|Home | About | Journals | Submit | Contact Us | Français|
In the Mediterranean basin, the Italian peninsula has been suggested to be one of the most important glacial refugia for temperate tree species. The orchid genus Epipactis is widely represented in the Italian peninsula by widespread species and several endemic, localized taxa, including selfing and outcrossing taxa. Here the phylogenetic and phylogeographic relationships in a group of closely related taxa in Epipactis are investigated with the aim of understanding the role of this refugial area for cladogenesis and speciation in herbaceous species, such as terrestrial orchids.
Ribosomal DNA (rDNA) was employed to assess phylogenetic relationships, and plastid sequence variation in the rbcL–accD spacer was used to reveal phylogeographic patterns among plastid haplotypes using a parsimony network.
Low genetic variation and shared ribotypes were detected in rDNA, whereas high levels of sequence variation and a strong phylogeographic structure were found in the examined plastid region. The parsimony plastid haplotype network identified two main haplotype groups, one including E. atrorubens/microphylla/muelleri/leptochila and the other including all accessions of E. helleborine and several localized and endemic taxa, with a combination of widespread and rare haplotypes detected across the Italian peninsula. A greater genetic divergence separated the Italian and other European accessions of E. helleborine.
Phylogenetic and phylogeographic patterns support a working hypothesis in which the Italian peninsula has only recently been colonized by Epipactis, probably during the most recent phase of the Quaternary age and, nevertheless, it acted as a remarkable centre of diversification for this orchid lineage. Changes in pollination strategy and recurrent shifts in mating system (from allogamy to autogamy) could have represented the mechanism promoting this rapid diversification and the observed high taxonomic complexity detected in the E. helleborine species complex.
The Mediterranean region is one of the major biodiversity hotspots (Médail and Myers, 2004; Thompson, 2005) and hosts much plant and animal diversity and endemism especially in mountain areas of Iberia, Italy, the Balkans and north Africa (Blondel and Aronson, 1999), probably because, as glacial refugia, these areas suffered only limited species extinction and favoured the emergence of new taxa (Comes and Kadereit, 1998). Although the role of refugia in the maintenance of biodiversity during glacial periods is well established (Magri et al., 2006), it is still unclear if they also contributed to anagenesis. In fact, the present composition of refugial biota not only includes species that experienced long stasis in the area but also taxa that secondarily colonized the regions as a consequence of climatic fluctuation and range expansion. Indeed, it has been shown that, in several cases, species occupying ancient refugial areas still show different plastid DNA lineages in which plastid haplotypes are not correlated, thus suggesting that they had already evolved before the migration towards the refugium (Thompson, 1999). In this scenario, the Mediterranean basin could be viewed both as a global refuge for relict plants and as an area that, by means of secondary contacts and hybridization, allowed floristic exchange and active speciation (Médail and Myers, 2009). In fact, the secondary meetings of lineages that have followed independent evolutionary trajectories along different colonization routes in interglacial periods may represent the opportunity for these glacial refugia, under some circumstances, to play a more creative role potentially leading to the differentiation of locally adapted endemic ecotypes (or even genetically, morphologically and ecologically diversified species), thus promoting incipient insurgence of reproductive isolation mechanisms (Comes and Kadereit, 1998). In this context, even if a glacial refugium is considered as an area where distinct genetic lineages have persisted through a series of Quaternary climate fluctuations owing to special, buffering environmental characteristics (Mèdail and Diadema, 2009), recent studies have suggested that these refugia were more complex and heterogeneous (reviewed in Hewitt, 2004; Gomez and Lunt, 2007) than expected and were themselves strongly sub-structured, thus potentially offering the opportunity for speciation.
However, the main limitation of such evolutionary trajectories is that the time elapsed in glacial refugia is generally short and thus the speciation process is unlikely to be completed in such situations. Based on the available information (fossil records of tree species) it appears that the dominant response, during this period, was extinction rather than speciation or stasis, especially for tree species (Bennett, 1997). Nevertheless, in some instances, in the Mediterranean basin, the isolation was sufficient to promote speciation for genetic diversification (Willis and Niklas, 2004). In fact some localized endemic ecotypes, or ambiguous complexes of taxa, can rapidly evolve through mechanisms that accelerated the biological diversification, e.g. self-pollination (Stebbins, 1974; Levin, 2000) that provides instantaneous reproductive isolation (and speciation) from other lineages (Levin, 2000; Grundt et al., 2006). The insurgence of a strong reproductive isolation can impede gene flow among formerly interbreeding individuals and allows them to follow independent evolutionary trajectories, including fixation of locally adaptive morphological traits (Charlesworth, 2006). This rapid diversification can generate fixed morphological differences even in the absence of a clear genetic differentiation and, as a consequence, can produce taxonomic complexity within a species or genus (Bateman, 2006).
Plant species, such as orchids, with strong ecological constrains (i.e. restricted to specific habitats or with specific ecological requirements) may be more prone to experience population fragmentation, with consequent low availability of conspecific potential mates or low availability of pollinators, and be subject to local selection promoting an increase in self-pollination (Charlesworth, 2006). In this context, the orchid genus Epipactis is an extremely morphologically variable aggregate of species (Bateman, 2006; Hollingsworth et al., 2006), widely represented in the Italian peninsula by widespread species and several endemic, localized taxa, including both selfing and outcrossing taxa (Ehlers and Pedersen, 2000; Squirrell et al., 2002; Talalaj and Brzosko, 2008), mainly pollinated by social wasps (Müller, 1873; Ivri and Dafni, 1977; Vöth, 1982). In Epipactis, active diversification and breeding system transitions are likely to have caused taxonomic complexity of selfing taxa apparently closely related to E. helleborine (Richards, 1982; Hollingsworth et al., 2006). For instance, the number of species of the genus in Europe varies from 25 to 60 depending on the treatment (Richards, 1982; Delforge, 1995, 2005; Bateman et al., 2005) and mainly based on the taxonomic status attributed to local endemic taxa. Generally, the outcrossing species show widespread distributions and high morphological variability (Richards, 1982), whereas the self-fertilizing species are extremely localized and often show little within-population morphological variation (Ehlers et al., 2002; Squirrel et al., 2002).
Natural populations of Epipactis occupy a wide range of forest habitats, being common in damp coniferous and mixed forests and, less frequently, in dry boreal forests or even on coastal dunes and in urban sites in northern Europe (Hollingsworth and Dickson, 1997; Bateman et al., 2005). The association between Epipactis and surrounding forest trees is further strengthened by the need for the orchid to form mycorrhizal associations with fungi that are simultaneously ectomycorrhial on roots of neighbouring trees (Bidartondo et al., 2004; Selosse et al., 2004; Bitartondo and Read, 2008; Ogura-Tsujita and Yukawa, 2008). This ecological constraint suggests that the phylogeographic dynamics of Epipactis can be linked to the well-known phylogeographic patterns of trees in Europe (Petit et al., 2003), making this group of orchids an interesting model for investigating the consequences of geographical fragmentation and quaternary glaciations on a herbaceous group with a strong ecological constraint, such as the strict association with the mycorrhizal community associated with trees.
Here, aiming to understand the role of the Italian peninsula in promoting diversification of taxa in Epipactis, we analysed nuclear internal transcribed spacer (ITS) and plastid sequence variation for the reconstruction of patterns of phylogenetic and phylogeographic relationships among closely related taxa belonging to Epipactis helleborine sensu lato (s.l.) and related lineages. Unlike biparentally inherited nuclear markers, plastid DNA is inherited maternally in orchids (e.g. Cafasso et al., 2005) and therefore provides a seed-specific marker. Thus, the joint use of plastid and nuclear markers, in a phylogeographic context, can enhance the power to detect effects of hybridization among closely related taxa and help disentangle the effects of pollen vs. seed flow in shaping the phylogeographic structure of the current populations (King and Roalson, 2009).
Specifically we addressed the following questions. (a) Do the closely related Epipactis taxa result from multiple independent colonization routes into the Italian peninsula? (b) Did E. helleborine and related lineages diverge only recently in the Italian peninsula? (c) Do morphological taxonomic treatments reflect underlying patterns of genetic variability? (d) Do the documented transitions to a self-pollination strategy contribute to shaping the current patterns of lineage diversification and distribution in E. helleborine?
In the present study we sampled the widespread species and several endemic and geographically isolated taxa with variable floral morphology and facultative self-mating systems that occur in the Italian peninsula. We refer to E. helleborine sensu stricto (s.s.) for all samples that have been attributed to the widespread species on morphological grounds and to E. helleborine s.l. for all those localized, frequently endemic, taxa that have been reported, on morphological grounds, as species different from, but related to, E. helleborine s.s. A total of 1004 individuals from 151 populations were sampled during four successive springs (2005–2008). To obtain a representative sample of the genus in the Italian peninsula and bordering areas, material was collected in southern France, Slovenia and Austria, plus additional collection from more distant European countries (England, Germany and Spain). Epipactis palustris accessions were also included in the sampling. Collection sites, number of samples per population and related geographical information are listed in Table 1 and andinin Supplementary Data Table S1 (available online). Leaf material was stored in silica gel or frozen at –80 °C and was used for subsequent molecular analyses.
After powdering the frozen or silica gel-dried leaves in a mixer mill (Vibration Mill, Retsch, Germany), total DNA was extracted using a modified version of the cetyltrimethyl ammonium bromide (CTAB) method (Doyle and Doyle, 1987).
Approximately 100 mg of each sample was macerated in 800 µL of standard CTAB buffer and 1·8 µL of β-mercaptoethanol, and incubated at 60 °C for 30 min. Subsequently, the mixture was extracted twice with an equal volume of chloroform/isoamyl alcohol (24:1) and centrifuged at 10 000 rpm for 10 min. DNA was precipitated in isopropanol, centrifuged at 13 000 rpm at 4 °C for 20 min, washed in cold 70 % ethanol (4 °C) for 5 min and air dried for 15–30 min. DNA pellets were resuspended in 30 µL of distilled water and used for the PCR template. The quality of extracted DNA was examined by electrophoresis on 0·8 % agarose gels.
The nuclear ribosomal DNA (rDNA) ITS1 spacer was selected as a nuclear marker because it was found to show greater nucleotide variation than ITS2 (Roy et al., 2009) and was amplified from all accessions using primers and polymerase chain reaction (PCR amplification) conditions as described in Aceto et al. (1999).
In a preliminary screening for plastid DNA variation, a subset of samples representing the most common species of Epipactis was used as template for PCR amplifications and sequencing of several putatively variable plastid regions, namely the psbA–trnH intergenic spacer (Sang et al., 1997), trnL intron (Taberlet et al., 1991), rpoB, rpoC1, ndhJ (primer sequences available at the Royal Botanic Gardens, Kew, website http://www.kew.org/ barcoding/protocol.htlm) and the rbcL–accD spacer (Yasui and Ohnishi, 1998) in order to detect potential hotspots for genetic diversity in the plastid genome of these orchids. From sequence alignment of this preliminary screening the highest genetic diversity among the tested accessions was found in the intergenic spacer of the rbcL– accD region. Because of its length (>2 kb) it was decided to design an internal set of primers, namely CP1F (5′-CGATTACATTCGAGTTCGAC-3′) and CP2R (5′-GATTCATTTGAATTCTGATGAG-3′), located at the end of the rbcL gene and immediately before the beginning of the accD gene, respectively, for selectively amplifying an internal portion of 770 bp of the rbcL–accD region that contained the variable intergenic spacer and is a suitable length to allow direct PCR sequencing. This region was then amplified and double strand sequenced in all examined accessions.
All PCRs were performed in a reaction volume of 50 µL using 10 ng of template DNA, 200 µm of each dNTP, 10 pmol of each of the two primers, 1× Taq buffer (50 mm KCl, 10 mm Tris–HCl pH 9·0), 1·5 mm MgCl2 and 0·5 U of Taq polymerase (GE Healthcare, UK). The reaction was programmed on a Thermal Cycler (Applied Biosystems, USA) as one cycle of denaturation at 94 °C for 4 min, then 35 amplification cycles of denaturation at 94 °C for 30 s, annealing at 50 °C for 1 min, and extension at 72 °C for 2 min, followed by a 7 min extension.
Amplification products were visualized on a 1·5 % agarose gel and photographed after ethidium bromide staining. All successfully amplified DNA fragments were purified using GFX PCR DNA and Gel Band Purification Kit (GE Healthcare, UK) following the manufacturer's instructions, and then sequenced in both directions using a modification of the Sanger dideoxy method as implemented in a double-stranded DNA cycle sequencing system with fluorescent dyes (Applied Biosystems, USA). The same primers were used for both amplification and sequencing. Purifications of sequencing reaction products followed the ethanol–sodium acetate precipitation protocol provided with the sequencing kit. Products were resolved on an ABI 3130 Avant automated sequencer (Applied Biosystems, USA). All sequence polymorphisms were visually checked from the chromatograms. The raw sequencing data were edited using Sequencer analysis version 3·0. All singletons (accessions with unique substitutions) were sequenced twice to exclude the possibility that mutations were caused by errors during amplification.
All sequences were visually inspected for the presence of heterozygosity, edited with Sequence Analyzing software 3·7 (Applied Biosystems, USA) and aligned with BioEdit version 5·09 (Hall, 1999) with additional minor manual adjustments. Sequences have been deposited in GenBank under the accession numbers listed in Table 2. The nuclear (nrDNA) and plastid sequences were analysed separately.
Sequences from the nuclear ribosomal ITS1 spacer were used to assess phylogenetic relationships and genetic variation among the analyzed Epipactis taxa. Phylogenetic relationships were reconstructed by maximum parsimony analyses in PAUP* 4·0b10 (Swofford, 2000). Aphyllorchis caudata (FJ454866·1) was used as the outgroup. The parsimony analysis consisted of a heuristic search with 1000 random addition replicates, tree–bisection–reconnection (TBR) branch swapping and the multrees option in effect. All characters were treated as unordered and equally weighted. Support for recovered nodes was inferred by bootstrap analysis (Felsenstein, 1985) (1000 replicates; search parameters same as above).
Phylogeographic relationships among plastid haplotypes were analysed using TCS version 1·21 (Clement et al., 2000). The TCS program constructs haplotype networks, thus allowing loops and polytomies, by implementing the statistical parsimony algorithm described by Templeton et al. (1992). For assessing phylogeographic relationships among plastid haplotypes, we used all informative sites but, in one case, large gaps in a sequence due to multiple base insertions/deletions (indels) were coded as single mutations to avoid theoretical intermediate haplotypes created by the program, which interprets each gap as an independent mutation event.
Plastid haplotypes were used to compute parameters of genetic diversity (hs and ht) and genetic differentiation (Gst and Nst) to test the presence of phylogeographic structure. These parameters were estimated following the methods described by Pons and Petit (1995, 1996) using the program Permut (Pons and Petit, 1996, available at http: //www.pierrton.intra.fr/genetics/labo/software/permut). Gst is based solely on allelic frequencies, whereas Nst also takes into account the genetic relationship among plastid haplotypes. An Nst higher than the Gst usually indicates the presence of phylogeographic structure (Pons and Petit, 1996; Petit et al., 2005), i.e. closely related haplotypes are more often found in the same area than less closely related haplotypes. Finally, these two parameters were compared using the U-test (Pons and Petit, 1996).
The ITS1 aligned data matrix, including E. palustris, consisted of 226 characters of which only ten were variable (two of which were potentially parsimony informative). Seven different ribotypes were found, some of which were private to some lineages (R2 in E. atrorubens, R3 in E. microphylla, R4 in E. nordeniorum, R5 in E. helleborine from Spain; R6 in E. helleborine from Germany; R7 in E. palustris). In contrast, the ribotype R1 was shared among several accessions, namely E. helleborine s.s. and members of E. helleborine s.l., E. leptochila, E. muelleri and part of the sampled populations of E. atrorubens and E. microphylla.
All ribotypes differ from each other in a single mutation step and were different by one (R2, R3 and R5) or two (R4 and R6) mutational steps from the widespread R1 ribotype. No heterozygotes were detected in any sequenced accession. Ten mutational steps separated the R7 ribotype of E. palustris. The resulting maximum parsimony tree has low resolution and low bootstrap support due to the low number of synapomorphic characters (Fig.1). Using A. caudata as the outgroup, E. palustris is sister to a clade containing all other Epipactis species that belong to a group with no internal phylogenetic resolution and/or low bootstrap support (Fig. 1).
Preliminary screening of six plastid DNA regions detected few or no interspecific variations in all except the rbcL–accD region. From sequence alignments of this region we detected DNA polymorphisms allocated in the intergenic spacer that was then selected for amplification with specific primers (CP1F–CP2R) in all examined accessions. A total of 1004 plastid DNA sequences was generated, and the resulting alignment of this data matrix consisted of 770 characters with 27 variable sites. Of these, 20 were indels and base substitutions and seven were variable poly(A), poly(T) and poly(C) repeats (see Table 2). This genetic variation allowed the identification of 18 plastid haplotypes (Tables 1 and and2),2), the geographical distributions of which are shown in Supplementary Data Fig. S1 (available online). Of a total of 151 examined populations, only 12 contained more than one plastid haplotype (ten with two haplotypes and two with three haplotypes), whereas all the remaining 139 populations were fixed for a single plastid haplotype (Supplementary Data Table S1, available online). Among these 12 populations, four populations contained private haplotypes (E13, E15, E4 or E6) that co-occurred with widespread haplotypes.
Epipactis palustris was characterized by the single plastid haplotype E9, E. atrorubens by haplotypes E5 and E8, E. microphylla by haplotypes E6, E7 and E8, E. leptochila by haplotype E3 and E. muelleri by haplotypes E10, E11, E12 and E15 (with E10 being the most common).
The E. helleborine group, including both E. helleborine s.s. and E. helleborine s.l., from the Italian peninsula and geographically proximate regions was characterized by plastid haplotypes E1, E2, E4, E13 and E16, whereas the distant accessions of E. helleborine from Spain, Germany and England were characterized by private E17, E18 and E19 plastid haplotypes, respectively. Within the Italian E. helleborine group, 89 populations out of 97 were characterized by the widespread plastid haplotype E1, with only two localized populations of E. aspromontana (an endemic member of E. helleborine s.l.) showing the unique plastid haplotype E2 (Table 3, Supplementary Data Table S1 and Fig. S1, available online).
The average gene diversity within populations (hs) was 0·045 ± 0·012 and the total gene diversity (ht) was 0·649 ± 0·041. The Gst (0·930) and Nst (0·966) values indicated significant geographic structuring of plastid DNA markers. The finding that Nst was greater than Gst (U = 5·42, P <0·05) indicated that plastid DNA variation contained phylogeographic information.
The parsimony plastid haplotype network (Fig. 2) can be split into two main haplotype groups (separated by a single change), one including E. atrorubens/microphylla/muelleri/leptochila and the other including all accessions of E. helleborine s.s. and E. helleborine s.l. In the first group the maximum number of steps for a 95 % parsimonious connection between plastid haplotypes (i.e. with a 95 % confidence level that no multiple substitution has occurred at some site) is only two mutation steps, whereas in the second group, all E. helleborine s.s. and E. helleborine s.l. accessions were connected by a single step, with the exception of distant accessions of E. helleborine (from Spain, Germany and England) separated by 3–9 mutation steps from the others. A comparable distance, 15 steps, separated E. palustris from E. muelleri haplotype E15. All the rare and localized haplotypes in both groups were always connected to the most common ones by a single mutation step.
Results of our phylogeographic and phylogenetic analyses of plastid DNA and nrDNA uncovered a complex evolutionary history of relationship within the Epipactis species complex in the Italian peninsula. All previous studies carried out on Epipactis spp. Also revealed a low level of plastid DNA diversity in this genus (Squirrel et al., 2002; Bateman et al., 2005; Roy et al., 2009). In the present study, for detection of the phylogeographic signal we employed a promising plastid region, located in the rbcL–accD intergenic spacer, that revealed high genetic variability and thus represents an unusual hotspot of genetic variation in the Epipactis plastid genome. Although this region has been helpful in revealing phylogeographic patterns, it has not been fruitful for species phylogeny. Indeed, in this region, by excluding microsatellite variation and a few terminal autoapomorphic changes, we found only a single base difference (at position 259 in the plastid haplotype alignment in Table 2) that represents a synapomorphic state between the two main groups of species (E. helleborine and E. atrorubens/microphylla/muelleri/leptochila). Low plastid DNA variation has also been found in several tree species, e.g. in Lauraceae (Chanderbali et al., 2001) and other tree families (Magri et al., 2007; Vendramin et al., 2008), and has been ascribed to potential low mutation rates and long generation times, the demographic stability of most taxa or relatively high gene flow (Petit and Hampe, 2006; Smith and Donoghue, 2008). However, in the case of these herbaceous orchids, taking into account also their widespread nature and frequently large population size, this low plastid DNA variation, at least in terms of base substitutions, is probably due to recent divergence among lineages rather than to low mutation rates or recurrent bottleneck events that strongly reduced plastid haplotype variability.
The phylogeographic pattern indicated a single main plastid haplotype for each species characterized by a wide geographical distribution in the Italian peninsula: E. helleborine (E1), E. muelleri (E10), E. microphylla (E7), E. leptochila (E3) and E. atrorubens (E5), and a few unique or rare haplotypes with localized distributions. The five widespread plastid haplotypes could be divided in the network analysis into two main groups, one including the four plastid haplotypes (E5, E7, E10 and E3) representing the E. muelleri/microphylla/atrorubens/leptochila alliance, and the other only including the E. helleborine (s.s. and s.l.) widespread plastid haplotype E1 (Fig. 2 and Supplementary Data Fig. S1).
The four widespread plastid haplotypes (E5, E7, E10 and E3) characteristic of the E. muelleri/microphylla/atrorubens/leptochila alliance are all closely related (one or two mutational steps) and related to other less widespread (E8) or geographically localized plastid haplotypes (E11, E12, E15 and E6).
Epipactis microphylla and E. atrorubens shared a common plastid haplotype (E8) that cannot be easily explained as a consequence of gene flow between the two species possessing distinct nuclear ribotypes. A potential alternative to explain this pattern is that plastid haplotype E8 represents the ancestral plastid haplotype of these two recently evolved species. This hypothesis is supported in the network analysis by the intermediate position of E8 between plastid haplotypes E5 (typical of E. atrorubens) and E7 (typical of E. microphylla; Fig. 2) and by the fact that individuals of E. microphylla and E. atrorubens carrying the E8 plastid haplotype always have the ancestral R1 ribotype (Fig. 1). Alternatively for explaining this non-assortative mating between nuclear and cytoplasmic genomes we could assume the presence of outcrossing and selfing lineages in populations of these two facultative allogamous species.
The close connections between Sardinian and central Italian plastid haplotypes can also be ascribed to Quaternary glacial cycles, which, due to the supposed connections at that time between the Corso-Sardinian block and present-day Tuscany, allowed enhanced biotic exchange (Mariotti, 1990) through the Tuscan archipelago. Present plastid haplotype diversity on Sardinia is represented by a combination of both unique haplotypes (E11 and E15) and haplotypes shared with the mainland (E1, E8 and E10). These latter cases concern the widespread plastid haplotypes E8–E10 (representative of E. muelleri/microphylla atrorubens in the Italian pensinsula) and E1 (characteristic of E. helleborine in the Italian pensinsula). Additionally, the two locally restricted (E11 and E15) Sardinian haplotypes are closely related to the mainland widespread haplotype E7 and to the rare haplotype E12 (by a single mutation step) only occurring in Tuscany and in the Tuscan archipelago (on Elba) that presumably acted as a land bridge with the main island.
The E. helleborine s.s. lineage is characterized by a unique widespread plastid haplotype (E1) distributed throughout the Italian peninsula, from near the Alps to Sicily. Closely linked to this main plastid haplotype (only one mutational step) there are few plastid haplotypes (E13, E2, E16 and E4) with a geographically localized distribution (E4 occurs only in the Sila mountains and E13 only on the Gargano promontory). Surprisingly, the widespread Italian E. helleborine plastid haplotype (E1) showed several differences compared with E. helleborine s.s. plastid haplotypes detected in both northern and southern Germany (E18, three mutation steps), England (E19, eight mutation steps) and Spain (E17, nine mutation steps; Fig. 2). The differences (ten mutational steps) among E. helleborine plastid haplotypes from Italy, England and Spain, confirming the presence of a significant phylogeographic signal in this plastid region, also suggest that they all represent lineages from different colonization routes that did not meet after the last post-glacial expansion. The lower number of differences between Italian and German E. helleborine plastid haplotypes (Fig. 2) probably indicates that the two lineages colonized central Europe and Italy, starting from a common ancestor.
All this phylogegraphic evidence indicates that only two main lineages, one including E. helleborine s.s. and E. helleborine s.l. (mostly characterized by the main plastid haplotype E1) and one including the E. muelleri/microphylla/atrorubens/leptochila alliance, are currently present in the Italian peninsula.
We presently do not know how much phylogenetic signal is contained in the examined hypervariable plastid region, but it is most likely that these two main lineages may have resulted from a recent diversification, probably datable during the Quaternary (within the last million years) rather than during any previous geological/geographical rearrangement and consequent climatic oscillations which occurred in the area during the Tertiary (Thompson, 2005). This is confirmed by the small number of mutation steps separating the two haplotype groups, indicating a close relationship and recent divergence, a finding also supported by the absence of any significant difference in the ITS1 region of members of both groups. Indeed, ITS1 sequences revealed little genetic divergence among lineages and the sharing of a main, presumably ancestral, ribotype R1. In fact, this common widespread ribotype characterized several taxa possessing different plastid haplotypes (E2, E3, E4, E6, E8, E10, E11, E12, E13, E15, E16 and E19) belonging to the E. muelleri/microphylla/atrorubens/leptochila alliance and E. helleborine lineages. Also, the distant accessions of E. helleborine, in spite of large differences in their plastid haplotype sequences (E17, E18 and E19), have the same (R1, in the English accession) or closely related unique ribotype sequences, namely R5 in Spain (one mutation step from the widespread ribotype R1) and R6 in Germany (one mutation step from the Spanish ribotype R5).
All the other rare ribotypes detected in the Italian peninsula were unequivocally associated with some specific and widespread plastid haplotypes. In particular, E. atrorubens with plastid haplotype E5 always has a unique ribotype sequence R2 (one mutation step from the widespread ribotype R1); and E. microphylla with plastid haplotype E7 always has the unique ribotype sequence R3 (one mutation step from the widespread ribotype R1). Finally, the E. nordeniorum accessions with the widespread plastid haplotype E1 have a unique ribotype sequence R4 (one mutation step from ribotype R3; Fig. 1). The sharing of the common ribotype R1 across presumed species boundaries and the occurrence of a few sporadic different ribotypes, always different only by single mutations, can be interpreted as persistent sharing of ancestral polymorphism due to the young age of lineages or as a consequence of recurrent gene flow between them (Muir and Schlötterer, 2005). Sharing of nuclear, presumably ancestral, polymorphisms due to recent lineage diversification is also supported by karyological data with both 2n = 38 and 2n = 40 among species and even populations of both lineages, and by the high homogeneity in their heterochromatin content and pattern which indicate the presence of a single genetically cohesive group (D'Emerico et al., 1999).
The plastid haplotype network showed that all rare plastid haplotypes detected in the Italian peninsula and the islands were only one/two mutational steps away from the dominant and widespread plastid haplotypes, with no signs of genetic discontinuity.
Additionally, all detected plastid haplotypes in the Italian peninsula were frequently found fixed in several populations in groups of both lineages and independently of population size (Table 3, Supplementary Data Table S1 and Fig. S1). In particular the widespread plastid haplotype E1 was basically fixed in almost all populations, suggesting a rapid and recent colonization of the peninsula from a colonizing population carrying this plastid haplotype, a pattern mirrored by E. palustris plastid haplotype E9 (although populations of this species are often clonal). Several studies on mycorhizzal interactions have highlighted that, in contrast to E. palustris, E. helleborine requires a mycorrhizal interaction with the roots of neighbouring forest trees (Bidartondo et al., 2004, Selosse et al., 2004; Bidartondo and Read, 2008; Ogura-Tsujita and Yukawa, 2008). The presence of a unique widespread plastid haplotype E1 throughout the Italian peninsula, therefore, sharply contrasts with the local plastid haplotype richness of associated forest tree vegetation previously investigated in other phylogeographic studies. Indeed, forest trees such as Quercus and Fagus (Fineschi et al., 2002; Vettori et al., 2004), frequently co-occurring with E. helleborine, have been found to possess several plastid haplotypes in the southern Italian peninsula, a pattern typically explained as a consequence of long stasis in a glacial refugium (Petit et al., 2003).
In contrast to tree phylogeography, the presence of a unique widespread plastid haplotype in E. helleborine suggests that this lineage was unlikely to have experienced any long refugial stasis in the area, but rather that it was able to colonize the peninsula, probably from the Alps, in a recent and comparatively short period, probably during or immediately after the last glaciation, to the extent that no phylogeographic signals are locally detectable. Indeed E. helleborine is one of the most widespread orchid species in the world, and it has radiated extensively throughout Eurasia and North Africa (Delforge, 1995). It seems to have the potential for rapid spread: E. helleborine was able to colonize large parts of North America in an extremely brief period after its introduction by humans at the end of the 19th century (Squirrell et al., 2001). Thus, it cannot be ruled out that it may have colonized the Italian peninsula only recently during or after climatic oscillations of the last glacial period, and that this colonization has been rapid due to the dust-like wind-dispersed seeds. Supporting the hypothesis of a recent colonization of the Italian peninsula is the striking number of differences detected among plastid haplotypes from distant accessions of E. helleborine across Europe, in sharp contrast to the virtual absence of variation found within and between the lineages in the Italian peninsula (see below).
The possible cause of the morphological diversification between E. helleborine and the E. muelleri/microphylla/atrorubens/leptochila alliance can probably be ascribed to different centres of origins or to a shift in pollination biology recently disclosed between these lineages. In fact, flowers of E. helleborine and E. purpurata (both characterized by plastid haplotype E1) emit green-leaf volatiles (GLVs), which are specifically attractive to foragers of the social wasps Vespula germanica and V. vulgaris, whereas E. atrorubens (belonging to the other plastid haplotype group) is visited by a broad spectrum of pollinators, mainly bumble-bees (Brodmann et al., 2008).
In spite of a potentially recent colonization time, the genus Epipactis, and the E. helleborine lineage in particular, has been subject to an extensive radiation in the Italian peninsula. This presumed, probably exaggerated, species richness contrasts sharply with the low or zero level of nuclear and organellar genetic diversity that reflects more geographical distribution rather than presumed taxonomic boundaries. Given the convergent genetic results, it can therefore be assumed that the wide species/taxon differentiation and taxonomic complexity, which has been described on the basis of morphological characters, is not mirrored in their genetic discontinuity.
In particular, it was found that the widespread Italian E. helleborine plastid haplotype (E1) is more closely related to the E. muelleri/microphylla/atrorubens/leptochila (E3, E7, E8 and E10) plastid haplotypes than to the conspecific Spanish, English and German E. helleborine plastid haplotypes (E17, E18 and E19). This finding is quite surprising and, if we exclude the possibility that this plastid haplotype pattern is a consequence of homoplasy or recurrent gene flow between the two main lineages in the Italian refugium (with E. helleborine plastid capture and subsequent selective loss of ancestral plastid haplotypes of the E. muelleri/microphylla/atrorubens/leptochila lineage), it can also point towards E. helleborine being paraphyletic, as already postulated by Bateman et al. (2005) and Bateman (2006). Future research avenues should thus investigate relationships among members of the continental E. muelleri/microphylla/atrorubens/leptochila lineage in order to understand whether they are related more to the continental E. helleborine or to their Mediterranean conspecifics.
Most of the endemic taxa of the E. helleborine s.l. lineage investigated in the present study possess the same ribotype and widespread plastid haplotype (E1) characterizing the E. helleborine s.s. lineage. In contrast, the floral morphology of these taxa varies considerably, e.g. the presence/absence of a functional and well-developed rostellum and viscidium, strictly involved in the shift from an allogamous to an autogamous reproductive system, and several species or presumed endemic species are characterized by selfing. Thus, the high level of genetic homogenization indicates that several taxa (species) may have diverged by recent speciation (or local adaptation) as a consequence of a shift in mating systems within E. helleborine. This shift can lead to an instantaneous reproductive isolation from its progenitor and can generate morphological difference between the selfers and their allogamous progenitors, as a consequence of an increased level of homozygosity (Ehlers and Pedersen, 2000; Squirrell et al., 2002; Talalaj and Brzosko, 2008).
It is more likely that a frequent and rapid change in breeding system (from allogamy to autogamy) could have represented the mechanism promoting this rapid diversification and the observed high taxonomic complexity in Epipactis, as discussed in Hollingsworth et al. (2006). The absence of detectable plastid sequence divergence within this species cluster suggests that the shift towards selfing may have occurred locally and independently multiple times starting from allogamous E. helleborine (Richard, 1982) and may have a simple genetic basis (Bateman et al., 2005), providing a sort of pre-adaptation for the shift to self-pollination in particular ecological conditions. A similar evolutionary pattern has been found in recent ‘post-glacial’ taxa of the E. helleborine species complex in the UK, where several selfing taxa have arisen from outcrossing taxa on numerous occasions in Epipactis (Hollingsworth et al., 2006).
One important aspect of reproductive assurance is the ability of selfing lineages to colonize new habitats from small founding populations. Consequently, natural selection for reproductive assurance can lead to major morphological evolution and speciation on relatively short evolutionary time scales (Foxe et al., 2009). However, in addition to the ecological effect of breeding system, the ecological factors must evidently be among the major influences on mating system evolution (Charlesworth, 2006). In this context, the ecological requirement for specialized mycorrhizal interactions for seed development may suggest that the species may frequently face pollinator limitations when colonizing suitable (from a mycorrhizal perspective) habitats, and this may have enhanced high flexibility in its reproductive system (selfing vs. allogamy) for ensuring reproductive success in conditions of severe pollination limitation such as in woody habitats.
Supplementary Data are available online at www.aob.oxfordjournals.org and consist of the following. Table S1. Further detailed information regarding the samples used in the study. Figure S1. Map of the geographical distribution of the ct haplotypes listed in Table 1.
We would like to thank T. Stuessy and R. Bateman for having significantly improved the manuscript. We thank M. Ayasse, S. Bernardos, R. Bateman and members of the GIROS group for their help with sample collections. We also thank S. Giorgio and G. Napolitano, members of the Naples Botanical Garden staff, for help in collecting Italian samples. The present study was supported by the PRIN programme of the Italian Ministry of the University and Scientific Research to S.C.