key: cord-328914-6zog3xf3
authors: Petroselli, Chiara; Montalbani, Elena; La Porta, Gianandrea; Crocchianti, Stefano; Moroni, Beatrice; Casagrande, Chiara; Ceci, Elisa; Selvaggi, Roberta; Sebastiani, Bartolomeo; Gandolfi, Isabella; Franzetti, Andrea; Federici, Ermanno; Cappelletti, David
title: Characterization of long-range transported bioaerosols in the Central Mediterranean
date: 2020-10-19
journal: Sci Total Environ
DOI: 10.1016/j.scitotenv.2020.143010
sha: 
doc_id: 328914
cord_uid: 6zog3xf3

Airborne bacteria were characterized over a 2-y period via high-throughput massive sequencing of 16S rRNA gene in aerosol samples collected at a background mountain European Monitoring and Evaluation Programme (EMEP) Network site (Monte Martano, Italy) located in the Central Mediterranean area. The air mass origin of nineteen samples was identified by air mass modelling and a detailed chemical analysis was performed. Four main origins (Saharan, North-western, North-eastern, and Regional) were identified, and distinct microbial communities were associated with these air masses. Samples featured a great bacterial diversity with Protobacteria being the most abundant phylum, and Sphingomonas followed by Acidovorax, Acinetobacter and Stenotrophomonas the most abundant genera of the dataset. Bacterial genera including potential human and animal pathogens were more abundant in European and in Regional samples compared to Saharan samples; this stressed the relevance of anthropic impact on bacterial populations transported by air masses that cross densely populated areas. The principal aerosol chemical characteristics and the airborne bacterial communities were correlated by cluster analysis, similarity tests and non-metric multidimensional scaling analysis, explaining most of the variability observed. However, the strong correlation between bacterial community structure and air mass origin hampered the possibility to disentangle the effects of variations in bacterial populations and in dust provenance on variations in chemical variables.

The presence and diffusion of bioaerosols (bacteria, viruses, fungi, and other dead or living organisms including biological debris) in the Earth atmosphere impact ecosystems, climate, and human health (Burrows et al., 2009; Burrows et al., 2009; Fröhlich-nowoisky et al., 2016; Pöschl and Shiraiwa, 2015) . The biosphere directly emits bioaerosols into the atmosphere, which subsequently enables their dispersion and transport even at long distances (Després et al., 2012; Womack et al., 2010) . In the course of atmospheric transport, bioaerosols may undergo further chemical and physical transformation, stress, and biological aging upon interaction with UV radiation, photo-oxidants, and various air pollutants like acids, nitrogen oxides, ozone, and aromatic compounds. All these processes can limit or even suppress the vitality of the living fraction of bioaerosol and therefore affect their capacity to diffuse and to colonize new ecosystems (Womack et al., 2010) . Due to the above challenges, the present knowledge on the ability of viruses and bacteria to spread in the air and diffuse infections and more in general diseases is still immature and demands a wide spectrum of investigation (Middleton, 2017; Morawska and Cao, 2020; Polymenakou, 2012) .

Most of the previous studies in the Mediterranean area have been limited to advections of air masses from the Sahara Desert only. The occurrence and impact of this type of air mass, very rich in desert dust, are frequent and well documented (Escudero et al., 2006; Formenti et al., circulation of air masses of different origin and distinguished by nature, type, quality, and extent of contributions (Cusack et al., 2012; Kallos et al., 2014 Kallos et al., , 2007 Petroselli et al., 2018) . Due to the very different characteristics of the source areas, these air masses are expected to carry different bacterial populations and specific chemical markers and pollutants. Moreover only a few studies have used molecular-based approaches to investigate the relationships of different air masses with the bacterial communities in the Mediterranean area. In such studies, bioaerosol characterization was conducted by a low-throughput approach (cloning and sequencing of 16S rRNA gene), while High-Throughput Sequencing (HTS) approaches were used in an even smaller number of cases. Most of the previous studies on aerosol-associated microbial communities in the Mediterranean area have been focused on intense Saharan intrusions sampled in the proximity of the dust sources (Gat et al., 2017; Katra et al., 2015; Mazar et al., 2016; Polymenakou et al., 2008) , or after a long-range transport over the Mediterranean basin (Federici et al., 2018; Rosselli et al., 2015; Sanchez De La Campa et al., 2013) . Much less is known about the specific characteristic of the bacterial communities transported by air masses from continental Europe.

In this frame, the present study aims at defining the patterns of the bacterial communities of atmospheric aerosol from distinct geographic regions reaching the Mediterranean. The samples were collected during different long-range transport events towards a background monitoring site hypotheses to be tested in this work are two-fold: (i) bacterial community structure associated with long-range transported aerosol in the Central Mediterranean area is significantly different based on the air mass provenance; (ii) there is a correlation between the main aerosol chemical characteristics and the airborne bacterial communities. To test these hypotheses, we investigated the chemical and microbial datasets by cluster analysis, similarity tests, and non-metric multidimensional scaling analysis.

All the aerosol samples analyzed in this work were collected at the EMEP regional background site of Monte Martano (MM) in Central Italy (42°48'19''N, 12°33'55''E) . MM has been established in a relatively undisturbed location, near a television antenna, on the ridge of a small mountain chain (1100 m asl), above the timberline and facing a completely free horizon (Moroni et al., 2015) . The site is equipped with aerosol, gaseous pollutants, and meteorological monitoring instrumentations (Moroni et al., 2015) . Due to its elevation, the low background concentrations and the 360° free horizon, the site is particularly suited for the assessment of long-range transport events of atmospheric aerosol (Federici et al., 2018; Petroselli et al., 2018a Petroselli et al., , 2018b . The importance of the site for the monitoring of Saharan dust advections was recognized

The sampled filters underwent a thorough chemical characterization that included the investigation of both the inorganic and organic fractions of particulate matter.

Major ion composition was determined by ion chromatography (DIONEX 2100) after 30 minutes ultra-sonication in ultrapure water (18 MΩ). The quantified analytes were: Li + , Na + , shaken for 1h at maximum speed, centrifuged for 30' at 10000 x g and then at 11500 x g for 15'

at 4C° to recover bacteria (Radosevich et al., 2002) . Supernatant was discarded and DNA was J o u r n a l P r e -p r o o f 

Each sequence was assigned to its original sample according to its index oligos and barcodes. After sorting the sequences, the reverse read of each paired-end sequence was reverse complemented and merged with the corresponding forward read. A quality cut-off was applied in order to remove the sequences that did not contain the barcode, those with an average base quality value (Q) lower than 30 and those that did not provide a perfect match in J o u r n a l P r e -p r o o f Journal Pre-proof the overlapping part between the two paired ends. The barcode was removed and sequences were sorted into Operational Taxonomic Units (OTUs) using the UPARSE-OTU algorithm (Edgar, 2013) . The minimum identity between each OTU member sequence and the representative sequence (i.e. the sequence that showed the minimum distance to all other sequences in the OTU) was set to 97%. The taxonomic classification of each OTU was carried on with the stand-alone version of RDP Bayesian Classifier (Wang et al., 2007) , using a 50% confidence level (Claesson et al., 2010) . Chloroplast sequences were not excluded by further analyses because their abundance can provide information on PM origin.

Three independent extractions, amplifications and sequencing on each sample were performed in order to test the robustness of the proposed experimental approach and the three replicates featured nearly identical OTU distribution profiles (data not shown).

Cluster analysis using the Bray-Curtis similarity index was applied to the bacterial communities belonging to the different aerosol samples. Similarity test (ANOSIM) was performed to detect differences in the bacterial community structure followed by the determination of discriminating genera by means of SIMPER routine. This analysis indicates the average contribution of each genus to the similarity and dissimilarity between groups of samples. Non-metric Multidimensional Scaling (NMDS) analysis was performed using the Bray-Curtis dissimilarity matrix and the first NMDS dimension was then plotted with chemical data in order to gain information from the correlation between abiotic and biotic components of dust samples.

Additionally, the chemical peculiarities of the samples based on similarities highlighted by the NMDS were interpreted by using principal component analysis (PCA). Statistical analyses and J o u r n a l P r e -p r o o f Journal Pre-proof graphical representations were carried out using the R statistical environment (Version 4.0.1 -R Core Team 2020) and ggplot2 package (Wickham, 2016) .

Nine Saharan dust advections and ten long-range transports from other geographical origin have been considered in this work. The air mass origins were identified on the basis of back-trajectory (BT) analysis. The BTs for the identified provenance groups are summarized in Figure 1 , for the 500 m endpoint. The other endpoints (50 and 1000 m above the ground) provided similar results and have been included in supplementary material ( Figure SM3 ). Saharan dust advection samples have been marked with the code SH. As for the other provenances, three main macroareas have been identified, namely regional (RG), North-western (NW) and North-eastern (NE).

RG air masses have been defined as those remaining over the terrestrial and marine sectors of central Italy for at least 48 h before sampling. Table 1 summarizes all the PM samples collected during 2014 and 2015.

As a general trend, the Saharan dust samples are characterized by higher aerosol mass concentrations with respect to the non-Saharan advections (see Table 2 ), i.e. an average +68.4%

for PM 10 and +85.3% for PM coarse , defined as PM 10 -PM 2.5 , and lower PM 2.5 /PM 10 ratio, reading 0.52±0.18 for SH and 0.76±0.12 for non-SH samples. The increase in the concentration of the coarse fraction is typical of natural crustal aerosol sources such as desert dust (Formenti et al., 2011) . Moreover, Ca and Fe, typical crustal markers resulted higher in Saharan dust on average ( Table 2 ). The insoluble fraction of Ca, defined as Ca tot -Ca 2+ , was close to 60% for Saharan dust, J o u r n a l P r e -p r o o f Journal Pre-proof slightly lower for RG air masses and much lower for the NW. This is consistent with both the source area mineralogy and the different atmospheric processes during the long-range transport (Avila et al., 2007) . Biomass burning markers such as ammonium and organic carbon (OC) were higher in non-Saharan samples, and particularly enriched in NE samples, possibly due to the frequent wildfires recorded in Eastern Europe regions. The latter have been found to exert a distinct impact on the Monte Martano site, as previously reported in . The average OC and EC values are in agreement with those reported in (Sandrini et al., 2014) at MM

for the year 2009. Total PAHs were on average the highest for RG followed by SH and NE and NW air masses. Benzo(a)Pyrene, the reference PAH for health effects, has the same order in abundance.

Sequencing of 16S rRNA gene fragments led to the recovery of 1286659 high-quality sequences, which clustered, across all samples, into a total of 10513 operational taxonomic units (OTUs) calculated at 97% of sequence similarity. The average number of OTUs per sample was 2239.

Although a considerable fraction of the total biodiversity (18.6% on average) could not be classified at genus level, a total of 879 different genera were identified across all samples.

Among them, a total number of 116 genera were found whose relative abundance was higher than 0.5% in at least one sample. These genera were considered abundant (abundant genera hereafter) and further analyzed. 32 bacterial genera manifested a relative abundance higher than 0.5% on average in all samples; they are shown in Figure 2 . Overall, the most abundant genera

were Sphingomonas (8.47%), followed by Acidovorax (3.89%), Acinetobacter (3.33%), Methylobacterium is known to be resistant to desiccation and to γ radiation together with

Arthrobacter, also abundant in our samples (Favet et al., 2013) . Microvirga was already found in desert-coming air-masses and some species of this genus can reduce nitrogen gas to ammonia (Favet et al., 2013; González-Toril et al., 2020) . It should be also noted that some of the 32 most abundant genera are human-and animal-associated bacteria and include known pathogens, such as Haemophilus, Staphylococcus, Streptococcus and Propionibacterium (Brock et al., 2012) .

Moreover, we retrieved some bacterial genera that, despite being ubiquitous in the environment, also contain many opportunistic pathogens and a few pathogens, such as the Pseudomonadales Acinetobacter and Pseudomonas, or Clostridium sensu stricto and Clostridium XI (Brown, 2014) . However, analyses based on 16S rRNA sequences do not allow to distinguish pathogenic from non-pathogenic species or strains.

The results of the cluster analysis on the database containing only the abundant genera, summarized by the dendrogram in the right panel of Figure 2 , were used for a data-driven visualization of the samples that are reported in the barplot following the dendrogram order. The average 1-D distances reported in the dendrogram revealed a high β-diversity among the J o u r n a l P r e -p r o o f Journal Pre-proof bacterial communities of the SH samples, which however generally clustered together. At genus level, the structure of bacterial communities clearly showed differences due to the sample provenance rather than to other factors such as seasonality. Interestingly, amongst the PM samples, the non-Saharan samples collected during regional movements of air masses (RG) and, to a lesser extent, during long-range intrusions (NE and NW), showed a high richness of genera with low abundance, indicating a highly diverse and even community. Conversely, PM during Saharan intrusions showed a lower richness of genera, indicating that these microbial communities were dominated by fewer typical phylotypes. This is in contrast with previous studies, which generally reported higher diversity during dust intrusion events compared to nondust events (González-Toril et al., 2020; Griffin, 2007; Mazar et al., 2016; Polymenakou et al., 2008; Sanchez De La Campa et al., 2013) . It may be hypothesized that, in the case of central Italy, both regional air masses and long-range intrusions from NE and NW mainly cross more heterogenous areas than Saharan intrusions, thus collecting a wider variety of microorganisms.

Particularly, when BTs indicated that regional air masses were prevalent, it is also possible that local sources played a major role in shaping bacterial communities. In fact, a wide variety of potential local sources, such as soil surface, leaf surface, water bodies and even animal faeces, To gain further insights about similarities and differences among all the aerosol samples and hypothesize possible effects of the air masses with different origins, a non-metric multidimensional scaling (NMDS) analysis was performed using the computation of Bray-Curtis distances between bacterial communities. This statistical method has been applied to the database containing only the abundant genera, and results are shown in Figure 3 . The results of J o u r n a l P r e -p r o o f NMDS analysis showed a good clustering (stress value < 0.15) of the samples according to their provenance group. In particular, the NMDS1 dimension (x-axis, Figure 3 ) seems to separate well the different clusters, and particularly the Saharan dust samples from the others, while NMDS2 describes the variability within each group.

NMDS is a helpful exploratory analysis but does not allow explaining the similarities or dissimilarities among samples, and additional information from other analysis is needed. For example, samples associated with RG and NW air masses show a partial overlap in the NMDS analysis, which is understandable based on the phenomenology of back trajectories (see Figure   1 ). In fact, RG air masses at MM tend to the terrestrial and marine western sectors of peninsular although it has been demonstrated that a number of diseases are linked to desert aerosols (Middleton, 2017) , the concern about Saharan intrusions might be reduced from a public health point of view in this context, since air masses from European and regional origin were more enriched in human-associated bacteria than Saharan air masses. A more detailed boxplot representation of the distribution of the 116 more abundant genera for the four air mass origins is reported in Supplementary Material (figure SM5 ).

Chemical and microbiological data were combined to check possible correlations between the variables and the sample provenances. In particular, some typical markers of Saharan dust, biomass burning, and industrial activities, the two latter being particularly enriched in non-Saharan samples, were identified amongst the chemical analytes. Moreover, the analysis of the β diversity showed that the microbial communities of long-range transported Saharan dust were significantly different from those sampled when other air masses were present, strongly supporting the hypothesis that desert dust can impact the bacterial composition of the aerosol at our latitudes (Gat et al., 2017; Mazar et al., 2016; Rosselli et al., 2015) . On the contrary, non-Saharan samples showed similar communities among each other, which in fact clustered together (Figures 2 and 3) . Nevertheless, as observed for the chemical characteristics, even if similarities existed within the PM samples sharing the same origin, the differences were not negligible and suggested that each event was independent of the others. This has been already observed in previous works, at least for dust events. In fact, significant differences in bacterial community J o u r n a l P r e -p r o o f Journal Pre-proof structure were reported during different dust events that impacted the same area, even when two events were very close in time (Federici et al., 2018; Yamaguchi et al., 2014) .

In order to combine information about microbiology and chemistry, the NMDS1 dimension from the statistical analysis on bacterial communities was correlated with the concentration of chemical variables, normalized against the PM values (w/w). Some of the statistically significant correlations (p = ***) are shown in Figure 5 .

Specifically, a significant correlation was found between NMDS1 and PM 2.5 (i.e. PM 2.5 /PM 10 ratio). PM ratio was lower for Saharan dust with the exception of the outlier SH_20141107 which corresponded to the weakest Saharan dust event, with a PM 10 concentration of 8.4 µg/m 3 .

The correlation was significant also between NMDS1 and PM coarse ., which was higher for SH because Saharan intrusions are constituted by coarser particles. Organic carbon (OC) content correlated significantly with NMDS1. OC in Saharan dust was lower than in non-Saharan samples because the latter can have a higher anthropogenic contribution. The highest OC/PM 10 values were found for SH_20140624 (which had also high sulphate concentration) and SH_20140522 of NW provenance. As stated above, also many bacterial genera that were significantly more abundant in non-SH than in SH samples (e.g. Lactobacillus, Streptococcus, Propionibacterium and Haemophilus) were generally related to anthropic and built environments. This confirms the relevance of the impact that densely populated areas may exert on bacterial populations transported by air masses.

Anthracene was the only PAH correlating significantly with NMDS1, being higher for SH samples. The sum of low molecular weight PAHs (LW) was also higher for SH samples.

Calcium concentrations showed no correlation with NMDS1, which was interpreted as due to the high local contributions of this element. Iron, on the other side, was richer in SH samples and J o u r n a l P r e -p r o o f Journal Pre-proof correlated negatively with NMDS1. Ammonium and sulphate concentrations were generally higher for non-Saharan air masses. Innocente et al. (Innocente et al., 2017) also reported that high ammonium and sulphate concentrations were associated with long-range transport from north-west in Milan (North Italy), and those air masses presented a high percentage of

Propionibacterium. This is in agreement with our SIMPER analysis, which indicated the genus Propionibacterium as significantly more abundant in NW than in SH samples ( Figure 4 ).

However, Innocente and colleagues also reported that this correlation was weak, and ionic composition of air masses was much more clearly related to air mass provenience than to bacterial community structure.

Indeed, also in our case, since NMDS1 was strongly correlated to air mass origin, it was not possible to fully understand whether variations in chemical variables were more correlated to variations in bacterial community structure or to dust provenance.

In this work, we have characterized the bacterial communities of 19 air masses of different origin, sampled as PM 10 at the remote site of Monte Martano, in Central Italy. This EMEP station is representative of the Central Mediterranean area. The main results of the present work can be summarized as follows:

 Four distinctive air masses were identified: previous similar work on this topic was substantially limited to Saharan (SH) dust air masses while in the present study we extended the characterization to regional (RG), North-Western (NW), and North-Eastern (NE) air masses.

 At genus level, the distribution of the bacterial populations in air masses clearly showed differences due to the sample provenance. In fact, PM 10 during Saharan intrusions J o u r n a l P r e -p r o o f Journal Pre-proof showed a relatively low number of genera, while non-Saharan samples, particularly those collected during regional movements of air masses (RG), showed a high number of different genera with low abundance, indicating a highly diverse and even community. Cluster analysis was performed on the 116 genera whose abundance was higher than 0.5% in at least one sample (abundant genera). Barplots represent only the 32 bacterial genera that showed a relative abundance higher than 0.5%.

Chloroplasts were also included in the analysis.

J o u r n a l P r e -p r o o f Table 1 . Sample characteristics in terms of provenance and aerosol mass concentration in the PM 10 and coarse (PM 10 -PM 2.5 ) fractions. Provenance classification is based on BTs analysis (see Figure S1 in the Supporting Information). 

Case Studies of Source Apportionment and

Variation of soluble and insoluble calcium in red rains related to dust sources and transport patterns from North Africa to northeastern Spain

Continental-scale distributions of dust-associated bacteria and fungi

Desert dust contribution to PM10 loads in Italy: Methods and recommendations addressing the relevant European Commission Guidelines in support to the Air Quality Directive

Microbial hitchhikers on intercontinental dust: catching a lift in Chad

Airborne bacteria and persistent organic pollutants associated with an intense Saharan dust event in the Central Mediterranean

Recent progress in understanding physical and chemical properties of African and Asian mineral dust

Persistent organic contaminants in Saharan dust air masses in West Africa, Cape Verde and the eastern Caribbean

Origin-Dependent Variations in the Atmospheric Microbiome Community in Eastern Mediterranean Dust Storms

Impacts of Saharan Dust Intrusions on Bacterial Communities of the Low Troposphere

Saharan dust storms: nature and consequences

Atmospheric movement of microorganisms in clouds of desert dust and implications for human health

Airborne microorganisms in the African desert dust corridor over the mid-Atlantic ridge

Microbial population structures in the deep marine biosphere

Influence of seasonality, air mass origin and particulate matter chemical composition on airborne bacterial community structure in the Po Valley

Long-Range Transport of Anthropogenically and Naturally Produced Particulate Matter in the Mediterranean and North Atlantic: Current State of Knowledge

Natural and anthropogenic aerosols in the Eastern Mediterranean and Middle East: Possible impacts

Richness and Diversity in Dust Stormborne Biomes at the Southeast Mediterranean. Sci. Rep. 4, 5265

Desert dust hazards: A global review

Airborne transmission of SARS-CoV-2: The world should face the reality

Ground-based measurements of long-range transported aerosol at the rural regional background site of Monte Martano (Central Italy)

Source identification of airborne bacteria in the mountainous area and the urban areas

Disentangling the major source areas for an intense aerosol advection in the Central Mediterranean on the basis of Potential Source Contribution Function modeling of chemical and size distribution measurements

R: A language and environment for statistical computing. R Foundation for Statistical Computing

Development of a high-volume aerosol collection system for the identification of air-borne microorganisms

Microbial immigration across the Mediterranean via airborne dust

Back-trajectories show export of airborne fungal spores (Ganoderma sp.) from forests to agricultural and urban areas in England

Chemical and microbiological characterization of atmospheric particulate matter during an intense african dust event in Southern Spain

Background levels of atmospheric hydrocarbons, sulphate and nitrate over the western Mediterranean

Intercontinental dispersal of bacteria and archaea by transpacific winds

Noaa's hysplit atmospheric transport and dispersion modeling system

Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy

Conservative fragments in bacterial 16S rRNA genes and primer design for 16S ribosomal DNA amplicons in metagenomic studies

Microbial aerosol characteristics in highly polluted and near-pristine environments featuring different climatic conditions

Biodiversity and biogeography of the atmosphere

ggplot2: Elegant Graphics for Data Analysis

Changes in the RG RG NW

CC: microbiological analysis:, DC: Conceptualization, Methodology, Data curation