On generic rank and phylogenetic relationships of Dorycnopsis Boiss . ( Leguminosae , Loteae )

Nuclear ribosomal ITS sequence data as well as morphological data show that Dorycnopsis gerardii (L.) Boiss. can not be placed in the genus Anthyllis L. The genus Dorycnopsis Boiss. includes two species, D. gerardii and D. abyssinica (A. Rich.) V.N. Tikhom. et D.D. Sokoloff (=Vermifrux abyssinica (A. Rich.) J.B. Gillett). Morphological similarity between Dorycnopsis gerardii and Anthyllis onobrychioides Cav. might be best explained by evolutionary parallelism. Anthyllis (including Hymenocarpos Savi but excluding Dorycnopsis and the monotypic Tripodion Medik.) is well-resolved as a highly supported monophyletic group in analyses of nrITS data set.

López González (2004) emphasized the substantial morphological similarity between Anthyllis onobrychioides and Dorycnopsis gerardii.In his opinion, these species differ mainly in flower colour and inflorescence morphology.He considered these differences as insufficient to maintain the genus Dorycnopsis as distinct from Anthyllis and included D. gerardii in Anthyllis.Consequently, he concluded that Vermifrux should be retained as a separate genus noting that molecular phylogenetic data by Allan & al. (2003) showed no close relationship between Vermifrux abyssinica and the genus Anthyllis.Allan & al. (2003) presented a phylogenetic analysis of nuclear ribosomal ITS data of more than 80 species of Loteae, including Vermifrux abyssinica and Anthyllis onobrychioides.In a maximum parsimony analysis of nrITS sequences, these species do not group together: A. onobrychioides is closely related to other Anthyllis species while Vermifrux grouped with Coronilla, although with only low bootstrap support.In Neighbour-Joining and Bayesian analyses Vermifrux does not group with Coronilla; it is also not resolved as a relative of Anthyllis (Allan & al., 2003;Degtjareva & al., 2003).These data clearly show that Vermifrux abyssinica cannot be placed in Anthyllis.However, the type species of Dorycnopsis, D. gerardii, has not yet been included in molecular phylogenetic analyses of Loteae (e.g., Allan & Porter, 2000;Kropf & al., 2002;Allan & al., 2003;Degtjareva & al., 2003;Nanni & al., 2004).Therefore, we have decided to produce and analyse nrITS sequences of Dorycnopsis gerardii to resolve its taxonomic position.

Materials and methods
Complete sequences of ITS1, ITS2 and the 5.8S region were generated for four accessions of Dorycnopsis gerardii and one accession of D. abyssinica.In addition, nrITS and 5.8S sequences were produced for four species of Anthyllis.Voucher information is presented in Table 1.Total genomic DNA was isolated from leaf tissue (in one case from fruit material) using the CTAB method of Doyle and Doyle (1987).PCR reactions were performed with universal primers (White & al. 1990).The sequence data were obtained through direct sequencing of double-stranded templates on an ABI 310 automatic sequencer.Both spacer regions   Table 1.GenBank accession numbers of sequences used and voucher information for newly produced sequences.
were sequenced in their entirety on both strands.Details of DNA extraction, PCR amplification (including primer locations and characteristics), and DNA purification and sequencing strategies used are the same as described by Valiejo-Roman & al. (2002).
The newly generated nrITS and 5.8S sequences were analysed together with a number of GenBank accessions (see Table 1).In some cases we have changed the nomenclature used in the original sequence publications.We here use generic limits and nomenclature of Loteae as proposed by Sokoloff (2003a).The data set used for the present analysis includes most genera of tribe Loteae and all taxa that were indicated as relatives of Anthyllis, Dorycnopsis and Vermifrux by different authors.Members of Robinieae (Robinia and Hebestigma) and Sesbanieae (Sesbania) were used as outgroups.
DNA sequences were aligned using CLUSTAL W 1.75 (Thompson & al., 1994) and corrected manually using the SED editor of the VOSTORG phylogenetic analysis package (Zharkikh & al., 1990).The alignment is available from the authors on request.Maximum parsimony analysis involved a heuristic search conducted with PAUP* 4.08b (Swofford, 2000) using TBR branch swapping, options mulpars, steepest descent, collapse, and acctran selected, with character states specified as unordered and equally weighted.1000 replicates with random addition of sequences were performed and all shortest trees were saved.All gaps were treated as missing data.Bootstrap values were calculated from 100 replicate analyses with random addition sequence of taxa.1000 most parsimonious trees from each replicate were saved.The GTR+I+Γ model of nucleotide substitution was selected by the AIC in Modeltest (Posada & Crandall, 1998).Maximum likelihood distances were computed using the selected model of molecular evolution.Distance trees were calculated using the Neighbour-Joining method as implemented in PAUP*.1000 bootstrap resamplings were performed; insertions and deletions were not taken into account.Bayesian inference of phylogeny was explored using the MrBayes program version 3.1 (Huelsenbeck & Ronquist, 2001).The GTR+I+Γ model of evolution was used.5×10 6 generations were performed; trees from the first 4×10 6 generations were discarded.The number of generations to be discarded was determined using a convergence diagnostic.

Results
All four newly produced nrITS and 5.8S sequences of D. gerardii are identical.These sequences are de-rived from material from different parts of the species range (Portugal, Spain and Corsica).The new sequence of D. abyssinica (based on material from Ethiopia) is identical to the sequence produced by Allan & al. (2003)  Numerous indels were introduced in the sequences during alignment, mainly in regions of ambiguous alignment.Such positions together with adjacent bases were excluded from analysis.The alignment of all 60 ITS sequences after exclusion of 220 ambiguous positions resulted in a matrix of 603 nucleotide positions.Of these, 306 positions are parsimony-informative, 247 invariable, and 50 variable but parsimonyuninformative.In the maximum parsimony analysis, 492 shortest trees (1631 steps) are found, with a consistency index of 0.41 and a retention index of 0.69.
Tree topology and bootstrap values found in the Neighbour-Joining analysis are similar to those found in the maximum parsimony analysis (Fig. 2).However, monophyly of subgen.Cornicina received higher bootstrap support in the Neighbour-Joining analysis (92%).

Discussion
Our molecular phylogenetic analyses show monophyly of both Anthyllis (incl.Hymenocarpos) and Do-rycnopsis (incl.Vermifrux).A morphological cladistic analysis also showed monophyly of these two genera, but the morphological data suggested that Dorycnopsis is closely related to Anthyllis (Sokoloff, 2003a,b; see also Sokoloff, 2006).
Our nrITS data clearly show that Dorycnopsis gerardii cannot be placed in the genus Anthyllis.The molecular data are also consistent with the placement of Vermifrux into synonymy of Dorycnopsis.In our opinion, D. abyssinica and D. gerardii are morphologically so similar to each other (Table 3) that it is not reasonable to maintain two monotypic genera, Dorycnopsis and Vermifrux.It is obvious that characters such as leaflet morphology, flower number per umbel, calyx teeth and standard blade length are not significant at the generic level.Basic chromosome numbers of 2n = 14 and 2n = 12 co-occur within Loteae (e.g., within Lotus and Anthyllis).There exist some differences in pollen morphology between the two species of Dorycnopsis, but this difference does not exceed variation found among species of Anthyllis (see Díez & Ferguson, 1990, 1994).
The fruits of D. abyssinica are morphologically very different from those of D. gerardii.However, fruits of D. gerardii and D. abyssinica share the same anatomical structure (Sokoloff, 1997;Tikhomirov & Sokoloff, 1997).In details of fibre orientation pattern, fruits of Dorycnopsis are very different from those of Anthyllis.In tribe Loteae, only members of Ornithopus have the same fibre orientation pattern as found in Dorycnopsis   ( Degtjareva & al., 2003).According to molecular phylogenetic data, Dorycnopsis is more closely related to Ornithopus than to Anthyllis (Degtjareva & al., 2003;this study).Main differences between fruits of D. abyssinica and D. gerardii are seed number (two vs. one) and fruit curvature.It is necessary to stress that in both species there are two ovules in the ovary (Tikhomirov & Sokoloff, 1997).Thus the difference in seed number is not relevant for identifying phylogenetic relationships.In some groups of Loteae (e.g., the North American Syrmatium) one-and few-seeded fruits as well as straight and strongly curved fruits also co-occur.We believe that all this argues for the broad circumscription of Dorycnopsis as proposed here.
Sequence divergence data among species of Anthyllis and Dorycnopsis (Table 2) also support a broad circumscription of Dorycnopsis.Sequence divergence between the two species of Dorycnopsis is much smaller than the maximum sequence divergence among Anthyllis species.Many authors suggest to segregate members of subg.Cornicina (Anthyllis circinnata, A. hamosa, A. cornicina, and A. lotoides) as the separate genus Hymenocarpos (e.g., Lassen, 1986;Greuter & al., 1989;Benedí González, 1998).Maximum ITS sequence divergence among species of Anthyllis s.str.(excl.Hymenocarpos) is 18.2% (A.vulneraria vs. A. terniflora), i.e., more than between the two species of Dorycnopsis.The nrITS sequence of A. vulneraria strongly differs from all other Anthyllis species included in the present study.The sequence divergence data suggest that A. vulneraria should also be placed in a separate genus if Hymenocarpos is segregated from Anthyllis.However, this would not be the best choice from a nomenclatural point of view because A. vulneraria is the type species of Anthyllis.It is more logical to place Hymenocarpos into synonymy of Anthyllis.Our data (in contrast to those of Nanni & al., 2004) show close relationships between all species that were segregated in Hymenocarpos.However, in a strict consensus of all shortest trees (Fig. 2), these species do not form a clade.This is another argument for placing all these species in Anthyllis rather than in a separate genus Hymenocarpos.
using material from Yemen.Therefore, D. gerardii and D. abyssinica represent single terminals in our trees.The length of the ITS1 region ranges from 226 to 233 bp in Anthyllis.It is 236 bp long in D. gerardii and 226 bp long in D. abyssinica.The length of the ITS2 region ranges from 201 to 214 bp in Anthyllis.It is 221 bp long in D. gerardii and 213 bp long in D. abyssinica.The length of the 5.8 S region ranges from 164 to 165 bp in Anthyllis.It is 165 bp long in D. gerardii and 164 bp long in D. abyssinica.Pairwise comparisons of percentage sequence divergence of the entire ITS region among species of Anthyllis and Dorycnopsis are summarised in Table 2.The sequence divergence between D. abyssinica and D. gerardii is 11%.Sequence divergence among species of Anthyllis ranges between 0% (A.hermanniae vs. A. lagascana) and 19.1% (A.vulneraria vs. A. circinnata).Sequence divergence between species of Anthyllis and Dorycnopsis ranges between 21.6% (A.hamosa vs. D. abyssinica) and 38% (A.vulneraria vs. D. gerardii).

Table 2 .
Pairwise comparisons of percentage sequence divergence of the entire ITS region among species of Anthyllis and Dorycnopsis.D.aby D.ger A.ono A.ram A.tej A.aur A.mon A.bar A.her A.lag A.ter A.cyt A.vul A.lot A.cir A.cor A.ham