Low genetic diversity and potential inbreeding in an isolated population of alder buckthorn (Frangula alnus) following a founder effect

Alder buckthorn (Frangula alnus) is one of Ireland’s rarest tree species, and in Northern Ireland the species is now restricted to a single population in Peatlands Park, Co. Armagh numbering ca. 140 mature trees. Genotyping of 95% of the trees at nine nuclear microsatellite loci revealed that levels of genetic diversity within this population were generally lower than those reported from larger populations in Spain. Analysis of six chloroplast microsatellite loci revealed no variation. The level of F IS was significantly higher than that in the Spanish populations, as well as in other populations across Europe, potentially indicating inbreeding. Spatial autocorrelation analysis indicated some evidence of fine-scale genetic structuring, most likely due to limited seed dispersal, but the overall level of differentiation between subpopulations was low, indicating high levels of gene flow, probably due to cross-pollination by bees. Our results are consistent with a gradual population expansion from a limited number of individuals. We suggest that more immediate conservation efforts might be best focused on ensuring suitable habitat for the continued recovery of this isolated population.

it is not known whether the original trees were remnants of a once larger population, or whether numbers have always been low due to an initial founder effect. In recent years, the use of polymorphic microsatellite markers 14 has allowed the testing of whether populations have gone through a bottleneck, based on theoretical expectations under mutation-drift equilibrium at a single point in time [15][16][17][18] . Consequently, the aim of the present study was to determine the levels of and patterns of genetic diversity in the remaining population to discover (1) whether there is any evidence for a genetic bottleneck, (2) if the establishment from a relatively limited number of individuals has been accompanied by a degree of inbreeding, and (3) whether there is any significant genetic substructuring within the population. In addition, as F. alnus is considered an invasive pest species in many countries 19,20 , our findings could also shed light on the genetic demography of this recently expanded population with respect to similar invasive populations.

Results
Current distribution of F. alnus in Northern Ireland. Surveys of sites where F. alnus had been recorded previously found that the species is now restricted to a single location; Peatlands Park, Co. Armagh. The sole remaining population exists as five discrete clusters of plants numbering between 3-98 individuals, each separated by between 100-1,300 m ( Fig. 1; Figure S1, Supplementary Material). In total, there are ca. 140 mature trees. The only other population recorded, at Drumawhey Bog on the northern edge of Strangford Lough, now appears to have been extirpated.
According to the CEH Land Class Map 2007 and site visits, Subpopulation A resides in a suburban area and is the only subpopulation outside the SAC of Peatlands Park (Fig. 2), Subpopulation B resides in grass-dominated bog, Subpopulation C occurs on the edge of deciduous woodland, Subpopulation D is split between grass-dominated bog, heather-dominated bog and scrub, and Subpopulation E occurs on the boundary between a coniferous and deciduous woodland. The majority of the trees in the population occur in bog (40.1%), divided between grass dominated bog (17.4%) and heather dominated bog (22.7%). The next most common habitat type for this population is scrub (33.3%), then deciduous woodland (16.7%), suburban areas (8.3%) and the least common habitat type these trees occur in is coniferous woodland (1.5%).
Levels of genetic diversity. No evidence of linkage disequilibrium was detected between the nine loci studied. Genotypes were obtained for 132 of the 139 trees (95%). Allele frequencies by locus and subpopulation are given in Table S1, Supplementary Material. Within-subpopulation levels of genetic diversity averaged across loci are given in Table 1 Table S2, Supplementary Material. No evidence of a genetic bottleneck was detected under any of the three mutation models, with two of the nine nuclear loci studied showing a heterozygote excess under all three mutation models ( Table 2) No diversity was observed for the chloroplast microsatellites, with all six loci being monomorphic.

Levels of genetic differentiation and spatial genetic structuring. Differentiation among subpopu-
lations calculated from the analysis of molecular variance (AMOVA; Table 3) was Φ ST = 0.0322 (p = 0.009). The BAPS analysis indicated that the first four subpopulations belonged to the same genetic cluster, with the fifth (Subpopulation E) being genetically distinct ( Figure S2, Supplementary Material). The spatial autocorrelation analysis did, however, suggest that there may be some fine-scale spatial genetic structuring within Subpopulation D, with the autocorrelation coefficient being significantly higher than zero at the smallest scale (<50 m; Fig. 2).

Discussion
The genetic analysis of the sole remaining population of Frangula alnus in Northern Ireland carried out in the present study suggests that establishment from a limited number of individuals has led to limited levels of genetic variation, accompanied by potential inbreeding during the recent expansion in census population size. A comparison with results from a study on F. alnus in Spain 21 suggests that the Peatlands Park population has significantly lower levels of genetic diversity than two of the three Spanish populations. The complete lack of genetic variation in the six chloroplast microsatellite loci studied is also consistent with a founder effect. Although comparable data for these markers are not available for the Spanish populations, chloroplast microsatellites represent the most variable regions of the chloroplast genome 22 , and have also been shown to be monomorphic in a previously well-documented population bottleneck in Torrey pine 23 . The observed differences in levels of diversity between Northern Ireland and Spain could also reflect longer-term historical factors, since the Spanish populations most likely represent refugial populations during the last glaciation, and that the low levels observed in Northern Ireland are consistent  with the founder effects associated with postglacial recolonization ("southern richness and northern purity") 24,25 . Interestingly, though, a further study on populations from Italy, France, Belgium and Sweden using 186 single nucleotide polymorphisms (SNPs) showed no decrease in genetic diversity with latitude 26 .    Table 3. Analysis of molecular variance (AMOVA). Significance of Φ ST : **P < 0.01.
Despite the low levels of genetic variation, the Wilcoxon test for heterozygote excess did not indicate the occurrence of a bottleneck. Seven of the nine loci indicated a heterozygote deficiency, which could be indicative of population stability or constant growth 15 . Although there are no records predating 1934, evidence suggests that the population remained relatively stable from this time until the end of the 1980s, after which numbers increased to the current census of ca. 140 individuals 10,13 . It may be unwise to take this at face value, though, since the power of the test is affected by several parameters, including number of loci used, number of generations since the bottleneck, the length of the bottleneck itself, and the magnitude of reduction in effective population size. Immigration, another potentially confounding factor, can most likely be ruled out, since the closest known F. alnus populations to that in Peatlands Park are found ca. 125 km away in Co. Westmeath, Ireland. Likewise, the mutation model is unlikely to have an effect, since the results were broadly consistent across the three models analyzed, and changing the parameters of the two-phase model from 90% single-stepwise mutations to 70% also had no effect. The generation time of F. alnus has been estimated to be between 5 years in Central Europe to around 20 years in the South 27 , so the maximum number of generations of growth is likely to be much less than 10, although the period of apparent population stability preceding this could be upwards of 10 generations. These time scales may limit the power of the heterozygote excess test to detect any possible bottleneck, but this may be balanced to some extent by the fact that, based on historical records, the effective population size (N e ) after the bottleneck is unlikely to be any higher than ca. 20 15 .
Removal of rhododendron from areas of the park since the designation of the Annagarriff Nature Reserve may have facilitated the increase in numbers of F. alnus (Keith Stanfield, personal communication), but this has apparently been accompanied by a degree of inbreeding, based on F IS values. Such inbreeding is most likely a result of the relatively limited genetic base of the population, particularly compared to the other populations of F. alnus. The mean value of F IS for the Peatlands Park population is significantly higher than those from the three Spanish populations studied previously 21 . It is also far higher than the range of values reported in F. alnus populations across Europe based on SNPs (F IS = −0.107-0.088, mean = −0.015) 26 . Reproductive dominance by a relatively low number of highly fecund individuals within a population could give rise to biparental inbreeding 28,29 , but since F. alnus possesses a self-incompatibility mechanism, as indicated by crossing studies 30 , this would be dependent on the diversity of S alleles in the population. Previous studies have indicated a breakdown in self-incompatibility following population bottlenecks 31, 32 , but such a possibility would have to be tested via controlled pollination experiments.
The absence of high levels of spatial structuring of genetic variation in the Peatlands Park population of F. alnus indicates a general lack of barriers to dispersal, which is unsurprising given the small spatial scale relative to the potential dispersal distances of pollinators. The overall level of population differentiation (Φ ST = 0.032) was lower than the average values for outcrossing species with seeds dispersed by gravity (0.152) or ingestion (0.200) quoted by Hamrick & Godt 33 , as well as the average value for biparentally inherited markers in angiosperms (0.184) quoted by Petit et al. 34 . The spatial autocorrelation analysis suggested some fine-scale genetic structure in Subpopulation D, probably due to limited seed dispersal. Seeds of F. alnus can be dispersed by birds 27,35 , but in the closely related (congeneric in some classifications) 36 Rhamnus cathartica, it has been shown that 90% of fruits fall beneath the mother tree 37 . This scenario could at least in part be responsible for the high F IS observed in Subpopulation D, by way of a Wahlund effect.
The area surrounding Peatlands Park has a number of apple orchards with large numbers of wild bees in the vicinity, and the high incidence of fruits in the Peatlands Park F. alnus population suggests substantial levels of cross-pollination. Bees have large foraging ranges, even across sub-optimal habitats, and are likely to be important in maintaining the connectivity observed between the fragments in the present study 38,39 . In contrast to the levels of fruiting found in Peatlands Park, where almost all trees had multiple fruits, a previous study on reproduction in populations of F. alnus from Cádiz, Spain found that only 2.8% of open-pollinated flowers set fruit 30 . Another study on southern range edge populations also indicated that the majority of ovule losses were due to cross-pollen limitation and extensive geitonogamy, and that seed output in the populations was limited to a few large trees 40 . The high percentage of fruiting trees observed in Peatlands Park suggests that this population is more similar to those found in Central European populations, which have a shorter generation time and higher levels of fruit production compared to Southern Iberian populations 27 . If the observed high F IS values reflect some degree of inbreeding, there has been no apparent impact on fitness, at least in terms of fruit production.
Although the BAPS analysis assigned Subpopulation E to a separate genetic cluster from the other four subpopulations, this should be taken with some degree of caution for several reasons. Firstly, the subpopulation numbers only three individuals, and thus allelic frequencies will be skewed. Secondly, it has been shown previously that the BAPS algorithm tends to over-estimate the true number of genetic clusters present in the data 41 . Finally, only a single private allele is present in Subpopulation E, with the majority of genetic differentiation being due to the aforementioned differences in allele frequencies at several of the loci studied.
Whilst it is true that knowledge of the evolutionary dynamics of natural populations with respect to demography and gene flow allows the management of threatened plant populations to go beyond simple "conservation gardening" 42 , in the case of F. alnus in Northern Ireland it would appear that edaphic and ecological factors are of greater importance. The land classes that are correlated with occurrence of alder buckthorn include bog, conifer and pastures, and this is consistent with the literature that claims this species likes open lowland areas with moist, fertile soils 43,44 . Nevertheless, there are large areas of apparently suitable habitat in Peatlands Park which have not been colonized during the recent expansion of the species in the nature reserve. This would suggest that although the low levels of genetic diversity and potential inbreeding revealed in the population are of some concern with respect to evolutionary potential, more immediate conservation efforts might be better focused on ensuring suitable habitat for the continued recovery of this isolated population. Additionally, consideration should be given to possible supplementation using material from the populations in the Republic of Ireland, which are larger and may harbor additional genetic diversity.

Methods
Study species. Frangula alnus Miller (syn. Rhamnus frangula L.) is a small tree or shrub found across temperate Europe from northern Scandinavia and Russia to the Mediterranean, where it is comparatively scarce 43 . In Britain and Ireland, the species tends to be found in damp, boggy habitats, but can also colonize drier ground outside of its native range, particularly in North America where the species is now considered an invasive alien 19,20,35 . Reproduction in F. alnus is exclusively sexual 30 . Flowers are hermaphroditic and well-adapted to insect pollination as they have nectar, colour and odour. They are generalised entomophiles, with orders Hymenoptera, Diptera and Coleoptera providing the main pollinators, and occasional pollination by Lepidoptera 40 . Controlled pollination studies revealed almost no selfing or geitonogamy, indicating the existence of self-incompatibility mechanisms, although occasional selfing in the absence of insect pollinators has been reported 30 . Field studies suggest that limited fruit initiation is primarily due to low levels of cross-pollination 30,40 . Seeds are dispersed either by gravity or via ingestion by birds 27, 35, 37 . Surveys and sampling. Surveys were carried out across Northern Ireland in the summer of 2007 at sites where F. alnus had been found previously based on records from the National Biodiversity Network (NBN) Gateway (http://data.nbn.org.uk) and the Centre for Environmental Data and Recording (CEDaR: http://www.habitas.org. uk/cedar/). Frangula alnus is restricted to the southern shores of Lough Neagh, where it exists as a fragmented population from a single location in Annagarriff Nature Reserve in Peatlands Park, County Tyrone (Fig. 1). All mature trees were numerically tagged with metal tags, GPS coordinates recorded, and a leaf sample from each obtained for genetic analysis. The GPS coordinates were loaded into ESRI ArcMap 10 and plotted on top of a CEH Land Cover Map 2007 45 to discover the main land class that the trees occupy. DNA was extracted from leaf material using the CTAB method 46 . Genotypes were successfully obtained for 132 of the 139 trees sampled.
Nuclear microsatellite analysis. We attempted to genotype all mature plants for sixteen previously described microsatellite loci for F. alnus 21 . Of these, four (FaA103, FaA125, FaA8 and FaB8) could not be consistently amplified, and three (FaA12, FaB102 and FaA116) were monomorphic, leaving nine polymorphic loci: FaB101, FaA110, FaB7, FaA104, FaB106, FaB4, FaA7, FaA3 and FaB9. All reactions were carried out on a MWG Primus thermal cycler. PCR was carried out in a total volume of 10 μl containing 100 ng genomic DNA, 10 pmol of dye-labelled forward primer (HEX), 1 pmol of tailed forward primer, 10 pmol reverse primer, 1x PCR reaction buffer, 200 μM each dNTP, 2.5 mM MgCl 2 and 0.25 U GoTaq Flexi DNA polymerase (Promega). PCR conditions were as described previously 19 . Genotyping was carried out on an AB3730xl capillary genotyping system. Allele sizes were scored using the GeneMapper software package (V5.0; Applied Biosystems) and LIZ-500 size standards, and were checked by comparison with previously sized control samples.
Chloroplast microsatellite analysis. Chloroplast microsatellite markers were developed by identifying mononucleotide regions of ten or more repeats in partial F. alnus chloroplast genome sequences either from GenBank or in regions amplified and sequenced de novo using universal chloroplast primers 22 . The trnT-trnF, atpH-atpI, atpI-rpoC2, rps18-clpp, psbC-trnS, trnS-trnfM and rpl16-rps3 (UCP6) regions were amplified and sequenced using the primers described in Grivet et al. 47 and Provan et al. 48 . Species-specific primers were designed using the Primer3 program to amplify six chloroplast microsatellites (Table S3, Supplementary Material). Primers were tailed as described previously, and PCR was carried out in 10 μl reactions as described previously using the following conditions: initial denaturation at 94 °C for 3 min followed by 30 cycles of denaturation at 94 °C for 30 s, annealing at 56 °C for 30 s, extension at 72 °C for 30 s and a final extension at 72 °C for 5 min. Genotyping was carried out as described previously.
Data analysis. Subpopulations of the Peatlands Park population were classed as distinct groups of trees with no other trees of this species growing within 100 m between groups, resulting in five distinct subpopulations (Fig. 1). GenePop V3.4 49 was used to test for linkage disequilibrium between nuclear loci. To estimate genetic diversity within the population, levels of observed (H O ) and expected (H E ) heterozygosity, and fixation indices (F IS ) were calculated using the Arlequin (V3.5.1.2) 50 and Fstat (V2.9.3.2) 51 software packages respectively. Significance of F IS was determined by 10,000 randomization steps. To test for the occurrence of a genetic bottleneck, the Wilcoxon test for heterozygote excess was performed under the infinite alleles model (IAM), the stepwise mutation model (SMM) and a two-phase model (TPM) incorporating 90% single-stepwise mutations using the program Bottleneck (V1.2) 52 . The Wilcoxon test was used as it is recommended for a relatively low number of loci.
To compare genetic diversity between the population analyzed in the present study, and those from three Spanish populations previously analyzed using the same nuclear microsatellites 21 , mean values for H E and F IS were calculated over the twelve loci successfully amplified in the present study, including the three loci which were monomorphic. Mann-Whitney tests were carried out to assess the significance of differences in H E and F IS .
The level of genetic differentiation between the five fragments was estimated using Φ ST , which gives an analogue of F ST 53 calculated within the analysis of molecular variance (AMOVA) framework 54 using Arlequin. To identify possible spatial patterns of gene flow, the software package BAPS (V5) 55 was used to identify clusters of genetically similar subpopulations using a Bayesian approach. Ten replicates were run for all possible values of the maximum number of clusters (K) up to K = 5, the number of subpopulations, with a burn-in period of 10 000 iterations followed by 50 000 iterations. Multiple independent runs always gave the same outcome. To further identify possible spatial patterns of gene flow, spatial autocorrelation analysis was carried out for Subpopulation D, the largest of the five subpopulations, using SPAGeDi (V1.4) 56 . Mean coancestry coefficients (θ xy ) 57 between pairs of individuals were calculated at 25 m distance class intervals, and plotted as a correlogram, with 95% confidence intervals calculated from 1,000 permutations of individuals within each distance class, and for estimates of θ xy using 1,000 permutations.