In this study we aimed to assess the genetic diversity of the contemporary Scottish pine marten population using microsatellites. Our goal was to gain insights into the impact past declines have on the current population, which is steadily expanding and recolonising its historical range. This is crucial for informing future conservation management and supporting the ongoing recovery of the species in Britain.
In this study, genetic diversity was assessed using a panel of 12 microsatellite markers. These markers, carefully selected and optimised over time for their variability and ability to amplify pine marten DNA from poor-quality samples, such as hair from non-invasive genetic studies (e.g. Mullins et al., 2010; Croose et al., 2019; Twining et al., 2022), are arguably biased towards this variability. Although microsatellite markers are a cost-effective and reliable method for assessing genetic diversity in various species, comparisons with SNP data from restriction-site-associated DNA sequencing (RADseq) suggest that SNPs may provide deeper historical population insights. This is due to their slower mutation rate compared to microsatellite regions, making them particularly useful in small, isolated populations with low diversity. Indeed, many newer studies, some of which are discussed below, take a genomic approach to provide genetic diversity assessments of recovering species such as the wolf (Canis lupus) (vonHoldt et al., 2023). While the cost implications and overall benefits of transitioning from a well-developed microsatellite panel should also be considered (Lemopoulos et al., 2019; Hauser et al., 2021), a genomic approach could be considered for future monitoring of the pine marten. The sampling method employed in this study was ad hoc, and some samples, now 10 years old, provide a genetic snapshot that is likely out-of-date, given the rapid expansion of the species. However, this study could serve as a valuable foundation for future studies, which should aim for more systematic sampling across the species' range.
In terms of nuclear genetic diversity, the number of alleles across the selected microsatellite loci in the Scottish population averaged 3.5, with Ho and He averaging 0.503 and 0.567. These figures are slightly lower than the values reported for the pine marten in Ireland, where the number of alleles averaged 4.3 and observed and expected levels of heterozygosity averaged 0.547 and 0.549, respectively (O’Reilly et al., 2021). Only the expected levels of heterozygosity are slightly higher in the Scottish population. O’Reilly et al., (2021) used 11 out of 12 of the same microsatellite markers, making comparisons reliable across these two studies. Similar levels of genetic diversity were noted in pine marten populations recorded in Spain (Ruiz-González et al., 2014), Denmark (Pertoldi et al., 2008), and France (Mergey et al., 2012). In North America averages of Ho and He were similar for M. americana (Ho = 0.57, He = 0.58) and M. caurina (Ho = 0.57, He = 0.56) (Lucid et al., 2020), suggesting that the results obtained in this study are similar across the range of this species and similar species’ ranges.
Notably, genetic diversity in the Cairngorms (Ho = 0.44, He = 0.42) and Dumfries and Galloway (Ho = 0.45, He = 0.44) was lower than average. This was anticipated in Dumfries and Galloway, where 12 pine martens were translocated from the Highlands in the 1980s, likely leading to genetic isolation. Genetic differentiation tests confirmed this region's uniqueness, as indicated by its early separation in the STRUCTURE analysis and differentiation in the PCOA. Significant Fst values further underscored this isolation, particularly when compared with other sampled zones. However, these values decreased closer to the Scottish Borders and Central Scotland, indicating some gene flow. Surprisingly, the highest Fst values were observed with the Highlands, the source of the translocation, suggesting genetic drift has made this population genetically distinct. STRUCTURE plots revealed admixture in some individuals at K = 3, indicating genetic mixing between animals from Dumfries and Galloway, Highlands, Central and Borders, but it remains unclear whether this reflects past or more recent gene flow. The genetic isolation observed in the Cairngorms, also supported by differentiation tests, may be attributed to geographical barriers, such as high terrain, the River Spey, and several lochs, alongside the presence of the A9 roadway, causing further isolation. Indeed, the rugged terrain of the Scottish Highlands, characterised by mountainous landscapes, numerous lochs (lakes), deep glens (valleys), and expansive moorlands, might have contributed to the fragmentation of species across Scotland. This fragmentation could have led to the formation of distinct genetic clusters, which are now mixing as individuals recolonise Scotland, producing a genetically mixed population in many areas.
Our results, analysed at individual level, suggest evidence of very weak isolation by distance within the study population, with geographically closer individuals tending to be more genetically similar. There was no association between genetic differentiation and geography when we analysed the data according to sampling zone, and it is likely that most of the genetic differentiation is influenced by other factors, which could include reintroductions, opportunistic ad hoc sampling selection, migration, genetic drift, or non-random mating patterns. Consequently, it is unclear how this genetic fragmentation of the population occurred. Beyond the known translocation to Dumfries and Galloway, there have been unofficial releases in the Scottish Borders, and possibly more that have gone undocumented, as evidenced by reports of unofficial introductions on the Isle of Mull (Solow et al., 2013). Such undocumented actions could have contributed to the observed levels of genetic mixing in this study. It is also likely that the genetic bottlenecks and inbreeding detected in our study have been influenced by unofficial translocations involving small numbers of animals. This could potentially create multiple, relatively recent founder effects in the population, thus complicating the use of conservation genetics to infer adaptive management strategies.
Five out of the 12 loci exhibited significantly positive Fis values. While positive Fis values can indicate inbreeding, they are also associated with genetic admixture, which is clear in this case. Analysing populations with underlying genetic structure for genetic diversity can artificially increase the level of homozygotes at the expense of heterozygotes, a phenomenon perceived as inbreeding. This is known as the Wahlund effect, which occurs when a population, composed of two or more subpopulations, is analysed as if it were a single, homogeneous population (Waples, 2015). However, in this study, defining populations for analysis was not straightforward, as most sampled localities exhibited genetic structure or sub structure that could not be easily separated for analysis. Indeed, it is rare in population genetics that 'ideal' populations are ever sampled from a statistical viewpoint. Therefore, the results should be interpreted with this consideration in mind (De Meeûs, 2018). A study of European rabbit populations across 17 sites in the East Anglian region of Britain, following a population crash due to a myxomatosis outbreak, found that the populations became genetically distinct with low effective population sizes. It suggested that this genetic divergence resulted from the myxomatosis-induced crash, combined with reproductive and social characteristics that influenced the genetics rather than from past historical events (Surridge et al., 1999). It is possible that a similar scenario exists in this study, where the pine marten population may retain genetic signatures of past population declines.
There was evidence of a genetic bottleneck across all tests implemented in this study. This was also confirmed by the mode-shift test. It is important to note that the Wahlund effect can confound these patterns, making it challenging to distinguish between the loss of alleles due to a bottleneck and the artificial reduction in heterozygosity due to population substructure. However, the mode-shift test is designed to detect shifts in the distribution of allele frequencies indicative of a bottleneck, making it less influenced by the presence of the Wahlund effect, but the results should be viewed with this implication. However, some evidence of a genetic bottleneck was also found in the Irish pine marten population, which was not impacted by genetic structure (O’Reilly et al., 2021). The mode shift test also indicated a shifted mode, which did not conform to the L-shaped distribution expected to occur in a non-bottlenecked population. The observed shift is associated with the loss of low-frequency alleles due to a decline in the Scottish pine marten population size, signifying a genetic bottleneck. When a population undergoes a bottleneck, there is a distinct mode-shift distortion in allele frequencies. Specifically, alleles with a low frequency (less than 0.1) become rarer compared to those in an intermediate range (e.g. 0.1–0.2). An L-shaped distribution suggests no recent bottlenecks for several generations; but a shifted mode, as seen in this study, signifies a recent genetic bottleneck (Luikart et al., 1998; Piry et al., 1999).
Similar genetic bottlenecks were not detected, when previously tested for, in France (Mergey et al., 2012) or Sardinia (Coli et al., 2011), but Ruiz-Gonzalez et al., (2015) did find significant support for historical reductions in effective population sizes in the north of Spain, which was attributed to habitat fragmentation and the presence of a competing mustelid, Martes foina, but there were also unknown factors that may have contributed to a reduction of gene flow. For example, it was proposed that the pine marten was expanding its range in parts of northern Spain, and a lag time, akin to what this study proposes, may have hindered landscape features from being reflected in the species' genetic structure. The detection of a potential bottleneck signature in the present study could reflect the pine marten’s past population retraction, as these signatures can remain for one hundred years or more, even in variable loci like microsatellites. Regardless of the direct cause of the result in this study, be it a past bottleneck or the impact of genetic structure, increasing contemporary gene flow within Scotland would benefit the diversity of the species.
The effective population size estimates from this study ranged from depressed (less than 50), as determined by the linkage disequilibrium method, to modest (less than 200), as determined by the sibship method, all falling below the putative 500-threshold required for long-term viability (Frankham et al., 2014). The variation in results can be attributed to the different underlying assumptions of the methods used to generate these estimates. The linkage disequilibrium method, which assumes observed linkage disequilibrium originates solely from genetic drift in a unified population, may underestimate the effective population size. This underestimation can occur due to the Wahlund effect, which inflates linkage disequilibrium and could misinterpret the extent of population size or drift in structured populations (Waples and Do, 2010). This scenario is very likely in our study, given the underlying genetic structure across most of Scotland. Conversely, the higher estimate derived from the sibship method offers a more reliable means for deriving effective population size estimates in cases like ours, where samples are not collected systematically and may include related individuals across different generations. Less impacted by the Wahlund effect compared to the linkage disequilibrium method, the sibship method focuses on identifying full and half-sibling groups within a sample based on genetic similarity among individuals. This approach avoids reliance on population-level allele frequencies or linkage disequilibrium patterns, making it more reliable for our study. Waples (2021) found that the linkage disequilibrium method outperformed the sibship method in terms of precision, but Gilbert & Whitlock, (2015) also warned that the accuracy of any method is dependent on the demography of the species. What we can say in this case is that the effective population was found to be modest using two independent methods, something which Waples (2021) says improves the overall precision of the estimates.
Estimates for effective population size for Martes americana and M. caurina, also derived using the same linkage disequilibrium method, were also generally modest (in the hundreds) across a 53,474 km2 area, encompassing portions of British Columbia, Idaho, Montana, and Washington (Lucid et al., 2020), and suggested that the long-term protection of the species in the region depended on corridor conservation efforts. Similar estimates were derived for the recovering pine marten in Ireland (O’Reilly et al., 2021). In Finland, the wolverine (Gulo gulo) is showing recovery, especially in the north and east, thanks to translocations in the late 20th century (Lansink et al., 2020). Two primary clusters displayed limited gene flow, with heterozygosity levels of He = 0.49 and He = 0.57, and average alleles of 3.8 and 4.0 – but the effective population size was under 50, indicating previous genetic bottlenecks and future viability concerns (Frankham et al., 2014). It appears that estimates of low to modest effective population size below 500 are not uncommon in mustelid species, particularly those that have had past and current population stresses.
For species whose populations have been anthropogenically reduced, such as the grey wolf, and are now expanding, the effective population size can be significantly smaller than the census population count. In 2021, the northern Rocky Mountains had a census size estimated at 3,354, and the western Great Lakes at 4,526, but when these values were converted to an effective population size, they ranged between 201 and 335 wolves for the northern Rocky Mountains, and between 272 and 453 for the western Great Lakes. Expressing the census population as a multiple of the effective population size reveals that the census population of grey wolves is estimated to be between 10 and 18 times larger than its effective population size. Given the strong skew in the effective-to-census size ratio in grey wolves, conservation practitioners aim to maintain larger wolf populations to ensure long-term adaptation and survival (vonHoldt et al., in 2023). However, achieving the required number is challenging, considering that the species' dispersal capabilities and success must be considered. To maintain and increase effective population size, vonHoldt et al., (2023) recommend that dispersers be granted protection. Indeed, there are many parallels in the pine marten population in this study, as this species was historically heavily persecuted and is now expanding. However, migrants too face barriers to dispersal, particularly roads. Based on the census estimate of 3,700 (Mathews & Harrower, 2020) and the effective population sizes derived from sibships, ranging from 83 to 137, the ratio of the census population to the effective population size in pine martens is estimated to be between 27 and 44 times larger. The strong skew in the effective-to-census size ratio in pine martens also implies that larger populations are necessary to ensure long-term adaptation and survival, as is the case with wolves. The higher multiplier in pine martens compared to grey wolves may stem from challenges encountered in accurately estimating effective population sizes as previously discussed, and could also reflect variations in social organisation, complicating cross-study comparisons. vonHoldt et al., (2023) recommended that studies should be carried out periodically to reassess the situation, which could also be applied in this case.
Given the uncertainty raised in this study related to the genetic recovery of the pine marten, which appears to be lagging behind its national demographic recovery, it is timely to suggest that long-term genetic population monitoring be adopted for the species throughout Britain including the re-establishing populations in England and Wales. Notably, such monitoring is already being planned for the restored population in Wales by Vincent Wildlife Trust (VWT) (J. MacPherson pers. comm.). In our experience, methods involving the collection of hair samples using tubes or feeders have consistently yielded good-quality DNA suitable for genotyping, as demonstrated in this study. Source populations in Scotland should undergo similar checks every decade. The feasibility of implementing a nationwide and standardised approach to long-term population and genetic monitoring should be evaluated and discussed among stakeholders. Methods utilised by Sheehy et al. (2018) in Scotland and O’Mahony et al. (2017a) in Ireland, along with density modelling techniques combined with genotyping as outlined in Twining et al. (2022), offer promising frameworks for such nationwide initiatives. By doing so, we can ensure the health and sustainability of these populations, using this study as a foundational reference for future efforts, and plan adaptive interventions accordingly. When effectively carried out, these efforts could shape future policy, legislation, and guidelines concerning the restoration of wildlife species. This precautionary approach seems wise considering the impacts of global heating and climate breakdown upon pine martens and the habitats upon which they depend. Given the parallels within Britain and Ireland (O’Reilly et al., 2021), there is an opportunity to collaborate on understanding the recovery of the pine marten. This collaboration could inform and establish best practices for other conservation recovery projects, especially as the field of genetic reinforcement is still in its early stages.