Environmental DNA (eDNA) degradation influences the effectiveness of eDNA-based biodiversity monitoring, but the factors that determine the rate of decay of eDNA in terrestrial environments are poorly understood. We assessed the persistence of vertebrate eDNA from a mock vertebrate community created with soil from zoo enclosures holding ten target species from different taxonomic classes (e.g. reptiles, birds and mammals) and of different biomass (e.g. little penguin and giraffe). We examined species detection rates resulting from eDNA metabarcoding, as well as relative eDNA concentrations via qPCR, from soil samples over eight time points (0 to 12 weeks), during exposure to three ambient temperatures (10, 25 and 40 °C) and three levels of ultraviolet B (UV-B) radiation (0%, 50% and 100% intensity). We recorded considerable variation in detectability between species, independent of temperature and UV-B effects. Quantitative PCR indicated degradation of eDNA over time for all temperature and UV treatments, although it was still possible to detect eDNA from some species after twelve weeks. Degradation rates were lowest for high UV-B treatments, presumably due to UV-B reducing bacterial metabolism. The temperatures investigated did not influence eDNA decay. Our results indicate that eDNA in soil can persist under a range of temperatures and high UV radiation for longer than expected. Sheltered sites with minimal UV-B radiation, which have previously been considered ideal sites for terrestrial eDNA collection, may not be optimal for eDNA persistence in some cases due to microbial decay. A better understanding of eDNA degradation in terrestrial environments is needed before eDNA metabarcoding of soil can be used as a stand-alone method for accurate surveying of terrestrial vertebrate communities.

Sample source

Soil was collected in mid-July 2021 from the enclosures of ten species at the Perth Zoo, South Perth, Western Australia. Soil samples (1.5 L), consisting of five equal subsamples from areas of recent animal activity (as determined by zoo staff) to maximise the likelihood of eDNA presence in a sample, were collected from each of four mammal, three bird and three reptile enclosures (Table 1): approximately 15 L of soil in total. Samples were taken from a depth of 0-50 mm, using sterilised sampling equipment and disposable nitrile gloves. Soil texture varied from sand (e.g. Cape porcupine) to sandy clay loam (e.g. Galápagos tortoise).

Target species (Table 1) did not occur in the Perth metropolitan region, meaning that any DNA detected from species in the samples must be derived from the original soil samples, not reintroduced through subsequent contamination. The emu (Dromaius novaehollandiae) was a possible exception, though contact with DNA from this species was deemed to be highly unlikely between sampling events.

A mock community of zoo vertebrate eDNA was created by homogenising the ten soil samples using a combination of soil sieves and mixing. Equal aliquots of the resulting soil mix were placed in aluminium trays (26.5 x 19.5 x 6.5 cm; approximately 550 mL per tray at 10 mm thickness) and exposed to three levels of ambient temperature (10, 25 and 40°C) and three levels of UV light intensity (0%, 50% and 100%), with three replicates for each treatment (27 trays in total). Modified plant growth cabinets (Arcus Refrigeration) maintained constant temperature and UV conditions for the duration of the experiment. Ultraviolet light was produced by 26-watt ExoTerra Reptile UVB200 bulbs, producing approximately 240 µW/cm² and 40 µW/cm² of UV-A and UV-B respectively (ExoTerra, 2021).  Temperature levels were selected to represent typical ambient air temperatures occurring in terrestrial settings across a variety of Australian climatic zones, after Jones et al. 2009. iButton temperature loggers showed that the temperature of soil in treatment trays was in equilibrium with climate control chamber ambient temperatures. Clear plastic wrap was used to cover the 100% UV treatment trays and did not affect UV transmission (measured using UV-vis spectrometry - see Supplementary Figure 1). 50% UV treatment trays were covered with clear plastic wrap and 50% UV-block shade cloth. Zero UV light treatment trays were covered with aluminium foil. All trays were well sealed on all edges between sampling points. Soil samples were collected in 50 mL falcon tubes from the surface soil (<10 mm) of each tray as the experiment commenced (Time 0), with subsequent sampling at 6 h, 24 h, 48 h, 1 week, 3 weeks, 6 weeks and 12 weeks later. Each sample was taken from a different area of the tray. Samples were stored at -21.0 °C until DNA extraction.

Sample Processing

DNA extraction

Soil DNA extraction and PCR were conducted at TrEnD (Trace and Environmental DNA) facilities at Curtin University, Perth, Western Australia. Samples were homogenised using a Qiagen TissueLyser, and DNA extracted from 250 mg of soil using the DNeasy PowerLyser Powersoil kit (Qiagen), followed by the Qiacube extraction platform (Qiagen). Extracts were kept at -20.0 °C until use. The final elution volume was 100 μL, and extraction controls (blanks, n = 10) were processed for every set of extractions.

PCR amplification

Quantitative Polymerase Chain Reaction (qPCR; Applied Biosystems, USA) was used to assess the quality and quantity of the DNA extracts and determine the optimum DNA input (Murray et al 2015).  The vertebrate-specific metabarcoding assay 12S-V5 (Riaz et al., 2011) was used, targeting a 98-bp region of the mitochondrial 12S gene with the primers F2 (5’-TAGAACAGGCTCCTCTAG-3’) and R2 (5’-TTAGATACCCCACTATGC-3’). Extracts were amplified on a StepOnePlus Real-Time PCR System (Applied Biosystems) under the following conditions: 95 °C for 5 minutes, and 50 cycles of 95 °C for 30 seconds, 58 °C annealing temperature for 30 seconds and 72 °C for 45 seconds. A melt curve of 95 °C for 15 seconds, 60 °C for one minute and 95 °C for 15 seconds was used, ending with ten-minute elongation at 72 °C. Each 25 µL qPCR mix contained 2.5mM MgCl₂ (Applied Biosystems), 1× PCR Gold buffer (Applied Biosystems), 0.25 mM dNTPs (Astral Scientific, Australia), 0.4 mg/mL bovine serum albumin (Fisher Biotec, Australia), 0.4 μmol/L forward and reverse primer, 1 U AmpliTaq Gold DNA polymerase (Applied Biosystems) and 0.6 μL of a 1:10,000 solution of SYBR Green dye (Life Technologies, USA). Cycle-threshold (C_t) values were recorded for all samples and dilutions. PCR reactions were performed with two dilutions (3 PCR reactions per extract at neat, 1:10 and 1:100), and the optimum level of DNA input was used for fusion tag PCR. Extraction controls, positive controls (Gallus gallus) and non-template controls were included in qPCR runs.

All PCR mixes were prepared in a dedicated laboratory to minimise contamination, and DNA extracts were added in a separate laboratory in dedicated UV cabinets. Samples containing amplifiable DNA as determined by qPCR were assigned a unique combination of fusion tag primers, containing unique multiplex identifier (MID) tags with lengths between 6 and 8 bp, the 12S-V5 primer and sequencing adapters (Illumina). A single step fusion protocol was used with no reuse of index combinations, and samples were amplified in duplicate using the same conditions as the qPCR protocol described previously, reducing the effect of PCR stochasticity (Murray et al., 2015). Duplicate samples were pooled alongside amplicons with similar ΔRn values. Pooled amplicons were cleaned using the QIAquick PCR purification kit (Qiagen) and quantified using the QIAxcel Advanced system (Qiagen).

DNA library preparation and sequencing

A DNA library was created by combining pools in equimolar ratios based on the quantification, and then size selected (160-400 bp) using Pippin Prep (Sage Sciences) to remove MID-tag primer-dimer products formed in amplification. The final library was cleaned using a QIAquick PCR purification kit (Qiagen) and eluted to 50 µL. A QuBit fluorometer (Thermo Fisher Scientific) was used to quantify the final DNA library, before sequencing on Illumina’s MiSeq platform (Illumina, San Diego, CA, USA) using a single-lane flow cell as per Illumina protocols for single-end sequencing (MiSeq v2 Reagent Kit 300 Cycles).

Sequence analysis

Raw sequence reads were processed using the eDNAFlow pipeline (Mousavi-Derazmahalleh et al., 2021). Sequences were quality-filtered using AdaptorRemoval v2 (Schubert et al., 2016) and FASTQC (Andrews, 2010), and demultiplexed using OBITools (Boyer et al., 2016). Reads were dereplicated, denoised and assigned a zero-radius operational taxonomic unit (ZOTU) using Usearch (Edgar, 2010), with sequences shorter than 60 bp removed. Taxonomic assignment was conducted by matching ZOTU sequences to Genbank, an online reference database (https://www.ncbi.nlm.nih.gov/genbank/), using a basic local alignment search tool (BLASTn) run on a high-throughput HPC cluster (Pawsey Supercomputing Centre Perth, WA, Australia). Final taxonomic assignments were conducted using a custom lowest common ancestor (LCA) algorithm in Python (Mousavi-Derazmahalleh et al., 2021), requiring 95% query coverage and 90% identity match with reference sequences. Where the LCA script was unable to assign a ZOTU to a species level or the species detected was not found in the study area, genus level was reported, and the ZOTU assessed manually based on known occupants of both the Perth Zoo and surrounding area. Samples with a low sequencing depth (less than 5000 total reads, as determined by rarefaction curves; Supplementary Figure 2) were removed from the data, and individual ZOTU detections with less than two sequence reads were removed. Due to the minimal contamination risk associated with single-fusion protocols through physical separation of pre and post-PCR areas, good laboratory practices and no reuse of fusion tag combinations (Taberlet et al., 2018), any detections above two identical sequence reads were considered ‘real’ detections. ZOTUs detected in positive or negative controls, as well as those from common PCR contaminants (e.g., human, ungulate species) were removed prior to analysis. C_t values from qPCR runs were recorded for each undiluted DNA extract and compared for sampling time and temperature and UV exposure. These values corresponded to the relative concentrations of vertebrate eDNA within each extract. Lower C_t values (less cycles to amplify DNA over a threshold) represented higher relative DNA concentrations, whilst higher C_t values signified lower relative DNA concentrations.