barrier between blood and the parenchymal space. Molecules in the blood can be transported across the endothelial barrier by two routes. Paracellular transport occurs through the junctions between cells and is limited to molecules of approximately 6 nm diameter or less; only microRNAs would be able to transit this barrier by this method. Caveolar-mediated transcytosis carries large proteins across the endothelium within vesicles of about 70 nm. In some tissues, such as liver and spleen, there are gaps or fenestrations of 100-200 nm diameter between the endothelial cells, which allow egress of larger molecules (Juliano, 2016). Endothelial permeability is also increased in sites of inflammation and in some tumours, in which increased permeability is associated with fenestrations between the endothelial cells that result from rapid and disorganized angiogenesis in the tumour. This is known as the EPR effect (enhanced permeability and retention), although some argue that this is mostly due to poor lymph flow in the area The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. and many minutes to several hours for the elimination phase; (ii) oligonucleotides are accumulated in most tissues, particularly in kidney and liver, but not in the central nervous system (probably due to the blood brain barrier); (iii) the major route of elimination is via the kidneys; and (iv) although the most detailed studies have been performed in rodents, the pharmacokinetic behaviour of the modified oligonucleotides in humans is similar to that observed in lower animals (Juliano et al., 2009) . The pharmacokinetics and biodistribution of siRNA duplexes is similar to that of single-stranded antisense molecules, with the highest uptake in the kidney followed by the liver (Juliano et al., 2009 ). siRNA and uncharged oligonucleotides do not bind extensively to plasma proteins, and are thus cleared by the kidney much more readily than modified oligonucleotides and tend to accumulate at lower levels in tissues (Juliano, 2016) . Typically, molecules with sizes of 3-6 nm or less can be ultrafiltered by the kidney; microRNAs could therefore be rapidly excreted by the renal route, unlike long ncRNAs. In fact, siRNA molecules intravenously injected can be visualized in renal filtrate within seconds of injection (Molitoris et al., 2009) . Plasma siRNA concentrations are reported to decline by more than 90% within 30 min (relative to levels at 5 min post dose) and decline by more than 98% within 2h (Thompson et al., 2012) . Only about 1-2% of the iv administered dose is absorbed by tissues and the majority of this uptake occurs in the kidney (Thompson et al., 2012) , most of it being cleared from tissues within 30 h of administration .

To reach their target, ncRNAs must enter mammalian cells and arrive at the appropriate subcellular location (the nucleus or the cytoplasm, depending where the target molecules are located). Like other large, polar and charged biological macromolecules, ncRNAs are internalized by endocytosis and then trafficked through multiple membrane-bound intracellular compartments. The ability of ncRNAs to reach their targets depends on both cellular internalisation (uptake) and intracellular trafficking. (Figure 12 ).

There are five major classes of endocytosis: (i) the clathrin-coated pit pathway; (ii) the caveolar pathway; (iii) the noncaveolar, clathrin-independent pathways (CLIC pathways); (iv) phagocytosis (mainly taking place in "professional phagocytes" such as macrophages and granulocytes); and (v) macropinocytosis (in which internalized macromolecules are simply dissolved in the ambient medium) (Juliano et al., 2009).

Non-coding RNAS (ncRNAs) can enter cells via different endocytic pathways, including micropinocytosis. Internalized ncRNAs can traffic from early endosomes (EE) to late endosomes (LE) and to lysosomes. ncRNAs must escape endosomal organelles to reach the cytosol and nucleus and act on the target molecules. Some proteins may direct ncRNA to non-productive pathways. The micropinocytosis pathway may represent a non-productive pathway for naked ncRNAs. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

The first four classes often involve a cell surface receptor and are collectively termed receptor-mediated endocytosis. These pathways are mainly utilized in cellular delivery of oligonucleotides. In general, receptor-mediated endocytosis includes three major steps.

(i) Receptor binding and internalization represents the primary barrier for oligonucleotide transport; the ligand-receptor binding determines which target cells and tissues oligonucleotides are delivered to. (ii) Sequential intracellular trafficking leads oligonucleotides into a variety of low pH endomembrane compartments, including early/sorting endosomes, late endosomes/multivesicular bodies, and lysosomes. In some cases, receptors/ligands can traffic to the Golgi complex. In many instances, receptor and ligand are dissociated in the low pH endosome environment. Vesicular trafficking can prevent ncRNAs from reaching their targets, for example, by sorting to secretory or lysosomal vesicles which may lead to export of ncRNAs out of cells or degradation in the lysosomes. (iii) ncRNAs must exit from the endosome to reach the site of action in the cytoplasm or nucleus. Endosomal trapping represents an important barrier for ncRNAs final functional objective. On the other hand, oligonucleotides are able to continuously shuttle between the nucleus and the cytoplasm mediated by nuclear pore structures, and do not require classical nuclear localization signals (Juliano et al., 2009 ).

The available data clearly describe several barriers to be considered when evaluating exogenous RNA absorption, circulation, cellular uptake or intracellular trafficking.

Exogenous RNA must overcome numerous biological barriers to reach its intended target within mammalian cells. Considering oral administration as the most relevant for exogenous plant ncRNAs, the first major limiting step to exogenous RNA bioavailability is the GI tract itself. The GI tract encompasses both extracellular and cellular barriers. The extracellular barriers include the presence of enzymes in the lumen, such as amylase or nucleases, a harsh environment with pH ranging from 1.5 to 7.9, and a net negative charged mucous layer with a very rapid turnover (50-270 min). The cellular barriers include a single layer of epithelial cells, the lamina propria and the muscularis mucosa, constituting the threelayer intestinal mucosa barrier. ncRNAs could cross the epithelium between cells but this is limited by the presence of tight junctions. Pore size in the human intestine would prevent passage of all ncRNAs other than miRNAs, the smallest ncRNAs. ncRNAs could also cross this epithelium through transcytosis by traversing the cells; this mechanism implies exposure to intracellular nucleases, recycling of ncRNAs back to the lumen and nuclear uptake.

Many biological barriers and environmental conditions between the GI epithelium and the target tissue are then encountered by ingested RNAs. ncRNAs would be exposed to nucleases both in the plasma and in the target tissues. Once inside the circulatory system, these RNAs would be subjected to distribution and elimination, accumulating mostly in the liver and kidney. Due to their small size, miRNAs are especially rapidly cleared by the kidney, unlike long ncRNAs. Therefore, a very small percentage of RNAs could actually be absorbed by tissues. Furthermore, ncRNAs would need to escape the reticuloendothelial system, the function of which is to clear the organism of foreign molecules. Then RNAs must cross the vascular endothelial barrier to reach the target tissue. For ncRNAs to exert their functions they must enter the target cells. RNA cellular uptake is achieved by endocytosis after which they would enter the intracellular trafficking system through multiple membrane-bound compartments. ncRNAs would need to exit these compartments to reach their functional sublocation within the cell, be it the cytosol or the nucleus, while simultaneously escape degradation in the lysosomes. EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

seeds, corn kernels and rice grains, all of which are common food and feed, and numerous endogenous plant sRNAs are reported to have perfect complementarity to genes in humans, and other animals . However studies in mammals and humans with siRNA indicates that the exposure levels required to produce regulatory effects exceed the levels achieved following ingestion .

All animal and plant-related foods and feeds contain naturally occurring coding (i.e. mRNAs) and ncRNAs. Early (Srivastava, 1965; Lassek and Montag, 1990 ) and more recent studies ) have estimated that the total RNA content in plants is about 1 mg/g plant tissue. Plants contain an array of ncRNAs, including highly abundant transfer RNAs (tRNAs) and ribosomal RNAs (rRNAs), single-stranded antisense RNAs, and the miRNAs and siRNAs that trigger RNAi and their precursor dsRNAs . Long dsRNAs from non-plant, exogenous sources are particularly common in plants, including plants used for food and feed, due to infection from RNA-containing viruses. Animalderived foods are generally richer in RNA than plant-derived foods, and are likely to contribute significantly to overall RNA consumption (Jonas et al., 2001) . Of the 1 mg/g tissue of total RNA content in plants, relative percentages (by weight) of the major RNA forms are approximately 80% rRNA, 3-5% mRNA, and 10-15% tRNA, with sRNAs accounting for less than 5% of total RNA in plants . sRNAs are present at levels of up to 1.61 µg/g of conventional soybean grain (average = 0.66 µg/g grain) and comparable amounts are present in the grains of conventional corn and rice . In tobacco plants engineered to overexpress a dsRNA under the control of a constitutive promoter, siRNAs levels in leaves reached about 1.5% of total RNA (Chau, B. L. and Lee, K. A., 2007) . Total exposure to construct-derived sRNAs from a putative biotech soybean product was estimated to be 45 µg/Kg/day in the general population and 82 µg/Kg/day for children aged six and under . Using these same assumptions for RNA expression levels in sweet corn, consumption of corn at the highest possible rate would result in an estimated intake of 106 µg/Kg/day in the general population and 170 µg/Kg/day for children aged In the absence of encapsulation or chemical stabilization to prevent degradation or without the addition of penetration enhancers, the absorption of RNA -including siRNAs -across the GI tract is described as negligible (Akhtar, 2009; Jain, 2008) . One study suggests that activity could be possible for certain highly expressed plant miRNAs following food intake (Zhang et al., 2012a) . This study reports that in mice fed a diet consisting entirely of cooked rice (i.e. human equivalent of about 33 kg/day of cooked rice) , several rice miRNAs were detectable in mouse serum and liver. Some of the results presented in this study were contested in later studies (Chen et al., 2013) . Petrick et al. clearly stated the unlikelihood of achieving such high concentrations of mature plant miRNAs in the serum, plasma and organs of humans or animals via food intake since the high levels of rice administered the experimental animals did not reflect anticipated dietary exposure levels. In another study, it was concluded that plant miRNAs identified in animal sRNA sequencing data can originate from artefacts of the sequencing process (Zhang et al., 2012b) . Even in the case of 20-mer oligonucleotides modified to increase their stability, oral bioavailability in rats was only 0.1% (Nicklin et al., 1998) and intestinal absorption was less than 1% in a model that bypasses the gastric acidic environment. Most of the labelled oligonucleotide was associated with the luminal epithelial cell membrane and very little EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

was localized intracellularly (Khatsenko et al., 2000) . Regarding the possibility of dietary siRNAs and other dsRNAs effects on gastrointestinal tissues, studies with radiolabelled or immunostained DNA oligonucleotides demonstrate that these are primarily located extracellularly within the intestinal tract (lumen and luminal wall) (Khatsenko et al., 2000; Nicklin et al., 1998) , and are thus presumed to be minimally absorbed (negligible bioavailability, <2%). It has also been observed that absorption decreased as the oligonucleotide length increased (from 6 to 22 nt) (Khatsenko et al., 2000) .

The only published quantitative data on sRNAs content in plant-derived foods ) (detailed above) indicate that 0.6 µg of sRNA is obtained per gram of soybean, rice seeds or corn kernels. As total RNA content in plant-derived foods appears to be around 1 mg/g tissue (Lassek and Montag, 1990), sRNAs could therefore represent 0.1-0.2% of plant seed materials. The amount of siRNAs in plants genetically modified to carry hairpin transgenes is estimated at about 50 ng siRNA/340 µg total RNA/g tobacco leaf tissue (Chau, Bess L. and Lee, Kevin AW, 2007) . The resulting estimated human dietary exposure to plant small RNAs from RNA-based biotech crops was derived from this (Table 22 ). . (b) Assuming 100% of the specific food is consumed from a biotech crop, with total RNA levels of ≈1 mg/g grain and 1.5% of these small RNAs derived from the transgene. For details see .

These doses are significantly lower than those required to elicit adverse effects in rodents or monkeys (800 or 1000 mg/Kg of chemically-stabilized siRNA, respectively) by iv administration (Thompson et al., 2012) . Moreover, high miRNA concentration is required to target suppression, and levels below 100 copies/cell miRNAs have little regulatory capacity (Mullokandov et al., 2012) . According to Snow et al., dietary plant miRNAs are present at less than one copy per cell in target organs assuming reliable quantification . In another study the calculated dietary exposure in humans of a specific dsRNA (DvSnf7) was estimated to be around 3.8 x 10 -5 µg/kg/day in the US population (Petrick et al., 2016b) . The same study reported that 100 mg DvSnf7/kg/day in a 28-day repeated dose oral (gavage) toxicity study in mice did not produce any toxic effects (Petrick et al., 2016a) . Levels for this specific dsRNA in this biotech crop were reported to be ≈ 0.104 x 10 -3 µg/g grain tissue on a dry weight basis for corn food, and ≈ 55.1 x 10 -3 µg/g in the whole plant on a dry weight basis .

Of note is that there is little information on the impact of intended RNAi in GM plants on the coding and non-conding plant RNA. Knowledge of which other RNAs are modified, or how other RNAs are compensated or desregulated in the GM plant needs to be experimentally evaluated for each specific case.

Different studies have evaluated the amount of plant-derived foods consumed by the general population. In a study of cohort of women in the UK, the difference between the mean intake of overall fruit and vegetables consumption was three times greater in high consumers vs. low consumers (Pollard et al., 2001) ; vegans and vegetarians were the highest consumers of fruit and vegetables. In an EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

evaluation of the Flemish population it was observed that vegans, while having the lowest total energy intake compared to omnivores, have the highest total fruit and total vegetables consumption. The intake of fruit, nuts and/or legume about twice in vegans, using either the healthy eating index (Clarys et al., 2014) or total mean intake (Clarys et al., 2013) . In the EPIC cohort study, which includes ten European countries, mean daily consumption was between 269 and 103 g/day vegetables and between 453 and 121 g/day fruit (Agudo et al., 2002) . Although not homogeneous among countries and centres, the ratio between the highest and lowest consumers was around 2.5 for vegetables and 3 for fruit (Agudo et al., 2002) . Countries in southern Europe had the highest vegetables and fruit consumption. For example, consumption of vegetables and fruit in healthy adults in Spain (EPIC data), considered higher than most European countries and the USA, was 273 g (3.4 servings) of vegetables and 348 g (4.4 servings) of fruit (Agudo et al., 1999) , equal to ≈600 g/d vegetables and fruit. In the United States (US) the total vegetable and fruit intake has increased slightly (up to ≈4.5 servings/day) in all consumer categories from the late 1990s (Krebs-Smith and Kantor, 2001) and remained stable over the last decade (Rehm et al., 2016) . Actual quantities, however, are around 2.4 servings/day (Rehm et al., 2016) , that is 1.12 for fruit and 1.19 for vegetables (Eaton et al., 2013) . In cup-equivalents, this is 1.7 cup equivalents for vegetables and 1.0 cup equivalent for fruit (Moore and Thompson, 2015) . Mean total vegetable intake in the US for vegetarians is reported to be around 250 g/day and that for non-vegetarians as 197 g/day. Total fruit intake was around 261 g/day for vegetarians and 159 g/day for non-vegetarians (Haddad and Tanzman, 2003) . In other words, there is less than two-fold difference in the amount of fruit and vegetables consumed between vegetarians and non-vegetarians in the general US population. In a specific population (i.e. the Adventist Health Study) in the US, daily mean consumption of fruits was higher (≈424 g/day) for vegans than for other types of vegetarians or non-vegetarians (≈319 g/day). The same trend was observed for daily mean consumption of fruit which was higher in vegans (≈483 g/day) than other vegetarians or non-vegetarians (≈298 g/day) (Orlich et al., 2014).

Based on these data, dietary exposure to RNAs from plants can in general be assumed to be higher in vegans and vegetarians than omnivores. Although total RNA content in plant-derived foods varies (Lassek and Montag, 1990), a person consuming an estimated daily dose of total fruit and vegetables ≈600 g/day would theoretically ingest 600 mg of total RNAs assuming about 1 mg/g of plant tissue. This value agrees with the dietary RNA intake, which typically ranges from 0.1-1 g/person/day (Jain, 2008) . Considering that sRNAs represent up to 0.2% of plant seed materials ), estimated daily intake in the general population would be around 1.2 mg sRNAs. Since vegans and vegetarians normally consume higher amounts of vegetables and fruits (see above), they could be exposed to an average of three times more plant exogenous RNA (3.6 mg/day). No pharmacokinetics studies were identified in these groups.

Changes in RNAse activity can occur in physiological or pathological conditions, this possibly impacting exposure after dietary intake. For instance, increased serum RNAse activity has been observed in pancreatic carcinoma, pancreatitis, and renal failure (Peterson, 1979 The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. et al., 2011) . Human Argonaute 2 (Ago-2) is a catalytic core component of RNAi machinery. Ago-2 mediated gene silencing has been proposed to be linked to the Mitogen Activated Proteine Kinase signalling pathway (Zeng et al., 2008) , which is activated in response to cellular stress. Since various kinds of stress on human cells induce formation of stress granules, Ago-2 can be recruited to these granules, modifying its intracellular localization and the efficacy of RNAi activity (Detzer et al., 2011) . Ago-2 is thus regulated at both the transcriptional and post-translational levels, suggesting that Ago-2 levels can influence cellular miRNA activity (Adams et al., 2009 ). It has also been proposed that inactivation of mitochondria could lead to a strong decrease in miRNA-mediated RNAi efficiency, and, to a lesser extent, to siRNA-mediated RNAi (Huang et al., 2011) . Due to the interaction of P-bodies with mitochondria, reduced Ago-2 activity could be due to changes in location of endogenous Ago-2 from Pbodies (Huang et al., 2011) . Association of Ago-2 with cytoplasmic RNA granules is known to regulate the translational repression activity of the protein, but phosphorylation of certain residues may prevent this inhibition, allowing the protein to remain active in the cytoplasm (Lopez-Orozco et al., 2015) . Overall, it seems clear that Ago-2 activity can be modified by cellular stress conditions. However, whether this can influence the function or the exposure to dietary ncRNAs remains unknown and would benefit from further research.

Most studies have been focused on miRNAs since these ncRNAs were the first to be described. These studies have consequently extended to transgenic dsRNAs as these have increasingly used to induce gene silencing. As mentioned previously, little is known about dietary exposure to other ncRNAs. Given the role of lncRNAs in plant physiology, it is important to better understand human and animal exposure to these ncRNAs. Further research is needed on how transgenic ncRNAs affect expression levels of other ncRNAs, and other RNAs such as mRNAs. Alteration of the expression levels of certain ncRNAs within the plant may modify the levels of other ncRNAs due to putative compensatory circuits and codifying RNAs. This in turn could lead to changes in protein and enzymatic content, with putative consequent alteration of the modified plant's nutritional value. Controlling for these changes would require complete transcriptome sequencing, and comparison of RNA levels in unmodified to those of the altered plant . Another point to consider is whether special diets (e.g. vegetarian or vegan) modulate ncRNAs uptake throughout the GI tract, possibly leading to increased exposure.

Humans and animals have been continuously exposed to naturally-occurring plant RNAs. Estimated amounts of ingested plant ncRNAs contrast with the much higher doses administered to experimental animals and in clinical trials to humans. Although some reports on the presence of plant miRNAs in serum, plasma and organs of humans or animals do not agree, the extremely low reported concentrations prevent them from being functional, even in the case of RNAs modified artificially to augment their bioavailability. Humans following special diets, such as vegetarians and vegans, have high intake of plant RNAs, but their increased estimated exposure to such RNAs is barely on average 3fold greater than in omnivorous diets. There are no specific studies determining the effects of increased dietary exposure to plant ncRNAs.

Adams BD, Claffey KP, White BA. 2009. Argonaute-2 expression is regulated by epidermal growth factor receptor and mitogen-activated protein kinase signaling and correlates with a transformed phenotype in breast cancer cells. Endocrinology 150, 14-23. EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

To exert a biological effect, exogenous ncRNAs need to overcome the physical and biological barriers present in both the GI tract and the circulatory system, reach the target tissue, and enter the appropriate intracellular pathway at sufficient dose. This section describes the obstacles and challenges encountered by exogenous ncRNAs following oral intake. Following an extensive literature search, and based on expert judgement in the topic, 21 publications were reviewed for this section.

RNA molecules are large, hydrophilic and negatively charged. Even the smallest RNA molecules (miRNAs) have a high molecular weight (∼13 kD) and their net charge is negative (Akhtar and Benter, 2007a; Akhtar and Benter, 2007b) . Due to these physicochemical properties, RNA molecules are highly water soluble with low membrane permeability, thus falling into Class III of the biopharmaceutical classification system (FDA, 2000; Amidon et al., 1995) . Like many other Class III compounds, they are not absorbed at all or only to a very small degree. Being highly charged species, RNA molecules resist partitioning across lipid bilayers, and higher molecular weights effectively restrict their movement by function of the tight junctions present in the GI tract epithelium (Tillman et al., 2008) . In vitro and in vivo preclinical studies have assessed the use of medium chain fatty acids (6-12 carbon atoms), as well as bile salts, as absorption enhancers to cross the intestinal mucosa (Gonzalez Ferreiro et al., 2002; Tillman et al., 2008; Raoof et al., 2002) .

Orally introduced RNA molecules encounter relevant obstacles in the GI tract, these precluding their absorption and activity. In general oligonucleotides are rapidly degraded in the harsh biological milieu of the acidic stomach and enzyme-rich GI tract, as described previously, and show poor transcytosis across the gut (Akhtar, 2009) . In humans the pH in the GI lumen can vary from 1 in the stomach to 8 in the intestine (Gamboa and Leong, 2013) . Exposure to these pH values can cause pH-induced oxidation or de-amination of RNA, resulting in loss of activity (Pridgen et al., 2015) . In addition, nucleases present in the GI lumen for digestion of biological molecules enzymatically degrade RNA molecules too (Gamboa and Leong, 2013; Kriegel et al., 2013) . The presence of a rich bacterial population, mostly in the intestines (The human microbiome project consortium, 2012) may also contribute to degradation of these molecules, as bacterial degradation of ribonucleic acids through the secretion of extracellular nucleases has been early described (Nishimura, 1960; Eaves and Jeffries, 1962) .

Finally, if the above-mentioned obstacles are overcome, RNA molecules still face the intestinal extrinsic and intrinsic barriers (see also 3.1.5).

The extrinsic barrier consists of the mucus layer covering the epithelial cells. This is a complex hydrogel material composed of proteins, carbohydrates, lipids, salts, antibodies, bacteria, and cellular debris (O'Neill et al., 2011; Pridgen et al., 2015) . It consists of loosely and firmly adherent layers that vary in thickness along the GI tract and can fluctuate based on diet. The mucus is secreted by cells presenting a rapid cell turnover, approximately 2-5 days in the small intestine in humans (Gamboa and Leong, 2013; O'Neill et al., 2011; Atuma et al., 2001) . Among other functions, the mucus layer protects www.efsa.europa.eu/publications 118 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

epithelial surfaces by trapping pathogens and foreign particulates, and rapidly clearing them. Penetration of this mucus barrier is necessary to reach the absorptive epithelial cells. Several mucuspenetrating and mucoadhesive materials can be used to enable penetration of molecule to the mucus layers, increasing residence time and contact of the delivered molecules to the epithelium (Pridgen et al., 2015) .

The intrinsic barrier consists of the epithelial cell monolayer, constituting the greatest barrier to material from the intestinal lumen to the lamina propria and bloodstream. Cells maintain this barrier by forming tight junctions, whose permeability can be modulated through specific combinations of different proteins (O'Neill et al., 2011) .

There are several possible pathways across the intrinsic (epithelial) barrier (Figure 13 ).

 The transcellular pathway passes through the apical and basolateral cell membranes, as well as the cytoplasm. This pathway is very restrictive to the passive flow of hydrophilic solutes such as RNA molecules, because of the lipid bilayer membrane and its impermeability to large molecules. Transport mechanisms for this pathway can be passive for hydrophobic molecules, or active with membrane pumps for specific molecules such as ions.  The paracellular pathway is the major passive permeation pathway and allows diffusion of small molecules in the space between epithelial cells. The tight junctions regulate permeability of this pathway based on molecules' size and charge.  Finally, transcytosis is an active transport pathway that relies on molecule-specific receptors guiding the molecule through the cell without entering degradation pathways. Because of their large hydrodynamic size, macromolecules such as RNAs are restricted to this pathway (Pridgen et al., 2015) .

The M cell transcytosis pathway is the most extensively studied for oral delivery of RNAs. This pathway, which is used to transport antigens across the epithelium for immune surveillance, is attractive because M cells lack mucus secretion and have a sparse glycocalyx, among other features. However M cells are closely associated with immune cells in the lamina propria and Peyer's patches, such as dendritic cells and macrophages, this from one side limiting the ability of the delivered molecule to reach the bloodstream and on the other side increasing the risk of triggering an immune response (Florence, 2004; Kriegel et al., 2013; Pridgen et al., 2015) . Moreover, M cells only make up a small percentage (5-10%) of the non-absorptive epithelium in humans. The surface properties of M cells, as well as the number of Peyer's patches vary among species, which could make difficult extrapolating animal model data to humans (Pridgen et al., 2015) .

An additional obstacle is represented by the immune system, which is intimately associated with the epithelium. Numerous types of immune cells patrol the lamina propria, including T cells, macrophages and dendritic cells. If the ingested RNA molecules reach the bloodstream, they must also evade the mononuclear phagocytic system (Reischl and Zimmer, 2009; Pridgen et al., 2015) .

The presence of endogenous and luminal nucleases, together with GI pH conditions, in general determine rapid degradation and low RNA molecules' biological half-life (Bhavsar and Amiji, 2007) The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

The paracellular pathway allows diffusion of molecules in the space between epithelial cells and is regulated by intercellular tight junctions. The transcellular pathway passes through the apical and basolateral cell membranes and cytoplasm. It is restricted to hydrophobic molecules or molecules transported by membrane pumps. The transcytosis pathway is an active transport system that relies on molecule-specific receptors guiding the molecule through the cell escaping degradation pathways. Transcytosis pathways are found in both epithelial and M cells. From Pridgen et al, 2015.

Transepithelial transport: paracellular, transcellular and transcytosis pathways.

First-pass elimination occurs when a compound is metabolised between its site of administration and the site of sampling for drug concentration measurement. This greatly affects the bioavailability of orally administered compounds since the concentration of the active substance reaching systemic circulation is decreased. The liver is usually assumed to be the major site of first-pass metabolism of an orally administered drug, but other sites are the GI tract, blood, vascular endothelium and lungs (Pond and Tozer, 1984) . In fact, the obstacles previously described for RNA molecule absorption in the GI tract can be considered a first-pass effect.

RNA molecules reaching the circulatory system must avoid filtration by the kidneys, accumulation in non-target reticuloendothelial system (RES) of the liver, kidney, lungs and spleen, degradation by endogenous nucleases, aggregation with proteins in the serum, and uptake by non-targeted cells such as phagocytes (Kriegel et al., 2013; Reischl and Zimmer, 2009 ). Once cellular uptake has taken place in the potential target tissue (surmounting the obstacle of crossing the vascular walls and the lipid bilayer cellular membranes), nucleic acids have to undergo efficient intracellular trafficking to ultimately produce the intended pharmacological effects. This includes endosomal transportation (escaping endolysosomal degradation), cytoplasm release, and efficient incorporation into the relevant intracellular pathways such as RISC loading (for siRNA, mRNA, and miRNA) or nuclear uptake (for lncRNA) Kriegel et al., 2013; Pouton and Seymour, 2001) .

www.efsa.europa.eu/publications 120 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

RNA molecules are large, hydrophilic and negatively-charged molecules. To achieve effect via the oral route, these molecules need to overcome the physical and biological barriers present in both the GI tract, where they encounter variable and harsh conditions along with degrading nucleases, and the circulatory system. RNA molecules must reach the target tissue, escaping endo-lysosomal degradation, and enter the appropriate intracellular pathway in sufficient doses to exert their effects.

www.efsa.europa.eu/publications 121 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Following a literature search as described in section 2.2.2.2. and based on the methodology described in section 2.2.1.2, a total of 110 documents were selected as relevant to a general review of RNA-based therapeutics administered by GI route. Of these documents, at least 24 report the use of specific vehicles for GI administration of exogenous ncRNAs (i.e. nanoparticles). Six of these documents simultaneously evaluated formulated exogenous RNAs and their naked version, and reported either rapid degradation under GI conditions, or reduced or absence of biological effect of the naked version compared to the formulated one (Table 23) . Twenty (20) of these documents provide specific examples of local effects of RNA-based drugs administered orally. In addition, 18 papers were selected as relevant to inform on administration of exogenous ncRNAs in either fish or birds and possibly exerting biological effects.

Orally delivered RNA has been used as a dietary supplement or empirical therapeutic agent since 1960s (Jain, 2008). For example, yeast RNA supplementation as a source of dietary ribonucleic acids has been used in different animals (Choudhury et al., 2005; Jha et al., 2007) including mammals (Sukumar et al., 1999; Heaf and Davies, 1976) and in human studies (Gianotti et al., 2002; Tepaske et al., 2001) . In human studies, up to 4 g per day (in divided doses) of exogenous RNAs have been administered (for 8-12 days), producing an increase in plasmatic uric acid levels (Zollner and Grobner, 1971; Zollner, 1982) .

The absence of attempts to study the GI absorption of pure RNA in these studies suggests that its effects, if any, could be related to RNA molecules as a source of nucleotides. Indeed, nucleotide supplementation in rats has been shown to affect the healing of ulcerative conditions, ameliorating indomethacin-induced ileitis and aggravating the severity of dextran sulfate sodium (DSS)-induced colitis (Sukumar et al., 1997 , Sukumar et al., 1999 . Also, a study of body fluid composition of rats orally administered yeast RNA, mixtures of its constituent nucleosides or its constituent bases and ribose was performed (Heaf and Davies, 1976) . The authors showed that ingestion of RNA increased intestinal levels of ribose, inorganic phosphate, uridine, pseudouridine, uracil, inosine or uric acid. The effect of orally administered mixed nucleosides on blood and urine composition was similar to that of RNA, while some differences were noted for some equivalent mixture of free bases (Heaf and Davies, 1976) . This study showed that the dietary RNA-phosphate passed to the urine from the gut, suggesting that most of the RNA-ribose was probably metabolized (Heaf and Davies, 1976) . It has been shown in a murine model of Staphylococcus aureus infection that the oral administration of RNA was ineffective while the intraperitoneal nucleoside-nucleotide mixture was more effective in maintaining host resistance against bacterial infection (Adjei et al., 1993) . Supplementation with nucleotides has also been shown to exert biological effects, including expression regulation of certain genes (Sanchez-Pozo and Gil, 2002; Gil, 2002) , although the mechanisms of action are still unknown. It is generally assumed that nucleotides are not essential nutrients, but under certain conditions (i.e. low dietary intake, tissue needs increased or stress) dietary nucleotides may play a role as conditionally essential nutrients (Jung and Batal, 2012; Carver and Walker, 1995) . In this context, dietary supplementation with yeast RNA promoted ulcer healing in a rat model of indomethacin-induced ulcerative ileitis (Sukumar et al., 1997) . Similar effects were observed when nucleosides and nucleotides were administered intravenously (Veerabagu et al., 1996) . In both cases, the possible mechanism of action was partly due to increased cell proliferation in the damaged tissue (Sukumar et al., 1997; Veerabagu et al., 1996) . Similar effects were observed when RNA was administered orally or its constituents (nucleotides) were injected iv, suggesting that the effects are due to the absorbed nucleotides. Dietary nucleotide supplementation in formula-fed infants EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

has been shown to improve gut microbiota composition (Singhal et al., 2008) . In studies on certain fish model, supplementation with dietary RNA at a range of concentrations (0.4-0.8% of nucleic acids in the diet compared to the non-suplemented control group) enhances immunological response (Choudhury et al., 2005; Jha et al., 2007) . In human studies, postoperative enteral diet supplementation with RNA, arginine and omega-3 fatty acids has been shown to modulate the postoperative immune response after surgery for upper gastrointestinal cancer (Senkal et al., 1995) , overcoming more rapidly the immunologic depression after surgical trauma (Kemen et al., 1995) . Also in humans, oral supplementation with RNA (from yeast) during 5 days before surgery improved host defence in patients undergoing cardiac surgery (Tepaske et al., 2001) or GI cancer surgery (Gianotti et al., 2002) . In both cases, RNA supplementation was also accompanied with arginine and omega-3 fatty acids administration. In summary, the specific contribution of orally administered RNA to these effects cannot be ascertained.

The instability of orally administered RNA molecules and the presence of a large array of RNAases within the GI tract (see sections 3.1.3 and 3.2.1) suggest low absorption, if any. Exposure to a highly active enzymatic environment in the GI tract, extreme pH conditions, and the existence of a mucosal epithelial barrier are the main challenges for oral delivery of RNA therapeutics (Martirosyan et al., 2014) . Although a considerable amount of food RNA is ingested, which varies widely between individuals but is typically in the range 0.1-1 g/person/day, it is assumed it is degraded and absorbed in minimal amount (Jain, 2008) . However, further data are required to document the extent to which this conclusion pertains to RNA molecules with complex structures. To overcome the multiple barriers encountered by exogenous RNAs for oral delivery, several strategies in the laboratory or clinical trials have been developed (see 3.2.2.1 below). The oral administration route is advantageous because it increases patient compliance and comfort over injection, provides for simple, repeatable administration, and offers a large surface area for absorption (Forbes and Peppas, 2012) . RNA delivery may also benefit from advances in the oral delivery of ASO for animal (Raoof et al., 2004; Raoof et al., 2002; . Indeed, polymeric nanoparticles containing alginate, dextran, chitosan, polyethylenimine, polyethylene glycol, polylactide, yeast derived β-glucans and/or several other natural or chemically synthesized molecules have been tested by the pharmaceutical/medicine field.

The literature describes several strategies for the oral delivery of RNA oligonucleotides. The target is to stabilize and protect RNAs during GI transit, enable cellular uptake in the intestine, and in some cases, promote endosomal escape and increase biological action.

Thioketal nanoparticles (TKNs) were formulated from a polymer, poly-(1,4-phenyleneacetone dimethylene thioketal) that degrades selectively in response to reactive oxygen species ( The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

(PLGA) nanoparticles were also tested in the same study, but with no efficacy in diminishing TNFα expression.

To obtain efficient mucus transportation, targeted cellular uptake and endosomal/lysosomal escape, different nanoparticles have been tested ( Nanoparticles of galactosylated trimethyl chitosan-cysteine containing siRNA against Map4k4 were tested and found to target activated macrophages in colonic tissue (Zhang et al., 2013b) . Compared to nanoparticle-protected siRNAs, naked siRNAs were completely degraded by the intestinal fluids. When administered orally (250 µg siRNA/kg) for 6 days, nanoparticles improved the DSS-induced model of colitis in mice (Zhang et al., 2013b) .

Other ternary polymeric nanoparticles formed by thiolated trimethyl chitosan with tripolyphosphate have been tested in mice and found to efficiently deliver siRNAs to the intestine and other tissues when administered orally (Zhang et al., 2013a The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

(4%, 13% and 21%) were also evaluated in trimethyl chitosan-cysteine conjugate nanoparticles . In a mouse model of ulcerative colitis, the mannose density of 4% was the most effective for siRNA knockdown of TNFα following oral administration (30 µg/kg) of 2'-O-methyl-modified TNFαspecific siRNA duplex . The use of supramolecular self-assembly nanoparticles (SSNPs) -which contain oleyl trimethyl chitosan, poly(γ-(4-(((piperidin-1-yl)ethyl)amino)methyl)benzyl-Lglutamate), oleyl-PEG-mannose, oleyl-PEG-cysteamine, and sodium tripolyphosphate -with specific functions of mucoadhesion, transepithelial permeation, membrane penetration and active targeting have also been tested ( However, not all nanoparticle types can efficiently deliver RNAs. In a study evaluating the ability of unassisted epithelial entry of nucleic acids as a consequence of nanocarrier contact with mucus, antisense oligonucleotides but not siRNA (naked) were shown to be delivered to mouse intestine when formulated with nanocarriers composed by chitosan (Martirosyan et al., 2016).

Multi-compartmental formulations -consisting of one or more internal compartments surrounded by protective external compartments -have been developed to surmount the many barriers to oral RNA delivery (Kriegel et al., 2013; . One example is the "nanoparticles in the microspheres oral system" (NiMOS), a solid-in-solid multi-compartmental system (Kriegel et al., 2013) . Gelatin nanoparticles entrapped in poly(epsilon-caprolactone) microspheres were prepared carrying TNFα siRNA (slightly chemically modified dTdT) to treat DSS-induced colitis. Oral administration of siRNA (1.2 mg/kg) to female Balb/c mice decreased colonic TNFα expression and supressed the expression of proinflammatory cytokines (Kriegel, C and Amiji, M, 2011) . When combined with siRNA against Cyclin D1 (naked siRNA, also at 1.2 mg/kg/day), the silencing effects were more potent than with TNFα siRNA alone (Kriegel C and Amiji MM, 2011) .

Dectin-1 recognizes beta-1,3 and beta-1,6 linked glucans rich particles and intact yeast and is particularly expressed on the monocyte/macrophage and neutrophil lineage (Herre et al., 2004) . β-1,3-D-glucan has been used to deliver exogenous RNA targeting macrophages (Aouadi et al., 2009 ). Delivery of β-1,3-D-glucan siRNA particles containing 20 µg/kg unmodified siRNA by oral gavage to mice for 8 days reduced Map4k4 expression in different macrophages from different tissues and protected them from lipopolysaccharide-induced lethality (Aouadi et al., 2009 The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

The CD40 gene was the target of the antigen-presenting dendritic cells and oral administration of the recombinant yeast. Short hairpin RNA delivered in Lactobacillus casei have also been tested to target the intestine (Kuwahara et al., 2007) . Other organisms, including attenuated Salmonella typhi, were used as vectors to deliver RNAs, even ribozymes (Bai et al., 2011) or siRNA (Jiang et al., 2007) , using oral administration in mice to target either local or systemic effects (Bai et al., 2011; Jiang et al., 2007) .

The The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Extracellular vesicles are a heterogeneous family of vesicles delimited by membranes. These can be classified by their size as i) exosomes (30-100 nm in diameter), which originate from the endosomal system, ii) microvesicles (100-2000 nm) formed by budding out of the plasma membrane, and iii) apoptotic bodies (>1000 nm) formed by blebbing of the plasma membrane during apoptosis (reviewed in Yanez-Mo et al., 2015; Colombo et al., 2014) . Increasing evidence suggests that extracellular vesiclemediated communication can take place in vivo, but the nature of the extracellular vesicles involved in these effects remains to be clarified (reviewed in Tkach and Thery, 2016; Pitt et al., 2016) .

Exosomes, which are formed in multivesicular compartments, are secreted when these compartments fuse with the plasma membrane (Kowal et al., 2014; Abels and Breakefield, 2016) . Exosomes have been shown to contain or transport a myriad of molecules including proteins, carbohydrates, lipids, and a variety of genetic material including DNA, mRNA and ncRNAs (reviewed in Choi et al., 2015; Abels and Breakefield, 2016) . Exosomes are also described as participating in intercellular communication (Costa-Silva et al., 2015; Pironti et al., 2015; Nojima et al., 2016) , delivering their parental cell-derived molecular cargo to a recipient cell. Some of these processes are thought to be mediated by ncRNAs, including miRNAs Thomou et al., 2017) . Detailed information on how all these biological processes occur is outside the scope of this review. Given the above characteristics of exosomes, these vesicles hold promise as delivery vehicles for therapeutics (Munagala et al., 2016; van der Meel et al., 2014) . Indeed, exosomes have been used to deliver siRNA to different tissues when administered systemically, including the brain (Alvarez-Erviti et al., 2011; Cooper et al., 2014; Didiot et al., 2016) . miRNAs have also been delivered via exosomes to different tissues including xenograft breast cancer tissue (Ohno et al., 2013), brain and other tissues when administered systemically or locally (Zhang, D et al., 2017) . Interestingly, oral administration of exosomes or extracellular vesicles has also been described in animal models Agrawal et al., 2017; Oliveira et al., 2017; Oliveira et al., 2016) . However, their biological effects were not related to transport of ncRNAs. Some extracellular vesicles have been shown to resist digestion under in vitro simulated GI conditions (Benmoussa et al., 2016; , particularly those from bovine milk (Vashisht et al., 2017) . Moreover, milk exosomes seem to exhibit cross-species tolerance and are not described to induce adverse immune and inflammatory responses (Munagala et al., 2016) . However, whether exosomes or other extracellular vesicles resist the harsh in vivo conditions of the GI tract and the digestive process remains poorly described and needs to be studied (Tomé-Carneiro et al., 2018) .

Exosome-like nanoparticles from plants have been isolated from different species including ginger, grape, grapefruit and carrot (Mu et al., 2014; Ju et al., 2013; Zhang et al., 2016) . Some authors indicated that the approximate amount of exosome-like nanoparticles in these edible plants were reported to be ≈350 mg/100 g edible plant (Mu et al., 2014) , while other authors reported only ≈50 mg/kg ginger . Most of these studies reported the presence of RNAs, either small or large, including ncRNAs (Mu et al., 2014; Zhang et al., 2016; Ju et al., 2013) . The amount of RNA present in these exosome-like nanoparticles was reported to be ≈ 4 µg RNA/100 mg exosomelike nanoparticles for grape and grapefruit, and ≈ 12 µg RNA/100 mg exosome-like nanoparticles for ginger and carrot (Mu et al., 2014) . The in vivo biological effects of these nanovesicles have been assessed. Oral delivery of ginger nanovesicles during 7 days at a dose of 0.3 mg/day was found to target the colon, to be taken up by epithelial cells and macrophages, and to reduce DSS-induced colitis in mice . Treatment for 18 weeks with the same dose also exhibited antiinflammatory activity in a mouse model of chronic colitis . Ginger-derived lipid vehicles have also been generated from ginger lipids and loaded with siRNA-CD98 (Zhang M et al., www.efsa.europa.eu/publications 130 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. 2017). Oral administration of these ginger-derived lipid vehicles targets the colon tissue and they are effective in treatment of induced ulcerative colitis in mice.

Grapefruit exosomes at a dose of 10 mg/kg per day for 7 days also ameliorated the DSS-induced colitis targeting intestinal macrophages . When administered orally at a dose of 2 mg of nanovesicles/day, grape exosomes-like nanoparticles were also found to protect mice from a DSSinduced model of colitis by induction of intestinal stem cells (Ju et al., 2013) . Local intestinal macrophages and stem cells from the small and large intestine were also found to be the target of these exosome-like nanoparticles when administered by oral gavage (2 mg/day) (Mu et al., 2014) . Even though miRNAs were reported to be present in these exosome-like nanovesicles (Ju et al., 2013; Mu et al., 2014; Zhang et al., 2016) , the biological effects of these ncRNAs were not evaluated. Whether these miRNAs resist the GI tract conditions remains unknown.

A main hurdle in RNA-based therapeutics is the successful delivery of naked RNA (non-chemically modified) molecules to specific (target) tissues in the GI tract. The literature contains few studies in which exogenous RNAs are administered in vivo using routes of administration other than oral, and most are in the field of inflammatory bowel disease.

An amphiphilic cationic cyclodextrin (CD) vector was developed for a complex siRNA against TNFα, where the complex was concentrated to 50 µg/100 µL in 5% glucose (McCarthy et al., 2013) . CD.TNFα.siRNA was found to be efficient against DSS-induced colitis in C57Bl/6 mice treated twice (day 2 and 4 post-DSS) by intrarectal administration with the test solution (100 µL). CD.siRNA administration reduced TNFα and IL-6 expression as compared to the non-silencing siRNA control or the naked TNFα siRNA (not delivered in formulated cyclodextrin).

Enteral delivery of siRNAs for systemic effect was also tested. A nuclease resistance and chemically modified siRNA against apolipoprotein B was conjugated with α-tocopherol and was administered (10 mg/kg) as lipid nanoparticles in the large intestine of mice. The lipid nanoparticle was composed of mixed micelles comprised of linoleic acid and PEG-60 hydrogenated castor oil. The siRNA efficiently reached the liver and other tissues using the chylomicron-mediated pathway via the lymphatic route (Murakami et al., 2015) . Delivery of siRNA into hepatocytes was markedly reduced when the siRNA was not bound to α-tocopherol, indicating that conjugation with α-tocopherol was essential to this delivery system using this route.

A calcium phosphate (CaP) core coated with siRNAs and encapsulated in poly(D,L-lactide-co-glycolide acid) (PLGA) nanoparticles containing an outer layer of polyethyleneimine (PEI) was developed for intrarectal delivery (Frede et al., 2016) . In a DSS-induced model of colonic inflammation, intrarectal administration (12 µg/day) of nanoparticles containing 114 ng siRNA against either TNFα, IP-10 or KC from day 2 to 5 after 4% DSS treatment, ameliorated the colitis. Indeed, while intestinal epithelial cells, dendritic cells, macrophages and T cells could uptake the TNFα siRNA nanoparticles, reduction of TNFα was only detected in epithelial cells and T cells (Frede et al., 2016) . The stability of siRNA within the CaP/PLGA/PEI nanoparticles against enzymatic nucleases under colonic conditions was evaluated. While the siRNA-loaded CaP/PLGA/PEI nanoparticles were detectable after 1 h incubation at 37ºC with colonic fluid or homogenate, the same siRNAs not formulated in nanoparticles was not detected and was assumed to be degraded. Even after exposure to colonic homogenate for 1 h at 37ºC, the nanoparticles retained biological activity. Since the supernatant of the homogenate subsequently incubated with a cell line, still reduced the expression of TNFα by 50% (Frede et al., 2016) , suggesting that siRNA-loaded nanoparticles retain its biological activity compared to that of nacked siRNAs (not formulated).

www.efsa.europa.eu/publications 131 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Other alternatives for RNA delivery have been tested. Ultrasound-mediated RNA delivery has been efficiently achieved in mice (Schoellhammer et al., 2017) . Chemically modified siRNA against TNFα (100 ng siRNA in 200 µL water) was delivered in the rectum using ultrasound exposure to 40 kHz for 0.5 s and after a 6-day treatment TNFα was efficiently repressed in colonic tissue of a mouse model of induced colitis. In the same study, a mRNA (≈ 950 kDa) complexed in lipid nanoparticles (LNPs) was administered in the colon (300 µL LNPs solution) with ultrasound and found to be successfully translated locally in the colonic tissue (Schoellhammer et al., 2017) .

Oral administration of exogenous RNAs to exert a biological effect, either locally or systemically, is successful if combined with specific delivery technologies, such as vehicles developed for pharmaceutical/medical purposes. Nanoparticles have been designed to release their cargo in specific regions of the GI tract (i.e. where inflammation occurs) or to resist the GI conditions. In contrast to other routes of administration, where normally highly chemically modified RNAs are used, oral delivery normally involves use of naked RNA or minimally modified RNAs formulated with complex delivery vehicles. Achieving the local desired effect (i.e. targeting inflammation in the colon) has contributed to the development of delivery technologies and the usage of other routes of administration different than the oral as described above. However, the available literature also suggests that naked or unmodified exogenous RNAs are rapidly degraded when exposed to the GI conditions without incorporation into delivery vehicles. Extracellular vesicles are natural lipid particles released by many cell types with the potential to mediate cell-to-cell communication. These extracellular vesicles have a tremendous potential as delivery vehicles for therapeutics. However, there are still very few studies evaluating their resistance to the harsh conditions of the GI tract following oral administration. Knowledge of their biological effects as transporters of exogenous ncRNAs is still very limited.

Nucleic acid-based therapeutics have the potential to treat numerous diseases by correcting abnormal expression of specific genes. As indicated previously, the oral route of drug administration poses serious delivery challenges due to the GI degrading environment, as well as the need to overcome other barriers preventing nucleic acid delivery. Because of these obstacles, efforts in developing oligonucleotide-based therapeutics using the oral route have focused on local gastrointestinal delivery, i.e. directly delivering the drug to the target tissue for localized effects, and, therefore, increasing local bioavailability and maintaining doses low while diminishing possible effects to non-target tissues Kriegel et al., 2013) .

Crohn's disease and ulcerative colitis are the two principal forms of IBD, characterised by being a chronic relapsing inflammatory condition of the GI tract. The pathogenesis of IBD is dependent on the interaction between local immune and environmental factors in genetically-susceptible individuals (O'Neill et al., 2011) . Inflammation in Crohn's disease is characterised by high production of the cytokines interferon (IFN)-gamma, IL-17 and TNF-alpha, representing a T helper1 (Th1)-Th17 response.

In ulcerative colitis, the immune response is characterised by Th17 and an atypical Th2, with high production of IL-5, IL-10 and IL-13 (O'Neill et al., 2011) . Conventional treatment consists of antiinflammatory and immune-suppressive drugs but, while some medications are effective to combat inflammation in the acute phase, they are ineffective in maintaining remission due to toxicity, dependency and higher relapse rates (Kriegel, C and Amiji, M, 2011) . The goal for IBD treatment is local delivery of therapeutics to intestinal immune cells. Initial approaches involved IL-10 supplementation www.efsa.europa.eu/publications 132 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

due to its huge potential in blocking proinflammatory cytokines and inflammatory tissue damage. But human clinical trials have not proven that systemic IL-10 supplementation can help prevent or improve the IBD symptoms, probably due to IL-10 low mucosal levels . Separate studies have used gene therapy to increase IL-10 mucosal availability by engineering IL-10 plasmid DNA, and achieving more success in rodent models (Barbara et al., 2000; Steidler et al., 2000; Lindsay et al., 2003; Bhavsar and Amiji, 2008) . Deregulation of tumour necrosis factor (TNF)-alpha production, a proinflammatory cytokine primarily produced by activated macrophages along with other cell types, is associated with the onset and progression of a number of inflammatory diseases (such as IBD, rheumatoid arthritis), Alzheimer's disease and psoriasis . The harmful effects of TNF-alpha are thought to be mainly mediated by activation of NF-kappaB, therefore regulating expression of over 200 genes. Systemic blockage of TNF-alpha is accompanied by several side effects that can be avoided if local inhibition occurs. RNAi-mediated local suppression of TNF-alpha can be beneficial for overcoming many of these side effects. Different groups have developed various delivery systems for siRNA molecules targeting TNF-alpha expression in the GI tract, assaying them both on cells and several rodent models for IBD (Wilson et al., 2010; Kriegel, C and Amiji, M, 2011; Kriegel, C and Amiji, MM, 2011; Laroui et al., 2011; Yin et al., 2013; He et al., 2013b, a; He et al., 2015) . Even though therapeutic RNAi development is in its infancy, current approaches and tools for delivery of these biomolecules via the oral route are improving. In fact, recent reports describe the effects of an oral SMAD7 antisense oligonucleotide in a randomized, placebo-controlled, double-blind, phase 2 clinical trial in patients with Crohn's disease (Monteleone et al., 2016; Monteleone et al., 2015) . Although in this instance the biomolecule used is an antisense oligonucleotide and not a siRNA, the design of the oral formulation could be very similar.

Iron is an essential metal required for numerous physiological functions, but in excess it is a well-defined risk factor in the pathogenesis of several diseases, including cardiovascular and neurodegenerative diseases. Since there is no recognised active pathway of iron excretion, disposal of excess iron from the body is the primary therapeutic goal of treating patients with iron overload. Use of iron chelators is limited due to nonspecific distribution in non-target tissues, which results in a number of serious side effects and toxicity . Another way to treat iron excess is by limiting absorption of exogenous iron. The divalent metal transporter 1 (DMT1) protein plays a well-established role in iron absorption. The primary site of DMT1 activity is the intestinal epithelium. DMT1 expression is regulated in response to body iron levels, with expression enhanced when iron stores are low and, conversely, reduced when iron stores are high. Oral delivery of siRNA-encapsulated NiMOS to selectively suppress intestinal DMT1 could decrease intestinal uptake of dietary iron, thereby mitigating iron overload and preventing iron-mediated toxicity .

Celiac disease is caused by a T-cell mediated immune response in the small intestine against deamidated cereal gluten peptides modified by the enzyme transglutaminase 2 (TG2). The only way to prevent celiac symptoms is strict adherence to a gluten-free diet. The pathophysiology of celiac disease involves a combination of environmental, genetic and immunological factors. Among several immune mediators, enzyme tissue TG2 and the proinflammatory cytokine interleukin-15 (IL-15) have emerged as key players in promoting inflammatory responses against dietary gluten. Given their important role in the pathophysiology of celiac disease, siRNA mediated silencing of intestinal TG2 and IL-15 (i.e. by oral administration using NiMOS) may result in neutralizing proinflammatory effects and could therefore alleviate disease symptoms . EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Colorectal cancer is one of the most common forms of cancer, and RNAi therapy has great potential in the treatment of local intestinal cancers. However, RNAi therapy faces major challenges since the genotypic footprints of the various cancers differ widely, even from individual to individual. Depending on the tumour, several different sets of genes may be misexpressed, making it extremely difficult to determine a common/unique "culprit" gene. Moreover, mice and rats do not generally spontaneously develop colon cancer, so induction of tumour growth is required. But these developed animal models do not reflect the genetic changes observed in human tumours (O'Neill et al., 2011) . Genetic mouse models that spontaneously develop intestinal cancer have also been developed. Mutations in the Apc gene are associated with development of colon cancer as well as a condition called familial adenomatous polyposis (FAP). FAP is an inherited condition characterised by development of numerous polyps in the colon. Some of these polyps can subsequently become malignant adenocarcinomas (O'Neill et al., 2011) .

Regarding FAP and human gene therapy, one study explored oral delivery of Escherichia coli expressing shRNA against beta-catenin, and found significant silencing activity in healthy mice (Xiang et al., 2006) .

The GI tract provides the possibility of acting at systemic level by targeting local resident cells, such as M cells. M cells can take up encapsulated biomolecules to be phagocytosed by macrophages. In fact, Aouadi et al. reported that glucan encapsulated siRNA particles (GeRPs) loaded with siRNA were phagocytosed by the underlying gut-associated lymphatic tissue (GALT) macrophages and then translocated to distal organs of the reticuloendothelial system such as the liver, spleen and lung (Aouadi et al., 2009) . Another report claims functional CD40 shRNA expression delivered into dendritic cells in mice by oral administration of recombinant yeast (Xu et al., 2016) . In this manner, the targeted cells located in the intestine can undergo siRNA-mediated gene silencing and migrate into tissues throughout the body.

When successfully administered, RNAi is a very useful therapeutic strategy for efficiently modulating gene expression. Orally administered local RNAi therapeutics have great potential due to the increasing bioavailability of the therapeutic biomolecule at the targeted tissue or cell type, coupled with simultaneous evasion of systemic side effects and immunogenic reactions. Most advanced studies are focused on inflammatory bowel disease, although other potential diseases are being explored. The current main challenge is development of suitable encapsulation methods to protect the RNA molecules from the harsh and varying conditions of the GI tract.

The literature on RNA delivery in fish and birds is scarce, with some examples of dietary delivery of exogenous RNAs. In birds, most of the studies have been done using central nervous system administration of siRNAs. For example, in the zebra finch bird, siRNAs (three 19-25 nt long sequences against tyrosine kinase B, TrkB) administration via direct injection into the brain was effective in reducing the volume of the high vocal centre by diminishing the robust nucleus of the arcopallium and the relative number of cells within it (Beach et al., 2016) . In another study, direct administration of siRNAs in the brain of white-crowned sparrows reduced resting time, spontaneous production of complex vocalizations, and stimulated brief agonistic vocalizations (Ubuka et al., 2013; Ubuka et al., 2012) . In the Japanese quail, central administration of the same siRNAs against gonadotropin-inhibitory hormone The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

siRNAs for RNA interference in chicken embryos has also been used to target Marek's disease virus in vivo (Chen et al., 2009 ).

In fish models, systemic exosome-mediated siRNA delivery to zebrafish resulted in effective crossing of the brain blood barrier and localization in the brain in vivo (Yang T et al., 2017) . Exosome-delivered siRNAs decreased fluorescence-labelled tumour cells in the brain more than fourfold by inhibiting vascular endothelial growth factor (VEGF) in a xenograft zebra fish brain tumour model (Yang T et al., 2017) . These results suggested that brain endothelial exosomes could be used to deliver exogenous siRNAs to the target site in the brain for treatment of brain cancer. The proposed higher delivery efficiency, lower immunogenicity, and better compatibility than existing foreign RNA carriers have promoted the possible use of exosomes for exogenous RNA delivery (Mei et al., 2018) . Several other studies have evaluated systemic delivery of siRNAs to fish animal models, including delivery to the heart using PEG-PLA nanoparticles (Diao et al., 2015) or using a neutralized non-charged polyethyleniminebased system . The literature describes other examples of siRNA delivery into zebrafish embryos to inhibit specific gene function (Gruber et al., 2005) .

In the context of other types of ncRNAs, direct injection of miR-26a mimic RNAs at the one-cell stage of zebrafish embryos resulted in inhibition of formation of the caudal vein plexus by means of targeting the endothelial cell bone morphogenic protein SMAD1 signalling (Icli et al., 2013) . Also in the zebrafish model, injection of a miR-722 mimic into embryos at the one-cell stage to deliver miR-722 ubiquitously resulted in protection against sterile inflammation (Hsu et al., 2017) . dsRNA microinjection administration to zebrafish embryos (1-2 cell stage) at a concentration of 15-60 pg RNA/embryo resulted in RNA interference, an effect that was diminished when ssRNA was used (Wargelius et al., 1999) . In cultured European sea bass, intramuscular injection (7 weeks, 40 µg of dsRNA per dose) followed by in vivo electrically mediated dsRNA delivery resulted in a reduction of myostatin gene expression (Terova et al., 2013) .

Delivery of exogenous RNAs has also been tested in lower aquatic organisms. In the crustacean red claw crayfish Cherax quadricarinatus, systemic injection of sequence specific dsRNAs against viral protein B2 of Macrobrachium rosenbergii nodavirus (MrNV) was conducive to RNAi effects that were able to functionally prevent and reduce mortality in infected individuals (Hayakijkosol and Owens, 2012) . dsRNAs (intramuscular injection) have been used to induce antiviral immunity in invertebrates as they can also recognize dsRNA as a virus-associated molecular pattern, resulting in activation of an innate antiviral response (Robalino et al., 2004) . Oral administration of encapsulated dsRNAs (inactivated bacteria expressing dsRNA) in the prawn Macrobrachium rosenbergii fed twice daily at a rate of 5% total biomass for 7 days resulted in their protection against white tail diseases caused by MrNV (Naveen Kumar and Karunasagar, 2013) . dsRNAs designed against the capsid and B2 genes of MrNV reduced the mortality of prawns by ≈70% due to sequence specific-mediated silencing of the viral genes (Naveen Kumar and Karunasagar, 2013) . Injection of viral protein 28-targeted dsRNA (≥12 µg dsRNA/g shrimp body weight) into shrimp resulted in a high level of 22 nt siRNAs mapping across the entire white spot syndrome virus genome after virus challenge, which resulted in protection against the virus (Nilsen et al., 2017) . In planarians, ingestion of bacterially-expressed dsRNAs or direct microinjection into the animals can inhibit gene expression through RNA interference (Newmark et al., 2003) . EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. Mongersen, an oral SMAD7 antisense oligonucleotide, and Crohn's disease. N Engl J Med 372, 1104-1113. Mu J, Zhuang X, Wang Q, Jiang H, Deng ZB, Wang B, et al. 2014 The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. 1927 -1943 . Zollner N. 1982 . Purine and pyrimidine metabolism. Proc Nutr Soc 41, 329-342. Zollner N, Grobner W. 1971 . Influence of oral ribonucleic acid on orotaciduria due to allopurinol administration. Z Gesamte Exp Med 156, 317-319.

Following a literature search as described in section 2.2.2.2 and based on the methodology described in section 2.2.1.2, a total of 64 documents were selected as relevant to reviewing the effect of dietary exogenous ncRNAs on the GI tract and annex glands. Very few studies have focused on the GI tract and their annex glands. Four documents specifically evaluated the resistance of plant-derived sRNAs to simulated in vitro digestion system, ex vivo digestion system or in vivo digestion samples (Philip et al., 2015; Yang, et al., 2017a; and reported that the percentage of unaltered miRNAs after digestion might be below 0.01%. Due to their relevance in early human nutrition, literature on breast milk ncRNAs was chosen to describe the biological effects of dietary exogenous ncRNAs of non-plant origin, for which 35 full-text reports were studied and reviewed.

Zhang et al. were the first to report on the transfer of miRNAs from food plants to the mammalian circulatory system causing effects on recipient cells . This study reported the detection by high-throughput sequencing of plant sRNAs in human serum, as well as in tissues of The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

humans, mice and calves. The plant miRNAs were distinguished from human miRNAs by their 2'-Omethyl modification, which made them resistant to periodate whereas the human miRNAs having free 2' hydroxyl were periodate sensitive. It was then assumed that the identified plant miRNAs were coming from the food uptake. Two rice miRNAs named osa-miR156a and osa-miR168a were particularly abundant in human serum samples, since rice is common in human diets. Additional in vitro experiments showed that these plant miRNAs can endure conditions that mimic the acidic environment of the gut, suggesting that plant miRNAs can survive being eaten and digested, pass through the mammalian GI tract and reach target organs. The authors also showed that diet-derived plant miR168a can target the Low Density Lipoprotein Receptor Adapter Protein 1 (LDLRAP1) mRNA based on sequence complementarity. This miRNA was able to reduce the LDLRAP1 protein level in the blood and liver of mice fed rice, and eventually increase their plasma low-density lipoproteins (LDL) content. This publication raised questions about whether food-derived sRNAs could play an active role in human/animal health. However, it was controversial and its results were questioned due to nutritional imbalances in the test-diet studies or lack of RNAi-mediated modulation of LDLRAP1 protein levels in mouse liver (Witwer and Hirschi, 2014; Chen et al., 2013; . Scientists from the Monsanto and miRagen companies were unable to reproduce these findings . Measuring LDLRAP1 protein levels in the liver of mice fed with miR168a-containing rice diets by ELISA instead of Western-blot analysis, Dickinson and colleagues did not detect any protein change modulated by miRNA168a-diets. Although changes in the LDL plasma levels were indeed detected, the authors concluded that LDL increases resulted from nutritional imbalances rather than from a miR168 consumption mediated effect. Additionally, plant miRNAs including the highly abundant and stable osa-miR168a were not detected in liver and plasma samples of mice fed under different rice-based regimes by sRNA sequencing with the HiSeq Illumina system .

Other independent research groups failed to detect dietary plant miRNAs in plasma and tissues when conducting controlled feeding studies in humans and in animal models using different detection methods, such as real time quantitative PCR (RT-qPCR), droplet digital PCR (ddPCR) and next generation sequencing (NGS) analysis . These studies provide contradicting evidence of miRNAs dietary absorption. In all the attempts, plant miRNAs were detected at substantial levels in the diets but were undetectable in animal fluids and tissues. Snow and colleagues studied three plant miRNA species (miR156a, miR159a, and miR169a) highly abundant in fruits routinely ingested by a group of ten healthy athletes. Examination of the athletes' plasma by RT-qPCR analysis using Taqman probes revealed high levels of endogenous miRNAs but not the three dietderived plant miRNAs . The same results were obtained with mice fed vegetarian diets rich in these 3 miRNAs. Furthermore, using a mir21 null mutant mouse they showed that none of the dietary miRNA including miRNA21 was detected in blood and body tissues, suggesting that dietderived miRNAs were not absorbed or maintained in mouse body fluids or organs. Another group reported negative results on dietary miRNA absorption in non-human primates . Using qRT-PCR analysis and the same timeframe of as in Zhang's report a, plant-specific miRNAs were undetectable in tissues from macaques given a vegetarian diet. Another feeding study in humans showed no significant differences in the levels of the plant miR167a and miR824 in the plasma after consumption of broccoli sprouts highly rich in these two miRNAs (Baier et al., 2014) . More recently, a different research group using by NGS analysis failed to detect plant miRNAs in the plasma of healthy volunteers that normally consume olive oil .

The possibility that the plant RNAs detected in human tissues and body fluids in some studies were contaminants remains a serious concern . By searching the composition of sRNA-seq data generated by the Zhang's team from amphioxus animals, www.efsa.europa.eu/publications 142 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

it was found that the amphioxus data contained rice miRNAs as previously reported for human serum. Given that amphibia have an exclusive algae-based diet, this strongly suggests that the rice miRNAs in both studies are the result of samples contamination . Spurious detection of such sequences could arise due to contamination during sample handling, library preparation or sequencing, or result from errors in data analysis. Since next-generation sequencing is very sensitive, just a few molecules could cause a false positive detection. More recently a comprehensive meta-study surveying for the presence of cross-species miRNAs in 824 sequencing data sets from various human tissues and body fluids indicated that xenomiRs (cross-species miRNAs) are technical artefacts rather than the results of dietary intake .

A bioinformatics study identified plant miRNAs in human and porcine milk exosomes . By searching publicly available NGS datasets, the authors identified several plant miRNA species, most of them belonging to evolutionarily conserved and highly abundant miRNA families. The target prediction suggests that these miRNAs may interact with mRNAs coding several transcription factors, protein receptors, transporters and immune-related proteins, thus potentially influencing biological functions in humans. Of note is that the plant miRNA profiles found in mammalian breast milk were similar to the composition in human blood presented by Zhang and colleagues .

Liang and colleagues reported positive results on the absorption of food-derived miRNAs from feeding studies in mice using cabbage (Brassica oleracea) . They showed that miR172, the most abundant miRNA in cabbage, was detected in blood and various organs after cabbage consumption. However, there was no absolute quantification of miR172 or a description of an endogenous control for normalization of qRT-PCR data. A more recent study by Zhang's group described how plant miRNA2911 from honeysuckle (Lonicera japonica) inhibited Influenza A virus replication in mice when entering through the GI tract, and they hypothesized that this miRNA might be the active component in the traditional Chinese medicine based on honeysuckle herb to treat influenza infection . The authors demonstrated that miRNA2911 is highly stable in honeysuckle decoction by Solexa sequencing and Northern blot analysis. They also showed that miR2911 levels increased in mouse peripheral blood and lung following continuous honeysuckle decoction intake (drinking) or gavage feeding, as measured by qRT-PCR using TaqMan miRNA probes. Additionally, luciferase reporter assays and in vivo experiments showed that MIR2911 was able to target the influenza A virus, and consequently inhibited its replication and reduced mice mortality. The same authors reported that plant miRNAs can be found in human umbilical cord blood and amniotic fluid, transferred from mother to the foetus, and that these influence foetal development and health . They showed that miR2911 content increased in the maternal plasma and foetal liver of mice fed honeysuckle. A fluorescently labelled siRNA was used to trace the transplacental transmission through feeding. Furthermore, they showed that after feeding mice with different amounts of siRNAs targeting the alphafetoprotein mRNAs the levels of this protein were down regulated in foetal tissues. These results suggest that dietary siRNAs delivered from the mother to the foetus could regulate foetal gene expression.

More recently, the anti-proliferative effect of a plant miRNAs on breast tumours was demonstrated in feeding experiments in mice (Chin et al., 2016) . The plant miR159, particularly abundant in broccoli, was found in sera from women, and its levels were inversely correlated with breast cancer morbidity and progression. In human sera, miR159 was detected in extracellular vesicles by qRT-PCR using a Taqman probe. Transfection of breast cancer cells with a synthetic mimic miR159 was shown to reduce their proliferative growth by targeting the Transcription Factor 7 mRNAs. More importantly, oral administration of synthetic miR159 significantly inhibited the growth of xerograph breast tumours in mice. These results agree with a previous report in which oral administration of a cocktail of three www.efsa.europa.eu/publications 143 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

tumour suppressor miRNAs that were 2'-O-methylated to mimic plant miRNAs reduced colon tumour burden .

Additional research is needed to address discrepancies among different studies and verify whether plant sRNAs derived from diet are absorbed at functional levels in animal species. Most of the available data suggests either that gastrointestinal absorption of dietary plant miRNAs does not occur in healthy consumers or miRNAs levels in blood and tissues of consumers are too low to have a biological impact. However, the biological activity of food miRNAs may be more dependent on their accumulation in exosomes than on their abundance per se, because of their reported low concentration. Development of sensitive sensors to help to detect the functionality of low-level absorbed miRNAs would facilitate assessment of the effects of dietary miRNAs in consumers. Future studies are needed to establish how exogenous ncRNAs absorption and tissue distribution occurs, as well as to address their bioavailability and their biological function. reported that increased MIR2911 was detected in sera and urine after consumption of particular foods (i.e. honeysuckle), and disappeared 48 h after the honeysuckle was removed from the diet; this supports the idea that these small RNAs were of dietary origin. The role of kidney damage in miRNA retention in the animals was also evaluated. Using two models of acute renal failure (Baliga et al., 1999) , i.e. the Cisplatin (a chemotherapeutic agent) and the glycerol-induced models, it was observed that only mice receiving Cisplatin exhibited measurable levels of dietary miRNAs in sera and urine. This suggests that kidney damage alone did not lead to enhanced miRNA retention in the mouse circulatory system . Of interest is that Cisplatin treatment, but not glycerol treatment or honeysuckle feeding, disrupted the organization of small intestine epithelial cells as shown by histological analysis. In this study, Yang et al. used droplet PCR to demonstrate that the amplified products were likely to be specific and not due, for example, to nonspecific amplification of endogenous RNAs . Whether particular foods and/or alterations in intestinal permeability could improve the capacity to absorb small RNAs from the diet or influence the biological effect within the GI tract, annex glands and systemically is poorly described, but still relevant for risk assessment of exogenous ncRNAs in a subset of population.

No studies specifically report the biological effects of dietary exogenous plant ncRNA on the GI tract. Only a few studies (described above) report the possible biological effects in their annex glands, including the liver . For gastrointestinal tissues or the liver, Liang et al. reported that ICR mice (a strain of albino mice) fed plant total RNA (10-50 µg) extracted from Brassica oleracea showed detectable levels of MIR172 in the stomach (up to ≈30000 copies), intestinal (up to ≈16000 copies) and faecal content (up to ≈12000 copies), with the maximum amount present at 2 to 4 h after feeding . Estimation of the proportion of orally administered RNA that survived digestion suggested that the stomach contained 0.4-4.5%, the intestines 0.2-2.4% and the faeces 0.3-1.8 %, while the blood contained about 0.2-1.3% and the spleen about 0.04-0.38% . The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

which was similar to that of their synthetic form (without 2'-O-methylated 3' ends) . Whereas plant miRNAs had a much slower degradation rate compared to their synthetic form without 2'-O-methylated 3' ends. It is important to note that extreme pH conditions are not the only physicochemical or biological condition within the GI tract (see section 3.2.1.1. for details).

Using a simulated human digestion system in vitro, Philip et al. evaluated the resistance of soybean and rice miRNAs to simulated gastric fluid early stage digestion (Philip et al., 2015) . The study found continual survivability of plant miRNAs in an in vitro simulated digestion system for over 75 min without any significant decrease in their levels (Philip et al., 2015) . miRNA levels have been analysed qualitatively by real-time PCR using TaqMan miRNA assays. Stability of the plant-derived small RNA miRNA2911 was also assessed using an artificial in vitro digestion system that simulates mammalian gastric and intestinal conditions . Yang et al. compared three samples which contained a) 1 mL cabbage extract with 10 pmol plant-derived MIR2911 (measured by qRT-PCR), and 10 pmol spiked-in synthetic 2'-O-methylated MIR168a and artificial miRNA termed C7; b) 1 mL phosphate buffered saline (PBS) with 10 pmol each of synthetic 2'-O-methylated MIR2911, MIR168a and C7; and c) 1 mL PBS with 10 pmol each of synthetic MIR2911, MIR168a and C7 without the 2'-O-methylation. Time course analysis during both gastric and intestinal phase digestion showed that most miRNAs displayed modest resistance in the acidic gastric environment (up to 60 min), while in the intestinal phase the levels of all miRNAs were drastically reduced after five minutes (Figure 14) . Regardless of synthesis origin and 2'-O-methylation, the digestive stability of MIR2911 was up to 100-fold higher than that of the other miRNAs. The most stable form of MIR2911 was the plant-derived form in the cabbage extract, since 0.044% survived, compared to 0.0059% for the 2'-O-methylated form, and 0.0037% for the non-modified form. Similar results were observed when 10 pmol of synthetic 2'-O-methylated MIR2911, MIR168a and C7 were digested in vitro using the ex vivo intestinal fluids ( Figure 14) . Indeed, MIR2911 appeared to be more stable after 2-hour digestion than the other tested miRNAs. Finally, by directly measuring MIR2911 levels in vivo in the small intestines of mice fed the plant-based diet (daily dietary MIR2911 intake 10 pmol) it was observed that this particular miRNA reached level more than 100-fold higher than those observed in the animals consuming a single 400 pmol dose of the synthetic MIR2911. MIR168a, a miRNA also present in the plant-based diets, was also detected in the small intestines at a level 100 times lower than that of MIR2911 . Using transgenic Arabidopsis lines expressing the artificial miRNA amiR-RICE, the murine miRNA mmu-miR-146a or their controls, Yang et al. studied gastrointestinal digestion and bioavailability of transgenic miRNAs (Yang et al., 2017b) . Levels of transgenic miRNAs in plants (qRT-PCR quantification) were similar (22.9 and 26.3 fmol/g of fresh weight for amiR-RICE and mmu-miR-146a, respectively) to that of sRNA MIR2911 (18-19 fmol/g in fresh shoot tissue). However, in diets stored at room temperature, the abundance of amiR-RICE and mmu-miR-146a decreased gradually by ≈10-fold (≈1 fmol/g diet) while that of the sRNA MIR2911 increased ≈85fold after 24h. In vitro gastric and intestinal simulated digestion of the synthetic forms of the plantbased transgenic miRNAs showed that amiR-RICE and MIR2911 had similar digestive stability, while mmu-miR-146a was significantly less stable. The surviving percentage in the intestinal phase after 75 min was 0.39% for mmu-miR-146a compared to ≈1.11% for MIR2911 or amiR-RICE (Figure 15 , frame A). The surviving percentage in the in vivo assays (gavage-fed 400 pmol of synthetic sRNAs) further decreased the stability of the miRNAs, reducing their level to 0.0014% for amiR-RICE and 0.00045% for MIR2911 (Figure 15, frame B) . However, the levels of miRNAs in the small intestines were higher for MIR2911 (238 fmol/mouse) than amiR-RICE (2.7 fmol/mouse), while mmu-miR-146a levels were undistinguishable from the host mouse miRNA (Figure 15 , frame C). amiR-RICE was not found in serum (neither are miRNAs reported in previous studies and present in transgenic plants) while MIR2911 was found only in the serum (28.5 fM, or 1.3 x 10 7 copies per mouse) of mice fed the plant transgenic diets (Yang, J. et al., 2017b) www.efsa.europa.eu/publications 145 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Altogether, these results suggest that the plant-derived MIR2911 might be more stable than other plantbased small RNAs. Whether this exceptional stability could be generalized to other ncRNAs is unknown. However, recent evidence, including some studies evaluating MIR2910 and miR2911, suggests that there may be some misidentified alternative miRNAs . Indeed, several other rarer but consistently mapped plant miRNAs also have 100% or near 100% matches to human transcripts or genomic sequences (i.e. rRNA), and some do not map to plant genomes at all, suggesting artefactual results of plant miRNAs in mammalian sequencing datasets.

This emphasizes the need for rigorous filtering strategies when assessing possible dietary exogenous miRNAs . Indeed, the small RNA MIR2911 is derived from 26S rRNA and does not undergo canonical miRNA processing, as it does not depend on DCL1 (Yang, J. et al., 2017a) . MIR2911 is also inefficiently assembled into the RISC complex in human cells and modestly regulates gene expression. This suggests that this exogenous sRNA is different from the canonical plant-based miRNAs (Yang, et al., 2017b) . These results highlight the need to evaluate other ncRNAs as they could originate from precise processing at the 5' or 3' end of mature or precursor RNAs generating other abundant small RNAs (Lee et al., 2009) . Also, the RNA degradome is a crucial component of the total cellular RNA pool of plants (Nowacka et al., 2013) . They can be stable degradation intermediates present in a high copy number, or they can derive from various RNA species including tRNA, rRNA, mRNA and snRNA (Addo-Quaye et al., 2008; Nowacka et al., 2013) .

Among the few examples in the literature of possible biological effects of exogenous non-plant origin ncRNAs, milk has been studied because it constitutes a rich source of secreted miRNAs.

Human breast milk is an essential source of nutrition for new-borns. In addition to its nutritional function, breast milk contains myriad biologically-active components that influence the development of the The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. more than 87% of the 234 unique miRNAs identified in porcine milk exosomes (Gu et al., 2012) or 75% in other studies (Chen et al., 2014) . In human exosomes, the ten most abundant miRNAs account for 62% of the 602 unique miRNAs identified . lncRNAs previously reported as being important for developmental processes were found in human breast milk extracellular vesicles (Karlsson et al., 2016) .

Maternal factors such as the mother's diet seem to influence miRNA expression in the breast milk fat globule (Munch et al., 2013) or exosomes (Sun J, 2017) . Maternal weight (Xi et al., 2016 ) also seems to influence miRNA expression. Certain miRNAs found in breast milk have also been found in placenta tissue (Munch et al., 2013) , although their physiological role is unknown. Other studies have shown that colostrum miRNAs (early lactation period) are more abundant than in the later lactation period (Gu et al., 2012; Xi et al., 2016; Sun et al., 2013) . Bovine milk miRNAs isolated from the colostrum whey fraction were more abundant than in the mature milk whey fraction (Izumi et al., 2012) . Immunerelated miRNAs expression is reported to be higher in colostrum than mature milk (Izumi et al., 2012; Na et al., 2015) . In vitro studies of human milk miRNAs using either RNase digestion, freeze-thaw cycles, low pH or other GI conditions have shown that these miRNAs are relatively highly stable (Kosaka et al., 2010; Liao et al., 2017) . Similarly high stability of miRNAs was found when using one or all of the above mentioned hard conditions for milk from cow (Hata et al., 2010; Izumi et al., 2012 ), pig (Gu et al., 2012 or panda . Indeed, it has been proposed that RNA in milk is present in microvesicles and is protected from RNases and the other GI conditions by surrounding membranes (Hata et al., 2010; Gu et al., 2012; . However, technological conditions such as pasteurization (Golan-Gerstl et al., 2017) have been shown to reduce miRNAs abundance in cow and goat milk by up to ≈40%, or up to ≈60% for certain miRNAs (Howard et al., 2015) , and in human exosome (personal communication, FASEB conference 2017). Microwave heating also reduces the amount of certain miRNAs (Howard et al., 2015) , while ultrasonication influences exosome membrane integrity and thus increases cargo instability (Sun et al., 2013) . Among dairy products, miRNAs concentrations varied considerably, but were generally lower than the concentration in pasteurized whole milk (Howard et al., 2015) . The exception was fresh cheese ("queso fresco") dip, which contained higher concentrations of miRNAs than observed in pasteurized milk (Table 24 ).

In vitro studies using miRNAs isolated from breast milk exosomes, either human or bovine, were found (Liao et al., 2017) , and CRL1831, K562 or Lim1215 (Golan-Gerstl et al., 2017) . When subjected to proteinase K treatment or when the vesicle structure was destroyed, uptake or transport of these exosomal miRNAs was compromised (Kusuma et al., 2016; Wolf et al., 2015; Sun et al., 2013) .

Different studies have attempted to quantify the amount of RNAs present in breast milk. RNA content in microvesicles isolated from cow milk has been estimated to be ≈1700 ng for colostrum and ≈980 for mature milk when isolated from 6 mL sample (Hata et al., 2010) . In human milk, total RNA was estimated to be ≈2600 ng/10 6 cells of the milk cell fraction and ≈30 ng/µl fat from milk fat (Alsaweed et al., 2016c) . Exogenous plant miRNAs have also been found in panda milk exosomes , the whole milk or exosomes of human breast milk The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. et al., 2012). However, levels of miRNAs or total RNAs were very low (Golan-Gerstl et al., 2017; Alsaweed et al., 2016c) compared to that of breast milk. Indeed, Alsaweed et al. reported that the total amount of RNA found in bovine milk-based infant formula was ≈0.1 ng/µL formula (45 human miRNAs detected) as compared to 69.9 ng/µL for a soy-based formula (22 human miRNAs detected) (Alsaweed et al., 2016c). In terms of possible breast milk miRNA absorption, plasma levels of miRNAs were found to be higher in colostrum-fed piglets compared to mature milk-fed piglets (Gu et al., 2012) . In humans, plasma miRNAs (miR-29b and miR-200c) increased when different doses of cow milk were consumed (Baier et al., 2014) . In contrast to the studies above, the literature also describes several studies that show either different or contradictory results regarding uptake of exogenous RNAs from breast milk. Auerbach et al. re-analysed a study in which miR-29b-3p and miR-200c-3p were found in plasma after bovine milk ingestion by healthy humans (Baier et al., 2014) and found no significant altered miRNA levels after milk ingestion (Auerbach et al., 2016) . These results were confirmed by qPCR and RNA-sequencing. Several technical issues including low miR-29b expression, use of control miRNAs (normalization), and variation level may have contributed to these discrepancies (Auerbach et al., 2016) . Piglets fed cow milk for 4 weeks showed very low levels of cow-specific miRNAs in the bloodstream, which were compared to the levels measured from animals fed maize, which also showed similar counts of cow-specific sequences , suggesting a lack of transfer of exogenous small RNAs.

Using a transgenic mouse model that overexpresses miR-30b precursor in the mammary epithelial cells, and thus increases levels of miR-30b in milk (≈134-fold), no differences were found in blood, liver, small intestine, kidney or lung tissues of pups from these dams when compared to pups from wild type dams (Laubier et al., 2015) . Interestingly, the level of miR-30b was ≈30-fold higher in the stomach of mice consuming the transgenic milk. Oral ingestion of 0.125 µg and 0.25 µg of total RNA from porcine milk exosomes administered daily during three weeks to mice, significalty increased villus height and crypt depth of the duodenum and jejunum relative to the control group, suggesting an effect improving the development of the GI tract in mice .

Using genetic knock-out (KO) mouse models for miRNAs miR-375 and miR-200c/141, Title et al., evaluated the uptake of maternal-milk derived miRNAs using the foster mother offspring exchange model to prevent the confounding effects of miRNAs derived from tissues of suckling offspring . Wild type (WT) offspring consuming miR-375 KO milk at day 3 exhibited basal levels of miR-375 in milk from the stomach, which likely comes from stomach epithelial cells. At day 14 of suckling, it was confirmed that most miR-375/miR-200c came from the milk itself rather than the offspring www.efsa.europa.eu/publications 149 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

genotype, because there was no significant difference between WT pups and 375KO pups receiving WT milk. Evaluating different parts of the GI tract (jejunum, ileum and colon), no evidence of uptake was observed because there was no measurable increase in miR-375 levels in enterocytes of 375KO pups receiving WT milk compared to 375KO milk. Moreover, no changes in plasma, liver or spleen levels of miR-375 were observed in 375KO pups receiving WT or 375KO milk . Similar effects were observed at day 3 of treatment and upon evaluating miR-200c on day 14. When the fate of milkderived miRNAs downstream of the stomach was analysed, the levels of miR-375 in the intestinal content of 375KO pups receiving WT milk was seen to dramatically decrease compared to that of the stomach contents. These data suggest that milk miRNAs are being degraded by the digestive system (since no evidence of uptake was observed). Indeed, when the spike in cell-miR-39 was tested, it was degraded more rapidly than milk miR-375, suggesting that milk miRNAs exhibited a certain degree of resistance (perhaps because of their exosome content). But less than 10% of miR-375 copies remained after 2 h incubation of intestinal content, suggesting a possible digestive enzymatic degradation of the milk-derived (exosomes) miRNAs .

Observational studies also suggest that human breast milk miRNAs are not related to preventing atopic dermatitis in infancy 3 months postpartum . In general, RNA contamination has been proposed as the main source of controversy in miRNAs studies reported in the field of breast milk (Bagci and Allmer, 2016) .

The literature contains very few studies on the biological effects of dietary exogenous ncRNAs in the GI tract and its annex glands. Indeed, these few studies only report a biological effect in the liver. The available literature suggests that dietary exogenous plant small ncRNAs seem to resist the harsh conditions of the GI tract. However, the stability of plant miRNAs under GI conditions in vitro, ex vivo or in vivo is low. Moreover, these studies suggest that the percentage of surviving miRNAs in the GI tract in the best-case scenario is around ≈1%, although this depends on the specific small ncRNAs evaluated. Several studies report the presence of dietary exogenous miRNAs in breast milk of different mammals, including humans. Some protected miRNAs (i.e. transported in extracelluar vesicles) have been found to be stable under in vitro degradative conditions. While some evidence suggests a possible absortion through oral feeding, other studies indicates a very limited biovailability or lack of accumulation within the GI tract or its annex glands.

Recent evidence supports the significant contribution of sRNAs to communication between hosts and some eukaryotic pathogens, pests, parasites, or symbiotic microorganisms. This silencing transfer phenomenon is very relevant to food and feed risk assessment of ncRNA GM plants. Although this section (EFSA Task 3) will focus on studies related to the possible trafficking of ncRNAs between plants and humans and animals, studies supporting the movement of RNA-silencing signals between plants and pathogens are also reviewed.

Virus-induced gene silencing represents the model of gene silencing in the host by exogenous siRNAs. In this case, the siRNA specificity determinant is derived from viral RNAs. Typically, with a discrete length of 22 nt they form perfect duplexes which are produced in plants by DCL2 and DCL4. Virusinduced gene silencing is a very effective defence system, and, consistently with the normal dynamics of host-pathogen interactions, all viruses encode silencing suppressors as a counter defence (Baulcombe, 2015) .

www.efsa.europa.eu/publications 153 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Cross-kingdom RNAi implies that a translocation of gene silencing signals occurs between hosts and these organisms. This implies a two-way traffic of siRNAs between pathogens and their plant hosts. Several studies report that RNAs produced in a host plant can be transferred to a fungus or oomycete, symbiont or pathogen, to induce RNA silencing in the interacting microorganism (Ghag et al., 2014; Koch et al., 2013; Vega-Arreguín et al., 2014; Helber et al., 2011) . Transgenic plants carrying a RNAi construct, usually with a sense-intron-antisense palindromic structure that produces dsRNAs and siRNAs, can induce silencing of the targeted transcripts in the interacting organism. Furthermore, treatment of fungal conidia with dsRNAs of essential genes led to fungal growth inhibition, demonstrating RNA interference from environmental silencing signals (Koch et al., 2013) .

Conversely, the RNA produced in a fungus can affect a host plant's defence system (Weiberg et al., 2013) . Remarkably, scientists have developed an effective disease control strategy, called host-induced gene silencing (HIGS), by generating transgenic plants that express exogenous RNAi triggers to successfully silence essential genes in pathogens and pests (Weiberg et al., 2015) . In this context, transgenic plants expressing dsRNAs or hairpin RNAs targeting vital fungal genes of Fusarium sp.

developed resistance to these important phytopathogens (Ghag et al., 2014; Koch et al., 2013) . Small RNAs or dsRNAs can also be transferred from plant to pests, such as insects eating leaves or nematodes infecting roots. Indeed, transgenic plants that express dsRNA homologous to essential genes of insect pests or nematodes became resistant to these specific parasites through the silencing activity produced within the target organism when sRNAs expressed from the plant transgenes is consumed (Baum et al., 2007; Fairbairn et al., 2007; Mao et al., 2007) .

Many aspects of these cross-kingdom RNA interference phenomena are still poorly understood. One of them is how these RNA molecules 'travel,' sometimes over long distances through diverse cellular boundaries between plants and interacting organisms. These silencing signals may utilize conserved cell-to-cell as well as systemic RNAi pathways present in plants and animals, and may also use organismspecific pathways. RNA-protective factors such as AGOs, other RNA-binding proteins, or encapsulation into extracellular vesicles likely play important roles in protecting mobile RNAs against degradation during transport (Weiberg et al., 2015; Lefebvre and Lecuyer, 2017) . Another important question is how these sRNAs use the target cell RNA interference machinery to convey the silencing effect. Based on current knowledge, RNA-mediated gene silencing seems to be an ubiquitous phenomenon that exists in almost all eukaryotes, and which always follows the principle of complementary nucleotide base-pairing between regulatory sRNA and mRNA sequences (Weiberg et al., 2015) . Despite the tremendous differences present in the structural features of regulatory RNAs, and the completely unrelated or highly divergent RNA gene-silencing mechanisms and pathways that have evolved in diverse organisms, complementary sequence matches seem to be sufficient enough to trigger cross-kingdom gene silencing, as exemplified by plants-parasites, bacteria-to-worms and other lower organism interactions (Weiberg et al., 2015) . 

Following a literature search as described in section 2.2.1.2 and based on the methodology described in the section 2.2.2.3, a total of 33 documents were selected for review of the topic molecular mechanisms of exogenous ncRNA uptake and function. The uptake of exogenous ncRNAs is described from the molecular mechanisms point of view (i.e. the presence of a specific transporter, if any). EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Although plant-derived exogenous ncRNAs are the main topic of this subsection, some other examples of general molecular mechanisms of exogenous ncRNAs uptake, intracellular trafficking and function are included as relevant to understanding the possible fate of plant-derived exogenous ncRNAs. Other general aspects of RNAs uptake are reviewed in chapter 3.1.5, and the uptake of exogenous RNAs from the pharmaceutical/medicinal area is covered in section 3.1.4.

As previously mentioned in section 3.1.5, cellular uptake of ncRNAs presents great challenges due to the physicochemical properties of these molecules. Their size and negative charges prevent their easy entry into mammalian cells. Cell membranes pose a major barrier to ncRNAs entrance since entry into cells is tightly controlled and regulated (McErlean et al., 2016) . The anionic lipophilic bilayer prevents entry of macromolecular anionic nucleic acids in their naked form, both restricting their binding to and passive diffusion across these lipophilic cell membranes . Furthermore, even though their internalization may depend on the endocytic pathway, endosomal entrapment and lysosomal degradation can be major issues because of the reduced accessibility of ncRNAs to their sites of action (nucleus, cytoplasm or mitochondria) (Won et al., 2011; .

Many different carrier molecules have been developed to achieve the entrance of ncRNAs (mostly miRNAs and dsRNAs) into target cells (Bolhassani, 2011; Lindgren and Langel, 2011; Mao et al., 2010; Ragelle et al., 2014; Rudzinski and Aminabhavi, 2010; Shum and Rossi, 2016) . Exogenous ncRNAs must be carried by polymers, synthetic cationic lipids or cell-penetrating peptides neutralizing the negative charges of the nucleic acids. Most of these cationic lipids tend to form liposomes when dispersed in an aqueous phase, such as blood. Endocytosis has been suggested as the main pathway for this nucleic acids-cationic lipid complex internalization by the cell (El Ouahabi et al., 1997) . Most reports suggest that so-called cell-penetrating peptides (CPPs) bind initially to negatively-charged proteoglycans at the cell surface and are internalized into endosomes (Juliano et al., 2012) .

There are multiple pathways of endocytosis (see also section 3.1.5).

i. The clathrin-coated pit pathway is the archetype of endocytosis pathways. Cell surface receptors and their associated ligands interact with adapter proteins and accessory factors, clustering the receptors into specialized membrane areas subtended by a network of clathrin triskelions. The clathrin network is invaginated by means of membrane curvature promoting specialized proteins, giving rise to a clathrin-coated endosome. This endosome quickly uncoats, generating an uncoated vesicle that will begin its intracellular journey (Juliano et al., 2012) . ii.

The caveolar pathway implies the presence of small cell membrane invaginations rich in cholesterol and sphingolipids containing caveolin1. Cavins, which coat proteins helping to stabilize caveolar structures, are present in these invaginations. There is controversy as to whether caveolae generate independent intracellular vesicles or whether they remain as tubular structures linked to the plasma membrane; however, some data suggest that caveolae generate vesicles able to participate in intracellular traffic, generally presenting a smaller size than other forms of endocytotic vesicles (Juliano et al., 2012) . iii.

A number of clathrin-and caveolin-independent pathways have been described. These pathways are often defined in terms of the morphologies of the vesicles they generate or in terms of the cargo that is preferentially internalized. For instance, the flotillin pathway involves the presence of flotillin-rich membrane microdomains, where flotillins are membrane-inserted proteins that may be involved in ordering lipid domains and subsequent endocytosis, similar EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

to caveolin (Juliano et al., 2012) . Another such clathrin-and caveolin-independent pathway is the CLIC/GEEC pathway, which seems to be particularly important for fluid phase endocytosis. CLIC/GEEC stands for Clathrin and Dynamin Independent Carriers/GPI-AP Enriched Early Endosomal Compartments. This pathway gives rise to high-volume tubular endosomes rich in GPI-proteins and that typically contain fluid phase markers such as dextrans (Juliano et al., 2012) . Additional clathrin-and caveolin-independent pathways exist, some involving dynaminmediated disjunction of vesicles from the plasma membrane (Juliano et al., 2012) . iv.

Macropinocytosis is a process by which cell protrusions pinch off large volumes of extracellular fluid and is therefore an important pathway in fluid phase endocytosis. It is also involved in internalization of clustered, activated receptor tyrosine kinases. v.

Phagocytosis and entosis are large-volume internalization mechanisms as well. They come into play in specialized cells or unusual circumstances but do not play much of a role in oligonucleotides processing in most cell types (Juliano et al., 2012) . vi.

The actin cytoskeleton plays an important role in most of the endocytotic processes described above, although certain arenaviruses enter cells by a pathway independent of clathrin, caveolin, dynamin and actin. It has been reported that phosphorothioate antisense oligonucleotides seem to enter the cells by this pathway as well (Juliano et al., 2012) .

There have been many attempts to identify endogenous receptors for antisense or siRNA molecules. However, no direct evidence for their involvement in oligonucleotide trafficking has been provided (Juliano et al., 2012) . Integrins of the beta-2 subclass as well as scavenger receptors have been suggested as candidates, but this is controversial. Toll-Like Receptor (TLR) family members seem to be the most convincing examples of cellular receptors for oligonucleotides; TLR9 binds DNA having CpG motifs, TLRs7/8 binds single-stranded RNA, while TLR3 binds double-stranded RNA. Although these TLRs are usually found within endosomes rather than at the cell surface, in some cases they seem to be able to assist in accumulation of oligonucleotides by cells (Juliano et al., 2012).

One interesting candidate as a receptor for oligonucleotides is the mammalian homolog of the doublestranded RNA (dsRNA) transport protein SID-1 found in Caenorhabditis elegans (Feinberg and Hunter, 2003) . The human homolog of this protein, SIDT1, has been described as facilitating rapid contactdependent intercellular small RNA transfer (Elhassan et al., 2012) and as selectively binding long doublestranded RNA (Li, W et al., 2015) . Another member of the SID1 transmembrane family, SIDT2, has been shown to take up extracellular double-stranded RNA in Drosophila S2 cells (McEwan et al., 2012) via endocytosis, although in mammalian cells this protein has been located in lysosomal membranes (Jialin et al., 2010) . In fact, some authors have proposed that SIDT2 could mediate cellular RNA degradation inside lysosomes through a novel type of autophagy called RNautophagy (Aizawa et al., 2016) . Another research group recently reported that SIDT2 is required to transport internalized dsRNA from endocytic compartments into the cytoplasm for immune activation (Nguyen et al., 2017), while others have described that this protein mediates "gymnosis", facilitating uptake of naked single-stranded oligonucleotides into living cells (Takahashi et al., 2017) . A more recent report indicates that both SIDT1 and SIDT2 not only do not transport RNA, but are involved in cholesterol transport (Mendez-Acevedo et al., 2017) .

Once an oligonucleotide has entered a cell in an endosome, it encounters a complex maze of intracellular pathways leading to multiple destinations and regulated by an intricate protein machinery. Key subcellular membrane bound compartments include early and recycling endosomes, late www.efsa.europa.eu/publications 156 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

endosomes/multi-vesicular bodies, lysosomes, the Golgi apparatus and the endoplasmic reticulum. Intracellular trafficking is not a random process, but rather a carefully orchestrated choreography that allows the cells to transport endogenous and exogenous materials to the most appropriate places. For nucleotides to exert their function, they need to leave the membrane bound compartments and access the cytosol and/or nucleus (Juliano et al., 2012) . Knowledge as to how endocytotic cargos are delivered to subcellular compartments is still partial. Many of the internalization pathways previously described converge at the stage of early endosomes, raising the question of how the differentially internalized components traffic to different destinations. Certain evidences suggest that membrane domains originating from different internalization pathways maintain their identity within early endosomes, thus providing the means for specific sorting and trafficking to distinct downstream destinations (Juliano et al., 2012) . All membrane traffic proceeds through the same basic steps:

i) a coated vesicle is pinched off from a larger donor membrane compartment; ii) the vesicle uncoats, allowing display of the tethering and fusion proteins; iii) the vesicle is carried to its destination along "tracks" provided by actin-or tubulin-based cytoskeletal structures; iv) the vesicle recognizes its target membrane compartment using tethering proteins and then utilizes SNARE proteins to complete the fusion process and deliver membrane and contents to the target compartment (Juliano et al., 2012) (Figure 16 ).

ncRNAs must escape from the endomembrane compartments and reach the cytosol to exert their function. As mentioned above, intracellular trafficking involves a highly dynamic flux of membrane vesicles engaged in a multitude of fusion and disjunction events. There are a few key points in these processes to be considered when designing non-viral oligonucleotide delivery strategies:

1) fusion involves localized stress on the fusion partners, including formation of non-bilayer lipid domains; 2) non-bilayer regions of membranes can be much leakier than bilayer regions; 3) many enveloped viruses fuse with cells via specialized membrane interacting proteins that, while differing in sequence, act in a manner similar to cellular SNARE proteins; in many cases these proteins can also induce increments in membrane permeability. Therefore, there is an intrinsic relationship between the fusion events inherent to intracellular trafficking and transient leakage of vesicular contents. Thus, the innate activity of oligonucleotides taken up by cells is likely due to a modest amount of continuous leakage from endomembrane compartments spontaneously occurring during intracellular trafficking (Juliano et al., 2012) .

Some ncRNAs function is exerted in the nucleus, although nuclear entry may not be the rate-limiting step for oligonucleotide action. Studies have shown that oligonucleotides, particularly those with phosphorothioate backbones, are able to continuously shuttle between the nucleus and cytoplasm. This is an active process mediated by nuclear pore structures but does not require classic nuclear localization signals. For phosphodiester oligonucleotides, both passive diffusion and active transport have been described as nuclear entry mechanisms (Juliano et al., 2012) .

In terms of the uptake and trafficking of "free" or "naked" oligonucleotides, the co-existence of both productive and non-productive uptake routes has been suggested. The non-productive pathway seems to involve trafficking to lysosomes, while the pathway that results in RNase H-dependent antisense effects eventually leads to interaction with cellular pre-mRNA (Koller et al., 2011) . Other studies have involved so-called "gymnotic" uptake of antisense oligonucleotides modified with LNA (locked nucleic acid) moieties (Stein et al., 2010; Zhang et al., 2011) . Subcellular distribution studies surprisingly suggested that the deoxy LNA compounds became associated with P-bodies that are usually thought to be siRNA action sites (Stein et al., 2010) . Unlike the study performed with phosphorothioate www.efsa.europa.eu/publications 157 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

oligonucleotides in which the concentrations used were in the nanomolar range, the antisense effects of "naked" LNA required micromolar concentrations (Juliano et al., 2012).

As described in section 3.1.5, exogenous ncRNAs must overcome a series of barriers to be functional at systemic level. In addition to the extracellular barriers specifically related to oral absorption, defensive humoral and cellular barriers efficiently prevent the intrusion of exogenous entities into the organism (Jeong et al., 2007) . Systemically-administered nucleic acids are rapidly degraded by nucleases in a few minutes and, even in nucleic acid drug therapy using positively-charged polyplexes, they readily interact with serum components to form larger aggregates, resulting in rapid clearance by RES or phagocytes (Jeong et al., 2007) . If the nucleic acids survive in the blood stream, they must still reach the appropriate tissues and enter the target cells. In systemic nucleic acid therapy, continuous endothelial walls in the microvasculatures are one of the main barriers limiting the access of oligonucleotides to target cells. In certain pathological cases, such as cancer, the presence of leaky vasculatures in the vicinity of highly penetrating solid tumours allows nanoparticulates to penetrate and accumulate in tumours due to the enhanced permeation and retention effect (Jeong et al., 2007) . Caveolae-mediated transcytosis has been considered one of the important mechanisms of macromolecules transport across the endothelium (Schnitzer, 2001) . Another possible limiting barrier is the extracellular matrix consisting of various proteoglycans, which are proteins covalently cross-linked with carboxylic or sulphated glycosaminoglycans (GAGs) (Ruponen et al., 2003) . Since GAGs, such as heparin sulphate, chondroitin sulphate and hyaluronic acid, are polyanions, they may repel the negatively-charged oligonucleotides, affecting mobility of the nucleic acids in the tissue extracellular matrix and therefore limiting their access to target cells. In the case that oligonucleotides reach the target cell membrane, they must enter the cell. The first barrier is the anionic plasma membrane itself, and the second is cellular uptake via endocytosis (see section 3.1.5). Without a specific ligand-receptor interaction mechanism, the entrance of ncRNAs results in a lack of cell specificity (Jeong et al., 2007) . Once taken up by cells, the nucleic acids are localized within the endosomal compartments, where the pH rapidly drops to about 5 by the action of membrane-bound ATP-driven proton pumps. The endosomes mature to lysosomes where the oligonucleotides can be degraded by various enzymes (Jeong et al., 2007) . As previously indicated, any surviving nucleic acids must then escape from the endosome to exert their functional effects. In nucleic acid drug therapy, certain endosomal disruptive agents, such as lysomotropic chloroquine, fusogenic peptides and pH-sensitive neutral lipid, are used to facilitate or promote endosomal escape through membrane disruption (Jeong et al., 2007) . After endosomal escape, ncRNAs should move through the cytoplasm to encounter their molecular targets, or to the peri-nuclear space where nuclear translocation takes place if the nucleus is the target compartment. Passive diffusion of high-molecular weight macromolecules is limited in the cytoplasm. A complex network of microfilaments and microtubules, highly concentrated proteins and various subcellular organelles provide the cytoplasm with a high fluid phase viscosity and mesh-like structure with 300 to 400 Å pores from, making the diffusion process size-dependent (Jeong et al., 2007) . In fact, microinjection of plasmid DNA into the cytoplasm showed that normal DNA was practically immobile in the cytoplasmic space (Dowty et al., 1995) . However, DNA molecules with less than 250 base pairs seem to undergo free diffusion in the cytoplasm (Lukacs et al., 2000) , suggesting that small oligonucleotides (i.e. miRNAs) do not face limitations regarding their intracellular trafficking, contrary to large ncRNAs (i.e. lncRNAs) which might encounter more restrictions reaching the peri-nuclear space. Finally, some ncRNAs need to enter the nucleus to be functional. The nuclear envelope consists of a double-layer membrane with nuclear pores. The nuclear pore complex allows free diffusion of ions and small molecules, but restricts passage of macromolecules with molecular EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. weights greater than 60 kDa, unless accompanied by a nuclear localization signal peptide (NLS) (Jeong et al., 2007) , which is not the case with lncRNAs Oligonucleotides are initially accumulated in an endomembrane compartment (the DONOR compartment, e.g. early endosomes) and are then trafficked by shuttle vesicles to various other endomembrane compartments (the RECIPIENT compartment, e.g. the trans-Golgi). The first step (1) involves disjunction ('pinching off') of a shuttle vesicle under the influence of a coat protein as well as other accessory proteins. At this stage there are non-bilayer regions at the junction between the membranes of the DONOR compartment and the shuttle vesicle. This provides an opportunity for some oligonucleotides to escape into the cytosol.

Step 2 involves uncoating of the vesicle; and Rab proteins can contribute to this step.

Step 3 comprises movement of the shuttle vesicle toward its destination along cytoskeletal tracks. Motor proteins such as various myosins (for the actin system) or dyneins or kinesins (for the microtubular system) propel the vesicle. Rab proteins are involved in forming the appropriate linkages to the cytoskeleton.

Step 4 entails recognition of the RECIPIENT ('target') compartment by the shuttle vesicle. Tether proteins work with Rab proteins to provide interaction specificity while v-SNARE proteins in the vesicle membrane interact with t-SNARE proteins in the RECIPIENT compartment membrane to provide firm bridging, as well as contributing to specificity. In step 5 the SNARE proteins undergo major conformational changes, and, with the assistance of accessory proteins, trigger fusion of the shuttle vesicle membrane with the membrane of the RECIPIENT compartment. At this stage non-bilayer regions exist at the junction between shuttle and RECIPIENT membranes potentially allowing oligonucleotide escape. Reprint with permission from (R L Juliano, X Ming, O Nakagawa. Cellular uptake and intracellular trafficking of antisense and siRNA oligonucleotides. Bioconjugate Chemistry 23:147-157). Copyrigth (2012) American Chemical Society. Figure 16 : Proposed mechanism of vesicular trafficking of oligonucleotides Some research has been done on cellular uptake, trafficking and tissue distribution of oligonucleotides without specific targeting or carrier mechanisms. In terms of tissue distribution, certain cell types exhibit a preferential in vivo uptake, particularly kidney proximal tubule cells and liver Kupffer cells for phosphorothioate (PS) antisense compounds and siRNA (Juliano et al., 2012) . However, only one report has studied systemic level uptake and function of orally administered miRNAs . These authors studied intestinal uptake of maternal milk-derived miRNAs by new-born mice employing genetically-modified models to distinguish endogenous miRNAs from milk-derived exogenous miRNAs. Their analysis of the intestinal epithelium, blood, liver and spleen revealed no evidence for miRNA www.efsa.europa.eu/publications 159 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

uptake, even though milk in lactating mothers has been shown to be a particularly rich source of secreted miRNAs. Sequencing and microarray analysis of miRNAs in the milk of various mammalian species has led to the discovery of hundreds of miRNAs in cow, pig, rat and human milk, which are derived from exosomes and cellular components. Furthermore, their study revealed rapid degradation of milk miRNA in intestinal fluid, indicating that miRNAs in the milk play a nutritional rather than a generegulatory role in new-born mice .

There are certain tissues containing special physical structures that can function as additional barriers to ncRNAs entry. Such is the case of the central nervous system (CNS), where delivery of nucleic acids is particularly challenging because of its anatomical and physiological complexities. The CNS is protected by the blood-brain barrier (BBB), which consists of tightly joined capillary endothelial cells, restricting access of large molecules into the brain. The CNS extends to include the spinal cord, which is protected by the blood-cerebrospinal fluid barrier (BCSFB), made up of choroid plexus epithelial cells limiting the free diffusion of molecules into the cerebrospinal fluid (CSF). The BBB and the BCSFB express numerous transporters and receptors which contribute to the transfer of essential nutrients such as glucose and amino acids into the CNS. Within the CNS, different cell types exist including neurons, astrocytes and other glial cells. Entry of nucleic acids into neurons is notoriously difficult for unclear reasons, although it is likely related to their post-mitotic nature, as well as the complex structures and intricacies of neuronal networking (O'Mahony et al., 2013) . For instance, variances in the uptake of siRNA lipoplexes (liposome and nucleic acid complexes) were reported at different parts of the neuron structure, with greater uptake efficiency at the cell soma compared to the neuritis, which may occur because of different membrane compositions in these domains (O'Mahony et al., 2013) . Circumventing the BBB is only possible by direct administration of ncRNAs to a target region of the brain or by administration to the spinal cord where trans-synaptic retrograde transport to the brain is possible. Some delivery systems exploit the receptor-mediated uptake of molecules such as transferrin, lactoferrin and insulin receptors by attaching a ligand for the receptor to the non-viral delivery system, or promoting a transient mechanical disruption of the BBB (O' Mahony et al., 2013) . In certain neurological diseases, and in particular in brain cancers, anatomical and physiological changes occur that can impact BBB integrity and the CNS accessibility. Delivery to the CSF to bypass the BBB presents its own obstacles, including rapid clearance from the CSF, the BCSFB and limited diffusion from the CSF into the brain parenchyma (O'Mahony et al., 2013) (Figure 17) . Another specialized barrier is the placenta. The placenta serves as the interface between the maternal and foetal circulatory systems, and regulates the transfer of oxygen, nutrients and waste products. When exogenous substances are present in the maternal bloodstream, the extent to which these affect the foetus is determined by transport and biotransformation processes in the placental barrier. This barrier is formed by fusion of the outer blastocyst trophoblast cells facing the uterine epithelium into multinucleated syncytiotrophoblast. Proliferation of syncytiotrophoblast is mediated by the fusion of precursor cytotrophoblast cells. Maternal blood in the intervillous space and foetal blood in the villous capillaries are separated by a continuous layer of syncytiotrophoblast, a discontinuous layer of cytotrophoblast cells, basal lamina, connective tissue, and foetal endothelial cells (Al-Enazy et al., 2017) . The placenta undergoes several changes as pregnancy progresses. Overall, the placental barrier becomes thinner over time, the distance separating the maternal circulation from the foetal circulation decreasing from 50-100 µm in the second month to only 4-5 µm at term. Passive diffusion across the placenta is favoured for lipophilic substances, allowing them to cross the phospholipid bilayer. Smaller compounds tend to cross the placenta more readily, with compounds having a molecular weight less than 500 Da crossing it most easily (Al-Enazy et al., 2017) . Since nucleic acids are polarized, negatively charged molecules, it is unlikely they would be able to cross the placenta by passive diffusion. Nevertheless, the placental syncytiotrophoblast is capable of EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

endocytosis (Al-Enazy et al., 2017) . No reports refer to exogenous ncRNAs entering the placental barrier. However, the presence of placental miRNAs has been described in the blood circulation of pregnant women. These miRNAs have been reported to be actively secreted into microvesicles and exosomes. These specific exosomal miRNAs may be important for embryo implantation, playing a significant role in the intercellular communication pathways that potentially contribute to placentation and development of maternal-foetal vascular exchange (Mitchell et al., 2015) . Since exosomes can cross the blood brain barrier, meaning they can cross several layers of cells, it has been speculated that exosomes could directly move from maternal to foetal tissues, and vice-versa, during pregnancy (Record, 2014).

The 

The molecular mechanisms of exogenous ncRNA cellular uptake have been inferred from studies performed when developing strategies for nucleic acid delivery in therapeutics. Few studies have been done with "naked" oligonucleotides, and even in these cases, the oligonucleotides were chemically modified. These molecular mechanisms of cellular uptake include the different types of endocytosis mechanisms known to date. The existence of receptor-mediated uptake of oligonucleotides through the SIDT1 and SIDT2 proteins has also been described, although their function has been contradicted since they have been described as cholesterol transporters. Among the numerous challenges ncRNAs must overcome once taken-up by the cell, escaping the endosomal compartments is the most relevant and significant, although reaching their site of action is also crucial to exerting their function. Specialized barriers such as the blood-brain barrier or the placental barrier present further obstacles to overcome.

www.efsa.europa.eu/publications 163 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Zhang Y, Qu Z, Kim S, Shi V, Liao B, Kraft P, et al. 2011 . Down-modulation of cancer targets using locked nucleic acid (LNA)-based antisense oligonucleotides without transfection. Gene Ther 18, 326-333. Zhou G, Zhou Y, Chen X. 2017 

Following a literature search as described in section 2.2.1.2. and based on the methodology in section 2.2.2.3., a total 20 documents were selected for reviewing the landscape of exogenous RNAs in biological fluids of human and animals. Using public databases of RNA sequencing datasets, several documents report the application of in silico bioinformatics analysis (Table 25 ) clearly describing the widespread presence of exogenous ncRNAs in biological fluids(from multiple sources, including diet) Wang et al., 2013; Beatty et al., 2014; Yeri et al., 2017; . However, four documents suggest that exogenous ncRNAs present in the public datasets may represent contamination from the laboratory environment since their abundance is extremely low, or they are of non-dietary origin, or could originate from technical artefacts or contamination Bagci and Allmer, 2016) . Five documents report the development of bioinformatics resources to either predict transportable miRNAs, predict functional analysis, or list and compare the exogenous miRNAs found in samples Chiang et al., 2015; Zhang et al., 2016; Zheng et al., 2017) .

The generalized use of highly parallelized next generation (NexGen) sequencing technologies has promoted the use of sequencing to study complex biological systems, including human and animal biological fluids and tissues.

Initial studies evaluating the possible presence of exogenous ncRNAs in biological fluids were done in silico using public databases of RNA-seq results. Several studies report the presence of exogenous RNAs (i.e. dietary plant ncRNAs) in the biological fluids of humans and animals. Initial studies of the sRNAs spectra in human plasma showed that a small fraction of sequence reads from plasma mapped to human miRNAs (≈1.5%) or human transcripts and human genome sequences (≈11%) . This fraction increased by up to ≈60% when a higher sequence mismatch tolerance was allowed (under two mismatch allowance). Surprisingly, a significant number of the unmapped reads aligned with various bacterial and fungal sequences (i.e. ribosomal RNAs and tRNAs). Also, many processed reads mapped to common food items, the most abundant being RNA sequences from corn (Zea mays) and rice (Oryza sativa). Compared to data from a serum sample from a Chinese individual, the sequence abundance between corn and rice was reversed , suggesting the influence of dietary habits.

Other sequences from other foods in human plasma samples included soybean (Glycine max), tomato (Solanum lycopersicum), grape (Vitis vinifera) and others. Including exogenous miRNAs, various common household insects were identified . Treating plasma samples with DNAse, protease, Triton X-100, and additional RNase reduced the total number of reads compared to the addition of RNase alone, suggesting that some of the exogenous RNA molecules are probably associated to circulating protein and/or lipid complexes . In the same study, a lower percentage of exogenous sequences was observed in mouse plasma samples compared to human samples . Another study evaluating a larger number of human plasma samples (n=183 young male athletes) reported exogenous sRNAs from different bacteria, but very few sRNAs were consistently detected in all samples (Yeri et al., 2017) .

In a study on public sRNA datasets from various tissues from mammals, chicken and insects, plant miRNAs were found in 63 out of 83 datasets analysed, with miR-168 being extremely over-represented EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. . In 19 datasets (2 humans, 14 mice, 1 pig and others) the plant miRNAs reads ranged from ≈0.053-0.456% of the total animal miRNA reads. Insects fed with plant foods containing miR-168, did not show accumulation of miR-168 when evaluated by Northern-blot. However, in a very few samples sequencing data detected plant miRNAs not present in the food source, including miR-1507, in addition to miR-168 at a moderate read, suggesting that plant miRNAs observed in some public sRNA databases may be artefactual and due to sequencing methodology . Plant miR-168a from monocots was also found to be the most abundant exogenous plant miRNA, in most cases, in other sRNA sequencing datasets studies with levels typically <1% of human miRNAs . But when analysis was performed in the absence of any known sources of plant contamination, plant miRNAs were not detected. This suggests that precaution should be taken to prevent cross-contamination of the samples due to the high sensitivity of next-generation sequencing methods .

Other human public sRNA sequencing datasets have also been questioned for the presence of exogenous plant miRNAs . Of 410 analysed human plasma samples, 1301 plant miRNAs were found when compared to the genome of 5 different plant model organisms (Arabidopsis thaliana, Triticum aestivum, Oryza sativa, Zea mays, and Brachypodium distachyon). At least one read was present in selected samples. Peu-miR-2910, found in all 410 samples, is a singular miRNA found in selected samples with more than 1000 counts. miR-2910 is found in Zea mays and conserved in fruits and vegetables including melon, sorghum, tomato, tea and oil palm. Other highly-expressed plant miRNAs found in up to hundreds of copies in selected samples include peu-miR-2916 (found in 379 samples), peu-miR-2914 (found in 359 samples), and tae-miR-2018 (found in 353 samples) . Plant miR-159 (found in 32 samples) and miR-168a (found in 7 samples) were detected in relatively small amounts (less than 4 copies). Using bioinformatics tools, the same study showed that seed sequences of miR-2910 and miR-2916 are similar to hsa-miR-4259 and miR-4715-5p for the former and has-miR-4652-5p for the latter, and thus potentially target several gene targets of their human miRNA counterparts .

Other studies have also reported in mouse plasma samples the presence of exogenous RNAs matching exogenous species including bacteria, fungi, viridiplantae and food items . These included miR-166d and miR-168a (corn or rice as the most likely source), and miRNAs from worms and insects . Abundant non-human RNAs were found in other healthy individuals human plasma samples (n=3) assigned to metazoan, bacteria or fungi (Beatty et al., 2014) . Several exogenous sRNAs were found that could potentially derive from food items including Fragaria vesca, Medicago truncatula, Lotus japonicus, Glycine max, Arabidopsis thaliana, and Solanum lycopersicum (Beatty et al., 2014) .

Twelve public sRNA sequencing databases were analysed for the presence of plant exogenous miRNAs. Breast milk exosomes from humans (4 samples, 86.37 million reads) and pigs (8 samples, 179.9 million reads) were analysed. Although 17 plant miRNA species belonging to 11 miRNA families (6 samples) were found, they showed low abundance (<25 read counts for the highest expressed miRNA). zma-miR-168a, zma-miR-156a, ath-miR-166a, ath-miR-319b and ptc-miR-319d showed the highest abundance levels, while in humans only 2 samples contained 35 plant miRNAs from 25 miRNA families (<194 read counts, for the highest expressed miRNA, average of 2 samples). ath-miR-166a, pab-miR-951, pc-miR-472a, bdi-miR-168, aly-miR-167d, osa-miR-444b.2 and zma-miR-156a showed the highest abundance level in human samples . However, Bagci and Allmer reanalysed the data from Lukasik and Zielenkiewicz study and found that exogenous plant-derived miRNAs in milk are likely to be contamination (Bagci and Allmer, 2016) . In a later observational study, Lukasik et al. evaluated plant miRNAs miR-166a, miR-156a, miR-157a, miR-172a and miR-168a and reported the presence of plant miRNAs in human (n=6) breast milk, both in whole milk and exosomes . Out of 5 miRNAs, only 2 were found in exosomes (miR-168a and miR-156a). While the human has-miR-148a was found in all samples in a concentration range of 6.5-50 pM for whole milk EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. and 1.2-12.5 pM in exosomes fractions, plant-derived miRNA miR168a was found in all samples at a concentration of 300-700 fM (≈100-fold higher than other plant miRNAs) . Compared to endogenous miRNAs, the levels of exogenous plant miRNAs were rather low and varied significantly between the samples .

Kang et al. analysed public sequencing sRNAs datasets from various human tissues and body fluids (n=824), including 215 from serum and plasma samples, 37 from exosomes, 180 from cerebrospinal fluids, 93 from liver, 197 from blood cells and 102 from brain . Of note is that exogenous miRNAs (xenomiRs) were only found in 17% of the tissue samples (from 392 total samples), with several studies being completely void of any xenomiRs. XenomiRs were absent from most human tissues samples, or present at very low abundances (with a median of 3.5 read counts). No enrichment of xenomiRs was found in tissues relevant to dietary intake (i.e. hepatocytes and blood cells). By contrast, out of 432 human body fluids datasets, xenomiRs were found in 69% of samples, although at very low levels (median of 5 read counts). Even though the blood brain barrier separates the cerebrospinal fluid from the bloodstream, xenomiRs were found in comparable quantities in both , suggesting no depletion (as expected by the presence of the barrier) of xenomiRs in body fluids separated from the main bloodstream. Moreover, the lack of co-occurrence of xenomiRs in cerebrospinal fluid and serum of the same individuals, do not support that xenomiRs detected in cerebrospinal fluid samples have entered through the bloodstream, and thus do not support a dietary origin of xenomiRs. The most common xenomiRs in body fluid samples belonged to the clades for rodents (34%), dicots (29%), insects (13%) and euphyllophytes (11%), suggesting that xenomiRs composition does not reflect human food sources. From 10 billion sequence reads (all samples), xenomiRs were in low abundance, comprising only 0.001% of the total reads (40,055). The same study reported a controlled feeding trial in rats and piglets (see section 3.3.3 below for details). Shu et al. (2015) used bioinformatics tools to characterize the structural characteristics of exogenous miRNAs that can contribute to cross-species transportation and thus be transferred into human circulation . Thirty-four thousand (34.000) miRNAs sequences were compiled from 194 species (from five kingdoms including Animal, Plantae, Fungi, Protista and Viruses), including dietary miRNAs (5217 miRNAs) from 15 types of common food species such as cow milk, breast milk, tomato, grape, and apple. Using a support vector machine (SMV)-based feature elimination strategy, 1102 features based on sequence, structure, and physicochemical properties were analysed to distinguish human circulating miRNAs from the remaining sequences. Predictions were made for 221 features (categorized into 8 groups) to better predict transportable miRNAs in human circulation. Known human plasma miRNAs were ranked among the top of the list (dominated by Animalia origin). Only 14 dietary miRNAs were ranked among the top 500 miRNAs, five of which have a sequence identical to that in humans (bta-miR-487b, bta-miR-421, bta-miR-216, gga-miR-29a-3p, gga-miR-20b-5p) . While only one miRNA from Plantae ranked among the top 500 miRNAs, 30 Plantae miRNAs ranked in the top 1000 as possibly transportable , suggesting that animal-borne miRNAs may be subject to more significant absorption in humans than in plant miRNAs . It is important to note that this study of evaluated 26,705 Animalia miRNAs and 7645 Plantae miRNAs, whether this discrepancy in the original number of evaluated miRNAs may have biased the SMV-based classifiers to mammalian transportable miRNAs was not addressed.

In the same context, a database called "Dietary miRNA Database" that describes different dietary miRNA parameters and functional analysis has been described and is freely available. Additional databases for exogenous miRNA discovery using RNA-seq data are also available in the scientific community . Using bioinformatics tools (i.e. RNAHybrid database), other studies have also predicted the potential function of dietary miRNAs in the human body . In a bioinformatics study of 25 plant miRNAs, some similar functions between human and plant target genes were identified, such as ion transport and stress response . EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Analysing 563 public sRNA sequencing datasets from humans (382), mice (128), pig (6), chicken (4), fly (8), C. elegans (16) and many other nematodes, protozoa or prokaryota, Zheng et al. developed a database (Exo-miRExplorer) for exploring and comparatively analysing exogenous miRNAs (Zheng et al., 2017) . Two hundred and thirty-seven (237) plant-derived miRNAs were detected in 382 sequencing samples. Common miRNA families included miR-168, miR-156, miR-166, miR-167 and miR-172. miR-168a was the most frequently found miRNA, and was detected in 95 human samples from 11 different studies (Zheng et al., 2017) . The average abundance of miR-168a was 22,114.9 reads per million. In total, miR-168a was found in 155 samples from multiple species. miR-156a is another frequently observed exogenous miRNA from plants, and was detected in 65 samples with an average abundance of 86.69 read per million (Zheng et al., 2017) .

Regarding other biological fluids, exogenous RNAs of bacterial origin were found in 204 (urine) and 46 (saliva) samples from young male athletes analysed for sRNA-seq (Yeri et al., 2017) . Plant small RNA miR-2911 was detected in urine of mice fed honeysuckle decoction (Yang et al., 2015b) . miR-168a has been also detected in serum when honeysuckle was supplemented with 400 pmol synthetic miR-168a. miR-2911 and miR-168a were detected in the urine of mice that had received cisplatin treatment (cisplatin administration produces acute renal failure) and consumed a honeysuckle decoction alone or supplemented with synthetic miR-168a. The increased detection of plant small RNAs was attributed to disruption of the organization of small intestine epithelial cells produced by cisplatin treatment or longterm consumption (pre-fed) of honeysuckle (Yang et al., 2015b) , by mechanisms as yet unknown. miR-2911 levels in urine reached ≈260 fM.

In tissues, the overall percentage of exogenous sequences in human lung tissue (normal lung RNA samples), was found to be less than 1% . In spermatozoa samples, using public sRNA sequences datasets, Tosar et al. found high amounts of exogenous sequences in three samples . However, in their own analysis of three spermatozoa samples they found no detectable traces of any plant miRNAs, suggesting that exogenous miRNAs present in the public datasets may have been caused by laboratory environment contamination .

In terms of the possible passage of plant miRNAs through the brain-blood barrier, Kang et al. analysed 180 samples from cerebrospinal fluid (CSF) (from one study), which is separated from the bloodstream by the brain-blood barrier, and 102 brain samples (from 7 studies) . In general, xenomiRs were absent from most of the human tissue samples, and, when present, they were found at very low abundance (17% of analysed tissues). Although the brain-blood barrier would relatively protect the brain from dietary molecules, xenomiRs were present in similar fractions in the exposed liver and in the relatively protected brain samples , suggesting, according to the authors, a nondietary origin . xenomiRs were present in more body fluids samples (≈69% of samples) than tissue samples (≈17% of samples) and were identified in comparable proportions in the serum and CFS. This study does not provide support for the dietary origin of xenomiRs (including plant exogenous RNAs) meaning it may originate from technical artefacts . EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. In addition to miRNAs reported by The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Very few studies have reported the influence of dietary or physiological factors on levels of exogenous ncRNAs in biological fluids. In a small study (n=3 per group), Wang and colleagues compared some abundant exogenous miRNAs species identified in plasma samples from healthy subjects, colorectal cancer and ulcerative colitis patients . Although ulcerative colitis patients apparently had increased levels of circulating miRNAs of certain species (i.e. from a Bantam family), a large variability of miRNAs levels were observed among subjects of the three groups , suggesting that health/disease status may influence the presence of circulating miRNAs in human plasma.

Tissue injury produced by excessive consumption of common pharmaceutical drugs can also influence exogenous RNAs levels in plasma . Wang et al. showed that acetaminophen overdose in a mouse model produced liver injury, and an increase of RNA concentration in plasma compared to the control group was observed (from ≈200 to ≈600 ng/ml of total RNA 8h after treatment). Plasma ncRNAs levels were affected by acetaminophen overdose, both endogenous and exogenous. Plantderived exogenous miRNAs levels were reduced, and this drop in plant RNA was related to the observation that mice lose appetite after acetaminophen injection , suggesting the influence of diet on plasma plant-derived miRNA levels.

Levels of exogenous RNAs can be affected in case of drug-induced kidney damage. Sera and urine from mice treated with cisplatin, a chemotherapeutic agent also producing small intestines epithelial cells disruption, showed increased plant exogenous ncRNAs (i.e. plant miRNAs), not observed in untreated mice (Yang et al., 2015b) . The effect was not observed in a glycerol-induced model of acute renal failure (which does not affect small intestine organization). Detection levels of exogenous plant miRNAs, including miR-2911 and miR-168a, were below background levels in mice pre-fed with chow diet devoid of honeysuckle. However, when mice were pre-fed with honeysuckle and honeysuckle was administered (alone or miR-168a-supplemented) dose-dependent levels of miR-168a was detected in serum and elevated serum levels of miR-2911 were observed (Yang et al., 2015b) . The study shows that long-term honeysuckle feeding potentiates absorption of dietary miRNAs through a mechanism that likely differs from cisplatin treatment, suggesting that particular diets can modify the capacity to absorb small RNAs, and that altered or damaged guts resulting from illness and/or therapeutic treatment may enhance dietary RNA uptake (Yang et al., 2015b) .

While comparing the presence of exogenous sRNAs in three human plasma samples, Beatty et al. reported that the largest proportion of reads matching plant sequences were found from one individual who reported following a vegetarian diet (Beatty et al., 2014) , suggesting that the diet can influence levels of exogenous sRNAs of food origin. In another study comparing human breast milk for miRNAs derived from plants, no significant differences were observed for exosomal breast milk plant-derived miRNAs (miR-168a and miR-156a) between vegetarian or non-vegetarian volunteers . In the whole milk an increase in plant miRNA miR-168a was observed in non-vegetarian versus vegetarian diet volunteers .

Although some studies (see above) have reported the presence of exogenous RNAs in human and animal circulation and tissues, very few have focused on quantitative data relevant to risk assessment. Wang et al. estimated the concentration of RNA in plasma to be <0.3 pM based on an average of 100 ng/mL of RNA in plasma with an average of 100 nt in length and a selected RNA (i.e. miRNA) sequence representing less than 0.01% of the total RNA population. After 7 days feeding of herbals, Yang et al. reported an increase of miR-2911 in the sera of mice from 25 fM (5-fold higher than control chow-fed animals) following ginseng intake to 207 fM (39-fold higher than control chow-fed animals) following honeysuckle intake . The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

The lack of enrichment of plant-derived small ncRNAs in samples to be analysed by RNA-sequencing (i.e. through periodate oxidation) might also influence the low enrichment of these exogenous sequences in biological fluids. However, although 2'-O-methyl modification of synthetic RNA oligonucleotides 3'-ends can affect ligation activity of the T4 ligase in sRNA library preparation and may potentially result in sRNA sequencing bias and under-representation of sRNA with 2'-O -methyl 3'-ends in quantification experiments (Munafo and Robb, 2010) , others suggest a broad capability to detect plant miRNAs during the sequencing procedure of small RNA library construction .

The available literature clearly suggests the widespread presence of exogenous RNAs from multiple species, including plants, microbiota and many others, in human and animal biological fluids. Evidence for plant-origin ncRNAs, possibly relevant to risk assessment of ncRNA-based GM, suggests that plantderived miRNAs can be found in the biological fluids of humans and animals. Whether these sequences are directly derived from dietary sources has not been clearly demonstrated. Their overall low abundance, lack of enrichment in tissues most exposed to dietary changes, and their high variation between samples suggest that some may have originated as technical artefacts or contamination, as indicated by some authors. Although few reports were available in this area, recent evidence also suggests that the levels of exogenous plant ncRNAs in biological fluids may be influenced by diet, subject physiological or pathological conditions, or certain therapeutic treatments.

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. 

Following a literature search as described in section 2.2.1.2. and based on the methodology described in the section 2.2.2.3., a total of 48 papers were selected as being relevant to review of the systemic effects of dietary exogenous ncRNAs. Very few of these studies (14) provide supporting evidence (in favour) that dietary exogenous ncRNAs can be detected in biological samples and/or exert a biological effect after dietary ingestion. By contrast, 10 studies provide contradicting evidence (against) that dietary exogenous ncRNAs are either bioavailable or exert a systemic biological effect. One study provides evidence of the possible transfer of small ncRNAs through the mammalian placenta . In addition, following a literature search as described in section 2.2.1.2. and based on the methodology described in the section 2.2.2.3, 12 papers were selected as relevant to review toxicological effects of dietary exposure to exogenous ncRNAs in humans or animals.

The first report of systemic effects of dietary exogenous miRNA in animals was published in 2012 by . Plant miRNAs were detected in human Chinese healthy donors, among which miR-156a and miR-168a were highly expressed. These miRNAs were also detected in calves and mice. Plant miRNA levels were relatively lower, but at a similar concentration range as endogenous mammalian miRNAs present in serum. Periodate oxidation, which oxidize miRNAs but not plant miRNAs (which are 2'-O-methyl modified), confirmed that these miRNAs were genuine plant miRNAs as most mammalian miRNAs in human sera were oxidized and failed to be sequenced by Solexa . The plant miRNAs miR-168a and miR-156a were also detected in different mice tissues, while plant miR-166a was not, although it was present in the sera. Both miR-168a and miR-156a were primarily detected in microvesicles in C57BL/6J mouse plasma. Foods were found to contain the plant miRNAs miR-168a, miR-156a and miR-166a, including a chow mice diet at a concentration of 0.43, 0.54 and 0.66 fmol/g, respectively. Fresh rice contained between 3 to 10-fold higher amounts than the chow diet. Plant miRNAs were detected in other foods including Chinese cabbage, wheat and potato. Plant miRNAs were found to be stable in cooked foods. Plant miR-168a and miR-156a were higher in the sera and liver of mice after 6h feeding with fresh rice. miR-168a also increased in the stomach and small intestines, but not in the kidney, of mice fed a chow diet or fresh rice. Mice fed with total RNA extracted from rice or synthetic miR-168a (methylated) showed an increased level of miR-168a in serum and liver 3h after feeding. In vitro stability assays suggested that plant miRNAs had a much slower degradation rate than their synthetic form (without 2'-O-methylated 3'ends), suggesting that methylation has a protective effect on the stability of plant miRNA . Of note EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

is that miR-168a targets the liver-enriched gene low-density lipoprotein receptor adapter protein 1 (LDLRAP1), which plays a role in facilitating the removal of LDL from the circulatory system (Garcia et al., 2001) . Direct binding of miR-168a to exon 4 of LDLRAP1, which is in the ORF (open reading frame), was demonstrated by luciferase reported assays. Transfection of both pre-miR-168a and mature miR-168a resulted in decreased LDLRAP1 protein expression (but not that of mRNA) in HepG2 cells. Moreover, transfection of the intestinal epithelial Caco-2 cell line with miR-168a produced microvesicles with plant miR-168a which can be taken up by HepG2 and reduce LDLRAP1 protein expression. miR-168a seemed to be associated to AGO2 protein and other cells like 293T cells which were also able to package miR-168a in microvesicles and then deliver miR-168a into HepG2 . Injection (iv) of antisense oligonucleotide against miR-168a into mice during fresh-rice feeding reduced miR-168a levels in the liver and reversed LDLRAP1 repression. Mice were finally fed with fresh rice for 7 days to evaluate the physiological function of food-enriched in plant miR-168a. Serum and liver miR-168a levels were induced already at 1 day of feeding and the liver LDLRAP1 protein was repressed after 1, 3 and 7 days. Consequently, LDL levels in mouse plasma were significantly elevated on days 3 and 7 after fresh rice feeding. Administration of an antisense oligonucleotide against miR-168a reduced plasma levels of miR-168a, repressed reduction of LDLRAP1 in the liver and blocked the rice-induced elevation of plasma LDL levels, suggesting that elevation of fresh-rice derived miR-168a in the mouse liver specifically decreased liver LDLRAP1 expression and that this caused elevated LDL levels in mouse plasma (Figure 18 ). This study provides the first overall evidence of cross-kingdom regulation of gene expression through dietary plant miRNAs. The authors further tested a mature mammalian miR-150 added to the chow diet for mice, and reported an increase in serum and liver levels of miR-150 after three days , leading to downregulation of liver c-Myc expression . The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

By analysing Western human sera samples from 42 female stage II-III breast cancer patients, Chin et al. detected multiple plant miRNAs from Arabidopsis thaliana and soybean (Glycine max), but with sparse counts (Chin et al., 2016) . Among the detected plant miRNAs, ath-miR-159a, gma-miR-159a-3p and gma-miR-159e-3p had the highest counts (≤37 read counts) and were over 6 times more abundant than other plant miRNAs. Circulating miR-159 was more abundant in female healthy donors (n=6) than breast cancer patients (n=30) as analysed by qRT-PCR. Moreover, circulating miR-159 was less abundant in patients who relapsed with metastatic disease, suggesting that human serum miR-159 levels inversely correlate with breast cancer incidence and progression (Chin et al., 2016) . miR-159 in human serum was detected in an extracellular vesicle-enriched serum fraction obtained by ultracentrifugation and was resistant to sodium periodate oxidation, suggesting its plant origin due to 2'-O-methylation of the 3' end (which makes plant miRNAs more resistant to periodate oxidation). In situ hybridization confirmed the presence of miR-159 in tumour samples, suggesting that miR-159 in human serum is capable of entering breast tissue (Chin et al., 2016) . The authors analysed several plant food items for the presence of miR-159 and found that broccoli is particularly rich in this miRNA and most of this miRNA was still present after cooking (boiling for 25 min, >50% remained stable, compared to fresh). In vitro studies showed that synthetic miR-159 (both double-stranded and 2'-O-methylated) reduced cell proliferation in breast cancer cell lines but not the non-cancerous breast cell line. Extracellular vesicles from healthy human donors also have a similar effect. Luciferase assays determined that miR-159 targets transcription factor 7 (TCF7) and binds to two different sites within the 3' UTR. TCF7 is a transcription factor of the Wnt signalling pathway and is upregulated in breast cancer. In vitro assays also showed that miR-159 inhibits breast cancer cell growth by targeting TCF7.

Oral gavage of synthetic 2'-O-methylated miR-159 (25 mg/kg miR-159 or scramble control oligos, daily) reduced tumour growth from day 16 of treatment until the end of xenograph experiments (5 weeks) (Figure 19) , an effect that was lost when tumour cells stably overexpress TCF7 without the 3' UTR (Chin et al., 2016) . miR-159 was detected in xenografts of tumours by in situ hybridization analysis.

Overall, these experiments suggest that plant miRNAs can inhibit cancer growth in mammals. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Liang et al. evaluated and quantified 16 plant miRNAs in human plasma after consuming a single dose of watermelon juice or a mixture of fruits (Liang et al., 2015b) . Using TaqMan probe-based qRT-PCR assays and appropriate non-template controls, sixteen plant miRNAs including ath-miR-156a, miR-157a, miR-158a, miR-159a, miR-160a, miR-162a, miR-163a, miR-166a, miR-167a, miR-168a, miR-169a, miR-172a, miR-390a, miR-528, miR-824 and miR-894 were evaluated and a standard curve generated for quantification. miRNAs concentration ranged from 6.66 pM for miR-163a to 39,846.54 pM for miR-894 in watermelon juice, and from 0.01 pmol/kg for miR-166a to 55,192.46 pmol/kg for miR-894 in the fruit mixture. Watermelon juice was orally administered to nine healthy male adult volunteers and plasma samples were taken 0.5, 1, 3, 6 and 9 h after drinking 2.5 L of juice. Different miRNAs followed different peak time or appearance kinetics in plasma after administration (Liang et al., 2015b) . Using six plant miRNAs ( Table 26 ) that showed typical kinetic absorption curves, the total amount of plant miRNAs absorbed by the body was calculated. Using the plant miRNAs concentration in watermelon juice and the areas under the plasma concentration-vs-time curves (AUC) of plant miRNAs, the total uptake amount, total absorption amount and absorption rate for each plant miRNA was calculated (Table 25) . Absorption rate ranged from 0.04% to 1.31%. miR-156a, miR-162a and miR-168a were found almost exclusively encapsulated in microvesicles. miRNAs were validated in watermelon juice or plasma by northern blotting. Using the same protocol, a fruit eating study was performed. Each subject consumed an equal portion of watermelon, apple, banana, orange, grape, mango and cantaloupe (2.5 kg in total). While miRNAs concentration varied, the kinetics of plant miRNAs after fruit administration appeared similar to the kinetics after juice administration (Liang et al., 2015b) . The authors also suggest that the real concentrations of plant miRNAs may be underestimated and, for instance, the real concentration of miR-156a in the watermelon juice might be six times higher and its plasma levels four times the previously calculated (Table 25) concentration (Liang et al., 2015b) . Adapted from (Liang et al., 2015b) In a comment (correspondence) to the editor, the results of above article were criticized as being either a result of technical artefacts or a contamination problem, due to the presence of miR-528 in monocots and not in the dicot watermelon (Witwer, 2015) . In a reply letter, the original authors indicated that miR-258 also exists in dicots and that the miRNAs display a dynamic physiological kinetic absorption curve in human plasma, supporting the concept of uptake of food-derived miRNAs by humans and animals (Liang et al., 2015a) . EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Analysing miR-2911 by qRT-PCR in dried flowers and herbs used in Chinese medicines or as tea, Yang et al. reported different levels in different products including: sophora (≈6736 fmol/g), honeysuckle (≈5000 fmol/g), chamomile (≈5037 fmol/g), blue mallow (≈1127 fmol/g), ginseng flowers (≈278 fmol/g) and as low as 0.2 fmol/g in hibiscus . Herbal feeding of male mice for 7 days increased plasma circulating levels of miR-2911 compared to that of chow-fed animals. Honeysuckle increased 39-fold (207 fM) in sera levels, while chamomile increased 27-fold (147 fM), sophora 22-fold (120 fM), lavender 13-fold (71 fM), blue mallow increased 12-fold (64 fM), ginseng increased 5-fold (25 fM) and no significant changes were detected in hibiscus (3 fM). Post-feeding urine levels of miR-2911 ranged from 160-fold higher (264 fM) for honeysuckle, and 82-fold higher (135 fM) for chamomile and to 17-fold higher (28 fM) for lavender . However, when chemically synthesized miR-2911 (2'-O-methylated, plant-specific) was administered after single dose feeding of 400 pmols, mice serum levels were elevated roughly 1-fold within 30 min after gavage but decreased to background levels ≈ 1h after gavage . miR-2911 was not associated with Ago2 in sera and remained in the unbound fraction. Clearance of miR-2911 and other plant-derived synthetic miRNAs (miR168a, miR156a and miR161) were analysed in mice by tail vein injection of 50 fmols. Sera collected from mice (5min, 30min, 1h, 3h and 24h post injection) showed that most miRNAs were rapidly cleared from circulation. Compared to the other miRNAs administered at equal doses, miR-2911 levels were substantially higher at 5 min after injection but after 3 h the apparent clearance of all sRNAs tested was complete . In the same study, gut microbiota did not seem to influence miR-2911 absorption despite its 100-fold increase in their microbiome titre after honeysuckle consumption. Honeysuckle feeding of mice was also found not to influence gut permeability, discarding its possible absorption due to changes in gut permeability .

Rapeseed (Brassica campestris) bee pollen was evaluated for the possible absorption of plant miRNAs in male mice after pollen oral feeding (10 g/kg). miRNAs were analysed by sRNA sequencing after 3 or 6 h in mice serum (200 µL). From ≈ 10 M clean reads, only 132 reads were plant miRNAs, which belonged to 34 miRNAs. Of the plant miRNAs, bna-miR-166a (35 reads) and bna-miR-159 (22 reads) showed the highest levels in mouse blood and were mapped to the rapeseed genome. The rest of the plant miRNAs had ≤ 8 reads, of which 23 miRNAs had ≤ 2 reads. Both miRNAs, miR-166a and miR-159, were also detected in the rapeseed bee pollen as evaluated by qRT-PCR, with higher abundance for miR-166a. Serum levels of miR-166a after 6h of rapeseed bee pollen ingestion versus a control were ≈4 pM for treated animals while in chow diet fed animals it was ≈1.5 pM, suggesting that plant miRNAs from rapeseed bee pollen can be absorbed by mice by oral ingestion and detected in systemic circulation . evaluated herbal medicine honeysuckle miRNAs before and after decoction (boiling for 30 min, ≈0.2 g/ml water). From ≈10 M reads, ≈0.5 M reads of 148 miRNAs were obtained, from which osa-miR-166g, peu-miR-2911 and peu-miR-2914 contained more than 10.000 copy numbers. Most miRNAs were degraded by decoction (i.e. miR-166 g degraded from 222.051 to 315 reads), while miR-2911 remained high (11.322 reads) reaching 70% of total miRNAs in the final honeysuckle decoction. Concentration analysis of miR-2911 showed a reduction from 0.34 pmol/g in honeysuckle to 0.06 pmol/ml in decoction when analysed by qRT-PCR, or from ≈0.40 pmol/g to 0.08 pmol/ml when assessed by Northern-blot . Compared to other plant miRNAs, the apparent high resistance to boiling of miR-2911 was also supported by its high resistance to RNase treatment, which was dependent on its unique sequence and high GC content. Oral administration (n=6 per group) of a single dose (500 µL) of honeysuckle decoction (miR-2911 concentration ≈0.12 pmol/ml) resulted in a maximum peak level in mice plasma and lung tissues 6h after ingestion (i.e. plasma levels increased from 242.0 to 1189.2 fM). Most plasma miR-2911 was found in cell-derived microvesicles and associated EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

to the AGO2 complex. Continuous drinking of decoction during three days (0.06 pmol/ml, ≈4 ml/day) increased plasma levels from 250.5 to 831 fM, while no changes were observed in the control mice (from 215.8 to 253.3 fM). Levels of miR-2911 in lung and liver tissues increased from 8 to 10-fold, while the intestines and kidneys remained unchanged. Synthetic miR-2911 administration (1 nmol/ml, 100 µl given once per mouse, n=5 per group) resulted in an increase at peak (3h) from 239.3 to 1655.7 fM in plasma and also an increase at peak (6h) in lung tissue, suggesting overall that atypical dietary miRNA miR-2911 is absorbed and delivered to tissues . Studying its possible biological function, Zhou et al. found influenza virus H1N1 encoded mRNAs targets of miR-2911, i.e. virus-encoded proteins PB2 and NS1, which are relevant for viral replication (Zanin et al., 2017) . Luciferase activity assays confirmed that miR-2911 binds to PB2 and NS1, which is lost by point-mutation of miR-2911 binding sites of the individual viral gene sequences. In vitro H1N1 replication was reduced by synthetic miR-2911 or total RNA extracted from honeysuckle decoction. The inhibitory effect was abolished by co-transfection with anti-miR-2911 antagomiR. Moreover, no effects in viral replication were observed when PB2 and NS1 mutant virus were used with synthetic miR-2911 or honeysuckle decoction RNAs . In vivo administration of honeysuckle decoction or synthetic miR-2911 (0.1 nmol per day) by gavage one day before and during 7 days after inoculation reduced weight loss and viral titre in lung tissues. These results were abolished when the anti-miR-2911 antagomir was used concomitantly. Moreover, weight loss and high viral titre were not affected when PB2 and NS1 mutant virus were used, and synthetic miR-2911 or honeysuckle decoction administered .

The authors also showed similar effects for synthetic miR-2911 or honeysuckle decoction with other influenza viruses, including H5N1 and H7N9, tested in vitro and in vivo (Figure 20) . The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Regarding this particular miRNA, Yang et al. observed a different bioavailability of plant-derived miR-2911 vs. its synthetic 2'-O-methylated counterpart . Oral administration of cabbage extract containing 12.5 pmol of miR-2911 per 500 µL gavage volume was compared with a 32-fold higher dose of synthetic 2'-O-methylated miR-2911 (400 pmoles per mouse). After gavage, mice plasma levels of cabbage miR-2911 reached an approximately 2-fold increase (≈23 fM) versus that of synthetic miR-2911 (≈10 fM), suggesting that miR-2911 derived from plants is more bioavailable than synthetic RNAs . Whether this observation can be generalized to other dietary plant miRNAs is unknown. The authors also showed that miR-2911 was not present in exosome vesicles that were proteinase-K resistant in circulation .

Du et al. evaluated the presence of sRNAs in Hong Jing Tian (HJT, Rhodiola crenulata), which is a Chinese herbal medicine. Male C57BL/6J mice were fed with HJT RNAs by gavage for three days (250 µg/kg per day) and lung tissue collected at 12, 24, or 48 h (n=3 mice per time). The sRNA sequences of lung tissues (top eight ranked) that were also present in the plant HJT had ≈200K read counts . One of the small RNA (HJT-sRNA-m7) was effective in reducing in vitro transfected gene expression vectors for α-SMA, fibronectin and collagen type I α1 (COL1A1) in a TGF-β1 induced fibrosis model. Although the type of sRNA is not clearly described, HJT-sRNA-m7 targets the 3' UTR of these three genes. Using the HJT-sRNA-m7 agomiR (cholesterol-conjugated sRNA), by intratracheal administration (8 mg/kg in 100 µL of saline) (n=10 mice per group), the authors showed that fibrosis was reduced in a bleomycin-induced fibrosis mouse model . Due to the relevance of liposomes in delivering RNA molecules, the authors evaluated the possible formation of liposomes during decoction (boiling for 35 min) with plant lipids. They found that phosphocholine lipid molecules were also present in the decoction and suggested that liposomes may be formed that might facilitate the entry of RNAs from plants to human cells and tissues; this was not tested .

In another study, Mlotshwa et al., described oral administration of a cocktail of tumour suppressor miRNAs (normally downregulated in colon cancer) and a reduction of tumour burden in the Apc Min/+ mouse (C57BL/6J) model of colon cancer . miR-34a, miR-143 and miR-145 were chemically synthesized with the same mice sequence, but with a methyl group on the 2' position of the ribose at the 3' end, which is a chemical characteristic of plant miRNAs (See section 3.1.1). For 28 days (starting at 5 weeks of age), mice (n=7 per group) received either total plant RNA spiked with the mix of the three miRNAs, total plant RNA alone (from A. thaliana) or water. Oral dose was 61 µg of total plant RNA alone or plus 4.5 to 7.4 µg (corresponding to 0.7 to 1.0 nmole) of each species of tumour suppressor miRNAs. Tumour burden was dramatically reduced in mice receiving the miRNA mixture. Moreover, mice receiving the plant RNA alone also showed a reduced tumour burden, although it was not statistically significant. After periodate oxidation, intestinal levels of miR-34a were 10-fold higher than that of the group receiving water alone . miR-143 and miR-145 intestinal levels failed to be detected due to high background of endogenous mice miRNAs after periodate oxidation. Although the study does not directly evaluate the miRNAs produced by the plant, miRNAs designed to mimic small RNAs produced in plants seemed to be taken up by the digestive tract of Apc Min/+ mice upon ingestion and exert a biological effect .

Oral ingestion of total RNA isolated from Brassica oleracea (50 µg) resulted in the detection of plant miR-172 (the most highly enriched exogenous plant miRNA in B. oleracea) in blood of ICR mice between 2 h and 72 h after feeding . miR-172 was also detected in the spleen (0.38-0.04% recovered of orally administered miR-172), and merely detected in the liver or kidney. No evidence of sex differences was reported . At lower doses of total RNA (30 µg or 10 µg), no miR-172 was detected in the blood of mice . EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Using pigs as experimental animals, Luo et al. evaluated the detection of plant-derived miRNAs in the serum and tissues after feeding fresh maize to pigs (Sus scrofa). Jinhua female pigs consumed a fresh maize diet and water ad libitum for 7 days (n=3), and blood and different tissue samples were obtained for plant-derived miRNA analysis. Eighteen maize miRNAs were evaluated (16 were detected) in serum and tissues, and exhibited relatively low abundance in the pancreas and longissimus dorsi muscle tissue . Periodate oxidation treatment of the samples confirmed the detection of zma-miR-164a-5p, zma-miR-167e-5p, zma-miR-168a-5p, zma-miR-319a-3p, and zma-miR-408a-3p (while the rest were completely degraded), suggesting that these miRNAs are bona fide plant miRNAs (Figure   21 ). In a separate experiment, female pigs were fed one meal with fresh maize (1 kg per pig) after overnight fasting. Serum was collected at different times (i.e. 0, 1, 3, 6, 12, 18 and 24h). After 24h, pigs were provided with fresh maize diet ad libitum, and serum collected at 1, 3, and 7 days and tissue collected after sacrifice at 7 days. The serum levels of some of these miRNAs (miR-164a-5p, miR-166a-3p, miR-167e-5p, miR-168a-5p and miR-319a-3p) gradually increased after one feeding of fresh meal, reaching peak values at 6 or 12 hours within the first 24 hours and maintaining stable detectable levels (by qRT-PCR) in the serum following 7 days of access to fresh maize. The changes in miRNA expression at all measured time-points in most cases were modest (≈2 to 3-fold increase). The five miRNAs were primarily present in serum exosomes (≈58.2% of concentration in the serum) ( Figure 21 ). Using bioinformatics, 50 potential targets were determined within the porcine mRNAs for zma-miR-164a-5p.

Using in vitro luciferase assays with porcine kidney cells, the authors showed that zma-miR-164a-5p

binds to three of these potential target genes including Cspg4, Otx1 and Plagl2, an interaction lost when their mutant constructions were used, suggesting overall that dietetically-absorbed miRNAs can target endogenous porcine mRNAs . The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Several of the above-mentioned studies have reported the presence of dietary plant miRNAs in circulation associated to extracellular vesicles or exosomes (Chin et al., 2016; Liang et al., 2015b; . This could be relevant for their possible biological effects, as extracellular vesicles can deliver a cargo to several tissues, including crossing the blood-brain barrier, and other biological barriers such as the placenta (Matsumoto et al., 2017; Yang, T et al., 2015; Jayabalan et al., 2017) . Their presence in microvesicles could modify their function, for example, it has been described that in some cases delivered naked miRNAs cannot exert the same effects as those exerted by delivered exosomal miRNAs (Thomou et al., 2017) . Also, exosomes as therapeutic drug carriers and delivery vehicles are gaining relevance in therapeutics (Aryani and Denecke, 2016; Ha et al., 2016) . Recently, it has been described that edible plant-derived exosome-like nanoparticles can be isolated from these plants (Ju et al., 2013) , can transport ncRNAs (i.e. miRNAs) (Mu et al., 2014) and can exert a biological effect (Raimondo et al., 2015) , highlighting their role as therapeutic nanoparticles . It is still unknown if these exosome-like nanoparticles contribute to the resistance of ncRNAs in the harsh conditions of the GI tract and to the absorption of plant ncRNAs or the formation of exosomes circulating in biological fluids.

Regarding other biological barriers, only one study has reported on the transfer of small ncRNAs through the placenta. Li et al. found that exogenous plant miRNAs were consistently detected in the umbilical cord blood and amniotic fluid of healthy Chinese pregnant woman . This belongs to foetal circulation system, suggesting that miRNAs may transfer through the placenta to the foetal side. Three hours after gavage feeding of synthesized mature miRNAs of the influenza virus to pregnant mice (at last 14-day pregnant and with mature placenta) increased miRNAs levels were detected in the maternal plasma. An increase of exogenous miRNAs levels in the foetal liver was also observed . To determine if mature plant miRNAs can pass through the placenta and efficiently enter foetal organs, the traditional Chinese herbal honeysuckle was tested. Mice were gavage fed with 0.5 mL honeysuckle decoction (0.25 pmol/mice of miR-2911) and maternal plasma and foetal liver collected 3 h after gavage feeding. Plant miR-2911 levels were elevated ≈3.5-fold in the maternal plasma (reaching 180 fM), ≈2-fold in the placenta and ≈2.5-fold in the foetuses, suggesting that plant miRNA miR-2911 in honeysuckle can efficiently transfer through the placenta to enter the foetal liver (Figure 22) . The amount of circulating miR-2911 increased only in microvesicles, suggesting that circulating miR-2911 was primarily present in microvesicles and that transplacental transmission may involve a microvesiclesmediated pathway . The same study also tested gavage fed synthetic siRNAs (5 nmol), which were found in the plasma and foetal liver. Protein targeting of the siRNA was dramatically reduced, suggesting that exogenous siRNAs delivered from the mother to the foetus by trasnplacental transmission may regulate foetal gene expression. Direct injection of microvesicles loaded with siRNAs also reduced their mRNA target in the foetus liver, suggesting overall that exogenous small ncRNAs can be transplacentally transmitted from the mother to the foetus.

Although outside the scope of this review, other non-plant derived exogenous ncRNAs have also been reported to exert systemic effects in animals. Bovine milk-derived extracellular vesicles (BMEVs) were orally administered to IL-1Ra -/mice on the BALB/c background, which spontaneously develop polyarticular arthritis, from week 5 to week 15. Drinking water was also supplemented and administered to a collagen-induced arthritis (CIA) DBA/1J mice model starting at week 1 before immunization to day 40. Mice receiving the higher dose of 28 x 10 6 BMEVs (≈1200 micrograms/ml) showed a considerable delay in disease onset in the IL-1 Ra -/mice model. Moreover, in the CIA model the higher dose administered in drinking water 14.3 x 10 6 /mL (≈115 microgram/mL) reduced arthritis incidence and severity, accompanied by reduced expression of inflammatory cytokine IL-6 and MCP-1 . Although BMEVs contained miRNAs miR-30a, miR-223 and miR-92a, it is unclear if they are EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

responsible for the observed biological effects. In a similar study, Oliveira et al. administered BMEVs (particle concentration of 4.7 x 10 6 /ml or 14.3 x 10 6 /mL) in the drinking water to female DBA/1J mice for 7 weeks and observed an increase in osteocyte numbers. Moreover, the highest concentration of BMEVs increased the percentage of woven bone compared to the PBS group, and a reduction of the adipocyte area in the bone marrow (Oliveira et al., 2016) . Although the BMEVs also contained miR-29a (Oliveira et al., 2016) , its implication in the biological effects is unknown. 

Exogenous plant-derived ncRNAs can pass the placenta and reach the foetus.

By analysing small RNAs components in royal jelly, honey, beebread and pollen collected during the cabbage (Brassica campestris) flowering stage, Zhu et al. found that honeybee sRNAs were present at a far higher level in royal jelly than in beebread and pollen, while the abundance of cabbage small RNAs gradually increased from royal jelly to honey, beebread and pollen . Several miRNAs were detected in the samples, the plant miRNAs occurring at far higher concentrations. Moreover, beebread and polled miRNAs were at a much higher concentration and their composition of individual miRNAs was highly similar. The 16 plant representative miRNAs with the highest concentration were miR-156a, miR-157a, miR-158a, miR-160a, miR-162a, miR-166a, miR-166g, miR-167a, miR-168a, miR-172a, miR-172c, miR-390a, miR-397a, miR-403, miR-824 and miR-845a. The same analysis in samples collected during the camellia (Camellia japonica) flowering stage showed similar results and 13 out of the 16 enriched miRNAs were similar, suggesting that plant miRNAs in beebread and pollen may not be very diverse between different sources . Absolute concentration of the selected plant miRNAs ranged from <0.1 fmol/µg of total RNA (almost undetectable) in royal jelly and honey, to a maximum of ≈10 fmol/µg total RNA in pollen and beebread. Feeding experiments to larvae with either RNA purified from cabbage pollen (levels of miRNAs similar to natural beebread miRNAs levels) or a EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

synthetic mix of the 16 plant miRNAs, increased whole body accumulation of the 16 representative plant miRNAs and developed worker bee-like characteristics (reduced weight and size, extended pre-adult development time and decreased ovary size). Ninety-six honeybee genes were predicted to be targeted by the 16 plant miRNAs, some of which are known to influence the developmental fate of honeybees (i.e. Apis millifera TOR, amTOR). Luciferase assays showed that miR-162a directly targets apTOR mRNA.

Dietary supplementation of larvae food with synthetic miR-162a decreased amTOR mRNA levels in honeybees and showed reduced body weight, size and ovary size . Similar experiments were performed in Drosophila melanogaster (dm) with total pollen RNA, a synthetic miRNA pool or miR-162a alone, and in all cases similar phenotypes were obtained (delayed Drosophila development, reduced final size, ovary size and fecundity). dmTOR was also a target of miR-162a. Overall, this study implies that plant RNAs can be transmitted between species of different kingdoms.

The literature also describes some in vitro studies of plant miRNAs and their possible effects on mammalian cells. For example, exposure (24h) of plant miRNAs extracted from a Glycyrrhiza uralensis decoction to human PBMCs produced apparent cell aggregation and increased cell number and HLA-DR+ cell proportion (Shao et al., 2015) . In silico studies have also attempted to predict possible plant miRNAs with potential targets in humans .

As described before (see section 3.1), in contrast to mammalian miRNAs, plant miRNAs possess a 2'-Omethylation at the 3'-end that stabilizes and provides resistance to RNases and oxidation. This characteristic has been used to improve the quantification method for mature exogenous miRNAs from plants in biological mammalian matrices . Thus, periodate oxidation and βelimination were introduced to evaluate the origin of exogenous miRNAs. Male C57BL/6 mice were fed with corn/soy-based rodent chow diet for 2 weeks. Plant miR156a, miR164a and miR167a were evaluated. These miRNAs were detected in abundant amounts in a corn/soy-based chow diet, but after periodate oxidation and β-elimination 30.6%, 38.4% and 23.2% of miR156a, miR164a and miR167a were determined to be in methylated forms (2.0, 0.19 and 21.4 fmol/g of diet, respectively). The control diet showed no differences between the diet and non-template control CT values after periodate oxidation/β-elimination. This study also suggests the use of exogenous RNAs (spike-in), both methylated and un-methylated forms of miRNAs, to ensure appropriate periodate oxidation/β-elimination . No detectable levels of plant miRNAs were found in plasma, liver or faeces samples from the mice fed the corn/soy-based chow diet, since no differences occurred between the diet and the non-template control after periodate oxidation/β-elimination . This suggests that false positive results can be resolved using treatment with periodate oxidation/β-elimination.

Plant miRNAs from extra virgin olive oil (EVOO) and beer have also been evaluated by small RNA sequencing. However, very few sequences (≤16 reads) were identified, and most aligned to soybean and orange . A feeding study of a single dose ingestion of 40 mL of EVOO in healthy adult volunteers (n=5) was performed. miRNAs analysis at baseline and after 2h ingestion was performed. No plant miRNAs were detected in plasma after single ingestion of EVOO . Only one read count was found for other species miRNA vvi-miR3623-3p from grape, while two read counts were found from Medicago trunculata, grape or Brachypodium distachyon when 2 mismatches were accepted . This suggests that beer or EVOO do not contain miRNAs that contribute to dietary intake, even though miRNAs were reported in the pulp of olive (Olea europaea) (Donaire et al., 2011) . EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

Dickinson et al. evaluated a controlled mouse feeding study with rice-containing chow diets (modified AIN93-G with 40.8% of rice or rice-based with 75% of rice) or with a purified synthetic chow devoid of plant grain or forage for 1, 3 and 7 days (Figure 23) . The "synthetic chow" had negligible levels of osa-miR-168a, the "balance rice chow" (40.8% of rice) contained 227 reads, while the rice grain contained 279 reads. The concentration of osa-miR-168a in each diet was 0.0025, 54 and 114 fmol/g of diet for synthetic chow, balance rice chow and rice chow, respectively. Plasma and liver analysis of mice fed the diets did not reveal measurable uptake of any rice grain miRNAs, including osa-miR-168a. Indeed, fewer than ten reads were detected in five out of eight samples from mice fed on rice-containing chow and four out of five samples from mice fed on synthetic chow . Since the synthetic chow contained no grain or forage from plants, this low number for rice miRNAs-mappable reads can only be explained by sequence errors or cross-contamination. In agreement with , the authors indicated that animals fed a chow containing high levels of uncooked rice (%) had significantly increased plasma low-density lipoproteins (LDL) levels at 3 and 7 days after treatment initiation. However, this increase in LDL was not observed in the balanced rice/chow group (40.8% rice), which contained 54 fmol/g osa-miR-168a. The increase of LDL was attributed to nutritional imbalances between the test and control groups . LDLRAP1 protein was also evaluated by ELISA in mouse liver samples, and no changes in levels were observed in any group at any time, suggesting that dietary intake of osa-miR-168a (from 54 to 114 fmol/g) does not produce RNAi-mediated modulation of LDLRAP1 protein levels in mouse liver . EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

A commercially available plant-based, plant miRNA-rich substance containing soy and fruit materials, but not animal products, was administered by gastric gavage (≈5% of estimated blood volume for each animal) to male macaques (Macaca nemestrina). Relative abundance of plant miRNAs miR-160, miR-156, miR-166, miR-167, miR-168 and miR-172 were analysed by qRT-PCR in the plant-based dietary substance. For instance, for miR-160 a consistent and efficient amplification through at least 35 PCR cycles was detected. Plant miRNAs were analysed in plasma samples of macaques (n=2) before and after 1, 4 and 12 h oral gavage of plant rich miRNAs. qRT-PCR results indicated late amplifications of some plant miRNAs, but results were highly variable. miR-172 did not amplify before 42 cycles, and miRs-166, -167, and -168 had a median Cq greater than 35. miR-168 non-template controls also showed regular amplification within the same Cq range. miR-160 exhibited a tendency to increase after ingestion, but was highly variable, and Cq cycles were greater than 35. This was not the case for endogenous animal miRNAs or the spike-in control, supporting the case that the late apparent amplification of plant miRNAs in plasma was non-specific . Moreover, droplet digital PCR analysis showed that counts from plasma were generally very low and of variable intensity, as occurs in non-specific amplifications. This even occurred in miR-160 (plant miRNA with relatively high apparent count by qRt-PCR), which showed large number of counts but with variable intensity plots ("rain" appearance) (Figure 24) , and is consistent with non-specific amplification . Overall, the study (though only 2 subjects were included) suggests that the detection of plant miRNAs in plasma does not support a general and consistent uptake of dietary plant miRNAs . 

: Substantial levels of miRNA in oral diets consumed by macaque but negligible steady-state presence of diet-derived miRNAs in plasma or organ tissues. Snow et al. (2013) analysed the presence of exogenous plant miRNAs after regular intake of a Western diet containing fruits in healthy human subjects, and miR-21 in plasma or organ tissues of miR-21 -/mice . The conserved plant miRNAs miR-156a, miR-159a and miR-169a were highly The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. expressed in many fruits common in the human diet such as ripened avocado, apple, banana, orange and cantaloupe. MiR-156a and miR-159 reached up to 1000,000 copy numbers/mg food. The conserved animal miRNA miR-21 was expressed at substantial levels (≈100,000 copies/mg food) in meat but not in fruits. Mouse vegetarian and soy-enriched diets also contained high quantities (up to 1000,000 copy numbers/mg food) of miR-156a and miR-159a, but not miR-21. In contrast, mouse diets enriched with animal products contained elevated levels of miR-21, suggesting that conserved plant and animal miRNAs are commonly present in oral diets for various animal organisms. Although the dietary record of 7 of the 10 healthy adults reported the intake of fruits replete in miR-156a, miR-159a and miR-169a the day prior to plasma harvest, these plant-derived miRNAs were undetectable in plasma samples. Following 4-week feeding with a custom animal lard diet containing miR-21 no detectable levels of miR-21 were found in plasma samples of miR-21 -/mice, while high quantities of this miRNA in plasma were seen in control wild type mice fed the same diet . In tissues (liver, lung, kidney and stomach) miR-21 was either not detected or detected at exceedingly low levels corresponding to less than one copy of miR-21 per cell (assuming that the mammalian cell expresses roughly 10 pg of total RNA). Wild type-mice showed only a non-significant trend toward increased plasma levels of miR-156a after consuming either a vegetarian or soy-enriched diet when compared to mice fed a lard diet for the same period (4-week) ( Figure 25) ; however, levels were very low. Indeed, miR-156a was detected in organ tissues (liver, lung, kidney and stomach) but at levels of less than one copy of miRNA per cell and regardless of diets. miR-159a and miR-169a were not detected at all. Fresh avocado containing three specific plant miRNAs was also fed to mice (4.5 g per mouse over 24 h). Even though the three miRNAs were found in the stomach content of mice at high levels, very low levels were found in plasma and organ tissues, reaching levels lower than one copy miRNA per cell in tissues. This provides no evidence of substantial steady-state presence of these plant miRNAs in organ tissues after dietary intake . Assuming that, on average, at least 100 copies per cell (>100 copies/pg sRNA per cell) are necessary to achieve canonical target gene repression (Brown et al., 2007) , at least 10 15 copies of a given plant miRNA must be ingested and delivered for widespread biological activity in humans (the average human body contains ≈10 13 cells). Based on the concentration in ripe cantaloupe (≈6 x 10 6 copies miR-156a/mg fruit), an individual would need to consume 1,670 kg of cantaloupe just to release this number of copies in the GI tract .

In terms of sRNAs, the literature describes the threshold for target gene regulation to be between 1000 and 10000 copies of mammalian miRNA per cell . In an analysis of activity for hundreds of miRNAs, only the most abundant miRNAs in cells were found to mediate targeted repression and a large amount of miRNAs had no discernible activity (Mullokandov et al., 2012) . This suggests that only a minimum amount of miRNAs need be reached to exert a biological activity, although this will also depend on the amount of target transcripts.

A study mentioned above The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

analyses of other common plant miRNAs reported by others to be bioavailable were not successful. This implies that although the transgenic miRNAs had resistance to degradation comparable to that of miR-2911 and high expression levels in plants, they were not readily bioavailable . Huang (Huang et al., 2018) analysed the possible bioavailability of orally-ingested corn miRNAs in mice when incorporated into a rodent diet and consumed for 14 days. In an initial experiment, male C57Bl/6 mice (n=10 per group) received either water (control), random nucleotides (25 µg, equivalent to the same amount of small RNAs) or purified small RNAs isolated from corn kernel (25 µg, equivalent to the same amout of random nucleotides). In the second experiment, mice received either a control diet (AIN-93M), AIN-93 + 3% autoclaved corn kernel powder or AIN-93 + 3% fresh corn kernel powder. miRNA levels in the isolated RNA contained 56.03 pg miR-156a, 13.2 pg miR-164a, and 1215.63 pg miR-167a per 25 µg corn sRNA isolates. The diet supplement with fresh corn powder contained 173 pg miR-156a, 65 pg miR-164a, and 515 pg miR-167a per gram of diet. Autoclaving the corn powder at 121ºC for 30 min reduced the amount of initial miRNAs by more than 96%. In the first study, liver and whole blood sample miRNAs (analysed by qRT-PCR), showed no differences in CT values between the three groups. After periodate oxidation followed by β-elimination, no differences were detected between the no-template control group and the experimental groups. Similar observations were made in the liver and whole blood in the second study using fresh corn powder. Also, after periodate oxidation followed by β-elimination, no differences were detected between the no-template control group and the experimental groups in the liver and blood samples. Analysis of corn miRNAs in cecal and faecal samples www.efsa.europa.eu/publications 186 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

showed similar results after periodate oxidation, although certain minor differences remained between the no-template control group and the experimental groups. In all cases, less than 0.1% of total the miRNA tested in both studies was recovered from the faecal samples. Moreover, analysis of the miRNAs in the content collected from different parts of the mouse GI tract from the animal studies administered via gavage or dietary intake showed that calculated recovery accounted for less than 0.3% of the content originally ingested in the stomach, less than 0.01% of that originally ingested in the intestine and faeces, and less than 0.01% of that originally ingested in the colon and cecum (Huang et al., 2018) .

Further experiments using in vitro digestion system suggested that after the gastric phase, over 97% of corn miRNA miR-167a from the diet or from the extract were degraded, suggesting overall that significant degradation of corn mRNAs occurred during digestion, which resulted in minimal uptake of corn miRNAs after oral intake (Huang et al., 2018) .

Witwer re-analysed the plasma data evaluated by Liu et al., and found that only one putative plant miRNA mapped above a median cut-off (200 reads), which was the plant miRNA peu-miR-2910 . This implies that all RNAs -including previously reported exogenous miRNAs (xenomiRs) such as miR-159a, miR-168a and the plant ribosomal degradation fragment miR-2911 -are below the level of background noise. It is not clear if miR-2910 is a real miRNA as it is found in the highly conserved and expressed subunit rRNA of plants . Moreover, the miR-2910 sequence is not only a fragment of plant rRNA, but also has a 100% coverage and 100% identity match with human 18S rRNA, while others do not map to plant genomes at all . Perhaps more analyses are needed when assessing possible xenomiRs.

After in silico xenomiRs analysis using public sequencing datasets (see section 3.3.2 for details), Kang et al. performed controlled feeding studies to detect the transfer of exogenous plant miRNAs to rat blood or from bovine milk sequences into piglet blood . Adult rats (n=3 per group) followed a 28-day controlled feeding study with supplementation in three different diets: monocot plant material (rice); dicot material (potatoes); and husbandry chow containing a defined mixture of grains, cereals, vitamins, minerals and fats. Small RNAs were isolated from serum and sequenced. Most samples showed either a lack of plant xenomiRs in rat serum, or three or less read counts in certain samples . These extremely low numbers suggested a probable false positive result. In the same study, piglets were fed either cow milk or soy milk for 4 weeks followed by 7 weeks of feeding with maize. Few cow-specific sequences were found (15 reads) in the piglets fed cow milk, while piglets fed maize as well did show cow-specific sequences (21 reads). This suggests likely false positive results. Both studies provide no evidence of dietary transfer of xenomiRs . The abovementioned studies are summarized in Table 27 .

Although outside the scope of this literature review, pollen plant miRNAs ingested as part of a typical diet of adult honey bees were not found to be robustly transferred across the epithelial barrier under normal conditions (Masood et al., 2016) . For example, despite the high levels of plant miR-156a in pollen or its detected levels in the bee digestive tract (0.5 copies per 10 pg total RNA in the midgut), less than 0.1 copy per 10 pg total RNA was found in abdominal tissue (Masood et al., 2016) .

www.efsa.europa.eu/publications 187 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. Large amount of plant miRNAs miR-156a, miR-159a and miR-169a detected in vegetarian, soy-enriched, or avocado diets (i.e. ≈1 x 1010 copies for vegetarian soy/diets or ≈1.1 x 109 copies for avocado diet of miR-156a after 24h), but very low levels found in plasma and less than 1 copy per/cell in tissues. After 4 weeks miR-21 rich diet in miR-21 KO mice, no plasma levels and less than one copy of miR-21 per cell detected in tissues. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. After periodate oxidation, no major differences observed in miR-156a, miR-164a, or miR-167a among treatment, control groups or no-template control in liver and blood. In cecal and fecal samples no major differences observed among control and treated groups. miRNA analysis in the GI content collected from different parts of mouse showed that the recovery (calculated) account for less than 0.3% of originally ingested in the stomach, less than 0.01% of originally ingested in the intestine and feces, and less than 0.01% of originally ingested in colon and cecum. In vitro digestion experiments suggested that after the gastric phase, over 97% of corn miRNA miR-167a from the diet or from the extract is degraded. (Huang et al., 2018) The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. Exogenous plant miRNAs found in umbilical cord blood and amniotic fluid of healthy Chinese pregnant women. Plant miR-2911 increased in plasma of the mothers, placenta and foetus liver 3h after oral gavage (2.5 nmol/mice). Exogenous siRNA gavage (5 nmol) increased plasma levels of mother and foetus liver. Protein target in foetus repressed. Direct injection of microvesicles loaded with siRNAs also increased its levels in foetal liver and reduced mRNA expression of its target. Exogenous small ncRNAs can be transplacentally transmitted from mother to foetus. n.d., not determined. KO, knock out; WT, wild type; miRNA, microRNA; sRNA, small RNA; dsRNA, double-stranded RNA; COL3A1, collagen type I α1; α-SMA, alpha smooth muscle actin; HJT, Hong Jing Tian;

www.efsa.europa.eu/publications 195 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s).

The available evidence presents several examples of systemic effects of plant-derived exogenous ncRNAs when administered orally. However, there are also several reports that do not support these findings and contradict the hypothesis of cross-kingdom regulation. In either case, the essential question concerning the existence of this possibility is still heavily debated. Important aspects such as the precise mechanism of transport of plant ncRNAs from food to the circulatory system, the amount of exogenous ncRNAs reaching tissues or the molecular mechanism of cellular uptake need to be understood. It remains unknown if such a transfer could be modulated in particular context (i.e. specific diet, disease status, medication).

The available evidence also suggests that plant ncRNAs can directly target mammalian genes through

RNAi effects when they are exposed in in vitro systems. Without considering other physiological parameters, if a synthetic plant miRNA is exposed to its mammalian target, it can effectively bind by base complementarity and exert a biological effect (i.e. gene repression). However, whether the exogenous plant ncRNA can bypass all the biological barriers to reach and interact with the possible target remains to be clarified. Apparently, exogenous plant-derived ncRNAs are found in exosomes or macrovesicles. It is not yet known how they reach these types of structure in biological fluids, or if novel plant exosome-like nanoparticles participate in any of these processes.

www.efsa.europa.eu/publications 201 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s).

immune cells were mainly from studies in human and mouse isolated lymphocytes. However, these studies were descriptive/not comprehensive, reporting specific effects of ncRNAs in proinflammatory cytokine production using human cells, and would not allow identifying specific endpoints that can be used to evaluate changes in the regulation of immune function and homeostasis by ncRNAs.

Retrieval of grey literature using the key words 'thesis plant RNA immune' provided numerous results (i.e. Google = 510.000), mostly regarding the function of miRNAs and RNA silencing in plant immunity. For instance, a document could be chosen based on title selection because it included the terms 'plant, double-stranded RNA and systemic immunity' (https://uknowledge.uky.edu/plantpath_etds/21). However, this study was not further considered since it refers to plant immunity.

After analysing the documents and refining by key questions, a large number of screened publications were discarded because they refer to the role of endogenous ncRNA (human and animals) as well as viral infections in relation to the immune function, which is outside the scope of this literature review.

As an example, within 1358 original in vivo publications related to the key words, only 30 studies were considered relevant for the proposed review questions. In summary, from the literature search, 91 papers were finally selected as relevant to review of the effects of ncRNAs on immunity. Very few studies analyse the direct relationship of exogenous plant ncRNA with immune function and homeostasis of humans or animals. Albeit controversial, exogenous ncRNAs derived from plants and microorganisms have also been described in human blood (detailed in section 3.3.2.). Overall, the results indicate that there is scarce information on the specific impact of plant ncRNAs on activation of specific immunocompetent cells. Of these 91 references, 64 documents were used for reviewing the general features of ncRNAs in immune function and immunity in humans and animals (section 3.4.1.), 22 were used for the effect of exogenous plant ncRNAs on immune system (section 3.4.2.) and 5 for the additional section on the effects of exogenous ncRNAs on gut microbial composition (section 3.4.3.) ncRNAs have been shown to play important roles in immune cell development and function in normal and disease conditions (Ansel, 2013) . Human and mouse studies demonstrate that miRNAs play a critical role in the regulation of immune cell development and their function. There is also evidence of the relevance of neonatal miRNA-mediated immune activation (Kosaka et al., 2010) . Here, miRNA molecules of maternal origin appear to be stable in the infant's gut conditions, allowing dietary intake and transfer of miRNA (Kosaka et al., 2010) . ncRNA have also been identified as regulators of metabolic shifts within microbial communities (Paul et al., 2015) . Thereby, ncRNAs may significantly influence the critical roles of gut microbiota on health and disease. This part of the literature review (EFSA Task 4) presents the current knowledge on the role of exogenous ncRNA molecules on the immune system of humans and animals and aims to provide information on the possible effets of dietary ncRNAs on the regulation and function of the immune system. In addition, since the gut microbiota influences the immune system, a review on the possible effects of exogenous ncRNAs as gut microbiota modulators is provided

This chapter provides background information on the modulatory effects of ncRNA on the immune system of humans and animals and serves as a general introduction to explore the possibility of dietary plant ncRNAs effect on the immune system after dietary intake. Sixty-four (64) documents were finally used to review this section.

When activated, the immune system tightly regulates innate and adaptive response(s), aiming at restoring homeostasis, and preventing autoimmunity. The 'amplitude' (i.e. intensity and duration) of the www.efsa.europa.eu/publications 202 EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s).

interaction with the T-(TCR) and B-(BCR) cell receptors determines cell fate decisions. In the periphery, regulatory T cells (Tregs) play a key role in restraining the activity of mature B and T (Th1, Th2 and Th17) cells, and preserving tolerance mechanism(s). These responses depend on transcriptional and epigenetic regulation including regulation by miRNAs. Emerging systems for measuring miRNA activity during immune activation revealed the complex network of genes that may be simultaneously targeted by miRNAs to tune distinct cell fate decisions (Wells et al., 2017) .

RNAs exhibit an intrinsic property of base pairs that is crucial in defining their structure and, thereby, their physiological role (Lu et al., 2016) . Plant miRNAs are 2'-O-methyl modified on their terminal nucleotide, making them more difficult to be ligated to the cloning adapter. The inhibitory effect of 2'-O-methyl modification at the 3' end of the RNA substrate for the mammalian nuclease family member NEF-sp has been described (Silva et al., 2017) . As described elsewhere, most mammalian miRNAs require 'pre-processing steps' to become physiologically active. Nucleotide sequence complementarity appears as a critical feature driving the predominant result exerted by miRNAs. Notably, it has been reported that a modification pattern involving alternating 2'-O-methyl RNA bases improves dsiRNA in vitro stability and evades activation of the innate immune system (Collingwood et al., 2008) . These data were obtained with isolated peripheral blood mononuclear cells (PBMCs) as a mixed 'lymphocyte'

population of immune receptors that are encountered in vivo. However, the authors do not provide any molecular mechanism for the observed effects.

siRNAs have been shown to be potent stimuli of interferon- production by plasmocytoid dendritic cells (Hornung et al., 2005) . This siRNA technology induced systemic immune responses in the same range as the Toll-like receptor (TLR)-9 ligand 'CpG', including activation of T cells and dendritic cells in spleen. Notably, immunostimulation by siRNA was absent in TLR7-deficient mice. There is a lack of data related to the effects of plant ncRNAs on PGE2 production, implicated in immunosuppression by TLR3 via COX2.

Several distinct classes of lncRNAs are transcribed from different DNA elements or are derived from long primary transcripts with noncanonical RNA processing pathways, generating new RNA species with unexpected formats. These lncRNAs can be processed by several mechanisms, including ribonuclease P (RNase P) cleavage to generate mature 3' ends, capped by small nucleolar RNA (snoRNA)-protein (snoRNP) complexes at their ends, or the formation of circular structures. Recently, it was reported that several lncRNAs mediate their regulatory effects through binding to specific RNA-binding proteins . In addition, expression of lncRNAs has been defined as highly cell-type specific (Guttman et al., 2010; Washietl et al., 2014) . This high specificity has been reported to be critical for the development and activation of immune lineages. However, lncRNAs-mediated regulation of innate immune activation (i.e. IL-6 production) has not been tested for function in vivo, but only evaluated in vitro (Roux et al., 2017) . The regulation of inflammatory mediators or cytokines by lncRNAs is still poorly understood.

Synthetic nucleic acids, such as dsRNAs, have been reported to be recognized by Toll-like receptor (TLR)-3 and can stimulate the innate immune system and trigger a Type I Interferon response (Marques et al., 2006; Mian et al., 2013) . In plants, it has been described that dsRNA with 5' overhangs contributes to endogenous and antiviral RNA silencing pathways (Fukunaga and Doudna, 2009) . Earlier reports defined a structural basis (i.e. 3' overhangs) to cause nonspecific effects on the ds-RNA-activated signalling pathways through the interferon responsive factor (IRF)-3 (Marques et al., 2006) . Similarly, immune activation by the subcutaneous route with a viral mimic dsRNA (Poly I:C) of C57Bl/6 mice during neonatal early or late brain development differentially affects depression-related behaviours developed in adolescent and adult mice (Majidi-Zolbanin et al., 2014) . The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s).

Specific molecular interactions of the exogenous RNA triggering immune responses have been reported. Several studies have demonstrated the role of components of innate immunity, including Toll-like receptors (TLRs), the retinoic acid-inducible gene I/melanoma-differentiation factor 5 (RIG-I/MDA5) and the interferon response effectors protein kinase R (PKR) and RNase L, as well as miRNAs in the recognition of viral or synthetic dsRNA (Malathi et al., 2007; Urcuqui-Inchima et al., 2017) . TLRs are a family of proteins recognizing different pathogen and damage-associated molecular patterns, PAMPs and DMAPs, respectively. RIG-I/MDA5, PKR and RNase L constitute cytosolic RNA sensing proteins. The pathways responsible for these defined signalling processes can be largely separated (Figure 26) , despite the existence of some degree of crosstalk between them where the engagement of interferon regulatory factors is a common feature.

TLRs play an essential role in innate immune responses in mammals. In humans, ten different TLRs recognize distinct molecular patterns, where TLR-3, -7, -8 and -9 stimulate innate immune responses upon interaction with nucleic acids Weber et al., 2012) . TLR3 is located on the plasmatic cell membrane, while TLR-7, -8 and -9 display endosomal localization. It has been reported that TLR3 recognizes dsRNA, viral or the synthetic poly I:C, and siRNA has also been identified (in the presence or absence of delivery systems) as a relevant ligand . Notwithstanding, RNA interference is considered an unlikely mechanism to be engaged for viral sensing (Cullen, 2006) . GUrich RNA from viruses or synthetic single-stranded (ss) oligoribonucleotides displays an apparent preference for stimulating human TLR7/8-mediated immune effects (Heil et al., 2004; Forsbach et al., 2008) . A minimum of four nucleotides, UUGU, found in the GU-rich region are reported as necessary to stimulate cytokine responses.

Endogenous or exogenous RNAs are internalized into the endosomal compartment. The RNA sensor system is composed of i) several innate immune 'Toll-like' receptors (TLRs) located at the endosomal compartment that trigger different signalling pathways engaging adaptor distinct molecules (MyD88 and TRIF), and ii) a cytosolic system (RIG-I/MDA-5) that coordinates its signalling with TLRs. Antigen processing will lead to rearrangement of MHC molecules to activate the adaptive immune response(s). Type I interferons (IFN-α/β) are expressed rapidly after infection and serve as a link between the innate and adaptive immune responses.

: Schematic representation of the eukaryotic RNA sensor (single-stranded RNA and short/long double-stranded RNA) and its relation to innate immunity. The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s).

Two classes of ssRNA motifs have been described that either preferentially activate TLR8-mediated or TLR7/8-mediated immune responses. AU-rich oligoribonucleotides stimulate TLR8-but not TLR7mediated immune activation (Forsbach et al., 2008) . RNA motifs activating TLR8-associated signalling fail to induce IFN- from TLR7-expressing plasmacytoid dendritic cells but induce the secretion of Th1like and proinflammatory cytokines from monocytes or myeloid dendritic cells. In addition, RNA motifs activating the TLR7/8-associated signalling pathway stimulate cytokine secretion from both TLR7-and TLR8-positive immunocytes (Forsbach et al., 2008) . Differences in the protein sequence of TLR-8 between different mammals (human, monkey, chimpanzee, cattle, porcine, mouse, and rat) can be found, and species-specific recognition of ssRNA via TLR7/8 had been described (Heil et al., 2004) . The TLR8-specific RNA sequences are unable to trigger cytokine responses from mouse, rat, and porcine immune cells (Forsbach et al., 2008) . A dual function has also been suggested for the murine coreceptor CD14 to enhance TLR-3, -7 and -9 signalling (Baumann et al., 2010; Lee et al., 2006) . However, further studies showed evidenced that human CD14 did not appear to function as a co-receptor for TLR-3 or TLR9 (Weber et al., 2012) .The innate immune system detects RNA lacking nucleoside methylation (m5C, m6A and m5U) or otherwise (i.e. pseudouridine or 2'-O-methyl-U) modified as a mechanism to selectively trigger immune responses to nucleic acids from necrotic tissues or of exogenous origin.

A recent systematic study on transcriptional and, especially, post-transcriptional regulation of stressresponsive lncRNAs in Oryza sativa showed that hundreds of lncRNAs with down-regulated polyadenylation are highly conserved in stresses . In the ascidian Ciona intestinalis a novel alternative polyadenylation signal activated by the prototypical TLR4 agonist (i.e. LPS) has been described (Vizzini et al., 2016) . Moreover, alternative polyadenylation has been identified as a regulatory mechanism during the innate antiviral immune response in macrophages (Jia et al., 2017) . These authors showed an enrichment of polyadenylated genes and mRNA abundance change in TLRs as well as RIG-I-like receptor, JAK-STAT and apoptosis-related signalling pathways.

Studies conducted on the crystal structure of the sensing TLR-3 ectodomain allowed definition of dsRNA recognition of at least 40 bp (Liu et al., 2008; Leonard et al., 2008) . It was concluded that TLR3 assembles on dsRNA as stable dimers and that the minimal signalling unit is one TLR3 dimer. Recognition by secondary structure 'alone' has been reported, independently of the sequence motif, by TLR3. This dsRNA length is larger than the typical helix occurring in normal miRNAs and siRNAs. TLR-7 and -8 have been found to recognize single-stranded RNA (ssRNA). The existence of a similar double horseshoe structure for TLR-7, -8 and -9 upon activation has been proposed, although their binding modes to their ligands remain unclear . Often, in a partial view, TLR7 is considered as the ssRNA-sensing TLR and as complementary to TLR3. However, higher structural features have been identified as important aspects in TLR7 RNA sensing (Jöckel et al., 2012) . In addition, small molecules that may be viewed as nucleoside analogues are able to activate TLR7 and/or TLR8 (Hemmi et al., 2002; Lee et al., 2003) . It has not been defined if the recognition mode for small molecules, including sRNAs, resembles that of TLR7 for RNA recognition.

Besides TLRs, RIG-I is a cytosolic multi-domain protein sensing viral RNA associated to N-terminal caspase activation and recruitment domains (CARDS) to transmit signalling but preventing it in the absence of RNA. The identification of RIG-I/MDA5 as cytoplasmic sensors for dsRNA, associated to facilitators such as PKR and 2'-5'-olygoadenylate synthetase, suggests that additional 'aspects' (i.e. dependence on ATP) other than dsRNA structure or short size influence innate immune responses to sRNAs (i.e. siRNAs and miRNAs). These aspects remain unclear. The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s).

Current information on miRNAs points to clear differences in cellular recognition of endogenous and exogenous dsRNA. Here, 5'-triphosphate and dsRNA represent molecular patterns enabling RIG-I to discriminate exogenous from endogenous RNA (Myong et al., 2009) . dsRNA products as short as 21-23 bp are sensed by RIG-I independently from the 3'-termini (Marques et al., 2006) . Similarly, the IFNinducible 2'-5'-olygoadenylate synthetase (OAS) requires dsRNA that must be at least 15 nucleotides long for activity, and no modification of the 2′-hydroxyl group is tolerated (Sarkar et al., 1999) . The OAS family requires synthesizing di-, tri-, and tetrameric 2'-5'-oligoadenylates (2-5A), which in turn bind and activate RNase Activation of RNase L together with PKR, IRF3 and c-Jun-N-terminal kinases, constituting pro-apoptotic mediators in response to dsRNA. Moreover, human OAS may also enhance RIG-I mediated signalling by mimicking polyubiquitin . Further structural and mechanistic studies identify a key role of domain duplication in the OAS family, thus revealing different functions of OAS-1 and OAS-3 in sensing dsRNA (Donovan et al., 2013; .

RNA interference (RNAi) as well as antisense oligonucleotides and aptamers have been postulated to exert immune stimulatory effects in mammals (Agrawal and Kandimalla, 2004; Cullen, 2006; Mustonen et al., 2017) . However, at present, there is still an important open debate about agents exerting physiologically relevant functions at systems level biology (Cullen, 2006; Kleinman et al., 2008) . Moreover, clinical trials of siRNA demonstrated that the only population carrying the 412FF coding variant for TLR3 was protected from siRNA-induced cytotoxicity (Kleinman et al., 2008) . This study concluded that, because multiple cells express surface TLR3, the therapeutic approach with siRNA might induce unanticipated vascular or immune effects. Here, anti-angiogenic innate immunity triggered by siRNA was found to occur without the induction of INF-α/β via TRIF activation being biased toward NF-ĸB rather than IRF-3. Side effects, for example, the immunosupressive effects of T regulatory cell function are inhibited by IFN-α-2β (Yu et al., 2016) , are still only vaguely understood in association with exogenous ncRNAs administration.

Naturally occurring mammalian RNAs do not display a stimulatory potential on innate immune responses through TLR-3, -7, -8 and -9 in dendritic cells and TLR-expressing cells (Karikó et al., 2005) . By contrast, it has been shown that major innate immune responses to chemically synthesized siRNAs are mediated by TLR7 and/or TLR8. Notably, replacement of uridines with their 2'-modified counterparts in the siRNAs, reduced immune activation (Sioud, 2010; Barik and Lu, 2015) . Studies conducted on synthetic siRNAs have shown results directly dependent on the nucleotide sequence (Judge et al., 2005 (Judge et al., , 2006 Hornung et al., 2005) , and not to silencing effects of a target gene. Putative immunostimulatory motifs allowing definition of IFN-α induction by β-gal control siRNA or its constituent sense (GU-rich) and antisense ssRNA oligonucleotides (Judge et al., 2005) have been identified. These studies were performed in PBMCs, and therefore lack a clear demonstration of these effects at the organism level.

Intravenous administration of encapsulated siRNA induced substantial dose-dependent (1-50 g siRNA) IFN-α responses in outbred Institute for Cancer Research (ICR) inbred mice. In this line, reports also exist of increased hepatotoxicity of ASOs containing locked nucleic acid modifications (Swayze et al., 2007) . In addition, ASOs carrying 2'-O-methoxyethylribose modifications showed no cytotoxic effects but were able to reduce targeted mRNA. By analysis of sequence variants of either ss or ds siRNA, uridine-rich immunostimulatory motifs within siRNA were identified in human peripheral blood cells via TLR7 (Sioud, 2010; Jurk et al., 2011; Judge et al., 2006) . There is a report of sequence-and targetindependent angiogenesis suppression induced by siRNA via TLR3 (Kleinman et al., 2008) . A minimum length for dsRNA of about 21-nucleotide or 23-nucleotide Luc siRNA, but not truncated versions The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s).

suppressed by choroidal neovascularization have been observed in wild-type mice (Kleinman et al., 2008) .

Recently, it has been established that nuclease-resistant aptamers can play a 'dual' role in immune signalling pathways (Gefen et al., 2017; Rajagopalan et al., 2017) . These aptamers seem to possess a unique feature acting as both agonist and antagonist depending on their degree of oligomerization (Nozari and Berezovski, 2017) . Aptamers are ss oligonucleotides (70 -100 nucleotides) capable of recognizing, in a specific manner and with high affinity, several types of target molecules by means of a three-dimensional folding of their chain. Preclinical studies in murine models have highlighted distinct regulatory points that may be influenced by oligonucleotide aptamers, including clusters of differentiation (CD) (Pastor et al., 2013) , cell-to-cell communication such as immunoglobulins (Gefen et al., 2017) and interleukin (IL) signalling (Rajagopalan et al., 2017) . The CD molecules can act in several ways, often acting as receptors or ligands important to the cell initiating signal cascades or serving as adhesion molecules. Here, CD28 is one of the main co-stimulatory receptors responsible for proper activation of T lymphocytes that could be manipulated to modulate the immune response (Pastor et al., 2013) . Further studies reported enhancement of T cell functions by specific T cell immunoglobulin-3 aptamers through negative regulation of IFN- secretion by CD4 + and CD8 + cells (Gefen et al., 2017) . Positive effects attributed to oligonucleotide aptamer have also been associated to attenuation of IL-2 signalling in CD8 + cells (Rajagopalan et al., 2017) .

While siRNA and dsRNA in cultured cells have been shown to trigger an interferon response (Alexopoulou et al., 2001; Sledz et al., 2003) , the administration of naked (dTdT) synthetic siRNAs (2.5 mg/kg) to mice (either intraperitoneally or iv via low pressure or high pressure injection) induced no interferon response (Heidel et al., 2004) . This suggests that siRNA administration in vivo may elicit little or no immune response. Also, modification of nucleotides within RNA molecules was shown to reduce immune response (Durbin et al., 2016) , which is relevant when applying nucleotide modifications to RNA therapeutics. For example, N-6-methyladenosine (m6A) and pseudouridine seems to reduce the retinoic acid inducible gene I triggering of the innate immune signalling (Durbin et al., 2016) . Thus, RNA modification seems to influence recognition of exogenous RNAs, for example, ribose 2'-O-methylation of mRNA at the 5'-end provides a molecular signature for the discrimination of self and non-self RNA (Züst et al., 2011) .

Prior to immune cells performing their immunological action on target cells, they need to receive a suitable 'maturation' signal to enhance clonal expansion and acquire the effector function. Activation signals occur in peripheral lymphoid organs through a concerted interaction with MHC molecules.

The major histocompatibility complex (MHC) regulates the cell-mediated immune response(s). There are recent reports on the regulatory role of miRNAs on MHC class I and II molecules (Wongfieng et al., 2017; Xie et al., 2017) . For instance, MHC class I chain related protein B present in natural killer (NK) cells was associated to cell-mediated antitumor immune response through engagement with the NKG2D receptor (Xie et al., 2017) . The expression level of this receptor has been shown to be regulated by exogenous miR-30c mimics in NK cells . Based on the temporal and spatial regulation which determines the expression patterns of miRNAs, that suggestion was made for the participation of complex regulatory networks composed of several transcription factors influencing the expression level of 'each' miRNA.

The MHC class II complex has been identified as a target in mammalian miRNAs regulation (Wongfieng et al., 2017) . In addition, some lncRNAs have been identified as candidates to be translated into peptide fragments (Guttman et al., 2013) suggesting, as several other 'indirect' lines of evidence, the The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s).

participation of ncRNAs-derived peptides on MHC complexes (Yewdell, 2007; Mellman et al., 2011) . Peptides bound to MHC molecules serve as recognizing fragments of antigen for the TCR. Here, it has been shown that miRNAs can influence the amplitude of TCR signalling modulating T cell fate decisions towards survival and maturation Wells et al., 2017) . Here it is worth noting the recently defined dual role of miRNA in maintenance of the naive phenotype of CD8 T cells or clonal expansion, where modulation of Let-7 miRNAs levels is required (Wells et al., 2017) . In a similar way to MHC class II for T cells, BCR determines the fate of developing B cells allowing their adequate maturation or the development of self-reactive B cells leading to immunodeficiency of B cell malignancy. The miRNAs set threshold for lymphocyte development targets several genes in the PI3K pathway (Simpson and Ansel, 2015) . It is also important to highlight that PI3K displays an inhibitory interaction on TLRs-induced production of IL-12, therefore limiting excessive Th1 polarization and suggesting a potential immunoregulatory role for miRNAs.

A few relatively recent studies address the potential regulatory role of 'specific' miRNAs in Treg functions (Jeker et al., 2012; Warth et al., 2015) . These studies have shown that inhibition of miRNA-10a in C57BL/6J mice led to reduced Foxp3 expression dependent on the levels of mediators such as TGF and/or retinoic acid (RA) (Jeker et al., 2012) . Another study reported induction of Treg cell differentiation regulated by a miRNA 'network' influencing the promoting factor mTOR (Warth et al., 2015) . These studies revealed that the underlying molecular mediators responsible for the miRNA regulation of T cell signalling pathways remain unknown. Unlike animals, plant miRNAs have not been associated to TGF nor RA. Intersecting the rapidly emerging field of Treg function, it has been discovered that RA controls both the homing and differentiation of Treg. In addition, computational and systems biology approaches have identified mTOR as a potential pathway to explain the cross-kingdom miRNAs-mediated regulatory potential of Camptotheca acuminata ).

Although many reports can be found regarding the immunomodulatory role of endogenous and viral ncRNAs, very few evaluated the impact of plant ncRNAs on immune function and homeostasis when ingested/absorbed by animals and/or humans. Thus, this section was reviewed using 19 scientific documents. However, no studies were found that directly evaluate the effect of plant exogenous ncRNA and their effects on the immune system following dietary intervention trials.

Recent reports highlight the shortcomings of existing miRNA measuring systems (i.e. qPCR) in comparison to sequencing procedures to explore miRNAs in biological fluids Chen et al., 2016a) . These studies point out the considerable number of unmapped ncRNAs as a major factor limiting the use of qPCR techniques. Notably, the abundance level of identified exogenous RNA sequences in biological fluids (i.e. human plasma and breast milk, and serum from mice) was rather low. In addition, the role of food matrix on ncRNA delivery to their cellular or tissue targets has not been defined. Although the identification of exogenous ncRNAs in biological samples led to considerations on their role as cross-kingdom regulators of immune-related processes, the impact (i.e. influence and regulatory role) of exogenous plant ncRNAs on immune function and homeostasis is inferential. Exogenous RNAs sequences have been found in plasma , serum (Chen et al., 2016b) and breast milk ) from humans and/or animals. These studies reported the presence of ncRNAs in biological fluids, although their impact on immune function remains to be fully clarified.

The recent discovery of blood circulating RNAs originating from foods seems to indicate that certain dietary plant miRNAs can be absorbed at significant levels . However, the potential contribution of distinct 'plant food matrices' on ncRNAs The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s).

absorption has not been addressed as it has been described for viral RNAs (Seljelid et al., 1973; Zou et al., 2010) . The existing studies show that adequate antigen-processing processes at the intestinal level can be avoided by ncRNAs bound to coadjuvant molecules (i.e. polycations, cationic lipids, phospholipids) (Seljelid et al., 1973; Zou et al., 2010; Vertzoni et al., 2004) . Notably, phospholipid homeostasis is central in the response to Toll Like Receptors (TLRs) activation through a type I interferon 'autocrine-paracrine' loop (Song et al., 2015) .

In biological samples, RNA sequences have been reported from most common GM food matrices, including corn (Zea mays), rice (Oryza sativa Japonica group), soybeans (Glycine max), tomato (Solanum lycopersicum) and grape (Vitis vinifera) . The identification of rice miR-168a at fM concentration level in various mouse tissues may indicate the possibility of a cellular active uptake system for circulating RNA . These data are in agreement with the role of defined food matrices as effective delivery systems protecting ncRNAs from interaction with macrophages, dendritic cells, B-lymphocytes, and T-lymphocytes found in Peyer's patches and other sites of gut-associated lymphoid tissue. However, several other studies contradict these results (see section 3.3.3. for details).

Enrichment of miRNAs with targets related to immune response has been observed in colostral milk . Colostrum represents a primordial food driving the first immunization of offspring and provides the nutritional needs of their immature organs in the earliest stages of life. The facts that immune-related miRNAs from host, and exogenous plant miRNAs (from bamboo diet to milk exosomes of giant panda) targeting synapse organization, neuron migration or axon guidance are enriched in giant breast milk of panda (restricted to a diet primarily comprising bamboo) indicate that breast milk may facilitate dietary intake of plant miRNAs by infants for possible regulation of postnatal development ).

Very few studies address feeding plant ncRNAs to animals (i.e. weaning, adult and old age) to evaluate induction of specific adaptive immune response(s) (i.e. Th1, Th2 or Th17). Similarly, there is a lack of data on the impact of exogenous plant ncRNAs on expression of MHC-II molecules to assess the potential allergenicity of these RNAs. Additionally, the different immune response(s) induced by the administration of distinct exogenous plant ncRNAs and well-known weak or strong immune inducers by different routes (intraperitoneal or intragastric) have not been tested. Understanding of the allergenic functions of exogenous plant ncRNAs is comparatively meagre in relation to proteins. The homology of distinct ncRNAs could elicit different immune response(s) as proteins showing homology values up to 86% on their amino acid sequences induce different IgG1 and IgG2 responses (Adel-Patient et al., 2011) . For example, the literature search identified one study on safety assessment of the possible effects of GM-triticale on the immune system using a murine C57BL/6 model (Krzyżowska et al., 2010) . Immunoblotting analysis in serum showed significantly increased IL-2 levels, but not those of IFN-γ. It was concluded that multigenerational use of feeds for rodents containing the GM-triticale leads to expansion of the B cell compartment in the secondary lymphoid organs not associated to allergic response(s) (Krzyżowska et al., 2010) . However, this study did not report the ncRNA content in GMtriticale.

Efforts over the last decade have focused on elucidating the role of ncRNAs in immune function. Although no immunogenic potential has been found, ncRNAs implications in pathways regulating the production of inflammatory mediators (Witwer et al., 2010) , the development of hematopoietic cells (Satpathy and Chang, 2015) , regulation of TLRs signalling (Carpenter et al., 2013) and changes in T cells needed to support clonal expansion and growth have been highlighted. Papers on ncRNAs-mediated regulation of the amplitude and quality of immune response(s) could not be found. The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s).

At the moment, knowledge on the role of ncRNAs in immune responses at systems-level remains incomplete.

Earlier reports defined significant differences in the effects of dsRNA (Poly I:C) in innate immune responses as a function of 'length' (0.5-1.5 kb and 2.0-8.0 kb), depending on cell type (i.e. splenocytes, PBMCs, RAW264.7 and THP-1) (Mian et al., 2013) . Notably, it was reported that the 'shorter' dsRNA caused higher immunomodulatory effects. Poly I:C, a synthetic analogue of viral dsRNA, has long been known as a strong inducer of innate immune responses by interacting with TLR3 in the endosome. Following ligand recognition, specific ligand cascades involving interferon regulatory proteins and nuclear factor kappa B (NF-B) are activated to produce inflammatory mediators (Démoulins et al., 2009) . Similarly, signal-specific lncRNAs have been reported in dendritic cells (DCs) stimulated with TLR4 agonists, 80% of which clustered with NF-B signalling components (Guttman et al., 2009) . Further approaches also identified upregulated responses in a long intergenic noncoding RNA (lincRNA) in response to TLR-1, -2, -7 and -8, but not TLR-3 in macrophages (Carpenter et al., 2013) . Specifically, lincRNA-Cox2 was found to mediate both activation and repression of immune response genes dependent on interactions of lincRNA-Cox2 with heterogeneous nuclear ribonucleoprotein A/B and A2/B1. A further study determined that lincRNA-Cox2 plays a key role as a coactivator of NF-B for the transcription of late-primary inflammatory response genes in macrophages through modulation of epigenetic chromatin remodelling (Hu et al., 2016) .

Due to the relevance of gut microbiota in immune system development and homeostasis, the possibility of modulating the gut microbiota through dietary exogenous ncRNAs was reviewed. As described in Section 1.3, during the preparatory phase this novel topic was identified as relevant to this literature review. This additional section serves as a general introduction to emphasize the relevance of the possible modulation of the immune system by dietary exogenous ncRNAs through the gut microbiota, and to identify gaps pointing to needs for future studies relevant to food/feed risk assessment of ncRNAbased GM plants. Specific details of the possible mechanisms of action, if any, are outside the scope of this review. Due to the novelty of this topic and the lack of information on the scientific literature, only 5 documents were reviewed for this section.

The intestinal tract harbours a complex community (microbiota), which has an enormous impact not only on the nutritional, but also immune status of the host. A balanced gut microbiota composition confers benefits to the host, while microbial imbalances are associated with metainflammatory disorders. The dynamic processes initiated upon birth define the microbiota composition that co-evolves with the host and its environment. Thus, colonization of the intestine in early life seems particularly important as it represents one of the major environmental stimuli for immune system maturation. Particularly, breast feeding is crucial to influencing early intestinal colonization. As previously indicated, development of emerging 'next generation sequencing technologies' has allowed identification of a significant fraction of the circulating RNA in human plasma that originated from exogenous bacterial and fungi species . This study, based on microarray profiling results, reported that plasma can contain ncRNAs from common foods influencing the expression profile of a number of genes in the cells. However, several other studies suggest that the quantitative amount of ncRNAs are irrelevant for influencing gene expression (see section 3.3.3. for details).

Though the interaction between humans and their environments, particularly microbes, raises the possibility of some sort of feedback signalling process, it is still not fully understood. The literature search showed how few publications address the role of ncRNAs in shaping gut microbiota. Emerging The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s). evidences point out interactions between environmental factors and inter-species gene regulation facilitating host control of the gut microbiota Choi et al., 2017) . However, the impact of exogenous plant ncRNAs on gut microbiota modulation remains elusive.

Increasing evidence demonstrates that a significant number of ncRNAs are found in the extracellular media, and it is apparent that mammalian cells release sRNA that might carry functions which transcend the confines of single cells (Valadi et al., 2007) . Few studies have identified changes in miRNA expression in human stool. Overall, they provide an inconsistent association between their level and dietary habits and lifestyle (Tarallo et al., 2014; . In this context, the survival of exogenous (orally administered) plant miRNAs in faeces from mice has been observed . Although the amount of particular miRNAs such as miR-172 that survived the GI tract reached a maximum of 4.5%, this study does not support consistent survival rates for distinct ncRNAs.

A recent study identified miRNAs secreted by intestinal epithelial cells in intestinal contents and supports their role in modulating gut microbiota composition . This study reports the potential of host miRNAs to enter bacteria and co-localize with microbial nucleic acids. Similarly, exogenous plant miRNAs that were present in animal faeces were indicated as being primarily acquired orally . Recently, bacterial RNAs comparable in size to miRNAs have been identified that could serve as signalling molecules mediating bacteria-to-human interaction (Choi et al., 2017) . These data were obtained from outer membrane vesicles produced by periodontopathogens and were able to exert 'slight' immunosuppressive effects on T cells cytokine production (IL-5, IL-13 and IL-15).

MiRNA expression levels and amount in stool appear to be modulated by diet (micro-and macronutrients, phytochemicals) without excluding the possibility that the results were affected by other lifestyle factors; however, the influence of GM plant exogenous ncRNAs has not yet been exhaustively studied in animals or humans.

Distinct innate immune mediators participating in the recognition and discrimination between the distinct nucleic acids have been identified; thus, additional definition of the dynamic interplay between the different 'sensors' (i.e. TLRs, RIG-I/MDA5, the interferon response effectors protein kinase R and RNase L) in response to ncRNAs have to be clarified.

The chemical modifications observed in plant ncRNAs highlight the need to address the detection and measurement of exogenous plant ncRNAs in body fluids, as well as the sensitivity and specificity of these RNAs to cellular and tissue targets.

Although many reports focusing on synthetic ss-and dsRNA mimics can be found in the literature, scientifically assessed direct hazardous impacts of food and feed on fauna and flora are scarce and conflicting. Notably, there is a need to evaluate the effects of exogenous plant ncRNAs at the system biology level and in human trials. Current knowledge on the immune functions of exogenous ncRNAs is comparatively lacking. Most studies use cell models based on isolated cells that could not mimick the whole set of players that determine immune response(s), which are necessary for mechanistic studies, but need approach cell-to-cell communication in target tissues. Additionally, these studies should focus on the relationship of immune responses to production of IFNs since TLRs signalling converge on the production of this immune regulatory factor.

To better understand the impact of exogenous ncRNAs in host microbiota composition and diversity, further studies are needed to establish the microbial targets and their effect on physiological conditions.

Understanding the basic biological aspects of plant-derived exogenous ncRNA (including their half-life and stability), gathering information on human and animal exposure by diet to these molecules, their subsequent bioavailability and possibility to exert local or systemic effects, and reviewing the experience from the development of RNA-based therapeutics are relevant aspects for the risk assessment of ncRNAbased GM food and feed.

The literature includes very few studies evaluating the half-life of plant-derived ncRNAs. While no specific studies on plant miRNAs and lncRNAs half-life were found in this review, the studies on plant circRNAs half-life suggest that they exhibit considerable stability as compared to linear RNAs (section 3.1.2.1). In mammals, the few reports related to ncRNAs half-life suggest that miRNAs and circRNAs are generally much more stable than mRNA, while lncRNAs have a half-life like that of mRNA. However, when assessing the stability of plant ncRNAs outside the plant (section 3.1.2.2.), compelling evidence exists that plant miRNAs are highly stable under different conditions including food storage, processing, cooking, or simulated digestion. Moreover, they seem to survive after long incubation in serum, or are detected in the gastric content of mice, suggesting that plant miRNAs are more resistant to degradation than synthetic or animal miRNAs. This high resistance has been attributed to the unique characteristics of plant miRNAs (and siRNAs), in contrast to miRNAs from mammalian and other organisms; the 2'-Omethylation at their 3' end which confers plant miRNAs more resistance to degradation (section 3.1.1.1). EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s).

The experience from development of ncRNA therapeutics (section 3.1.3) shows that unmodified nucleic acids (i.e. RNAs) exhibit very low stability in biological media and are subjected to rapid nuclease mediated degradation. Indeed, a plethora of RNases are encoded within most genomes, often with overlapping activities, making redundancy a general feature of RNA degradation systems. Consequently, different chemical modifications or conjugations have been introduced to potential RNA-based therapeutics (section 3.1.3.2) to improve their RNA-binding affinity, their in vivo nuclease stability and their pharmacokinetic and pharmacodynamic properties. The available literature also suggests that naked or unmodified RNAs parenterally administered (e.g. iv) are rapidly cleared from circulation and their presence in the kidney and urine has been reported immediately after administration (section 3.1.4). Similarly, following oral administration naked or unmodified exogenous RNAs are rapidly degraded when exposed to the GI harsh conditions, and different delivery vehicles for exogenous RNAs have been developed. In general, and in contrast to chemically modified or formulated RNAs, no major biological effects have been observed for naked RNAs, which has probably limited further evaluation in humans (section 3.1.4.7). Although certain disease conditions may influence the pharmacokinetics of exogenous RNAs, there are no studies that specifically evaluate this for naked unmodified exogenous ncRNAs.

To exert a biological effect, an exogenous ncRNA first needs to reach the intended target tissue in sufficient quantity levels, this implying the necessity to overcome many biological barriers (section 3.1.5). Considering the oral intake as the main possible entry point of exogenous plant ncRNAs into both animals and humans, the first major barrier is the mammalian GI tract, which encompasses a group of extracellular and cellular barriers. Extracellular barriers include the presence of different enzymes (i.e. nucleases), the harsh environment and a net negative charged mucous layer. The cellular barriers include a three-layer (epithelial cells or enterocytes, the lamina propria and the muscularis mucosa) barrier known as the intestinal mucosa. Possible ncRNAs passage between cells is limited by the presence of tight junctions and the pore size in the human intestine would prevent the passage of all but miRNAs, which are the smallest ncRNAs. Although crossing the cells by transcytosis may be possible, this mechanism implies facing new intracellular barriers, such as nucleases, recycling of ncRNAs back to the lumen and nuclear uptake. However, M-cells, present in the gut epithelium interspersed between enterocytes, present certain characteristics that would make them more amenable for RNA uptake. A second barrier is represented by a set of obstacles during the passage from the intestine to the target tissue, which encompasses plasma and tissue nucleases. Once in the circulatory system, exogenous ncRNAs are subjected to distribution and elimination. For example, small RNAs would be rapidly cleared by the kidney due to their small size. Furthermore, ncRNAs need to escape the reticuloendothelial system, the function of which is to clear foreign pathogens. Finally, they would have to cross the vascular endothelial barrier to reach the target tissue, facing hurdles like those in the enterocytes. Packaging or incorporation of ncRNAs into extracellular transport systems (i.e. exosomes/microvesicles or inclusion in ribonucleoprotein complexes) would confer certain resistance to nucleases and other barriers.

ncRNAs must enter the target cells to exert their functions. The molecular mechanisms of exogenous ncRNA cellular uptake have been inferred from studies performed when developing strategies for nucleic acids delivery as therapeutics and is also largely derived from invertebrates, with little data reported in mammals. Descriptions exist of receptor-mediated uptake of oligonucleotides through the SIDT1 and SIDT2 proteins, although this is now in doubt since they have been described as cholesterol transporters. Due to the charged nature of RNAs, cellular uptake could be achieved by endocytosis after which they would enter the system of intracellular trafficking through multiple membrane-bound compartments. ncRNAs must exit these compartments to reach their functional sublocation within the EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s).

cell, either the cytosol or the nucleus, while simultaneously escaping degradation in the lysosomes. Specialized barriers such as the blood-brain barrier or the placental barrier represent additional obstacles to overcome.

While RNA content in plant-derived foods varies (≈1 mg/g of tissue), it could be estimated that when consuming a daily dose of total fruit and vegetables of ≈600 g/day, humans theoretically ingest 600 mg of total RNAs (dietary intake of plant RNA may typically range from 0.1-1 g/person/day) (section 3.1.6.1). Humans following special diets, such as vegetarians and vegans, have higher intake of plant origin RNAs, but estimated exposure to such RNAs implies only a 3-fold increase on average (section 3.1.6.2). It has been estimated that small RNAs make up far less than 5% of total RNA in plants, suggesting a very low exposure to plant ncRNAs under normal conditions. Whether sufficient levels of exogenous plant ncRNAs can be consumed in the diet to exert a biological effect is also another aspect that needs to be evaluated. Some examples in the literature suggest that for certain plant miRNAs (i.e. miR-156a) a person would need to consume 1670 kg of cantaloupe to reach the 100 copies of miRNA per cell possibly determining an effect. These aspects should be evaluated for each plant ncRNA (on a case basis), considering that some plant-derived miRNAs have shown unexpectedly high resistance to degradation and greater bioavailability.

The available literature indicates a widespread presence of exogenous RNAs (including plant derived ncRNAs, i.e. small RNAs) in the biological fluids of humans and animals (section 3.3.2). Whether these RNAs are derived from dietary intake is unclear. Their generally low abundance and lack of enrichment in tissues mostly exposed to dietary changes (i.e. liver) suggest that some of may have be originated as technical artefacts or through contamination.

Very few studies address biological effects of dietary exogenous ncRNAs in the GI tract and its annex glands (e.g. liver) (section 3. can directly target mammalian genes through RNAi effects, but still clarification is needed if exogenous plant ncRNAs can overcome all the biological barriers to reach and interact with their possible targets. Exogenous plant-derived ncRNAs have been found in exosomes or macrovesicles. How they reach these types of structures in biological fluids is unknown. In summary, supporting and contradicting evidence concerning the existence of systemic effects of dietary plant-derived exogenous ncRNAs is heavily debated. Important aspects such as the precise mechanism/s of transport of plant ncRNAs from food into the systemic circulation, the amount of exogenous ncRNAs reaching tissues or the molecular mechanisms of cellular uptake need to be determined.

While several in vitro studies cover the participation of exogenous ncRNAs in the dynamic network to modulate innate immunity responses, few tackle the immunomodulatory effects of plant ncRNAs in adaptive immune response (section 3.4). None of these studies have considered the potential impact of exogenous dietary plant-derived ncRNAs at the systemic level.

Since the first publication suggesting a possible cross-kingdom effect by dietary ncRNA (through rice miR-168a) the knowledge on exogenous ncRNAs, and particularly plant-derived exogenous ncRNAs, on EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s).

their fate when ingested has increased. However, the knowledge useful to support the assessment of dietary ncRNA such as ncRNA-based GM food and feed still faces several gaps.

An important aspect needing further investigation is the stability of plant ncRNAs. The meagre available literature contains a very few studies of the half-life of plant ncRNAs (i.e. miRNAs, lncRNAs and circRNAs) both inside the plant cell and outside the plant. Specifically, the stability of circRNAs should be assessed due to their apparently high stability. Moreover, very few reports focus on assessment of plant ncRNAs stability in mammalian systems, when ingested. Understanding the molecular basis that confers plant ncRNAs stability is of great importance and needs to be further evaluated. For instance, plant miRNAs and siRNAs carry modifications that do not occur in mammals. These modifications, such as the presence of a methyl group located on the 3' nucleotide ribose, could hypothetically have an impact on the plant miRNAs stability in mammalian systems/cells due to either the lack of the appropriate enzymes for recognition and degradation of plant miRNAs by mammalian cells, or the increased stability to degradation by mammalian RNases. The circularized structure of plant circRNAs would hypothetically confer them increased stability to mammalian RNAses. If this is the case, plant miRNAs and circRNAs could exhibit a much longer half-life in mammalian systems than expected, especially when compared to mammalian miRNA or circRNAs. This would increase their chances of reaching the appropriate target tissue and encountering suitable target molecules within the cells. The half-life of plant ncRNAs within mammalian cells is also an important aspect to consider. Experiments are needed with labelled plant ncRNAs, first in transfected mammalian cells (in vitro studies) and then orally administered to experimental animals (in vivo studies). Radioactive probes or molecules of very low molecular weight, such as biotin or fluorescein, should be used, in order to prevent distorting the actual size of the studied plant ncRNA; this is important when studying miRNAs stability due to their small size.

Another aspect possibly relevant for risk assessment is to understand whether and how ingested GM plant-derived ncRNAs can affect the expression levels of other ncRNAs and other RNAs, due to putative compensatory circuits and codifying RNAs. Whole transcriptome sequencing, comparing the RNA levels of unmodified (wild type) to those of modified (transgenic) plants coul be useful to this aim.

Although available information suggests that many biological barriers exist in mammalian system to ingested plant exogenous ncRNAs preventing these to exert a possible local or systemic effect, several gaps are still present. Putative pathways of dietary exogenous ncRNAs after ingestion and the many transporters potentially involved (i.e, in cellular uptake, or barrier passage transport pathways) in the transit from the gut to the target tissues should be characterized using standard molecular biology gainof-function and loss-of-function approaches. Also, the many routes of mammalian intracellular trafficking to the specific intracellular target should be studied. To this aim, ncRNAs isolated from human diets should be used preferentially, since most of the incomplete current knowledge comes from the RNA-therapeutics field or is derived from the invertebrate world. Which molecular mechanisms support the possible passage through specialized barriers (i.e. blood-brain barrier or the placental barrier), is another critical aspect requiring further studies.

While there are in vitro studies describing the interaction between plant miRNAs and mammalian miRNA silencing complexes possibly leading to target repression, these findings need to be experimentally validated in vivo. Therefore, more studies are needed to understand the possibility of processing plant ncRNAs in mammalian cells. It is also unknown if ingested plant miRNAs reaches or needs to reach a necessary level to exert a biological effect. For instance, studies in mammals suggest that the threshold for target gene regulation is >100 copies per cell for a certain study; however, another study suggests around 1000 to 10.000 copies of mammalian miRNA per cell. Clearly, a minimum amount of small RNAs EFSA Supporting publication 2019:EN-1688

The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s).

would be required to achieve biologically relevant effects on gene expression. On a case by case basis it would be useful to evaluate the amount of consumption of a given ncRNA in a normal diet. It is also relevant to study if exposure to these plant ncRNAs could change under certain pathological conditions (i.e. compromised intestinal permeability or renal function) or dietary patterns (i.e. vegetarian or vegan diets).

Although tissue accumulation of dietary plant exogenous ncRNAs has not been reported to date, it seems clear that plant miRNAs and other exogenous RNAs are present in biological fluids from humans and animals. The biological significance of their presence is still unknown and needs to be addressed. Methods that do not require amplification for detection should be used. More quantitative approaches to determine the levels of plant-derived exogenous ncRNAs should be determined both in biological fluids and tissues to understand the magnitude of presence. For example, this could be partially obtained from studies reporting the presence of exogenous RNAs in biological fluids using public small RNA-seq databases. In the case of plant-derived miRNAs, the use of sodium periodate oxidation of samples could be a good start to verifying the presence of genuine plant-derived miRNAs from diet in human and animal biological fluids.

The immunomodulatory effects of exogenous RNAs have been widely described in the literature.

However, there is no in vivo information regarding the immunomodulatory effects of dietary exogenous ncRNAs. These types of studies should be performed both in vitro and in vivo to determine if the different types of ncRNAs consumed in the diet can affect the immune system.

In the case of target gene repression by plant ncRNAs investigations in specific animal models could be considered. Expression levels of other ncRNAs and RNAs, which may be modified due to putative compensatory circuits could be evaluated in these models, as well as information on the stability and half-life of the specific ncRNA in specific cells and systemically following ingestion. 

Genome-Wide High-Resolution Mapping of Exosome Substrates Reveals Hidden Features in the Arabidopsis Transcriptome

Genome-Wide Identification of Circular RNAs in Arabidopsis thaliana

A lariat-derived circular RNA is required for plant development in Arabidopsis

Pattern formation via small RNA mobility

Small RNAs are on the move

Literature review of baseline information on RNAi to support the environmental risk assessment of RNAi-based GM plants

Identification of Circular RNAs from the Parental Genes Involved in Multiple Aspects of Cellular Metabolism in Barley

A group I plant intron accumulates as circular RNA forms with extensive 5' deletions in vivo

Toward a Consensus on the Binding Specificity and Promiscuity of PRC2 for RNA

A stepwise pathway for biogenesis of 24-nt secundary siRNAs and spreading of DNA methylation

Site-Specific Phosphorylation Profiling of Arabidopsis Proteins by Mass Spectrometry and Peptide Chip Analysis

Hierarchical Action and Inhibition of Plant Dicer-Like Proteins in Antiviral Defense

Characterization of stress-responsive lncRNAs in Arabidopsis thaliana by integrating expression, epigenetic and structural features

Identification and characterization of human testis derived circular RNAs and their existence in seminal plasma

The RNA-binding proteins HYL1 and SE promote accurate in vitro processing of pri-miRNA by DCL1

Genome-wide Discovery of Circular RNAs in the Leaf and Seedling Tissues of Arabidopsis Thaliana

Intra-and intercellular RNA interference in Arabidopsis thaliana requires components of the microRNA and heterochromatic silencing pathways

DICER-LIKE 4 is required for RNA interference and produces the 21-nucleotide small interfering RNA component of the plant cell-to-cell silencing signal

Plant Mobile Small RNAs

The Arabidopsis thaliana doublestranded RNA binding protein DRB1 directs guide strand selection from microRNA duplexes

Naturally occurring variations in sequence length creates microRNA isoforms that differ in argonaute effector complex specificity

The Xist lncRNA Exploits Three-Dimensional Genome Architecture to Spread Across the X Chromosome

Identification of Nuclear Dicing Bodies Containing Proteins for MicroRNA Biogenesis in Living Arabidopsis Plants

Phased, Secondary, Small Interfering RNAs in Posttranscriptional Regulatory Networks

Antisense RNA polymerase II divergent transcripts are P-TEFb dependent and substrates for the RNA exosome

Long Noncoding RNAs in Cell-Fate Programming and Reprogramming

Target mimicry provides a new mechanism for regulation of microRNA activity

Location of a Possible miRNA Processing Site in SmD3/SmB Nuclear Bodies in Arabidopsis

Partially Redundant Functions of Arabidopsis DICER-like Enzymes and a Role for DCL4 in Producing trans-Acting siRNAs

Short Integuments1/suspensor1/carpel Factory, a Dicer Homolog, Is a Maternal Effect Gene Required for Embryo Development in Arabidopsis

Oncogenic Role of Fusion-circRNAs Derived from Cancer-Associated Chromosomal Translocations

Topological organization of multichromosomal regions by the long intergenic noncoding RNA Firre

Natural RNA circles function as efficient microRNA sponges

Vernalization-mediated epigenetic silencing by a Long Intronic Noncoding RNA

Flowering Locus C's Lessons: Conserved Chromatin Switches Underpinning Developmental Timing and Adaptation

Transitivity-dependent andindependent cell-to-cell movement of RNA silencing

Specific interactions between Dicer-like proteins and HYL1/DRB-family dsRNA-binding proteins in Arabidopsis thaliana

Phosphate utilization efficiency correlates with expression of low-affinity phosphate transporters and noncoding RNA, IPS1, in Barley

Gene silencing by microRNAs: contributions of translational repression and mRNA decay

Cyclophilin 40 facilitates HSP90-mediated RISC assembly in plants

In Vitro Assembly of Plant RNA-Induced Silencing Complexes Facilitated by Molecular Chaperone HSP90

A rice cis-natural antisense RNA acts as a translational enhancer for its cognate mRNA and contributes to phosphate homeostasis and plant fitness

The Nuclear Pore Protein AtTPR Is Required for RNA Homeostasis, Flowering Time, and Auxin Signaling

Dicer-1 and R3D1-L catalyze microRNA maturation in Drosophila

Molecular insights into miRNA processing by Arabidopsis thaliana SERRATE

Covalent circularization of exogenous RNA during incubation with a wheat embryo cell extract

Small RNAs as Guardians of the Genome

Fast-Forward Genetics Identifies Plant CPL Phosphatases as Regulators of miRNA Processing Factor HYL1

The evolution and diversification of Dicers in plants

Passenger-Strand Cleavage Facilitates Assembly of siRNA into Ago2-Containing RNAi Enzyme Complexes

Circular single-stranded RNA replicon in Saccharomyces cerevisiae

RNA-directed DNA methylation: an epigenetic pathway of increasing complexity

Intercellular and systemic movement of RNA silencing signals

Circular RNAs are a large class of animal RNAs with regulatory potency

Sorting of small RNAs into Arabidopsis Argonaute complexes is directed by the 5' terminal nucleotide

MORC Family ATPases Required for Heterochromatin Condensation and Gene Silencing

Small Silencing RNAs in Plants Are Mobile and Direct Epigenetic Modification in Recipient Cells

Specificity of ARGONAUTE7-miR390 Interaction and Dual Functionality in TAS3 Trans-Acting siRNA Formation

Evolution of animal and plant dicers: early parallel duplications and recurrent adaptation of antiviral RNA binding in plants

Literature review of baseline information to support the risk assessment of RNAi-based GM plants

Systemic acquired silencing: transgene-specific post-transcriptional silencing is transmitted by grafting from silenced stocks to non-silenced scions

MicroRNA399 is a long-distance signal for the regulation of plant phosphate homeostasis

Evidence for Nuclear Processing of Plant Micro RNA and Short Interfering RNA Precursors

Nuclear processing and export of microRNAs in Arabidopsis

CARPEL FACTORY, a Dicer Homolog, and HEN1, a Novel Protein, Act in microRNA Metabolism in Arabidopsis thaliana

RDE-4 preferentially binds long dsRNA and its dimerization is necessary for cleavage of dsRNA to siRNA

RNA Exosome-Regulated Long Non-Coding RNA Transcription Controls Super-Enhancer Activity

Prediction of trans-antisense transcripts in Arabidopsis thaliana

Genome-wide identification of long noncoding natural antisense transcripts and their responses to light in Arabidopsis

Identification and characterization of circRNAs in Pyrus betulifolia Bunge under drought stress

Analysis of noncoding transcriptome in rice and maize uncovers roles of conserved lncRNAs associated with agriculture traits

A long noncoding RNA maintains active chromatin to coordinate homeotic gene expression

Arabidopsis noncoding RNA mediates control of photomorphogenesis by red light

Identification of Circular RNAs and Their Targets in Leaves of Triticum aestivum L. under Dehydration Stress

Importin 8 Is a Gene Silencing Factor that Targets Argonaute Proteins to Distinct mRNAs

Noncoding Transcription by RNA Polymerase Pol IVb/Pol V Mediates Transcriptional Silencing of Overlapping and Adjacent Genes

Expression of Arabidopsis MIRNA Genes

Genetic and functional diversification of small RNA pathways in plants

Identification and characterization of wheat long non-protein coding RNAs responsive to powdery mildew infection and heat stress by using microarray analysis and SBS sequencing

Bidirectional promoters generate pervasive transcription in yeast

Functions of long intergenic non-coding (linc) RNAs in plants

Widespread noncoding circular RNAs in plants

A Systemic Small RNA Signaling System in Plants

WAVY LEAF1, an Ortholog of Arabidopsis HEN1, Regulates Shoot Development by Maintaining MicroRNA and Trans-Acting Small Interfering RNA Accumulation in Rice

VIGS, HIGS and FIGS: small RNA silencing in the interactions of viruses or filamentous organisms with their plant hosts

Control of coleopteran insect pests through RNA interference

The expanding world of small RNAs in plants

Genome-wide analysis of long noncoding RNA stability

Structural variations and stabilising modifications of synthetic siRNAs in mammalian cells

Circular RNAs are long-lived and display only minimal early alterations in response to a growth factor

Host-delivered RNAi: an effective strategy to silence genes in plant parasitic nematodes

Analysis of microRNA turnover in mammalian cells following Dicer1 ablation

Polysome arrest restricts miRNA turnover by preventing exosomal export of miRNA in growth-retarded mammalian cells

A Versatile Monosaccharide Transporter That Operates in the Arbuscular Mycorrhizal Fungus Glomus sp Is Crucial for the Symbiotic Relationship with Plants

The Drosophila RNA Methyltransferase, DmHen1, Modifies Germline piRNAs and Single-Stranded siRNAs in RISC

Uridylation of mature miRNAs and siRNAs by the MUT68 nucleotidyltransferase promotes their degradation in Chlamydomonas

Circular RNAs are abundant, conserved, and associated with ALU repeats

Regulation of small RNA stability: methylation and beyond

Methylation Protects miRNAs and siRNAs from a 3' End Uridylation Activity in Arabidopsis

Assessing the survival of exogenous plant microRNA in mice

Effective detection and quantification of dietetically absorbed plant microRNAs in human plasma

Adenylation of plant miRNAs

Covalent circularization of exogenous RNA during incubation with a wheat embryo cell extract

Silencing a cotton bollworm P450 monooxygenase gene by plant-mediated RNAi impairs larval tolerance of gossypol

RNA-specific ribonucleotidyl transferases

Oral delivery of siRNA and antisense oligonucleotides

Toxicogenomics of non-viral drug delivery systems for RNAi: potential impact on siRNA-mediated gene silencing activity and specificity

Nonviral delivery of synthetic siRNAs in vivo

A theoretical basis for a biopharmaceutic drug classification: the correlation of in vitro drug product dissolution and in vivo bioavailability

Oral nucleic acid therapy using multicompartmental delivery systems

The adherent gastrointestinal mucus gel layer: thickness and physical state in vivo

Polymeric nano-and microparticle technologies for oral gene delivery

Isolation and properties of an extracellular nucleases of serratia marcescens

Guidance for Industry on Waiver of In Vivo Bioavailability and Bioequivalence Studies for Immediate-Release Solid Oral Dosage Forms Based on a Biopharmaceutics Classification System

Issues in oral nanoparticle drug carrier uptake and targeting

In vitro and in vivo models for the study of oral delivery of nanoparticles

Alginate/poly-L-lysine microparticles for the intestinal delivery of antisense oligonucleotides

Multi-compartmental oral delivery systems for nucleic acid therapy in the gastrointestinal tract

Extracellular ribonucleases from Bacillus subtilis. I. Crystallization and specificity

Intestinal delivery of non-viral gene therapeutics: physiological barriers and preclinical models

First-pass elimination. Basic concepts and clinical consequences

Key issues in non-viral gene delivery

Polymeric nanoparticle drug delivery technologies for oral delivery applications

Effect of sodium caprate on the intestinal absorption of two modified antisense oligonucleotides in pigs

Drug delivery of siRNA therapeutics: potentials and limits of nanosystems

Oral delivery of antisense oligonucleotides in man

The human microbiome project consortium. 2012 Structure, function and diversity of the healthy human microbiome

Milk-derived exosomes for oral delivery of paclitaxel

Delivery of siRNA to the mouse brain by systemic injection of targeted exosomes

Orally delivered siRNA targeting macrophage Map4k4 suppresses systemic inflammation

Oral administration of bovine milk derived extracellular vesicles attenuates arthritis in two mouse models

Oral nucleic acid therapy using multicompartmental delivery systems

Oral delivery of RNase P ribozymes by Salmonella inhibits viral infection in mice

Protection and Systemic Translocation of siRNA Following Oral Administration of Chitosan/siRNA Nanoparticles

Interleukin 10 gene transfer prevents experimental colitis in rats

Inhibition of TrkB limits development of the zebra finch song system

Commercial Dairy Cow Milk microRNAs Resist Digestion under Simulated Gastrointestinal Tract Conditions

Oral IL-10 gene delivery in a microsphere-based formulation for local transfection and therapeutic efficacy in inflammatory bowel disease

The role of nucleotides in human nutrition

Retroviral delivery of RNA interference against Marek's disease virus in vivo

Biogenesis, secretion, and intercellular interactions of exosomes and other extracellular vesicles

Systemic exosomal siRNA delivery reduced alpha-synuclein aggregates in brains of transgenic mice

Pancreatic cancer exosomes initiate pre-metastatic niche formation in the liver

Proteomics of extracellular vesicles: Exosomes and ectosomes

Dietary yeast RNA supplementation reduces mortality by Aeromonas hydrophila in rohu (Labeo rohita L.) juveniles

Effects of mannose density on in vitro and in vivo cellular uptake and RNAi efficiency of polymeric nanoparticles

PEG-PLA nanoparticles facilitate siRNA knockdown in adult zebrafish heart

Exosome-mediated Delivery of Hydrophobically Modified siRNA for Huntingtin mRNA Silencing

Oral delivery of small RNA and DNA

Colonic gene silencing using siRNA-loaded calcium phosphate/PLGA nanoparticles ameliorates intestinal inflammation in vivo

A randomized controlled trial of preoperative oral supplementation with a specialized diet in patients with gastrointestinal cancer

A novel class of small RNAs: tRNA-derived RNA fragments (tRFs)

Small non-coding RNAs transfer through mammalian placenta and directly regulate fetal gene expression

Assessing the survival of exogenous plant microRNA in mice

Effective detection and quantification of dietetically absorbed plant microRNAs in human plasma

Human milk exosomes and their microRNAs survive digestion in vitro and are taken up by human intestinal cells

Detection of Plant miRNAs Abundance in Human Breast Milk

In silico identification of plant miRNAs in mammalian breast milk exosomes--a small step forward?

Detection of dietetically absorbed maizederived microRNAs in pigs

Exosomal microRNAs in giant panda (Ailuropoda melanoleuca) breast milk: potential maternal regulators for the development of newborn cubs

Unsuccessful Detection of Plant MicroRNAs in Beer, Extra Virgin Olive Oil and Human Plasma After an Acute Ingestion of Extra Virgin Olive Oil

A novel chemopreventive strategy based on therapeutic microRNAs produced in plants

Annotation of the goat genome using next generation sequencing of microRNA expressed by the lactating mammary gland: comparison of three approaches

Expressional analysis of immune-related miRNAs in breast milk

Identification of stable, high copy number, medium-sized RNA degradation intermediates that accumulate in plants under non-stress conditions

Determination of the potential bioavailability of plant microRNAs using a simulated human digestion process

Human Breast Milk miRNA, Maternal Probiotic Supplementation and Atopic Dermatitis in Offspring

The effect of dietary ginseng polysaccharide supplementation on the immune responses involved in porcine milk-derived esRNAs

Immune modulatory function of abundant immunerelated microRNAs in microvesicles from bovine colostrum

Uptake and Function Studies of Maternal Milk-derived MicroRNAs

Mining of public sequencing databases supports a nondietary origin for putative foreign miRNAs: underestimated effects of contamination in NGS

Literature review of information supporting food/feed risk assessment of RNAi-based GM plants www.efsa.europa

This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject

The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document

Alternative miRNAs: Human Sequences Misidentified As Plant miRNAs In Plant Studies And In Human Plasma

Transfer and functional consequences of dietary microRNAs in vertebrates: concepts in search of corroboration: negative results challenge the hypothesis that dietary xenomiRs cross the gut and regulate genes in ingesting vertebrates, but important questions persist

Real-time quantitative PCR and droplet digital PCR for plant miRNAs in mammalian blood provide little evidence for general uptake of dietary miRNAs: limited evidence for general uptake of dietary plant xenomiRs

The Intestinal Transport of Bovine Milk Exosomes Is Mediated by Endocytosis in Human Colon Carcinoma Caco-2 Cells and Rat Small Intestinal IEC-6 Cells

The levels of human milk microRNAs and their association with maternal weight characteristics

Detection of dietary plant-based small RNAs in animals

Anomalous uptake and circulatory characteristics of the plant-based small RNA MIR2911

The atypical genesis and bioavailability of the plant-based small RNA MIR2911: Bulking up while breaking down

Bioavailability of transgenic microRNAs in genetically modified plants

Exogenous plant MIR168a specifically targets mammalian LDLRAP1: evidence of cross-kingdom regulation by microRNA

Analysis of plant-derived miRNAs in animal small RNA datasets

Immune-related microRNAs are abundant in breast milk exosomes

Honeysuckle-encoded atypical microRNA2911 directly targets influenza A viruses

One Step Forward, Two Steps Back; Xeno-MicroRNAs Reported in Breast Milk Are Artifacts

Small RNAs from plants, bacteria and fungi within the order Hypocreales are ubiquitous in human plasma

Dietary MicroRNA Database (DMD): An Archive Database and Analytic Tool for Food-Borne microRNAs

Lack of detectable oral bioavailability of plant microRNAs after feeding in mice

Survey of 800+ data sets from human tissue and body fluid reveals xenomiRs are likely artifacts

Plant miRNAs found in human circulating system provide evidences of cross kingdom RNAi

Detection of Plant miRNAs Abundance in Human Breast Milk

In silico identification of plant miRNAs in mammalian breast milk exosomes--a small step forward?

Optimization of enzymatic reaction conditions for generating representative pools of cDNA from small RNA

Computational Characterization of Exogenous MicroRNAs that Can Be Transferred into Human Circulation

Mining of public sequencing databases supports a nondietary origin for putative foreign miRNAs: underestimated effects of contamination in NGS

The complex exogenous RNA spectra in human plasma: an interface with human gut biota?

The spectrum of circulating RNA: a window into systems toxicology

Detection of an Abundant Plant-Based Small RNA in Healthy Consumers

Oral administration of bovine milk derived extracellular vesicles attenuates arthritis in two mouse models

Exosomes as a Nanodelivery System: a Key to the Future of Neuromedicine?

Endogenous microRNA can be broadly exploited to regulate transgene expression according to tissue, lineage and differentiation state

Identification of Dietetically Absorbed Rapeseed (Brassica campestris L.) Bee Pollen MicroRNAs in Serum of Mice

Lack of detectable oral bioavailability of plant microRNAs after feeding in mice

High-throughput sequencing of RNA silencingassociated small RNAs in olive (Olea europaea L.)

Plant-derived phosphocholine facilitates cellular uptake of anti-pulmonary fibrotic HJT-sRNA-m7

Autosomal recessive hypercholesterolemia caused by mutations in a putative LDL receptor adaptor protein

Exosomes as therapeutic drug carriers and delivery vehicles across biological membranes: current perspectives and future challenges

Extensive degradation and low bioavailability of orally consumed corn miRNAs in mice

An improved method to quantitate mature plant microRNA in biological matrices using modified periodate treatment and inclusion of internal controls

This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document

Cross Talk between Adipose Tissue and Placenta in Obese and Gestational Diabetes Mellitus Pregnancies via Exosomes

Grape exosome-like nanoparticles induce intestinal stem cells and protect mice from DSS-induced colitis

Survey of 800+ data sets from human tissue and body fluid reveals xenomiRs are likely artifacts

Cross-Kingdom Regulation of Putative miRNAs Derived from Happy Tree in Cancer Pathway: A Systems Biology Approach

Small non-coding RNAs transfer through mammalian placenta and directly regulate fetal gene expression

Assessing the survival of exogenous plant microRNA in mice

Reply to Dr. Witwer's letter to the editor

Effective detection and quantification of dietetically absorbed plant microRNAs in human plasma

Plant miRNAs found in human circulating system provide evidences of cross kingdom RNAi

Detection of dietetically absorbed maizederived microRNAs in pigs

Negligible uptake and transfer of diet-derived pollen microRNAs in adult honey bees

The transport mechanism of extracellular vesicles at the blood-brain barrier

Unsuccessful Detection of Plant MicroRNAs in Beer, Extra Virgin Olive Oil and Human Plasma After an Acute Ingestion of Extra Virgin Olive Oil

A novel chemopreventive strategy based on therapeutic microRNAs produced in plants

Interspecies communication between plant and mouse gut host cells through edible plant derived exosome-like nanoparticles

High-throughput assessment of microRNA activity and function using microRNA sensor and decoy libraries

Milk extracellular vesicles accelerate osteoblastogenesis but impair bone matrix formation

Corn rootwormactive RNA DvSnf7: Repeat dose oral toxicology assessment in support of human and mammalian safety

A 28-day oral toxicity evaluation of small interfering RNAs and a long double-stranded RNA targeting vacuolar ATPase in mice

Citrus limon-derived nanovesicles inhibit cancer cell proliferation and suppress CML xenograft growth by inducing TRAIL-mediated cell death

This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document

Ineffective delivery of diet-derived microRNAs to recipient animal organisms

Adipose-derived circulating miRNAs regulate gene expression in other tissues

Uptake and Function Studies of Maternal Milk-derived MicroRNAs

Contamination or artifacts may explain reports of plant miRNAs in humans

Alternative miRNAs: human sequences misidentified as plant miRNAs in plant studies and in human plasma

Real-time quantitative PCR and droplet digital PCR for plant miRNAs in mammalian blood provide little evidence for general uptake of dietary miRNAs: limited evidence for general uptake of dietary plant xenomiRs

Detection of an Abundant Plant-Based Small RNA in Healthy Consumers

Anomalous uptake and circulatory characteristics of the plant-based small RNA MIR2911

Bioavailability of trasngenic microRNAs in genetically modified plants

Exosome delivered anticancer drugs across the blood-brain barrier for brain cancer therapy in Danio rerio

Molecular basis of mammalian transmissibility of avian H1N1 influenza viruses and their pandemic potential

Exogenous plant MIR168a specifically targets mammalian LDLRAP1: evidence of cross-kingdom regulation by microRNA

Honeysuckle-encoded atypical microRNA2911 directly targets influenza A viruses

Plant microRNAs in larval food regulate honeybee caste development

To evaluate oral toxicity of exogenous ncRNAs in mice, Petrick et al. administered for 28 days a repeated oral dose of siRNAs and dsRNA and evaluated several parameters including body weight, food consumption, clinical observations, clinical chemistry, haematology, gross pathology, or histopathology endpoints

It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s). signs (mortality, abnormalities, and sign of pain and distress), body weight and food consumption, haematology, organ weight, or pathology results. The minor differences observed in certain parameters were limited to single intervals, were not dose-related, and were attributed to interanimal variability

Torula yeast RNA (RNA negative control) at 100 mg/kg/day or a control vehicle were used. Ten animals per sex and per group were used. No treatment-related effects were observed on clinical signs (mortality, or sign of pain and distress), body weight and food consumption, haematology, organ weight, or pathology results. Minor changes in selected parameters in selected groups were observed, but these were attributed by the authors to normal variability and not treatment-related effects

This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document

Ecological risk assessment for DvSnf7 RNA: A plant-incorporated protectant with targeted activity against western corn rootworm

Literature review of baseline information on RNAi to support the environmental risk assessment of RNAi-based GM plants

A comparative evaluation of the regulation of GM crops or products containing dsRNA and suggested improvements to risk assessments

Endogenous small RNAs in grain: semi-quantification and sequence homology to human and animal genes

Literature review of baseline information to support the risk assessment of RNAi-based GM plants

Corn rootwormactive RNA DvSnf7: Repeat dose oral toxicology assessment in support of human and mammalian safety

A 28-day oral toxicity evaluation of small interfering RNAs and a long double-stranded RNA targeting vacuolar ATPase in mice

RNAi technologies in agricultural biotechnology: The Toxicology Forum 40th Annual Summer Meeting

No impact of DvSnf7 RNA on honey bee (Apis mellifera L.) adults and larvae in dietary feeding tests

Microstructure and ultrastructure of highamylose rice resistant starch granules modified by antisense RNA inhibition of starch branching enzyme

A 90-day toxicology study of high-amylose transgenic rice grain in Sprague-Dawley rats

A three generation reproduction study with Sprague-Dawley rats consuming high-amylose transgenic rice

High-amylose rice improves indices of animal health in normal and diabetic rats

Studies performed with human cells were mainly carried out ex vivo using lymphocytes and, to a significant extent, macrophages. From the 1730 studies retrieved during the screening phase on immune cells in relation to exogenous ncRNA

7%) from rat and 348 (36.5%) from mouse. And within the 776 studies on immune cells other than lymphocytes (i.e. macrophages), the studies were allocated into 378 (48.7%) from human, 91 (11.7%) from rat, and 307 (39.6%) from mouse. All these data clearly show that studies retrieved in relation to ncRNAs and www

This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document

Immunological and metabolomic impacts of administration of Cry1Ab protein and MON 810 maize in mouse

Role of Toll-like receptors in antisense and siRNA

Recognition of double-stranded RNA and activation of NF-kappaB by Toll-like receptor 3

RNA regulation of the immune system

Therapy of respiratory viral infections with intranasal siRNAs

CD14 is a coreceptor of Toll-like receptors 7 and 9

A long noncoding RNA mediates both activation and repression of immune response genes

Identification of Dietetically Absorbed Rapeseed (Brassica campestris L.) Bee Pollen MicroRNAs in Serum of Mice

Identification of dietetically absorbed rapeseed (Brassica campestris L.) bee pollen miRNAs in serum of mice

Chemical modification patterns compatible with high potency dicer-substrate small interfering RNAs

Is RNA interference involved in intrinsic antiviral immunity in mammals?

Poly (I:C) induced immune response in lymphoid tissues involves three sequential waves of type I IFN expression

Structural basis for cytosolic double-stranded RNA surveillance by human oligoadenylate synthetase 1

Structural mechanism of sensing long dsRNA via a noncatalytic domain in human oligoadenylate synthetase 3

RNAs Containing Modified Nucleotides Fail To Trigger RIG-I Conformational Changes for Innate Immune Signaling

This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document

Identification of RNA sequence motifs stimulating sequence-specific TLR8-dependent immune responses

dsRNA with 5′ overhangs contributes to endogenous and antiviral RNA silencing pathways in plants

A TIM-3 Oligonucleotide Aptamer Enhances T Cell Functions and Potentiates Tumor Immunity in Mice

Ab initio reconstruction of cell type-specific transcriptomes in mouse reveals the conserved multi-exonic structure of lincRNAs

Ribosome profiling provides evidence that large noncoding RNAs do not encode proteins

Chromatin signature reveals over a thousand highly conserved large noncoding RNAs in mammals

Lack of interferon response in animals to naked siRNAs

Species-specific recognition of single-stranded RNA via toll-like receptor 7 and 8

Small anti-viral compounds activate immune cells via the TLR7 MyD88-dependent signaling pathway

Sequencespecific potent induction of IFN-alpha by short interfering RNA in plasmacytoid dendritic cells through TLR7

LincRNA-Cox2 Promotes Late Inflammatory Gene Transcription in Macrophages through Modulating SWI/SNF-Mediated Chromatin Remodeling

MicroRNA 10a marks regulatory T cells

The role of alternative polyadenylation in the antiviral innate immune response

The 2'-O-methylation status of a single guanosine controls transfer RNA-mediated Toll-like receptor 7 activation or inhibition

Design of noninflammatory synthetic siRNA mediating potent gene silencing in vivo

Sequence-dependent stimulation of the mammalian innate immune response by synthetic siRNA

Immunostimulatory potential of silencing RNAs can be mediated by a non-uridine-rich toll-like receptor 7 motif

Suppression of RNA recognition by Toll-like receptors: the impact of nucleoside modification and the evolutionary origin of RNA

Sequence-and target-independent angiogenesis suppression by siRNA via TLR3

microRNA as a new immune-regulatory agent in breast milk

This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document

The effect of multigenerational diet containing genetically modified triticale on immune system in mice

Cross-Kingdom Regulation of Putative miRNAs Derived from Happy Tree in Cancer Pathway: A Systems Biology Approach

Double-stranded RNA-mediated TLR3 activation is enhanced by CD14

Molecular basis for the immunostimulatory activity of guanine nucleoside analogs: activation of Toll-like receptor 7

The TLR3 signaling complex forms by cooperative receptor dimerization

miR-181a is an intrinsic modulator of T cell sensitivity and selection

Assessing the survival of exogenous plant microRNA in mice

Structural basis of toll-like receptor 3 signaling with double-stranded RNA

The Host Shapes the Gut Microbiota via Fecal MicroRNA

In silico identification of plant miRNAs in mammalian breast milk exosomes--a small step forward?

Exosomal microRNAs in giant panda (Ailuropoda melanoleuca) breast milk: potential maternal regulators for the development of newborn cubs

MicroRNA-30c promotes natural killer cell cytotoxicity via up-regulating the expression level of NKG2D

Neonatal immune activation during early and late postnatal brain development differently influences depressionrelated behaviors in adolescent and adult C57BL/6 mice

Small self-RNA generated by RNase L amplifies antiviral innate immunity

A structural basis for discriminating between self and nonself double-stranded RNAs in mammalian cells

Cancer immunotherapy comes of age

Length of dsRNA (poly I:C) drives distinct innate immune responses, depending on the cell type

Oligonucleotide-based pharmaceuticals: Non-clinical and clinical safety signals and non-clinical testing strategies

Cytosolic viral sensor RIG-I is a 5'-triphosphate-dependent translocase on double-stranded RNA

Aptamers for CD Antigens: From Cell Profiling to Activity Modulation

CD28 aptamers as powerful immune response modulators

Influences of diet and the gut microbiome on epigenetic modulation in cancer and other diseases

This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document

Aptamer-Targeted Attenuation of IL-2 Signaling in CD8+ T Cells Enhances Antitumor Immunity

Catalog of Differentially Expressed Long Non-Coding RNA following Activation of Human and Mouse Innate Immune Response. Front Immunol 8, 1038

Long noncoding RNA in hematopoiesis and immunity

The effect of poly-L-lysine on the uptake of reovirus doublestranded RNA in macrophages in vitro

Characterization of the mammalian RNA exonuclease 5/NEF-sp as a testis-specific nuclear 3' → 5' exoribonuclease

MicroRNA regulation of lymphocyte tolerance and autoimmunity

Advances in RNA sensing by the immune system: separation of siRNA unwanted effects from RNA interference

Activation of the interferon system by short-interfering RNAs

Type I Interferons Function as Autocrine and Paracrine Factors to Induce Autotaxin in Response to TLR Activation

Antisense oligonucleotides containing locked nucleic acid improve potency but cause significant hepatotoxicity in animals

MicroRNA expression in relation to different dietary habits: a comparison in stool and plasma samples

Interplay between dengue virus and Toll-like receptors, RIG-I/MDA5 and microRNAs: Implications for pathogenesis

Exosome-mediated transfer of mRNAs and microRNAs is a novel mechanism of genetic exchange between cells

Dissolution media simulating the intralumenal composition of the small intestine: physiological issues and practical aspects

LPS injection reprograms the expression and the 3' UTR of a CAP gene by alternative polyadenylation and the formation of a GAIT element in Ciona intestinalis

The complex exogenous RNA spectra in human plasma: an interface with human gut biota?

Induced miR-99a expression represses Mtor cooperatively with miR-150 to promote regulatory T-cell differentiation

Evolutionary dynamics and tissue specificity of human long noncoding RNAs in six mammals

Toll-like receptor (TLR) 3 immune modulation by unformulated small interfering RNA or DNA and the role of CD14 (in TLRmediated effects)

Homology modeling of human Tolllike receptors TLR7, 8, and 9 ligand-binding domains

Modulation of let-7 miRNAs controls the differentiation of effector CD8 T cells

This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document

microRNA regulation of IFN-beta protein expression: rapid and sensitive modulation of the innate immune response

2017. 5'-UTR and 3'-UTR Regulation of MICB Expression in Human Cancer Cells by Novel microRNAs

microRNA-889 is downregulated by histone deacetylase inhibitors and confers resistance to natural killer cytotoxicity in hepatocellular carcinoma cells

ncRNA-and Pc2 methylation-dependent gene relocation between nuclear structures mediates gene activation programs

Plumbing the sources of endogenous MHC class I peptide ligands

INFalpha-2b inhibitory effects on CD4(+)CD25(+)FOXP3(+) regulatory T cells in the tumor microenvironment of C57BL/6 J mice with melanoma xenografts

Stress-responsive regulation of long noncoding RNAs' polyadenylation in Oryza sativa

Exogenous plant MIR168a specifically targets mammalian LDLRAP1: evidence of cross-kingdom regulation by microRNA

Antiviral activity of human oligoadenylate synthetases-like (OASL) is mediated by enhancing retinoic acid-inducible gene I (RIG-I) signaling

Lipid-mediated delivery of RNA is more efficient than delivery of DNA in non-dividing cells

Ribose 2'-Omethylation provides a molecular signature for the distinction of self and non-self mRNA dependent on the RNA sensor Mda5

The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s). (Zhou et al., 2014) . Twenty female and ten male rats per group in the first generation (F0) were fed the different diets. Their pups, 20 weanling/female/group and 10 weanling/male/group were randomly selected as an F1 generation. The F2 generation was acquired by using the same procedure described above. At weaning, F3 generation rats were chosen, and 10 rats/sex/group provided the corresponding diets and observed for 13 weeks. No major differences in animal survival, health status, behaviour, body weight, food consumption, or reproductive capacity were observed with the different diets. Some statistically significant differences were observed in animals given the transgenic rice diets compared to those receiving the standard diet for certain clinical chemistry parameters, in certain generations (ALT, ALKP, LDH, cholesterol, HDLC, LDLC). These changes were not considered adverse or biologically significant (Zhou et al., 2014) . Moreover, no evidence of altered incidence or altered severity of background changes was observed in any organ or tissues of the rats fed the three diets (Zhou et al., 2014) . Although minor changes in the mean relative weight of epididimides in male F3 adults was observed in rats fed the rice diets compared to the control group, no difference was observed between the transgenic rice fed rats and their near-isogenic line controls, suggesting overall that the transgenic line was unlikely to cause any risk to rat health in reproduction or development, even when consumed for up to three generations (Zhou et al., 2014) . However, these studies did not evaluate any aspect related to RNAi.Heinemann et al. compared the history of risk assessment of GMOs producing dsRNAs in Australia, New Zealand and Brazil, with a focus on regulatory context (Heinemann et al., 2013) . The authors suggested some processes to properly assess the safety of dsRNA-producing GM plants before their release or marketing. These include i) bioinformatics analysis to identify any likely, unintended targets of the dsRNA in human and animals; ii) experimental procedures that would identify all new intended and unintended dsRNA molecules in the GM product; iii) testing animal and human cells in tissue cultures for a response to intended and unintended effects of dsRNAs from the product; iv) long-term testing on animals; and possibly v) clinical trials on human volunteers (Heinemann et al., 2013) . In response to this article, the regulatory agency Food Standards Australia New Zealand (FSANZ) published a document addressing regulation of GM crops and foods developed using gene silencing (http://www.foodstandards.gov.au/consumer/gmfood/Pages/Response-to-Heinemann-et-al-on-theregulation-of-GM-crops-and-foods-developed-using-gene-silencing.aspx). This document criticised some of the comments of the above risk assessment review. Some key points included the lack of weight of scientific evidence (published up to 2013) to support the view that small dsRNAs in foods are likely to have adverse consequences in humans; the lack of a scientific basis for suggesting that small dsRNAs present in GM foods have different properties than naturally-occurring ones; or the adequate acknowledge of many barriers (in the uptake, distribution and targeting) during oral development of small dsRNA therapies, among others. The overall suggestion was that there was no need to consider additional studies as proposed by Heinemann et al. (2013) .Commentary documents reviewing scientific meetings in the context of RNAi technologies in agricultural biotechnology discuss some of the aspects reviewed in this document .Additional relevant documents of baseline information to support the risk assessment and environmental risk assessment of RNAi-based GM plants can be obtained from EFSA external scientific reports 

The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the author(s).