Intrahabitat Differences in Bacterial Communities Associated with Corbicula fluminea in the Large Shallow Eutrophic Lake Taihu

ABSTRACT

The Asian clam Corbicula fluminea is a keystone zoobenthos in freshwater ecosystems. However, its associated microbiome is not well understood. We investigated the bacterial communities of this clam and its surrounding environment, including sediment and water simultaneously, in a large lake by means of 16S rRNA gene sequencing. Approximately two-thirds of the bacterial operational taxonomic units (OTUs) associated with clams were observed in the surrounding environment and mostly from particle-associated samples. The associated bacterial communities were site specific and more similar to environmental bacteria from the same site than those at other sites, suggesting a local environmental influence on host bacteria. However, the significant differences in bacterial diversities and compositions between the clam and the environment also indicated strong host selection pressure on bacteria from the surrounding environment. Bacteria affiliated with Firmicutes, Spirochaetes, Tenericutes, Bacteroidetes, Epsilonbacteraeota, Patescibacteria, and Fusobacteria were found to be significantly enriched in the clams in comparison to their local environment. Oligotyping analyses of the core-associated bacterial OTUs also demonstrated that most of the core OTUs had lower relative abundances and occurrence frequencies in environmental samples. The core bacterial OTUs were found to play an important role in maintaining the stability of the bacterial community network. These core bacteria included the two most abundant taxa Romboutsia and Paraclostridium with the potential function of fermenting polysaccharides for assisting host clams in food digestion. Overall, we demonstrate that clam-associated bacteria were spatially dynamic and site specific, which were mainly structured both by local environments and host selection.

IMPORTANCE The Asian clam Corbicula fluminea is an important benthic clam in freshwater ecosystems due to its high population densities and high filtering efficiency for particulate organic matter. While the associated microbiota is believed to be vital for host living, our knowledge about the compositions, sources, and potential functions is still lacking. We found that C. fluminea offers a unique ecological niche for specific lake bacteria. We also observed high intrahabitat variation in the associated bacterial communities. Such variations were driven mainly by local environments, followed by host selection pressure. While the local microbes served as a source of the clam-associated bacteria, host selection resulted in enrichments of bacterial taxa with the potential for assisting the host in organic matter digestion. These results significantly advance our current understanding of the origins and ecological roles of the microbiota associated with a keynote clam in freshwater ecosystems.

INTRODUCTION

The Asian clam Corbicula fluminea is one of the most invasive freshwater species and has expanded its distribution in the last decades from Asia to Europe, America, and Africa (1). The great invasive and reproductive capacities of C. fluminea make this species an important factor in widespread mussel declines (2). High densities of C. fluminea in the benthic ecosystem influence aquatic systems in several ways, including alterations in species composition, species-species interactions, community structure, and ecosystem properties, through the production of durable shells, excretion, and filter-feeding behavior (3–5). This clam feeds mainly on particulate organic matter by filtering water at high rates, and it is also an important food source for fish, suggesting that C. fluminea plays an important role in connecting primary producers and terrestrial organic carbon to higher trophic levels (3, 6–8). Although the ecological roles of this species in freshwater ecosystems are important, comprehensive investigations of its associated microbiota, which not only may play important roles in host development, immunity, metabolism, behavior, and numerous other processes (9–13) but also may carry out biogeochemical transformations in its living environment (11, 14, 15), are still lacking.

The core microbiota framework aims to identify potentially crucial microbes within microbial communities based on the persistence of the microbes within the host and within a niche across spatial and temporal boundaries (16). In previous studies, the core bacterial community has been defined as operational taxonomic units (OTUs) with high occurrence frequencies, with the threshold ranging from 30% to 100% (17). The core microbiota is thought to play a key role in providing essential metabolic functions or beneficial adaptations to the host, and identifying the core microbiota of hosts is one of the necessary steps to deduce keystone microbial metabolic functions for the host (18–22). Highly diverse microbiota and significant variation of microbial community structure were found among different organs (tissues) of bivalves (23–26). Due to the filter-feeding ecology of bivalves, most of their microbial colonists are probably transient, i.e., just passing through the gill and gut (27). Identifying the core microbiota is a way to exclude the influence of opportunistic and transient microbiota and contamination on the understanding of symbioses (28).

Site-specific differences in host microbial associates have been reported in some studies across a wide range of organisms (27, 29–32). Having host sampling from the same site with a similar genotype background is a key factor in explaining the site effect of the host microbiota (9, 29, 30). Genetic diversity among hosts is known to generate variation in host ecology and, subsequently, the host microbiota. For example, the immune responses may differ among hosts, which in turn may influence the filtering of microbial colonists (27). Moreover, host ecology is also influenced by local environmental factors such as food type and temperature; thus, differences in local environmental parameters could also create site-specific differences in the host microbiota (24, 33, 34). In addition to internal sorting processes dependent on the host genotype and local environment, due to differential exposure to environmental microbes, site-specific differences in the host-associated microbiota are also influenced by colonization processes (27, 35).

In this study, we collected 64 C. fluminea tissue samples and 62 environmental samples from 7 sampling sites across the large shallow Lake Taihu in China. C. fluminea dominates the benthic biomass in most parts of this lake. The distance between two sampling sites ranged from 10 to 51 km. There were large environmental variations among the seven sampling sites, which included water chemical and physical properties, environmental microbes, and potential food sources for C. fluminea. Our study aimed to characterize the diversity and community structure of bacteria associated with C. fluminea. We particularly wanted to reveal if there are site-specific differences in C. fluminea-associated bacterial composition and mechanisms underlying these differences. We also wanted to know if C. fluminea clams from this large lake share a core bacterial community and if these core taxa are beneficial to the host clam in terms of their potential functions.

RESULTS

Diversity and community composition of bacteria associated with C. fluminea.

The operational taxonomic unit (OTU) richness and phylogenetic diversity were significantly lower in samples of C. fluminea tissue (638 ± 229 and 51 ± 19) than in free-living bacteria (1,319 ± 168 and 91 ± 8), particle-associated bacteria (1,519 ± 416 and 112 ± 26), and sediment bacteria (4,088 ± 197 and 251 ± 13), respectively (Fig. 1a) (P < 0.001 for all). Totals of 7,525, 5,249, 7,302, and 14,971 OTUs were found in C. fluminea tissues, free-living bacteria, particle-associated bacteria, and sediment bacteria, respectively. A total of 4,956 OTUs were shared between C. fluminea and the environmental samples; these shared OTUs accounted for approximately two-thirds of the total OTUs associated with C. fluminea (see Fig. S1 in the supplemental material). Fast expectation maximization for microbial source tracking (FEAST) was used to identify potential sources of microbes present in C. fluminea, and all of the environmental samples from different sampling sites were pooled as potential sources. Source contributions of bacteria associated with C. fluminea were mostly from particle-associated bacterial samples (35% ± 9%), followed by free-living bacteria (26% ± 9%) and sediment bacteria (14% ± 6%) (Fig. 1d).

FIG 1.

Comparison of the bacterial communities in the Corbicula fluminea (CF), free-living (FL) and particle-associated (PA) bacterioplankton, and sediment (S) samples. (a) Alpha diversity: OTU richness and phylogenetic diversity. (b) Beta diversity: NMDS analysis based on Bray-Curtis dissimilarity. (c) Relative abundances and taxonomic compositions of the microbiota in the Corbicula fluminea, free-living, particle-associated, and sediment samples at the phylum level. A to G indicate the sampling sites. (d) FEAST analysis depicting the contributions of bacteria from free-living, particle-associated, and sediment samples to C. fluminea. The boxes in panels a and d represent the upper and lower quartiles, horizontal lines indicate the medians, whiskers show the 95% ranges, and points are outliers. The letters above the columns indicate significant differences between different groups (P < 0.05 by two-sided Wilcoxon rank sum tests with Benjamini-Hochberg correction [95% confidence interval {CI}]). The number of samples follow: CF (n = 64), FL (n = 21), PA (n = 21), and S (n = 20).

Nonmetric multidimensional scaling (NMDS) ordination using Bray-Curtis dissimilarity matrices revealed that microbial communities clustered by sample type (i.e., C. fluminea tissue, free-living bacteria, particle-associated bacteria, and sediment bacteria) (Fig. 1b). Bacteria affiliated with Acidobacteria, Actinobacteria, Chloroflexi, Cyanobacteria, and Planctomycetes constituted a large proportion of the environmental microbial communities, whereas these groups were much less abundant in the samples from C. fluminea (Fig. 1c and Fig. S2). In contrast, the bacteria associated with C. fluminea were dominated by OTUs from members of the phyla Proteobacteria (31.3%), Firmicutes (22.9%), Spirochaetes (17.5%), Bacteroidetes (14.1%), Tenericutes (7.8%), and Epsilonbacteraeota (3.4%), whereas the latter two were almost absent in the environment (Fig. 1c and Fig. S2). Linear discriminant analysis effect size (LEfSe) analysis suggested that Firmicutes, Spirochaetes, Tenericutes, Bacteroidetes, Epsilonbacteraeota, Patescibacteria, and Fusobacteria were significantly enriched in C. fluminea samples (Fig. S3). A circular phylogenetic tree with detailed legends including all taxa enriched in C. fluminea samples is presented in Fig. S4.

According to function prediction by functional annotation of prokaryotic taxa (FAPROTAX), microbial functions, including chemoheterotrophy, fermentation, nitrate reduction, nitrate respiration, nitrogen respiration, and chlorate reduction, were significantly enriched in C. fluminea bacteria in comparison to the environmental microbial communities from the water column and sediment. Phototrophy, photoautotrophy, oxygenic photoautotrophy, and photosynthetic cyanobacterial functions were significantly enriched in environmental microbial communities (Fig. S5).

Community composition and oligotype composition of core bacteria associated with C. fluminea.

Among the 7,585 bacterial OTUs of C. fluminea, 106 OTUs were classified as core taxa (i.e., OTUs present in more than 70% of the C. fluminea samples); the core taxa accounted for 72% of the total bacterial sequences of C. fluminea samples (Fig. 2b). The relative abundance of each core OTU ranged from 0.01% to 8.1% (Table S3). The core bacterial communities were composed of 9 phyla, and the most abundant phyla were Firmicutes, Spirochaetes, and Proteobacteria, with averages of 26.2, 23.5, and 22.5% of all sequence reads, respectively (Fig. 2a). There were 12 genera with a relative abundance of >1% of all sequence reads, and these genera accounted for more than 50% of the total sequences (Fig. 2a). The potential functions of the core bacterial taxa determined by FAPROTAX included chemoheterotrophy, fermentation, aromatic compound degradation, nitrate reduction, chlorate reduction, and hydrocarbon degradation (Fig. S6).

FIG 2.

Taxonomic composition, phylogenetic tree, and distribution pattern of the core bacteria. (a) Community composition of C. fluminea core bacteria at the phylum and genus levels. (b) Contributions of group I and group II to the total number of OTUs and total sequences in samples from C. fluminea. Due to the low numbers and abundances of OTUs in group I, the red color in the key cannot be readily seen in the pie chart. (c) Distributions of 106 core OTUs in C. fluminea (CF), free-living (FL) and particle-associated (PA) bacterioplankton, and sediment (S) samples. The inner strip indicates the phylum. The strip colors associated with CF, FL, PA, and S indicate the average reads of each OTU in C. fluminea, free-living, particle-associated, and sediment samples. The hollow star in panel a indicates OTUs in group I, and the solid star indicates OTUs in group II. The number of samples follow: CF (n = 64), FL (n = 21), PA (n = 21), and S (n = 20).

Oligotyping analysis was used to further investigate the differences in core bacterial OTUs between C. fluminea and environmental samples. In total, 105 core OTUs were successfully identified as oligotypes, and the number of oligotypes in each core OTU ranged from 2 to 46 (Table S3). The core OTUs of C. fluminea were divided into two groups by determining whether the environmental bacteria were the main potential sources of these OTUs in the C. fluminea samples (Fig. S7). There were 11 OTUs in group I possessing higher numbers of oligotypes in environmental samples (10 to 32; 16 on average) than in the C. fluminea samples (6 to 24; 12 on average). The average relative abundance of these core OTUs in the environmental samples reached 0.2% (0.05% to 0.65%), with few unique oligotypes of these OTUs being observed in the C. fluminea samples (Fig. 3a and b and Table S3). Environmental samples had the potential to serve as a source for these 11 OTUs in C. fluminea. On the other hand, there were 92 core OTUs in group II that had fewer oligotypes in the environmental samples (0 to 16; 2.5 on average) than in the C. fluminea samples (2 to 46; 13.9 on average); these OTUs also had lower relative abundances in the environmental samples (0 to 0.2%; 0.01% on average) (Fig. 3a and Table S3). Moreover, the C. fluminea samples contained large numbers of unique oligotypes of these OTUs (Fig. 3b).

FIG 3.

(a) Relative abundances and numbers of oligotypes of each core OTU from group I and group II in the environmental (red box plots) and C. fluminea (green box plots) samples (paired two-sided Wilcoxon rank sum test [95% CI]). (b) Venn diagram showing the shared and unique oligotypes of the core OTUs from group I and group II for the environmental and C. fluminea samples. (c) Bray-Curtis dissimilarity based on oligotype compositions between individual C. fluminea samples and individual environmental samples. The oligotype compositions of the core OTUs in C. fluminea more closely resembled the oligotype compositions from their own site (red box plot) than those from foreign sites (green box plot). (d) Intrapopulation and interpopulation Bray-Curtis dissimilarity based on oligotype compositions. The interpopulation variation (green box plots) in the oligotype compositions was significantly greater than the intrapopulation variation (red box plots) for both group I and group II. The boxes in panels a, c, and d represent the upper and lower quartiles, horizontal lines indicate the medians, whiskers show the 95% ranges, and points are outliers (two-sided Wilcoxon rank sum tests with Benjamini-Hochberg correction [95% CI]) environment (n = 62) and C. fluminea (n = 64).

The OTUs in group I, which accounted for 0.8% of the total sequences in the C. fluminea samples, belonged to the phyla Actinobacteria, Nitrospirae, Proteobacteria, and Verrucomicrobia (Fig. 2c). Six of the 11 OTUs in group I were significantly enriched in environmental samples compared to C. fluminea samples, and 1 OTU in group I was significantly enriched in C. fluminea samples (P < 0.05) (Table S3). In contrast, the OTUs in group II accounted for 71.7% of the total sequences and were composed of taxa belonging to Actinobacteria, Bacteroidetes, Epsilonbacteraeota, Firmicutes, Proteobacteria, Spirochaetes, and Tenericutes (Fig. 2c).

Intrahabitat differences in the bacterial communities associated with C. fluminea.

Permutational multivariate analysis of variance (PERMANOVA) revealed that the bacterial communities present in C. fluminea from the seven different sampling sites were distinct from each other (R² = 0.33; P < 0.001) (Fig. 4a). Site-specific differences were also found in the oligotype compositions of core OTUs as well as the OTUs in group I and group II (R² = 0.34, R² = 0.31, and R² = 0.34, respectively; P < 0.001) (Fig. 4b and Fig. S8). The differences in the bacterial communities between two sampling sites were further examined using PERMANOVA (Table S4).

FIG 4.

Site-specific differences in the bacterial communities of C. fluminea. Bray-Curtis dissimilarity based on the total OTU composition and oligotype composition of core OTUs present in C. fluminea was visualized using NMDS. Differently colored symbols represent the 7 different sampling sites in Lake Taihu. The significance of the site factor on community dissimilarity was tested by PERMANOVA based on Bray-Curtis distances (total OTU composition, R² = 0.32 and P < 0.001; oligotype composition of the core OTU, R² = 0.34 and P < 0.001) (n = 64).

Furthermore, we performed a Wilcoxon rank sum test using Bray-Curtis distances between the oligotype compositions of individual C. fluminea samples and the corresponding individual environmental samples. We found a greater resemblance of the oligotype compositions of OTUs from group I in the C. fluminea samples to the bacteria from their local environments than to bacteria from foreign (other sampling sites) environments (P < 0.001) (Fig. 3c).

Cooccurrence network of bacteria associated with C. fluminea.

The cooccurrence network, which included 94 core OTUs (21.9% of all nodes), was composed of 2,864 edges and 429 nodes (Fig. 5a). The core OTUs from group II had significantly higher degree and closeness centrality values than did the noncore OTUs in the network (P < 0.001) (Fig. 5d and Table S5). We further defined hub species as OTUs in the network with high degree (>60) and closeness centrality (>0.3) values. Thus, 16 hub species composed of 7 core OTUs and 9 noncore OTUs were identified; all hub species were affiliated within the genus Romboutsia (Fig. 5 and Table S5). The network was grouped into three major (>10% of all network nodes) ecological clusters (modules), which comprised >50% of all whole-network nodes (Fig. 5c). Module 1 was composed of core and noncore OTUs and was represented by Romboutsia and Paraclostridium. Module 2 was mainly composed of noncore OTUs and was represented by Vogesella. All nodes of Nitrospira were shown in module 3.

FIG 5.

(a to c) Nodes of the network colored according to subcommunity (a), bacterial genus (b), and module (c). The sizes of the nodes reflect the degrees of connection. (d) Average degree and closeness centrality values of OTUs in group I and group II and noncore OTUs. Core unknown in panel a indicates that the core OTUs cannot be classified as group I or group II. The boxes in panel d represent the upper and lower quartiles, horizontal lines indicate the medians, whiskers show the 95% ranges, and points are outliers. The letters above the columns indicate significant differences between different groups (P < 0.05 by two-sided Wilcoxon rank sum tests with Benjamini-Hochberg correction [95% CI]) (n = 64).

DISCUSSION

The freshwater clam C. fluminea offers a unique ecological niche for specific lake bacteria.

The environment is a reservoir for microbial taxa associated with plants and animals (5, 36). In our study, approximately two-thirds of the bacterial OTUs associated with C. fluminea were found in the environment (see Fig. S1 in the supplemental material), and source tracking analysis also predicted that the environment contributed up to 86% of the taxa of bacteria associated with C. fluminea at the genus level (Fig. 1d). Interestingly, although the species diversity was higher in the sediment than in the other environmental sample types (Fig. 1a), the C. fluminea-associated bacteria were derived mostly from particle-associated samples (Fig. 1d), which is congruent with the filter-feeding ecology of the host. Moreover, although some bacterial taxa were shared among the sample types, the C. fluminea-associated bacteria do not merely represent the bacterial community of the surrounding environment (14, 22). The large decrease in bacterial alpha diversity from the environmental samples to the C. fluminea samples and the significant differences in bacterial community composition between the C. fluminea samples and the environmental samples indicated that clam hosts impose ecological selection pressure on the bacterial diversity from the surrounding environment (37).

Oligotyping analysis was used to further contrast the bacterial community compositions between the C. fluminea samples and the environmental samples. In group I, the environmental bacteria could be the source of the bacterial OTUs in the C. fluminea samples; few unique oligotypes of bacterial OTUs from this group were observed in the C. fluminea samples, and there was a higher similarity of the oligotype compositions of these OTUs in the C. fluminea samples with their local environmental samples (Fig. 3). The presence of these 11 OTUs in C. fluminea was likely due to horizontal transmission from the environment. We found that most of the core-associated bacterial OTUs (group II) have lower abundances and occurrence frequencies in environmental samples (Fig. 2c, Fig. 3a, and Table S3). A large number of the oligotypes of OTUs in group II were undetectable in environmental samples, and the C. fluminea samples contained a large number of unique oligotypes of these OTUs (Fig. 3b).

Intrahabitat variations in the C. fluminea microbiota.

Consistent with previous studies on host symbionts, a site effect was also found for bacteria associated with C. fluminea (Fig. 3d and Fig. 4). Although the filter-feeding ecology of bivalves makes them highly exposed to environmental microorganisms, necessary measures have been taken to avoid contaminant bacteria from water and sediment during sample collection, i.e., rinsing tissues several times with sterile water to get rid of loosely attached foreign bacteria. After further excluding opportunistic and transient bacteria and contamination by identifying core taxa, site-specific differences were still found in core bacteria (oligotype level) (Fig. 4b).

In our study, we observed that the core C. fluminea-associated bacteria belonging to group I more closely resembled the environmental bacteria from the host’s site than those from other sites. This suggested that the among-site oligotype differences in the group dominated by horizontal transmission may be due to differential exposure to environmental microbes. Previous studies reported that dispersal is expected to be high for horizontally transmitted bacteria and that stochastic effects on the colonization of environmental bacteria will manifest themselves as differences in the associated microbial composition among hosts (38–40).

Potential ecological functions of C. fluminea-associated core bacteria.

According to the cooccurrence network analysis, the core OTUs, especially those in group II, had significantly higher degree and closeness centrality values than did noncore OTUs (Fig. 5d), suggesting that core taxa play a key role in maintaining the network stability of the associated microbiota.

In our study, the same taxa (genus) tended to group into the same modules (Fig. 5), which is consistent with previous studies showing that members of a network module often exhibit similar habitat adaptations and ecological functions (41, 42). Romboutsia and Paraclostridium are the most abundant genera in the core bacteria of C. fluminea and belong to the family Peptostreptococcaceae (Fig. 2a), which was found to be the primary taxon in module 1 (accounting for 90% of the nodes in module 1) of the network. These two genera are common in the guts of animals and are capable of fermentation, according to function prediction. In this study, we were not able to pick up the digestive microbiome independently due to the small size of the clams. Romboutsia and Paraclostridium are probably localized in the digestive system of C. fluminea and are involved in fermenting polysaccharides, resulting in short-chain fatty acids that can be further assimilated by the host (13). Thus, we speculate that the complex relationships of Romboutsia and Paraclostridium from module 1 in the network are due to the utilization of metabolites that they produce in food digestion processes.

A large variety of benthic invertebrates were found to be actively involved in the aquatic nitrogen cycle, i.e., excreting inorganic nitrogen at the form of feces and emitting nitrous oxide (43–45). We found some genera involved in the nitrogen cycle among the core bacteria of C. fluminea. Bacteria affiliated with Aeromonas, Azospira, Azoarcus, and Dechloromonas have the potential for nitrate reduction according to FAPROTAX predictions, and some strains of Acidovorax, Halomonas, and Acinetobacter were previously reported to be capable of heterotrophic denitrification of nitrate (46–51), which is consistent with previous studies showing that the high content of organic matter in the anoxic gut makes it an ideal habitat for denitrifiers (52–54). In addition, we also found genera related to nitrification (Nitrospira) among the C. fluminea core bacteria. The surrounding environment of benthic invertebrates is usually abundant in ammonium and dissolved organic N released by the host itself, which causes the enrichment of nitrifiers in the mantle, shell, and surrounding sediment of the bivalve (25, 55, 56). The presence of genera associated with the N cycle in conserved core bacteria indicates that C. fluminea may be involved in essential biogeochemical processes in aquatic ecosystems through host-bacterium interactions.

Conclusions.

In this study, we characterized bacteria associated with C. fluminea through a 16S rRNA gene sequencing approach. We found that C. fluminea offers a unique ecological niche for specific lake bacteria. A total of 106 core OTUs were identified in bacteria associated with C. fluminea from Lake Taihu. Most of the core OTUs have lower relative abundances and occurrence frequencies in environmental samples. Site-specific differences were found in all and core bacteria associated with C. fluminea. Moreover, we also found that a small set of core OTUs may be horizontally derived from environmental bacteria, and this colonization process could be site specific in C. fluminea-associated bacteria. We found that core taxa play an important role in maintaining network stability. These conserved bacterial associates include taxa with the potential for assisting host clams in food digestion. Altogether, these results significantly advance our current understanding of the origins and ecological roles of the microbiota associated with a keynote clam in freshwater ecosystems.

MATERIALS AND METHODS

Sample collection and preparation.

In July 2020, water, sediment, and C. fluminea samples were collected from 7 sampling sites across Taihu Lake (30°55′40″ to 31°32′58″N, 119°52′32″ to 120°36′10″E) (Fig. 6). Taihu Lake is a typical large, shallow eutrophic lake in China with an area of 2,427.8 km² and an average depth of 1.9 m. The northwestern portion of the lake is hypertrophic, while the eastern portion is less eutrophic and covered by various species of aquatic macrophytes (57, 58). Chlorophyll and nutrient concentrations gradually decrease from the northwest to the southeast (57). Sampling sites A, B, and C are located in the most eutrophic part of the lake, which exhibits a cyanobacterium-dominated turbid state; sites D and G are located in the transitional area with lower eutrophication; and sites E and F are located in the submersed macrophyte-dominated region (Fig. 6; see also Table S1 in the supplemental material). Characterization of the water physicochemical parameters at each sampling site can be found in Table S1. Surface water samples were collected at a depth of 0.5 m with a 5-L Schindler sampler. The C. fluminea and sediment (0 to 10 cm) samples were collected with a Peterson grab sampler (1/16 m²) at each sampling site. All collected samples were transported to a laboratory for further processing within 3 to 8 h. At each sampling site, we took three replicated samples of water and sediment as well as 6 to 14 individuals of C. fluminea.

FIG 6.

Sampling sites in Lake Taihu, China. Red stars indicate the seven sampling sites.

Approximately 500 mL of water from each sampling site was filtered sequentially with 5-μm and 0.22-μm filters (Isopore membrane filters; Merck Millipore) for analyses of the diversity of particle-associated and free-living bacteria, respectively. All of the filter membranes were stored at −40°C until microbial DNA extraction. The widths of the C. fluminea individuals were measured with a vernier caliper. The shells of all C. fluminea clams (width of 1 to 3.3 cm) were scrubbed and rinsed to remove debris. Each individual was stored overnight for about 12 h in a 100-mL transparent plastic bottle with sterile water to evacuate the gut contents. As the C. fluminea clams were too small and the organs cannot be separated independently without contamination from other tissues during dissection, soft tissues, including adductor muscle, mantle, foot, gill, and visceral tract, of C. fluminea were thus dissected using sterile knives and scissors as a whole. After rinsing with sterile water several times, the soft tissue was placed into a 1.5-mL tube for further treatments. Sediment and tissue samples were freeze dried and stored at −40°C until microbial DNA extraction. In total, we collected 64 C. fluminea tissue samples, 21 samples for particle-associated bacteria, 21 samples for free-living bacteria, and 21 samples for sediment bacteria.

DNA extraction and high-throughput sequencing.

DNA was extracted from the filter membranes, sediments, and C. fluminea tissues using the DNeasy PowerSoil Pro DNA isolation kit (Qiagen, Germanton, MD, USA) according to the manufacturer’s instructions. The soft tissue of C. fluminea was homogenized in lysis buffer provided with the DNA isolation kit. Primers 515F (5′-GTGYCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACNVGGGTWTCTAAT-3′) targeting the V4 region of bacterial 16S rRNA genes were selected for PCR amplification (59). The PCR protocol consisted of an initial denaturation step at 95°C for 2 min followed by 25 cycles of denaturation at 95°C for 30 s, annealing at 55°C for 30 s, and elongation at 72°C for 30 s, with a final extension step at 72°C for 5 min. Each reaction mixture contained 4 μL of 5× FastPfu buffer, 2 μL of 2.5 mM deoxynucleoside triphosphates (dNTPs), 0.8 μL each of 5 μM forward and reverse primers, 0.4 μL of FastPfu polymerase, and 10 ng of template DNA. Triplicate PCR products were pooled, purified, and used for library construction according to Illumina’s genomic DNA library preparation procedure and sequenced using the Illumina MiSeq platform (Shanghai Biozeron). Negative controls for DNA extraction and PCR amplification were also performed to exclude potential contamination. We did not find any potential contamination for the negative controls. A total of 126 samples were sequenced successfully and considered for downstream analysis; 1 sediment sample was removed from further analysis because there were only 8,760 reads for this sample (Table S2).

Sequence analysis.

Raw fastq sequences were demultiplexed, quality filtered, and assembled in QIIME (version 1.9.1) as previously described (60). Operational taxonomic units (OTUs) were clustered with 97% similarity using UCLUST and compared against the SILVA (SSU138.1) 16S rRNA database using a confidence threshold of 80% (61). A total of 5,646,394 high-quality reads were obtained, and all samples were randomly subsampled to 18,226 reads prior to downstream analysis. Alpha diversity, including OTU richness and phylogenetic diversity, was calculated using the vegan and picante packages in R (http://www.r-project.org) (62, 63). In this study, we defined the core microbiota as OTUs that were present in at least 70% of the C. fluminea tissue bacterial communities (64, 65).

Oligotyping analysis.

The intra-OTU microdiversity of 106 core OTUs was analyzed by oligotyping, according to the pipeline described by the developers (http://merenlab.org/software/oligotyping/). This analysis identifies variable sites in sequences based on Shannon entropy values, generating oligotypes and helping resolve closely related taxa (66). Five entropy positions (-c option) were used to assess the oligotypes of each core OTU. Entropy positions were not identified in 1 OTU, and 105 core OTUs were retained for further oligotyping analysis. To reduce the noise, only the oligotypes that appeared in at least 3 samples (-s option) were considered for downstream analysis.

Microbial network analysis.

The OTUs with low frequencies (i.e., those present in <21 C. fluminea tissue samples) were removed from the network analysis to avoid introducing spurious correlations (41). The network analysis was conducted using the CoNet routine (42) in Cytoscape v3.9 (67) by selecting Spearman and Pearson correlation measures, and the inferred correlations were restricted to those having correlations greater than 0.7 or less than −0.7 (P < 0.05). The network was then visualized with Gephi (v0.9.2) (68). The topology characteristics (degree and closeness centrality) and modularity of the network were also calculated using Gephi. Hub OTUs were defined as nodes with high degree (>60) and high closeness centrality (>0.3) values (37).

Statistical analysis.

Fast expectation maximization for microbial source tracking (FEAST) was used to predict the contributions of the environment to the microbiota in individual C. fluminea clams using the FEAST R package (69). A phylogenetic tree was constructed using MEGAX (v10.1.8) based on the maximum likelihood method and annotated and visualized using iTOL software (70). Bacterial functional profiles were predicted using functional annotation of prokaryotic taxa (FAPROTAX) (71). Functional differentiation of the bacterial communities in C. fluminea and environmental samples was performed and graphics were created using STAMP (v2.1.3) through Welch’s t test (72), and P values were adjusted with the Benjamini-Hochberg false discovery rate (FDR) multiple-test correction. Bray-Curtis dissimilarity matrices were used for nonmetric multidimensional scaling (NMDS) to assess differences in beta diversity by PERMANOVA (permutational multivariate analysis of variance), using the adonis function of the vegan R package. Linear discriminant analysis effect size (LEfSe) tests were conducted to detect discriminative taxa enriched in the associated bacteria of C. fluminea (73). Nonparametric statistical tests were run to evaluate the alpha diversities, source contributions, Bray-Curtis dissimilarities, occurrence frequencies of core OTUs, numbers of oligotypes of core OTUs, relative abundances of core OTUs, and node-level topological feature (degree and closeness centrality) differences among different groups (Kruskal-Wallis test or Wilcoxon test) using wilcox.test and kruskal.test of the stas R package. Graphs were created using the ggplot2 R package.

Data availability.

Raw data files have been made available at the NCBI Sequence Read Archive under BioProject accession number PRJNA766455.