Skip to main content

Thermal stratification and fish thermal preference explain vertical eDNA distributions in lakes

Cite this dataset

Littlefair, Joanne et al. (2020). Thermal stratification and fish thermal preference explain vertical eDNA distributions in lakes [Dataset]. Dryad.


Significant advances have been made towards surveying animal and plant communities using DNA isolated from environmental samples. Despite rapid progress, we lack a comprehensive understanding of the “ecology” of environmental DNA (eDNA), particularly its temporal and spatial distribution and how this is shaped by abiotic and biotic processes. Here, we tested how seasonal variation in thermal stratification and animal habitat preferences influence the distribution of eDNA in lakes. We sampled eDNA depth profiles of five dimictic lakes during both summer stratification and autumn turnover, each containing warm- and cool-water fishes as well as the cold-water stenotherm, lake trout (Salvelinus namaycush). Habitat use by S. namaycush was validated by acoustic telemetry and was significantly related to eDNA distribution during stratification. Fish eDNA became “stratified” into layers during summer months, reflecting lake stratification and the thermal niches of the species. During summer months, S. namaycush, which rarely ventured into shallow waters, could only be detected at the deepest layers of the lakes, whereas the eDNA of warm-water fishes was much more abundant above the thermocline. By contrast, during autumn lake turnover, the fish species assemblage as detected by eDNA was homogenous throughout the water column. These findings contribute to our overall understanding of the “ecology” of eDNA within lake ecosystems, illustrating how the strong interaction between seasonal thermal structure in lakes and thermal niches of species on very localised spatial scales influences our ability to detect species.


eDNA metabarcoding

Sampling was conducted at the IISD Experimental Lakes Area (IISD-ELA), a remote research and monitoring facility in north-western Ontario, Canada. We sampled two lakes in summer and autumn of 2017 and repeated the summer and autumn sampling in five lakes in 2018. There are 14 species of fish across all the study lakes (mean 8, range 6-10 species per lake). All lakes have overlapping community compositions, including Salvelinus namaycush, a cold-water top predator, in every lake. Water samples were taken at six depths, dispersed vertically throughout the water column at the deepest centre point of each lake. The sampling points were distributed at six evenly spaced intervals, but because the lakes were different depths, absolute measurements differ between the lakes. We refer to the shallowest depth as sampling point one and the deepest depth as point six. Four 500 ml replicate water samples were taken per depth (for a total of 24 samples per lake per season) using an electrical pump and Jayflex PVC tubing (Winnipeg Johnston Plastics, MB, Canada) secured to a weight. In total, 336 samples were taken throughout the entire study (24 samples x 2 lake states x seven lake replicates). Water was filtered onto 47 mm 0.7 μm pore GF/F filters using an electric vacuum pump and filtering manifold (Pall Corporation, ON, Canada). The filters were immediately stored in screw-cap tubes at -20 ⁰C and then shipped on dry ice to McGill University, Montréal for molecular analysis. DNA was extracted from filters using the Qiagen Blood and Tissue kit. All samples were treated with the OneStep PCR Inhibitor Removal Kit (Zymo Research, Irvine, California). DNA was amplified in triplicate 12.5 µl reactions using 12S MiFish-U primers selected to target fish assemblages tagged with Illumina adapters. PCR replicates from each sample were combined and cleaned with a 1 : 0.875 ratio of AMPure beads. Samples were dual-indexed with v2 Nextera DNA indexes (Illumina). The samples were cleaned again with AMPure beads, quantified and equimolarised to 3 ng/μl for sequencing. Sequencing was conducted using 2 x 250 bp Illumina MiSeq at Génome Québec, Montréal. We used custom scripts to remove adapters, merge paired sequences, check quality and generate amplicon sequencing variants (ASVs). Samples were received as demultiplexed fastq files from Génome Québec. Non-biological nucleotides were removed (primers, indices and adapters) using cutadapt. Paired reads were merged using PEAR. Quality scores for sequences were analysed with FASTQC. Amplicon sequencing variants (ASVs) were generated using the UNOISE3 package, which uses a denoising pipeline to remove sequencing error and to cluster sequences into single variants (100% similarity). The full pipeline is available from After ASVs were generated, we assigned taxonomy using BLAST+ and BASTA, a last common ancestor algorithm. We used a custom reference database which contained only fish known to exist in the Lake of the Woods region (Ontario, CA), downloaded from the NCBI database on 12 August 2018.

Acoustic telemetry

Salvelinus namaycush were captured by angling and surgically implanted with coded, acoustic, pressure-sensing telemetry tags (model V13P-1L; Vemco, Innovasea, Bedford, NS). Between five and ten tagged adults were monitored in each lake during the study period. The pressure sensor on each tag was calibrated in the lake it was deployed in prior to implantation to ensure accurate depth readings (resolution: 0.08-0.15 m). The tags randomly emitted signals every 120-300 s (lakes 373, 626 and 239) or every 110-250 s (lakes 223 and 224). A number of data logging receivers (VR2W, 69 kHz; Vemco, Innovasea, Bedford, NS) were deployed under water at specific locations in the lake such that the “listening radius” of each receiver (spherical volume ~350 m diameter) overlapped slightly with the other receivers, resulting in maximum coverage of the lake. Each receiver was attached to a floating buoy and suspended ~2 m below the water’s surface or ~2-4 m above the bottom of the lake (dependent on mooring apparatus design). The receivers logged acoustic signals emitted by the tags through an omnidirectional hydrophone. Data (fish ID, date, time, pressure sensor reading) were continuously collected except when receivers were removed from the lake and downloaded (~8 h duration per lake, semi-annually). The pressure sensor data were converted to depth information using Vemco VUE software for each detection for the duration of the study (yielding ~200-700 depth detections for each fish in a typical 24-hr period). After downloading, duplicate detections (single tag signals detected by more than one receiver) were removed. In order to assess whether different time periods of cumulative eDNA persistence in the lakes affected the relationship between eDNA counts and telemetry data, we grouped telemetry data for each fish at different temporal scales, ranging from the day of eDNA sample collection, as well as one week, and one month prior to sample collection. The total number of detections of all fish were grouped into depth intervals reflecting the vertical distribution of the eDNA sampling (6 intervals per lake). We adjusted for varying depth interval size and variation in the total amount of telemetry detections for each lake over the relevant time period.

Usage notes

All instructions for use are found in the README file.


Canada Research Chairs, Award: 230517

Canada Research Chairs, Award: 237170

Mitacs, Award: IT08003

Natural Sciences and Engineering Research Council, Award: 04331-2017

Natural Sciences and Engineering Research Council, Award: 2016-04016

Natural Sciences and Engineering Research Council, Award: 523760-2018

Quebec Centre for Biodiversity Science

WSP Montreal Environment Department

Quebec Centre for Biodiversity Science

WSP Montreal Environment Department