Stopover habitat selection drives variation in the gut microbiome composition and pathogen acquisition by migrating shorebirds

Abstract Long-distance host movements play a major regulatory role in shaping microbial communities of their digestive tract. Here, we studied gut microbiota composition during seasonal migration in five shorebird species (Charadrii) that use different migratory (stopover) habitats. Our analyses revealed significant interspecific variation in both composition and diversity of gut microbiome, but the effect of host identity was weak. A strong variation in gut microbiota was observed between coastal and inland (dam reservoir and river valley) stopover habitats within species. Comparisons between host age classes provided support for an increasing alpha diversity of gut microbiota during ontogeny and an age-related remodeling of microbiome composition. There was, however, no correlation between microbiome and diet composition across study species. Finally, we detected high prevalence of avian pathogens, which may cause zoonotic diseases in humans (e.g. Vibrio cholerae) and we identified stopover habitat as one of the major axes of variation in the bacterial pathogen exposure risk in shorebirds. Our study not only sheds new light on ecological processes that shape avian gut microbiota, but also has implications for our better understanding of host–pathogen interface and the role of birds in long-distance transmission of pathogens.


Introduction
The impact of gut microbiota on the physiology and fitness of hosts is a topic of e v er-gr owing inter est in studies on animals.Vertebr ate gastr ointestinal tr act is inhabited by a plethor a of bacterial taxa (Waite and Taylor 2014 , Bodawatta et al. 2022a ), which are essential for health of the host organism and, consequently, for its fitness .T he gastr ointestinal micr obiota modulates the functions of an organism in several wa ys , being crucial in the digestion and assimilation of nutrients (Ste v ens and Hume 1998 ), as well as breaking down nondigestible or toxic compounds (Godoy-Vitorino et al. 2008, Zepeda Mendoza et al. 2018 ).The gut microbiota also affects imm unity, constantl y stim ulating both ada ptiv e and innate components of the immune system and playing an essential role in the development of immunity during the early stages of host ontogeny (Simon et al. 2016, Thaiss et al. 2016, Evans et al. 2017 ).The gastrointestinal bacterial community also contributes to individual health by r estr aining the presence of pathogenic microbes via competiti ve exclusion, effecti vely preventing them from causing disease (Kamada et al. 2013 , McLaren andCallahan 2020 ).Finall y, the micr obiota may affect host beha viour (Ma yer 2011 , Cusick et al. 2021 ), which has been mostly demonstrated in model organisms under laboratory conditions (Desbonnet et al. 2014, Slevin et al. 2020 ), but it remains to be studied whether, and to what degr ee, the micr obiota may determine ada ptiv e behavioural and cognitive performance of wild vertebrates (Davidson et al. 2020 ).
In nonresident birds, seasonal migration between breeding and wintering grounds represents one of the most important challenges in the annual cycle, whic h sha pes their behavior, physiology, and anatomy (Alerstam 1990, Newton 2008, Butler 2016 ).During pr emigr atory period man y species c hange their diet and use carbohydr ates-ric h food to accumulate sufficient fat reserves essential to cross geographical barriers (Bairlein and Simons 1995, Ottich and Dierschke 2003, McWilliams et al. 2004 ).Some internal organs can be reduced, whereas others are expanded to meet physiological r equir ements of long-distance flights or hyper phagia, which allows rapid accumulation of fat reserves (McWilliams andKar asov 2001 , Bauc hinger et al. 2005 ).Other physiological tr aits, suc h as oxygen-carrying capacity of blood or intensity of ener gy pr oduction within cells, can also be ada ptiv el y adjusted during the migratory period (Weber 2009, Yap et al. 2019 ).Finally, migratory species can use various habitats or food resources when they stop and refuel at stopover sites scattered along the migration route (Aamidor et al. 2011, Quinn and Hamilton 2012, Lewis et al. 2016 ).Taking all this into account, it can be expected that migration per se and migration-related adaptations should have an impact on the composition of gut microbiota or its rearrangement during the migratory period.For example, diet is considered one of the most important determinants of gut microbiota diversity (Matheen et al. 2022 ) and fattening during pr emigr atory period or habitat-related shifts in diet composition en route should enhance alterations of gut microbiota in migratory species.Despite these theoretical considerations, our knowledge on the factors that gov ern micr obiome composition in nonmodel long-distance migratory bird species is still scant (e.g.Wu et al. 2018, Tr e v elline et al. 2023 ).
Apart from environmental and ecological factors, composition of the gut microbiota should be shaped by basic individual (intrinsic) tr aits, suc h as sex or a ge (Matheen et al. 2022 ).So far, most information on this intrinsic variation originates from poultry r esearc h (v an Dongen et al. 2013 ), but field studies on wild birds supported substantial role of vertical microbe transfer during feeding between hatchings and their parents as the main factor responsible for establishment of gut symbionts (Diez-Méndez et al. 2023 ).At the same time, little is known about how bacterial communities within the gastrointestinal tract change with age in wild birds after fledging period, but scant evidence suggest that first y ear/immature bir ds host less diverse bacterial communities than adult/mature individuals (Kohl et al. 2019, Hernandez et al. 2021 ).
Since gut microbiota is shaped by a complex network of extrinsic and intrinsic factors, similar pr ocesses ar e likel y to driv e acquisition and maintenance of pathogenic bacteria in the host digestiv e tr act.In fact, c hanges of diet or habitats during migr ation can expose migratory birds to an array of novel pathogens.Empirical studies documenting pr e v alence of enter opathogenic bacteria within intestinal flora of wild birds are still sparse and mostly focus on the occurrence of specific strains that present a potential health threat to humans or domestic animals (Hubálek 2004, Benskin et al. 2009 ).Most data on the pr e v alence of bacterial pathogens in wild birds come from studies on disease outbreaks yielding high mortality rate (e.g.Kirkwood et al. 1995, Pedersen et al. 2003, Niedringhaus et al. 2021 ).Pathogenic gastrointestinal flora and the processes of disease transmission have been extensiv el y studied in commerciall y br ed poultry (Evans and Sayers 2000, Bang et al. 2003, Kursa et al. 2022 ), whereas little is known about the source and pr e v alence of the microbial pathogens in wild-living bird species (Benskin et al. 2009 ).Hence, the role of wild migratory birds as vectors of disease could be crucial and underestimated.Some of the most extreme long-distance migrants belong to the clade of Charadrii shorebirds (Battley et al 2012 ), which thus have a great potential for pathogen transmission at large, intercontinental geogr a phical scales (Jourdain et al. 2007, Altizer et al. 2011 ).
Man y P alaearctic shor ebird species migrate between breeding areas in the High Arctic and wintering grounds in southern Africa (e.g.different Calidris species and wood sandpiper Tringa glareola ), although some others spend winter in the areas that ov erla p with their br eeding gr ounds (e.g. common sandpiper Actitis hypoleucos , common snipe Gallinago gallinago , and lapwing Vanellus vanellus ) (Piersma 2007 ).Similarl y, some shor ebirds cov er the entir e migration distance in only a few long-distance flights, adopting a time-minimizing, but ener geticall y demanding migr ation str ategy ( Calidris species), while the others use a large number of stopover sites, flying short distances and using the energy-minimization strategy (snipe and lapwings) (van de Kam et al. 2004 ).Migrating shor ebirds pr efer differ ent habitat types (e .g. coastal vs .freshwater) and their choice might not only be related to species-specific pr efer ences, but can also be modulated by age, sex, or body condition (van Gils et al. 2000, Piersma 2003, Kober and Bairlein 2009, Allen et al. 2020 ).As marine/br ac kish and fr eshwater habitats can str ongl y differ in the amount and type of available food supply (Colwell andLandrum 1993 , Piersma et al. 1993a ), pathogen or pr edation pr essur e (Mendes et al. 2005, Rosa et al. 2006 ), and stability of environmental conditions (Verkuil et al. 1993, Piersma 2012 ), habitat choice during migration may be crucial not only for shorebird behavior, but also physiology.Howe v er, e v en within similar habitats, shorebirds exploit species-specific food resources (hard to digest biv alv es and soft bodied larvae of insects or polyc haetes), as differ ent shor ebir d species sho w unique adaptations in the morphology of foraging apparatus , e .g. bill shape and length (Barbosa and Moreno 1999 ).Differences in forag ing strateg ies during migration can also have important fitness consequences , e .g. species exploiting different food resources within similar habitat types can vary in pathogen exposure and infection risk (Clark et al. 2015, Minias et al. 2016 ).
The main aim of our study was to compr ehensiv el y inv estigate how host species, habitat, and age influence gut microbiomes and bacterial pathogens in migratory shorebird species.To do so, we collected faecal samples from five shorebird species at one coastal site (sandy seashore) and two inland sites (artificial reservoir mudflats and natural riverbed) in Central Europe (Poland).We used bacterial 16S rRN A metabar coding to obtain data on the composition and diversity of gut microbiome across all our study species and our analyses were targeted at four specific aims.First, we aimed to c har acterize interspecific differences in gut microbiota, likel y r eflecting taxonomic v ariation and species-specific differences in migration and for a ging str ategies.Second, we c har acterized intraspecific variation in gut microbiota between different stopover sites and examined whether microbiome diversity and composition were better explained by habitat selection than host taxonomy.Thir d, w e tested for a ge-r elated v ariation in shor ebird microbiome and we hypothesized that the gut microbiota of y oung (immature) bir ds should be less diverse and significantly different in composition from the microbiome of adult individuals (Somers et al. 2023 ).Finally, we focused on the presence of pathogenic taxa within the gut microbiota to assess variation in infection risk related to taxonomy, stopover habitat selection, and host age.

F ield works
The field w ork w as carried out in two consecutive seasons of shor ebird autumn migr ation (June-September 2020 and 2021).Shorebir ds w ere captured at three important stopover sites for migratory w aterbir ds in Poland: Vistula river mouth (54 • 21 N, 18 • 57 E), Jeziorsk o r eservoir (51 • 47 N, 18 • 40 E), and Rembeza's Island (51 • 57 N, 21 • 16 E; Fig. 1 ).The first study area (Vistula river mouth) was located on the southern Baltic Sea coast, where birds use sandy beaches and small brackish ponds within sandy dunes during for a ging and r esting.T he second site (J eziorsko) was located at an artificial dam r eservoir, wher e the water le v el decr eases gr aduall y in autumn, exposing extensiv e ar eas of m udflats rich in benthic in vertebrates .T he third site (Rembeza's Island) was located in the natural Middle Vistula river valley, characterized by the presence of sandy islets, large areas of mud d y riv er banks, and locall y shallow riv erbed.T hus , each sampling site was associated with a different habitat type (i.e.sea coast, artificial inland mudflats, and natural river valle y), re presenting k e y stopover habitats used by migrating waders in Central Europe.Bir ds w ere captured in walk-in traps designed for small and middle-size shorebirds (Busse and Meissner 2015 ), aged using pluma ge c har acteristics (Demongin 2016 ), and ringed to avoid repeated sampling of r eca ptur ed individuals.Afterw ar ds, each bir d was k e pt in a small cage (48 cm × 31 cm × 33 cm) with a sterilized lining.Bir ds w er e r eleased fr om the ca ge either after 5 minutes, if no faeces were deposited, or immediately upon defecation.Faecal samples wer e tr ansferr ed to sterile Eppendorf tube with 96% ethanol in 2020 or ATL buffer (QIAGEN GmbH, Germany) in 2021.Samples were cooled in portable r efriger ators befor e tr ansportation to the labor atory, wher e they were kept at −18 • C until DNA isolation.
Fiv e differ ent shor ebir d species w ere sampled: w ood sandpiper (TRI GLA), common snipe (GAL GAL), common sandpiper (ACT HYP), dunlin Calidris alpina (CAL ALP), and common ringed plover Charadrius hiaticula (CHA HIA).Although we originally aimed to collect samples from each species across different sampling sites, it was not feasible because of different species-specific capture rates at each stopover.We set a threshold of 10 samples collected per species per site and this criterion was satisfied across all three sampling sites for only one species (common sandpiper), across two sites (sea coast and river valley) for another two species (common ringed plover and dunlin), and at only one site (artificial inland mudflats) for the remaining two species (wood sandpiper and common snipe).It was possible to collect good quality samples from the two distinguishable age classes (juvenile vs. adult) only for the dunlin (50% of adults) and common ringed plover (36% of adults).Samples from the remaining species were collected only for juv eniles, whic h ar e m uc h mor e abundant during autumn migration than adults (Meltofte 2008 ).

DN A extr action, sequencing, and bioinformatics
Bacterial DNA for microbiome metabarcoding was extracted from faecal samples using the DNeasy Po w erSoil Pro kit (QIAGEN GmbH) following the manufacturer's protocol.For the first subset of faecal samples, we used 96% ethanol as a stor a ge buffer, but due to the low efficiency of DNA extraction (only 24% of extracted faecal samples yielded > 10 ng/ μl DN A concentration), w e decided to store all the remaining samples in the ATL tissue lysis buffer (QIAGEN GmbH), whic h r esulted in a marked impr ov ement of DNA extr action r ate using the same extr action kit (99% samples showing > 10 ng/ μl DNA concentration).All samples yielding poor DNA extraction success were excluded from downstream processing.Ho w ever, since storage conditions may affect the composition of bacterial strains detected in avian faecal samples (Zhou et al. 2019 ), we controlled for sample storage buffer in the anal yses of micr obiome v ariation.Sequencing of the v ariable V3 and V4 regions of the 16S rRN A gene w as conducted using primers 515F-806R applied in Earth microbiome project (Klindworth et al. 2013 ) and standard Illumina protocol, following the Illumina 16S Meta genomic Sequencing Libr ary Pr epar ation Guide .T he amplicons wer e m ultiplexed with a unique dual-barcode combination used for each sample and sequenced using two 250-bp paired end runs on Illumina MiSeq platform in the BioBank Laboratory, University of Lodz.In total, 180 samples were sequenced, including 159 unique samples, 15 technical replicates (independent amplicons from the same samples replicated within and between sequencing runs, eight and se v en samples, r espectiv el y), and six negativ e contr ols (amplicons fr om the buffer with no faecal material).
Raw MiSeq sequences were trimmed using Trim Galore v0.6.7.Process of quality filtering, merging, denoising and filtering c himer as was performed with D AD A2 algorithm implemented in the QIIME2 v2022.2pac ka ge (Bol yen et al. 2019 ).Sequences were categorized to amplicon sequence variants (ASVs) based on 100% similarity.Bacterial species wer e r ecognized based on 97% sequence similarity using Silva 138 SSURef NR99 database (Quast et al. 2013, Robeson et al. 2021 ).Identification of putative pathogens was conducted using two databases, Silva 138 SSURef NR99 and FAPROTAX script (Louca et al. 2016 ) and we assigned our bacterial taxa to four ecological groups (human pathogens, intr acellular par asites, pr edatory or exopar asitic, and animal parasites or symbionts).All bacterial species assigned to these groups wer e v erified a gainst classifications pr ovided by Benskin et al. ( 2009 ) and other av ailable liter atur e to confirm their pathogenicity in avian hosts.Putative nonbacterial sequences (mitochondrial, c hlor oplast, and c y anobacteria) w er e r emov ed fr om the dataset using QIIME2 filtering plugins.Processing of negative controls revealed little evidence of contamination [on average 292.6 ± 84.8 (SE) reads per sample].Samples with < 1000 reads were removed fr om downstr eam anal yses.Pr ocessing and anal ysis of tec hnical replicates indicated excellent replicability of gut microbiome composition.In fact, the r elativ e ASV abundance within eac h sample (proportion of raw reads of specific ASVs to total number of reads) sho w ed very high and significant intraclass coefficients (ICC), both within and between sequencing runs (ICC > 0.99, all P < .001).ICC were calculated using the irr R pac ka ge (Gamer et al. 2019 ) de v eloped for the R statistical envir onment (R De v elopment Core Team 2021 ).In total, 136 samples passed our quality control process and the final sample size r anged fr om 17 (common snipe) to 42 (common sandpiper) samples [on av er a ge 16 ± 1.5 (SE) samples per species per site] [Fig. 1 ; Table S1 ( Supporting Information ) in the Electronic Supplementary Material ].
Alpha diversity (ASV richness and Shannon diversity index H') and beta diversity (pairwise J accar d distances) indices were calculated using the BIOM table in QIIME2 to e v aluate ric hness and diversity of bacterial species.Microbiome composition was characterized at different levels of taxonomic resolution (phylum, class, and genus) using the mean number of reads assigned to each taxon r elativ e to the total number of reads.We used r ar efaction (as implemented into QIIME2) to normalize data (diversity measures) for variation in sequencing depth (i.e.number of sequences) (Weiss et al. 2017 ).
General diet composition of our study shorebird species was c har acterized using data summarized by Cramp et al. ( 1998 ).For each species, we compiled information on the presence or absence of main prey types in the diet and this c har acterization was conducted on the higher (class) and lo w er (family) taxonomic level.We used J accar d distance to quantify pairwise diet differentiation between species.

Sta tistical anal ysis
To compare the gut microbiome alpha diversity between species, sites, and age classes, we entered ASV richness and Shannon index H' as the response variables in separate generalized linear models (GLMs).First, we ran GLMs testing for interspecific differences in alpha diversity and species identity was entered as a fixed factor in these models.In the second step, species-specific models wer e de v eloped for thr ee species that wer e sampled acr oss differ ent stopov er sites (common sandpiper, dunlin, and common ringed plover).In all these models, site was entered as a fixed factor, wher eas a ge was enter ed as another fixed factor onl y for dunlin and common ringed plover (all sampled common sandpipers were juveniles).In all GLMs storage buffer type was included as a fixed factor to control for any possible variation in micr obiome div ersity and composition attributed to different sample stor a ge pr otocols.ASV ric hness was anal ysed using Poisson distribution models and a logit link function, whereas H' was analysed using Gaussian distribution models.To compare the gut microbiome composition between species , sites , and age classes, we entered beta diversity indices (J accar d distance matrices based on the presence/absence of ASVs) into permutational multivariate analysis of variance (PERMANOVA).Three separate models were run with species , site , and age (where possible) as explanatory v ariables.All PERMANOVA anal yses wer e conducted using the QI-IME2 pac ka ge.Statistical significance was inferr ed based on the pseudo-F statistic and the r esulting P v alues wer e corr ected for m ultiple testing ( q -v alues) using Benjamini and Hoc hber g method (Benjamini and Hoc hber g 1985 ). Differ ential abundance anal ysis was conducted using analysis of composition of microbiomes in QIIME2 to identify significantly and differentially abundant ASVs in each sampling location and age group (Mandal et al. 2015 ).To test for differences in the prevalence of putative pathogenic gut bacteria between species , stopo ver sites , and age classes , we ran GLMs using binomial distribution and a logit link function.An occurrence of each putative pathogen was entered as a binary response v ariable (pr esence/absence) in a separ ate model.We performed GLM anal yses onl y for pathogens with > 20% total pr e v alence across all samples .T he results of full models are presented for eac h anal ysis .T he models were fitted using restricted maximum likelihood and run with glm function in R statistical environment.
In order to c har acterize the r ole of major pathogens in the gut micr obial comm unities we anal ysed micr obial co-occurr ence netw orks (Co yte et al. 2015 ).Specifically, we estimated the number of inter connections betw een each major pathogen and other bacterial species that sho w ed significant co-occurrence within samples (Spearman's correlation coefficient > 0.3).The analysis was performed using networkX and pandas libraries in Python v3.9.2.To compare and visualize composition of microbiome similarity across species and stopover sites, we used nonmetric multidimensional scaling (NMDS) based on J accar d distances matrix (Hancock 2004 ).The analysis was performed only for statistically significant relationships, as previously inferred with GLMs models.For the purpose of analysis, we used metaNMDS function from the vegan R pac ka ge.Associations between pairwise differ entiation (J accar d distances) of microbiome and diet composition were analysed with Pearson product-moment correlation coefficient.All values are reported as means ± SE.

Alpha diversity
Ther e wer e significant interspecific differ ences in the gut microbiome alpha diversity, as quantified with ASV richness and H' ( Table S2 , Supporting Information ).The highest ASV richness (185.0 ± 59.3) was found in the common ringed plover and it was significantly higher when compared to all the other species (range 68.9-164.6,all P < .001; Figure S1A , Supporting Information ).An anal ysis of H' r e v ealed a slightl y differ ent pattern, showing similar alpha diversity between common ringed plover, dunlin, and common sandpiper ( Table S2 and Figure S1B , Supporting Information ), while significantly higher H' was found in dunlin compared to wood sandpiper and common snipe .T he lo w est ASV richness w as found in the common sandpiper (68.9 ± 8.9), whereas wood sandpiper had lo w est H' (2.45 ± 0.18).
At the intraspecific level, we found significant between-site differences in ASV richness in two shorebird species: common sandpiper and common ringed plover (Table 1 ).Common sandpipers ca ptur ed at the inland artificial reservoir had lo w er micr obiome ASV ric hness compar ed to birds fr om coastal ar eas, but higher than individuals ca ptur ed in natural river valley (Table 1 ).In contrast, common ringed plovers from the coastal site had significantly lo w er ASV richness compared to plovers fr om natur al riv er v alley (Table 1 ).An analysis of the Shannon index did not support significant between-site variation in the common sandpiper, although it was nearly significant in the common ringed plov er, r e v ealing the pattern of variation consistent with ASV richness ( Table S3 , Supporting Information ).No significant between-site differences in microbiome alpha diversity (ASV richness and H') were found in the dunlin (Table 1 ; Table S3 , Supporting Information ).
We also found significant a ge-r elated v ariation in the microbiome ASV richness of both common ringed plover and dunlin (i.e. the only two species with both age classes sampled).The patterns of v ariation wer e consistent across the species, as adult bir ds sho w ed significantl y higher ASV ric hness in comparison to juveniles (Table 1 ).Ho w ever, no significant age-related variation in the H' was found in either plovers or dunlins ( Table S3 , Supporting Information ).
found in the common ringed plover versus common snipe (0.956 ± 0.001) and wood sandpiper (0.957 ± 0.001), while the lo w est J accard distances were found in the dunlin versus common sandpiper (0.935 ± 0.001) and ringed plover (0.941 ± 0.001).Despite these significant differences, no clear clustering by species was r e v ealed by NMDS a ppr oac h ( Figur e S2 , Supporting Information ).We also found a significant effect of stopover habitat on beta diversity in all shorebird species that were sampled across multiple locations (PERMANOVA: all P < .001).When we looked at differences between habitats in the dunlin and common ringed plover, J accar d distances betw een riv er v alley and sea coast were significantly higher than within-habitat distances (all q < 0.01).In the common sandpiper, the highest J accar d distances were found between sea coast and inland reservoir (0.949 ± 0.001), whereas the lo w est J accar d distances w ere found betw een sea coast and riv er v alley (0.894 ± 0.004).All pairwise between-habitat distances wer e significantl y higher than r espectiv e within-habitat distances in this species (all q < 0.05).Ov er all, mean between-habitat Jaccard distances were higher in the common ringed plover (0.946 ± 0.001) compared to dunlin (0.935 ± 0.001) and common sandpiper (0.923 ± 0.002).Consistentl y, differ ential abundance anal ysis r ev ealed the pr esence of habitat-specific bacteria onl y in the common ringed plover, including three taxa from Bacilli ( Bacillacae ) and Gammaproteobacteria ( Steroidobacteraceae and Sutterellaceae ) significantl y mor e abundant in samples from ri ver valle y than sea coast.NMDS a ppr oac h sho w ed r elativ el y good clustering of the gut micr obiome comm unities by stopov er site within all thr ee species (Fig. 4 ).Finall y, beta div ersity was associated with age in the dunlin (PERMANOVA: pseudo-F = 1.61,P = .006),although differential abundance analysis failed to identify any specific bacterial taxa significantly contributing to these variation.No age-related variation in beta diversity was detected in the common ringed plover (PERMANOVA: pseudo-F = 1.04,P = .28).

Diet and microbiome composition
The strongest differentiation of diet composition (prey types characterized at the family level) was observed between the common ringed plover and three other species , i.e .common snipe , wood sandpiper, and common sandpiper (Jaccard distance range = 0.90-0.92),while the weakest differentiation was found between

Pathogenic bacteria within the gut microbiome
Based on classification from Benskin et al. ( 2009), we identified 26 species of putative avian pathogenic bacteria in the gut mi-cr obiome acr oss all fiv e shor ebird species ( Table S4 , Supporting Information ).Additional 33 putative pathogens were identified based on FAPROTAX database ( Table S5 , Supporting Information ), although two of them ( Citrobacter freundii and Esc heric hia coli ) wer e excluded fr om downstr eam anal ysis despite high pr e v alences (46.3% and 56.6%, r espectiv el y), as showing onl y opportunistic or facultative pathogenic activity.Nearly half of all identified pathogens ( n = 26) were present only in a single sample, but eight bacterial species were found in all study shorebird taxa ( Tables S4 and S5 , Supporting Information ).Three pathogenic bacteria ( Campylobacter lari , Mycoplasma iowae , and Enterobacter cloacae ) sho w ed r elativ el y high pr e v alence ( > 30%) in the gut microbiome across all hosts ( Table S6 , Supporting Information ; Fig. 5 ).
Ther e wer e significant interspecific differ ences in the pr e v alence of three pathogenic bacteria, i.e. C. lari , E. cloacae , and Vibrio cholerae ( Table S6 , Supporting Information ).Pr e v alence of C. lari was highest in the common sandpiper and common snipe (64.3% and 47.1%, r espectiv el y), showing significant differences from the remaining shorebird species ( Table S6 , Supporting Information ).
In contr ast, pr e v alence of E. cloacae and V. cholerae was significantly higher in the dunlin and common sandpiper, compared to the other three species ( Table S6 , Supporting Information ).Mycoplasma iowae was identified to be the most embedded into the cooccurrence netw ork, sho wing numerous ( n = 92) positive interconnections with nonpathogenic gastrointestinal bacteria ( Figure S4 , Supporting Information ).Enterobacter cloacae sho w ed m uc h fe wer interconnections with other members of bacterial community (some shared with M. iowae ), representing either positive ( n = 4) or negative ( n = 4) associations.Vibrio cholerae and C. lari also sho w ed few interconnections (all positive), but formed separate clusters fr om M. iow ae and E. cloacae ( Figur e S4 , Supporting Information ).We found considerable between-site variation in the pr e v alence of pathogenic bacteria, but these patterns were hostspecific.The pr e v alence of E. cloacae differ ed significantl y between stopover habitats in only one species, the common sandpiper, as birds from natural river valley were more often infected than birds from artificial reservoir ( Table S7 , Supporting Information ).The pr e v alence of M. iow ae differ ed significantl y between stopover habitats in the common sandpiper and dunlin ( Table S8 , Supporting Information ), being higher in common sandpipers from the inland artificial reservoir (61.5%) compared to sea coast (16.7%) and in dunlins fr om natur al riv er v alley (69.2%) compared to the coastal stopover site (5.6%).There were also significant between-site differences in the prevalence of V. cholerae in the common sandpiper, but the pattern was opposite to M. iowae , as more individuals from the sea coast were infected (50.0%) compared to artificial reservoir (7.7%) ( Table S9 , Supporting Information ).The pr e v alence of C. lari did not show any significant differences between sites ( Table S10 , Supporting Information ).We found significant differences in infection rate between age classes for only one putative bacterial pathogen in only one shorebird species, as the prevalence of V. cholerae was significantly higher in adult than juvenile dunlins (81.2% vs. 6.6%, r espectiv el y; Table S9 , Supporting Information ).

Discussion
Our study not onl y r e v ealed significant interspecific differences in the composition and diversity of the gut microbiota among fiv e shor ebird species during migr ation, but also pr ovided e vidence for intraspecific variation in the gut microbiome between stopover sites and age classes.We also detected the presence of nearl y 60 putativ e bacterial pathogens in shor ebird faecal sam- ples, also showing site-and a ge-r elated differ ences in pr e v alence.Our study reinforces the view that an application of metabarcoding a ppr oac hes to c har acterize micr obiomes of nonmodel v ertebrate species may still provide novel insights into the complex interface between bacterial symbionts/pathogens and their hosts.
Shor ebirds ar e a div erse gr oup showing a high le v el of v ariation in terms of ecology, feeding behavior, and habitat preferences (del Ho y o et al. 1996 ).Their interspecific variation in fora ging tec hniques driv es str ong differ ences in diet and e v entuall y may promote interspecific variation in the composition of gut micr obiota, e v en at the same geogr a phical locations .T he speciesspecific composition of gut microbial community in similar environmental conditions was previously found in two shorebird species, the red knot Calidris canutus and rud d y turnstone Arenaria interpres migr ating thr ough Delawar e Bay, United States (Grond et al. 2014 ).In fact, only 5 of 46 bacterial genera were shared betw een these tw o species, ho w e v er, the r esults wer e based on material collected only from three individuals (Grond et al. 2014 ).Similar effect of host taxonomy on the composition of gut microbiota was reported in a wide range of other avian lineages, including ducks , owls , penguins , and passerines (Dewar et al. 2013, Bodawatta et al. 2018, Hird et al. 2018, Maraci et al. 2021, Bartlow et al. 2022 ).In man y cases, differ ences in the gut micr obiota wer e still a ppar ent despite sympatric occurrence and similar habitat types used by different species (Yang et al. 2016, Cho and Lee 2020, Lu et al. 2022 but see Grond et al. 2019 ), which is consistent with our results.It has also been suggested that diet and feeding preferences are among the most important drivers of interspecific variation in the composition of gut microbiota, not only in shorebirds, but also in other avian linea ges (Gr ond et al. 2018, Matheen et al. 2022, Sun et al. 2022 ).This hypothesis has been supported by the observations of diet-related shifts in the gut microbiome composition, as reported for both nonpasserine (e.g.Siberian crane Grus leucogeranus , common crane Grus grus , and great bustard Otis tarda dybowskii ; Li et al. 2021, Wang et al. 2023 ) and passerine birds (e.g.great tit Parus major and house sparrow Passer domesticus ; Davidson et al. 2020, Teyssier et al. 2020 ).In our study, we did not found a clear statistical support for associations between differentiation in diet and microbiome composition (although some posi-tiv e tr ends wer e detected), but our quantification of diet composition was rather rough and based on published species-specific information.We acknowledge that direct empirical data on individual variation in diet composition could enhance the po w er to detect microbiome-diet associations (Bodawatta et al. 2022b ).The str ongest differ entiation of the micr obiome composition (as expressed by pairwise Jaccard distance) was found between the common snipe and common ringed plover, birds that use completel y differ ent feeding tec hniques (v an de Kam 2004de Kam et al. 2004 ) and exploit different food resources (as shown here with large Jaccard distances observed for main prey types).The common snipe has the longest bill and feeds mainly on large invertebrates (such as earthw orms) b y pr obing in m ud or soil, wher eas the common ringed plover has short pointed bill and feeds on small insects, catc hing them activ el y fr om the surface of water and land (Cramp et al. 1998, Kozik et al. 2022 ).Despite these differences, we found that many gut microbial taxa were still shared between the hosts and clustering methods did not allow for any clear separation of microbiota according to host taxonomy, suggesting that these effects are relatively weak.
Apart from the taxonomic effects, we observed a substantial en vironmental component (i.e .stopo ver site habitat-related variation) in the composition of gut microbiota in all three shorebird species sampled at different locations.Wetland habitats exploited b y shorebir ds r epr esent a br oad gr adient of v arying envir onmental conditions, either in terms of sediment quantity, percentage of organic matter, or water quality .Consequently , different wetland ecosystems are inhabited by different invertebrate communities, whic h r equir es adjustments in shor ebird diet.In gener al, dam reservoirs located on large lowland rivers in Central Europe are rich in organic sediments and their benthic fauna is often dominated by Diptera larvae (mainly Chirominidae ) and oligochaetes (Grzybkowska andDukowska 2001 , Pozna ńska et al. 2009 ).Natur al riv er v alleys ar e mor e v ariable in terms of oxygen conditions and sediments, so their benthic fauna is often more diverse compar ed to anthr opogenic r eservoirs or r egulated riv ers (Horsák et al. 2009, Jones 2013 ).Finally, coastal areas dominated by brackish water and sandy sediments are expected to show high divergent inv ertebr ate faunas fr om inland wetland ecosystems, being dominated by molluscs, pol yc haetes, aquatic arthr opods, and crustaceans (Kotwicki 1997, Hansen et al. 2008, Henseler et al. 2019 ).Differ ences in av ailable food r esources may not onl y impact diet composition of shorebirds that use different wetland ecosystems along their migr atory r oute, but birds at different stopo ver sites ma y also ingest different food-associated microorganisms, which could colonize their gastrointestinal tract (Grond et al. 2018 ).So far, tight relationships between diet, food niche, and micr obiome composition wer e r e v ealed in a compar ativ e study of 21 tropical bird species (Bodawatta et al. 2022b ).If diet shapes the gut microbial community, bird species that use different habitats and variable food resources are expected to show higher plasticity in microbiome composition than species adapted to a single habitat type and having narr ow for a ging nic hes.We can assume that shorebirds using different stopover sites during migration should r espond quic kl y to local conditions and adjust composition and diversity of their gut microbiota.Extraordinary capacity of bir ds, shorebir ds in particular, for a quic k ada ptation to local conditions during migration is well acknowledged (Alerstam 1990, Berthold et al. 2003 ).Migr ating shor ebirds hav e been reported to rapidly remodel many physiological processes, such as the rate of nutrient circulation in blood (Jenni-Eiermann and Jenni 2003, Klaassen et al. 2012, Araújo et al. 2022 ) or fat accum ulation r ate (Maillet andWeber 2006 , Araújo et al. 2019 ), but they also show r a pid mor phological adjustments , e .g. temporal shifts in the size of internal organs (Piersma et al. 1993b, Battley et al. 2000 ).It was also shown that r a pid body mass gain during migration in shorebirds was associated with changes in their gut microbiome composition (Grond et al. 2023 ).An active migration was associated with changes in the gut microbiota in Calidris shorebirds .For example , migrating red-necked stints Calidris ruficollis had higher pr e v alence of Corynebacterium bacteria than resident conspecifics and similar adjustments were also reported for migr ating curle w sandpiper Calidris ferruginea (Risel y et al. 2018 ).So far, the majority of studies on habitat-related variation in the gut bacterial comm unity wer e based on the samples collected at a single location or during the breeding season (e.g.Lewis et al. 2016, Góngora et al. 2021, Drobniak et al. 2022 ), although some exceptions occur.Between-habitat comparison of the gut microbial community in nine species of Darwin's finches sho w ed that lowland versus highland habitat variation was a primary determinant the microbiome composition (Loo et al. 2019 ).Interspecific comparisons also indicated that local environmental conditions may be considered a crucial driver of the gut microbiota composition in birds and their importance may e v en exceed the effects of phylogenetic relationships (Hird et al. 2014, Grond et al. 2019, Skeen et al. 2023 ).In a gr eement with this hypothesis, we observed a r elativ el y good clustering of individual microbiome composition by the stopover site habitat (within species), and this clustering was mor e a ppar ent that the clustering by species.Similar v ariation was observed in the microbiome alpha diversity measures, as the common ringed plover and common sandpiper sho w ed differ ent ASV ric hness between all sampling locations .T hus , habitat type and diet may not only affect taxonomic composition of gut microbiota, but also its overall level of diversity.
So far, a ge-r elated differ ences in the gut micr obiota wer e detected in various bird species, but the majority of studies were based on comparisons between nestlings and adult birds (van Dongen et al. 2013, Barbosa et al. 2016, Bartlow et al. 2022 ).In gener al, bacterial comm unity in c hic ks is often less abundant and mor e tempor aril y unstable than in adult individuals (Sun et al. 2022 ).Highl y div erse comm unities of micr obial commensals can impr ov e health status of their host by protecting it from the colo-nization of intestinal tract by pathogens (Grond et al. 2018 ).Adult birds with a well-de v eloped imm une system and a better r esistance to disease than immature individuals are expected to have a div erse comm unity of bacterial commensals.Higher diversity of cloacal microbiota during the breeding season was found in adult (after second year) female tree sw allo ws Tach ycineta bicolor compared to young individuals (second-year) (Hernandez et al. 2021 ).The analysis of the gut microbiome composition in mute swans Cygnus olor from different age classes also provided evidence for lo w er diversity of bacterial community in the first year birds compared to older individuals, but these differences were only weakly pronounced (Hill et al. 2023 ).In our study, first-year dunlins and common ringed plovers had lo w er gut bacterial diversity in comparison with adult birds, although these differences were apparent only at the level of ASV richness (but not H').Interestingly, not onl y alpha div ersity, but also the composition of gut microbiota sho w ed a ge-r elated differ ences in the dunlin.It suggests that not only the number of gut bacterial taxa increases with age, but some taxonomic r earr angement of micr obiota composition may also be observed during the ontogen y.The intensiv e pr ocesses of micr obiome comm unity formation take place pr obabl y shortl y after fledging, when full y gr own juv eniles start to for a ge independentl y and their diet gr aduall y becomes similar to the diet of adult birds.Ho w e v er, str ong a ge-r elated differ ences in for a ging efficiency or for a ging habitats ar e often observ ed betw een first-y ear and older shor ebirds, whic h can maintain some variation in diet long into the postfledging period (e.g.Goss-Custard andDurell 1987 , Cresswell 1994 ).As a r esult, a ge-r elated v ariation in the gut microbiome diversity and composition may still be detectable during the autumn migration period, as found in this study.
We observed a relatively large number and diversity of putative bacterial pathogens within intestinal tract of all shorebird species migr ating thr ough our study sites in Poland.It is well acknowledged that wild birds act as reservoir and vectors of different bacterial and viral diseases (Dhama et al. 2008, Chung et al. 2018 ).Ov er the last decades a gr eat effort has been invested to study avian pathogens that cause epidemiologic threat to poultry or humans (e .g. a vian influenza virus, West Nile virus, or Salmonella ) (Brittingham et al. 1988, Martinez-De La Puente et al. 2018 ).Ho w e v er, an adv ent of next-gener ation sequencing er a brought an easy access to methodology that allows to detect a broad spectrum of pathogens from different taxonomic groups, pr oviding an unpr ecedented opportunity to gain a better understanding of interactions between avian hosts and their pathogens (Benskin et al. 2009, Konicek et al. 2016, Rajapaksha et al. 2019, Michel et al. 2021 ).For example, an application of metabarcoding a ppr oac hes r e v ealed se v er al potentiall y pathogenic bacteria in the gut microbiota of two passerine species, Swainson's thrush Catharus ustulatus and gray catbird Dumetella carolinensis , migrating along the coast of the Gulf of Mexico, including Shigella as the most abundant pathogenic taxon (Lewis et al. 2016 ).The investigation of the gut microbiota in the Gala pa gos penguin Spheniscus mendiculus r e v ealed the presence of five putative pathogenic taxa, including one taxon ( Clostridium perfringens ) with extr emel y high pr e v alence (95% samples) (Rohrer et al. 2023 ).The analysis of the gut microbiota in three nonpasserine bird species (great bustard, common crane, and common coot Fulica atra ) during the winter period r e v ealed the pr esence of 13 potentiall y pathogenic genera, but all sho w ed rather lo w ( < 4%) r elativ e abundance (Lu et al. 2022 ).In our study we detected nearly 60 species of putative pathogenic bacteria in faecal samples from shorebirds migrating though Poland and eight of them were found in all study hosts .Moreo ver, C. lari , M. iowae , and E. cloacae were present in more than 30% of all sampled individuals (across species).The majority of associations between pathogens and other members of bacterial community were positive, indicating that some nonpathogenic bacterial taxa may promote colonization of digestive tract by pathogenic agents.Taking all this into account, it seems that shorebirds may be considered as relatively important reservoirs of pathogens, which could be primarily associated with their feeding ecology.The pr efer ences for wetland habitats, where the risk of exposure to diverse bacteria is relatively high, especially when pathogens spread through water contaminated by se wa ge or a gricultur al pr actices (Cabr al 2010 ), makes shorebirds specificall y pr one to act as hosts for a br oad spectrum of bacterial pathogens.Between-and within-species transmission of bacterial pathogens may also be promoted by gregariousness during the migr atory period, whic h is typical for most shorebirds, as they usually gather in huge interspecific flocks at favorable stopover sites (van de Kam 2004et al. 2004, Dhama et al. 2008, Kole ček et al. 2021 ).
The common sandpiper and common snipe were identified as the main reservoirs of C. lari in our study, showing significantly higher pr e v alence than the r emaining thr ee shor ebird species.In contrast, the dunlin and common sandpiper sho w ed significantly higher pr e v alence of E. cloacae and V. cholerae than the common ringed plo ver, common snipe , and wood sandpiper.Interspecific differences in pathogen prevalence that were observed in our study could be attributed to the div er gent for a ging tec hniques or differences in diet between shorebirds, as feeding ecology appears to be a k e y determinant of bacterial acquisition (Benskin et al. 2009 ).Our pr e vious r esearc h sho w ed that differences in for a ging nic he could hav e a major impact on exposur e to avian botulism caused by Clostridium botulinum in two shorebird species, common snipe and wood sandpiper (Minias et al. 2016 ).Similarly, the presence of Campylobacter bacteria within the intestinal tract was driven by feeding behavior and varied amongst ecological guilds of birds, as insectivores and granivores were only rarely identified as hosts for this pathogen (Waldenström et al. 2002 ).Despite a ppar ent interspecific v ariation in the pr e v alence of some important avian pathogens, an ov er all species composition of pathogenic gut bacterial fauna was r elativ el y similar between hosts .In consequence , shor ebirds ar e pr obabl y exposed to a gener all y similar arr ay of pathogens and the fine-scale interspecific differ ences ar e likel y to become mor e a ppar ent at the le v el of pr e v alence, r ather than the taxonomic composition of pathogenic bacterial communities.
The differences in prevalence of two bacterial pathogens ( M. iowae and V. cholerae ) between stopover sites could be related to local conditions at each stopover site or habitat pr efer ences of migrating individuals.Birds that use locations with high microbial contamination le v el (e.g.dam r eserv oirs) could sho w higher pr e v alence le v el of bacterial pathogens than individuals using mor e natur al or less contaminated habitats (e.g.coastal sites).Consistent with this prediction, we observed significantly lo w er pr e v alence of M. iowae in both common sandpipers and dunlins migrating along the sea coast ( < 20%) than through artificial inland reservoir or natural river valley (60%-70%).Ho w ever, prevalence of V. cholerae sho w ed the opposite pattern (higher at the sea coast than inland reserv oir), so w e ackno wledge that there could be many more external or intrinsic factors that shape exposure of shorebirds to pathogens.Vibrio cholerae is one of the most important bacterial pathogens in terms of public health, as it can be transmitted to humans, causing acute diarrheal illness (Lipp et al. 2002, Vezzulli et al. 2010 ).Available information suggest that V. cholerae may be occasionally present in wild birds, mostly in species found in aquatic habitats, including shorebirds (Ogg et al. 1989, Hubálek 2004, Ayala and Ogbunugafor 2023 ).Consistently with these observations, we detected V. cholerae in all our study species fr om differ ent gener a, but the pr e v alence in some species (e.g.45% in the dunlin) was clearly higher than usually reported for wild birds (Hubálek 2004 , Ayala andOgbunugafor 2023 ).
In conclusion, our stud y effecti v el y decomposed taxonomic, environmental (habitat), and intrinsic (age) components of variation in the diversity and composition of gut microbiota among five species of shorebirds .T he results indicated that local stopover habitat may play a k e y role in shaping the gut microbiota of migr ating shor ebirds , and these local effects ma y likely dominate over the effects of taxonomic variation.We also provided empirical evidence for the significant role of shorebirds as reservoirs of bacterial pathogens and r e v ealed a complexity of mechanisms that could determine pathogen exposure risk in this group of birds .T hus , our study not only sheds new light on ecological processes that shape avian gut microbiota, but also has implications for our better understanding of host-pathogen interface and the role of birds in long-distance transmission of pathogenic bacteria.At the same time, we acknowledge that any active pathogenicity of gastrointestinal bacteria was not confirmed in our study and, thus, an y infer ences on the occurr ence and r ole of gut bacterial pathogens in shorebirds should be made with caution.

Figure 4 .
Figure 4. NMDS plot based on Jaccard distances showing the clustering of gut microbiota communities by stopover site for the common sandpiper (A), common ringed plover (B), and dunlin (C).

Figure 5 .
Figure 5. Pr e v alence of four putative avian pathogenic bacteria in the gut microbiome across five shorebird species migrating through Poland.Species are marked with acronyms: common ringed plover (CHA HIA), dunlin (CAL ALP), common sandpiper (ACT HYP), common snipe (GAL GAL), and wood sandpiper (TRI GLA).

Table 1 .
Habitat and a ge v ariation in ASV richness of the gut microbiome in three shorebird species.Sample storage buffer was included as a fixed factor.Significant predictors are marked in bold.