Concentrations of environmental DNA (eDNA) reflect spawning salmon abundance at fine spatial and temporal scales

Highlights

• We measured salmon DNA in water samples collected daily during the spawning season

• This environmental DNA (eDNA) closely tracked the rise and fall of salmon numbers in the creek

• Concentrations of eDNA changed over fine spatial and temporal scales

• A statistical model including fish and environmental variables accounted for >75% of the variance in recovered salmon eDNA

• Both dead and live fish shed DNA into the water, but model results suggest that the shedding rate is much greater after death

Abstract

Developing fast, cost-effective assessments of wild animal abundance is an important goal for many researchers, and environmental DNA (eDNA) holds much promise for this purpose. However, the quantitative relationship between species abundance and the amount of DNA present in the environment is likely to vary substantially among taxa and with ecological context. Here, we report a strong quantitative relationship between eDNA concentration and the abundance of spawning sockeye salmon in a small stream in Alaska, USA, where we took temporally- and spatially-replicated samples during the spawning period. This high-resolution dataset suggests that (1) eDNA concentrations vary significantly day-to-day, and likely within hours, in the context of the dynamic biological event of a salmon spawning season; (2) eDNA, as detected by species-specific quantitative PCR probes, seems to be conserved over short distances (tens of meters) in running water, but degrade quickly over larger scales (ca. 1.5 km); and (3) factors other than the mere presence of live, individual fish — such as location within the stream, live/dead ratio, and water temperature — can affect the eDNA-biomass correlation in space or time. A multivariate model incorporating both biotic and abiotic variables accounted for over 75% of the eDNA variance observed, suggesting that where a system is well-characterized, it may be possible to predict species' abundance from eDNA surveys, although we underscore that species- and system-specific variables are likely to limit the generality of any given quantitative model. Nevertheless, these findings provide an important step toward quantitative applications of eDNA in conservation and management.

Introduction

All organisms shed bits of DNA into their surrounding environments, leaving behind a genetic footprint comprised of skin, scales, waste, and other tissues, collectively referred to as environmental DNA or eDNA. In aquatic environments, eDNA remains suspended in the water, where it can be collected, extracted, and sequenced/amplified, to reveal the habitat's species composition (Díaz-Ferguson and Moyer, 2014; Rees et al., 2014; Thomsen and Willerslev, 2015). Over the past decade, the use of eDNA technologies in the conservation and management of animal populations has evolved from a novelty to a valuable tool for scientists and resource managers (Jones, 2013; Kelly et al., 2014b). In comparison with traditional means of sampling taxa, eDNA technologies may be more efficient with regard to both time and money, and can overcome the practical and regulatory challenges associated with capturing rare or endangered species (Evans et al., 2017; Shaw et al., 2016). Indeed, in some cases, eDNA can generate substantially more information on species in less time than traditional survey methods (Ji et al., 2013; Valentini et al., 2015), such as alpha and beta diversity (e.g. Kelly et al., 2016). Furthermore, the rapid reduction in costs associated with genetic research over the past decades makes genetic approaches increasingly cost-effective.

At present, eDNA sampling is being applied extensively for the detection of invasive (Fujiwara et al., 2016; Jerde et al., 2011; Takahara et al., 2013) and endangered (Laramie et al., 2015a; Thomsen et al., 2012; Wilcox et al., 2013) species, and for the estimation of biodiversity across terrestrial, aquatic, and marine environments (Kelly et al., 2014a; Lodge et al., 2012; Port et al., 2016). These applications of eDNA all rely on its ability to detect the presence or absence of one or more species in a given environment. As such, the usefulness of eDNA methods remain limited because effective conservation of threatened species and management of exploited populations often requires estimates of abundance in addition to knowledge of presence and distribution. Consequently, development of eDNA sampling and analysis protocols that allow for estimation of relative or absolute abundance will greatly expand the applicability of the technology for conservation and management, and remains an important frontier in eDNA research (Kelly, 2016).

Recent examples of quantitative eDNA applications highlight both the promise and challenge of abundance estimation using these technologies. Kelly et al. (2014a) and Evans et al. (2016) both found positive correlations between eDNA concentration and biomass of multiple species in mesocosm settings using a multispecies metabarcoding approach. However, in both cases, despite known abundance and biomass, the relationships were generally weak and varied markedly between taxa. Targeted analyses using species-specific probes and qPCR avoid some of the potential biases of multi-species methods, and recent field studies have reported significant relationships between abundance or biomass and eDNA quantity using this approach. For example, Lacoursière-Roussel et al. (2016) found a weak, but positive relationship between catch per unit effort of lake trout (Salvelinus namaycush) and concentrations of the species' eDNA as detected using qPCR in several large lakes in Canada. Doi et al. (2017) detected a significant, positive relationship between snorkel-survey counts of the stream-dwelling fish Plecoglossus altivelis and eDNA concentrations in the Saba River, Japan. The direction of the relationship remained constant across seasons, but the slope increased markedly in the fall, highlighting the potential for temporal variability in the processes of eDNA production, transport or degradation. Buxton et al. (2017) also found a positive relationship between great crested newt (Triturus cristatus) abundance and eDNA concentrations in a series of ponds. However, they detected seasonal dynamics in eDNA concentration independent of abundance (likely a result of breeding activity). The results of these and other recent studies (e.g., Baldigo et al., 2017; Pilliod et al., 2013; Takahara et al., 2012) suggest that estimating abundance of aquatic organisms with eDNA approaches may indeed be possible.

Quantitative applications of eDNA would be particularly valuable in conservation of salmon and other anadromous fishes. Productive populations of these species are often exploited at high rates and managed using data-intensive approaches. Other populations are severely depleted and of significant conservation concern with many listed as endangered or threatened (Quinn, 2005). In both cases, estimates of abundance at various stages of the life cycle provide valuable data on survival, movement, habitat quality, and population productivity that directly inform management and conservation efforts. Traditional methods of salmon enumeration include weirs, counting towers, mark-recapture studies, float/walking/aerial surveys and hydroacoustics (Enzenhofer et al., 1998; Holt and Cox, 2008). These and other counting methods can be expensive and are often labor-intensive, limiting the spatial and temporal extent of monitoring by management agencies with finite budgets. The extent or resolution of monitoring efforts would be greatly increased if robust abundance estimates could be derived from eDNA samples, benefiting fish populations and the ecosystems, individuals, and communities that rely upon them. However, despite recent progress in quantitative applications of eDNA sampling, many uncertainties remain regarding the generation, transport and degradation of DNA in the environment (Andruszkiewicz et al., 2017; Barnes and Turner, 2015; Civade et al., 2016; Deiner and Altermatt, 2014; Jerde et al., 2016; Sassoubre et al., 2016; Shogren et al., 2016), and thus the spatial and temporal resolution of eDNA-based abundance estimates. For example, shedding rates may vary seasonally (Buxton et al., 2017) or with age (Maruyama et al., 2014) and degradation rates may be correlated with environmental factors such as temperature and microbial activity (Lance et al., 2017). Many of these issues are particularly acute for spawning salmonids occupying dynamic stream environments and undergoing rapid physiological, behavioral and morphological changes.

While the ultimate goal of quantitative eDNA methodologies is to estimate animal abundance from DNA concentration, the studies described above demonstrate such correlations to be both variable and uncertain. Reversing this model, and instead examining biological and environmental factors that influence the amount of DNA detectable in the environment is a key step in improving the predictive power of eDNA-based quantification. In this study, we sought to address some of these key uncertainties by exploring the quantitative relationship between eDNA and the abundance of sockeye salmon (Oncorhynchus nerka) in Hansen Creek, a small tributary of Lake Aleknagik near Bristol Bay, Alaska, USA. Counts of live and dead fish in the creek have been conducted daily during the spawning season – typically mid-July through August – for over 20 years. As such, the timing and spatial extent of the spawning run is well understood, and there is an established methodology for assessing the salmon population. Moreover, sockeye salmon are semelparous, so all adults die at the end of the spawning season rather than remaining in the stream, as would be the case for most fishes. These features, combined with physical characteristics making it amenable to visual surveys of salmon, make Hansen Creek attractive for exploring the relationship between eDNA and fish abundance in a natural setting. In order to examine the relationship between sockeye salmon abundance and the amount of sockeye salmon eDNA in the water we tested three hypotheses:

• eDNA does not substantially degrade or settle, or become entrained in sediment along the length of Hansen Creek, and therefore the furthest-downstream sampling site will consistently record the highest concentrations of eDNA.

• eDNA from tributaries will be additive, so that eDNA sampled on separate tributaries upstream of a confluence should approximately equal eDNA sampled just downstream of the confluence.

• eDNA concentration is linearly related to the total number of fish in Hansen Creek and inversely related to temperature.