The microbial community metabolic regime adapts to hydraulic disturbance in river–lake systems with high–frequency regulation
JieweiDING1
WeiYANG1,2,4✉Phone159-118- 1787Emailyangwei@bnu.edu.cn
XiaoxiaoLI3
XinyuLIU1
JiayueZHAO1
TaoSUN1,2
HaifeiLIU1
1State Key Laboratory of Regional Environment and Sustainability, School of EnvironmentBeijing Normal University100875BeijingChina
2Yellow River Estuary Wetland Ecosystem Observation and Research StationMinistry of EducationDongyingChina
3School of BiosciencesUniversity of SheffieldS10 2TNSheffieldUnited Kingdom
4No. 19 Xinjiekouwai St., Haidian District100875BeijingChina
Jiewei DINGa, Wei YANGa,b*, Xiaoxiao LIc, Xinyu LIUa, Jiayue ZHAOa, Tao SUNa,b, Haifei LIUa
a State Key Laboratory of Regional Environment and Sustainability, School of Environment, Beijing Normal University, Beijing 100875, China
b Yellow River Estuary Wetland Ecosystem Observation and Research Station, Ministry of Education, Dongying, China
c School of Biosciences, University of Sheffield, Sheffield, S10 2TN, United Kingdom
*Corresponding author: yangwei@bnu.edu.cn (Wei YANG), Tel. 159-118-1787, Fax 010-5880-5053
Corresponding author at: No. 19 Xinjiekouwai St., Haidian District, 100875 Beijing, China
A
Abstract
Background
River–lake ecosystems are crucial for the rational allocation of water resources, but frequent water diversion can destabilize water quality due to hydraulic disturbance. Microbial communities can respond rapidly to such external perturbations and influence these systems through the effects on nutrient metabolism. Therefore, understanding how microbial communities respond to hydraulic shocks in aquatic systems and whether they can adapt to such disturbances is essential for maintaining the health of river–lake systems. We used 16S rRNA and metagenomic sequencing technologies to examine the metabolic regimes of microbial communities during water regulation and non- regulation periods in river–lake systems.
Results
We found that hydraulic disturbance tended to drive the microbial community toward homogenized selection, thereby weakening its stability. Flow velocity (V) and the nitrate (NO3−) concentration significantly affected microbial community composition and abundance, with clear threshold effects. We established low (V = 0.284 m/s, NO3− = 0.031 mg/L) and high (V = 0.461 m/s, NO3− = 0.055 mg/L) thresholds. The microbial community enhanced its nitrogen metabolism by promoting denitrifying microbial genera (e.g., Pseudomonas and Flavobacterium) to counterbalance the impact of high V and NO3−, which increased NO3− levels. In addition, we predicted microbial community abundance using an artificial neural network to validate the thresholds we identified.
Conclusions
Our study provides theoretical support for understanding how microbial communities adapt to high-frequency hydraulic disturbances and offer valuable insights for managers to adjust water diversion strategies in a timely manner, thereby safeguarding the integrity of river–lake ecosystems.
Graphical Abstract
Keywords:
River–lake ecosystems
Hydraulic disturbances
High-frequency regulation
Microbial communities
Metabolic regimes
Threshold effects
Background
River–lake systems, as globally distributed hydrological continuums [1–2], serve as critical water conservation hubs that play a vital role in hydrological regulation while substantially contributing to socioeconomic development [3–5]. However, frequent water diversion activities and the construction of water conservation facilities, such as gates and dams, have complicated the hydrodynamic conditions within these systems [6]. Hydraulic engineering management may induce significant hydraulic disturbance, potentially triggering sediment resuspension and consequent water quality deterioration [7]. These anthropogenic impacts on river–lake ecosystems have emerged as a critical environmental attention worldwide [8].
Water diversion will also affect a wide range of organisms within river–lake systems. Artificial regulation activities can fragment the habitats in these ecosystems, disrupt food chains, and threaten the survival of animals and plants [9–10]. Moreover, the water quality degradation caused by regulation may trigger algal blooms, which can disrupt the ecological balance through impacts on phytoplankton and other communities [11–12]. Furthermore, changes in the sediment deposition pattern during water diversion can lead to uneven nutrient distribution in the water. This poses a challenge for microbiomes that depend on specific nutrient ratios, as they must adapt to these altered conditions [13–14]. It should be noted that among the many organisms affected by water diversion, microorganisms are sensitive and responsive to environmental changes. Changes in the composition and structure of microbial communities therefore provide clues to the overall state of an ecosystem [15–16]. This makes it necessary to clarify the impact of water regulation on microbial communities and determine whether they can adapt to the hydraulic disturbance caused by water regulation and can maintain the river–lake system’s ecological health.
As fundamental components of river–lake ecosystems, microorganisms drive nutrient cycling, and this cycling is essential for maintaining the stability of the aquatic ecosystems [17–19]. Microorganisms can respond to perturbations in their environment and can adapt by adjusting their metabolic pathways [20–22]. Ren et al. [23] studied the river–lake system of China’s Poyang Lake, and found that the deposition of organic matter and the input of exogenous nitrogen shaped microbial communities to promote organisms with strong metabolic potential for carbohydrates and nitrogen, both in the water and sediment. Similarly, Yuan et al. [24] found that microorganism in river channels were affected by concentrations of nitrite and dissolved oxygen. These environments can lead to proliferation of microorganisms that are efficient at denitrification and organic pollution degradation, thereby adapting to environmental disturbances caused by seasonal runoff. In addition, a recent study demonstrated that the death of algae in lakes can introduce sources of carbon, nitrogen, and phosphorus to surface sediments. In response to this environmental change, the microbial community can adjust its metabolic potential, contributing more genes related to the conversion of nitrogen, phosphorus and carbon [25].
However, most previous studies focused on static water in river and lake systems. The lack of research on dynamic systems is an important gap in our knowledge, as there are complicated and dynamic hydrodynamic conditions in river–lake systems, and particularly in those that are primarily used for water regulation [3]. On the one hand, the hydraulic shock created by flow management may induce the migration and succession of the original microbial community, which leads to instability of community composition and structure [26]. On the other hand, water diversion and hydraulic shocks may introduce additional pollution sources, such as nitrogen and phosphorus, that affect microbial community stability and alter the community’s metabolic trends [14]. How microbial communities respond to these multiple forms of disturbance caused by complex hydrodynamic conditions, whether they can adapt to this environmental disturbance, and by what mechanism remain unclear.
In the present study of a Chinese river–lake system, our objectives were to (1) elucidate the community assembly patterns and stability of microbial communities during regulation (R) and non-regulation (NR) periods; (2) identify the key factors that affected the stability or changes of microbial communities; (3) explore the composition and metabolic patterns of microbial communities during the R and NR periods; and (4) reveal the key mechanisms by which microbial communities adapt to hydraulic disturbance. We hypothesized that the change between R and NR periods would drive changes in the water environment that would cause the microbial community to change in response. Our results provide insights into the identification and prediction of changes in microbial communities during different regulation periods, and will support the maintenance of stable river–lake ecosystems during high-frequency hydraulic shock periods.
2. Materials and methods
2.1 Study area and sampling area division
Our study area was the Dongping river–lake ecosystem. Dongping Lake is the last lake used to regulate water flows in the eastern route of China's South-to-North Water Diversion Project (SNWDP) (Fig. 1). It experiences an extended and high-frequency period of water transfer each year as water is channeled through diversion routes into the river–lake system [27]. At the same time, it receives inflows from the Dawen River on the eastern side of the lake [28] and the Liuchang River on the southern side of the lake, and the lake sustains an aquaculture area on the northern side of the lake, near its outflow zone. Therefore, its internal hydraulic conditions are complex. During periods of regulation, the changing hydraulic conditions will alter the original microbial community composition, potentially weakening the community’s stability [26]. In addition, internal pollution caused by hydraulic disturbances also poses challenges to microorganism survival [29]. Therefore, it is imperative to identify the microbial community composition, its metabolic regime, and the mechanisms driving its shifts in both R and NR periods to ensure effective scientific management of the river–lake system.
Fig. 1
Study area and locations of sampling points. AC, aquaculture zone; DI, Dawen River inflow zone; LC, lake center zone; LI, Liuchang River inflow zone; LO, lake outflow zone.
We divided the study area into five distinct zones for sampling, based on the functions of different zones in the Dongping river–lake system and the design of the South-to-North Water Diversion Eastern Route. These zones were the Dawen River inflow zone (DI), the Liuchang River inflow zone (LI, which serves as the inflow channel from the SNWDP), the lake center zone (LC), an aquaculture zone (AC), and the lake’s outflow zone (LO). Importantly, our zoning method also accounted for the flow field characteristics of the Dongping river–lake system (Fig. 1) [26]. This approach enhanced the scientific rigor and rationality of our sampling design.
2.2 Sampling and sequencing
A
Samples were collected in April and August 2023. According to the schedule of the SNWDP, the period the end of June was designated as the regulation period, and the period from August fell within the non-regulation period [26]. In each zone, we established 10 sampling points to collect water samples. We detected Secchi disk transparency SD, pH, dissolved oxygen (DO), flow velocity (V) and water depth (D) in situ. In addition, we detected the total phosphorus (TP), phosphate (PO43−), total nitrogen (TN), ammonia nitrogen (NH4+) in water samples. nitrate nitrogen (NO3−), nitrite nitrogen (NO2−), total organic carbon (TOC), and chemical oxygen demand (COD) of the water samples. The specific sampling method and detection index are shown in the supplementary materials and the specific testing methods are presented in Table S1. We integrated 16S rRNA sequencing and metagenomic sequencing technologies to identify the microbial community’s species composition and performed gene annotation of the microbial communities during different periods. Details are provided in the supplementary materials.2.3 Statistical analyses
β nearest-taxon index (βNTI) based on a null model by the Majorbio cloud platform (https://cloud.majorbio.com/) was used to assess the relative importance of certainty and randomness in microbiome assembly [30–31]. See supplementary materials for a specific description. We calculated the average variation degree (AVD) to evaluate microbial community stability [32]. See supplementary materials for specific calculations. Mantel test was utilized to identify the driving factors that exhibited a significant association with the AVD [26]. To further explore the relationships between AVD and the factors and to assess the goodness of fit, we utilized the mgcv package for the R software to construct a generalized additive model and the ggplot2 package for visualization [33]. We used the rug plot feature in the Origin software with a fitted regression line and a contour line plot to represent the distribution of microbial abundance in relation to environmental factors and to identify potential environmental threshold ranges. The critical points where significant shifts in species abundance occur are defined as ecological thresholds in our study [34–35]. We used the Kruskal-Wallis non-parametric test to statistically validate the robustness of these threshold delineations [36]. Based on stochastic matrix theory, we conducted molecular ecological network analysis to construct a microbial co-occurrence network [3]. Genera with the top 1% in terms of the degree were classified as keystone microorganisms [26]. We used the neuralnet package for R to build an artificial neural network (ANN) model, and validated our microbial abundance predictions using the trained ANN model [37]. R2 = 0.62 indicated that the model fitting effect was good [38]. In addition, we used ggplot2 to generate contour line plots that helped us to identify potential threshold points for the environmental factors.
3. Results
3.1 Assembly patterns and stability of the microbial communities
The results showed that during the R period, the primary assembly mechanism of the microbial community was (HoS) (Fig. 2A). The relative importance of HoS was 0.56 ± 0.11 (mean ± SE) across all zones, which was 1.51 times the value observed during the NR period (0.37 ± 0.14) (Fig. 2C). The mean βNTI value during the R period was 2.92 ± 1.71 (Fig. 2B), which illustrates the main role of HoS during this period. During the NR period, DR was the predominant assembly mechanism, with a mean relative importance of 0.49 ± 0.14, which was 0.11 points higher than that during the R period (0.39 ± 0.09). In addition, the βNTI values of all zones except the LI zone during the NR period fell between − 2 and 2 (Fig .2D), indicating that DR was the primary driver of community dynamics. It is noteworthy that even during the NR period, HoS remained the dominant community assembly mechanism in the LI zone, where water transfer into the lake occurred, with a relative importance of 0.64.
Fig. 2
Microbial community assembly patterns in the (A, B) R and (C, D) NR periods. Homogeneous selection (HoS), heterogeneous selection (HeS), homogenizing dispersal (HD), dispersal limitation (DL) and drift (DR).
The mean AVD index of the microbial community (Fig. 3A) was significantly higher during the R period (0.53 ± 0.03) than during the NR period (0.48 ± 0.02) (p<0.05). Specifically, during the R period, the highest AVD values were observed in the LI zone (0.59 ± 0.06), followed by DI (0.58 ± 0.03), LO (0.50 ± 0.07), LC (0.49 ± 0.01), and AC (0.47 ± 0.02) had the lowest value. Even during the NR period, the LI zone maintained the highest AVD value, with the overall ranking in the order LI (0.51 ± 0.06) >AC (0.50 ± 0.02) >DI (0.48 ± 0.05) >LO (0.46 ± 0.04) >LC (0.44 ± 0.03). Interestingly, the trends of V (Fig. 3B) and NO3− (Fig. 3C) were consistent with the AVD values of the microbial community, and the other environment factors were not significantly correlated with AVD (Fig. S1). Therefore, determining key factors that affect microbial community AVD values was necessary, which we will discuss in the following section.
Fig. 3
(A) Average variation degree of microbial community and the changes in the (B) flow velocity and (C) nitrate concentration in the R and NR periods.
3.2 The key factors affecting the stability of microbial communities
We identified nitrate (NO3−), flow velocity (V), water depth (D), and pH as the key factors that influenced the AVD of the microbial community. During the R period (Fig. S2A), there was a significant positive correlation between NO3− and AVD (r = 0.24, p<0.05), as was V (r = 0.35, p<0.01). In the NR period (Fig. S2B), NO3− and V continued to show significant positive correlations with AVD (r = 0.22 and 0.27, respectively; both p<0.05). In addition, the correlation between pH and AVD showed a significant positive trend (r = 0.28, p<0.05), but D was significantly negatively correlated with AVD (r = 0.55, p<0.01). In addition, V and NO3− were strongly and significantly positively correlated during both the R period (r = 0.73, p<0.001) and the NR period (r = 0.79, p<0.001).
We further analyzed these variables using generalized additive models (Fig. 4). The results revealed weak but significant correlations between V and AVD (R2 = 0.29, p<0.05; Fig. 4A) and between NO3− and AVD (R2 = 0.21, p<0.05; Fig. 4C). In contrast, there were no significant correlations between D and AVD (Fig. 4B) or between pH and AVD (p>0.05; Fig. 4D). However, the correlation between V and NO3− was significant and well-fitted (R2 = 0.66, p<0.05; Fig. 4E). Overall, V and D most strongly influenced AVD, which suggests that AVD may be influenced by a combination of these variables rather than being controlled by a single variable.
Fig. 4
Predicted relationships derived from generalized additive models for the relationships between (A) flow velocity (V) and the average variation degree (AVD), (B) water depth (D) and AVD, (C) nitrate (NO3−) concentration and AVD, and (D) pH and AVD. (E) Relationship between V and the NO3− concentration.
3.3 Microbial community regimes and metabolic potential during different regulation periods
Our study revealed a distinct bimodal distribution of microbial abundance in response to changes in V (Fig. 5A) and NO3− (Fig. 5B). Specifically, microbial abundance was high at both low and high values of V and NO3−. In addition, the critical points of the regions with significant changes in abundance were shown in Fig. 5C. The impacts of V and NO3− on microbial abundance had significant threshold effects, with low (V = 0.284 m/s, NO3−=0.031 mg/L) and high (V = 0.461 m/s, NO3− = 0.055 mg/L) thresholds. The microbial abundance was higher below the low threshold and above the high threshold, but decreased greatly between these two thresholds. We defined three microbial community regimes based on these thresholds: regime 1 (R1) below the low threshold, regime 2 (R2) between the two thresholds, and regime 3 (R3) above the high threshold. Kruskal-Wallis tests demonstrated significant differences in the abundance of microbial communities among regimes (p<0.001; Fig. S3), confirming the validity of our regime division. We found substantial differences in the composition of the microbial communities among these regimes (Fig. 5D). In R1, the microbial community was dominated by Pirellula, Bacillus, Cavicella, Dinghuibacter, and Sporichthyaceae. In R2, the dominant taxa were hgcl_clade, Arenimonas, Brevundimonas, Aeromonas, and Chloroplast was relatively abundant. In R3, the dominant microorganisms were Flavobacterum, Comamonadaceae, Rhodoferax, Pseudomonas and Acinetobacter. These differences reveal clear differences in the composition of microbial communities across the three regimes. In particular, the microbial community composition in R3 exhibited remarkable similarity to the dominant microbial community observed during the R period (Fig. S4).
Fig. 5
Distribution of microbial community abundance under the different regimes: R1, below the low threshold, R2, between the two thresholds; and R3, above the high threshold with respect to (A) the flow velocity (V) and (B) the nitrate (NO3_) concentration. (C) Simultaneous effects of the two variables. (D) The community composition under the three regimes.
We found that the microbial co-occurrence network showed different characteristics between R and NR periods (Table 1). Specifically, the microbial co-occurrence network during the NR period exhibited more connections (326) than in the R period (276). In addition, the average clustering coefficient (5.46), modularity (0.72), and average path length (5.46) were higher during the NR period. Furthermore, during the R period, the keystone microbial taxa were primarily clustered in three modules, which we designated as modules 1, 2, and 3 (Fig. 6B). These modules accounted for 11.8, 8.0, and 8.9% of the total microbial co-occurrence network, respectively. Module 1 included the keystone taxa Aeromonas, Pseudomonas, Flavobacterium, Acinetobacter, Arthrobacter, and Arenimonas (Fig. 6A). Module 2 was composed of Fluviicola, Desulfobulbus, Sediminibacterium, and Thiobacillus. Module 3 consisted of Pseudoxanthomonas, Streptomyces, Thauera, and Hyphomicrobium. In contrast, during the NR period, the keystone microorganisms were mainly clustered into four modules, which we designated as Modules 4, 5, 6, and 7. The proportions of the keystone microbial degrees for these four modules in the whole network were 7.2, 10.3, 5.1, and 3.8%, respectively (Fig. 6C). The potential metabolic trends associated with the keystone taxa in each module are presented in Table S2.
Number of points | Number of links | Average degree | Average clustering coefficient | Modularity | Average path length | |
|---|---|---|---|---|---|---|
Regulation (R) period | 171 | 276 | 4.700 | 4.43 | 0.70 | 4.43 |
Non-regulation (R) period | 160 | 326 | 3.223 | 5.46 | 0.72 | 5.46 |
Fig. 6
(A) Composition of keystone taxa and (B and C) the proportions of connectivity within each module of the microbial co-occurrence networks during the R and NR periods, respectively. The number represents the microbial OUT serial number. Modules with similar functional potential are assigned the same color. OTU, operational taxonomic unit.
The abundance of denitrification-related genes was significantly higher during the R period (Fig. 7A). This was particularly evident for the genes narB, nasA, nirK, nirS, nirA, nirB, and nirD, which exhibited notable abundances values of 1544, 6534, 4908, 2122, 1002, 6082, and 1118 (OTUs), respectively. These genes are associated with key processes involved in nitrogen cycling. Among them, narB is related to the coding for the NO3− assimilation system (Nas) and dissimilatory NO3− reductase (Nar), and nasA is linked to the NO3− assimilation system (Fig. 7B). In addition, nirA, nirB, nirD, nirK, and nirS are all involved in the synthesis of nitrite reductase (Nir). Furthermore, we identified several microbial taxa (Pseudomonas, Curvibacter, Acidovorax, and Flavobacterium) that collectively contributed three or more genes related to nitrogen metabolism (Fig. 7C). These taxa may play a crucial role in driving community-level nitrogen cycling during the R period.
Fig. 7
(A) Nitrogen metabolism of the microbial communities during the water R and NR periods. Anammox, anaerobic ammonium oxidation; ANRA, assimilatory nitrate (NO3−) reduction to ammonium; DNRA, dissimilatory NO3− reduction to ammonium. (B) important enzymes involved in nitrogen metabolism during the water regulation period: Nap, periplasmic nitrate reductase; Nar, nitrate reductase; Nas, assimilatory nitrate reductase; Nir, nitrite reductase; Nor, nitric-oxide reductase; and Nos, nitric-oxide synthase. (C) microorganisms that contributed genes related to nitrogen metabolism during the regulation period.
3.4 Identification and prediction of microbial community status
We developed an ANN model to predict and evaluate the structure of the microbial community and abundance of various taxa. The input layer consisted of two key variables (V and the NO3− concentration), which we had previously identified as significantly influencing microbial abundance. The input layer comprised two neurons, the hidden layer comprised six neurons, and the output layer represented microbial community abundance (Fig. 8A). The model’s predictions showed a good correlation with the actual values of microbial abundance (R2 = 0.62 Fig. 8A) (r = 0.69, p < 0.05; Fig. 8B). In addition, we used the trained model to simulate microbial abundance values under various V and NO3− conditions to identify potential critical threshold points. We detected three possible thresholds (Fig. 8C), which we labeled as T1 (V = 0.325 m/s, NO3− = 0.038 mg/L), T2 (V = 0.448 m/s, NO3− = 0.034 mg/L), and T3 (V = 0.453 m/s, NO3− = 0.043 mg/L) (Table 2).
Fig. 8
(A) Structure of the artificial neural network model. (B) Prediction of the microbial abundance (MA) using an artificial neural network model. (C) MA as a function of the nitrate (NO3−) concentration and flow velocity (V). T1 to T3 represent three potential thresholds.
Threshold point | V (m/s) | NO3− (mg/L) |
|---|---|---|
T1 | 0.325 | 0.038 |
T2 | 0.448 | 0.034 |
T3 | 0.453 | 0.043 |
4. Discussion
4.1 Flow velocity significantly affected the microbial community composition
Our research revealed that hydraulic disturbances led to a shift in the microbial community assembly mechanisms towards homogeneous selection, thereby making the community composition more similar. High V reduced the stability of the microbial community. During the water regulation period, a substantial influx of water from the SNWDP increased flow velocity, thereby creating hydraulic disturbance and altering the existing water levels [39]. This short-term increase in V led to sediment resuspension, which in turn cause elevated nitrogen concentrations in the water [26]. This resulting high nitrogen loads created a homogenous selection environment for microorganisms, and potentially promoted species adaptation or elimination within the microbial community [40]. This process led to a more similar microbial community composition in areas affected by the hydraulic disturbance.
Even during the non-regulation period, microbial communities in the Liuchang River inflow zone were still dominated by homogeneous selection processes, which confirmed the influence of high V on community composition. In addition, we found that during the non-regulation period, the microbial community assembly in most zones was mainly driven by drift; that is, random events such as birth, death, and reproduction of microbial individuals affected the community composition [41]. This may be due to the relatively stable water flow during this period, which would have exerted lower environmental selection pressure on the microbial community. As a result, species would exhibit similar competitiveness for resource acquisition [42]. Therefore, the species turnover and community composition changes were largely dominated by the stochastic processes related to drift.
During the regulation period, the microbial communities had high AVD values, but this means that community stability was low [43]. In addition, the microbial communities exhibited shorter average path lengths under high V, which suggested closer interactions among species and a rapid response to external disturbance [44]. However, this also indicated that community composition was highly dynamic, and that the original structure was unable to persist in the face of intense hydraulic disturbances, resulting in weak community stability [45]. In addition, during the regulation period, high V increased nitrogen loads in the water body, shortened the contact time between microorganisms and pollutants, and affected the discharge of microbial metabolites [46]. These changes likely favored the persistence of genera that can tolerate or benefit from high nitrogen concentrations while eliminating those with low tolerance. Furthermore, the microbial community might change to increase the number of new genera that could adapt to or take advantage of elevated nitrogen loads [47], leading to significant shifts in community composition and reduced community stability. Therefore, the effect of V on microbial community composition is substantial and should not be overlooked.
4.2 Microorganisms adapt to hydraulic pulses through changes in the metabolic balance
We found that microbial communities can adapt to water disturbance by balancing their metabolic processes, with V and the NO3− concentration emerging as key factors that influenced these metabolic processes. During the regulation period, hydraulic shocks increased V and the NO3− concentration. Confronted with high V and NO3 levels, microbial communities may exhibit changes in their metabolic pathways to enhance nitrogen resource utilization efficiency [48–50]. This adjustment lets them achieve metabolic balance, improve survival and development, and better adapt to the new environmental conditions. It’s worth emphasizing that, compared to the non-regulation period that is characterized by low V, microbial communities during the regulation period exhibited distinct metabolic patterns, with significant changes in community composition and function.
In addition, the dominant and keystone species showed stronger nitrogen metabolic potential, especially with the appearance of genera capable of denitrification. The microbial community also showed high expression of genes related to denitrification processes. During the regulation period, several dominant microorganisms in the community, including Flavobacterium [51], Rhodoferax [52], Comamonadaceae [53], Arenimonas [54], Pseudomonas [55], and Thiobacillus [56] have been confirmed to exhibit an ability to remove nitrogen. Importantly, Arenimonas, Pseudomonas, and Flavobacterium also served as keystone microorganisms during this period. In addition, Pseudomonas and Flavobacterium were important contributors of genes related to nitrogen metabolism, as they provided nirA, nirB, nirD, and nasA, which are associated with denitrification. This agrees with similar previous studies [51, 57–58]. These genes are involved in the synthesis of Nir and Nas enzymes that promote NO3− metabolism [59–60], and they help microorganisms to metabolize excess nitrogen and adapt to the hydraulic disturbance caused by water diversion.
Our results also revealed that V and the NO3− concentration exerted significant threshold effects on the microbial community abundance. Specifically, lower and higher V thresholds corresponded to two distinct NO3− concentration thresholds. This suggests that at lower V, most microorganisms can adapt and maintain normal function, resulting in no clear metabolic preference by the community for any particular substance. For instance, at low V, the community comprises a diverse array of microorganisms involved in a range of metabolic processes: Module 4 is associated with carbon metabolism, Module 5 with sulfur metabolism, Module 6 with iron reduction, and Module 7 with nitrogen and phosphorus conversion. However, once V exceeded the higher threshold, those microorganisms that could not adapt quickly to the high V and nitrogen load would be eliminated, leaving only highly resilient genera to survive [47].
Consistent with this, we found that as the V and NO3− concentration increased, the abundance of microorganisms decreased significantly between the low and high threshold intervals (i.e., in R2). When flow V and the NO3− concentration reached the higher threshold, the competitive dynamics among microorganisms shifted to favor the proliferation of dominant genera capable of resisting external stresses. For example, at the higher threshold (R3), genera such as Flavobacterum, Comamonadaceae, Pseudomonas, and Rhodoferax emerged, driving an increase in the microbial community abundance. These highly resilient microorganisms drove the whole community’s nitrogen cycling towards denitrification through adjustments of their metabolic patterns. This, in turn, decreased the external NO3− concentrations, and this explains why the NO3− concentration in our study did not consistently increase with increasing V.
In addition, we developed an ANN model based on actual measurements and used it to predict community abundance under various combinations of V and NO3− concentration. The model confirmed the threshold effects of the two variables on microbial abundance, with predicted results closely aligning with measured results. This further enhances our understanding of the microbial community dynamics. Our findings provide valuable insights for river–lake system managers, as they will help them to tailor water diversion strategies to account for nutrient loads. This approach can maximize microbial community metabolism and stabilize the aquatic environment.
5. Conclusions
In our study, we examined the metabolic dynamics of microbial communities during different regulation periods in the Dongping river–lake system. We confirmed our hypothesis that the environmental change between regulation and non-regulation periods changed the water environment in ways that caused the microbial community to change in response. During these periods, high-frequency hydraulic disturbances caused increased NO3− concentrations. These external environmental perturbations drove microbial community assembly patterns toward homogenization, which decreased the stability of the community composition.
In addition, we found that the microbial community abundance was primarily influenced by V and the NO3− concentration, with distinct thresholds observed for their effects. Specifically, microbial abundance was higher in regions with low and high threshold values, but lower in regions between these thresholds. The variation in abundance was likely due to the emergence of genera such as Pseudomonas and Flavobacterium, which possess high nitrogen metabolic capacities. These genera proliferated in response to disturbances caused by high V and NO3− concentrations. Finally, we employed an artificial neural network to predict microbial abundance based on V and the NO3− concentration, thereby validating the identified thresholds. Our study provides novel insights into assessing the health of river–lake ecosystems through the lens of microbial metabolism.
Data availability
We have uploaded the 16S rRNA gene amplicon sequences and metagenomic sequences to the NCBI Sequence Read Archive (SRA) with the accession number PRJNA1281048 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1281048). And other data were all provided in in this study.
Contributions
A
Jiewei Ding: Writing – original draft, Data curation, Visualization, Investigation. Wei Yang: Writing – review & editing, Supervision, Funding acquisition. Xiaoxiao Li: Investigation, Methodology. Xinyu Liu: Investigation. Jiayue Zhao: Data curation. Tao Sun: Writing – review & editing, Resources. Haifei Liu: Methodology, Software.A
Acknowledgement
The authors are grateful to Majorbio Co., Ltd. (Shanghai, China) for assisting in the bioinformatics analysis of this manuscript.
A
Funding
This work was supported by the National Key Research and Development Program of China (No. 2023YFC3209003), and by the Major Scientific and Technological Innovation Projects in Shandong Province (2021CXGC011201).
Electronic Supplementary Material
Below is the link to the electronic supplementary material
A
Author Contribution
Jiewei Ding: Writing – original draft, Data curation, Visualization, Investigation. Wei Yang: Writing – review & editing, Supervision, Funding acquisition. Xiaoxiao Li: Investigation, Methodology. Xinyu Liu: Investigation. Jiayue Zhao: Data curation. Tao Sun: Writing – review & editing, Resources. Haifei Liu: Methodology, Software.All authors reviewed the manuscript.
A
Data Availability
We have uploaded the 16 S rRNA gene amplicon sequences and metagenomic sequences to the NCBI Sequence Read Archive (SRA) with the accession number PRJNA1281048 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1281048). And other data were all provided in this study.
References
1.
Battin TJ, Lauerwald R, Bernhardt ES, Bertuzzo E, Gener LG, Hall R Jr, et al. River ecosystem metabolism and carbon biogeochemistry in a changing world. Nature. 2023;613:449–59. https://doi.org/10.1038/s41586-022-05500-8.
2.
Jia J, Gao Y, Qin B, Dungait JAJ, Liu Y, Lu Y, et al. Evolving geographical gross primary productivity patterns in global lake systems and controlling mechanisms of associated phytoplankton communities since the 1950s. Earth-Sci Rev. 2022;234:104221. https://doi.org/10.1016/j.earscirev.2022.104221.
3.
Ding J, Yang W, Liu X, Zhao Q, Dong W, Zhang C, et al. Unraveling the rate-limiting step in microorganisms’ mediation of denitrification and phosphorus absorption/transport processes in a highly regulated river-lake system. Front Microbiol. 2023;14:1258659. https://doi.org/10.3389/fmicb.2023.1258659.
4.
Wang S, Ran F, Li Z, Yang C, Xiao T, Liu Y, et al. Coupled effects of human activities and river-lake interactions evolution alter sources and fate of sedimentary organic carbon in a typical river-lake system. Water Res. 2024;255:121509. https://doi.org/10.1016/j.watres.2024.121509.
5.
Wu S, Dong Y, Stoeck T, Wang S, Fan H, Wang Y, et al. Geographic characteristics and environmental variables determine the diversities and assembly of the algal communities in interconnected river-lake system. Water Res. 2023;233:119792. https://doi.org/10.1016/j.watres.2023.119792.
6.
Yang X, Zhang S, Tang C, Wu C, Ge Y. Impact of Three Gorges Dam construction on spatiotemporal variations in the hydrodynamic regime of Poyang Lake (China). J Hydrol. 2025;646:132302. https://doi.org/10.1016/j.jhydrol.2024.132302.
7.
Cofalla C, Hudjetz S, Roger S, Brinkmann M, Frings R, Woelz J, et al. A combined hydraulic and toxicological approach to assess re-suspended sediments during simulated flood events-part II: an interdisciplinary experimental methodology. J Soils Sediments. 2012;12:429–42. https://doi.org/10.1007/s11368-012-0476-2.
8.
Scanlon BR, Fakhreddine S, Rateb A, de Graaf I, Famiglietti J, Gleeson T, et al. Global water resources and the role of groundwater in a resilient water future. Nat Rev Earth Environ. 2023;4:87–101. https://doi.org/10.1038/s43017-022-00378-6.
9.
He F, Zarfl C, Tockner K, Olden JD, Campos Z, Muniz F, et al. Hydropower impacts on riverine biodiversity. Nat Rev Earth Environ. 2024;5:755–72. https://doi.org/10.1038/s43017-024-00596-0.
10.
Lu X, Ren J, Gao D, Yin L, Dou J, Li H, et al. Impacts of hydrodynamic variation on submerged macrophytes in lakes: a review. Acta Ecol Sin. 2022;42:4245–54. https://doi.org/10.5846/stxb202012243261. (in Chinese with English Abstract).
11.
Chen L, Wan N, Liu B, Liu G, Zhang Y. Effect of water diversion on phytoplankton in the Jihongtan reservoir revealed by eDNA. Environ Sci Technol. 2024;47:80–8. https://doi.org/10.19672/j.cnki.1003-6504.1978.23.338. (in Chinese with English Abstract).
12.
Hou X, Hu X, Li Y, Zhang H, Niu L, Huang R, et al. From disruption to adaptation: response of phytoplankton communities in representative impounded lakes to China’s South-to-North Water Diversion Project. Water Res. 2024;261:122001. https://doi.org/10.1016/j.watres.2024.122001.
13.
Liang J, Yan M, Zhu Z, Lu L, Ding J, Zhou Q, et al. The role of microorganisms in phosphorus cycling at river-lake confluences: insights from a study on microbial community dynamics. Water Res. 2025;268:122556. https://doi.org/10.1016/j.watres.2024.122556.
14.
Tang M, Chen Q, Xiao X, Lyu Y, Sun W. Differential impacts of water diversion and environmental factors on bacterial, archaeal, and fungal communities in the eastern route of the South-to-North water diversion project. Environ Internat. 2025;195:109280. https://doi.org/10.1016/j.envint.2025.109280.
15.
Rieder J, Kapopoulou A, Bank C, Adrian-Kalchhauser I. Metagenomics and metabarcoding experimental choices and their impact on microbial community characterization in freshwater recirculating aquaculture systems. Environ Microbiome. 2023;18:8. https://doi.org/10.1186/s40793-023-00459-z.
16.
Martiny JBH, Martiny AC, Brodie E, Chase AB, Rodriguez-Verdugo A, Treseder KK, et al. Investigating the eco-evolutionary response of microbiomes to environmental change. Ecol Lett. 2023;26:S81–90. https://doi.org/10.1111/ele.14209.
17.
Tang X, Xie G, Shao K, Hu Y, Cai J, Bai C, et al. Contrast diversity patterns and processes of microbial community assembly in a river-lake continuum across a catchment scale in northwestern China. Environ Microbiome. 2020;15:10. https://doi.org/10.1186/s40793-020-00356-9.
18.
Bhatnagar S, Cowley ES, Kopf SH, Perez Castro S, Kearney S, Dawson SC, et al. Microbial community dynamics and coexistence in a sulfide-driven phototrophic bloom. Environ Microbiome. 2020;15:3. https://doi.org/10.1186/s40793-019-0348-0.
19.
Almog G, Rubin-Blum M, Murrell C, Vigderovich H, Eckert W, Larke-Mejía N, et al. Survival strategies of aerobic methanotrophs under hypoxia in methanogenic lake sediments. Environ Microbiome. 2024;19:44. https://doi.org/10.1186/s40793-024-00586-1.
20.
Paquette AJ, Bhatnagar S, Vadlamani A, Gillis T, Khot V, Novotnik B, et al. Ecology and biogeochemistry of the microbial underworld in two sister soda lakes. Environ Microbiome. 2024;19:98. https://doi.org/10.1186/s40793-024-00632-y.
21.
Liu Y, Mohamad O, Gao L, Xie Y, Abdugheni R, Huang Y, et al. Sediment prokaryotic microbial community and potential biogeochemical cycle from saline lakes shaped by habitat. Microbiol Res. 2023;270:127342. https://doi.org/10.1016/j.micres.2023.127342.
22.
Sierra MA, Ryon KA, Tierney BT, Foox J, Bhattacharya C, Afshin E, et al. Microbiome and metagenomic analysis of Lake Hillier Australia reveals pigment-rich polyextremophiles and wide-ranging metabolic adaptations. Environ Microbiome. 2022;17:60. https://doi.org/10.1186/s40793-022-00455-9.
23.
Ren Z, Qu X, Peng W, Yu Y, Zhang M. Functional properties of bacterial communities in water and sediment of the eutrophic river-lake system of Poyang Lake, China. PeerJ. 2019;7:e7318. https://doi.org/10.7717/peerj.7318.
24.
Yuan X, Wang M, Guo X, Wu D. Analysis of the seasonal changes in planktonic microbial diversity in urban river supplied with reclaimed water: a case study of the North Canal River. Environ Sci. 2022;43:4097–107. https://doi.org/10.13227/j.hjkx.202112023. (in Chinese with English Abstract).
25.
Chen Y, Li D, Liu S, Song X, Li Z, Sun J, et al. Deposited dead algae influence the microbial communities and functional potentials on the surface sediment in eutrophic shallow lakes. Environ Res. 2025;271:121072. https://doi.org/10.1016/j.envres.2025.121072.
26.
Ding J, Yang W, Liu X, Zhao J, Fu X, Zhang F, et al. Hydraulic conditions control the abundance of antibiotic resistance genes and their potential host microorganisms in a frequently regulated river-lake system. Sci Total Environ. 2024;946:174143. https://doi.org/10.1016/j.scitotenv.2024.174143.
27.
Sun R, Wei J, Zhang S, Pei H. The dynamic changes in phytoplankton and environmental factors within Dongping Lake (China) before and after the South-to-North Water Diversion Project. Environ Res. 2024;246:118138. https://doi.org/10.1016/j.envres.2024.118138.
28.
Zhang S, Li X, Ren Z, Zhang C, Fang L, Mo X, et al. Influence of precipitation and temperature variability on anthropogenic nutrient inputs in a river watershed: implications for environmental management. J Environ Manage. 2025;375:124294. https://doi.org/10.1016/j.jenvman.2025.124294.
29.
Song H, Li Z, Du B, Wang G, Ding Y. Bacterial communities in sediments of the shallow Lake Dongping in China. J Appl Microbiol. 2012;112:79–89. https://doi.org/10.1111/j.1365-2672.2011.05187.x.
30.
Zhou X, Lennon JT, Lu X, Ruan A. Anthropogenic activities mediate stratification and stability of microbial communities in freshwater sediments. Microbiome. 2023;11:191. https://doi.org/10.1186/s40168-023-01612-z.
31.
Zhou J, Ning D. Stochastic community assembly: does it matter in microbial ecology? Microbiol. Mol Biol Rev. 2017;81:e00002–17. https://doi.org/10.1128/MMBR.00002-17.
32.
Xun W, Liu Y, Li W, Ren Y, Xiong W, Xu Z, et al. Specialized metabolic functions of keystone taxa sustain soil microbiome stability. Microbiome. 2021;9:35. https://doi.org/10.1186/s40168-020-00985-9.
33.
Gleich SJ, Cram JA, Weissman JL, Caron DA. NetGAM: using generalized additive models to improve the predictive power of ecological network analyses constructed using time-series data. ISME Commun. 2022;2:23. https://doi.org/10.1038/s43705-022-00106-7.
34.
Blake C, Barber JN, Connallon T, McDonald MJ. Evolutionary shift of a tipping point can precipitate, or forestall, collapse in a microbial community. Nat Ecol Evol. 2024;8:2325–35. https://doi.org/10.1038/s41559-024-02543-0.
35.
Shang J, Li Y, Zhang W, Ma X, Niu L, Wang L, et al. Hysteretic and asynchronous regime shifts of bacterial and micro-eukaryotic communities driven by nutrient loading. Water Res. 2024;261:122045. https://doi.org/10.1016/j.watres.2024.122045.
36.
Ma R, Tian Z, Zhao Y, Wu Y, Liang Y. Response of soil quality degradation to cultivation and soil erosion: A case study in a Mollisol region of Northeast China. Soil Tillage Res. 2024;242:106159. https://doi.org/10.1016/j.still.2024.106159.
37.
Liu X, Nie Y, Wu X. Predicting microbial community compositions in wastewater treatment plants using artificial neural networks. Microbiome. 2023;11:93. https://doi.org/10.1186/s40168-023-01519-9.
38.
Shi L, Deng Q, Lu C, Liu W. Prediction of PM10 mass concentrations based on BP artificial neural network. J Cent South Univ. (Sci Technol.). 2012; 43: 1969–1974. https://doi.org/CNKI:SUN:ZNGD.0.2012-05-056. (in Chinese with English Abstract).
39.
Li M, Yang X, Wang K, Di C, Xiang W, Zhang J. Exploring China’s water scarcity incorporating surface water quality and multiple existing solutions. Environ Res. 2024;246:118191. https://doi.org/10.1016/j.envres.2024.118191.
40.
Sun M, Li M, Zhou Y, Liu J, Shi W, Wu X, et al. Nitrogen deposition enhances the deterministic process of the prokaryotic community and increases the complexity of the microbial co-network in coastal wetlands. Sci Total Environ. 2023;856:158939. https://doi.org/10.1016/j.scitotenv.2022.158939.
41.
Vellend M. Conceptual synthesis in community ecology. Q Rev Biol. 2010;85:183–206. https://doi.org/10.1086/652373.
42.
Dini-Andreote F, Stegen JC, van Elsas JD, Salles JF. Disentangling mechanisms that mediate the balance between stochastic and deterministic processes in microbial succession. Proc Natl Acad Sci U S A. 2015;112:E1326–32. https://doi.org/10.1073/pnas.1414261112.
43.
Chen L, Qin X, Wang G, Teng M, Zheng Y, Yang F, et al. Oxygen influences spatial heterogeneity and microbial succession dynamics during Baijiu stacking process. Bioresour Technol. 2024;403:130854. https://doi.org/10.1016/j.biortech.2024.130854.
44.
Guseva K, Darcy S, Simon E, Alteio LV, Montesinos-Navarro A, Kaiser C. From diversity to complexity: microbial networks in soils. Soil Biol Biochem. 2022;169:108604. https://doi.org/10.1016/j.soilbio.2022.108604.
45.
Guo B, Zhang L, Sun H, Gao M, Yu N, Zhang Q, et al. Microbial co-occurrence network topological properties link with reactor parameters and reveal importance of low-abundance genera. NPJ Biofilms Microbiomes. 2022;8:3. https://doi.org/10.1038/s41522-021-00263-y.
46.
Wang W, Ma C, Liu H, Fan Y, Liu G, Zhang K. Effects of hydraulic loading rate on the removal of pollutants from an integrated biological settling tank. Environ Sci. 2016;37:4727–33. https://doi.org/10.13227/j.hjkx.201605214. (in Chinese with English Abstract).
47.
Rong X, Zhou X, Li X, Yao M, Lu Y, Xu P, et al. Biocrust diazotrophs and bacteria rather than fungi are sensitive to chronic low N deposition. Environ Microbiol. 2022;24:5450–66. https://doi.org/10.1111/1462-2920.16095.
48.
Fang W, Tian W, Yan D, Li Y, Cao A, Wang Q. Linkages between soil nutrient turnover and above-ground crop nutrient metabolism: the role of soil microbes. iMetaOmics. 2025;2:e55. https://doi.org/10.1002/imo2.55.
49.
Jia Y, Hu X, Kang W, Dong X. Unveiling microbial nitrogen metabolism in rivers using a machine learning approach. Environ Sci Technol. 2024;58:6605–15. https://doi.org/10.1021/acs.est.3c09653.
50.
Liu Y, Wei H, Dai J, Liu S, Qiu D. Microbial nitrogen removal and the molecular mechanisms underlying modulation and switching of dissimilatory nitrate reduction pathways in Shewanella strains. Acta Microbiol Sin. 2024;64:4656–68. https://doi.org/10.13343/j.cnki.wsxb.20240716. (in Chinese with English Abstract).
51.
Abdelhamed H, Nho SW, Karsi A, Lawrence ML. The role of denitrification genes in anaerobic growth and virulence of Flavobacterium columnare. J Appl Microbiol. 2021;130:1062–74. https://doi.org/10.1111/jam.14855.
52.
Chen H, Qi L, Chen J, Xia Z, Li Q, Ao Z, et al. Mechanisms and differences of N2O emission characteristics in typical wastewater treatment processes. Environ Sci. 2025;45:718–26. https://doi.org/10.19674/j.cnki.issn1000-6923.2025.0030. (in Chinese with English Abstract).
53.
Liao H, Song C, Wan L, Shi S, Wang X. Effect of chelated iron on nitrogen removal efficiency and microbial community structure in the anaerobic ferric ammonium oxidation. Environ Sci. 2021;42:4366–73. https://doi.org/10.13227/j.hjkx.202012216. (in Chinese with English Abstract).
54.
Huang S, Yu D, Chen G, Wang Y, Tang P, Liu C, et al. Realization of nitrite accumulation in a sulfide-driven autotrophic denitrification process: simultaneous nitrate and sulfur removal. Chemosphere. 2021;278:130413. https://doi.org/10.1016/j.chemosphere.2021.130413.
55.
Cheng W, Yin Y, Li Y, Li B, Liu D, Ye L, et al. Nitrogen removal by a strengthened comprehensive floating bed with embedded pellets made by a newly isolated Pseudomonas sp. Y1. Environ Technol. 2024;45:208–20. https://doi.org/10.1080/09593330.2022.2102940.
56.
Yang Y, Gerrity S, Collins G, Chen T, Li R, Xie S, et al. Enrichment and characterization of autotrophic Thiobacillus denitrifiers from anaerobic sludge for nitrate removal. Process Biochem. 2018;68:165–70. https://doi.org/10.1016/j.procbio.2018.02.017.
57.
Fenn S, Dubern JF, Cigana C, De Simone M, Lazenby J, Juhas M, et al. NirA is an alternative nitrite reductase from Pseudomonas aeruginosa with potential as an antivirulence target. Mbio. 2021;12:e00207. https://doi.org/10.1128/mBio.00207-21.
58.
Luo Y, Luo L, Huang X, Jiang D, Wu X, Li Z. Characterization and metabolic pathway of Pseudomonas fluorescens 2P24 for highly efficient ammonium and nitrate removal. Bioresour Technol. 2023;382:129189. https://doi.org/10.1016/j.biortech.2023.129189.
59.
Shen C, Liu S, Su J, Tian P, Dai J. Rhizosphere bacterial community structure and function of Caragana korshinskii in semiarid desert area. Genomics Appl Biol. 2021;40:3508–17. https://doi.org/10.13417/j.gab.040.003508. (in Chinese with English Abstract).
60.
Wang Z, Wang S, Liu Y, Feng K, Deng Y. The applications of metagenomics in the detection of environmental microbes involving in nitrogen cycle. Biotechnol Bull. 2018; 34: 1–14. https://doi.org/10.13560/j.cnki.biotech.bull.1985.2018-0024. (in Chinese with English Abstract).
Total Word Count: 6804.
Number. of figures: 8.
Number. of tables: 2.
Figure captions
Figure 1. Study area and locations of sampling points. AC, aquaculture zone; DI, Dawen River inflow zone; LC, lake center zone; LI, Liuchang River inflow zone; LO, lake outflow zone.
Figure 2. Microbial community assembly patterns in the (A, B) R and (C, D) NR periods. Homogeneous selection (HoS), heterogeneous selection (HeS), homogenizing dispersal (HD), dispersal limitation (DL) and drift (DR).
Figure 3. (A) Average variation degree of microbial community and the changes in the (B) flow velocity and (C) nitrate concentration in the R and NR periods.
Figure 4. Predicted relationships derived from generalized additive models for the relationships between (A) flow velocity (V) and the average variation degree (AVD), (B) water depth (D) and AVD, (C) nitrate (NO3−) concentration and AVD, and (D) pH and AVD. (E) Relationship between V and the NO3− concentration.
Figure 5. Distribution of microbial community abundance under the different regimes: R1, below the low threshold, R2, between the two thresholds; and R3, above the high threshold with respect to (A) the flow velocity (V) and (B) the nitrate (NO3_) concentration. (C) Simultaneous effects of the two variables. (D) The community composition under the three regimes.
Figure 6. (A) Composition of keystone taxa and (B and C) the proportions of connectivity within each module of the microbial co-occurrence networks during the R and NR periods, respectively. The number represents the microbial OUT serial number. Modules with similar functional potential are assigned the same color. OTU, operational taxonomic unit.
Figure 7. (A) Nitrogen metabolism of the microbial communities during the water R and NR periods. Anammox, anaerobic ammonium oxidation; ANRA, assimilatory nitrate (NO3−) reduction to ammonium; DNRA, dissimilatory NO3− reduction to ammonium. (B) important enzymes involved in nitrogen metabolism during the water regulation period: Nap, periplasmic nitrate reductase; Nar, nitrate reductase; Nas, assimilatory nitrate reductase; Nir, nitrite reductase; Nor, nitric-oxide reductase; and Nos, nitric-oxide synthase. (C) microorganisms that contributed genes related to nitrogen metabolism during the regulation period.
Figure 8. (A) Structure of the artificial neural network model. (B) Prediction of the microbial abundance (MA) using an artificial neural network model. (C) MA as a function of the nitrate (NO3−) concentration and flow velocity (V). T1 to T3 represent three potential thresholds.
Table captions
Table 1 Characteristics of microbial co-occurrence networks.
Table 2 Thresholds (T1 to T3) predicted by the artificial neural network model for flow velocity (V) and nitrate (NO3−) concentration.