Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://herba.msu.ru/shipunov/author/shipunov2005a.pdf Äàòà èçìåíåíèÿ: Sat Apr 9 23:02:26 2005 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 22:17:00 2012 Êîäèðîâêà:

Original Article EVOLUTION OF DACTYLORHIZA BALTICA A. P. SHIPUNOV ET AL.

Botanical Journal of the Linnean Society, 2005, 147, 257 ­ 274. With 7 figures

Evolution of Dactylorhiza baltica (Orchidaceae) in European Russia: evidence from molecular markers and morphology
ALEXEY B. SHIPUNOV*, MICHAEL F. FAY and MARK W. CHASE
Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3DS, UK
Received November 2003; accepted for publication September 2004

Four plastid markers, four nuclear markers and 14 morphometric characters were used in this study to investigate the evolution of Dactylorhiza baltica (Orchidaceae) in European Russia. In total, 98, 214 and 775 samples from 85, 112 and 121 populations were involved in the combined and separate molecular and morphometric analyses, respectively. In most cases, morphometric measures were done on exactly the same plants that were used for DNA studies. Dactylorhiza baltica plants from European Russia are most probably the products of several recent and mostly local hybridization events between the diploids D. fuchsii and D. incarnata, which have each been the maternal parent on different occasions. Considerable introgression into the parental diploids via the allopolyploid D. baltica is also hypothesized. Several morphological characters, such as length of the lip lateral lobe and the length of longest leaf, were found to be robust and could be useful in identification of D. baltica. This study demonstrates the advantage of `combined' techniques, especially in the case of taxonomically complex taxa. © 2005 The Linnean Society of London, Botanical Journal of the Linnean Society, 2005, 147, 257­274.

ADDITIONAL KEYWORDS: allotetraploids ­ microsatellites ­ morphometrics ­ systematics.

INTRODUCTION
Dactylorhiza Necker ex Nevski (Orchidaceae) is, along with Epipactis and Ophrys, one of the most taxonomically controversial orchid genera in Europe. There is great instability in the accepted number of species and infraspecific taxa. The borders between many species are unclear, and there are considerable difficulties in the determination of single plants ( Averyanov, 1990; Reinhard, 1990; Delforge, 1995; Stace, 1997; Bateman, 2001). Many of the most problematic taxa are allotetraploids, most of which are believed to be the result of multiple hybridization events between two broadly defined parental species, D. fuchsii (Druce) SoÑ and D. incarnata (L.) SoÑ (Heslop-Harrison, 1968; HedrÈn, 2002; Devos et al., 2003; Y. Pillon, unpubl. data). Multiple lines of evidence indicate that this complex is an `. . . unusually dynamic system of polyploid speciation

*Corresponding author. E-mail: a.shipunov@kew.org

and extinction in which polyploids evolve continuously from the same set of broadly defi ned parental lineages' (HedrÈn, 2003: 2678). Furthermore, the limits of the diploid parental taxa are sometimes made less c lear by the exchange of genetic material, hypothesized to be via allotetraploids (HedrÈn, Fay & Chase, 2001; HedrÈn, 2003), in spite of the differences in their ploidy. One good example of such polyploid species and at the same time a less well known member of this complex is D. baltica (Klinge) Orlova, for which the distribution, unlike other named allotetraploids of this complex (which occur principally in western Europe), is restricted to the eastern part of Germany, Poland, the Baltic countries, southern Finland and Russia. The eastern parts of its distribution are less defi nite; some authors (Nevski, 1935; Smoljaninova, 1976) have argued that it is restricted to the western parts of European Russia (Pskov and Leningrad regions) together with some localities in the northern Urals and southern Siberia, whereas others expand the European portion across all of European Russia

© 2005 The Linnean Society of London, Botanical Journal of the Linnean Society, 2005, 147, 257 ­ 274

257


258

A. P. SHIPUNOV ET AL. cies was later upgraded to species rank by Nevski (1935) because there are some obvious morphological differences between D. majalis s.s. and D. baltica, especially in leaf form and lip shape ( Table 1). Most authors now accept D. baltica as a separate species, some suggesting that D. praetermissa (Druce) SoÑ and D. purpurella (T. & T.A.Stephenson) SoÑ are its closest relatives (e.g. Vermeulen, 1947; Senghas, 1968; Averyanov, 1990). Morphological characters advocated for distinguishing D. baltica vary among authors (Table 1), although most descriptions mention the long, pointed leaves, short inflorescence and relatively wide lip with small lateral lobes. Recent morphometric investigations ha ve shown that Dactylorhiza allotetraploids often have morphological character states that are generally intermediate between D. incarnata and D. fuchsii (e.g. Tyteca & Gathoye, 1993). Biochemical and molecular methods can highlight molecular markers that are able to reveal inheritance, parentage and possible introgression between taxa. Studies of allozyme markers (HedrÈn, 2002), AFLPs (HedrÈn et al., 2001), plastid markers and ITS alleles (Y. Pillon, unpubl. data) and plastid RFLPs (HedrÈn, 2003) showed that: (1) most allotetraploids have indeed originated from hybridization between D. incarnata and D. fuchsii or D. maculata (L.) SoÑ; (2) they have originated several times (and are most likely still being generated); (3) most western European allotetraploids are easily distinguished by molecular characters from their parental species; (4) most allotetraploids have inherited plastid markers from either D. fuchsii or D. maculata rather than from D. incarnata, indicating that D. fuchsii and D. maculata are more often maternal parents; and (5) some allotetraploids ha ve acquired markers thus far not found in parental taxa ( HedrÈn et al. 2001; HedrÈn, 2003; Devos et al., 2003; Y. Pillon, unpubl. data; Shipunov et al., 2004).

(between the Arctic Circle and 50N latitude) to the Urals (SoÑ, 1980; Averyanov, 1990). The most recent evidence is, however, that `D. baltica' populations in the southern Urals have been misidentified and should be assigned to D. fuchsii (Kulikov & Filippov, 1999a). Thus, current opinion limits the distribution of D. baltica in European Russia to between 50 and 60 N latitude (with two exceptions in the northern Urals) and west of 60 longitude (Fig. 1). The epithet baltica was first used by Klinge (1895, 1898) for a subspecies of `Orchis' latifolia L., nom. illeg. [= Dactylorhiza majalis (Reichb.) P.F.Hunt & Summerhayes], a species with a western European distribution, long believed to be another member of the polyploid complex ( Averyanov, 1990). This subspe-

0 200 400

Figure 1. The putative European distribution of Dactylorhiza baltica (Averyanov, 1990; Kulikov & Filippov, 1999a). Each collection site is labelled with an abbreviated region name (see Appendix 1).

Table 1. The most diagnostic morphological characters of Dactylorhiza baltica in this study compared with those from three previous studies (all measurements in millimetres) Characters Plant height Leaf length Leaf width Leaf spots (1 light, 2 heavy) Length of inflorescence Length of lowest bract Spur length Lip length Lip width Length of lip middle lobe (from the base of sinuses) Klinge (1898) 250­700 100­250 15­35 1 20­80 ­ 6­9 6­7 8­12 <3 Nevski (1935) 300­600 90­200 20­32 1 30­95 20­30 7.5­9 7­8.5 9­10 2.5­3.5 Delforge (1995) 250­700 100­250 15­40 1 30­100 ­ 6­9 6­9 8­13 ­ This study 250­700 90­250 15­40 1­2 20­100 >20 6­9 6­9 7­13 <4

© 2005 The Linnean Society of London, Botanical Journal of the Linnean Society, 2005, 147, 257 ­ 274


EVOLUTION OF DACTYLORHIZA BALTICA However, no molecular analysis has yet been performed on D. baltica, which is unique among other Dactylorhiza allotetraploids due its eastern distribution and relative isolation from other allotetraploids . Moreover, there are few morphometric studies of Russian dactylorchids (Kulikov & Filippov, 1999a, b). Our recent investigation of European Russian Dactylorhiza showed good agreement between morphometric characters and molecular markers suc h as plastid microsatellites and ITS alleles ( Shipunov et al., 2004). Plastid microsatellites had previously been shown to be useful for revealing geographical patterns, the maternal parentage of hybrids, and even some relationships among populations (Y. Pillon, unpubl. data; Shipunov et al., 2004). ITS alleles, on the other hand, generate clear phylogenetic patterns and thereby help to distinguish species ( Bateman et al., 2003), but their biparental inheritance and especially ITS conversion following allopolyploid events (Chase et al., 2003) can blur species boundaries. Pillon (Y. Pillon, unpubl. data) studied the D. maculata complex (primarily D. maculata s.s. and D. fuchsii), the D. incarnata complex and their allotetraploid derivatives throughout their ranges , but particularly in western and northern Europe . We have made use of the markers (plastid microsatellites and ITS sequences) identifi ed in this study. Shipunov et al. (2004) used the same markers to study general patterns of these same species and allotetraploid complexes in Russian Europe . The goal of this study is to explore diversity in detail in one of the Russian allotetraploid taxa, D. baltica, via morphological and molecular markers in the context of its likely origin via hybridization between the D. fuchsii and D. incarnata aggregates (both are treated as broadly defi ned species for simplicity). To the two sets of markers developed in Shipunov et al. (2004) and Y. Pillon (unpubl. data), we have added a set of two nuclear microsatellite markers, which we hope will be more variable than ITS and thus reveal more structure among populations of the putative parental taxa. We chose to focus on this allotetraploid taxon because it appeared to us that it w as likely to be operating locally as a `bridge' between the diploid taxa and would therefore make an appropriate subject for a more detailed study to determine whether we could detect evidence of this phenomenon though the study of both morphological and molecular markers.

259

investigation (see Appendix). All incoming samples were initially identified and assigned to a priori species by experts in regional fl oras (G. Konechnaja and I. Kucherov in Botanical Institute, Saint-Petersburg; N. Reshetnikova in Main Botanical Garden, Moscow and M. Vakhrameeva in Moscow University). In total, 98, 214 and 775 samples from 85, 112 and 121 populations were involved in simultaneous combined and molecular and morphometric analyses, respectively. For most analyses, we used a subset of samples that consisted of the allotetraploid species together with the putative parental species, D. fuchsii s.l. and D. incarnata s.l. One herbarium sample of D. baltica (Smolensk region, A. Averyanov, 2000, LE) and one of D. traunsteineri (Saut. ex Reichb.) SoÑ (Karelia, I. Kucherov, 1999, LE) were used as yardsticks for morphological comparison and also for DNA extraction. One sample of D. baltica from Estonia in the RBG K ew DNA Bank (Chase 9485) was also used for sequencing.

MOLECULAR

MARKERS

MATERIAL AND METHODS
Some of the (Shipunov et pean Russia regions) and samples were used in a previous study al., 2004), but many samples from Euro(mostly from central and north-western Britain were newly collected for this

Samples for DNA extraction were dried in silica gel (Chase & Hills, 1991). DNA was extracted by the 2 ¥ CTAB protocol (Doyle & Doyle, 1987 but without an RNA treatment). PCR was performed with a set of primers designed by Y. Pillon & M. F. Fay (unpubl. data) to amplify four polymorphic plastid loci: Orch1, Msf, Ms1 and Ms2, located in three plastid DNA regions: the trnS-trnG spacer, trnL intron and trnLtrnF spacer. Two pairs of specific primers were also used to amplify length-variable regions of ITS ribosomal DNA that, taken together, indicate which ITS alleles are found in each sample (Shipunov et al., 2004; Y. Pillon, unpubl. data). To identify other molecular makers that are sufficiently polymorphic to reveal interpopulational structure, we have developed several nuclear microsatellites, two of which proved useful for this study. To develop these markers, we used a strategy proposed by Fisher, Gardner & Richardson (1996), which employs a degenerate primer PCT4 ( Brachet et al., 1999) that contains a (CT)6 repeat at its 3¢ end. The conditions for PCR amplifi cation were those of Fisher et al. (1996). Several PCR products were c loned using the Promega pGEM-T Easy Vector System. These were reamplified from transformed bacterial colonies by touching them with a sterile toothpic k and using that sample as the template in a further round of PCR. Primers for this PCR were located on the vector. Amplified DNA fragments were purifi ed using QIAquick PCR mini-columns (QIAGEN, Inc.), following the manufacturer's protocols, and sequenced on a 3100 genetic analyser (Applied Biosystems Inc .), following the manufacturer's protocols (we again used the primers that annealed to sites on the vector).

© 2005 The Linnean Society of London, Botanical Journal of the Linnean Society, 2005, 147, 257 ­ 274


260

A. P. SHIPUNOV ET AL.

Table 2. Nuclear microsatellite loci used Locus Repeat Annealing temperature, C Size range in base pairs (bp) No. of alleles Primer sequence (5¢-3¢) Forward Reverse Da963_1-2 (CAG)5 52 103­120 5 TCCATATCCCCCTTCCTCAA CTCTCTCTCTCTTGTCTTTA Ds3978_2-1 (TTA)6 52 126­153 10 GAGATATATAGAGTGGTGGT TATGCGTTGGTATTGGGAGT

Sequence editing and assembly of the two complementary strands used SequenceNa vigator and AutoAssembler (Applied Biosystems Inc.) software. Several pairs of specific primers were subsequently designed to amplify the most promising microsatellite loci. The resulting fragments were checked to determine whether they revealed any polymorphisms , and two loci were then chosen for this investigation ( Tables 2, 3). The size of each fragment was determined using GeneScan and Genotyper software (Applied Biosystems Inc.). The subsequent unweighted pair-group (UPGMA) tree construction used PAUP* version 4.0b10 (Swofford, 2000). For most statistical analyses, the calculation of distances between samples w as based on the proportion of shared alleles.

Table 3. Nuclear microsatellites alleles most typical for some Dactylorhiza species Typical lengths (base pairs) and names for alleles Locus Da963­1-2 Ds3978­2-1 D. incarnata 110 (da110) 126 (ds126) D. fuchsii 114 118 130 153 (da114), (da118) (ds130), (ds153) D. maculata 103 (da103) 143 (ds143), 147 (ds147)

Windows (Venables, Smith & R Development Core Team, 2002).

MORPHOLOGY
We used the set of 14 morphological c haracters, slightly modified from previous work ( Shipunov et al., 2004). These characters were measured in nature on either the same plants that were used for DNA extractions or, on a few occasions (e.g. for D. praetermissa and several populations of D. baltica), we measured neighbouring plants in the same population. We used principal component analysis (PC A) and multidimensional scaling (MDS) of individual and population data. In the latter case, population medians (because these are usually more robust than means; Fowler, Cohen & Jarvis, 1999) were used. The analysis of population data was wider than the analysis of individual data because we inc luded some species and populations for whic h DNA sampling and morphometric measurements were made on different plants. We have also analysed correlation from individual data (all species included) for all morphological measurements and nuclear DNA markers, and used recursive partitioning analysis, which is the model-based version of discriminant analysis , describing which character values best predict the existing classification (Breiman et al., 1984). Statistical calculations used the R program, version 1.8 for

RESULTS
DNA
MARKERS

Most D. baltica plants have the A haplotype (the unique combinations of plastid fragment lengths , in this case four, from different regions of the plastid genome, henceforth termed `haplotypes'), which is typical of D. fuchsii (see Appendix), but several populations also contained the E and H haplotypes from D. incarnata. Most plants have more than one ITS allele, with the D. incarnata and D. fuchsii alleles being most common (several samples also ha ve the D. maculata ITS allele). The nuclear microsatellite alleles are also mostly those of D. incarnata (da110, sometimes ds126) and/or D. fuchsii (da114, da118, ds130, ds153). For most population samples, multiple alleles were amplified (see Appendix); putative diploids displayed 1­2 alleles, and several of the allotetraploids had up to four alleles. Some samples collected as D. fuchsii (a diploid) have 3­4 alleles, but these plants are in fact `northern tetraploids' ( Shipunov et al., 2004) and are morphologically intermediate between D. maculata and D. fuchsii. Several D. incarnata samples also displayed three alleles; most of them belong to populations 242 and 215, and

© 2005 The Linnean Society of London, Botanical Journal of the Linnean Society, 2005, 147, 257 ­ 274


EVOLUTION OF DACTYLORHIZA BALTICA bear D. fuchsii alleles ds153, and da114­da118, respectively. In addition, plants from population 215 have the A plastid haplotype (see below). We can guess that these samples belong to triploids of likely hybrid origin. Although the repeats selected were all in triplets (Table 2), the lengths of the fragments produced indicates that some of the variation we detected occurred in the flanking regions rather than just in the repeats; we did not verify this hypothesis by sequencing some of the variants. Nonetheless, some of the length variants (alleles) detected were diagnostic of the taxa studied, which made them useful markers. Analysis reveals that for nuclear markers overall polymorphism in D. baltica and D. fuchsii is higher than in D. incarnata (F = 2.89 and 2.39, respectively; P < 0.05). The distribution of the nuclear microsatellite alleles is complex, and it is difficult to say which, if any, are typical for the parental taxa. Some are predominant in one species, e.g. da110 in D. incarnata, but then these also show up in other species occasionally (D. fuchsii) and commonly in D. baltica, which is expected since D. baltica has D. incarnata as one of its parents. When they occur in the other parent, they are often associated with the ITS allele of the other species as well, which thus provides two lines of evidence for introgression. Other microsatellite alleles are found in only some populations of one of the parental taxa (e.g. ds153 in some D. fuchsii) and then in some D. incarnata (often again with the D. fuchsii ITS allele) and some D. baltica. Such patterns show that local populations of all species, parental and allotetraploids, are likely to have similar microsatellite alleles. This pattern emerges particularly c learly in the UPGMA analysis (see below).

261

2

4

b

b b bb b b b b b bb b bbb bb b b b

-2

Dim 2 0

b

-4

fuchsii incarnata maculata traunsteineri

b b

-6 -4

-2

0

2 Dim 1

4

6

8

Figure 2. Multidimensional scaling of morphological data from individuals (species highlighted). All points marked with the letter `b' correspond to Dactylorhiza baltica.

MULTIVARIATE

ANALYSES

Both PCA and MDS of the morphological data revealed similar patterns. There are three overlapping groups (Fig. 2) consistent with haplotype and ITS allele distributions and species descriptions ( D. fuchsii + D. maculata; D. baltica and D. incarnata) in general. Plants of the presumed autotetraploid D. maculata were not clearly separated from those of D. fuchsii. Dactylorhiza baltica plants are located not between the two putative parents but rather, are offset to the bottom right-hand corner of the graph. Several plants of D. baltica, however, overlap with the two parental groups. Similar results were obtained by Tyteca & Gathoye (1993) for D. majalis s.s., whereas D. praetermissa samples in their multivariate plot were located directly between the presumed parental species (again, D. fuchsii and D. incarnata). Dactylorhiza baltica plants with haplotypes A and E are distributed closer to D. fuchsii (A haplotype) and

D. incarnata (E haplotype), respectively (not shown). The percentages of the D. incarnata ITS allele vary among D. baltica individuals; plants possessing more copies of this allele are usually located c loser to D. incarnata. Many plants of D. maculata have other haplotypes, B, N or X, which is typical of this species throughout its range. Population-level analysis revealed similar groups (Fig. 3), but for D. baltica, in this case, the offset was less and the diversity greater. We were able to include some additional species in this analysis so it could be seen that British populations of D. praetermissa and D. purpurella were located near the D. baltica points. Addition of other morphological c haracters would be likely to improve the separation of these allotetraploid taxa; the characters we selected were those that appeared to be good for separating D. fuchsii from D. incarnata and D. baltica. A population of D. incarnata ssp. coccinea (Pugsley) SoÑ from Wales was marginal to the D. incarnata group. In both cases the most important c haracters (which have relatively high loadings in the fi rst component, PC1) are for individuals, plant heights, all leaf characters and inflorescence lengths, and for populations, bract lengths, stem diameters and leaf lengths. Simultaneous analysis of morphology, ITS alleles and nuclear microsatellites produced a less ambiguous structure (Fig. 4), both for individuals and populations (the latter not presented), demonstrating that there is agreement between these kinds of data. An analysis of individuals combining all characters, including the uniparentally inherited plastid

© 2005 The Linnean Society of London, Botanical Journal of the Linnean Society, 2005, 147, 257 ­ 274


262
4

A. P. SHIPUNOV ET AL.
8

99 2

fuchsii incarnata maculata traunsteineri b b b b bb b b

6

PC2 (13.2% of variance) 0 2 4

PC2 (25.6% of variance)

98

216 219 100 218 213 243 111 coccinea fuchsii incarnata maculata praetermissa purpurella traunsteineri 231 18 5 306 304

b b b b bbb b b bb bb b b b

114 0 -2

-4

-6

-4

-2

0

2

4

-4 -8

-2

-6

PC1 (31.0% of variance)

-4 -2 0 PC1 (24.6% of variance)

2

Figure 3. Principal component analysis of the morphological data for populations (species highlighted). All points marked with population numbers correspond to Dactylorhiza baltica.

Figure 5. Simultaneous principal component analysis of all characters (morphology, plastid DNA markers, ITS alleles and nuclear microsatellites) from individuals. All points marked with the letter `b' correspond to Dactylorhiza baltica.

b PC2 (16.5% of variance) 0 2 4 b

fuchsii incarnata maculata traunsteineri b

b b bb bb bb b b bb bb b b b

b bb b

-4

-2

0 2 4 PC1 (19.7% of variance)

6

8

Figure 4. Simultaneous principal component analysis of characters with biparental origin (morphology, ITS alleles and nuclear microsatellites) from individuals. All points marked with the letter `b' correspond to Dactylorhiza baltica.

sequences, changed the picture completely ( Fig. 5). Dactylorhiza baltica specimens were divided into two groups, each corresponding to their contrasting haplotypes and consequently, to their putative maternal parents. Some D. incarnata samples appeared close to

D. fuchsii; these individuals, from population 215, have a D. incarnata morphology, but most have the A haplotype typical of D. fuchsii with a low frequency of the D. incarnata ITS allele. The corresponding analysis of populations generated a similar result (not presented). To represent better some interpopulation relationships and possible geographical patterns, we constructed a UPGMA tree in PAUP for the nuclear DNA data (presence/absence of molecular markers) for populations of D. baltica and its putative parental species. This analysis showed that most D. baltica populations have clear relationships with their putative parents , either D. fuchsii, D. incarnata or even, in some cases, both. The tree (Fig. 6) demonstrates that most D. baltica populations share the same terminal c lusters with nearby populations of D. incarnata and/or D. fuchsii (i.e. from adjacent regions, the same regions, or even the same collection sites). The tree cannot clearly distinguish between D. fuchsii and D. incarnata, but these species have been distinguishable in other molecular analyses (Bateman et al., 2003; Y. Pillon, unpubl. data). The morphological characters formed three correlation groups: (1) most of the vegetative characters, including bract and inflorescence length but not leaf spots, (2) leaf spots and (3) fl oral characters in which the largest significant correlation is between lateral lobe and mid-lobe lengths ( r = 0.83, P << 0.05). The DNA characters most correlated with species parti-

-2

6

© 2005 The Linnean Society of London, Botanical Journal of the Linnean Society, 2005, 147, 257 ­ 274


EVOLUTION OF DACTYLORHIZA BALTICA
fuchsii 19 Sm1 fuchsii 22 Ar1 fuchsii 23 Ar1 fuchsii 30 Mu4 fuchsii 33 Mu4 fuchsii 34 Mu4 fuchsii 10 Tv1 fuchsii 14 Tv2 fuchsii 24 Ar1 fuchsii 35 Mu4 fuchsii 42 Mu2 fuchsii 31 Mu4 fuchsii 90 Mu3 fuchsii 91 Mu3 fuchsii 253 Mu3 fuchsii 9 Tv1 fuchsii 60 Mu4 fuchsii 77 Mo2 traunsteineri 110 Kr1 traunsteineri 130 Mu1 fuchsii 73 Mo2 fuchsii 38 Mu4 fuchsii 92 Mu3 fuchsii 76 Ka1 BALTICA 111 Sm1 fuchsii 305 Mo3 fuchsii 93 Mu3 fuchsii 241 Ka1 fuchsii 321 Mo3 fuchsii 2011 En2 fuchsii 310 Mo3 fuchsii 307 Mo3 BALTICA 99 Or1 fuchsii 12 Tv1 fuchsii 312 Mo3 fuchsii 313 Mo3 fuchsii 206 En2 fuchsii 314 Mo3 fuchsii 41 Mu2 fuchsii 233 Sa4 traunsteineri 78 Mu1 fuchsii 254 Mu3 fuchsii 221 Sa3 fuchsii 222 Sa3 fuchsii 302 Mo3 fuchsii 303 Mo3 BALTICA 304 Mo3 BALTICA 243 Ka1 fuchsii 301 Mo3 incarnata 215 Ps1 fuchsii 308 Mo3 fuchsii 322 Mo3 fuchsii 97 Or1 BALTICA 98 Or1 BALTICA 218 Ps1 BALTICA 231 Sa2 BALTICA 100 Or1 fuchsii 101 Or1 BALTICA 216 Ps1 fuchsii 311 Mo3 incarnata 81 Mo2 praetermissa 201 En2 incarnata 309 Mo3 incarnata 25 Ar1 incarnata 70 Ta1 BALTICA 219 Ps1 fuchsii 325 Tv1 incarnata 252 Mu3 coccinea 2042 En2 incarnata 8 Tv2 incarnata 15 Tv2 incarnata 16 Tv2 BALTICA 18 Sm1 BALTICA 5 Tv1 incarnata 17 Sm1 incarnata 4 Tv1 incarnata 6 Tv1 incarnata 72 Mo2 incarnata 232 Sa4 incarnata 242 Ka1 incarnata 251 Mu3 incarnata 208 Mo2 incarnata 209 Mo2 incarnata 210 Mo2 incarnata 217 Ps1 incarnata 211 Mo2 fuchsii 307 Mo2 BALTICA 213 Sa2 BALTICA 306 Mo3 fuchsii 326 Tv1 fuchsii 40 Mu2

263

Figure 6. UPGMA tree derived from nuclear DNA data of investigated populations. All labels contain the species epithet, population number and code for the collection site (see Fig. 1).
© 2005 The Linnean Society of London, Botanical Journal of the Linnean Society, 2005, 147, 257 ­ 274


264

A. P. SHIPUNOV ET AL. D. incarnata ITS allele (the plants with less than 3.5% are D. fuchsii), (b) leaf spots (plants with unspotted leaves belong to D. incarnata) and (c) leaf length (plants with leaves shorter than 136 mm are D. baltica). Analysis of the morphological c haracters (misclassification error rate = 8.2%) alone added to this set another character (d): length of the lateral lobes of the lip (plants with lateral lobes less than 2 mm are mostly D. incarnata).

tioning are percentage of D. incarnata ITS alleles, which formed a correlation group (4) with da110 and da118 nuclear microsatellite alleles. This group also has a significant correlation with groups (2) and (3). Recursive partitioning revealed that the three c haracters most important in the analysis of data for D. fuchsii, D. incarnata and D. baltica (Fig. 7; misclassification error rate = 5.2%) were: (a) percentage of

i.inc < 3.5 |

DISCUSSION
The position of D. baltica plants between D. fuchsii and D. incarnata in all analyses provides general support for the hypothesis of origin of D. baltica by hybridization between these two species. Although D. maculata and D. fuchsii are difficult to separate morphometrically, they are clearly distinct on a molecular basis, particularly plastid haplotypes, and it is clear that the markers in D. baltica are those of D. fuchsii and D. incarnata, with occasional markers (ITS alleles and nuclear microsatellites) from D. maculata. Some populations of D. maculata have the A haplotype of D. fuchsii rather than B and other haplotypes related to B that are typical of D. maculata throughout its range (Shipunov et al., 2004; Y. Pillon, unpubl. data). In our sampled plants of D. maculata, the A haplotype is more common than the B haplotype , whereas no sample of D. fuchsii has the B haplotype; so, based purely on the results of our study, we could not clearly state that A is the haplotype of D. fuchsii and B that of D. maculata, although we know this to be the case from other studies. The UPGMA results also demonstrate that D. baltica is related to D. incarnata and D. fuchsii (not D. maculata), but it is clear that in this region the two

LEAF.L < 156.5 fuchsii fuchsii

LEAF.SP < 1.5

i.fuch < 32.5 incarnata incarnata

LEAF.L < 135.5

P.HIGH < 327.5 fuchsii fuchsii baltica

Figure 7. Tree of binary recursive partitioning analysis (a dichotomous key) of morphological and nuc lear DNA characters in the model. The nodes are marked with character codes (see Table 4). Analysis performed on data from individuals of three species (Dactylorhiza incarnata, D. fuchsii and D. baltica).

Table 4. Morphological characters used (all measurements in millimetres) Label P.HIGH LEAF.L LEAF.W L.WPOS LEAF.SP ST.DIAM INFL.L SPUR.L LIP.L LIP.W MIDD.L LATER.L BR.L LIP.COL Description Plant height, from the ground to the top of infl orescence Length of longest leaf Width of longest leaf Position of maximal width (the distance from leaf base to the place of maximal width) Leaf spots (0 none, 1 light, 2 heavy) Stem diameter (measured just above the node of longest leaf) Length from the lowest bract to the top of infl orescence Spur length, measured from lower side of spur Lip length, from the base to the top of middle lobe Lip maximum width Length of middle lobe of the lip, from the base of the sinus to the top apex of lobe Length of lateral lobe of the lip, from the base of the sinus to the top apex of lobe Length of lowermost bract Lip colour (1 white or nearly white, 2 pink, 3 dark pink)

© 2005 The Linnean Society of London, Botanical Journal of the Linnean Society, 2005, 147, 257 ­ 274


EVOLUTION OF DACTYLORHIZA BALTICA species of the D. maculata complex are difficult to distinguish, and pure populations of either are rare . At the same time, the D. baltica points are offset in all graphs, as if some independent evolution of this hybrid has occurred and is refl ected in a morphological bias. HedrÈn et al. (2001) also noticed in their AFLP study the same type of offset pattern for Dactylorhiza allotetraploids relative to their putative parents . It is also possible that this offset pattern could be gener ated by introgression from another (or other) species , but we favour the former explanation because we do not find markers of any other species of