Evaluating the utility of environmental DNA for detecting a large Critically Endangered lizard in tropical northern Australia

Introduction

Rare or elusive species can be challenging to detect using traditional survey methods, particularly when they are found in remote areas, where surveys are logistically complex, time-consuming, and expensive. The use of environmental DNA (eDNA) presents a promising tool for species detection, involving the collection and analysis of DNA that is shed into the environment through the excretion of bodily fluids, skin, and hair. eDNA can be extracted from water, soil, or even air samples, allowing for non-invasive detection of species without the need for visual sightings or trapping (Thomsen et al. 2012; Taberlet et al. 2018).

The application of eDNA methods in wildlife conservation and management is growing rapidly (Beng and Corlett 2020; Nordstrom et al. 2022). Two main approaches include: (1) metabarcoding, which uses general primers to amplify DNA from multiple species simultaneously; and (2) species-specific primers, which target DNA specific to a single species. Species-specific approaches can be more sensitive and are useful for targeted species surveys (Beng and Corlett 2020). However, they require reference sequences to create a reliable assay, making them unsuitable for species with little or no genomic data. While generally less sensitive, metabarcoding can detect rare or elusive species from a single sample (Holdaway et al. 2017) and offers a diverse range of applications for biodiversity monitoring, from specific species groups to wider biodiversity studies (Ruppert et al. 2019).

The recent advancements in eDNA methods may offer a more cost-effective alternative to traditional survey methods (Fediajevaite et al. 2021). This potential allows for rapid and economical surveying of expansive areas, reducing the need for extensive and labour-intensive fieldwork, particularly when dealing with rare or elusive species that require specialised expertise or significant survey effort (Biggs et al. 2015; Davy et al. 2015; Akre et al. 2019). However, the application of eDNA methodologies requires initial investment in quality control and assurance. For example, it is crucial to address the challenge of discriminating between false positives and false negatives for ensuring the accuracy and reliability of eDNA-based assessments through rigorous sampling protocols and validation techniques (Ficetola et al. 2015).

Here, we piloted the use of eDNA metabarcoding to detect the Arnhem rock skink (Bellatorias obiri), an extremely cryptic lizard found in rugged and remote rocky gorges and boulder fields. The species is listed as Critically Endangered (Gillespie et al. 2018), and is considered one of Australia’s most imminently threatened vertebrates (Garnett et al. 2022). The Arnhem rock skink is a large (snout-vent length ~200 mm), robust skink with a restricted distribution centred on Kakadu National Park, in Australia’s Northern Territory (Jolly et al. 2023). Detecting the Arnhem rock skink via active searches or live trapping poses significant challenges due to its elusive behaviour, rugged habitat, and remote distribution. Since scientists first became aware of the species in 1974, there have been fewer than 40 confirmed records of the species, primarily incidental, underscoring the difficulty and time-consuming nature of this endeavour (Cogger 1974; Sadlier 1990; Armstrong and Dudley 2004). In the 50 years since, there has been up to a decade between detections of live specimens, despite considerable survey effort. Given the difficulties of detecting this elusive species using traditional search methods, combined with the challenges and costs of accessing and surveying in a remote location, eDNA was a desirable method to trial for the detection of B. obiri.

By using eDNA sampling, we aimed to assess the effectiveness of detecting this Critically Endangered lizard species from water and soil samples collected from the only confirmed location where the species was known to persist. Our study trialled a metabarcoding approach as no reference sequences existed for the target species, preventing the design of a species-specific assay. If effective, this method could be employed more widely to survey areas with historical records of the species and to search new potential habitats. Given that eDNA is not yet widely used in vertebrate surveys in northern Australia, we suggest that our results will also provide valuable information for agencies seeking to efficiently and cost-effectively survey for B. obiri and other rare or cryptic reptile species.

Materials and methods

Sample collection

In October 2022, we collected samples from the only site where Bellatorias obiri was known to persist; a spring-fed, rocky gorge in Kakadu National Park, Northern Territory (approximately 13.6°S, 132.3°E), where it inhabits deep rock crevices (Fig. 1). We collected both soil and water samples, as these materials have previously been used to detect skink eDNA (Kyle et al. 2022; West et al. 2023).

Fig. 1.

Location of the study site (black circle) in Kakadu National Park, Northern Territory, Australia and the study species, the Arnhem rock skink (Bellatorias obiri) (inset). Photo credit: C. Jolly.

We collected two soil samples from each of 12 rock crevices, where pre-existing camera trap data allowed for quantification of B. obiri activity levels (NT Government and Kakadu National Park, unpublished data). Bellatorias obiri activity was defined as the average number of motion-trigger camera images in which the species was detected in a crevice during the preceding build-up and wet season (from August 2021 to February 2022). To investigate how effective eDNA methods may be for detecting the species at different levels of activity, we stratified the sampled crevices into three categories based on the number of detections from camera trapping: (1) high activity; (2) low activity; and (3) no activity (see Supplementary material S1, Table S1). An average was used as each crevice had one or two cameras that were operational for the entire sampling duration. We arbitrarily defined crevices with high activity as those with >80 detections, low activity as crevices with 15–29 detections, and no activity as crevices with zero B. obiri detections. Soil samples were collected from rock crevices while wearing clean gloves and using sterile containers for each sample. A new clean plastic spatula was used to scrape the upper soil layer into a 50-mL centrifuge tube. Tubes were not filled greater than halfway (25 mL). The lid of the tube was screwed on and then wrapped with parafilm. Tubes containing soil samples were immediately placed into a cooler with ice bricks and then frozen at −20°C within eight hours.

We also collected two water samples from each of three waterbodies that were adjacent to the crevices where B. obiri was known to occur (Main Pool, Upper Pool, and Lower Pool). All three waterbodies had little flow at the time of collection, as sampling occurred in October, which is late in the dry season during the build-up. An incidental scat sample was also collected from a crevice where B. obiri had been detected recently (in the preceding 9 months). A reference tissue sample from a B. obiri specimen (stored in ethanol at room temperature) was obtained from the Museum and Art Gallery of the Northern Territory (Museum Sample ID: TS0052) and was included to validate that the assay used would successfully amplify B. obiri DNA from a higher quality sample (refer to Laboratory analysis below and Supplementary material S2 for molecular methods).

Water samples were also collected while wearing clean gloves and using sterile plastic containers. Containers were initially rinsed internally by half-filling with site water, and were then capped and shaken, before the water was discarded away from the sample collection site. Water samples (1 L) were collected by submerging the container below the water surface, and then containers were sealed and placed immediately into a cooler with ice bricks. Water samples were filtered within 4 h of collection in the field using 0.45 μm mixed cellulose ester with a Sentino peristaltic pump (PALL Corporation, Port Washington, USA) to capture eDNA present in the water. Each sample was filtered for up to 30 min and the volume of water filtered for each sample was recorded. The filter was then removed and folded in half, with the filtrate on the inward side, and placed in a labelled sample bag using clean forceps. Filter papers were stored at −20°C. Common equipment (e.g. forceps) was cleaned with a 10% bleach solution for 10 min and then rinsed thoroughly before re-use. A blank control sample from the cleaning solution was also processed and sent for analysis to detect any cross-contamination.

All samples were stored at −20°C before and after transportation to the eDNA Frontiers Laboratory at Curtin University, Perth, Western Australia, for preparation, DNA extraction, sequencing, and bioinformatics analysis.

Results

We retained 2,689,909 reads after the initial read filtering steps, producing 1090 taxonomic units. Of these, 56 matched to a GenBank entry at ≥95% similarity after clustering. Once the filters for non-eukaryotes, environmental contaminants, and non-vertebrates had been applied, and duplicated taxa were merged, we retained 16 genus or species level taxonomic units (including the museum-collected Arnhem rock skink sample). These included taxa from five animal classes, including 12 vertebrate genera in the water samples (six samples from three pools), and five vertebrate genera in the soil samples (24 samples from 12 crevices) (Table 1). Of the 15 taxa identified from the water or soil samples, we were able to identify 10 taxa to species, including a range of non-target vertebrate groups (mammal, bird, amphibian, and fish species) (Table 1).

Despite successfully amplifying B. obiri DNA from the tissue and incidental scat samples, we were unable to detect the DNA of our target species, B. obiri, in any of the in situ water or soil samples. Laboratory extraction controls were all negative and no contamination was found in the field blank sample. The only reptile detected in the water or soil samples was Macleay’s water snake (Pseudoferania polylepis), which was identified in a single water sample. The introduced cane toad (Rhinella marina), which is widespread and common in the region, was detected in two of the three sampled waterbodies.

Discussion

We trialled eDNA metabarcoding for detecting the Critically Endangered B. obiri in soil and water samples from a site known to be occupied by the species. Our aim was to trial the approach before applying it more widely as a detection method across the species’ known historical range in Kakadu National Park. While the species’ DNA was successfully identified from a preserved B. obiri tissue sample and a suspected scat sample, it could not be detected in any of the soil or water samples collected in the field. It is acknowledged that a species-specific qPCR assay is a more sensitive technique than metabarcoding for detecting a single target species and therefore may have displayed improved performance in this study; however, this technique requires a comprehensive reference sequence database (ideally spanning the genetic diversity of the target species) in order to design a robust assay, and reference sequences are not currently available for B. obiri. Additionally, the eDNA method trialled here did not detect nine additional reptile species that were known to be present in the crevices in the preceding 2 months (i.e. detected by camera trapping; see Supplementary material S1, Table S2). The lack of the target species’ DNA in the soil and water samples, as well as the lack of DNA from any other terrestrial reptile species, suggests that the eDNA techniques trialled here may not be a reliable tool for detecting terrestrial reptiles in the study region using these types of substrates and the 16S rRNA fragment that we used.

Although possible, it is unlikely that B. obiri was inactive at the crevices in our study area between the time of camera trapping and substrate sampling (~7 weeks). Subsequent camera surveys documented use at all high activity crevices in the weeks immediately following the eDNA sampling, including activity at one crevice on the same day as the camera trap deployment (3 days after the eDNA sampling was conducted). This supports that our results are primarily due to false negative errors. One likely explanation for this is that B. obiri, like many other terrestrial reptiles, sheds little DNA into the environment (Nordstrom et al. 2022). Previous studies have suggested that eDNA detection success rates may be lower for animals with hard keratinised outer layers, such as reptiles, due to differences in DNA shedding rates (Adams et al. 2019; Allan et al. 2021). However, Matthias et al. (2021) successfully detected DNA of a small snake in soil to which an individual was exposed for only 10 minutes, and terrestrial skinks have been detected in surface water samples from north-eastern Australia (West et al. 2023), suggesting eDNA methods can be effective in some circumstances.

Other explanations for our results may be that our choice of target DNA fragment was inappropriate, or that high levels of microbial activity and UV solar radiation in tropical environments results in rapid degradation of eDNA, reducing the chances of successful detection (Zulkefli et al. 2019). However, as our chosen DNA fragment has been widely used for eDNA metabarcoding in a variety of taxa and we were able to detect numerous non-target species, we suggest that these explanations are less relevant.

We were unable to conduct an ex situ experiment using a captive population as no individuals exist in captivity and the species was only known from a single site. However, further research using captive individuals of one or more surrogate skink species and different substrates and target DNA fragments may help to improve the use of eDNA methods for detecting reptiles in northern Australia. For example, swabbing for surface DNA on artificial cover objects has successfully detected terrestrial reptiles (Matthias et al. 2021; Kyle et al. 2022). Kyle et al. (2022) found surface swabbing using a paint roller performed better than collecting soil for detection of a terrestrial skink. This ‘roller’ method could be used to sample larger surface areas of rock crevices or from artificial surfaces (e.g. metal) that may have higher detection probability (Kyle et al. 2022). However, the cost of alternative eDNA trials should be considered versus other potentially viable detection methods, such as camera trapping.

Our findings highlight the challenges of applying eDNA metabarcoding techniques for detecting rare or cryptic terrestrial reptiles – in this case, the Critically Endangered lizard, B. obiri – emphasising the need for further refinement and exploration of this or alternative methods to improve species detection in terrestrial reptiles and tropical environments.

Supplementary material

Supplementary material is available online.