Three sampling strategies reveal spatial and methodological patterns in macroinvertebrate diversity in freshwater habitats

Summary

Assessing macroinvertebrate diversity is crucial for freshwater ecosystem management. Environmental DNA (eDNA) metabarcoding has emerged as a powerful, non-invasive tool that complements conventional morphological methods. In this study, we compared morphological sampling, ethanol-derived eDNA from preserved specimens, and water-based eDNA to evaluate macroinvertebrate diversity in the Burhatong and Hailan Rivers of the Tumen River Basin, China. Water-derived eDNA revealed the highest species richness and α-diversity, underscoring method-specific detection patterns. Despite these biases, strong correlations between metabarcoding read abundance and morphological data supported the reliability of all three approaches for within-habitat (α) diversity assessment. Consistent β-diversity patterns across methods further validated their comparability. Among environmental factors, total suspended solids (TSS) emerged as the dominant driver, with α-diversity peaking at moderate TSS levels, consistent with the intermediate disturbance hypothesis. By integrating morphological identification, ethanol-eDNA, and water-eDNA metabarcoding, this study demonstrated that despite method-specific biases, all the approaches captured consistent ecological gradients in riverine macroinvertebrate communities along an East Asian international border.

Graphical abstract

Highlights

• •Three sampling methods capture distinct but complementary macroinvertebrate diversity
• •Water-based eDNA detects the highest taxon richness and α diversity
• •Strong correlations found among morphological and eDNA-based diversity metrics
• •Total suspended solids significantly shape macroinvertebrate community structure

Aquatic science; Biological sciences; Zoology; Molecular biology; Aquatic biology

Introduction

Human activities, climate change, and environmental pollution have significantly contributed to the global loss of biodiversity, particularly in freshwater habitats,^1,2,3,4 where populations decline by more than 80% in forty years.⁵ Freshwater ecosystems are crucial for supporting a variety of aquatic species, maintaining ecosystem functions, as well as providing essential services for human beings.^6,7 To conserve and manage freshwater ecosystems, it is essential to first understand the distribution patterns of organisms and the factors shaped them.

Macroinvertebrates are integral components of aquatic food webs and widely used as bioindicators of the health of corresponding freshwater ecosystems, based on their community composition and diversity patterns.^8,9 In particular, insects of the orders Ephemeroptera, Plecoptera, Trichoptera, and Diptera (EPTD), which often constitute a significant proportion of freshwater macroinvertebrate communities, are highly sensitive to environmental change and provide valuable signals of ecological condition.^10,11 Traditional surveys routinely rely on morphological sampling, using kick nets or sediment samplers, which potentially cause disturbance to local habitats.^12,13,14,15 Moreover, morphological identification STAR Methods are time-consuming and skill-challenging, particularly for organisms in varying larval stages.^16,17

Environmental DNA (eDNA) refers to the genetic material obtained from biological metabolic waste, damaged tissue, or shed skin cells that can be recovered from environmental samples (e.g., water, soil, air, sediment).¹⁸ By applying a metabarcoding approach, that is, extracting total DNA, PCR-amplifying standardized barcode regions, and conducting high-throughput sequencing, researchers can efficiently characterize species composition and biodiversity in complex ecosystems. The emergence of eDNA technology has gained prominence as a complementary tool to conventional biodiversity assessment STAR Methods, offering high-resolution, non-invasive detection of species, and more cost-effective surveys.^18,19,20 For instance, McInerney and Rees (2016,2018) found that eDNA and traditional STAR Methods produced similar results in assessing macroinvertebrate communities in Australian streams, but eDNA approach incurred only one-sixth of the cost of traditional techniques.^19,20 Subsequent studies have confirmed that eDNA metabarcoding is a powerful tool for elucidating biodiversity and inferring mechanisms of ecological evolution. Moreover, it provides high-resolution taxonomic information that overcomes many limitations inherent in traditional morphological surveys.^21,22,23 Practical applications include the detection of endangered or invasive species; for example, eDNA analysis successfully detected the presence of the critically endangered Swinhoe’s softshell turtle (Rafetus swinhoei) in Vietnam’s Xuan Khanh Lake.²⁴ eDNA has also proven effective for tracking ecological disturbances and augmenting water quality evaluation. Fernández et al. (2018), for instance, reported that eDNA-based taxonomic profiles strongly correlated with conventional physicochemical stress indicators and served as a robust complementary tool for water quality evaluation.²⁵ Clark et al. (2020) demonstrated high concordance between eDNA-based and morphological assessments in detecting benthic communities responding to nutrient loading.²⁶

Compared with morphological sampling, eDNA collected from natural water samples (water-eDNA) often detects a greater number of species.^11,27,28 However, it may reflect DNA transported from upstream habitats, complicating local macroinvertebrate communities inference.²³ Thus, some researchers suggested that water-eDNA is more suitable for large-scale biodiversity assessments.²⁹ In contrast, preservative used in bulk invertebrate collections captures DNA (ethanol-eDNA) directly from cells and tissues released by organisms, often showing higher congruence with local assemblages and species-level diversity.^23,30,31 It is particularly useful for detecting species in targeted collections, though its performance may be influenced by the sampling effort and preservation conditions (e.g., ethanol concentration). Sediment and bulk tissue samples are also widely used in macroinvertebrate eDNA studies. Nevertheless, tissue samples require destructive processing that is incompatible with preservation standards mandated by national regulatory agencies and may overlook rare species due to uneven biomass contributions.^31,32 Sediment samples, by contrast, demand labor-intensive processing and may contain historical DNA, thereby complicating interpretations of present-day community composition.^33,34,35

Despite growing use of molecular approaches in freshwater biodiversity monitoring, we prioritized water-eDNA and ethanol-eDNA due to their non-invasive, cost-effective, and scalable, while enhancing the detection of locally abundant species. However, few studies have systematically compared these approaches or evaluated their consistency in revealing community-environment relationships. To address this gap, we integrated conventional morphological sampling with ethanol- and water-eDNA metabarcoding to assess macroinvertebrate diversity in the Burhatong and Hailan Rivers of the Tumen River Basin, Northeast China. Specifically, we focused on (1) how do taxonomic detection patterns differ among three STAR Methods, (2) do β-diversity patterns and environmental associations remain consistent across approaches, and (3) which environmental drivers shape community composition and diversity. By combining methodological and ecological perspectives, our study provided new insights into the reliability and complementarity of eDNA-based biodiversity monitoring in freshwater ecosystems.

Results

Macroinvertebrate community composition

A total of 2,398 individuals were collected from the Buerhatong River and Hanlan Rivers using morphological sampling, representing 189 taxa across 3 phyla, 4 classes, 15 orders, 55 families, and 114 genera. Among these, 99 taxa were confidently identified to the species level, with the remainder assigned to genus or higher taxonomic ranks. Among the 189 taxa identified morphologically, 140 (74.1%) had corresponding COI barcode records in the GenBank database used for taxonomic assignment (Data S1).

Ethanol-eDNA metabarcoding produced a total of 2,669,728 sequence reads, identifying 174 taxa from 3 phyla, 5 classes, 18 orders, 56 families, and 107 genera, including 147 valid species. In the Water-eDNA sample, 131,641 reads were obtained, detecting 212 taxa across 3 phyla, 5 classes, 18 orders, 66 families, and 137 genera, with 176 valid species assigned. The three phyla identified were consistent across all sampling types, and the five classes identified in both eDNA approaches were identical (Table 1). From order to species level, taxa became progressively less shared among STAR Methods (Figure 2). Overall, the two eDNA approaches were more similar to each other than either was to morphological sampling across sampling sites (Figures 2 and S1).

Table 1.

Order-level macroinvertebrate taxa and their abundances detected with the LCO1490/ZBJ-ArtR2c primer pair across the two eDNA and morphological approaches

Class	Order	Morphology (N/Individuals)	Ethanol-eDNA (N/Reads)	Water-eDNA (N/Reads)
Clitellata	Arhynchobdellida	3/12	–	–
–	Crassiclitellata	–	2/57	1/1039
–	Enchytraeida	–	1/52	1/41
–	Rhynchobdellida	1/1	–	1/66
–	Tubificida	3/17	6/23572	4/581
Arachnida	Araneae	–	1/1523	1/456
Collembola	Entomobryomorpha	–	2/119	–
–	Poduromorpha	–	–	1/3
Gastropoda	Caenogastropoda	4/14	–	1/7
–	Heterobranchia	8/21	1/3666	1/54
–	Stylommatophora	–	1/3	1/74
Insecta	Coleoptera	5/8	8/11976	9/3956
–	Diptera	83/542	83/768337	91/72678
–	Ephemeroptera	32/926	11/1088378	7/2593
–	Hemiptera	1/3	1/69	2/48
–	Hymenoptera	–	2/151	4/962
–	Lepidoptera	1/1	20/50872	56/10699
–	Mecoptera	–	1/911	1/144
–	Megaloptera	1/1	–	–
–	Odonata	8/21	1/64	–
–	Plecoptera	15/193	10/374888	12/20670
–	Psocoptera	–	2/116	–
–	Trichoptera	22/624	21/344947	18/17570
Malacostraca	Amphipoda	2/14	–	–
Total		189/2398	174/2669780	212/131641

Figure 2.

Taxa richness detected by three STAR Methods across taxonomic levels.

The Kruskal-Wallis test results revealed significant variation in the number of species across EPTD orders and among non-EPTD macroinvertebrates across three sampling types. Pairwise comparisons further showed that both eDNA approaches consistently detected significantly more species than morphological observation in all other groups (Figure 3), except for Ephemeroptera, where morphological identification detected more species than the two eDNA metabarcoding STAR Methods. No significant species number differences within or across EPTD orders were detected between the two eDNA approaches, except in the detection of non-EPTD macroinvertebrates (Figure 3).

Figure 3.

Species richness detected by three sampling methods

Richness of detected species within and among EPTD orders and other macroinvertebrates by three sampling types (∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.001; and ∗∗∗∗, p < 0.0001; ns, p ≥ 0.05).

Macroinvertebrate diversity

In terms of alpha diversity, all indices estimated from water-eDNA were significantly higher than those from ethanol-eDNA (Figure 4; Table S1). Both eDNA STAR Methods showed significantly higher species richness, ACE, and Chao1 indices than the morphological method (p < 0.05). However, Shannon-Wiener and Simpson indices from water-eDNA did not differ significantly from morphological sampling, whereas those from ethanol-eDNA metabarcoding were significantly lower (p < 0.05).

Figure 4.

Comparison of the alpha diversity of macroinvertebrate communities among three sampling types

(∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.001; and ∗∗∗∗, p < 0.0001; ns, p ≥ 0.05).

Significant correlations were found between morphological counts and read abundance from both eDNA metabarcoding STAR Methods (Figure S2). Scatterplot matrices further revealed significant correlations (p < 0.05) between Shannon-Wiener, species richness, and Chao1 indices derived from ethanol-eDNA metabarcoding reads and morphological taxa (Figure S3). Water-eDNA metabarcoding reads showed a significant correlation with taxa number identified by morphology only for species richness (p < 0.05). No other alpha diversity indices were observed to be significantly correlated among three sampling types (p > 0.05).

Procrustes analysis of NMDS ordinations based on Bray-Curtis distance revealed significant consistency in macroinvertebrates community structures of beta diversity across all three STAR Methods (Procrustes: p < 0.05; Figure 5).

Figure 5.

Procrustes analysis on non-metric multidimensional scaling (NMDS) ordinations of beta diversity based on Bray-Curtis distance

Congruence between (A) water-eDNA metabarcoding and morphological method, (B) congruence between ethanol-eDNA metabarcoding and morphological method, and (C) congruence between the two eDNA metabarcoding STAR Methods. Lower M2 values indicate greater similarity between the two configurations, and significance was assessed using a permutation test.

Effects of environmental variables on macroinvertebrate communities

Hierarchical partitioning (HP) analysis showed that among the eight environmental variables, total suspended solids (TSS) generally made the most independent contribution to macroinvertebrate community composition across all STAR Methods (Figure 6; Table S2). TSS and total nitrogen (TN) accounted for significant and independent contributions of 16.77% and 16.83% (p < 0.01), respectively, in morphologically identified communities. For ethanol-eDNA and water-eDNA metabarcoding-detected communities, TSS independently explained 18.07% (p < 0.01) and 13.59% of variation, respectively. Significant independent contribution of TSS reached 47.74% (p < 0.01) for alpha diversity across all STAR Methods, whereas no other environmental variable was statistically significant (Figure 6).

Figure 6.

Independent importance of environmental factors for macroinvertebrate communities

Independent importance (%) of environmental factors assessed by hierarchical partitioning (HP) analysis for macroinvertebrate communities observed in (A) morphological sampling, (B) ethanol-eDNA metabarcoding, and (C) water-eDNA metabarcoding, as well as for (D) alpha diversity metrics across the three STAR Methods. (Chl-a, chlorophyll a; TN, total nitrogen; TP, total phosphorus; TSS, total suspended solids; NO3.N, nitrate nitrogen; NH, ammonia nitrogen; PO, phosphate^-; COD, permanganate index. Abbreviations are also defined in the method details section for clarity. ∗, p < 0.05 and ∗∗, p < 0.01).

Redundancy analysis (RDA) with permutation tests confirmed that environmental factors statistically significantly constrained alpha diversity metrics of macroinvertebrates (p = 0.015), explaining 42.5% of the total variance. The two primary axes, RDA1 and RDA2, accounted for 68.32% and 18.35% of the total variance, respectively (Figure 7A). RDA1 was negatively associated with TSS, NH₃-N, and TN (biplot scores = -0.9590, -0.5445, and -0.3564), and corresponded with ACE, Chao1, and species richness from water-eDNA (scores = 0.3488, 0.3510, 0.4030) and morphological taxa (scores = 0.4250, 0.4814, 0.3719). RDA2 was positively associated with TSS and PO₄^3- (scores = 0.2403 and 0.3435), while negatively correlated with NH₃-N (biplot scores = -0.2049), and explained variation in ACE, Chao1, and species richness from ethanol-eDNA (scores = -0.5686, -0.5505, and -0.5237). Among all environmental variables, only TSS had a significant negative influence on macroinvertebrate alpha diversity indices (p < 0.05).

Figure 7.

Environmental effects on macroinvertebrate alpha diversity

Response of macroinvertebrate alpha diversity across the three STAR Methods to (A) eight environmental factors, and (B) TSS gradients. (Morph, morphological taxa; ethanol, ethanol-eDNA metabarcoding; water, water-eDNA metabarcoding; Chl-a, chlorophyll a; TN, total nitrogen; TP, total phosphorus; TSS, total suspended solids; NO3.N, nitrate nitrogen; NH, ammonia nitrogen; PO, phosphate^-; COD, permanganate index).

Generalized additive models (GAMs) fitted to NMDS ordinations were used to visualize macroinvertebrate alpha diversity index distribution patterns across three STAR Methods along a gradient of environmental variables.³⁶ NMDS results overlaid onto TSS surfaces with GAMs, revealing that TSS account for 60.15% of the variation in macroinvertebrate alpha diversity indices across three sampling approaches (p < 0.01). Species richness, Chao1, and ACE from morphological data were associated with lower TSS values, whereas the Shannon-Wiener and Simpson indices from morphology, together with all alpha diversity indices from both eDNA metabarcoding STAR Methods, peaked at intermediate TSS levels (Figure 7B).

Discussion

Variation in detected macroinvertebrate communities across sampling strategies

Overall, the results revealed generally consistent macroinvertebrate compositions across sampling strategies at higher taxonomic levels, but divergences emerged at finer taxonomic resolution, particularly at the species level. This may suggest that each sampling strategy captures complementary aspects of the community.^23,37

As expected, water-eDNA metabarcoding identified more taxa than morphological identification, aligning with previous studies.^22,38 The ability of water-eDNA to capture DNA from rare, cryptic, or early-life-stage organisms likely explained its broader taxonomic coverage than morphological sampling.^39,40 However, notable exceptions occurred; for example, Ephemeroptera were better represented by morphological observation than eDNA metabarcoding. This likely reflected primer bias in metabarcoding, inefficient amplification of specific taxa observed in morphological samples, or gaps in reference sequences in public databases.^41,42,43,44

Consistent with a pattern in prior studies, fewer macroinvertebrate taxa were detected using ethanol-derived eDNA metabarcoding than morphological identification.^30,43 This may be due to the variable amount of DNA released into ethanol solution, which varies depending on the organism’s body structure. Taxa with protective shells (e.g., Caenogastropoda) or hard exoskeletons (e.g., Odonata) are likely to release fewer cells or less free DNA, reducing their detection probability compared to soft-bodied organisms.^30,31,45 Additional factors such as pre-sorting bias during sample handling or DNA source and fate, PCR primer bias, sampling and filtration efficiency, or the absence of corresponding reference sequences in public databases may also contribute to reduced detection.^41,42,43,44 Among these, primer bias was a particularly plausible contributor to the differences observed across sampling STAR Methods. Primer-template mismatches and taxon-specific amplification efficiencies can suppress detection of specific taxa, thereby reducing apparent overlap with morphological identification. We therefore interpret differences cautiously and recommend multi-primer, multi-marker designs (e.g., BF2/BR2, fwhF2/fwhR2n, mlCOIintF/jgHCO2198) in future work to mitigate primer bias.

Notably, in addition to differences in taxonomic detection, a pronounced discrepancy in sequencing depth was observed between ethanol-eDNA and water-eDNA samples. Both eDNA STAR Methods underwent identical workflows, that is, DNA extraction, PCR amplification, sequencing, and bioinformatic workflows, reflecting intrinsic differences in DNA quantity rather than methodological bias. Water-eDNA generally contains lower concentrations of degraded, diluted, or particle-bound DNA.^11,28 In contrast, ethanol-eDNA benefits from the accumulated cellular material and tissue fragments released during preservation.^23,30 Despite fewer reads in water-eDNA, rarefaction analyses confirmed sufficient sequencing depth to capture robust α- and β-diversity inference. Together, these findings highlighted the complementary nature of ethanol- and water-eDNA approaches, with ethanol-eDNA providing higher sensitivity to locally abundant taxa, while water-eDNA reflecting broader environmental DNA pools directly.

Macroinvertebrate diversity patterns

Alpha diversity indices differed significantly among STAR Methods (Figure 4), reinforcing that detection strategy strongly influenced diversity estimates. Nevertheless, significant correlations between morphological and eDNA-derived abundance measures indicated that all three approaches captured macroinvertebrate within-habitat diversity patterns (Figure S2). In particular, the significant correlations of the Shannon-Wiener index, species richness, and Chao1 between ethanol-eDNA and morphological samples (Figure S3) highlighted the consistency of these two STAR Methods. Within this context, eDNA metabarcoding offered a practical alternative for macroinvertebrate surveys, reducing costs in both field and laboratory work and requiring less taxonomic expertise.^46,47 Indeed, several studies have suggested that read abundance from eDNA data provided a more informative measure of macroinvertebrate biodiversity than presence-absence data from morphological counts.^22,33,48 These findings reinforced the view that eDNA metabarcoding can serve as a convenient alternative to morphology-based identification, offering cost- and labor-efficiency. However, alpha diversity estimated from eDNA metabarcoding may remain susceptible to methodological biases, including DNA degradation, variation in extraction efficiency, PCR primer bias, and gaps in reference databases, which can lead to under- or over-estimation of detected taxa.

At the beta diversity level, the significant consistency in community structure across sampling types suggested that all three STAR Methods reliably revealed large-scale macroinvertebrate community patterns. Previous studies have reported such consistency between ethanol-eDNA metabarcoding and morphological sampling in large-scale,^22,31 and our results extended this finding by demonstrating that water-eDNA metabarcoding also detected community patterns similar to those identified by morphological sampling and ethanol-eDNA metabarcoding. This consistency may arise due to NMDS reducing random noise in the community matrix and accentuating underlying ecological gradients.^38,49 Notably, while methodological biases inherent to eDNA metabarcoding may substantially affect alpha diversity estimates, their influence on beta diversity appeared more limited, as all three sampling strategies consistently detected community-level ecological gradients.

Environmental driving force

Assessing how macroinvertebrate communities’ response to environmental change is central to understanding ecosystem stability and diversity.⁵⁰ Hierarchical partitioning (HP) analysis indicated that total suspended solids (TSS) was the primary environmental driver of both community composition and diversity (Figure 6). Additionally, redundancy analysis (RDA) further confirmed that TSS exerted a significant influence on alpha diversity along the Buerhatong River, while generalized additive models (GAMs) indicated that alpha diversity peaked at intermediate TSS levels (Figure 7). This pattern aligned with the intermediate disturbance hypothesis,⁵¹ which posited that moderate levels of disturbance promoted species coexistence by reducing competitive exclusion while avoiding the loss of sensitive taxa. Previous studies have shown that elevated TSS can alter aquatic habitats, clog respiratory systems, and cause abrasion or mechanical damage, thereby reducing feeding efficiency and ultimately shaping species distributions.^52,53 Consequently, habitat heterogeneity may increase at intermediate TSS, allowing tolerant and sensitive taxa to coexist, whereas high TSS imposes physical stress that excludes sensitive species, and low TSS favors competitive dominants, both leading to reduced diversity.

Interestingly, species richness, Chao1, and ACE derived from morphological samples peaked at lower TSS levels, likely reflecting methodological limitations. Elevated TSS impaired specimen detectability and integrity in morphological surveys, whereas eDNA approaches were less affected and could still recover DNA fragments.^54,55 Thus, TSS influenced macroinvertebrate diversity not only through ecological mechanisms but also through interactions with the sampling method.

Limitations of the study

While our study offered valuable insights into the use of multiple sampling STAR Methods for assessing macroinvertebrate diversity, several limitations should be acknowledged when interpreting our findings. First, the limited availability of genetic reference data for macroinvertebrates in the Tumen River Basin led to underrepresentation of regional taxa in the GenBank database, thereby reducing the efficiency of taxonomic assignment. Second, primer choice (LCO1490/ZBJ-ArtR2c) and a relatively low sequencing depth per sample (below the recently recommended ≥200,000 reads⁵⁶) may have constrained the detection of rare taxa, although rarefaction analyses indicated sufficient coverage to capture dominant diversity patterns. Moreover, metabarcoding read counts should be interpreted as relative proxies for taxon abundance rather than direct measures of organism abundance, because sequence reads are influenced by multiple biological and methodological factors, including DNA shedding rates, environmental degradation, PCR amplification efficiency, primer bias, sequencing depth, and bioinformatic processing. Consequently, quantitative comparisons based on read abundance should be interpreted with appropriate caution. Third, the restriction of our sampling to two rivers within the basin and a single time point (June 2022) resulted in limited spatial and temporal coverage, which may constrain the generalizability of our findings to other seasons or regions. Finally, although total suspended solids (TSS) were identified as a key environmental driver of macroinvertebrate community structure, the present study cannot fully disentangle the effects of environmental variation from methodological differences. Future research with optimized primer sets, deeper sequencing, improved reference databases, and broader spatiotemporal sampling will enhance the resolution and applicability of eDNA-based biodiversity assessments.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to the lead contact, Jie Han (jiehan@bnu.edu.cn).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •The sequencing data that support the findings of this study are publicly available at Mendeley Data: https://doi.org/10.17632/x9x4d95j7x.2.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This research was supported by the National Science and Technology Basic Resources Survey Program of China (2019FY101700).

Author contributions

Conceptualization, J. H., and L-J.L.; methodology, L-J.L., Q-R.W., and T-X.P.; investigation, L-J.L., X-W.Y., and T-X.P.; writing – original draft, L-J.L.; writing – review and editing, L-J.L., Q-R.W., T-X.P. X-W.Y., and J.H.; funding acquisition, X-W.Y. and J.H.; supervision, J.H. All authors have read and agreed to the published version of the manuscript.

Declaration of interests

The authors declare that they have no conflict of interest.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Biological samples
31 morphological samples	Burhatong and Hailan Rivers, Tumen River Basin, Northeastern China	N/A
31 water samples	Burhatong and Hailan Rivers, Tumen River Basin, Northeastern China	N/A
31 ethanol-based samples	Burhatong and Hailan Rivers, Tumen River Basin, Northeastern China	N/A
Chemicals, peptides, and recombinant proteins
2 × Rapid Taq Master Mix	Vazyme Biotech	Cat# 7E572A1
Bovine serum albumin (BSA)	Sigma	Cat# 102273805
Trans2K DNA Marker	TransGen Biotech	Cat#R30523
SYBR™ Gold	Thermofisher	Cat# S33102
Biowest Regular Agarose G-10(Agarose Gel)	Biowest	Cat# 111860
Critical commercial assays
Mixed cellulose esters membrane filter	Millipore	Cat# RAWP04700
DNeasy Blood and Tissue Kit	QIAGEN	Cat# 172030100
High-throughput sequencing	BGI	BGI Group Official Website
Deposited data
Raw sequencing data	This paper	Available at Mendeley data https://doi.org/10.17632/x9x4d95j7x.2
Oligonucleotides
LCO/ZBJ primers – Forward: LCO1490: 5′-GGTCAACAAATCATAAAGATATTGG-3’; Reverse: ZBJ-ArtR2c: 5′-WACTAATCAA TTWCCAAATCCTCC-3′	Folmer et al.,57, Zeale et al.58	N/A
Software and algorithms
QIIME2	Bolyen et al.59	https://qiime2.org/
GenBank database	N/A	https://www.ncbi.nlm.nih.gov/
R version 4.1.1	Team et al.60	https://cran.r-project.org/bin/windows/base/old/

Experimental model and study participant details

Omitted as our study does not involve experimental biological models or human participants.

Method details

Study area and sampling

Sampling was conducted at 31 sites along the Burhatong and Hailan Rivers, covering an area of approximately 7,064 km² in the Tumen River Basin, northeastern China, in June 2022 (Figure 1). The elevation of the sampling sites ranged from 200 to 1,300 m. The upper reaches of the rivers are predominantly mountainous, characterized by forested landscapes, narrow river channels, and substrates primarily composed of stones and pebbles. In contrast, the middle and lower reaches are mostly plains, dominated by cropland and impervious surfaces, with wider and shallower riverbeds, slower water flow, and substrates mainly consisting of gravel and sand.

Figure 1.

Map of the sampling sites in the Tumen Basin, northeast China.

Sampling typically began at the upstream sites and progressed downstream. To avoid cross-contamination, water-eDNA samples were collected immediately upon arrival at each sampling site. The sampler wore disposable sterile masks and gloves throughout the process. Disposable sterile plastic bottles were used to collect three 1 L water samples^29,61 at each site - one each from the left, center, and right sides of the river surface at a depth of approximately 5 cm. The collected water samples were stored in ice boxes and transported to the laboratory, where they were filtered the same day using a vacuum pump (Millipore, Tianjin) with mixed cellulose ester filters of 1.2 μm pore size (Merck Millipore Ltd., Carrigtwohill, Ireland).⁶²

An additional 2 L water sample was collected using disposable sterile plastic bottles, stored in a cooler at 4 °C, and transported to the laboratory. Eight parameters, including total suspended solids (TSS), ammonia nitrogen (NH₃-N), nitrate nitrogen (NO₃-N), phosphate (PO₄^3-), chemical oxygen demand based on permanganate measurement (COD), total nitrogen (TN), total phosphorus (TP), and chlorophyll a (Chl-a), were measured following the standard STAR Methods outlined in the Chinese Surface Water Quality Standards (Table S3).⁶³

After water sample collection, a 30 cm × 30 cm D-frame net with 40 mesh (0.25 mm pore size) was used to collect macroinvertebrates from a 100 m stretch of the river by moving against the current. The collected sample was sieved through a circular sieve (40 mesh) to remove debris, and the organisms were transferred to a white porcelain tray for field sorting. The sorted specimens were preserved in 99% ethanol in 500 mL wide-mouth plastic bottles and transported to the laboratory. In the laboratory, specimens were morphologically sorted and identified to the lowest possible taxonomic level using a dissecting stereo microscope (ZEISS Discovery. V8, Germany) and microscope (SOPTOP EX20, China), following freshwater invertebrate identification guidelines.^64,65,66

After 20 days of preservation in ethanol, 10 mL of the preservative solution was filtered for macroinvertebrate ethanol-eDNA analysis.³⁸ Morphological and ethanol-eDNA samples were collected as whole bulk samples without further biological replication. Water-eDNA and ethanol-eDNA filters were stored in a freezer at -20 °C and extracted within one month of filtration.

eDNA metabarcoding

Water-and ethanol-eDNA from each sampling site were extracted following the protocols described in Li et al.³⁷ and Zhang et al.⁶² Specifically, filtration units and laboratory tools were thoroughly rinsed with a 10% bleach solution and double-distilled H₂O (ddH₂O) between sampling sites to remove potential contamination. Negative controls--filtered (one laboratory blank per sample) and extracted (one extraction blank per batch) -- were prepared separately for water-eDNA and ethanol-eDNA during the filtration and DNA extraction processes to monitor potential contamination. The extracted DNA was quantified using NanoDrop One (Thermo Fisher Scientific, USA) and then diluted 5-fold to reduce the PCR inhibitors, creating the working solution.⁶²

Three primer pairs frequently used to amplify invertebrates were tested to improve the amplification efficiency of eDNA metabarcoding and the combination primers were found to perform best.^57,58,67 We further tested the mlCOIintF/HCO2198 primer pair to validate our primer choice; however, it did not yield greater taxonomic recovery compared to LCO1490/ZBJ-ArtR2c (Appendix Figure S1). Consequently, A pair of universal primers (LCO1490: 5′-GGTCAACAAATCATAAAGATATTGG-3’; ZBJ-ArtR2c: 5′-WACTAATCAATTWCCAAATCCTCC-3′) was used to amplify a ∼178 bp fragment of the COI gene, serving as the metabarcode for macroinvertebrate eDNA throughout this study.^57,58 Two sample-specific oligonucleotide DNA tags, 6–8 bases each, were respectively attached to the two PCR primers to facilitate sequence demultiplexing during raw sequencing data processing.³⁷

PCR amplifications for the metabarcode were performed on each eDNA sample in a 20 μL reaction mixture containing 4 μL of DNA template (∼2–40 μg of eDNA extract per reaction), 0.6 μL (10 μM) of each primer, 10 μL of 2 × Rapid Taq Master Mix (Nanjing Vazyme Biotech Co., Ltd.), 3.8 μL of ddH2O, and 1 μL (0.4 μg/μL) of Bovine Serum Albumin (Sigma-Aldrich, Australia). The PCR protocol as follows: an initial denaturation at 94 °C for 3 min; 16 cycles of denaturation at 94 °C for 30 s, annealing at 55 °C for 30 s, and extension at 72 °C for 30 s; followed by 16 cycles of denaturation at 94°C for 30 s, annealing at 58 °C for 30 s, and extension at 72 °C for 30 s; then 24 cycles of denaturation at 94 °C for 30 s, annealing at 60 °C for 30 s, and extension at 72 °C for 30 s; and a final extension at 72 °C for 10 min.

Three PCR replicates were conducted for each eDNA sample; three negative control (in which the DNA template replaced with ddH₂O) and one positive control per PCR run to monitor potential contamination. The PCR amplicons were verified by electrophoresis on a 2 % agarose gel (Biowest Regular Agarose G-10, Spanish), and no contamination was detected in any of the negative controls. A library of the pooled PCR amplicons was prepared using the MGIEasy PCR-Free DNA Library Prep Kit (Guangdong, China) and sequenced on an MGISEQ-2000 platform (BGI-Shenzhen, China) with PE150 sequencing read length and PCR-free library type.

Bioinformatics analysis

Raw sequencing data were processed using the QIIME2 platform.⁵⁹ The processing steps included: (1) demultiplexing the sequence reads; (2) trimming the primers and tags; (3) denoising and dereplicating paired-end sequences, as well as filtering chimeric sequences; (4) generating the feature table and corresponding feature sequences; (5) matching the resulting amplicon sequence variants (ASVs) to the GenBank database (downloaded in January 2024), which at the time contained 766,587 unique species; (6) filtering ASVs with an abundance of fewer than 10 or present in fewer than two samples from the feature table. To avoid false positives or tag jumps, sequences with a maximum identify of less than 98% or with counts lower than those observed in the PCR negative control were removed.³⁷ The step-by-step process and results are included in Notes S1 and S2; Data S2, and rarefaction curves (Figures S4 and S5) indicated that sequencing depth was saturated. To evaluate reference database coverage, all unique taxonomic units identified by morphological approach were compared against the GenBank database used for taxonomic assignment. The proportion of morphologically identified taxa represented in GenBank was calculated to estimate the theoretical detectability of the morphological community by eDNA metabarcoding. A complete presence/absence matrix of all detected taxa is provided in Data S3.

After quality control and bioinformatic filtering, the resulting ASV read counts were used for downstream analyses of community composition and diversity. All statistical analyses were performed using R version 4.1.1.⁶⁰ Venn diagrams were created with the R package VennDiagram to compare the numbers of macroinvertebrate taxa observed in the three types of samples. The Kruskal-Wallis test was conducted using the R package ggpubr to compare the number of species within each EPTD order, across EPTD orders, and the number of other macroinvertebrates detected in the three sampling types. Pairwise comparisons were performed using the function wilcox.test when significant variation was revealed by the Kruskal-Wallis test.

Quantification and statistical analysis

The alpha diversity indices of macroinvertebrates, including species richness, ACE,⁶⁸ Chao1,⁶⁹ Shannon-Wiener⁷⁰ and Simpson⁷¹ index, were calculated using the R package vegan.⁷² Pairwise correlations between macroinvertebrate abundance (log-transformed) detected in the three types of samples were tested using generalized linear models (GLMs) with the R package ggplot2. Scatterplot matrices illustrating the correlations between alpha diversity indices estimated from the three types of samples were created using the function chart.Correlation from the R package Performance Analytics. Beta diversity of macroinvertebrate communities, based on Bray-Curtis distance (Hellinger-transformed), was analyzed using non-metric multidimensional scaling (NMDS) with the R package vegan, and the consistency of community structures detected in the three types of samples was evaluated using Procrustes analysis.^73,74

The independent contributions of eight environmental variables to macroinvertebrate community composition in each of the three sampling types, as well as to overall alpha diversity, were estimated using hierarchical partitioning (HP) with the R package rdacca.hp.⁷⁵ The relationships between alpha diversity metrics from each of the three sampling types and the environmental variables were assessed using redundancy analysis (RDA) with the R package vegan. The statistical significance of each variable’s effect was examined using a permutation test with 999 randomizations. Using a generalized additive model (GAM), NMDS results were overlaid onto the significant environmental variable surfaces to visualize macroinvertebrate alpha diversity distribution patterns across the three sampling types along the gradient of the significant environmental variable.³⁶

Additional resources

Scientific and medical research data

Data associated with the article:

Frontiers | Validation and Development of COI Metabarcoding Primers for Freshwater Macroinvertebrate Bioassessment

Bioinformatic analysis with Freshwater Biology is a freshwater ecology journal publishing innovative papers on the aquatic biology of freshwater environments.