Intestinal microbiota of Nearctic- Neotropical migratory birds vary more over seasons and years than between host species

Seasonal migration of Nearctic- Neotropical


| INTRODUC TI ON
Vertebrates harbour diverse microbial communities within their gastrointestinal tract which influence numerous host-associated functions, including dietary specialization (Gunasekaran et al., 2021;Kohl et al., 2016), immune system functioning (Broom & Kogut, 2018), metabolic capacity (Apajalahti & Vienola, 2016), and behaviour (Slevin et al., 2020). The complex symbiosis between a host and its microbiome is often a dynamic relationship, varying in response to both intrinsic and extrinsic factors (Adair & Douglas, 2017;Capunitan et al., 2020;Hird et al., 2015). In birds and many other nonmammalian vertebrates, an individual's microbiota is shaped in large part by the ecology, diet, and environment of the host, often more so than by the host species identity (Capunitan et al., 2020;Fleischer et al., 2020;Grond, Santo Domingo et al., 2019;Michel et al., 2018;Song et al., 2020). A key objective of research on host-associated microbiota is to determine how the composition of microbes varies across time within host species (Parfrey & Knight, 2012;Videvall et al., 2019). While many studies focus on within-individual temporal variation, studies spanning multiple years examining intraspecific variation may yield additional understanding of host characteristics and environmental factors shaping the composition of microbiota.
By characterizing the microbiota of animals in their natural habitats within and across years, we can identify broadscale patterns that may not be identifiable at single sampling periods. This is especially important in studies of wild animals for which annually recurring biological events or seasonally fluctuating environments may have significant impacts on the diversity and abundance of microbiota (Bobbie et al., 2017;Maurice et al., 2015;Skeen et al., 2021).
Approximately 40% of all bird species take advantage of resources that appear in seasonally fluctuating environments by completing a biannual migration. (Alerstam et al., 2003;Dingle & Drake, 2007;Winger, Auteri et al., 2019). To facilitate long-distance movements between breeding and nonbreeding areas, migratory birds may exhibit flexible phenotypes across the annual cycle that enhance migratory performance (McWilliams & Karasov, 2005;Piersma & Van Gils, 2011). Seasonal changes in physiology are well documented within the gastrointestinal tract, where organs including the stomach or gizzard may dramatically change in size to accommodate additional food intake (McWilliams & Karasov, 2001), increase metabolic capacity (McKechnie, 2008), and enhance flight efficiency (Piersma, 1998). Additionally, many species of migratory birds have adapted to seasonal variation in diet, which allows them to exploit periodic abundance of different food sources (Bairlein, 2002). Flexibility in phenotype and diet across the annual cycle aids in optimization of migratory performance (Bauchinger et al., 2005;Gómez et al., 2018;Hedenström, 2008).
Given the pronounced physiological and dietary seasonal changes often exhibited by migratory birds, corresponding shifts in the abundance and composition of gastrointestinal microbiota are expected. Several studies have examined the factors that impact the microbiome of migratory birds, including the effects of variable breeding habitats and migratory period (Grond, Santo Domingo et al., 2019;Li et al., 2021;Turjeman et al., 2020). Composition of microbiota has been shown to vary at different stages of the migratory cycle within the same population of birds, correlating with change in food sources (Skeen et al., 2021;Wu et al., 2018). Movement ecology of the host, including local foraging behaviours and longdistance migratory movements, has also been shown be associated with altered microbial diversity Skeen et al., 2021;Wu et al., 2018). In actively migrating birds, the host-associated microbiota can rapidly acclimate to the local environmental microbial pools, including on stopover sites (Lewis et al., 2016;Zhang et al., 2020). Avian gut microbial communities may include a substantial portion of environmental bacteria ingested through food sources (Bodawatta et al., 2022). These bacteria are probably transient and provide little to no benefit to the host (Kreisinger et al., 2017). As environmental microbial communities vary over time and space (De Gruyter et al., 2020), this may be reflected in avian gut microbiota.
Owing to the dynamic nature of seasonal migration, major questions addressing how host-associated microbiota vary within and across years in migratory birds remain largely unanswered.
Challenges associated with sample collection, such as logistical obstacles to collecting across broad geographic regions with variable environments, are often prohibitive to generating the data sets necessary to assess temporal and geographic variation of host-associated microbiota. Yet these studies are critical for a comprehensive understanding of the flexibility and resilience of an animal's microbiota in response to fluctuating ecological and physiological conditions. In this study, we seek to identify if and how avian microbiota vary across multiple time scales in migratory birds. We characterized the intestinal microbiota of four closely related species of migratory Catharus thrushes during spring and fall migration through a single stopover sight in Chicago, IL USA over a period of 3 years (2017-2019), C. fuscescens (Veery), C. guttatus (Hermit Thrush), C. minimus (Grey-cheeked Thrush) and C.
ustulatus (Swainson's Thrush). For three of these species, we were able to compare the migratory samples to those from individuals collected on the breeding grounds at the beginning of the breeding season during the same years.
The genus Catharus (Family Turdidae, Order Passeriformes) contains 13 species, including five species of migrants to North America and eight nonmigratory species in the Neotropics (Clements & Principe, 2000). The co-occurrence of multiple morphologically K E Y W O R D S annual cycle, avian microbiome, catharus, migration, temporal variation similar migratory species has made this genus central to understanding the effect of migration on microevolution (Ruegg & Smith, 2002;Winker & Pruett, 2006: Delmore et al., 2016Everson et al., 2019), and can in turn provide insight into the effect of migration on the microbiomes. All four focal species in this study co-occur on migration. Their breeding ranges span northern Canada to the southern Appalachians of the United States. Three of the species (C. fuscescens, C. guttatus, and C. ustulatus) have predominantly sympatric breeding grounds in boreal and temperate forests while C. minimus has a more distinct breeding range across the subarctic treeline. Their nonbreeding distributions are largely allopatric, spanning from the southern United States to southern Brazil and northern Argentina (Heckscher et al., 2020;Mack & Yong, 2020).
The four Catharus species in this study primarily consume insects and berries or other fruits, with the proportion of insects or fruit varying throughout the annual cycle (Heckscher et al., 2020;Mack & Yong, 2020;Whitaker et al., 2020). Numerous host traits, including change in diet (Kreisinger et al., 2017;Meena et al., 2014), moult (Giorgio et al., 2018), age (Videvall et al., 2019), and sex (Escallón et al., 2019) have been shown to correlate with variation in bacteria abundance and diversity in other species of birds but remain uninvestigated in Catharus thrushes.
Here, we test the hypothesis that variation in gut microbiota is driven by shared ecological variables more than by host species identity ("shared ecology" hypothesis). To test this hypothesis, we collected luminal contents of the lower intestines from birds in North America during three periods of the annual cycle: spring migration, during which the birds are flying towards the breeding grounds; breeding season, a stationary period during which the birds reproduce; and fall migration, as the birds are migrating from the breeding sites to their overwintering ranges. We tested for consistency of patterns by replicating across 3 years. The microbiota of the lower intestines represents downstream mixing from the previous regions of the gastrointestinal tract and therefore can be used to assess general community composition of gut microbiota of the host (Drovetski et al., 2018;Wilkinson et al., 2016;Yan et al., 2019).
Specifically, the shared ecology hypothesis predicts variation in microbiota to be correlated with change in physiology, diet, and habitat that all species experience throughout the annual cycle. It also predicts similarities in microbial composition will be found across species and differ between seasons, owing to shared ecological variables at stopover sites. For example, all four thrush species flying south through Chicago during fall migration may experience more similar habitats and food resources with other fall migrants than they do to birds flying north on spring migration. In that case, fall birds would have a microbial composition more similar to fall birds of other species than when compared to spring migratory birds, even of the same species. Further, if ecological variables on breeding and nonbreeding ranges have a significant impact on the microbiota, then C. minimus will have a distinct microbial composition from the other three species during fall migration due to its largely allopatric breeding range, and we would observe distinction between the microbiota all four species during spring migration due to their distinct nonbreeding ranges. Alternatively, if the intense physical demands of long-distance flight, compared to the relatively lower requirements of stationary birds, are the primary factor impacting microbiota structure, then spring and fall birds would be more similar to each other than when compared to the summer breeding birds. In our evaluations, we also considered that different bacterial groups may be impacted differentially and hence show different patterns, for example, with some responding to seasonal change, some to age groupings, and some being host-species specific.

| Bird collection
We leveraged ongoing specimen collection occurring for other research purposes to sample the intestinal contents of the study spe- this study are primarily nocturnal migrants (Winker & Pruett, 2006) and because the CBCM conducted daily checks, birds were probably recovered the morning after collision with buildings. Based on that assumption, all individuals included in this study were processed or frozen within 24 h of death. Fall migrants were aged based on skull ossification and categorized as Hatch Year (HY -birds that hatched the previous breeding season and were migrating for the first time) or After Hatch Year (AHY -birds that hatched prior to the previous summer) (Pyle, 1997). Sex was determined from the gonads. In some cases physical damage from the collision prevented age and/or sex determination. A total of 687 individuals were collected throughout the spring and fall migratory periods of 2017-2019, with voucher specimens deposited in The Field Museum (Table S1).
The intestinal contents from birds on their breeding grounds (n = 60 total) were sampled for C. fuscescens, C. guttatus, and C.  (Table S1). Field collection was approved by the University of Michigan Institutional Animal Care and Use Committee and all local, state, and federal permitting authorities (see Acknowledgements).

| Intestinal sample collection
We collected the luminal contents of the lower intestine and stored them on Flinders Technology Associates cards (FTA cards; GE Whatman). Previous studies have shown that results from FTA cards are comparable to those resulting from long term ultra-cold storage (Song et al., 2016;Wang et al., 2018). We used sterilized instruments to detach the lower intestines from the cloaca. We then expressed the contents of 4-8 cm of the posterior end of the lower intestines.
We noted food materials visible in the luminal contents, such as seed or fruit. We transferred the sample to the FTA Cards using a sterile swab. We air dried the FTA cards and stored them in airtight containers with desiccants. The spring and fall migrant specimens are housed at The Field Museum.

| DNA isolation and sequencing
We transferred approximately 1 cm 2 of the FTA cards to extraction plates. We randomized samples across extraction plates so that plates included samples from all species, seasons, and years, to ensure potential differences in microbial composition were not due to laboratory work bias. Following the manufacturer's extraction protocol, we used the Qiagen DNeasy PowerSoil kit (Qiagen). We included 16 negative controls, two per extraction plate, which included no sample or sample preservation materials, for quality control and to account for possible contamination during extraction and PCR. We used the Earth Microbiome Project universal primers 515F/806R to amplify the V4 region of the 16S rRNA genetic marker (Caporaso et al., 2011(Caporaso et al., , 2012. We then used the Illumina MiSeq Platform to obtain paired-end 150 base pair reads (Kozich et al., 2013). We used four sequencing lanes and loaded 188 samples and four controls per lane. Subsampling and DNA isolation took place in the Pritzker Laboratory at The Field Museum using a specialized fume hood to reduce possible contamination. All subsequent sample processing and sequencing took place at the Argonne National Laboratory (Lemont, Illinois, USA).

| Sequence processing
We processed raw sequence data with the program quantitative insights into microbial ecology [QIIME2] version 2021.4 (Bolyen et al., 2019). Following standard demultiplexing and quality filtering, we generated amplicon sequence variants (ASVs) using divisive amplicon denoising algorithm (DADA2; Callahan et al., 2016). We classified ASV taxonomies using the SILVA reference database (version 132; Quast et al., 2012). After classification we removed all ASVs identified as chloroplasts and mitochondria. We aligned sequencing using MAFFT and then built a phylogenetic hypothesis for all bacterial sequences using FastTree (Katoh & Standley, 2013;Price et al., 2010). Reads that did not align to any known bacterial phylum were blasted to confirm their nonbacterial sources and removed from the final data set. We identified bacterial contaminants with the R package decontam using the prevalence-based contaminant determination (Davis et al., 2018). We used the 16 extraction blanks that were processed in parallel with the other samples as controls. Using a threshold of 0.5 (from a possible range of 0 to 1), decontam identified 120 contaminant ASVs that were found in a higher fraction of negative controls than in Catharus samples. We subsequently removed these ASVs from all libraries ( Figure S1, Table S2).

| Investigation of potential batch effects
To ensure that biases were not introduced during sample collection or processing, we compared alpha and beta diversity measures across three possible batch effect categories: sample collector (four people), extraction plate (eight plates), and if samples were taken from fresh birds or those that had been frozen prior to processing.
Comparisons between collectors and between fresh versus frozen birds were conducted within the same year so that variation between years did not confound the assessment of batch effects.
Samples were randomized across extraction plates so quality control measures were analysed across the full data set. No significant differences were observed with any of the potential batch effect categories (Table S3).

| Normalization of microbial data
Following sequence processing, we analysed libraries using the R package phyloseq (McMurdie & Holmes, 2013). Sequenced libraries were substantially variable in size (as detailed in Results).
Therefore, we rarefied libraries at two depths, 500 reads and 5000 reads, to ensure that we retained as many libraries as possible to accurately captured bacterial diversity while assessing the robustness of our results (Cameron et al., 2021;Weiss et al., 2017). This resulted in the removal of 67 and 279 libraries out of 747 libraries, respectively. The majority of results were consistent across analyses at both levels of normalization. We discuss the results of the libraries normalized at 500 reads and note when results differ at 5000 reads.

| Alpha diversity
We estimated alpha diversity of rarefied libraries using both richness and the Shannon diversity Index. The Shannon diversity index was approximately normally distributed but we log transformed the observed richness measures to meet assumptions of normality. Due to the low level of shared ASVs across individuals (see species-specific common microbes results) and possible functional redundancies (Li et al., 2021;Shade, 2017), we did not conduct alpha diversity analyses at the ASV level but did so at every other taxonomic level. We conducted ANOVAs ("aov" function in the stats R package) with post hoc comparisons using Tukey's HSD test. We tested for and found no significant interaction between year and season, so all variables were modelled as independent factors. These variables include year (2017,2018,2019), season (Spring, Summer, Fall), and species (C. fuscescens, C. guttatus, C. minimus, and C. ustulatus). We also compared alpha diversity of host sex (male or female) and age (HY or AHY) independently on reduced data sets, omitting samples where the host metadata was unable to be obtained and, in the case of age, only on fall birds as all spring birds are considered AHY. For age and sex variables we conducted a Kruskal-Wallis test to use as a nonparametric pairwise comparison of alpha diversity measures.

| Beta diversity
We compared beta-diversity between years, seasons, and host species separately, using the Bray-Curtis dissimilarity and weighted UniFrac metrics (Beals, 1984;Lozupone et al., 2011). We also compared beta-diversity between species in spring migratory birds, between species in fall migratory birds, and between years within each season (spring migration, summer breeding, and fall migration). We visualized the resulting using nonmetric multidimensional scaling (nMDS) of weighted UniFrac distances setting the number of dimensions to four. We determined significance using analysis of similarities (ANOSIM) with 9999 permutations (Clarke, 1993). The R test statistic derived from the ANOSIM test compares the mean of ranked dissimilarities between and within groups. R values closer to 1.0 reflect increased levels of dissimilarity between groups while R values close to 0 reflect a distribution of ranks that is similar within each group. A significance level of p < .05 was applied to test the null hypothesis of no differences between microbial communities of different categories. We conducted similar analyses for sex and age, on reduced data sets.

| Differential abundance
To identify genus and phylum level taxa which differ in abundance across years, seasons, and host species, we used the ANCOM-BC method (analysis of composition of microbiomes with bias correction; Lin & Peddada, 2020). ANCOM-BC estimates changes between groups using the log-transformed values of absolute sequence counts; therefore we used all unrarefied libraries of at least 500 reads. This method accounts for the compositional nature of microbiome data by using a linear regression framework to estimate and eliminate bias introduced by differences among sampling fractions, while controlling false discovery rate. We set a significance cutoff of p adj < .05 with a Bonferroni correction.

| Species-specific common microbes
We quantified microbial profiles common to Catharus and within each species as microbial ASVs and genera recovered from >50% of all individuals (Grond et al., 2017;Risely, 2020). We quantified year and season specific lineages as being present in >50% of all individuals within each subset. We analysed shared microbes at the ASV and genus level using unrarefied libraries of at least 500 reads using the microbiome R package (Lahti & Shetty, 2018). Additionally, we tested for shared ASVs at lower prevalence within the full data set to determine if and in what proportion the majority of ASVs become common across all individuals.  Table 1 includes a breakdown of the samples by species, sampling period, age, and sex.
Planctomycetes (22%, Phylum Planctomycetes) was the most abundant class, followed by Cyanobacteria (18%, Phylum Cyanobacteria) and Alphaproteobacteria (16%, Phylum Proteobacteria). As discussed below, the abundance of Planctomycetota and Cyanobacteria recovered in this study is high, relative to previously published research (Ambrosini et al., 2019;Dewar et al., 2014;Hird et al., 2015;Trevelline et al., 2020). Exploratory plots illustrating relative abundance of phyla suggest variation by host species, season, and year (Figure 2, Figure S5) which we explored more formally in the next section.

| Differential abundance
We determined bacterial genera and phyla that exhibited significant variation in abundance between host species, seasons and years using the ANCOM-BC method on unrarefied data sets of at least 500 TA B L E 1 Breakdown of samples by species (C. fuscescens, C. guttatus, C. minimus, C. ustulatus), sampling Season (Spring, Summer, Fall), Year (2017 Table S7). At the level of genus, we identified 28 bacterial genera that were significantly enriched in specific years ( Figure S6).
This includes Aliterella (Phylum Cyanobacteria) as significantly more abundant in 2019 than in 2017 or 2018.
When comparing between seasons, twelve phyla exhibited significant variation between seasons ( Figure S7, Table S8).
Myxococcota and Dependentiae had highest relative abundance in the summer, Proteobacteria and Campilobacterota in the fall, and Plactomycetota and Fibrobacterota in both the spring and the fall.
At the level of genera, our ANCOM-BC analyses identified 45 genera to be differentially abundant across seasons ( Figure S7, Table S8).
Several genera containing common pathogenic microbes were significantly enriched in specific sampling periods, such as Escherichia-Shigella in the fall, Neochlamydia in the spring and Diplorickettsiaceae in the summer.

| Shared microbial profiles
Eleven genera and three ASVs within those genera were identified as present in more than 50% of all libraries ( and an unnamed genus in the family Geminicoccaceae (Phylum F I G U R E 2 Relative abundance box plot of most abundant phyla with libraries rarefied to 500 reads, representing the variation seen in relative abundance species, seasons, and year. Individual points represent the relative abundance of each phyla per individual bird. Colours of box plots correspond to host species. Proteobacteria). Notably, Aliterella was the most common genus in the data set, found in 77% of individuals. This prevalence was driven by a single ASV, whose species identity has not yet been described.

| Alpha diversity
Consistently, across both levels of rarefaction, all taxonomic levels, and both diversity metrics, year and season showed significant differences in alpha diversity ( Table 2, Table S10). The differences across  Figure S8). There were no significant differences, at either level of rarefaction, any taxonomic level, or diversity metric between host species, sex, or age, with the exception of a comparison between hatch year and after hatch year birds at the phylum level (SD: p = .02, Kruskal-Wallis test). Older birds showed slightly elevated alpha diversity compared to younger birds. There were no significant differences in pairwise comparisons between species for either alpha diversity metric (Table 2).

| Beta diversity
Community-level analysis revealed sharp distinctions in the beta diversity of gut microbiota in birds between years, when comparing across the full data set and within specific seasons, with both Bray-Curtis dissimilarity (R = 0.371, p < .001) and weighted UniFrac distances (R = 0.311, p < .001) ( Figure 5, Table 3, Table S11). Comparisons of host species also revealed significant shifts in microbial composition, however low global R values indicate that this significance may be due to dispersion of samples, rather than true differences in community composition of microbes (Chapman & Underwood, 1999) (BC: R = 0.032, p = .003; WU: R = 0.047, p < .001). Visual inspection of the ordination plot shows no clear clustering by species ( Figure 5).

Log fold change
The results of the ANOSIM analysis of microbial beta diversity in spring migratory birds as well as in fall migratory birds show low support for distinct variation in microbial community composition between Catharus species for either migratory period (Table S11C).

| DISCUSS ION
Our results highlight that the microbiome is dynamic over time, with both year and season significantly impacting the overall composition of thrush microbiota. We find that temporal variation over years and seasons has a more observable impact on the diversity and composition of microbiota than host species, age, or sex. Migratory birds have evolved numerous physiological adaptations that enable them to complete long distance flights (Battley et al., 2000;Bauchinger et al., 2005;Piersma, 1998). These adaptations, as well as processes associated with migration itself, may impact host-associated microbiota (Hedenström, 2008;Song et al., 2020

| Community composition
The high-level composition of Catharus intestinal microbiota is generally similar to that previously reported in numerous species of birds, with Proteobacteria, Actinobacteroita and Firmicutes representing a large portion of the overall composition (Ambrosini et al., 2019;Dewar et al., 2014;Hird et al., 2015;Trevelline et al., 2020). However, unlike in previous studies, Planctomycetota and Cyanobacteria represent a substantial portion of the overall microbiota in our sampling.
Additionally, Bacteroidota, found in relatively higher abundances in avian microbial studies which used faecal matter or cloacal swabs, was often absent or in low abundance in the intestinal samples used in this study (Hird et al., 2014;Turjeman et al., 2020;Videvall et al., 2019).
The low relative abundance of Bacteroidota reported here, though inconsistent with several previous surveys of bird microbiota, may be a true characteristic of migratory thrushes and not an artefact of sample type, as an analysis of C. ustulatus faecal microbiota on stopover in Louisiana reported similarly low abundances (Ambrosini et al., 2019;Dietz et al., 2020;Grond et al., 2017;Lewis et al., 2016).

| Migration
Migration is a physically taxing endeavour which may increase pathogen susceptibility through decreased immune function due to the stress of migration or exposure to novel pathogen pools at stopover sites (Altizer et al., 2011;Owen & Moore, 2008). Additionally, seasonal variation in immune capacity, where stationary birds may allocate more resources to mounting an immune response than those on migration, can lead to increased pathogen susceptibility during nonbreeding relative to breeding periods (Altizer et al., 2011;Valdebenito et al., 2021). Though host associated microbiota has been shown to be variable throughout the breeding season (Escallón et al., 2019), there is evidence that cloacal microbial pathogen prevalence is lower in the breeding period than during nonbreeding periods (Poiani, 2010). While pathogenicity was not directly assessed in this study, we observed increased relative abundance of several bacterial genera that contain known pathogen strains in fall or spring birds when compared to summer birds. In particular, Neochlamydia, Esherichia-Shigella and Coxiella were all significantly enriched in actively migrating birds during either spring or fall migration. Bacterial genera which may include well known pathogenic species are generally not composed solely of disease-causing strains.
For example, most Yersinia in migratory birds have been identified as nonpathogenic (Niskanen et al., 2003). One study of migratory passerines on stopover observed an increased abundance of bacterial genera which contain potentially pathogenic strains, but found no evidence of illness within the host, suggesting the genera may actually act more as commensals, possibly providing some type of benefit to the host (Lewis TA B L E 2 Results of analyses of alpha diversity values for natural log of observed amplicon sequence variants (ASV) richness and Shannon diversity index compared across bacterial genera of libraries rarefied at 500 reads.  In previous studies, increased abundance of genus Corynebacterium has been correlated with migration, as it has been found in heightened levels in three species of migratory birds compared to closely related, nonmigratory conspecifics Risely et al., 2017). It has been hypothesized that Corynebacterium may enable increased fat deposition or may be associated with an immune response brought on by the stress of migration (Risely et al., 2017;Zhang et al., 2021). In this study, this genus appears in less than 20% of the individuals in this study, and we found no significant enrichment of Corynebacterium abundance in actively migrating birds compared to those on the breeding grounds, suggesting the role of Corynebacterium may be variable across host species and not a major factor in the four species of Catharus studied here.
Our results suggest that actively migrating birds may have reduced microbial diversity compared to birds during a stationary period of the annual cycle. Generally, there were weak differences between the four species of thrush in this study during spring and fall migration, which became slightly more pronounced when compared to birds on the breeding grounds. Additionally, summer birds consistently exhibited higher alpha diversity on the breeding grounds compared to fall or spring birds. Phenotypic flexibility associated with migration induces numerous changes to the birds' digestive system, including atrophication of the intestinal tract (McWilliams & Karasov, 2001;Piersma & Gill Jr, 1998). These changes may reduce the diversity of resident gut microbiota and promote increased presence of bacteria from the local environmental pool, as suggested in a study of migratory passerines on stopover after crossing the Gulf of Mexico (Lewis et al., 2016). This causes different species of birds co-occurring at the same stopover sites to exhibit similar composition of the gut microbiota. Our results are consistent with these previous observations and support hypotheses that the migration process limits intestinal microbiome diversity and homogenizes intestinal microbiota across species. This study is the first to find this pattern to be consistent across both spring and fall migratory periods across multiple years.

| Host species
Overall, our results indicate weak differences in the overall community structure between species. Few bacterial phyla or genera were significantly more abundant in any of the four sampled species of migratory thrushes, relative to another. Additionally, no distinct variation between host species microbial community composition were observed, in the full data set or within specific migratory periods. The migratory species within Catharus overwinter in minimally overlapping ranges. Should nonbreeding habitat have a sustaining impact on microbiota we would expect a significant difference between the species observed in spring migratory birds. Similarly, we would expect C.
minimus to have a distinct microbiota in the fall migratory period, as they breed largely in allopatry to the other three species, which have more overlapping breeding ranges. The results from the differential abundance analyses as well as the insignificant variation in community composition between species imply that the four species of migratory thrushes do not exhibit species-specific microbial profiles due to differing physiologies or ecologies. Previous research has indicated that environment and diet are more influential than host genetics in shaping avian gut microbiota (Grond et al., 2018;Song et al., 2020) and that host taxonomy plays a weakly significant role compared to abiotic factors (Capunitan et al., 2020;Hird et al., 2015). Our results support this and further suggest that a bird's microbiota reflects recent environment, such as stopover sites, with little carryover from breeding or nonbreeding ranges evident when birds are actively migrating.

| Moult
Several components of the microbiome may be directly tied to host processes and characteristics, including the annual feather moult.
Moult occurs when old feathers are shed and replaced by new feathers, which has been suggested as an adaptation to microbial control (Burtt Jr & Ichida, 1999;Giorgio et al., 2018). Microbes found on feathers may be transferred to the gut, most plausibly due to incidental ingestion during preening of feathers premoult. Bacillus is a genus which includes feather-degrading bacteria found naturally occurring on many species of birds and may play a role in the timing of the annual moult birds undergo as part of the annual cycle (Gunderson, 2008). The species of thrushes studied here all moult prior to, or at the beginning of, fall migration (Cherry, 1985;Pyle, 1997). We found Bacillus to be significantly more abundant in the summer birds than spring or fall birds.
The enrichment of Bacillus in summer is consistent with the model of Gunderson (2008) where, preceding moult, birds show high levels of Bacillus which are then reduced through the moulting process.

| Age
Changes in microbial diversity and community structure between adults and chicks has been well documented (Grond et al., 2017;Kreisinger et al., 2017;Videvall et al., 2019). In contrast, comparisons between age classes of adult wild birds are relatively few. In one previous study of Tree Swallows (Tachycineta bicolor), the microbiota of females of this species was assessed during the breeding season revealing that older birds had significantly higher diversity than birds in their first breeding season, possibly due to increased opportunities for mating and therefore increased contact with other birds (Hernandez et al., 2021).
We observed a slight, although not significant, increase in alpha diversity in the after hatch year fall migrants compared to the hatch year fall migrants. The increased diversity we observed in the older birds may be due to increased contact with other birds during the mating season. The increased diversity may also be a result of the older birds foraging far from the nest while rearing the hatchlings, leading to more exposure to local environments, which has been shown to increase microbial diversity .

| Diet
Variation in diet is known to influence the microbiome (Grond, Perreau et al., 2019;Li et al., 2021;Song et al., 2020). Many species of birds consume different food sources throughout the annual cycle. For example, C. ustulatus consume more insects than fruit during spring migration and breeding seasons but tend to consume more fruit during fall migration (Parrish, 1997). In general, frugivory in migrants is more prevalent in fall than in spring (Bairlein, 2002).
However, no bacteria known to aid in the digestion of fruit materials, such as those associated with complex carbohydrate degradation, were identified as more abundant in the fall or any other period of this study. Rather, Paenibacillus, a genus which contains several chitinolytic bacteria, was significantly more abundant in fall birds and is consistent with an insect-rich diet (Meena et al., 2014). Whether rates of frugivory in the fall are decreasing in these thrush species, or whether frugivory has little impact on the gut microbiome of the thrushes, is an open question that may be addressed by future observational studies of diet and its impact on microbiota.

| Environmental effect
Annual differences in climate can affect the composition and turnover of environmental microbes (Averill et al., 2019;De Gruyter et al., 2020;Guo et al., 2018). A study of zebra finches (Taeniopygia guttata) demonstrated that birds may acquire as much as 25% of the gut microbiota from environmental sources, driving some of the variation observed between years (Chen et al., 2020).  Lewis et al., 2017). Community composition of thrush microbiota within seasons and years was more similar than microbiota of thrushes from different seasons or years. This may suggest environmental influences from stopover sites more strongly influenced the observed thrush microbial communities, rather than longterm carryover from breeding or nonbreeding areas.

| Shared microbial profile
Although some surveys of avian microbiota identify shared taxonomic units of up to 50% , the majority of studies report a much lower percentage of ASVs recovered as shared across the majority of samples in the data set (Escallón et al., 2019;Grond, Santo Domingo et al., 2019;Jose et al., 2021). In thrushes, only three of the 26,895 total ASVs were found in at least 50% of all individuals, which is exceptionally low. Those three ASVs as well as the 11 shared microbial genera have no described functions known to be associated with host processes within the bird, such as facilitating nutrient uptake or breakdown of food materials. Additionally, common intestinal flora, such as Faecalbacterium, are reported as core microbes in many host species (Grond, Santo Domingo et al., 2019;Escallón et al., 2019;Skeen et al., 2021). The shared microbes across and within Catharus species contained no common intestinal flora.
The functions of the shared genera of Catharus are generally unknown. Functional characterization of the microbiome provides a complementary view of variation in microbiota between and within groups (Cadotte et al., 2011;Escalas et al., 2019). A study of migratory sympatric overwintering birds revealed that gut microbiota functions are more conserved than bacterial diversity structure, indicating that different bacteria function in similar ways (Li et al., 2021). Therefore, although the thrushes in this study share exceptionally few ASVs across all individuals or within specific subsets, the inferred functional composition of the microbiota may reveal a more similar structure across individuals and can lead to further exploration into the impact of gut bacteria on migratory birds.

| CON CLUS ION
This study adds to a growing body of literature demonstrating that the diversity and community structure of host-associated microbiota of many, but not all, migratory bird species significantly varies throughout the annual cycle (Risely et al., 2018;Skeen et al., 2021;Wu et al., 2018). Additionally, we advocate for the necessity of interpreting results within the context of the time period from which the samples were collected. Here, we characterized Catharus intestinal microbiota from spring and fall migratory birds as well as on their breeding grounds. Surprisingly our annual replicates revealed that between-year variation was significantly higher than across seasons.
Additionally, we describe weak host-species specific impacts on the composition and diversity of the microbiota and identify correlations between specific host processes and the microbiome. Finally, we note that the physiological changes associated with migration may have important effects on microbiota and further research is needed in this area.
One challenge of studying wild birds under natural conditions is untangling the large number of uncontrolled variables that can influence host microbial communities. By characterizing the microbiome of four closely related Catharus thrushes at three separate portions of the annual cycle replicated over 3 years, we are able to identify components of the microbiome that vary geographically and temporally, including specific bacterial taxa and overall community composition. We highlight the necessity of temporal sampling of species to gain a fuller understanding of how the microbiome can vary over time and to better identify specific components of the microbiome that are likely to be associated with physiological processes affecting host ecology, evolution, and conservation.

AUTH O R CO NTR I B UTI O N S
Heather R. Skeen designed the project with input from Shannon J.

FU N D I N G I N FO R M ATI O N
None.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The code used in this study has been made available at https:// github.com/skeen hr/catha rus_micro biome. All sequence data generated for this study are available at the NCBI Sequence Read