key: cord-0313587-xw6sm2q7
authors: Adiliaghdam, Fatemeh; Amatullah, Hajera; Digumarthi, Sreehaas; Saunders, Tahnee L.; Rahman, Raza-Ur; Wong, Lai Ping; Sadreyev, Ruslan; Droit, Lindsay; Paquette, Jean; Goyette, Philippe; Rioux, John; Hodin, Richard; Mihindukulasuriya, Kathie A.; Handley, Scot A.; Jeffrey, Kate L.
title: Human enteric viruses shape disease phenotype through divergent innate immunomodulation
date: 2021-10-20
journal: bioRxiv
DOI: 10.1101/2021.10.14.464404
sha: d8f26deafc4d541ee1be2dea8f3a847a7f06921b
doc_id: 313587
cord_uid: xw6sm2q7

Altered enteric microorganisms in concert with host genetics shape inflammatory bowel disease (IBD) phenotypes. However, insight is limited to bacteria and fungi. We found virus like particles (VLPs) enriched from normal human colon resections, containing eukaryotic viruses and bacteriophages (collectively, the virome), actively elicited atypical anti-inflammatory innate immune programs. Conversely, IBD patient VLPs provoked inflammation, which was successfully dampened by healthy VLPs. The IBD colon tissue virome was perturbed, including enriched Picornovirus Enterovirus B, not previously observed in fecal virome studies. Mice with humanized healthy colon tissue viromes had attenuated intestinal inflammation while those with IBD-derived viromes exhibited exacerbated inflammation in a nucleic acid sensing-dependent fashion. Furthermore, there were detrimental consequences for IBD-associated MDA5 loss-of-function on patient intestinal epithelial cells exposed to healthy or IBD viromes. Our results demonstrate that innate recognition of either healthy or IBD human viromes autonomously influences disease phenotypes in IBD. Harnessing the virome may offer therapeutic and biomarker potential. One Sentence Summary Human viromes divergently shape host immunity and disease

Altered enteric microorganisms in concert with host genetics shape inflammatory bowel disease (IBD) phenotypes. However, insight is limited to bacteria and fungi. We found virus like particles (VLPs) enriched from normal human colon resections, containing eukaryotic viruses and 5 bacteriophages (collectively, the virome), actively elicited atypical anti-inflammatory innate immune programs. Conversely, IBD patient VLPs provoked inflammation, which was successfully dampened by healthy VLPs. The IBD colon tissue virome was perturbed, including enriched Picornovirus Enterovirus B, not previously observed in fecal virome studies. Mice with humanized healthy colon tissue viromes had attenuated intestinal inflammation while those with 10 IBD-derived viromes exhibited exacerbated inflammation in a nucleic acid sensing-dependent fashion. Furthermore, there were detrimental consequences for IBD-associated MDA5 loss-offunction on patient intestinal epithelial cells exposed to healthy or IBD viromes. Our results demonstrate that innate recognition of either healthy or IBD human viromes autonomously influences disease phenotypes in IBD. Harnessing the virome may offer therapeutic and biomarker 15 potential.

Main Text: 20 Viruses are the most abundant biological entity on earth and infect all types of life. Generally considered obligate pathogens, viruses are a constant and emerging threat to human health, as exemplified by the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) pandemic. Yet, paradoxically, many viruses persistently reside within asymptomatic individuals. These include a range of eukaryotic viruses, as well as bacteriophages (1-4), that exist both within their 25 bacterial host, as well as free virions embedded within mucus layers of the intestine (5) . These collective resident viruses, along with bacteria, archaea, fungi, helminths and protozoa form the human metagenome that have intimate functional relationships with their host. The human intestinal virome is established at birth and is dominated by bacteriophages, while eukaryotic viruses gradually emerge after birth, as measured in feces that contains ~10 9 virus like particles 30 (VLPs) per gram (2, 6) . Furthermore, despite multiple remaining hurdles in identifying what constitutes a normal human virome (7) , robust fluctuations of the virome have been reported in IBD (1, [8] [9] [10] , colorectal cancer (11, 12) , type I diabetes (13) , nonalcoholic fatty liver disease (14) , cystic fibrosis (15) , graft-versus-host-disease (16) and HIV infection (17) and a range of other diseases (18) . However, this rapid cataloging of viral genomes in human tissues by sequencing has 35 vastly outpaced any mechanistic understanding we have of virome biology. Viruses inhabiting a healthy host (19, 20) , and host innate sensing of viruses by gut-resident Toll-like Receptors (TLRs) or RIG-I like Receptors (RLRs) (19, (21) (22) (23) (24) (25) (26) provide homeostasis in the mouse intestine. However, functional studies of the human healthy or diseased enteric virome are currently lacking. Although, examples of direct and distinct bacteriophage-human immune interactions are emerging (5, 27, 40 28). Furthermore, loss-of-function variants in the gene IFIH1 that encodes the host virus receptor melanoma differentiation associated protein 5 (MDA5) were shown to significantly associate with IBD by genome wide-association studies, GWAS (29) . However, how abrogation of MDA5 function may contribute to intestinal inflammation in IBD, especially in the context of the virome, is unclear. 45

To determine whether the human enteric virome has homeostatic or pathophysiological properties, we purified virus-like particles (VLPs) from fresh colon resections taken from non-inflamed/non-IBD, Ulcerative Colitis (UC) or Crohn's disease (CD) patients, post-surgery (Fig 1A, Table S1 ). We reasoned that since viruses are obligate intracellular pathogens that rely on a host organism for 5 successful replication, and bacteriophages reside within mucosal layers (5) and are readily taken up by eukaryotic cells (27, 30, 31) certain viruses may be better represented in upper intestinal tissue than in feces. Tissue from non-IBD individuals was confirmed to be not inflamed and UC and CD tissue was confirmed to be inflamed by a blinded clinical pathologist (Table S1) . We quantified our VLP preparations by confocal microscopy using nucleic acid-binding SYBR Gold, 10 and found an average of 2.8x10 7 VLPs per mg of colon tissue from non-IBD, UC or CD individuals (Fig 1B) . Our preparations were comprised of prokaryotic and eukaryotic viruses, as visualized by electron microscopy (Fig S1A) , and importantly, human colon-derived VLP preparations were both endotoxin-and cytokine-free (Fig S1B-F) . 15 Our understanding of the host innate immune response to virus stems predominantly from studies of resolving acute infection. This mammalian antiviral paradigm centers on the detection of virus moieties by cytosolic or endosomal receptors and production of type I or III interferon (IFNs) and hundreds of IFN stimulated genes (ISGs) that aim to eliminate virus, or the cell that it infects (32) . The rules governing if and how innate immune cells chronically respond to persistent eukaryotic 20 or prokaryotic viruses in the intestine, as well as in other systemic niches (33), and how this shapes host physiology, are still poorly understood. Macrophages are one of the most abundant leukocytes in the intestinal mucosa where they are essential for maintaining homeostasis but are also implicated in the pathogenesis of IBD (34) . Macrophages are also critical sensors of virus, including endocytosed bacteriophage (27) . Therefore, to identify the precise innate immune 25 response initiated by eukaryotic or prokaryotic viruses residing in healthy or diseased human colon, we exposed primary human peripheral blood derived macrophages to collective VLPs isolated from non-IBD, UC or CD patient colon resections (Fig 1A) . Fluorescently labeled colon resection-derived VLPs demonstrated dose-dependent internalization by primary human macrophages that was equal for non-IBD, UC or CD-derived viromes (Fig 1C, D) . Furthermore, 30 virome uptake was dependent on endocytosis as treatment with brefeldin A a vesicular transport inhibitor, greatly reduced intracellular accumulation (Fig. S2) . Thus, the human virome, that includes both eukaryotic viruses and bacteriophages, is readily internalized by human immune cells via endocytosis for pathogen associated molecular pattern (PAMP) recognition, consistent with other reports (27, 30) . To assess the transcriptional pathways induced by the human colon 35 tissue virome in macrophages, we performed unbiased RNA sequencing. Multi-dimensional scaling (MDS) analysis found non-IBD colon-resident viromes initiated a distinct transcriptional program, that was highly reproducible between 5 different patient VLP sources (Fig 1E) . Conversely, UC or CD human enteric viromes from multiple patients induced macrophage transcriptional programs that were distinct from non-IBD virome-induced programs but almost indistinguishable from one 40 another (Fig 1E) . Interestingly, viruses resident to normal colon tissue predominantly suppressed the host innate immune response as more transcripts were down-regulated in macrophages exposed to non-IBD virus than were up-regulated (Fig S3) . On the other hand, both UC and CD viromes triggered a prominent upregulation in gene expression of macrophages and these transcriptional programs were almost identical (Fig S3) . A striking feature of the non-IBD virome induced 45 macrophage transcriptional signature was a broad downregulation of the classical antiviral response. Non-IBD viromes significantly downregulated an array of virus receptors, signalling adapters and IFN stimuated genes (ISGs) such as TLR7, TLR8, IFIT2, IFIT3, RSAD2 (also known as Viperin) and IFI44L (Fig 1F and Fig S3D) . This atypical host response to non-IBD viromes was also exemplified by the prominent induction of anti-inflammatory and pro-survival genes such as IL4R, CISH, BCL2, SOCS2 (Fig 1F) and multiple genes that define a homeostatic/resolving 5 macrophage state (35, 36) (Fig S3C) . In stark contrast, UC and CD viromes both robustly promoted inflammatory pathways in macrophages including many cytokines (IL6, IL1B, IL8, IFNG) and chemokines (CCL1, CCL20, CCL24, CXCL3, CXCL5, Fig 1G, H) . Gene-set enrichment analysis defined the significant pathways downregulated by non-IBD viromes to be apoptosis, inflammatory and antiviral (Fig 1I) while those pathways that were upregulated were 10 pro-survival and pathways known to polarize macrophages toward homeostatic/resolving states such as WNT-b Catenin signaling (Fig 1I) . In constrast, both UC and CD viromes robustly promoted inflammatory and apoptotic pathways (Fig 1J,K) . We next confirmed the divergent macrophage activation states induced by non-IBD and IBD viromes using ELISA and qPCR with VLPs isolated from additional patient colon resections. Cytokines IL-10, IL-22, TGFb or negative 15 regulators of inflammation CISH or SOCS2, hallmark genes of anti-inflammatory and tissue remodeling type macrophages (35, 36) were all consistently and significantly elevated in macrophages exposed to VLPs isolated from 12 different non-IBD patient colon resections ( Fig  1L, Fig S4) . Notably, IL-22 induction is consistent with recent reports of mice mono-associated with apathogenic eukaryotic enteric viruses (37). Conversely, macrophages that internalized 20 viromes derived from UC or CD intestinal resections from 8-12 separate patients, repeatedly displayed significantly enhanced production of cytokines TNF, IL-6, Type I and II IFNs, IL-1b and IL-12 that define pro-inflammatory type macrophages as well as other cytokines such as IL-15, IL-8 and IL-31 with designated roles in intestinal inflammation and IBD (38, 39) (Fig 1M, Fig  S4) . Importantly, inhibition of endocytosis or heat inactivation of VLPs, failed to induce 25 macrophage cytokine responses (Fig S5) . However, crosslinking of VLP preparations did not significantly reduce innate responses indicating that virus replication is not an absolute requirement for virome stimulatory capacity (Fig S5) . Moreover, we observed similar divergent macrophage responses with virus isolated from non-IBD, UC and CD ileum lumen contents ( Fig  1L,M, Fig S4, S6 , Table S2 ). Taken together, these results demonstrate that human enteric viruses 30 residing in colon tissue or ileum content autonomously invoke host innate immune responses. The macrophage response to enteric viruses is atypical to the classical antiviral response normally characterized by the induction of type I IFNs and ISGs. Enteric viruses from a healthy intestine instead dampened the classical antiviral response and promoted anti-inflammatory and prosurvival programs. Conversely, enteric viruses existing within an inflamed UC or CD intestine 35 independently and robustly promoted a state of inflammation.

Intestinal epithelial cells (IECs) express innate immune receptors that recognize and respond to commensal microorganisms. Furthermore, bacteriophages are readily internalized by IECs (30, 31) . We therefore also tested the ability of healthy or IBD enteric viruses to affect epithelial barrier 40 function and the immune response of IECs. We utilized a trans-well Caco-2 cell monolayer system and exposed IEC monolayers with viruses isolated from non-IBD, UC or CD colon resections and observed significantly more barrier integrity damage induced by both UC and CD-derived enteric viruses in a TLR4-independent manner, as measured by transepithelial electrical resistance (TEER, Fig S7A,B) . Disrupted barrier integrity elicited by IBD viruses was also associated with a greater 45 downregulation of transcripts for tight junction proteins Zonula occludens-1 (ZO-1) and Occludin (Fig S7C,D) . Importantly, IFNl which controls both acute and persistent viral infections in the intestine (40, 41) and confers protection from intestinal inflammation (20) was preferentially induced by non-IBD viruses (Fig S7E) . In contrast, viruses isolated from either UC or CD colons promoted production of pro-inflammatory cytokines such as TNF and IL-15 (Fig S7F,G) that play crucial roles in intestinal tissue destruction and the pathogenesis of IBD (38, 39, 42) . Of note, IL- 5 15 was also recently shown to be induced by the murine enteric virome in dendritic cells to promote expansion of innate epithelial lymphocytes (IELs) (23) . Thus, similar to macrophages, IECs sense and divergently respond to non-IBD and IBD enteric viruses whereby IBD viromes promote both inflammatory responses and disruption of epithelial barrier function. 10 Fecal microbiota transplants (FMT) to restore disrupted microbiota are proving beneficial in both experimental and clinical settings (43) but the virome content of these preparations is currently not considered. We wondered if isolated healthy enteric viruses could overcome the inflammatory response induced by IBD viruses. To this end, we mixed increasing ratios of non-IBD colon resection derived viruses with IBD viruses while maintaining the same overall number of VLPs 15

and exposed human macrophages to them. Increasing amounts of non-IBD viruses were capable of suppressing the ability of IBD-derived viruses to induce pro-inflammatory cytokines such as IL-6, TNF and IL-15 in a dose-dependent manner (Fig 1N, Fig S8A,B) . Moreover, increasing the ratio of non-IBD viruses could partially restore the ability of macrophages to produce antiinflammatory cytokines such as IL-22 (Fig S8C) . To test if the product of the immune response to 20 healthy viruses in the form of secreted soluble factors could also suppress the IBD virome responses, we performed supernatant transfer experiments. In brief, human macrophages were exposed to non-IBD VLPs for 24 hours, then supernatant was collected, clarified, and placed on fresh cells in the presence of IBD colon-derived VLPs. Supernatant from non-IBD VLP treated macrophages was capable of suppressing the production of IL-6, TNF, IL-15 by IBD VLPs (Fig  25  1O, Fig S8D, E) . Furthermore, heat inactivation ameliorated suppressive activity of non-IBD virome-induced macrophage supernatant but UV crosslinking did not (Fig 1O, Supp Fig 8D,E) demonstrating soluble proteins, but not residual viruses in the supernatant, were eliciting antiinflammatory activity. Thus, this data demonstrates the potential utility of viruses in a healthy intestine, or the anti-inflammatory immune response to it, to suppress the inflammatory capacity 30 of viruses in a diseased intestine.

To gain a mechanistic understanding of the divergent immunomodulation by non-IBD or IBD derived human colon tissue or ileostomy fluid-derived viromes, we profiled the virus types in each cohort since only fecal VLPs have been examined to date. DNA and RNA were purified from 35 preparations of VLPs from surgical colon resections or ileostomy fluid and characterized by metagenomic sequencing. After filtering out human and other contaminant (primers and adapters) and low-quality data, sequences were assigned a eukaryotic viral or bacteriophage taxonomic lineage. Specific eukaryotic viral families were selected for subsequent analysis based on the alignment length and percent identity for each family that separated the high-quality alignments 40 from potentially spurious alignments (Table S3) . Total reads aligning to eukaryotic viruses were similar for non-IBD, UC or CD patients, however there was a significant increase in bacteriophage reads in CD patients (Fig 2A) . Of the high-quality reads for subsequent analysis, we found Papillomaviridae, Picornoviridae, Coronoviridae, Orthomyxoviridae and Anelloviridae animal viruses, Virgiviridae plant viruses and a number of unknown eukaryotic virus families in colon 45 tissue (Fig 2B) . Furthermore, we found the majority of VLP classifications in human colon tissue to be bacteriophage families consisting of dsDNA Caudovirales order including Siphoviridae, Myoviridae, Podoviridae and unknown; dsDNA families Herelleviridae, Akkermanniviridae, Lipothrixviridae, as well as ssDNA phage family Microviridae. Quantification of viral species richness showed no significant difference between groups (Fig S9A) but differential analysis showed a significant elevation in the Picornoviridae family member Enterovirus B, in both UC 5 and CD colon tissue compared to non-IBD controls (Fig 2C) . Enterovirus For bacteriophages, CD colon tissue exclusively demonstrated a significant expansion, consistent with reports of CD fecal viromes (1). We found significant elevations in total phage aligned reads (Fig 2A) , a distinct separation of CD from non-IBD and UC by NMDS analysis (Fig S9B) and a multitude of significantly increased genus of Caudovirales members Siphoviridae, Myoviridae and Podoviridae (Fig 2D-F, Fig S9C) . Interestingly, ileostomy fluid VLPs showed significantly 25 reduced viral diversity compared to colon derived VLPs (Fig S10) underscoring tissue rather than luminal contents to be the abundant source of virus in the intestine. Detectable eukaryotic families in ileostomy fluid were limited to Annelloviridae, Papillomaviridae, Picornoviridae and Virgiviridae eukaryotic viruses and the bacteriophage families were Caudovirales family members Siphoviridae, Myoviridae, Podoviridae and unknown; dsDNA families Herelleviridae, 30 Akkermanniviridae, Lipothrixviridae as well as ssDNA phage family Microviridae (Fig S10) .

We next determined the virus receptor requirement for host responses to human enteric viruses. Virus-derived RNA structures are recognized by pattern recognition receptors (PRRs) such as cytosolic (RIG-I)-like receptors (RLRs), or endosomal Toll-like receptor (TLR)3, 7, 8 and viral 35 DNA is mainly recognized by cyclic GMP-AMP (cGAMP) synthase (cGAS) or TLR9 (47). So far, endosomal TLR3, TLR7, TLR9 and cytosolic RIG-I have all been shown to be essential for sensing of the healthy murine virome (19, 23, 28) . We generated bone marrow-derived macrophages from mice deficient in the RLR signaling adapter mitochondrial antiviral signaling protein (MAVS), DNA sensor cGAS, TLR adapter protein MyD88 or endotoxin sensor TLR4, and 40 exposed them to human colon VLPs (Fig 3A,B) . Indeed, MAVS, cGAS or MyD88 deletion resulted in attenuated non-IBD virome responses, as measured by pro-survival gene Cish (Fig 3A) .

Conversely, deletion of endotoxin receptor TLR4 had little effect (Fig 3A) , underscoring recognition of virus nucleic acid in VLP-driven responses. Similarly, pro-inflammatory cytokine IL-6 production by colon tissue derived IBD VLPs was also significantly diminished in MAVS-,  45 cGAS-, and MyD88-deficient macrophages but was unchanged in TLR4-deficient macrophages ( Fig. 3B) . Notably, ileostomy fluid-derived VLPs, that had significantly reduced virus diversity compared to colon tissue (Fig S10) and was mostly bacteriophages and animal viruses of the Annelloviridae family (Fig S10) , required DNA sensor cGAS exclusively for cytokine induction. Certainly, a dominance of viral-derived DNA was observed in the ileum content whilst the colonic tissue virome was a mix of RNA and DNA (Fig S11) . This data suggests there is recognition 5 redundancy for normal/non-IBD or IBD enteric viruses but qualitative and quantitative differences in the downstream immune responses induced by them. RLRs RIG-I (retinoic acid-inducible gene I; encoded by DDX58) and MDA5 (melanoma differentiation-associated gene 5; encoded by IFIH1) are activated by distinct viral RNA structures to differentiate between viral and cellular RNAs. MDA5 senses long double-stranded (ds)RNA, while RIG-I responds to blunt-ended 10 dsRNA bearing a triphosphate (ppp) moiety (48). Transfection of purified RNA isolated from non-IBD VLPs, triggered more IL-10 than equal concentrations of RNA from UC or CD VLPs ( Fig  3C) . In contrast, UC or CD VLP RNA triggered more IL-6 production (Fig 3D) demonstrating divergent macrophage responses were not due to differences in viral tropism or levels. Importantly, digestion of single-stranded or double-stranded VLP RNA with RNase I or III, respectively, or 15 removal of phosphates with Calf Intestinal Phosphatase (CIP), eliminated this response (Fig 3C,  D) demonstrating the existence of "non-self" RNA moieties within the human enteric virome and the requirement for both RIG-I and MDA5 in their recognition. Divergent anti-or proinflammatory responses by non-IBD or IBD enteric viruses were also induced by transfection of VLP-derived DNA, in a cGAS-dependent manner (Fig 3E, F) . Thus, altered RNA or DNA 20 moieties that are sensed by host virus receptors RIG-I, MDA-5 or cGAS (and likely other receptors), rather than an adaptive function of viral proteins on the host, contribute to the divergent activation thresholds and immunomodulation by normal/non-IBD and IBD enteric viromes. This deviation in RNA and DNA viral ligands within the diseased virome is attributed to the fluctuations of certain viruses we observed (Fig 2) . However, it is also appealing to speculate that akin to some 25 bacteria (49), normally symbiotic viruses may also become pathobionts when in an inflamed environment.

Given the existence of genetic mutations within MDA5 that were identified to associate with IBD incidence by GWAS (29) and the discovery of elevated Enterovirus B of the Picornoviridae family 30

in IBD colon tissue, that is exclusively sensed by MDA5 (50), we next examined the consequence of these mutations on responses to the human non-IBD and IBD enteric virome. IFIH1 gain-offunction has been reported as a cause of a type I interferonopathy encompassing a spectrum of autoinflammatory phenotypes including Systemic Lupus Erythematosus and Aicardi-Goutières syndrome (51). Yet, intriguingly, all IBD-associated IFIH1 variants have predicted loss-of- 35 function biological effects, either truncating the protein (rs35744605, E627X), affecting essential splicing positions (rs35732034 in a conserved splice donor site at position +1 in intron 14) or altering a highly conserved amino acid (rs35667974/I923V; Fig 3G) within the C-terminal domain (CTD) which recognizes and binds to RNA (52). We transfected HEK-293T cells with wild-type MDA5 or E627X or I923V MDA5 coding variants and found the wild-type and MDA5 mutants 40

were expressed at comparable levels although the E627X truncation, that lacks a part of the helicase domain and the entire CTD in which I923 resides, showed slightly reduced MDA5 protein levels (Fig S12) . Wild-type MDA5 conferred similar responsiveness to the synthetic RNA ligand poly(I:C) and RNA isolated from healthy human colon tissue VLPs but was not responsive to self RNA (Fig S12) . However, both IBD-associated MDA5 variants exhibited a loss of function 45 phenotype to poly(I:C) (Fig S12) in agreement with a previous report (53). Importantly, a loss of function phenotype to human colon derived viral RNA was observed for both IBD-associated coding MDA5 variants (Fig S12) . We next examined the functional consequence of a loss of MDA5 function on intestinal health. MDA5 is predominantly expressed in epithelial cells ( Fig  S13) so we examined functional consequences of loss-of-function MDA5 by converting induced pluripotent stem cells (iPSCs) derived from a compound heterozygote UC patient carrying a single 5 allele of rs35732034/IVS14+1 and a single allele of rs35744605/E627X into human intestinal epithelial cells (hIECs, Fig 3H, Fig S14) . Compared to non-affected individuals, hIECs from patients carrying loss-of-function MDA5 E627X/IVS14+1 displayed similar morphology and barrier integrity at steady state, as measured by confocal microscopy for F-actin, ZO1 and E-cadherin ( Fig  3I) . However, unlike iPSC-derived hIECs cells bearing wild-type MDA5, MDA5 E627X/IVS14+1 10 iPSC-derived hIECs were severely impaired in their ability to maintain barrier integrity when exposed to enteric viruses derived from non-IBD colon resections (Fig 3J) . Furthermore, damage to MDA5 E627X/IVS14+1 hIECs was exacerbated when exposed to IBD derived viromes containing elevated Enterovirus B (Fig 3J) . Moreover, IBD-risk MDA5 E627X/IVS14+1 iPSC-derived hIECs displayed significantly abrogated virome-induced IFNl production compared to wild-type cells 15

( Fig 3K) . Thus, abolished innate responses to the virome due to human genetic variation within virus receptor MDA5 has detrimental consequences for intestinal epithelial cell function and integrity. Moreover, intestinal damage is intensified in patients bearing IBD-risk MDA5 variants responding to an IBD virome, underscoring the ability of environmental cues to uniquely impact IBD phenotypes in the context of host genetics. 20

We next assessed the ability of non-IBD or IBD human colon-resident viruses to shape local immunity and intestinal disease in vivo. To achieve this, we first diminished the murine enteric virome of C57BL/6 using an antiviral (AV) cocktail by oral gavage for 10 days (19) and then administered human colon resection-derived VLPs 3 times intragastrically (Fig 4A) . To 25 complement our functional data in macrophages, we first assessed the influence of healthy or diseased enteric viromes on gut resident mononuclear phagocytes (MNPs), especially those bearing the fractalkine receptor CX3CR1 (CX3CR1 + MNPs), since they have established roles in the induction of immune responses toward enteric bacteria (54) and more recently intestinal fungi (55). In parallel, we examined the influence of human enteric viromes on subsets of dendritic cells 30

(DCs) differentially expressing integrins CD11b and CD103 that initiate immune tolerance (56, 57). Mice with a healthy humanized virome showed no changes in lamina propria CX3CR1 + MNPs compared to wild-type or antiviral treated mice (Fig 4B,C) but had elevated tolerizing CD11b + CD11c + phagocytes expressing CD103 (Fig 4B,C, Fig S15) . In contrast, mice with an IBD humanized virome displayed a significant accumulation of MNPs expressing CX3CR1 and a 35 reduction in CD11b + CD11c + MNPs expressing CD103 in the colonic lamina propria, both in number and as a fraction of the total CD11b + phagocyte compartment (Fig 4B,C, Fig S15) . Thus, IBD-associated enteric viruses can promote an inflammatory state in the intestine in vivo by spontaneously influencing the ratio of CX3CR1 + and CD103 + lamina propria MNPs. 40

The ability of the human enteric virome to contribute to disease phenotype in a dextran sulfate sodium (DSS)-induced colitis model was then assessed. Since this model is more closely associated with UC, mice diminished of their murine virome were administered human enteric viruses derived from colon resections of non-IBD or UC or CD patients 3 times intragastrically (Fig 4D) . We assessed the abundance of isolated virus-like particles (VLPs) in each treatment 45 group by staining fecal or colon tissue derived VLPs with SYBR. We found a mean of 1.1x10 7

VLPs/mg of feces and 1.1x10 6 VLPs/mg of colon tissue in naïve C57BL/6 mice (Fig 4E,F) . After AV cocktail treatment, fecal VLPs declined ~20-fold to a mean of 6x10 5 VLPs/mg in accordance with previous publications in Balb/c mice (19) and VLPs in colon tissue declined ~12 fold to 9x10 4 VLP/mg. Quantification of viruses in feces or colon tissue of humanized virome mice revealed an equal abundance of ~4x10 6 VLPs/mg of feces and ~4x10 5 VLPs/mg of colon tissue from mice 5 humanized with healthy, UC or CD viromes (Fig 4E, F) . Consistent with recent publications (19, 20) , depletion of viruses in healthy mice lead to significantly exacerbated DSS induced intestinal damage (Fig 4G-I) . Remarkably, mice repopulated with a healthy humanized virome conferred robust protection against colitis, including less shortening of the colon (Fig. 4G) , rescued stool lipocalin levels (Fig. 4H) , rescued colon IL-6 levels (Fig 4I) . Thus, normal/non-IBD human 10 enteric viromes actively immunodulate and incite protection from intestinal inflammation in vivo, consistent with our in vitro results. In contrast, both UC or CD human colonic viromes were incapable of rescuing antiviral treated mice and promoted intestinal disease in vivo (Fig 4G-I) , results that were confirmed by histological analysis (Fig 4J,K) . Moreover, confirming our in vitro data, both UC or CD intestinal viromes resulted in reduced intestinal barrier integrity in vivo, as 15 measured by gavaged FITC-dextran in the serum (Fig 4L) . We had demonstrated an autonomous immunodulatory capacity of intestine-derived viromes in cells, however, to assess any viromemicrobiome community interactions vivo, we performed 16S sequencing of host microbiota. Bacterial diversity and richness were not significantly affected by antiviral treatment or gavage administration of human healthy or IBD viromes (Fig S16A) . However, in UniFrac-based PCoA 20

analysis, groups were clustered separately dependent on virus depletion or human virome source (P<0.05, PERMANOVA, Fig S16B) . By proportion (Fig S16C) and LEfSe analysis, the bacterial groups most changed by AV cocktail treatment was a reduction in the phyla Firmicutes and an elevation certain genera of the Ruminococcaceae family (Fig S16D) . Both healthy and UC humanized virome mice displayed similar changes in murine microbiota (Fig S16E) suggesting 25 microbiome changes are unlikely to be the driver of divergent human virome-induced phenotypes. Furthermore, CD humanized virome mice had unique changes in the murine microbiome ( Fig  S16F) reflecting the distinct bacteriophage alterations we had observed in CD colon tissue. Importantly however, virus nucleic acid sensing by the host immune system was an absolute requirement for immunomodulation by healthy or IBD human viromes in vivo. Mice deficient in 30 MAVS, the adapter protein of RNA sensors RIG-I and MDA5, or mice deficient in the DNA sensor cGAS, displayed no change in colon length, lipocalin or colon IL-6 production following administration of non-IBD, UC or CD human enteric viromes, compared to antiviral-treated controls (Fig 4M-R) . Thus, an ability of human enteric viromes to directly shape intestinal disease phenotype in vivo was via host innate immune recognition of viruses. Moreover, it demonstrates 35 that fluctuations in the virome are not correlative with disease but rather there are direct functional consequences on host immune and intestinal state due to active host sensing of an altered virome in IBD (1, 8, 10) (Fig 2) While disruption of microbiome and virome has been observed in IBD and other diseases, a 40 functional role for the virome has remained elusive. Our data demonstrate that viruses resident to normal human colon tissue are readily endocytosed and elicit an immunomodulatory transcriptional program in macrophages and intestinal epithelial cells, consistently promote production of anti-inflammatory cytokines, and can maintain homeostasis and protect against induced colitis in vivo. Conversely, viromes from inflamed intestines, and specifically UC and CD, 45 triggered pro-inflammatory macrophage and IEC barrier loss phenotypes, and intestinal inflammation in vivo upon transfer. These functional differences were dependent on innate viral sensors, indicating active and unique recognition of the intestinal virome by the host. Altogether, our findings provide evidence that viruses in a normal/non-inflamed intestine can help build gut immunity in humans -a contrast to the misconception that all viruses are harmful. An alteration in the intestine-resident virome, or disrupted virome sensing due to genetic variation in the virus 5 receptor MDA5, has detrimental consequences for the human intestine. It is appealing to speculate that viruses play a role in the protection mediated by fecal transplants, as well as in the pathology of intestinal inflammation. Therapeutic modulation of the virome through targeted elimination or replacement of disease-or health-driving intestinal viruses presents an novel strategy to treat intestinal and immunological diseases more broadly. 

The authors declare no competing interests. 5

Additional information

Correspondence and requests for materials should be addressed to Kate L. Jeffrey KJeffrey@mgh.harvard.edu 10 Data and materials availability: The data presented in this study are tabulated in the main text and supplementary materials. RNA sequencing raw and processed data has been deposited at Gene Expression Omnibus (GEO) and is accessible using accession number GSE135223. Human colon resection and ileostomy fluid virus like particle (VLP) sequencing are deposited at European 15 Nucleotide 

Colitis and Crohn's disease patients. VLPs were assessed using VLP DNA and RNA 5 metagenomic sequence data. Sequences were assigned a eukaryotic viral or phage taxonomic lineage using primary searches against virus sequence databases and subsequently confirmed in secondary searches using reference databases with additional non-viral taxonomic lineages. Specific eukaryotic viral families were selected for subsequent analysis by gating on the alignment length and percent identity for each family that separated the high-quality alignments 10 from potentially spurious alignments (Table S3 ). (A) Violin plots represent the distribution of individual datasets for total reads aligning to eukaryotic viruses or bacteriophages from non-IBD, UC or CD cohorts; samples were compared using two-sided Wilcoxon signed-rank tests. ( Table S1 ). Study research coordinators obtained consent, and medical history was obtained and confirmed by review of the electronic medical record. Inflammation in UC or CD colon resection samples was confirmed grossly and microscopically by a clinical pathologist. Non-IBD controls included individuals undergoing colon resection surgeries for colorectal cancer, familial adenomatous polyposis (FAP) or Diverticulitis. Lack of inflammation in non-IBD control resection samples was confirmed by a clinical pathologist. Ileostomy fluid was collected from patients enrolled in the "Quantifying the Pro-Inflammatory Nature of Human Fluid Collection Study" that was approved by the Committee for the Protection of Human Subjects at MGH (IRB#2011P002755 , Table S2 ).

Virus Like Particle (VLP) Isolation 100 mg of colon tissue or centrifuged ileostomy fluid was resuspended in 500 µL of saline magnesium buffer (SM) buffer (8 mM MgSO4, 100 mM NaCl, 50 mM Tris-HCl pH 7.4, and 0.002% (w/v) gelatin, passed through a 0.02µm Whatman filter), homogenized using sterile metal beads, and centrifuged. The supernatant was then passed through a 0.45 μm pore-sized membrane, followed by 0.22 μm pore-sized membrane (Millipore). The filtrates were treated with lysozyme (10 µg/mL; Sigma) for 30 min at 37 °C, followed by incubation with 0.2 volumes of chloroform for 10 min to degrade any remaining bacterial and host cell membranes and then centrifuged at 2,500 × g for 5 min at room temperature. The aqueous phase was collected and incubated with DNase I (3U/200µL, Sigma) for 1 h at 37 °C to remove any non-virus protected DNA. Enzyme activity was inactivated by incubation at 65°C for 15 min as described (1). To remove any potential residual endotoxin, samples were treated with Polymyxin B (10 µg/mL, Sigma) for 30 minutes. Undetectable endotoxin was confirmed using Limulus Amebocyte Lysate (LAL) assay kit (GenScript) following the manufacturer's instructions.

Virus-like particles (VLPs) were diluted 10-fold serially, stained for 30 minutes with 10× SYBR Gold (Thermo Fisher Scientific) for nucleic acid or Dialkylcarbocyanine (DiI, Thermo Fisher Scientific) for lipid bilayers of enveloped viruses, and imaged using a Zeiss LSM510 laser scanning confocal microscope. Images were captured using Zeiss software (ZEN). Particles <0.5 μm in diameter were regarded as VLPs. SM buffer was used as a negative control. 10 VLP images were captured per sample and VLPs were counted using an image analyzer (InnerView™).

VLP sequencing Total RNA and DNA was extracted from VLPs on a MagNA Pure 24 instrument (Roche), according to the manufacturer's instructions. In order to evaluate samples for both RNA and DNA viruses, the total nucleic acids were randomly amplified as described previously (2, 3) using barcoded primers consisting of a base-balanced 16-nucleotide-specific sequence upstream of a random 15-mer and used for NEBNext library construction (New England BioLabs). The libraries were multiplexed on an Illumina NextSeq (Washington University Center for Genome Sciences) using the paired-end 2 × 150 protocol.

Unprocessed paired-end reads were processed through a multistage quality-control procedure to remove primers and adapters, human and other contaminant and low-quality sequence data. Exact duplicate sequences were removed reserving a single copy and all remaining sequence dereplicated allowing for 4 substitutions. These high-quality / low-redundancy sequences were systematically queried against protein or genomic reference databases using MMseqs2 (4) translated and untranslated search strategies. Sequences assigned a eukaryotic viral or phage taxonomic lineage were first identified using primary searches against virus sequence databases and subsequently confirmed in secondary searches using reference databases with additional non-viral taxonomic lineages. Specific eukaryotic viral families were selected for subsequent analysis by gating on the alignment length and percent identity for each family that separated the high-quality alignments from potentially spurious alignments (Table S3) . Only high-quality reads were used for subsequent analysis. Specific phage reads were selected for subsequent analysis by gating on an alignment length of 75 nt, to account for the current state of phage taxonomy. In order to determine which phage and eukaryotic viral reads are differentially abundant in the three cohorts, we analyzed the reads at the family, genus and species levels. The non-parametric Kruskal-Wallis test was used to assess the difference in median abundance of eukaryotic viral and phage taxa in the Non-IBD, UC and CD cohorts. The Dunn's post-hoc test was used to correct for multiple comparisons and only those taxa with an adjusted p-value < 0.05 were considered significant. The mean abundance of viral or phage reads, scaled by library size, was determined for each sample in the colon and ileostomy samples. The non-parametric Welch's paired t-test was used to determine which cohorts had significantly different numbers of viral or phage reads.

Human peripheral blood derived macrophages Peripheral Blood Mononuclear Cells (PBMCs) were isolated from 20-30 mL blood buffy coats from healthy human volunteers (MGH blood components lab). Briefly, mononuclear cells were isolated by density gradient centrifugation of PBS-diluted buffy coat/blood (1:2) over Ficoll-Paque Plus (GE Healthcare). The PBMC layer was carefully removed and washed 3 times with PBS, mononuclear cells were collected as PBMC. In order to obtain macrophages, PBMCs were resuspended in X-VIVO medium (Lonza) containing 1% penicillin/streptomycin (Gibco) and incubated at 37 o C, 5% CO2 for 1h to adhere to the tissue culture dish. After 1h, adherent cells were washed 3 times with PBS and differentiated in complete X-VIVO medium containing 100 ng/mL human M-CSF (Peprotech) for 7 days at 37 o C, 5% CO2. On day 4, cultures were supplemented with one volume of complete X-VIVO medium containing 100 ng/mL human M-CSF.

Delivery of VLPs to macrophages VLPs or SM buffer (8 mM MgSO4, 100 mM NaCl, 50 mM Tris-HCl, pH 7.4, and 0.002%, w/v, gelatin) were pre-treated for 30 minutes with Polymyxin B (Sigma) to ensure endotoxin-free conditions. Media was removed from 1×10 6 adherent macrophages and VLPs (10 VLPs per cell) or SM buffer ("mock") was added drop-wise onto the cells at a fixed volume sufficient to cover the well surface. VLPs were allowed to adsorb to adherent cells for 1 hour at 37 o C with occasional rocking. After 1 hour, VLPs were aspirated, cells were gently washed with PBS, and media was replaced (this was considered time = 0). For heat inactivation experiments, VLPs were first inactivated for 1 hour at 60 o C. For UV crosslinking of VLPs, VLPs were placed in a UV Stratalinker 2400 from Stratagene on ice to UVB-irradiate three times with 200 mJ/cm2 for 15 minutes. For in vitro ratio experiments, non-IBD colon resection-derived VLPs were mixed with CD colon resection-derived VLPs at increasing ratios to the final concentration of 10 VLPs per cell and VLPs were delivered to macrophages as described above. For supernatant in vitro experiments, the supernatant from macrophages that were exposed to the non-IBD VLP 24 hours prior was mixed with IBD VLPs and added drop-wise onto the cells. VLPs and supernatant mix were allowed to adsorb to adherent cells for 1 hour at 37 o C with occasional rocking. Heat inactivated (95 o C for 10 min) or UV crosslinked (three times 200 mJ/cm2 for 15 minutes) supernatant were used as controls.

Fluorescent labeling of VLPs and uptake assays Purified, quantified colon resection derived VLP preparations were labeled with 0.01 µM SYTO TM 16 Green Fluorescent Nucleic Acid Stain (Thermo Fisher Scientific, #S7578) according to manufacturer's instructions. Labeled VLPs were then separated from unincorporated dye using PD10 Sephadex G-25 desalting columns (GE Healthcare, #17085101) according to manufacturer's instructions. Uptake assays of SYTO16-labeled VLPs were performed with human peripheral blood monocyte derived macrophages. Cells were seeded at a density of 10 5 cells/well in a 24-well plate. Cells were delivered 10 6 or 10 7 non-IBD, UC or CD VLPs or SM buffer control by adsorption in 100 µL for 1h, washed and left for 3 hours at 37°C for uptake analysis or 4°C for adsorption analysis. Cells were removed from the culture plate surface with cold PBS, washed in flow cytometry buffer twice, stained for live-dead discrimination and acquired on an LSR II flow cytometer (BD Biosciences). For the inhibition assays, cells were treated with 3.0 mg/mL of Brefeldin A (Invitrogen, #00-4506) for 30 min at 37°C before addition of labeled VLPs and during the adsorption. Data was analyzed using FlowJo v10 software (TreeStar).

Poly(A)-mRNA was enriched using the NEBNext Poly(A) mRNA magnetic isolation module (NEB), according to manufacturer's instructions. Libraries were then generated using the NEBNext Ultra II RNA Library Prep Kit for Illumina (NEB). To allow for multiplexing of samples, libraries were dual indexed using indices from NEBnext Multiplex Oligos for Illumina (NEB). Paired-end sequencing was performed using a NovaSeq 6000. Samples were run in replicate across three separate lanes. Transcriptome mapping was performed with STAR version 2.5.4a(5) using the NCBI37 assembly gene annotations. Read counts for individual genes were produced using the unstranded count feature in HTSeq 0.9.1(6). Differential expression analysis was performed using the EdgeR package(7) after normalizing read counts and including only those genes with count per million reads (cpm) greater than 1 for one or more samples (8) .

Differentially expressed genes (DEG) were defined based on the criteria of minimum 2-fold change in expression value and false discovery rate (FDR) less than 0.05. Gene Set Enrichment Analysis (GSEA) (9) was adopted for functional annotation analysis. Hospital) . For creating the monolayer, wildtype or TLR4-null Caco2 cells were seeded onto collagen-coated polyethylene terephthalate (PET) filter supports (1 µm pore size, Corning) at a concentration of 3×10 5 cells/well and were incubated (37°C, 5% CO2) for 2-3 weeks. Both the apical and the basolateral side of the trans-wells were fed with DMEM supplemented with 10% FBS and 1% P/S solution. After 2-3 weeks the integrity of the monolayer was measured by transepithelial electrical resistance measurements (TEER) using the Millicell-ERS electrical resistance measuring system (Millipore). The electrodes were immersed in a way that the shorter electrode was in the inner well and the longer electrode was in the outer well. 200-220 Ω.cm 2 resistance was indicated as a confluent monolayer. When TEER peak was reached, and maintained for 3 days, the monolayer was used for subsequent experiments. SM buffer and LPS (100 ng/mL) were used as negative and positive controls, respectively. Apical compartment surface was incubated with VLPs whereby VLPs were allowed to adsorb to adherent cells for 1 hour at 37 o C (10 VLPs per cell). After 1 hour, VLPs were aspirated, cells were gently washed with PBS, and fresh media was replaced (this was considered time = 0). The cells were then returned to the cell culture incubator (37°C, 5% CO2) and TEER measured every 6 hours up to 48 hours. All experiments were performed in triplicate.

HEK293T IFNβ luciferase reporter assay HEK293T cells were maintained in 48-well plates in Dulbecco's modified Eagle medium (Cellgro) supplemented with 10% FBS and 1% penicillin/streptomycin. Cells were transfected with 10 ng of indicated FLAG-MDA5 plasmids (kindly provided by the T. Fujita, Kyoto University), IFNBpromoter driven firefly luciferase reporter plasmid (100 ng) and a constitutively expressed Renilla luciferase reporter plasmid (pRL-CMV, 10 ng) by using Lipofectamine2000 (Life Technologies), according to the manufacturer's protocol. Empty vector was used to maintain a constant total amount of DNA and lipofectamine. Media was changed 6 hours after the first transfection and cells were additionally transfected with 1μg/mL of RNA isolated from either non-IBD virus like particles (VLPs), polyinosinic:polycytidylic acid (poly I:C, Sigma), or RNA isolated from the cytoplasm of HEK-293T cells (self RNA). Cells were lysed 24 h post-stimulation and IFNβ promoter activity was measured using the Dual Luciferase Reporter assay (Promega). Firefly luciferase activity was normalized against Renilla luciferase activity.

Differentiation of human iPSC into intestinal epithelial cells Two lymphoblastoid cell lines (LCL) were selected from the NIDDK IBDGC Repository based on their IFIH1 genotypes: (1) an LCL line carrying wild-type IFIH1 alleles from a 39 year-old Female Caucasian and (2) an LCL line that is a compound heterozygote for rs35732034 (IVS14+1) and rs35744605 (E627X) from a Caucasian 47 year-old male Ulcerative Colitis patient. Specifically, the risk allele at rs35732034 (IVS14+1) is a splice donor site substitution at position +1 in intron 14 that leads to exon 14 skipping and the truncation of the C-terminal domain, and the risk allele at rs35744605 (E627X) is a nonsense variant leading to the loss of the C-terminal half of the protein (10, 11) . The two LCL lines were then reprogrammed to human induced pluripotent stem cells (iPSCs) as described (12) via nucleofection with four episomal reprogramming plasmids (pCE-hUL, pCE-hSK, pCE-hOCT3/4, and pCE-mp53DD) and selection of TRA-1-81 positive reprogrammed hiPSC colonies. LCL-derived hiPSC were then used to generate polarized epithelial monolayers using a stepwise differentiation protocol involving a conversion to definitive endoderm (DE) using the STEMdiff Definitive Endoderm kit (StemCell Technologies). This was followed by the formation of hindgut (HG) structures with a treatment of 3 µM CHIR99021 and 250 ng/ml FGF4 for 3-4 days, as previously described (13) . HG structures were harvested and then centrifuged at 100g for 5 min, embedded in Matrigel (Corning) and cultured for 15 days in Intestinal Growth Medium with feeding every 2-3 days as described (14) but using human recombinant inducers rhWnt3a (Cedarlane), rhEGF (Cedarlane), rhNoggin (Cedarlane), rhR-Spondin-1 (Millipore-Sigma) and rhHGF (Cedarlane). Following this, the spheroids were cultivated in human organoid culture medium (HOCM) and passaged every 6-8 days for maturation into 3D Intestinal epithelial cell (IEC) cultures of spheroids (SPH) and organoids (ORG) as described (13, 14) . Intestinal epithelial cell structure and function was then assessed (see Supplementary Fig 14) . Voltohmmeter (EVOM2, World Precision Instruments, LLC) until a plateau was reached (similar TEER value on three consecutive measurements). Cells on the apical compartment surface were infected with VLPs (MOI 10) as described above. VLPs were allowed to adsorb to adherent cells for 1 hour at 37 o C. After 1 hour, VLPs were aspirated, cells were washed with PBS, and fresh media was replaced. SM buffer and poly I:C (50 µg/mL) were used as negative and positive controls, respectively. The cells were then returned to the cell culture incubator (37°C, 5% CO2) and the integrity of the monolayer was measured by transepithelial electrical resistance measurements (TEER) every 6 hours up to 24 hours.

Mouse intestines were harvested and pre-digested in 1µM ethylenediaminetetraacetic acid (EDTA) and 3µM dithiothreitol (DTT) at 37°C for 15 minutes with shaking (400 rpm) to isolate epithelial cells. Remaining tissue was minced and digested with collagenase type II (Gibco) for 45 minutes at 37 o C with constant agitation followed by Percoll density-gradient centrifugation to isolate single cells from the lamina propria. Cells were incubated with biotinylated anti-CD45 to purify the hematopoietic cell population. Beads were gently washed, and bound cells isolated by MACS mini-column purification using streptavidin magnetic beads. Epithelial cells and lamina propria hematopoietic cells were lysed in 200 µL of RIPA buffer (Tris-HCl, pH 7.4, 50 mM, NaCl 150mM, NP-40 1%, sodium deoxycholate 0.5%, SDS 0.1%) with a cocktail of protease and phosphatase inhibitors (1%, Roche). 20 µg of protein was loaded per lane and separated on a 4-12% SDS/PAGE gel and transferred to PVDF. Ponceau stain (Sigma) confirmed protein transfer, and the membrane was blocked for 1 h at 25°C with 5% skim milk in TBS/0.1% Tween-20 (TBS-T), and then incubated overnight at 4°C with the indicated primary antibodies (diluted 1:1000 v/v in TBS-T with 3% BSA). Primary antibody was washed off and the membrane was incubated with the indicated secondary antibodies (diluted 1:10000 in TBS-T). The following primary antibodies were used: anti-MDA5 antibody (Sigma), b-actin (Sigma) was used as a loading control, and Villin (Santa Cruz) was used to confirm purity of epithelial cells.

The ileum was fixed with 4% paraformaldehyde. After antigen retrieval (Sigma, Citrate Buffer, #C9999), sections were treated with autofluorescence blocking MaxBlock, blocked in 5% normal goat serum with 1% BSA in PBS (Blocking buffer) for 1 hour at RT, and stained with the primary antibody anti-MDA5 (Sigma; #SAB3500356) overnight at 4°C at a final concentration of 1:500. Slices were washed and incubated with goat anti-rabbit secondary antibody for 1 hour at RT at a final concentration of 1:500. The slices were washed, counterstained with DAPI (Vector), and mounted for confocal microscopy analysis. All images were collected using a confocal microscope (Nikon A1) (400×).

Five-to six-week-old female C57BL/6 mice, MAVS-/-, cGAS-/-or TLR4-/-(C3H/HeJ) were purchased from The Jackson Laboratory. All mice were housed in specific pathogen-free conditions according to the National Institutes of Health (NIH), and all animal experiments were conducted under protocols approved by the MGH Institutional Animal Care and Use Committee (IACUC), and in compliance with appropriate ethical regulations. For all experiments, agematched mice were randomized and allocated to experimental group, with 5-8 mice per group, and repeated three independent times. No statistical method was used to determine sample size. For "humanized" virome mice, mice were first depleted of viruses using an antiviral cocktail (Acyclovir 20mg/kg, Lamivudine 10mg/kg, Ribavirin 30mg/kg, Oseltamivir 10mg/kg) by daily gavage for 10 days as described (15) . AV-treated mice were then gavaged with 200 μL of SM buffer as a control or 4×10 8 VLPs isolated from non-IBD, UC or CD patient fresh colon resections. Mice received the human VLP gavage every other day for three times in total. To ensure that the gastric pH would not affect the viability of the gavaged VLPs, each mouse was gavaged with 100 μL of 1M NaHCO3 15 min prior to VLP gavage to keep gastric pH at 7.4. Quantification of viruses in different treatment groups was performed by confocal imaging of VLPs isolated from feces and colon tissue.

Bone marrow-derived macrophages Bone marrow-derived macrophages (BMDMs) were produced from MAVS-deficient, cGASdeficient or TLR4-deficient mice (obtained from The Jackson Laboratory) and their respective littermate controls. Briefly, bone marrow was flushed from tibia and femur and allowed to adhere to a non-treated tissue culture plate for 1 day. Non-adherent cells were then differentiated to macrophages in DMEM containing 10% FBS, 1% L-Glutamine, 1% penicillin/streptomycin, 0.1% b-mercaptoethanol, 5 ng/mL of interleukin 3 (IL-3, Peprotech) and Macrophage Colony Stimulating Factor (M-CSF, Peprotech) for 7 days. Macrophage maturity was assessed by surface expression of CD11b and F4/80 with flow cytometry. MyD88-/-immortalized macrophages were obtained from Kate Fitzgerald (UMass Medical School).

Colons were harvested from mice and caeca were removed. The tissue was placed in RPMI containing 5% fetal bovine serum (FBS) and fat was removed by careful dissection followed by gentle rolling along a moist paper towel. The tissue was washed gently then inverted and rinsed to remove loose fat and fecal matter. Pieces of the colon were transferred to a solution of RPMI containing 2% FBS, 1µM ethylenediaminetetraacetic acid (EDTA), 3µM dithiothreitol (DTT) and incubated at 37 o C for 15 minutes with shaking (400 rpm). Next, tissue pieces were rinsed with fresh FBS-containing media to wash off EDTA/DTT and tissue was again rolled gently along a moist paper towel. Tissue was cut into very small pieces and transferred to a digestion medium: each sample was placed in 25mL of RPMI media containing 2% FBS and mixed with 12.5mg Dispase II (Gibco) and 37.5mg type IV collagenase (Worthington). Digestion was allowed to take place for 45 minutes at 37 o C with constant agitation, and halfway through this period, tissue was disrupted further by pipetting. After 45 minutes, the digestion solution was pipetted again and smashed through a 100µm mesh. Digestion was quenched by addition of 25 mL of RPMI containing 5% FBS. The suspension was spun down for 10 minutes at 450g and the resulting pellet was resuspended in fresh media, passed through a 40 µm mesh, washed with additional fresh media and spun down again. The final pellet represented the lamina propria fraction of the colon. Cells were blocked, stained, and washed for acquisition on an LSRII flow cytometer (BD). Antibodies used for flow cytometry can be found in Supplementary Table 5 . All data was analyzed using FlowJo v10 software (TreeStar).

Dextran Sodium Sulfate induced colitis (DSS) Mice were administered 2.5% Dextran Sodium Sulfate Salt (DSS, MW = 36,000-50,000 Da; MP Biomedicals) in drinking water ad libitum for 7 days (freshly prepared every other day), followed by regular drinking water for 5 days. Mice were sacrificed at day 12. Colon length was measured and fecal samples were homogenized in PBS, spun down and supernatants were collected and assayed for Lipocalin-2 by ELISA according to manufacturer's instructions (R&D). To assess intestinal permeability, mice were given FITC-dextran tracer (Sigma, 4 kDa, 0.6 mg/g body weight) intragastrically in 0.2 mL PBS and FITC levels were measured 4h later in hemolysis-free serum using a Fluorescence Spectrophotometer. For cytokine measurements, day 12 colon tissue (approximately 1 cm) was washed in PBS and then placed in a 24 well plate containing 750 μL of RPMI medium with 1% penicillin/streptomycin and incubated at 37°C with 5% CO2 for 24 hours. Supernatants were collected and centrifuged for 10 min at 4°C and IL-6 levels were assessed by ELISA according to manufacturer's instructions (R&D). For histology, tissue was collected from the distal part of the colon and hematoxylin and eosin staining was performed. Histologic evaluation performed in a blinded fashion and scored based on the criteria outlined previously (16) . Three main categories were used to reflect the severity of histopathology: (i) inflammatory cell infiltrates, (ii) epithelial changes and (iii) mucosal architecture.

16S sequencing and analysis QIAamp DNA Stool Mini Kit was used to extract DNA from feces. The hypervariable region (V4) of the 16S rRNA gene sequences were amplified by using PCR with adaptor primer set containing 12bp multiplex identifier sequences (Integrated DNA Technologies, Coralville, IA). Each reaction contained template DNA, each primer (0.2 μM), and 2× Platinum™ Hot Start PCR Master Mix (Invitrogen, Carlsbad, CA). PCR conditions were as follows: 94°C for 3 min, followed by 22 cycles at 94°C for 45 sec, 52°C for 1 min, and 72°C for 1.5 min and a final extension step at 72°C for 10 min. Triplicate reactions were prepared, pooled, and purified using SpeedBead™ carboxylate magnetic bead solution (GE Healthcare, Marlborough, MA). The amplicons were quantified using KAPA Library Quantitation kit (KAPA, Cape Town, South Africa) and an equimolar amount of each sample was sequenced on the MiSeq System (Illumina, San Diego, CA) with 2x150 paired-end run parameters. Microbiome bioinformatics analysis was performed with QIIME 2 release 2018.11 (17) . In brief, cutadapt was used to trim 515F/806R primers in the paired end sequences. Subsequently, the sequences were stripped of noise by using the DADA2 (18) denoiser program. The denoised, trimmed and higher quality amplicon sequence variants (ASVs) were aligned with MAFFT (19) and used to construct a phylogeny with fasttree2 (20) . Alphadiversity metrics (Shannon's diversity index), beta diversity metrics (UniFrac (21)) and Principle Coordinate Analysis (PCoA) were estimated using q2-diversity after samples were rarefied (subsampled without replacement) to 69009 sequences per sample. All ASVs were assigned taxonomy using the q2-feature-classifier (22) . In brief, a naïve Bayes classifier was trained on the Greengenes 13_8 99% OTUs reference sequences (23) to assign taxonomy to each ASV. The compositional differences among the groups were determined by a linear discriminant analysis using LEfSe (24) with a threshold of 2.0 on the logarithmic score using Galaxy (https://huttenhower.sph.harvard.edu/galaxy/).

Results are shown as mean ± s.e.m. Visual examination of the data distribution as well as normality testing demonstrated that all variables appeared to be normally distributed. Comparisons and statistical tests were performed as indicated in each figure legend. Briefly, for comparisons of multiple groups over time or with two variables, a two-way analysis of variance (ANOVA) was used and corrected for multiple post hoc comparisons, comparing all groups to each other, all groups to a control, or selected groups to each other. For comparisons of multiple groups with only one variable, a one-way ANOVA was performed and corrected for multiple post hoc comparisons. For comparisons of two groups, two-tailed paired or unpaired t tests were used, except where indicated. Statistical analyses were performed in the GraphPad Prism 8 software. The nonparametric Kruskal-Wallis test with a Dunn's post-hoc test was used to assess the difference in median abundance of eukaryotic viral and phage taxa in the Non-IBD, UC and CD cohorts. P values denoted throughout the manuscript highlight biologically relevant comparisons. A P value of less than 0.05 was considered significant, denoted as *P ≤ 0.05, **P ≤0.01, ***P ≤ 0.001, and ****P < 0.0001 for all analyses. FDR correction was performed for all RNA-seq data. VLP were assessed using VLP DNA and RNA metagenomic sequence data. Sequences were assigned a eukaryotic viral or phage taxonomic lineage using primary searches against virus sequence databases and subsequently confirmed in secondary searches using reference databases with additional non-viral taxonomic lineages. Specific eukaryotic viral families were selected for subsequent analysis by gating on the alignment length and percent identity for each family that separated the high-quality alignments from potentially spurious alignments (Table S3) . (A) VLP species richness was calculated using the spec number function of the Vegan package. Significance was assessed via Kruskal-Wallis test with a Dunn's post-hoc test. (B) Non-metric multidimensional scaling (NMDS) using Bray-Curtis dissimilarity was conducted using the metaMDS function in Vegan and PERMANOVA analysis (permutational multivariate analysis of variance) was performed using the pairwise Adonis package to assess significance. There were no significant differences in beta diversity between the groups. (C) Significantly (P-adjusted < 0.05, Kruskal-Wallis test with a Dunn's post-hoc correction) differentially abundant eukaryotic and prokaryotic viruses in UC or CD colon resections. Data are the mean of 5-7 biological replicates. Distribution of operational taxonomic units (key) in feces; each operational taxonomic unit (selected at a sequence identity of 99%; QIIME software) is presented as relative abundance in each sample. (D) The compositional differences between control and antiviral treated mice, (E) antiviral treated compared to non-IBD virome humanized mice, (F) antiviral treated compared to UC virome humanized mice and (G) antiviral treated compared to CD virome humanized mice were determined by a linear discriminant analysis using LEfSe with a threshold of 2.0 on the logarithmic score using Galaxy (https://huttenhower.sph.harvard.edu/galaxy/). 

Disease-specific alterations in the enteric virome in inflammatory 15 bowel disease

The stepwise assembly of the neonatal virome is modulated by breastfeeding

Global phylogeography and ancient evolution of the widespread human gut virus crAssphage

Massive expansion of human gut bacteriophage diversity

Bacteriophage adhering to mucus provide a non-host-derived immunity

Early life dynamics of the human gut virome and bacterial microbiome in infants

Illuminating the human virome in health and disease

Whole-Virome Analysis Sheds Light on Viral Dark Matter in 30 Inflammatory Bowel Disease

Murine colitis reveals a disease-associated bacteriophage community

The dynamics of the stool virome in very early onset inflammatory bowel disease

Diagnostic Potential and Interactive Dynamics of the Colorectal Cancer Virome

Alterations in Enteric Virome Are Associated With Colorectal Cancer and Survival Outcomes

Intestinal virome changes precede autoimmunity in type I diabetessusceptible children

Intestinal Virome Signature Associated With Severity of Nonalcoholic Fatty Liver Disease

Metagenomic analysis of respiratory tract DNA viral communities in cystic fibrosis and non-cystic fibrosis individuals

The eukaryotic gut virome in hematopoietic stem cell transplantation: new clues in enteric graft-versus-host disease

Altered Virome and Bacterial Microbiome in Human 5

Effects of Intestinal Fungi and Viruses on Immune Responses and Inflammatory Bowel Diseases

Enteric Viruses Ameliorate Gut Inflammation via Toll-like Receptor 3 10 and Toll-like Receptor 7-Mediated Interferon-beta Production

IFN-lambda suppresses intestinal inflammation by non-translational regulation of neutrophil function

A Map of Toll-like Receptor Expression in the Intestinal Epithelium 15 Reveals Distinct Spatial, Cell Type-Specific, and Temporal Patterns

Rig-I-/-mice develop colitis associated with downregulation of G alpha i2

Commensal viruses maintain intestinal intraepithelial lymphocytes via 20 noncanonical RIG-I signaling

An enteric virus can replace the beneficial function of commensal bacteria

IFN-I and IL-22 mediate protective effects of intestinal viral infection

Mitochondrial antiviral signaling protein (MAVS) monitors commensal bacteria and induces an immune response that prevents experimental colitis

Bacteriophage trigger antiviral immunity and prevent clearance of bacterial infection

Expansion of Bacteriophages Is Linked to Aggravated Intestinal Inflammation and Colitis

Fine-mapping inflammatory bowel disease loci to single-variant resolution

Bacteriophage Transcytosis Provides a Mechanism To Cross Epithelial 35 Cell Layers

Bacteriophage uptake by mammalian cell layers represents a potential sink that may impact phage therapy

Type I and Type III Interferons -Induction, Signaling, Evasion, and Application to Combat COVID-19

The virome in mammalian physiology and disease

Intestinal macrophages: well educated exceptions from the rule

Macrophage activation and polarization: nomenclature and 45 experimental guidelines

Disease-specific alterations in the enteric virome in inflammatory bowel disease

Identification of a novel astrovirus (astrovirus VA1) associated with an outbreak of acute gastroenteritis

Viral discovery and sequence recovery using DNA microarrays

MMseqs2 enables sensitive protein sequence searching for the analysis of massive data sets

STAR: ultrafast universal RNA-seq aligner

HTSeq--a Python framework to work with highthroughput sequencing data

Bioconductor package for differential expression analysis of digital gene expression data

Count-based differential expression analysis of RNA sequencing data using R and Bioconductor

Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles

Fine-mapping inflammatory bowel disease loci to single-variant resolution

Medical relevance of protein-truncating variants across 337,205 individuals in the UK Biobank study

Utility of Lymphoblastoid Cell Lines for Induced Pluripotent Stem Cell Generation

Intestinal Commitment and Maturation of Human Pluripotent Stem Cells Is Independent of Exogenous FGF4 and R-spondin1

Reciprocal Inflammatory Signaling Between Intestinal Epithelial Cells and Adipocytes in the Absence of Immune Cells

Enteric Viruses Ameliorate Gut Inflammation via Toll-like Receptor 3 and Toll-like Receptor 7-Mediated Interferon-beta Production

A guide to histomorphological evaluation of intestinal inflammation in mouse models

Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2

High-resolution sample inference from Illumina amplicon data

MAFFT: a novel method for rapid multiple sequence alignment based on fast Fourier transform

FastTree 2--approximately maximum-likelihood trees for large alignments

UniFrac: a new phylogenetic method for comparing microbial communities

Optimizing taxonomic classification of marker-gene amplicon sequences with QIIME 2's q2-feature-classifier plugin

An improved Greengenes taxonomy with explicit ranks for ecological and evolutionary analyses of bacteria and archaea

Metagenomic biomarker discovery and explanation