Fresh macrophyte, V. natans seedlings were collected from our aquatic plant breeding tanks containing 10 cm-deep aquarium black soil substrates (Anubias, Japan). These tanks have been used to cultivate V. natans since 2020. The original V. natans plants were transplanted from Honghu Lake (N29.827°, E113.476°) in Hubei Province, China. One microcystin (MC)-producing Microcystis strain (FACHB-915) was obtained from the Freshwater Algae Culture Collection at the Institute of Hydrobiology, the Chinese Academy of Sciences.

The selected plant seedlings (12 ± 1 cm high) and Microcystis cells were pre-cultured separately. Microcystis was cultured in 1/10 diluted BG-11 medium in controlled conditions with a temperature of 22 ± 3 °C, a 12:12 light: dark cycle, and a light intensity of 25 μmol photon (PAR) m⁻² s⁻¹. The Microcystis cells in the exponential growth phase were used for experiments.

The indoor microcosm experiments included three groups: one treated group where V. natans is exposed to Microcystis (TSV), one plant control group with only a monoculture of V. natans (V), and one Microcystis control group with only a monoculture of Microcystis (TS) (Figure 1). In the TSV group, V. natans was initially cultivated at a fresh weight of 2.1 ± 0.06 g L^-1. The experimental process for this group consisted of three periods: a pre-stress period (0–7 days) for macrophyte acclimation, a stress period initiated on day 8 by the introduction of Microcystis at an initial cell density of 3.94 ± 0.32 × 10⁶ cells mL^-1, and a subsequent recovery period after the suppression of Microcystis cells. The entire experiment lasted 33 days. On day 14, when Microcystis cells were no longer detectable, a second addition of Microcystis cells was introduced into the TSV group. Based on the integrated analysis of Microcystis cell density (Figure 2A), and the corresponding growth, physiological and metabolic responses of V. natans (Figure 3), the experimental timeline was divided in to a stress phase (days 8 to 19) and a recovery phase (day 20 onward). The stress phase was characterized by the coexistence of Microcystis and V. natans, accompanied by a gradual decline in Microcystis cell density. By the end of this phase, V. natans exhibited significant growth inhibition, elevated oxidative stress, and pronounced disruption in metabolic pathways. In contrast, the recovery phase began following the complete suppression of Microcystis, during which V. natans transitioned to a monoculture-like state. Throughout this phase, the plants showed marked growth recovery, alleviation of oxidative damage, and restoration of metabolic homeostasis.

The V and TSV groups each consisted of 12 replicates, whereas the TS group had 6 replicates. There was a total of 30 beakers (9 cm in diameter and 15 cm in height) with a 2 cm layer of black soil substrate (Anubias, Japan). The culture conditions during the stress period were the same as those during the preculture period. Three out of four replicates from the V and TSV groups were collected on days 7, 19 and 33 to harvest plants and phyllosphere samples. This collection enabled the investigation of plant and epiphytic microbial responses at three stages. In the TSV and TS groups, four replicates of Microcystis samples were collected every two days after inoculation to monitor temporal variation in cell density and extracellular microcystin concentrations. To maintain consistent experimental conditions and compensate for evaporation, an equal volume of 1/10 BG11 medium was added to each treatment group after each sampling.

A 5 mL of Microcystis samples was collected to determine the optical density at a wavelength of 680 nm (OD₆₈₀) by using a spectrophotometer (SPECORD210Plus, Jena, Germany). Another 1 mL of Microcystis samples was centrifuged at 10,000×g for 15 minutes. The supernatant was collected and diluted for the determination of extracellular microcystin (MC) concentration using an MC-LR ELISA kit (Institute of Hydrobiology, CAS). It has a detection sensitivity of 0.1 μg L⁻¹.

Fresh weight (FW) and leaf length were measured to observe plant biomass changes. Non-targeted metabolomics were employed to measure the changes of metabolites and metabolic process of V. natans in V and TSV groups. A subsample of fresh plant shoots (200 mg) was homogenized with liquid nitrogen for the determination of peroxidase (POD) and malondialdehyde (MDA) with assay kits from Nanjing Jiancheng Company, China (Gao et al., 2022). Exactly 50 mg of plant shoots were placed into a 2 mL grinding tube and extracted with 0.5 mL of methanol-aqueous solution. The detailed pretreatment and subsequent untargeted metabolomics analysis were the same as those in previous studies (Li et al., 2024). Raw metabolomics data have been uploaded to the BIG Submission under the project ID PRJCA053329.

An accurate weight of 1.0 g of plant leaves was collected and transferred to a sterile centrifuge tube containing 50 mL of 0.1 mol L^-1 PBS (pH 7.0). The tube was then placed on a shaker platform and agitated at 180 rpm for 20 minutes, vortexed for 1 min. This process was repeated twice, and the wash fluids obtained from both repetitions were collected to form suspensions containing epiphytic microbes. Subsequently, 100 mL of suspensions were filtered using a 0.22 μm cellulose acetate filter membrane, and the filter membrane was collected and placed in a 15 mL sterile centrifuge tube, then stored in a -80 °C freezer for further DNA extraction and analysis of epiphytic microbial communities (Gao et al., 2022; Wei et al., 2024) The bacterial V3–V4 hypervariable region of the 16S rRNA genes were amplified by PCR using primers 341 F(CCTAYGGGRBGCASCAG) and 806R (GGACTACNNGGGTATCTAAT). The 18S rRNA gene was amplified using the primers SSU0817F (TTAGCATGGAATAATRRAATAGGA) and 1196R (TCTGGACCTGGTGA GTTTCC). The data were analyzed through the free online platform of major bio cloud platform (Ren et al., 2022). All obtained raw sequence datasets have been uploaded to the NCBI Sequence Read Archive (SRA) with the accession number PRJNA1377903 and PRJNA1377893.

All data were analyzed with Microsoft Excel and SPSS 22.0. The p-value < 0.05 was considered statistically significant. Origin 2023 and GraphPad Prism software were used for data visualization and graph processing. Principal Coordinate Analysis (PCoA) was employed to visualize the differences in microbial communities based on the Bray-Curtis distance matrix. KEGG pathway enrichment analysis was performed using Python software (version: 1.0.0). The dominant microbial network was visualized utilizing the “reshape” package within R and Gephi software. The correlation between dominant microbial genera was analyzed using Spearman’s correlation coefficient. The normalized stochasticity ratio (NST) was calculated using the R package to quantify the relative contributions of stochastic and deterministic processes to microbial community assembly, and correct sampling bias through the Modified Stochasticity Ratio (MST). Deterministic process dominated the community assembly if the MST value is less than 0.5, otherwise, stochastic process dominated (Ning et al., 2019).

To explore the relationship between algicidal bacteria and metabolites, we utilized random forest regression analysis. All analyses were carried out using the “random forest” package in R software (version 4.5.1). For each metabolite, its concentration was taken as the response variable, and the abundance data of algicidal bacteria were used as the predictor variables to construct the model. Considering the limited sample size (n = 12), in order to make full use of the data for feature importance assessment, we adopted a full - dataset modeling approach. The model parameters were set as follows: the number of decision trees was set to 1000, and the number of variables randomly considered at each split was set to one third of the total number of features (i.e., 7). To evaluate the goodness of fit of the model to the training data, we computed the coefficient of determination (R²) and the root mean square error (RMSE). To assess the overall statistical significance of the model, we conducted a permutation test: by randomly shuffling the response variable (metabolite concentration) 100 times, reconstructing the model each time and calculating the R² value, a null distribution was established. The actually observed R² value was compared with the null distribution to compute the p value.

The cell density of Microcystis, as indicated by OD₆₈₀ values, exhibited a marked decline following the initial introduction to the macrophyte (TSV treatment) on day 8 and after the subsequent introduction on day 14. From day 19 onwards, the OD₆₈₀ value the water containing Microcystis decreased to 0.01 and remained at this low level (Figure 2A). In contrast, the OD₆₈₀ values in the TS group (Mono-culture of Microcystis cells as control) increased from 0.24 on day 8 to 0.44 on day 33.

Despite the inhibition of Microcystis cells, the extracellular MC concentrations continued to increase, reaching the maximum value during the late stage of stress phase in the TSV group. This concentration was 5.5 times that of the control group during the same period. Subsequently, it gradually decreased and dropped to a quarter of the initial concentration on the 25th day, remaining at this low level. In the TS group, the extracellular MC concentrations gradually increased and reached 5.8 times the initial concentration by the end of the experiment (Figure 2B).

The growth status of V. natans in the TSV group, as indicated by the fresh weight of plant tissues and length of plants, was significantly lower in the treatment with Microcystis (p < 0.05). However, during the recovery period following the decrease of Microcystis, the biomass and length of V. natans was significantly higher compared to the V group, with plants only (p < 0.05, Figures 3A, B). During the stress period, V. natans in the TSV group exhibited significantly stronger oxidative stress than those in the V group, as evidenced by a marked increase in peroxidase (POD) activity and malondialdehyde (MDA) concentration (p < 0.05). In the recovery period, both POD activity and MDA levels remained elevated relative to the control, although the differences were no longer statistically significant (Figures 3C, D).

An enrichment analysis of the KEGG pathways, based on the non-target metabolomics of V. natans, demonstrated that 16 metabolic pathways of the plants that were altered during the stress period in the TSV group compared to those in the V group. This included glutathione metabolism, ABC transporters, the biosynthesis of cofactors, Poantothenate and CoA biosynthesis, as well as glycine, serine, and threonine metabolism (Figure 3E). During the subsequent recovery period, the number of metabolic pathways decreased, to four, including nucleotide metabolism, ABC transporters, pyrimidine metabolism, and fatty acid biosynthesis, all of which exhibited significant differences between the TSV and V groups (Figure 3F).

The alpha-diversity of epiphytic bacteria and eukaryotic algae, as measured by the Shannon and Pielou_e indices, increased during the stress period following the introduction of Microcystis cells, and subsequently decreased during the recovery period in the TSV group. However, the Shannon and Pielou_e indices of other eukaryotes mainly including fungi and protozoa in the TSV group decreased significantly during the stress period (p < 0.05), subsequently increasing to levels exceeding those in the V group (Figures 4A, B). The variation in the Faith’s Phylogenetic diversity index exhibited similar trends as the Shannon and Pielou’s evenness indices, but the differences between the TSV and V groups were not pronounced (Figure 4C).

The PCoA demonstrated obvious dissimilarity of the community composition of all epiphytic microbiomes between the TSV and V groups, as well as between the stress period and recovery period in the TSV group. Among the three groups, the temporal variation of epiphytic bacteria community was much bigger than that of eukaryotes, including algae, in the V group (Figures 4D–F).

The phyla Proteobacteria, Actinobacteriota and Bacteroidota dominated the epiphytic bacteria of V. natans (Supplementary Figure S1A). At the genus level, Pseudomonas (0.08%~76.06%), Phreatobacter (0.06%~19.96%), Reyranella (0.04%~9.54%), and Streptomyces (0.52%~9.87%) were dominant. The relative abundance of Reyranella and Streptomyces in the TSV group increased after the introduction of Microcystis cells, and was much higher than that in the V group in the same period (Supplementary Figure S1D). The attached eukaryotic algae were dominated by diatoms, with a relative abundance ranging from 99.1% to 99.6% (Supplementary Figure S1B). The relative abundance of the genera, Cocconeis and Sellaphora exceeded 80% in both the TSV and V groups, showing no significant differences between the two groups (Supplementary Figure S1E).

Other eukaryotes were predominately composed fungi, i.e., Ascomycota and Aphelidea, and protozoa, i.e, Ciliophora (Supplementary Figure S1C). The relative abundance of Ascomycota, dominated by the genus Thysanophora, decreased substantially in the TSV group compared to that in the V group during the stress and recovery period. However, the relative abundance of the genus Paraphelidium belonging to Aphelidea, increased to 54.88% in the TSV group at the end of the stress. The genus Stentor was the most dominant within Ciliophora in the V group, but its relative abundance in the TSV group decreased markedly (Supplementary Figure S1F).

The modified stochasticity ratio (MST) analysis was applied to reveal the community assembly process of the epiphytic microbiomes (Supplementary Figure S2). With the exception of the other epiphytic eukaryotes in the V group, the MST values for epiphytic microbiomes in the V and TSV groups were lower than 0.5, indicating the dominance of the deterministic process (Supplementary Figures S2A–C). The deterministic process contributed 94%, 89% and 86% to the community assembly process of epiphytic bacteria, eukaryotic algae and other eukaryotes in the TSV group (Supplementary Figures S2D–F). For other eukaryotes, the stochastic process was dominant with the contribution ratio of 78% in the V group (Supplementary Figures S2C, F).

To explore the interactions among epiphytic bacteria, eukaryotic algae and other eukaryotes within the V and TSV groups across three stages, co-occurrence networks were constructed using the dominant genera with the relative abundance >0.1% from each microbial group. The modularity values were higher than 0.4 in both groups, indicating the networks of epiphytic microbes have obvious modular structures with intensely connected network nodes. The network properties were altered, with an increase in total nodes and edges, as well as network density and modularity in the TSV group. This suggests that the complexity of the epiphytic microbial community was enhanced and stronger interactions among epiphytic microbes occurred when inhibiting Microcystis cells (Supplementary Table S1, Figure 5B). The hub microorganisms (nodes highly connected to other members in a module) in the TSV group belonged to the bacteria, but were not same as the dominant ones in the V group. The contribution of epiphytic bacteria in the TSV group accounted for up to 65.31%, highlighting the significant roles of epiphytic bacteria in the Microcystis-inhibition process by host plants.

Network analysis identified significantly more hub taxa (weighted degree> 90% of maximum) in the TSV group than in the V group. All hub taxa in the TSV group were bacteria, predominantly affiliated with the phyla Proteobacteria, Actinobacteriota, Cyanobacteria, Bdellovibrionota. Key genera included Candidatus_Megaira, Cupriavidus, Pseudomonas, Amycolatopsis, Sphaerotilus, Phreatobacter, Ancylobacter, and 7 unclassified genera. In contrast, hub taxa in the V group comprised both bacteria and eukaryotes. The bacterial hubs (Phreatobacter, SWB02, and Sphingorhabdus) belonged exclusively to Proteobacteria. Eukaryotic hubs consisted of the Choanoflagellate Monosiga and one unclassified fungus (Figure 5A).

We further screened 22 genera of bacteria known to be algicidal from the epiphytic microbiomes of V. natans (Supplementary Table S2). The relative abundance of the algicidal bacteria reached up to 13.4% in the TSV group, significantly higher than that in the V group (5.8%, p < 0.01) at the end of the stress period. However, this disparity between the TSV group and V group diminished during the recovery period (Figure 6A). The enriched algicidal bacteria in the TSV group predominantly belonged to the phyla Actinobacteriota, Proteobacteria, Firmicutes, Bacteroidota, and Bdellovibrionota. The first dominant algacidal bacteria was Streptomyces belonging to Actinobacteriota, accounting for 73%. The genera Pseudomonas and Chryseobacterium were second dominant, accounting for 12% and 10% respectively (Figure 6B).

The random forest model screened out top 10 metabolites that were significantly correlated with the algicidal bacterial community (p < 0.05), from 117 major differential metabolites in the treatment and control groups during the stress period. They are catechin 7-O-apiofuranoside, 2-(4-Methyl-3-cyclohexen-1-yl)-2-propanyl 6-O-(6-deoxy-L-mannopyranosyl) -D-glucopyranoside, jasmonic acid, clothianidin, fungichromin, n-(1,1-dioxotetrahydro-1H-1lambda-6-–thiophen-3-yl)-5-(2-pyridinyl)-2-thiophenecarboxamide, {(2R,4S,5R)-5-[1-Methyl-3-(2-naphthyl)-1H-pyrazol-5-yl]-1-azabicyclo[2.2.2]oct-2-yl}methyl [3-(trifluoromethyl)phenyl] carbamate, n-Acetyl-L-carnosine, propiomazine (Supplementary Figure S3). We also found that the relative abundance of metabolites (2R,3R)-3-Methylglutamyl-5-semialdehyde-N6-lysine, 1H-Indole-3-carboxaldehyde, 3-[3-(beta-D-Glucopyranosyloxy)- 2-hydroxyphenyl] propanoic acid and 2,5-di-tert-Butylhydroquinone in the TSV group was higher than that in the V group during the stress period, but not during the recovery period (p < 0.05, Supplementary Figure S4).

While previous research has established that Microcystis can influence epiphytic bacterial communities on submerged macrophytes (Jiang et al., 2019; Li et al., 2020; Gao et al., 2022), temporal dynamics and the response of the entire microbial communities-including both prokaryotic and eukaryotic components- remain less understood. This study provides, to our knowledge, the first time-resolved evidence that the interaction with microcystin-producing Microcystis altered the structure and diversity of the entire epiphytic microbial community on the phyllosphere of the submerged macrophyte V. natans, with distinct responses observed between prokaryotic and eukaryotic microbial components. It has previously been reported that V. natans can reconstruct their epiphytic microbiome as a “cry for help” mechanism in response to acute ammonium stress (Hu et al., 2023). These findings suggest that aquatic plants may also utilize the “cry for help” strategy to defend against biotic and abiotic environmental stressors like terrestrial plants (Hu et al., 2018; Jiang et al., 2025; Zeng et al., 2025).

Taxon-specific responses of epiphytic microbiomes have been observed to differ between prokaryotic and eukaryotic communities across various life stages of aquatic plants and in response to diverse environmental changes, including warming, nutrient enrichment, pollutants, and hydrodynamic disturbances (Chen et al., 2022; Peng et al., 2024; Xia et al., 2020; Yang et al., 2025; Zhan et al., 2021). However, this is the first study to reveal that the diversity and community structure of epiphytic bacteria, algae, and other eukaryotes (mainly fungi and protozoa) on submerged macrophytes exhibit contrasting response dynamics to the emergence of microcystin-producing Microcystis. It is not only closely linked to their sensitivity to stress from Microcystis and microcystins, but is also shaped by the roles they played in facilitating plant stress resilience (Zeng et al., 2025).

The increased diversity of epiphytic bacteria and algae, coupled with reduced diversity of other epiphytic eukaryotes (Figures 4A, B, C), coincided with the complete disappearance of Microcystis cells, maximum extracellular microcystins and adverse effects on the growth and metabolism of V. natans at the end of the stress period in the TSV group (Figures 2, 3). The decomposition process of Microcystis cells releases a substantial amount of inorganic and organic matter, including secondary metabolites such as microcystins (Peng et al., 2024). This provides a diverse array of nutritional resources for epiphytic bacteria and algae (Cui et al., 2025; Liu et al., 2024). Concurrently, the diverse planktonic bacteria surrounding the Microcystis cells may have chances to colonize the phyllosphere of submerged macrophytes (Gong and Xin, 2021; Gao et al., 2022; He et al., 2023; Tsai et al., 2025). These factors could collectively account for the significant increase in bacterial and algal community diversity on leaves by the end of the stress period. In contrast, other epiphytic eukaryotes-including fungi and protozoa-likely experience a marked reduction in diversity. This decline may be attributed to the toxic effects of microcystins released in large quantities into the aquatic environment (Mehinto et al., 2021), combined with alterations in the interaction network including cross-feeding, parasitism, symbiosis, and predatism among phyllosphere eukaryotes and bacteria (Manirakiza et al., 2022; Shi et al., 2025). During the subsequent recovery phase of the TSV group, the diversity of the epiphytic bacterial and eukaryotic community diversity recovered to a level close to that of the control group. This shift occurred alongside a rapid decrease in extracellular microcystin concentrations and the restoration of plant growth and metabolic activity. These findings indicate that as the Microcystis -induced stress subsided, the plant-phyllosphere microbiome symbiont exhibited considerable resilience, which likely supports more stable plant colonization in degraded eutrophic waterbodies (Hu et al., 2023).

At both the phylum and genus levels, the structural shifts in the entire epiphytic microbial community were significantly more pronounced in the TSV treatment group than in the control (Supplementary Figure S1). The ambient environmental changes induced by Microcystis stress likely constitute a key deterministic process driving the reassembly of the epiphytic microbiota, a conclusion further supported by community assembly mechanism analysis (Supplementary Figure S2). All hub taxa of co-occurrence network of epiphytic microbiomes in the TSV group is bacteria, which confirmed the important roles of epiphytic bacteria during the inhibition of host plant on Microcystis. Proteobacteria was the most dominant phylum of the epiphytic bacteria in the study, which is consistent with other survey results (Schlechter et al., 2019; Zhen et al., 2020b; Yu et al., 2022; Wang et al., 2024). However, the relative abundance of Actinobacteria in the TSV group increased significantly during both the stress and recovery stages. There are reports of some algicidal bacteria belonging to Actinobacteria, such as the genera Rhodococcus and Streptomyces (Morón-López et al., 2024). Interestingly, the relative abundance of Streptomyces in the TSV group significantly increased at the end of stress period, whereas it decreased markedly during the recovery period. Its primary mechanisms involved in algicidal impact have been shown to be mycelia-mediated cell-to-cell contact, which facilitates the formation of flocs as well as the secretion of active algicidal substances that lead to Microcystis cell death (Yu et al., 2019; Kong et al., 2020; Zeng et al., 2021). A higher abundance of Reyranella belonging to the phylum Proteobacteria was also observed during the stress and recovery period in the TSV group compared to the V group. Reyranella exhibits denitrification capabilities and an enhanced ability to degrade chemical oxygen demand (Duan et al., 2023; Zhang et al., 2025), which is likely involved in the degradation of organic carbon and nitrogen released from lysed Microcystis cells.

While submerged plants have been often demonstrated to be negatively impacted by epiphytic algae due to competition for limited light and nutrients (Song et al., 2017; Wijewardene et al., 2022), our experimental results revealed a distinct pattern. The structure of the epiphytic eukaryotic algal community (primarily diatoms) remained stable throughout the experiment, and its contribution to the microbial co-occurrence network was consistently low (< 10% in both groups, Supplementary Table S1). This suggests that eukaryotic algae were less responsive to Microcystis stress than the other epiphytic microbial community under the present conditions. Borrego-Ramos et al. (2019) observed higher diatom richness on dead macrophyte stems compared to living tissues. Eukaryotic algae in our study exhibited the lowest diversity among the three major epiphytic microbial groups. This low diversity and stable structure may correlate with reduced competitive pressure on the host plant for phyllosphere resources, thereby minimizing adverse impacts on V. natans (Jones et al., 2002).

The limited response of eukaryotic algae could be linked to top-down regulatory pressure from eukaryotic parasitoids. Notably, we observed a marked increase in the relative abundance of the genus Paraphelidium in the TSV group at the end of the stress (Supplementary Figure S1). Paraphelidium belongs to Aphelidea, a group of intracellular parasitoids known to infect eukaryotic algae such as diatoms and increase host mortality by up to 70%, thereby preventing excessive algal biofilm formation on submerged surfaces (Karpov et al., 2014; Traver-Azuara et al., 2025). We therefore hypothesize that Aphelidea may act as a key regulator of biofilm stability by suppressing the overgrowth of epiphytic algae, which in turn helps maintain colonization space for epiphytic bacteria. Nevertheless, this regulatory role requires further experimental validation.

Within the epiphytic bacterial community identified in this study, 22 bacterial genera previously reported to possess algicidal potential was detected (Deng et al., 2010; Yang et al., 2020; Coyne et al., 2022; Morón-López et al., 2024). Notably, the total relative abundance and diversity of these algicidal bacteria significantly increased during the stress period. During the recovery phase, their relative abundance decreased back to levels comparable with the control. This significant dynamic response, combined with our preliminary experiments and the detection of algicidal bacteria (Wei et al., 2024), leads us to propose that submerged macrophytes may recruit algicidal bacteria to enhance the suppression capability on Microcystis. Building upon the traditional perspective that submerged macrophytes directly suppress Microcystis via the release of allelochemicals (Gao et al., 2017; Gross et al., 2007; Hilt and Gross, 2008; Nakai et al., 2012), this study proposes an indirect inhibitory mechanism mediated by the recruitment of algicidal bacteria. Consistent with the key roles of epiphytic microbiota in regulating biological processes of terrestrial plants, such as nutrient acquisition and resistance to abiotic and biotic stresses (Gong and Xin, 2021), the epiphytic microbiota of submerged macrophytes also plays a non-negligible role in mediating host plants’ inhibitory effects against Microcystis.

It is well-established that terrestrial plants employ a “cry for help” strategy by exuding various primary and secondary metabolites. These compounds recruit beneficial environmental microbes, which in turn enhance the host plant’s stress tolerance and growth (Rizaludin et al., 2021; Rolli et al., 2021). Non-targeted metabolomics analysis exhibited at least 117 different metabolites and 16 differential metabolic pathways of V. natans in the TSV group compared to those in the V group at the end of the stress period. Among the 16 significantly perturbed metabolic pathways, glutathione metabolism and ABC transporter pathways were extensively documented in relation to plant detoxification and stress resistance (Dorion et al., 2021; Zhou et al., 2024). We speculate that metabolic diversification within the plant kingdom may provide a basis for communication and recognition that enables the sculpting of microbiota tailored to the needs of the host (Huang et al., 2019).

Jasmonic acid, identified as one of the key metabolites strongly associated with the epiphytic algicidal bacteria (Supplementary Figure S3), is a pivotal phytohormone. Beyond its well-established roles in regulating plant growth and development, jasmonic acid is a central mediator of plant defense signaling and plant-microbe interactions (Fresno and Munné-Bosch, 2021). A recent study has documented a significant positive correlation between jasmonic acid and the recruitment of beneficial drought-enriched bacteria that enhance plant fitness (Xiang et al., 2025). More specifically, jasmonic acid has been shown to directly regulate the growth, development, and specialized metabolism of Streptomyces (van der Meij et al., 2023), a genus that was notably enriched on the phyllosphere of V. natans during the stress phase in our experiment. This suggests that jasmonic acid likely functions as a crucial signaling molecule, coordinating the active recruitment and physiological activation of specific beneficial epiphytic bacteria under stress. However, the precise mechanisms by which plant-derived metabolites shape the structure and function of epiphytic microbial communities remain an important area for future investigation (Xu et al., 2022).

The screened algicidal bacteria mainly belonged to phyla phyla Proteobacteria, Firmicutes, Bacteroidota and Actinobacteriota. It has been reviewed that a large number of algicidal bacteria belong mainly to class Gammaproteobacteria (47%) and class Bacilli (31%) (phylum Firmicutes) (Morón-López et al., 2024). The dominant genera Streptomyces, Pseudomonas and Chryseobacterium have been reported to exert algicidal effects through direct and indirect modes (Guo et al., 2015; Yu et al., 2019; Zhang et al., 2019; Kong et al., 2020; Zeng et al., 2021). Some screened metabolites with the highest abundance at the end of the stress period in the TSV group, such as 1H-Indole-3-carboxaldehyde, have been identified from algicidal bacteria as algicidal substances (Kong et al., 2022). The algicidal modes and mechanisms of epiphytic bacteria from submerged macrophytes still need further studies on the basis of the isolation of algicidal bacteria.

We also found several algicidal bacteria had capability to degrade MCs, including Pseudomonas, Sphingomonas, and Stenotrophomonas belonging to Phylum Proteobacteria, Bacillus belonging to Phylum Firmicutes, Chryseobacterium belonging to phylum Bacteroidota, Rhodococcus belonging to phylum Actinobacteriota, Streptomyces belonging to phylum Actinobacteriota (Dziga et al., 2013; Kormas and Lymperopoulou, 2013; Zhang et al., 2016; Kumar et al., 2018). This suggests that epiphytic bacteria may play a role in the degradation of microcystins, although further experimental evidence is required to confirm their involvement.

Based on our experimental findings, we propose that the submerged macrophyte V. natans restructures its epiphytic microbial communities and recruits algicidal bacteria to suppress Microcystis. Specifically, during the stress period—characterized by the gradual disappearance of Microcystis cells—the alpha-diversity of epiphytic bacteria increased significantly, while that of epiphytic eukaryotes (including fungi and protozoa) decreased markedly. In the subsequent recovery phase, the alpha-diversity of the epiphytic microbial community returned to a level comparable to that of the control plants. Throughout these phases, the structure and diversity of epiphytic microbiomes (with the exception of eukaryotic algae) shifted dynamically. Notably, epiphytic bacteria contributed most substantially to the interactions among the three major epiphytic microbial groups in response to Microcystis stress. The enrichment of algicidal bacteria on the phyllosphere during the stress phase likely played a key role in the successful suppression of Microcystis by V. natans (Figure 7).