Revealing the Structure and Biodiversity Patterns of Fish Communities in River Networks Based on Environmental DNA

Abstract

Revealing taxonomic, functional, and phylogenetic dimensions of biodiversity is critical for the effective conservation and management of aquatic organisms in freshwater ecosystems subjected to multiple threats. Fish biodiversity patterns in river ecosystems exhibit complex spatiotemporal variation influenced by hydrological connectivity and the dispersal ability of species within the river network. This study utilized eDNA metabarcoding to investigate fish communities in three subtropical mountain rivers, aiming to uncover the community structure and spatiotemporal dynamics of the multidimensional biodiversity of fish communities across varying positions within the river network. The three biodiversity indices of the fish communities at the confluence were significantly greater than those of the mainstream community in the dry season, whereas the phylogenetic diversity of the fish community at the tributaries during the wet season was significantly greater than that in the mainstream. This likely reflected the unique spatial configuration of confluences and the seasonal dispersal of fish in the river network. The results of this study highlighted the important role of confluences and tributaries in river networks and the fact that they need to be prioritized to conserve riverine fish diversity.

1. Introduction

Freshwater ecosystems support a disproportionately high level of global biodiversity, with fish communities playing essential ecological and economic roles [1,2]. However, these ecosystems are facing severe anthropogenic pressures, leading to marked reductions in fish diversity and abundance [3,4,5]. The efficient monitoring of fish communities and a comprehensive understanding of biodiversity patterns are thus critical for informing conservation priorities and guiding adaptive river management [6].

Biodiversity comprises multiple dimensions [7,8,9], including taxonomic diversity (TD), which captures species richness and composition [10,11]; functional diversity (FD), which describes the variation in species’ ecological roles based on trait differences [12]; and phylogenetic diversity (PD), which reflects the evolutionary relationships among species [13]. Although TD is widely used in biodiversity assessments, neglecting FD and PD can obscure critical aspects of ecosystem functioning and evolutionary distinctiveness, thereby hindering the development of effective conservation and management strategies [14,15,16]. An integrated assessment of TD, FD, and PD offers a more holistic understanding of community assembly processes and ecosystem resilience [17,18].

In river ecosystems, fish biodiversity varies systematically with their position in the river network [6]. The river continuum concept suggests that fish diversity in headwater streams is generally lower than in downstream reaches due to environmental filtering effects, such as fast-flowing water, that limit the suitability of headwater habitats for many species [19,20]. Confluences are key junctions where tributaries and mainstreams converge, generating complex spatial structures and distinct environmental conditions that enhance local habitat heterogeneity [21,22]. Confluences usually present high species richness and unique community compositions because they are important pathways for the dispersal of organisms in adjacent mainstreams and tributaries [23,24]. For example, Czeglédi et al. [25] reported that compared to those at other tributary sites, fish communities at tributary confluences presented the highest species richness. Similarly, the phylogenetic diversity of fish communities at confluences is greater than that in tributaries and mainstem streams [6]. However, fish biodiversity patterns within river networks can exhibit marked seasonal variation [25,26,27]. However, the taxonomic, functional, and phylogenetic diversity patterns of fish communities across different positions within river networks (e.g., tributary, mainstream, and confluence), as well as their seasonal dynamics, remain insufficiently understood.

Recently, eDNA metabarcoding, which identifies species based on genetic fragments released by organisms into water bodies, has been widely used for monitoring fish communities and diversity because it is labor-saving and noninvasive to ecosystems [9,28]. Along with fish taxonomic information, eDNA also has high efficiency and accuracy in detecting the functional diversity and phylogenetic diversity of fish communities [29,30,31]. In this study, based on eDNA metabarcoding technology, a cross-seasonal fish community investigation was conducted at different river network positions (tributary, mainstream, and confluence) in three subtropical mountain rivers. For fish communities in three different habitats, we hypothesized that (1) the diversity of the fish communities in tributaries is low because the habitat structure of tributaries is simple and fish are mostly specialized species adapted to this environment, and (2) the fish community at the confluence has the highest biodiversity as species from tributaries and the mainstream may appear at the confluence. This study revealed the patterns of fish diversity in river networks and may contribute to the conservation and management of river fish diversity.

2. Materials and Methods

2.1. Study Area and Sampling Sites

The Wannan Mountain region (WMR) is located in the southern Anhui, China, where most mountains and hills are 500–1000 m above sea level. Influenced by the subtropical monsoon climate, the annual precipitation in the mountainous areas of southern Anhui varies between 1100 and 2500 mm, 60% of which is concentrated between May and August. The average annual temperature in the WMR is between 15.5 and 16 °C. The main river systems in the WMR include the Qiupu River, the Qingyi River, and the Shuiyang River, which are the main tributaries in the lower reaches of the Yangtze River (Table S1). The gradient of these rivers ranges from 1 to 5‰, and the bottom is mostly rocky. In this study, eDNA surveys were conducted in the Qiupu River, Qingyi River, and Shuiyang River Basins in August 2023 and January 2024. The eight sites in each basin were distributed across tributaries, main streams, and confluences, and 24 sites were investigated in the three basins each season (Figure 1). Based on the method of Strahler [32], tributary and mainstream sampling sites were located on first- or second-order streams and on fourth- or higher-order streams, respectively. Confluence sites were defined as the intersections of second- and third-order streams, or third- and fourth-order streams.

2.2. Water Sample Collection

2.2.1. Collection and Processing of eDNA

To minimize the potential influence of hydrological variability on eDNA detection, all sampling was conducted under baseflow conditions, with sites subject to recent heavy rainfall or sudden discharge events systematically excluded. Sampling was consistently performed between 09:00 and 16:00 to mitigate potential biases arising from diel fluctuations in water temperature. Before collecting water samples, the sampling bottles and hydrophore sampler were cleaned with 10% bleach, and the operators changed to fresh sterilized gloves when new water samples were collected while sampling. Composite samples were obtained for a comprehensive survey of fish species at each sampling site. Specifically, 1 L water samples were collected from the left, middle, and right sections of the river. In the wadable river sections, we also considered different microhabitats, such as riffles and pools. Three replicates were set for each sampling site. The water samples obtained from different replicates at each site were mixed with 3 L of water and used for subsequent filtration and eDNA extraction. A blank control was established with 3 L of distilled water to assess potential contamination by field collection. The water samples were concentrated on mixed cellulose filter membranes (Jinteng 0.45 µm, Tianjin, China) using a vacuum pump within 4–8 h. To avoid cross-contamination between sampling sites, all equipment was sterilized before and after each water sample was filtered [33]. The filter membranes enriched with eDNA were transferred to cryopreservation tubes and stored at −80 °C until further processing.

Total DNA was extracted using PowerWater DNA Isolation Kits (Qiagen, Shanghai, China) following the manufacturer’s instructions. Following the amplification system and PCR procedure described by Zhang et al. [34], the mitochondrial 12S rRNA region was amplified by PCR using the primers Tele02_F (5′-AAACTCGTGCCAGCCACC-3′) and Tele02_R (5′-GGGTATCTAATCCCAGTTTG-3′) [35]. Moreover, the negative and blank controls were amplified using the same PCR protocol to assess potential contamination. The PCR products were analyzed by 2% agarose gel electrophoresis, and the target bands were purified and recovered. The samples were subsequently submitted to Lingen Biotechnology Co., Ltd. (Shanghai, China) for sequencing on the Illumina MiSeq platform. The original sequences were filtered by Trimmomatic v.0.36 [36] to remove low-quality and chimeric sequences. Based on the overlap between the original reads, paired reads were merged into sequences using FLASH v.1.2.11 [37]. Using de novo and reference sequence methods, chimeras were removed using the Usearch software [38] and the Gold database. Sequence clustering was performed using Usearch v.10 based on a similarity threshold of ≥97% [39]. The most abundant sequences were selected as the representative sequences for each operational taxonomic unit (OTU), which refers to a cluster of similar DNA sequences used as a proxy for a species in metabarcoding studies. Preliminary taxonomic identification was conducted by comparing representative sequences with those in the NCBI GenBank database using the Blastn tool, with parameters set to a similarity of ≥97%, coverage of ≥0.9, and e-value ≤ −10⁻⁵ [40]. Further species classification was performed based on a self-built database [41]. The self-built 12S rRNA database was constructed using muscle samples from previously captured fish specimens, along with public sequences from NCBI for species listed above that have not yet been collected. Our self-built database encompasses nearly all the fish species recorded in the drainage area. In cases of discrepancies between the results obtained from the self-built and NCBI databases, the findings from the self-built database were prioritized. OTUs with a relative abundance below 0.001% were excluded [42], and species unrelated to the study area were also removed (e.g., native fish species that are only found in the upper reaches of the Yangtze River). Using this process, the species composition in each basin was determined.

2.2.2. Functional Traits

To assess the functional composition of fish communities in different habitats (i.e., tributary, mainstream, and confluence), 9 morphological metrics that reflect feeding, migration, and habitat use behaviors were selected (Table S2). These morphological metrics for most species are based on individual measurements collected during historical surveys [43,44]. For species that have not been measured historically, morphological metrics were obtained from published data. Based on these 9 morphological metrics, we identified and calculated seven functional traits for subsequent analysis: relative eye size, mouth gape position, relative gut length, eye position, body shape, body height, and Caudal peduncle depth (Table S2).

2.2.3. Phylogenetic Tree Construction

The representative sequences of the most abundant OTUs assigned to each species were used to infer phylogenetic relationships among species. The maximum likelihood (ML) method was used to construct the phylogenetic tree, and the nearest neighbor interchange (NNI) algorithm was applied to optimize its topology [45,46]. The bootstrapping (n = 1000) method was subsequently used to assess the stability and confidence of the tree [47]. The phylogenetic tree was constructed and evaluated using the MEGA 7.0 software [48].

2.2.4. Data Analysis

A Venn analysis was performed to determine the differences in species composition among the three habitats, and the VennDiagram package was used for the Venn analysis [49]. A permutational multivariate analysis of variance (PERMANOVA) was applied to assess variations in fish community composition across the three habitat types. To identify specific species driving these differences among habitat types, a similarity percentage analysis (SIMPER) was performed. Both PERMANOVA and SIMPER analyses were conducted using the vegan package [50]. Based on habitat preferences, fish species were classified into three ecological types: (1) stream species, typically characterized by smaller body size and primarily inhabiting streams with moderate flow velocity and clear water conditions; (2) riverine species, predominantly occurring in deeper-water habitats within the main river channel; and (3) generalist species, exhibiting flexible distributions across both tributary and main channel environments. Based on OTU relative abundance data, we examined the distribution patterns of these three ecological fish types across the three distinct habitat categories.

Alpha diversity was calculated based on the relative abundance of species to assess the differences in community diversity within each habitat. The Shannon index and species richness were used as surrogates for the TD of the fish communities. The functional richness (FRic), which reflects the volume of the convex hull formed by the species in the communities, was used to reflect the functional diversity of the fish communities [51]. Phylogenetic diversity is reflected by the sum of the shortest branch lengths of the phylogenetic tree that corresponds to all species in the community [52]. The TD of the fish communities across different habitats was calculated using the Vegan package in R [50]. The mFD package was used to calculate the functional richness, and the picante package was used to evaluate the phylogenetic diversity of the fish communities in the WMR [53,54]. The differences in the diversity indices between different river habitats were determined by one-way analysis of variance (ANOVA) [55]. All differences were considered to be statistically significant at p < 0.05. ANOVA was conducted using the stats package. All analyses were performed using R 4.4.0 [56].

3. Results

In total, 48 samples were collected during the two seasons. No bands were detected in the negative controls. After the samples were amplified, high-throughput sequencing was performed to generate 26,766,078 raw reads. After quality control and filtering, 9,374,315 reads were retained. During the two seasons, 63 species were identified in the WMR, representing five orders, 15 families, and 47 genera. Specifically, 60 species were observed during the wet season, whereas 51 species were recorded during the dry season. Cypriniformes and Perciformes accounted for 63.36% and 10.62%, respectively. A total of 34 species were shared by all three types of habitats, accounting for 54.0% of the fish species recorded while sampling (Figure 2a). Across the three river basins, a total of 30 fish species were shared, accounting for 47.6% of the total species richness. Qiupu River and Qingyi River each supported six unique species, whereas Shuiyang River harbored two species exclusive to its basin.

The results of PERMANOVA revealed a statistically significant difference in fish community structures across the three habitat types (p = 0.045). The SIMPER identified Cyprinus carpio, Carassius auratus, Zacco platypus, and Aristichthys nobilis as the primary species contributing to the observed differences in fish community composition, collectively accounting for more than 40% of the variation among habitats (Supplementary Material S1).

Fish belonging to different ecological groups showed distinct distribution patterns across the three habitat types (Figure 3 and Figure 4). Riverine fish represent the dominant group in the mainstream, comprising 50.4% and 42.5% of the assemblage during the dry and wet seasons, respectively. Generalist species constituted a notable proportion in each habitat, with an average relative abundance of 43.6%. Stream fish represented the smallest proportion among all ecological groups, with a mean relative abundance of 26.1%, and were primarily concentrated at river confluences during both seasons (Figure 3). In different habitats, the dominant species of riverine and generalist fish exhibited relative consistency. For instance, Aristichthys nobilis was the dominant species within riverine fish group, while Cyprinus carpio and Carassius auratus were prevalent in the generalist fish group, all showing high relative abundance across the three habitats (Figure 4, Figures S1 and S2). In the stream fish group, the dominant species vary across habitat types. For example, the Vanmanenia stenosoma was the dominant species in tributaries and confluences but is rarely observed in the mainstream.

The biodiversity analysis revealed that although confluence sites generally exhibited higher diversity, the magnitude and statistical significance of this pattern varied across rivers (Table S3). In the Qiupu River, both confluence and tributary sites exhibited significantly higher phylogenetic diversity than mainstream sites. In the Qingyi River, phylogenetic diversity was also significantly higher at tributary sites compared to mainstream segments. In contrast, in the Shuiyang River, taxonomic diversity was significantly greater at confluence sites than at mainstream locations.

The Shannon diversity across various river network positions showed consistent patterns in both seasons, with the highest values observed at river confluences, followed by tributary sites, and the lowest values at the mainstream locations (Figure 5). Additionally, during the dry season, a significant difference (p = 0.019) was found in the Shannon index values of the fish communities between the main stream and the confluence. The functional richness of the community in the tributaries was the highest during the wet season, followed by that in the confluence, while that in the mainstream was the lowest (Figure 5). In contrast, during the dry season, the highest functional richness was observed at the confluence, followed by the tributaries, and the mainstream had the lowest functional richness. A significant difference was found between the confluence and mainstream communities (p = 0.027). For phylogenetic diversity, a significant difference was found between the tributaries and the main stream during the wet season (p = 0.039). The phylogenetic diversity of the community was the highest in the tributaries, followed by the confluence, with the lowest diversity in the main stream. In the dry season, the phylogenetic diversity of the confluence community was the highest, followed by that of the tributaries, with the main stream showing the lowest diversity. A significant difference in phylogenetic diversity was observed between the main stream and the confluence (p = 0.033) (Figure 5).

4. Discussion

The biodiversity pattern of fish communities in river networks is an important research topic in river ecology, but the multidimensional diversity distribution pattern of fish communities in different river network locations is unclear. In this study, using the eDNA metabarcoding method, we analyzed the fish community structure in tributaries, main streams, and confluences of mountain rivers, as well as the taxonomic, functional, and phylogenetic diversity distribution patterns of the fish communities.

4.1. Comparison of eDNA Metabarcoding and Traditional Monitoring in the WMR

In this study, the eDNA metabarcoding results were not entirely consistent with previous findings from traditional survey methods, such as electrofishing and gillnetting [28]. Among the 63 species detected based on eDNA in this study, except for two alien species (Coptodon zillii and Micropterus salmoides), all other species have been recorded in previous fish investigations, which documented 42 species [43,57]. These results underscore the high efficiency of eDNA metabarcoding in fish surveys. However, the discrepancy may also be attributed to the focus of previous traditional surveys on stream headwaters, which likely resulted in the underrepresentation of fish species inhabiting the middle and lower reaches [43].

Compared to the more comprehensive list of fish species in the WMR [58], which recorded a total of 104 species, many fish species were not detected in this study based on eDNA metabarcoding. The eDNA monitoring results suggested that it was less effective in distinguishing some closely related fish species, such as Acheilognathus species. There is a possibility that multiple cryptic species exist within this genus [59]; however, the limited availability of molecular reference data constrained the taxonomic resolution of the eDNA-based approach, preventing accurate species-level discrimination. In addition, detecting small and demersal fish using eDNA extracted from water samples in streams is difficult because their genetic segments may be more concentrated in the bottom sediment [41,60,61]. Some species, such as Odontobutis potamophilus, exhibit specialized spawning behaviors by depositing eggs in crevices between riverbed stones or in burrows within the substrate, which may further limit the release of detectable eDNA into the surrounding water. However, in this study, eight fish species were detected only at the confluence, including some small benthic fish, such as Odontobutis potamophilus and Leptobotia bellacauda. The genetic fragments of these fish from nearby tributaries were likely released into the water environment at confluences due to specific sediment transport processes and hydrological conditions, such as turbulent water flow caused by the separation zone and shear layer [24]. Despite the efficiency of eDNA metabarcoding, methodological limitations remain critical for accurate biodiversity assessments [28]. Our study highlights potential biases against detecting small-bodied, benthic fish species, or closely related fish species. To mitigate these limitations, future efforts should prioritize the enrichment of local genetic reference databases, development of taxon-specific primers, and inclusion of sediment-based eDNA sampling approaches [61,62].

Due to potential false-positive and false-negative errors, the results of eDNA metabarcoding may deviate from the actual scenario [63,64]. For example, eDNA may detect fish species that are not present in the watershed or fail to identify certain fish due to eDNA degradation [65]. Establishing a local database of fish in the study area can enable accurate classification and identification [9,66]. Several studies have reported the composition and distribution of fish in the WMR, and in another study, we obtained sufficient fish samples to construct a local database of fish [43,57,67]. In this study, we detected the endangered fish species Pseudorasbora elongata in the Qiupu River based on the self-built database. This species is characterized by a small wild population and a narrow distribution range [68]. eDNA monitoring offers a novel approach for identifying the distribution range of rare species and implementing targeted conservation strategies for their populations.

4.2. Variations in Fish Community Composition and Biodiversity Across Different Habitats

The compositions of the fish communities in the mainstream and tributaries usually differ greatly because of their distinct habitats [21]. The confluences can serve as an environmental filter and limit the dispersal of fish from tributaries to the mainstream [25]. However, in this study, the fish communities in the mainstream and tributaries shared 25 species, accounting for 78.1% and 69.4% of the total species in the fish communities in the mainstream and tributaries, respectively. The PD of the fish communities in the tributaries was significantly greater than that in the mainstream in the wet season. These results did not support our first hypothesis, and we probably underestimated the dispersal process of fish in the river network [69,70]. We found several typical mainstream species in the fish communities of the tributaries, such as Culter mongolicus, Aristichthys nobilis, and Mylopharyngodon piceus. Owing to the unidirectional mobility of eDNA in rivers, these results indicated that mainstream fish can migrate upstream in tributaries, highlighting the role of tributaries in providing reproduction, foraging, and refuge for fish [70,71].

Mainstreams and tributaries provide diverse habitats to fish throughout different stages of their lives. For example, the greater water depths of the mainstream provide overwintering habitats for fish, whereas tributaries provide habitats for fish to forage and reproduce [72,73]. The confluence is the transition zone between the main stream and tributaries, in which the hydrological regime and material transport are regulated by the tributaries and the main stream, resulting in a complex and seasonal habitat [24,74]. In this study, the TD of the fish community in the confluence area was significantly greater than that in the mainstream area during the dry season, which was consistent with our second hypothesis. This may be related to the life history of the fish in the river network. Yuan et al. 2023 reported that the TD of fish communities at the confluence of the Yangtze River and Poyang Lake was greater in spring than in autumn because greater connectivity allowed fish from the lakes to enter the confluence [24]. We hypothesized that in the rivers of the WMR, fish from tributaries disperse to the confluence during the dry season to overwinter, thus increasing the TD of the fish communities at confluences. The changes in FD and PD during the dry season also showed similar patterns, which also supported our hypothesis to some extent that fish from tributaries increased the biodiversity of the fish communities at the confluence.

Our results underscore the importance of tributaries and confluences as critical biodiversity nodes within river networks, a finding consistent with habitat heterogeneity theory and metacommunity dynamics [23,74]. The elevated multidimensional biodiversity observed at confluences during the dry season suggests these areas function as ecological sinks by integrating species pools from adjacent habitats through seasonal dispersal, an observation aligning well with the “confluence effect” described in dendritic riverine systems [22]. Such confluence zones play pivotal roles as connectivity corridors, facilitating critical seasonal fish migrations between tributary spawning sites and mainstream overwintering habitats [69]. These findings advocate strongly for prioritizing the conservation of confluence areas, particularly through measures aimed at maintaining hydrological connectivity and mitigating habitat fragmentation, such as dam construction and removal strategies. Tributaries, characterized by elevated biodiversity during wet seasons, serve as refugia for specialized species (e.g., Vanmanenia stenosoma), further emphasizing the need for targeted sediment management and microhabitat preservation. Consequently, river management strategies should explicitly include protection and restoration of these habitats, recognizing their disproportionate contribution to regional biodiversity. Integrating eDNA-based biodiversity assessments into adaptive river management frameworks could operationalize these insights, enabling dynamic and responsive habitat protection strategies tailored to seasonal variability.

5. Conclusions

There are seasonal differences in the diversity distribution patterns of fish communities at different river network locations, which may be related to the hydrological regime of the river and the life history of fish. This study highlighted the importance of confluences and tributaries in river networks, which should be prioritized in the monitoring, conservation, and management of river fish.