Abstract

INTRODUCTION

Valerianaceae (Dipsacales) is made up of some 400 species of almost cosmopolitan distribution, the only exceptions being Australia and the Pacific islands. Approximately 40 genera have been described. The suggested number of genera varies from one (Linné, 1753) to 16 (Graebner, 1906). Recent revisions accept 13 (Weberling, 1970) or eight genera (Eriksen, 1989).

The first genus described in the family was Valeriana L. (1753). As is often the case with Linnean genera, the concept of Valeriana was wide and included all 16 species of Valerianaceae known at that time. Miller (1754) segregated the species lacking a pappus into a new genus, Valerianella. Later on, three new genera were split from Valerianella: Fedia Gaertn. emend. Moench, nom. cons., Patrinia Juss. and Plectritis (Lindl.) DC. De Candolle (1815) separated the species with only one stamen from Valeriana, thus creating the genus Centranthus. Then, De Candolle (1830) segregated the species with a five‐lobed accrescent calyx from Patrinia, and formed the new genus Nardostachys.

Despite the amendments by Miller (1754) and De Candolle (1815), Valeriana remains heterogeneous, especially because of the great diversity in the South and Central American species. Some of them lack a pappus, which could indicate that they should be placed in Valerianella, although none of them has ever been placed there. They have been placed in Valeriana or in other solely South American genera on the basis of their life history (perennials rather than annuals), ovary anatomy and inflorescence type. This heterogeneity among the American species has led to the description of many taxa, sometimes included in Valeriana, sometimes placed in other genera. The frequent reclassification of the species in this group indicates the unsuitability of morphology alone for clarifying systematic assignments.

Most modern authors (Graebner, 1906; Weberling, 1970) agree on the classification of Valerianaceae into three tribes, as shown in Table ​Table11 and listed as follows.

Table 1.

Classification of the Valerianaceae according to Weberling (1970)

Tribe Patrinieae Höck

Patrinia Juss.

Nardostachys DC.

Tribe Triplostegiae Höck

Triplostegia Wall. ex DC.

Tribe Valerianeae Höck

Subtribe Fediinae Graebn. (emend. Weberling)

Plectritis (Lindl.) DC.

Valerianella Mill.

Fedia Gaertn.

Subtribe Valerianinae Graebn. (emend. Weberling)

Valeriana L. (incl. Phuodendron Graebn.)

Astrephia Dufr.*

Stangea Graebn.*

Aretiastrum (DC.) Spach*

Phyllactis Pers.*

Belonanthus Graebn.*

Subtribe Centranthinae Graebn.

Centranthus DC.

Open in a separate window

* Eriksen (1989) considers these exclusively South American genera as sections of Valeriana.

Tribe Triplostegieae Höck, with a single genus from Asia: Triplostegia Wall., a genus which De Candolle (1830) placed in Valerianaceae, a position accepted by Höck (1902), Graebner (1906), Weberling (1970), Backlund and Bremer (1997) and Backlund and Nilsson (1997). However, recent molecular analyses placed Triplostegia as sister to the Dipsacaceae (Bell et al., 2001; Wen‐Heng et al., 2003).

Tribe Patrinieae Juss., formed by two Asian genera, Patrinia and Nardostachys. Most authors agree in assigning a basal position in the family to both genera (e.g. Weberling, 1975). Molecular analyses indicate the paraphyletic nature of the tribe (Bell et al., 2001; Pyck et al., 2002).

Tribe Valerianeae Höck, formed by the genera Valeriana, Centranthus, Fedia, Plectritis and Valerianella, with the addition of some small American genera segregated by some authors. Valeriana s.l. and Valerianella, with 200 and 50 species, respectively (Judd et al., 1999), are the largest genera in the family.

The tribe Valerianeae encompasses most of the diversity of the family, on morphological, geographical and numeric grounds. It is also the tribe with the most interesting and vexing taxonomic problems. They are reviewed briefly here.

MATERIALS AND METHODS

RESULTS

matK analysis

The matK alignment for 41 taxa consisted of 960 positions and contained 228 phylogenetically informative substitutions and 45 phylogenetically informative gapped positions. The major insertion/deletion event (indel) appears as a deletion of 39 bp (position 197–235) in all the sequences of Valeriana (apart from V. hardwickii Wall.), Centranthus and Plectritis. Mean pairwise distance (as calculated by PAUP) within the ingroup varied from 0 % [between the Andean species Valeriana bractescens (Hook.) Höck, V. parviflora (Trevir.) Höck and V. phylicoides (Turcz.) Briq. and between Centranthus angustifolius (Mill.) DC., C. ruber DC. and C. trinervis (Viv.) Bég.] to 11·6 % (between Patrinia triloba Miq. and Valerianella locusta L.). Mean pairwise distance between ingroup and outgroup varied from 2·7 % (between Abelia chinensis R. Br. and Nardostachys jatamansi (D. Don) DC.) to 17·3% (between Sambucus nigra L. and Valerianella locusta).

The only change resulting from the different treatments of indels concerned Triplostegia, which was placed as sister to Dipsacaceae with strong support when the indels were coded as missing data or fifth base. When the indels were coded as ‘total omission’ this node collapsed in the analysis. The reason for this is that the informative positions coincide with a gap in one species and are therefore eliminated by this method. Finally, it was decided to code the indels as ‘missing data’.

The parsimony analysis yielded eight MPTs of 556 steps in one island. The strict consensus of all trees is shown in Fig. ​Fig.1;1; the CI was 0·7115 and the RI was 0·9294 (Table ​(Table33).

Open in a separate window

Fig. 1. Strict consensus tree of the eight most parsimonious trees generated by the matK matrix. Numbers above branches represent bootstrap values and numbers below indicate Bremer supports.

Table 3.

Comparison of results from the ITS and matK

Data set	matK	ITS
Number of taxa	41	18
Total characters	960	522
Informative substitutions	228	187
Number of MPTs	8	6
Number of steps	556	357
Consistency index (CI)	0·7115	0·7437
Retention index (RI)	0·9294	0·8321
Mean pairwise distances, ingroup (%)	11·6	25·2

Open in a separate window

The consistency and homoplasy indexes are calculated excluding uninformative characters.

Valerianaceae (excluding Triplostegia) are monophyletic (BS = 91%, DI = 3) (Fig. ​(Fig.1).1). The association of Nardostachys and Patrinia and their position as sister to the remainder of the family are strongly supported (BS = 93%, DI = 4 and BS = 100%, DI = 13, respectively). Patrinia is monophyletic with moderate support (BS = 88%, DI = 2). The Asian species Valeriana hardwickii is sister to the remainder of Valerianeae (BS = 100%, DI = 22). Within this group, Fedia and Valerianella form a strongly supported clade (BS = 100% and DI = 16) and are strongly supported as sister group to a clade consisting of Valeriana (excluding V. hardwickii), Centranthus and Plectritis (BS = 100% and DI = 12). These genera share the previously mentioned 39 bp deletion. Even though we coded indels as missing data (and thereafter lost the phylogenetic information conveyed by this large indel), support for this branch is extremely high. This clade is divided into two groups. The first group is a moderately supported clade formed only by species of Valeriana (BS = 78%, DI = 2), divided into two robust clades, one with Valeriana officinalis L. subsp. tenuifolia Schüber & Martens, V. dioica and V. jatamansi Jones (BS = 100%, DI = 12), and the other with V. pyrenaica L., V. saliunca All., V. montana and V. apula Pourr. (BS = 98%, DI = 6). The second group is a weakly supported (BS = 64%, DI = 1) clade formed by the genera Centranthus and Plectritis and some Valeriana spp. This heterogeneous clade is divided into two branches: a well‐supported clade (BS = 99%, DI = 6) with Valeriana longiflora Willk. and the genus Centranthus, which appears to be monophyletic (BS = 96%, DI = 3) with C. calcitrapae (L.) Dufr. sister to the rest of the species (BS = 98%, DI = 4). The second clade is weakly supported (BS = 67%, DI = 1), and encompasses the American species and the genus Plectritis. Finally, the American group is poorly resolved, the only exception being the clade formed by Valeriana parviflora, V. bractescens, V. phylicoides and V. rosaliana C. A. Meyer with moderate support (BS = 92%, DI = 2), with V. rosaliana as sister to the three other species (BS = 86%, DI = 2).

For Valeriana montana, V. pyrenaica and V. jatamansi, two different matK sequences were determined, starting with the same amplification product obtained always from a single individual. The two matK sequences were separated by using different sequencing primers. The translation of these sequences into proteins revealed the existence of many stop codons in one of the sequences for each one of the three species. A phylogenetic analysis including both types of sequences shows all the sequences with stop codons grouped together within Valerianaceae. The possibility of long‐branch attraction was excluded by carrying out a neighbour‐joining analysis (Fig. ​(Fig.2).2). Our conclusion is that the sequences with stop codons correspond to non‐functional genes or pseudogenes. Regarding Fedia graciliflora Fisch. & C. A. Meyer, only a single matK sequence was obtained which groups together with the Valeriana pseudogenes; this sequence also contains many stop codons and thus corresponds to a pseudogene. Our hypothesis is confirmed by the results of Bell et al. (2001), who sequenced the matK region of Fedia cornucopiae (L.) Gaertn., and this sequence is placed in a conventional position in our analysis.

Open in a separate window

Fig. 2. Neighbour‐joining phylogram generated by the matK matrix including the pseudogene (PS) sequences.

Wendel and Doyle (1998) explained the presence of pseudo genes by means of an ancestral polymorphism that was retained throughout speciation processes. As pseudogenes are not homologous to normal genes, we have removed them from the final analyses involving the matK gene.

ITS analysis

Our DNA extractions were often contaminated with alien DNA that, at least in one case (Valeriana tripteris L.), was identified as the endophytic fungus Colletotrichum destructivum O’Gara. The idenfication was carried out by comparison with the sequences from GenBank (ITS sequences for approx. 15 species of fungi were obtained). This contamination was so extensive that it precluded further sequencing of the ITS region without resorting to cloning.

The ITS alignment for 18 taxa consisted of 522 positions and contained 187 phylogenetically informative substitutions. The mean pairwise distance (as calculated by PAUP) within the ingroup varied from 0 % (between Valeriana bractescens and V. parviflora, and between Centranthus trinervis and C. angustifolius) to 25·2 % [between Centranthus angustifolius and Patrinia intermedia (Hornem.) Roem. & Schult.]. Mean pairwise distance between the ingroup and outgroup varied from 16·9 % [between Scabiosa africana L. and Patrinia villosa (Thunb.) Juss.] to 25·5 % (between Scabiosa uniseta Savi and Centranthus angustifolius).

The ITS alignment shows numerous indels, often combined with substitutions of nucleotides among the studied species. We coded these indels as missing data because this seems to be the more adequate approach where the sequences contain important insertion–deletion zones (Raymúndez et al., 2002).

The parsimony analysis yielded six MPT of 357 steps. The tree obtained is shown in Fig. ​Fig.3;3; the CI was 0·7437 and the RI was 0·8321 (Table ​(Table33).

Open in a separate window

Fig. 3. Strict consensus tree of the six most parsimonious trees generated by the ITS matrix. Numbers above branches represent bootstrap percentages and numbers below indicate Bremer supports.

Patrinia is monophyletic (BS = 100%, DI = 21) and is strongly supported as sister to Valerianeae (BS = 100%, DI = 20). Fedia and Valerianella form a robust clade (BS = 91%, DI = 4), sister to an unsupported group formed by Valeriana and Centranthus (BS = 59%, DI = 2). Valeriana celtica L. is sister to the remainder of Valeriana and Centranthus (BS = 97%, DI = 5). Centranthus is monophyletic (BS = 100, DI = 24), and sister to a weakly supported group formed by Valeriana (excluding V. celtica). In this group, only the clade formed by V. parviflora and V. bractescens is strongly supported (BS = 100, DI = 11).

DISCUSSION

Our results are basically coincident with the biogeography of Valerianaceae, with four groups being identified: Asian, Mediterranean, Eurasian and American.

The Asian group is formed by the genera Patrinia and Nardostachys, with the addition of Valeriana hardwickii. As was the case in the molecular analysis by Bell et al. (2001), Triplostegia glandulifera Wall. ex DC. falls outside Valerianaceae with the species of Dipsacales used as the outgroup. This position contradicts not only morphological evidence, but also chemical data: Triplostegia contains valepotriates, a family of chemical compounds only known from Valerianaceae (Backlund and Moritz, 1998). Further studies are needed to clarify this contradiction.

Patrinia and Nardostachys, with strong morphological affinities, are sister to the remainder of the family in a well‐supported clade (tribe Patrinieae) in the matK analysis (Fig. ​(Fig.1).1). Bell et al. (2001) and Pyck et al. (2002), on the other hand, placed Nardostachys as derived in relation to Patrinia.

The unexpected position of Valeriana hardwickii, as sister to the remainder of Valerianeae, makes the genus Valeriana paraphyletic. However, V. hardwickii is an Asian species, and Valerianaceae are generally considered to have an Asian origin (Höck, 1902; Eriksen, 1989). Our results also suggest an Asian origin for the Valerianeae. This hypothesis needs confirmation, especially because of the profound taxonomic and nomenclatural implications of the paraphyly of the genus Valeriana.

The Mediterranean group is made up of two different clades. The first group is formed by the genera Fedia and Valerianella and conforms to the subtribe Fediinae sensu Weberling (1970), with the exclusion of Plectritis. Both ITS (Fig. ​(Fig.3)3) and matK (Fig. ​(Fig.1)1) analyses strongly support this association, as did those of rbcL (Bell et al., 2001) and atpB‐rbcL (Raymúndez et al., 2002). The ITS analysis supports the monophyly of Fedia as for atpB‐rbcL (Raymúndez et al., 2002). The subtribe Fediinae is sister to the rest of the tribe Valerianeae (excluding V. hardwickii).

A second clade is formed by a monophyletic Centranthus and Valeriana longiflora. This unexpected association opens up an avenue for elucidating the origin of Centranthus, which could be merged into Valeriana. The most important morphological trait shared by these taxa is the long corolla tube. Relationships of this group and the Eurasian and American ones are obscure in the analyses.

The Eurasian group is formed by two robust clades in matK analyses. One includes Valeriana officinalis subsp. tenuifolia, V. dioica and V. jatamansi, and the other one V. montana, V. pyrenaica, V. apula and V. saliunca. Neither of these groups coincides with the classification suggested by Höck (1882). We find here once again the lack of useful morphological traits among the Eurasian species of Valeriana. To date, we have been unable to detect any character for defining either group on morphological grounds. The ITS analysis places V. celtica as sister to Valerianinae, which makes Valeriana paraphyletic.

The American group is made up of species of Valeriana from both American subcontinents with the addition of a North American species of Plectritis. This group of species is not supported, or only weakly supported, in both the matK (Fig. ​(Fig.1)1) and ITS (Fig. ​(Fig.3)3) analyses. The position of Plectritis within the American clade has never been suggested before. The low resolution of this group in the matK tree does not allow any inference on its precise position, but the exclusion of Plectritis from the subtribe Fediinae seems to be confirmed. Its origin should be sought within Valeriana.

Our results confirm the paraphyly of Valeriana in its present delimitation, for four reasons. Firstly, V. hardwickii is placed at the base of the tribe Valerianeae. Secondly, V. longiflora is sister to Centranthus. Thirdly, Plectritis is firmly nested in the clade of American Valeriana spp. Fourthly, V. celtica is sister to the subtribe Valerianinae. Certainly, a redefinition of the subtribal and generic boundaries and some nomenclatural changes will be necessary to reconcile the classification of the family with this new evidence. However, our sampling of species and sections of Valeriana is not complete: American species are not well represented, which makes our results somewhat provisional. Prior to any changes, more comprehensive representation of the genus is necessary.

The matK region has proved to be an excellent tool: the matK tree is finely resolved and the nodes are well supported, both in deep and terminal branches. The exceptions (polytomies and weakly supported branches in the American and Eurasian clades) must correspond to incomplete sampling of these groups, which it is intended to correct in further studies, or insufficient variation in matK especially in the cases of rapid radiation and recent speciation.

Regarding the ITS results, for the first time a nuclear region has been used for phylogenetic analysis of Valerianaceae, whilst five different plastid ones have already been used. The ITS results are concordant with the matK phylogeny, and confirm the monophyly of the subtribe Fediinae which was unclear in the ndhF analysis (Bell et al., 2001).

Discovery of pseudogenes in this group brings new and interesting perspectives. The literature includes only a few examples of plastid DNA polymorphisms crossing species boundaries (Wendel and Doyle, 1998). According to GenBank data, matK pseudogenes are reported here for the first time within Dicotyledones. Moreover, species possessing these pseudogenes in our analysis form a monophyletic group within Valerianaceae. This implies that species with pseudogenes share a common polymorphic ancestor. Why was this pseudogene only found in a few descendants of this taxon? Two hypotheses could be put forward. It could be the result of incomplete lineage sorting (by which the polymorphism of the common ancestor was lost in most species). Alternatively, all descendants of the polymorphic ancestor could be polymorphic, but we have sequenced only one type of matK sequence for each species (the efficient one for most of them).

In view of the excellent results obtained with the matK region, a new analysis involving this gene should be considered for the purposes of resolving some problems revealed by our research. Perhaps the most interesting questions that remain open are the origin of Centranthus, the position of Plectritis within Valeriana, and the reasons for the unexpected placement of V. hardwicki and V. celtica.

ACKNOWLEDGEMENTS

This research was partly funded by projects BOS2001‐3041‐C02‐02, Ministerio de Ciencia y Tecnología, Spain, and 2001SGR00125, Generalitat de Catalunya. Oriane Hidalgo benefited from a 9‐month grant from the Ministerio de Asuntos Exteriores, Spain. We thank E. Desmarais, J.‐Y. Dubuisson, N. Garcia‐Jacas, M.‐B. Raymúndez and R. Vilatersana for their help in the molecular procedures. Special thanks are extended to N. Xena de Enrech for his permanent collaboration. We are grateful to F. Barrie, A. Fridlender, L. Garraud. S. Ghishmire, Pr. Hiroshi, J. Luque, T. Navarro, J. El Oualidi, C. Roché, O. Salvador, P. Schäfer, Y. Thomas and M. Veny for their assistance with collections. We also thank the two anonymous reviewers for critically reading the manuscript.

Notes

Received: 29 October 2002;; Returned for revision: 12 May 2003; Accepted: 24 November 2003

References