2.1 qPCR Primer-Probe Development

We designed a species-specific hydrolysis probe-based qPCR in the mtDNA Female-type (F-type) COI region (i.e., the maternally inherited mtDNA of the Unionidae and several other bivalve groups that exhibit a doubly uniparental mode of mitochondrial inheritance; Breton et al. 2009) of R. sumatrensis. We initially used Primer-BLAST (http://www.ncbi.nlm.nih.gov/tools/primer-blast/) (Ye et al. 2012) to identify species-specific primer-binding sites 80–250 bp apart in the COI mtDNA sequence of the target species using all R. sumatrensis F-type COI sequences available on GenBank, resulting in a set of 20 sequences from 13 sites and five river basins, and covering intraspecific variation across the species range (Table 1). Sequences were aligned by ClustalW in MEGA11 (Tamura et al. 2021) to generate a consensus sequence. The developed primers Rec_sumF and Rec_sumR (Table 2) amplify a 204 bp fragment within the COI region of R. sumatrensis, situated at the 3′ end of COI, and overlap with the section amplified by Leray et al.'s (2013) DNA metabarcoding primers. While Primer-BLAST is integrated into BLAST (Basic Local Alignment Search Tool Tool; GenBank, www.ncbi.nlm.nih.gov/blast/) and finds species-specific primers based on available sequences on GenBank, potential affinity of the primers with non-target species was additionally checked by comparing developed primers with all available sequence data with BLAST, and via manual comparison to related mussel species added into the master R. sumatrensis alignment.

Using the alignment generated above as a starting point, a TaqMan probe was designed in Primer3Plus (Untergasser et al. 2012). The probe design was constrained to the 5' region of the species-specific amplicon, where R. sumatrensis intraspecific variation was low and mismatches were present with respect to other mussel taxa potentially present in the study region. The developed primers Rec_sumF_475 and Rec_sumR_545 amplify a 90 bp fragment of the COI region of R. sumatrensis (Table 2).

2.2 Assay Optimisation and Validation

The qPCR assay was optimized and validated in vitro using DNA extracts of nine sumatrensis tissue samples collected from 6 sites and four river basins, including the same site where specimens for the eDNA experiments were collected (i.e., Sungai Sekerubong, Sarawak, Malaysia, see details below) (Table 1). The specificity of the assay was additionally tested against four other freshwater mussel species from the study region (i.e., Sarawak), representing all unionid genera reported from Sarawak (Zieritz et al. 2024) (Table 3). Genomic DNA for this purpose was extracted using the Qiagen DNeasy Blood & Tissue Kit following the manufacturer's protocol. Total DNA concentration was measured using a Qubit Fluorometer (Thermo Fisher Scientific) with a dsDNA HS assay.

Optimal annealing temperature for the qPCR primer pair (Table 2) was determined using a gradient PCR on nine 1:10 diluted genomic R. sumatrensis DNA samples (Table 1) in duplicate. Reactions consisted of 2x GoTaq PCR Mastermix (Promega), 0.1 mM of each Rec_sumF_475 and Rec_sumR_545 primers (Table 2), 2 μL template DNA, and nuclease-free water to 20 μL. Thermal cycling conditions were: initial denaturation of 5 min at 95°C, followed by 35 cycles of 45 s at 95°C, 45 s at annealing temperature ranging from 55°C to 66°C in 1°C intervals, and 45 s at 72°C, and a final extension of 10 min at 72°C. Amplicons were run on a 2% gel electrophoresis (25 min at 100 V) and reviewed visually.

Sensitivity of the qPCR assay was tested on a 10-fold dilution series of R. sumatrensis genomic DNA (Table 1; 6.38 ng/μL to 6.38 ng/μL × 10⁻⁹), run with 2 replicates for dilutions 6.38 ng/μL to 6.38 ng/μL × 10⁻², 22 replicates for dilutions 6.38 ng/μL × 10⁻³ to 6.38 ng/μL × 10⁻⁹, and four negative PCR controls (nuclease-free water).

Specificity of the qPCR assay was tested using the four non-target freshwater mussel species (Table 3). qPCRs were run with a 10-fold dilution series of each of these species and R. sumatrensis (all 2.5 ng/μL to 2.5 ng/μL × 10⁻²) run with 2 replicates, 2 positive eDNA controls, and 4 negative PCR controls (nuclease-free water).

All subsequent qPCR reactions were setup in 96-well MicroAmp Fast Optical Plates (Thermo Fisher Scientific Inc., Waltham, MA) with reactions containing 2x GoTaq Probe qPCR Master Mix (for qPCR primer-probe; containing GoTaq Hot Start Polymerase, MgCl₂, dNTPs and a proprietary reaction buffer and 2 μL/mL CXR Reference Dye, Promega), 0.5 mM of each primer, 0.25 mM of the Rec_sumP_495 probe (Table 2), 2 μL template DNA, and nuclease-free water to a total reaction volume of 10 μL. Thermal cycling conditions were: initial denaturation of 2 min at 95°C, followed by 55 cycles of 3 s at 95°C and 30 s at 60°C. Reactions were analyzed on a 7500 Fast Real-Time PCR System version 2.3 (Applied Biosystems).

2.3 Experimental Setup

Tank experiments to test the performance of a range of protocols to capture, preserve, and extract eDNA were conducted on 21 and April 22, 2022 at the Aquatic Biology laboratory at the Faculty of Resource Science and Technology, Universiti of Malaysia Sarawak, Kota Samarahan, Borneo, at 25°C–27°C room temperature. To test the performance of eDNA protocols under different environmental conditions, on 17/04/2022, mussel eDNA-free water was collected using a pump and 25 L-containers from two sites with differing land use and water chemistry, i.e., the man-made Prau Lake situated within the Universiti Malaysia Sarawak campus (1.4602 N, 110.4534 E) and Sungai (River) Beri Gru'Ong situated within a forest reserve (1.3512 N, 110.3864 E). Sites were known to be devoid of freshwater mussels, including upstream sections of the river. In the laboratory, four 80 L-capacity glass tanks each were set up with 40 L of lake and river water, respectively (Figure 1A). All equipment, including tanks, containers, and pump, had beforehand been sterilized with 2% sodium hypochlorite (NaOCl) active (bleach) and thoroughly rinsed with water. Tanks were aerated with an airstone until the start and throughout the experiment.

On 19/04/2022, 12 specimens of R. sumatrensis were collected from Sungai Sekerubong, Rajang River basin, Kanowit, Sarawak, Borneo, Malaysia, by hand, and the following day, transported to the laboratory in original river water and immediately placed in an additional aerated tank with river water. Mussels were arranged in two groups of “small” and “large” specimens, respectively.

The eDNA release experiment with the “Lake” water was conducted on April 20, 2022, and the “River” water experiment 24 h later. “Lake” and “River” experiments were conducted in different tanks and with different mussel specimens. On each occasion, one small and one large R. sumatrensis specimen was placed in each of three “Lake” or “River” tanks to attain similar mussel eDNA release (average total mussel wet weight per tank = 19.23 ± 1.96 g). All mussels were verified to be actively filtering during the incubation and were removed after 16 h using a sterilized net. One control tank per experiment was left without mussels (Figure 1A).

After incubation, four sample types were generated from both “Lake” and “River” experiments (Figure 1A). In each case, 100 mL of water was filtered (1) directly from the mussel incubation tanks (“high density” treatment, i.e., reflecting a population density of 2 individuals per 40 L, and thus, 50 individuals m⁻³ or m⁻² assuming 1 m water depth); (2) after a 1:10 dilution of the mussel incubation water (“low density” treatment, i.e., reflecting a population density of 2 individuals per 400 L, and thus, 5 individuals m⁻³ or m⁻² assuming a 1 m water depth); (3) from the incubation tank with no mussels (“controls”); and (4) from a source of distilled water (“distilled water controls”).

In total, 18 different eDNA protocols were tested on replicate 100 mL samples of each of the 16 water sources, respectively (Figure 1). Not all variables were tested in all combinations, due to factors such as the unavailability of all pore-size membranes in all membrane materials. The variables tested are summarized in Figure 1B and varied with respect to (1) eDNA capture method, i.e., filtration through (1a) a Thermo Scientific Nalgene Reusable Filter Unit and vacuum pump using 47 mm-diameter filter membranes (ca. 17.35 cm² surface area), or (1b) a Terumo 50 mL Luer lock syringe using Sartorius 25 mm diameter (ca. 4.9 cm² surface area; polycarbonate) filter holders and filters; (2) filter type, i.e., (2a) Mixed Cellulose Ester (MCE; all MF-Millipore) or (2b) Cellulose Nitrate (CN; 47 mm diameter—all Whatman, 24 mm diameter—all Sartorius); (3) filter membrane pore size (0.2, 0.45, 0.8, 3 and 5 μm); (4) pre-filters (i.e., Whatman Nucleopore Hydrophilic Polycarbonate Membrane 12 μm (PC) or 5 μm (Sartorius CN)) vs. no pre-filter; (5) preservative, i.e., 1 mL of 100% ethanol at −20°C vs. 1 mL Longmire's buffer (hereafter, “buffer”) at 4°C (constant storage temperatures following Mauvisseau et al. (2021)); and (6) DNA extraction kit, i.e., Qiagen DNeasy Blood & Tissue Kit vs. Qiagen PowerSoil ProKit (Figure 1).

The Nalgene filtration unit and tubing were sterilized between each sample by submerging in 2% NaOCl active (bleach) for 5 min, followed by thorough rinsing in water and subsequent sterilization under UV light for 20 min on each side. Syringes and filter holders were sterile and not reused. Ethanol-preserved filters were stored at −20°C, and buffer-preserved filters at 4°C until processing, with the exception of a ~36 h period of transportation from Malaysia to the University of Nottingham Life Science laboratories, UK, where samples were processed.

The 5 μm (pre-filter) and 0.45 μm membranes generated with the syringe filtration method were stored and extracted separately. After extraction, a subsample of eDNA from each size fraction extract was pooled equimolar and analyzed alongside the extracts from separate size fractions. The much larger 12 μm pre-filter membranes generated with the vacuum pump filtration method were not extracted.

2.4 eDNA Extraction

Ethanol-preserved filter membranes were removed from the ethanol and placed in sterile 1.5 mL Eppendorf tubes covered with pierced tin foil in an incubator in a fume hood at 56°C for 12 h to allow ethanol to evaporate while minimizing potential cross-sample contamination. The original ethanol collection tubes were retained at −20°C until the following day (i.e., the day of extraction) when they were centrifuged at 10,000 g and 4°C for 40 min to pellet remaining DNA in solution. Ethanol was poured out, and filter membranes were placed back in their original tubes and into the incubator for a further 3 h to remove any residual ethanol. 47 mm-diameter filter membranes were thereby cut into smaller pieces using sterilized microscissors upon replacement in their original tubes to optimize DNA lysis (also see e.g., Baudry et al. 2021). eDNA from buffer-preserved filter membranes was extracted without prior treatment except for cutting 47 mm-diameter filter membranes as explained above.

DNA extraction from filter membranes using the Qiagen DNeasy Blood & Tissue Kit followed the manufacturer's protocol with following modifications: The only modification for ethanol-preserved membranes was a 3-fold increase in ATL lysis buffer volume to completely cover filter membranes. For samples preserved originally in 1 mL of Longmire's buffer, buffers ATL and AL were omitted. Instead, to increase GuHCL concentration in the buffer-preserved samples, 8 M GuHCL (50 μL per 100 μL volume of Longmire's buffer) and 100% ethanol (100 μL per 100 μL Longmire's buffer) were added. Custom Proteinase K solution (10 μL per 100 μL Longmire's buffer) was used to achieve active unit concentration equivalent to kit specification. For all sample types, DNA was eluted twice with 50 μL of AE buffer with 5 min incubations at 37°C. DNA extraction using the Qiagen PowerSoil ProKit followed the manufacturer's protocol with Longmire's buffer replacing the CD1 buffer in buffer-preserved samples.

All samples were extracted in a dedicated eDNA extraction-only room in a UV-sterilized laminar flow hood to minimize cross-contamination. The qPCR Master Mix preparation, eDNA addition, and qPCR analysis were conducted in separate hoods to further control and minimize contamination. Negative laboratory controls were extracted regularly (i.e., with each batch) and amplified via qPCR as below.

2.5 qPCR For eDNA Detection

To amplify R. sumatrensis eDNA, qPCRs were carried out as detailed above (Section 2.2). Five technical replicates were run for each eDNA extract. For Lake-Syringe kit samples only, we additionally tested the performance of combined eDNA extracts from 5 μm and 0.45 μm pore-size filter membranes. For that purpose, equal parts of the two paired DNA extracts per sample were combined and concentrated by including 4 μL of pooled template per reaction. Reaction volumes were adjusted to retain the same concentrations as above.

To ensure comparability of data across plates, a dilution series of R. sumatrensis genomic DNA (Table 1; 6.38 ng/μL to 6.38 × 10⁻⁴ ng/μL) was run in duplicate, along with two replicate eDNA samples (positive controls; one from each low density-lake and river tank, respectively; both captured with Nalgene filter unit, preserved in buffer, and extracted with Qiagen DNeasy Blood & Tissue Kit) and two negative controls (nuclease-free water; all 2 μL) per plate.

2.6 Inhibition Tests

Inhibition of the qPCR assay was tested using an exogenous internal positive control (IPC) following Carraro et al. (2017) on 14 eDNA extracts (i.e., five lake samples, five river samples, and four distilled water-control samples), covering all filter pore sizes, preservation, and extraction methods tested (Figure 1), run in duplicates and with four negative PCR controls (nuclease-free water). Reactions were run with 5 μL 2x GoTaq Probe qPCR Master Mix, 0.05 mM of each MIMf and MIMr primer, 0.25 mM of the IPC probe, 1 μL of the IPC template (1 × 10⁻⁷ mM), 2 μL DNA, and nuclease-free water to equal a total reaction volume of 10 μL. Thermal cycling regime and reaction analysis followed Carraro et al. (2017). Inhibition was considered significant if the amplification of the IPC template in a reaction with an eDNA sample added showed a shift in the quantification cycle number (C_q) of three or more relative to the reactions with only PCR-grade water added (Hartman et al. 2005).

2.7 Data Analysis

Limit of detection (LOD) and limit of quantification (LOQ) were calculated based on C_q-values obtained from qPCRs on the dilution series of R. sumatrensis genomic DNA (see above). LOQ was determined as the lowest template concentration that could be reliably quantified at 95% confidence (Armbruster and Pry 2008). For this purpose, two log10-regression models were fitted to the DNA concentration vs. C_q-values obtained from the genomic dilution series and compared. Firstly, the “predictive” model was obtained by fitting data for dilutions 6.38 ng/μL to 6.38 × 10⁻² ng/μL and projected to 6.38 × 10⁻⁹ ng/μL. Secondly, the “observed” model was obtained by fitting data across the whole dilution series. LOQ was set where the observed values started to divert from the predictions and fell outside of the corresponding 95% confidence interval. Beyond this limit, the assay can no longer reliably quantify samples with a lower concentration.

To assess the LOD, we examined the success rate of amplifications across all replicates at a given dilution. Any observed amplification was classed as a success, and no amplification was regarded as a failure. The absolute LOD was set at the concentration where at least one of five replicates was successfully amplified (20%). We then ran another qPCR using the same conditions and only the tissue dilutions 6.38 × 10⁻⁶ ng/μL and 6.38 × 10⁻⁷ ng/μL, alongside a 1:5 dilution of 6.38 × 10⁻⁶ ng/μL for a concentration between these two dilutions, with 20 replicates per dilution. The LOD was set where at least 5% of these replicates were successfully amplified. The efficiency of the assay was calculated as [10^(−1/slope) – 1] × 100%.

A sample was considered as positively detecting R. sumatrensis eDNA if (1) at least 80% (i.e., four of five) of qPCR replicates amplified above the LOD (i.e., C_q < 38.4; see Section 3); (2) amplification curves showed a regular and clear transition into a steady exponential phase from a uniform and stable baseline (with an even normalization to the reference dye (ROX)); and (3) there was no amplification in the corresponding laboratory or qPCR controls. Detection rates for samples amplifying below the LOD (i.e., at C_q > 38.4) are also reported below as “sub-LOD” amplifications.

To assess statistically significant differences in C_q-values among different water sources, mussel densities, and methodologies of DNA capture and extraction, General Linear Models (GLMs) were run on data collected from buffer-preserved samples in R-version 3.6.3, fitting C_q-values (including those from sub-LOD amplifications) as the response variable and (1) DNA extraction method, (2) filter membrane pore size, (3) water type, and/or (4) mussel density as factors with two levels, respectively.

3.1 qPCR Assay Development, Optimization, and Validation

Gradient PCR revealed strong amplification across the dilution series of genomic DNA. qPCRs were therefore run with a 60°C annealing temperature, which is the optimal annealing temperature recommended for the polymerase/Master Mix formulation used. There was no evidence for a shift in C_q value in IPC amplification when eDNA was introduced to the reactions, indicating that eDNA samples were not inhibiting the assay.

There were considerable numbers of mismatches between the developed primer/probe sequences and available COI sequences of non-target taxa (Table 3). No amplification of non-target taxa occurred above the LOD in any of the tissue extractions, although sub-LOD amplification occurred in all four species in one to four of six replicates (Table 3). Based on our occurrence criteria, none of these detections would have been regarded as positive for R. sumatrensis.

qPCR assay efficiency was 102%. The LOQ for the assay was determined as 6.38 × 10⁻⁶ ng/μL or a C_q of 36.72. The LOD was determined as a concentration of 6.38 × 10⁻⁷ ng/μL, referring to a C_q of 38.4.

3.2 Variation in Rates of False Positives, eDNA Detection Rates and C_q-Values Across Protocols

No amplification was observed in any of the negative laboratory controls.

Only one out of the 120 samples for which eDNA was captured with the Nalgene unit using one of the 10 protocols involving an ethanol-preservation step (at −20°C) fulfilled the requirements for positive R. sumatrensis eDNA detection (i.e., amplifying at C_q < 38.4 in at least four out of five qPCR replicates) (Table 4). In addition, one qPCR replicate of a corresponding distilled water sample, which had been preserved in ethanol and was extracted with the Qiagen DNeasy Blood & Tissue Kit, amplified above the LOD threshold (Table 4).

Buffer-preservation in Nalgene-unit samples resulted in a considerably higher eDNA detection rate than ethanol-preservation and no false positives above the LOD (Table 4). Although buffer-preservation was tested only on low-density samples, R. sumatrensis eDNA was positively detected in 19 out of 24 samples (79%). A 100%-detection rate was achieved across all buffer-preserved samples extracted with the Qiagen DNeasy Blood & Tissue Kit, for which eDNA was collected either with a 12 μm-PC pre-filter followed by a 0.45 μm-MCE filter or a 0.8 μm-MCE filter without pre-filter. However, three qPCR replicates of one of the DW controls (Lake, 0.8 μm-MCE) that were extracted with the Qiagen DNeasy Blood & Tissue Kit also amplified, albeit below the LOD threshold (Table 4). No amplification in negative controls was observed in buffer-preserved samples extracted with the Qiagen PowerSoil ProKit, which recovered 50% and 67% detection rates in 0.45 μm-MCE and 0.8 μm-MCE samples, respectively.

Across all buffer-preserved samples of the Nalgene experiment, R. sumatrensis eDNA amplification success rate was higher in Tissue Kit- compared to Soil Kit-extracts, River- compared to Lake-samples, and 0.8 μm-MCE compared to pre-filtered 0.45 μm-MCE samples, respectively (Table 5a). The type of DNA extraction method, filter membrane pore size, and water type all significantly affected C_q-values (GLM: F_3,105 = 43.25, p < 0.0001, R²_adjusted = 0.54; see Table 6a for results on each factor). Significantly lower C_q-values were recovered for samples (1) for which DNA was extracted using the Tissue- compared to the Soil Kit, (2) for which eDNA was collected with a pre-filtered 0.45 μm compared to a 0.8 μm-MCE filter membrane, and (3) from river compared to lake water (Table 5a, Figure 2a). Thus, although a greater proportion of 0.8 μm-MCE compared to pre-filtered 0.45 μm-MCE samples amplified above the LOD, those 0.45 μm-MCE samples that amplified above the threshold did so earlier (i.e., at lower C_q-values).

All samples processed with the syringe kit were preserved in buffer. R. sumatrensis eDNA was positively detected in 37 out of 48 single-membrane samples (77%; i.e., not including pooled 5 μm- and 0.45 μm-CN membrane extracts), with no false positives above the LOD (Table 4). A 100%-detection rate was achieved across samples collected with a 5 μm-CN filter and extracted with the Qiagen DNeasy Blood & Tissue Kit (Table 4). However, at least one qPCR replicate of one or two of the negative control samples also amplified—albeit below the LOD—for both protocols involving the Qiagen DNeasy Blood & Tissue Kit and one protocol involving the Qiagen PowerSoil ProKit, respectively. No amplification in negative controls was observed in the 5 μm-CN—Qiagen PowerSoil ProKit combination, which recovered an 83% detection rate across all samples.

At high mussel densities, detection rates for pooled 5 μm- and 0.45 μm-CN membrane extracts (i.e., 5-CN + 0.45-CN samples in Table 4) were similar to extracts from > 5 μm- (i.e., 5-CN Main filter) and 5–0.45 μm fractions (i.e., 5 μm prefilter and 0.45 μm main filter). At low mussel densities, the pooled extracts failed to detect R. sumatrensis in all replicate tanks, with detection rates being considerably lower than for the > 5 μm-fraction and similar to the 5–0.45 μm fraction (Table 4).

Across all samples of the Syringe experiment (not including pooled 5 μm- and 0.45 μm-CN membrane extracts), R. sumatrensis eDNA amplification success rate was higher in high-density compared to low-density samples, Tissue Kit compared to Soil Kit extracts, River compared to Lake samples, and 5 μm-CN compared to 0.45 μm-CN samples, respectively (Table 5b). C_q-values were significantly affected by mussel density and water type but not by the type of DNA extraction method or filter membrane pore size (GLM: F_4,217 = 100.1, p < 0.0001, R²_adjusted = 0.64; see Table 6b for results on each factor). Significantly lower C_q-values were recovered for samples (1) from high-density compared to low-density treatments, and (2) river compared to lake water (Table 5b, Figure 2b,c).

4.1 A qPCR Protocol for Detecting the Tropical Freshwater Mussel Rectidens sumatrensis

Our study presents the first qPCR assay for a tropical freshwater mussel, i.e., the Southeast Asian species R. sumatrensis. Our validation procedure has shown that the assay is specific, efficient, and able to detect and quantify R. sumatrensis DNA at concentrations above 6.38 × 10⁻⁷ ng/μL and 6.38 × 10⁻⁶ ng/μL, respectively.

4.2 Effects of eDNA Capture, Preservation, and Extraction Methods on eDNA Detection Rates

Of the 18 different protocols for capturing, preserving, and extracting freshwater mussel eDNA tested in our study, eight resulted in a high eDNA detection rate by subsequent qPCR with no to minimal occurrence of positive controls. The single unifying characteristic across these successful protocols was the preservation of filter membranes in Longmire's buffer rather than absolute ethanol. In general, the amplification success rate was particularly high and C_q-values were particularly low for filter membrane pore sizes > 0.45 μm (without pre-filtering) and when extracting DNA with a Tissue- rather than Soil Kit. The protocols performed well across both tropical water types tested, albeit generally better in river- compared to lake samples, and at densities as low as 5 individuals m⁻³ or m⁻² assuming a 1 m water depth. While this may still be considered a relatively high density for many freshwater mussel species (see e.g., Smith 2006; Simeone et al. 2021), considering that only a comparatively low amount of water was filtered in our experiments (i.e., 100 mL), our protocols can be expected to perform well at even lower population densities when filtering larger water volumes (commonly 0.5 L up to > 30 L; e.g., Dysthe et al. 2018; Gasparini et al. 2020; Preece et al. 2021; Prié et al. 2021; Marshall et al. 2022; Prié et al. 2023). Importantly, the protocols presented here are cost- and time-efficient, utilizing equipment that can be at least partially reused and is readily available in most tropical countries.

Our study applied two different types of filtering equipment, i.e., a reusable filter unit with a vacuum pump vs. a Luer lock syringe with filter holders. Due to differences in pre-filter pore sizes (i.e., 12 μm for filter unit and 5 μm for syringe-system) and material in our experiment, our results do not allow for a direct comparison of the performance of these systems. However, both performed well, achieving detection rates of 77%–79% for buffer-preserved samples. A major advantage of the syringe approach compared to the filter unit approach is that sterilized syringe kits (including syringes, a pair of gloves, filter holders, filter membranes, forceps to handle filter membranes, and vials with buffer) can be pre-prepared for each field outing, thereby reducing the post-fieldwork time required to process collected samples and minimizing the risk of cross-contamination of samples from different sites.

Our results indicate the advantages of using filter-pore sizes of 0.8 μm for detecting tropical freshwater mussels through eDNA. Filter pore sizes > 0.45 μm (commonly > 1 μm) were also found to be highly effective in detecting freshwater mussel eDNA in, for example, North American rivers (Currier et al. 2018; Dysthe et al. 2018; Marshall et al. 2022) and Japanese lakes (Togaki et al. 2020; Sugawara et al. 2022). A trend of larger filter membrane pore sizes retrieving higher eDNA concentrations was also observed for Margaritifera margaritifera in Norwegian rivers, although this result did not account for the reduced water volume filtered through smaller pore sizes (Wacker et al. 2019). Regardless, considering the commonly high sediment loads and turbidity in tropical rivers and lakes (Dudgeon 1999) and the excellent performance of 0.8 and 5 μm filter membranes in our study, we recommend using large pore-size filters for future tropical freshwater mussel eDNA surveys, which will allow for filtering comparatively large water volumes, thus further reducing rates of false negatives.

Our study indicates that preservation in Longmire's buffer is by far the preferred method of preserving eDNA of tropical freshwater mussels compared to ethanol preservation. This result is in accordance with Majaneva et al. (2018), Kumar et al. (2020), and Huerlimann et al. (2020), who recommend buffer- rather than ethanol-preservation for tropical and other aquatic eDNA. Apart from resulting in a higher eDNA detection rate, buffer-preservation has a number of additional advantages to ethanol-preservation, including easier handling and transport (including via airplane) and removing the ethanol-evaporation step in the laboratory. Majaneva et al. (2018) highlighted that this step may increase the risk of contamination, which may have been the reason for the false positive in one qPCR replicate of one of our ethanol-preserved, Tissue Kit-extracted samples. Our results also confirm that buffer-preservation at 4°C maintains high-quality eDNA, with some studies showing evidence for effective protection against degradation even at temperatures of up to 45°C for up to 2 weeks (Renshaw et al. 2015; Majaneva et al. 2018). This provides a further argument for applying buffer-preservation on eDNA samples collected under challenging fieldwork conditions and remote locations in the tropics. An alternative preservation methodology not tested here is drying with silica gel, which has shown promising results in other studies (Dysthe et al. 2018; Majaneva et al. 2018).

The Qiagen DNeasy Blood & Tissue Kit is arguably the most commonly used kit for freshwater mussel eDNA extraction (see above). Our results confirm the excellent performance of this kit for extracting tropical freshwater mussel eDNA preserved in Longmire's buffer. This kit performed better than the Qiagen PowerSoil ProKit in terms of amplification success rates and C_q-values and positively detected R. sumatrensis across all mussel tanks in all buffer-preserved filter-unit samples and all buffer-preserved syringe-system samples captured with 5 μm-filter membranes. However, instances of amplification in distilled water and control tanks—albeit all below the LOD—were also more than twice as high in Tissue Kit- compared to Soil Kit samples (i.e., 7 vs. 2 qPCR replicates). Thus, the Qiagen PowerSoil ProKit, which includes an Inhibitor Removal Technology that increases the purity of isolated DNA (Qiagen 2013–2024b), maybe the preferred extraction method in some instances due to its lower susceptibility to false positives whilst retaining relatively high amplification and detection rates, particularly in combination with membrane pore sizes ≥ 0.8 μm (positively detecting R. sumatrensis in 14 out of 18 (78%) buffer-preserved samples). However, the costs per sample are more than twice as high for the Qiagen PowerSoil ProKit (Qiagen 2013–2024b) compared to the Qiagen DNeasy Blood & Tissue Kit (Qiagen 2013–2024a).

4.3 Effects of Water Type and Inhibition

No PCR inhibition was observed in our samples, which were taken from a tropical lake and river, respectively. However, PCR inhibition is a common problem in eDNA from tropical water bodies, due to organic compounds from phytoplankton and/or terrestrial input, particularly during the monsoon season (Matheson et al. 2010; Robson et al. 2016; Huerlimann et al. 2020). A commonly applied strategy in these instances is the dilution of eDNA extracts (Cantera et al. 2019; Cilleros et al. 2019; Gasparini et al. 2020), which, however, reduces eDNA concentrations in the PCR. Alternatives include post-extraction inhibitor-cleanup processes, cleanup columns (McKee et al. 2015; Robson et al. 2016), and the Zymo OneStep PCR inhibitor (Zymo Research) (Dokai et al. 2023). The Qiagen PowerSoil ProKit or similar kit eliminates inhibitors already in the course of the eDNA extraction process.

Whilst eDNA detection rate was higher and C_q-values were lower in river- compared to lake-water, respectively, it is difficult to identify the reason(s) for this discrepancy. Potential explanations for this pattern include differences in eDNA shedding rates and differences in eDNA particle binding to suspended solids in the two water types (Stoeckle et al. 2017). Specifically, eDNA shedding rates were potentially higher in river water because it likely provided conditions more similar to the natural habitat of R. sumatrensis than the lake water (Zieritz and Lopes-Lima 2018).

4.4 Conclusions and Outlook

Whilst our study provides the first eDNA protocols specifically targeted towards detecting tropical freshwater mussels, many questions remain unanswered. Due to the requirement of live specimens in tank experiments, for ethical reasons, we here focused on the most common species in Borneo. Thus, the application of the eDNA protocols on other tropical mussel species will require the development of new qPCR assays. Multi-species approaches, such as eDNA metabarcoding (e.g., Prié et al. 2021) or multiplex assays (e.g., Rodgers et al. 2020) may thereby be more cost-effective than species-specific approaches depending on the study system and aim. Field studies comparing eDNA detection rates to data from traditional hand surveys will be needed to validate eDNA protocols in the field. These should be conducted at different times of the year, as eDNA concentrations and degradation rates of freshwater mussels have been shown to vary significantly across seasons, partly due to temporal patterns in larval release periods (Wacker et al. 2019; Lor et al. 2020; Sugawara et al. 2022). Freshwater mussel eDNA transport has been shown to be extremely variable across previous studies in temperate freshwater systems, ranging from < 1 m to dozens of km (downstream) downstream of populations (Deiner and Altermatt 2014; Stoeckle et al. 2016; Lor et al. 2020; Steiner et al. 2022), but no data are available for tropical systems. The spatial resolution of sampling sites and the number of replicate samples per site will need to be adjusted depending on the specific conditions at the site and the season of sampling. An exciting prospect for future surveying of tropical freshwater mussel eDNA is passive samplers, which have been shown to adsorb mussel eDNA from temperate freshwater environments (Kirtane et al. 2020). Whilst eDNA is an exceptionally promising tool for tropical freshwater mussel and other biodiversity monitoring, traditional surveys using hand sampling will still be required to monitor population demography and certain other conservation-relevant data (Stoeckle et al. 2016).