HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Supplementary Data Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via ISI Web of Science (1) Citing Articles via Google Scholar Google Scholar Articles by Hedrén, M. Articles by Hansson, S. Search for Related Content PubMed Articles by Hedrén, M. Articles by Hansson, S. Agricola Articles by Hedrén, M. Articles by Hansson, S.

Patterns of polyploid evolution in Greek marsh orchids (Dactylorhiza; Orchidaceae) as revealed by allozymes, AFLPs, and plastid DNA data¹

Plant Ecology and Systematics, Department of Ecology, Lund University, Sölvegatan 37, SE-223 62 Lund, Sweden; Department of Horticultural Plant Breeding, Swedish University of Agricultural Sciences, Fjälkestadsvägen 123–1, SE-291 94 Kristianstad, Sweden; Botanical Garden & Museum, Natural History Museum of Denmark, University of Copenhagen, Gothersgade 130, DK-1123 Copenhagen K, Denmark; Almvägen 50, SE-187 34 Täby, Sweden

Received for publication March 20, 2006. Accepted for publication May 9, 2007.

ABSTRACT

Polyploidy is common in higher plants, and speciation in polyploid complexes is usually the result of reticulate evolution. We examined variation in nuclear AFLP fingerprints, nuclear isozymes, and hypervariable plastid DNA loci to describe speciation patterns and species relationships in the Dactylorhiza incarnata/maculata polyploid complex (marsh orchids; Orchidaceae) in Greece. Several endemic taxa with restricted distribution have been described from this area, and to propose meaningful conservation priorities, detailed relationships need to be known. We identified four independently derived allopolyploid lineages, which is a pattern poorly correlated with prevailing taxonomy. Three lineages were composed of populations restricted to small areas and may be of recent origins from extant parental lineages. One lineage with wide distribution in northern Greece was characterized by several unique plastid haplotypes that were phylogenetically related and evidently older. The D. incarnata/maculata polyploid complex in Greece has high levels of genetic diversity at the polyploid level. This diversity has accumulated over a long time and may include genetic variants originating from now extinct parental populations. Our data also indicate that the Balkans may have constituted an important refuge from which northern European Dactylorhiza were recruited after the Weichselian ice age.

Key Words: AFLP • allozymes • Dactylorhiza • Greece • Orchidaceae • plastid DNA • polyploid evolution • systematics

Polyploidy, the presence of more than two chromosome sets in single individuals, is widespread among the angiosperms and contributes significantly to increased diversity in physiological traits, reproductive ecology, and habitat preference within many plant genera (Stebbins, 1971 ; Thompson and Lumaret, 1992 ). Furthermore, in allopolyploids, in which the genomes of somewhat divergent parental lineages are combined in the same species, polyploidization results in a hybrid genome and in effect reticulate evolution. After formation of the raw allopolyploid, the genome may undergo extensive restructuring with selection for genes from either parent at different loci (Leitch and Bennett, 1997 ; Wendel, 2000 ). Accordingly, the allopolyploid may be seen as an important arena at which adaptive combinations of genes of different origins may be tested and selected for. Allopolyploid speciation is a rapid process. It may be seen as a form of hybrid speciation, but it is clearly more important than hybrid speciation at the diploid (homoploid) level because chromosome doubling circumvents many of the problems associated with hybrid fertility and reproduction (Grant, 1981 ).

Dactylorhiza is a moderately species-rich genus of terrestrial orchids with diversity centers in Europe and western Asia. Most taxa are members of a polyploid complex within which allotetraploid taxa have evolved on repeated occasions from hybridization between two broadly defined parental lineages, D. incarnata s.l., the diploid marsh orchids, and D. maculata s.l., the spotted orchids (Hedrén et al., 2001 ; Devos et al., 2003 ; Hedrén, 1996c , 2003 ). Dactylorhiza incarnata s.l. is variable and includes distinct color morphs, which may be sympatric, and also geographical races, which are often treated as subspecies or separate species. However, allozyme analyses (Hedrén, 1996c ; Pedersen, 1998a ), AFLP analysis (Hedrén et al., 2001 ), or plastid DNA variation (Hedrén, 2003 ; Pillon et al., in press ) reveal that D. incarnata has astonishingly little genetic variation and that morphologically defined taxa are poorly correlated to genetic differentiation patterns. In contrast, D. maculata s.l. is variable both in morphology and at molecular marker loci, and it includes diploid as well as autotetraploid cytotypes. The tetraploid D. maculata subsp. maculata and the mostly diploid D. maculata subsp. fuchsii are widespread in Europe but are replaced by the diploid D. maculata subsp. saccifera in southeastern Europe. The former two are sometimes subdivided into additional taxa (cf. Sóo, 1980 ; Buttler, 1991 ; Delforge, 2001 ).

Chromosome counts combined with allozyme analyses (Hedrén, 2002b ) indicate that about 50 species of Dactylorhiza are allotetraploids originating from hybridization between D. incarnata and D. maculata. Given that only two parental lineages are involved, the number of recognized allotetraploid species is much higher than in other polyploid complexes studied by molecular tools, such as Tragopogon (Cook et al., 1998 ), Draba (Brochmann et al., 1992 ), or Heuchera (Segraves et al., 1999 ), in which multiple independent formation of polyploids has been demonstrated. The high number of species may be the combined result of repeated independent polyploidization events over time and space (which may each involve somewhat different variants of parental taxa), divergent speciation at the tetraploid level, and secondary hybridization after formation of the allotetraploid. The number of species, however, may be inflated because of narrow species concepts adopted by many European orchidologists (cf. Pedersen, 1998b ; Bateman, 2001 ).

The allotetraploids often have small distribution areas (cf. Baumann and Künkele, 1988 ; Delforge, 2001 ), and many of them are restricted to regions in northern or western Europe that were completely ice covered during the Weichselian glaciation. Accordingly, it could be postulated that many of these allotetraploids have evolved after the ice age and in the areas where they grow today, a hypothesis that is generally in agreement with molecular data, including allozymes (Hedrén, 1996a – c ; Pedersen, 2004 ) and AFLPs (Hedrén et al., 2001 ). However, studies of plastid DNA (Devos et al., 2003 ; Pillon et al., in press ; M. Hedrén, S. Nordström and D. Ståhlberg, unpublished) have disclosed some variants in the allotetraploids that have not been encountered in the extant parental lineages, indicating that the allotetraploid complex may also include taxa of higher age.

A study of the D. incarnata/maculata polyploid complex in Greece may be of particular interest in this context, because the Balkans were considerably less affected by the Pleistocene ice ages than northern Europe (Hewitt, 1996 ) and may have served as an important refuge for the many plant and animal groups that were extirpated from the northern ice-covered areas during glacial maxima (Taberlet et al., 1998 ). Accordingly, the Balkans may house some old allotetraploid lineages that have survived one or several glacial cycles and that are not distributed elsewhere in Europe. If so, Balkanic allotetraploids may include unique genetic material absent from present day representatives of the parental lineages. But it is also possible that allotetraploid Dactylorhiza on the Balkans have recent origins and are not older than their northern European counterparts. If this is true, we would expect that the genomic contents of the Balkanic allotetraploids correspond closely to those of the present day representatives of the parental lineages growing in the same area.

The D. incarnata/maculata polyploid complex constitutes a very diverse group on the Balkans (cf. Baumann and Künkele, 1988 ; Delforge, 2001 ). In the present study we investigated the complex in Greece, where taxonomic diversity appears to be most pronounced. About six narrowly distributed allotetraploids occur in the northern part of the country, and another allotetraploid, D. pythagorae, is confined to the island of Samos in the southeast Aegean Sea. The delimitation of many of these allotetraploids is uncertain, and orchid floras differ in the number of species recognized. In addition to the allotetraploids with narrow distributions, we also included several populations of D. majalis subsp. cordigera, which is widespread in southeastern Europe and which is sometimes regarded as closely related to the Central European D. majalis s.s.

To understand the significance of polyploidy in relation to Pleistocene climate cycles and to work out an appropriate taxonomy, the parentage and pathways leading to polyploid Dactylorhiza need to be described. Because of the potential mosaic of processes leading to the formation of existing polyploids, three different molecular data sets were compiled in this study, two describing variation in the nuclear genome and one describing variation in the plastid (chloroplast) genome. The main objectives of the study were (1) to verify the allotetraploid nature of the Greek taxa, (2) to describe the parentage of the Greek allotetraploids, (3) to describe relationships and variation patterns among Greek Dactylorhiza, and (4) to compare the Greek taxa with taxa in other parts of Europe.

MATERIALS AND METHODS

Classification of plant material
Based on various floras (Baumann and Künkele, 1988 ; Buttler, 1991 ; Delforge, 2001 ) or published reports for individual localities (Hölzinger and Künkele, 1988 ; Willing and Willing, 1988 ; Thiele and Thiele, 2001 ), the material examined in the present study was placed in the following taxa: D. majalis subsp. cordigera (Fr.) Sundermann, D. baumanniana Hölzinger & Künkele, D. smolikana B. & E. Willing, D. graeca H. Baumann, D. kalopissii E. Nelson, D. macedonica Hölzinger & Künkele, D. pindica B. & E. Willing, D. pythagorae P. Gölz & H.R. Reinhard, and D. maculata subsp. saccifera (Brongn.) Dikli. We initially also screened material of the diploid D. iberica (Willd.) Soó for all molecular characters studied here. However, this species is not regarded as closely related to the remaining taxa investigated, and because it was evident from the initial analyses that it was different in all data sets, we excluded it from subsequent data analysis and further consideration. We were unable to obtain any material of D. incarnata from Greece, because it is rare in the country, but when needed we have used samples from other areas (mainly northern Europe and Turkey) and data from previous studies as reference material. For AFLP analysis, we also included reference material of D. majalis subsp. majalis and D. maculata subsp. fuchsii from areas outside Greece. Chromosome counts indicate that D. maculata subsp. saccifera and D. iberica are diploids with 2n = 40 and that the remaining taxa are tetraploids with 2n = 80. Allozyme studies agree with these reports (Hedrén, 2002a , b). Much of the morphological variation studied here is accordingly expressed at the tetraploid level. Studied sites and taxa are reported in Table 1, and the position of sampling sites is indicated in Fig 1. Vouchers (preserved flowers) were deposited in the Lund University Botanical Museum (LD), except for vouchers from localities 11 and 19, which were deposited in the Copenhagen University Botanical Museum (C).

View larger version (30K): [in this window] [in a new window]   Fig. 1. Map of Greece showing sampling sites and haplotype composition in populations analyzed for plastid data. Pie diagrams illustrating haplotype composition in the diploid Dactylorhiza maculata subsp. saccifera are along the left margin, whereas those for allotetraploid populations are spread out. Taxa from mixed populations are connected by a common baseline. The size of each pie diagram is roughly proportional to sample size, and the number of samples belonging to different haplotypes is given. Population codes are given in Table 1

Plastid DNA analysis
The plastid genome was initially screened for variations in restriction sites and fragment sizes by PCR-RFLP analysis, using the same primer pairs and conditions reported in Hedrén (2003) . However, because relatively little variation was detected in the Greek material, we also included the primer pair c/d of Taberlet et al. (1991) , which amplifies a portion of the trnT-trnL intergenic spacer, the trnL exon I, and the trnL intron. Following digestion of the PCR product with various restriction endonucleases, we separated the fragments on agarose gels. Some size variation was observed, but fragment sizes were still very similar, and variants could not be separated with sufficient certainty. Accordingly, we also sequenced the region, and ultimately, we designed specific primers for the amplification of short fragments, including the two most size-variable loci identified. The entire material was subsequently screened for size variation, and the fragments were separated on acrylamide gels in an automated sequencer for accurate size determination.

All primer pairs and PCR conditions used in the study are summarized in Table 3. Primers trnL5/Cy5trnLR5 amplified three size-variable loci adjacent to and partially overlapping with each other. The first locus consists of a 6-bp fragment that may appear as a single or two tandemly repeated copies. The second locus is a polyA mononucleotide microsatellite locus located just upstream of the 5' end of trnL exon I. This locus has previously been found to be informative in the related orchid Gymnadenia conopsea (Soliva and Widmer, 1999 ) and also in other European terrestrial orchids (Fay and Cowan, 2001 ). The third locus is a 6-bp duplication located at the border of the trnL exon I, including also a 2-bp fragment of the polyA locus. The primer pair Cy5trnLF7/trnLR7 amplified a ca 150- to 164-bp portion of the trnL intron containing a 7-bp repeat region, which was found to be duplicated two, three, or four times. In addition to the two primer pairs amplifying portions of the trnL region, we also designed primers for the amplification of the central, most size-variable fragment of the psbC-trnS pseudospacer (cf. Hedrén, 2003 ) and determined the exact fragment size on acrylamide gels.

Amplification and size determination of fragments
Size-variable fragments were amplified by an initial round of denaturing at 94°C for 2 min; followed by 40 cycles of denaturing at 94°C for 1 min, annealing at 54.4–57.0°C (depending on primer pairs; Table 3) for 1 min, and extension at 72°C for 1.5 min; and ended by a final extension at 72°C for 10 min. PCR reactions were performed in a reaction volume of ca. 5 µL containing 3.9 µL ddH₂O, 0.5 µL 10x reaction buffer (100 mM Tris-HCl pH 8.3, 500 mM KCl, 15 mM MgCl₂), 0.1 µL dNTPs (10 mM of each nucleotide), 0.0625 µL Cy5-labelled primer (1.5 pmol/µL), 0.025 µL unlabelled complementary primer (25 pmol/µL), 0.023 µL AmpliTaq Gold polymerase (5 units/µL; Applied Biosystems, Stockholm, Sweden), and 0.4 µL template DNA (14 ng/µL). One of the primers used in each primer pair was Cy5-labelled, which made it possible to visualize the amplified fragments on the sequencer. Prior to loading on the sequencer, the PCR product from each reaction was diluted in 20 µL formamide and mixed with appropriate size standards to enable exact size determination of the amplified fragments. To reduce costs, we used homemade PCR fragments as size standards, the size of which had already been determined by comparison to a commercially available 50- to 500-bp ladder (Amersham Biosciences).

AFLP analysis
Amplified fragment length polymorphism (AFLP) analysis (Vos et al., 1995 ) was carried out by means of an AFLP kit developed for normal-size plant genomes (Gibco BRL, Life Technologies, Täby, Sweden). The restriction, ligation, and pre-amplification steps were performed in accordance with the manufacturer's protocol, whereas in the selective amplification only half the amounts of reagents were used. The EcoRI selective primers were Cy5-labelled to enable detection of fragments on an ALFexpress II automated sequencer (Amersham Pharmacia Biotech, Uppsala, Sweden). All material was screened for five different primer combinations of selective primers; -ACT/-CTT, -ACT/-CAT, -ACC/-CAC, -ACC/-CAC, and -AGG/-CAA for the EcoRI and MseI primer extensions, respectively. The first three of these were also used in Hedrén et al. (2001) , whereas the latter two were added in the present study to obtain greater resolution among potentially closely related allotetraploid populations.

A data matrix with all AFLP fragments ranging from 50 to 500 base pairs was compiled with the help of the computer program ALFwin Fragment Analyzer (Amersham Pharmacia Biotech, 1998 ). Fragments with a signal less than 2% of full detection level were excluded, as were samples with very low amounts of reaction products. All fragments falling within the same base pair unit were considered as the same allele. Recognized bands were sorted to the nearest base pair and scored as present (1) or absent (0).

Data analysis
The allozyme data were summarized as a table of mean allele frequencies within populations (Table 2). Rogers' genetic distances (Rogers, 1972 ) were calculated between all pairs of populations, and the resulting distance matrix was used as starting point for a principal coordinates analysis (PCO; Gower, 1966 ). Calculations of genetic distances and PCO were computed with the program NTSYSpc version 2.2 (Rohlf, 2005 ).

AFLP data were summarized as a table of mean band frequencies within populations (Appendix S1; see Supplemental data accompanying the online version of this article). Assuming that band frequencies corresponded to allele frequencies, we then used this table for calculating Nei's (1972) genetic distances between pairs of populations. Nonmetric multidimensional scaling analysis (MDS; Kruskal, 1964a , b ) as well as UPGMA cluster analysis were applied to the resulting distance matrix. The MDS was performed with NTSYSpc version 2.2 (Rohlf, 2005 ), whereas the UPGMA cluster analysis was performed with PAUP* version 4 (Swofford, 2001 ).

Plastid data for the five analyzed loci were summarized as a series of haplotypes (Table 4). A distance matrix was then constructed by hand to summarize the number of mutational steps between every pair of haplotypes. The four loci (1, 1B, 1C, and 2) were regarded as microsatellite loci changing according to a stepwise mutational model (Ohta and Kimura, 1973 ), i.e., as ordered characters with character states ordered according to repeat number. Character changes at locus 6 were ordered by analysis of known sequences of the various fragments (summarized in Hedrén, 2003 ). Thus, the two shortest fragments of 177 and 191 bp, respectively, were regarded as being independently derived from the 217-bp fragment, whereas the remaining fragments at this locus were ordered in a series according to size. All haplotypes were then connected by hand in a minimum spanning network (MSN), which includes all alternative links of minimum length.

Finally, the distance matrices of population differentiation were compared with each other and with geographical distances between populations in a series of Mantel tests (Mantel, 1967 ). Table 5 contains correlation coefficients based on comparison of the data points in any two distance matrices along with significance levels based on the Mantel statistic z, which is the proportion of permutations of one of the distance matrices resulting in a higher correlation coefficient than that derived from the two original matrices. The z value was derived from 10 000 permutations in each test. Mantel tests were performed with Arlequin 2.000 (Schneider et al., 2000 ).

Ordination of the allozyme data by means of PCO (Fig. 2) separated the allotetraploid populations to the left from all populations of D. maculata subsp. saccifera to the right. The first and second principal coordinate axes accounted for 72% and 17% of the total variance, respectively. Among the allotetraploids, the population of D. pythagorae to the upper left was separated from the remaining populations to the lower left, apparently because of the dominance of alleles Pgi-c and Pgd-h, which were rare or absent from the other populations (Table 2). The population of D. macedonica was also somewhat separated from the rest of the allotetraploids. Some of the remaining allotetraploids were clustered in groups according to geographic origin. Thus, populations from sites 2 and 3 in central Greece, comprising D. majalis subsp. cordigera and D. kalopissii, were close to each other, and the populations from western Greece with D. pindica, D. baumanniana, and D. smolikana at sites 4 and 6 were also close. Populations from northern Greece mostly comprising D. cordigera, one population of D. graeca, and one that was regarded as intermediate between D. cordigera and D. baumanniana were more dispersed.

View larger version (7K): [in this window] [in a new window]   Fig. 2. Principal coordinates analysis (PCO) on Rogers' genetic distances between populations based on allozyme data. The first two axes account for 72% and 17% of the total variation, respectively. Population codes are composed of population number and three-letter taxon code (see Table 1)

View larger version (5K): [in this window] [in a new window]   Fig. 3. Nonmetric multidimensional scaling analysis (MDS) ordination plot of Nei's genetic distances between allotetraploid population samples based on AFLP data. Stress = 0.23. Population codes are given in Table 1

View larger version (6K): [in this window] [in a new window]   Fig. 4. UPGMA tree showing phenetic relationships between populations based on AFLP data. Bootstrap support values (based on 1000 replicates) are given above branches. Population codes are given in Table 1

View larger version (22K): [in this window] [in a new window]   Fig. 5. Minimum spanning network (MSN) connecting the 20 plastid haplotypes found in Greek Dactylorhiza. Plastid haplotypes were defined by the combined variation pattern at five plastid loci, cf. Table 4. Squares enclose haplotypes encountered in Dactylorhiza maculata subsp. saccifera; circles enclose those in allotetraploid taxa. The triangle encloses haplotype T, found in the allotetraploid D. macedonica and also in D. incarnata s.l. outside Greece. Triplet lines connect haplotypes in the same population

View larger version (10K): [in this window] [in a new window]   Fig. 6. Nonmetric multidimensional scaling analysis (MDS) ordination of Greek Dactylorhiza populations based on plastid haplotype data. The mean number of differences between haplotypes in every pair of population samples was calculated, and the MDS was based on the resulting distance matrix. Stress = 0.23. Population codes are given in Table 1

DISCUSSION

Genome characterization of Greek allotetraploids
Allozyme data compiled here support the view that the Greek populations initially collected as allotetraploids indeed belong to allotetraploid taxa. The tetraploid condition is supported by the fact that whenever three or more alleles were encountered at any locus within any of these populations, almost invariably some individual plants in the populations expressed three or four alleles at the same locus. In contrast, in D. maculata subsp. saccifera, no individual ever expressed more than two alleles at any locus, though more than two alleles were often encountered at single loci within populations of this taxon. Furthermore, the observation that allotetraploid populations were often variable at allozyme loci, with some individuals containing alternative alleles and others being heterozygous for these alleles, also suggests that these taxa are largely sexual and outbreeding. Moreover, the allele composition at allozyme loci supports the view that all the allotetraploids in Greece originated from hybridization between the D. maculata s.l. and D. incarnata s.l. lineages. All allotetraploid populations from Greece contained alleles recorded in D. maculata s.l. to frequencies around 50% at each locus (Table 2). This was also true for individual plants within populations (data not shown). The remaining alleles in the allotetraploids are mostly alleles that dominate in D. incarnata s.l. in northern Europe or alleles encountered in D. incarnata s.l. from Asia Minor (Table 2). At Pgi for instance, European material of D. incarnata s.l. is fixed for the allele Pgi-e. In some of the allotetraploid populations from Greece, the same allele is present at frequencies around 50%, but if the frequency of this allele is lower, other alleles also present in D. incarnata s.l. in Asia Minor are found and add up to frequencies around 50% (Table 2). In Asia Minor a third parental lineage in Dactylorhiza has contributed to polyploid formation, viz. D. euxina (Hedrén, 2002). This lineage differs from D. incarnata s.l. at several allozyme loci and has apparently not contributed to any of the Greek allotetraploids.

Relationships of haplotypes found in Greek taxa
The MSN (Fig. 5) shows several alternative pathways of haplotype evolution, but because we did not include any outgroup taxon in the analysis, we cannot deduce the direction of character evolution from the network itself. By comparison with other studies in Dactylorhiza based on plastid DNA data and by consideration of the taxonomic distribution of the various haplotypes found here, three major groups of haplotypes can be distinguished and possible evolutionary scenarios can be proposed.

The first group is composed of the single haplotype T, which was found in D. macedonica only. This haplotype differs from all other haplotypes studied here by having a 368-bp fragment at locus 6. Because all material of D. incarnata s.l. but no other taxa studied so far also have fragments of the same length at this locus (Hedrén, 2003 ; M. Hedrén, unpublished data), we infer that haplotype T must have been inherited from the D. incarnata s.l. lineage. The second and third groups of haplotypes are characterized by shorter fragments at locus 6, which indicates that they should be derived from haplotypes belonging to the D. maculata s.l. parental lineage (cf. Hedrén, 2003 ). The second major group of haplotypes comprises haplotypes A–E, which are directly linked to each other and which were all present in D. maculata subsp. saccifera. These haplotypes are characterized by a single repeat at locus 1B and short fragments (217 bp or shorter) at locus 6. Haplotypes C–E were also found in some allotetraploid populations. The third group of haplotypes consists of haplotypes F–S, which were restricted to allotetraploid material. They share a 6-bp duplication at locus 1B, which is a duplication that we have not encountered so far in any material outside Greece (data not shown) and which may be regarded as a synapomorphy for the group.

Based on their position in the haplotype network, two potentially monophyletic but mutually exclusive subgroups can be identified within the third major group. The first subgroup is composed of haplotypes MNOPQRS, which are characterized by a 320-bp fragment at locus 6, and the second by haplotypes GHLPQ, which are characterized by a unique duplication at locus 1C. The potential monophyly of the first subgroup is supported by geographical data because its constituent haplotypes are restricted to populations in northern Greece. The haplotypes in the second subgroup have more scattered distributions.

Systematic position of D. maculata subsp. saccifera
The Greek material of D. maculata subsp. saccifera differed little in allozyme composition from that of D. maculata s.l. reported from other areas (Table 2). Also, one of the common plastid haplotypes in the Greek material, D, was indistinguishable (based on the marker loci studied here) from haplotype C of Hedrén (2003) , which was a widespread haplotype in D. maculata s.l. as well as in many of the allotetraploids in northern Europe. However, some differences between D. maculata s.l. from Greece and from outside Greece were also recorded. First, two of the fairly common alleles in D. maculata s.l., Pgi-a, Mnr-b, were not encountered in the Greek material, and judging from allozyme variation at all loci, Greek material of D. maculata subsp. saccifera has a rather narrow genetic basis. On the other hand, the Greek material contains haplotypes A, B, and C in moderate to high frequencies, none of which have been found in material outside Greece. More detailed analyses of plastid DNA variation in Dactylorhiza provides further evidence that most of the Greek material of D. maculata subsp. saccifera differs from northern and central European material in haplotypes, although some overlap is still evident (M. Hedrén, S. Nordström and D. Ståhlberg, unpublished). Accordingly, based on molecular data, D. maculata subsp. saccifera could be included in a widely circumscribed D. maculata s.l., but it still may be appropriate to treat it as separate subspecies, at least until morphological variation patterns in the entire D. maculata s.l. have been studied in sufficient detail. Finally, allozyme variation patterns in Greek populations of D. maculata subsp. saccifera and previous findings based on material from Turkey (Hedrén, 2002) indicate that this is a diploid taxon.

Multiple origins of Greek allotetraploids
The three molecular data sets compiled in the present study subdivide allotetraploid material in different ways. Comparison of the three data sets gives a minimum of four independently derived groups of Greek allotetraploids, but additional polyploidization events may have occurred.

1. All studied allotetraploids had plastid haplotypes that must have been derived from the D. maculata s.l. ancestral lineage, except for D. macedonica, which had a haplotype corresponding to those found in the D. incarnata s.l. ancestral lineage. Because the plastids are inherited from the seed parent in orchids (Corriveau and Coleman, 1988 ), this pattern shows that D. macedonica must have originated from a separate polyploidization event in which D. incarnata s.l. served as seed parent and D. maculata s.l. as pollen parent, whereas the opposite pattern must be true for the remaining allotetraploids. Other studies of plastid genome variation in Dactylorhiza have shown that the D. majalis s.l. allotetraploids are generally characterized by the plastid genome of the D. maculata s.l. parental lineage. Some odd individuals within populations of allotetraploids sometimes contain D. incarnata s.l. haplotypes, which may have entered into the allotetraploids because of local backcrossing and introgression with the D. incarnata s.l. parental lineage (Hedrén, 2003 ; Pillon et al., 2006 ; M. Hedrén, S. Nordström and D. Ståhlberg, unpublished), but this is the first study to show that a whole population of allotetraploids is fixed for a D. incarnata s.l. haplotype.
   
2. Dactylorhiza pythagorae differs from the other allotetraploids in that it is fixed for haplotype D. It also differs from the other allotetraploids at two allozyme loci, which contain alleles that are fairly common in D. incarnata s.l. in Asia Minor but which are absent from European D. incarnata s.l. Furthermore, it is strongly differentiated from the other allotetraploids in AFLP markers. Considering also the isolated distribution area on Samos in the eastern Aegean Sea (Fig. 1), the combined results also point to a separate origin of this taxon. Haplotype D also occurs in present day D. maculata s.l., which indicates that the taxon may be of relatively recent origin. It is still possible, however, that D. pythagorae may be derived from other allotetraploid Dactylorhiza distributed in Asia Minor. The nearest taxon is D. nieschalkiorum, which has a scattered distribution in northeast Turkey. This taxon is like D. pythagorae in that it contains haplotype D, but it contains different alleles at allozyme locus Pgd (Hedrén, 2002).
   
3. The allotetraploids found at sites 2 and 3, constituting all investigated material of D. kalopissii and some material sorted as D. majalis subsp. cordigera, were characterized by haplotypes C and E, which were also found in D. maculata subsp. saccifera in the same area. The data indicate that the allotetraploids at sites 2 and 3 may have arisen recently and within the same general area where they occur today. That two different haplotypes are found in these allotetraploids suggests that they originated from two separate polyploidization events, but given that the same haplotypes are also found in local material of D. maculata subsp. saccifera and that some hybrids were observed at these sites (cf. Willing and Willing, 1988 ), it is equally as likely that this haplotype variation resulted from secondary introgression from subsp. saccifera into the regional allotetraploids.
   
4. Finally, we conclude that the allotetraploids containing haplotypes F–S must have one or several separate origins. These haplotypes were restricted to the allotetraploids, and accordingly, we cannot decide whether they evolved from each other by mutations among existing allotetraploids or entered into the group of allotetraploids by separate polyploidization events from a now extinct ancestral lineage. However, the observation that variants differing by a single character often occur together in the same population indicates that at least some of these variants have evolved by mutation at the allotetraploid level and that the number of polyploidization events may be lower than the number of haplotypes identified in this group.
   

Two lines of reasoning suggest that allotetraploids carrying haplotypes F–S must be relatively old, and given the high levels of accumulated haplotype diversity within and between populations, they may well predate the Weichselian glaciation. First, the putative synapomorphy for haplotypes F–S, the duplication at locus 1B, has not been found in analyzed samples of D. maculata s.l. from Greece or from other parts of the distribution area. The duplication has also not been found in any allotetraploid samples from outside Greece. Second, the character uniting haplotypes M–S, the 320-bp fragment at locus 6, appears to be rare in samples from other areas, but it has been identified in some scattered populations of D. maculata subsp. maculata from southernmost Sweden and continental Europe (D. Ståhlberg, unpublished data) and in the allotetraploid D. elata from southwest Europe (Hedrén, 2003 ), although haplotypes in these taxa differ from the Greek haplotypes at other loci. Based on a combination of ITS sequences and plastid microsatellite loci, Pillon et al. (in press ) concluded that D. maculata subsp. maculata must have served as the maternal parent to D. elata. Assuming that all haplotypes characterized by the 320-bp fragment at locus 6 are linked to each other, we hypothesize that the lineage that gave rise to the Greek allotetraploids was related to the present-day D. maculata subsp. maculata rather then to the diploid subsp. saccifera. The former is not known from Greece but is present in northern parts of the Balkans (Soó, 1980 ). Extended sampling from this region may give further clues to the origin of the Greek allotetraploids.

Correlation between data sets
The conclusion that the allotetraploids must have evolved through several separate polyploidization events—and probably during different periods—may also help to explain the observations from the Mantel tests that genetic distances between allotetraploids are overall poorly correlated with geographical distances. The only significant correlation was that between geographical distances and allozymes when all allotetraploid populations were included, and this correlation may simply reflect that the most geographically distant population (i.e., D. pythagorae) differed from the other allotetraploids at two allozyme loci. For the lowest level of comparisons, when only allotetraploid populations characterized by haplotypes F–S were included, the lack of significant correlations indicates that differentiation between populations is not simply due to isolation by distance but may also involve historical processes, including migration. In contrast, all comparisons between molecular distance data sets were either significant or nearly significant, which would be expected when independently derived groups of allotetraploids are included in the same analysis. Furthermore, the finding that correlations between molecular data sets were still significant or nearly significant at the lowest level of analyses convinces us that different molecular data sets are capable of tracking the same underlying differentiation patterns.

Taxonomic implications
We found that the allotetraploid populations from sites 2 and 3 probably have arisen independently from the remaining allotetraploids. All material of D. kalopissii was collected at these sites, which supports the circumscription of D. kalopissii as a separate entity. However, at site 3 plants of D. kalopissii (characterized by unspotted leaves, cf. Baumann and Künkele, 1988 ) were growing mixed with D. majalis subsp. cordigera (with spotted leaves). A large proportion of individuals at this site had mixed character combinations, and furthermore, each of the three molecular data sets indicated that all individuals at this site constituted a single panmictic population. We conclude that the circumscription of D. kalopissii must include also individuals with spotted leaves, as already stated by Nelson (1976) and Willing and Willing (1988) . Still, it remains to be seen whether other morphological characters could be found that would support separating this taxon from D. majalis subsp. cordigera.

The allotetraploids D. pindica and D. baumanniana were growing together at site 4, and molecular data revealed that all the allotetraploid material formed a single panmictic population. Furthermore, molecular data grouped these allotetraploids together with D. smolikana from the nearby site 6. Allotetraploids from these two sites have haplotypes characterized by the duplication at locus 1B, which also characterizes most populations of the widespread D. majalis subsp. cordigera. From a molecular standpoint, we could recognize D. pindica, D. baumanniana, and D. smolikana as a southwestern taxon related to D. majalis subsp. cordigera, but again, it is unclear whether such a treatment would be supported by morphological data.

Dactylorhiza macedonica and D. pythagorae were only sampled at single sites but appear to be distinct, at least in some of the molecular data sets. Dactylorhiza macedonica differed from all other allotetraploids in being fixed for a haplotype originating from the D. incarnata s.l. parental lineage. However, D. macedonica has been reported from several additional sites in northern Greece and FYR Macedonia (Hölzinger and Künkele, 1988 ), and material from these sites should also be investigated to confirm the pattern. Furthermore, D. macedonica is morphologically similar to D. kalopissii, and some authors treat them as the same taxon (e.g., Buttler, 1991 ). Dactylorhiza pythagorae from Samos appears to be distinct from allotetraploids growing in northern mainland Greece, but as stated, its relationships to allotetraploid taxa growing in Turkey needs to be studied further.

Based on morphological observations, Willing and Willing (1991) interpreted D. graeca as a hybrid swarm between D. majalis subsp. cordigera and D. macedonica, which occur sympatrically on Vrontous. (The latter was actually given as D. kalopissii by Willing and Willing, but according to Hölzinger and Künkele [1988 ], populations from northeastern Greece should be separated as D. macedonica, as we have done here.) In the PCO plot of allozyme data (Fig. 2), the D. graeca population is located close to the D. macedonica population together with populations of D. majalis subsp. cordigera from northern Greece; and in the MDS ordination plot of Nei's genetic distances based on AFLP data (Fig. 3), D. graeca is located intermediate between D. macedonica and the populations of D. majalis subsp. cordigera from the north. On the other hand, D. graeca was embedded in D. majalis subsp. cordigera in the UPGMA tree based on AFLP data (Fig. 4), suggesting that D. graeca is more genetically similar to this taxon than to D. macedonica. Furthermore, D. macedonica was fixed for the distinct haplotype T, which was not present in any of the other allotetraploids sampled from Greece.

From molecular data compiled here, we cannot confirm that D. graeca is a hybrid swarm, but we could also not dismiss this hypothesis. If it originated by hybridization, however, we propose a modified scenario in which D. graeca is basically a population of D. majalis subsp. cordigera modified by limited and unidirectional gene flow from D. macedonica pollen. Such a scenario would explain why the D. macedonica plastid haplotype is absent from D. graeca and why D. macedonica clustered with D. majalis subsp. cordigera in the UPGMA tree. We emphasize, however, that we included only a single population of D. macedonica in our analysis and that populations from Vrontous may carry other haplotypes.

We have shown in this study that the genetic differentiation pattern among Greek Dactylorhiza is largely incongruent with existing taxonomic treatments of this group. The results indicate that most if not all Greek allotetraploids may be treated as regional or local variants of the southeastern European D. majalis subsp. cordigera (cf. Pedersen et al., 2003 ). However, a revised taxonomy needs to be based on good morphometric data. Such a data set is now being collected (H. Æ. Pedersen, in preparation), but for now we refrain from proposing any new taxonomic combinations to avoid further confusing the taxonomy of the D. incarnata/maculata polyploid complex.

Conservation aspects
The allotetraploid populations in Greece are highly divergent from each other and contain unique genetic material that seems to be absent from extant parental lineages or allotetraploids growing elsewhere in Europe. However, the existing taxonomic treatment of the Greek allotetraploids does not reflect population differentiation. Rather, we found a pattern of high similarity among populations growing in the same areas, suggesting that populations from all major regions must be considered to preserve as much diversity as possible in Greece.

Conclusions
Our results suggest that the Greek allotetraploids include some old genetic variants no longer present in the parental lineages that gave rise to these allotetraploids. This finding indicates that the Balkan area may harbor allotetraploid populations with other genes or haplotypes that are no longer present in the parental lineages but that have dispersed further to the north during postglacial recolonization. Such a scenario may explain why some of the plastid haplotypes in Scandinavian allotetraploids today have never been found in any of the parental lineages, even though thousands of specimens has now been examined from other parts of Europe (Pillon et al., in press ; D. Ståhlberg, unpublished). This scenario would not exclude the possibility that northern Europe also has many, or even a majority, of allotetraploids that are of postglacial origin and from parental lineages present in the north today, but it would help to explain the relatively broad genetic diversity in the northern group of allotetraploid taxa and the success of this group as indicated by its wide habitat spectrum. Such a model of polyploid evolution—amalgamation at the polyploid level of gene diversity accumulated through successive time periods of parental lineage evolution—may serve as a general model to explain the success of polyploid complexes in temperate areas.

¹

⁵ Author for correspondence (mikael.hedren{at}ekol.lu.se )

LITERATURE CITED

Amersham Pharmacia Biotech.. 1998. ALFwin fragment analyser, version 1.00, user manual Amersham Pharmacia Biotech, Uppsala, Sweden.

Ashton G. C. Braden A. W. H.. 1964. Serum ß-globulin polymorphism in mice. Australian Journal of Biological Sciences 14: 248-254.

Bateman R. M.. 2001. Evolution and classification of European orchids: insights from molecular and morphological characters. Journal Europäischen Orchideen 33: 33-119.

Baumann H. Künkele S.. 1988. Die Orchideen Europas Kosmos Naturführer, Stuttgart, Germany.

Brochmann C. Soltis D. E. Soltis P. S.. 1992. Electrophoretic relationships and phylogeny of Nordic polyploids in Draba (Brassicaceae). Plant Systematics and Evolution 182: 35-70.[CrossRef][ISI]

Buttler K. P.. 1991. Dactylorhiza. In A. Strid and K. Tan [eds.], Mountain flora of Greece 873-880 Edinburgh University Press, Edinburgh, Scotland.

Chase M. W. Hills H. G.. 1991. Silica gel: an ideal material for field preservation of leaf samples for DNA studies. Taxon 40: 215-220.[CrossRef][ISI]

Cook L. M. Soltis P. S. Brunsfeld S. J. Soltis D. E.. 1998. Multiple independent formations of Tragopogon tetraploids (Asteraceae): evidence from RAPD markers. Molecular Ecology 7: 1293-1302.[CrossRef]

Corriveau J. L Coleman A. W.. 1988. Rapid screening method to detect potential biparental inheritance of plastid DNA and results for over 200 angiosperm species. American Journal of Botany 75: 1443-1458.[CrossRef][ISI]

Delforge P.. 2001. Guide des orchidées d'Europe, d'Afrique du Nord et du Proche-Orient Delachaux et Niestlé, Paris, France.

Demesure B. Sodzi N. Petit R. J.. 1995. A set of universal primers for amplification of polymorphic non-coding regions of mitochondrial and chloroplast DNA in plants. Molecular Ecology 4: 129-131.[Medline]

Devos N. Tyteca D. Raspé O. Wesselingh R. A. Jacquemart A.-L.. 2003. Patterns of chloroplast diversity among western European Dactylorhiza species (Orchidaceae). Plant Systematics and Evolution 243: 85-97.[CrossRef][ISI]

Doyle J. J Doyle J. H.. 1990. Isolation of plant DNA from fresh tissue. Focus 12: 13-15.[Medline]

Fay M. F. Cowan R. S.. 2001. Plastid microsatellites in Cypripedium calceolus (Orchidaceae): genetic fingerprints from herbarium specimens. Lindleyana 16: 151-156.

Gower J. C.. 1966. Some distance properties of latent root and vector methods used in multivariate analysis. Biometrika 53: 325-338.[Abstract/Free Full Text]

Grant V.. 1981. Plant speciation Columbia University Press, New York, New York, USA.

Hedrén M.. 1996a. The allotetraploid nature of Dactylorhiza praetermissa (Orchidaceae) confirmed. Watsonia 21: 113-118.

Hedrén M.. 1996b. Electrophoretic evidence for allotetraploid origin of Dactylorhiza purpurella (Orchidaceae). Nordic Journal of Botany 16: 127-134.[ISI]

Hedrén M.. 1996c. Genetic differentiation, polyploidization and hybridization in northern European Dactylorhiza (Orchidaceae): evidence from allozyme markers. Plant Systematics and Evolution 201: 31-55.[CrossRef][ISI]

Hedrén M.. 2002a. Speciation patterns in the Dactylorhiza incarnata/maculata polyploid complex (Orchidaceae): evidence from molecular markers. Journal Europäischer Orchideen 34: 707-731.

Hedrén M.. 2002b. Systematics of the Dactylorhiza euxina/incarnata/maculata polyploid complex (Orchidaceae) in Turkey: evidence from allozyme data. Plant Systematics and Evolution 229: 23-44.[CrossRef][ISI]

Hedrén M.. 2003. Plastid DNA variation in the Dactylorhiza incarnata/maculata polyploid complex and the origin of allotetraploid D. sphagnicola (Orchidaceae). Molecular Ecology 12: 2669-2680.[CrossRef][Medline]

Hedrén M. Fay M. F. Chase M. W.. 2001. Amplified fragment length polymorphisms (AFLP) reveal details of polyploid evolution in Dactylorhiza (Orchidaceae). American Journal of Botany 88: 1868-1880.[Abstract/Free Full Text]

Hewitt G.. 1996. Some genetic consequences of ice ages, and their role in divergence and speciation. Biological Journal of the Linnean Society 58: 247-276.[CrossRef][ISI]

Hölzinger J. Künkele S.. 1988. Dactylorhiza macedonica sp. nova. Mitteilungsblatt Arbeitskreis Heimische Orchideen, Baden-Württenberg 20: 185-195.

King J. N. Danzik B. P.. 1983. Inheritance and linkage of isozymes in white spruce (Picea glauca). Canadian Journal of Genetics and Cytology 25: 430-436.[ISI]

Kruskal J. B.. 1964a. Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis. Psychometrika 29: 1-27.[CrossRef][ISI]

Kruskal J. B.. 1964b. Nonmetric multidimensional scaling: a numerical method. Psychometrika 29: 28-42.[Medline]

Leitch I. J. Bennett M. D.. 1997. Polyploidy in angiosperms. Trends in Plant Science 2: 470-476.[CrossRef][ISI]

Lönn M. Prentice H. C.. 1990. Mosaic variation in Swedish Petrorhagia prolifera (Caryophyllaceae): the partitioning of morphometric and electrophoretic diversity. Biological Journal of the Linnean Society 41: 353-373.[CrossRef][ISI]

Mantel N. A.. 1967. The detection of disease clustering and a generalized regression approach. Cancer Research 27: 209-220.[ISI][Medline]

Nei M.. 1972. Genetic distance between populations. American Naturalist 106: 283-292.[CrossRef][ISI]

Nelson E.. 1976. Monographie und Ikonographie der Orchidaceen-Gattung, III. Dactylorhiza Verlag Speich AG, Zurich, Switzerland.

Ohta T. Kimura M.. 1973. The model of mutation appropriate to estimate the number of electrophoretically detectable alleles in a genetic population. Genetics Research 22: 201-204.

Pedersen H. Æ.. 1998a. Allozyme variation and genetic integrity of Dactylorhiza incarnata (Orchidaceae). Nordic Journal of Botany 18: 15-21.[ISI]

Pedersen H. Æ.. 1998b. Species concept and guidelines for infraspecific taxonomic ranking in Dactylorhiza (Orchidaceae). Nordic Journal of Botany 18: 289-311.[ISI]

Pedersen H. Æ.. 2004. Dactylorhiza majalis s.l. (Orchidaceae) in acid habitats: variation patterns, taxonomy, and evolution. Nordic Journal of Botany 22: 641-658.[ISI]

Pedersen H. Æ. Hedrén M. Bateman R. M.. 2003. (1600) Proposal to conserve the name Orchis majalis against O. elata, O. vestita, and O. sesquipedalis (Dactylorhiza: Orchidinae: Orchidaceae). Taxon 52: 633-634.[CrossRef]

Pillon Y. Fay M. F. Hedrén M. Bateman R. M. Devey D. Shipunov A. van der Bank M. Chase M. W. In press Insights into the evolution and biogeography of western European species complexes in Dactylorhiza (Orchidaceae). Taxon.

Rogers J. S.. 1972. Measures of genetic similarity and genetic distance. Studies in Genetics 7: 145-153.

Rohlf F. J.. 2005. NTSYS-pc, numerical taxonomy and multivariate analysis system, version 2.2 Exeter software, Setauket, New York, USA.

Schneider S. Roessli D. Excoffier L.. 2000. Arlequin version 2.000: a software for population genetics data analysis. Genetics and Biometry Laboratory, University of Geneva, Geneva, Switzerland http://cmpg.unibe.ch/software/arlequin3.

Segraves K. A. Thompson J. N. Soltis P. S. Soltis D. E.. 1999. Multiple origins of polyploidy and the geographic structure of Heuchera grossulariifolia. Molecular Ecology 8: 253-262.[CrossRef]

Soliva M. Widmer A.. 1999. Genetic and floral divergence among sympatric populations of Gymnadenia conopsea s.l. (Orchidaceae) with different flowering phenology. International Journal of Plant Sciences 160: 897-905.[CrossRef][ISI][Medline]

Soltis D. E. Haufler C. H. Darrow D. C. Gastony G. J.. 1983. Starch gel electrophoresis of ferns: a compilation of grinding buffers, gel and electrode buffers, and staining schedules. American Fern Journal 73: 9-27.[CrossRef][ISI]

Soó R. de.. 1980. Dactylorhiza. In T. G. Tutin, V. G. Heywood, N. A. Burgess, D. M. Moore, D. H. Valentine, S. M. Walters, and D. A. Webb [eds.], Flora europea, vol. 5 333-337 Cambridge University Press, Cambridge, UK.

Stebbins G. L.. 1971. Chromosomal evolution in higher plants Edward Arnold, London, UK.

Swofford D.. 2001. PAUP*: phylogenetic analysis using parsimony (*and other methods), version 4 Sinauer, Sunderland, Massachusetts, USA.

Taberlet P. Fumagalli L. Wust-Saucy A.-G. Cosson J.-F.. 1998. Comparative phylogeography and postglacial colonization routes in Europe. Molecular Ecology 7: 453-464.[CrossRef][Medline]

Taberlet P. Gielly L. Pautou G. Bouvet J.. 1991. Universal primers for amplification of three non-coding regions of chloroplast DNA. Plant Molecular Biology 17: 1105-1109.