3.1 Analysis of water quality
The water quality parameters of each sampling point are shown in Fig. S1. Among them, the temperature, pH and DO in TT region were higher than those in MS region (p < 0.05). In addition, the distribution range of temperature, pH and DO in TT was significantly larger, compared with the MS and RC. The average COD concentration in MS was highest in three area, which may be due to more production and living activities along the MS. The Fig. S1, population living map around the sampling, also supported that. Therefore, human production and living activities will increase the concentration of pollutants in water [32]. When COD diffuses from the bank to the center of the river, the COD concentration will be reduced by a series of physical and chemical reactions, resulting in the COD concentration in the river center is lower than that on the bank. There were no obvious differences in other water quality parameters among the three regions. Overall, the significant differences in water quality parameters mainly occurred in the MS and TT.
3.2 Bacterial diversity and abundance probed by 16S rRNA and pufM gene
Through Illumina high-throughput sequencing platform, we generated 55,233 and 22,476 high-quality sequences from the bacterial and AAPB libraries of 40 sediment samples, respectively (Table S1). In bacterial community, 13,874 OTUs were identified, which classified into 67 phyla, 202 classes, 481 orders, 796 families, and 1,560 genera; while for AAPB communities, 5438 OTUs were identified, which included 11 phyla, 16 classes, 34 orders, 54 families, and 107 genera (Table S2). In the bacterial communities, Proteobacteria (26.37%) was the dominant bacterial phylum, followed by Actinobacteriota (15.28%), Acidobacteriota (13.23%), Chloroflexi (11.48%), Firmicutes (6.42%) and Planctomycetota (5.37%). However, the most abundant was an unclassified phylum (57.91%) of pufM gene in sediments of TGR, followed by Proteobacteria (41.32%) and Chloroflexi (0.6%) (Fig. 2b). The sum of abundance of other phyla, were less than 1% in the sediments of TGR (Fig. 2b). Meanwhile, to understand the abundance of AAPB community in sediments of TGR, the qPCR of pufM gene was measured. In this study, the abundance of AAPB ranged between (9.07±0.55) × 104 and (4.17±0.65) × 107 pufM gene copies/g in MS, between (2.21±0.44) × 104 and (9.98±0.30) × 107 copies/g in TT, between (1.02±0.04) × 106 and (6.21±0.54) × 107 copies/g in RC (Fig. S2). No significant difference was found in three regions in sediment of the TGR, but the abundance changes greatly in TT area.
3.3 Spatial differences in bacterial and AAPB communities
This study assessed the differences of microbial community richness and diversity among different sediments in the TGR by measuring Chao1 and Shannon indices [33]. The results of Chao1 and Shannon indexes showed that there was no obvious difference in bacterial and AAPB communities in the sediments of the three regions (p > 0.05) (Fig. S3). It indicated that richness and diversity of microbial communities was no significantly affected by regionalization. For AAPB community, Chao1 and Shannon indices decreased with the increase of sampling distance in the mainstream, indicating that the diversity and richness decreased in the downstream (Fig. S4). However, the qPCR of AAPB tended to increase in the downstream (Fig. S4). Therefore, in this study, the abundance of AAPB did not decrease or even increase with the decrease of biodiversity.
Through the investigation of unique and shared microorganisms in sediments. ~60% bacterial OTUs were shared across the sediment samples from different regions (Fig. 3a), while only ~38% pufM gene OTUs were shared in that (Fig. 3b). It indicated that microbial similarity was higher based on bacterial OTUs than that based on pufM gene OTUs. In addition, the unique OTUs were the most in the TT. At class level, γ- and α-proteobacteria, Vicinamibacteria, Actinobacteria, and Anaerolineae dominated in total bacterial community; while for AAPB community, β- and α-proteobacteria were dominant bacteria. Regardless of the bacterial community or the AAPB community, only α-proteobacteria were more abundant in the TT region, and the abundance of other bacterial taxa was basically similar in different regions (Fig. 3cd). For bacterial communities, Comamonadaceae, Sphingomonadaceae and Beijernckiaceae were obviously higher in TT area than that in the MS and RC area (p < 0.05) (Fig. 3e). As shown in Fig. 3f, it was a heatmap based on the top 15 dominant genera of AAPB communities (excluding unclassified genera). The main genera abundance of AAPB community, included Methylibium spp., Methylobacterium spp., Novosphingobium spp., Hydrogenophaga spp., Skermanella spp. and Bradyrhizobium spp., were higher in TT and RC area than that in MS area. All dominant genera came from Proteobacteria (except Roseiflex spp.). Through the above analysis, the phenomenon of differences in the composition of microbial community in the sediments of the TGR was found.
3.4 Effects of environmental variables on changes of bacterial community
Regional and temporal are two important factors influencing microbial community changes, but potential environmental factors remain to be further studied. In previous studies, the change of bacterial community was closely related to many environmental factors, such as T, TDS, DO, pH and TN [10, 18 34]. Mantel test was performed to find correlations between microbial communities and environmental factors in their respective growth areas. In Fig. 4, we found that bacteria were significantly correlated with TDS (p < 0.05) and AAPB was significantly correlated with T (p < 0.05). Meanwhile, the correlation between AAPB abundance and water quality parameter was determined by Pearson correlation analysis. It can be seen from Fig. 4 that the abundance of AAPB was significantly positively correlated with COD, and COD was also positively correlated with T, pH, DO, NO3–-N, TP and the number of people (NP), indicating that the abundance of AAPB was affected by environmental factors and human production activities (p < 0.05). Interestingly, the AAPB community had no relationship with the qPCR of AAPB.
3.5 Distance-decay dissimilarity of bacterial and AAPB communities
Bacterial and AAPB community geographic differences in TGR area were evaluated by analyzing PCoA; the results were shown in Fig. 5. The first two PCs of Bray-Curtis distance accounted for 30.18% and 29.38% of the community variations in the bacterial and AAPB communities, respectively. It was worth noting that only AAPB community composition has significant spatial changes by PERMANOVA tests. In addition, only value of PC1 had obvious difference in AAPB communities. The results of PERMANOVA showed that 6.9% and 10.5% of the total variation in the bacterial and AAPB communities were explained by geographic location. The distance-decay pattern was further performed to assess community similarity based on the geographic distances and environmental variables (Fig.S5). We observed that the community similarity of bacterial and AAPB communities in the TGR did not change with distance for geographic and environmental variables (r2 < 0.05). According to the above results, geographical distance and environmental variables contributed little to the change of microbial community.
3.6 Ecological assembly process of bacterial and AAPB communities
The results of VPA indicated that the contribution of environmental factors and spatial factors was less than 5% to the impact of sediment benthic communities (Fig. 6). Environmental and spatial factors could not explain about 95% of the changes in bacterial and AAPB communities. The null model was to further explore the microbial community composition structure. The results showed that the dominant driver of bacterial and AAPB community formation in TGR sediments was stochastic processes (61.92% and 94.36% for bacteria and AAPB, respectively, Fig. 6 and Fig. S6). This suggested that dispersal limitation factor was the dominant factor affecting the bacterial and AAPB communities. It was worth noting that drift also affects the communities of AAPB assembly. Although the contribution of the determinate processes was relatively low, the heterogenous selection of bacterial communities was distinctly higher than that of AAPB communities (38.08% vs 5.26%). Interestingly, pure spatial variation without environmental components (diffusion limitation) contributed little in VPA, but the main factor in the biological assembly process was dispersal limitation based on null model. In summary, the assembly process of bacterial and AAPB communities was mainly controlled by stochastic processes in the TGR.
3.7 Relationships between bacterial communities explored by network approach
A total 2 network analysis were established to study the potential interactions between microorganisms of 16S rRNA and pufM gene in different habitats. Table 1 contained the topological characteristic parameters such as average degree and total nodes. The nodes and edges of the network based on 16S rRNA were obviously higher than those of pufM gene network. The size of each edge represented the strength of correlation; specifically, stronger the correlation, thicker the edge [35]. The size of the node represented the number of edges connected; as the number of connected edges increases, the node became larger [3]. The average degree indicates the closeness of the relationship between each node and other nodes. Therefore, the connection between total bacteria was closer. The high modularity (> 0.4) of two networks proved that their differentiated modular structures were not randomly constructed [24]. The high modularity showed that the microbial network was more stable, and the correlation of the network was measured by clustering coefficient. Therefore, the stability of the microbial network of 16S rRNA gene was lower than that of pufM gene (Table 1).
The most closely connected nodes in each module were defined as a “hubs”, which served as indicators of the potential metabolisms of other members in the same module [36]. In pufM gene microbial network, the hubs were Tabrizicola spp. and Gemmobacter spp. for modules I and Ⅵ, respectively. They were from α-proteobacteria. However, it was difficult to identify the hub in modules in 16S rRNA gene network because many genera were closely linked. Meanwhile, we observed that rare species were also hub in the module of 16S rRNA gene network. In the pufM gene microbial network, unclassified had positive/negative correlation with multiple genera in modules I and Ⅱ.
Table 1
Network properties of bacterial communities in the TGR
Parameters
|
16S rRNA gene
|
pufM gene
|
Nodes
|
553
|
45
|
Edges
|
6662
|
46
|
Average degree
|
24.092
|
2.044
|
Average weighted degree
|
16.377
|
1.554
|
Network diameter
|
11
|
7
|
Graph density
|
0.044
|
0.046
|
Modularity
|
0.415
|
0.844
|
Average clustering coefficient
|
0.472
|
0.824
|
Average path length
|
3.254
|
1.782
|