Introduction

Understanding patterns and processes at range margins is important for the conservation management of declining species. Near the edge of a species’ geographical limit, populations are typically smaller and more unstable compared to the range centre, and show a more scattered local distribution (Gaston 2009; Sexton et al. 2009). Of particular interest are occurrences away from a species’ continuous range, at distances exceeding natural dispersal abilities. They can be interpreted as signatures of range contraction, for example caused by human habitat destruction or longer-term climatic oscillations. Alternatively, they can also be the result of human-assisted colonisation. From a conservation perspective, it is important to address whether local occurrences are native or not, as indigenous populations usually warrant significantly more protection efforts than populations stemming from introductions.

Genetic tools can trace back the demographic history of populations, and are therefore also able to distinguish between native and introduced populations when historical knowledge on the origin of local occurrences is absent or ambiguous (Saltonstall 2002; Fitzpatrick et al. 2012; Avery et al. 2013; Pfenninger et al. 2014). In taxa with low powers of dispersal, such as amphibians, colonisation is expected to proceed along a broad front rather than through isolated populations separated by large distances from the main range (e.g. Arntzen and Wallis 1991; Urban et al. 2008; but see also White and Shine 2009). While dispersal abilities of amphibians are now thought to be higher than assumed in the past (Smith and Green 2005; Jehle and Sinsch 2007), they are on average several orders of magnitude less mobile than for example mammals or birds (Wells 2007). Focusing on isolated occurrences, information drawn from genetic markers has already proven invaluable to identify the source of known introductions (Arntzen and Thorpe 1999; Ficetola et al. 2008), to discern between adjacent native and introduced populations (Johnson et al. 2011), to demonstrate that putatively native occurrences actually stem from introductions (Dubey et al. 2014), and to demonstrate that potentially introduced populations are actually likely to be native (Zeisset and Beebee 2001; Arntzen et al. 2010).

The great crested newt Triturus cristatus has a wide distribution across northern parts of Europe, including England and Scotland (e.g. Jehle et al. 2011). In the Scottish Highlands, occurrences around Inverness are separated by more than 80 km of unfavourable habitat from the northern limits of its main British distribution in the central belt of Scotland. As a consequence, they were in the past regarded as stemming from introductions (e.g. Langton and Beckett 1995), resulting in a situation where legislation would make it in effect illegal to re-release animals, for example those caught during surveys (Wildlife and Natural Environment (Scotland) Act 2011). However, a recently formalised alternative biogeographic scenario outlines a possible colonisation route from the south between 3000 and 7000 years ago (O’Brien and Hall 2012). Corroborated by evidence from historical records (Harvie-Brown and Buckley 1895) and the existence of several clusters of populations divided by tens of kilometres in fragmented landscapes, this scenario suggests that the majority of the Scottish Highland populations might actually be native. Here, we present data from genetic markers (mtDNA and microsatellites) across the distribution of T. cristatus in the UK (Scottish Highlands, central Scotland, central England), to provide evidence that the populations around Inverness are indeed indigenous. T. cristatus is an important wetland flagship species protected under European law (Annex II and Annex IVa of the European Habitats Directive), and this finding has profound consequences for the species’ local conservation management. It also extends the known natural range of the species at its northwestern part of the range.

Materials and methods

Twelve sites were sampled in total: eight in the Scottish Highlands, two in central Scotland (representing the northern limit of the continuous distribution of T. cristatus in the UK), and two in central England (Leicestershire, representing populations from the core area of T. cristatus in the UK, Fig. 1). Fieldwork took place between 23 May and 29 May 2012 (Scotland), and in the early 1990s (Leicestershire). DNA sampling was based on the collection and immediate storage of embryos (eggs) in ethanol in the field. Egg sampling is a standard method to retrieve DNA for newts (e.g. Jehle et al. 2005a); it has minimal effects on population dynamic processes (each female lays about 200–400 eggs per season at high natural mortality rates, e.g. Jehle et al. 2011) and avoids pond perturbation through trapping and netting for DNA sampling from adults. The Highland sites included at least one site from each of O’Brien and Hall’s (2012) proposed clusters of populations; for more details of the study area in Leicestershire, see Scribner et al. (2001).

Fig. 1
figure 1

Distribution of the great crested newt (Triturus cristatus) in the UK (NBN 2014), and location of sampling sites. The different colours of T. cristatus locations represent genetic clusters as identified using the software BAPS (for more details see text)

Full size image

DNA extractions were performed using standard phenol–chloroform procedures (Bruford et al. 1998). For an analysis based on nuclear microsatellites, we genotyped individuals at six polymorphic loci (Tcri27, Tcri29, Tcri35, Tcri36, Tcri43 and Tcri46; Krupa et al. 2002). Polymerase chain reactions (PCRs) were carried out in 10-μL reaction volumes under the following conditions: 1 or 1.5 mM MgCl2, 0.1 mM dNTPs, 0.5 U polymerase (GoTaq) and 0.1 mM of each primer in the manufacturer’s buffer. Thermal profiles for Tcri27, Tcri35, Tcri36 and Tcri43 were 39 cycles of 30 s at 93 °C, 30 s at the primer-specific annealing temperature, 45 s at 72 °C; for Tcri29 and Tcri46, we used a ‘touchdown’ thermal profile (for more details, see Krupa et al. 2002). Primers were labelled with fluorochromes AT550, HEX and FAM. PCR products were separated on an ABI3130 Genetic Analyser (Applied Biosystems) and scored using Peakscanner v1.0 (Applied Biosystems) and Tandem v1.08 (Matschiner and Salzburger 2009). Individuals for which more than two loci failed to amplify were discarded.

Allele frequencies, Hardy–Weinberg equilibria and measures of genetic differentiation (F ST) were calculated using Genepop on the Web (Rousset 2008). The possible existence of past population bottlenecks was investigated with the software Bottleneck (Luikart et al. 1998), using the Wilcoxon sign rank test and a mixed model of stepwise and infinite allelic mutations of 95 and 5 %, respectively (as determined previously for the employed T. cristatus microsatellites; Jehle and Arntzen 2002). Because multiple tests were conducted across study sites and loci, Bonferroni-adjusted significance levels were applied when applicable. For the Scottish Highland populations, isolation-by-distance scenarios were tested using Mantel tests to correlate F ST and log-transformed geographic distances as implemented in the software IBD (Bohonak 2002); geographic distances were calculated along shorelines if applicable, as dispersal through salt water is impossible for T. cristatus. The partitioning of genetic variation among populations was addressed using the hierarchical Bayesian approach implemented in the software BAPS v3.2 (Corander et al. 2004), which clusters a predefined substructure (sites) into a more meaningful structure as reflected in the dataset.

We also characterised two random individuals from each population from Scotland (Highland and central Scotland) by sequencing a 658-bp-long ND4 (mtDNA) fragment using existing PCR primers and protocols (F: AGCGCCTGTCGCCGGGTCAATA, R: GTGTTTCATAACTCTTCTTGGT, Wielstra et al. 2010). PCR reactions were carried out in a 10-μL reaction volume with the following constituents: 10 ng genomic DNA, 1 μL GoTaq × 5 reaction buffer (1.5 mM MgCl2 in ×1 concentration), 1 μM of each primer, 0.25 mM of each dNTP and 1 unit of GoTaq DNA Taq polymerase. Amplified products were cleaned using a standard ethanol/salt precipitation method and commercially sequenced in both directions. Sequences were examined using Molecular Evolutionary Genetics Analysis (MEGA) software (Tamura et al. 2013).

Field work and sample collecting was conducted under licences issued by Scottish Natural Heritage (Licence 13374) in Scotland.

Results

In total, 253 individuals across 12 sites and three British regions were analysed (Table 1). Five populations showed significant deviations from the Hardy–Weinberg equilibrium at one locus each (largely heterozygote deficiencies: D, MO, NS, LV, CC, Table 2). As these deviations are not based on specific loci, we argue that it is unlikely that they are caused by locus-specific non-amplifying alleles.

Table 1 Details of sample sites
Full size table
Table 2 Descriptive population genetic parameters for the sites under consideration
Full size table

The amount of genetic variation found for each population was markedly lower in Scottish populations compared to central England. The central England populations displayed about three times higher amounts of alleles per locus than populations from Scotland; observed heterozygosities of the two central Scotland populations were between the values of the eight Highland and the two central England populations (Table 2). In the Highlands, otherwise polymorphic loci were observed as fixed for specific populations in several instances (Tcri27: BW, Tcri29: PH, CL, Tcri36: D, PH; due to low sample size we do not consider population FGC here), whereas all loci were polymorphic in central Scotland and central England. In line with their higher level of genetic variation, the central England populations also possessed the highest numbers of private alleles, which were rare for Highland populations (Table 2). There was no consistent signature of past population bottlenecks across the study populations; the lowest p value shown by population P in central Scotland (0.02) is non-significant after applying a Bonferroni-adjusted threshold (Table 2).

Measures of genetic differentiation (F ST) revealed that most of the study populations were highly genetically differentiated from each other (Table 3). F ST values between Highland and central Scotland, and those between Highland and central England sites, were above 0.24 throughout. F ST between the two central Scotland and the central England sites were 0.11 and 0.03, respectively. F ST values between Highland sites had a wide range between 0.03 and 0.52, with a marginally significant pattern of isolation by distance (Z = 11.314, r = 0.338, p = 0.055). The hierarchical analysis implemented in BAPS v3.1 revealed an optimum partition comprising six clusters. Both central England and central Scotland sites were assigned to one cluster each (PF/P and CC/G); the remaining clusters consisted of the merged Highland populations, CL/MO/LV and D/FGC/NS, and the single Highland populations BW and PH (Fig. 1).

Table 3 Pairwise genetic differentiation (F ST values) across sampling sites
Full size table

All of the 20 Scottish individuals examined possessed the single ND4 haplotype previously found in British populations (GenBank accession numbers GU982383, Wielstra et al. 2010, and HQ697273, Wielstra and Arntzen 2011), confirming that, in the UK, T. cristatus comprises a single phylogeographic lineage depauperate in variation at the mtDNA level. Given that the haplotype is widespread across large parts of the species’ range (including its glacial refugia in Southeastern Europe, Wielstra et al. 2013), it is, however, of limited diagnostic value to infer the native status of the Highland populations.

Discussion

The primary purpose of this study was to use molecular means to investigate the origin of isolated T. cristatus populations in the Scottish Highland region. Although these occurrences were previously assumed to be introduced (with, however, unknown sources, e.g. Langton and Beckett 1995), we argue that the most parsimonious interpretation of the gathered population genetic evidence is that they actually represent native populations.

The examined populations showed no evidence of a recent founder event followed by a demographic expansion, which would be expected for populations which have experienced an introduction based on a small propagule pool. The employed method is able to detect a bottleneck signature for up to about ten generations after its occurrence (Luikart et al. 1998). Typical average generation times for T. cristatus were estimated to be around 4–5 years (for a summary, see Jehle et al. 2011) and, due to the colder climate, are expected to be higher at northern range limits. Thus, our approach covered a period of about four to five decades back in time (~1960s). However, more than six loci would be necessary for a high statistical power to detect past bottlenecks (Luikart et al. 1998). Nevertheless, there were no noticeable differences between Highland populations and known native populations from central Scotland and England, suggesting that all groups of populations have experienced comparable recent demographic histories.

The discounting of a recent introduction based on demographic analyses is supported by records of the species in the region of Inverness dating back to the end of the nineteenth century (Harvie-Brown and Buckley 1895). That the Highland populations possessed a common (and the only British) T. cristatus ND4 mtDNA haplotype did not entirely rule out an introduction from other phylogeographic lineages in mainland Europe, and allows the possibility of an introduction by, e.g. Victorian naturalists earlier in the same century. This, however, lacks historical evidence, as only two introductions of T. cristatus, both of which are over 75 km away from the populations near Inverness, are reported (Scottish Natural Heritage, unpublished). Assisted colonisations have high failure rates for T. cristatus (even when specifically conducted for conservation purposes, Oldham and Humphries 2000), and isolated populations are predicted to have high long-term extinction rates (Halley et al. 1996). The success of amphibian translocations also rises with the size of the donor population regardless of the adaptive genetic background (Zeisset and Beebee 2013), and it is unlikely that deliberate introductions are based on a large number of propagule individuals. Taken together, this suggests that tens if not hundreds of introductions would have been necessary for the local occupancy we see today, which is highly unlikely.

A main finding of the present study is the pronounced decline in microsatellite genetic diversity from central England northward to the Scottish Highlands. While the general pattern is in line with previous knowledge of the distribution of genetic diversity across the range of T. cristatus (Babik et al. 2009; Wielstra et al. 2013), a roughly three-fold decrease in microsatellite variation within a distal (island) area about 2000 km away from the species glacial refugium is difficult to explain by post-glacial range expansion alone. Low levels of neutral variation for Highland populations would fit the pattern expected of an introduction. However, an introduction from core areas of the species’ range (UK or Europe) would result in marked differences in the amount of genetic variation between the northern limit of the species’ continuous range and the Highland populations, which was not observed. We argue that the observed patterns are rather a result of demographic processes in peripheral populations, which include reduced gene flow, small population size and higher extinction–recolonisation turnover (see also, e.g. Peterman et al. 2013 for North American frogs). The two populations from central England showed amounts of neutral variation equal to or higher than that of populations in central and Western Europe (Jehle et al. 2005a; Babik et al. 2009; Schön et al. 2011), reinforcing the interplay between local population processes and biogeographic history in shaping the amount of genetic variation at the level of microsatellites.

Within the Highland populations, several loci were monomorphic in some populations but polymorphic in others. The observed pairwise F ST values for Highland T. cristatus exceed the values documented during studies conducted elsewhere at comparable spatial scales (0.07–0.11: Jehle et al. 2005a, 0.074–0.141: Schön et al. 2011). This confirms that a scenario of assisted colonisation of the Highlands is only plausible if multiple successful introductions have taken place, which we deem highly unlikely. Instead, we argue that local forces, such as low local population density combined with a terrestrial environment around ponds which is unfavourable for dispersal, lead to high degrees of genetic drift and fragmentation of otherwise natural populations, which is also reflected in the weak but existing pattern of isolation by distance. It is also noteworthy that the genetic data match with the barriers to dispersal proposed by O’Brien and Hall (2012). With an F ST value of 0.03, NS and D have the lowest genetic differentiation between them, and indeed are situated in close geographic proximity. F ST between MO and CL is also low (0.05), and nineteenth century maps suggest that they would formerly have been connected by favourable terrestrial habitat (O’Brien and Hall, unpublished data). That the algorithm in BAPS merged the two central Scotland and central England sites into one cluster each, and separated the ten Highland sites into four further clusters, once again supports the hypothesis that the Highland occurrences are of native origin. As found for a set of T. cristatus populations in Western Europe (Jehle et al. 2005b), high random genetic drift in the absence of connectivity between specific pairs of populations is likely to be responsible for clusters partially, but not entirely, representing the actual spatial settings. Due to our hierarchical sampling and problematic effects of isolation by distance (Frantz et al. 2009), we refrained from a further analysis of population assignments using, e.g. the software Structure.

The weight of evidence suggests that T. cristatus is native to the Highland region, and this has already given impetus to efforts to conserve the species, particularly site protection and habitat creation to link previously isolated populations. The genetic data further highlight that Highland populations might be at risk from loss of fitness arising from low genetic diversity (microsatellite heterozygosity is correlated with body size and age in T. cristatus, Herdegen et al. 2013). Ongoing fieldwork has already discovered several new sites (Miró, O’Brien and Hall, unpublished) and suggests that habitat requirements of Highland populations resemble those of T. cristatus at the northern range margins in Scandinavia (e.g. Skei et al. 2006). Taken together, this work has contributed to our set of case studies showing how genetic research can be used to clarify the origin of local populations, and has extended the known northwestern range of an important wetland flagship vertebrate.