Drivers of Microbial Community Structure in Surface Sediment Aross Bohai Sea

Background: Microbial spatial distribution has been widely investigated in sediment. However, there is poorly available information on microbial distribution patterns in sediment of Bohai Sea coastal zone. Results: Here, we investigated the bacterial community composition and diversity in riverine and marine surface sediment around and in the Bohai Sea using high-throughput sequencing. Bacterial communities mainly comprised Proteobacteria, Bacteroidetes, and Firmicutes. Salinity, dissolved oxygen, pH, and magnetic susceptibility played the main role in determining bacterial α-diversity and community composition in this region. Of the total bacterial community composition variation, environmental factors (explained 29.41% of the total microbial community composition variation) played a more important role than spatial variables (explained 3.03%) in conditioning the bacterial community composition. Meanwhile, the signicantly pure spatial effect and distance-decay tendency suggested that dispersal limitation was also an inuential factor in shaping the bacterial biogeographical pattern. The presence of magnetite center might shape the geographical distribution of ve genera Lactococcus, Clostridium, Caulobacter, Gillisia and Sphingomonas probably by affecting their iron-related geochemical cycle. Conclusion: Our results may provide a better understanding of present-day bacterial biogeography and the correlation between microbial communities and key environmental variables in a typical coastal area. Depending on these information, coastal resources could be eciently predicted, assessed and used.

Based on the high-throughput sequencing data of the 16S rRNA gene, microbial diversity analysis of 24 locations (sediment A and B) (Table S2) provided 417,896 high quality bacterial sequences, averaging between 7,083 and 64,605 sequences per sample.
Signi cant effects of environmental variables on the α-diversity of microbial communities were distinguished by correlation analysis (Table 1). Sediment salinity was signi cantly correlated with the Chao1 index (r = 0.254, P = 0.012), S index (r = 0.303, P = 0.005), and H' index (r = 0.234, P = 0.017). The Chao1 index (r = 0.188, P = 0.034) and S index (r = 0.246, P = 0.014) showed a positive correlation with the DO content of sea water collected near the sea oor. The Fe content (r = 0.236, P = 0.016) also showed a signi cant correlation with H' index. There were no signi cant correlations between other measured characteristics and indices.
NMDS analysis of the OTUs from the 19 locations showed a considerably high amount of scatter and signi cant differences in bacterial community composition between estuarine locations (Fig. 2b). Signi cant differences between estuary and marine sediment separated by salinity were also observed.
Individuals from marine sediment with high salinity tended to be clustered, indicating a high degree of microbial community similarity among marine sediment.
Environmental determinants of bacterial community composition RDA showed that sediment salinity, DO, pH, and χlf have strong effects (longer arrows) in shaping bacterial community composition (Fig. 2b). TOC exhibited an opposite trend to χlf. The rst canonical axis (RDA1) explained 22.2% of the variation and the second canonical axis (RDA2) explained a further 18.6% of the variation. These results suggest that sediment salinity, DO, χlf, TOC, and pH could be good predictors of bacterial community composition variation.
Mantel tests showed that β-diversity of the total community was signi cantly correlated with salinity (r = 0.680, P = 0.001), pH (r = 0.462, P = 0.001), DO (r = 0.528, P = 0.001), and χlf (r = 0.388, P = 0.006) ( Table 1), even when the geographic distance was controlled in partial Mantel tests (Table 1). Salinity, DO, pH, and χlf were the strongest environmental drivers of β-diversity, which was con rmed by both multiple regressions on a distance matrix and permutational analyses of variance on a distance matrix ( Table 2). As in the Mantel test results outlined in table 1, total community structure was driven mainly by salinity.
Because χlf values are coupled to iron-reducing bacterial activity [21], we also analyze their relation in correlation assay. Differences in χlf between sediment samples showed a close relationship to the relative abundance of total identi ed Iron-Reducing Bacteria (IRB) (all reads classi ed as potential iron reducers) [22,23] (Fig. 5b). Further, regression analysis showed that sediment χlf had a signi cant correlation with the relative abundances of total IRB (r² = 0.334, P = 0.013) (Fig. 5c). This implies that magnetism in the environment could affect the diversity of total IRB.

Effects of geographic distance on microbial community composition
The relationship between geographic distance and Bray-Curtis dissimilarity between stations showed a clear distance-decay pattern (R 2 = 0.0142, P < 0.001) (Fig. 6). This showed that community composition was remarkably correlated with geographic distance.
The composition of the microbial community is not just affected by the distance between the points, but also likely by the distance from the center of the magnetite. Sedimentary magnetite distribution in the gulf of Bohai Sea showed the highest contents of magnetite at site (120.4°E, 39°N) [10,11] located in the joint of three fault zones (the Tanlu, Liaocheng-Lankao, and Zhangjiakou-Penglai fault zones) within the Bohai Sea. In order to analyze the in uence of the presence of magnetite center on the microbial community structure, we analyze the relationship between the relative abundances of top 15 genera of 2 cross-sections samples ( Fig. 7a and b) and the distances from these samples to the center of magnetite.

Variation partitioning of microbial community composition
Variation partitioning analysis showed that environmental and spatial variables could explain 37.45% of the total variation in microbial community composition (Fig. 8). Pure environmental effects explained 29.41% of the variation (P < 0.001), which was markedly higher than pure spatial effects (3.03%, P < 0.05). Approximately 5.01% of the variation was attributed to spatially structured environmental variation (the fraction jointly explained by environmental and spatial factors). The residual 62.55% of the variation was unexplained.

Discussion
The important role of environmental variation in shaping bacterial α-diversity In our study, microbial α-diversity was highly correlated with salinity, DO levels and total Fe content. Previously, increased salinity has been re ected by decreased microbial activity in the surface sediments of the Qinghai-Tibetan lakes [3], corroborating our ndings that salinity shapes microbial diversity. DO was also documented to be the primary driving force in mine drainage habitats, related to metabolism associated with oxygen [24], and DO may play a crucial role in taxonomic diversity on a small (vertical) scale [5,25,26]. Thus, salinity and DO may have the same profound impacts on shaping microbial αdiversity in this study. Although total Fe content had a weaker effect than salinity and DO in our research regions, it is considered one of three factors which signi cantly affected the diversity index, implying the potential importance of Fe-containing mineral (eg. high χlf magnetite) for shaping microbial diversity in sediments.

Environmental variation in uenced microbial community composition
In this study, community composition was correlated with environmental factors, and was affected by salinity, pH, DO, and χlf. From these ndings we can infer that environmental variables may be crucial to shaping the microbial community composition. Salinity is well known to be a major contributor to microbial community structure and function [25,26]. The important role of salinity in bacterial communities has been found across globally heterogeneous environments, and in the sedimentary ecosystems of Hypersaline Laguna Tebenquiche [5]. To be consistent with those ndings, we observed the most signi cant correlations between salinity and bacterial community structure (Tables 1 and 2). Previous research has shown that bacterial communities in transition zones greatly vary geographically owing to sharp salinity gradients [27,28]. Thus, a possible reason for the in uence of salinity on bacterial distribution is that our research sites focused on bacterial communities distributed in a geographic area with a wide salinity gradient, including estuaries, coastal margins, and an open sea, where salinity values differ greatly. Additionally, all living things need energy and metabolize. Salinity is related to osmotic pressure, which changes the intracellular membrane structure and affects microbial energy cost and metabolic pathways [29,30]. In this study, different energy cost and metabolic pathways are more adaptable to different salinity. Then certain bacteria are selected and they survive in this region with de nite salinity. Taken together, these ndings suggest that salinity is a universal predictor of bacterial distribution.
After salinity, sediment DO concentration was also found to be critical in shaping microbial community structure ( Fig. 2c; Tables 1 and 2). The dominant phyla in this region were Proteobacteria and Bacteroidetes (Fig. 2a), which also represent the dominant phyla in the sediments of the eastern Mediterranean Sea [31]. In this study, Proteobacteria, Bacteroidetes, Chloro exi, Actinobacteria, and Cyanobacteria were strikingly correlated with salinity and DO (Figs. 3 and 4). The tendency of the relative abundances of these ve phyla to vary with salinity was opposite to their tendency to change with DO. The relative abundances of Proteobacteria and Bacteroidetes increased with elevated salinity and decreased with increased DO. This implied that the dominant phyla, Proteobacteria and Bacteroidetes, in the tested sedimentary regions may prefer sediment environments with high salinity and low DO. The relative abundances of Chloro exi, Actinobacteria, and Cyanobacteria decreased with elevated salinity and decreased with the declining DO. This indicated that these three phyla prefer to inhabit shallowestuary sediment with low salinity and high DO.
In this study, we also found the pH values of the Bohai Sea to be another important environmental variable in uencing bacterial community structure. In arctic soils [32] and lake sediments [4], pH has also been strongly related to the bacterial community structure. Here, the relative abundances of Planctomycetes, Actinobacteria, and Fibrobacteres decreased with elevated pH values. The relative abundances of WS3 and Caldithrix increased with elevated pH values. The pH values from the deepest water samples showed a signi cant correlation with these ve phyla, indicating that the pH value is a predictor of bacterial community structure. Therefore, these ndings indicate that environmental heterogeneity (including salinity, DO, and pH) was the predominant factor shaping the community composition of benthic bacteria in the Bohai Sea. pH would increase during the initial degradation of algal derived organic matter [33][34][35]. The RDA assay showed that TOC was positively associated with pH, possibly because surface sediments contain a higher proportion of labile algal derived aliphatic organic matter and more anions [33][34][35]. Meanwhile, sea sediments contain considerable clay. Restricted aeration in clay reduces the rate of organic matter oxidation and protects it from decomposition [36]. These indicated that the studied sediment in our study might contain lots of fresh organic matter with little decomposition because of limited aeration.
In our study, Proteobacteria was the most abundant group of bacteria. When surface sediments contain higher proportions of organic matter, Proteobacteria and Bacteroidetes are often prominently detected during the initial degradation of algal derived organic matter in marine sediments [33][34][35]. The dominant members of Bacteroidetes in the surface sediments of our study site have been consistently enriched, as seen in previous studies [34,37]. Because Cyanobacteria is documented to be the main participant and contributor to primary productivity of the global carbon cycle [38,39]. The occurrences of Bacteroidetes and Cyanobacteria in the surface sediment suggest that they may better survive in areas rich in fresh organic matter.
In the current study, χlf affected community composition of many phyla including Firmicutes, Cyanobacteria, Actinobacteria, and Proteobacteria. Cyanobacteria showed a positive correlation with χlf.
That indicated χlf could affect the growth patterns of ecologically important species such as Cyanobacteria, the main participants of primary productivity in the global carbon cycle. In sediment, available Fe is momentous for Cyanobacteria growth and Fe oxide minerals have the largest release potential [40]. Therefore, growth of Cyanobacteria is heavily in uenced by Fe availability in all water bodies [41]. Then organic matter and the carbon cycle are also affected. But a negative relationship between TOC and χlf or Fe were observed in the RDA assay indicated that organic matter produced were consumed. Numerous bacteria that are dependent on organic matter produced by Cyanobacteria will be in uenced. χlf values are coupled to iron-reducing bacterial activity [21]. In our study, we also found relative abundances of iron-reducing bacteria genera changed with χlf values and were signi cant correlated with χlf values (Fig. 5). Thus, in Bohai Sea organic matter could stimulate microbial iron related metabolism, and in return, lead to a microbial-driven change in magnetic susceptibility [42,43] as showed by measurable χlf values.

The roles of dispersal limitation in conditioning bacterial biogeography
Of all measured environmental variables, only Fe was spatially correlated (Mantel tests, Table 4).
However, important drivers in this study (salinity, pH, DO, and χlf) were not signi cantly spatially correlated. These results indicated that local environmental conditions were not shaped spatially.
In addition, despite the fact that the magnetite content in the center of the Bohai Basin is high, the magnetic susceptibility does not show a correlation with this distance, which may be because sedimentary settling happens in a vertical manner from surface to deep, whereas magnetic susceptibility is determine as iron minerals form based on location on the Earth, in relation to magnetic north. But in top 15 genera (Fig. 7) ve genera Lactococcus, Clostridium, Caulobacter, Gillisia and Sphingomonas showed a clear correlation with the distance from the center of the magnetite (Table 3). This implies that the presence of magnetite might shape the geographical distribution of these genera, and most likely by affecting the iron-related geochemical cycle these genera participate in. Lactococcus is reported that it could participate in Fe (III) reduction during the external electron transfer mediated by sodium anthraquinone-2,6-disulphonate (AQDS) [44]. The mechanism may be possible that Lactococcus sp. uses a very small portion of regenerated reducing power NADH for the reduction of external electron acceptor Fe (III) to Fe (II) in anaerobic lactic fermentation [45]. Then Lactic acid produced by Lactococcus was used by Clostridium for Fe (III) reduction, because Clostridium could act as a lactic fermenter and Fe reducer [46,47]. Caulobacter is known to be involved in metal oxidation [48] by biosorption and metabolism against Fe [48]. That is one of the reasons why Caulobacter distribution is also affected. Sphingomonas during the decay of Cyanobacteria was identi ed as a microcystin-degrading bacterium [49]. Thus, its distribution might be adjusted according to iron presence. Gillisia was detected as siderophore producer in seawater or sand samples [50]. Siderophores are the metal-chelating agents that primarily function to capture the insoluble ferric iron from different habitats [50,51]. Numerous bacteria could not produce siderophores but have siderophore acceptors [50,51]. Gillisia might assist other bacteria such as Cyanobacteria, Lactococcus, Clostridium and Caulobacter without siderophore generation capability for bioavailable iron release or absorption from Fe-containing minerals. Therefore, its geographical distribution also was impacted by the presence of magnetite. Of course, further work is needed to con rm this.
Environmental variables explained 29.41% (P < 0.001) of the total microbial community composition variation higher than spatial factors 3.03% (P < 0.05) in a variation partitioning analysis (Fig. 8) indicating their signi cant and dominant contribution.
In the present study, dominant OTU patterns and whole community composition were markedly (P < 0.001) correlated with geographical distance (Fig. 6). The results of the distance-decay pattern indicated that dispersal limitation may be another in uential factor driving microbial biogeography. Dispersal could eliminate the distance-decay relationship by counteracting microbial compositional differentiation [52]. Limited dispersal should strengthen the distance-decay relationship [52], and the strength of correlation between dispersal limitation and microbial community composition relies on geographical distance [53] and organism size [54]. Limitations of microbial dispersal have been demonstrated at large [55] or intermediate (10-3000 km) spatial scales [53]. Dispersal limitation may exist in intermediate spatial scale the Bohai Sea (approximately 100 km). Powerful dispersal limitations exist with increased cell size [56], the bacteria in the current study occurred within a relatively narrow size range, from 2-5 µm. This could help explain why the distance-decay curve inclined slightly, which is evidence of community variation purely constrained by spatial factors (3.03%) (Fig. 8). This demonstrated that dispersal limitation was associated with microbial community composition, but was not the dominant factor in shaping microbial biogeography in the Bohai Sea.
A large unexplained fraction (62.55%) was still in the variation partitioning analysis, and included unmeasured environmental variables, local arti cial effects, and random factors. However, it is unclear how these processes work as portions of variation.

Conclusions
Salinity was the crucial environmental factor for bacterial α-diversity in riverine and marine surface sediment around and in the Bohai Sea. Salinity, DO, pH, and χlf were strikingly correlated with bacterial community composition in this region. Dispersal limitation made an important contribution in shaping the bacterial biogeographical pattern. Environmental factors (explained 29.41% of the total microbial community composition variation) played a more important role than spatial variables (explained 3.03%) in varying microbial community composition.The presence of magnetite shapes the geographical distribution of ve genera Lactococcus, Clostridium, Caulobacter, Gillisia and Sphingomonas by affecting the iron-related geochemical cycle these genera participate in.
Our study makes a signi cant contribution to the literature because it not only provides novel information about key environmental variables that in uence the microbial distribution and community structure in a typical coastal area, it examines the effect of magnetic in uence on microbial community composition as well in an area with special geological structures in earthquakes natural fault lines. Our work provided valuable information for studying the relationship between geological fault lines and microbial activities.

Study area, sampling, and environmental variables
Within the study region, we selected six large rivers, namely the Daliao River, Liao River, Liugu River, Shi River, Yellow River, and the Sha River. Fourteen marine sediment samples were taken at various locations across the sea to capture the areas in uenced by these rivers, alongside a further 10 riverine sediment samples (Fig. 1a). These samples were used for the main microbial community analysis. In addition to these samples, a further 124 marine sediment samples were collected across the Bohai Sea to spatially map DO, pH, salinity, and low-frequency magnetic susceptibility (χlf) (Fig. 1b).
During peak rainy season (Aug 23-29, 2014), marine surface sediment (0-20cm depth) were collected using a stainless steel grab sampler deployed off the side of the "Yi Xing" research vessel. Riverine sediment (0-20cm depth) was collected using a stainless corer from areas of extensive sediment deposition. Representative samples were achieved by mixing three independent subsamples collected within a 5 m 2 area. All sediment samples were split into two parts and stored at -20 °C immediately after collection. Upon coming back to the laboratory, one part of each sample was stored in the dark and freeze-dried prior to chemical analysis. The other was stored at -80 °C until DNA extraction.
Throughout the survey, a global positioning system was utilized to map all the sampling sites. The temperature, DO, pH, and salinity measurements for the seawater overlying each sediment sample were obtained in situ using a SBE 25plus Sealogger CTD (Sea-Bird Scienti c Ltd., Bellevue, WA, USA) tted with 10 L Niskin bottles. The biogeographic data and kriging maps were analyzed with ArcGIS v.10.0 spatial analyst tools (ESRI Inc., Redlands, CA, USA).

Chemical analysis and magnetic characterization
Chemical analyses were carried out on the Analysis and Testing Center of the Yantai Institute at Coastal Zone Research, Chinese Academy of Sciences. After digestion with aqua regia, sediment total iron content was determined using an ELAN DRC II ICP-MS (PerkinElmer, Waltham, MA, USA). After pretreatment with 1 M HCl to remove carbonates, total organic carbon (TOC) and total nitrogen (TN) were determined using a Vario MACRO cube elemental analyzer (Elementar Analysensysteme GmbH, Langenselbold, Germany). χlf was measured using a MS2B magnetic susceptibility meter (Bartington Instruments Ltd., Witney, UK). High-throughput sequencing data processing PCR and amplicon library preparation for high-throughput sequencing was performed as previously reported [32,57]. The bacterial 16S rRNA gene data were processed using the Quantitative Insights Into Microbial Ecology (QIIME) 1.9.1-dev pipeline [58] (http://www.qiime.org) with the default parameters, unless otherwise noted according to procedures in a previous study [10]. We assessed microbial αdiversity using four metrics caculated with R software v3.4.4 (https://www.r-project.org): the Shannon index (H') [59], the Chao1 index [60], and the observed OTU richness (S) according to previously published procedures [10].

Statistical analyses
Redundancy analysis (RDA) and non-metric multidimensional scaling (NMDS) were used to relate community structure and sediment/water properties for the different locations. Regression analysis was conducted using Origin v8.1 software (OriginLab Corp., Northampton, MA, USA). The Spearman's rank correlations between environmental variables and geographic distance were derived from a Mantel test with 999 permutations [61]. Simple and partial Mantel tests (based on 999 permutations) were analyzed the relationship between environmental variables (Euclidean distance) and community composition (Bray-Curtis dissimilarity) with controlled geographic distance [62].
The contribution of each environmental variable on community composition was separately calculated using a permutational multivariate analysis of variance (PERMANOVA, 'adonis' function in vegan R package with 9999 random permutations), and multiple regression on distance matrices (MRM, 'MRM' function in ecodist R package with 9999 permutations) based on Bray-Curtis dissimilarity [63].
The relative importance of spatial and environmental variables in driving the microbial community was determined by variation partitioning [64]. The roles of spatial factors were estimated by the principal coordinates of neighbor matrices (PCNM) method [65]. We used the 'pack for' library forward-selection procedure [66] to select environmental variables. Then, the variation of the community composition was partitioned between the extracted PCNM spatial variables and selected environmental variables using a partial redundancy analysis, which decomposed community composition into fractions explained by pure environmental variables, pure spatial factors (PCNM variables), spatially structured environmental variation (shared fraction), and unexplained variation. All statistical analyses were performed using R software v3.4.4 (https://www.r-project.org).

Availability of data and materials
The datasets generated during and/or analyzed during the current study are available from the corresponding author on request.
Authors' contributions LC performed the experiments, analyzed the data and wrote the paper. YL, MW, WS and JT carried out the microbial community analysis. DLJ participated in the revision of manuscript. FL designed the experiments. All authors read and approved the nal manuscript.
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.             Table S1. c Regression analysis in between the relative abundance of identi ed DIRB genera and χlf .  Table S1. c Regression analysis in between the relative abundance of identi ed DIRB genera and χlf .

Figure 6
Relationship between the Bray-Curtis similarity of the microbial community and geographic distance between sampling stations. The solid red line indicates the t between geographic distance and Bray-Curtis similarity.

Figure 6
Relationship between the Bray-Curtis similarity of the microbial community and geographic distance between sampling stations. The solid red line indicates the t between geographic distance and Bray-Curtis similarity.  Table 4.  Table 4. Variation partitioning of bacterial community composition in coastal sediments. The explanatory power of the pure and shared fractions of environment (Env), and spatial factors are indicated as adjusted R2.
ANOVA tests were carried out on the variation explained by the pure fraction. and represent statistically signi cant differences of P < 0.001 and < 0.05.

Figure 8
Variation partitioning of bacterial community composition in coastal sediments. The explanatory power of the pure and shared fractions of environment (Env), and spatial factors are indicated as adjusted R2.
ANOVA tests were carried out on the variation explained by the pure fraction. and represent statistically signi cant differences of P < 0.001 and < 0.05.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.