Developing and Evaluating Aquatic Passive Sampling of Environmental DNA for Microbial Community Profiling
Cheng Qian, Gert‐Jan Jeunen, Wu Han, Tsz Ying Chan, Yan Jiang, Weiqi Fu, Mathew Seymour
TL;DR
A new passive sampling method for collecting environmental DNA in water is shown to be more effective than traditional methods for studying microbial diversity.
Contribution
The study introduces and validates a passive sampling method for eDNA collection that outperforms active filtration in microbial biodiversity monitoring.
Findings
Passive sampling for 24 hours with enzymatic extraction yielded significantly more eDNA and higher biodiversity than shorter durations and mechanical extractions.
Passive sampling outperformed active filtration with over 100% higher eDNA yields and 50% higher taxonomic and phylogenetic diversities.
Passive sampling detected more environmental factors and bioindicators compared to active filtration.
Abstract
Environmental DNA (eDNA) metabarcoding has transformed biodiversity monitoring across taxa from bacteria to mammals, yet sample collection remains a major bottleneck. Passive sampling via adsorption and entrapment has emerged as a promising alternative to overcome the limitations of conventional active filtration. However, the performance of passive sampling for microbial biodiversity monitoring remains unknown. Here, we developed passive sampling‐based microbial community profiling by testing five submersion times and three common eDNA extraction methods in mesocosms, and comprehensively evaluated it by comparing results with active filtration in estuarine and coastal environments. We found that passive sampling for 24 h with enzymatic extraction yielded significantly more eDNA and higher biodiversity than shorter durations and mechanical extractions. Passive sampling consistently…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
FIGURE 1
FIGURE 2
FIGURE 3
FIGURE 4
FIGURE 5
FIGURE 6- —Marine Conservation Enhancement Fund
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsEnvironmental DNA in Biodiversity Studies · Microbial Community Ecology and Physiology · Bacteriophages and microbial interactions
Introduction
1
Environmental DNA (eDNA) refers to DNA obtained from environmental samples, comprising both intracellular DNA and extracellular DNA (Taberlet et al. 2012; Pawlowski et al. 2020). By leveraging advanced molecular technologies, such as quantitative PCR (qPCR) and high‐throughput sequencing (HTS), eDNA can be effectively analyzed to characterize microorganisms and macroorganisms present in various environments (Zinger et al. 2012; Deiner et al. 2017). In contrast to traditional biological surveys (e.g., microscopic identification and visual census), which require considerable manpower and morphological expertise, eDNA metabarcoding offers a cost‐efficient approach with higher detection sensitivity and taxonomic resolution (McColl‐Gausden et al. 2021; Seymour et al. 2021), enhancing biodiversity monitoring, environmental assessment, and resource management (Kelly et al. 2014; Thomsen and Willerslev 2015; Valentini et al. 2016; Seymour et al. 2025).
Active filtration, the most used method for aqueous eDNA collection, involves forcing water through a filter membrane to retain eDNA (Tsuji et al. 2019; Wang et al. 2021). However, this method has several notable limitations. Active filtration is a labor‐intensive and time‐consuming process, requiring transport of bulky pumps and power supplies to field sites and filtration of large water volumes to ensure eDNA integrity, which restricts the number of samples available for subsequent analysis (Minamoto et al. 2021). Additionally, filter clogging, particularly in turbid water, hampers eDNA enrichment and introduces PCR inhibitors (Kumar et al. 2022). Moreover, active filtration is a static method, providing only a snapshot of a natural system, meaning that multiple samplings are required to adequately capture spatiotemporal dynamics (Jensen et al. 2022). Furthermore, numerous disposable consumables (e.g., plastics) and cleaning chemicals (e.g., bleach) are required for active filtration in order to limit contamination, posing a potential environmental risk (Bruce et al. 2021). Passive sampling is a promising alternative to active filtration, directly addressing the aforementioned limitations (Kirtane et al. 2020; Bessey et al. 2021). It primarily utilizes artificial materials, such as glass fiber, cellulose, or activated carbon in membrane, sponge, or granular forms, to accumulate eDNA through adsorption and entrapment over a defined period (Bessey et al. 2022; Chen et al. 2022; Jeunen et al. 2022; Maiello et al. 2022; Verdier et al. 2022). Compared with natural eDNA samplers, such as living sponges and bivalves, which obtain eDNA via filter‐feeding (Mariani et al. 2019; Weber et al. 2023), passive sampling with artificial materials offers a non‐invasive approach with greater accessibility and consistency for biomonitoring, eliminating the influence of habitat preference and seasonality associated with filter‐feeder sampling.
Fundamental knowledge gaps persist regarding passive sampling. Previous studies have demonstrated that passive sampling is effective for extracellular DNA collection and metazoan community profiling (Van Der Heyde et al. 2023; Chen et al. 2024; Jeunen et al. 2024; Yan et al. 2024; Zhang et al. 2024). However, its performance for intracellular DNA collection and microbial community profiling remains unknown. Microbes, including prokaryotic bacteria and archaea, as well as microeukaryotic protists and fungi, are the most abundant, diverse, and ubiquitous organisms in nature, playing important roles in food webs and biogeochemical cycles, however these communities are difficult to routinely assess (Azam et al. 1983; Fuhrman 2009). The development of an effective passive sampling for microbial biodiversity is essential for accurate community profiling, which includes consideration for several factors. Passive sampling for a short submersion time may lead to insufficient eDNA yields, whereas a long submersion time could result in increased contaminants affecting downstream analysis (Bessey et al. 2021; Chen et al. 2022). Yields of passively sampled eDNA also vary substantially with different extraction methods, such as enzymatic lysis and mechanical disruption (Kirtane et al. 2020). Hence, determining the appropriate submersion time and extraction method is critical for developing passive sampling. In addition, prior studies focused on biodiversity monitoring and community composition when evaluating passive sampling (Cananzi et al. 2025; Von Ammon et al. 2025), with no exploration of environmental assessment due to limited research objectives and sampling effort. Microbial communities differ significantly across environments due to environmental heterogeneities in several key factors, such as temperature (Xiao et al. 2018), salinity (Rath et al. 2019), dissolved oxygen (Qian et al. 2023), and turbidity (Goosen et al. 1999). Microbes also respond rapidly to environmental changes, making them ideal environmental bioindicators (Aylagas et al. 2017). Thus, comparisons of microbial communities between distinct environments could enable more comprehensive evaluations of the developed passive sampling method.
In this study, we developed passive sampling‐based microbial community profiling by testing various submersion times and extraction methods, and comprehensively evaluated it by comparing results with active filtration in estuarine and coastal environments. We tested three hypotheses: (1) passive sampling yields more eDNA and higher taxonomic and phylogenetic diversity than active filtration, (2) passive sampling is more suitable for spatiotemporal detection of community than active filtration, and (3) passive sampling has greater sensitivity in identifying key environmental factors and potential environmental bioindicators than active filtration.
Materials and Methods
2
Mesocosm Experimental Design
2.1
The mesocosm experiment was designed to develop passive sampling‐based microbial community profiling by testing five different submersion times and three different extraction methods in mesocosms (Figure 1A). Passive samplers (Figure S1) consisted of commercial artificial sponges (Cellulose Speci‐Sponge, B01245, Whirl‐Pak, USA), which are specifically manufactured for collecting environmental microparticles and have been shown to be effective for passive eDNA sampling (Jeunen et al. 2022, 2024). All procedures for constructing the passive samplers were conducted in a laminar flow hood. All materials and tools were sterile or sterilized before use. First, artificial sponges were cut into small pieces (30 mm length × 7 mm width) using scissors. Passive samplers were assembled by securing the sponge pieces to a metal frame with cable ties. Lead weights and ropes were then installed to enable suspension in water columns. Each passive sampler was finally placed under ultraviolet light for 30 min as an additional disinfection step to ensure sterilization. The mesocosm experiment was conducted in six 2042 L tanks (Figure 1). These mesocosms were maintained outdoors and filled with seawater from the adjacent marine environment (Si et al. 2025). Passive samplers were deployed ~0.5 m below the water surface in each mesocosm. A total of 90 passive samples (i.e., artificial sponges) were obtained during the mesocosm experiment. Of these, 18 passive samples were retrieved at the end of each submersion time, including 10 min, 1 h, 6 h, 12 h, and 24 h, with 30 passive samples allocated to each of the three following eDNA extraction methods. Upon removal from the mesocosms, passive samples were gently wrung out using two sterilized tweezers, placed into sterile 2 mL microcentrifuge tubes (Eppendorf, Germany), and immediately stored at −20°C until eDNA extraction.
(A) Design of mesocosm and field experiments. (B) Map of experimental sites. The mesocosm experiment was conducted at the Swire Institute of Marine Science (SWIMS), indicated by the red dot. The field experiment was conducted at three estuarine sites (ES1, ES2, ES3) and three coastal sites (CS1, CS2, CS3), indicated by the yellow and blue dots, respectively. Key environmental factors between the estuarine and coastal environments including salinity, turbidity, and chlorophyll are shown.
Field Experimental Design
2.2
The field experiment was designed to comprehensively evaluate the developed passive sampling method by comparing results with active filtration in estuarine and coastal environments (Figure 1A). We utilized the distinct environments present in Hong Kong waters, including a western estuarine environment directly influenced by freshwater runoff from the Pearl River, and an eastern coastal environment primarily oceanic due to the influence of the South China Sea (Figure 1B; Lai et al. 2016). Environmental differences were determined based on long‐term water quality monitoring conducted monthly by the Hong Kong Environmental Protection Department (EPD). We selected three estuarine sites (ES1, ES2, ES3), characterized by turbid, brackish, and high biomass (Figure 1B). In contrast, three coastal sites (CS1, CS2, CS3) were selected, characterized by clear, saline, and low biomass (Figure 1B). At each site, a calibrated EXO2 multiparameter sonde (YSI, USA) was used to measure temperature (°C), salinity (ppt), turbidity (NTU), pH, dissolved oxygen (mg/L), and chlorophyll (μg/L) (Table S1). Both active filtration and passive sampling were conducted at all field sites. For active filtration, subsurface water (~1 m below the surface) was collected using a Van Dorn water sampler (Wildco, USA), immediately before passive sampling. Triplicate samples were collected at each site, with each sample consisting of 1 L of water filtered through a 0.22 μm enclosed filter (Sterivex‐GP polyethersulfone pressure filter unit, SVGPL10RC, Millipore, USA) using a peristaltic pump (Geotech, USA). In addition, 1 L of autoclaved Milli‐Q water was filtered at each site as a sampling negative control. Upon completion of active filtration, active samples (i.e., enclosed filters) were individually sealed in zip lock bags. For passive sampling, a passive sampler (identical to that used in the mesocosm experiment) equipped with triplicate artificial sponges was deployed ~1 m below the water surface at each site for 24 h. This submersion time was chosen because it outperformed other treatments based on the mesocosm results. In addition, an artificial sponge was immersed in a sterile 1 L bottle (Nalgene, USA) filled with autoclaved Milli‐Q water at each site for 24 h as a sampling negative control. After 24 h, passive samples were retrieved, gently wrung out using two sterilized tweezers, and placed into sterile 2 mL microcentrifuge tubes. In total, 24 active samples and 24 passive samples were collected, immediately placed on ice, and transferred to the laboratory, where they were stored at −20°C until eDNA extraction.
eDNA Extraction and Purification
2.3
All eDNA extractions were performed in a designated clean room under strict contamination control at the eDNA & eEcology Lab, the University of Hong Kong. Prior to extraction, the surfaces of sample containers and all equipment were treated with 70% ethanol and 10% bleach. For passive samples collected during the mesocosm experiment, three common eDNA extraction methods were applied, including the DNeasy PowerWater Kit, the DNeasy PowerSoil Pro Kit, and the DNeasy Blood & Tissue Kit modified by Spens et al. (2017) (Qiagen, Germany). Extraction protocols were followed with slight modifications to enable proper mechanical disruption or enzymatic lysis of passive samples: (1) artificial sponges were cut into ~5 mm^3^ cubes using sterilized scissors before bead‐beating or lysis, and (2) buffer solutions were retrieved using sterile membrane‐free spin columns after bead‐beating or lysis. Active samples collected from the field experiment were extracted using the DNeasy Blood & Tissue Kit modified by Spens et al. (2017), a widely used method for eDNA extraction from filter membranes. Passive samples collected from the field experiment were also extracted using the DNeasy Blood & Tissue Kit modified by Spens et al. (2017), with the modifications used in the mesocosm experiment, as it outperformed other treatments based on the mesocosm results. Finally, eDNA were eluted in 100 μL volumes of TE buffer (Invitrogen, USA). Then, eDNA were purified to remove potential PCR inhibitors using the DNeasy PowerClean Pro Cleanup Kit (Qiagen, Germany), following the purification protocol and stored at −20°C until further processing.
eDNA Quantification
2.4
A series of quantitative measurements were conducted in the mesocosm and field experiments. Total eDNA concentration was quantified in triplicate for each sample using a Qubit 4 fluorometer (Invitrogen, USA) with the dsDNA HS assay kit (Invitrogen, USA), following the manufacturer's instructions. Linear standard curves were constructed for fluorometric measurements across a concentration range of 5 pg/μL to 120 ng/μL. To quantify the prokaryotic and microeukaryotic eDNA concentrations, qPCR was performed in triplicate for each sample using a QuantStudio 7 Pro Real‐Time PCR System (Applied Biosystems, USA) with the 515F/806R primer pair targeting the V4 region of the prokaryotic 16S rRNA gene (Caporaso et al. 2011) and 1380F/1510R primer pair targeting the V9 region of the microeukaryotic 18S rRNA gene (Amaral‐Zettler et al. 2009). Each qPCR plate (Applied Biosystems, USA) included triplicate qPCR standards, that is, six 10‐fold serial dilutions of synthetic gBlocks Gene Fragments (IDT, USA; Table S2) containing either the partial 16S rRNA gene (sequence of Escherichia coli , ranging from 6.25 × 10^3^ to 6.25 × 10^8^ copies/μL) or the partial 18S rRNA gene (sequence of Skeletonema costatum , ranging from 5.03 × 10^3^ to 5.03 × 10^8^ copies/μL). Each qPCR plate also included three qPCR negative controls, with no amplification observed in any of them. Each qPCR was conducted in a 10 μL reaction volume containing 5.0 μL PowerUp SYBR Green Master Mix (Applied Biosystems, USA), 2.0 μL template, 2.0 μL UltraPure Distilled Water (Invitrogen, USA), 0.5 μL forward primer (10 μM; IDT, USA), and 0.5 μL reverse primer (10 μM; IDT, USA). The prokaryotic qPCR conditions consisted of an initial denaturation at 95°C for 5 min, followed by 40 cycles of denaturation at 95°C for 45 s, annealing at 50°C for 60 s, and extension at 72°C for 90 s, with a final extension at 72°C for 10 min. The microeukaryotic qPCR conditions consisted of an initial denaturation at 95°C for 5 min, followed by 40 cycles of denaturation at 95°C for 30 s, annealing at 57°C for 60 s, and extension at 72°C for 90 s, with a final extension at 72°C for 10 min. Initial copy numbers in each sample were determined based on the standard curves, using the linear relationships between the Ct values obtained from the qPCR standards and the corresponding logarithmic values of the copy numbers.
Library Preparation and Sequencing
2.5
Partial mesocosm samples and all field samples were sequenced. Mesocosm samples used for metabarcoding were restricted to those extracted with the Blood & Tissue method because they yielded the highest eDNA concentrations and were directly related to the extraction method used for the field experiment. Prokaryotic and microeukaryotic dual indexed libraries were constructed using a two‐step PCR protocol (Seymour et al. 2020) with the primer pairs of 515F/806R (Caporaso et al. 2011) and 1380F/1510R (Amaral‐Zettler et al. 2009). A Biomek i5 Automated Workstation (Beckman Coulter, USA) was used during library preparation. The first‐round amplifications were performed in triplicate in 25 μL reaction volumes containing 12.5 μL AmpliTaq Gold 360 Master Mix (Applied Biosystems, USA), 10.5 μL UltraPure Distilled Water, 1.0 μL template, 0.5 μL forward primer with Illumina overhang adapter (10 μM; IDT, USA), and 0.5 μL reverse primer with Illumina overhang adapter (10 μM; IDT, USA). The first‐round PCR conditions followed those used in qPCR, except that the number of cycles was reduced from 40 to 30. The pooled first‐round PCR products were purified using AMPure XP beads (Beckman Coulter, USA). The second‐round amplifications were performed in 25 μL reaction volumes containing 12.5 μL KAPA HiFi HotStart Ready Mix (Roche, Switzerland), 5.0 μL Illumina DNA UD Indexes (Illumina, USA), 5.0 μL purified first‐round PCR products, and 2.5 μL UltraPure Distilled Water. The second‐round PCR conditions consisted of an initial denaturation at 95°C for 3 min, followed by 8 cycles of denaturation at 98°C for 30 s, annealing at 55°C for 30 s, and extension at 72°C for 30 s, with a final extension at 72°C for 5 min. The second‐round PCR products were purified using AMPure XP beads and quantified using a Varioskan LUX Microplate Reader (Thermo Fisher Scientific, USA) with the Quant‐iT dsDNA HS Assay Kit (Invitrogen, USA). Subsequently, all samples were pooled in equimolar quantities and fragment sizes of pooled libraries were verified using a Fragment Analyzer (Agilent, USA). In addition, PCR negative controls were included in each library to check potential contamination. The prokaryotic and microeukaryotic libraries were sequenced separately on two dedicated flow cells in a MiSeq platform (Illumina, USA) with 2 × 300 bp paired‐end read configuration at the Centre for PanorOmic Sciences, the University of Hong Kong.
Bioinformatic Processing
2.6
Raw sequence reads generated by HTS were imported into QIIME 2 v2025.7 (Bolyen et al. 2019) after demultiplexing and adapter removal using the “qiime tools import” function. Sequences were inspected using the “qiime demux summarize” function and denoised using the “qiime dada2 denoise‐paired” function (Callahan et al. 2016). Representative amplicon sequence variants (ASV) and count tables were then generated using the “qiime feature‐table tabulate‐seqs” and “qiime feature‐table summarize” functions, respectively. Rooted and unrooted phylogenetic trees were built with the representative ASVs using the “qiime phylogeny align‐to‐tree‐mafft‐fasttree” function. Reference sequences of 16S and 18S rRNA genes and their corresponding taxonomy were extracted from Greengenes2 v2024.09 (McDonald et al. 2024) and PR^2^ v5.0.0 (Guillou et al. 2013), respectively, according to the primer pairs used in this study, using the “qiime feature‐classifier extract‐reads” function. Naive Bayes classifiers were subsequently trained based on these extracted information using the “qiime feature‐classifier fit‐classifier‐naive‐bayes” function. Finally, representative ASVs were taxonomically assigned by the trained classifiers using the “qiime feature‐classifier classify‐sklearn” function. Bioinformatic processing was performed in the HPC2021 System provided by the Information Technology Services, the University of Hong Kong.
Statistical Analysis
2.7
Unassigned ASVs were removed from both prokaryotic and microeukaryotic communities. ASVs assigned as mitochondria and chloroplasts were removed from the prokaryotic communities, and ASVs assigned as metazoa and plants were removed from the microeukaryotic communities. Observed ASV counts in the sampling and PCR negative controls were subtracted from the corresponding samples to account for potential contaminations. Rarefaction curves were visualised (Figure S2), and ASV counts were rarefied to the minimum sequencing depths using the “vegan::rrarefy” function (Oksanen et al. 2025). Shapiro–Wilk test was performed to check data normality using the “stats::shapiro.test” function (R Core Team 2025). Levene's test was performed to check homogeneity of variance using the “car::leveneTest” function (Fox et al. 2024). Number of ASVs was calculated to estimate taxonomic diversity using the “vegan::specnumber” function. Faith's phylogenetic diversity (Faith's PD) was calculated to estimate phylogenetic diversity using the “picante::pd” function (Kembel et al. 2020). Analysis of variance (ANOVA) and post hoc Tukey's honest significant difference test (Tukey's HSD test) were implemented to examine the difference in eDNA yield and biological diversity among submersion times and extraction methods using the “stats::aov” and “stats::TukeyHSD” functions. Generalised additive model (GAM) was constructed using smooth function (thin plate regression splines) to fit patterns between submersion time and eDNA yield/biological diversity using the “mgcv::gam” function (Wood 2025). Paired t‐test was implemented to examine the difference in eDNA yield and biological diversity between active filtration and passive sampling using the “stats::t.test” function. Community dissimilarity between samples and partitioned components of turnover and nestedness were computed based on Jaccard distance using the “betapart::beta.pair” function (Baselga and Orme 2012). Permutational multivariate analysis of dispersion (PERMDISP) was performed to check homogeneity of dispersion using the “vegan::betadisper” function. Permutational multivariate analysis of variance (PERMANOVA) was applied to test the significance of community dissimilarity between active filtration and passive sampling using the “vegan::adonis2” function. Nonmetric multidimensional scaling (NMDS) was performed for ordination using the “vegan::metaMDS” function. Similarity percentages analysis (SIMPER analysis) was applied to calculate the contributions of taxa to community dissimilarity using the “vegan::simper” function. Environmental dissimilarity between sites was computed based on Euclidean distance using the “vegan::vegdist” function. Partial Mantel test was conducted to assess the correlation between community dissimilarity and environmental dissimilarity using the “vegan::mantel.partial” function. Linear discriminant analysis (LDA) effect size (LEfSe) with a threshold of 2 in LDA score was used to identify differential species between estuarine and coastal environments, using the “lefser::lefser” function (Khleborodova et al. 2024). Data manipulation and graphical visualisation were performed in R v4.5.1 (R Core Team 2025) using the “tidyverse” package (Wickham et al. 2019). Statistical significance was set at 0.05.
Results
3
eDNA Concentration, Number of ASVs, and Faith's PD Across Submersion Times and Extraction Methods in Passive Sampling
3.1
In the mesocosm experiment, mean total, prokaryotic, and microeukaryotic eDNA concentrations yielded by passive sampling varied with submersion time and extraction method, ranging from 0.15 to 21.24 ng/μL, 0.07 to 2.97 million copies/μL, and 0.05 to 2.10 million copies/μL, respectively (Figure 2). The 24‐h submersion time with Blood & Tissue extraction method yielded on average 147‐fold, 47‐fold and 41‐fold higher total, prokaryotic, and microeukaryotic eDNA concentrations, respectively, compared with the 10‐min submersion time with PowerWater extraction method (Figure 2). Three eDNA concentrations were significantly affected by submersion time, extraction method, and their interaction (two‐way ANOVA, all p < 0.001; Figures 2 and S3). The 24‐h submersion time, the Blood & Tissue extraction method, and their combination yielded significantly higher eDNA concentrations compared with other individual or combined treatments (Tukey's HSD tests, all p < 0.05; Figures 2 and S3). Total, prokaryotic, and microeukaryotic eDNA concentrations increased continuously with submersion time (GAMs, all p < 0.001; Figure 3A). In addition, mean number of prokaryotic and microeukaryotic ASVs and prokaryotic and microeukaryotic Faith's PD yielded by passive sampling varied with submersion time, ranging from 825 to 1774 ASVs, 150 to 313 ASVs, 72.84 to 117.49, and 39.76 to 76.90, respectively (Figure S4). The 24‐h submersion time yielded on average 2.4‐fold, 2.3‐fold, 1.7‐fold, and 2.0‐fold higher number of prokaryotic and microeukaryotic ASVs and prokaryotic and microeukaryotic Faith's PD, respectively, compared with the 10‐min submersion time (Figure S4). Number of ASVs and Faith's PD were significantly affected by submersion time (one‐way ANOVA, all p < 0.001; Figure S4). The 24‐h submersion time yielded highest number of ASVs and Faith's PD, whereas the former did not differ significantly among 6‐h, 12‐h, and 24‐h submersion times (Tukey's HSD tests, all p > 0.05; Figure S4A,B) and the latter did not differ significantly between 12‐h and 24‐h submersion times (Tukey's HSD tests, all p > 0.05; Figure S4C,D). Number of prokaryotic and microeukaryotic ASVs and prokaryotic and microeukaryotic Faith's PD saturated gradually with submersion time (GAMs, all p < 0.001; Figure 3B,C).
(A) Total, (B) prokaryotic, and (C) microeukaryotic eDNA concentrations across five submersion times (10 min, 1 h, 6 h, 12 h, 24 h) and three extraction methods (PowerWater, PowerSoil, Blood & Tissue) in the mesocosm experiment. Bars represent the mean values across the six replicates. Error bars represent the standard errors of the mean. Statistical significances of two‐way ANOVA are shown and lowercase letters represent the statistical significances between treatments assessed by post hoc Tukey's HSD tests.
(A) Total, prokaryotic, and microeukaryotic eDNA concentrations, (B) number of prokaryotic and microeukaryotic ASVs, and (C) prokaryotic and microeukaryotic Faith's PD across five submersion times (10 min, 1 h, 6 h, 12 h, 24 h) in the mesocosm experiment. Dots represent the values of the replicates. Smooth lines were fitted by GAMs, with shaded areas indicating the 95% confidence intervals. Statistical significances and R2 of GAMs are shown.
eDNA Concentration, Number of ASVs, and Faith's PD Between Active Filtration and Passive Sampling
3.2
In the field experiment, passive sampling yielded significantly higher eDNA concentrations than active filtration at every estuarine and coastal site, with an average increase of 162% in total eDNA concentration, 296% in prokaryotic eDNA concentration, and 114% in microeukaryotic eDNA concentration (paired t‐tests, all p < 0.05; Figure 4A–C). Mean total eDNA concentrations yielded by active filtration ranged from 6.65 to 22.44 ng/μL, compared with 10.74 to 40.54 ng/μL for passive sampling (Figure 4A). Mean prokaryotic eDNA concentrations yielded by active filtration ranged from 3.15 to 15.65 million copies/μL, compared with 4.46 to 70.02 million copies/μL for passive sampling (Figure 4B). Mean microeukaryotic eDNA concentrations yielded by active filtration ranged from 0.23 to 3.51 million copies/μL, compared with 0.46 to 4.75 million copies/μL for passive sampling (Figure 4C). In addition, passive sampling yielded significantly higher number of ASVs and Faith's PD at every estuarine and coastal site, with an average increase of 92% in number of prokaryotic ASVs, 57% in number of microeukaryotic ASVs, 51% in prokaryotic Faith's PD, and 53% in microeukaryotic Faith's PD (paired t‐tests, all p < 0.05; Figure 4D–G). Mean number of prokaryotic ASVs yielded by active filtration ranged from 459 to 1132 ASVs, compared with 786 to 2231 ASVs for passive sampling (Figure 4D). Mean number of microeukaryotic ASVs yielded by active filtration ranged from 387 to 1238 ASVs, compared with 521 to 1858 ASVs for passive sampling (Figure 4E). Mean prokaryotic Faith’ PD yielded by active filtration ranged from 49.03 to 101.20, compared with 69.68 to 139.52 for passive sampling (Figure 4F). Mean microeukaryotic Faith’ PD yielded by active filtration ranged from 34.58 to 84.58, compared with 52.31 to 147.36 for passive sampling (Figure 4G).
*(A) Total, (B) prokaryotic, (C) microeukaryotic eDNA concentrations, number of (D) prokaryotic and (E) microeukaryotic ASVs, and (F) prokaryotic and (G) microeukaryotic Faith's PD between active filtration and passive sampling in the field experiment. Squares (estuarine sites: ES1, ES2, ES3) and circles (coastal sites: CS1, CS2, CS3) represent the mean values across each of the triplicates per site. Error bars represent the standard errors of the mean. Grey lines connect the samples taken from the same sites. Asterisks represent significant differences between active filtration and passive sampling assessed by paired t‐tests, , p < 0.05.
Community Dissimilarity, Turnover, and Nestedness Between Active Filtration and Passive Sampling
3.3
In the field experiment, prokaryotic communities yielded by active filtration and passive sampling were significantly different (PERMANOVA: estuarine environment, R^2^ = 0.67, F = 33.16, p < 0.001; coastal environment, R^2^ = 0.52, F = 17.63, p < 0.001; Figures 5A and S5A,B). Prokaryotic community dissimilarities between active filtration and passive sampling across field sites ranged from 0.7995 to 0.9029 (mean 0.8617; Figure 5B). Partitioned components of turnover and nestedness across field sites ranged from 0.5900 to 0.7618 (mean 0.6821) and from 0.1005 to 0.2660 (mean 0.1796), respectively (Figure 5C,D). Microeukaryotic communities yielded by active filtration and passive sampling were significantly different (PERMANOVA: estuarine environment, R^2^ = 0.38, F = 9.61, p < 0.001; coastal environment, R^2^ = 0.41, F = 11.11, p < 0.001; Figures 5E and S6A–B). Microeukaryotic community dissimilarities between active filtration and passive sampling across field sites ranged from 0.7233 to 0.8521 (mean 0.7754; Figure 5F). Partitioned components of turnover and nestedness across field sites ranged from 0.4497 to 0.6990 (mean 0.6041) and from 0.1082 to 0.3185 (mean 0.1713), respectively (Figure 5G,H). In addition, top 5 taxa contributing to prokaryotic community dissimilarities between active filtration and passive sampling were: phylum level, Pseudomonadota (30.1%), Cyanobacteriota (21.4%), Bacteroidota (18.2%), Actinomycetota (6.9%), Thermoplasmatota (6.9%); class level, Gammaproteobacteria (35.7%), Alphaproteobacteria (14.2%), Bacteroidia (13.3%), Cyanobacteriia (12.4%), Poseidoniia (5.1%); order level, Pseudomonadales (29.3%), Cytophagales (12.4%), Cyanobacteriales (5.2%), Flavobacteriales (5.1%), PCC‐6307 (4.3%); family level, Cellvibrionaceae (28.7%), Cyclobacteriaceae (11.5%), Coleofasciculaceae (4.9%), Cyanobiaceae (4.0%), Poseidoniaceae (3.2%) (SIMPER analyses; Figure S5C–F). Top 5 taxa contributing to microeukaryotic community dissimilarities between active filtration and passive sampling were: subdivision level, Dinoflagellata (24.2%), Gyrista (21.4%), Ciliophora (19.6%), Bigyra (10.8%), Chlorophyta (5.8%); class level, Mediophyceae (13.5%), Bacillariophyceae (13.2%), Dinophyceae (12.5%), Phyllopharyngea (10.5%), Spirotrichea (9.1%); order level, Nanomonadea (12.5%), Thalassiosirales (11.9%), Suctoria (5.6%), Gymnodiniales (5.5%), Oligotrichida (4.1%); family level, Thalassiosiraceae (7.6%), Acinetidae (7.3%), MAST‐3 (5.6%), Bacillariaceae (3.8%), Stephanodiscaceae (3.8%) (SIMPER analyses; Figure S6C–F).
Ordinations of (A) prokaryotic and (E) microeukaryotic communities yielded by active filtration and passive sampling in the field experiment. Dots represent the triplicate communities at each site and ellipses represent the 95% confidence intervals. Ordinations were completed by NMDS. Community dissimilarities between the two collection methods were assessed by PERMANOVA. Statistical significances of PERMANOVA are shown. Prokaryotic (B) community dissimilarities between active filtration and passive sampling and partitioned components of (C) turnover and (D) nestedness. Microeukaryotic (F) community dissimilarities between active filtration and passive sampling and partitioned components of (G) turnover and (H) nestedness. Solid lines represent the medians and dotted lines represent the means.
Environmental Correlation and Differential Species Between Active Filtration and Passive Sampling
3.4
In the field experiment, prokaryotic and microeukaryotic community dissimilarities between estuarine and coastal environments yielded by active filtration were significantly correlated with environmental dissimilarities in salinity and turbidity (partial Mantel tests, both p < 0.001, both r > 0.3; Figure 6A). Prokaryotic and microeukaryotic community dissimilarities between estuarine and coastal environments yielded by passive sampling similarly showed significant correlations with salinity and turbidity, and were also significantly correlated with chlorophyll (partial Mantel tests, all p < 0.001, all r > 0.3; Figure 6D). In addition, 12 differential prokaryotic species between estuarine and coastal environments were identified in active filtration‐yielded communities (LEfSe, all LDA score > 2; Figure 6B), whereas 28 such species were identified in passive sampling‐yielded communities (LEfSe, all LDA score > 2; Figure 6E), with 4 overlapping species: Marisediminitalea mangrovi, UBA3478 sp., GCA‐2686945 sp., and Caldora sp. (Figure 6B,D). Similarly, 8 differential microeukaryotic species between estuarine and coastal environments were identified in active filtration‐yielded communities (LEfSe, all LDA score > 2; Figure 6C), whereas 12 such species were identified in passive sampling‐yielded communities (LEfSe, all LDA score > 2; Figure 6F), with 4 overlapping species: Skeletonema sp., MAST‐12A sp., Chaetoceros tenuissimus and Tripos fusus (Figure 6C,F).
Correlations between community dissimilarities and environmental dissimilarities in (A) active filtration and (D) passive sampling in the field experiment. Line colors represent Mantel's p (statistical significances) and line widths represent Mantel's r (correlation coefficients) assessed by partial Mantel tests. TEM, temperature; SAL, salinity; TUR, turbidity; pH, pH; DO, dissolved oxygen; CHL, chlorophyll. Differential prokaryotic species between estuarine and coastal environments yielded by (B) active filtration and (E) passive sampling. Differential microeukaryotic species between estuarine and coastal environments yielded by (C) active filtration and (F) passive sampling. LDA scores represent size of differentiation assessed by LEfSe with a threshold of 2.
Discussion
4
Influences of Submission Time and Extraction Method
4.1
Our results show that 24 h was optimal for eDNA collection among the five submersion times tested (Figures 2 and S3A–C), with concentrations following an increasing pattern over time (Figure 3A). Notably, 24 h was the maximum duration in this study, indicating that a longer submersion time might further enhance eDNA collection. This finding aligns with two previous studies that reported continued eDNA increases within 4 h (Kirtane et al. 2020) or 72 h (Chen et al. 2022). We speculate that the similar eDNA concentrations observed between short and long submersion times in other previous studies in an aquarium (Bessey et al. 2022) and at a groundwater site (Van Der Heyde et al. 2023) may be due to the extensive use of detergents in the aquarium or the relatively complex environmental conditions in groundwater that largely degraded eDNA (Barnes et al. 2014). The results show that 24 h was also optimal for species detection among the five submersion times tested (Figure S4), with taxonomic and phylogenetic diversity following a saturating pattern over time (Figure 3B,C). In contrast to prior studies reporting that as little as 4 h (Zhang et al. 2024) or 8 h (Chen et al. 2022) is sufficient to profile fish communities, our finding denotes that a longer submersion time is needed to profile microbial communities. This difference could be attributed to the much higher biodiversity of microbes (Locey and Lennon 2016) than fish (Su et al. 2021) in natural systems. Considering both aspects, although eDNA yield may continue to increase beyond 24 h, biological diversity saturates within 24 h, indicating that species detection should be prioritised over eDNA collection to avoid increased contaminants or sample loss originating from prolonged submersion time. These findings suggest that 24 h is an appropriate submersion time for passive sampling‐based microbial community profiling.
Extraction methods can significantly affect the quantity and quality of eDNA, impacting downstream molecular analyses and ecological interpretations (Deiner et al. 2018). Enzymatic extraction utilizes enzymes to digest proteins and cellular structures for DNA isolation (Salazar and Asenjo 2007), whereas mechanical extraction relies on bead‐beating to break cells and release DNA (De Boer et al. 2010). We found that an enzymatic extraction method (i.e., Blood & Tissue) was optimal among the commonly used eDNA extraction methods tested here, outperforming other two mechanical extraction methods (i.e., PowerSoil and PowerWater) (Figures 2 and S3D–F). The superior performance of enzymatic extraction over mechanical extraction likely arises from following two aspects. First, unlike bead‐beating, which primarily acts on eDNA adsorbed on the surface of passive samplers, enzymatic solution can effectively act on both exterior‐adsorbed and interior‐entrapped eDNA through external washing and internal penetration, thereby potentially increasing eDNA yields. Second, the applied Blood & Tissue enzymatic extraction method involves an extended reaction time of 24 h, as outlined in Spens et al. (2017), compared with the shorter stated reaction times used in the mechanical extraction methods (5 min for PowerWater and 10 min for PowerSoil), likely contributing to improved eDNA yields. These findings suggest that enzymatic extraction is an appropriate extraction method for recovering passively sampled eDNA.
Performance of Passive Sampling in Biodiversity Monitoring
4.2
Our results show that passive sampling using a small artificial sponge (30 × 7 mm) with a 24‐h submersion time outperformed conventional active filtration of 1 L water (0.22 μm pore size) in eDNA collection (Figure 4A–C) and species detection (Figure 4D–G). The superior performance of passive sampling over active filtration is likely due to the high thickness and porosity of the material used in this study (ca. 15 mm thickness; Figure S1), which provides a vast internal surface area and numerous cavities for eDNA adsorption and entrapment. This result contrasts with studies using filter membranes as passive sampling materials, which reported similar or lower eDNA yield and biological diversity compared with active filtration (Bessey et al. 2022; Van Der Heyde et al. 2023; Chen et al. 2024). Notably, a previous study (Chen et al. 2022) found that glass fiber membranes (0.38 mm thickness) collected more eDNA than thinner membranes (0.007–0.17 mm thickness), supporting the role of material thickness in enhancing eDNA collection. These findings suggest using thick and porous materials for passive sampling. Additionally, despite substantial differences between prokaryotic and microeukaryotic traits (Keeling and Campo 2017) and between estuarine and coastal environments (Figure 1B), passive sampling consistently outperformed active filtration regardless of biological characteristics and physicochemical properties (Figure 4). These findings suggest that passive sampling is an effective method for biodiversity monitoring across taxa and environments.
Performance of Passive Sampling in Spatiotemporal Detection
4.3
We found that active filtration and passive sampling yielded significantly different prokaryotic and microeukaryotic community compositions (Figure 5A,B,E,F), driven primarily by turnover rather than nestedness (Figure 5C,D,G,H), indicating differential sampling strategies of the two methods. Passive sampling, deployed for 24 h in this study and influenced by tides and currents, continuously collects eDNA from day to night and from deep or remote sources, enabling the capture of spatiotemporal dynamics in microbial communities (Kelly et al. 2019; Louca 2022; Xiong et al. 2025). In contrast, active filtration, an instantaneous method, reflects the community state only at a specific sampling moment and site, thereby limiting detection of spatial and temporal variations. This is supported by the observed substantial contributions of photosynthetic and planktonic taxa to community differences (e.g., Cyanobacteriota, Mediophyceae and Bacillariophyceae; Figures S5C–F and S6C–F). Moreover, unlike active filtration, which can thoroughly collect and reveal nearly all microbes present in a given water sample, passive sampling could be selective in capturing microbes with motility. This is evident by the observed important differential taxa between the two methods that are generally flagellated or ciliated (e.g., Dinoflagellata, Ciliophora and Nanomonadea; Figure S6C–F). Furthermore, prolonged exposure of passive sampling material could be a concern, as it may serve as a substrate for biofilm formation (Figures S5C–F and S6C–F). These findings suggest that passive sampling is suitable for spatiotemporal detection of community but may introduce biases for motile microbes, necessitating further methodological optimization.
Notably, prokaryotic and microeukaryotic community similarities among three biological replicates did not differ significantly between active filtration and passive sampling (paired t‐test, both p > 0.05; Figure S7). This finding denotes that, despite continuous 24‐h exposure, triplicate community compositions yielded by passive sampling were highly consistent, with variability among replicates similar to that observed across active filtration replicates.
Performance of Passive Sampling in Environmental Assessment
4.4
Three key environmental factors: salinity, turbidity, and chlorophyll distinguish the estuarine and coastal environments examined here, as confirmed by long‐term water quality monitoring from the Hong Kong EPD and our in‐situ measurements (Figure 1B). Previous studies have shown that microbial communities in these two environments differ significantly due to environmental heterogeneity (Liu et al. 2019; Lee et al. 2024). We found that active samples were not sufficient for differentiating environments related to chlorophyll (Figure 6A), whereas passive samples showed significant community differences across all three key environmental factors (Figure 6D). Unlike salinity and turbidity, which remain seasonally stable, chlorophyll is associated with phytoplankton communities that show dramatic diurnal and diel variations driven by light intensity and availability (Prézelin 1992; Flöder et al. 2002). Passive sampling, with its 24‐h deployment in this study, likely captured temporal dynamics of phytoplankton, thereby further identifying the key environmental factor of chlorophyll. Additionally, more potential environmental bioindicators were identified by passive sampling than active filtration (Figure 6B,C,E,F), probably due to its capability to more thoroughly profile microbial communities through spatiotemporal detection. These findings support that passive sampling can effectively identify key environmental factors and potential environmental bioindicators, highlighting its enhanced sensitivity for environmental assessment.
Conclusion
5
Here we developed and comprehensively evaluated passive sampling‐based microbial community profiling. Specifically, we demonstrate that 24‐h submersion time with enzymatic extraction is optimal for eDNA collection and species detection. We also highlight that passive sampling outperforms active filtration in biodiversity monitoring across taxa and environments. Passive sampling shows enhanced capability for spatiotemporal detection of communities from individual sampling events, compared with active filtration. Additionally, passive sampling identifies more key environmental factors and potential environmental bioindicators than active filtration, underscoring its greater sensitivity for environmental assessment. Effectively implementing passive sampling provides a robust, comprehensive and reliable method for microbial biodiversity monitoring and environmental assessment in aquatic environments.
Author Contributions
Cheng Qian, Gert‐Jan Jeunen, and Mathew Seymour designed research. Cheng Qian, Tsz Ying Chan, Yan Jiang, and Mathew Seymour performed research. Cheng Qian and Wu Han analyzed data. Cheng Qian, Gert‐Jan Jeunen, Wu Han, Tsz Ying Chan, Yan Jiang, Weiqi Fu, and Mathew Seymour wrote and revised the paper.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Table S1: Measurements of environmental factors. Table S2: Information of gBlocks Gene Fragments. Figure S1: Images of the passive sampler. Figure S2: Rarefaction curves. Figure S3: eDNA concentrations across submersion times and extraction methods. Figure S4: Number of ASVs and Faith's PD across submersion times. Figure S5: Prokaryotic community compositions. Figure S6: Microeukaryotic community compositions. Figure S7: Community similarities among biological replicates.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Amaral‐Zettler, L. A. , E. A. Mc Climent , H. W. Ducklow , and S. M. Huse . 2009. “A Method for Studying Protistan Diversity Using Massively Parallel Sequencing of V 9 Hypervariable Regions of Small‐Subunit Ribosomal RNA Genes.” P Lo S One 4, no. 7: e 6372. 10.1371/journal.pone.0006372.19633714 PMC 2711349 · doi ↗ · pubmed ↗
- 2Aylagas, E. , Á. Borja , M. Tangherlini , et al. 2017. “A Bacterial Community‐Based Index to Assess the Ecological Status of Estuarine and Coastal Environments.” Marine Pollution Bulletin 114, no. 2: 679–688. 10.1016/j.marpolbul.2016.10.050.27784536 · doi ↗ · pubmed ↗
- 3Azam, F. , T. Fenchel , J. Field , J. S. Gray , L. A. Meyer‐Reil , and F. Thingstad . 1983. “The Ecological Role of Water‐Column Microbes in the Sea.” Marine Ecology Progress Series 10: 257–263. 10.3354/meps 010257. · doi ↗
- 4Barnes, M. A. , C. R. Turner , C. L. Jerde , M. A. Renshaw , W. L. Chadderton , and D. M. Lodge . 2014. “Environmental Conditions Influence e DNA Persistence in Aquatic Systems.” Environmental Science & Technology 48, no. 3: 1819–1827. 10.1021/es 404734 p.24422450 · doi ↗ · pubmed ↗
- 5Baselga, A. , and C. D. L. Orme . 2012. “Betapart: An R Package for the Study of Beta Diversity.” Methods in Ecology and Evolution 3, no. 5: 808–812. 10.1111/j.2041-210X.2012.00224.x. · doi ↗
- 6Bessey, C. , Y. Gao , Y. B. Truong , H. Miller , S. N. Jarman , and O. Berry . 2022. “Comparison of Materials for Rapid Passive Collection of Environmental DNA.” Molecular Ecology Resources 22, no. 7: 2559–2572. 10.1111/1755-0998.13640.35570323 PMC 9544503 · doi ↗ · pubmed ↗
- 7Bessey, C. , S. N. Jarman , T. Simpson , et al. 2021. “Passive e DNA Collection Enhances Aquatic Biodiversity Analysis.” Communications Biology 4, no. 1: 236. 10.1038/s 42003-021-01760-8.33619330 PMC 7900116 · doi ↗ · pubmed ↗
- 8Bolyen, E. , J. R. Rideout , M. R. Dillon , et al. 2019. “Reproducible, Interactive, Scalable and Extensible Microbiome Data Science Using QIIME 2.” Nature Biotechnology 37, no. 8: 852–857. 10.1038/s 41587-019-0209-9.PMC 701518031341288 · doi ↗ · pubmed ↗