Introduction

Molecular markers are now routinely used to contribute towards programmes concerned with the conservation of endangered species. Conservation genetics research has proved valuable in assessing the extent and consequences of genetic erosion (Young et al, 1996; Saccheri et al, 1998), in clarifying taxonomic issues to provide guidelines on which entities to preferentially conserve (Soltis and Gitzendanner, 1998), and in gaining insights into modes of reproduction and patterns of contemporary and historical dispersal (Sork et al, 1999).

This latter topic, essentially using molecular markers to investigate reproductive ecology, is relevant to many species of conservation concern, as halting and reversing the decline of a given species ultimately requires successful reproduction and dispersal. The mode of reproduction and the spatial scales of effective dispersal impact on both management strategies of individual sites and the maintenance of networks of populations at the landscape level. In plant species that can reproduce both sexually (via pollen and seed) and asexually (via vegetative spread), it can be exceedingly difficult to gauge the relative importance of each mode from field observations alone. Likewise, physical tracking of pollen and seed dispersal is also problematical beyond extremely local scales. Polymorphic genetic markers are well suited to tackling these issues, and hence are increasingly being applied in a conservation context.

Spiranthes romanzoffiana (Irish Lady's-tresses) is a terrestrial, herbaceous, perennial orchid, widespread and relatively frequent in North America, yet confined to the western fringes of the British Isles within Europe (Preston et al, 2002; Figure 1). Its localised European distribution has led to increasing conservation interest; it is now listed in the UK Biodiversity Action Plans as a conservation priority species (UK Biodiversity Group, 1999), and recognised by the IUCN Orchid Specialist group as having ‘critically low populations’ in Europe (IUCN/Orchid Specialist Group, 1996).

Figure 1
figure 1

Distribution of S. romanzoffiana in Europe. Closed triangles: 1987–1999 records. Open triangles: pre-1987 records (data from Preston et al, 2002). The map grid represents 100 km squares.

Full size image

One of the major factors behind this conservation concern relates to the mode of reproduction of European populations of S. romanzoffiana. Pollination of this species in North America occurs via unspecialised, medium-sized long-tongued bees, which are attracted to nectar produced from two bosses near the base of the labellum. Similar potential pollinator species occur within the plant's habitat in Europe, and flower visits have been observed (J Robarts, 2002, personal communication). Despite this, no seed set has been recorded in the European range, leading to suggestions that the plant's persistence there is attributable to vegetative reproduction (Horsman, 1994). Plants produce lateral buds in the autumn, which over-winter and develop into the following year's aerial parts. The only documented mechanism for vegetative reproduction is the occasional production of twin (rarely three or four) lateral buds, which can lead to the formation of multiple aerial parts; these may eventually divide to form different ramets (Summerhayes, 1968). However, while this mechanism may lead to clonal spread over very small areas, it is difficult to imagine this mode of reproduction being effective over larger scales (eg >1 m).

A second area of conservation concern relates to the populations showing a high level of demographic instability; for instance, no population recorded in Scotland prior to 1981 is now known to be extant (UK Biodiversity Group, 1999). New populations have been recorded, but this frequent extirpation of populations leads to difficulties in assessing the distribution and conservation status of this species.

There is thus something of a paradox about European populations of S. romanzoffiana. On the one hand, populations appear to have a high turnover, with known populations disappearing and new populations being discovered. On the other hand, there is no satisfactory explanation for how the plant reproduces and disperses. Seed set has not been observed, and although vegetative reproduction has been invoked, the species' apparently limited capacity for clonal spread makes this an unlikely mode of the formation of new populations. We have used chloroplast microsatellites (cpSSRs; Provan et al, 2001) and amplified fragment length polymorphisms (AFLPs; Vos et al, 1995) to establish whether reproduction is primarily sexual or asexual, and to examine population differentiation in order to infer the spatial scales over which effective gene flow is occurring.

Materials and methods

Plant material and isolation of genomic DNA

To assess genetic variation in European populations of S. romanzoffiana, individuals were sampled from 17 populations: eight from Scotland and nine from Ireland (Table 1; see also Figure 2a). Sampling was designed to assess variation at the within-population, among-population and regional levels. In 10 populations, between 10 and 25 individuals were sampled and examined for both AFLP and cpSSR variation. While this sampling primarily consisted of individuals separated by at least 5 m, in eight of the populations five individuals were sampled within a 1 m2 quadrat to look for evidence of local clonal growth. From a further seven populations, between four and seven individuals were sampled for cpSSR analyses only (Table 1). For all molecular analyses, a section of leaf material approximately 30 mm long was harvested from single leaves and stored in silica gel.

Table 1 Accession details of 17 European populations of S. romanzoffiana
Full size table
Figure 2
figure 2

(a) Distribution map showing the location of the sampled populations. Numbers correspond to the population names in Table 1, and symbols correspond to different chloroplast haplotypes (squares A20, southern group; triangles A19, northern group). The map grid represents 100 km squares. (b) Sample electropherogram and partial sequences showing cpSSR variability. The peak traces correspond to the difference in poly-A repeat number.

Full size image

DNA isolation was performed with DNeasy® Plant Mini Kits (Qiagen Ltd, UK). DNA quality and concentration was assessed by running samples alongside Hyperladder concentration marker (Bioline, UK) on a 1% agarose gel in 1 × tris borate/EDTA buffer, and visualised by staining with ethidium bromide.

Chloroplast microsatellite analysis

Five cpDNA regions (Table 2) were screened for microsatellite loci among 11 accessions representing a range of populations. Amplification of chloroplast regions was performed via polymerase chain reaction (PCR) on an MJ Research PTC-200 DNA Engine thermal cycler in 50 μl reactions containing 5 μl 10 × NH4 reaction buffer (Bioline, UK), 5 μl 2 mM deoxynucleotide triphosphate (dNTP), 2.5 μl 50 mM MgCl2, 1.5 μl of each primer (10 mM), 33.5 μl dH2O, 1.25 U Biotaq polymerase (Bioline, UK) and 1.0 μl DNA template. The thermocycle profile was: initial denaturation for 4 min at 94°C, followed by 30 cycles of 45 s at 92°C, 45 s at 53–62°C (Table 2) and 3 min at 72°C, with a final extension for 10 min at 72°C. PCR products were run on 1% agarose gels to check for amplification success and quality.

Table 2 cpDNA regions sequenced to search for polymorphic microsatellite loci in S. romanzoffiana
Full size table

Amplified fragments were purified using QIAquick™ PCR purification kits (Qiagen Ltd, UK) following the manufacturer's protocol, and sequenced using the dideoxy chain-termination method. Cycle sequencing was performed in 20 μl reactions containing 4 μl Thermo Sequenase II (Amersham Pharmacia, UK), 0.5 μl primer (10 mM), 13.5 μl dH2O and 2 μl of purified PCR product, under the following PCR conditions: 25 cycles of 96°C for 10 s, 50°C for 5 s and 60°C for 4 min. Sequencing PCR products were purified following the manufacturer's instructions, then run on an ABI Prism™ 377 DNA Sequencer, before editing and manual alignment using Sequence Navigator™ 1.0.1 (Applied Biosystems Inc., USA).

Initial screening revealed eight mononucleotide repeat loci >8 bp in length, located in four of the five cpDNA regions amplified (atpB-rbcL (T)10, (T)14, (T)15, (A)20; psbC-trnS (T)9; trnH-trnK (A)13; trnL-trnF (T)17, (A)19 or 20). Underlined numbers indicate repeat motifs with single base-pair interruptions. Only the poly-A repeat located in the trnL intron showed any polymorphism among the 11 accessions studied. Forward and reverse primers were designed to amplify this microsatellite region (forward primer: 5′-GGTAACTTCCAAATTCAGA-3′; reverse primer: 5′-ACAGCTTCCGTTGAGTCTC-3′) using Primer 3 (www.basic.northwestern.edu/biotools/Primer3.html). Amplification was carried out in 10 μl reactions containing 0.5 μl of each of the two primers (10 mM), 1 μl 10 × NH4 reaction buffer, 0.5 μl 50 mM MgCl2, 1 μl 2 mM dNTPs, 6 μl dH2O, 0.2 U Biotaq polymerase and 0.5 μl DNA template. PCR conditions were as follows: 7 min at 95°C, followed by 30 cycles of 15 s at 95°C, 15 s at 50°C and 15 s at 72°C, with a final extension for 15 min at 72°C. Amplification products were diluted 130-fold in dH2O and analysed on 5% Long Ranger® acrylamide gels (BMA, USA) on an ABI Prism™ 377 DNA sequencer. Fragment size was determined by comparison with Rox 500 GeneScan size standard using Genotyper 2.0® software (Applied Biosystems Inc., USA).

Amplified fragment length polymorphism

Accessions were analysed following the plant mapping protocol of the PE Applied Biosystems AFLP™ kits. In all, 13 primer combinations were assessed for clarity, reproducibility and polymorphism among 12 accessions representing a range of populations, with three primer combinations selected for further analysis (A=EcoRI-ACT/MseI-CTA, B=EcoRI-ACT/MseI-CAG, C=EcoRI-ACA/MseI-CAG). Selective amplification products were run on 2% agarose gels to check for amplification success, and then electrophoresed on 5% Long Ranger® acrylamide gels; multilocus profiles were visualised using GeneScan Analysis® 3.1.2 software (Applied Biosystems Inc., USA). Markers were scored as present/absent using Genotyper® 2.0, and a binary data matrix constructed encompassing markers in the range 50–400 bp. Markers scoring below 50 arbitrary units of fluorescence were disregarded. Reproducibility of amplification was assessed by replicate extractions and amplifications of a subset of the samples. Only repeatable and unambiguous markers were included in the analysis.

Data analysis

Standard descriptive statistics were obtained to assess patterns of within-population variation: P, the proportion of loci that are polymorphic; PD, the proportion of distinguishable genotypes; APD, the average number of pairwise differences of markers among individuals. To establish the apportionment of variation into within- and between-population components, an analysis of molecular variation (AMOVA) was undertaken using Arlequin version 2.0 (Schneider et al, 2000), and the significance of FST analogues (φST) was assessed by permutation tests.

Results

cpSSRs

Polymorphism at the trnL microsatellite locus revealed two haplotypes differing by a single base pair (A19 or A20), resulting in fragments of 139 or 140 bp, respectively (Figure 2b). No intrapopulation variation was detected, but there was clear geographically structured interpopulation variation (Figure 2a). The A19 repeat occurred in all accessions from Coll, Vatersay and Barra (hereafter referred to as the northern group), and the A20 repeat occurred in all accessions from Colonsay and Ireland (hereafter referred to as the southern group).

AFLP analysis

A total of 138 unambiguous AFLP markers were generated from the 205 individuals that gave clear amplification profiles. Of these 138 markers, 86 were polymorphic (P=0.62) either within or among populations (Table 3).

Table 3 Within-population AFLP diversity in S. romanzoffiana
Full size table

Estimation of differentiation among all populations was high (φST=0.892, P<0.01). A hierarchical AMOVA analysis was conducted based on the geographical regions corresponding to the two chloroplast haplotypes (northern and southern groups). Between-group variation accounted for 85.8% of the total variation, with 3.5% of variation among populations within groups and 10.8% among individuals within populations. In total, 10 markers were differentially fixed between the northern and southern groups. Average pairwise distances among multilocus genotypes were also high when comparing populations from the two groups (Table 4).

Table 4 Average pairwise distances (APD) within and among populations of S. romanzoffiana based on 138 AFLP loci
Full size table

The northern populations (Coll, Barra and Vatersay) showed lower levels of differentiation than that among all populations (φST=0.193), but differentiation was still significant (P<0.01). Little differentiation was detected among the southern populations (Colonsay and Ireland), although one population on Colonsay showed a fixed difference for a single marker in all individuals from the other southern populations. Obtaining meaningful estimates of population differentiation within the southern group was hampered by the paucity of polymorphism within and among these populations (see below).

There is a striking difference in the levels of intrapopulation diversity in northern vs southern populations (Table 3). The vast majority (90/92) of individuals from the northern populations harboured unique multilocus genotypes (PD=0.98), the only exception being a pair of accessions from Vatersay, and shared multilocus genotypes among single accessions from Vatersay and Barra (SC-BA1). All multilocus genotypes were unique to single populations except this latter pair. All 20 samples collected within 1 m2 quadrats (five plants from each of four northern populations) had different multilocus genotypes, consistent with sexual reproduction rather than local clonal spread. In contrast, only 12 unique multilocus genotypes among 113 individuals (PD=0.11) were detected from the southern populations, most being present only at low frequencies. The majority of these multilocus genotypes differed from each other by only a single marker (APD=0.37, range 0–4). Artefactual differences attributable to subtle variations in AFLP reaction conditions cannot be excluded as the source of these differences.

Discussion

Regional geographic structure

Both AFLPs and cpSSRs provide strong evidence for a marked genetic–geographic divide in European populations of S. romanzoffiana. Differential fixation of chloroplast haplotypes and 10 AFLP markers suggests a lack of contemporary gene flow among the northern (Coll, Barra, Vatersay) and more southerly (Colonsay, Ireland) populations. This split was previously unsuspected, and does not correlate with any documented morphological or ecological differences between these population groups.

Within-region population genetic structure (northern populations)

The northern populations (Barra, Coll, Vatersay) show a moderate degree of population differentiation (φST=0.193) but high levels of intrapopulation genotypic diversity. Although seed set has not been observed in European populations of S. romanzoffiana, the genotypic diversity in these northern populations suggests that seed set and sexual recruitment has taken place. The lack of individuals with shared multilocus genotypes (even within 1 m2 quadrats) indicates that vegetative reproduction does not play a major role in the perpetuation and dispersal of individuals.

It could be argued that the high levels of genotypic diversity, coupled with an absence of observed seed set, could be attributable to trans-atlantic seed dispersal (seed set is frequently observed in North American populations). We feel that this is exceedingly unlikely given (a) the considerable number of distinct genotypes and plants involved, and (b) the marked regional differentiation, which would require differential migration of seeds with the different chloroplast haplotypes into the northern and southern regions.

Instead, the most parsimonious explanation of our data is that sexual reproduction has taken place in situ. This reproduction could be historical, or current-but-infrequent. It is difficult to distinguish between these two scenarios, but it is worth noting that an absence of observed seed set does not necessarily equate to a genuine absence of seed set. Sexual reproduction could take place via occasional overlooked fertilised flowers; a typical individual terrestrial orchid flower can produce 1000–10 000 seeds (Arditti, 1992). Another potential theory is that a low level of seed set occurs within individual capsules, such that the capsules do not swell and appear fertile, but nevertheless contain a few viable seeds (J Robarts, 2002; M Light, 2003, personal communication). Detailed observations of flowering spikes and capsules, along with pollen viability, seed viability and crossing experiments, are needed to establish the precise conditions under which reproduction occurs.

Until such studies are completed, grazing controls should be established in these northern populations of S. romanzoffiana during the flowering and fruiting period. Sheep can have devastating effects on S. romanzoffiana flower spikes, and entire populations can be stripped of all flowering spikes in a matter of hours or days when flocks are moved into the orchid-containing pastures. As these populations were believed to reproduce entirely vegetatively (Horsman, 1994), the reproductive consequences of this grazing have been considered to be small. However, as the genetic data from this study suggest that reproduction is predominantly sexual, and yet field observations suggest that seed set is rare, the flowering spikes should be protected from herbivory to maximise the chances of successful seed production and periodic sexual recruitment.

Within-region population genetic structure (southern populations)

In contrast to the data from the northern populations, the more southerly populations are typified by extremely low levels of genotypic diversity, with identical multilocus genotypes being detected in geographically disparate populations. Such genetic uniformity within and among populations could be attributed to efficient clonal growth and dispersal. In clonal plants, if propagules from a single genotype are able to disperse, large areas can be covered by ramets of a single genet (eg Hollingsworth and Bailey, 2000). However, given the lack of evidence for clonal spread in the morphologically similar northern populations, this seems an unsatisfactory explanation. This is particularly evident when one considers the substantial geographical distances involved.

Agamospermy has been reported in some Spiranthes species from North America (Catling, 1982; Schmidt and Antlfinger, 1992). Although this has never been observed in S. romanzoffiana, several species that usually reproduce via sexual out-crossing have been observed to develop agamospermous races in colonising populations near the edge of their ranges (Catling and Catling, 1991). This offers a potential explanation for the patterns of genotypic similarity among populations in Ireland and Colonsay.

An alternative explanation is that reproduction in the southern populations is sexual, but that the genotypes involved are homozygous at the vast majority of the study loci. An extreme genetic bottleneck could lead to genetic uniformity among sexually reproducing individuals. Alternatively, self-pollination can lead to a rapid reduction in heterozygosity. Autogamy has been observed in other Spiranthes species (eg Catling, 1983a, 1990; Sipes and Tepedino, 1995; Sun, 1997). However, Catling's (1983b) observations on North American populations of S. romanzoffiana suggest that it is unlikely in this species. This is due to protandry, and also the column structure which prevents the pollinia from coming into contact with the stigmatic surface.

Data from variable codominant markers (eg nuclear microsatellites) are needed to distinguish between these alternative explanations. Hypervariable codominant markers could establish whether the observed genetic uniformity is associated with homozygosity (selfing), or random mating from a very narrow genetic base (bottleneck), or whether there is fixed heterozygosity (indicative of agamospermous reproduction).

Population differentiation

The overall estimate of population differentiation for European populations of S. romanzoffiana is high (φST=0.892). However, this figure includes two groups of populations fixed for different markers (apparently not exchanging genes), and potentially showing different reproductive strategies. In terms of evaluating the genetic connectivity of populations, it is perhaps more meaningful to examine patterns of differentiation among the sexual northern populations. Here, there is a smaller, but still significant, level of population differentiation (φST=0.193). It is also worth noting that if one considers pairwise differences among northern populations, an estimate of φST (0.217) is obtained between the two populations on Barra (separated by only 0.5 km), which is a similar order of magnitude to the estimate for differentiation among the northern populations per se (involving interisland distances of 70 km between Coll and Barra). This differentiation, evident even over small scales, may be related to the infrequent production of seeds, which will inevitably reduce opportunities for interpopulation gene flow.

This significant deviation from panmixia among proximal populations of S. romanzoffiana could be considered high when compared to average levels of differentiation among orchid populations. Hamrick and Godt (1996) summarised allozyme data from 16 orchid population genetic studies and obtained a mean estimate of GST for orchids of 0.087. They noted that orchids ‘had an exceptionally low mean GST… perhaps due to the species-specific pollinators characteristic of orchids, and to their tiny wind-borne seeds. Both of these traits could produce high rates of gene flow among populations’ (Hamrick and Godt, 1996, p 1297). However, it is worth stressing that allozyme work by Sun and Wong (2001, p 2186) led to exactly the opposite conclusion, that ‘gene flow appears to be much more restricted in wild orchids than in other plants’.

These opposing views triggered our own review of levels of differentiation among orchid populations (Table 5). To the best of our knowledge, this is the most comprehensive summary of orchid GST (or FST) estimates to date. Our estimates of population differentiation range from GST=0.012 to 0.924, with mean values of GST=0.187 (all studies), GST=0.219 (excluding studies examining less than five populations) and GST=0.161 (further excluding studies examining dominant marker data sets). These mean estimates are somewhat higher than the ‘exceptionally low mean GST’ of 0.087, and within the range recorded for several other plant families by Hamrick and Godt (1996). In this respect it is worth noting that while orchid species share biological attributes such as small, mostly dust-like seeds, they are also a large, heterogeneous and diverse group (Arditti, 1992; Rudall and Bateman, 2002). The Orchidaceae contains ca 19 000 species encompassing an exceptionally wide range of reproductive strategies, variation in generation times, and includes both narrow endemic and widespread species, as well as species with relatively continuously distributed ranges and species with only isolated populations. As such, they should be expected to show, and do show, considerable variation in the levels of population differentiation.

Table 5 Genetic diversity among populations of orchid species based on literature survey
Full size table