Particle-attached bacterial communities are more susceptible to seasonal environmental fluctuations in mesotrophic than eutrophic tropical reservoirs

Abstract

Particle-attached bacterial (PAB) communities play pivotal roles in water organic matter decomposition, nutrient cycling, and the natural self-purification processes. However, we know little about their responses to seasonal environmental fluctuations, under eutrophication in reservoir ecosystems. In this study, we studied the shifts of PAB communities to seasonal environmental fluctuations in tropical China. Trophic state index (TSI) indicated that the studied reservoirs ranged from mesotrophic to eutrophic state with a gradual increase in TSI from 31 to 58. In eutrophic reservoirs, Cyanobacteria, especially Raphidiopsis raciborskii, significantly increased in its relative abundance from the wet to dry season, but Synechococcales and Microcystaceae decreased. In contrast, the relative abundance of Clostridia, Bacilli, Coriobacteriia, Enterobacteriales, and Vibrionales were more susceptible to seasonal environmental fluctuations in mesotrophic than eutrophic reservoirs. PAB co-occurrence relationships in mesotrophic reservoirs varied more greatly in response to seasonal environmental fluctuations, compared with eutrophic reservoirs, in terms of topological properties of connectedness, average degree, robustness, and vulnerability. Our results further demonstrated that the seasonal stability of PAB co-occurrence relationships was strongly correlative with TSI through mediating key bacterial taxa and community biodiversity. We proposed that eutrophication dramatically reduced the seasonal variation of PAB community compositions and co-occurring relationships in reservoir ecosystems.

Introduction

Tropical reservoirs play significant roles in preserving biodiversity (Marques et al. 2018), supporting aquatic food webs (Barbosa et al. 2012), regulating climate conditions (Winton et al. 2019), and participating in carbon sequestration (Borges et al. 2015, Yang et al. 2018). They also contribute to the hydrological cycle through the regulation of water flow and runoff (Nhiwatiwa et al. 2021), and are vital for water purification, as they filter pollutants and regulate water quality (Becker et al. 2009). In addition to ecological functions, such reservoirs promote social-economic development, for they can provide citizens with drinking water, agricultural irrigation water, and electrical energy (Beaulieu et al. 2019). Considering the ecological and economic importance of these ecosystems, it is vital for understanding the ecosystem stability of tropical reservoirs and their underlying mechanisms.

Particle-attached bacterial (PAB) communities are integral to aquatic ecosystems, playing pivotal roles in organic matter decomposition, nutrient cycling, pollutant removal, and the natural self-purification processes of aquatic ecosystems (Lyons and Dobbs 2012, Liu et al. 2019, Yu et al. 2023). Particle-attached bacteria have been shown to have a more significant impact on biogeochemical cycles, particularly in nutrient-limited conditions, where they can dominate over free-living bacteria in terms of activity and abundance (Hu et al. 2020, Shen et al. 2022). They also exhibit different metabolic capabilities, with some studies highlighting their roles in processes such as denitrification and phosphorus cycling (Bižić‐Ionescu et al. 2015, Tang et al. 2015). The diversity and stability of PAB communities are influenced by multiple environmental factors, including nutrient availability and water quality, so they serve as sensitive indicators of water body health, reflecting environmental pressures and changes (Newton et al. 2011, Crespo et al. 2013). These PAB communities were found to be sensitive to environmental fluctuations, and their community compositions showed significant dissimilarities across time (Urvoy et al. 2022). Understanding the temporal stability of these communities is crucial for managing aquatic ecosystems, particularly in tropical reservoirs, where they contribute to ecological balance and water quality maintenance.

Eutrophication, primarily induced by human activities, may lead to a reduction in beta-diversity of PAB communities by reducing environmental heterogeneity (Monchamp et al. 2018, Geng et al. 2022, Xie et al. 2023) or cause increased heterogeneity of PAB community compositions by large variations in phytoplankton community assembly across eutrophic reservoirs (Zhang et al. 2021b). Different patterns of microbial communities driven by eutrophication may affect their responses to seasonal environmental fluctuations in reservoir ecosystems. In tropical reservoirs, where seasonal variations of climate and environmental conditions are pronounced, the impact of eutrophication on PAB communities may be further exacerbated during seasonal transitions. These aquatic ecosystems experience significant changes in water temperature (T) and nutrient loads, which, combined with eutrophication, could further alter the structure of PAB communities (Parveen et al. 2011, Savio et al. 2015). However, the impact of eutrophication on the stability of PAB communities in response to seasonal transitions has yet to be thoroughly investigated in tropical reservoir ecosystems.

To address this gap, our study examines the stability of PAB communities in response to seasonal transitions from dry to wet conditions in both mesotrophic and eutrophic tropical reservoirs, utilizing data from eight typical reservoirs. We aimed to explore the impact of eutrophication on (i) the variability of PAB community compositions in response to seasonal environmental fluctuations; (ii) the stability of the co-occurring bacterial taxa in response to seasonal environmental fluctuations; and (iii) the mechanisms maintaining the seasonal stability of PAB communities.

Materials and methods

Sampling site

In this study, we investigated eight reservoirs located in Guangdong Province, China (Fig. 1) in both wet (August) and dry (December) seasons, 2019. The eight are Dashahe Reservoir (DSH, 112.38°E, 22.58°N), Zhenhai Reservoir (ZH, 112.56°E, 22.56°N), Gaozhou Reservoir (GZ, 111.01°E, 22.14°N), Liuxihe Reservoir (LXH, 113.78°E, 23.76°N), Duobao Reservoir (DB, 116.36°E, 24.85°N), Huangzhuping Reservoir (HZP, 116.23°E, 24.72°N), Dongfanghong Reservoir (DFH, 115.69°E, 23.89°N), and Yitang Reservoir (YT, 115.61°E, 23.94°N). Dashahe Reservoir (in the upper reaches of the Dasha River) and Zhenhai Reservoir (in the upper reaches of the Zhenhaishui River) are located in Kaiping, Guangdong Province. Gaozhou Reservoir is situated in the upper reaches of the Jianjiang River of Maoming, while Liuxihe Reservoir is located in the lower reaches of the Liuxi River of Guangzhou, Guangdong Province. Duobao Reservoir (in the middle reaches of the Songyuan River), Huangzhuping Reservoir (in the upper reaches of the Shiku River), and Yitang Reservoir (in the lower reaches of the Tanxia River) are located in Meizhou, Guangdong Province. Dongfanghong Reservoir is located in Jiangmen City, Guangdong Province. These reservoirs were built around 1960s and 1970s, with general functions of irrigation, flood control, power generation, water supply, aquaculture, and so on.

According to the shapes and water flows of the reservoirs, six sampling points were set in each reservoir (Fig. 1). Using the YSI-6600V2 multi-parameter water quality monitor, water T, pH, and dissolved oxygen (DO) concentration of the water bodies were directly measured. Transparency (SD) was determined in situ using a Secchi disk. We used the repeated freeze-thaw-infiltration method to measure water chlorophyll a (Chla) (Kwartiningsih et al. 2021). According to the national water quality monitoring standard method (GB3838-2002, https://www.mee.gov.cn/ywgz/fgbz/bz/bzwb/shjbh/shjzlbz/200206/t20020601_66497.shtml), the concentrations of ammonia (NH₄⁺-N), nitrite (NO₂⁻-N), nitrate (NO₃⁻-N), total nitrogen (TN), and total phosphorus (TP) were determined.

We collected 1 L surface water samples by HQM-1 water collector, and six samples were collected from each reservoir as replicates. The samples were packed in sterile polyethylene plastic bottles, brought back in an ice box, and filtered through a Millipore filter membrane (particle-attached bacteria) with a pore size of 3 µm, immediately (Jain et al. 2019). After suction filtration, the filter membrane was stored at –70 °C for subsequent analyses.

DNA extraction, amplification, sequencing, and data processing

We extracted the genomic DNA from the samples according to the standard steps described in the instructions of the water DNA extraction kit. Used 1% agarose gel electrophoresis to detect the integrity and purity of sample DNA, and used NanoDrop One to identify the concentration of sample DNA. Each sample was polymerase chain reaction (PCR) amplified three times in a 50 µl reaction mixture, which contained 25 µl 2x PCR Premix Taq, 10 mM barcode-16S V4 region primers (515F/806R), 60 ng of genomic DNA, and 20 µl of nuclease-free water. The amplicons were sequenced using a NovaSeq6000 platform at Novogene Bioinformatics Technology Co., Ltd. (Beijing, China).

By filtering the raw sequencing data, low-quality data that interfered with subsequent analysis were removed. First, each sample data was separated from the offline data based on the Barcode sequence and PCR amplification primer sequences. After truncating the Barcode and primer sequences, we used FLASH (V1.2.7) software to splice the reads of each sample, and the resulting spliced sequences were the original Tags data (Raw Tags) (Magoč and Salzberg 2011). The raw tags obtained by splicing were strictly filtered by Qiime (V1.9.1) software to obtain high-quality Tags data (Clean Tags) (Caporaso et al. 2010). The tags obtained after processing were processed to remove chimera sequences. The tags sequences were compared with the species annotation database to detect the chimera sequences, and the chimera sequences were removed to obtain the final effective data.

Uparse (v7.0.1001) software was used to perform cluster analysis (Edgar 2013). By default, the sequences were clustered into operational taxonomic units (OTUs) with 97% similarity, and the sequences with the highest frequency of occurrence in each OTU were selected as the representative sequences of each OTU. The Mothur method and the SSUrRNA database of SILVA132 (http://www.arb-silva.de/) were used to analyze the species annotation of the OTU sequence (the threshold value was set at 0.8∼1), and the taxonomic information was obtained at each taxonomic level. Finally, the data of each sample were normalized based on the standard with the smallest amount of read counts of all samples (i.e. 42 000 sequences), so that the obtained normalized data could be used for subsequent analyses.

Statistical analysis

The water body’s physical and chemical factors were drawn through GraphPad Prism 9.0. Furthermore, a trophic state index (TSI) inspired by Carlson (1977) was calculated based on the concentration of TP, TN, Chla, and SD in each reservoir. Alpha diversity (species richness and Shannon index) of bacterial communities was calculated based on the normalized OTU abundance table using the “vegan” package in R software . Based on the Bray–Curtis distance, non-metric multidimensional scaling analysis of PAB community structure was performed (stress < 0.2). The method of the least squares was used to establish the linear regression between TSI and Bray–Curtis distance. For visualization of PAB community compositions, we performed heat tree by using the “metacoder” package in R (http://www.cloud.biomicroclass.com/CloudPlatform) (Wang et al. 2024).

We uploaded the OTU table to the iNAP platform (http://mem.rcees.ac.cn:8081) for online analysis of PAB co-occurring networks (Feng et al. 2022). The molecular ecological network (MEN) analyses pipeline was carried out to construct PAB co-occurring networks based on random matrix theory, an effective method for detecting non-randomness and system specificity embedded in complex communities (Zhou et al. 2010).We sequentially performed data standardization, Pearson correlation estimation, and random matrix theory-based method to determine adjacency matrix, the MENs, and network characteristics. MEN robustness is defined as the proportion of the remaining species in this MEN after random or targeted node removal, while vulnerability measures the relative contribution of the nodes to the global efficiency (Yuan et al. 2021). We calculated the MEN robustness when 50% of random nodes were removed and the vulnerability of MEN was indicated by the maximal vulnerability of nodes in the network (Yuan et al. 2021). Finally, Gephi was used to visualize the MENs.

We used redundancy analysis (RDA) in the linear model to relate the PAB community structure with the water environmental factors because axis lengths of gradients were <3 in detrended correspondence analysis. Before RDA, we conducted forward selection through ordiR2step automatic selection program to choose the best subset of explanatory variables in the vegan package in R. Partial least squares - path modeling (PLS-PM) analysis was carried out to build a framework to assess the effects of abiotic variables (water nutrition and TSI), and biotic factors (key taxa and Bray–Curtis distance) on PAB community stability by using R package “plspm” (Hulland 1999).

Results

Seasonal variations of environmental factors

TSI indicated that the studied reservoirs ranged from mesotrophic to eutrophic state with a gradual increase in TSI from HZP (31∼34, mesotrophy), YT (35∼36, mesotrophy), LXH (36∼38, mesotrophy), and GZ (44∼45, mesotrophy) to ZH (50∼54, eutrophy), DB (51∼53, eutrophy), DSH (51∼55, eutrophy), and DFH (51∼58, eutrophy) (Table S1). TSI in the four eutrophic reservoirs (ZH, DB, DSH, and DFH) were higher in wet than dry season but showed inverse trend in the four mesotrophic reservoirs (HZP, YT, LXH, and GZ). Meanwhile, the eutrophic and mesotrophic reservoirs in different seasons showed significantly different water properties (t test: P < .05 in all cases, Fig. S1). We found that the concentration of DO (from 7.99 to 8.75 µmol·l⁻¹) and phosphate (PO₄³⁻-P, from 3.8 to 4.1 µg·l⁻¹) showed gradually increasing trends from mesotrophic to eutrophic reservoirs. Moreover, compared with mesotrophic reservoirs, significant higher chlorophyll a (Chla, from 6.85 to 35.56 mg·l⁻¹), TN (from 0.45 to 0.98 mg·l⁻¹), and TP (from 0.016 to 0.038 mg·l⁻¹) and significant lower nitrite (NO₂⁻-N, from 0.2 to 0.3 µg·l⁻¹) and nitrate (NO₃⁻-N, from 0.07 to 0.21 mg·l⁻¹) were found in eutrophic reservoirs (t test: P < .05 in all cases, Fig. S1).

Seasonal variations of PAB community compositions

In both mesotrophic and eutrophic reservoirs, we found PAB alpha diversity (i.e. OTU richness and Shannon index) was not significant differences from wet to dry season (Tukey's multiple comparison in one-way ANOVA: P > .05 in all cases, Fig. 2A). However, PAB community structure of mesotrophic reservoirs (HZP, YT, GZ, and LXH) was obviously further between wet and dry seasons, while were much closer of eutrophic reservoirs (DSH DB, DFH, and ZH) (Fig. 2B). That implied that PAB community structure were more susceptible to seasonal environmental fluctuations in mesotrophic reservoirs than in eutrophic reservoirs. We then estimated PAB beta-diversity of each reservoir based on the Bray–Curtis distance (Fig. S2). It showed a gradually increasing trend from eutrophic to mesotrophic reservoirs and reservoir TSI and its PAB beta-diversity followed a negative linear relationship (R² = 0.29; P < .001, Fig. 2C), indicating eutrophication caused homogenization of reservoir PAB community compositions.

We found although the dominant phyla of PAB communities were similar in all investigated reservoirs, their relative abundances showed great differences (Fig. S3). The relative abundance of Planctomycetes and Alphaproteobacteria were totally increased from wet to dry season in all reservoirs, while Gammaproteobacteria (from 23.1% to 10.7%) and Actinobacteria (from 17.0% to 4.5%) decreased from the mesotrophic to the eutrophic reservoirs, but Bacteroidetes (from 5.4% to 35.0%), Cyanobacteria (from 12.2% to 27.2%), and Alphaproteobacteria (from 7.7% to 12.4%) increased (Fig. S3). We further performed heat tree to display the seasonal stability of PAB taxonomic classification in both mesotrophic to eutrophic reservoirs (Fig. 3). In eutrophy, Cyanophyceae, especially Raphidiopsis raciborskii, significantly increased from wet to dry season, but Synechococcales and Microcystaceae decreased (Fig. 3). Obviously, Firmicutes (especially Clostridia, Bacilli, and Negativicutes), Bacteroidales of Bacteroidetes (i.e. Prevotella, Alloprevotella, Paraprevotella, Bacteroides, Alistipes, Porphyromonas, and Parabacteroides), Atopobium of Actinobacteria, and some genera of Gammaproteobacteria (i.e. Shewanella, Serratia, Arsenophonus, Arenimonas, and Pseudoxanthomonas) were more susceptible to seasonal environmental fluctuations in mesotrophic than in eutrophic reservoirs, which obviously increased from dry to wet season.

Seasonal variations of the topological features of PAB ecological networks

To identify the impact of seasonal environmental fluctuations on potential PAB interactions in reservoirs with different trophic status, we constructed two seasonal MENs of mesotrophic and eutrophic reservoirs, respectively (Fig. 4). We found the PAB network complexity in eutrophic reservoirs was higher than that in mesotrophic reservoirs (Table S2). In mesotrophic reservoirs, the number of total nodes ranged from 372 to 501 and the total links ranged from 569 to 633 in two seasons, while the eutrophic networks owned 485–487 total nodes and 1709–2675 total links. For the network modules in eutrophic reservoirs, we focused on the large modules with at least 44 nodes, which contained a range of 73.82%∼82.53% of nodes, while the mesotrophic large modules only contained ~41% of nodes. Among these modules, the highest proportion of positive links was observed in mesotrophic wet networks (99.3%), suggesting that correlated associations among PAB communities greatly increased, while it reduced in the dry season (77.73%). Compared with the mesotrophic seasonal changes, from 99.3% in wet season to 77.73% in dry season, the positive links seemed to be more stable in eutrophic networks, the proportion of which were 76.07% and 75.48%, respectively.

PAB MEN patterns in mesotrophic reservoirs varied more greatly in response to seasonal environmental fluctuations, compared with eutrophic reservoirs, as indicated by multiple topological properties of the observed networks. We compared the node-level topological features [i.e. connectedness, average degree (avgK), robustness, and vulnerability] among the eight networks (Fig. S4). The eutrophic MEN's connectedness ranged from 0.740 in dry season to 0.769 in wet season, while those in mesotrophic MENs were only 0.067 in wet and turned into 0.440 in dry season. Similarly, avgK values in eutrophic reservoirs were all higher than those in mesotrophic reservoirs in both seasons. However, the mesotrophic network in wet season was the most stable, considering robustness (0.353) and vulnerability (0.043), while it became the most vulnerable in dry season with low robustness (0.264) and high vulnerability (0.255). Instead of such great seasonal fluctuation, eutrophic MEN stability showed little changes between wet and dry seasons, with robustness (0.332 in wet and 0.314 in dry) and vulnerability (0.057 in wet and 0.116 in dry). Obviously, the mesotrophic networks were greatly less complex than eutrophic networks and more susceptible to seasonal fluctuations.

The compositions at phylum level of large modules were visualized using pie charts (Fig. S5). Eutrophic and mesotrophic networks both had eight large modules. Briefly, Alphaproteobacteria, Bacteroidetes, Cyanobacteria, Gammaproteobacteria, and Actinobacteria dominated a large proportion of eutrophic large modules in both two seasons, while there was a great difference in mesotrophic dominant phyla in different seasons. In wet season, Firmicutes, Bacteroidetes, Actinobacteria, and Gammaproteobacteria dominated a large proportion of mesotrophic large modules, while that turned into Cyanobacteria, Gammaproteobacteria, Alphaproteobacteria, and Bacteroidetes in dry season. Moreover, a total of 29 module hubs and 6 connectors were identified from eight networks based on their topological roles (Fig. 5). The taxonomic information of these keystone taxa and their relative abundances were listed in Supporting Material (Table S3; Fig. S6). We found that most keystone taxa belonged to the predominant phyla which were almost identified in large modules in eutrophic networks of both wet and dry seasons, that is Bacteroidetes (i.e. Saprospiraceae), Alphaproteobacteria (i.e. Rhizobiales), unclassified Gammaproteobacteria, Cyanobacteria (i.e. Synechococcales), and Actinobacteria (i.e. Acidimicrobiia), while in mesotrophic networks, keystone taxa mainly belonged to unclassified Cyanobacteria in wet season, and affiliated with Bacteroidetes (i.e. Ignavibacteria) and Planctomycetes (i.e. Planctomycetales) in dry season (Table S3; Fig. S6).

The seasonal stability of PAB communities associated with abiotic and biotic variables

To comprehensively understand the mechanisms underlying seasonal shifts of PAB biodiversity and MEN stability in reservoirs, we linked the environmental factors to the key taxa that played dominant roles in seasonal turnover and network compositions, including Cyanobacteria, Planctomycetes, Bacteroidetes, Firmicutes, Verrucomicrobia, Acidobacteria, Alphaproteobacteria, and Gammaproteobacteria in mesotrophic and eutrophic reservoirs (Fig. 6A). The environmental variables controlling the major ecological clusters of bacteria were distinct between eutrophic and mesotrophic reservoirs. Eutrophic network compositions exhibited strong relationships with PO₄³⁻-P and TP, while mesotrophic community variations showed strong relationships with both nitrogen and phosphorus nutrients (especially NH₄⁺-N, NO₂⁻-N, NO₃⁻-N, TN, and PO₄³⁻-P) in two seasons. Bacterial seasonal variation was significantly affected by pH and DO. Moreover, PAB variation was significantly more correlated with TSI in mesotrophic reservoirs, comparing with eutrophic ones. Meanwhile, the environmental drivers of eutrophic and mesotrophic PAB compositions in different seasons were further analyzed by RDA (Fig. 6B). In the eutrophic reservoirs, TSI, chlorophyll a, and TN were the most significant drivers of community composition, while NO₂⁻-N, and NO₃⁻-N played dominant roles in mesotrophic reservoirs. Besides, T, pH, and TP were the most significant factors influencing both eutrophic and mesotrophic community compositions in wet season.

We then constructed a PLS-PM (Goodness of Fit value = 0.621) with water nutrition (NO₂⁻-N, NO₃⁻-N, TN, PO₄³⁻-P, and TP), TSI, key taxa, beta diversity, and networks stability (robustness and vulnerability) (Fig. 6C). It was found that PAB networks stability was significantly and directly impacted by water nutrition (path coefficient = 0.216) and PAB beta diversity (path coefficient = 0.5449). The key taxa had a significantly positive correlation with water nutrition (path coefficient = 0.9379). The key taxa (path coefficient = 0.6599) and TSI (path coefficient = -0.8748) had strong contributions to reduce beta diversity, while water nutrition (path coefficient = 0.7839) significantly increased beta diversity. Overall, these results indicate that different water environmental conditions under mesotrophic and eutrophic states (i.e. nitrogen and phosphorus nutrients, and TSI) played important roles in mediating the PAB biodiversity and network stability.

Discussion

PAB community compositions were more susceptible to seasonal shifts in mesotrophic than eutrophic reservoirs

Our results showed that the structure of PAB communities changed more significantly with seasons in mesotrophic reservoirs, and the characteristic differences in community composition were mainly reflected in changes in PAB relative abundance. This is consistent with observations from other reservoirs, including the Qingcaosha reservoir (Xu et al. 2018) and the reservoirs in Zhuhai City, Guangdong Province, China (Zhang et al. 2021b). In this study, we found the relative abundance of Firmicutes (especially Clostridia, Bacilli, and Negativicutes), Bacteroidales of Bacteroidetes, Coriobacteriia of Actinobacteria, and some taxa of Gammaproteobacteria (i.e. Enterobacteriales, Xanthomonadales, and Vibrionales) were more susceptible to seasonal environmental fluctuations in mesotrophic than eutrophic reservoirs. These variations in PAB community compositions can be attributed to environmental parameters (e.g. nitrogen and phosphorus nutrients) driven by seasonal alternations (Liu et al. 2020, Chen et al. 2022). In our study, water T, DO, and TSI were the main factors shaping the seasonal changes in PAB community compositions (Fig. 6). TSI significantly influences the distributions of bacterial community compositions in aquatic reservoir ecosystems (Zhang et al. 2023b). PAB communities in less eutrophic environments had been shown to be more specialized in degrading complex organic matter (Kiersztyn et al. 2019, Hu et al. 2020). As TSI increases, eutrophication can lead to a dominance of bacterial groups (e.g. Raphidiopsis raciborskii) that thrive in nutrient-rich environments (Shi et al. 2019). We found Raphidiopsis raciborskii had higher relative abundance during the wet season in eutrophic reservoirs, possibly because it can tolerate stratified and mixed conditions and can dominate during mixed periods (Berger et al. 2006, Bonilla et al. 2012, Soares et al. 2013). Recent studies had pointed out specific associations between heterotrophic bacterial taxa and cyanobacteria, which had been attributed to the different quality and quantity of dissolved organic matter produced by different cyanobacteria (Eiler and Bertilsson 2004, Bagatini et al. 2014, Louati et al. 2015). Bacteroidetes and Gemmatimonadetes often reached high relative abundances during cyanobacterial blooms and contributed significantly to material cycling and energy flow in freshwater ecosystems (Sonnenburg et al. 2010, Li et al. 2015, Steffen et al. 2017). Bacteroidetes, especially Sediminibacterium and Sphingobacteriales, are often found in high abundance following phytoplankton blooms, either adjacent or attached to phytoplankton (Allgaier and Grossart 2006, Li et al. 2011, Bagatini et al. 2014). However, other mechanisms may influence changes in PAB beta diversity in reservoirs under different trophic states. For example, seasonal differences, including changes in precipitation, T, and DO, may dominate bacterial community assembly by affecting stochastic and deterministic processes (Zhang et al. 2021a; Zhang et al. 2023b).

In this study, the lowest PAB diversity was all recorded in reservoirs with higher eutrophication (Fig. 2). A similar pattern was observed in bacterial communities in the eutrophic Changhuai River (Geng et al. 2022). Three phenomena may help explain the lower diversity. The first is environmental filtering, which is important for bacterial community structure by limiting or removing taxa that cannot compete under given conditions. In this case, environmental filtering may occur through organic nitrogen limitation, since TN concentration was significantly elevated in eutrophic reservoirs (Fig. S1). Second, the number of modules was lower in eutrophication (Table S2), indicating higher similarity between bacterial communities. Finally, tropical reservoirs are prone to cyanobacterial blooms, but some reservoirs are small in size and have limited self-purification capabilities (Donchyts et al. 2022, Yan et al. 2023). In the meantime, eutrophication may reduce PAB diversity through enhancing community homogenization (Geng et al. 2022, Shen et al. 2024).

PAB community network stability was more susceptible to seasonal shifts in mesotrophic than eutrophic reservoirs

In this study, compared with the mesotrophic MENs, the eutrophic MENs had more nodes and edges (Table S2), indicating that there were more ecological connections in the eutrophic network and contained more key taxa (Table S3). The lower complexity of the PAB community in mesotrophic reservoirs may be related to its higher alpha diversity (Goyal 2022). The ecological functions of key taxa in PAB communities were related to the supply and recycling of nutrients, and weaker complex symbiotic patterns were maintained in habitats with lower nutrient supply (Cao et al. 2018, Yu et al. 2023). Our observations in tropical reservoirs were therefore consistent with the described phenomena. However, robustness and vulnerability indicated that eutrophic reservoirs maintained greater stability across seasonal fluctuations than mesotrophic reservoirs (Fig. S4). Different symbiosis patterns and key taxa drove the community composition and network stability of PAB (Zhang et al. 2023a). For this study, the above complexity and stability analysis can be summarized as PAB communities with high diversity but low complexity and connectedness remained less stable (Yonatan et al. 2022). There is an insurance hypothesis that communities with high diversity may consist of multiple species that have redundant functions but differ in their ability to withstand disturbance, so that if the function of one species is reduced, other redundant species can substitute for that function (Yachi and Loreau 1999, Huelsmann and Ackermann 2022). Our results were not in line with this insurance hypothesis on diversity and stability. In contrast, they were consistent with May's diversity-stability paradox, which postulates that diversity is detrimental to an ecosystem because as the number of species increases, the ecosystem becomes less stable (May 1972).

Compared with the eutrophic network, more interaction changes under different seasons were observed in the mesotrophic network (Fig. 4). This was significantly correlated with changes in the composition of key bacterial communities in the network (Fig. 3; Fig. S5), indicating that the impact of environmental filtering on the community was manifested in changes in the abundance of key taxa (Tancula et al. 1992, Xu et al. 2003). In this study, Bacteroidetes (i.e. Saprospiraceae), Alphaproteobacteria (i.e. Rhizobiales), unclassified Gammaproteobacteria, Cyanobacteria (i.e. Synechococcales), and Actinobacteria (i.e. Acidimicrobiia) were the key taxa of eutrophic reservoirs in both seasons. However, Cyanobacteria were the key taxa in mesotrophic reservoirs in wet season, while they turned into Bacteroidetes (i.e. Ignavibacteria) and Planctomycetes (i.e. Planctomycetales) in dry season. The roles of these groups were interconnected and varied depending on environmental conditions, underscoring the complexity and adaptability of microbial communities (Yonatan et al. 2022, Shen et al. 2024).

TSI played an important role in shaping the seasonal stability of PAB community compositions and co-occurring relationships

TSI was a key indicator of nutrient enrichment in reservoir ecosystems and significantly affected the seasonal stability of PAB community compositions and co-occurring relationships (Fig. 6). Higher TSI, accompanied with eutrophication, can lead to a shift in bacterial community composition toward taxa that thrive in nutrient-rich conditions. This may lead to a decrease in PAB community diversity, as some specific species may out-compete more generalists (Hollibaugh et al. 2000, Rooney-Varga et al. 2005). Furthermore, it was supposed that the stability of microbial networks may be destabilized by environmental stressors, such as those induced by high TSI (Hernandez et al. 2021). However, we found the eutrophic network had higher stability, while the network became most vulnerable in mesotrophic reservoirs, which may be because increasing TSI could keep the balance of positive and negative interactions within bacterial communities to some degree, potentially enhancing their resilience and resistance to environmental perturbations. On the other hand, differences in the taxonomic affiliation of community members also influenced microbial resilience because they determined the cumulative number and identity of traits present in the bacterial communities (Philippot et al. 2021). The role of TSI in shaping PAB communities was further emphasized by the findings of Hu et al. (2020), who demonstrated that eutrophication, indicated by high TSI, led to convergent succession in aquatic communities. This suggests that TSI was a key factor mediating the transition of bacterial communities along a trophic state gradient. Our results suggest that eutrophication reduced the seasonal variations of PAB communities in reservoir ecosystems and dramatically changed the compositional succession of reservoir ecosystems.

Conclusion

In this study, we found the PAB community compositions in mesotrophic reservoirs were more susceptible to seasonal environmental fluctuations than those in eutrophic reservoirs. The relative abundance of Firmicutes (especially Clostridia, Bacilli, and Negativicutes), Bacteroidales of Bacteroidetes, Coriobacteriia of Actinobacteria, and some taxa of Gammaproteobacteria (i.e. Enterobacteriales, Xanthomonadales, and Vibrionales) were more susceptible to seasonal environmental fluctuations in mesotrophic than eutrophic reservoirs. Moreover, PAB co-occurrence relationships in mesotrophic reservoirs varied more greatly in response to seasonal environmental fluctuations, compared with eutrophic reservoirs, including the topological properties of connectedness, avgK, robustness, and vulnerability. The mesotrophic networks were less complex than eutrophic networks and more susceptible to seasonal fluctuations. Our results demonstrated that the stability of PAB communities was strongly correlated with TSI through mediating co-occurring bacterial taxa and community biodiversity.

Acknowledgments

We were grateful to thank Yuanyuan Liu and Jiexiang Zhang for their assistance with the experimental sampling and data analyses.

Author contributions

Jingyi Xiao (Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – original draft, Writing – review & editing), Sijia Hao (Methodology, Writing – review & editing), Li-Juan Xiao (Funding acquisition, Project administration, Writing – review & editing), Yang Yang (Writing – original draft, Writing – review & editing), Qinglong L. Wu (Writing – review & editing), Dan He (Investigation, Methodology, Writing – review & editing), Lijun Zhou (Project administration, Writing – review & editing), Ren Hu (Writing – review & editing), and Lijuan Ren (Conceptualization, Data curation, Formal analysis, Methodology, Project administration, Supervision, Writing – original draft, Writing – review & editing).

Conflict of interest

The authors declare no conflict of interest.

Funding

This work was supported by the National Natural Science Foundation of China (General Program) [grant number 32171517] and the Natural Science Foundation of Jiangsu Province, China [BK20211398].

References