Sediment microbial communities contribute to shrimp intestine microbiota in cultural pond ecosystems

Microorganisms in a shrimp cultural pond ecosystem (SCPE) play important roles for animal and environmental health. While the microbial diversity and assembly mechanism, and how communities of multiple habitats contribute to the SCPE metacommunity are poorly understood. Here, we analyzed phylogenetic structure of 792 samples of three habitats (water, shrimp intestine and sediment) from 88 shrimp cultural ponds across six regional sites in China. We show that the microbial communities of three habitats in the SCPE contained 15,197 operational taxonomic units (OTUs), and the microbial community of sediment had the highest OTUs (14,857). Despite the high microbial diversity, the SCPE has small core taxa (13 OTUs) among three habitats, and 30, 15, 53 core OTUs in water, shrimp intestine, and sediment, respectively. Especially, some opportunistic pathogens or functional microbes were core taxa in shrimp intestine or water and sediment. Of the SCPE metacommunity, the microbial diversity of the SCPE was mainly contributed by‾αLocalCommunities (66.0%), and followed by βInterHabitats (29.0%). The sediment microbial communities had a larger contribution (56.8%) to shrimp intestine microbiota compared to water communities (18.1%), which was also corroborated by the Sloan neutral community model analysis. Further, microbial assembly of three habitats appeared to be largely driven by stochastic processes (over 78%) in the SCPE. Our results demonstrate that those core microbial taxa of three habitats were distinct with each other, that sediment communities mainly contributed to shrimp intestine communities, and that the microbial assembly was largely driven by stochastic processes.

ecosystems. Therefore, it is necessary to understand the microbial ecology of aquaculture ecosystems for the sustainable outputs of aquaculture products.
The microbial communities of multiple habitats had high diversity [17,[21][22][23]and varied concurrently with some key environmental and geographic factors (e.g., host developmental stages, environmental factors, geographical distance) in aquaculture ecosystems [14,16,24,25]. Despite recent advances in understanding microbial ecology of aquaculture ecosystems, their microbial assembly mechanisms remain unclear. In general, the mechanisms shaping the microbial diversity among species are considered as ecological processes [26]. Recently, our knowledge about ecological processes in shaping the microbial structure has been enriched substantially, and the importance of ecological stochasticity has also been emphasized [27]. For aquatic ecosystems (e.g., lakes), deterministic processes generally play a primary role in shaping the water or sediment microbial community structure [28,29]. Of aquatic animal intestines, a dominant role of stochastic processes was observed for microbial assemblages, and deterministic processes became less important as hosts developed [18,24,30]. More importantly, microbial communities of multiple habitats (e.g., water, animal intestine, sediment habitats) constitute a metacommunity in aquatic ecosystems [31][32][33][34], and such close relationships were observed among them, which is important for aquatic animal productivity and health [15,35]. However, how microbial communities of multiple habitats (water, animal intestine, sediment) in aquaculture ecosystems contribute to the metacommunity is poorly understood. Especially, it is essential to understand the interactions among microbial communities of animal intestine and surrounding environments for healthy aquaculture [35].
In this study, we aimed to understand the microbial assembly mechanisms for a metacommunity of three habitats (water, shrimp intestine and sediment) in the shrimp cultural pond ecosystem (SCPE) of Litopenaeus vannamei, and specifically focusing on three ecological questions: (i) Are there core microbial taxa among three habitats across the SCPE or in each habitat of the SCPE? (ii) What is the contribution of microbial communities from each habitat to the SCPE metacommunity? (iii) What ecological processes shape the microbial community structures in three habitats? To address those questions, we hypothesized that there would be core microbial taxa among communities of three habitats across the SCPE and in each habitat, and that sediment communities would have a decisive role to intestine microbiota of L. vannamei as it is one of benthic animals and has the characteristics of feeding from sediments. To test those hypotheses, we collected 792 samples from 88 cultural ponds in six regional sites of China (Additional file 1: Table S1), and analyzed microbial communities and the contribution of ecological processes to their assembly as well as the relationship among communities of three habitats in the SCPE based on 16S rRNA gene sequences. We found that those core microbial taxa of three habitats were distinct with each other, that sediment communities mainly contributed to shrimp intestine communities, and that the microbial assembly was largely driven by stochastic processes. Our study provides new insights into the understanding of microbial assembly mechanisms and the developed framework will facilitate the metacommunity analysis in the SCPE, significantly advancing microbial ecology of aquaculture systems and shrimp culture strategies.

Overview of microbial diversity of three habitats in the SCPE
To understand the microbial diversity in the SCPE, we sequenced 16S rRNA gene amplicons from 792 samples in 88 shrimp cultural ponds across six regional sites. A total of 38,662,478 high quality sequences (30,580 to 87,563 sequences for each sample) were obtained from all samples, and clustered into 15,197 operational taxonomic units (OTUs) with the highest number (i.e. 14,857) of OTUs in the sediment (Additional file 1: Tables S2   and S3). Such sequencing efforts were enough to capture a majority of microbial communities in all samples (Additional file 1: Table S3). The Shannon index was the highest in sediment (6.5 ± 0.3), and followed by water (4.4 ± 0.5) and shrimp intestine (3.2 ± 1.2), which significantly (P < 0.001) differed among those three habitats, and Chao1 index showed similar results ( Fig. 1a; Additional file 1: Table S3). To further evaluate the overall differences among three habitats, non-metric multidimensional scaling (NMDS) analysis showed that microbial communities were clustered by water, shrimp intestine and sediment habitats, which was further corroborated by analysis of similarity (ANOSIM), revealing that the microbial community structure significantly (r = 0.8115, P < 0.001) differed between any two of compared habitats (Fig. 1b).

Core OTUs among three habitats or of each habitat in the SCPE
To examine if core microbial taxa exist among three habitats in the SCPE, we defined that core microbial taxa should occur in ≥ 90% of all 792 samples tested (see Methods for details). About 0.08% (13 out of 15,197 OTUs) constituted core taxa among three habitats that accounted for the relative abundance of 7.98% of all sequences ( Fig. 2a and 2b). The core OTUs belonged to Cyanobacteria (6 OTUs and a relative abundance of 5.63%), Actinobacteria (3 OTUs and 0.96%), Proteobacteria (3 OTUs and 0.98%), and Verrucomicrobia (1 OTU and 0.41%) (Additional file 1: Table S4).

Comparison of the microbial community composition in three habitats
To understand the composition of microbial communities from three habitats, we compared them at the OTU level using Venn analysis. The results showed that a total of 5,997 OTUs were commonly present in three habitats, and the number of OTUs was found to be in any two habitats: 6,094 (water, 81.6%) or 7,279 (sediment, 97.5%) out of 7,466 intestine OTUs, 6 094 (intestine, 77.2%) or 9,625 (sediment, 97.5%) out of 9,874 water OTUs, and 7,279 (intestine, 49.0%) or 9,625 (water, 64.8%) out of 14,857 sediment OTUs (Fig. 3). Most OTUs were in the sediment habitat, and a high percentage (~97.5%) of OTUs in shrimp intestine and water habitats were shared with the sediment habitat. For each pond and all six regional sites, similar trends were observed (Additional file 2: Figure S1).
We further compared the composition of microbial communities among three habitats at the phylum and genus levels, showing that all detected phyla and genera were always present in any of the three habitats (Additional file 2: Figures S2 and S3), but their relative abundances significantly (P < 0.001) differed (Additional file 1: Tables S9 and S10; Additional file 2: Figure S4a and S4b). Some opportunistic pathogens, such as Vibrio, Photobacterium and Candidatus Bacilloplasma, were detected in three habitats, and their relative abundances were significantly (P < 0.001) higher in the shrimp intestine than in the water or sediment habitat (Additional file 1: Tables S10 and S11; Additional file 2: Figure S4b and S4c).

Sediment communities mainly contribute to shrimp intestine microbiota compared to water community in the SCPE metacommunity
In order to evaluate the contribution of microbial communities of three habitats to the SCPE metacommunity diversity, we used additive diversity partitioning of diversity across scales to determine whether the microbial diversity observed at the ecosystem level (ɤ Ecosystem ) was mainly from a high microbial dissimilarity among habitats (β InterHabitats ), a high dissimilarity among communities within each habitat (β IntraHabitats ) or from a high microbial diversity within each local community ( LocalCom m unities , i.e., water, shrimp intestine or sediment sample). The results showed that the contribution of α LocalCom m unities to the metacommunity diversity (ɤ Ecosystem ) was 66.0 ± 11.2%, outweighing β InterHabitats (29.0 ± 11.0%) and IntraHabitats (5.0 ± 2.4%) in their contributions to ɤ Ecosystem (Fig. 4a). This high contribution of α-diversity and β-diversity to ɤ Ecosystem revealed that LocalCom m unities and β IntraHabitats were important for generating the microbial diversity in the SCPE.
To test if sediment communities have a decisive role to shrimp intestine microbial communities, we evaluated the contribution of different source communities to water, shrimp intestine and sediment habitats in the SCPE by SourceTracker analysis. For water microbial communities, the most dominant potential source was sediment (an average of 53.6%), followed by shrimp intestine (35.0%); for shrimp intestine communities, the most dominant potential source was also sediment (56.8%), followed by water (18.1%) (Fig. 4b).
In contrast, for sediment communities, the source from water or shrimp intestine only attributed 17.9% or 15.4 %, respectively (Fig. 4b). These results indicated that each microbial community could be the source for the other two communities, and especially, sediment was found to be the most important source of water and shrimp intestine communities, and such similar trends were observed across all six regional sites (Additional file 2: Figure S5), indicating a possible general pattern in the SCPE.
The Sloan neutral community model was further applied to analyze the shared OTUs between the surrounding sediment or water and shrimp intestine. That is, neutral distribution (black points) accounted for 37.8% in sediment, and 30.8% in water microbial communities (Fig. 4c). In contrast, the proportion of under-represented (green points) and over-represented (red points) OTUs was 15.0% or 5.2%, and 44.7% or 45.7% in sediment or water, respectively (Fig. 4c). Thus, the proportion of shared and neutrally distributed OTUs between shrimp intestine and sediment was higher than these in water, suggesting that communities of shrimp intestine were tended to colonize from sediment rather than from water. More obvious similar trends were observed for the six locations across a regional scale (Additional file 2: Figure S6).

The microbial assemblies of three habitats in the SCPE were largely controlled by stochastic processes
In order to understand microbial assembly mechanisms in three habitats of the SCPE, we used the null model-based approach to calculate stochastic ratios with taxonomic (Bray-Curtis, abundance-weighted and unweighted) and phylogenetic (weighted and unweighted UniFrac) metrics. The results showed that the average stochastic ratios based on taxonomic and phylogenetic metrics were higher than 78.4% in water, 80.8% in shrimp intestine and 80.3% in sediment habitats (Fig. 5a), suggesting that stochastic factors were more important than deterministic factors in influencing microbial community composition of three habitats in the SCPE.
Variation partitioning analysis (VPA) was performed to discern the relative importance of and geographic distance showed a significant correlation (P < 0.001) with the water or sediment microbial structure. These variables explained 29.5% or 24.9% of the observed variation in water or sediment habitats, respectively, leaving 70.5% or 75.1% of the variation unexplained ( Fig. 5b and 5c). The water or sediment environmental factors explained 21.7% or 14.2% (P < 0.001), and geographic distance alone explained 2.5% or 9.0% (P < 0.001) variations with 5.3% or 1.7% interaction effect detected, respectively ( Fig. 5b and 5c).

Discussion
Microbial communities of multiple habitats constitute the metacommunity in aquatic ecosystems. However, a more comprehensive understanding of the microbial assembly mechanism and the relationships among communities of animal intestine and surrounding environments in aquatic ecosystems remains unclear. In the present study, we analyzed the microbial assembly of water, shrimp intestine and sediment habitats to understand how their microbial communities contribute to the SCPE metacommunity, and the results generally support our hypotheses.
Our first hypothesis is that there would be core taxa among microbial communities of three habitats across the SCPE and in each habitat. Core microbial taxa provide information on putatively important microorganisms for ecosystem functioning [36].
Previous studies showed that there were 9, 6, 2, 1 and 28 bacterial OTUs identified as core taxa in soil, human feces, air, freshwater and wastewater treatment plants, identified as core taxa in a sediment habitat, reflecting their importance for sulfur cycling and organic matter degradation in aquatic ecosystems [42,43]. Thus, core microbial taxa were among three habitats across the SCPE or in each habitat, and such core taxa in shrimp intestine may play key roles for shrimp and environment health in the SCPE.
Our second hypothesis is that sediment microbial communities would have a decisive role for shrimp intestine microbiota in the SCPE. Generally, multiple habitats constitute a metacommunity for the overall microbial diversity in aquatic ecosystems [31], and surrounding environments are the main sources of microbes colonizing aquatic animal intestines, and the host animal drives, in a large part, the selection of microorganisms [44,45]. Aquatic animals are theoretically microorganism-free at birth, and all postnatally acquired intestine microorganisms should migrate from their surroundings [24]. As all activities carried out by aquatic animals (e.g., feeding, defecation) take place in water or/and sediment habitats, interactions between host and their environments may be much more direct than those between terrestrial animals and their environments, thus the assembly of aquatic animal intestine microbial communities is directly influenced by their environmental communities [36,44] dispersal is a key factor influencing the metacommunity and its associated community structure [48]. In this study, we observed that a high percentage of dispersal rates from sediment to the other two habitats in the SCPE, and especially, compared with water communities, shrimp intestine communities were more closely related to sediment communities. The explanations can be largely due to shrimp lifestyles and sediment features. It is well-known that L. vannamei is a planktobenthos (mainly lives a benthic and sometimes floats in water), and their activities are more related to the sediment habitat.
Also, L. vannamei has the characteristics of feeding from the sediment and ingestion of particulate matter into its intestine. Our results indicate that sediment communities mainly contribute to shrimp intestine microbiota, but such mechanisms need to be further investigated.
We also found that each habitat harbored distinct microbial communities, indicating that different taxa have obvious preferences in three habitats. For example, Vibrio,

Photobacterium, Candidatus Bacilloplasma, Shewanella, Spongiimonas, Rhodobacter and
Aeromonas were enriched in shrimp intestine. On one hand, Vibrio, Photobacterium and Candidatus Bacilloplasma are known as opportunistic pathogens. Previous studies indicated that these pathogens in shrimp intestine could be from their surrounding environments, and they are widespread in cultural pond ecosystems [14,18,49]. The crucial question is if the aquatic animal intestine offers a best-suited micro-environment for such rare taxa from the environment to become dominant in shrimp intestine. On the other hand, those opportunistic pathogens may also spread into their environments through the excretion of aquatic animals, making the control of proliferation of opportunistic pathogens extremely challenging [13]. For example, they may be excreted by aquatic animals and spread into the environment as "seeds", when these pathogens in the environment are killed by disinfection [50]. As aquatic animals are very important for maintaining the microbial diversity in aquatic environments [51], more microbial ecological management strategies should be developed to restrain opportunistic pathogens in aquaculture.
Additionally, further understanding of underlying microbial assembly mechanisms was attempted, showing that the stochastic processes could play more important roles in influencing the microbial community structure than deterministic processes. A possible reason may explain such observations: ecological drift (e.g., stochastic processes of birth, death, colonization) becomes stronger due to high percentages of dispersal rates [52]. A recent study of activated sludge communities from wastewater treatment plants at a global scale indicated that microbial spatial turnover was largely driven by stochastic processes [37]. Similarly, in other natural and engineering ecosystems, such as temperate forest [53], grassland (under warming condition) [54], bioreactor [55]and groundwater system (perturbed by adding emulsified vegetable oil for uranium immobilization) [52], stochastic processes played larger roles than deterministic ones in explaining the microbial assembly. Our present study is largely consistently with those previous studies, suggesting that the microbial assembly was largely driven by stochastic processes in the SCPE.

Conclusion
In summary, the importance of microbial communities in aquaculture ecosystems has been widely recognized. Here, we systematically evaluated the microbial community composition of water, shrimp intestine and sediment habitats in the SCPE as a metacommunity, and revealed their relationships and possible assembly mechanisms ( Fig. 6). Specifically, we found that core microbial taxa were among three closely related habitats and in each habitat, and sediment communities dominated intestine microecosystem of L. vannamei, and microbial variations were largely controlled by stochastic processes in the SCPE. This study provides new insights into microbial assembly mechanisms and a framework for metacommunity analysis in the SCPE, which will have important implications for developing new strategies for shrimp healthy culture.

Sample collection and physicochemical analysis
A total of 792 water, shrimp intestine and sediment samples were collected from six Paired-end sequences were merged using FLASH [56], and merged sequences were processed following the Quantitative Insights Into Microbial Ecology pipeline (QIIME version 1.9.0) [57]. In brief, the sequences with ambiguous bases or truncated at any site of more than three consecutive bases receiving a Phred quality score (Q) < 20 were removed. Chimeric sequences were discarded using the UCHIME algorithm [58]. Sequences with a distance-based identity of 97% or greater were grouped into OTU using UCLUST [59]. The most abundant sequence from each OTU was selected as representative and then was taxonomically assigned against the Silva SSU database 128 using the RDP Classifier algorithm (http://rdp.cme.msu.edu/), which enables each identified OTU to have a close relative. The core OTU was defined based on multiple reported measures: OTU with an occurrence frequency in more than 90% of all samples [60,61]. Following the same criteria as described above, the core OTUs was identified for all 792 samples among three habitats or each habitat (e.g., 264 samples each in water, intestine and sediment habitats) in the SCPE.

Relationships among microbial communities of water, intestine and sediment habitats
The relationships among microbial communities in water, shrimp intestine and sediment habitats of each pond and all sites were firstly analyzed using Venn analysis [62]. An additive partitioning framework was applied to separate the total microbial diversity at the ecosystem level (ɤ Ecosystem ) into contributions at smaller scales from habitats to local communities [63]. More precisely, total ecosystem microbial diversity was expressed as the sum of inter-habitat difference in the community diversity, the mean intra-habitat difference and mean local community diversity with: ɤ Ecosystem = β InterHabitats + IntraHabitats + LocalCom m unities . The ecosystem level (ɤ Ecosystem ) may arise from a high microbial dissimilarity among habitats (β InterHabitats ), a high dissimilarity among communities within each habitat (β IntraHabitats ) or from a high diversity within each local community ( LocalCom m unities ; i.e., each water, shrimp intestine or sediment sample). To further evaluate the relationships among microbial communities in water, shrimp intestine and sediment habitats, the different sources were used to estimate their contributions to microbial community composition of the shrimp cultural ponds using SourceTracker based on Bayesian algorithm [64], which was run through QIIME with default settings and with one habitat as the sink and the other two habitats as sources. The Sloan neutral community model [65] was used to analyze the OTUs that were shared between the shrimp intestine and surrounding water/sediment, in which the microbial community in water or sediment was the source of intestine microbiota. This model predicts that the probability of detecting an OTU in shrimp intestine due to dispersal is directly proportional to its abundance in the corresponding water/sediment community. OTUs were sorted into three categories depending on whether they occur more frequently (over-represented), less frequently (under-represented) or within (neutrally distributed) the 95% confidence interval of the neutral model predictions.

Stochasticity of microbial assembly
We assessed microbial community-assembly stochasticity with a null-model-based index, stochasticity ratio, as described previously [52,54,66]. Since null-model approaches usually require adequate replicates, each sampled regional site in this study had more than 8 cultural ponds samples. We calculated stochasticity ratios using both taxonomic and phylogenetic metrics. When using dissimilarity index (abundance-weighted and unweighted Bray-Curtis), the stochasticity ratio was calculated based on typical null-model algorithms for taxonomic metrics [67,68]. When using weighted and unweighted UniFrac, the stochasticity ratio was calculated based on typical null-model algorithms for phylogenetic metrics [68,69]. Samples within each regional site were considered sharing the same regional species pool in null model algorithms.

Statistical analysis
A ternary plot was applied to reveal the distribution of the dominant genera (> 0.1%) among water, shrimp intestine and sediment habitats using the package "ggtern" in R 3.3.2 [70]. Welch's t-test was used to compare the microbial diversity indices among water, shrimp intestine and sediment habitats. NMDS and ANOSIM were performed to evaluate the overall differences in microbial communities of water, shrimp intestine and sediment using weighted UniFrac distance [14]. Then, the differentially abundant taxa among three habitats were identified using one-way analysis of variance [34].  Figure 1 Microbial diversity among water, shrimp intestine and sediment habitats in the SCPE. a Statistical significance of the -diversity indices among water, shrimp intestine, and sediment habitats were based on the Welch's t-test (**: P < 0.01).
b β-diversity of water, shrimp intestine, and sediment microbial communities in the SCPE analyzed by NMDS and ANOSIM based on weighted UniFrac distance.

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to download.