ODOnsqiplpmtafetmridflasaoppdm

tcgS

ts

Molecular Phylogenetics and EvolutionVol. 13, No. 1, October, pp. 67–76, 1999Article ID mpev.1999.0628, available online at http://www.idealibrary.com on

Phylogeny and Evolution of Orchis and Allied Genera Based on ITSDNA Variation: Morphological Gaps and Molecular Continuity

Serena Aceto,* Paolo Caputo,† Salvatore Cozzolino,‡ Luciano Gaudio,* and Aldo Moretti†

*Dipartimento di Genetica, Biologia Generale e Molecolare, Universita Degli Studi di Napoli Federico II, Via Mezzocannone 8,I-80134 Naples, Italy; and †Dipartimento di Biologia Vegetale e ‡Orto botanico, Universita Degli Studi di Napoli Federico II,

Via Foria, 223, I-80139 Naples, Italy

Received July 10, 1998; revised December 14, 1998

e1rReR

rwiiteor1ra

stsictra1gt

alri

oa1t

Phylogenetic relationships among members of genusrchis and allied genera Aceras, Anacamptis, Barlia,actylorhiza, Gymnadenia, Himantoglossum, Neotinea,phrys, Platanthera, and Serapias were inferred fromucleotide sequence variation in the internal transcribedpacer (ITS) regions of nuclear ribosomal DNA. Se-uences were subjected to various alignments by chang-ng the gap opening and extension parameters. After areliminary parsimony analysis, the alignment with the

owest homoplasy indicators was chosen as optimal. Thehylogenetic analysis, carried out on the optimal align-ent by using Gennaria as an outgroup and a total of 31

axa, showed that all the genera considered in this studyre nested in Orchis despite their distinct morphologicaleatures. Genus Orchis is divided into two major clades,ach of which includes one or more of the other genera inhis study. The resulting phylogenetic hypothesis does notatch previous conclusions based on vegetative and flo-

al morphology of the taxa involved but is congruent withsoenzyme, karyological, and chloroplast DNA restrictionata. Our results indicate that floral morphology is highlyexible and current generic and infrageneric limits arertificial. Even if some floral characters closely corre-pond to the molecular data, most are highly homoplasticnd thus unsuitable for phylogenetic reconstruction. Vari-us traits pertaining to floral morphology may be inter-reted as a result of ecological convergence related toollinator-mediated selection; such characters can un-ergo drastic modifications without correspondingly dra-aticgeneticchanges. r 1999 Academic Press

Key Words: internal transcribed spacer; floral evolu-ion; molecular systematics; Orchidaceae; Aceras; Ana-amptis; Barlia; Dactylorhiza; Gymnadenia; Himanto-lossum; Neotinea; Ophrys; Orchis; Platanthera;erapias.

INTRODUCTION

Orchis L. (Orchidaceae), in its present circumscrip-ion, is a rather homogeneous genus with over 30

pecies centered in the Mediterranean basin and outli- r

67

rs in the Northern regions of Europe andAsia (Dressler,993). Orchis is clearly related to other, mainly Mediter-anean, genera, i.e., Aceras R. Br., Anacamptis L. C. M.ichard, Barlia Parl., Dactylorhiza Necker, Gymnad-

nia R. BR., Himantoglossum Koch, and Neotineaeichenb. fil.The striking beauty of these taxa, their wide occur-

ence in countries with a long botanical tradition, asell as their endangered status has fostered much

nterest in these plants which, accordingly, have beenntensively studied over the past 100 years. Despitehese studies, the pervasive parallel evolution andcological convergence which characterize Europeanrchids (Dressler, 1981, 1993) as well as a tendency toeticulate evolution in some groups (e.g., Ehrendorfer,980) greatly hinder a clear understanding of theelationships both within Orchis and between it and itsllied taxa.Floral features have traditionally been employed in

egregating groups within orchids. However, floralraits usually show strong discontinuities, often corre-ponding to unique apomorphies, which are useless fornferring phylogenetic relationships. Moreover, floralharacters are often inconsistent compared to otheraxonomic evidence (e.g., chromosome number). As aesult, the few generic-level taxonomic treatments avail-ble (e.g., Vermeulen, 1972; Cauwet-Marc and Balayer,984) and the scanty but novel circumscriptions for theenus (e.g., Love and Love, 1972) are often contradic-ory.

In addition, studies based on isozymes (Schlegel etl., 1989; Rossi et al., 1994) have shown a pervasiveack of correspondence between floral traits and geneticelationship, indicating that some of the generic andnfrageneric boundaries may be artificial.

More recently, nucleic acid data have been used inrder to infer relationships among members of Orchisnd allied genera (Caputo et al., 1995; Cozzolino et al.,998). These studies, based on chloroplast DNA restric-ion analysis, have shown that both Aceras and Dactylo-

hiza Necker are nested within Orchis. An exploratory

1055-7903/99 $30.00Copyright r 1999 by Academic PressAll rights of reproduction in any form reserved.

atrfsptvIme(Sia

qdhhb

P

AcrbehOS(lsmppOslS

cLm

D

np

p(5

TTrTfcae9tf

1dmasrtFr

D

WSawsoVUfiom

Iss(eiiTsdeatsccPsc

68 ACETO ET AL.

nalysis of the Internal Transcribed Spacer 1 (ITS1) ofhe nuclear ribosomal DNA in five Orchis species wasecently conducted (Cozzolino et al., 1996) to test theeasibility of ITS1 sequencing as a tool to infer relation-hips in the genus. The evidence obtained from thisreliminary investigation showed good correlation withhe isozyme and chloroplast DNA data. Among thearious sequences employed in phylogenetic inference,TS1 and 2 appear to have a mutation rate whichakes them suitable (Baldwin et al., 1995 and refer-

nces therein) for studies of intergeneric relationshipse.g., Suh et al., 1993; Susanna et al., 1995; Bogler andimpson, 1996) or of infrageneric relationships in rap-

dly evolving taxa (e.g., Kim and Jansen, 1994; Bayer etl., 1996).In this paper we present evidence gathered by se-

uencing ITS1 and 2 for 31 orchid taxa. Data will beiscussed in the light of the selective pressures whichave probably acted on floral morphology, consequentlyindering the recognition of the phylogenetic patternased upon floral morphology.

MATERIALS AND METHODS

lant Material

The following taxa were employed in this study:ceras anthropophorum (L.) R. Br. ex Aiton fil., Ana-amptis pyramidalis (L.) L. C. M. Richard, Barliaobertiana (Loisel.) Greuter, Dactylorhiza romana (Se-astiani) Soo, Gennaria diphylla (Link) Parl., Gymnad-nia conopsea (L.) R. Br. ex Aiton fil., Himantoglossumircinum (L.) Spreng., Neotinea maculata (Desf.) Stearn,phrys tenthredinifera Willd., O. collina Banks etolander ex A. Russel, O. coriophora L. subsp. fragrans

Pollini) Sudre, O. italica Poir., O. lactea Poir., O.axiflora Lam., O. longicornu Poir., O. mascula (L.) L.ubsp. mascula, O. militaris L., O. morio L. subsp.orio, O. palustris Jacq., O. papilionacea L. subsp.

apilionacea, O. patens Desf., O. pauciflora Ten., O.rovincialis Balbis ex Lam. et DC., O. purpurea Huds.,. quadripunctata Cyr. ex Ten., O. simia Lam., O.

pitzelii Sauter ex Koch, O. tridentata Scop., O. ustu-ata L., Platanthera chlorantha (Custer) Reichenb., anderapias lingua L.All specimens were field collected by the authors and

ultivated at the Botanical Garden of Naples, Italy.eaves were collected at flowering time. Voucher speci-ens of the examined plants are deposited at NAP.

NA Extraction and Sequencing

Leaves (1 g per sample) were ground in liquiditrogen and total DNA was extracted following therocedure described in Caputo et al. (1991).ITS1 and 2 were amplified by using two pairs of

rimers which anneal in the 38 region of the 18S58-GGAGAAGTCGTAACAAGGTTTCCG-38) and in the

8 region of the 5.8S (58-ATCCTGCAATTCACACCAAG- s

ATCG-38), or in the 38 region of the 5.8S (58-TGCAGAATCCCGTGAACCATCG-38) and in the 58egion of the 26S (58-CCAAACAACCCGACTCG-AGACAGC-38), respectively. PCRs were carried out

or 30 cycles in a Perkin–Elmer–Cetus 9600 thermocy-ler. Initial conditions were as follows: 1 min denatur-tion at 94°C, 1 min annealing at 55°C, and 45 sxtension at 72°C. Samples were denatured for 5 min at4°C before the beginning of the first cycle; extensionime was increased by 3 s/cycle; and extension wasurther prolonged for 7 min at the end of the last cycle.

PCR fragments were then purified by using Microcon00 microconcentrators (Amicon, Danvers, MA) andouble-strand sequenced in both directions by using aodification of the Sanger dideoxy method (Sanger et

l., 1977) as implemented in a double-strand DNA cycleequencing system with fluorescent dyes. Sequenceeactions were then loaded into a 373A Applied Biosys-ems Automated DNA sequencer (Applied Biosystems,oster City, CA). Various sequencing experiments wereepeated to solve all uncertainties.

ata Analysis

A raw alignment was accomplished by using Clustalver. 1.6 (Thompson et al., 1994) with default settings.

equences were then reduced to only ITS1 and ITS2 byligning them with the 38 termini of 18S and 5.8S andith the 58 termini of 5.8S and 26S of various monocot

equences available in the literature. When differencesf starting and ending points occurred, the sequence ofanilla planifolia Andrews (GenBank Accession no.66819, Mai and Coleman, unpublish.) was used. Sixctitious unknowns (N’s) were added at the 38 terminusf both ITS1 and ITS2 in all taxa to prevent terminalisalignments.The optimal alignment was searched separately for

TS1 and ITS2 by employing Clustal W. An alignmentoftware was preferred to sequential pairwise compari-on (Swofford and Olsen, 1990) used by other workerse.g., Oxelman et al., 1995; Susanna et al., 1995; Bayert al., 1996; Sang et al., 1996) because of the difficulty innterpreting a fairly large amount of small (1–5 bp)nsertions/deletions (indels) in various ingroup taxa.he approach used in Clustal W is two-stepped: a firsttep in which pairwise alignments are carried out toetermine which are the closest sequences in each andvery possible pair and a second step in which multiplelignments are calculated according to the results ofhe first step. In each step, gaps are opened to maximizeequence similarities. Gap opening and extension areontrolled by four parameters: PWGAPOPEN (whichontrols the gap opening cost in the first step), PWGA-EXT (which controls the gap extension cost in the firsttep), and GAPOPEN and GAPEXT (which are theorresponding parameters for the second step).The optimal alignment was determined using a

trategy similar to that described in Bogler and Simp-

saem1sG

ecg

musewtcrwaaIbprmsnosai

mteamaampprtmBtiiEtori

aotgHaTl

eiwrtsc16

moni

B

tAbsaoafI0p2l(iOl

ftc2aw

69ITS DNA VARIATION IN Orchis

on (1996), which employs homoplasy indicators (C.I.nd R.I.) as optimality criteria, and gap opening andxtension parameters were varied across every align-ent with unit increments from 4 to 15 and from 4 to

0, respectively. We also imposed the following con-traint: PWGAPOPEN 5 GAPOPEN $ PWGAPEXT 5APEXT within any single alignment.In all the tests, transitions and transversions were

qually weighted, and the parameter MAXDIV (whichontrols the delay of the alignment for the most diver-ent sequences) was set to 85%.Each alignment was then converted to numeric for-at by means of a word processor macro, fictitiousnknowns were removed, and the resulting matrix wasubjected to a heuristic parsimony analysis (mh*; bb*;)mploying the Hennig86 software (Farris, 1988). Gapsere coded as missing data and all characters were

reated as unordered (nonadditive). For each analysis,onsistency and retention indices (C.I. and R.I.) wereecorded. The optimal alignments were chosen as thoseith highest C.I. 3 R.I. values. In case of ties, all thelignments with identical C.I. 3 R.I. values werenalyzed further. The optimal alignments for ITS1 andTS2 were then united (for a total of eight matricesecause of ties within ITS1) and subjected again to aarsimony analysis and to the same optimality crite-ion. Ties in this case were resolved by choosing theost parsimonious solution in terms of the number of

teps. This yielded a single optimal alignment. Becauseumber of steps is positively correlated with the gappening and extension costs (see also Bogler and Simp-on, 1996), we checked that all equally homoplasiouslignments yielding the minimal number of cladogramsndicated the same topology.

Because the gap parameters of the optimal align-ents, although yielding the least homoplasious solu-

ions, are arbitrary per se, in order to verify to whatxtent ambiguous gap positions may influence topologynd to avoid the problem that a single set of parametersay influence the final result, we used the elision

pproach, as described by Wheeler et al. (1995). Thispproach consists in ‘‘eliding’’ various individual align-ents into a single combined alignment on which a

hylogenetic analysis is carried out. The elision ap-roach, although afflicted by theoretical difficultieselated to homology assessment (Wheeler et al., 1995),ends to give more resolved solutions compared to otherethods of the same family (e.g., Gatesy et al., 1993).ecause merging all the matrices obtained by varying

he gap parameters would have been computationallynfeasible, we sampled our set of matrices at widentervals according to the variation of gap parameters.ight matrices were sampled and merged in such a way

hat a single combined matrix of 4260 characters wasbtained. This matrix was then analyzed by using aecent version of Nona (Goloboff, 1993), with the follow-

ng parameters: hold 1000; hold/25; mult*25; max*. O

Once the optimal alignment was found as indicatedbove, the calculation of pairwise distances was carriedut independently for ITS1 and ITS2 using the ‘‘dis-ance matrix’’ option of Clustal W and correcting foraps but not for multiple substitutions. A heuristicennig86 analysis was run (mh*; bb*;) on the optimallignment by using Gennaria diphylla as an outgroup.he resulting cladograms were examined and manipu-

ated with the Clados software package (Nixon, 1992).Bootstrap percentages (Felsenstein, 1985) and pres-

nce of phylogenetic signal (Permutation Tail Probabil-ty test as implemented by Faith and Cranston, 1991)ere calculated by using the SEQBOOT (1000 and 500

eplicates, respectively) and DNAPARS programs ofhe Phylip 3.57 package (Felsenstein, 1993). Bremerupport (Bremer, 1988, 1994; Kallersjo et al., 1992) wasalculated by using a recent version of Nona (Goloboff,993) that explored the trees up to six steps longer (BS) than the most parsimonious solutions.Finally, although we preferred to treat indels asissing data and not as a fifth state in order to avoid

verweighting, we also carried out a heuristic phyloge-etic analysis on the optimal data matrix by coding

ndels as a fifth state.

RESULTS

asic ITS Information

The lengths and GC contents of both the ITS’s for allaxa in the study are reported in Table 1 (EMBLccession nos.: Z94059 to Z94120; the increasing num-ers correspond to the alphabetical order of the genus/pecies initial, Z94059 being the ITS1 sequence of A.nthropophorum and Z94120 being the ITS2 sequencef S. lingua). ITS1 length ranged from 226 to 250 bpnd GC content from 39.6 to 50.4%. ITS2 length rangedrom 225 to 247 bp and GC content from 41.3 to 51.3%.TS1 sequence divergence in the ingroup ranged from.4% (pairwise distance between O. mascula and O.rovincialis and between O. purpurea and O. simia) to7.5% (pairwise distance between O. lactea and S.ingua); the minimum intergeneric distance was 2.8%pairwise distance between A. anthropophorum and O.talica) and the maximum infrageneric distance withinrchis was 25.5% (pairwise distances between O. col-

ina and O. patens).ITS2 sequence divergence in the ingroup ranged

rom complete identity (O. mascula and O. provincialis)o 25.8% (pairwise distance between O. morio and P.hlorantha). The minimum intergeneric distance was.1% (pairwise distance between A. anthropophorumnd O. italica) and the maximum infrageneric distanceithin Orchis was 23.6% (pairwise distances between

. papilionacea and O. patens).

P

aPiGlcys

bdMto

oati

oi

tlpHOoSgtt

maptOAibspzo

eCtoscp

(CCIpsrlsti

smtptortP

AABDGGHNOOOOOOOOOOOOOOOOOOOOOPS

70 ACETO ET AL.

hylogenetic Analysis

The cladistic analysis carried out on the optimallignment (ITS1: PWGAPOPEN 5 GAPOPEN 5WGAPEXT 5 GAPEXT 5 4, consensus length 261 bp,

nformative characters 44.8%; ITS2: PWGAPOPEN 5APOPEN 5 PWGAPEXT 5 GAPEXT 5 4, consensus

ength 262 bp, informative characters 40.8%; totalonsensus length 523, informative characters 42.8%)ielded eight equally parsimonious cladograms (839teps, C.I. 5 0.60, R.I. 5 0.69).The consistency index of our cladograms is quite low

ut altogether comparable with those obtained whenealing with large genera (e.g., Yuan et al., 1996).oreover, as has been demonstrated by Goloboff (1991),

he C.I. is negatively correlated with both the numberf terminal taxa and the number of characters.The consensus tree of two of the eight topologies

btained as well as the consensus among all topologiesre shown in Fig. 1. The cladogram(s) of Fig. 1 showhat all the genera considered in this study are nested

TABLE 1

ITS Length and GC Content for the Taxa in Study

Taxon

ITS1 ITS2

Length(bp) GC%

Length(bp) GC%

ceras anthropophorum 249 45.78 240 45.00nacamptis pyramidalis 240 39.58 234 47.86arlia robertiana 243 46.91 243 50.62actylorhiza romana 246 48.78 243 46.09ennaria diphylla 248 43.95 247 42.51ymnadenia conopsea 247 46.56 240 48.33imantoglossum hircinum 244 45.08 241 47.72eotinea maculata 240 45.42 230 44.78phrys tentrediniphera 237 45.15 239 46.03rchis collina 241 40.66 238 46.22. coriophora 241 42.74 237 48.95. italica 249 46.59 240 49.16. lactea 248 41.13 225 44.00. laxiflora 241 41.67 240 46.67. longicornu 243 46.91 236 50.42. mascula 250 47.60 240 49.58. militaris 243 43.21 238 46.64. morio 245 46.94 234 51.28. palustris 246 41.06 240 46.67. papilionacea 226 42.92 226 47.35. patens 244 50.41 239 49.79. pauciflora 249 46.59 240 47.08. provincialis 250 48.00 240 50.00. purpurea 249 42.97 240 47.08. quadripunctata 244 48.36 239 49.37. simia 249 43.37 240 46.67. spitzelii 241 49.38 239 50.20. tridentata 248 42.34 230 44.35. ustulata 247 41.30 229 45.85latanthera chlorantha 250 42.40 240 41.25erapias lingua 240 46.25 237 48.10

n Orchis (except Gennaria, which was used as an w

utgroup). In particular, the ingroup appears dividednto two major clades.

One clade is composed of Anacamptis, Barlia, Himan-oglossum, Ophrys, Orchis collina, O. coriophora, O.axiflora, O. longicornu, O. morio, O. palustris, O.apilionacea, and Serapias. In this clade, Barlia andimantoglossum are a sister group and, together withphrys, are in a sister group relationship with the restf the listed species. This latter group consists oferapias plus a not fully resolved clade in which sisterroup relationships between O. laxiflora and O. palus-ris, between O. longicornu and O. morio, and betweenhe latter two and O. papilionacea can be observed.

The other major clade is composed in turn of twoonophyletic groups. The first group has Dactylorhiza

nd Gymnadenia in a sister group relationship, with aectinate sequence of Platanthera, Neotinea, O. triden-ata, O. ustulata, and O. lactea. The second group has. italica in a basal position to a clade composed ofceras as sister group to a unit in which a trichotomy

ncluding O. militaris, O. purpurea, and O. simia isasal to a poorly resolved monophylum. In this clade,ister group relationships between O. mascula and O.rovincialis, between O. quadripunctata and O. spit-elii, and between the latter two and O. patens can bebserved.Repeated successive weighting (Farris, 1969) on the

ight cladograms yielded two cladograms (3336 steps,.I. 5 0.83, R.I. 5 0.86) with a greater resolution. These

wo topologies are a subset of the eight obtained for theriginal matrix. For this reason, we choose the consen-us tree between them, topologically identical to theladogram in Fig. 1, as a working hypothesis of thehylogenetic relationships in our ingroup.ITS1 and ITS2 data, when analyzed independently

mh*; bb*;), yielded 180 cladograms (437 steps,.I. 5 0.58, R.I. 5 0.70) and 77 cladograms (393 steps,.I. 5 0.64, R.I. 5 0.70), respectively. Neither ITS1 nor

TS2 topologies are fully congruent with those ex-ressed by the total data set. In particular, the consen-us tree for ITS1 shows Serapias as sister group to theest of the ingroup and Ophrys as sister group to O.axiflora and O. palustris. The consensus tree for ITS2hows Barlia and Himantoglossum at the base of theree in a sister group relationship to the rest of thengroup.

Bootstrap analysis on the complete data set (boot-trap values .50% are shown in Fig. 1) generated aajority-rule consensus tree topologically similar to

he cladogram in Fig. 1. The only difference is in theosition of Dactylorhiza and Gymnadenia compared tohe rest of their clade. Despite the fact that the majorityf the clades have bootstrap values above 50%, theelationships among the three major clades into whichhe ingroup is split as well as those between Ophrys,latanthera, and their respective sister groups, are

eakly supported.

btmnrslolI1

craab

tsod

ca

oswtrcsema

P

det

is

71ITS DNA VARIATION IN Orchis

The evaluation of phylogenetic signal as carried outy the PTP test (Faith and Cranston, 1991) showedhat no tree obtained from any of the 500 fictitiousatrices was comparable in length to the most parsimo-ious cladograms mentioned above. The length of theesulting 3342 cladograms ranged from 1331 to 1369teps (mean value 1350.0, SD 6.65), that is 37 to 39%onger than the most parsimonious trees found for theriginal data set, and at least 30% longer than theongest topology obtained in the preliminary analyses.n such cases, PTP equals 0.01 (Faith and Cranston991) and therefore the null hypothesis was rejected.Bremer support (Fig. 1) is high for the majority of the

lades. Notable exceptions are the clade with Dactylo-hiza and Gymnadenia at the base, the clade with A.nthropophorum at the base, the clade immediatelybove the latter, the clade with O. papilionacea at thease, and the clade above O. patens.The consensus among the four cladograms obtained

hrough the elision approach described above (7002teps long) is extremely similar to the consensus treebtained by the original data set (Fig. 1), the only

FIG. 1. Consensus tree between two of the eight equally parsimondentical to the consensus topology obtained after successive weightinupport; bootstrap support above 50% is indicated below each branch.

ifference being the pectinate position of O. papiliona- n

ea, Anacamptis, and O. collina, which is not present inny of our original cladograms.The consensus tree between the four most parsimoni-

us cladograms obtained by coding indels as a fifthtate (1032 steps, C.I. 5 0.63, R.I. 5 0.71) is congruentith that obtained in the original matrix (Fig. 1), with

he only difference being a greater resolution in theelative positions of Anacamptis, O. collina, and O.oriophora (this resolution is present also in the consen-us tree shown in the boxes of Fig. 1). This is furthermpirical proof of the fact that different indel treat-ents affect ITS topologies only minimally (Baldwin et

l., 1995).

DISCUSSION

hylogenetic Relationships

The analyses presented here address various long-ebated issues in the study of European orchid phylog-ny. Perhaps the most relevant result of this study ishe indication that all the genera considered here are

s topologies obtained (839 steps, C.I. 5 0.60, R.I. 5 0.69). This tree isf the original matrix. Numbers above the branches indicate Bremer’sxes: general consensus topology.

ioug oBo

ested in Orchis. Indications of the inclusion of Aceras

(1etlPsptteatfBOsscOOei

dcgtnOoHOARreowbnw

C

htiawanoomV

awlgoaLaortrsw

moc2s3iaCod

72 ACETO ET AL.

Rossi et al., 1994; Caputo et al., 1995; Cozzolino et al.,998) and Dactylorhiza (Caputo et al., 1995; Cozzolinot al., 1998) in Orchis were already available, but inhis case nesting is hypothesized for Anacamptis, Bar-ia, Gymnadenia, Himantoglossum, Neotinea, Ophrys,latanthera, and Serapias. However, the inclusion ofeveral of these genera in Orchis is only weakly sup-orted (i.e., bootstrap percentages are below 50% forhe basal clades of our analysis). In order to evaluatehe strength of our hypothesis more accurately, westimated the increase in step number by forcing thebove-mentioned genera outside Orchis. The results ofhese attempts show that the above mentioned generaall into two groups. The first group is composed ofarlia–Himantoglossum, Dactylorhiza–Gymnadenia,phrys, Platanthera, and Serapias. The separate exclu-

ion of each of them increases the tree length by 6–8teps. The second group is composed of Aceras, Ana-amptis, and Neotinea, which when forced outsiderchis increases tree length by 25–35 steps. Makingrchis strictly monophyletic costs over 50 steps, andxcluding only the genera of the first group causes anncrease of 23 steps.

This substantial increase in the step number mayepend upon how far removed from the ingroup is thehosen outgroup. More closely related possible out-roups were discarded after preliminary results showedhat the selected taxa, all belonging to subtribe Orchidi-ae (namely, Anacamptis, Barlia, Himantoglossum,phrys, Platanthera, and Serapias), were included inur ingroup. We therefore chose a member of subtribeabenariinae Bentham (the only other subtribe of triberchideae according to Dressler, 1993) as an outgroup.lso a previous analysis based on chloroplast DNAFLPs (Cozzolino et al., 1998) demonstrated that chlo-

oplast DNA in our orchids was less conserved thanxpected and thus led us to avoid selecting far removedutgroups (e.g., Cephalanthera L. C. M. Richard, whichas the outgroup in that analysis, or any other taxa notelonging to Orchideae) because ITS sequences areormally less conserved than chloroplast DNA as ahole.

orrelation with Previous Evidence

Morphology. The phylogenetic hypothesis shownere does not match in entirety previous knowledge onhe gross morphology of the taxa involved. Floral traits,n particular the shape of the labellum, the presence ofspur, and the convergence of outer tepals into a galea,hich have been traditionally used to discriminatemong species and related genera within orchids, doot appear to be completely congruent with the patternf relationships depicted by ITS’s (Fig. 2). Therefore,ur proposal of phylogenetic relationships does notatch any previous treatment based on morphology.

ermeulen (1972), in the last taxonomic revision avail- c

ble for Orchis, divided the genus into four sectionshich are at least paraphyletic if not altogether polyphy-

etic according to our data (Table 2 and Fig. 1). Also, theroups recognized by Cauwet-Marc and Balayer (1984)n chromosomal and morphological grounds (Table 2),s well as the segregated genera Vermeulenia Love andove (Love and Love, 1972) and Anteriorchis E. Kleinnd D. Strack (Strack et al., 1989), are not supported byur analysis. In contrast, some small groups alreadyecognized in previous works (e.g., Vermeulen’s subsec-ions) are congruent with our data. Notable in thisespect are subsect. Moriones (Rchb. f.) Parl. andubsect. Laxiflorae Nevskij. Despite the incongruenciesith previous work based on morphology, some of the

FIG. 2. Plotting of the chromosome numbers and of the keyorphological characters traditionally used for diagnostic purposes

nto the cladogram of Fig. 1. Characters are as follows: 1, haploidhromosome number (1.0 5 17; 1.1 5 20; 1.2 5 21; 1.3 5 18; 1.4 5 16);, shape of the tuberoids (2.0 5 entire; 2.1 5 palmately lobed); 3,hape of the labellum (3.0 5 three equal lobes; 3.1 5 entire;.2 5 three lobes, central one with lobules; 3.3 5 five tapering lobes,n the shape of a standing man); 4, spur (4.0 5 present; 4.1 5 absent);nd 5, galea formed by the outer tepals (5.0 5 absent; 5.1 5 present).haracter 3 has been scored following the traditional views ofrchidologists; states 3.2 and 3.3 may be also scored as an indepen-ent binary character. Terminal polymorphic states are not shown.

lades in our analysis show evident morphological

sn

peti1tcPG1DTa2Hm

a11nHntrirg

(

paudpsiia

amfhtfc(

prAscaeahapec

V

I

I

73ITS DNA VARIATION IN Orchis

ynapomorphies. For example, Dactylorhiza and Gym-adenia share palmate tuberoids (Fig. 2).Karyology. Our hypothesis (Fig. 1), albeit not com-

atible with evidence from gross morphology, is, how-ver, broadly congruent with the cytogenetic reports forhe species in this study (Fig. 2). For all taxa includedn the Ophrys and Serapias clades, reports indicate n 58, with the exception of Serapias lingua (n 5 36, one ofhe few tetraploid species of the genus) and O. papiliona-ea (n 5 16). All taxa included in the O. italica andlatanthera clades are n 5 21 and Dactylorhiza andymnadenia are n 5 20 (Cauwet-Marc and Balayer,984; Bianco et al., 1989; Del Prete et al., 1991;’Emerico et al., 1992, 1996; D’Emerico, pers. comm.).he only karyological ambiguities in this frameworkre related to O. patens and O. spitzelii for which n 51, 40 and n 5 20, 21, respectively, have been reported.owever, we chose to report only the apparently plesio-orphic number 2n 5 42 in Fig. 2.For Gennaria, reports indicate n 5 17 for Sardinia

nd n 5 20 for Canary Islands (Dolcher and Dolcher,961; Scrugli, 1978; Sundermann and Von Der Bank,977). If n 5 20 is accepted as the plesiomorphicumber for the species (the most probable case forabenariinae, according to Dressler, 1993) n 5 18 and5 21 would be apomorphic for the involved clades. In

his case, Dactylorhiza and Gymnadenia would haveetained the plesiomorphic feature. On the other hand,f n 5 17 is regarded as the plesiomorphic state,econstruction of chromosome number onto the clado-ram would be equivocal.Previous molecular evidence. Data from isoenzymes

TAB

Synopsis of the Groups Rec

ermeulen’s (1972) sectionsLabellointegrae LabellotrilobataeO. caspia O. laxifloraO. collina O. masculaO. papilionacea O. morio

O. pallensO. palustrisO. patensO. paucifloraO. provincialisO. quadripunctata

nformal groups by Cauwet-Marc and Balayer (1984)O. coriophora O. morio O.O. collina O. papilionacea O.

nformal groups by Rossi et al. (1994)O. morio O. italicaO. papilionacea O. simiaO. coriophora A. anthropophorumO. sancta O. purpurea

Schlegel et al., 1989; Rossi et al., 1994) give only s

artial support to the ITS-based phylogeny. Theseuthors recognize a group composed of O. morio and O.stulata which is not supported by the present evi-ence. The same authors, however, implicitly define theositions of Dactylorhiza and Gymnadenia in a mannerimilar to ours. Also the grouping by Rossi et al. (1994)s not entirely congruent with ours (Table 2). However,t is worth mentioning that the latter authors placed A.nthropophorum within Orchis.It is regrettable that no previous cladistic analysis is

vailable for morphological, karyological, and enzy-atic characters. Consequently this has prevented us

rom carrying out a combined analysis, which couldave provided further insights into the phylogeny ofhese orchids in a total evidence framework. We there-ore resorted to simply plotting the relevant karyologi-al and morphological characters on our consensus treeFig. 2) under a delayed transformation model.

Finally, a chloroplast DNArestriction fragment lengtholymorphism analysis carried out in our laboratory onepresentatives species of Orchis, Aceras, Dactylorhiza,nacamptis, and Serapias (Cozzolino et al., 1998)howed that Orchis is paraphyletic because it alsoontains Aceras and Dactylorhiza. The genus Orchisppeared divided into two major clades which, to thextent of the species common to both studies, arelmost completely congruent with the results shownere. The only relevant difference between the twonalyses was in the positions of Anacamptis and Sera-ias, which were external to Orchis and sister group toach other in the former study. As far as Serapias isoncerned, we interpret this discrepancy as due to

2

nized by Previous Authors

Coriophorae MilitaresO. coriophora O. italicaO. sancta O. lactea

O. militarisO. punctulataO. purpureaO. simiaO. tridentataO. ustulata

ustris O. pallens O. tridentataiflora O. mascula O. ustulata

O. provincialis O. simiaO. militarisO. purpurea

O. laxiflora O. quadripunctataO. palustris

LE

og

pallax

ampling paucity in the chloroplast analysis that in-

vrttct

dmeSPcathotoatccfrmcpEm

essdeccp

tgccflttt

thip

mp

tOpslsrOtconflcphnrvgldl(Nfgl

fmittdftp(otsToc1tuoitDr

74 ACETO ET AL.

olved only 14 taxa. In fact, the position of this genus isather basal even in the present analysis, and forcinghe genus out of Orchis causes only a slight deteriora-ion of our hypothesis in terms of step number. Ana-amptis, in contrast, is highly nested in our ingroup inhe present analysis.

The discordance between the pattern of descentepicted by ITS’s and chloroplast DNA in this genusay be interpreted as a possible chloroplast capture

vent (at least, for the populations of Central andouthern Italy used in the chloroplast DNA study).lastid capture has been demonstrated to be a ratherommon occurrence in some groups of plants (Soltisnd Kuzoff, 1995; Schilling and Panero, 1996), and inhis case the event may be hypothesized as followingybridization either with a member of the stem lineagef Serapias or with some extant relative of Serapias notested in that analysis. Besides plastid capture, somether kind of event can be hypothesized as a factorcting in the production of the floral organizationypical of Anacamptis. A natural hybrid between O.ollina (the sister group of Anacamptis according to theladogram in Fig. 1) and a species of Serapias, recentlyound in Apulia (Italy) (Bianco et al., 1990), does notesemble Anacamptis. This monotypic genus, whoseorphology is strikingly dissimilar from that of its

losest relatives (it has a very long and thin spur), isollinated mainly by Zygaenidae moths (Nazarov andfetov, 1993) and this condition does not occur in anyember of its clade.Unfortunately, the chloroplast DNA data (Cozzolino

t al., 1998) are representative of less than 50% of thepecies employed in the present study, do not includeeveral of the genera studied here, and are rooted by aifferent outgroup. This prevented us from properlyxploring incongruence issues a posteriori and becausehloroplast DNA is strictly matrilinear in orchids aombined analysis would perhaps not have been appro-riate.

CONCLUDING REMARKS

Heterogeneity in pollination syndromes or in pollina-ion-related floral characters is widespread in the studyroup (e.g., Aceras, Neotinea, Ophrys, and Platantheraompared to their close relatives). Orchidaceae areonsidered a paramount example of evolution throughoral diversification which includes changes of pollina-or, variations in flower color, and pollinia specializa-ion representing only some of the mechanisms poten-ially involved in orchid speciation.

Our results indicate that while some floral charac-ers closely correspond to the molecular data, most areighly homoplastic and thus unsuitable for reconstruct-

ng phylogeny. Various traits pertaining to floral mor-

hology may be interpreted as a result of pollinator- t

ediated selection and thus have more ecological thanhylogenetic implications.As already shown on molecular grounds for Platan-

hera pro parte (Haperman et al., 1996a,b) and forncidium and related taxa (Chase and Palmer, 1992),ollinator-mediated selection may undermine the under-tanding of phylogenetic relationships through morpho-ogical analysis of floral characters, because of wide-pread convergence. However, floral traits may beelevant in understanding isolation mechanisms withinrchidaceae. An issue which emerges in this study is

hat isolations appear to have occurred mainly inonnection with pollinator specificity. The captivationf new pollinators (or the onset of elaborate mecha-isms of deception) may have resulted in divergentoral morphologies that quickly segregated new spe-ies and species groups. We are therefore inferring thatollinator-mediated morphological selection was some-ow punctuated and in some instances abruptly so. Aumber of markedly distinct lineages (i.e., Dactylo-hiza, Ophrys, Platanthera, and Serapias types) di-erged more or less simultaneously from an ancestralroup and little evidence (i.e., few apomorphies) waseft in the ribosomal DNA of these plants from this firstynamic but short evolutionary period. Successive simi-ar events led to the divergence of more derived groupsi.e., Aceras, Anacamptis, Barlia, Himantoglossum, andeotinea). It is interesting to note that taxa in the

ormer group (i.e., those placed basally in the clado-ram of Fig. 1) are multispecific, whereas those in theatter group tend to be pauci- or monotypic.

In addition to pollinator-mediated processes, anotherrequently suggested mechanism of production of newodels of floral organization, i.e., interspecific hybrid-

zation, albeit commonly observed, does not seem to behe key to the origin of new bodyplans in the taxa inhis study (with the possible exception of Anacamptis,iscussed above). In fact, interspecific hybrids areound regularly but rarely. Most of the hybridization inhe genus Orchis is usually restricted to a few hybridlants, and hybrid swarms have rarely been observedDafni and Ivri, 1981; Dafni and Baumann, 1982; fieldbservations). Moreover, among all the sequences ob-ained by us, we only occasionally found paralogousites, very few of which were of apparent hybrid origin.his is in contrast with similar findings in other groupsf angiosperms, for which paralogy seems to be aonsequence of past hybridization (e.g., Sang et al.,996). We are not, of course, suggesting that hybridiza-ion or the other kinds of mechanisms usually calledpon as involved in speciation do not play any role inrchid evolution. For example, Hedren (1996) clearlyndicates that allopolyploidy, and therefore hybridiza-ion, is a key factor in the genetic differentiation withinactylorhiza. We are only implying that they are not

esponsible for the production of new floral organiza-

ions.

oriwp

otGwltit1omqmlrsaom

p

aosJ(Il

B

B

B

B

B

B

B

C

C

C

C

C

C

D

D

D

D

D

D

D

D

E

F

F

F

F

F

75ITS DNA VARIATION IN Orchis

In conclusion, we think that the paraphyletic statusf genus Orchis has been demonstrated beyond anyeasonable doubt. Orchis may be regarded as represent-ng a very successful, probably plesiomorphic body planhich has been rather frequently modified by selectiveressures.Classical grouping of species within the genus has

ften been based on convergence or parallelism in floralraits, which has been misconstrued as synapomorphy.iven the specialization in reproductive morphologyithin the family, the tendency to emphasize morpho-

ogical discrepancy has possibly been the main cause ofhe taxonomic inflation besetting the group. The find-ngs reported here, merged with previous investiga-ions on different orchid groups (Chase and Palmer,992; Haperman et al., 1996a,b), indicate that conspicu-us morphological changes may be a consequence ofinimal DNA rearrangements. Most probably, se-

uences in which minimal modifications cause drasticorphological changes are regulatory ones. In particu-

ar, mutations in floral homeotic genes have beenepeatedly demonstrated to generate major floral re-tructuring (Weigel and Meyerovitz, 1994; Theissennd Saedler, 1995) and may well be involved in therigin of the strongly divergent floral shapes present inany European orchids.

ACKNOWLEDGMENT

The authors acknowledge a MURST 1997 grant by which theresent research was partly funded.Note added in proof: The authors would like to communicate that,

fter sending the manuscript, they became aware of the publicationf the following paper, whose authors independently reached resultsimilar to theirs: Pridgeon A. M., Bateman R. M., Cox A. V., Hapeman. R., and Chase M. W., 1997. Phylogenetics of subtribe OrchidineOrchidoideae, Orchidaceae) based on nuclear ITS sequences. 1.ntergeneric relationships and polyphyly of Orchis sensu lato. Lind-eyana 12: 89–109.

REFERENCES

aldwin, B. G., Sanderson, M. J., Porter, M. J., Wojciechowski, M. F.,Campbell, C. S., and Donoghue, M. J. (1995). The ITS region ofnuclear ribosomal DNA: A valuable source of evidence in angio-sperm phylogeny. Ann. Missouri Bot. Gard. 82: 247–277.

ayer, R. J., Soltis, D. E., and Soltis, P. S. (1996). Phylogeneticinferences in Antennaria (Asteraceae: Gnaphalieae: Cassiniinae)based on sequences from nuclear ribosomal DNA internal tran-scribed spacer (ITS). Am. J. Bot. 83: 516–527.

ianco, P., D’Emerico, S., Medagli, P., and Ruggiero, L. (1990).XOrchiserapias nelsoniana Bianco, D’Emerico, Medagli et Ruggi-ero, hybr. nat. nov. della Puglia. Webbia 44: 315–322.

ianco, P., D’emerico, S., Medagli, P., and Ruggiero, L. (1989).Karyological studies of some taxa of the genus Ophrys (Orchi-daceae) from Apulia (Italy). Caryologia 42: 57–63.

ogler, D. J., and Simpson, B. B. (1996). Phylogeny of Agavaceaebased on ITS rDNA sequence variation. Am. J. Bot. 83: 1225–1235.

remer, K. (1988). The limits of amino acid sequence data in

angiosperm phylogenetic reconstruction. Evolution 42: 795–803. G

remer, K. (1994). Branch support and tree stability. Cladistics 10:369–372.

auwet-Marc, A. M., and Balayer, M. (1984). Les genres Orchis L.,Dactylorhiza Necker ex Newski, Neotinea Reichb. et Traunsteri-nera Reichb.: Caryologie et proposition de phylogenie et d’evolu-tion. Bot. Helvetica 94: 391–406.aputo, P., Stevenson, D. W., and Wurtzel, E. T. (1991). A phyloge-netic analysis of American Cycads (Cycadales) using chloroplastDNA restriction fragment polymorphisms. Brittonia 43: 135–145.aputo, P., Aceto, S., Cozzolino, S., Nazzaro, R., and Gaudio, L.(1995). Relazioni tra Orchis L. (Orchidaceae) ed alcuni generiaffini. Giorn. Bot. Ital. 129: 116.

hase, M. W., and Palmer, J. D. (1992). Floral morphology andchromosome number in subtribe Oncidiinae (Orchidaceae): Evolu-tionary insights from a phylogenetic analysis of chloroplast DNArestriction site variation. In ‘‘Molecular Systematics of Plants’’(P. S. Soltis, D. E. Soltis, and J. J. Doyle, Eds.), pp. 324–339.Chapman & Hall, New York.ozzolino, S., Aceto, S., Caputo, P., Nazzaro, R., and Gaudio, L.(1998). Phylogenetic relationships in Orchis and some relatedgenera: An approach using chloroplast DNA. Nord. J. Bot. 18:79–87.

ozzolino, S., Caputo, P., Aceto, S., Rossi, W., and De Luca, P. (1996).Testing the usefulness of ITS1 sequence as a tool to infer relation-ships in Orchis L. Delpinoa 33–34: 77–85.afni, A., and Ivri, Y. (1981). Floral mimicry between Orchis isra-elitica Baumann and Dafni (Orchidaceae) and Bellevalia flexuosaBoiss. (Liliaceae). Oecologia 49: 229–232.afni, A., and Baumann, H. (1982). Biometrical analysis on popula-tions of Orchis israelitica, O. caspia, and their hybrids (Orchi-daceae). Pl. Syst. Evol. 140: 87–94.el Prete, C., Mazzola, P., and Miceli, P. (1991). Karyological differen-tiation and speciation in C. Mediterranean Anacamptis (Orchi-daceae). Pl. Syst. Evol. 174: 115–123.’Emerico, S., Bianco, P., and Medagli, P. (1992). Karyological studieson Orchidaceae. Tribe Ophrydeae, subtribe Serapiadinae. Caryolo-gia 45: 301–311.’Emerico, S., Pignone, D., and Bianco, P. (1996). Karyomorphologi-cal analyses and heterochromatin characteristics disclose phyleticrelationships among 2n 5 32 and 2n 5 36 species of Orchis(Orchidaceae). Pl. Syst. Evol. 200: 111–124.olcher, E., and Dolcher, T. (1961). Sulla cariologia di Gennariadiphylla (Link) Parl. (Coeloglossum diphyllum Fiori). Gion. Bot.Ital. 68: 59–63.ressler, R. L. (1981). ‘‘The Orchids: Natural History and Classifica-tion,’’ Harvard Univ. Press. Cambridge, MA.ressler, R. L. (1993). ‘‘Phylogeny and Classification of the OrchidFamily,’’ Cambridge Univ. Press, Cambridge, UK.

hrendorfer, F. (1980). Hybridisierung, Polyploidie und Evolution beieuropaischmediterranen Orchideen. Jahresber. Naturwiss. Ver. Wup-pertal 33: 15–34.

aith, D. P., and Cranston, P. S. (1991). Could a cladogram this shorthave arisen by chance alone?: On permutation tests for cladisticstructure. Cladistics 7: 1–28.

arris, J. S. (1969). A successive approximation approach to charac-ter weighting. Syst. Zool. 18: 374–385.

arris, J. S. (1988). Hennig86 ver.1.5 reference manual. Published bythe author. Port Jefferson Station, NY.

elsenstein, J. (1985). Confidence limits on phylogenies: An approachusing the bootstrap. Evolution 39: 783–791.

elsenstein, J. (1993). PHYLIP (Phylogeny Inference Package) ver-sion 3.5c. Distributed by the author. Department of Genetics, Univ.of Washington, Seattle.

atesy, J., DeSalle, R., and Wheeler, W. (1993). Alignment-ambigu-

G

G

H

H

H

L

K

K

N

N

O

R

S

S

S

S

S

S

S

S

S

S

S

T

T

V

W

W

Y

76 ACETO ET AL.

ous nucleotide sites and the exclusion of systematic data. Mol.Phylogenet. Evol. 2: 152–157.oloboff, P. A. (1991). Homoplasy and the choice among cladograms.Cladistics 7: 215–232.oloboff, P. A. (1993). Nona. Instruction manual. Distributed by theauthor.aperman, J. R., Inoue, K., and Thomas, J. G. (1996a). Phyhyloge-netic relationships, biogeography and speciation in Platanthera(Orchidaceae): Inferences from morphological and nuclear ribo-somal DNA sequence data. Am. J. Bot. 83: 161.aperman, I. R., Thomas, J. G., and Inoue, K. (1996b). Floralevolution, pollination ecology and speciation in Platanthera (Orchi-daceae): Inferences from nuclear ribosomal DNA sequence dataand implications for the analysis of floral evolution in the Orchi-daceae. Am. J. Bot. 83: 161.edren, M. (1996). Genetic differentiation, polyploidization andhybridization in northern European Dactylorhiza (Orchidaceae):Evidence from allozyme markers. Pl. Syst. Evol. 201: 31–55.

ove, A., and Love, D. (1972). Vermeulenia—A new genus of orchids.Acta Bot. Neerl. 21: 553–554.allersjo, M., Farris, J. S., Kluge, A. G., and Bult, C. (1992).Skewness and permutation. Cladistics 8: 275–287.im, Y. D., and Jansen, R. K. (1994). Comparison of phylogenetichypotheses among different data sets in dwarf dandelions (Krigia,Asteraceae): Additional information from internal transcribedspacer sequences of nuclear ribosomal DNA. Pl. Syst. Evol. 190:157–185.azarov, W., and Efetov, K. A. (1993). On the role of Zygaenidae(Lepidoptera) in pollination of Anacamptis pyramidalis (Orchi-daceae). Zool. Z. 72: 54–67.ixon, K. C. (1992). Clados version 1.2. Published by the author.Trumansberg, NY.xelman, B., and Liden, M. (1995). Genetic boundaries in the tribeSileneae (Caryophyllaceae) as inferred from nuclear rDNA se-quences. Taxon 44: 525–542.ossi, W., Corrias, B., Arduino, P., Cianchi, R., and Bullini, L. (1994).Multilocus electrophoresis and European orchid systematics: Thegenus Orchis and related genera. Proc. 14th World Orchid Conf.Glasgow. HMSO: 78–83.

ang, T., Crawford, D. J., and Stuessy, T. F. (1995). Documentation ofreticulate evolution in peonies (Paeonia) using internal transcribedspacer sequences of nuclear ribosomal DNA: Implication for bioge-ography and concerted evolution. Proc. Natl. Acad. Sci. USA 92:6813–6817.

anger, F. S., Nicklen, S., and Couson, A. R. (1977). DNA sequencing

with chain terminating inhibitors. Proc. Natl. Acad. Sci. USA 74:5463–5467.

chilling, E. E., and Panero, J. L. (1996). Phylogenetic reticulation insubtribe Helianthinae. Am. J. Bot. 83: 939–948.

chlegel, N., Steinbruk, G., Hanh, K., and Rottger, B. (1989).Interspecific relationship of ten European orchid species as re-vealed by enzyme electrophoresis. Pl. Syst. Evol. 163: 107–119.

crugli, A. (1978). Numeri cromosomici per la Flora Italiana. Inf. Bot.Ital. 10: 414–421.

oltis, D. E., and Kuzoff, R. K. (1995). Discordance between nuclearand chloroplast phylogenies in the Heuchera group. Evolution 49:727–742.

track, D., Busch, E., and Klein, E. (1989). Anthocyanin patterns inEuropean orchids and their taxonomic and phylogenetic relevance.Phytochemistry 28: 2127–2139.

uh, Y., Thien, H. E., and Zimmer, E. A. (1993). Molecular evolutionand phylogenetic implications of internal transcribed sequences ofribosomal DNA in Winteraceae. Am. J. Bot. 80: 1042–1055.

undermann, H., and Von der Bank, C. (1977). Neue Beitrage zurZytotaxonomie der Erdorchideen II (Ophris, Orchis, Neotinea,Steveniella und Gennaria). Orchideen 28: 167–175.

usanna, A., Garcia Jacas, N., Soltis, E. E., and Soltis, P. S. (1995).Phylogenetic relationships in tribe Cardueae (Asteraceae) based onITS sequences. Am. J. Bot. 82: 1056–1068.

wofford, D. L., and Olsen, G. J. (1990). Phylogeny reconstruction. In‘‘Molecular Systematics’’ (D. M. Hillis and C. Moritz, Eds.), pp.411–501. Sinauer, Sunderland, MA.

heissen, G., and Saedler, H. (1995). MADS-box genes in plantontogeny and phylogeny: Haeckel’s biogenetic law revisited. Curr.Opin. Genet. Dev. 5: 628–639.

hompson, J. D., Higgins, D. G., and Gibson, T. J. (1994). CLUSTALW: Improving the sensitivity of progressive multiple sequencealignment through sequence weighting, positions-specific gap pen-alties and weight matrix choice. Nucleic Acids Res. 22: 4673–4680.

ermeulen, P. (1972). Ubersicht zur Systematik und Taxonomie derGattung Orchis s. str. Jahresber. Naturwiss. Ver. Wuppertal 25:22–36.eigel, D., and Meyerowitz, E. M. (1994). The ABCs of floral homeoticgenes. Cell 78: 203–209.heeler, W. C., Gatesy, J., and DeSalle, R. (1995). Elision: A methodfor accommodating multiple molecular sequence alignments withalignment-ambiguous sites. Mol. Phylogenet. Evol. 4: 1–9.

uan, Y.-M., Kupfer, P., and Doyle, J. J. (1996). Infrageneric phylog-eny of the Gentiana (Gentianaceae) inferred from nucleotidesequences of the internal transcribed spacers (ITS) of nuclear

ribosomal DNA. Am. J. Bot. 83: 641–652.