Dongping Lake is an important hub of the world's largest artificial water conservancy project, China's South-to-North Water Diversion Project. It receives its primary inflow from the Dawen River in the east and discharges water to the Yellow River through the two manmade Chenshankou and Qinghemen sluice gates in the north (Figure 1), with a high frequency of regulation (e.g., flow releases and storage). It is a typical river–lake transition area with complex hydrological conditions and drastic changes in water quality (Chen et al., 2014). Especially when the flood was discharged by these artificial gate dams, the huge impact of water flow led to an unstable sedimentary environment and fluctuating water quality.

Based on the strategic position and functional zoning, we divided this lake–river system into four zones: the Dawen River inflow zone (the main inflow river of the lake, DR), the lake center zone (natural ecological protection center area, LC), the aquaculture zone (there are artificial release and naturally raised fish, AZ), and the Yellow River outflow zone (transferring water storage through the sluice gate and providing drinking water source area, YR). The samples were collected in July 2022, during which Dongping Lake is in the artificial flood discharge and regulation period. At each sampling site, we collected five water samples and five sediment samples with triplicate samplings. We collected 1.0-L water samples at a depth of 0.5 m in polyethylene bottles and collected sediment source samples to a depth of 5 cm with a Peterson mud sampler (TC-600BD, Qingdao, China). Samples used to determine physical and chemical properties were immediately sent to a laboratory for testing. Water samples for microbial determination were immediately filtered, and the filter membranes were securely placed at −80°C. Sediment samples were securely placed at −80°C for <48 h until we could extract microbial determination.

We measured the water temperature (T), dissolved oxygen (DO), pH, specific conductance of the water (SPC), and the oxidation–reduction potential (ORP) at the study sites using a HACH HQ30d portable measuring instrument (HACH, Loveland, CO, USA). We measured the flow velocity (V) using a Doppler current meter (SF-6526J-21, Beijing, China). We measured the water's transparency (SD) using a Secchi disk in situ. For the chemical properties, we measured the chemical oxygen demand (COD), total nitrogen (TN), ammonia nitrogen (NH4+-N), nitrate nitrogen (NO3--N), nitrite nitrogen (NO2--N), total phosphorus (TP), and phosphate (PO43-) values according to the standard methods defined by China's Ministry of Ecology and Environment (Supplementary Table S1). We measured all environmental parameters three times to mitigate random measurement errors, and we used the average value to represent each sample in our subsequent analyses. All environmental parameters that required laboratory testing were measured within 48 h.

Microbial bacterial 16s rRNA and metagenomic sequencing were conducted in our study. The 16s rRNA data were used for microbial community structure-related analysis, while metagenomic data were used for microbial gene abundance and key biochemical step analysis. The relevant sequencing processes and analysis methods are shown in the Supplementary Text description of the microbial 16S rRNA and metagenomic sequencing. Our sequencing data has been uploaded to the National Center for Biotechnology Information (NCBL) database. The accession number is SRP457080.

The data were analyzed using the SPSS software, version 22.0 (https://www.ibm.com/spss). We used a one-way analysis of variance (ANOVA) with location as the level to test for significant differences. Where the ANOVA results were significant, we used least-significant difference (LSD) tests to identify pairs of values that differed significantly, with significance defined as p < 0.05. The independent sample t-tests were used for significant differences in topological values for the empirical and random networks (for specific analysis, see Supplementary Text description of the data analysis tools and methods).

In the water column (Table 1), the nutrient concentrations of nitrogen (TN, NH4+-N, and NO3--N) and phosphorus (TP and PO43-), which were major pollutants in the water environment (Dong et al., 2021), showed the order of YR > DR > AZ > LC. In particular, the values of NH4+-N, NO3--N, TP, PO43- in the YR and DR zones were significantly higher than those of AZ and LC (p < 0.05). The concentration of COD also showed significant differences (p < 0.05), with the highest concentration in YR (19.0 ± 3.16), followed by DR (17.8 ± 1.17), AZ (16.0 ± 1.40), and LC (15.0 ± 2.00). In addition, although the flow velocity values did not show significant differences in different zones, they also showed a higher trend in the YR and DR. These results showed that the water quality in the lake zone was significantly better than that in the zones where rivers flow in and out (especially in the zones with artificially constructed gates and dams) during the regulation period of the river–lake system. In sediments, the concentration of nitrogen and phosphorus varies in different zones, but the difference was not significant. The COD content (0.58 ± 0.10 mg/g) was highest in the YR, by 0.02, 0.08, and 0.26 mg/g compared with the other zones, but the difference was only significant for the DR.

It was worth noting that the TN was strongly and significantly correlated with NO3--N concentration (R = 0.72, p < 0.01), as well as TP and PO43- concentration (R = 0.86, p < 0.01) (Supplementary Figure S1). Therefore, the regulation of nitrate and phosphate concentrations was the key to maintaining the steady state of water quality during regulation in this study, and microbial-mediated pathways related to nitrate and phosphate should also be a major focus of this study.

Hierarchical clustering and PCoA results showed that all samples in our study can be well classified and clustered into three categories: zones where rivers flow in and out (DRW and YRW), lakes (LCW and AZW), and sediments (DRS, LCS, AZS, and YRS) (Figure 2A; Supplementary Figure S2). In this study, these three categories were defined as river–water, lake–water, and sediment. ANOSIM results showed significant differences (R = 0.669, p = 0.0001) among these three groups (Supplementary Figure S3), which implied the clustering was reasonable and reliable (Shang et al., 2023). Based on this result, the composition and structure of the microbial community were analyzed (Figure 2B). Their values are shown in Supplementary Table S2.

At the phylum level, there were significant differences in the relative abundance of the dominant microorganisms in the river, lake, and sediment (p < 0.05). The relative abundance of Proteobacteria (47.9 ± 5.4%, mean ± SD; 40.9 ± 5.1%), Actinobacteriota (16.6 ± 3.3%; 20.2 ± 3.3%), and Firmicutes (8.3 ± 3.3%; 18.2 ± 3.3%) in rivers and lakes was significantly higher than that in sediments (p < 0.05), with percentages of 1.5 and 1.3, 3.6 and 4.4, and 1.5 and 3.3 times those in the sediment, respectively. In the sediment, the relative abundances of Chloroflexi (12.0 ± 3.3%), Desulfobacterota (6.3 ± 0.8%), and Acidobacteriota (7.1 ± 2.3%) were significantly higher than those in rivers and lakes (p < 0.05), with abundances of 9.7 and 11.2%, 5.3 and 5.9%, and 4.5 and 6.7%, respectively.

At the genus level, hgcI_clade, Pseudomonas, Comamonadaceae, Acinetobacter, Hydrogenophaga, and Exiguobacterium had a higher relative abundance in river water and lake water, but their abundances in the sediment were <1%. However, Thiobacillus, Anaerolineaceae, and Steroidobacteraceae had a high abundance (>2.0%) in the sediment but were rarely detected in the river and lake. The abundance of Arenimonas was significantly higher in the lake water (5.3 ± 1.3%) (p < 0.05), 4.7 and 5.0% more than in the river water and sediment, respectively. The Pseudomonas abundance was significantly greater in the lake water (4.7 ± 1.3%), followed by the river water (2.2 ± 0.6%) and sediment (0.5 ± 0.1%) (p < 0.05). The relative abundance of Comamonadaceae in the river water (4.7 ± 2.2%) was slightly higher than in the other zones (p < 0.05). The Acinetobacter relative abundance was 9.3 ± 2.7% in the lake water, which was 5 and 9.3% higher than in the river water and the sediment zone, respectively (p < 0.05). In addition, the relative abundance of Exiguobacterium in the lake-water zone reached 12.1 ± 3.6%, which was significantly higher than in the river water (3.9 ± 0.8%) and sediment (0.1 ± 0.0%) (p < 0.05).

Table 2 summarizes the characteristics of microbial molecular ecological networks. The microbial molecular ecological network constructed by our research was scientific and reasonable (see Supplementary Text description of the topological characteristics of the molecular ecological network). All ecological networks were dominated by positive correlations, with values ranging from 75.4 to 83.2% (Figure 3; Supplementary Table S3). It is worth noting that there was a significant difference among the results of average path length (p < 0.05), and the order was as follows: lake water (8.392) > river water (6.432) > sediment (6.077). The modularity of microbial communities showed significant differences (p < 0.05), and the order was lake water (0.728) > river water (0.726) > sediment (0.635). The keystone taxa can be used as indicators of the microbial community (Banerjee et al., 2018), and the keystone taxa were identified in this study (Supplementary Table S4; Figure 3).

The specific functions of microbial communities (such as nitrogen and phosphorus conversion) were closely related to metabolic pathways (Xiong et al., 2023). The results showed that the KO pathway (Supplementary Figures S4, S5) involved in the nitrogen and phosphorus cycle was highly expressed in lake water. Microbial functional genes were crucial for the transformation of nutrients (Li et al., 2022) to identify the genes that play key roles in the pathway of this river–lake system, we analyzed the expression levels of genes related to nitrogen and phosphorus pathways (Supplementary Figure S6) and the contribution of key genes (Figure 4). The results showed that the nirB (3.6%), narG (5.5%), narH (2.5%), and nasA (2.8%) genes related to the nitrate cycle and the ppk (6.5%), pstA (4.3%), pstB (4.2%), pstC (4.4%), and pstS (6.4%) genes related to the phosphate cycle were highly expressed in all samples (Supplementary Figure S6), and most (nitrate: nirB/narG/narH/nasA; phosphorus: pstA/pstB/pstC/pstS) were expressed significantly higher in the lake water than in the river water and sediment (Figure 4). These genes are closely related to nitrate reduction, phosphate absorption, and transport (Yu et al., 2018; Li et al., 2023; Liu et al., 2023). On this basis, the main microorganisms contributing these genes with high expression in lakes were identified in this study through the combination of species annotation and gene abundance calculation (Supplementary Figure S7; the detailed descriptions are shown in the Supplementary Text description of the analysis of key microorganisms contributing nitrogen and phosphorus genes). The results showed that Pseudomonas, Acinetobacter, Nocardioides, Microbacterium, Agromyces, and Aeromonas had a relatively high contribution of genes involved in the nitrate-phosphate cycle, especially in lake water. These microorganisms were defined as the key microorganisms that we should focus on in our study.

The results showed that there was a significant positive correlation between the abundance of key microbial abundance and the genes promoting nitrogen reduction (0.836, p < 0.001) and phosphorus absorption and transport (0.873, p < 0.001, Figure 5). The effect of key gene abundance on nitrogen (−0.898, p < 0.001) and phosphorus (−0.242, p < 0.001) concentration showed a highly significant negative trend. The environmental factors showed strong positive effects on the concentrations of nitrogen (0.697) and phosphorus (0.333), and the effect on nitrogen was significant (p < 0.05). There was a significant negative effect (−0.648, p < 0.001) of environmental factors on the abundance of key microbial species. In the environmental factors, the factor load of V (0.727) and COD concentration (0.839) was obviously high and significant (p < 0.001), and showed strong significant positive correlations with the concentrations of nitrate and phosphate and negative correlations with key microbial abundance (Supplementary Table S5), respectively.

Our study has revealed that flow velocity and COD concentration in river water were obviously higher than in lake water. In the Dongping river–lake system, the water surface in the lake zone was wide and the flow velocity was relatively slow (Liu et al., 2022), while in the inflow area, the water surface was narrow and the flow velocity was fast (Liu et al., 2016). During the regulation period, there was often a huge water impact, which undoubtedly increased the flow velocity in the outflow area (Zhang et al., 2012). The Dawen River mainly passes through rural areas, and organic matter in soil, farm compost, and decaying plants will enter the river with rainwater and irrigation water, resulting in a high COD concentration (Lv et al., 2018). In addition, some of the pollutants generated by human activities were discharged into the river and accumulated at the gates. When the flood was discharged, the gates were opened, and the pollutants were released along with the flow (Piirsoo et al., 2018; Watkins et al., 2019). This may also lead to higher COD concentration in the outflow zones.

We found that high flow velocity in river water zones has negative effects on the stability of microbial communities. In lake water zones with relatively low flow velocity, the microbial network has the longest path length, which suggests that when the environment is disturbed (e.g., through water inflows or releases), the response rate is the slowest for microorganisms in the lake zone, and the impact of environmental disturbance will not spread quickly to the whole microbial molecular ecological network (Jeanbille et al., 2016). The modularity of the microbial community in the lake water was the highest, which was similar to the study of Khu et al. (2023), who found that the low flow velocity conditions were more conducive to microbial growth and formed a higher degree of modularity. The high degree of modularity indicated that the function of the microbial community was more complex, and the microbial community has a strong ability to resist environmental stress (Feng et al., 2017). In addition, the numerous positive correlations in lake water represented that microorganisms were more inclined toward cooperative relationships, which promoted the stability of microbial communities (Layeghifard et al., 2017). The influence of higher flow velocity and COD concentration on the abundance of key microorganisms was negative in our study. When the water flow velocity increased, the desorption force generated by flow shear reduced the adsorption of microorganisms to suspended matter in the water, inhibiting the formation of biofilm (Tian et al., 2017). High COD concentration indicated that the water accepts more organic pollutants, and the biodegradation of these organic substances will consume DO (Liu et al., 2021). Most of the key microorganisms found in our study were aerobic denitrifying bacteria and phosphorus-accumulating organisms (PAOs), which need DO for their own reproduction (Zhao et al., 2022; Shen et al., 2023), so high COD concentration will inhibit the abundance of these key microorganisms.

We found that total nitrogen and total phosphorus concentrations depend on NO3--N and PO43- concentrations in the water column. Therefore, we conclude that controlling NO3--N and PO43- concentrations was the key to maintaining water quality stability during water regulation. Previous studies have also confirmed our findings about nitrate nitrogen and phosphate, which are often used as key indicators in water quality monitoring to judge the deterioration of water quality for subsequent control (Rashid and Romshoo, 2013; Bhurtun et al., 2019). Hence, microorganism-mediated steps that can affect nitrate and phosphate conversion become the focus of this study, which may also be a key step in water quality stability during water regulation.

We found that Pseudomonas, Acinetobacter, and Microbacterium were dominant and keystone taxa in the lake-water (Figures 2, 3) zones. The dominant microorganisms with high biomass can affect some broad community processes, and rare keystone taxa are more visible in narrower processes; they are the drivers of microbial community structure and function, such as a particular one in the nitrogen and phosphorus cycles (Banerjee et al., 2018). In addition, we found that Pseudomonas mainly contributes nirB, narG, and narH genes, while Acinetobacter can contribute nirB, pstB, pstC, and pstS genes. Microbacterium contributed the nasA, pstA, and pstB genes. For the nitrogen cycle, nirB is the gene corresponding to the nitrite assimilation reductase, narG/narH corresponds to denitrification, and nasA corresponds to nitrate assimilation reduction (Wang et al., 2018). For the phosphorus cycle, pstA, pstB, pstC, and pstS were the corresponding genes for phosphate absorption and transport, and these were the key to the function of PAOs (Singleton et al., 2021). In addition, previous studies have confirmed our findings: Pseudomoas was confirmed to be aerobic denitrifying bacteria (Shen et al., 2023), and the presence of Pseudomonas and Acinetobacter can accelerate the rate of nitrate reduction in sewage (Hu et al., 2022; Xiao et al., 2023). Pseudomonas and Acinetobacter are typical PAOs, that can realize excessive phosphate absorption through the aerobic phosphorus absorption process (He et al., 2020). Microbacterium was found to be denitrifying bacteria (Tang et al., 2023) and have good phosphorus removal potential (Wang et al., 2016). Hence, these microorganisms contribute significantly to the lower nitrogen and phosphorus concentrations in the lake zone. Hence, these dominant and keystone taxa may dominate the denitrification, phosphate absorption, and transport pathways to become key rate-limiting steps for water quality stabilization in this river–lake system.

We found that the inhibition of key genes on nitrogen and phosphorus concentrations was significant. This also showed that the key genes defined in this study can indeed promote the processes of denitrification and phosphorus absorption and transport. In addition, the increase in the abundance of key microorganisms was significantly beneficial to the abundance of nitrogen and phosphorus genes. These further confirmed the importance of microorganism-mediated denitrification, phosphorus absorption, and transport pathways for water quality stability. However, we found that high flow velocity and COD concentration may not be conducive to the stability of water quality during the regulation period. Hydrological disturbance, especially the increase in flow velocity, will cause the exchange of sediments and water in shallow lakes (Tang X. et al., 2020; Luo et al., 2022), resulting in the release of nitrogen and phosphorus from the sediments, which further increases the concentration of pollutants in the water phase (Huang et al., 2015; Peng et al., 2021). The higher COD concentration may be unfavorable to the propagation of aerobic denitrifying microorganisms and PAOs (as mentioned in Section 4.1), resulting in higher nitrogen and phosphorus concentrations in water.

In summary, in this study, maintaining the abundance of key microorganisms (especially the dominant and keystone taxa) and controlling the regulation flow as well as the concentration of pollutants accepted by the water body were crucial to maintaining the stability of water quality during hydrological regulation. Based on this, important strategies were proposed to protect water quality in this study (Figure 6A). On the scale of hydrological regulation, we suggest adopting the mode of cross-peak flood regulation to reduce the kinetic energy of flow to a certain extent and weaken the exchange between water and sediment. In terms of the water quality supervision scale of the accepted water bodies, we suggest increasing the monitoring of the incoming water quality of the Dawen River, removing the sewage behavior along the river, and regularly cleaning up the plant spoilage litter in the river to reduce the organic matter content. On the scale of the microbial community, we propose to add key microbial complex agents containing Pseudomonas, Acinetobacter, and Microbacterium in the inflow and outflow zones during the regulation period. Through the above strategies, the ideal microbial community gene function model was finally achieved (Figure 6B). It can promote nitrogen and phosphorus removal to maintain water quality stability.