Plant Syst. Evol. 246: 163–175 (2004) DOI 10.1007/s00606-004-0154-y Molecular phylogenetics of Dipsacaceae reveals parallel trends in seed dispersal syndromes P. Caputo, S. Cozzolino, and A. Moretti Dipartimento di Biologia vegetale, Università degli studi di Napoli Federico II, Napoli, Italy Received January 22, 2002; accepted February 22, 2004 Published online: July 13, 2004 Ó Springer-Verlag 2004 Abstract. Phylogenetic relationships among 17 taxa of Dipsacaceae were inferred from nucleotide sequence variation in both the internal transcribed spacer (ITS) regions of nuclear ribo(UAA) somal DNA and the chloroplast trnL intron sequences. The combined phylogenetic analysis, carried out by using two taxa from Valerianaceae as an outgroup yielded a single most parsimonious tree, in which Dipsacaceae are divided into two major clades: one including Lomelosia and Pycnocomon, both in a sister group relationship with a clade containing Pterocephalus, Scabiosa and Sixalix; the other including Pseudoscabiosa, Succisa and Succisella is sister group to Knautia, Pterocephalidium, Dipsacus and Cephalaria. The results obtained here greatly differ from previous ones based on classical morphology, but are congruent with recent findings on epicalyx differentiation and with pollen characters. In particular, our results would confirm on molecular grounds the recently restricted circumscription for Scabioseae proposed by other authors. Our phylogenetic hypothesis indicates that adaptations to seed dispersal have been a very strong driving force in Dipsacaceae evolution, with similar selective pressures causing the onset of similar epicalyx shapes and dispersal modes in a parallel fashion in various taxa. For this reason, the gross morphology of the involucel is deceptive in inferring relationships. Key words: Dipsacaceae, Scabioseae, phylogeny, seed dispersal, ITS, trnL. Dipsacaceae Juss. (Dipsacales Lindley) is a dicotyledonous family including 12–13 genera and 250–350 species, depending on circumscription, distributed in Eurasia and Africa, with the great majority of the taxa mainly centred around the Mediterranean basin. The family constitutes a monophyletic group (Ehrendorfer 1964a; Verláque 1977a, 1984a,b; Caputo and Cozzolino 1994), whose characteristics are represented by an epicalyx, which encases the fruit and whose shape, symmetry and ornamentations are taxonomically relevant, and by an involucrate head. The family is a derived member of the order Dipsacales, where it is sister group to Valerianaceae (Backlund and Bremer 1997). Delimitation of taxa within the family has always been subject to controversy; consequently, circumscription of genera and tribes has repeatedly changed over time, because of the overall morphological similarity among the taxa in the family and of their diversity in structural detail. In particular, the classical concepts of Scabiosa L. s.l. and Pterocephalus (Vaill.) Adans. s.l. included species which are 164 presently attributed to eight separate genera, most of which with entirely independent phylogenetic histories. In fact, genus Scabiosa was traditionally divided into five sections. After the studies by Verláque (1984a, 1985a, 1986a, b) and Devesa (1984a, b), which demonstrated independent evolutionary histories for its sections, Greuter and Raus (1985) and Greuter et al. (1986) designated Lomelosia Rafin. for the species traditionally belonging to sect. Trochocephalus Mertens et Koch, Scabiosa s.str. for the species belonging to sect. Sclerostemma Mertens et Koch, Sixalix Rafin. for those formerly belonging to sect. Cyrtostemma Mertens et Koch, Pseudoscabiosa Devesa for several archaic species formerly isolated in sect. Asterothrix Font Quer; in the same period the name Pycnocomon Hoffmanns. et Link was resumed in order to refer to the atypical species Scabiosa rutifolia Vahl, formerly placed in Scabiosa sect. Pycnocomon. Similarly, the genus Pterocephalus (Vaill.) Adans., has included in the past various atypical species which have been recently segregated in Pycnocomon [Pycnocomon intermedium (Lag.) Greuter et Burdet ¼ Pterocephalus intermedius (Lag.) Coutinho], Pterocephalidium G. López [Pterocephalidium diandrum (Lag.) G. López ¼ Pseudoscabiosa diandra Devesa ¼ Pterocephalus diandrus (Lag.) Lag.] and Pterocephalodes V. Mayer et Ehrend. (name adopted for three South-Eastern Asian taxa). As in the case of some of the taxa formerly lumped in a loose concept of Scabiosa, the taxa segregated from Pterocephalus are not close to the former genus from a phylogenetic point of view. Correspondingly, the hypotheses of tribal relationships within the family have been greatly modified during time. The main differences among the various phylogenetic hypotheses produced in the recent past (Ehrendorfer 1964a,b, 1965; Verláque 1977a,b, 1984a,b, 1985a,b, 1986a,b; Caputo and Cozzolino 1994, 1995), consisted in the positions of the archaic genera Succisa Necker, Succisella Beck P. Caputo et al.: Molecular phylogeny of Dipsacaceae and Knautia (L.) Coult. as compared to the rest of the family and in the relative positions of the above mentioned segregates of Scabiosa s.l. However, all the authors listed above concurred towards the idea that tribe Scabioseae, as traditionally circumscribed (i.e. Lomelosia, Pseudoscabiosa, Pterocephalidium, Pterocephalus, Pycnocomon, Scabiosa, Sixalix, Succisa, and Succisella) was monophyletic. The recent papers by Mayer and Ehrendorfer (1999, 2000) showed that a monophyletic circumscription of Scabioseae should include less taxa than previously suggested. The latter authors also indicated that the homology of parts of the epicalyx, with special reference to a differentiation of its distal region (the corona, which is a membranous expansion in various taxa), has been often misinterpreted, as a consequence of widespread modifications in response to adaptative pressures to fruit dispersal which have fostered rampant homoplasy in both the corona and the calyx. Therefore, Mayer and Ehrendorfer (1999) suggested that a monophyletic circumscription of Scabioseae should include only Lomelosia, Pterocephalus s.str., Pycnocomon, Scabiosa s.str. (Scabiosa sect. Scabiosa in their study), and Sixalix (Scabiosa sect. Cyrtostemma in their study). This more restricted circumscription of Scabioseae would exclude Pseudoscabiosa, Pterocephalidium, Succisa, Succisella, as well as Pterocephalodes. This paper aims at a molecular verification of the pattern of descent among dipsacaceous genera, in the light of the aforementioned uncertainties, also in order to test to which extent epicalyx and calyx modifications (which are functional towards seed dispersal) developed homoplasiously. To this purpose, Internal Transcribed Spacer (ITS) regions of nuclear (UAA) inribosomal DNA and chloroplast trnL trons were sequenced for representative taxa within the family. These sequences were chosen as they are regarded to have a mutation rate which makes them suitable for studies of relationships below the family level (Baldwin et al. 1995, Gielly and Taberlet 1996). P. Caputo et al.: Molecular phylogeny of Dipsacaceae Material and methods The taxa employed in this study are reported in Table 1. The selection includes representatives from all recognized genera of Dipsacaceae with the exception of Pterocephalodes, Scabiosiopsis Rech. f., and Tremastelma Rafin. (the latter two clearly being, however, derived members of genus Lomelosia). All specimens were either field collected by the authors or planted from seeds and cultivated at the Botanical Garden of Naples, Italy. Leaves were collected at flowering time. Voucher specimens of the examined plants are deposited at NAP. Total DNA was extracted following either the procedure described in Caputo et al. (1991) or that of Doyle and Doyle (1990). ITS 1 and 2 were amplified by using the primers described in Aceto et al. (1999). The (UAA) chloroplast trnL intron was amplified using the two primers reported by Taberlet et al. (1991). PCR conditions were as described in Aceto et al. (1999). PCR amplification products were purified by using Microcon 100 microconcentrators (Amicon, Danvers, MA, U.S.A.) and double-strand sequenced in both directions by using a modification of the Sanger dideoxy method (Sanger et al. 1977) as implemented in a double strand DNA cycle sequencing system with fluorescent dyes. Sequence reactions were then loaded into a 373A Applied Biosystems Automated DNA sequencer (Applied Biosystems, Foster City, CA, U.S.). Sequences were then reduced to the appropriate length by aligning them with various sequences available in the literature. Six fictitious unknowns (N’s) were added at the 3’ terminus of ITS1 in all taxa to prevent terminal misalignments. Alignments were carried out by using Clustal W ver. 1.81 (Thompson et al. 1994) with default settings, except for the parameter TRANSWEIGHT (which controls the transition/transversion ratio), which was set to one and MAXDIV (which controls the delay of the alignment for the most divergent sequences), which was set to 80%. Aligned sequences were then visually inspected to correct gap distributions devoid of biological meaning, aiming to reduce the number of gaps. The complete alignment used for all further analyses is available upon request to the senior author. Cladistic analyses were carried out on the ITS and trnL intron data separately, as well as on a combined matrix. Three different sets of analyses were conducted in relation to indel treatment: with 165 indels included in the analyses (scoring gaps as missing data, dataset A), with equivocal indels (i.e. indels susceptible of different reconstructions) excluded (dataset B), and with unequivocal indels given unit weight (nested indels were scored as different states of a multistate character, dataset C). In order to verify to which extent ambiguous gap positions may influence topology, the elision approach, described by Wheeler et al. (1995), was used. This technique consists in ‘‘eliding’’ various individual alignments into a single combined alignment on which a phylogenetic analysis is carried out. To this aim, six alignments were carried out for ITS data and for trnL intron data separately. These alignments were carried out by using Clustal W ver. 1.81 with the same parameters as above, but with variable gap opening and extension costs at wide intervals in each alignment (PWGAPOPEN= GAPOPEN from 10 to 20 and PWGAPEXT= GAPEXT from 4 to 8). The six alignments obtained plus the original alignment were combined, for each dataset, in a single matrix. These elided ITS and trnL intron matrices were analyzed both separately and in combination, in this case always scoring gaps as missing data. All the manipulations of the matrices, as well as the cladistic analyses, were carried out by using using the cladistic software environment Winclada (Nixon 1999), running Nona (Goloboff 1999) as a daughter process, with the following parameters: hold 100000; hold/100; mult*100; max. The resulting cladograms were investigated with Winclada, which was also used to evaluate congruence between the ITS and trnL intron matrices according to the ILD test of Farris et al. (1994), to calculate bootstrap percentages (out of 1000 replicas) and branch support (up to trees 10 steps longer), as well as to plot morphological characters onto the tree shown in Fig. 1. Comparison of sequence substitution rates between clades was carried out by using the method of Robinson et al. (1998), as implemented in the RRTree software (Robinson-Rechavi and Huchon 2000). The following tests were carried out separately for the ITS and trnL intron matrix: the two major clades resulting from the previous analysis, the four major clades, all the genera, four groups of taxa homogeneous in terms of seed dispersal strategies. Significance threshold was calculated as 0.05/numbers of groups in each test. 166 P. Caputo et al.: Molecular phylogeny of Dipsacaceae Table 1. Species, length (bp) and Genbank accession nos. of the sequences used in this study. In the columns labelled ‘‘Genbank accession no.’’, the three values correspond to ITS 1, ITS2 and trnL intron, respectively Species ITS1 length (bp) ITS2 length (bp) trnL intron length (bp) Genbank accession nos. Cephalaria leucantha (L.) Roem. & Schult. 249 248 513 Cephalaria syriaca (L.) Roem. & Schult. 235 213 512 Dipsacus sylvestris Huds. 223 213 513 Knautia arvensis (L.) Coult. 290 224 526 Lomelosia argentea (L.) Greuter et Burdet 222 228 516 Lomelosia caucasica (MB.) Greuter et Burdet 222 227 519 Pseudoscabiosa limonifolia (Vahl) Devesa 238 230 510 Pterocephalidium diandrum (Lag.) G. López 240 229 523 Pterocephalus perennis Coult. 223 227 547 Pycnocomon rutifolium (Vahl) Hoffmanns. & Link 227 233 537 Scabiosa africana L. 222 226 537 Scabiosa japonica Miq. 225 227 537 Scabiosa uniseta Savi 223 225 537 Sixalix atropurpurea (L.) Greuter et Burdet subsp. maritima (L.) Greuter et Burdet Sixalix farinosa (Cosson) Greuter et Burdet 224 226 541 223 229 541 AJ426523 AJ426524 AJ427376 AJ426525 AJ426526 AJ427377 AJ426527 AJ426528 AJ427378 AJ426529 AJ426530 AJ427379 AJ426531 AJ426532 AJ427380 AJ426533 AJ426534 AJ427381 AJ426535 AJ426536 AJ427383 AJ426537 AJ426538 AJ427382 AJ426539 AJ426540 AJ427384 AJ426541 AJ426542 AJ427385 AJ426543 AJ426544 AJ427386 AJ426545 AJ426546 AJ427387 AJ426547 AJ426548 AJ427388 AJ426549 AJ426550 AJ427389 AJ426551 P. Caputo et al.: Molecular phylogeny of Dipsacaceae 167 Table 1 (continued) Species ITS1 length (bp) ITS2 length (bp) Succisa pratensis Moench 243 226 512 Succisella inflexa (Kluk) G. Beck 239 220 513 Patrinia intermedia L. 220 224 514 Valeriana officinalis L. 226 237 497 Results The lengths of the ITS regions and of the trnL introns for all taxa in study, as well as their Genbank accession numbers, are reported in Table 1. ITS1 length ranged from 222 to 290 bp; ITS2 length ranged from 213 to 247 bp; trnL intron length ranged from 497 to 541 bp. The combined matrix including all indels (consensus length 1154 bp, dataset A) produced one most parsimonious cladogram (length ¼ 775, CI ¼ 0.69, RI ¼ 0.60; by excluding uninformative characters, length ¼ 551, CI ¼ 0.57, RI ¼ 0.60). The combined matrix obtained excluding equivocal indels (consensus length 1042 bp, dataset B) produced 2 equally parsimonious cladograms (length ¼ 609, CI ¼ 0.69, RI ¼ 0.62; by excluding uninformative characters, length ¼ 442, CI ¼ 0.57, RI ¼ 0.62). The combined matrix obtained excluding equivocal indels and scoring unequivocal indels as multistate characters (consensus length 1042 bp, dataset C) produced 2 equally parsimonious cladograms (length ¼ 656, CI ¼ 0.70, RI ¼ 0.62; by excluding uninformative characters, length ¼ 470, CI ¼ 0.59, RI ¼ 0.62). 59 (2) 7 (>10) 33 Genbank accession nos. trnL intron length (bp) AJ426552 AJ427390 AJ426553 AJ426554 AJ427391 AJ426555 AJ426556 AJ427392 AJ426557 AJ426558 AJ427393 AJ426559 AJ426560 AJ427394 Valeriana officinalis Patrinia intermedia Pseudoscabiosa limonifolia 95 (10) 22 Succisella inflexa 84 (4) 12 29 7 Succisa pratensis 8 (1) Knautia arvensis 49 (1) Pterocephalidium diandrum 5 23 5 Dipsacus sylvestris 88 (6) 14 55 (1) 7 14 Cephalaria leucantha 6 C. syriaca 12 84 (5) 9 37 95(5) 6 92 (6) 8 82 (3) 5 21 Pycnocomon rutifolium Lomelosia argentea 8 L. caucasica 8 Pterocephalus perennis Sixalix atropurpurea 7 Sixalix farinosa 8 Scabiosa japonica 61 (2) 14 Scabiosa africana 95 (7) 5 3 8 Scabiosa uniseta 100 (>10) 22 (3) 77 5 5 Fig. 1. Single maximum parsimony cladogram obtained for the combined data set including all indels (dataset A). Length ¼ 775, CI ¼ 0.69, RI ¼ 0.60 (by excluding uninformative characters, length ¼ 551, CI ¼ 0.57, RI ¼ 0.60). Numbers below branches represent synapomorphies; numbers above branches indicate bootstrap percentages above 50% (out of 1000 replicas, rounded to the unit); numbers in parentheses above branches indicate branch support (up to 10 steps longer). Support data for the outgroup are not shown 168 P. Caputo et al.: Molecular phylogeny of Dipsacaceae Table 2. Consensus length (bp), number of most parsimonious trees, with length, CI’s and RI’s for the data sets used in this paper. The second row in the datasets contains the values after removal of uninformative characters. Dataset A = alignment after visual inspection, indels interpreted as missing values; dataset B = all equivocal indels excluded, unequivocal indels interpreted as missing values; dataset C = same as B, unequivocal indels given unit weight DATA SET A CL n. L (bp) MP trees DATASET B Cl RI CL (bp) DATASET C n. L MP trees CI RI CL n. L (bp) MP trees CI RI ITS 562 21 157 671 0.66 0.55 503 0.55 0.55 472 3 126 512 0.65 393 0.55 0.57 0.57 472 3 131 547 0.67 0.58 417 0.57 0.58 TrnL 592 30 2 103 0.90 0.87 47 0.78 0.87 570 2 29 95 0.89 47 0.78 0.87 0.87 571 2 34 107 0.90 0.88 51 0.80 0.88 1154 187 1 775 0.69 0.60 1042 2 551 0.57 0.60 156 609 0.69 442 0.57 0.62 1042 2 0.62 165 656 0.70 0.62 470 0.59 0.62 Combined All the results of the ITS, trnL intron, and combined cladistic analyses are reported in Table 2. In all three datasets, the test for incongruence (Farris et al. 1994), run for 20 replicas, showed that the ITS and trnL intron matrices were not significantly incongruent. The combined analyses for the three datasets provided cladograms entirely congruent with each other, and the single most parsimonious cladogram obtained for dataset A (Fig. 1) is one of the two equally parsimonious cladograms obtained for dataset B and dataset C. The elision approach generated a 3919 character matrix (of which 1182 informative) for ITS data, and a 4144 character matrix (of which 246 informative) for the trnL intron data. The combined elided dataset (8063 characters, of which 1428 informative) yielded a single cladogram whose topology is very similar to those obtained from the combined datasets A, B, C. The single most parsimonious cladogram obtained for dataset A (Fig. 1) shows that Dipsacaceae are divided into two major clades. The first clade includes Pseudoscabiosa, Succisa and Succisella, in a sister group relationship with Knautia, Pterocephalidium, Dipsacus and Cephalaria; in this latter clade, Knautia and Pterocephalidium behave as sister groups, and Dipsacus is sister to Cephalaria. The other major clade includes Lomelosia in a sister group relationship with Pycnocomon; this clade is in turn sister to a clade containing Pterocephalus, sister group to Scabiosa and Sixalix. Bootstrap support out of 1000 replicas (in Fig. 1 only the percentages > 50% are shown) is high for the majority of the clades; the only notable exceptions are the clade including Knautia and Pterocephalidium and the one immediately below (<50%). Branch support has a distribution which is, not surprisingly, similar to bootstrap values, and the weakest areas of the cladogram are the same ones characterized by low bootstrap percentages (Fig. 1). Comparations of sequence substitution rates (Robinson et al. 1998, Robinson-Rechavi and Huchon 2000) in the ITS and trnL intron datasets showed that no one of the tested lineage evolved significantly faster than any other. The trees from the different analyses of the combined datasets (Table 2) are very similar in their basic structure. In fact, the extra tree obtained for datasets B and C has the relative positions of the Pseudoscabiosa clade (clades will be indicated by using the name of their topmost taxon as shown in Fig. 1) and of the Knautia clade reversed (data not shown); more- P. Caputo et al.: Molecular phylogeny of Dipsacaceae over, the elided cladogram from the combined dataset has a topology which differs from that shown in Fig. 1 only for the position of Pterocephalidium, which is in this case sister group to the Pseudoscabiosa clade. From this, we may infer that the positions of the gaps, or the different ways in which they are scored, are not crucial for these datasets. For this reason, we choose the topology corresponding to dataset A and shown in Fig. 1 as a working hypothesis. The trees obtained for the separate ITS and trnL intron analyses (Table 2), albeit statistically congruent, are different. Notably, the ITS datasets (excluding the ITS dataset A, for which the consensus tree is greatly unresolved), show the Pycnocomon clade as sister group to Scabiosa, in a more nested position than in the cladogram of Fig. 1, as well as inversions of the relative positions of the Pseudoscabiosa and Knautia clades (data not shown). The trnL datasets alternately show the Pycnocomon clade either in the position of Fig. 1 or as sister group to the other major clade in the analysis. In all the trnL analyses, extensive collapses occur among the Pseudoscabiosa, Knautia, and Dipsacus clades as a consequence of lack of resolution in the data (data not shown). The elided ITS and trnL intron datasets obviously show similar discrepancies. In addition, the two cladograms from the elided trnL intron dataset show Knautia as the most basal taxon in the family and Pseudoscabiosa sister group to the clade including Scabioseae sensu Mayer and Ehrendorfer (1999). Discussion The results shown here address various controversial issues in Dipsacaceae relationships, in particular for what attains to the circumscription of tribes and to the monophyly of genera. Our proposed pattern of descent (Fig. 1) greatly differs from previous ones based on classical morphology, and, in particular, is different from the hypotheses formulated by Ehrendorfer (1964a,b, 1965), Verláque (1977b, 169 1984a, 1985a) and Caputo and Cozzolino (1994). The hypothesis presented here, however, is entirely congruent (to the extent of the taxa present) with recent findings on epicalyx differentiation (Mayer and Ehrendorfer 1999). In particular, our results confirm on molecular grounds the restricted circumscription for Scabioseae recently proposed by these just mentioned authors. In fact, the clade including Lomelosia, Pterocephalus, Pycnocomon, Scabiosa, and Sixalix (i.e. Scabioseae sensu Mayer and Ehrendorfer) is one of the most supported (Fig. 1, 92% bootstrap value, 6 BS value). Similarly, the sister group relationship between Pycnocomon and Lomelosia, which has been postulated by Mayer and Ehrendorfer (1999), but entirely excluded by previous authors, is well supported (84% bootstrap, 5 BS). Other well supported clades are those including Sixalix and Scabiosa (77% bootstrap, 3 BS) and the two just mentioned genera plus Pterocephalus (82% bootstrap, 3BS). Outside Scabioseae, the Pseudoscabiosa clade (95% bootstrap, 10 BS) and the Dipsacus clade (88% bootstrap, 6 BS) may also be mentioned in this regard. The least supported clades (bootstrap values <50%, collapsing in any one-step longer tree) are the clade including Knautia and Pterocephalidium and that immediately above. The position of Knautia is the one of the most widely varying in the present study. The branch leading to Knautia is by far the longest in the cladogram (49 autapomorphies, 26 of which non homoplasious), in spite of the fact that we included a species belonging to the most basal subgenus Trichera Schrad. (Caputo and Cozzolino 1994). Moreover, removal of Knautia from the combined dataset A causes an increase of 100% (from 2 to 4) in the branch support in the clade including Pseudoscabiosa, Succisa, Succisella, Pterocephalidium, Dipsacus, and Cephalaria. We suspect that the position of Knautia in the cladogram of Fig. 1 (and, to a lower extent, that of Pterocephalidium) may be influenced by low taxonomic sampling in that area of the family. Future studies including representatives of the Eastern sect. Sphaero- 170 dipsacus Lange of genus Dipsacus, of genus Pterocephalodes, as well as some, putatively more archaic, Eastern European species of Knautia (Ehrendorfer 1981) and the aberrant species Pterocephalus centennii M. J. Cannon, will possibly contribute to a slightly different placement of Knautia in the cladogram. Despite the fact that we regard the place of Knautia in the cladogram as a possible artifact and, therefore, we restrain from extensive comments, still, it is worth mentioning that the position of Knautia and Pterocephalidum, as well as their sister group relationship, is a novel interpretation emerging from our analysis. All four genera for which at least two species were included appear to be monophyletic (Fig. 1). Given the small number of representatives per genus, testing monophyly at genus level was not a goal of this investigation; however, the species selected for this investigation were chosen in order to maximise intrageneric diversity and/or differences in distribution ranges. For this reason the perennial Cephalaria leucantha, belonging to the archaic, Western Mediterranean subg. Fimbriatocarpus Szabó, was chosen together with the annual C. syriaca, belonging to the more widespread subg. Cephalaria. Similarly, the chamaephyte Sixalix farinosa, a narrow Tunisian endemic appearing as quite archaic in the genus according to Verláque (1986a), was selected together with the biennial S. atropurpurea, a widespread, more derived species. The perennial Japanese endemic Scabiosa japonica (regarded as a member of sect. Prismakena Bobrov, a group including the most archaic, Asian members of the genus) was selected, along with the chamaephytic South African endemic S. africana and with the perennial S. uniseta, an Italian representative of the widespread S. columbaria species group. Interestingly enough, genus Scabiosa as circumscribed for the past, even in its narrowest delimitation (i.e. including Lomelosia, Sixalix and Scabiosa s.str.) is not monophyletic, as Pterocephalus and Pycnocomon would be nested in it. On the contrary, the nomenclatural proposal by Mayer and Ehrendorfer (1999), i.e. P. Caputo et al.: Molecular phylogeny of Dipsacaceae to keep genus level for Lomelosia, and to use sectional rank within genus Scabiosa for Sixalix and Scabiosa would correctly represent the pattern of descent hypothesized here (Fig. 1). Our phylogenetic hypothesis suggests that adaptations to seed dispersal have been one of the strongest driving force in Dipsacaceae evolution, with similar selective pressures towards a particular dispersal mode causing the onset of similar epicalyx shapes, and correspondingly, of similar dispersal syndromes, in a parallel fashion in various taxa of the family (Fig. 2A). Among the taxa used in this analysis, the great majority is anemochorous. Lomelosia and Scabiosa have evolved the distal part of the epicalyx, above the epicalyx tube (the corona) into a membranous, multinerved, expanded rim, which helps the diaspore to be carried by the wind (pterochorous dispersal, i.e. wind dispersal depending on the presence of a wing). Furthermore, various species of Lomelosia (as well as some derived species of Sixalix) have a calyx which protrudes from the corona and is covered with short hairs, so exploiting the opportunity of epizoochorous dispersal. Pycnocomon and Sixalix are psammophylous (with the exclusion of the widespread S. atrorpurpurea subsp. maritima) and disperse by rolling on sand (Verláque 1986b). Both genera show a very reduced membranous corona, which is substituted by a more solid expansion of the distal part of the epicalyx tube, funnel-shaped in Sixalix and almost straight in Pycnocomon. However, this diminutive corona is homologous to the expanded corona of Lomelosia and Scabiosa (Mayer and Ehrendorfer 1999). The reduction of the corona in Pycnocomon and Sixalix occurred homoplasiously, under any character transformation model. Also Pseudoscabiosa limonifolia has a conspicuous membranous corona; however, as Mayer and Ehrendorfer (1999) point out, the distal part of the epicalyx is variable in this genus, with P. grosii (Font Quer) Devesa lacking a corona. By plotting presence or absence of the corona onto the cladogram P. Caputo et al.: Molecular phylogeny of Dipsacaceae Valeriana officinalis Patrinia intermedia Pseudoscabiosa Succisella Succisa Knautia Pterocephalidium Dipsacus Cephalaria 171 Dispersal syndrome Pterochorous Pogonochorous Autochorous Myrmecochorous Unspecialized Pycnocomon Ambiguous Lomelosia Pterocephalus Sixalix A Scabiosa Valeriana officinalis Patrinia intermedia Pseudoscabiosa Succisella Succisa Knautia Pterocephalidium Dipsacus Cephalaria Membranous corona Absent Present Ambiguous Pycnocomon Lomelosia Pterocephalus Sixalix Scabiosa Valeriana officinalis Patrinia intermedia Pseudoscabiosa Succisella Succisa Knautia Pterocephalidium Dipsacus Cephalaria B Plumose multiplied calyx bristles Absent Present Ambiguous Pycnocomon Lomelosia Pterocephalus Sixalix Scabiosa C Fig. 2. Morphological characters and ecological adaptations in the studied taxa plotted onto a simplified version of the cladogram in Fig. 1 (unambiguous transformations only). A dispersal syndromes (only those which, based on external evidence, are clearly plesiomorphic in the single terminals are shown). B membranous corona (a very small number of species in Dipsacus and Cephalaria have a membranous corona, and are not shown here). C plumose multiple calyx bristles. For simplicity’s sake, characters which are inapplicable in the outgroup (e.g. the characters related to the epicalyx) have been reconstructed as absent 172 (Fig. 2B), under all character transformation models, the wide, membranous corona is homoplasious in P. limonifolia, as compared to Lomelosia, Pycnocomon, Scabiosa, and Sixalix. Other Dipsacaceae have developed a membranous corona, whether expanded and evidently nervate or minute and without apparent nerves (Mayer and Ehrendorfer 1999, 2000): among them, a genus absent in our analysis, Pterocephalodes p.p. (most conspicuous in P. bretschneideri (Batal.) V. Mayer et Ehrend.) and Pterocephalus p.p. (e.g. P. pyrethrifolius Boiss. et Hohen.). In the latter genus, the epicalyces show a great diversity in structural detail, in spite of their general similarity (Mayer and Ehrendorfer 2000). Such variability, especially in the presence and shape of the corona, may be related to the fact that the corona does not seem to have an evident adaptative value towards dispersal and, as a consequence, its presence and shape may not be strictly constrained by selection. The species selected for the present investigation, the Balkanic P. perennis, shows a minute membrane joining the teeth present at the top of the epicalyx tube. However, the presence of a corona is regarded as an apomorphic feature in Pterocephalus, and the most archaic species of the genus lack it (Mayer and Ehrendorfer 2000, Fig. 25). Other genera, i.e. Pterocephalidium and Pterocephalus (as well as Pterocephalodes and Pycnocomon intermedium, both absent from this analysis) have multiple calyx bristles with a feathery indumentum, aiming at pogonochorous dispersal (i.e. wind dispersal depending on the presence of a pappus). Within the limits of the taxa present in this study, also this character developed homoplasiously (see also Mayer and Ehrendorfer 1999, 2000), irrespectively of the character transformation model chosen (Fig. 2C). The other major clade observed in the cladogram of Fig. 1, i.e. that including Cephalaria, Dipsacus, Knautia, Pseudoscabiosa, Pterocephalidium, Succisa, and Succisella does not show such a high elaboration in the P. Caputo et al.: Molecular phylogeny of Dipsacaceae epicalyx morphology. Conversely, the clade shows a greater variety of diaspore dispersal syndromes, not always linked to morphological differentiations in the epicalyx or in the calyx. Notable exceptions in this respect are Pseudoscabiosa, Pterocephalidium and Knautia. In Pseudoscabiosa, as already said, a membranous corona is either present [in P. limonifolia and P. saxatilis (Cav.) Devesa], or absent (in P. grosii). Interestingly, the lack of corona in the latter species corresponds to an increase in length, number and hairiness of the calyx bristles, as compared to the other two species, indicating a possible incipient transition from pterochory to pogonochory (or vice-versa). Pterocephalidium, the only genus in the family having two stamens instead of four, is clearly pogonochorous, with a plumose, pappus-like calyx. However, in this case the distal region of the epicalyx tube is elaborated into a single, very long, flexible awn of uncertain function (López Gonzáles 1987). In Knautia the pedicel of the dorsiventrally flattened epicalyx has developed into an elaiosome, and its dispersion is myrmecochorous (Fig. 2A). In the other genera, the epicalyx seems to be less directly involved in dispersion (Verláque 1985a, 1986b). In fact, the majority of the species of Dipsacus and Cephalaria rely on a combination of rigid, often acuminate, involucral/receptacular bracts in the capitulum, and of elastic stems, hereby allowing short-distance projection of diaspores. In some cases these diaspores subsequently adhere to animal furs, indicating either an autochorous or an epizoochorous dispersal syndrome (Fig. 2A), or a combination of both (Verláque 1985a, 1986b). For Succisa and Succisella, no correspondence between any morphological feature and a dispersal syndrome is observed; both genera are for most part confined to humid environments and have perhaps simplified their epicalyx by loosing features which were less valuable in the that environment. These taxa have been interpreted by various authors (e.g. Verláque 1986b, Caputo and P. Caputo et al.: Molecular phylogeny of Dipsacaceae Cozzolino 1994) as basal within Scabioseae sensu lato; however, they are clearly external to that tribe both according to Mayer and Ehrendorfer (1999) and to the data presented here, which show the two genera in a sister relation (with very high bootstrap value and decay index) with Pseudoscabiosa. Despite the differences in dispersal strategies, which have apparently contributed to cause adaptative radiation in various taxa and, as a consequence, may have influenced the evolution rate, comparisons of sequence substitution rates (Robinson et al. 1998, Robinson-Rechavi and Huchon 2000) showed that no one of the tested lineages evolved significantly faster than any other; in particular, not even a comparison among pterochorous, pogonochorous, autochorous, and myrmecochorous groups did show significant differences. However, our choice of taxa is unbalanced under this respect (e.g. only a single species of Knautia, i.e. a single myrmecochorous terminal), and also annuals (which constitute derived species in various genera and which often have a higher evolution rate) are underrepresented in our data set. This unbalance may have influenced the results of the test. For all what said above, also our study, after the work by Mayer and Ehrendorfer (1999) would indicate that the gross morphology of the epicalyx is deceptive in inferring relationships, as the whole family is beset by rampant homoplasies originated by selective constraints towards several diaspore dispersal syndromes. The presence of a diaphragma, i.e. a protrusion of the distal part of the inner side of the epicalyx tube, which, by enclosing the calyx stalk, secures a better encasing of the achene in the epicalyx, is much more promising in assessing relationships (Mayer and Ehrendorfer 1999, 2000). This structure, together with the corresponding diversification of an epidiaphragma (Mayer and Ehrendorfer 1999), is present in virtually all members of Lomelosia, Scabiosa, Sixalix, and in various species of Pterocephalus (including P. perennis, 173 which was used for this study), but absent from all other Dipsacaceae. However, the diaphragma, together with the membranous corona, may be regarded as synapomorphic for Scabioseae sensu Mayer and Ehrendorfer (1999). This character may be either reverted more than once (loss in the immediate ancestors of Pterocephalus and novel development within the genus) or occurrence in Pterocephalus is plesiomorphic. The latter hypothesis, however, is excluded by Mayer and Ehrendorfer (2000). Palynological and karyological evidences (Verláque 1985a, 1986b; Caputo and Cozzolino 1994; Mayer and Ehrendorfer 1999) are much more supportive of our phylogenetic hypothesis. Dipsacaceae pollen (Verláque 1985a, 1986b) is either triporate (in Knautia, Lomelosia, Pterocephalidium, Pycnocomon, and in the aberrant annual Scabiosa parviflora Desf.) or tricolpate (in Cephalaria, Dipsacus, Pseudoscabiosa, Pterocephalus, Scabiosa, Sixalix, Succisa, and Succisella). Tricolpate pollen, also present in all Valerianaceae, is unequivocally reconstructed as plesiomorphic in Dipsacaceae (data not shown); triporate pollen has independently evolved more than once, in Knautia/Pterocephalidium, in Lomelosia/Pycnocomon, and in S. parviflora. Chromosome numbers in the family are less amenable to an unequivocal reconstruction. In fact, in the majority of the large genera, chromosome numbers show dysploid variation. However, the haploid numbers appearing as plesiomorphic in the genera in which they occur are n ¼ 8 in Pterocephalidium, Scabiosa, Sixalix (Verláque 1985b, 1986a,b), n ¼ 9 in Cephalaria, Dipsacus, Pycnocomon, Pterocephalus and Lomelosia (Verláque 1977a, 1985a), n ¼ 10 in Knautia, Pseudoscabiosa, Succisa and Succisella (Ehrendorfer 1962, 1981; Verláque 1985b, 1986a). By plotting chromosome haploid numbers on the cladogram (data not shown), it appears that n ¼ 8 independently originated twice in the family (in Pterocephalidium, as well as in Scabiosa and Sixalix); n ¼ 9 is ambiguous in reconstruction; regardless, it is reconstructed 174 as plesiomorphic in the family under both an accelerated and a delayed character transformation model (data not shown). However, the character state n=9 is synapomorphic for the two groups in which it appears only under a delayed transformation model. The state n ¼ 10 is also ambiguous: it appears as uniquely derived in both the Pseudoscabiosa clade and Knautia only under an accelerated transformation model. In spite of all the homoplasies described above, it may be noted that our present phylogenetic hypothesis requires a smaller number of homoplasious changes in the presence of diaphragma, pollen morphology and chromosome numbers as compared to previous ones (e.g. Verláque 1986b, Caputo and Cozzolino 1994). Furthermore our evolutionary reconstruction is almost entirely congruent with the novel insights on epicalyx morphology shown by Mayer and Ehrendorfer (1999). The few possible problems of our proposed pattern of descent (e.g. the position of Knautia) will be probably solved by both the inclusion of some critical taxa (members of Pterocephalodes; basal species of Dipsacus, representatives of the derived subgenera of Knautia) and the extension of the molecular dataset. References Aceto S., Caputo P., Cozzolino S., Gaudio L., Moretti A. (1999) Phylogeny and evolution of Orchis and allied genera based on ITS DNA variation: morphological gaps and molecular continuity. Mol. Phylogenet. Evol. 13: 67–76. Backlund A., Bremer B. (1997) Phylogeny of the Asteridae s. str. based on rbcL sequences, with particular reference to the Dipsacales. Plant Syst. Evol. 207: 225–254. Baldwin B. G., Sanderson M. J., Porter M. J., Wojciechowski M. F., Campbell C. S., Donoghue M. J. (1995) The ITS region of nuclear ribosomal DNA: a valuable source of evidence in angiosperm phylogeny. Ann. Missouri Bot. Gard. 82: 247–277. Caputo P., Cozzolino S. (1994) A cladistic analysis of Dipsacaceae (Dipsacales). Plant Syst. Evol. 189: 41–61. P. Caputo et al.: Molecular phylogeny of Dipsacaceae Caputo P., Cozzolino S. (1995) Cladogeny in Dipsacaceae: dispersal and vicariance models. Boll. Soc. Sarda Sci. Nat. 30: 233–243. Caputo P., Stevenson D.W., Wurtzel E.T. (1991) A phylogenetic analysis of American cycads (Cycadales) using chloroplast DNA restriction fragment polymorphisms. Brittonia 43: 135–145. Devesa J. A. (1984a) Revision of the genus Scabiosa in Spain and Balearic Islands. Lagascalia 12: 143–212. Devesa J. A. (1984b) Pseudoscabiosa, genero nuevo de Dipsacaceae. Lagascalia 12: 213–221 Doyle J. J., Doyle J. L. (1990) Isolation of plant DNA from fresh tissue. Focus 12: 13–15. Ehrendorfer F. (1964a) Cytologie, Taxonomie und Evolution bei Samenpflanzen. In: Turril W. B. (ed.) Vistas in Botany. 4. The MacMillan Company, New York, pp. 99–186. Ehrendorfer F. (1964b) Über stammesgeschichtliche Differenzierungsmuster bei den Dipsacaceen. Ber. Deutsch. Bot. Ges. 77: 83–94. Ehrendorfer F. (1965) Evolution and karyotype differentiation in a family of flowering plants: Dipsacaceae. Genetics Today (Proc. XI International Congress of Genetics, The Hague, The Netherlands, 1963) 2: 399–407. Ehrendorfer F. (1981) Neue Beiträge zur Karyosystematik und Evolution der Gattung Knautia (Dipsacaceae) in den Balkanländern. Bot. Jahrb. Syst. 102: 225–238. Farris J. S., Källersjö M., Kluge A. G., Bult C. (1994) Testing significance of incongruence. Cladistics 10: 315–319. Gielly L., Taberlet P. (1996) A phylogeny of European gentians inferred from chloroplast trnL (UAA) intron sequences. Bot. J. Linn. Soc. 120: 57–75. Goloboff P. (1999) Nona. Instruction manual. Published by the author, S. M. de Tucumán, Argentina. Greuter W., Burdet R. (1985) Dipsacaceae. In: Greuter W., Raus T. (eds.) Med-Checklist Notulae 11. Willdenowia 15: 71–76. Greuter W., Burdet H. M., Long G. (eds.) (1986) Med-Checklist. Vol. 2. Editions des Conservatoire et Jardin botaniques de la Ville de Genève, Genève, pp. 176–198. López Gonzáles G. (1987) Pterocephalidium, un nuevo género ibérico de la familia Dipsacaceae. Anales Jard. Bot. Madrid 43: 245–252. Mayer V., Ehrendorfer F. (1999) Fruit differentiation, palynology, and systematics in the Scabi- P. Caputo et al.: Molecular phylogeny of Dipsacaceae osa group of genera and Pseudoscabiosa (Dipsacaceae). Plant Syst. Evol. 216: 135–166. Mayer V., Ehrendorfer F. (2000) Fruit differentiation, palynology, and systematics in Pterocephalus Adanson and Pterocephalodes, gen. nov (Dipsacaceae). Bot. J. Linn. Soc. 132: 47–78. Nixon K. C. (1999) Winclada (beta) ver. 0.9.9. Published by the author, Ithaca, NY. Robinson M., Gouy M., Gautier C., Mouchiroud D. (1998) Sensitivity of the relative-rate test to taxonomic sampling. Mol. Biol. Evol. 15:1091– 1098. Robinson-Rechavi M., Huchon D. (2000) RRTree: Relative-Rate Tests between groups of sequences on a phylogenetic tree. Bioinformatics 16: 296– 297. Sanger F., Niklen S., Coulson A. R. (1977) DNA sequencing with chain terminating inhibitors. Proc. Natl. Acad. Sci. USA 74: 5463–5467. Taberlet P., Gielly L., Pautou G., Bouvet J. (1991) Universal primers for amplification of three noncoding regions of chloroplast DNA. Plant Mol. Biol. 17: 1105–1109. Thompson J. D., Higgins D. G., Gibson T. J. (1994) CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, positions-specific gap penalties and weight matrix choice. Nucl. Acids Res. 22: 4673–4680. Verláque R. (1977a) Rapports entre les Valerianaceae, les Morinaceae et les Dipsacaceae. Bull. Soc. bot. Fr. 124: 475–482. Verláque R. (1977b) Importance du fruit dans la determination des Dipsacaceae. Bull. Soc. bot. Fr. 124: 515–527. 175 Verláque R. (1984a) A biosystematic and phylogenetic study of the Dipsacaceae. In: Grant R. (ed.) Plant biosystematics. Academic Press, Toronto, pp. 307–320. Verláque R. (1984b) Etude biosystématique et phylogénétique des Dipsacaceae. I. – Délimitation des Dipsacaceae à l’intérieur des Dipsacales, rapports avec les autres familles de l’ordre. Rev. Gen. Bot. 91: 81–121. Verláque R. (1985a) Etude biosystématique et phylogénétique des Dipsacaceae. II - Caractères gènèraux des Dipsacaceae. Rev. Cytol. Biol. Veg. Le Botaniste 8: 117–168. Verláque R. (1985b) Etude biosystématique et phylogénétique des Dipsacaceae. III – Tribus des Knautieae et des Dipsaceae. Rev. Cytol. Biol. Veg. Le Botaniste 8: 171–243. Verláque R. (1986a) Etude biosystématique et phylogénétique des Dipsacaceae. IV – Tribus des Scabioseae (phylum n° 1, 2, 3). Rev. Cytol. Biol. Veg. Le Botaniste 9: 5–72. Verláque R. (1986b) Etude biosystématique et phylogénétique des Dipsacaceae. V - Tribus des Scabioseae (phylum n° 4) et conclusion. Rev. Cytol. Biol. Veg. Le Botaniste. 9: 97–176. Wheeler W. C., Gatesy J., DeSalle R. (1995) Elision: a method for accommodating multiple molecular sequence alignments with alignmentambiguous sites. Mol. Phylog. Evol. 4: 1–9. Addresses of the authors: Paolo Caputo (corresponding author; e-mail: pacaputo@unina.it), Salvatore Cozzolino, Aldo Moretti, Dipartimento di Biologia vegetale, Università di Napoli Federico II, Via Foria 223, I-80139 Napoli, Italy.