Illumina sequencing‐based analysis of sediment bacteria community in different trophic status freshwater lakes

Abstract Sediment bacterial community is the main driving force for nutrient cycling and energy transfer in aquatic ecosystem. A thorough understanding of the community's spatiotemporal variation is critical for us to understand the mechanisms of cycling and transfer. Here, we investigated the sediment bacterial community structures and their relations with environmental factors, using Lake Taihu as a model system to explore the dependence of biodiversity upon trophic level and seasonality. To combat the limitations of conventional techniques, we employed Illumina MiSeq Sequencing and LeFSe cladogram to obtain a more comprehensive view of the bacterial taxonomy and their variations of spatiotemporal distribution. The results uncovered a 1,000‐fold increase in the total amount of sequences harvested and a reverse relationship between trophic level and the bacterial diversity in most seasons of a year. A total of 65 phyla, 221 classes, 436 orders, 624 families, and 864 genera were identified in the study area. Delta‐proteobacteria and gamma‐proteobacteria prevailed in spring/summer and winter, respectively, regardless trophic conditions; meanwhile, the two classes dominated in the eutrophication and mesotrophication lake regions, respectively, but exclusively in the Fall. For LEfSe analysis, bacterial taxon that showed the strongest seasonal or spatial variation, majority had the highest abundance in spring/summer or medium eutrophication region, respectively. Pearson's correlation analysis indicated that 5 major phyla and 18 sub‐phylogenetic groups showed significant correlation with trophic status. Canonical correspondence analysis further revealed that porewater NH 4 +‐N as well as sediment TOM and NO x‐N are likely the dominant environmental factors affecting bacterial community compositions.

. Because the spatial and temporal distribution of these microbes is controlled by physiochemical conditions of the sediments, temperature, nitrogen level, and organic matter in particular (Haller et al., 2011;Song, Li, Du, Wang, & Ding, 2012), a shift in sediment bacterial communities can provide important insights into environmental changes in the local ecosystem.
Bacterial community is characterized by its structure and biodiversity which have been well studied so far by conventional experimental techniques such as polymerase chain reaction-denaturing gradient gel electrophoresis (PCR-DGGE) and clone library techniques. These investigations helped to establish a broad understanding concerning microbial community's temporal and spatial distribution patterns.
For example, a PCR-DGGE/clone library study of the bacterial community in Sitka stream, Czech Republic found that most of the mcrA gene clones showed low affiliation with known species and probably represented genes of novel methanogenic archeal genera/species (Rulik et al., 2013). Another research in the Yangtze Delta (Huang, Xie, Yuan, Xu, & Lu, 2014), using the same technique found the number of total cultivable bacteria in an estuary reservoir was significantly lower than that of the main river. Despite the advancement, the knowledge obtained by these studies may have its limitation because the low-throughput methods employed often underestimate the overall diversity and lack the ability to detect rare species in complicated environmental systems. For example, Berdjeb, Pollet, Chardon, and Jacquet (2013) used the similar methods to examine the archaeal community structure in two neighboring peri-alpine lakes of different trophic status but found no spatiotemporal dynamics in their study area, suggesting the potential inadequacy of the conventional techniques to probing the complexity of biodiversity and community structure in natural environments.
Compared to the conventional methods, high throughput sequencing has the advantage of being able to generate multi-million sequences and thousands of Operational Taxonomic Units (OTUs) in environmental samples. For example, Conrad et al. (2014) used pyrosequencing to obtained more than 1000 bacterial OTUs in the sediment of Amazon region and found that rewetting of the sediments resulted in a dramatic increase of the relative abundance of Clostridiales. The chosen study area, Lake Taihu (2,338 km 2 ), is highly heterogeneous in the trophic levels due to the difference in river input to different regions. As such, the water body in the lake can be divided into different ecological types based upon trophic status and plankton community structure. Spatial variation of bacterial communities in the lake sediments was documented by a number of researchers but no consensus has been reached so far. For example, Liu et al. (2009) reported the absence of Actinobacteria in the eutrophied area of the lake, but was contradicted by Chen et al. (2015), where the authors detected as much as 5% abundance for this phylum. Similar inconsistency can be found for the spatial distribution of Cyanobacteria, alpha-proteobacteria, and Planctomycetes upon comparing the results by Shao et al. (2011) andChen et al. (2015).
On the vertical dimension, Ye et al. (2009) reported similarity of bacterial communities in different layers of sediments taken from Meiliang Bay, but Shao et al. (2013) in a later work discovered the variation of bacterial community and an overall decrease of biodiversity with depth in Meiliang Bay. A literature review indicates such disagreement may have originated largely from the limitations of the clone library because most of these previous studies employed the conventional analytical techniques. In this study, we assessed the sediment bacterial community in a lake with known trophic gradient used a high-throughput sequencing method (Illumina MiSeq) to circumvent the technical limitations of the traditional methods.
For data processing, we used Linear discriminative analysis Effect Size (LEfSe) to recover the spatiotemporal variations of the bacterial community. The aim of this study is to dissect the bacterial community, using the high-throughput sequencing technique (1) to determine the relations of sediment bacterial taxa with the trophic status of the lake water and sedimentary environmental factors (2) and to provide powerful evidences for further elucidation of the nutrients cycle and accumulation mechanism driven by bacteria in aquatic ecosystem.

| Sampling site and procedure
The study area ( Figure 1) is located at the north to east side of the Lake Taihu with total nitrogen decreasing from Meiliang Bay (region A-1, north), to Gonghu Bay (region A-2, northeast), and finally to Xukou Bay (region A-3, east). Area A-1 is highly enriched in nutrients and has frequent algal blooming incidents. In contrast, the low nutrient waterbody in A-3 is characterized by submersed vegetation and diverse communities of fishes and invertebrates and, in fact, is a drinking water source for local communities. The water in A-2 was similar to that in A-1 till about 15 years ago but has since improved its quality due to the interference of the local government.
Sample collection was carried out in the Fall of 2014, and in Winter, Spring, and Summer of 2015. For sediments, loose sediment samples in the depth of less than 5 cm was collected using a 1/16 m 2 Petersen grab sampler. Triplicate samples from three separate grabs were homogenized to generate one composite sample in each sampling site. Water samples were taken together at the same locations.
All samples were immediately stored in an icebox and transported back to the lab within 3 hr. Once in the lab, an aliquot of the sediment samples was placed in a 15 ml sterile centrifuge tube at −80°C until DNA extraction was carried out. The remaining portion was further processed (freeze-dried to collect sediment particles, and centrifuged to collect the pore water) for physicochemical analyses.

| Physicochemical analyses
Seventeen physicochemical parameters of the overlying water, pore water, and freeze dried sediments were analyzed ( Table 1).

| DNA extraction and purification
Total genomic DNA of each sediment sample was extracted using   (Caporaso et al., 2012). The conditions for amplification are as follows: 95°C for 2 min; 27 cycles of 95°C for 30 s, 55°C for 30 s, followed by 72°C for 45 s, with a final extension 72°C for 10 min. The PCR products were gel-purified, using the UltraClean PCR Clean-Up Kit (Mo Bio laboratories) and quantified, using a Qubit system (Invitrogen). Equimolar amounts of purified amplicons were pooled and stored at −20°C until sequenced. Library construction and sequencing were performed commercially (Beijing Genomics Institute).

| Sequences data analyses
Illumina sequence reads were processed using MOTHUR version 1.27.0 (Schloss et al., 2009). Briefly, upon completing sequencing by the Illumina MiSeq platform, the reads from the original DNA fragments were merged, using FLASH (V1.2.7, http://ccb.jhu.edu/software/FLASH/), and quality filtering of reads was performed according to the literature (Caporaso et al., 2011). Chimeric reads were removed by checking against a chimera-free database of 16S rRNA gene sequences, using UCHIME (DeSantis et al., 2006). Sequences were assigned to the OTUs with a maximum distance of 3%, using MOTHUR (Schloss et al., 2009). Community diversity indices and rarefaction curve of each sample were generated, using the UPARSE pipeline (Edgar, 2013). The RDP classifier was used to assign taxonomic identity to the representative sequence for each OTU. T A B L E 1 Physicochemical parameters of the overlying water, pore water, and freeze dried sediments

| Statistics analysis
The Trophic Status Indices (TSI) (Aizaki, 1981) of all sampling sites were calculated using the measured Chl-a, W-TP, W-TN, COD, and SD by the following expression: where TSI(∑) is the completed TSI; w j is the relative weight of TSI of the j parameter; and TSI (j

| Physicochemical properties of the samples
Measured TSI in the study area decreased in the direction of A-1,

| Bacterial community structures via Illumina MiSeq sequencing
The similarity of sediment bacterial communities within individual lake regions was first analyzed by PCR-DGGE. The dendrograms ( Figure S1) indicated that the communities in each region can be grouped into 2 defined clusters corresponding to winter and summer.
Each cluster can be further divided into two sub-clusters. Guided by this understanding, we selected two sites in each region, one from each sub-cluster, and performed further analysis by Illumina MiSeq sequencing.
A total of 1,918,768 high quality sequences (average length 253 bp) were obtained by Illumina MiSeq sequencing at which the rarefaction curves of Shannon diversities approached a plateau, suggesting a complete capture of the bacterial community at each site.
Based on a 97% sequence similarity cutoff, these sequences yielded a bacterial OTU number that ranged from 2279 to 4331 (Table S2).
Of the three regions, the sites in A-2 showed the highest diversity in Fall, while the sites in A-3 reached a peak in the other three seasons. Seasonality-wise, the lowest diversities were observed in fall and winter.  (Table S6). At the genera level, Acinetobacter, with the relative abundance ranging from 0.01% to 61.8%, was the most dominant. GOUTA19 and LCP-6 were the other abundant genera and were present in all sediment samples with the relative abundance of 0.5-6.0% and 0.8-5.8%, respectively (Table S7).

| The spatial-temporal distribution of bacterial communities
Spatial-temporal variation of the bacterial community can be evaluated either by a direct comparison of the relative abundance of individual taxa, or by LeFSe algorithm. Direct comparison found that, for the dominant phyla Proteobacteria, its major classes varied greatly with trophic status and seasonal change. For example, delta-proteobacteria and gamma-proteobacteria prevailed in spring/summer and winter, respectively, regardless trophic conditions; meanwhile, the two classes dominated in the eutrophication and mesotrophication lake regions, respectively, but exclusively in Fall ( Figure 5 and Table S4).
The strongest seasonal dependence may be manifested by gammaproteobacteria whose abundance showed a greatest decrease from winter and fall to summer and spring ( Figure 5). The spatial variation may be exemplified by the behavior of Planctomycetes, Chloroflexi, and For those that showed the strongest seasonal variation, the majority had the highest abundance in spring and summer (Table 2); for those that showed the strongest spatial variation, a majority had the highest abundance in region A-1 (Table 3). Some bacterial taxonomy levels (from phylum to family or genus levels) had consistent variation among F I G U R E 5 The temporal and spatial variations characteristics of bacterial community structure in different bacterial taxonomical levels (only shown the sequence of bacteria >1% of all sequences). The size of the circle represented the relative abundance of bacteria at each site, and the color of the circle represents bacterial taxonomical levels, red is phylum, blue is class, black is order, family and genus

| Relationship between bacterial community structure and environmental variable
The overall level of biodiversity in the study area appeared to be

| DISCUSSION
The bacterial OTUs and Shannon diversity obtained in the present study are more than two orders of magnitude and twofold higher than the results acquired via low-profiling biology techniques for the same/ similar eutrophication lakes (Zhao et al., 2013;Szabó et al., 2011), which is similar to that found by the high-throughput pyrosequencing method , suggesting that high-throughput

| The characteristics of bacterial community structure
The bacterial communities observed in this study were dominated by gamm-, delta-, beta-proteobacteria, a pattern similar to those found in soils (Liu, Zhang, Zhao, Zhang, & Xie, 2014) and other fresh water lake sediments , but distinct from those found in salt water lake sediments (Xiong et al., 2012) and marine coastal waters (Fortunato et al., 2013). It is known that the phylum Proteobacteria might be involved in a variety of biogeochemical processes in aquatic ecosystems (Zhang, Zhang, Liu, Xie, & Liu, 2013;Liu et al., 2014).
For example, numerous studies, either through conventional approach or high-throughput method, have shown the predominance of Proteobacteria in sediments of various lakes, with a large shift in the composition of major classes and relative proportions (Ye et al., 2009;Haller et al., 2011;Song et al., 2012;Bai et al., 2012). At the class level, both gamma-proteobacteria (Sinkko et al., 2013;Liu et al., 2014) and delta-proteobacteria (Rodionov, Dubchak, Arkin, Alm, & Gelfand, 2004;Lehours, Evans, Bardot, Joblin, & Gerard, 2007) were observed to occur in organic-rich lacustrine sediments. Beta-proteobacteria, a major class in most of the samples in this study, occurs almost exclusively in freshwater environments (Hempel, Blume, Blindow, & Gross, 2008) and is seen as the most abundant group in the sediments of eutrophication lakes (Bai et al., 2012).
The predominance of Proteobacteria's class and the observed strong correlation between these bacteria and nitrogen conversion (Table 4) in the present study suggest that they were actively involved in the functioning and processes of lake sediment ecosystems (Song et al., 2012). Numerous studies point to a linkage between nitrogen conversion with Proteobacteria's class. For example, Zhang et al.

| Spatial and seasonal variations in bacterial community structure
The spatial variation of bacterial community is characterized by the dominance of delta-proteobacteria in the eutrophication regions (A-1 and A-2) and gamma-proteobacteria in the mesotrophication region (A-3) in Fall. Such pattern might be due to the regional differences in the sediment organic matters at different trophic levels. In F I G U R E 6 Cladograms indicating the phylogenetic distribution of bacterial lineages associated with the 4 seasons of a year. The phylum, class, order, family, and genus levels are listed in order from inside to outside of the cladogram and the labels for levels of order, family, and genus are abbreviated by a single letter. The green, blue, red, and purple circles represent the bacteria enriched in the sediment of spring, summer, fall, and winter, respectively, whereas the yellow circles represent the taxa with no significant differences between 4 seasons of a year | 11 of 15 WAN et Al.
the mesotrophication region, the sediment organic matter is derived mainly from decomposing and dead residues of large vascular plants; in comparison, the sediment organic fraction of the eutrophication regions originated primarily from the organic remains of algae (Qin, Xu, Wu, Luo, & Zhang, 2007). Our results differ from previous research (Shao et al., 2011) where the authors reported that delta-proteobacteria was the prevailing class in the macrophyte-flourishing areas while

| High abundance bacterial phyla at eutrophic conditions
The high abundance of five phyla at the eutrophication region may be an indication that these microbes have specific nutritional or environmental preference. For example, the observed relation between TP and Chloroflexi assemblage (Table 4), along with previous studies in a different lake (Song et al., 2012), may suggest a possible role of phosphorus in promoting the growth of Chloroflexi. In addition, it was reported that this phylum was a predominate taxa (57-82%) in the sediment of copper mine (Lucheta, Otero, Macias, & Lambais, 2013). Following this lead, we hypothesize that the high abundance of Chloroflexi in region A-1 may be due to the discharge of phosphorus and heavy metal-containing industrial wastewater in this area. The observed high TP concentration in the overlying water and sediment in region A-1 (Figure 2) provide additional support for this view point.
For Verrucomicrobia, the high abundance may be due to the prosthecate morphology of these bacteria which renders a unique ability for nutrient uptake (Zwart et al., 1998). Verrucomicrobia, which was able F I G U R E 7 Cladograms indicating the phylogenetic distribution of bacterial lineages associated with the sediments of 3 lake regions. The phylum, class, order, family, and genus levels are listed in order from inside to outside of the cladogram and the labels for levels of order, family, and genus are abbreviated by a single letter. The green, red, and blue circles represent the bacteria enriched in the sediment of Meiliang Bay (A-1), Gonghu Bay (A-2), and Xukou Bay (A-3), respectively, whereas the yellow circles represent the taxa with no significant differences between the sediments of 3 lake regions  to take advantage of nutrient-rich environments, had been found in eutrophic ponds and lakes such as those in recreational parks where visitors feed waterfowl (Schlesner, 2004). Chlorobi are photosynthetic bacteria and hence require the presence of adequate light penetration in water (Vila, Abella, Figueras, & Hurley, 1998). On the contrary, Region A-3 is teemed with a great many submersed vegetation or aquatic plants. The dense leaves of macrophyte, in particular, can effectively block the transmission of light to the surface of sediments, resulting in an opaque condition that leads to slow growth for Chlorobi. Nitrospirae is a known significant group related to the nitrite oxidation in freshwater lake sediments (Bartosch, Hartwig, Spieck, & Bock, 2002). Consequently, these bacteria will flourish in high nitrogen condition such as regions A1 and A2. Lastly, positive relation between Planctomycetes and eutrophication may be understood from genome analysis (Gloeckner et al., 2003) which revealed the microbes' ability to derive energy from the degradation of sulfated polysaccharides of algal origin. Planctomycetes was present at high levels in diatom blooms (Morris, Longnecker, & Giovannoni, 2006)

| Factors affecting bacterial community structures
Agreeing with previous studies in other similar eutrophication freshwater lake (Zeng et al., 2008;Dang et al., 2010;Macalady, Mack, Nelson, & Scow, 2000), CCA results from our analyses (Figure 7) showed that pore water NH 4 + -N as well as sediment TOM and NO  (Zhong et al., 2015). The content of TOM is more than 45 g kg −1 in winter and less than 20 g kg −1 in other three seasons (Figure 2), the increase of TOM in winter directly influence the abundance of Cloacibacterium (Figure 7), because these bacteria participate in organic matter degradation, and TOM provide nutrient for the growth of Cloacibacterium (Bauer et al., 2006).

| CONCLUSIONS
High throughput Illumina MiSeq sequencing method was used to investigate the biodiversity and bacterial community structure in Lake Taihu. More than 1,910,000 sequences were analyzed in the context of changing environmental conditions to evaluate the impact of trophic status on bacterial community, and the results showed significant correlation with trophic status in 5 major phyla and 18 sub-phylogenetic groups. Findings from this investigation can be summarized as follows: 1. The diversity of bacterial community is inversely related to the trophic levels of water body in most seasons of a year.

2.
The bacterial taxa, delta-proteobacteria and gamma-proteobacteria, that dominated, respectively, in the eutrophication and mesotrophication regions showed the strongest seasonal variation.