Simple, sensitive and species-specific assays for detecting quagga and zebra mussels ( Dreissena rostriformis bugensis and D. polymorpha ) using environmental DNA

Early detection is paramount for attempts to remove invasive non-native species (INNS). Traditional methods rely on physical sampling and morphological identification, which can be problematic when species are in low densities and/or are cryptic. The use of environmental DNA (eDNA) as a monitoring tool in freshwater systems is becoming increasingly acceptable and widely used for the detection of single species. Here we demonstrate the development and application of standard PCR primers for the detection of two freshwater invasive species which are high priority for monitoring in the UK and elsewhere: the Dreissenid mussels; Dreissena rostriformis bugensis (Andrusov, 1987) and D. polymorpha (Pallas, 1771). We carried out a rigorous validation process for testing the new primers, including DNA detection and degradation experiments in mesocosms, and a field comparison with traditional monitoring protocols. eDNA from single individuals of both mussel species could be detected within four hours of the start of the mesocosm experiment. In field trials, the two mussel species were detected at all sites where the species are known to be present, and eDNA consistently outperformed traditional kick-net sampling for species detection. These results demonstrate the applicability of standard PCR for eDNA detection of freshwater invasive species.


Introduction
The rate of biological invasions has increased rapidly over the last 25 years due, at least in part, to increasing globalisation (Sutherland et al. 2013;Hulme 2009;Gallardo and Aldridge 2013a). Concern over the increasing number of invasive non-native species (INNS) has led to a number of horizon scanning studies aimed at identifying and prioritizing the threat of potential INNS (Gallardo and Aldridge 2013a, b;Roy et al. 2014). Roy et al. (2014) for example, concluded the potential impact, risk of arrival and risk of establishment of quagga mussels, Dreissena rostriformis bugensis more desirable approach than PCR due to its increased sensitivity for species detection (Thomsen et al. 2012a;Nathan et al. 2014) and, when using a probe-based assay, its added specificity. However, for many applications, the sensitivity of standard PCR may be quite adequate, and PCR may be preferable as it is cheaper and less technically challenging. Moreover, studies that have directly compared the two approaches have indicated that PCR can be more robust to PCR inhibitors than qPCR, which is important for avoiding false negatives (De Ventura et al. 2017).
In this study, we focussed on the detection of quagga and zebra mussels, which are high priority species for monitoring in a number of countries, including the UK; the focus of our sampling. These species originate from the Ponto-Caspian area and have spread rapidly throughout their invasive ranges via boat transportation, canals and river basin connections (Timar and Phaneuf 2009;Bij de Vaate et al. 2002). The rapid spread has been aided by both human interaction and their unique ecology (Timar and Phaneuf 2009). Like other mussels, Dreissenids have a free-floating planktonic veliger life stage, during which young can be dispersed over a large area downstream of parental populations (Ricciardi et al. 1995;Karatayev et al. 2002Karatayev et al. , 2015. Compared to many other mussel species, Dreissenids exhibit unique abilities to colonise new environments by using protein-based byssal strands formed inside their shell to secure to hard surfaces, which can be a significant aid to transportation and establishment (Ricciardi et al. 1998;Karatayev et al. 2002;Aldridge et al. 2004;Timar and Phaneuf 2009;Peyer et al. 2009). Colonization of new areas and establishment has been facilitated by the ability of Dreissenids to survive out of water for up to 15 days (Ricciardi et al. 1995) and survive a wide range of environmental extremes (Gallardo and Aldridge 2013b). Both mussels are described as "ecological engineers" (Karatayev et al. 2002(Karatayev et al. , 2007Roy et al. 2014) having influences on all trophic levels. In some instances, mussels provide increase in shelter and habitat for benthic macroinvertebrates (Karatayev et al. 2002), however they also compete for food and decrease diversity, and have been directly linked to declines in native Unionid mussels (Ricciardi et al. 1996). Dreissenid feeding behaviour also has negative effects on phytoplankton and has been linked to greater numbers of cyanobacteria blooms (Karatayev et al. 2002). Economic impacts are also widely documented for both species. For example, between US$161-US$467 million was spent by water treatment and electric power facilities in North America on the control and removal of D. polymorpha between 1989polymorpha between -2004polymorpha between (Connelly et al. 2007. The traditional monitoring of D. r. bugensis and D. polymorpha poses a challenge due their morphological similarity (Peyer et al. 2011). eDNA is therefore a promising complementary tool for Dreissenid monitoring.
Previous studies have designed and tested primers for detection of Dreissenid eDNA using either standard PCR (Bronnenhuber and Wilson 2013;De Ventura et al. 2017) or qPCR (Peñarrubia et al. 2016;De Ventura et al. 2017;Gingera et al. 2017;Amberg et al. 2019). However, the performance of these assays were not assessed in controlled mesocosm experiments or compared directly to traditional sampling methods in the field, and (in the case of Peñarrubia et al. 2016) do not distinguish the two species. eDNA assays based on microfluidic chip , light transmission spectroscopy (LTM) technology (Egan et al. 2013) are also available for quagga mussels, and on loop-mediated isothermal amplification (LOOP) for both Dreissena species (Williams et al. 2017) although these tools, at present, are arguably less widely applicable. Our objective was to design and test species-specific standard PCR assays for detection of eDNA from quagga and zebra mussels, that can be deployed for routine monitoring in a simple and cost-effective way by both regulators and researchers. Our framework for development consisted of: 1. in silico and in vitro primer testing of de novo primer pairs; 2. mesocosm experiments to evaluate eDNA detection and degradation over time at three different densities (one, five and twenty individuals of each species); and 3. testing the efficiency of the eDNA assays compared to traditional kick-net sampling in the field.

Specimen sampling and tissue DNA extraction
Dreissena polymorpha is widespread and common in the UK, having arrived in the 1820s potentially via the timber trade (Bij de Vaate et al. 2002;Quinn et al. 2014). Dreissena rostriformis bugensis is a much more recent invader, with the first UK record from 2014 in the River Wraysbury (Mills et al. 2017). This reservoir is subject to water transfers within the region and the mussel was subsequently found in neighbouring reservoirs. Specimens of both Dreissena species were collected at sites with known populations, two weeks prior to the beginning of each mesocosm experiment. These sites were as follows: D. r. bugensis: Wraysbury River, UK Grid. Ref.
Specimens were kept in tanks with continuous aeration and fed dried Cyclotella ad libitum. Samples from the commonly co-occurring native mussel species (Sphaerium corneum and Anadonta anatina) were also collected for tissue DNA extraction and primer testing. Tissue samples from both individuals of each invasive species and native species were extracted using the DNeasy Blood and Tissue Kit® (Qiagen, Hilden, Germany) according to the manufacturer's protocol.

Species-specific primer development
Species-specific primers were designed and tested in silico with Primer BLAST (Ye et al. 2012) using all available COI reference sequences from GenBank (D. r. bugensis -7 sequences, D. polymorpha -31 sequences, see Supplementary material Appendix 1 Table S1 for details of GenBank accession numbers). Suitable primers for in vitro testing were selected based on target species amplification, similarity of forward and reverse primer melting temperature, > 50% GC content and primers with an amplicon size of < 200 bp were preferentially selected to be suitable for amplification of degraded eDNA (Deagle et al. 2006;Jerde et al. 2011;Bronnenhuber and Wilson 2013;Mächler et al. 2014;Ardura et al. 2015). In total four primer pairs were tested in vitro for each species (see Appendix 1 Table S2 for all primer pairs tested). In vitro testing was carried out on tissue samples of the target INNS and three non-target taxa i.e. the congeneric INNS and two native taxa which are likely to co-occur in the same habitat. Serial dilutions of neat tissue-extracted target-species DNA (at approximate concentrations of 5 ng/μl, 0.5 ng/μl, 0.05 ng/μl, 0.005 ng/μl) were carried out to establish the Limits of Detection (LoD) for each primer pair (see Appendix 1 Table S3). PCRs were carried out in 25 μl volumes with 12.5 μl MyTaq Red Mix (Bioline, UK), 0.4 μM of each primer and 2 μl of DNA template. Temperature gradient PCRs were carried out on an Applied Biosystems Veriti Thermal Cycler in order to determine the correct PCR conditions. The following profile was selected after testing: 94 °C for 3 min, followed by 37 cycles of 94 °C for 30 s, 65 °C for 1 min and 72 °C for 1 min 30 s, with a final extension time of 10 min at 72 °C. PCR products from tissue samples were visualised by gel electrophoresis and stained with GelRed (Cambridge Bioscience Ltd, UK).
Based on PCR optimisation and the performance of the primer pairs, one primer pair from each species was selected. Four products using the selected primer was then. Sanger sequenced by Macrogen Europe in the forward direction. Sequences were compared with the National Centre for Biotechnology Information (NCBI, http://www.ncbi.nlm.nih.gov) using BLAST to confirm species identification (see Appendix 1 Figures S1 and S2 for sequence alignments).

Mesocosm experiments
Mesocosm experiments were carried out from January to April 2016 to test the sensitivity of the selected primer pairs under controlled densities and to investigate the rate of DNA accumulation and degradation. Each experiment was conducted in 15 L plastic tanks with fitted lids. Tanks were located in a climate controlled facility where temperature averaged 16 °C (range 14-18 °C) with light:dark cycles of 16 h:8 h. All tanks, aeration equipment and sampling equipment was sterilized in 10% commercial bleach solution for 10 minutes, then rinsed with 10% MicroSol detergent (Anachem, UK) and purified water prior to the experiment. Sampling and filtering equipment were also cleaned using the above method between each sampling event. Each tank was filled with water collected from Hotham Beck (SE 89133 32489) which has no prior record of Dreissenid mussels at the site or at any location within the waterbody. Field-collected water was used in order to provide natural feeding opportunities for the mussels and limit stress caused. Tanks were supplied with constant air via sterile tubing and aeration stones for 48 hours prior to the start of the experiment and covered for the duration of the experiment with a fitted lid.
For each species, the experiment consisted of 10 tanks representing three replicates of three treatment densities (one, five and twenty individuals) and a control tank with no individuals. Specimens of similar total biomass were used in the density replicates in order to minimise any influence of different biomass (see Appendix 1 Tables S4 and S6 for biomass information gathered pre-and post-mesocosm experiments). Room temperature, control tank and water temperature were recorded prior to each sampling event (see Appendix 1 Tables S5 and S7 for temperature measurements taken during the mesocosm experiments). Tank water temperature was kept below 10 °C to minimise any potential spawning events. Before the specimens were added to the tanks, a water sample was collected and filtered to ensure no contamination from the target taxa; this sample was recorded as 0 hours. Tanks were sampled over 42 days at 4 hrs, 8 hrs, 24 hrs, 7 days, 15 days and 21 days with the species present. On day 21, the specimens were removed from the tanks and sampling continued at 22 days, 28 days, and 42 days.
A total of N = 100 water samples were collected per species. For each sampling event, the tank water was homogenised by stirring with a sterile spatula before collecting 200 ml water from each tank. Samples were vacuum filtered through sterile 47 mm diameter 0.45 μm cellulose nitrate membrane filters with pads (Whatman, GE Healthcare, UK) immediately after collection, using Nalgene filtration units (Thermo Fisher Scientific) in combination with a vacuum pump (15~20 in. Hg, Pall Corporation) in a dedicated eDNA laboratory at the University of Hull, UK. Filter papers were then placed in sterile petri dishes, sealed with parafilm and stored at −20 °C until extraction. The filtered water was then returned to the tank to maintain the water volume. This process was completed within one hour. The filtration units were cleaned with 10% commercial bleach solution and 10% MicroSol (Anachem, UK), and then rinsed thoroughly with deionized water after each filtration to prevent cross-contamination. All DNA extractions were carried out using a protocol modified from Brolaski et al. (2008) (for the full extraction protocol, see Appendix 2). Mesocosm samples were PCR amplified using the species-specific primers and visualised with conditions previously described. Three PCR products from each species/mesocosm experiment were Sanger sequenced to confirm primer specificity.

Field trials
Water samples were collected at sites with previous records of the target INNS to test and verify the efficiency of each INNS assay in the field (Figure 1). We collected a total of 54 samples per INNS from their respective river catchments (River Wraysbury for D. r. bugensis and River Welland for D. polymorpha). For each waterbody we collected 6 samples from three selected UK Environment Agency macroinvertebrate monitoring sites. Each sample consisted of 500 ml of water collected across the width of the river and was replicated 3 times (N samples = 18 per site). Each replicate was filtered and extracted independently. Sample bottles filled with ddH 2 O were taken into the field as sample blanks. Samples were processed within 24 hours using the same method as the mesocosm samples. Each replicate was PCR amplified three times. PCR products from a total of 4 field samples for each species were Sanger sequenced to confirm primer specificity. Sites were surveyed after eDNA sample collection using standard 3-minute kick-net samples also (Murray-Bligh 1999). Using the same approach as eDNA sampling we collected 6 × 3-minute kick-net samples at each of the selected sites.

Data Analysis
Positive detection of the target INNS in both mesocosm experiment and field trial was determined by a band found on the electrophoresis gel at the correct size. The bands strength was not used to determine a positive detection, as we believe any positive detection (including a faint band strength) found using the primers described in this paper will result in implementation of management strategies to confirm detection visually and carry our control measures. Binomial Generalized Linear Models (GLMs) with a logit link function were used to investigate the influence of density or total biomass and time since the start of the experiment (until the taxa were removed from the mesocosm) on the detection in the mesocosms. Models were checked by testing whether the residual deviance fitted a chi squared distribution. The best supported model was identified by the lowest AIC value, and models with Δ AIC < 2 were also considered equivalent (Burnham and Anderson 2002). All data analyses were performed in R v.3.3.1. (R Core Team 2017), with GLMs performed using the MASS package (Venables et al. 2002) and plots created in ggplot2 (Wickham 2016).
To ensure full reproducibility of this study the raw data and code can be accessed (https://github.com/RosettaBlackman/Blackman_et_al_Dreissenidae).

Primer specificity
One primer pair for each species (DRB1 for D. r. bugensis and DP1 for D. polymorpha, Table 1) Table S8). Note that the forward primer pair selected for D. polymorpha shares a 16 bp overlap with DpoCOI3F designed by Bronnenhuber and Wilson (2013) but our primer pair, DP1, amplifies a much shorter sequence (73 bp, as opposed to 164 bp). Species-specific primer testing on target tissue samples yielded positive PCR amplification of a single band at the expected size for both selected assays (Figure 2) following the above prescribed PCR chemistry and PCR settings. It should be noted that using less stringent conditions (e.g., lower annealing temperature) may result in non-target amplification or and must   Figure 1A and B). No bands of the expected size were obtained in the cross amplification tests; however, much larger, non-specific bands, were seen in non-target species for DP1 ( Figure 1B). Due to the substantial size difference these non-specific bands are easily distinguishable from the target band size and will not lead to false positive detections. Sequences generated from PCR products from all tissue, mesocosm and field samples were verified as being from the correct target species for D. rostriformis bugensis. However due to the short amplicon length of the DP1 product, we were unable to determine a good quality sequence length greater than the reverse primer (see Appendix 1 Figures S1 and S2) and therefore all positive detection for D. polymorpha was determined on band size alone.

Mesocosm experiments
We had no contamination of tanks prior to target species being added (see Figure 3 and also Appendix 3 Figures S1 and S2). Both Dreissenid mussel primers showed positive detection of their target species at all three density treatments at the four hour sampling event (Figures 3 and 4). Dreissena polymorpha was detected at all three replicates for each density treatment after 4 hours, whereas D. r. bugensis was detected in 1/3, 2/3 and 3/3 replicates for the 1, 5, and 20 individual treatments respectively. At least one positive replicate was obtained for every sampling point over the first 21 days. However, the number of detections did not increase linearly over time or density. The 20 individual density treatment was quite consistent over time, with positive detections in 3/3 replicates from 4 h to 21 days for both species, with just one exception (D. r. bugensis after 7 days). For D. r. bugensis, time and total biomass significantly influenced the detection by standard PCR. Of these two measures, total biomass was the more significant predictor in GLMs and generated the lowest AIC (GLM, z = 2.262, P = 0.023, AIC 55.368). After removal of D. r. bugensis, DNA was only detected in tanks with the highest mussel density (20 individuals) 24 hours after removal. DNA from these tanks was no longer detected at day 28 (7 days after removal). For D. polymorpha, both time and density were significant predictors of detection. Of the two measures, density was the strongest predictor with the lowest AIC (GLM, z = 1.969, P = 0.049, AIC 32.823). DNA from D. polymorpha persisted to day 42 (21 days after removal) in two of the three density treatments (see Appendix 1 Tables S9 and 10 for full GLM breakdown).

Field trials
Dreissenid mussels: D. r. bugensis specimens were found by kick-net sampling at all 3 sites surveyed, but the number of individuals found decreased with distance along the River Wraysbury from the main source population at Wraysbury Reservoir (Table 2). Detection by kick-net sampling was 33% (6 samples out of 18). Positive eDNA detections were obtained for every sampling replicate at each of the three sites along the River Wraysbury, hence eDNA detection was 100% (Table 2 and Appendix 3 Figure S5). Dreissena polymorpha was found by kick-net sampling in only one of three sites (Duddington, Table 2) although the species is known to be present throughout the sampled catchment. The number of positive detections for kick-net sampling was 11% (2/18 samples). Positive eDNA detections for D. polymorpha were obtained in 77.7% (14/18) of samples and 53.7% (29/54) of the PCR replicates, including in sites where specimens of D. polymorpha were not found (Table 2 and Appendix 3 Figure S6). Both primer sets showed positive detections of their respective target taxa in field, however for detection of D. polymorpha we do not see the clear bright bands observed in the mesocosm experiments, this could be due to inhibition in the samples.

Discussion
Rapid, cost-effective tools are needed for detection of newly invading, or spreading invasive non-native species. Here, we designed and tested PCR primer pairs for two highly invasive non-native species: D. r. bugensis and D. polymorpha, which are high priority for monitoring. Primers were tested in silico and in vitro, then in a series of mesocosm experiments and field trials. The two primer pairs amplify target tissue at a low concentration (0.005 ng/μl) which is broadly in line with other eDNA species-specific standard PCR primer assays, (e.g. detection limit of 0.00046 ng/μl (Davison et al. 2016) to 0.4 ng/μl (Ardura et al. 2015), with no cross-species amplification with each other, or with two native mussel species present in the UK. Both species were detected from eDNA collected from water samples in both laboratory and field trials. eDNA could be detected in mesocosms within 4 hours of the start of the experiment and detection at this first time point was possible from just one individual. Dreissenid eDNA was detected at every sampling point at all three densities in the mesocosms, and outperformed kick-net sampling for detection in the field. Below we highlight the range of factors that likely interact to determine the success of eDNA detection in real-world applications.

Mesocosm trials
Mesocosm experiments have been advocated (De Ventura et al. 2017), and performed by previous studies (Dejean et al. 2011;Thomsen et al. 2012b;Sansom and Sassoubre 2017) to allow information on species-specific DNA production rates, persistence and degradation over time, which informs users whether the method is appropriate for the detection of target taxa. Here, both abundance variables (density and total biomass) were significant predictors of detection for both species. Hence, there is a positive relationship between abundance and detection, as found in previous studies (e.g. Thomsen et al. 2012b). Mesocosm experiments also demonstrated the depletion of DNA once the specimens were no longer present in the tank; 7 days after removal for Dreissena rostriformus bugensis in agreement with similar studies (Dejean et al. 2011;Thomsen et al. 2012b). However, DNA for D. polymorpha was still detecting at the end of the experiment, 21 days after removal of the species. This indicates there are differences between species in terms of detectability and DNA did not accumulate or degrade in a linear fashion over time, as discussed below.
The mesocosm experiments performed in this study were useful for determining the assay sensitivity and for identifying differences in detectability between species. Our experiments revealed that the Dreissena primers are sensitive and robust -being able to detect single individuals within four hours and then consistently throughout the course of the experiment. The high rate of detection for Dreissenid mussels is likely due to the fact they were able to continuously filter feed on algae and phytoplankton present in the water column during our experiments, as they would in the wild, enabling them to maintain an active metabolism. However, it is clear from our study and others that DNA production and its availability in the water column is a complex topic and can vary substantially even between closely related species.
We might expect that as long as DNA production rate is greater than the degradation rate, (as seen in models produced by Thomsen et al. 2012a), eDNA availability should increase over the course of the experiment. Under this prediction, we expect the DNA concentration and the number of positive detections to increase over time, and for there to be an interaction with density. Alternatively, DNA concentrations may increase at first and then plateau, when an equilibrium is reached between DNA production and degradation (Klymus et al. 2015;Sansom and Sassoubre 2017;Nevers et al. 2018). As we are using standard PCR, rather than qPCR we are unable to determine DNA concentration, however we do see an increase in band strength in both Dreissenid mesocosm experiments between 4 and 24 hours. However, overall the number of positive detections fluctuates rather than showing an accumulation or a plateau over time. Time was also a significant predictor of detection probability for both species. The fluctuation in the number of detections over time may be due to a combination of the activity of the organisms, the balance between DNA production and degradation, and/or changes in the concentration of PCR inhibitors. In Dreissenid mussels, filter feeding may both release and uptake DNA, so the amount of DNA present in a controlled environment may reach an equilibrium. Further experiments with a quantitative method such as qPCR or ddPCR are needed to fully understand the dynamics of DNA concentration over the course of the experiment, and the influence of feeding and other behaviours on the rate of DNA production.

Field application
In the field tests, eDNA outperformed kick-net sampling for detection of both D. r. bugensis (100% eDNA vs 33% kick-net samples) and D. polymorpha (89% vs 11%). Below we discuss the reasons for the discrepancies between eDNA and kick-net samples, and the performance of both assays in the field.
There are numerous influences on the persistence of eDNA in waterbodies, that have been well documented such as: pH, microbial activity and transportation Jane et al. 2015;Jerde et al. 2016;Shogren et al. 2017, see Barnes et al. 2014 for further discussion). In our study, we reported higher detection rates from eDNA compared to kick-net samples in both species. These samples were collected from lotic systems, and therefore detection is likely to come from both local populations and eDNA being transported from upstream sources Jane et al. 2015). However, to what extent the DNA is being transported is still largely unknown. Previous work on river morphology states substrate type and the related flow regime, are huge influences on DNA transportation, substrate retention and subsequent resuspension (Shogren et al. 2017;Jerde et al. 2016).
Our results show no decrease in band strength for the detection of D. r. bugensis across the population density gradient along the River Wraysbury (2 km). In part this is likely to be down to transported DNA. However, it is also likely to be attributed to an increase in water mixing caused by rainfall before the samples were collected. As eDNA is not uniformly distributed through a river (Macher and Leese 2017) the rainfall is likely to increase the dispersion of eDNA in a waterbody (Shogren et al. 2017). We therefore see a greater number of positive detections. Similarly, variation in the eDNA detection throughout the river for D. polymorpha may be due to the relatively lower flows during these surveys which have caused a reduction in DNA distribution across the river. However, there is a fluctuation in the number of detections across the samples at each site for this species and we detect DNA at sites where they were not physically collected (as we also see with D. r. bugensis). This is further evidence of eDNA being transported down the catchment rather than a false positive result. This greater variability in detection due to the lower flow conditions is likely to demonstrate the true variation encountered when surveying lotic systems for target species.
In this study we were unable to determine successful detection of D. polymorpha by Sanger sequencing the PCR product generated from our primer pair, DP1. This was due to the short amplicon length (73 bp).
Although several studies may demonstrate species detection with small amplicon length (< 200 bp, Deagle et al. 2006;Jerde et al. 2011;Bronnenhuber and Wilson 2013;Mächler et al. 2014;Ardura et al. 2015;Takeuchi et al. 2019) there may be an optimal size to use when considering species detection from eDNA samples. As we know eDNA from water samples is known to be a combination of cellular and extracellular DNA  we must therefore opt for species determination over detection and aim to design primers based on a larger size amplicon (> 100 bp) for successful species determination (Meusnier et al. 2008).

Conclusion
This study provides targeted eDNA assays for two priority invasive nonnative species, and demonstrates a simple framework for assay development that can be used by regulatory bodies with responsibility for invasive species monitoring. It is important to highlight that by using eDNA it not only allows detection but also easily differentiate between cryptic species, such as the ones in this study. Standard PCR outperformed established kick-net sampling for both target species, and provides a simple and effective detection method without significant investment in qPCR, ddPCR or Next Generation Sequencing facilities. However, more quantitative methods are needed to provide deeper insights into the rate of DNA accumulation and degradation in both mesocosms and field experiments. This study also highlights some of the challenges for designing and implementing eDNA assays for different species, emphasizing the need to understand the dynamics of DNA production and degradation even by closely related species.

Supplementary material
The following supplementary material is available for this article: Appendix 1. Primer design and GLM testing Table S1. GenBank accession numbers used for designing species specific primers for INNS target taxa. Table S2. Species specific primer design. Table S3. Tissue samples used in species specific primer testing. Table S4. Specimen biomass pre and post mesocosm experiment, D. r. bugenisis. Table S5. Temperature measurements during mesocosm experiment, D. r. bugenisis. Table S6. Specimen biomass pre and post mesocosm experiment, D. polymorpha. Table S7. Temperature measurements during mesocosm experiment, D. polymorpha. Table S8. Species specific primer mismatches. Table S9. D. r. bugensis mesocosm GLM. Table S10. D. polymorpha mesocosm GLM. Figure S1. AliView alignment for D. polymorpha sequencing. Figure S2. AliView alignment for D. r. bugensis sequencing.