2.1 Sampling

Three sediment and water samples and four snails for gastrointestinal microbiome samples were collected from each sampling site in Denmark (DK), Estonia (EE), and Germany (DE) (Table 1 and Figure 1). The sampling stations in each country were closer to each other than to any sampling station in the other countries. Sediment was collected in sterile 2 mL tubes by carefully scraping the top 1 cm of the sediment (macrophytes or litter were removed if present). Water was collected using 100 mL syringes and pumped through Sterivex GP 0.2 µm filters on site. Sediment and water filters were stored cooled (4°C) for transport (2–3 h) and were kept frozen at −20°C until DNA extraction. Snails were placed into tap water for 24 h to reduce the number of food-derived bacteria in the microbiome and frozen at −20°C until dissection. A. balthica samples included average snails of age, size, and health status. Before dissecting the gastrointestinal tract, the shell was treated with 70% ethanol to avoid external contamination. The dissection process was performed under sterile conditions on Petri dishes. The snail's soft body was carefully removed from the shell with tweezers, and the gastrointestinal tract was isolated from the rest of the body.

2.2 DNA extraction

Different DNA extraction protocols were used to optimize the yield of genomic DNA from each sample. For DNA extraction, the frozen snails were melted for 10 min at room temperature, and after dissection, DNA extraction was performed based on a modified protocol from Lueders et al. (2004) and Weinbauer et al. (2002) as described in Kivistik et al. (2020). In brief, gastrointestinal tracts were incubated at 65°C for 1 h and bead-beaten for 3 min at 2000 rpm. From the supernatant, DNA was extracted using a mixture of phenol:chloroform:isoamyl alcohol (25:24:1). The DNA extract was incubated with RNase (100 mg/mL) (Qiagen), washed with isopropanol and with 95% ethanol. The pellet was resuspended in 50 μL AE buffer (Qiagen). For water bacterial community analysis, complete filters were removed from the Sterivex capsules and cut into six pieces under sterile conditions. DNA extraction was performed by using the DNeasy PowerWater Kit (Qiagen) following the producer's instructions. Sediment (250 mg) was transferred into bead-beating columns provided in the PowerSoil Pro Kit (Qiagen) and DNA was extracted according to the manufacturer's protocol.

2.3 Bacterial community analysis

For bacterial community analysis, the DNA was polymerase chain reaction (PCR) amplified using V3–V4 primer amplifying the 16S ribosomal RNA (rRNA) gene according to the protocol of Herlemann et al. (2011) and Kivistik et al. (2020). The amplicons were purified using PCR Kleen (Bio-Rad), tags were added, and sent to FIMM, University of Helsinki, Finland for Illumina (MiSeq) sequencing using PE250 chemistry.

2.4 Phylogenetic analysis of A. balthica

For the phylogenetic assignment of the snails, the internal transcribed spacer 2 (ITS2) was analyzed. PCR on genomic DNA from snails was performed according to a protocol from (Vinarski et al., 2011) using the primers LT1 (Bargues et al., 2001) and ITS2-Rixo (Almeyda-Artigas et al., 2000). The temperature profile used was 94°C 4 min, 35 × (94°C 30 s, 56°C 30 s, 72°C 1 min), 72°C 7 min, 8°C store. The amplicons were purified using PCR Kleen (Bio-Rad) and Sanger sequenced by the sequencing facility of the University of Tartu, Estonia.

2.5 Bioinformatic processing of the sequences

The raw sequence data from Illumina MiSeq sequencing were processed with the QIIME2 microbiome analysis package (Bolyen et al., 2019) using default settings. Quality filtering and Chimaera identification were performed using the DADA2 plugin for QIIME2 (Callahan et al., 2016) with SILVA 138.1 as the database. Sequences assigned to chloroplasts, mitochondria, eukaryotes, and Archaea were excluded because the primer set employed in the analysis had very limited coverage of these groups. Samples with insufficient amplicon products for sequencing, lower quality sequencing results, and samples with less than 1000 reads after removal of chloroplasts, mitochondria, eukaryotes, and Archaea were excluded from the analysis. This resulted in a total of 3,301,247 reads assigned to 4787 amplicon sequencing variants (ASVs) for 63 sediment samples, 62 water samples, and 88 snail samples. The raw reads were deposited at the NCBI in bioproject PRJNA1028832 under the accession numbers SAMN38726042–SAMN38726255.

Water and gastrointestinal samples from Ratzeburg and Lütau did not yield sufficient high-quality reads and were therefore excluded from the analysis. The sequence names of SILVA 138.1 were changed according to the Genome Taxonomy Database (Parks et al., 2013). Verrucomicrobia were named Verrucomicrobiota, Fusobacteria were named Fusobacteriota, Bacteroidetes were named Bacteroidota, Cyanobacteria were named Cyanobacteriota, Actinobacteria were named Actinomycetota, Comamonadaceae were named Burkholderiaceae. Firmicutes were named Bacillota, Chloroflexi were named Chloroflexota, Rhodobacteriaceae were named Rhodobacteraceae, Dechloromonas were named Azonexus, and Tychonema were named Microcoleaceae.

The Sanger sequences from the ITS2 region of the snail specimen were quality-checked using Chromas software, merged using Bioedit (version 7.7.1.0), and submitted to NCBI GenBank under the accession numbers OR288013–OR288075 (Supporting Information Figure 1).

Sequences from this study and selected reference sequences from Kivistik et al. (2022a) were imported to the ARB Program suite (Ludwig, 2004) to calculate a maximum likelihood phylogenetic tree (PhyML). The topology of the tree was tested separately by neighbor-joining and parsimony analysis (DNAPARS) with a bootstrapping algorithm (seqboot; 1000 bootstraps). The original sequence definitions in the GenBank database were replaced with a consistent nomenclature including accession number, species, and geographic origin.

2.6 Statistical analysis

The PAST software package version 3.22 (Hammer et al., 2001) was used for statistical analysis. Samples for beta diversity analysis were normalized using the Centered log-ratio (CLR) transformation. CLR values are scale-invariant such that the same ratio is obtained regardless of differences in read counts (Gloor et al., 2017). For Chao1 estimations, Explicet (Robertson et al., 2013) was used, which performs a rarefaction-based analysis through bootstrapping. A Kruskal–Wallis test and a post hoc Dunn's pairwise test were used to calculate significant differences between the numbers of ASVs in the samples. Differences in the bacterial community composition were tested by permutational multivariate analysis of variance (PERMANOVA) combined with a pairwise PERMANOVA test between all pairs of groups as a post hoc test corrected by the sequential Bonferroni method. The similarity between bacterial communities was visualized by principal coordinates analysis (PCA, based on Euclidean distance) in combination with CLR-transformed data at an Aitchison distance. The relative overlap of the ASV in the specific substrates and countries was calculated using a multiple list comparator (https://molbiotools.com/) and visualized as Venn diagrams. To test for correlations between community similarity and geographic distances between samples, the data were split by substrate. The Dice/Sørenson index was calculated for the bacterial community distance in PAST, and the geographic distance between sampling points was estimated using the spherical law of cosines formula, which accounts for the spherical nature of the Earth. The distance-decay rate was assessed by slope of linear fitting (lm() in R) between Dice/Sørenson index and geographical distance, with statistically significant difference between slopes as tested by analysis of covariance (lstrends() in R). In addition, we tested the correlation between Dice/Sørenson index and geographical distances matrices using a Mantel test (mantel.rtest() in R).

3.1 Bacterial community composition

On a broad phylogenetic level (Figure 2a), abundant taxa were present in snail gastrointestinal samples, sediment, and water. However, abundances differed between the substrates and between the different countries. In the gastrointestinal tract, Gammaproteobacteria (31%–34%) and Alphaproteobacteria (14%–22%) were the most abundant classes, followed by Planctomycetota (11%–17%) and Verrucomicrobiota (6%–11%). Fusobacteriota had the highest relative abundance in samples from Germany (5%) and the lowest in samples from Denmark (1%). In the sediment samples Gammaproteobacteria (18%–21%), Alphaproteobacteria (11%–17%), Bacteroidota (10%–12%), and Cyanobacteriota (0%–17%) were also the most abundant. These abundant phyla were irregularly distributed between countries, including Actinomycetota (9%–16%) and Bacillota (2%–11%). Cyanobacteriota had a high abundance in the samples from Estonia (17%) and were missing in Denmark (0%). Except for the Gammaproteobacteria (19%–21%), the abundant phyla/classes in the water samples were also irregularly distributed, including Actinomycetota (18%–31%), Bacteroidota (9%–27%), and Alphaproteobacteria (8%–17%). Bacteroidota had a high abundance in Estonia (27%) but low in Germany (9%). In addition, at the finest taxonomic level (ASV), a clear difference in the abundance of ASVs between the three substrates and countries was obvious (Figure 2b). Unclassified Burkholderiaceae were most abundant in snails, with the lowest abundances being in Denmark. The genus Flavobacterium was almost absent in snails but highly abundant in water samples, especially in Estonia. Unclassified Microbacteriaceae and Candidatus Nanopelagicus were only found in the water samples, whereas the highest abundances of Candidatus Nanopelagicus were found in Germany and unclassified Microbacteriaceae in Estonia. Unclassified ASVs of the genera Pirellula, Luteolibacter, Cutibacterium, and Candidatus Saccharimonas were found in 90% of the A. balthica samples, of which Pirellula and Luteolibacter were abundant (Figure 2b).

Accordingly, the PCA of all samples showed a separation of the bacterial community depending on the substrate, that is, sediment, water, and gastrointestinal tract (Figure 3). PERMANOVA analysis supported a significant difference between water versus sediment versus gastrointestinal tract-associated bacterial communities (Supporting Information Table 1) and pairwise PERMANOVA between all samples indicated p < 0.01 for all comparisons. Therefore, the three substrates were analyzed separately in the following analysis. Comparison of the separated gastrointestinal, sediment, and water bacterial community composition between countries showed a significant difference between the country-specific bacterial communities (Figure 4 and Supporting Information Tables 2 and 4–6) also in the pairwise comparison. The bacterial communities within a substrate varied between the sampling stations of a country (Supporting Information Table 3). However, the more conservative pairwise comparison did not reveal significant differences between the sites of a country (Supporting Information Tables 7–15), suggesting a greater difference in the bacterial community composition between the countries compared with that within countries.

To explore biogeographic differences in microbial communities along the longitudinal gradient, community similarity was plotted as a function of geographic distance between sampling points (Supporting Information Figure 2). The geographic distance was, on average, 625 km, between the farthest sampling points was 1245 km and the smallest distance was 2 km. The slopes between the host-associated and free-living bacterial communities differed significantly and gastrointestinal tract bacterial community distance had the lowest r value in the Mantel test (Supporting Information Tables 16 and 17). The turnover rate (i.e., the rate at which species composition changes with geographic distance (Soininen et al., 2007)) was the highest for water samples (0.051 per 1000 km) and sediment (0.027 per 1000 km) and smallest for snail gastrointestinal microbiome samples (0.013 per 1000 km). However, the distance was not continuous, and samples between 300 and 900 km were missing (Supporting Information Figure 2).

3.2 Number of bacteria within one substrate/country

The average Chao1 diversity of the gastrointestinal microbiome was significantly lower than the average sediment and the average water Chao1 diversity except for water from Germany (Supporting Information Table 17 and Figure 5). The water bacterial Chao1 richness from Estonia and Germany differed significantly from that of sediment. To differentiate substrate/country-specific ASVs, Venn diagrams were prepared. There was a significant difference between the relative abundance of ASVs in Denmark, Estonia, and Germany within the water, snail gastrointestinal tract, and sediment. Also, a significant distance existed between the gastrointestinal tract bacterial community, water, and sediment in Denmark, Estonia, and Germany.

The snail gastrointestinal, water and sediment bacterial communities shared 30%, 29%, and 36% of ASVs between the three countries (Figure 6a–c). Within a country, the snail gastrointestinal tract bacterial community composition always had the fewest specific ASVs (Denmark 7%, Estonia 6%, and Germany 10%), whereas sediment always had the highest percentage of specific ASVs (Denmark 32%, Estonia 27%, and Germany 35%) (Figure 6d–f). In addition, the overlap of ASV between the substrates in one country was relatively low (Denmark 9%, Estonia 14%, and Germany 10%). The highest number of ASVs overlapping with the gastrointestinal bacterial community was the water bacterial community (Denmark 15%, Estonia 13%, and Germany 15%). In addition to the difference between country and substrate, we observed differences in the bacterial communities depending on whether the sample was taken in flowing water (river) or standing water (lake, pond) in all samples (PERMANOVA, F = 3.23, p < 0.01). PCA and statistical tests between standing and flowing water of the gastrointestinal tract (PERMANOVA, F = 2.1, p < 0.01), sediment (PERMANOVA, F = 2.2, p < 0.01), and water bacterial community (PERMANOVA, F = 4.1, p < 0.01) supported this result (Supporting Information Figure 3). However, within a water type (flowing or standing water) the country and substrate bacterial community composition were significantly different (standing water, two-way PERMANOVA country F = 3.18 p < 0.01; substrate F = 5.17 p < 0.01; flowing water, two-way PERMANOVA country F = 4.13 p < 0.01; substrate F = 15.2 p < 0.01). Because the two environments were relatively evenly distributed in the countries (Table 1), the effect of flowing and standing water did not influence the country-dependent analysis.

3.3 Host identification through phylogenetic marker genes

To support the morphological identification and exclude cryptic species in the different geographic regions of A. balthica, we randomly analyzed the ITS2 biomarker regions of 62 snail samples. Our analysis confirmed that the snails sampled in Germany, Denmark, and Estonia belong to A. balthica with highly identical ITS sequences (99% identity, Supporting Information Figure 1) and a comparable genetic host basis can be expected.