A molecular phylogeny and classification of Leptochloa (Poaceae: Chloridoideae: Chlorideae) sensu lato and related genera

Abstract

Background and Aims

Leptochloa (including Diplachne) sensu lato (s.l.) comprises a diverse assemblage of C₄ (NAD-ME and PCK) grasses with approx. 32 annual or perennial species. Evolutionary relationships and a modern classification of Leptochloa spp. based on the study of molecular characters have only been superficially investigated in four species. The goals of this study were to reconstruct the evolutionary history of Leptochloa s.l. with molecular data and broad taxon sampling.

Methods

A phylogenetic analysis was conducted of 130 species (mostly Chloridoideae), of which 22 are placed in Leptochloa, using five plastid (rpL32-trn-L, ndhA intron, rps16 intron, rps16-trnK and ccsA) and the nuclear ITS 1 and 2 (ribosomal internal transcribed spacer regions) to infer evolutionary relationships and revise the classification.

Key results

Leptochloa s.l. is polyphyletic and strong support was found for five lineages. Embedded within the Leptochloa sensu stricto (s.s.) clade are two Trichloris spp. and embedded in Dinebra are Drake-brockmania and 19 Leptochloa spp.

Conclusions

The molecular results support the dissolution of Leptochloa s.l. into the following five genera: Dinebra with 23 species, Diplachne with two species, Disakisperma with three species, Leptochloa s.s. with five species and a new genus, Trigonochloa, with two species.

INTRODUCTION

Leptochloa P.Beauv. sensu lato (s.l.) (including Diplachne P.Beauv.) comprises a diverse assemblage of C₄ [nicotinamide adenine dinucleotide co-factor malic enzyme (NAD-ME) and phosphoenolpyruvate carboxykinase (PCK)] grasses with approx. 32 annual or perennial species (Snow, 1997, 2003). The native range of the genus is pantropical into warmer temperate regions, with several species being weedy and widely distributed. The diversity of taxa in Leptochloa increases towards the tropics on all continents. The richest areas include: Louisiana, Texas and western Mexico in North America; Ethiopia through Tanzania in Africa; the Chacó in South America; and eastern Queensland in Australia (Snow, 1997). Asia has relatively few taxa and there are only three wide-ranging species reported for China (Chen and Phillips, 2006). Evolutionary relationships and a modern classification of Leptochloa spp. based on the study of molecular characters have been investigated only superficially.

The range of morphological variation within many Leptochloa spp. is significant, but they can be characterized as having membranous ligules, flat leaf blades, a paniculate inflorescence of spicate racemes with the branches inserted alternately to sub-whorled (sub-digitate), appressed pedicels, secund spikelets that are generally in two distinct rows, disarticulation above the single-nerved glumes and membranous lemmas that are three- (rarely four- or five-) nerved with various pubescence patterns. The presence or absence of a prominent sulcus and adnation of the pericarp in the caryopsis may possibly be of phylogenetic utility in Leptochloa spp. (Snow, 1998a), whereas micromorphological features of the lemma apparently vary little among these species (Snow, 1996).

Considerable controversy has surrounded the generic placement of Leptochloa spp. since Palisot de Beauvois (1812) described it and Diplachne in the same publication. There was even confusion regarding the use of Leptochloa until Niles and Chase (1925) designated a lectotype [based on L. virgata (L.) P.Beauv.] that effectively stabilized the nomenclature. Many genera have been erected to accommodate the species that now reside in Leptochloa, most notably including: Diplachne [based on L. fusca subsp. fascicularis (Lam.) N.Snow], Oxydenia Nutt. [based on L. panicea subsp. brachiata (Steud.) N.Snow] and Disakisperma Steud. [based on L. dubia (Kunth) Nees].

The lack of monophyly of Leptochloa s.l. has been demonstrated in multiple studies, including some based on morphological and anatomical data (Snow, 1997). Based on a numerical analysis of morphological characters, Phillips (1982) found that species placed in Diplachne and Leptochloa overlapped with one another in a principal coordinates scatter plot.

The monophyly of Leptochloa has not been tested in breadth using molecular data, and the relationship between Leptochloa and other genera is far from certain (Columbus et al., 2007; Peterson et al., 1997, 2007, 2010a). For example, in a restriction fragment analysis of plastid DNA using a few New World representatives, Leptochloa (L. dubia) did not form a particularly strong clade with the other genera, and instead shared a common ancestor in a clade that included the Muhlenbergiinae, Dasyochloa Willd. ex Rydb, Munroa Torr., Erioneuron Nash, Scleropogon, Sporobolus R.Br, Eleusine Gaertn., Tridens Roem. & Schult., Tripogon Roem. & Schult. and Eustachys Desv. (Duvall et al., 1994). In a phylogenetic study of Chloridoideae based on matK sequences, Leptochloa (L. dubia) formed a clade with Coelachyrum Hochst. & Nees, and Astrebla F.Muell. emerged as sister to these (Hilu and Alice, 2001). Other genera in Hilu and Alice's (2001) C1 clade included Brachyachne (Benth.) Stapf, Chloris Sw., Cynodon Rich., Dinebra Jacq., Eleusine, Enteropogon Nees, Eustachys, Lepturus R.Br., Lintonia Stapf, Microchloa R.Br., Oxychloris Lazarides, Tetrapogon Desf. and Trichloris E.Fourn. & Benth. In another phylogenetic study of Chloridoideae, based on analysis of combined trnL-F and internal transcribed spacer (ITS) sequences, Leptochloa spp. [L. dubia, L. fusca (L.) Kunth and L. panicea (Retz.) Ohwi] appear in separate clades, suggesting a polyphyletic origin (Columbus et al., 2007). In a phylogenetic analysis of rps16 sequences Leptochloa (L. dubia) forms a clade with Eleusine [E. coracana (L.) Gaertn.], whereas in a tree based on analysis of waxy sequences, Leptochloa forms a clade with Eleusine (E. coracana) and Dactyloctenium Willd. [D. aegyptium (L.) Willd. and D. radulans (R.Br.) P.Beauv.] (Ingram and Doyle, 2007).

In a large phylogenetic study of Chloridoideae based on seven DNA sequence markers (ITS, ndhA intron, ndhF, rps16-trnK, rps16 intron, rps3 and rpl32-trnL) Leptochloa was clearly polyphyletic, representing three separate lineages (Peterson et al., 2010a). Leptochloa viscida (Scribn.) Beal was sister to L. filiformis (Pers.) P.Beauv. (= L. panicea subsp. brachiata) and Dinebra retroflexa (Vahl) Panz. in one strongly supported clade [bootstrap support (BS) = 90, posterior probability (PP) = 1·00]; L. dubia was sister to Sclerodactylon mascrostachyum (Benth.) A.Camus. in another clade (PP = 0·59); and L. uninervia (J.Presl) Hitchc. & Chase was sister (PP = 0·81) to all remaining members of subtribe Eleusininae (Peterson et al., 2010a, fig. 3). In summary, the monophyly of Leptochloa has not been corroborated with molecular data, and phylogenetic relationships of Leptochloa with other genera remain obscure (Columbus et al., 2007; Peterson et al., 2007, 2010a).

Fig. 3.

Phylogram of maximum-likelihood tree from analysis of combined plastid and ITS sequences. Numbers above branches represent bootstrap values; numbers below branches represent posterior probabilities; colour indicates native distribution (see legend); roman numerals indicate clades discussed in the text and correspond to recognized genera: I = Trigonochloa, II and III = Dinebra, IV = Diplachne, V = Leptochloa and VI = Disakisperma. Scale bar = 10 % sequence divergence.

Using an analysis of plastid and nuclear DNA sequences, we present a new phylogenetic analysis for 28 of the 32 species that occur in Leptochloa. We estimate the phylogeny among the members of Leptochloa based on the analysis of six molecular markers (nuclear ITS and plastid rpL32-trn-L, ndhA intron, rps16 intron, rps16-trnK and ccsA DNA sequences). Peterson et al. (2010a) considered four species of Leptochloa in their large study investigating 246 species of Chloridoideae using seven molecular markers. We include an expanded survey of the subtribe Eleusininae by sampling an additional 18 species of Leptochloa for these six markers (314 new sequences), which is a significant advance over Peterson et al. (2010a). We compare the ITS and plastid-based phylogenetic trees with previous classifications by Clayton and Renvoize (1986), Watson and Dallwitz (1992), Snow (1997), Hilu and Alice (2001), Columbus et al. (2007) and Peterson et al. (2010a) (see Table 1 for a comparison). In addition, we seek morphological and anatomical characters supporting relationships in the molecular trees and propose changes to the classification.

Table 1.

Comparison of recent classifications of the genera indicating tribe and subtribe placement

Taxon	Clayton and Renvoize (1986)	Watson and Dallwitz (1992)	Snow (1997)	Hilu and Alice (2001)	Columbus et al. (2007)	Peterson et al. (2010a)	This paper
Dinebra Jacq.	Eragostideae, Eleusininae; 3 spp.	Chlorideae; 3 spp.	not treated	Eragrostideae, Eleusininae	Cynodonteae	Cynodonteae, Eleusininae	Chlorideae, Eleusininae; 23 spp. (incl. Drake-brockmania)
Diplachne P. Beauv.	–	Chlorideae; 18 spp.	–	Eragrostideae, Eleusininae	–	Cynodonteae, Eleusininae	Chlorideae, Eleusininae; 2 spp.
Disakisperma Steud.	–	–	–	–	–	–	Chlorideae, Eleusininae; 3 spp.
Drake-brockmania Stapf	Eragrostideae, Eleusininae; 2 spp.	Chlorideae; 2 spp.	not treated	not treated	not treated	unplaced in subfamily	placed in Dinebra
Leptochloa P. Beauv.	Eragrostideae, Eleusininae; 40 spp. (incl. Diplachne)	Chlorideae; 27 spp.	Cynodonteae, Eleusininae; 32 spp. (incl. Diplachne)	Eragrostideae, Eleusininae	Cynodonteae (incl. Diplachne)	Cynodonteae, Eleusininae	Chlorideae, Eleusininae; 5 spp. (incl. Trichloris)
Trichloris E. Fourn. ex Benth.	Cynodonteae, Chloridinae; 2 spp.	placed in Chloris	not treated	not treated	Cynodonteae	Cynodonteae, Eleusininae	placed in Leptochloa
Trigonochloa P.M. Peterson & N. Snow	–	–	–	–	–	–	Chlorideae; 2 spp.

The number of species of each genus is included if known. A dash indicates a taxon was not recognized.

MATERIALS AND METHODS

Taxon sampling

The taxon sampling consists of 130 species of grasses, of which 126 are included in subfamily Chloridoideae; these are partitioned to represent the following five tribes: Centropodieae with one species, Triraphideae with two species, Eragrostideae with nine species, Zoysieae with five species and Chlorideae with 109 species. Within Chloridoideae all tribes and subtribes are evenly represented. Our sampling was primarily focused on genera that are morphologically similar and possibly phylogenetically related to Leptochloa; therefore, we have a large sample within the subtribe Eleusininae where the genus has been recently aligned (Peterson et al., 2010a). The data set for Leptochloa includes 69 % (22 of 32) of the species currently placed in the genus and four infraspecific taxa (Snow, 1997, 1998b, 2000; Snow and Simon, 1997). A complete list of taxa, voucher information, and GenBank numbers can be found in Supplementary Data Table S1. Outside Chloridoideae, two species of Danthonioideae, one species of Aristidoideae and one species of Panicoideae [Chasmanthium latifolium (Michx.) H.O.Yates, phylogenetic root] were chosen as outgroups. Two representatives of subfamily Danthonioideae [Danthonia compressa Austin and Rytidosperma penicellatum (Labill.) Connor & Edgar] were included as this subfamily was shown to be sister to Chloridoideae (Peterson et al., 2010a, 2011). The following regions were chosen to emphasize the native distribution of Leptochloa spp.: North America, South America, Central Africa (Tanzania, Kenya), South Africa, Australia and the Marquesas Islands.

DNA extraction, amplification and sequencing

All procedures were performed in the Laboratory of Analytical Biology (LAB) at the Smithsonian Institution. DNA isolation, amplification and sequencing of rpL32-trnL spacer and ndhA intron (small single copy region), rps16-trnK spacer and rps16 intron (large single copy region), ccsA (encoding region), and ITS were accomplished following procedures outlined in Peterson et al. (2010a, b). We specifically targeted four of the plastid regions which proved to be most informative in our previous studies on chloridoid grasses (Peterson et al., 2010a, b, 2011). Forty per cent (314) of the sequences used in our study are newly reported here and in GenBank, and only 9 % (70) are missing.

The ccsA marker in our study is nearly a complete sequence of the gene encoding biogenesis of c-type cytochromes, located on the positive DNA strand in the small single copy unit of the plastid genome (Xie and Merchant, 1996). To the best of our knowledge, this is the first application of this region as a marker in molecular phylogenetic analysis of grasses. The length of the sequenced region for Chloridoideae varies from approx. 850 to 950 bp. The region was amplified and sequenced with a newly designed set of primers: ccsA1F (5′- ATGCTATTTGCAACTTTAGAACA-3′) used for forward and ccsA930R (5′-TAAACGAACCATAACTATGTAA-3′) for reverse. The amplification parameters were as follows: initial denaturation phase of 4 min at 94 °C, followed by 35 cycles of denaturation at 94 °C for 40 s, annealing at 51 °C for 40 s and extension at 72 °C for 90 s, and a final extension at 72 °C for 10 min.

Phylogenetic analyses

We used Geneious 5·3·4 (Drummond et al., 2011) for contig assembly of bidirectional sequences of rpL32-trnL, ndhA intron, rps16 intron, rps16-trnK, ccsA and ITS regions, and we used MUSCLE (Edgar, 2004) to align consensus sequences and adjust the final alignment. We identified models of molecular evolution for the plastid DNA and nrDNA regions using jModelTest (Posada, 2008). We applied maximum-likelihood (ML) and Bayesian searches to infer overall phylogeny. The combined data sets were partitioned in accordance with the number of the markers in use. Nucleotide substitution models selected based on Akaike's information criterion (AIC), as implemented in jModelTest v.0·1·1, were specified for each partition (Table 2). The ML analysis was conducted with GARLI 0·951 (Zwickl, 2006). The ML bootstrap analysis was performed with 1000 replicates, with ten random addition sequences per replicate. The output file containing trees of ML found for each bootstrap data set was then read into PAUP* 4·0b10 (Swofford, 2000) where the majority-rule consensus tree was constructed. BS values of 90–100 % were interpreted as strong support, 70–89 % as moderate and 50–69 % as weak.

Table 2.

Summary of the five plastid and nrDNA ITS regions used in the maximum-likelihood and Bayesian searches indicated by Akaike's Information Criterion (AIC)

Characteristic	rpL32-trnL	ndhA intron	rps16 intron	rps16-trnK	ccsA	Combined plastid data	ITS	Overall combined data set
Total aligned characters	1233	1323	1074	1040	916	5586	832	6418
Sequencing success (%)	97·7	79·2	90·0	94·6	89·2		95·4	91·0
Range of likelihood scores across 88 models from (–lnL)	12 261·1–11 209·8	11 062·4–9923·4	8369·5–7634·5	10 152·0–9313·0	7615·5–6667·1		26 404·4–22 821·9
Maximum-likelihood scores (–lnL)	11 136·1355	9874·0937	7634·4594	9313·9542	6667·1345		22 937·3535
Number of substitution types	6	6	6	6	6		6
Model for among-site rate variation	Gamma	Gamma	Gamma	Gamma	Gamma		Gamma
Substitution rates	0·8439	1·3761	2·0082	1·7449	1·0000		1·3687
	1·8504	2·9593	2·3904	3·6185	3·5360		3·0126
	0·4460	0·5922	1·0000	1·0000	0·6972		1·5695
	1·2388	2·4171	2·0082	1·7449	0·6972		0·9582
	1·4202	2·9593	3·6308	3·2233	3·5360		5·1859
	1·0000	1·0000	1·0000	1·0000	1·0000		1·0000
Character state frequencies	0·3822	0·3647	0·3644	0·2843	0·3060		0·2433
	0·1428	0·1252	0·1045	0·1424	0·1657		0·1869
	0·1271	0·1445	0·1800	0·1586	0·1641		0·2513
	0·3479	0·3655	0·3511	0·4146	0·3642		0·3185
Proportion of invariable sites	0·0	0·0	0·0	0·0	0·0		0·2289
Substitution model	GTR + I + G	TVM + G	TIM3 + I + G	TIM3 + G	TPM1uf + I + G		GTR + I + G
Gamma shape parameter (α)	0·7030	0·5850	0·5550	0·8680	0·2980		0·9944

Bayesian PP values were estimated using a parallel version of MrBayes v3·1·2 (Huelsenbeck and Ronquist, 2001; Ronquist and Huelsenbeck, 2003) where the run of eight Markov chain Monte Carlo iterations was split between an equal number of processors. Bayesian analysis was initiated with random starting trees and was initially run for four million generations, sampling once per 100 generations. The analysis was run until the value of standard deviation of split sequences dropped below 0·01 and the potential scale reduction factor was close to or equal to 1·0. The fraction of the sampled values discarded as burn-in was set at 0·25. PPs of 0·95–1·00 were considered as strong support.

The incongruence length difference (ILD) test (Farris et al., 1994) was conducted to assess the homogeneity of the data sets and to detect possible incongruence among six partitions of the plastid and nrDNA ITS sequence data. The pairwise tests were performed using the program WinClada v.1·00·08 (Nixon, 2002) with 1000 replications. The significance level was adjusted to P ≤ 0·01.

RESULTS

Phylogenetic analyses

A total of 323 sequences representing 40 species are newly reported in GenBank (Supplementary Data Table S1). Total aligned characters for individual regions are noted in Table 2. Plastid rpL32-trnL had the highest sequencing success of 97·7 % of taxa recovered across the entire data set. Recovery in other regions ranged from 79·2 to 95·4 %. An average of 9 % of data (70/780) was missing across the entire data set.

The pairwise ILD tests among the six partitions indicated acceptable levels of congruence between a majority of the data sets (P = 0·0779–0·9750), with the exception of ndhA–rps16–trnK (P = 0·0040) and rpl32–trnL–ITS (P = 0·0070) pairs. As each region in the detected ‘incongruent’ pairs was congruent with all other partitions in the data set, we attribute these outcomes as a limitation of the method, which is prone to type I error (erroneously rejecting the hypothesis of congruence) when there was a high level of heterogeneity in the data (Barker and Lutzoni, 2002; Darlu and Lecointre, 2002). Therefore, we combined the plastid sequences and plastid–ITS sequences in our analysis.

Analysis of plastid sequences

The ML tree from the combined analysis of the five plastid regions (rpL32-trnL, ndhA intron, rps16 intron, rps16-trnK and ccsA) is well resolved with strong to moderate support for each of the tribes and most of the subtribes of Chloridoideae (Fig. 1). The species currently placed in Leptochloa do not align in a single clade and are found in seven principal clades (Fig. 1, I–VI). The principal lineages of Leptochloa are depicted as follows: clade I (BS = 97, PP = 1·00) includes two accessions of L. uniflora Hochst. ex A.Rich. that are sister to Mosdenia–Toliara–Lopholepis–Perotis (BS = 98, PP = 1·00); clade II (BS = 100, PP = 1·00) includes Dinebra retroflexa, L. chinensis (L.) Nees, L. marquisensis (F.Br.) P.M.Peterson & Judz., a monophyletic L. panicea with three accessions (BS = 99, PP = 1·00), L. squarrosa Pilg. and L. xerophila P.M. Peterson & Judz.; clade IIIA (PP = 0·93) includes L. caudata (K.Schum.) N.Snow, a monophyletic L. decipiens (R.Br.) Stapf ex Maiden (BS = 93, PP = 1·00), L. ligulata Lazarides, L. nealleyi Vasey, L. neesii (Thwaites) Benth., L. panicoides (J.Presl) Hitchc., L. scabra Nees, L. southwoodii N.Snow & B.K.Simon and two accession of L. viscida that are sister; clade IIIB (PP = 0·81) includes L. coerulescens Steud. that is sister to Coelachyrum–Eleusine; clade IV (BS = 100, PP = 1·00) includes a monophyletic L. fusca with L. fusca subsp. muelleri (Benth.) N.Snow and two accessions of L. fusca subsp. uninervia (J.Presl) N.Snow that is sister to all other Eleusininae; clade V (BS = 86, PP = 1·00) includes L. digitata (R.Br.) Domin, and three accessions of L. virgata in a grade (BS = 100, PP = 1·00), and two Trichloris spp. (BS = 100, PP = 1·00); and clade VI (BS = 100, PP = 1·00) includes two sister accessions of L. dubia with L. eleusine (Nees) Cope & N.Snow and L. obtusiflora Trin. ex Steud. The Eleusininae clade is weakly supported (BS = 65, PP = 1·00).

Fig. 1.

Phylogram of maximum-likelihood tree from analysis of combined plastid sequences. Numbers above branches represent bootstrap values; numbers below branches represent posterior probabilities; colour indicates native distribution (see legend); roman numerals and letters indicate clades discussed in the text. Scale bar = 10 % sequence divergence.

Analysis of ITS sequences

The ML tree based on data from the nrITS region is not as well resolved and the species are found in six principal clades (Fig. 2, I–VI): (I) is a single accession of L. uniflora sister to Mosdenia–Perotis–Toliara (BS = 100, PP = 1·00); (II) is similar to the plastid tree but lacks L. chinensis which is sister (BS = 52, PP = 0·77) to both clades II and III (BS = 96, PP = 1·00); (III) is similar to the plastid tree, but the subspecies of L. decipiens appear as a grade and L. coerulescens is sister to L. caudata (BS = 64, PP = 0·56); (IV) depicts a monophyletic L. fusca (BS = 100, PP = 1·00) that is a member of a grade between the Coelachyrum poiflorum–Eleusine indica and L. digitata–Trichloris–L. virgata clades (V); and (V) and (VI) are similar to the plastid tree. Branch support for Eleusininae is lacking.

Fig. 2.

Phylogram of maximum-likelihood tree from analysis of ITS sequences. Numbers above branches represent bootstrap values; numbers below branches represent posterior probabilities; colour indicates native distribution (see legend); roman numerals indicate clades discussed in the text. Scale bar = 10 % sequence divergence.

Analysis of combined plastid and ITS sequences

The ML tree based on data from the combined plastid and ITS regions is well resolved and is remarkably similar to the plastid tree and has six principal clades (Fig. 3, I–VI). Differences between the combined plastid–ITS tree and the plastid tree are: (1) overall support for Eleusininae is slightly higher (BS = 79, PP = 1·00) in the former tree; (2) clade IV (L. fusca) is not sister to remaining genera in Eleusininae but is placed in a grade with clades V and VI; (3) clades II and III are strongly supported (BS = 95, PP = 1·00) and, as in the ITS tree, L. coerulescens is included in clade III; and (4) clade V is strongly supported (BS = 100, PP = 1·00) in the former tree. In the following sections we will refer to the combined plastid–ITS tree when discussing evolutionary scenarios.

Other novelties

Our trees portray the evolutionary relationships for non-Leptochloa s.l. and related taxa that were previously found in Peterson et al. (2010a), except for the following. In this study, Desmostachya bipinnata (L.) Stapf is aligned within Tripogoninae (BS = 99, PP = 1·00) but was earlier reported as incertae sedis (Peterson et al., 2010a). The new monotypic genus, Toliara arenacea Judz., is aligned with Perotis and will need to be investigated further (Judziewicz, 2009). Sclerodactylon macrostachyum (Benth.) A.Camus, although based on only two markers (rpL32-trnL and ccsA), is aligned within Tridentinae as sister to Triplasis purpurea (Walter) Chapm. and is not a member of Eleusininae as earlier reported (Peterson et al., 2010a).

DISCUSSION

Our results indicate that Leptochloa, as currently circumscribed, is polyphyletic and our phylograms show that at least five major clades are necessary to portray the evolutionary relationships among the species (Columbus et al., 2007; Peterson et al., 2010a). Monophyly of Eleusininae is moderately supported (BS = 79, PP = 1·00) by the combined plastid–ITS tree, whereas earlier support for this subtribe was based entirely on posterior probabilities (Peterson et al., 2010a). We attribute the slightly higher support value for Eleusininae to increased taxon sampling and more polymorphic markers that we selected from different regions (spacers, introns and coding regions).

Clade I

Leptochloa uniflora (clade I, Figs 1–3) and L. rupestris C.E.Hubb. share a unique character of a deep, vertically walled (or nearly so) sulcus on the hilar side of the trigonous caryopsis (Phillips, 1974b; Snow, 1997, 1998a). This was suggested by Snow (1998a) to be one of two useful phylogenetic characters of the caryopsis within Leptochloa spp. (the other being loose pericarps). Leptochloa uniflora does not align with other species of Eleusininae but is sister to the Mosdenia–Toliara–Lopholepis–Perotis clade, providing further evidence for an independent origin. The latter four genera have an inflorescence of a single raceme whereas L. uniflora and L. rupestris have numerous racemose branches inserted along the main axis (Clayton et al., 2006; Snow, 1997). Apparently, the inflorescence structure is environmentally plastic and easy to alter in the evolutionary history of these African genera, a situation that recurs in other chloridoid genera such as Muhlenbergia, which also displays a wide range of branching types (Peterson et al., 2010b).

Clades II and III

The majority of the species of Leptochloa s.l. (Fig. 3) occur in two moderately supported clades (II and III, each with BS = 88, PP = 1·00). The first clade includes Dinebra retroflexa, the Marquesas Island endemics L. marquisensis and L. xerophila, L. panicea, L. chinensis and L. squarrosa; the second clade comprises Drake-brockmania somalensis on a long branch sister to L. caudata, L. coerulescens, L. decipiens, L. ligulata, L. nealleyi, L. neesii, L. panicoides, L. scabra and L. southwoodii. Morphologically, Dinebra is similar to Leptochloa and the two genera have long been linked (Phillips, 1973; Clayton and Renvoize, 1986). Phillips (1973) indicated that Dinebra is closely related to the larger and more widespread Leptochloa and shares a racemose inflorescence of one- to several-flowered spikelets which disarticulate between the florets, three-nerved and keeled lemmas with pilose nerves, and lemmas with entire or toothed apices. Our molecular data clearly place Dinebra retroflexa (type species of the genus) within these core species of Leptochloa, despite its spikelet morphology being atypical compared with most others. A similar result was obtained by Columbus et al. (2007) where Dinebra retroflexa formed a clade in the combined trnL-F–ITS tree with L. panicea (BS = 100). Drake-brockmania somalensis is unusual in having subcapitate inflorescences with two to six ovate spikes, 11- to 17-nerved upper glumes (mostly one- but occasionally two- or three-nerved in Leptochloa) half as long to longer than the spikelet, and five- to seven-nerved lemmas (mostly three- but occasionally five-nerved in Leptochloa) (Phillips, 1995). The only other species placed in Drake-brockmania, D. haareri (Stapf & C.E.Hubb.) S.M.Phillips, was originally placed in a separate genus, Heterocarpha Stapf & C.E.Hubb., based on possessing three-nerved lemmas, but otherwise these two species are probably closely related (Phillips, 1974a).

Even though we have not sampled L. aquatica Scribn. & Merr., L. simoniana N.Snow and L. srilankensis N.Snow, based on their morphological characteristics we believe they share a common ancestor with other species in clades II and III. Leptochloa simoniana, known only from Papua New Guinea, most closely resembles L. coerulescens in having narrow panicles with flexuous branches, leaf blades that are densely scabrous and upper glumes with margins that are densely scabrous (Snow, 2000). Leptochloa srilankensis is morphologically similar to L. decipiens subsp. asthenes, as both species have sparsely to moderately pilose leaf sheaths (the hairs usually tuberculate at base), rather short (1–11 cm long) narrowly to ovate leaf blades that are 1–3 mm wide, panicles 6–40 cm long with glabrous branch axils, and spikelets that are two- to four- (five-)flowered (Snow, 1997). Leptochloa aquatica most closely resembles L. panicoides by having an annual habit, long (4–20 cm) leaf blades that are 4·0–8·3 mm wide, long (20–36 cm) panicles that are relatively narrow (2–12 cm wide), four- to six-flowered spikelets, lemmas that are 2·4–3·5 mm long, and caryopses that are 1·1–1·5 mm long (McVaugh, 1983; Snow, 1997).

The placement of L. coerulescens is not congruent between the plastid and ITS trees (Figs 1 and 2). In the plastid tree (Fig. 1, IIIB) L. coerulescens is in a separate clade sister to Coelachyrum–Eleusine whereas in the ITS tree (Fig. 2, III) it is sister to L. caudata. We attribute the uncertain position to possible DNA shuffling (introgression) and the mixing of phylogenetic signals from two or more distant progenitors. We hope to address this question by applying low copy nuclear markers to taxa of Chlorideae.

Clade IV

The controversy surrounding the use of Diplachne or Leptochloa was summarized by McNeill (1979), who discussed the diagnostic characters used to separate these taxa and provided a key to those that occur in North America. Our study does not support this division. Preliminary data using plastid restriction site banding patterns (N. Snow, unpubl. res.) of several North American Leptochloa spp. segregated the single North American species (L. fusca) from several others of the genus. However, there is strong support for recognizing the widely distributed L. fusca (Fig. 3, clade IV) in Diplachne as the type is included in this complex. Leptochloa fusca subsp. fusca is a polymorphic palaeotropical taxon that is adventive in a few areas of the New World (Nicora, 1995); L. fusca subsp. muelleri is known from much of the interior portions of eastern Australia, particularly the Northern Territory; L. fusca subsp. uninervia is native to the Neotropics but adventive elsewhere; and L. fusca subsp. fascicularis is native to the temperate and tropical regions of the New World (Snow, 1997). Also included in our interpretation of Diplachne is Leptochloa gigantea (Launert) Cope & N.Snow (not sampled), an obligate wetland species restricted to parts of eastern and southern Africa, given its possession, with L. fusca, of distinguishing characters such as long spikelets, apically attenuate ligules (before mechanical laceration) mostly 5–12 mm long and long (6–12 mm) florets (Snow, 1997).

Clade V

The inclusion of the American Trichloris crinita and T. pluriflora in clade V (Figs 1–3) has not been previously documented in molecular studies. Previous molecular results indicated that the two Trichloris spp. were sister and not closely related to Chloris where they often have been placed (Anderson, 1974; Watson and Dallwitz et al., 1992; Peterson et al., 2010a). Likewise, the inclusion of Enteropogon chlorideus (J.Presl) Clayton in Leptochloa sensu stricto (s.s.) has not been definitively documented using molecular studies. However, Columbus et al. (2007) did show strong support for a clade containing T. crinita and E. chlorideus. To test these relationships we included E. chlorideus (EF153044) and E. mollis (Nees) Clayton (EF153045) ITS sequences in our analysis (results not shown). Enteropogon chlorideus formed a clade with Saugetia fasciculata Hitchc. & Chase and Tetrapogon villosus Desf., whereas E. mollis was aligned with Leptochloa s.s. As we have a single ITS marker supporting the inclusion of E. mollis within Leptochloa s.s. and no plastid sequences, it is premature to transfer this species to Leptochloa. We hope to sample many species of Chloris, Enteropogon and Eustachys to clarify generic boundaries.

The wide ranging neotropical Leptochloa virgata, the southern South American L. chloridiformis (Hack.) Parodi (not sampled), the Australian endemic L. digitata and the two Trichloris spp. possess an inflorescence with digitate or sub-digitately inserted racemes usually with two or more branches per node. Historically, agrostologists have placed great importance on whether the lemma is awned, and, if so, how many times. A distinguishing feature of T. crinita and T. pluriflora is that the lemmas are three-awned. Apparently this trait is not important when determining phylogenetic relationships as L. digitata has unawned lemmas, L. virgata has single-awned or mucronate lemmas, and L. chloridiformis is at most mucronate (Snow, 1997). Additional characters shared by L. digitata, L. chloridiformis and L. virgata include a perennial habit, solid culms and short, apically truncate (but slightly fimbriate) ligules.

Clade VI

Many taxonomists have recognized Leptochloa dubia as a unique, polymorphic, widespread taxon as it has deeply bifid or notched lemmas with short, glabrous lateral nerves (Steudel, 1854; Philippi, 1870; Valls, 1978; McNeill, 1979). Clade VI (Figs 1–3) of our study includes L. dubia distributed in temperate North and South America, L. eleusine from South Africa, and L. obtusiflora from central Africa. Leptochloa eleusine and L. obtusiflora are atypical within Leptochloa s.l. by possessing clavicorniculate macrohairs on the lemmas (Snow, 1996, 1997). Hairs having a clavicorniculate apex have been reported to function as salt glands in the panicoid genus Eriochloa (Arriaga, 1992). Along with a few other species in Leptochloa, L. eleusine and L. obtusiflora have a weakly adnate pericarp on the caryopses (Snow, 1997). The adhesion of the pericarp has been shown to be an important character defining natural groups in Chloridoideae (Sporobolinae; Peterson et al., 2010a). Future workers should sample additional Coelachyrum spp. as they have clavicorniculate microhairs (Snow, 1996) and a spikelet structure that closely resembles those of L. eleusine and L. obtusiflora (Snow, 1997).

TAXONOMY

Because our molecular analysis renders Leptochloa s.l. polyphyletic, we propose recognizing five genera (Dinebra, Diplachne, Disakisperma, Leptochloa and Trigonchloa), each of which was found in a separate, strongly supported clade (Fig. 3, BS values ≥95). Given their overall similarities to other species, Dinebra perrieri (A.Camus) Bosser, Dinebra polycarpha S.M.Phillips, Drake-brockmania haareri (Stapf & C.E.Hubb.) S.M.Phillips, Leptochloa aquatica, L. chloridiformis, L. gigantea, L. rupestris, L. simoniana and L. srilankensis can be confidently placed among the identified clades. Leptochloa divaricatissima S.T.Blake probably belongs in Dinebra, but we are withholding transfer pending additional work. The placement of L. longa Griseb., L. malayana (C.E.Hubb.) Jansen ex Veldkamp and L. tectoneticola (Backer) Jansen ex Veldkamp also will require additional sequencing studies, because morphologically they are at odds with other species of Leptochloa s.l. (Snow, 1997). We provide a preliminary key to the newly circumscribed genera and list of accepted species in Table 3. The species are listed below: new combinations appear in bold and species not included in our DNA analysis are preceded by an asterisk (*).

Table 3.

Key to the genera

1	Leaf blades thin, membranous, broader than linear; caryopses trigonal in transverse section; sulcus present, deep with nearly vertical walls; one or two lateral bundle sheath cells on each side significantly enlarged	Trigonochloa
1′	Leaf blades thicker, more or less linear; caryopses lenticular to terete in cross section, sulcus absent or if present then shallow; enlarged lateral bundle sheath cells lacking	2
2	Ligules 4–8 (–15) mm long, apex acute to attenuate, lacerate only by tearing	Diplachne
2′	Ligules 0·2–8·0 (–15·0) mm long, apex usually truncate to obtuse and somewhat erose	3
3	Apex of the lemmatal hairs clavicorniculate, ovate to broadly ovate; base of lemma often indurate and sometimes five-veined; plants perennial; ligules 0·8–2·2 mm long, apex erose	Disakisperma
3′	Apex of the lemmatal hairs ovate to acute, never clavicorniculate; base of lemma soft and always three-veined; plants annual or perennial; ligules (0·2–) 0·5–5·5 (–7·0) mm long, apex usually entire	4
4	Panicle branches digitate to sub-digitate inserted along the rachis usually with two or more branches per node; lemmas one- or three-awned or unawned; plants perennial; culms solid; ligules 0·5–3·0 mm long, apex ciliate	Leptochloa
4′	Panicle branches racemosely inserted along the rachis, rarely digitate to sub-digitate, usually with a single branch per node; lemmas unawned; plants annual or perennial; culms solid or hollow; ligules (0·2–) 0·5–5·5 (–7·0) mm long, apex erose or entire but never ciliate	Dinebra

Dinebra Jacq., Fragm. Bot. 77. 1809. (Fig. 3, clades II & III) Type: D. arabica Jacq. [≡D. retroflexa (Vahl) Panz.]. = Drake-brockmania Stapf, Bull. Misc. Inform. Kew 1912: 197. 1912. Type: D. somalensis Stapf.

1. *Dinebra aquatica (Scribn. & Merr.) P.M.Peterson & N.Snow, comb. nov. Basionym: Leptochloa aquatica Scrinb. & Merr., Bull. Div. Agrostol., U.S.D.A. 24: 26. 1901.

2. Dinebra caudata (K.Schum.) P.M.Peterson & N.Snow, comb. nov. Basionym: Diplachne caudata K.Schum., Pflanzenw. Ost-Afrikas 113. 1895. ≡Leptochloa caudata (K.Schum.) N.Snow.

3. Dinebra chinensis (L.) P.M.Peterson & N.Snow, comb. nov. Basionym: Poa chinensis L., Sp. Pl. 1: 69. 1753. ≡Leptochloa chinensis (L.) Nees.

4. Dinebra coerulescens (Steud.) P.M.Peterson & N.Snow, comb. nov. Basionym: Leptochloa coerulescens Steud., Syn. Pl. Glumac. 1: 209. 1854.

5. Dinebra decipiens (R.Br.) P.M.Peterson & N.Snow, comb. nov. Basionym: Poa decipiens R.Br., Prodr. 181. 1810. ≡Leptochloa decipiens (R.Br.) Stapf ex Maiden.

5a. Dinebra decipiens subsp. asthenes (Roem. & Schult.) P.M.Peterson & N.Snow, comb. nov. Basionym: Poa asthenes Roem. & Schult., Syst.Veg. 2: 574. 1817.

5b. Dinebra decipiens subsp. peacockii (Maiden & Betche) P.M.Peterson & N.Snow, comb. nov. Basionym: Diplachne peacockii Maiden & Betche, Agric. Gaz. New South Wales 15: 925. 1904.

6. *Dinebra haareri (Stapf & C.E.Hubb) P.M.Peterson & N.Snow, comb. nov. Basionym: Heterocarpha haareri Stapf & C.E.Hubb., Bull. Misc. Inform. Kew 1929: 263. 1929. ≡Drake-brockmania haareri (Stapf & C.E.Hubb.) S.M.Phillips.

7. Dinebra ligulata (Lazarides) P.M.Peterson & N.Snow, comb. nov. Basionym: Leptochloa ligulata Lazarides, Brunonia 3(2): 259. 1980.

8. Dinebra marquisensis (F.Br.) P.M.Peterson & N.Snow, comb. nov. Basionym: Eragrostis marquisensis F.Br., Bernice P. Bishop Mus. Bull. 84: 81. 1931. ≡Leptochloa marquisensis (F.Br.) P.M.Peterson & Judz.

9. Dinebra nealleyi (Vasey) P.M.Peterson & N.Snow, comb. nov. Basionym: Leptochloa nealleyi Vasey, Bull. Torrey Bot. Club 12: 7. 1885.

10. Dinebra neesii (Thwaites) P.M.Peterson & N.Snow, comb. nov. Basionym: Cynodon neesii Thwaites, Enum. Pl. Zeyl. 371. 1864. ≡Leptochloa neesii (Thwaites) Benth.

11. Dinebra panicea (Retz.) P.M.Peterson & N.Snow, comb. nov. Basionym: Poa panicea Retz., Observ. Bot. 3: 11. 1783. ≡Leptochloa panicea (Retz.) Ohwi.

11a. Dinebra panicea subsp. brachiata (Steud.) P.M.Peterson & N.Snow, comb. nov. Basionym: Leptochloa brachiata Steud., Syn. Pl. Glumac. 1: 209. 1854.

11b. Dinebra panicea subsp. mucronata (Michx.) P.M.Peterson & N.Snow, comb. nov. Basionym: Eleusine mucronata Michx., Fl. Bor.-Amer. 1: 65. 1803.

12. Dinebra panicoides (J.Presl) P.M.Peterson & N.Snow, comb nov. Basionym: Megastachya panicoides J.Presl, Reliq. Haenk. 1(4–5): 283. 1830. ≡Leptochloa panicoides (J.Presl) Hitchc.

13. *Dinebra perrieri (A.Camus) Bosser.

14. *Dinebra polycarpha S.M.Phillips.

15. Dinebra retroflexa (Vahl) Panz.

16. Dinebra scabra (Nees) P.M.Peterson & N.Snow, comb. nov. Basionym: Leptochloa scabra Nees, Agrostogr. Bras. 2: 435. 1829.

17. *Dinebra simoniana (N.Snow) P.M.Peterson & N.Snow, comb. nov. Basionym: Leptochloa simoniana N.Snow, Novon 10: 328. 2000.

18. Dinebra somalensis (Stapf) P.M.Peterson & N.Snow, comb. nov. Basionym: Drake-brockmania somalensis Stapf, Bull. Misc. Inform. Kew 1912: 197. 1912.

19. Dinebra southwoodii (N.Snow & B.K.Simon) P.M.Peterson & N.Snow, comb. nov. Basionym: Leptochloa southwoodii N.Snow & B.K.Simon, Austrobaileya 5(1): 138. 1997.

20. Dinebra squarrosa (Pilg.) P.M.Peterson & N.Snow, comb. nov. Basionyn: Leptochloa squarrosa Pilg., Bot. Jahrb. Syst. 45: 210. 1910.

21. *Dinebra srilankensis (N.Snow) P.M.Peterson & N.Snow, comb. nov. Basionym: Leptochloa srilankensis N.Snow, Novon 8: 183. 1998.

22. Dinebra viscida (Scribn.) P.M.Peterson & N.Snow, comb. nov. Basionym: Diplachne viscida Scribn., Bull. Torrey Bot. Club 10(1): 30. 1883. ≡Leptochloa viscida (Scribn.) Beal.

23. Dinebra xerophila (P.M.Peterson & Judz.) P.M.Peterson & N.Snow, comb. nov. Basionym: Leptochloa xerophila P.M.Peterson & Judz., Taxon 39: 659. 1990.

Diplachne P.Beauv., Ess. Agrostogr. 80. 1812. (Fig. 3, clade IV) Type: D. fascicularis (Lam.) P.Beauv. [≡Leptochloa fusca subsp. fascicularis (Lam.) N.Snow]

1. Diplachne fusca (L.) P.Beauv. ex Roem. & Schult. ≡Leptochloa fusca (L.) Kunth.

1a. Diplachne fusca subsp. fasciculata (Lam.) P.M.Peterson & N.Snow, comb. nov. Basionym: Festuca fascicularis Lam., Tabl. Encycl. 1: 189. 1791.

1b. Diplachne fusca subsp. muelleri (Benth.) P.M.Peterson & N.Snow, comb. nov. Basionym: Diplachne muelleri Benth., Fl. Austral. 7: 619. 1878.

1c. Diplachnea fusca subsp. uninervia (J.Presl) P.M.Peterson & N.Snow, comb. nov. Basionym: Megastachya uninervia J.Presl, Reliq. Haenk 1(4–5): 283. 1830.

2. *Diplachne gigantea Launert.

Disakisperma Steud., Syn., Pl. Glumac. 1: 287. 1854. (Fig. 3, clade VI) Type: D. mexicana Steud. (≡ D. dubia).

1. Disakisperma dubia (Kunth) P.M.Peterson & N.Snow, comb. nov. Basionym: Chloris dubia Kunth, Nov. Gen. Sp. 1: 169. 1816. ≡Leptochloa dubia (Kunth) Nees.

2. Disakisperma eleusine (Nees) P.M.Peterson & N.Snow, comb. nov. Basionym: Diplachne eleusine Nees, Fl. Afr. Austral. Ill. 255. 1841. ≡Leptochloa eleusine (Nees) Cope & N.Snow.

3. Disakisperma obtusiflora (Hochst.) P.M.Peterson & N.Snow, comb. nov. Basionym: Leptochloa obtusiflora Hochst., Flora 38: 203. 1855.

Leptochloa P.Beauv., Ess. Agrostogr. 71. 1812. (Fig. 3, clade V) Type: L. virgata (L.) Benth. = Trichloris E.Fourn. ex Benth., J. Linn. Soc., Bot. 19: 102. 1881. Lectotype: Trichloris pluriflora E.Fourn.

1. *Leptochloa chloridiformis (Hack.) Parodi.

2. Leptochloa digitata (R.Br.) Domin.

3. Leptochloa virgata (L.) Benth.

4. Leptochloa crinata (Lag.) P.M.Peterson & N.Snow, comb. nov. Basionym: Chloris crinata Lag., Varied. Ci. 2(4): 143. 1805.

5. Leptochloa pluriflora (E.Fourn.) P.M.Peterson & N.Snow, comb. nov. Basionym: Trichloris pluriflora E.Fourn., Mexic. Pl. 2: 142. 1886.

Trigonochloa P.M.Peterson & N.Snow, gen. nov. (Fig. 3, clade I) Type: Leptochloa uniflora Hochst. ex A.Rich.

Diagnosis

Trigonochloa differs from Perotis Aiton by having an inflorescence with several to numerous unilateral, secund racemes scattered along a central axis, a three-veined lemma and a trigonous caryopsis with a deep sulcus on the hilar surface.

Description

Plants annual to short-lived perennial, sprawling or clambering, often stoloniferous. Culms 30–150 cm long, mostly decumbent. Leaf sheaths ½ to almost as long as the internodes above, glabrous; ligule membranous, irregularly lacerate with age; leaf blades 3–12 cm long, 2–18 mm wide, broad, lanceolate–oblong to narrowly oblong or narrowly lanceolate, glabrous and smooth, often thin and flaccid, apex acute. Inflorescence 6–45 cm long, open oblong to narrowly oblong to narrowly elliptic, composed of several to numerous unilateral, secund racemes scattered along a central axis, each spikelet oriented laterally to the axis; rachis semiterete; racemes 2–9 cm long, ascending, straight or slightly arching. Spikelets 1·6–2·8 mm long, one-flowered, laterally compressed, sub-sessile, overlapping; disarticulation above the glumes; glumes 1·6–2·4 mm long, as long or longer than the floret, sub-equal, linear to linear lanceolate, one-nerved, apex acute to acuminate; lemmas 1·5–2·1 mm long, narrowly elliptic to elliptic oblong, three-nerved, minutely hairy along the nerves, apex acute, entire; paleas keels ciliolate, two-nerved. Caryopsis 1·0–1·2 mm long, narrowly elliptic, trigonous in cross section, deeply sulcate on the hilar side; pericarp adherent. 2n = 36 (for T. uniflora).

Leaf anatomy

A probable synapomorphy for Trigonochloa is the presence of greatly enlarged lateral cells in the primary vascular bundles, i.e. the presence of one or two cells on each side of the wreath that are significantly larger than the others. We include a summary of the anatomical characteristics for the two species presented in Snow (1997).

Leaf blades generally thinly membranous. Keels present but lacking lacunae. Outer sheath of primary bundles continuous adaxially, continuous or interrupted abaxially. Extension cells above primary bundles present or absent. Rib size of primary bundles of normal size. Secondary outer bundle continuous abaxially. Primary bundle adaxial sclerenchyma absent, or present as a girder. Primary bundle abaxial sclerenchyma present as a girder. Secondary bundle adaxial sclerenchyma absent. Secondary bundle abaxial sclerenchyma present as girders. Adaxial cells of primary bundle sheath cells not enlarged. Abaxial cells of primary bundle sheath cells not enlarged. Primary bundles not projecting adaxially. Primary bundles not projecting abaxially. Secondary bundles not projecting adaxially. Secondary bundles not projecting abaxially. Primary and secondary bundles height nearly equal. Colourless cells absent between primary and secondary bundles. Chlorenchyma continuous between adjacent bundles. Continuous abaxial band of sclerenchyma below epidermis absent. Phloem not interrupted by abaxial sclerenchyma. Vascular bundles with greatly enlarged lateral cells. [Vouchers, all South Africa: Snow et al. 6978 (MO), Davidse & Ellis 5925 (MO); Ellis 3635 (PRE); Ellis 4534 (PRE)].

Distribution and habitat

The two species are distributed in Africa: west tropical, west-central tropical, north-east tropical, east tropical, southern tropical, and south and Asia-temperate: Arabia. Asia-tropical: India. Plants occur in forests and shady areas on hillsides, bushlands, on well-drained and often sandy soils and damp rocks along streams in disturbed Yemen and Eritrea south to Kenya; woodlands, hillsides, bushland and on damp rocks along streams and riverine areas in disturbed habitats at 0–1800 m.

Notes

The generic name emphasizes the unique, trigonous caryopses with a deep sulcus on the hilar (ventral) surface.

1. *Trigonochloa rupestris (C.E.Hubb.) P.M.Peterson & N.Snow, comb. nov. Basionym: Leptochloa rupestris C.E.Hubb., Kew Bull. 1941: 195. 1941.

2. Trigonochloa uniflora (Hochst. ex A.Rich.) P.M.Peterson & N.Snow, comb. nov. Basionym: Leptochloa uniflora Hochst. ex A.Rich., Tent. Fl. Abyss. Tent. 2: 409–410. 1851.

Currently unplaced species

1. Leptochloa divaricatissima S.T.Blake

2. Leptochloa longa Griseb.

3. Leptochloa malayana (C.E.Hubb.) Jansen ex Veldkamp

4. Leptochloa tectoneticola (Backer) Jansen ex Veldkamp

CONCLUSIONS

We have performed a multi-gene phylogenetic analysis based on six molecular markers (nuclear ITS and plastid rpL32-trn-L, ndhA intron, rps16 intron, rps16-trnK and ccsA DNA sequences) of 22 Leptochloa spp. Previously, only four Leptochloa spp. had been included in a molecular analysis and they appeared polyphyletic (Peterson et al., 2010a). Our results indicate that Leptochloa is polyphyletic and our phylograms show that at least five major clades are necessary to portray the evolutionary relationships among the species. Our molecular results support the dissolution of Leptochloa s.l. into five genera: Dinebra, Diplachne, Disakisperma, Leptochloa s.s. and Trigonochloa. Based on the interpretation of our phylogenetic trees and for consistency in rank we provide the necessary changes in the classification of these species. We recognize an expanded Dinebra that includes 23 species, two of which were formerly placed in Drake-brockmania and 19 species in Leptochloa; Diplachne includes two species; Disakisperma includes three species; Leptochloa includes five species, two of which were formerly placed in Trichloris; and a new genus, Trigonochloa, includes two species. There are still 13 species of Leptochloa s.l. (including Dinebra) that need to be surveyed with molecular markers, although nine can be classified with a high degree of confidence given previous studies of morphology (Phillips, 1973; Anderson, 1974; Snow, 1997), stem and leaf anatomy (Snow, 1997; N. Snow, unpubl. data) and lemma micromorphology (Snow, 1996). A much larger molecular sample of species now included in Chloris, Enteropogon and Eustachys is needed to clarify affinities.

SUPPLEMENTARY DATA

Supplementary data are available on line at www.aob.oxford-journals.org and consist of Table S1: list of specimens sampled, voucher (collector, number and where the specimen is housed), country of origin and GenBank accessions for DNA sequences; and a sequence alignment for each of the six DNA markers (provided as an Excel file).