3.2. Community characteristic analysis of periphytic biofilms
3.2.1. Periphytic biomass and the relationships with environmental factors
Ash-free dry weight (AFDW) and Chlorophyll-a (Chl-a) are important assessment indicators of periphyton biofilms, reflecting the current and later microbial structural characteristics, and are closely related to the water ecological environment(Huang et al., 2018). AFDW (total biomass) and Chl-a (algal biomass) of periphytic biofilms had a statistically significant difference among all samples (P < 0.01, Fig. 3a, Fig. 3b, Fig.S2). The total biomass ranged from 0.462 to 11.292 mg/cm2 in AFDW and algal biomass ranged from 0.741 to 16.277 µg/cm2 in Chl-a (Fig.S2). The AFDW in A sections was obviously higher than that of other sampling points. The difference in periphytic biomass accumulation may contribute to properties of water physicochemical characteristics, a higher periphytic biomass was most likely due higher stability(Zelnik and Susin, 2020). The thicker periphytic biofilms also coincided with higher pollution or flow intermittency, resulting in increased accumulation of organic matter(Smeti et al., 2019). The content of Chl-a was relatively lower in the range of oligotrophic U samples, it was likely that algal growth was resources limited under these low concentrations of nutrients(Weitere et al., 2021). Obviously, the contents of Chl-a had similar levels except for U samples, light and nutrients may improve enrichment of autotrophic biofilms(Bengtsson et al., 2018; Weitere et al., 2021). The Chl-a by contrast performed some advantages in characterizing water quality and local growth conditions. Such results shown that traditional descriptors for periphytic biofilms are relevant to reveal general alterations of the water quality dynamics and significant associations with several stressors in different aquatic ecosystems.
Table 1
Results of Spearman correlation analysis for the dependency of periphyton-related metrics on different predictor variables (triplicate times per site). The correlation was used to assess between variables and confidence level was 95%. (⁎ p < 0.05, ⁎⁎ p < 0.01, ⁎⁎⁎ p < 0.001)
Periphytic biomass | Temperature | pH | TN | TP | \({\text{N}\text{H}}_{4}^{+}-\text{N}\) | \({\text{N}\text{O}}_{3}^{-}-\text{N}\) | \({\text{N}\text{O}}_{2}^{-}-\text{N}\) | SRP | DOC |
AFDW (mg/cm2) | −0.2906 0.0619 | 0.0805 0.6122 | 0.4874 0.0011** | 0.4560 0.0024** | 0.2855 0.0669 | 0.3400 0.0276* | 0.5154 0.0005*** | 0.5239 0.0004*** | −0.2220 0.4448 |
Chl-a (µg/cm2) | −0.0739 0.6419 | −0.4055 0.0077** | 0.5421 0.0002*** | 0.3643 0.0177* | 0.3269 0.0346* | 0.4385 0.0037** | 0.3948 0.0097** | 0.2756 0.0771 | −0.2571 0.3740 |
Spearman's correlation analysis revealed a significant correlation of biomass with parameters such as TN, TP, \({\text{N}\text{O}}_{3}^{-}-\text{N}\), \({\text{N}\text{O}}_{2}^{-}-\text{N}\), and pH (P < 0.05, Table 1). Agricultural activities might play a key role as a nitrate source in explaining the correlations of AFDW and nutrients(Weitere et al., 2021). Feeding fertilizers used on agricultural reaches increased phosphorus concentration in water columns, thus contributing to an increase in AFDW(Vilches et al., 2013). However, it should be noted that such relationships were rather weak (correlation coefficients ranged from 0.3269 to 0.5421, Table 1), which can be ascribed to light intensity(Gu and Wyatt, 2016), organic pollutants(Weitere et al., 2021) or unmeasured parameters. Strong correlation between Chl-a and TN (P < 0.001) was observed, nitrogen inputs increased the cell density and Chl-a of the periphytic algae, but it also showed that pH was significant negatively related to Chl-a. The results of the correlation analysis confirmed that TN and TP may jointly promote the growth of periphytic microorganisms, this was consistent with the conclusions of previous studies(Huang et al., 2018; Stelzer and Lamberti, 2001; Weitere et al., 2021). AFDW indicated positive association with SRP rather than Chl-a, this can be contributed to light availability(Bengtsson et al., 2018) or temperature(Wijewardene et al., 2021). DOC served as a nutrient for the heterotrophic community but high DOC concentration may lead to limited light availability for autotrophs and for their photosynthesis(Levesque et al., 2017). The few relationships between DOC and biomass might not accurately reflect the influence of DOC on microbial community. Within this overall situations, such correlations were found between variables and Chl-a as well AFDW, which were further mostly explained by differences of environmental parameters.
The structural equation model (SEM) was conducted to further examine the effects of environmental explanatory variables on biomass of periphyton. The SEM including possible relationships between the biomass (AFDW and Chl-a) and experimental parameters was constructed. We also trimmed the initial SEM through removing nonsignificant pathways until most of path ways were significant(Allen et al., 2021). The goodness of fit for the SEM was assessed mainly by the CFI, Chi square and RMSEA, which indicating a good fit of the model to the original data.
Above all, we built multiple models to evaluate reliability and applicability of the model (Fig. 3c, Fig.S3). In the resulting model (Fig.S3b), we found that total biomass (AFDW) and algal biomass (Chl-a) were significantly directly or indirectly affected by TP, \({\text{N}\text{O}}_{3}^{-}-\text{N}\), and pH in this study. The SEM showed a strong positive effect of TP, \({\text{N}\text{O}}_{2}^{-}-\text{N}\), and TN on total biomass not only directly but also indirectly through the positive effect of TP on algal biomass (Fig. 3c, Fig.S3a). Furthermore, consistent with previous researches(Xia et al., 2020; Yan et al., 2019), nutrients such as nitrogen and phosphorus were also important factors affecting total biomass. The significant effect of TP on algal biomass was confirmed in another research(Chen et al., 2022). Indeed, algal biomass was an important part of total biomass of periphyton, TP increased algal biomass which in turn raised total biomass of periphyton finally leading to periphytic development. \({\text{N}\text{O}}_{3}^{-}-\text{N}\) indicated directly negative (p < 0.001, Fig. 3c) effect on total biomass. It can be that agricultural water utility associated nitrates and pesticides with variable intensity negatively effected total biomass(Allen et al., 2021). The SEM revealed that TN and \({\text{N}\text{O}}_{2}^{-}-\text{N}\) also contributed significantly to the shifts of total biomass, suggesting that nitrogen was an key factor, which affected the cell density and community structures of periphytic algae(Chen et al., 2022). The effect of TP on total biomass seemed to be more apparent than for algal biomass (Fig. 3c, Fig.S3b). This may be that phosphorus has been proven to limit primary productivity, algal biomass and the formation of harmful algal blooms(Paerl et al., 2016). pH was a critical environmental variable that directly influences the photosynthetic rate of algae, protoplasmic ion balance and the solubility of numerous chemicals including inorganic carbon(Agrawal, 2012). Briefly, the SEM further shown that significant directly relationships between biomass and environmental variables, this metric was suitable as bioindicators to identify multiple stressors. The results provided a set of specific stressors and response variables as relevant candidates for an experimental testing of interactions in multi-factorial experiments.
3.2.2. Microbial diversity and community richness within periphytic biofilms and plankton
Sequencing of 16S rRNA genes and 18S rRNA genes at the whole community level provided an almost unbiased profiling strategy for characterizing the changes in biodiversity in different samples (Table S4). Rarefaction curves and average values of Good’s coverage beyond 0.9995 directly reflected that microbial community could be well-represented at the present sequencing depth, almost all of samples reached their saturated stage (Table 2, Fig.S1). A total of 485,538 clean sequences were produced from 14 samples for plankton community. 14 samples with 699,668 clean sequences and 13 samples with 1,109,925 clean sequences were also generated for prokaryotes and eukaryotes in periphytic biofilms, respectively. The results shown that the number of average eukaryotic sequences were higher than that of prokaryotes, while ASVs were lower than these of prokaryotes which was in agreement with previous studies(Guo et al., 2021; Ragon et al., 2012), indicating that community richness composed of eukaryotes were relatively higher within different samples (Table 2). The alpha diversity of the planktonic communities in different water samples also exhibited certain differences. In the constitutive structure of the prokaryotes, the diversity such as Simpson, Shannon, Chao1, ACE and ASVs of the prokaryotic communities were significantly higher than that of their corresponding water columns among all aquatic habitats in this study (Table 2), these values obviously indicated that submersed substrates may provide microhabitats for periphytic prokaryotes adhesion and promote product secretions such as extracellular polymeric substances(Stelzer and Lamberti, 2001; Yan et al., 2019; Yu et al., 2022).
Table 2
Summary of alpha diversity of prokaryotes and eukaryotes in water and periphytic biofilms. Mean value of the different types samples.
Types | Samples | Average Sequences | Average ASVs number | Average Good’s coverage | The species richness and diversity |
Simpson | Chao1 | ACE | Shannon |
Prokaryotes | U | 48218.67 | 1851.33 | 0.9998 | 0.9974 | 1852.4425 | 1852.3775 | 6.8771 |
A | 46310.67 | 1817.00 | 0.9999 | 0.9966 | 1817.1255 | 1817.4131 | 6.8264 |
WN | 52629.00 | 1763.00 | 0.9999 | 0.9948 | 1763.2020 | 1763.4712 | 6.6411 |
MH | 52912.33 | 1742.00 | 0.9999 | 0.9954 | 1742.7084 | 1742.9984 | 6.6246 |
R | 49728.00 | 1778.00 | 0.9999 | 0.9969 | 1780.2222 | 1779.1421 | 6.7876 |
UW | 34868.00 | 1122.00 | 0.9997 | 0.9837 | 1124.3485 | 1123.5041 | 5.6966 |
AW | 33130.67 | 1248.33 | 0.9995 | 0.9711 | 1255.0224 | 1250.7944 | 5.7505 |
WNW | 34141.00 | 1367.67 | 0.9993 | 0.9603 | 1375.4602 | 1371.6450 | 5.7406 |
MHW | 35900.67 | 1174.67 | 0.9994 | 0.9446 | 1182.3544 | 1177.9465 | 5.3421 |
RW | 35708.50 | 1126.50 | 0.9996 | 0.9815 | 1131.1857 | 1128.5804 | 5.6963 |
Eukaryotes | U | 84672.00 | 1275.33 | 0.9999 | 0.9829 | 1276.4595 | 1276.7877 | 5.6291 |
A | 85438.00 | 851.00 | 0.9999 | 0.9645 | 851.3849 | 851.9160 | 4.7689 |
WN | 85661.33 | 921.33 | 0.9999 | 0.9609 | 922.6311 | 923.4488 | 4.8307 |
MH | 85600.67 | 924.67 | 0.9999 | 0.9168 | 925.4930 | 926.2628 | 4.5508 |
R | 85623.50 | 931.50 | 1.0000 | 0.9757 | 931.6071 | 931.8962 | 5.1317 |
Notably, the diversity indexes of U samples were significantly higher than these of the other four groups for eukaryotes (Table 2), indicating that the richness and evenness of species were relatively higher and ecological environment improved the value of the biodiversity. The eukaryotic biodiversity of samples except for U samples with different nutrients level suggested slight differences in periphytic biofilms. The high concentration of nitrogen could promote the homogenization of microbial communities in different sections, and the disappearance or replacement of species may decline biodiversity(Wang et al., 2022). Besides, it is well known that nutrient inputs below WWTP effluents and domestic sewage discharge may induce changes in species compositions(Drury et al., 2013; Esser et al., 2023). These findings suggested that biodiversity of community differed from various habitats and environmental ecological differences may have a significant impact on microbial community diversity in aquatic ecosystems.
To illustrate the differences of microbial structure, principal coordinates analysis (PCoA) and ANOSIM analysis based on Bray-Curtis dissimilarity were explored. Both PCoA and ANOSIM analysis indicated that prokaryotes (Fig. 4a) and eukaryotes (Fig. 4b) communities in all samples formed clear separate clusters. The distance feature reflected a clear separation in prokaryotes between plankton and periphytic biofilms according to samples groups (ANOSIM R = 0.8317, P = 0.001) and the two principal components explained 28.51% of the variations. The PCoA of the eukaryotic community also had a significant amount of statistically variations of the total variance (ANOSIM R = 0.8833, P = 0.001) and two axis totally accounted for 41.49% variations.
The PCoA analysis revealed that periphytic prokaryotes and eukaryotes communities could be significantly distinguished by different ecological habitats (Fig.S4). Densely populated urban area and near industrial WWTP sewage indicated similar influence to eukaryotic community. The ANOSIM analysis implied that eukaryotic communities (Fig. 4b) were less similar than prokaryotic communities (Fig.S4b) (Reukaryotes = 0.8833 > Rprokaryotes = 0.7160), implying that prokaryotic community dissimilarity may increase more slowly with environmental changes than prokaryotic community dissimilarity. Consistent with recent study(Xia et al., 2020), it may reveal that eukaryotes shown a stronger response to environmental variations than prokaryotes in periphytic biofilms(Liu et al., 2015). According to previous study, prokaryotic communities performed the highest dispersal abilities, followed by plankton, and that zooplankton showed the lowest(Soininen et al., 2011). The PCoA of prokaryotic community shown that differences between periphytic and planktonic community may due to the effects of selection and periphyton physicochemical properties of substrates types on the periphytic community(Yan et al., 2019; Yu et al., 2022). The previous studies also indicated that the similarity of planktonic communities was higher than that of eukaryotic communities, the reason was that smaller species may have greater niche plasticity than larger species, short-lived and small microorganisms were less tolerant than larger microbes in a dynamic colonization extinction equilibrium(Wang et al., 2015; Xia et al., 2020). Overall, such results underlined the fact that a combination of local, regional and biogeographic factors might influence the natural variations of community structure. Ecological habitats resulted in differences of microbial communities of prokaryotes and eukaryotes within periphytic biofilms.
3.2.3. Community structures and compositions of periphytic biofilms
Overall, a total of 15,500 ASVs were taxonomically assigned to 466 genera belonging to 38 phylum, 105 classes, 227 orders and 286 families within prokaryotic biofilms. 4,506 ASVs were separated to 436 genera belonging to 40 phyla, 105 classes, 215 orders and 332 families in eukaryotes. Additionally, 11,098 ASVs were classified into 257 genera belonging to 24 phylum, 53 classes, 132 orders and 182 families in plankton community.
Regarding dominant prokaryotes at phylum level (Fig. 5a, Fig. 5c), Proteobacteria (average relative abundance: 42.56%), Bacteroidota (21.57%), Cyanobacteria (9.22%), Verrucomicrobiota (9.18%), and Planctomycetota (3.38%) were identified to be abundant in periphytic biofilms. The abundances of these dominant phyla were different among aquatic habitats. Moreover, Flavobacterium (20.67%), Luteolibacter (11.35%), and Rhodobacter (5.08%) with distinct abundance were representative genera. Proteobacteria was the most abundant prokaryotic phyla and participated in the degradation of organic matter of wastewater treatment and the reduction of nitrate and nitrite, which at most of the sampling groups sever as typical members of aquatic microbial communities(Esser et al., 2023; Wang et al., 2019; Yu et al., 2022). It was reported that Verrucomicrobia was enriched in potassium-rich habitats(Mohiuddin et al., 2019). Flavobacterium was characteristic for periphytic biofilm and plankton especially in MH samples (Fig.S5a), it also was most abundant heterotrophs involving in various biogeochemical processes and the cycling of carbon in ecological habitats, commonly associating with urban environments and wastewater-polluted waters(Carles et al., 2022; Esser et al., 2023; Williams et al., 2013).
Within abundant eukaryotes (Fig. 5b, Fig. 5d), eukaryotic microbes phyla were mainly dominated by Bacillariophyta (29.64%), Chlorophyta (19.38%), Annelida (12.77%), Ciliophora (9.17%), and Rhizaria (7.83%). At the same time, dominant genera including Surirella (14.66%), Gomphonema (9.44%), Nais (7.81%), and Chaetogaster (6.23%) held a higher proportion, these exhibited significant abundance differences at the phylum and genus level within all samples. The relative abundance of Bacillariophyta in U samples and A samples was lower than that of other samples. Aquaculture pond mainly changed eukaryotic compositions, with a significant increase in Chlorophyta, and a significant decrease in Bacillariophyta. It might be caused by nutrient effects on eukaryotes composition that were stronger than grazing pressure from fish or aquatic predators. The higher abundance of Cyanobacteria and Chlorophyta in periphyton might be because the Chl-a was higher in A samples. Chlorophyta and Annelida were relatively higher in A samples and WN samples, indicating that diatoms phyla may be barely tolerant in high nutrient levels or less polluted rivers(Chen et al., 2022), these Chlorophytes have been linked to eutrophication(Nozaki et al., 2003) and communities dominated by autotrophs(Esser et al., 2023). Dominant Bacillariophyta can improve the metabolism of some heterotrophic microbes by secreting substances such as carbohydrates(Battin et al., 2016). Heterotrophic Ciliophora was significant bacterivores in eutrophic habitats and could directly stimulate the bacterial colonies as well biofilms development by grazing activity(Boehme et al., 2009). Nais and Chaetogaster were found that dominated significantly in A samples, followed by U and WN samples (Fig. 5b), these eukaryotic genera may be able to tolerant in aquaculture ponds or heavy eutrophication environments. These observations suggested significant alterations of microbial communities due to habitats differences.
The planktonic communities by contrast were dominated by the phyla Bacteroidota (29.30%), Proteobacteria (27.92%), Cyanobacteria (21.82%), and Actionbacteriota (12.08%) (Fig.S5). It can be seen that Flavobacterium (35.18%), Rhodoferax (8.78%), hgcI_clade (7.04%), Limnohabitans (6.82%), and Luteolibacter (6.21%) had higher percentage, yet with greater sample to sample variation. This dominance of Bacteroidota, Proteobacteria, Cyanobacteria, and Actionbacteriota in aquatic environments (water and sediment) were in accordance with previous studies(Wang et al., 2022; Wang et al., 2019). The abundance of Proteobacteria was much higher than that of plankton compared to periphytic prokaryotes. While relative abundance of Cyanobacteria comparatively was higher than that of periphytic biofilms, it was more likely to enriched in water body without plants and caused water eutrophication(Chang et al., 2020). High abundant species existing in varied nutrients level such as Rhodoferax and Flavobacterium indicated that they can adapt to high organic nutrients brought by effluents and runoffs from agriculture activities(Yan et al., 2019) (total phosphorus, ammonium and total nitrogen). hgcI_clade was the main parts of plankton in lakes and can consume nitrogen-containing compounds(Ghylin et al., 2014). Previous study have suggested that periphytic biofilms derived from surrounding water columns, distinct and shared microbial community exist between periphyton and plankton communities(Belles-Garulera et al., 2016; Yu et al., 2022). Planktonic prokaryotes and periphytic prokaryotes thus were very similar at phylum level, while planktonic communities in water seemed to change faster than periphytic prokaryotes. Some species of Bacteroidota were associated with the gut microbiota of many mammals and humans, which have been proposed as effective alternative fecal indicators(Wery et al., 2008). These findings may be attributed to environmental stressors (i.e., inorganic nutrients) that shifted the microbial dynamics at phylum level(Manirakiza et al., 2022). The results demonstrated that obvious compositional variations of prokaryotic communities in plankton and periphytic biofilms among different habitats, moreover some specific taxa were enriched in periphyton due to environmental properties.
To determine the spatial variability of microbial taxa among all sample types, the microbial taxa of prokaryotes and eukaryotes responsible for the differentiation were further identified applying indicator species analysis (Fig. 6). Prokaryotic genera (467 genera, including unclassified genera) and eukaryotic genera (637 genera) were explored indicator genera in periphytic biofilms.
Specifically, with regard to periphytic prokaryotes (Fig. 6a), 16 indicator genera such as Tabrizicola (0.88% of the average relative abundance), Oscillatoria_SAG_1459-8 (0.71%), and Candidatus_Aquirestis (0.52%) were included in U samples, these differentially abundant genera mainly affiliated to Proteobacteria (17.5% of all indicator genera) and Cyanobacteria (10%). 14 genera such as Dechloromonas (3.47%), Terrimonas (3.17%), and Geitlerinema_PCC-7105 (2.63%) were included in A samples, mainly assigned to Proteobacteria (15%), Actinobacteriota (5%), and Cyanobacteria (5%). The WN, MH and R samples with 2 genera mainly including Altererythrobacter (0.29%) and Gordonia (0.05%), 5 genera including Arcicella (2.04%) and Sphaerotilus (0.69%), and 3 genera including Pseudorhodobacter (5.86%) and Lactobacillus (0.15%) presented relatively greater abundance, respectively. These indicator genera mainly belonged to Proteobacteria (15%) and Firmicutes (5%). Within the plankton (Fig.S6), 4 genera including hgcI_clade (17.18%) and Pseudarcicella (2.67%) belonged to UW samples. 22 genera mainly including CL500-29_marine_group (13.99%), CL500-3 (3.34%) and Dinghuibacter (1.87%) existed in AW samples. 5 genera including Luteolibacter (16.03%) and Exiguobacterium (0.40%), 5 genera including Pseudomonas (3.60%) and GKS98_freshwater_group (1.39%) presented in WNW and RW samples, respectively. CL500-29_marine_group thriving in high nutrients has been reported that could degrade different forms of carbon-based compounds such as the dead phytoplankton, thus diminishing the carbon cycling and nutrient transfer(Lindh et al., 2015).
For eukaryotes sections (Fig. 6b), 12 genera such as Chaetophora (4.11%), Oophila (2.97%), and Uroleptus (0.40%) could serve as indicator genera in U samples, which mainly belonging to Chlorophyta (14.24%) and Ciliophora (8.16%). 20 indicator genera including Frontonia (5.77%), Desmodesmus (4.73%) and Coelastrum (2.88%), which mainly assigning to Chlorophyta (26.53%) and Ciliophora (4.08%). Moreover, 6 genera including Nais (21.95%) and Eucyclops(6.29%), 5 genera Carteria (1.58%) and Holosticha (1.02%), and 6 genera Aegagropilopsis (0.43%) and Sellaphora (0.43%) could be found in the WN, MH and R environments, respectively. All of these genera mainly were classified in Chlorophyta (8.16%), Ciliophora (6.12%), Arthropoda (6.12%), and Bacillariophyta (6.12%). Above all, the massive enrichment of indicator genera with stronger differences caused by ecological habitats existed in periphytic biofilms, they were well adapted to different aquatic environments.
3.2.4. Co-occurrence patterns within periphytic biofilms
Co-occurrence networks of periphytic biofilms were employed to investigate the possible interactions and understand the manifestation of biological interrelationships among the genera of prokaryotes and eukaryotes in each habitat (Fig. 7). Overall, a total of 904 genera (nodes) including prokaryotic genera (467 nodes) and eukaryotic genera (437 nodes) were used to construct the co-occurrence network, which were grouped into different ecological modules comprised of genera strongly co-occurring with each other. In specific environments, the microbial modules containing the largest number of genera were selected based on strong (Spearman |r| > 0.7) and significant (P < 0.001) correlations(Wang et al., 2022). Notably, the modules were observed to contain relatively high proportions of genera among all sampling groups. The nodes and edges of network pattern were calculated (Fig. 7). There were 30,701 (63.12%), 59,478 (50.89%), 18,197 (61.45%), 16,208 (64.05%), and 61,566 (50.97%) positive correlations among U, A, WN, MH, and R samples, respectively. The large number of correlations implied that prokaryotes and eukaryotes co-exist under ecological habitats can interact, whereas functional interdependencies may exist between these two biotic groups. The production of dissolved organic matter by eukaryotes had generally been reported to be a significant carbon source that was sufficient for prokaryotes to survive(Liu et al., 2014).
The average degree, clustering coefficient, average path, and graph density can illustrate the degree of aggregation and complexity within microbial community(Manirakiza et al., 2022) (Table S5). Results suggested that interactions were more complex in U groups compared to WN and MH groups due to higher average degree and modularity of U samples. The limitation of samples number of A and R samples however may not be calculated to reflect accurate topological properties. The interconnections between microbial genera may be attributed to the habitats distinctive, food and energy sources, niches, allelopathy, and seasons(Liu et al., 2019). Periphytic biofilms were tightly coated with extracellular polymers substance, which created a physical barrier for substances transfer and formed a closed microhabitats for closer biotic interactions(Seymour et al., 2017). Moreover, previous studies have suggested that most algae cannot directly synthesize nitrogen and thus depended on nitrate or ammonium produced by nitrogen-fixing bacteria, algal growth was likely to benefit from associations with bacteria that provided and delivered regenerated nitrogen, as well as bacteria that provided phosphorus via remineralization, synthesize vitamins or enhanced the bioavailability of micronutrients(Durham et al., 2015; Foster et al., 2011). In summary, intimately related taxa with higher the number of edges connections in prokaryotes and eukaryotes were inclined to be closer interconnected and clustered into independent modules among different microbial groups, complementary genera with variations further reflected the influence of ecological environments on periphytic biofilms.
3.3. Relationships between microbial communities and environmental parameters
The co-occurrence network was performed to elucidate the relationship between periphytic genera and most-explained environmental variables based on strong (Spearman |r| > 0.65) and significant (P < 0.05) correlations(Wang et al., 2022) (Fig. 8a, Fig. 8b). For prokaryotes, the network consisted of significant edges between 88 prokaryotic genera and 9 environmental variables, which were dominated by negative (63.77%) associations. pH, TN, and TP were the biggest nodes separately related with 25, 23, and 22 associations. Furthermore, there were significant correlations between prokaryotic genera and the four dominant species of Proteobacteria (42.05%), Cyanobacteria (14.77%), Verrucomicrobiota (6.82%), and Bacteroidota (6.82%), which formed the strong correlation nodes, respectively. Within eukaryotes, 79 eukaryotic genera and 9 environmental parameters were found and dominated by negative (66.38%) associations. pH, TN, \({\text{N}\text{H}}_{4}^{+}-\text{N}\), and \({\text{N}\text{O}}_{3}^{-}-\text{N}\) separately related with 31, 16, 15, and 14 associations exhibited strong correlations with the majority of eukaryotic genera. Significant correlations between eukaryotic genera and the three dominant species of Chlorophyta (26.14%), Ciliophora (14.77%), and Stramenopiles (11.36%) were found and formed strong correlation nodes, respectively. Previous studies have revealed that abiotic factors such as pH(Guo et al., 2021; Liu et al., 2014) and various nutrients(Li et al., 2022) could influence the distributions of eukaryotic communities. It can therefore be inferred that inorganic nutrients could be essential factors for eukaryotes variations(Xu et al., 2020). Besides, to disentangle the potential main drivers for microbial compositions, we constructed the relative importance of environmental predictors to shape the microbial structure by the random forest analysis(Zhang et al., 2021) (Fig. 8c, Fig. 8d). In this study, TP was found to be significantly important determinants in structuring microbial community, pH contributed mostly to prokaryotes but no significant difference. Also, pH, TN, and \({\text{N}\text{H}}_{4}^{+}-\text{N}\) had greater effects on eukaryotes despite no significance. Consistent with previous research, nutrients such as nitrogen and phosphorus were important stressors affecting microbial community(Yan et al., 2019). To sum up, pH and nutrients were relatively more important in affecting prokaryotes and eukaryotes compared to other physicochemical variables according to the network. pH, which is one of the main drive factors, was also the major factor driving variations in periphytic prokaryotes and plankton (Fig. 8a, Fig.S7). The effect of pH on eukaryotes was more apparent than for prokaryotes (Fig. 8c, Fig. 8d). There are several possible explanations for some negative correlations in microbial community. Firstly, the respiration of algae may lead to an increase in the pH of the periphytic biofilms that may have been harmful to bacterial dynamics(Wijewardene et al., 2021; Yan et al., 2019). Secondly, heterotrophic prokaryotes competed with eukaryotes for inorganic nutrients, which were more efficient at utilizing nitrogen from organic compounds than eukaryotes(Yan et al., 2019). Thirdly, some algae can release toxic chemicals to destroy bacteria, whereas bacteria in turn were able to produce molecules that kill algae(Amin et al., 2012). Significant negative effects of pH on prokaryotic genera was in line with the study(Guo et al., 2021), despite other authors found positive relationships between water pH and benthic bacterial diversity in rivers ecosystems(Zhang et al., 2021). Briefly, we found that the interactions between environmental factors and prokaryotes as well eukaryotes, implying that the periphytic microbial communities were affected by both water columns physicochemical factors and abiotic factors.