Introduction

Habitat loss and fragmentation is one of the greatest threats to global biodiversity (Wilcox and Murphy 1985; Saunders et al. 1991). The potentially deleterious ecological effects of fragmentation on species and communities include changes in resource availability, reduction in population numbers and loss of connectivity leading to population isolation. Within species, fragmentation of natural populations can have negative effects at the genetic level, thus threatening their evolutionary potential (Young et al. 1996). Theoretical and empirical population genetic studies have predicted that fragmentation will lead to a loss of genetic diversity due to inbreeding (Keller and Waller 2002), population isolation and restricted gene flow (Schaal and Leverich 1996; Couvet 2002) and small effective population sizes (Ellestrand and Elam 1993) and that these may lead to a decline in fitness or even, ultimately, extinction (Newman and Pilson 1997; Frankham and Ralls 1998; Keller and Waller 2002).

Many of the negative genetic impacts of population fragmentation can be ameliorated by gene flow (Allendorf 1983). In tree species, which are generally believed to harbour high levels of within-population genetic diversity, studies on impacted populations have yielded conflicting results regarding the effects of fragmentation. It has been suggested using both direct and indirect estimates of gene flow that the removal of potential physical barriers to pollen movement allows for high or even increased levels of gene flow in wind-pollinated tree species (Foré et al. 1992; White et al. 2002; Dick et al. 2003; Bacles et al. 2005) but other studies have suggested that this is not always the case (Sork et al. 2002; Koenig and Ashley 2003; Jump and Peñuelas 2006).

In this study, we have examined the genetic diversity in fragmented populations of juniper (Juniperus communis) in Ireland. Coniferous trees are generally highly heterozygous, outbreeding and wind-pollinated, and thus should exhibit high levels of intrapopulation genetic diversity but low levels of genetic differentiation between populations. Information on the genetic diversity of extant juniper populations and how this diversity is partitioned is important for conservation purposes, since many extant populations exhibit a highly fragmented distribution. This is exemplified by the distribution of juniper in Ireland, where the majority of populations are restricted to the extreme western regions of the island (Fig. 1). The species is one of only three native conifers in Britain, the others being yew (Taxus baccata) and Scots pine (Pinus sylvestris), and one of only two in Ireland since natural populations of Scots pine became extinct on the island several hundred years ago (Bradshaw and Browne 1987). Juniper exhibits a variety of morphological forms ranging from prostrate and creeping to erect, tree-like shrubs and two subspecies, ssp. communis and ssp. nana, are currently believed to exist in the UK, although previous molecular and biochemical analyses failed to discriminate between the two (Vines 1998; Filipowicz et al. 2006). Plants are dioecious, with wind-pollinated female cones, or “berries”, producing seeds that are primarily dispersed by birds. Despite the potential for high levels of dispersal of both pollen and seeds, the species has shown a serious reduction in distribution across the UK and Ireland and populations are believed to have declined by up to 60% since 1960 (Ward 1973; Preston et al. 2002; Thomas et al. 2007). This decline can be attributed to a wide range of factors including climate change, intensification of agriculture, especially grazing, and urbanisation (Clifton et al. 1997; Sanz-Elorza et al. 2003; Verheyen et al. 2005). Recruitment levels appear to be low, with a recent survey of juniper in Northern Ireland finding an age structure highly skewed towards mature and old trees with very little evidence of berries (Preston et al. 2007). As a consequence of population decline, juniper is protected under Section 8 of the Wildlife and Countryside Act in Britain and corresponding legislation in Northern Ireland.

Fig. 1
figure 1

Map showing locations of populations sampled. Inset map shows distribution of juniper in Ireland

Full size image

The aims of the present study were to use biparentally inherited nuclear markers and paternally inherited chloroplast markers (Neale and Sederoff 1989; Neale et al. 1991; Wagner 1992) to elucidate the levels and patterns of genetic diversity in juniper in Ireland to inform conservation and management strategies. We utilised a combination of nuclear microsatellites and a cheap, high-throughput method of analysing single nucleotide polymorphisms (SNPs) in the chloroplast genome to assess the effects of gene flow patterns in shaping the present-day genetic architecture of extant juniper populations throughout its known Irish range.

Materials and methods

Sampling and DNA extraction

Samples were obtained from 19 populations in 12 regions representing the majority of the distribution of juniper in Ireland (Table 1). Where sample numbers within populations are small, these reflect small numbers of accessible plants. Samples were stored at −20°C and DNA was extracted from needle tissue 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.

Table 1 Locations of sampled sites and sample numbers
Full size table

Nuclear microsatellite analysis

All samples were genotyped for nuclear microsatellite loci JC16, JC32 and JC35. Primer sequences and PCR protocols are given in Michalczyk et al. (2006). The other two primers described in the same paper, JC31 and JC37, could not be reliably amplified and thus were not used in the present study. PCR was carried out on an MWG thermal cycler in a total volume of 10 μl containing 100 ng genomic DNA, 10 pmol of 32P-end labelled forward primer, 10 pmol of reverse primer, 1× PCR reaction buffer (5 mM Tris–HCl (pH 9.1), 1.6 mM [NH4]2SO4, 15 μg/μl BSA), 2.5 mM MgCl2 and 0.5 U Taq polymerase (Genetix). Products were resolved on 6% denaturing polyacrylamide gels containing 1× TBE and 8 M urea after addition of 10 μl of 95% formamide loading buffer. Gels were run at 70 W constant power for 2 h, transferred to 3MM Whatman blotting paper and exposed to X-ray film overnight at −20°C. In all cases, previously analysed samples were included as controls to compare product sizes across gels.

Chloroplast single nucleotide polymorphism allele-specific PCR (SNP AS-PCR) analysis

An initial screen for chloroplast variation was carried out using a single individual from each of the populations studied (Table 2). The following eight regions were analysed: trnT-trnF (Taberlet et al. 1991); trnD-trnT, psbC-trnS (Demesure et al. 1995); atpH-atpI, atpI-rpoC2, petB-petD (Grivet et al. 2001); trnV intron (Wang et al. 2003); trnG-trnS (Zhang et al. 2005). PCR was carried out on a MWG Primus thermal cycler using the following parameters: initial denaturation at 94°C for 3 min followed by 35 cycles of denaturation at 94°C for 1 min, annealing at 55°C for 1 min (48°C for petB-petD), extension at 72°C for 2 min and a final extension at 72°C for 5 min. PCR was carried out in a total volume of 20 μl containing 200 ng genomic DNA, 20 pmol of forward primer, 20 pmol of reverse primer, 1× PCR reaction buffer (7.5 mM Tris–HCl (pH9.0), 2.0 mM [NH4]2SO4, 5.0 mM KCl, 2.0 mM MgCl2) and 2.0 U BIOTOOLS DNA polymerase. Five micro litre PCR product was resolved on 1.5% agarose gels and visualised by ethidium bromide staining and the remaining 15 μl sequenced commercially (Macrogen, Korea). Sequences were aligned using the clustalw program in the BioEdit software package.

Table 2 Juniper chloroplast SNP allele-specific PCR (AS-PCR) primers
Full size table

To facilitate inexpensive, large-scale genotyping of SNPs, mutations detected in the chloroplast sequences were converted into allele-specific PCR (AS-PCR) primer sets. These mainly used the nested competitive primer approach of Soleimani et al. (2003) but a pair of specific PCR primers was also used to screen for length variation in the trnT-trnD region using standard PCR. For nested competitive primer design, the selective primer was designed so that the 3′ nucleotide of the primer was the SNP position and had an annealing temperature of 58°C. Compatible flanking primers, also with annealing temperatures of 58°C, were designed approximately 100 bp upstream and downstream of the SNP. In total, five SNPs were assayed in all samples using these approaches (Table 2). The AS-PCR protocol was as follows: initial denaturation at 94°C for 3 min followed by 11 touchdown cycles of denaturation at 94°C for 60 s, annealing at 65°C for 60 s (−0.7°C per cycle), extension at 72°C for 60 s followed by 24 cycles of denaturation at 94°C for 60 s, annealing at 58°C for 60 s, extension at 72°C for 60 s and a final extension at 72°C for 5 min. PCR was carried out in a total volume of 10 μl containing 100 ng genomic DNA, 10 pmol of forward primer, 10 pmol of reverse primer, 10 pmol SNP-selective primer, 1× PCR reaction buffer (5 mM Tris–HCl (pH9.1), 1.6 mM [NH4]2SO4, 15 μg/μl BSA), 200 μM each dNTP, 2.5 mM MgCl2 and 0.5 U Taq polymerase (Genetix). The trnT-trnD PCR protocol was as follows: initial denaturation at 94°C for 3 min followed by 35 cycles 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 both assays, PCR products were resolved on a 2% agarose gels and visualised by ethidium bromide staining.

Data analysis

Nuclear microsatellite allele sizes were scored using a 10 bp ladder and were checked by comparison with previously sized control samples. Levels of polymorphism measured as allelic richness (A R) and expected heterozygosity (H E) were calculated using the fstat software package (V2.9.3.2; Goudet 2001) and the popgene software package (V1.32; Yeh et al. 1997) respectively. Polymorphisms at the five chloroplast SNPs were combined to give multi-locus haplotypes. For both nuclear and chloroplast markers, interpopulation differentiation and differentiation between regions (see Table 1) were estimated from allele and haplotype frequencies using Φ-statistics, which give an analogue of F-statistics (Weir and Cockerham 1984) calculated within the analysis of molecular variance (AMOVA) framework (Excoffier et al. 1992), using the arlequin software package (V3.01; Excoffier et al. 2005). To facilitate comparisons with future studies, we also calculated a standardized value of population differentiation, F′ST(N), from the nuclear microsatellite data set, as this statistic is independent of the levels of variation detected within populations (Hedrick 2005). Population pairwise estimates of gene flow based on nuclear microsatellites were calculated using the private alleles method (Slatkin 1985; Barton and Slatkin 1986) as implemented in the genepop software package (V3.4; Raymond and Rousset 1995). Population pairwise F ST values were also calculated using genepop and significance of population differentiation was estimated using the genic differentiation option in genepop after sequential Bonferroni correction for multiple tests. To further identify possible spatial patterns of gene flow, the software package baps (V3.2; Corander et al. 2003) was used to identify clusters of genetically similar populations using a Bayesian approach. Ten replicates were run for all possible values of the maximum number of clusters (K) up to K = 19, the number of populations sampled in the study, with a burn-in period of 10,000 iterations followed by 50,000 iterations. Multiple independent runs always gave the same outcome. Finally, a test for isolation by distance (IBD; Rousset 1997) was carried out using a Mantel test to assess the relationship between genetic distance, measured as F ST/(1−F ST), and geographical distance in genepop.

Results

Levels of within-population genetic variation

The three microsatellite loci used in this study were moderately to highly polymorphic, with numbers of alleles ranging from 8 (JC16) to 35 (JC32). Within-population levels of expected heterozygosity averaged across loci ranged from 0.460 in the ROS population to 0.765 in the PNO population (Table 3). Levels of allelic richness averaged across loci ranged from 2.733 in the MOY population to 4.265 in the PNO population.

Table 3 Nuclear microsatellite diversity statistics: A R—allelic richness; H E—expected heterozygosity
Full size table

Analysis of a total of 4,735 bp of sequence from eight regions of the chloroplast genome in one individual per population revealed only four substitutions and two indel mutations (Table 2). AS-PCR analysis of these mutations in the complete sample gave rise to six haplotypes (Fig. 2). All populations were variable and within-population chloroplast diversity values ranged from 0.200 in the BAR population to 0.600 in the MOY population (Table 4).

Fig. 2
figure 2

Distribution of chloroplast AS-PCR haplotypes

Full size image
Table 4 Distribution and frequency of chloroplast AS-PCR haplotypes. H – gene diversity
Full size table

Population structuring and levels of gene flow

The analysis of molecular variance (AMOVA) revealed significant differences between populations for both nuclear (Φ ST(N) = 0.0957; P < 0.001) and chloroplast (Φ ST(C) = 0.2491; P < 0.001) markers (Table 5). The standardised estimate of population differentiation based on nuclear microsatellite markers, F′ST(N), was 0.429. The three-level AMOVA suggested that the majority of between-population variation for nuclear markers was due to differences between regions (Φ CT(N) = 0.0755; P < 0.001) but that between-region differentiation was not a significant factor for chloroplast markers (Φ CT(C) = 0.0526; NS). Despite this, there was evidence of some geographical substructuring of chloroplast haplotypes: Haplotype 3 was found only in the three populations from the Mournes area in the northeast (CAS, GUL and ANN), Haplotype 4 was restricted to the far northwest populations (FAN and PNO) and Haplotype 6 was only found in one population (CAP) from the far southwest.

Table 5 Analysis of molecular variance (AMOVA)
Full size table

Levels of gene flow between pairs of populations calculated from private alleles at nuclear microsatellite loci ranged from 0.38 (GUL/BAR) to 3.95 (FAN/MON) with a mean value of 1.15 (Table 6) and a global value (i.e. across all populations) of 1.09. Although it has been suggested that calculation of Nm values gives an indirect estimate of historical, rather than contemporary, levels of gene flow, the approach has been widely used and comparison with other studies in outcrossing coniferous tree species may be informative (see Discussion). Over half (87 of 171) of the values were less than 1.00, which represents the theoretical threshold for population differentiation due to genetic drift (Wright 1951). 162 of 171 population-pairwise F ST values were significantly different from zero, with values ranging from 0.002 (CAP/CAR) to 0.453 (ROS/CUR) and a mean of 0.103. Six of the nine non-significant F ST values were between populations from the same region. No evidence for isolation by distance was detected.

Table 6 Above diagonal: population pairwise estimates of gene flow (Nm) calculated from nuclear microsatellite data using the private alleles method of Barton and Slatkin (1986). Below diagonal: population pairwise F ST values calculated from nuclear microsatellite data. NS—non-significant F ST value
Full size table

The BAPS analysis identified nine genetic clusters (Fig. 3a). In general, populations from the same region were assigned to the same cluster with the exception of the Lough Derg populations, where the PMN population was assigned to a cluster of its own whereas the CAR population was grouped with the FAN and MON populations. The Voronoi tessellation (Fig. 3b) further highlights the spatial organisation of the genetic clusters, with clusters containing multiple populations usually comprising geographically proximal populations. The only exceptions to this are the grouping of the PNO and CUR populations, and the grouping of the CAR, FAN and MON populations as described above.

Fig. 3
figure 3

(a) Colour-coded assignment of populations to nine clusters using the BAPS software package. (b) Voronoi tessellation showing spatial organisation of populations in nine clusters delineated by BAPS. Colours as in (a)

Full size image

Discussion

Allele-specific PCR as a tool for population and conservation genetics

To our knowledge, this represents the first population genetics study to utilise AS-PCR for high-throughput screening of SNP variation. SNP genotyping techniques range from simple, PCR-based assays that can be resolved on standard agarose gels such as PCR-RFLP, to more complicated methods requiring the use of fluorescently labelled primers and/or dideoxynucleotides and polyacrylamide gel or capillary electrophoresis such as single base extension (SBE) or allele-specific primer extension (ASPE; Morin et al. 2004). Although PCR-RFLP approaches are cheap and technically simple, only a small fraction of SNPs give rise to restriction site changes. Whilst not as amenable to multiplexing as other SNP assays and not as straightforward when applied to diploid nuclear genes, the three-primer AS-PCR technique allows reliable and cost-effective genotyping of organellar SNP variation for large-scale population genetic analyses, particularly where the SNP does not result in a restriction site gain or loss.

Population differentiation and restricted gene flow

Although levels of gene flow in outcrossing, wind-pollinated tree species such as juniper are expected to be high, the findings of the present study are contrary to this. Our value for population differentiation based on nuclear loci (Φ ST(N) = 0.0957) is slightly higher than the average value for outcrossing gymnosperm species (0.073) quoted by Hamrick and Godt (1996) but slightly lower than the mean value quoted for biparentally inherited markers in conifers (0.116) by Petit et al. (2005). The standardised value of population differentiation, F′ST(N), was much higher (0.429), reflecting the high levels of within-population variation detected by microsatellites. Population differentiation based on chloroplast markers was also high (Φ ST(C) = 0.2491). Previous studies using chloroplast markers in conifers have tended to find around 10% or less of the total genetic variation partitioned between populations (e.g. Provan et al. 1998 (Φ ST(C) = 0.032 in Pinus sylvestris); Vendramin et al. 2000 (R ST(C) ≈ 0.1 in Picea abies); Richardson et al. 2002 (Φ ST(C) = 0.046 in Pinus albicaulis); Robledo Arnuncio et al. 2005 (Φ ST(C) = 0.031 in Pinus sylvestris); Naydenov et al. 2005 (Φ ST(C) = 0.110 in Pinus banksiana), 2006 (Φ ST(C) = 0.061 in Pinus nigra)) and in cases where high levels of population differentiation have been reported, these inflated values tend to be the result of long-term isolation of populations (e.g. Vendramin et al. 1998 (Φ ST(C) = 0.254 in Pinus pinaster); Jaramillo-Correa et al. 2006 (Φ ST(C) = 0.295 in Picea chihuahuana)). Where data are available for both the nuclear and chloroplast genomes in gymnosperms, as is the case in this study, differentiation between populations is expected to be more marked for chloroplast markers than for nuclear markers (Ennos 1994; Hu and Ennos 1997, 1999). Empirical studies, however, have generally found comparable levels of differentiation in both classes of markers which have been attributed to the high dispersal capabilities of pollen in conifers (Dong and Wagner 1994; Latta and Mitton 1997; Viard et al. 2001; Ribeiro et al. 2002). In the present study, differentiation based on chloroplast markers (Φ ST(C) = 0.2491) was much higher than that calculated for nuclear markers (Φ ST(N) = 0.0957), which is consistent with the action of genetic drift on the smaller effective population size of the uniparentally transmitted, haploid chloroplast genome. The limited dispersal suggested by both the nuclear and chloroplast Φ ST values is also reflected in the baps analysis, which delineated nine genetic clusters that are largely congruent with the spatial organisation of populations studied.

To date, there have only been two published population-level genetic studies on Juniperus communis. Oostermeijer and de Knegt (2004) used allozymes to assess the levels and distribution of genetic diversity in twelve populations from fragmented heathlands in the Netherlands and found low (F ST(N) = 0.026) levels of population differentiation. In contrast, a study using AFLPs on eight populations from England and Wales suggested a high level of genetic structuring, although summary statistics for population differentiation (e.g. F ST/G ST/Φ ST) were not calculated (van der Merwe et al. 2000). Of particular note, though, is a study on the congeneric J. przewalskii, which revealed very high levels of differentiation using chloroplast markers (G ST = 0.772; Zhang et al. 2005). Although no evidence of isolation by distance was evident at the global scale (i.e. across all populations) in the present study, suggesting the predominance of genetic drift over gene flow, individual values for inter-population differentiation and gene flow suggest that there may be adequate gene flow at local scales to prevent population divergence. Six of the nine non-significant pairwise F ST values were between populations from the same geographical region and the average value of Nm between populations from the same region (1.94) was almost double that of the average figure between populations from different regions (1.09). Values of Nm in conifers tend to be much higher, with values of Nm > 3 being the norm (Ledig 1998). Although Nm values give an indication of historical gene flow, the decline in juniper populations over the last few hundred years means that these values probably overestimate contemporary levels of gene flow and thus the degree of connectivity between extant populations is even lower. Zhang et al. (2005) reported that field studies on J. tibetica revealed no wind-mediated pollen dispersal beyond 2 km and in all six cases in the present study where populations were separated by less than this distance (BRO versus TRI, GOR versus TRI, BRO versus GOR, GUL versus CAS, ANN versus CAS and GUL versus ANN), population-pairwise F ST values were non-significant. Seeds in juniper are primarily dispersed by thrushes of the genus Turdus (Livingston 1972; Snow and Snow 1988) but a study on thrush communities in fragmented Juniperus thurifera populations has suggested that a decrease in abundance of frugivorous birds from smaller patches of woodland has had a negative impact on dispersal and seedling recruitment (Santos and Telleria 1994; Santos et al. 1999). Taken together, the potentially limited capacity for dispersal within and between fragmented populations via both pollen and seeds may explain the high levels of genetic differentiation found in Irish juniper populations. Ennos (1994) described a method to calculate the relative rates of interpopulation seed and pollen flow using a combination of maternal and biparentally inherited markers. In conifers, this generally uses data from the mitochondrial genome since the chloroplast genome is almost always transmitted paternally, unlike in angiosperms where maternal inheritance of the chloroplast genome occurs in the vast majority of taxa. Such calculations are unlikely to be feasible for the present study, however, since previous evidence suggests that the mitochondrial genome may be paternally inherited in the Cupressaceae, which includes juniper. Neale et al. (1989) described paternal inheritance of mitochondrial DNA in the coast redwood Sequoia sempervirens and cytological studies have shown the cytoplasmic inheritance of paternal mitochondria in other members of the Cupressaceae (Camefort 1970; Chesnoy 1973).

Conservation implications

Juniper populations in both the UK and Ireland have been in decline for many years now and one of the goals of the Species Action Plan is to maintain and re-establish natural populations. Information from population genetic studies is now considered an integral part of conservation programmes (Haig 1998) and the findings of this study are particularly relevant to the conservation of juniper in Ireland, where populations tend to be highly fragmented. The relatively high levels of genetic differentiation between populations and the apparent geographical structuring of this variation coupled with the occurrence of geographically localized haplotypes suggest that the concept of provenance should be taken into account when formulating conservation strategies for Irish populations of juniper. One obvious starting point for the designation of distinct management units would be the genetic clusters identified by the BAPS analysis which tend to reflect the limited levels of gene flow at larger geographic scales as described above. Of particular note for conservation purposes are the populations from the Mournes area: the region is geographically distinct and isolated from the remainder of the populations in Ireland, which have a predominantly western distribution, and almost a third (18 of 59) plants studied from this area exhibited an endemic chloroplast haplotype.

One of the main perceived threats to juniper populations is the lack of recruitment from seed (reviewed in Thomas et al. 2007) and establishment of seedlings has been shown to be negatively affected by both grazing (Ward 1973; Gilbert 1980) and climatic factors (Rosen 1988, 1995; Garcia et al. 1999). Overgrazing may present a particular problem to many of the populations examined in this study, particularly those occurring in montane and rough pasture hillsides where effective fencing is problematic. Coastal populations tend to be out of the reach of many grazing animals but their persistence on cliff faces exposes them to windthrow and, consequently, many of these populations comprise limited numbers of stunted trees. These threats to seedling establishment are further exacerbated by low levels of seed viability coupled with limited dispersal. Verheyen et al. (2005) showed that only 3% of seeds collected from a managed nature reserve in Belgium were viable and seed viability may be even more limited for Irish populations, which have an age structure skewed towards mature and old plants, since older stands tend to have lower reproductive capacity (Dearnley and Duckett 1999; Preston et al. 2007). The limited dispersal suggested by the findings of the present study is reflected by field observations which suggest that thrushes responsible for seed dispersal in juniper tend to favour larger, berry-rich populations over smaller, isolated populations even where individual plants within smaller populations produce large numbers of berries (Garcia et al. 2001). Consequently, the small, isolated populations comprising mainly senescent plants examined in this study may be at particular risk of ongoing loss of diversity and extinction. Conservation efforts aimed at the maintenance and reintroduction of these populations may be most effective when ex-situ management of seed and seedlings is implemented.