The correlation analyses of bacterial community composition and spatial factors between freshwater and sediment in Poyang Lake wetland by using artificial neural network (ANN) modeling

Abstract

As one of the most important components of the lake ecosystem, microorganisms from the freshwater and sediment play an important role in many ecological processes. However, the difference and correlation of bacterial community between these two niches were not clear. This study investigated the diversity of microbial community of freshwater and sediment samples from fifteen locations in Poyang Lake wetland. The correlation between the bacterial community and physicochemical property of Poyang Lake wetland was analyzed by artificial neural network (ANN). Our results demonstrated that the freshwater and sediment bacterial community were dominated by groups of the Bacteroidetes (23.33%) and β-Proteobacteria (22.54%) separately, whereas, Canalipalpata, Bacillariophyta, Gemmatimonadetes, and Verrucomicrobia were detected in freshwater niches only. Phylogenetic analysis further indicated that bacterial composition in freshwater significantly differed with the sediment niches. There are 34 unique species accounted for 85% in fresh water samples and 28 unique species accounted for 82% in sediment samples. Cluster analysis further proved that all the samples from freshwater niches clustered closely together, far from the rest sediment samples. ANN analysis revealed that the freshwater with high N and P nutrients will greatly increase the diversity of the bacterial communities. In general, both environmental physicochemical properties, not each factor independently, contributed to the shift in the bacterial community structure. The five tributaries (Gan, Fu, Xin, Rao, Xiu Rivers) play a vital role in shaping the bacterial communities of Poyang Lake. This study provides new insights for understanding of microbial community compositions and structures of Poyang Lake wetland.

Introduction

Wetland ecosystem, an integral realm, consisted with water, sediment, aquatic plant, plankton, and microorganism, which is considered the most biologically diverse ecosystem [1]. The river-lake network is an important channel in the transformation of terrigenous substance into the wetland ecosystem [2]. It also becomes a natural accumulation hotspot of nutrients due to the violent change of hydrological condition and peculiar characteristics of sedimentary environments [3]. Moreover, the accumulation and fast circulation of nutrients contribute to the high biological productivity and biodiversity in river-lake wetland ecosystem [4]. The microbial community is the foundation of biogeochemical cycles in wetland ecosystems [5]. Specifically, microorganisms are among the most important contributors to the transformation of complex organic compounds and minerals in wetland [6], serve as a source of food for consumers, and assimilate inorganic nutrients [7, 8]. Overall, exploring the correlation between microbial communities and environmental factors in river-lake wetland ecosystem is meaningful to understand the original transformation process of nutrients and the succession of microbial community.

Poyang Lake, the largest freshwater river-lake wetland in China, is one of the most important wetland nature reserves in the world. It receives about 87% of water volume from five tributaries (Xiushui, Raohe, Ganjiang, Fuhe, Xinjiang), which constitutes five river-lake systems [9]. The lake offers important spawning grounds and food for various fishes, which is also the natural habitat of migratory birds [10]. It is important for water supply, flood control, tourism, recreation, shipping, and aquaculture. However, nowadays, due to human disturbance and the increasingly polluted inflow rivers, the water qualities deteriorated gradually, which posed a serious threat to the stability of the overall ecological environment around Poyang Lake [11]. Abundant nutrients, various bacteria, and high metabolic activity of living micro-organisms made the Poyang lake wetland hot issue in biogeochemical cycling and natural ecological balance [12]. The microorganisms living in Poyang lake wetland are ubiquitous and play key roles in ecosystem functioning, including cycling of biologically active elements. Several investigations on the water quality and biodiversity of fish, birds, and aquatic plants have been conducted [13, 14]. However, systematic exploration of bacterial structures through the simultaneous consideration of stereoscopic spatial distribution (fresh water layer and sediment layer) is relatively limited, resulting in a poor understanding of how environmental factors shape bacterial communities in Poyang Lake wetland ecosystems. Consequently, it is of great interest to disclose the function of environmental parameters on microbial communities’ stereoscopic spatial distribution in Poyang river-lake wetland ecosystems.

Recently, analyses of bacterial community structure that do not depend on cultivation have been widely utilized [15], enabling researchers to analyze multiple samples simultaneously, and many studies have been conducted to obtain outlines of the bacterial communities associated with environmental, seasonal, spatial, and geographical variability [5]. Several preliminary studies have been reported on the composition and dynamics of bacterial communities in Poyang and try to explore the relationships between bacterial community and environment factors [1, 16–18]. However, few sampling sites in the previous studies were incomplete to represent the real status of bacterial community. Furthermore, the level of microorganism community composition and possible regulating factors in this lake were under-evaluated. Most of the research focused on the bacterial communities of water [18, 19] or sediment [1, 2, 16, 17] separately; little is currently known about the comparison of the microbial diversity and activity between the freshwater and sediment ecological niches. Therefore, the deeper study needs to be explored to understand the relationships of microbial community and environmental parameters between the freshwater and sediment niches.

As many different environmental variables appear to influence the structure of microbial communities simultaneously, analyzing the relationships between environmental variables and microbial composition is a difficult challenge due to the large numbers of irrelevant variables [20]. Artificial neural network (ANN) is a powerful method that can be trained to extract nonlinear patterns that exist in large complex data sets without requiring a priori hypotheses to guide the model development. These models may thereafter be utilized to predict how different combinations of variables may affect microbial community structures [21]. In this study, we describe the application of ANN modeling methods for determining the relationships between bacterial diversity indices of microbial communities from sediment and freshwater and their correspondence with environmental chemical variables that shape the community structures.

To better understand the importance of the microorganism in Poyang lake wetland, we investigated the spatial heterogeneity of bacterial communities from freshwater layer and sediment layer simultaneously. The main objectives of this study were as follows: (1) to investigate the environmental parameters and nutrient level in different river-lake systems, (2) to determine the bacterial diversities and community compositions in freshwater and sediment niches, (3) to explore the relationship between environmental parameters and the diversity and structure of bacterial communities. This work would provide beneficial information for further understanding of the aquatic ecosystem structure and function of Poyang Lake.

Materials and methods

In this study, we performed comparative analyses of the bacterial diversity from fifteen sampling sites and explored the relationship of the bacteria from freshwater layer and sediment layer in Poyang Lake wetland. By directly extracting the environmental DNA from freshwater and sediment and conducting 16S rDNA-DGGE analysis, we described the geographical succession in the bacterial community structure and illustrated environmental factors regulating their diversity and spatial distribution by ANN analysis.

Site description and sampling

Poyang Lake, as the largest fresh water lake in China, is located in the northern part of the Jiangxi province. It is a fluvial lake formed about 6000 years ago. Poyang Lake covers 3585 km² during the rainy season (28° 22′-29° 45′ N and 115° 47′-116° 45′ E), with a maximum depth of 29.2 m and an average depth of 5.1 m. It receives inflows from five tributaries, namely Gan, Xiu, Fu, Xin, and Rao Rivers, and discharges into the Yangtze River at Hukou (Fig. 1). However, the water in the Yangtze River may flow into Poyang Lake to reduce its flood discharge during heavy rainfall periods. The water level changes significantly over the course of a year, resulting in large variation of wetland area [9]. Being the largest natural lake under mesotrophic conditions, the water quality of Poyang Lake is classified as Class IV [11]. It plays an important role in the maintenance of the unique biota of the Yangtze floodplain ecosystem [22].

Fig. 1.

Location of the fifteen sampling sites of Lake Poyang

Poyang Lake can be divided into several regions on the basis of physical, chemical, and biotic characteristics. The fifteen sampling sites were chosen at the different representative regions in Poyang Lake wetland (Fig. 1). Fresh water samples were collected from surface water (top 20 cm) with 2-L sterilized plastic bottles. Sediment layer samples were collected simultaneously from the same sites (top 5 cm) with an Eckman grabber. Each sample was collected in triplicate. The physicochemical properties of the different sampling sites in Poyang Lake, such as coordinates, water depth, transparency, water temperature, and water turbidity, were measured on location (Table 1). The samples were immediately transported to the laboratory and processed. For molecular analysis, each sample was filtered through a 0.22-μm pore size filter (Millipore, USA). Filters were stored at − 80 °C until nucleic acids were extracted. Samples used for analysis of physical and chemical properties were then stored in a refrigerator at 4 °C until further processing within 1 month [5]. According to the methods reported previously [12, 16, 23], the physical and chemical properties were analyzed by Shanghai Micro-spectrum Chemical Technology Services Co., Ltd. Water temperature and pH were determined using a five-star portable multi-parameters measuring instrument (Orion, Co, USA). Transparency was measured by a Secchi disk. Chlorophyll a (Chl a) was analyzed by using hot-ethanol methods [24]. Total nitrogen (TN), total phosphorus (TP), ammonium (NH₄), nitrite (NO₂), and nitrate (NO₃) contents were analyzed according to APHA [25].

Table 1.

Freshwater physicochemical properties of the different sampling sites in Poyang Lake

No.	TRA	NTU	DO	CON	pH	ORP	PAR	CHL	TDS	SAL
1	0.20h	139.4d	0.28f	70.5hi	8.2cd	339d	760f	2.7f	0.045f	0.02bc
5	0.15i	174.8c	0.28f	80.2g	8.1cde	298g	2539a	2.2g	0.051ef	0.03bc
11	0.20h	117.0e	6.89a	351.6a	8.5b	363b	652g	1.1h	0.225a	0.17a
14	0.30g	51.4f	6.89a	110.7c	8.1cde	380a	132l	1.4h	0.070c	0.04b
17	0.30g	41.6g	4.55d	216.7b	7.4g	380a	347i	3.4e	0.139b	0.01c
21	0.10j	340.2a	6.33c	69.9hi	7.9ef	364b	20m	2.7f	0.044f	0.02bc
22	0.10j	328.2b	6.44b	68i	7.7f	364b	140l	2.7f	0.043f	0.02bc
47	0.90b	11.1i	0.24f	94.8d	8.9a	349c	326j	6.7a	0.060de	0.04b
54	0.60e	16.7h	0.25f	89e	8.5b	327f	824d	11.7a	0.056ef	0.03bc
80	0.80c	11.4i	2.29e	67.9i	8.3bc	348c	1582b	4.1c	0.043f	0.02bc
82	0.90b	16.7h	0.29f	94.9d	8.3bc	350c	1025c	3.8de	0.060de	0.04b
85	0.40f	50.6f	0.26f	72.2h	8.2cd	330ef	784e	4.1c	0.046ef	0.02bc
89	0.60e	16.6h	0.21f	94.3d	8.3bc	351c	259k	3.8de	0.060de	0.03bc
102	0.70d	18.8h	6.4bc	89e	8.1cde	346c	814d	3.4e	0.056ef	0.03bc
104	1.00a	8.6i	6.31c	84.3f	8.0de	335de	609h	6.1b	0.053ef	0.03bc
No.	TN (mg/L)		TP (mg/L)	NO3−-N (mg/L)		NH4+-N (mg/L)	NO2−-N (mg/L)		PO43−-P (mg/L)
1	1.293c		0.074c	0.575e		0.113d	0.012def		0.007e
5	1.160d		0.049de	0.430f		0.080e	0.016de		0.011d
11	1.484b		0.060d	1.907a		0.018h	0.023c		0.025b
14	1.076f		0.044ef	0.796c		0.079e	0.031b		0.015c
17	3.044a		0.157a	1.338b		0.722a	0.358a		0.074a
21	1.498b		0.110b	0.376g		0.182b	0.014de		0.004de
22	1.122e		0.099b	0.337h		0.158c	0.010efg		0.005efg
47	1.049g		0.038efg	0.318i		0.020h	0.017cd		0.001cd
54	0.792k		0.049de	0.208j		0.024h	0.012def		0.001def
80	0.815j		0.024g	0.081m		0.018h	0.005g		0.001g
82	0.923i		0.036efg	0.312i		0.022h	0.017cd		0.005ef
85	1.483b		0.044ef	0.664d		0.067f	0.012def		0.006ef
89	0.961h		0.029ef	0.128l		0.018h	0.007fg		0.002gh
102	0.955h		0.036efg	0.423f		0.021h	0.030b		0.002gh
104	0.751l		0.045e	0.195k		0.04g	0.014de		0.010d

For each parameter, values sharing the same letter are not significantly different (Duncan’s test, P < 0.05), (mean, n = 3). TRA, transparency (m); NTU, turbidity (ntu); DO, dissolved oxygen (mg/L); CON, conductivity (s/m); ORP, oxidation-reduction potential (mV); PAR, photosynthetically active radiation (μmol/m² s); CHL, chlorophyll content (μg/L); TDS, total dissolved solid (g/L); SAL, salinity (%)

DNA extraction and PCR amplification

Total DNA was extracted from bacterial cells trapped on the filter by using a Power Soil™ DNA Isolation Kit (MOBIO, USA). Total DNA of sediment soil samples (500 mg) was extracted by the same kit. The procedures were performed as described in the manufacturer’s instructions. PCR amplification was performed using the following reaction mixtures: 5 μL of 10× PCR buffer, 3.2 μL of dNTP (2.5 mM), 0.4 μL of rTaq (5 U/μL), 1 μL of each primer (20 mM), 1 μL (100 ng) of template DNA, and nucleases-free water to give a final volume of 50 μL. To amplify the V3 region of the bacterial 16S rRNA gene, the primers used for PCR were the 338F (5′-CCT ACG GGA GGC AGC AG-3′) and 518R (5′-ATT ACC GCG GCT GCT GG-3′) with GC-clamps (CGC CCG GGG CGC GCC CCG GGG CGG GGC GGG GGC GCG GGG GGC CTA CGG GAG GCA GCAG) attached to the forward primer. Thermal cycling was carried out using the following conditions: initial denaturation at 94 °C for 5 min, followed by 30 cycles at 94 °C for 30 s, 55 °C for 30 s, 72 °C for 30 s, and a final extension at 72 °C for 10 min. The size of the PCR product was confirmed by electrophoresis in 1% agarose gel and with Gold view Nucleic Acid Gel Stain (10,000×) added.

Denaturing gradient gel electrophoresis analyses

Denaturing gradient gel electrophoresis (DGGE) was conducted using a Dcode™ Universal Mutation Detection System (Bio-Rad, USA). Samples of PCR product were loaded onto 8% polyacrylamide gel with liner denaturing gradient (35–55% denaturant; 100% denaturant is 7 mol/L urea and 40% (V/V) formamide). Electrophoresis was performed in 1× TAE buffer at 150 V and 60 °C for 240 min. Gel was stained for 20 min using SYBR Gold (1:1000) dilutions (Invitrogen, USA) and photographed using a Bio Rad Gel Doc 2000 Gel Documentation System (Bio-Rad, USA). The Shannon index and peak intensity was determined based on analyses of the photograph using the Bio Rad Quantity One software (Bio-Rad, USA). PCR products were purified with Poly-Gel DNA Extraction Kit (OMEGA, USA) and ligated into pEASY-T vector kit (Promega, USA) according to the instructions of the manufacturer [26].

Phylogenetic analysis

All sequences were compared with those in the GenBank database (www.ncbi.nlm.nih.gov/BLAST) by using the BLAST program for a first phylogenetic affiliation. After automatic and manual sequence alignment, phylogenetic trees were constructed by the neighbor-joining method and bootstrap analyses for 1000 replicates were performed [27]. The 16S rRNA genes were sequenced with an ABI 3730 DNA Analyzer, and the sequences were submitted to the GenBank database (http:// www. ncbi. nlm. nih. gov) with the following accession numbers: S1-S71: KX258475-KX258532; W1-W90:KX258533-KX258614.

Statistical analysis

Analysis of the variance (ANOVA) was performed using IBM SPSS Statistics 22 to test significant differences of environmental physicochemical properties. The level of significance was determined using Duncan’s test at P < 0.05. Migration and intensity of DGGE bands were analyzed with Quantity One image analysis software (Bio-Rad, USA). The bands which shared identical migration position were considered the same species. Cluster analysis was conducted based on the dice coefficient method and the unweighted pair group method with arithmetic mean (UPGMA) dendrogram method. BLAST tools were used to evaluate the similarity of sequences obtained from the DGGE bands with other 16S rRNA gene sequences in the NCBI database. The neighbor-joining method in the MEGA 6.0 software packages was used for phylogenetic analysis. The Shannon-Weaver index (H), Pielou Evenness (E), and Simpson index (D) of the bacterial community were calculated by Eqs. (1), (2), and (3).

$$H=-{\sum }_{\mathrm{i}=1}^{\mathrm{S}}\mathit{Pi}ln\mathit{Pi}$$ 1

$$E=\raisebox{1ex}{$H$}\!\left/ \!\raisebox{-1ex}{$lnS$}\right.$$ 2

$$D=1-\sum {\mathit{Pi}}^2$$ 3

where H is the value of the Shannon index, Pi is the ratio of the specific band intensity to the total intensity of all bands in a lane, and S is the number of bands in the sample. The richness (S) of the bacterial community was determined from the number of bands in each lane.

Artificial neural network (ANN) identification was performed with the program Synapse (Peltarion Inc.) to generate a Kohonen self-organizing map (KSOM). Artificial neural network (ANN) analysis was performed using the bacterial diversity indices and physicochemical variables, which is a particularly useful tool for extracting patterns among complex nonlinear relationships between input and output data sets that cannot be detected with traditional methods for multivariate analyses [28]. Kohonen self-organizing maps (KSOM), a subtype of ANN, present the data in a visual way that provides a means for easy interpretation of complicated dimensional data sets [21]. The KSOM does not replace existing statistical tools, but instead complements our ability to examine relationships between disparate types of variables in a visual presentation of the data. Two separate parts were displayed by KSOM. These include the unified matrix (U-matrix), and the component panels, which represent individual input or output variables. Low values are indicated by blue colors and high values are indicated by red colors. The relationships between each of the variables are visualized by comparing the color patterns for individual maps. Similar color areas within the panels indicated positive correlations between variables, whereas opposite colors in the same area indicated inverse relationships. In this manner, the relationships between all of the variables in the model can be examined simultaneously or in pair-wise combinations.

Results and discussion

Environmental physicochemical properties

Samples were collected at 15 sites located in different areas of the Poyang Lake wetland. Physicochemical properties of each site are presented in Table 1. Total nitrogen (TN) contents and total phosphorus (TP) were extremely high in the sampling sites 17 with the value of 3.044 ± 0.01 mg/L and 0.157 ± 0.01 mg/L respectively. The lowest TN contents were 0.792 ± 0.00 mg/L of site 54. The lowest TP contents were 0.024 ± 0.00 mg/L of site 80. This phenomenon may be due to the site 17 being located at the entrance of Rao River to Poyang Lake. According to the previous report, Rao River was mainly affected by the agricultural activities and copper/phosphorite mines [12], which could cause the high-concentration value of TN and TP that occurred in the river mouth area of Poyang Lake. However, the sites 54 and 80, located in the middle of Poyang lake, were in mesotrophic condition. The data of site 11 showed higher dissolved oxygen (DO) (6.89 ± 0.03), conductivity (CON) (351.60 ± 0.80), total dissolved solid (TDS) (0.23 ± 0.01), salinity (SAL) (0.17 ± 0.03) and NO₃^—N (1.91 ± 0.00) value than other sites, which indicated that site 11 located at the entrance of Yangtze River contains more dissolved solid and salinity compare with other places.

DGGE, denaturing gradient gel electrophoresis; WD, water depth; TRA, transparency; TEM, water temperature; NTU, turbidity; DO, dissolved oxygen; CON, conductivity; ORP, oxidation-reduction potential; PAR, photosynthetically active radiation; CHL, chlorophyll content; TDS, total dissolved solid; SAL, salinity

DGGE profiles of the bacterial community from the water column and sediment niches

In order to clarify the species of bacterial community from 15 different sampling sites in Poyang Lake wetland and examine whether or not bacteria from fresh water might match the bacteria from the sediment, PCR-DGGE was used for further research. PCR-DGGE profiles of the bacterial community from each sampling sites are shown in Fig. 2a (freshwater samples) and Fig. 2b (sediment samples). The identities of bacterial species that were represented by individual DNA OTUs in the DGGE gels were determined by cutting the bands and DNA sequencing.

Fig. 2.

Denaturing gradient gel electrophoresis (DGGE) of fresh water samples (a) and sediment samples (b) in fifteen different sampling sites. Band number 1–90: predominant DGGE bands from fresh water samples; band number 1–71: predominant DGGE bands from sediment samples; lane number 1–104: fifteen sampling sites

The results showed that there was great bacteria diversity in the fresh water and sediment of Poyang Lake wetland. There are 30 to 61 kinds of dominant microflora isolated from freshwater microenvironment, and they respectively belonged to 90 species, 59 genuses, and 12 classes. In contrast, sediment microenvironment isolated 40 to 68 kinds of dominant microflora and they respectively belonged to 71 species, 49 genuses, and 10 classes. For the sediment samples, the band 28 (Flavobacterium xinjiangense) and 60 (Sphingomonas sanxanigenens) were detected among all the 15 samples. In contrast, there was only one band 30 (Azospira sp.) detected among freshwater sample. This result indicated that community structure from different regions and niches of Poyang Lake has a certain extent similarity, but each region has the unique dominant bacterium groups, probably influenced by the different micro domain environment. The DGGE banding patterns revealed a remarkable spatial heterogeneity which was closely related to their geographical positions. Among the freshwater samples, the site 1 (central part) was the least diverse and the site 17 (Rao River estuaries) was the most diverse, and this phenomenon was probably related to the nutrient status in Poyang Lake, especially contents of TP and TN (Table 1). These findings indicated that high loads of nutrients may relate to the increases in bacterial diversity. This is in agreement with the results reported in previous research, which showed positive correlations between sedimentary bacterial abundance and levels of organic matter and nutrients [17, 29].

Spatial variability in bacterial community structure

Microbial community plays an important role in the earth’s biogeochemical cycles, such as decomposing, transforming of complex organic compounds, and demineralizing nutrients in freshwater and sediment niches [27, 30]. Consequently, investigating microbial structure and function in freshwater and sediments is vital for acquiring a better general understanding of wetland ecosystems [31]. Identifications of the dominant bacteria were conducted by analyses of the DNA sequences cut from diverse bands in the DGGE gels. The phylogenetic distribution of prominent 16S rRNA gene sequences from samples of the freshwater and sediment niches is shown in Table 2 and Table 3. Taxonomic analysis showed that all the DGGE bands from fresh water niches were classified into fourteen groups. The ninety bands sequenced were affiliated with Bacteroidetes (23.3%), β-Proteobacteria (21.1%), α-Proteobacteria (16.7%), γ-Proteobacteria (14.4%), Actinobacteria (4.4%), and δ-Proteobacteria (2.2%) (Fig. 4). The rest sequences were belonging to Acidobacteria, Firmicutes, Verrucomicrobia, Gemmatimonadetes, Chloroflexi, and Nitrospirae. In contrast, all the DGGE bands from sediment sequenced were classified into eleven groups. The seventy-one bands sequenced were affiliated with β-Proteobacteria (22.5%), Bacteroidetes (21.1%), γ-Proteobacteria (16.9%), α-Proteobacteria (9.9%), δ-Proteobacteria (8.5%), Firmicutes (8.5%), and Chloroflexi (4.2%). The rest sequences were belonging to Acidobacteria, Actinobacteria, and Nitrospirae (Fig. 3). This phenomenon indicates that the bacterial communities from freshwater niches were more diverse than the bacterial communities from sediment niches, since the fresh water niches changed greatly due to the season, climate, weather, and influent [32, 33]. It was noticeable that Canalipalpata, Bacillariophyta, Gemmatimonadetes, and Verrucomicrobia were detected in fresh water niches only. Kolmonen et al. [34] pointed out that Verrucomicrobia are relatively more prevalent in eutrophic lakes than in oligo- or mesotrophic lakes. In our research, Verrucomicrobia was detected in sites 11, 17, and 21, which are located at the entrance of Hu Kou, Rao River, and Fu River respectively. The total nitrogen content was also ranked top three among all the sample sites. Therefore, Verrucomicrobia may be utilized as indicators of environmental change.

Table 2.

Sequence results specially identified from fresh water samples

ID	Blast search results	Accession number	Score*	Taxon
1	Pseudomonas poae	HQ701805	99%	γ–Proteobacteria
2	Pseudomonas toyotomiensis	KF984313	100%	γ–Proteobacteria
4	Flavobacterium sp.	JX290476	99%	Bacteroidetes
6	Bacteroides heparinolyticus	JN867284	94%	Bacteroidetes
7	Uncultured Thioalkalivibrio sp.	GQ366384	96%	γ–Proteobacteria
8	Pseudomonas mohnii	HF585477	100%	γ–Proteobacteria
10	Caulobacter sp.	JX861096	99%	α–Proteobacteria
11	Variovorax paradoxus	HF584859	99%	β–Proteobacteria
12	Bacteroidetes bacterium	JN817712	95%	Bacteroidetes
14	Uncultured Bacteroidetes bacterium	EU639798	99%	Bacteroidetes
15	Uncultured Desulfobacterales bacterium	KC006394	100%	δ–Proteobacteria
16	Uncultured Heliobacterium sp.	GU556467	94%	Firmicutes
19	Uncultured Verrucomicrobia bacterium	GQ243077	98%	Verrucomicrobia
21	Uncultured Gemmatimonas sp.	HQ132388	97%	Gemmatimonadetes
22	Uncultured Dehalococcoides sp.	KC785915	98%	Chloroflexi
23	Uncultured Geodermatophilus sp.	JQ401510	100%	Actinobacteria
24	Streptomyces scabrisporus	KF317986	100%	Actinobacteria
26	Flavobacterium xanthum	AB680727	99%	Bacteroidetes
29	Shewanella sp.	AB792775	95%	γ–Proteobacteria
30	Uncultured Azospira sp.	GQ183381	96%	β–Proteobacteria
34	Janthinobacterium sp.	GU213396	99%	β–Proteobacteria
37	Altererythrobacter marensis	KF876909	100%	α–Proteobacteria
38	Uncultured Reyranella sp.	JQ177500	100%	α–Proteobacteria
39	Uncultured Rickettsiales bacterium	KF583165	100%	α–Proteobacteria
40	Herminiimonas sp.	JX983165	98%	β–Proteobacteria
41	Undibacterium sp.	HM149217	98%	β–Proteobacteria
42	Uncultured Bacteroidetes bacterium	JX548549	96%	Bacteroidetes
43	Amphora coffeaeformis	KF698705	100%	Bacillariophyta;
46	Uncultured Riftia sp.	JQ177983	93%	Canalipalpata;
47	Uncultured β-proteobacterium	KC603153	98%	β–Proteobacteria
49	Uncultured Herbaspirillum sp.	JQ685008	99%	β–Proteobacteria
50	Stenotrophomonas sp.	JX977085	100%	γ–Proteobacteria
51	Lysobacter brunescens	KF911330	97%	γ–Proteobacteria
52	Uncultured Methyloversatilis sp.	AM990019	98%	β–Proteobacteria
53	Aquabacterium commune	NR024875	100%	β–Proteobacteria
55	Uncultured Rhodoferax sp.	HQ008584	98%	β–Proteobacteria
56	Uncultured Methylophilus sp.	GU472577	100%	β–Proteobacteria
57	Uncultured Cytophaga sp.	GU269394	98%	Bacteroidetes
58	Sphingomonas hunanensis	KF923436	99%	α–Proteobacteria
61	Brevundimonas sp.	JX861623	99%	α–Proteobacteria
62	Brevundimonas intermedia	KF923439	100%	α–Proteobacteria
63	Brevundimonas alba	KC789788	100%	α–Proteobacteria
64	Uncultured Hyphomicrobium sp.	JQ791683	96%	α–Proteobacteria
65	Uncultured actinobacterium	KF543146	99%	Actinobacteria
66	Rhodobacter sp.	JF792117	100%	α–Proteobacteria
67	Flavobacterium dankookense	NR108738	98%	Bacteroidetes
68	Acinetobacter brisouii	AB859735	95%	γ–Proteobacteria
69	Uncultured Acetobacteraceae bacterium	GQ242733	100%	α–Proteobacteria
70	Uncultured Oxalobacteraceae bacterium	EU640693	99%	β–Proteobacteria
72	Prevotella maculosa	AB547690	93%	Bacteroidetes
73	Flavobacterium saliperosum	NR043481	100%	Bacteroidetes
74	Paenibacillus sp.	JX464210	96%	Firmicutes
75	Flavobacterium hydatis	JX657044	98%	Bacteroidetes
76	Uncultured Nannocystis sp.	EU809678	98%	δ–Proteobacteria
77	Muricauda ruestringensis	NR074562	100%	Bacteroidetes
78	Uncultured Bacteroidetes bacterium	KF827177	99%	Bacteroidetes
79	Brevundimonas intermedia	KF923439	100%	α–Proteobacteria
80	Flavobacterium tiangeerense	KF528718	100%	Bacteroidetes
82	Uncultured Zoogloea sp.	GQ420883	99%	β–Proteobacteria
83	Uncultured Holophaga sp.	AJ519667	95%	Acidobacteria
85	Paucibacter toxinivorans	AY515384	100%	β–Proteobacteria
86	Uncultured bacterium	HM452222	100%	Environmental samples
90	Uncultured Methylocapsa sp.	KF956743	98%	α–Proteobacteria

Table 3.

Sequence results specially identified from sediment samples

ID	Blast search results	Accession number	Score*	Taxon
1	Nitrosomonas aestuarii	NR104818	97%	β–Proteobacteria
2	Flavobacterium johnsoniae	EU860081	99%	Bacteroidetes
3	Pedobacter composti	NR041506	96%	Bacteroidetes
5	Bacteroides barnesiae	NR041446	92%	Bacteroidetes
8	Massilia niabensis	NR044571	97%	β–Proteobacteria
9	Gibbsiella dentisurs	NR108121	97%	γ–Proteobacteria
12	Uncultured bacterium	HQ395801	100%	Environmental samples
14	Uncultured Duganella sp.	EU300443	100%	β–proteobacteria
17	Pedobacter ruber	HQ882803	98%	Bacteroidetes
18	Psychrosinus fermentans	DQ767881	100%	Firmicutes
19	Paenibacillus sp.	JX464210	96%	Firmicutes
20	Uncultured Pedobacter sp.	KC172295	99%	Bacteroidetes
21	Uncultured Phenylobacterium sp.	KC166861	100%	α–Proteobacteria
22	Bdellovibrio exovorus	NR102876	99%	δ–Proteobacteria
24	Pseudomonas koreensis	KJ009424	99%	γ–Proteobacteria
25	Pseudomonas migulae	KJ127238	100%	γ–Proteobacteria
26	Desulfosporosinus burensis	NR109421	99%	Firmicutes
27	Pseudomonas syringae	CP007014	100%	γ–Proteobacteria
28	Flavobacterium xinjiangense	KF911333	100%	Bacteroidetes
29	Uncultured Sphingobacterium sp.	KC453702	98%	Bacteroidetes
30	Uncultured Steroidobacter sp.	HE974825	98%	γ–Proteobacteria
31	Uncultured Chloroflexi bacterium	JQ795274	99%	Chloroflexi
32	Uncultured Sphingopyxis sp.	JX530076	100%	α–Proteobacteria
33	Paenisporosarcina macmurdoensis	KC921194	100%	Firmicutes
36	Uncultured Kouleothrix sp.	JQ071842	92%	Chloroflexi
38	Polaromonas sp.	JX949585	100%	β–Proteobacteria
39	Acidovorax avenae	KF498647	98%	β–Proteobacteria
40	Uncultured Ferruginibacter sp.	JQ308722	99%	Bacteroidetes
41	Rhodoferax sp.	AY788954	98%	β–Proteobacteria
42	Uncultured Syntrophus sp.	GU556344	97%	δ–Proteobacteria
43	Uncultured Bacteriovorax sp.	HQ691859	98%	δ–Proteobacteria
44	Uncultured Chloroflexi bacterium	AB293392	99%	Chloroflexi
45	Uncultured Steroidobacter sp.	HE648175	96%	γ–Proteobacteria
47	Terrimonas sp.	HM124372	98%	Bacteroidetes
50	Ferruginibacter lapsinanis	NR044589	99%	Bacteroidetes
51	Uncultured Ferruginibacter sp.	JQ723682	99%	Bacteroidetes
54	Uncultured Rhizobiales bacterium	JQ087229	100%	α–Proteobacteria
55	Methylosoma difficile	NR043562	95%	γ–Proteobacteria
56	Uncultured delta proteobacterium	JN038592	96%	δ–Proteobacteria
57	Uncultured Sideroxydans sp.	GU556436	98%	β–Proteobacteria
58	Uncultured Aquaspirillum sp.	DQ646801	98%	β–Proteobacteria
60	Sphingomonas sanxanigenens	KF981586	100%	α–Proteobacteria
61	Sphingomonas jaspsi	KF580874	100%	α–Proteobacteria
63	Uncultured Anaeromyxobacter sp.	JQ183093	98%	δ–Proteobacteria
64	Uncultured Streptomyces sp.	GU000479	92%	Actinobacteria
65	Uncultured Clostridium sp.	JQ684900	96%	Firmicutes
66	Uncultured Gallionella sp.	KF851155	97%	β–Proteobacteria
67	Methylosinus trichosporium	KC353469	99%	α–Proteobacteria
68	Bradyrhizobium elkanii	KF749003	100%	α–Proteobacteria
70	Polaromonas naphthalenivorans	KF528728	99%	β–Proteobacteria
71	Megasphaera micronuciformis	KC632239	94%	Firmicutes

Fig. 4.

Heat map analysis of environmental factors and diversity index of different sampling sites. SS, species richness from sediment samples; SH, Shannon index from sediment samples; SD, Simpson index from sediment samples; SE, Pielou Evenness index from sediment samples; FS, species richness from freshwater samples; FH, Shannon index from freshwater samples; FD, Simpson index from freshwater samples; FE, Pielou Evenness index from freshwater samples)

Fig. 3.

Relative abundance of bacterial communities in Poyang Lake niche

While compared with the fresh water samples, the relative abundances of δ-Proteobacteria, Firmicutes, and Chloroflexi in sediment sample were one times higher. Previous studies indicated that Proteobacteria, Firmicutes, Actinobacteria, Acidobacteria, Chloroflexi, Bacteroidetes, Planctomycetes, Verrucomicrobia, and Nitrospirae were the major components of sediment bacterial communities in Poyang Lake. Among them, Proteobacteria was the most dominant phylum, followed by Firmicutes [16]. Ding et al. pointed out that δ-Proteobacteria were more abundant at depths less than 30 cm in Poyang Lake sediment and the Chloroflexi relative abundance increased along the distance gradient [1]. Apparently, Acidobacteria and Actinobacteria were more abundant at freshwater system than in the sediment niches. Proteobacteria accounted for 54.4% and 57.7% of the total bacterial community in freshwater and sediment niches separately, which is the most abundant phyla in Poyang Lake. At species level, except the uncultured bacterium, the freshwater and sediment samples shared 6 species in common (Table 4), while most of the species were different from the two niches. Nitrospira was detected both in freshwater and sediment niches of Poyang Lake, which has an effect on nitrification and nitrite oxidation, and mainly distributed in the environment influenced by human activities [35, 36]. Therefore, nitrification is intensively involved in the nitrogen cycle of lake sediment [37, 38]. There are 34 unique species accounted for 85% in fresh water samples (Table 2) and 28 unique species accounted for 82% in sediment samples (Table 3). Consequently, the composition and abundance of microbes at freshwater niches obviously differed with those at sediment sites. Our results revealed that freshwater and sediment niches with different physicochemical properties may harbor metabolically diverse microorganisms.

Table 4.

Identical sequence results from band excised from freshwater samples and sediment samples

Figure 2a	Figure 2b	Blast search results	Accession number	Score*	Taxon
3	4	Prevotella buccalis	JN867261	97%	Bacteroidetes
9	7	Pedobacter lentus	NR044218	98%	Bacteroidetes
13	11	Uncultured Acidobacterium sp.	JX945558	100%	Acidobacteria
17	13	Acidovorax avenae subsp. Avenae	KF498647	94%	β–Proteobacteria
18	15	Uncultured Thiobacter sp.	GQ287535	97%	β–Proteobacteria
25	34	Uncultured Nitrospira sp.	JQ795206	99%	Nitrospirae
27	16	Uncultured gamma proteobacterium	AY528805	100%	γ–Proteobacteria
33	35	Uncultured Tetrasphaera sp.	JN866993	100%	Actinobacteria
48	37	Massilia aurea	HE648123	98%	β–Proteobacteria
60	49	Uncultured Xanthomonadaceae bacterium	JQ086973	97%	γ–Proteobacteria
71	10	Uncultured Sphingobacteria bacterium	EF520590	97%	Bacteroidetes
81	52	Luteimonas sp.	JQ349045	98%	γ–Proteobacteria
84	53	Uncultured bacterium	GU270493	90%	Environmental samples
89	69	Propionivibrio militaris	EU849004	97%	β–Proteobacteria

Freshwater niches harbor many bacterial groups that appear to be phylogenetically distinct from soil samples, where α-, β-, and γ-Proteobacteria, Actinobacteria, Bacteroidetes, Cyanobacteria, Actinobacteria, Bacteroidetes, Cyanobacteria, Verrucomicrobia, and Planctomycetes are mostly detected [39] [40]. According to previous reports, the most dominant bacterial taxa are β-Proteobacteria in freshwater ecosystems [41–43], whereas α- and γ-Proteobacteria appear to be predominant indigenous community in marine [5]; however, some studies have shown that Bacteroidetes may be crucial bacterial groups in freshwater ecosystems [44]. In our research, phylogenetic analysis of the sequences obtained from the DGGE revealed that most of the OTUs from freshwater belonged to Bacteroidetes (23.33%); β-Proteobacteria (21.11%) represented the second most abundant division. However, most of the OTUs from sediment belonged to β-Proteobacteria (22.54%); the second most abundant division is Bacteroidetes (21.13%). The results showed the coincidence with the previous research. A previous study also suggested that Bacteroidetes and β-Proteobacteria are the most abundant bacterial groups in eutrophic freshwater ecosystems [45]. Actinobacteria is another bacterial class that is consistently present in freshwater niches. Sekiguchi et al. [5] collected samples in Hukou and reported that Actinobacteria was regarded as the most significant lineage in the Poyang Lake. Moreover, they pointed out that the dominant bacterial groups changed from β-Proteobacteria to Actinobacteria along the Yangtze River. In our result, Actinobacteria just comprised 4.44% of all 16S rRNA gene sequences. The reason possibly is Poyang Lake and Yangtze River connected with each other at Hukou, and the bacterial community structure in Hukou was affected by water quality both from Poyang Lake and Yangtze River. Therefore, location was considered to be the key factors driving the changes in bacterial community composition.

Although, not examined for the individual species identified here, prior report has revealed that partial of the bacterial species related to degradation, eutrophication, nitrification, denitrification, and ammonification function [45–47]. These include species identified from the DGGE results such as Sphingomonas sp., Flavobacterium sp., Caulobacter sp., Lysobacter sp., and Acinetobacter sp., which have the potential of degrading organic contaminant and participating the nitrogen geochemical cycle [15, 48]. A plausible explanation for the phenomenon is that even though the pollutant and nutrient were discharged into Poyang Lake, the diverse microorganisms were able to maintain the aquatic ecosystems in a sustainable way. Xiao’s research was also proved that microbial community carbon metabolic profiles changed with the shift of bacterial community structure, and pollution enhanced the sediment bacterial diversity [49].

Diversity index analyses

The structural diversity of the bacterial community was examined by the species richness (S), Shannon-Wiener diversity index (H), Simpson index (D), and Pielou Evenness Index (E) (Table 5). Greater Shannon index represents higher community diversity, while greater Simpson index value represents higher community diversity. When species richness is the same, community with greater species evenness has more diversity. For the sediment samples, richness and Shannon-Wiener diversity index ranged from 40 to 68 and 2.43 to 3.87 respectively. And Simpson and Pielou Evenness Index ranged from 0.71 to 0.98 and 0.59 to 0.93 respectively. For the fresh water samples, richness and Shannon-Wiener diversity index ranged from 30 to 61 and 2.79 to 3.79, respectively. And Simpson and Pielou Evenness Index ranged from 0.90 to 0.97 and 0.78 to 0.93 respectively. This phenomenon indicated that the microbial diversity and community composition are varying with the surrounding environment.

Table 5.

Bacterial diversity indices are based on DGGE analysis of 16S rRNA gene fragments of sediment samples and fresh water samples from fifteen sampling sites

Lane	Sampling sites	Sediment samples				Fresh water samples
		Sa	Hb	Dc	Ed	S	H	D	E
1	1	63	2.43	0.71	0.59	30	2.79	0.93	0.81
2	5	64	3.85	0.97	0.92	48	3.40	0.95	0.88
3	11	52	3.64	0.97	0.92	51	3.49	0.96	0.89
4	14	55	3.60	0.97	0.90	46	3.45	0.96	0.90
5	17	58	3.60	0.96	0.89	51	3.81	0.98	0.95
6	21	50	3.65	0.97	0.93	49	3.47	0.96	0.89
7	22	58	3.80	0.97	0.93	46	3.00	0.90	0.78
8	47	40	3.13	0.94	0.85	58	3.52	0.96	0.89
9	54	48	3.56	0.96	0.92	59	3.79	0.97	0.93
10	80	45	3.05	0.93	0.80	61	3.76	0.97	0.91
11	82	61	3.69	0.96	0.90	59	3.58	0.96	0.88
12	85	57	3.76	0.93	0.93	52	3.41	0.95	0.86
13	89	68	3.87	0.98	0.93	55	2.17	0.79	0.57
14	102	62	3.81	0.97	0.92	54	3.53	0.96	0.88
15	104	54	3.72	0.97	0.93	44	3.43	0.95	0.88

^aS species richness, ^bH Shannon index, ^cD Simpson Index, ^dE Pielou Evenness Index

Among the sediment samples, the site 89 sample showed the most complex profiles with highest Shannon index (H = 3.87) and Simpson Index (D = 0.97), which indicated that the sediment of 89 sampling site contains a great variety of bacterial species. Site 89 is located at the entrance of Gan River; Gan River flows across the Nan Chang City, mainly affected by domestic sewage and industrial activities of the city, which means more nutrients and pollutants from Gan River may be brought to the Poyang Lake [19]. In contrast, samples from site 1 showed the lowest Shannon index (H = 2.43) and Simpson Index (D = 0.71). Site 1 is located at the middle place of Poyang Lake, which can keep a natural situation without disturbing by external artificial factor, and the content of nutrients and pollutants were less than the place close to the river bank. This phenomenon indicated that the bacterial community structure and diversity were greatly influenced by the external environmental factors.

For the fresh water samples, the site 89 exhibited an interesting result with the lowest Shannon index (H = 2.17), Simpson Index (D = 0.79), and Evenness (E = 0.57), whereas, the sediment samples of site 89 showed the most complex profiles with highest Shannon index (H = 3.87), Simpson Index (D = 0.97), and Evenness (E = 0.93). Wastewater from many mines located at the upstream of Gan jiang River resulted in high heavy metal level in Gan jiang River-lake system. This phenomenon may be due to the fresh water from Gan Jiang containing more pollutants (Cd, Mn, W, Cu, Xn, As) which may have a negative effect to the diversity of microbes [2, 3]. However, the sediment layer was stable and full of nutrients, which can provide abundant carbon and nitrogen source for the sediment bacteria growth [50]. Samples from site 17 showed the highest Shannon index (H = 3.81) and Simpson Index (D = 0.98) and Evenness (E = 0.95), which indicated that the bacterial communities’ diversity was greatly influenced by the Rao River with high total nitrogen and total phosphorous. In general, the diversity index indicated that the bacterial community’s structure and diversity were unstable, which was easily changed by complex hydrological condition and location.

Heat map analysis

In order to illustrate the relationship between the fifteen sampling sites, the heat map analysis was conducted by using hierarchical average linkage as clustering method. According to the heat map analysis of environmental factors and diversity index in Fig. 4, the sampling sites were classified into different groups. Interestingly, site 1 was distinguished from other branches probably due to the central location in the Poyang Lake, which maintains the most natural environment compared with the other sites. Site 17 was separated alone from the other rest sites. The main reason may be due to the location at the entrance of the Rao River which is full of N and P nutrients. Site 11 was also noticeably different from other group mainly because its location is close to the Hu Kou area, which is the entrance of the Poyang Lake water to Yangtze River. Site 89 was outstanding from other group may be due to its location at the entrance of Gan River. 104 and 102 were clustered as one group, which means the water quality from Gan River was stable. Similar phenomenon was also observed at sites 21 and 22, which are located at the entrance of Xin River. Site 47 and site 80 grouped together may be due to its location far from the entrance of tributaries and formed a relatively stable environment. Site 5 is located at the entrance of Xiu River; interestingly, it was also separated from other sample sites.

The heat map illustrated that the physicochemical properties and bacterial diversity of the different sampling sites in Poyang Lake were determined by the five tributaries’ characteristics. River-lake systems are sites where rivers connect to lakes. These systems play an important role in the transformation of terrigenous material. Dramatic changes in the hydrological condition of river-lake systems are conductive to the settling of suspended particulate matter as well as the attached metal pollutants to the sediment [3]. The five tributaries (Xiushui, Raohe, Ganjiang, Fuhe, Xinjiang) that constitute the whole basin feeding Poyang Lake contribute 87% of its water volume. Meanwhile, large amounts of heavy metal pollutants are also transferred to the lake by these rivers. Raohe was mainly affected by the agricultural activities and copper/ phosphorite mines, etc., which could cause the high-concentration value of TP that occurred in the river mouth area of Poyang Lake [19]. Moreover, previous research indicated that Raohe had high concentration of Cd, Cu, Zn, and As, but had a relatively low TOC content; the main reason was caused by the Dexing copper mine, the largest open-cast copper mine in Asia, which is located in upstream of Raohe [3]. Ganjiang River was mainly affected by domestic sewage and industrial activities of the city, as a result of population increase, The amount of TN discharged into the Ganjiang River has increased in recent years, which caused a richness of nitrifying bacteria [16]. The concentrations of W and Sn were higher in the Xiushui sediment than in the other sediments [2]. On the whole, due to the effect of Yangtze River and five tributaries, microbial resources in different geographic space of Poyang Lake were both related and varied with different environments.

Cluster analysis

To assess changes in the genetic diversity of bacterial communities with changing locations, DGGE banding patterns were analyzed by cluster analysis. Similarities between the banding patterns generated by PCR-DGGE of the fresh water samples and sediment samples were analyzed using the dice coefficient. The unweighted pair group method using neighbor-joining cluster method was applied to the DGGE profiling with the purpose of clustering the samples with similar patterns. In our research, 30 samples from fresh water and sediment were conducted using DGGE analysis. The resulting dendrograms (Fig. 5) revealed apparent differences in the microbial communities among different locations. Interestingly, all the samples from freshwater niches clustered closely together, the rest samples from sediment formed another two groups. Samples from fresh water and sediment demonstrated distinct clusters, indicating that the bacterial community structure differed significantly at the two niches. This result showed consistence with the previous species level analysis, while most of the species were different from the two niches except 6 species in common (Table 4). Freshwater samples from fifteen sites were separated into four main groups; in contrast, for the sediment samples, five branches were developed from the fifteen sites. Freshwater and sediment samples from site 5 were clustered closely together, which indicated the bacterial species share the similar information between the freshwater and sediment niches. However, the freshwater and sediment samples from site 1 grouped loosely away from the main cluster as another single branch. This phenomenon explained that the bacterial communities in central part of Poyang Lake were greatly different from the other sites maybe due to the middle area being less affected by human activity. The fresh water and sediment samples from sites 102 and 104 were closely grouped together, which indicated that these two sites both are located at the entrance of Gan River, and shared the similar bacterial community information. Interestingly, the similar phenomenon was also found between the fresh water samples from sites 21 and 22. It was noticeable that sediment sample from site 17 was separately grouped as a single branch, probably because the location at the entrance of Rao River which contains more nutrients and pollutants. These results showed the coincidence with the data from heat map analysis, which means the geographical features play an important role in shaping the bacterial communities’ distribution. Moreover, bacterial abundance and community composition might be mainly affected by estuaries’ inputs of both nutrients and pollutants, which were caused by domestic waste, and agricultural and industrial activities.

Fig. 5.

Hierarchical cluster analysis results of all the DGGE profiles from samples demonstrated graphically as an neighbor-joining dendrogram. F1–F104: fresh water samples; S1–S104: sediment samples

Bacterial community composition in relation to physicochemical properties

By using a KSOM model, the plots were generated through machine learning to extract and recognize patterns (Fig. 6), which allows visual comparison of the correlations between all of the variables. There were 26 component panel maps representing 8 bacterial diversity indices (S species richness, H Shannon index, D Simpson Index, E Pielou Index from S sediment samples, and F freshwater samples), and 16 physicochemical parameters (TRA, transparency; NTU, turbidity; DO, dissolved oxygen; CON, conductivity; ORP, oxidation-reduction potential; PAR, photosynthetically active radiation; CHL, chlorophyll content; TDS, total dissolved solid; SAL, salinity (%); TN; TP; NO₃-N; NH₄⁺-N; NO₂-N; PO₄³-P). The value of the KSOM analysis was to observe interrelationships that exist between these 27 variables and provide a basis for generating hypotheses that can be experimentally examined. In the U-matrix that was generated, there are four groups indicating that there were four major determinants of structure in the data set.

Fig. 6.

Kohonen self-organizing maps of variables from bacterial diversity indices and physicochemical parameters. (Peltarion synapse) Color patterns in each box indicate ranges for each variable (red = high, blue = low). Comparison of patterns between boxes revealed simultaneous correlations between all variables as determined using an artificial neural network model

Visual inspection of the KSOM component panels shows common patterns between many sets of parameters. In this study, we were concerned primarily with environmental variables of different sampling sites and microbial diversity index, which we hypothesized the environmental variables will shape microbial community structure and affect the biogeochemical function. Comparison of the map generated for management across the entire display showed that there were four similar color distribution patterns with maps for group 1 (FH; FD; FE), group 2 (SH; SD; SE) and group 3 (TN; TP; NH₄⁺-N; NO₂⁻-N; PO₄³⁻-P), and group 4 (CON; TDS; NO₃⁻-N). Specially, the TRA and NTU showed obvious negative correlation; the similar relationship was also found between ORP and PAR. This phenomenon indicated that the high value of the photosynthetically active radiation will decrease the oxidation-reduction potential, whereas CON and TDS showed significant positive correlation with NO₃^—N concentration. Noticeably, the conductivity of aqueous solution is directly proportional to the dissolved solids concentration, and the higher the solids concentration, the higher the conductivity. Differences in environmental factors largely determine bacterial composition and diversity, such as C and N availability, temperature, pH, and sediment structure characteristics [51–53]. The chemical factor pH is a significant parameter which can shape the microbial diversity. In our result, sediment richness (SS) showed significant negative correlation with pH value, which illustrates that the higher the pH, the lower the bacterial richness of the sediment. The highest values of group 1 (FH; FD; FE) and group 3 (TN; TP; NH₄⁺-N; NO₂⁻-N; PO₄³⁻-P) showed consistency with each other. This result indicated that the freshwater with high N and P nutrients will greatly increase the diversity of the bacterial communities. Noticeably, the data from Rao River showed consistency with this conclusion. Above all, different types of environmental changes occurred and induced alterations of the bacterial diversity indices and species distribution.

In aquatic ecosystems, nutrient loading increase may cause chain reaction in food web structure, increased growth of destructive algal blooms, and the interruption of important ecosystem functions [54, 55]. Heterotrophic bacteria respond to nutrient enrichment promptly, resulting in variation in microbial species and community composition [56]. Similarly, in our research, there was a direct correlation between the main environmental indicators and the microbial community diversity index. This result indicated that the trophic state and the physiochemical properties of lake play roles in sustaining bacterial community structure.

Conclusion

In conclusion, this is the overall report on comparison of bacterial community from freshwater and sediment samples in the Poyang Lake. The results obtained in this study demonstrated that the location characteristics of fresh water and sediments may lead to different bacterial communities. Phylogenetically, both the bacteria from freshwater and sediment samples were dominated by α-, β-, γ-Proteobacteria and Bacteroidetes sequences. However, the bacteria proportions varied at individual sites, which was mainly affected by different tributaries’ characteristics. Due to the unique hydrological regime and special geographical conditions, the bacterial community structure differed significantly between freshwater and sediment niches. Horizontal differences associated with the water quality indicated that the trophic state and the physiochemical properties of lake play integrate roles in sustaining bacterial community structure. In general, microbial components in different geographic space were both related and varied with different environments. These results are important, since microbial community in freshwater and sediment niches plays a vital role in nutrient cycling, which is necessary to maintain aquatic ecosystem health. The information acquired in this study would be helpful to elucidate the structure and function of microbes in the aquatic ecosystem of Poyang Lake wetland.

Abbreviation

DGGE

Denaturing gradient gel electrophoresis

WD

Water depth

TRA

Transparency

TEM

Water temperature

NTU

Turbidity

DO

Dissolved oxygen

CON

Conductivity

ORP

Oxidation-reduction potential

PAR

Photosynthetically active radiation

CHL

Chlorophyll content

TDS

Total dissolved solid

SAL

Salinity

Funding information

This work was financially supported by the National Natural Science Foundation of China (Program No. 41601338), Natural Science Basic Research Plan in Shaanxi Province of China (Program No. 2018JQ4019 and 2020JM-110), the Fundamental Research Funds for the Central Universities (Program No. 3102018zy042), and National Training Programs of Innovation and Entrepreneurship for Undergraduates (S201910699176).