Retrospective genetic monitoring of the threatened Yellow marsh saxifrage (Saxifraga hirculus) reveals genetic erosion but provides valuable insights for conservation strategies

Study sites

Only two populations of S. hirculus occur in Ireland; the most threatened population consisted of only 13 individuals at the Garron Plateau, County Antrim during 2011 whilst the other is substantially larger occurring at 13 sites (each with c. 100–200 individuals) near Bellacorcik, Co. Mayo. Herbarium samples show that the species occurred at much greater abundance at the Garron Plateau in the past and also occurred at another site near Rasharkin, County Antrim. Two further populations, one near Coleraine, County Derry and one near Lisclogher Co. Westmeath are now extinct.

Study species

Chromosome numbers for S. hirculus include 2n = 16, 2n = 24 and 2n = 32, although all populations outside the Arctic Circle studied to date, including those in Europe, are tetraploid (2n = 32). Reproduction is both sexual and asexual. Flowers are markedly protandrous, but self-compatible, and are pollinated by a wide range of insects, with different pollinators in different parts of the species' range (Olesen & Warncke, 1989; Warncke et al., 1993). Seeds lack special adaptations for dispersal and are generally deposited close to the parent plant (Olesen & Warncke, 1989). Vegetative reproduction occurs via rhizomes and is believed to be an important mechanism of propagation (Olesen & Warncke, 1990).

Sampling and DNA extraction

A total of 319 present-day samples were obtained during 2011. All 13 plants from the Garron Plateau, Co. Antrim were sampled, and 306 samples were collected from all 13 sites included within the Co. Mayo population (24 individuals from each with the exception of site SHE-D, for which 18 individuals were analysed). The species is protected under the Wildlife Acts 1976–2010 (Ireland) and Schedule 8 of the Wildlife (Northern Ireland) Order (1985), and it was an offence to pick, uproot or destroy the plant. Consequently, a single leaf was taken from each plant under Government licence. Samples were stored in silica gel for transportation. A total of 103 historical samples were obtained from individual plants on herbarium sheets representing both extant locations in Counties Antrim and Mayo, as well as from extinct populations at Rasharkin, Co. Antrim, Coleraine, Co. Derry and Lisclogher, Co. Westmeath. The sampling regime of herbarium samples and their distribution are given in Table 1 and Figure 1 (for herbarium codes see Table S1).

DNA was extracted from all samples using the Qiagen DNeasy Plant Mini Kit, after an initial 8 min grinding at 30 Hz using a Retsch MM300 mixer mill. DNA was quantified visually on 1% agarose gels stained with ethidium bromide and diluted to a concentration of 50 ng μl⁻¹ for subsequent PCR. All DNA extractions from herbarium samples were carried out in a laboratory where no previous S. hirculus work had been performed.

Microsatellite genotyping

All samples were genotyped for six microsatellite markers developed for S. hirculus using the ISSR cloning method outlined in Provan & Wilson (2007). Because of difficulties associated with amplifying longer fragments from herbarium samples, primers were designed to amplify products of less than c. 200 bp (Table 2). The polyploid nature of S. hirculus in Ireland represented a further problem in scoring allele sizes accurately where stutter bands were present. Consequently, we were limited to using two trinucleotide microsatellites and two tetranucleotides, which generally display minimal stuttering, and two dinucleotides with relatively low stuttering. Forward primers were modified by the addition of a 19 bp M13 tail (5′-CACGACGTTGTAAAACGAC-3′) and reverse primers were modified by the addition of a 7-bp tail (5′-GTGTCTT-3′). PCR was carried out in a total volume of 10 μl containing 100 ng genomic DNA, 10 pmol of HEX-labelled M13 primer, 1 pmol of M13-tailed forward primer, 10 pmol reverse primer, 1× PCR buffer, 200 μm each dNTP, 2.5 mm MgCl₂ and 0.25 U GoTaq Flexi DNA polymerase (Promega, Sunnyvale, CA, USA). PCR was carried out on a MWG Primus thermal cycler using the following conditions for all loci with the exception of SH-3-B11: initial denaturation at 94 °C for 3 min followed by 10 touchdown cycles of denaturation at 94 °C for 30 s, annealing at 68 °C for 30 s (−1 °C per cycle), extension at 72 °C for 30 s followed by 30 cycles (40 for herbarium samples) of denaturation at 94 °C for 30 s, annealing at 58 °C for 30 s, extension at 72 °C for 30 s and a final extension at 72 °C for 5 min. For locus SH-3-B11, the following conditions were used: initial denaturation at 94 °C for 3 min followed by 40 cycles (50 for herbarium samples) of denaturation at 94 °C for 30 s, annealing at 52 °C for 30 s, extension at 72 °C for 30 s and a final extension at 72 °C for 5 min. Blank negative controls were routinely used, and c. 25% of herbarium samples were genotyped twice to check for artefacts resulting from low DNA quality, which were not observed. Genotyping was carried out on an AB3730xl capillary genotyping system. Allele sizes were scored using LIZ-500 size standards and were checked by comparison with previously sized control samples.

Genetic analysis

No chromosome counts are available for Irish S. hirculus, but the species is believed to be polyploid in the majority of its non-Arctic range (Hedberg, 1992), and more than two bands were observed at all six loci analysed, suggesting polyploidy. Thus, it was not possible to score genotypes based on allele frequencies, and as a result we could not carry out many standard population genetic analyses (e.g. calculation of allelic richness, AMOVA). Consequently, the patterns of alleles observed for each locus in an individual were scored as phenotypes. Within-population phenotype diversity was estimated for samples with N ≥ 5 using the total number of alleles, observed heterozygosity (H_O), namely the proportion of observed heterozygous individuals in a population, and the Gini–Simpson diversity index, analogous to Nei's gene diversity, averaged over loci (Jost, 2006). Genetic clustering of individuals was assessed using a Bayesian procedure implemented in the Structure software package (V2.3.3; Pritchard et al., 2000), which can accommodate ambiguous co-dominant markers for polyploid species. The programme was run using no prior knowledge and the admixture ancestry model. Five independent runs were carried out for each value of K, the number of genetic clusters, up to K = 10, because log-likelihood values reached a peak at K = 8 and decreased thereafter. Each Markov chain Monte Carlo analysis used a burn-in period of 10,000 followed by a further 100,000 iterations. The most likely value for K was estimated using the ΔK statistic of Evanno et al. (2005) implemented in the Structure Harvester software package (V0.6.1; Earl & vonHoldt, 2012).

Species distribution modelling

A presence-only maximum entropy approach was used to predict landscape suitability for the S. hirculus throughout Ireland, from a sample set of known occurrences and spatially explicit environmental parameters. Maximum entropy has been shown to frequently outperform other presence-only modelling techniques particularly at very low sample sizes (Elith & Graham, 2009). Environmental parameters were described at a 500-m cell resolution (Table S2). The software package MaxEnt was used (V3.3.3k; Phillips et al., 2006). To maximize model flexibility, we considered linear, quadratic, threshold and hinged functions for all environmental parameters (Phillips & Dudík, 2008). Due to the paucity of records, it was not possible to segregate the data set into a training and test sets, thus only a training set was used (Farren et al., 2010). Jackknife resampling analysis was used to determine a heuristic estimate of the relative contribution of each variable based on the performance of the global model (known as the regularized gain) without the variable of interest compared with the influence of that variable in isolation (derived from a univariate model only). Global model performance was judged using the area under the receiver operating characteristic (ROC) curve (Liu et al., 2005). Marginal response curves of the predicted probability of species occurrence were graphed for each explanatory variable. A map of landscape favourability was generated using ArcGIS 9.3 (ESRI, California, USA), and the 10th percentile training presence was used as the threshold.

Saxifraga hirculus genetic erosion over time

Maintaining genetic variation in threatened populations, which by their nature tend to be small and/or fragmented, is one of the central tenets of conservation genetics (Allendorf & Luikart, 2007; Schwartz et al., 2007). The retrospective genetic monitoring afforded by the comparison of a critically threatened extant population with historical samples from the same area suggests that levels of population genetic diversity of S. hirculus in the critically endangered Garron Plateau, Co. Antrim are not only far lower than those in the other, larger extant populations in Co. Mayo, but also appreciably lower than the average values from historical samples (pre-1958) from the same area, although there is evidence of some fluctuation in these values through time. This loss of genetic diversity has been accompanied by the replacement of one lineage identified by the Bayesian clustering analysis (shown in Fig. 2, bottom, in light blue) by another (shown in light yellow), to the extent that the new lineage accounts for over 60% of the total genetic diversity. These changes are probably the result of a combination of factors, foremost among which are stochastic fluctuations in allele frequencies due to the greatly exaggerated effects of genetic drift in very small populations. These changes may also have been accompanied by a founder effect, because the Garron Plataeu population was recorded as extinct after 1920, but subsequently ‘refound’ in 1955 (Kertland, 1956). The extinction of the population at Rasharkin, Co. Antrim at the end of the 19th century would have contributed further to the loss of this lineage, both directly and indirectly via the cessation of possible gene flow between the two Co. Antrim populations. Nevertheless, although the levels of phenotypic diversity were significantly lower in the extant population than in the historical samples, the number of alleles was not. Given that the number of alleles is correlated with the ability to respond to selection, this does not appear to be as a serious concern as potential inbreeding depression.

Conservation implications

Although information from population genetic studies can inform best-practice conservation strategies, the long decline in S. hirculus population numbers highlights a further conservation dilemma. Numbers of recorded individuals at the Garron Plateau have fallen from 130 during 1999 to 13 during 2011 (Georgina Thurgate, pers. comm.). Despite the importance of vegetative reproduction in the spread and persistence of S. hirculus (Olesen & Warncke, 1990), we detected only two potentially clonal individuals (i.e. 15% shared identical multilocus phenotypes). Nevertheless, the extremely small number of individuals comprising the extant population at the Garron Plateau leaves it extremely vulnerable to sudden, stochastic extinction. Thus, it is clear that some sort of augmentation programme is necessary. One of the goals of conservation genetics is to ensure that the provenance of individuals used for such programmes is closely aligned with the population (Lesica & Allendorf, 1999), but the Garron Plateau individuals are generally associated with a genetically distinct group that is not found at any significant level elsewhere in the remaining populations in Ireland. Re-establishment or augmentation of a population using genetically divergent individuals results in a trade-off between increasing population numbers at the risk of outbreeding depression (Edmands, 2007). The Bayesian clustering analysis suggests that if a source population is required for augmentation, it should be the SHA population in Co. Mayo, which not only has the highest percentage assignment to the predominant cluster found in the Garron Plateau, but also contains a mixture of several different lineages. This represents a further important aspect of strategic conservation based on molecular genetic approaches, namely the fact that data from a low number of putatively neutral loci do not necessarily provide insights into the relative fitness of genotypes, particularly in differing habitats (Ennos et al., 1997; Hollingsworth et al., 1998). Augmentation or introduction of a range of genotypes, including those closest to genotypes found in the extant Garron population, as is the case for the Co. Mayo SHA population, will provide a balance between ‘genotype matching’ and sufficient variation for natural selection to operate on. Of course, reintroduction need not be limited to material from a single source population, and the results of the genetic clustering analysis could be used to identify the widest possible range of genetic diversity for reintroduction, if so desired.

Saxifraga hirculus occurrence was associated with bog, fen, marsh and swamp typically on high-altitude plateaus with low maximum temperatures, high precipitation and low seasonality, that is, consistently cool and wet. The loss of the Coleraine, Co. Derry and Lisclogher, Co. Westmeath populations could be attributed to loss of such habitat, because the species distribution model did not identify these areas as currently suitable. A number of areas in the Garron Plateau, Co. Antrim were identified as potentially suitable for establishing new populations (via ex situ conservation) to further supplement and expand the extant population. A single area of potentially suitable habitat was identified near the site of the now extinct population at Rasharkin, Co. Antrim. Insights into the historical genetic makeup of the now extinct Rasharkin population afforded by the analysis of the herbarium samples mean that a controlled reintroduction programme, based on recreating the genetic make-up of the original population from individuals extant elsewhere (i.e. Co. Mayo), could maximize the potential success of re-establishing this lost population.

The genetic changes revealed by the retrospective genetic monitoring indicate the need to implement such approaches as soon as possible. Regular censuses of the population at the Garron Plateau began during 1999 when there were 130 plants. If genetic monitoring had commenced at the same time there would have been more chance of developing a successful ex situ conservation programme to maximize genetic diversity than at present. As it is, the current scenario further highlights the need for conservation practitioners to move away from a ‘fire-fighting’ mentality (Mace & Purvis, 2008). Nevertheless, the findings of our study can be used to inform any potential reintroduction/augmentation programmes. Results from a previous re-establishment programme in Scotland indicate that ex situ propagation of seedlings followed by transplantation is a more successful method than simply sowing seeds directly onto potential recovery sites (Welch, 2002). Based on the information from the current study, genetic analysis of ex situ individuals could be used to select individuals most representative of the current extant gene pool, whilst aiming to maximize genetic diversity.