Fully terrestrial eDNA sampling approaches offer a potentially powerful addition to biodiversity monitoring efforts23,24. However, protocols for using eDNA-based methods to characterize terrestrial biodiversity, and vertebrate communities in particular, are still nascent27,28,30. In this study, we show for the first time that an eDNA metabarcoding approach can be used to broadly characterize tree-dwelling mammal communities by sampling tree trunks and surrounding soil. Our findings add to recent work (e.g., for reptiles22,24) showing that surface eDNA collection methods, which are relatively untested compared with soil-based eDNA methods, can also be effective at detecting terrestrial vertebrates. Further, we demonstrate that supplementing metabarcoding detection with qPCR-based methods can greatly improve sensitivity, a potentially important consideration for monitoring schemes focused on rare taxa (e.g., Refs.11,12). Together, our results have significant implications for global biodiversity conservation as the broader guild of arboreal vertebrates includes highly threatened5,48,49, as well as invasive alien species50, that are often cryptic, inhabit inaccessible locations, and are therefore challenging to monitor.
Our methods captured over 60% of the mammalian diversity expected at the sites, and a similar fraction of the subset of arboreal species, despite sampling only 21 trees. Species accumulation curves suggest that more species would likely have been added with increased sampling effort. These results broadly agree with those of Leempoel et al.23 who found that soil eDNA metabarcoding well characterized mammal communities in California chaparral. However, in both our study and that of Leempoel et al.23, some conspicuous absences were evident. Bats comprised all of the arboreal species that we expected but failed to detect at our sites using metabarcoding (Fig. 1A). Leempoel et al.23 also noted a lack of bat detections (2 of 14 possible taxa detected), which they suggested could be due to low efficiency of either the 12S primer set or of their soil sampling methods for that order. While both reasons could also apply to the lack of bats detected in our study (discussed further below), the performance of the 12S primer set very likely contributed to our lack of American black bear detections as MiMammal-U primers are known to be ineffective at amplifying bear DNA38. These challenges highlight the reality that false negatives and varying detectability among species are common issues to all survey approaches, including eDNA metabarcoding. Our study represents a rare example among metabarcoding studies in that it uses repeated sampling and community occupancy models to quantify false negative rates. This quantitative approach, coupled with continued experimentation with different molecular techniques and survey methods (e.g., Refs.23,27), will be vital to helping researchers decide how eDNA metabarcoding methods will fit into existing biodiversity monitoring efforts moving forward.
Although our results suggest that sampling for tree-roosting bats using eDNA metabarcoding still requires further research and optimization, our approach likely has application to characterizing communities in a much broader range of arboreal species globally. Geographic regions with multiple elusive arboreal mammals of management interest—for example, gliders and tree kangaroos in Australasia, or primates in the global tropics—may be particularly suited for a metabarcoding approach for community-level assessments4,8,9,49. It may be especially useful for rapid biodiversity assessments (RBAs51) in remote forested environments, where the ability to collect multiple samples relatively rapidly without regard to time of day would be a key advantage27. Existing survey methods to monitor arboreal mammals tend to be optimized for particular groups of species, often segregated by body size and behavior, with no suitable single method available to characterize all members of the guild4,8,9,16,49. Diurnal and nocturnal species, for example, often require separate survey methods or timing8. While camera traps capture both diurnal and nocturnal species, they typically miss smaller species16,23. The need for multiple methods to survey for nocturnal and diurnal, or large and small, species separately raises the cost of sampling and can result in datasets that are difficult to compare across sites because of inherent sampling biases8. Excluding bats, we found encouraging results for both diurnal and nocturnal arboreal species of a broad range of body sizes, detecting all seven expected species (Fig. 1A). While more work is needed to assemble robust genetic reference libraries before global arboreal mammal monitoring with eDNA metabarcoding will be broadly feasible, a clear advantage of the technique remains the power to detect a broad swath of species, with widely varying morphologies and behaviors, with a single method23,27,28,51.
The promise of eDNA metabarcoding approaches for at least some arboreal guilds is well illustrated by our results for southern flying squirrel, Glaucomys volans. Like other flying squirrels (Tribe: Pteromyini), this species is strictly nocturnal, highly arboreal, and tends to get injured in live traps, making it difficult to directly observe and monitor48,52. Yet G. volans eDNA was readily detectable using metabarcoding in our study, occurring in 19–26% of soil samples and 47–52% of roller samples across both sites. Our similarly encouraging results for detecting other squirrels (Sciuridae) also bode well for management applications. For example, the methods would enable fine-scale mapping of habitat use in places such as the United Kingdom where native red squirrels (Sciurus vulgaris) are outcompeted by eastern gray squirrels, or the Delmarva peninsula (USA) to support the conservation efforts for the Delmarva fox squirrel (Sciurus niger cinereus)53. Further research is needed to determine the extent to which our results for squirrels generalize to other taxa with similar active tree-climbing lifestyles (e.g., gliders4, primates49).
Our finding that soil samples revealed fewer species, had lower detection probability, and had lower read counts than roller samples, even for some non-arboreal species like white-tailed deer, likely reflects multiple factors. First, soil and tree bark represent markedly different biological and chemical environments that likely differ in eDNA quantity by species, eDNA persistence rates54,55, and microorganism abundance. The latter may be especially pertinent to our study as we observed a relatively large drop in the number of reads after removal of microorganism reads, especially for soil samples. This suggests that performing additional purification steps prior to sequencing could boost the ability of both methods, and especially soil eDNA, to detect target species by increasing mammalian sequencing depth. Other in-lab factors, such as method of extraction23 or choice of primers, similarly have the potential to influence the recovery and amplification of target species’ DNA and should be the focus of future research.
Next, our focal trees were not chosen to occur near any special attractants or areas of multi-species use, such as saltlicks or water sources, which has proven successful in other vertebrate eDNA studies18,25,31,32,56. It is possible that adding a broader range of soil sampling sites, including some targeted towards other guilds (e.g., burrow users32,56), would have yielded a more complete inventory. Nevertheless, both soil and surface methods have advantages over the much more commonly-used metabarcoding approaches that rely on natural water bodies for assessing mammal communities16,17,27,28,29,30,38 as they are not limited to where these features occur. Our study is the first to suggest that surface eDNA metabarcoding methods can be a powerful supplement to established soil-based methods of characterizing mammal communities, especially for arboreal species.
As noted, bats were especially lacking from our eDNA metabarcoding results, with only two of six likely species detected. Notably, our metabarcoding species list lacked two of the bat species that our sampling scheme was designed around (eastern red bat and northern long-eared bat) and for which we had confirmed recent presence at the sites (Table 1). The lack of northern long-eared bat detections may directly relate to recent precipitous population declines (~ 99%) caused by white-nose syndrome57. However, the lack of eastern red bat detections was especially surprising as roosting of this species was suspected based on telemetry in 17 of our 21 target trees. Reasons for this omission may relate to the fact that eastern red bats roost singly on small twigs and in leaf clusters, and therefore may not leave much DNA on tree trunks. Another possibility is low efficiency of the 12S primer set for bats, although we were unable to find information about this in the literature. It is notable that Leempoel et al.23 had a similarly poor representation of bats with comparable soil-based methods. However, our metabarcoding results did indicate that we are capable of detecting even uncommon, or at least unexpected, bat species with our methods. Eastern small-footed bat, which is typically viewed as a rock-roosting species and is considered endangered by the International Union for Conservation of Nature (IUCN)58, was detected in both soil and surface eDNA samples from Morristown National Historic Park. This species was not otherwise confirmed as present at the site until a year later, in spring 2022, when it was caught in a mist net (BM, unpublished data). Our results with respect to bat detections, along with those of others23, underscore the need for further research to adapt eDNA metabarcoding methods to this vulnerable group, which could contribute much needed demographic and distribution information. This is especially urgent as 18% of bat species are listed as “data deficient” by the IUCN, while 57% lack basic population trend information5,58.
Our comparison of qPCR to metabarcoding detection methods for big brown bat represents a hopeful result for the use of eDNA to monitor rare vertebrates that are of particular conservation interest. It is well-known that qPCR-based eDNA surveys targeted towards individual species return higher detection probabilities and have greater power at low abundance, than metabarcoding approaches59. Our results agree, showing for the first time that adding a qPCR step in the analysis of surface and soil eDNA samples can be effective for detecting bats in forested environments. The addition of a qPCR step opens the door for developing species-specific assays to increase detection power for endangered or elusive bat species, or other cryptic arboreal mammals49. Emerging molecular detection approaches such as droplet digital PCR have the potential to increase this sensitivity even further59. Like other eDNA-based tools and survey tools in general, careful consideration of sampling effort, the natural history of target species, and the configuration of different field and molecular methods will be key to optimizing our approach to characterize mammal communities, or to target a particular species, in different regions.
Although eDNA surveys are not inexpensive given the need for both fieldwork and molecular analyses, they can be cheaper than conventional approaches, especially if such approaches require many hours of fieldwork or expensive equipment60. Thus, the relative cost-effectiveness of surface or soil eDNA surveys will depend heavily on the mammal communities of interest, the mix of methods that must be employed to effectively sample them, and the purpose of the sampling efforts. However, even if costs are increased, eDNA surveys can reduce field time to the extent that they can improve detection rates, either by replacing or supplementing conventional sampling methods (e.g., as a supplement to visual observations). With higher detection rates, fewer visits are required to achieve the same results. This operational efficiency would be especially advantageous when field conditions present safety risks, are intrusive to sensitive habitats, or are challenging to access. For example, adding surface eDNA sampling to existing visual surveys of eastern wood rats (Neotoma floridana), a cryptic mammal that inhabits steep, rocky slopes in the eastern US, could likely increase detection power, thereby reducing the need for additional risky and costly sampling visits. More studies involving direct comparisons among methods (e.g., Refs.23,24,30,60), in a variety of ecoregions, are needed to determine the extent to which incorporating our methods into existing vertebrate monitoring workflows would increase efficiency.
Finally, we detected other vertebrates, including seven birds and one salamander, in soil and surface eDNA samples, despite our use of a mammal-specific primer set. This is similar to results from Leempoel et al.23 in California using the same primer set, in which six bird species were detected. We found that surface eDNA detected more bird species than soil, perhaps for the same reasons as for mammals (above). Our results provide evidence that surface eDNA surveys, with taxon-specific primers, could be used to survey bird communities, or used to target particularly rare species in forested ecosystems (e.g., Ref.61). Our detection of a salamander, coupled with recent promising research into reptile detection using surface eDNA methods22,24 suggests a broader potential for applications with other vertebrates as well. Finally, both surface and soil eDNA metabarcoding can be expanded beyond forests, providing insight into their effectiveness in other habitats (e.g., caves17 or talus slopes). Our study and others highlight that the potential of coupling surface and soil eDNA methods for detecting and monitoring mammalian biodiversity, and terrestrial organisms generally, has yet to be fully realized.
Source: Ecology - nature.com
