HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via ISI Web of Science (4) Citing Articles via Google Scholar Google Scholar Articles by Kenicer, G. J. Articles by Murata, J. Search for Related Content PubMed Articles by Kenicer, G. J. Articles by Murata, J. Agricola Articles by Kenicer, G. J. Articles by Murata, J.

Systematics and biogeography of Lathyrus (Leguminosae) based on internal transcribed spacer and cpDNA sequence data¹

²Botanical Gardens, Graduate School of Science, University of Tokyo, 3-7-1 Hakusan, Bunkyo, Tokyo 112-0001, Japan; ³Royal Botanic Garden Edinburgh, 20A Inverleith Row, Edinburgh EH3 5LR, UK

Received for publication September 8, 2004. Accepted for publication April 14, 2005.

ABSTRACT

Lathyrus (Leguminosae; Papilionoideae) is the largest genus in tribe Fabeae and exhibits an intriguing extratropical distribution. We studied the systematics and biogeography of Lathyrus using sequence data, from accessions representing 53 species, for the internal transcribed spacer plus 5.8S-coding region of nuclear ribosomal DNA as well as the trnL-F and trnS-G regions of chloroplast DNA. Our results generally supported recent morphology-based classifications, resolving clades corresponding to sections Lathyrus and Lathyrostylis, but question the monophyly of the large, widespread section Orobus sensu Asmussen and Liston. Sections Orobus, Aphaca, and Pratensis form a predominantly northern Eurasian–New World clade. Within this clade, the North American and eastern Eurasian species, including both Holarctic species (L. palustris and L. japonicus), form a transberingian clade of relatively recent origin and diversification. The South American Notolathyrus group is distant from this transberingian lineage and should be reinstated as a distinct section within the northern Eurasian–New World clade. The Notolathyrus lineage reached the New World most probably through long-distance dispersal from Eurasia. The remaining sections in the genus are centered on the Mediterranean region.

Key Words: Bering land bridge • biogeography • cpDNA • extratropical distribution • internal transcribed spacer • Lathyrus • Leguminosae • Notolathyrus

Lathyrus L. (the sweet peas) is the largest genus in the economically important tribe Fabeae (Adans.) DC. Members of Lathyrus include food and fodder crops, ornamentals, soil nitrifiers, dune stabilizers, important agricultural weeds, and model organisms for genetic and ecological research. Tribe Fabeae is nested within the "temperate herbaceous" papilionoid group of the Leguminosae (Gunn, 1969 ; Polhill, 1981 ; Wojciechowski et al., 2000 ).

The genus has a predominantly extratropical distribution pattern in both the northern and southern hemispheres (i.e., antitropical sensu Humphries and Parenti, 1999 ). Its approximately 160 species (Kupicha, 1983 ; Tsui, 1984 ; Asmussen and Liston, 1998 ) are distributed throughout the Northern Hemisphere, with their primary center of diversity in the seasonally dry Mediterranean basin and neighboring western Irano-Turanian region. Secondary centers of diversity are in North America and temperate areas of South America. A few species reach tropical East Africa and, as with those of the Neotropics, are typically found in more temperate (often montane) habitats. Most members of Lathyrus are mesophytes of open woodlands, forest margins, and roadside verges, but littoral, alpine, and more drought-tolerant species are also represented.

Most treatments arrange Lathyrus in 12 or 13 sections (Czefranova, 1971 ; Kupicha, 1983 ; Asmussen and Liston, 1998 ; International Legume Database and Information Service, 2002 ) (Fig. 1). Developing primarily the work of Bässler (1966 , 1973 , 1981 ), Ball (1968) , Davis (1970) , and Czefranova (1971) , Kupicha's 1983 sectional treatment is the most comprehensive of recent accounts. She attempted to reflect natural groups and to provide a functional framework for horticultural, agronomic, and biogeographic discussion. Subsequent studies have generally taken Kupicha's classification as a starting point and focused on limited taxonomic or regional groupings. Throughout this article, we have adopted her system as modified by Asmussen and Liston (1998) .

View larger version (50K): [in this window] [in a new window]   Fig. 1. Classifications of Lathyrus. Sections are shown in plain type, genera in boldface type. Hatched areas represent groups not studied. Adapted from Fig. 1 of Asmussen and Liston (1998) with permission

Sectional classifications of Lathyrus by Bässler (1966 , 1971 , 1973 , 1981 ), Czefranova (1971) , and Kupicha (1983) attempted to account for convergence in characters and possible reversal of character states. The groups defined by these authors are based on combinations of character states in which one or more states may be absent for some taxa within a group. Such reliance on a preponderance of shared characters rather than on diagnostic synapomorphies hinders the demonstration of sectional monophyly based on morphology. One aim of this paper is to assess the monophyly of Lathyrus sections based on DNA sequence data.

Five molecular phylogenetic studies of Lathyrus have been published: Asmussen and Liston (1998) , Croft et al. (1999) , Chtourou-Ghorbel et al. (2001) , Badr et al. (2002) , and Ben Brahim et al. (2002) . All except that of Asmussen and Liston are of limited geographic or taxonomic scope; they focus on Mediterranean taxa, particularly within section Lathyrus, a group containing many economically and ecologically important species.

The study by Asmussen and Liston (1998) , which mapped cpDNA restriction fragment length polymorphism (RFLP) data, is the largest molecular investigation to date. They sampled 42 western Eurasian and New World species, and their study included representatives of all but one section. Asmussen and Liston's results generally agree with the sectional classifications of Kupicha (1983) , but they suggested merging some groups: the monotypic sections Orobon and Orobastrum with section Lathyrus, and the South American section Notolathyrus with the Holarctic section Orobus (Fig. 2). Their results support the close relationship between sections Pratensis and Aphaca proposed by Kupicha (1974 , 1983 ) but were inconclusive for other sections.

View larger version (45K): [in this window] [in a new window]   Fig. 2. Asmussen and Liston's (1998) strict consensus of 18 500 equally parsimonious trees based on rpoC and inverted-repeat–negative-cpDNA RFLP data. Filled bars represent major length mutations. Open bars (in Notolathyrus) represent reversals of the same mutations at the base of the Orobus clade. The numbers above branches are percentage bootstrap values. Reproduced from Fig. 5 of Asmussen and Liston (1998) with permission

As a caveat to their conclusions, Asmussen and Liston (1998) emphasized the need for further research on section Orobus. Their study included only two exclusively Eurasian species from this section, and the key nodes representing the intercontinental disjunctions received low (<50%) bootstrap support. Wider sampling of Eurasian species from section Orobus using higher resolution molecular sources is necessary before any more robust biogeographic conclusions can be made.

Developing suggestions by Burkart (1937) and Bässler (1973) , Kupicha proposed that the South American section Notolathyrus derived from North American taxa (Kupicha, 1974 , 1983 ), but she did not suggest a possible sister group from among the extant North American species. Asmussen and Liston's results (1998) supported Kupicha's interpretation (Fig. 2). This pattern—of South American taxa derived from North American taxa—was interpreted as support for the idea, also implied by Bässler (1973) , that Lathyrus is a member of the temperate element successor to the mid-Tertiary boreotropical flora in the Northern Hemisphere (see Wolfe, 1975 ; McKenna, 1983 ). Asmussen and Liston (1998) viewed the Notolathyrus group as a Holarctic-derived element of the modern Andean flora (see van der Hammen and Cleef, 1985 ; Burnham and Graham, 1999 ). Unfortunately, the lack of any reliable fossil evidence for tribe Fabeae precludes confirmation of the sequence and timing of putative colonization events.

By providing a test that is relatively independent of morphological and RFLP-based classifications, DNA sequence data may help us to understand the systematics of this genus. Furthermore, investigating multiple genomes (e.g., nuclear and chloroplast) and regions facilitates recognition of patterns and processes such as homoplasy and hybridization events. We used sequence data from the nuclear ribosomal internal transcribed spacer (ITS) plus 5.8S-coding regions of nuclear ribosomal DNA, as well as the trnL intron plus flanking spacer (trnL-F) and trnS (GCU)-trnG (UCC) intergenic spacer (trnS-G) regions of cpDNA, to estimate the phylogeny of Lathyrus worldwide. The internal transcribed spacer has been used to reconstruct phylogenies at the species level in Lupinus L. (Aïnouche and Bayer, 1999 ), Lotus L. (Allan and Porter, 2000 ), and Lens L. (Mayer and Bagga, 2002 ); trnL-F has been used in Vicia L. (Fennell et al., 1998 ), Astragalus L. (Wojciechowski et al., 1999 ), and Genista L. (de Castro et al., 2002 ). The trnS-G region is less widely used but has proved informative in a sample set of Glycine subgenus Soja (Xu et al., 2001 ). This implies that the trnS-G region can resolve relationships in closely related taxa of legumes; an initial pilot survey in Lathyrus proved encouraging, showing similar levels of variation in ITS and trnS-G.

Sampling from within section Orobus received particular attention because of the group's wide distribution, morphological diversity, and important position within the genus.

MATERIALS AND METHODS

We obtained total genomic DNA from fresh or silica gel-dried leaf material from wild, cultivated, and herbarium specimens. Many of these specimens were donations, including most of the accessions used by Asmussen and Liston (1998) . Accessions representing 53 species of Lathyrus and four other members of tribe Fabeae (Vicia cracca L., V. nipponica J. Matsumura, V. unijuga A. Braun, and the Pisum sativum L. cv. Tomyo) were used (Appendix). Where possible, we sequenced multiple accessions of the same species for each region to verify identification and to investigate potential intraspecific variation, especially in wide-ranging and morphologically variable species. Although two accessions were sequenced for ITS and trnL-F from L. angulatus L. and L. sphaericus Retz., the trnS-G region was only obtained from one accession from each of these species.

DNA extraction followed the cetyltrimethylammonium bromide (CTAB) protocol of Doyle and Doyle (1987) . We ran polymerase chain reactions (PCR) as a 25-µL system with a final concentration of 0.1–0.2 µM forward and reverse primers. Primer pairs were universal: ITS plus 5.8S (from White et al., 1990 ); ITS5 (forward), ITS4 (reverse), and trnLF (from Taberlet et al., 1991 ); trnLc (forward), trnFf (reverse), and trnSG (from Hamilton, 1999 ).

Our PCR protocol followed a standard program, adjusted only for annealing temperatures. (1) Initial denaturation was conducted for 90 s at 95°C. (2) Thirty cycles of denaturation were conducted for 45 s at 95°C, annealing for 60 s at a temperature depending on the region, and elongation at 72°C. The elongation time began at 90 s and increased by 2 s per cycle. (3) A final elongation step of 15 min at 72°C was performed. Annealing temperatures were 51°C for ITS, 53–54°C for trnL-F, and 61°C for trnS-G. Some herbarium specimens produced low initial yields for the trnS-G region, in which case we did a second round of PCR.

The PCR product was purified using the Geneclean III kit (Bio101, Carlsbad, California, USA), using two-thirds volumes of reagents for economy but otherwise following the manufacturer's instructions.

Cycle sequencing used the same primers as those for PCR. The reverse reaction verified each nucleotide position. Lathyrus gmelinii Fritsch, L. littoralis (Nutt.) Endl., L. rigidus T. White, and L. setifolius L. needed internal primers, ITS2g and ITS3p (White et al., 1990 ). Thus the 5.8S-coding region was not completely sequenced for these three species.

The cycle sequencing reaction used the BigDye Terminator Cycle Sequencing kit, version 3.0 or 3.1 (Perkin Elmer, Foster City, California, USA) in an iCycler thermal cycler, version 1.280 (BioRad, Hercules, California, USA) with a total reaction volume of 5 µL. Initial denaturation (1 min at 96°C) was followed by 25 cycles of denaturation (10 s at 96°C), annealing (5 s at 50°C), and elongation (4 min at 60°C).

Samples were sequenced on an ABI Prism 377 automated sequencer (Perkin Elmer). Sequences were initially aligned using ClustalX (Thompson et al., 1997 ), then adjusted manually as Nexus-format text files. We used PAUP*, version 4.0b10 (Swofford, 2001 ) for phylogenetic analysis. The sequences were lodged with GenBank.

Where insertion–deletions ("gaps" or "indels") were shared by two or more taxa and could be aligned unequivocally, they were treated as potentially phylogenetically informative to maximize phylogenetic information. We coded such indels as a binary "1" or "0" state for all data matrices, following Simmons and Ochoterena (2000) and Simmons et al. (2001) . Each putative event was treated as independent of overlapping gaps (see Fig. 1 of Eriksson et al., 2003 ). Once indels were coded, all gap regions were excluded using the "exclude gapped" command in PAUP. Long indels sometimes spanned areas containing potentially phylogenetically informative nucleotide or indel sites in other sequences. In such cases, "N" coding was used in the gap-containing sequence to ensure the inclusion of that site in the analysis. Consequently, it is possible for the number of parsimony-informative sites to exceed the total number of sites minus gap-containing sites (Table 1). To eliminate the primer site and areas of poor read at the start and end of each sequence, we excluded these portions from the matrices. In the combined matrix, this corresponded to positions 1–49 (beginning of ITS), 687–803 (between ITS and trnL-F), 1521–1595 (between trnL-F and trnS-G), and 2708– 2797 (end of trnS-G). Where two accessions of the same species showed identical sequence data, we usually used only one in the sequence analysis.

View larger version (31K): [in this window] [in a new window]   Fig. 3. Strict consensus of 512 equally most parsimonious trees produced by the analysis of combined ITS, trnL-F, and trnS-G sequence data and gap characters. Numbers above branches are percentage bootstrap values. Sectional classifications according to Asmussen and Liston (1998) are shown at right; section Notolathyrus and Bässler's (1973) series Lutei and Verni are also shown. Letters below branches denote the clades identified in the Discussion

View larger version (30K): [in this window] [in a new window]   Fig. 4. A randomly chosen tree from the 512 equally most parsimonious trees produced by the analysis of combined ITS, trnL-F, and trnS-G sequence data and gap characters. Letters above branches denote the clades identified in the Discussion. Distributions are shown at right, and annual species are marked with an asterisk

Ingroup delimitation and outgroup choice
The combination of an adaxial pollen brush on the style and non-brochiododromous leaflet venation (i.e., veins reaching the margins) is considered synapomorphic for the Lathyrus– Pisum group, distinguishing it from Vicia (Kupicha, 1974 , 1981 ; Gunn and Kluve, 1976 ). Data from the matK gene support this arrangement (Steele and Wojciechowski, 2003 ). We included three Vicia species in the analysis as multiple outgroups to test the hypothesis that the Lathyrus–Pisum group is monophyletic.

RESULTS

Sequence characteristics
Table 1 summarizes sequence characteristics. The ITS and trnL-F sequences aligned readily, with insertion of a few gaps, while the trnS-G region contained more extensive indels.

Visual inspection of the strict consensus trees derived from individual analyses of ITS, trnL-F, and trnS-G (trees not shown) indicated many common clades, with differences reflecting lack of resolution rather than conflicting groupings with strong bootstrap support. The trnL-F and ITS phylogenies resolved few major clades, but several smaller groups such as the taxa corresponding to section Lathyrus and Bässler's series Verni (1973) were consistently resolved in each of the single-region trees (ITS, trnL-F, and trnS-G). Inclusion of gap characters generally bolstered support for clades present in analyses without gaps (data not shown).

The incongruence length difference test result (P = 0.1) indicates an acceptable degree of congruence among the three components of the combined data set (Farris et al., 1995 ; Johnson et al., 2001 ). These lines of evidence justified the combined analysis including gaps.

The combined parsimony analysis, with gaps included, produced 512 equally most parsimonious trees of 776 steps, with a consistency index of 0.558 (homoplasy index 0.442), retention index of 0.773, and rescaled consistency index of 0.432. Figure 3 shows the strict consensus of these trees. A single most parsimonious tree is shown in Fig. 4.

Phylogeny
The strict consensus tree (Fig. 3) is more resolved than that derived from the cpDNA RFLP analysis of Asmussen and Liston (1998) (Fig. 2). The Lathyrus–Pisum clade is well supported (100% bootstrap) as distinct from the species of Vicia included in our analysis. In Lathyrus, clades corresponding to section Lathyrus, section Lathyrostylis, the Notolathyrus group, and Bässler's series Verni received strong (99% or 100%) bootstrap support, as did a large clade within section Orobus (95%). A number of other clades were also resolved, with more limited support. Among these are two large clades (labeled A and B in Fig. 3) that are important in the interpretation of systematics and biogeography in section Orobus and allied taxa but do not correspond to any recognized entities in the genus.

When we analyzed multiple accessions of the same species, only L. vernus and L. palustris showed any variation. These represent two of the most widespread and variable species in western Eurasia and the Northern Hemisphere, respectively (Czefranova, 1971 ; Bässler, 1973 ).

DISCUSSION

Phylogeny and sectional classification
Our analysis confirms the generally accepted view that Lathyrus and Vicia are distinct (Kupicha, 1974 , 1983 ; Gunn and Kluve, 1976 ; Steele and Wojciechowski, 2003 ). Early classifications, including Species Plantarum and a revision by Roskov et al. (1998) , treated the etendrillous species of these two genera as a third, intermediate genus, Orobus; however, our data clearly refute this view.

When we included sections Nissolia, Clymenum, and Neurolobus in the analysis, bootstrap support was weak (<50%) for the separation of Pisum from Lathyrus. However, the two genera can be separated on the basis of stylar characters and leaflet ptyxis (Kupicha, 1981 ). The recent matK sequence analysis of Fabeae (as Vicieae) by Steele and Wojciechowski (2003) also distinguished these genera, but their study included only species from sections Lathyrus, Orobus, and Aphaca. Representatives of the other sections, particularly Nissolia, Clymenum, and Neurolobus, should be included in any future analyses of intergeneric relationships within the Fabeae.

Section Lathyrus
Our tree places L. tingitanus in a weakly supported (bootstrap <50%) position as sister to the rest of a well-supported (bootstrap 99%) section Lathyrus (Fig. 3), giving a similar topology to that retrieved by Badr et al. (2002) from amplified fragment length polymorphism (AFLP) data.

Lathyrus setifolius, formerly Kupicha's monotypic section Orobastrum, is sister to the delicate annual members of section Lathyrus, represented here by L. sativus L. and L. cicera L. This clade is sister to a second clade (96% bootstrap support) that is composed of perennials and a few robust annuals such as L. odoratus (Fig. 3). These results support those of Asmussen and Liston (1998) (Fig. 2), who were unable to decide whether to treat these two groups as separate sections. However, our sequence data draw together the results of previous RFLP and AFLP analyses and provide support for a circumscription of section Lathyrus in Asmussen and Liston's broader sense, containing both L. setifolius and L. tingitanus (Fig. 3).

Sections Lathyrostylis and Linearicarpus
Bässler's section Lathyrostylis is a morphologically uniform group of approximately 20 erect, perennial species. Members of Lathyrostylis share many characters with some species in section Orobus, which prompted Czefranova (1971) and many authors before her to treat them as part of a broader section (or genus) Orobus. Bässler claimed the two sections to be distinct, describing section Platystylis in 1966 (later renamed Lathyrostylis in 1971 and 1981). Kupicha (1983) agreed with his circumscription. The species of section Lathyrostylis we sampled form a monophyletic clade (100% bootstrap support) distinct from section Orobus, supporting Bässler's (1981) and Kupicha's (1983) interpretations (Fig. 3).

Our strict consensus tree (Fig. 3) shows L. sphaericus Retz. and L. angulatus L. of Kupicha's (1983) section Linearicarpus to be sisters to the section Lathyrostylis clade, although with low bootstrap support (76% and <50%, respectively). The morphometric analysis of Dogan et al. (1992) , as well as molecular data from Asmussen and Liston (1998) and Badr et al. (2002) , questioned the monophyly of section Linearicarpus, and our results appear to support this. However, further data for these difficult sections are needed before any firm systematic decisions can be made.

The Notolathyrus group
The South American Notolathyrus group forms a well-supported (100% bootstrap support) monophyletic clade, agreeing with morphological (Kupicha, 1983 ), cpDNA RFLP (Asmussen and Liston, 1998 ), and karyological (Senn, 1938 ; Seijo and Fernandez, 2003 ) results. In our study, the group's position is unresolved in the clade A polytomy (Fig. 3). Our results contradict those of Asmussen and Liston's 1998 study, in which Notolathyrus is nested deep within section Orobus; however, bootstrap support for their arrangement was low and invoked reversals of two large cpDNA structural mutations (Asmussen and Liston, 1998 ) (Fig. 2). We suggest that the group should be reinstated as Kupicha's section Notolathyrus until more conclusive evidence of its relationships can be found. Our sequence data do not suggest any sister group (Fig. 3); more complete sampling of the South American species and potential relatives would be useful. The trnS-G region of the South American L. pubescens Hook. et Arn. contained deletions that made it unalignable with the other taxa. However, analysis of the combined ITS and trnL-F regions alone (G. J. Kenicer, unpublished data) placed this species with the Notolathyrus group.

Section Orobus
Our results place all members of section Orobus in a clade with L. aphaca L. and L. laxiflorus (Desf.) Kuntze (Fig. 3, clade A) that receives only moderate bootstrap support (75%). The monophyly of section Orobus sensu Kupicha (1983) is equivocal, reflecting the wide range of morphological variation within the section. However, some clades within the section were strongly resolved.

One group of Eurasian species forms a well-supported clade (100% bootstrap) comprising Bässler's series Verni plus his monotypic series Tuberosi (L. linifolius) (Bässler, 1966 ) (Fig. 3). This group is equivalent to Czefranova's subsection Tuberosi plus L. alpestris (Czefranova, 1971 ). All species in this clade share similar, purple-flowered inflorescences; broad-deltoid lower calyx teeth; semisagittate, foliacious stipules; aristate rachises; and three parallel primary leaflet veins. However, these characters are also common in section Lathyrostylis, which made it difficult for Czefranova (1971) to separate the members of the series Verni plus Tuberosi clade from section Lathyrostylis, as previously discussed.

The remaining Eurasian members of section Orobus form a reasonably well-supported (91% bootstrap) "core Orobus group" with the North American species (Fig. 3). Within this group, clade B is composed of species that lack any clear morphological synapomorphies indicating subgroups. Because of this difficulty, Czefranova (1971) and Bässler (1973) were unable to suggest how the Eurasian members of this clade may be interrelated. Our results show the same problem: the taxa in this clade form a polytomy. Inclusion of the North American species further compounds this lack of resolution. However, members of clade B are undoubtedly closely related, all being eastern Eurasian, North American, or Holarctic (Fig. 4). The sister group to this clade plus its sister, Lathyrus aureus (Steven) Brandza, correspond to a morphologically uniform group of stocky, erect, yellow-flowered species from western and central Eurasia. This group was treated as a series, Lutei, by Bässler, although he also included the East Asian, purple-flowered L. vaniotii in this group. The ITS region failed to amplify in both our accessions of L. niger, using any combination of primers; we therefore excluded the species from the combined analysis. However, the strict consensus trees generated for trnS-G (with and without gaps) alone suggested L. niger was sister to L. aureus, although with only low bootstrap support (<50% data not shown).

Major gaps and inversions between the primers precluded satisfactory alignment of the trnS-G region for L. quinquenervius (Miq.) Litv. (East Asia) and L. delnorticus C. Hitchc. (North America). When they were included in the analysis of the combined ITS and trnL-F regions alone (G. J. Kenicer, unpublished data), results were as predicted from morphological and geographic affinities, as well as from the molecular data of Asmussen and Liston (1998) . Thus L. quinquenervius formed a clade with the East Asian accessions of L. palustris, and L. delnorticus was in clade B.

Sections Aphaca and Pratensis
In our combined analysis, section Aphaca (one or two species) is represented by L. aphaca and section Pratensis (approximately six species) by L. laxiflorus. Members of these two sections share a distinctive wing-petal architecture and sagittate stipules supplied by an unusual vascular arrangement. Such robust synapomorphies are rare in Lathyrus and prompted Kupicha (1974 , 1975 , 1983 ) to suggest that the two sections are closely related, a proposal supported by cpDNA RFLP data (Asmussen and Liston, 1998 ) (Fig. 2) and our sequence data (Fig. 3). Despite this, we agree with the conclusion of Kupicha and of Asmussen and Liston: these sections should be retained as separate because of their morphological distinctiveness. Suspected deletions or other structural rearrangements prevented amplification of the trnS-G region for L. pratensis L., but it appeared as sister to L. laxiflorus in the combined analysis of trnL-F and ITS.

Sections Clymenum, Neurolobus, and Nissolia
The relationships among these taxa receive little bootstrap support, although they are consistently resolved in the component trees of the strict consensus.

Lathyrus neurolobus and L. nissolia are sister taxa and are collectively sister group to the rest of Lathyrus. Kupicha (1983) placed them in monotypic sections based on morphology and habit and suggested that they both occupy isolated positions within the genus. Our sequence data support this view, and the branches leading to each of the two taxa have many autapomorphies (Fig. 4).

Lathyrus articulatus L. is often considered to be synonymous with L. clymenum, and this pairing is sister to L. gloeospermus Warb. et. Eig. (Fig. 3), all of which are traditionally placed in section Clymenum on the basis of their phyllodic leaves. Our finding contrasts with the weighted RFLP data analysis of Asmussen and Liston (1998) , which placed L. gloeospermus distant from other members of section Clymenum. We suggest that L. gloeospermus should be provisionally retained in section Clymenum because of nomenclatural stability and ease of diagnosis, although this does not reflect monophyly.

Suspected deletions or other structural rearrangements prevented amplification of the trnS-G region for L. ochrus (L.) DC. (also section Clymenum), although it appeared as sister to L. clymenum in a combined analysis of trnL-F and ITS.

Biogeography: Northern Hemisphere
Section Orobus had been thought of as the most unspecialized group in Lathyrus because of its perennial life cycle, the relative complexity of leaves (multijugate and often tendrillous), several or many-flowered inflorescences, and northern Eurasian (i.e., mesophytic) distribution (Simola, 1968 ; Bässler, 1973 ; Kupicha, 1974 , 1983 ). Kupicha proposed that Lathyrus arose from perennial taxa similar to section Orobus in northern Eurasia during the early Tertiary (Kupicha, 1974 , 1983 ). A later "southward shift of emphasis in the evolution of Vicia and Lathyrus" (Kupicha, 1974 ) occurred as the Mediterranean phytochorion developed.

Our tree topologies contradict Kupicha's hypothesis because the Mediterranean taxa (L. neurolobus, L. nissolia, section Clymenum, and Pisum) appear at the base of the tree (Fig. 4). Except for L. neurolobus, these species are annuals adapted to the seasonally dry climate. This pattern repeats in the predominantly Mediterranean clade containing sections Lathyrostylis, Linearicarpus, and Lathyrus. In this clade, the modern perennial lineages occupy seemingly derived positions because they are nested within groups of annual species. Morphometric (Dogan et al., 1992 ), RFLP (Asmussen and Liston, 1998 ), and AFLP (Badr et al., 2002 ) studies showed similar results, with annual species as sisters to perennial clades.

Using dated phylogenies based on rbcL and matK data, Lavin and colleagues recently estimated the origin of the Fabeae to be approximately 17.5 million years ago (Mya; mid Miocene) (Lavin et al., in press ). This estimate disagrees with Kupicha's early Tertiary origin for Lathyrus (Kupicha, 1983 ). Our ITS data provide independent support for Lavin and coworkers' more recent origin. We calculated mean substitutions per site across the Lathyrus–Pisum clade using the program DNAsp 3.53 (Rozas and Rozas, 2001 ), and in the absence of any fossil data for calibration, we used absolute substitution rates for ITS from previously published studies of herbaceous legumes with similar life histories. Based on the rates given for the combined ITS1 and ITS2 regions in Astragalus (Wojciechowski et al., 1999 ), the Lathyrus–Pisum crown group is estimated at 5.4–6.3 Mya, while the rates for Lupinus (Käss and Wink, 1997 ) give an estimate of 6.4–8.2 Mya (ITS1) and 3.5–4.3 Mya (ITS2). It thus seems likely that the modern diversity of Lathyrus stems from relatively recent radiations.

In the light of this timing, our results require a different biogeographic explanation from that proposed by Kupicha (1974 , 1983 ). We suggest that Lathyrus originated in the eastern Mediterranean region during the mid- to late Miocene rather than dispersing into this area from northern Eurasian Eocene or Oligocene lineages, as Kupicha proposed. From the late Miocene, the establishment of the modern, seasonally xeric climatic rhythm (Suc, 1984 )—coupled with major tectonic upheavals, including the Alpine, Caucasus, and Zagros mountain orogenies (Meulencamp and Sissingh, 2003 )—would have acted as major engines for evolution, with xerophytic annuals being selected in the eastern Mediterranean and mesophytic perennial lineages in the incipient mountains and northern forests of the Euro-Siberian region (Ramstein et al., 1997 ). In our strict consensus tree, at least four independent transitions from annuality to perenniality are inferred: in L. neurolobus, section Lathyrus, section Lathyrostylis, and the clade A lineages (Fig. 4). Accompanying increases in complexity of leaf form and inflorescence would also have been necessary. Alternatively, as Kupicha (1974 , 1983 ) suggested, if Lathyrus originated as mesophytic perennials, extensive extinctions in the basal lineages would be required to explain the topology seen in our trees.

The North American–East Asian polytomy (Fig. 3, clade B) shows no evidence for a monophyletic origin of the North American species. Several morphologically distinct lineages exist within the clade (S. Broich, Department of Human Services, State Government, Oregon, personal communication). Kupicha's proposal that modern North American taxa derived from a "primitive ancestral stock [from Eurasia] having characteristics of sect. Orobus" (Kupicha, 1983 , p. 242) is well supported. However, her estimate of a Cretaceous or early Tertiary colonization of the New World (Kupicha, 1974 ) is clearly awry. The transberingian distribution of this complex reflects its probable center of diversification and suggests that the Bering land bridge was the main route by which taxa have been exchanged between the two continents.

Biogeography: South America
Burkart (1966) and Kupicha (1983) suggested that the South American species of Lathyrus dispersed into the region from North America via the Andes, a scenario supported by Asmussen and Liston (1998) . Taking the opportunity presented by cooler historical climates, the closure of the isthmus of Panama, and the incipient Andean orogeny (Burnham and Graham, 1999 ), these lineages were thought to have dispersed over land. However, our sequence data place the Notolathyrus clade clearly outside the transberingian clade B, which contains all the extant North American species sampled for this study (Figs. 3 and 4). This suggests other scenarios for the colonization of South America.

Transoceanic dispersal
The cladogram topology (Fig. 3) and age estimates for Lathyrus suggest that the most likely scenario is direct, long-distance dispersal of taxa to South America from Eurasia, most probably as sea-drifted seeds. Burkart (1937) claimed that the coastal Chilean populations of L. japonicus are relatively recent seaborne arrivals. Although L. japonicus is distant from the Notolathyrus group in our tree, it illustrates that such dispersal between the Northern and Southern hemispheres is possible. However, the sea-dispersal potential of seeds from species in section Notolathyrus is untested.

A relationship between the South American clade and Eurasian species is also suggested by morphology. The South American taxa are morphologically more similar to taxa outside section Orobus than they are to those in clade B (Kupicha, 1974 , 1983 ), most notably the members of section Pratensis. Indeed, L. pusillus Elliott shares striking similarities in stipule and leaf vasculature with L. pratensis and L. laxiflorus, although this may be a parallelism (Kupicha, 1974 ). Interestingly, the South American species of Vicia are also thought to be morphologically closer to western Eurasian species than they are to most North American ones, although V. americana Muhl. may provide a link (Kupicha, 1974 , 1976 ). This relationship is corroborated by stylar morphology (Endo and Ohashi, 1997 ) and matK data (Steele and Wojciechowski, 2003 ), but such evidence is still patchy and a targeted study of the phenomenon remains to be undertaken. Given the striking similarities in evolutionary history between Vicia and Lathyrus, this is an intriguing area for future investigation of the two genera.

Extinction of North American lineages
An alternative hypothesis necessitates extra inferences of extinction. Under this scenario, an early lineage dispersed from Eurasia into North America, with subsequent dispersal into South America (possibly via the temperate Andes), followed by extinction of the lineage in North America. A similar scenario has been proposed for the Eurasian–South American disjunction in Chrysosplenium (Saxifragaceae), a group in which long-distance dispersals are thought to have been unlikely (Soltis et al., 2001 ). Testing this possibility in the absence of Lathyrus fossils is difficult.

Sampling artifact
Although the North American species not sequenced in our study may prove to contain relatives of section Notolathyrus, this seems unlikely based on morphology. The only exception to this is L. pusillus, a species that is well characterized as a member of the Notolathyrus group (Kupicha, 1983 ; Seijo and Fernandez, 2003 ; R. Ridley, unpublished data), but that is found throughout temperate South America (often as L. crassipes Gillies) and disjunctly in the southwestern United States. The remaining North American taxa form a morphologically uniform group of taxa similar to those in clade B (Fig. 4). They have multiple leaflets with reticulate venation and foliolaceous, semisagittate stipules. Although the South American species show a greater range of morphological variation, almost all are unijugate with parallel veins and many species have sagittate stipules (including L. pusillus). In addition, the broad styles with sometimes bifid stigmas and the lanceolate lower calyx teeth found in some South American taxa are not found in North American taxa, but are characteristic of some west Eurasian species.

Conclusions
Current sectional classifications of Lathyrus are generally well supported by our sequence data. Most sections in Kupicha's genus-wide classification, as modified by Asmussen and Liston (1998) , are monophyletic. Therefore Kupicha and her contemporaries' sectional classification based on character suites remains the most satisfactory and convenient way to treat the patterns of diversity in Lathyrus.

In contrast to the conclusions of Asmussen and Liston (1998) , our results suggest that the South American Notolathyrus group should be retained as the section proposed by Kupicha (1983) . Also, studies based on chloroplast RFLP (Asmussen and Liston, 1998 ) and AFLP (Badr et al., 2002 ) and our sequences (Fig. 3) differ on the relative positions of L. nissolia, L. neurolobus, and sections Clymenum and Linearicarpus. These taxa are evidently phylogenetically difficult and offer intriguing possibilities for further investigation.

Our DNA sequence data resolve the discrepancies in classification of section Orobus between Czefranova (1971) and Bässler (1966 , 1973 ). The lineages within section Orobus show several patterns of dispersal and diversification and include monophyletic groups corresponding to Bässler's series Verni (if L. linifolius is included), and a transberingian group of North American–East Asian species. The position of section Notolathyrus relative to section Orobus remains uncertain but appears likely to be derived from Eurasian rather than from extant North American lineages.

Lathyrus has evidently been profoundly influenced by climatic and tectonic changes in the Mediterranean, events that are responsible for the great diversity of species in the region. The Mediterranean annual taxa may hold the key to understanding the relationships between Lathyrus and Pisum as well as those among the difficult sections Lathyrus and Lathyrostylis. Section Orobus and the Pratensis–Aphaca group appear to represent a northern lineage distinct from the Mediterranean species and with a complex biogeographic history. Future studies should certainly aim to use broad sampling from both these geographical groups whenever possible.

Further investigation of the poorly resolved nodes within the section Orobus clade will provide important insights into the interrelationships of each of these lineages and consequently the intercontinental biogeography of the genus.

Appendix. Accessions of Fabaceae sequenced for nuclear internal transcribed spacer and chloroplast trnL-F and trnS-G regions. Sectional classification and species numbers are from Kupicha (1983) with modifications based on Hara and Williams (1979), Nelson and Nelson (1983), Tsui (1984) , Broich (1986), Zhu and Meng (1986), Maxted and Goyder (1988), Iseley (1992), and Asmussen and Liston (1998) . General species distributions are shown as well as more specific localities for wild-collected specimens. Numbers in boldface are author's accession numbers. Locations of voucher specimens by source: Asmussen & Liston, University of Aarhus, Denmark; J.-Y. Lee, Inha University, Incheon, South Korea; S. Norton, National Council for the Conservation of Plants and Gardens Lathyrus collection, West Wickham, Cambridgeshire, UK; G. Kenicer, RBGE, Herbarium, Royal Botanic Garden Edinburgh, Edinburgh, UK. GenBank accession numbers are given under each of the regions;—designates a region unable to be sequenced or not alignable.

Genus

Section (no. spp.); Operational taxonomic unit; Distribution (wild origin collection locality); Sources, accessions and vouchers; GenBank accession numbers: ITS; trnL-F; trnS-G.

Lathyrus

Aphaca (2); L. aphaca L.; Europe & Mediterranean (Turkey); 118 S. Norton 1991-008; AY839345; AY839413; AY839489.

Clymenum (3/4); L. ‘articulatus’ L. = clymenum?; Mediterranean; 1 Asmussen & Liston; AY839346; AY839414; AY839528. L. clymenum L.; Mediterranean; 2 Asmussen 1994-2; AY839349; AY839417; AY839529. L. gloeospermus Warb. & Eig.; Mediterranean; 3 Asmussen 1994-3; AY839356; AY839424; AY839527. L. ochrus (L.) DC; Mediterranean; 4 Asmussen 1994-4; AY839376; AY839444;—.

Lathyrostylis (20); L. digitatus (M. Bieb.) Fior.; S.E. Europe (Crimea); 6 Asmussen 1994-5; AY839352; AY839420; AY839514. L. digitatus (M. Bieb.) Fior.; S.E. Europe (Turkey); 129 RBGE Lampinen 7635; AY839353; AY839421;—. L. pallescens (M. Bieb.) K. Koch; N.E. Mediterranean (Turkey); 124 RBGE D.M. Brown 572; AY839378; AY839445; AY839515. L. filiformis (Lam.) Gay; S.W. Europe (Spain); 133 RBGE Brummit, Gibbs & Ratter 670; AY839354; AY839422; AY839536. L. spathulatus Celak.; Turkey (Turkey); 125 RBGE Coode & Jones 1200; AY839392; AY839459; AY839513.

Lathyrus (34); L. annuus L.; Mediterranean; 8 Asmussen 1994-07; AY839344; AY839412; AY839525. L. cicera L.; Mediterranean; 9 Asmussen 1994-08; AY839348; AY839416; AY839518. L. ‘heterophyllus’ L. = latifolius?; C. Europe; 12 Asmussen & Liston (cult. Oregon); AY839358; AY839425; AY839532. L. odoratus L.; Sicily (Garden origin); 16 Asmussen 1994-14; AY839377; AY839474; AY839533. L. rotundifolius Willd.; W. Asia (Turkey); 89 S. Norton 2000-443; AY839388; AY839455; AY839535. L. sativus L.; E. Mediterranean to Iran; 18 Asmussen 1994-15; AY839389; AY839456; AY839517. L. setifolius L.; S. Europe; 27 Asmussen 1994-27; AY839391; AY839458; AY839516. L. sylvestris L.; Europe; 19 Asmussen 1994-16; AY839398; AY839465; AY839523. L. tingitanus L.; North Africa; 20 Asmussen 1994-17; AY839399; AY839466; AY839519. L. tuberosus L.; Europe; 21 Asmussen 1994-18; AY839401; AY839468; AY839524.

Linearicarpus (7); L. angulatus L.; Mediterranean; 23 Asmussen 1994-19; AY839342; AY839410; AY839522. L. angulatus L.; Mediterranean (Portugal); 132 RBGE Sales & Hedge 96/12; AY839343; AY839411;—. L. sphaericus Retz.; Europe; 24 Asmussen 1994-20; AY839393; AY839460; —. L. sphaericus Retz.; Europe (Cyprus); 122 RBGE Edmondson & McClintock 2847; AY839394; AY839461; AY839512.

Neurolobus (1); L. neurolobus Boiss. & Heldr.; (endemic to Crete); 25 Asmussen 1994-21; AY839373; AY839440; AY839521.

Nissolia (1); L. nissolia L.; W. and C. Europe; 26 Asmussen 1994-22; AY839375; AY839443; AY839520.

Pratensis (6); L. laxiflorus (Desf.) Kuntze; S. and E. Europe; 29 Asmussen 1994-47; AY839367; AY839434; AY839482. L. pratensis L.; W. Eurasia; 30 Asmussen 1994-48; AY839384; AY839451;—.

Orobus (50; Eurasia 21); L. alpestris (Waldst. & Kit.) Kit.; Balkans (Slovenia); 99 S. Norton 2001; AY839341; AY839409; AY839476. L. aureus (Steven) Brandza; E. Europe; 100 S. Norton 1992-122 (cult. Cambridge); AY839347; AY839415; AY839481. L. davidii Hance; E. Asia (S. Korea); 79 Kenicer-38 (Kangwando Province); AY839350; AY839418; AY839492. L. gmelinii Fritsch; C. Asia (SW Siberia); 113 RBGE Elias, Shetler & Murray 7468; AY839357; AY839475; AY839537. L. humilis (Ser.) Sprengel; C. and E. Asia (Mongolia); 128 RBGE R.J. Allen 1008; AY839360; AY839427; AY839494. L. japonicus Willd.; Worldwide (Korea); 135 Kenicer-52 (Kangwando Province); AY839361; AY839428; AY839495. L. komarovii Ohwi; C. and E. Asia (Altai); 120 RBGE Elias, Shetler & Murray 7091; AY839363; AY839430; AY839478. L. laevigatus subsp. laevigatus (Waldst & Kit.) Kit.; C. Europe (Dolomites); 101 S. Norton; AY839364; AY839431; AY839497. L. laevigatus subsp. laevigatus (Waldst & Kit.) Kit.; C. Europe (Slovenia); 102 S. Norton 1997-583; AY839365; AY839432; AY839498. L. linifolius (Reich.) Bässler; N. Europe; 48 Asmussen & Liston (cult. Oregon); AY839368; AY839435; AY839477. L. niger (L.) Bernh.; W. Eurasia (Hungary); 50 Asmussen 1994-43;—; AY839442; AY839488. L. palustris L.; Northern Hemisphere; 51 Asmussen 1994-44; AY839379; AY839446; AY839502. L. palustris subsp. pilosus; Northern Hemisphere (S. Korea); 107 Ji.-Y. Lee (Kangwando Province); AY839380; AY839447; AY839503. L. palustris subsp. pilosus; Northern hemisphere (Japan); 108 Kenicer-61 (Hakone BG, from Sapporo); AY839381; AY839448; AY839504. L. pisiformis L.; W. Eurasia (Kirgizistan); 116 RBGE Maxted & Sperling 8251; AY839382; AY839450; AY839491. L. quinquenervius (Miq.) Litv.; E. Asia (S. Korea); 109 Ji.-Y. Lee (Inha University); AY839386; AY839453;—. L. transsylvanicus (Sprengel) Reichenb. f.; E. Europe; 110 S. Norton 1992-212; AY839400; AY839467; AY839509. L. vaniotii Léveillé; E. Asia (S. Korea); 117 Kenicer-30 (Kangwando Province); AY839402; AY839469; AY839511. L. vernus (L.) Bernh.; W. Eurasia; 54 Asmussen & Liston (cult. Oregon); AY839403; AY839470; AY839479. L. vernus (L.) Bernh.; W. Eurasia; 85 RBGE living collection (19881239); AY839404; AY839471; AY839480.

Orobus (N. America 29); L. delnorticus C. Hitchc.; Oregon; 33 Asmussen 1994-29; AY839351; AY839419;—. L. glandulosus Broich; N. California; 34 Asmussen & Liston; AY839355; AY839423; AY839493. L. holochlorus (Piper) C. Hitchc.; (Oregon); 35 Asmussen & Liston; AY839359; AY839426; AY839490. L. jepsonii E. Greene; N.W. USA; 36 Asmussen 1994-40; AY839362; AY839429; AY839496. L. lanszwertii Kellogg; N.W. USA (Oregon); 37 Asmussen & Liston; AY839366; AY839433; AY839499. L. littoralis (Nutt.) Endl.; N.W. USA (Oregon); 38 Asmussen 1994-42; AY839369; AY839436; AY839500. L. nevadensis S. Watson; N.W. USA (Oregon); 39 Asmussen 1994-41; AY839374; AY839441; AY839501. L. polyphyllus Nutt.; N.W. USA (Oregon); 40 Asmussen 1994-45; AY839383; AY839449; AY839505. L. rigidus T. White; N.W. USA; 41 Asmussen & Liston; AY839387; AY839454; AY839506. L. splendens Kellogg; S. Californian endemic; 98 S. Norton 1992-245 (cult. Cambridge); AY839395; AY839462; AY839507. L. sulphureus Brewer; N.W. USA; 43 Asmussen 1994-46; AY839397; AY839464; AY839508. L. vestitus Nutt.; N.W. USA; 44 Asmussen & Liston; AY839405; AY839472; AY839510.

Notolathyrus (23); L. multiceps D. Clos.; S. America (Chile); 31 RBGE-19912326 (Kirkpatrick 377); AY839370; AY839437; AY839484. L. magellanicus Lam.; S. America (Chile); 94 S. Norton 1999-684; AY839371; AY839438; AY839485. L. subandinus Philippi; S. America (Chile); 95 S. Norton 1999-676; AY839396; AY839463; AY839483. L. nervosus Lam.; S. America (Uruguay); 32 Asmussen 1994-24; AY839372; AY839439; AY839486. L. pubescens Hook. & Arn.; S. America (Chile); 96 S. Norton 1992-247; AY839385; AY839452;—L. hookeri G. Don.; S. America (Chile); 97 S. Norton 1999-679; AY839390; AY839457; AY839487.

Pisum

Pisum; Pisum sativum L.; Pantemperate crop (cultivated); 136 Kenicer-65 (cult. Tokyo University BG); AY839340; AY839473; AY839526.

Vicia

Cracca (40); V. cracca L.; Eurasia (Denmark); 57 Asmussen 1994-49; AY839339; AY839406