Compartment-Specific Niche Filtering Shapes the Structure and Nutrient-Cycling Potential of Bacterial Communities in Eutrophic Waters with Hydrilla verticillata

Abstract

Submerged aquatic macrophytes and their microbiomes can help mitigate eutrophication, yet how microbial communities and functions differ across specific plant-associated and surrounding niches remains unclear. Here, we profiled bacterial community composition (16S rRNA gene sequencing) and quantified nitrogen and phosphorus cycling genes (narG, nirK, nirS, nosZ, phoD by qPCR) across eight distinct compartments associated with the submerged macrophyte Hydrilla verticillata in a eutrophic freshwater wetland. The niches spanned open water, bulk sediment, rhizosphere, and plant phyllosphere (leaf/stem surfaces) and endosphere (leaf/stem/root interiors). Alpha diversity differed significantly among niches: sediments (non-rhizosphere and rhizosphere) exhibited the highest Operational Taxonomic Unit (OTU) richness and diversity, whereas leaf-associated niches (phyllosphere and endosphere) had the lowest. Beta diversity showed clear separation by niche, indicating strong habitat filtering. Community composition also varied markedly: the water column was dominated by Bacteroidota (~51% of sequences), plant-associated communities were enriched in Pseudomonadota (43–90%), and sediment niches were dominated by Firmicutes (23~48%). Functional gene abundances showed pronounced niche partitioning. Nitrate/nitrite reduction genes (narG, nirK, nirS) were most enriched on leaf phyllosphere, with narG abundance equally high in the water, whereas the N₂O reductase gene nosZ peaked in sediment niches. The alkaline phosphatase gene phoD had its highest copy numbers in leaf biofilms, with significantly lower levels in internal plant tissues. Overall, neutral processes explained ~61% of community variation, but deterministic assembly was evident in the well-connected water and leaf surface niches. These findings reveal strong niche differentiation in plant-associated microbiomes and suggest that compartmentalized microbial functional capacity within the H. verticillata holobiont enhances nitrogen removal and phosphorus cycling in eutrophic waters.

1. Introduction

Eutrophication, driven by excessive nitrogen (N) and phosphorus (P) inputs, is a pervasive threat to freshwater ecosystems and water quality worldwide [1]. Nutrient over-enrichment fuels harmful algal blooms, oxygen depletion, and biodiversity loss in lakes and wetlands, undermining ecosystem services and potable water supplies [2,3]. Managing eutrophic conditions remains a critical challenge, requiring not only the reduction of external nutrient inputs but also a better understanding of internal nutrient cycling processes within these water bodies. In this context, aquatic microorganisms play a pivotal role by mediating key nutrient transformations such as nitrification, denitrification, N-fixation, and P mineralization [4,5]. Through these microbial processes, excess nutrients can be converted, removed, or sequestered, thereby regulating nutrient availability and influencing the severity of eutrophication. Indeed, harnessing the nutrient-cycling capacity of microbial communities has become a cornerstone of sustainable water quality management strategies, as exemplified by the use of biofiltration and constructed wetlands in eutrophic waters [6,7].

Submerged aquatic macrophytes are widely recognized as natural allies in the mitigation of eutrophication and maintenance of water clarity. These underwater plants directly assimilate dissolved N and P from the surrounding water and sediments, effectively stripping nutrients that would otherwise fuel algal growth [8,9]. In addition, their presence stabilizes sediments and reduces resuspension by dampening wave action and water turbulence, which helps increase light penetration and sustain clear-water conditions [10]. Crucially, submerged macrophytes host complex microbiomes on their leaves, stems, and roots that further contribute to nutrient cycling [11,12]. The biofilms and rhizosphere communities associated with these plants can perform complementary processes—for example, oxygen released by macrophyte shoots and roots supports nitrifying bacteria, while anoxic microzones in dense periphyton or root sediments foster denitrification [13]. Through such plant–microbe interactions, submerged macrophytes and their microbiota synergistically remove or immobilize nutrients, thereby curbing eutrophication. Accordingly, dense stands of submerged vegetation are often employed in lake restoration and constructed wetland systems to improve water quality. Hydrilla verticillata (L.f.) Royle is one such submerged macrophyte, known for forming extensive stands in nutrient-rich freshwater wetlands; it not only uptakes substantial amounts of N and P but also provides habitat for diverse microbial communities that can influence nutrient dynamics [14,15]. A body of recent research has consistently demonstrated the efficacy of H. verticillata in removing nutrients and fostering complex microbial communities in eutrophic systems [16,17]. However, it is noteworthy that this species is invasive in many non-native regions, where it causes ecological disruption, incurs management costs, and may host neurotoxin-producing cyanobacteria [18,19]. Therefore, elucidating the mechanisms of its interactions with niche-specific microbiomes is crucial. This knowledge is essential to inform both its potential applications in water remediation and the necessary ecological risk assessments.

Despite the recognized importance of aquatic macrophytes and their associated microbiota in nutrient cycling, the mechanisms underlying microbial community assembly and functional differentiation across the various plant-linked and environmental niches remain poorly understood. Previous studies have typically examined microbial communities in isolated compartments—for instance, focusing on bacterioplankton in the water column, bulk sediment communities, or the rhizosphere of aquatic plants—but few have compared the full spectrum of niches associated with a single macrophyte simultaneously [20,21]. As a result, it is unclear how microbial composition is partitioned among distinct spatial compartments such as open water, surface sediment, leaf surfaces (phyllosphere epiphytes), and internal plant tissues (endospheres), nor how each of these communities differs in its capacity for N and P cycling. Understanding these patterns is important for elucidating the processes that drive community assembly (e.g., environmental filtering by oxygen gradients or plant exudates in each niche) and for determining how nutrient-cycling functions are distributed within a macrophyte-associated ecosystem [12,22]. In particular, knowledge is lacking on whether certain niches (for example, the leaf biofilm versus the root interior) harbor specialized microbes or genes that confer heightened roles in nitrogen removal or phosphorus transformation. This gap in understanding limits our ability to predict or manage the contributions of plant-associated microbiomes to water quality improvement in eutrophic systems.

To address these knowledge gaps, the present study investigates microbial community differentiation and nutrient-cycling gene distribution across eight ecological niches associated with H. verticillata in a eutrophic freshwater wetland. These niches span both plant-associated habitats—including the leaf surface biofilm, leaf endosphere, root surface (rhizoplane), and root endosphere—and the adjacent environmental compartments of the water column and sediment. The study had two main objectives: first, to elucidate how bacterial community structure and diversity vary across these distinct compartments; and second, to characterize the spatial distribution of key functional genes involved in nitrogen and phosphorus cycling within each niche. By integrating community profiling with functional gene analysis across multiple spatial niches, this work aims to reveal how the compartmentalization of an aquatic plant system shapes microbial assemblages and their nutrient-cycling potential. Ultimately, these insights will advance our understanding of the role of macrophyte-associated microbiomes in mediating nutrient dynamics and inform more effective strategies for mitigating eutrophication in freshwater ecosystems.

2. Results

2.1. Variations in Bacterial Community Diversity Among Different Niches

Sequencing depth adequacy was confirmed by rarefaction curves and Good’s coverage indices, all exceeding 97% across niches (Figure 1A). Alpha diversity analyses revealed significant variations (p < 0.001) among niches. Specifically, OTU richness was highest in sediment niches (S and RR), while the phyllosphere (LP) and water column (W) exhibited the lowest richness. OTU evenness and diversity indices also demonstrated significant differences across niches, with rhizosphere and non-rhizosphere sediments generally presenting higher diversity than plant endospheric niches (Figure 1C). Hierarchical clustering analysis based on abund_jaccard similarity clearly differentiated microbial communities by niche type, illustrating distinct separation particularly between water, sediment, and plant-associated niches (Figure 1B). The principal coordinate analysis (PCoA) based on Bray–Curtis distance further supported these distinctions, highlighting significant differences in community structure (p < 0.001), with clear niche-specific clustering (Figure 1D).

2.2. Variations in Bacterial Community Composition Among Different Niches

The bacterial community composition differed markedly among the eight sampled niches as shown by the phylum-level heatmap (Figure 2A). Pseudomonadota did not show strong niche preference and were present at comparable relative abundances across compartments, whereas other major phyla exhibited pronounced niche-specific patterns. For example, Bacteroidota exhibited a very high relative abundance in the water samples but was far less abundant in all plant-associated niches. In contrast, Gemmatimonadota and Fusobacteriota were conspicuously depleted in the water sample, while Gemmatimonadota and Dependentaiae reached their highest proportions in the leaf surface community. The phyla TA06 and Latescibacterota were most enriched in the rhizosphere. Based on these composition patterns, the niches can be categorized into three broad community types defined by their dominant phyla. The water niche was overwhelmingly dominated by Bacteroidota (~51% of sequences in W). The plant-associated niches—including all endospheres (RE, SE, LE) as well as the leaf surface (LP)—were enriched in Pseudomonadota (approximately 43–90% in those communities). In external sediment and surface environments (S, RR, SU), Firmicutes were predominant (around 20% of the community in S and RR, with a similarly high proportion in SU). A circos diagram (Figure 2C) reinforced these niche-specific dominance patterns: Bacteroidota were almost exclusively associated with the water sample, Pseudomonadota dominated the plant interior and phyllosphere samples, and Firmicutes were most prominent in the soil and other external (surface) samples.

LEfSe analysis of the communities (Figure 2B) identified specific bacterial taxa as high-confidence biomarkers for each niche (LDA score > 4, p < 0.05). In the bulk sediment (S), members of Chloroflexi (notably class Anaerolineae) and Firmicutes (orders Clostridiales and Bacillales) were significantly enriched. The rhizosphere sediment (RR), in contrast, was characterized by an abundance of Firmicutes (class Clostridia, e.g., orders Peptostreptococcales-Tissierellales), along with Pseudomonadota (family Steroidobacteraceae) and diverse Verrucomicrobiota. The root endosphere (RE) exhibited enrichment of sulfate-reducing Deltaproteobacteria (orders Desulfobacterales and Geobacterales) alongside fermentative Bacteroidales. Above-ground plant niches showed clear differentiation between internal tissues and external surfaces. In the stem, the endosphere (SE) was dominated by Pseudomonadota (especially order Burkholderiales and class GammaPseudomonadota), whereas the stem surface (SU) had a high abundance of Firmicutes (class Bacilli, including the genus Exiguobacterium). A similar pattern was observed in the leaf niches: the leaf endosphere (LE) was enriched in AlphaPseudomonadota (order Acetobacterales) and Bacteroidota (order Sphingobacteriales), whereas the leaf surface (LP) was enriched in Pseudomonadota such as Sphingomonadaceae (family) and the orders Rhizobiales and Rhodobacterales. Finally, the aquatic water sample (W) was dominated by Flavobacteriales (phylum Bacteroidota) and unclassified Actinobacteria.

Briefly, the community composition shifted decisively across niches. The water column was dominated by Bacteroidota (~56% relative abundance), while Pseudomonadota were the major group in plant-associated niches (43–90%). Sediment and surface environments were enriched with Firmicutes. This tripartite division, supported by phylum-level abundances and genus-level biomarkers (e.g., Sphingomonadaceae on leaves, Clostridiales in sediment), underscores a systematic turnover in taxa from planktonic to plant-associated with sediment-dwelling communities.

2.3. Niche-Driven Variations in Community Assembly Processes

Neutral community model (NCM) analysis showed that stochastic processes accounted for 61.08% of the total community variation in the overall water–plant–sediment system (Figure 3A,E). Partitioning this system into aboveground (including the water column) and belowground (including non-rhizosphere sediment) compartments revealed slightly higher stochastic contributions in both subsets (71.36% and 69.41%, respectively) compared to the whole-system average. This difference was minimal (Δ = 1.95%) and not statistically significant, indicating that aboveground and belowground compartments were similarly governed by stochastic assembly processes. However, focusing exclusively on plant-associated niches (i.e., excluding water and sediment) revealed a clear divergence in assembly patterns. Endospheric communities exhibited substantially higher stochasticity (77.48% of variation) compared to epiphytic communities (59.96%), a difference of 17.52%. This discrepancy corresponded with differences in the model-estimated migration rates: endospheric habitats had a higher community migration rate (m = 0.3971) than epiphytic habitats (m = 0.2798), and similarly the belowground niches showed a higher m (0.3896) than the aboveground niches (m = 0.2539).

Consistent with the NCM results, βNTI analysis indicated that deterministic processes (|βNTI| ≥ 2) operated only in W and LP communities, whereas all other niches were governed predominantly by stochastic assembly (|βNTI| < 2) (Figure 3F). Bray–Curtis-based RC decomposition further resolved the deterministic fraction and showed that homogeneous selection was the main process in W, LP, and the LE, with its relative contribution decreasing in the order W > LP > LE, indicating environmentally driven community convergence in these niches. In contrast, along the gradient from stem-associated with root-associated with sediment niches, the influence of homogeneous selection declined progressively, while the contributions of stochastic processes (drift and homogenizing dispersal) and heterogeneous selection increased (Figure 3G). Heterogeneous selection, reflecting selection imposed by divergent environmental conditions, was detected only in RE and S, with a stronger effect in S than in RE.

2.4. Niche-Specific Functional Gene Variation and Plant-Mediated Microbial Recruitment Mechanisms

Across the submerged macrophyte niches, the distribution of these nitrogen (N) and phosphorus (P) cycling genes varied markedly by microhabitat (Figure 4). For the nitrogen cycling genes, distinct niche-specific patterns were evident. The nitrate reductase gene narG showed the highest abundance in W and LP niches, with no significant difference between those two environments. Its abundance then progressively declined in the S, LE, SU, RR, RE, reaching a minimum in SE. The nitrite reductase genes exhibited contrasting preferences: nirK peaked in LP (followed by RR, RE, SU, S, and W in descending order), whereas nirS was most abundant in W and LP, with a secondary peak observed in RR. Similarly, the nitrous oxide reductase gene nosZ was highest in RR and S, showed intermediate levels in W, LP, SU, and RE, and was lowest in LE and SE. In the case of phosphorus cycling, the alkaline phosphatase gene phoD also displayed a strongly niche-dependent distribution. phoD was most abundant in LP, followed sequentially by high levels in RR, S, SU, RE, and W. The lowest phoD copy numbers were detected in the LE and SE niches. It is noteworthy that across all measured plant-associated niches (LP, LE, SU, SE, RR, RE), the highest detected abundances of the target functional genes (narG, nirK, nirS, nosZ, phoD) were consistently recorded in the LP.

3. Discussion

3.1. Niche Differentiation as the Driver of Bacterial Community Structuring

Our results demonstrate that each compartment in the plant-associated aquatic system harbors a distinct bacterial community, consistent with strong niche differentiation driving community assembly. Alpha diversity differed markedly by niche: sediments (both bulk and rhizosphere) sustained the highest OTU richness and diversity, whereas the plant interior compartments—especially the leaf endosphere—showed the lowest diversity. These patterns align with broader observations that soil or sediment environments generally support more diverse microbiota than the highly filtered endosphere and phyllosphere of plants [23,24]. Indeed, terrestrial studies have similarly found that rhizosphere communities are richer and more complex than those inside leaves or roots, reflecting the strong selective pressures imposed by the plant host on internal colonizers [25,26]. In our study, the water column also showed intermediate diversity, higher than the leaf-associated niches but lower than sediments. This gradient in alpha diversity likely reflects the degree of environmental heterogeneity and resource availability: the sediments provide a complex, nutrient-rich and spatially heterogeneous habitat supporting many microbial niches, whereas the interior of H. verticillata leaves is a constrained environment accessible only to specialized taxa that tolerate plant defenses and low oxygen, resulting in lower richness.

Beta diversity analyses confirmed that community composition is structured primarily by habitat type (water vs. sediment vs. various plant-associated niches). Both hierarchical clustering and PCoA ordinations revealed complete separation of samples by niche, indicating that microbial communities cluster according to their compartment and not by random chance. This compartmentalization underscores deterministic “species sorting” by local environment: even within the same ecosystem and over small spatial scales, the unique physicochemical conditions of each niche (e.g., oxygen levels, carbon sources, plant-derived substrates, physical exposure) filter the regional species pool and select for distinct communities [27,28]. The clear niche-specific clustering we observed mirrors these findings and indicates that, in our eutrophic H. verticillata system, environmental selection pressures in each compartment are strong enough to outweigh any homogenizing influence of dispersal among compartments.

Taxonomic compositional shifts across compartments further illustrate niche filtering at work. We found Pseudomonadota to be highly abundant in all plant-associated niches (endospheres and epiphytes), whereas Bacteroidota dominated the open-water bacterioplankton (~56% of sequences in the water column), and Firmicutes were predominant in external sediments and on plant surfaces. The water column, being eutrophic and rich in phytoplankton detritus, favored Bacteroidota (notably Flavobacteriales). Such Flavobacteria are well-known for thriving in nutrient-rich freshwaters and often form blooms constituting a large fraction of lake bacterioplankton [29]. In contrast, the plant interior (roots, stems, leaves) and phyllosphere were disproportionately enriched in Pseudomonadota. This aligns with many studies reporting Pseudomonadota as the dominant endophytic and phyllospheric bacteria across diverse plant species [30]. Rapidly growing copiotrophic Pseudomonadota (e.g., Burkholderiales, Rhizobiales) may be adept at using plant exudates and resisting plant defenses [31,32], allowing them to colonize internal tissues in high numbers. The sediment compartments, on the other hand, showed a strong prevalence of Firmicutes (especially anaerobic spore-formers like Clostridiales), reflecting anoxic, organic-rich conditions in lake mud that select for fermentative and anaerobic bacteria. Such Firmicutes and Chloroflexi (Anaerolineae) thriving in sediments are typical of oxygen-depleted environments where they decompose organic matter and cycle nutrients under anaerobic conditions [33,34]. LEfSe biomarker analysis provides further evidence that each niche selects for particular taxa adapted to those conditions. These niche-specific taxa illustrate how H. verticillata creates a mosaic of microhabitats—from oxygenated, sunlit leaf surfaces to anoxic interior roots—each favoring microorganisms with appropriate metabolic and life-history traits.

Overall, the pronounced differences in community structure across water, sediment, and plant-associated niches indicate that niche filtering is a dominant force shaping bacterial community composition in this eutrophic lake ecosystem. The environmental parameters tied to each compartment (such as redox status, organic carbon availability, plant secondary metabolites, and physical disturbance) act as strong selectors.

3.2. Stochastic versus Deterministic Processes in Community Assembly

While niche differentiation clearly underlies much of the community structuring, our assembly process analyses indicate that both stochastic (neutral) and deterministic forces act in concert to shape these bacterial assemblages—with the balance between them varying by niche. Neutral community model (NCM) fitting suggested that a majority (~61%) of the variation in community composition across all samples could be attributed to stochastic processes consistent with neutral theory (e.g., random dispersal, ecological drift). This implies that, superimposed on the niche filtering discussed above, there is a substantial role of chance and dispersal in determining which taxa end up in a given sample. Microbial ecologists increasingly recognize that community assembly lies along a continuum from purely deterministic selection to purely stochastic drift, with most communities influenced by both to some degree [35]. Our findings align with this consensus view—strong habitat filtering is evident, yet neutral processes also make a significant contribution to community assembly in the water–plant–sediment system.

Importantly, the relative influence of deterministic vs. stochastic processes was not uniform across compartments. When we partitioned the data by different subsets, clear differences emerged. For instance, considering only plant-associated niches, we found that endosphere communities were assembled in a more stochastic manner (approximately 77% of variation explained by neutral processes) compared to epiphytic (surface-associated) communities (~60% stochastic). In other words, bacterial assemblages living inside H. verticillata tissues appear to be closer to random draws from the regional species pool (within the constraints of host filtering) than those living on the plant surface. One possibility is that colonization of internal tissues involves stochastic “bottlenecks” and priority effects—for example, which specific bacteria happen to colonize a young root or leaf first can be a matter of chance [36], and those early arrivals then shape the developing endophytic community [37]. By contrast, the epiphytic communities (e.g., leaf surface biofilms) are in direct contact with the surrounding water and are more exposed to constant immigration of microbial propagules, as well as more homogenizing environmental conditions (light, temperature, water chemistry) [38,39]. These factors can enforce a more consistent community composition across replicates (hence a relatively stronger deterministic signature). Indeed, our β-diversity null model analysis (βNTI) found that the only niches showing statistically significant deviation from random assembly (|βNTI| ≥ 2) were the water column (W) and the leaf phyllosphere (LP)—precisely those exposed, well-mixed habitats.

By contrast, as one moves from the well-connected aboveground compartments to the more isolated or heterogeneous belowground compartments, stochasticity and heterogeneous deterministic processes appear to play greater roles. Along the gradient from stem surface (SU) to root-associated niches (rhizosphere and bulk sediment), we observed a progressive decline in the influence of homogeneous selection and a rise in stochastic processes (drift, dispersal limitation) and heterogeneous selection. Heterogeneous selection—which is deterministic selection driven by spatially variable environmental conditions—was detected primarily in the root endosphere (RE) and in the bulk sediment (S). In these niches, β-diversity was higher than expected by neutral drift, consistent with variable environmental conditions between samples selecting for different taxa. This makes ecological sense: sediment microhabitats can differ markedly over small distances in organic content, redox conditions, or plant detritus, causing divergent community composition—i.e., each sediment core harbors a somewhat unique microbiome shaped by its micro-environment [40]. Similarly, each root endosphere sample (likely coming from different individual plants or roots) may experience idiosyncratic host factors (e.g., differences in root exudate profiles, immune responses, or colonization history), leading to greater compositional variability among root interiors [41,42]. Thus, rather than converging to one “optimal” community, the RE and S communities appear to be deterministically pulled in different directions by local conditions—hence variable selection—superimposed on a background of high stochasticity.

3.3. Plant-Mediated Functional Zonation and Its Implications for Eutrophication Control

We observed that across all plant-associated niches, leaf phyllosphere (LP) showed the consistently highest gene copy numbers for nearly every target gene. This study observed that the leaf phyllosphere exhibited the highest copy numbers for nearly all target genes. This functional enrichment aligns with recent findings that the phyllosphere of submerged macrophytes sustains heightened microbial activity and nutrient removal rates [43]. The phyllosphere is recognized as a site for substantial biofilm accretion, a pattern pronounced in upper leaves and influenced by light availability [44]. Such an environment likely benefits from the distinct chemical milieu shaped by plant exudates, which can select for and support dense, metabolically active, and cooperative microbial assemblages [45], as observed here. In contrast, the highest abundance of nosZ (nitrous oxide reductase) was found in the rhizosphere sediment (RR) and bulk sediment (S). Meanwhile, the interior niches (stem and leaf endospheres) showed uniformly low abundances of all these functional genes, suggesting that the bulk of microbial nutrient-cycling potential resides in communities at the plant surfaces or in immediately adjacent sediment/water, rather than within plant tissues. This compartmentalized functional capacity suggests that H. verticillata and its microbiome form an integrated nutrient-processing unit, wherein different parts of the holobiont specialize in different steps of nutrient cycling. The leaf surface biofilms, bathed in the water column, appear to function as a primary interface for capturing and processing nutrients from the surrounding water. High narG and nirK/nirS gene content in the LP communities indicates a strong potential for nitrate and nitrite reduction on the surfaces of H. verticillata leaves [46]. This implies that as water flows through H. verticillata beds, nitrate can be rapidly assimilated or denitrified by epiphytic bacteria on leaves. Indeed, recent work has shown that the phyllosphere of submerged macrophytes can harbor diverse nitrifying and denitrifying microbes and even exhibit higher potential denitrification rates than the surrounding water column [44,47]. Furthermore, the dominance of nirK over nirS in the leaf biofilms (nirK peaked in LP, whereas nirS was also high in LP but shared dominance with water column) hints at a selection for particular denitrifier communities on plant surfaces. Prior studies have noted that nirK- and nirS-type denitrifiers often occupy different niches and respond to environmental factors differently [48,49]. The enrichment of nirK-type denitrifiers on leaves might be related to plant-derived carbon compounds or microaerobic conditions within biofilms that favor nirK-bearing taxa. In contrast, the water column, with nirS genes abundant, might support a different guild of denitrifiers. Such niche partitioning between nirK and nirS groups has been observed, for example with nirK communities sometimes predominating in organic-rich, plant-influenced environments while nirS types flourish in open-water or sediment niches [50,51]. Thus, the presence of H. verticillata may shift the balance of denitrifier types in its vicinity, adding to overall denitrification capacity.

The significantly higher nosZ gene abundance in root-associated and sediment communities (compared to leaves or water) suggests these niches host microbes (often Pseudomonadota like Bradyrhizobium or Anaeromyxobacter, or Firmicutes like Clostridia) capable of scavenging N₂O and converting it to N₂ [52,53]. In practical terms, this means that N₂O potentially produced by denitrifiers on the oxygen-rich leaf surfaces (where nosZ was comparatively low) can be further reduced in the anoxic rhizosphere or sediment before escaping to the atmosphere. This division of labor—with upper parts of the plant-associated microbiome performing the initial steps (nitrate → nitrite → N₂O) and lower parts completing the final step (N₂O → N₂)—could maximize overall nitrogen removal while minimizing N₂O emissions [54,55]. The alkaline phosphatase gene phoD showed a strongly niche-dependent distribution, with the highest levels on leaf surfaces (LP), followed by the rhizosphere (RR) and bulk sediment, and minimal representation in internal tissues. phoD encodes an enzyme that liberates inorganic phosphate (PO₄³⁻) from organic phosphorus compounds [56,57]. High phoD abundance in epiphytic and sediment communities indicates active organic P mineralization at those locations. In eutrophic lakes, a substantial portion of total P can be in organic form or bound in particulate matter; microbial phosphatases are crucial for converting that into bioavailable phosphate [58]. Submerged macrophytes are known to uptake nutrients directly from the water through leaves as well as from sediment via roots [8]. Thus, any phosphate released by epiphytic phosphatases on the leaf surface is likely quickly assimilated by either the microbes or the plant, rather than accumulating in the water column. In this way, the H. verticillata a holobiont can act as a phosphate sink—intercepting organic P and incorporating it into biomass. Supporting this idea, researchers have observed that submerged macrophytes like H. verticillata can significantly reduce water phosphorus levels by sequestration in plant tissue and periphyton.

Our findings highlight that the microbial communities associated with H. verticillata exhibit functional stratification, a trait that synergistically enhances nutrient uptake by the plant, thereby boosting the overall efficiency of nutrient removal in aquatic environments. This plant-microbe synergy has direct implications for ecological engineering and lake management. Submerged macrophytes are already a cornerstone of many bioremediation and lake restoration efforts due to their ability to stabilize sediments, oxygenate water, and compete with algae for nutrients [59,60]. What our study adds is a finer understanding that the microbiome of these plants amplifies those benefits by carrying out critical biochemical transformations. The phyllospheric microbes effectively “pre-process” the water’s nutrients—for example, converting nitrate to N₂ (denitrification) and breaking down organic phosphorus—before those nutrients can fuel algal blooms. The rhizosphere microbes, meanwhile, finalize the removal of nitrogen (consuming residual N₂O) and can mobilize sediment nutrients in tandem with root uptake. In holistic terms, H. verticillata functions as an ecosystem engineer not only through its own physiology but also via the metabolic potential of its microbial partners. This concept of macrophyte holobionts mitigating eutrophication is supported by growing evidence that epiphytic microorganisms significantly contribute to nutrient cycling [11].

4. Materials and Methods

4.1. Site Description and Field Sampling

To elucidate the functional roles and niche differentiation of microbial communities associated with submerged plants in the remediation of eutrophic water, a comprehensive field investigation was conducted during the active growth phase of Hydrilla verticillata in August 2022. Sampling was carried out within a constructed wetland system designed for purifying agricultural tailwater, located in the Datong Lake area (29°04′–29°22′N, 112°17′–112°42′E). The wetland system completed its vegetation configuration and planting in February 2022 and was put into operation in May of the same year. During the construction process, establishment was achieved through artificial sowing of axillary turions [61] of H. verticillata in the ponds. Subsequently, the plants formed dense communities via asexual reproduction. To systematically investigate microbial communities across distinct ecological niches, samples were classified into eight categories based on their respective habitats within the water-plant-sediment continuum: (1) Water (W), (2) Non-rhizosphere sediment (S), (3) Leaf endosphere (LE), (4) Leaf phyllosphere (LP), (5) Stem endosphere (SE), (6) Stem surface (SU), (7) Root endosphere (RE), and (8) Rhizosphere sediment (RR) (Figure 5).

Field sampling was conducted using a standardized five-point method. Sampling points were established at the center of the constructed wetland and in the east, south, west, and north directions, with an interval of 5 m between adjacent points. approach to ensure representative sample collection. Submerged plants (H. verticillata) were carefully retrieved as entire individuals from a depth of approximately 0.5 m using an ethanol-sterilized hook. For each of the eight ecological niches, plant materials collected from the five points were pooled, rinsed gently with in situ water, and composited into a single representative homogenized sample per niche. This composite sample was then divided into two portions. One portion was allocated to assess plant growth indices, including biomass (determined using a high-precision electronic balance), plant height, and root length (measured using a calibrated ruler). The second portion, designated for microbial extraction, was preserved in sterile polyethylene bottles filled with 400 mL of sterilized phosphate-buffered saline solution (PBS, 50 mM, pH = 7.4).

Sediment samples designated as non-rhizosphere were collected using a Peterson grab sampler from locations at least 10 cm distant from plant roots to minimize rhizospheric influence. Larger particles, such as dry roots and rocks, were removed using a 2 mm sieve, after which approximately 0.5 g of the sieved sediment was stored in sterile 15 mL centrifuge tubes. Water samples were meticulously collected prior to plant and sediment sampling to avoid contamination and turbidity. A 500 mL composite water sample was taken at each sampling point and divided into two aliquots: one was immediately analyzed in situ for key physicochemical parameters using a portable multiparameter meter (HQ30D, HACH, Loveland, CO, USA), including temperature (T), dissolved oxygen (DO), total dissolved solids (TDS), salinity (SAI), oxidation-reduction potential (ORP), and pH. The measured values are detailed in

Table 1. Plant growth indicators and water quality parameters.

Water Quality
T (°C)			DO (mg/L)			C (-μS/cm)			TDS (mg/L)
31.63 ± 0.12			3.90 ± 0.13			419.57 ± 1.39			254.83 ± 10.43
SAI (ppl)			PH			ORP (mV)			Twb (JTU)
0.18 ± 0.00			8.16 ± 0.41			125.67 ± 6.24			13.3 ± 0.45
TN (mg/L)			NO3− (mg/L)			NH4+ (mg/L)			COD (mg/L)
4.67 ± 0.09			0.63 ± 0.05			10.3 ± 0.14			26.67 ± 3.68
TP (mg/L)			TDP (mg/L)			PP (mg/L)			PO4+ (mg/L)
0.65 ± 0.00			0.54 ± 0.01			0.11 ± 0.00			0.24 ± 0.00
Plant
Weight (g)				height (cm)				Root length (cm)
66.08				21.33 ± 7.66				16.32 ± 10.48

; the other aliquot was filtered using 0.22 μm membrane filters, and the filter membranes were subsequently stored in sterile 50 mL centrifuge tubes for microbial analyses.

Five individual plants were randomly selected and dissected into roots, stems, and leaves, each weighing approximately 10 g fresh weight. Each plant part was then transferred separately to sterile polyethylene bottles containing 400 mL PBS, subjected to sonication for 10 min, and subsequently filtered through 0.22 μm membrane filters. The resulting filter membranes were preserved in sterile 50 mL centrifuge tubes. The sonicated plant tissues (roots, stems, and leaves) were individually stored in sterile 50 mL centrifuge tubes filled with PBS.

A total of 32 samples for microbial analysis were obtained from the eight ecological niches. Each corresponding composite sample was subdivided into four analytical replicates to account for technical variability, thus resulting in 32 samples in total (8 niches × 4 replicates). Rigorous aseptic techniques were maintained throughout the process to prevent potential contamination. All collected samples were rapidly frozen at −80 °C to preserve microbial DNA integrity until further analysis.

Upon return to the laboratory, comprehensive water quality analyses were performed following standardized protocols outlined in China’s “Environmental Quality Standards for Surface Water” (GB3838-2002). Parameters measured included total nitrogen (TN), total phosphorus (TP), phosphate (PO₄⁺), and total dissolved phosphorus (TDP). Concentrations of ammonia nitrogen (NH₄⁺), nitrate nitrogen (NO₃⁻), and chemical oxygen demand (COD) were quantified using standardized HACH reagent assays (HACH, Loveland, CO, USA).

4.2. 16S rRNA Gene Amplicon Sequencing and Analysis

Total genomic DNA was extracted from 0.25 g of biofilm sample with the DNeasy PowerBiofilm Pro Kit (Qiagen, Hilden, Germany). Integrity of the extracted DNA was verified by 1% agarose gel electrophoresis prior to storage at −80 °C. The V3–V4 hypervariable regions of the bacterial 16S rRNA gene were amplified in 20 μL reactions containing the primer pair 341F/806R [62] and TransStart FastPfu DNA Polymerase (TransGen Biotech, Beijing, China). Thermal cycling conditions comprised an initial denaturation at 95 °C for 3 min, followed by 24 cycles of 95 °C for 30 s, 50 °C for 30 s, and 72 °C for 45 s, with a final extension at 72 °C for 10 min. Amplification success was confirmed on 2% agarose gels. The PCR products were then purified with the AxyPrep DNA Gel Recovery Kit (Axygen Biosciences, Union City, CA, USA), and their concentrations were determined using the QuantiFluor™-ST system (Promega, Madison, WI, USA). Purified amplicons from each sample were pooled in equimolar ratios for library preparation. Paired-end sequencing (2 × 300 bp) was performed on an Illumina MiSeq platform at MajorBio Bio-Pharm Technology Co., Ltd. (Shanghai, China). Processing of raw sequencing data included clustering reads into operational taxonomic units (OTUs) at a 97% similarity threshold. Taxonomic assignment was carried out by comparing representative sequences against the SILVA reference database (Release 138, accessed on 20 July 2023, https://www.arb-silva.de/documentation/release-138/).

4.3. Quantitative PCR (qPCR) of N-Cycle and P-Cycle Genes

The abundances of key microbial genes involved in nitrogen and phosphorus cycling were determined by quantitative PCR (qPCR). The assay targeted the following genes with their respective primer pairs: narG (primers narG-517F/narG-773R) [63], nirK (primers nirK-876F/nirK-1040R) [64], nirS (primers nirS-cd3aF/nirS-R3cd) [65], nosZ (primers nosZ-1126F/nosZ-1381R) [63], and phoD (primers ALPS-F730/ALPS-1101) [66]. Each 20 μL qPCR reaction contained 10 μL of 2× ChamQ SYBR Color qPCR Master Mix (Vazyme Biotech, Nanjing, China), 0.8 μL each of forward and reverse primers (5 μM), 1 μL of template DNA, and 7.4 μL of sterile ddH₂O. Reactions were run in triplicate on an ABI 7300 Real-Time PCR System (Applied Biosystems, MA, USA). The thermal cycling protocol comprised an initial denaturation at 95 °C for 5 min, followed by 35 cycles of denaturation at 95 °C for 30 s, annealing at 58 °C for 30 s, and extension at 72 °C for 1 min, ending with a melt curve analysis to verify amplification specificity. For absolute quantification, standard curves were generated from ten-fold serial dilutions of linearized plasmid DNA containing the target gene fragment. Amplification efficiency (E), correlation coefficient (R²), and full primer sequences for each assay are provided in Supplementary Tables S1–S3, with amplification plots shown in Supplementary Figures S1 and S2.

4.4. Statistical Analyses and Data Visualization

Significant differences (p < 0.001) in alpha diversity indices (ACE, Pielou, Shannon) and functional gene abundances among niches were determined using Kruskal–Wallis tests followed by Dunn’s post hoc analysis via the ‘agricolae’ package. Rarefaction curves and Good’s coverage were evaluated using the Mothur program (v.1.30.2) to assess sequencing depth sufficiency. Hierarchical clustering of β diversity, based on abund_jaccard distance, was performed with Qiime 2020.2.0 and visualized in R to illustrate microbial community distribution across niches. Principal coordinates analysis (PCoA) based on Bray–Curtis distance matrices, alongside ANOSIM analysis using the ‘vegan’ package, assessed significant structural differences among microbial communities (p < 0.001). The relative abundance of dominant genera was visualized using heatmaps generated by the ‘pheatmap’ package. Circos plots, constructed using Circos-0.67-7 software, depicted shifts in species abundances across niches. Microbial biomarkers were identified using linear discriminant analysis effect size (LEfSe, LDA > 4.0) with a stringent all-against-all comparative strategy. Neutral community model (NCM) analyses were conducted using the ‘Hmisc’, ‘minpack.lm’, and ‘getopt’ packages, and the ‘icamp’ package was employed to evaluate community assembly processes. All results were presented as means ± standard deviations of four replicates. Statistical analyses and data visualizations were conducted using R software (version 4.3.1).

5. Conclusions

H. verticillata creates a spatial mosaic of microbial niches—from leaf surface to root zone—each specializing in particular nutrient transformations. This compartmentalized functional capacity likely maximizes the removal or sequestration of excess nitrogen and phosphorus in eutrophic waters. Our findings underscore the importance of considering the whole plant-microbe system (the holobiont) in ecological functioning. The structure of the bacterial community is not an idle pattern; it has direct repercussions on ecosystem services like water purification. By shaping its microbiome through niche provisioning (oxygenation, exudation of carbon, etc.), H. verticillata effectively “recruits” microbes that help alleviate the very nutrient overload that threatens aquatic ecosystems. This mutualistic interplay between plant and microbes offers a promising avenue for nature-based solutions to eutrophication: harnessing the inherent nutrient-cycling potential of compartmentalized microbiomes in and around aquatic plants. Several considerations contextualize these findings. First, our comparative analysis was based on tissue mass rather than habitat area, meaning abundance patterns reflect differences per unit mass of sample. Second, the single sampling time point precludes insight into seasonal dynamics. Furthermore, the study was conducted in a turbid agricultural tailwater wetland where light penetration was attenuated; thus, the potential diurnal influence of light (PAR/UV) on niche partitioning may differ from that in clear-water systems. Finally, any long-term application must account for the management of accumulated plant-derived organic matter in sediments. Future research integrating spatial scaling, temporal monitoring, and activity-based measurements such as metatranscriptomics or metabolomics will be key to quantifying actual process rates and advancing sustainable application. These mechanistic insights into nutrient-cycling potential should also be considered within the broader ecological context of H. verticillata. Beyond its remediation capacity, this species is a recognized invasive plant in many non-native ecosystems. Therefore, any potential application in restoration must be grounded in careful site-specific risk assessment and adaptive management, ensuring that functional benefits are balanced with responsible environmental stewardship.