Pl. Syst. Evol. 264: 135–156 (2007) DOI 10.1007/s00606-006-0479-9 Printed in The Netherlands Plant Systematics and Evolution The phylogeny of the austral grass subfamily Danthonioideae: Evidence from multiple data sets N. P. Barker1, C. Galley2, G. A. Verboom3, P. Mafa3,*, M. Gilbert1, and H. P. Linder2 1 Molecular Ecology and Systematics Group, Dept. Botany, Rhodes University, Grahamstown, South Africa 2 Institute for Systematic Botany, Zurich, Switzerland 3 Dept. Botany, University of Cape Town, Cape Town, South Africa Received March 6, 2006; accepted September 6, 2006 Published online: February 12, 2007 Ó Springer-Verlag 2007 Abstract. The grass subfamily Danthonioideae is one of the smaller in the family. We utilize DNA sequence data from three chloroplast regions (trnL, rpoC2 and rbcL) and one nuclear region (Internal Transcribed Spacer; ITS) both singly and in combination to elucidate the relationships of the genera in the subfamily. The topology retrieved by the ITS region is not congruent with that of the plastid data, but this conflict is not strongly supported. Nine wellsupported clades are retrieved by all data sets. The relationships at the base of the subfamily are clearly established, comprising a series of three clades of Merxmuellera species. The earliest diverging clade probably does not belong in Danthonioideae. The other two clades are centered in the tropical African mountains and Cape mountains respectively. A clade of predominantly North and South American Danthonia species as well as D. archboldii from New Guinea is retrieved, but the African and Asian species of Danthonia are related to African species of Merxmuellera, thus rendering Danthonia polyphyletic. The relationships of the Danthonia clade remain equivocal, as do those of the two Cortaderia clades, the Pseudopentameris and Rytidosperma clades. * Paseka Mafa was tragically killed in a vehicle accident in July 2001. This paper includes information he collected during the course of his MSc in Systematics and Biodiversity Science at the University of Cape Town. Key words: Danthonioideae, Poaceae, phylogeny, cpDNA, ITS, Internal Transcribed Spacer, Danthonia, Cortaderia, Merxmuellera. The subfamily-level phylogeny of the grass family (Poaceae) has recently been resolved using eight molecular data sets, as well as a morphological analysis (GPWG 2001). As a consequence of these data, the Danthonioideae was recognized as a subfamily, comprising more or less those taxa that were previously included in the tribe Danthonieae, subfamily Arundinoideae. The combined data analysis presented by the GPWG (2001) suggest that the Aristidoideae is sister to the Danthonioideae, but Barker et al. (1995) resolve the Chloridoideae as the sister group, albeit with very weak bootstrap support. Indeed, there is no robust solution for the relationships among the subfamilies of the PACCAD (Panicoideae, Arundinoideae, Chloridoideae, Centothecoideae, Aristidoideae, Danthonioideae) clade (Barker et al. 1995, 1999; Hilu et al. 1999). However, the subfamily Danthonioideae is undoubtedly monophyletic, characterized by the presence of haustorial synergids (Verboom et al. 1994), 136 N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) as well as several other morphological features such as bilobed and awned lemmas, a festucoid spikelet structure, ciliate ligules, and festucoid leaf anatomy (GPWG 2001). Comprising approximately 330 species, the Danthonioideae is one of the smaller subfamilies of the Poaceae. It is an austral group, with almost all species found in the Southern Hemisphere (almost equally well represented on all southern continents), and only a few species of Danthonia found in the northern Hemisphere (Linder and Barker 2000). Preliminary molecular dating studies suggest that the subfamily is about 10 Myr old (Linder and Barker 2005). The relationships among the genera in the subfamily have been investigated using several molecular and morphological markers, but the latter are too homoplasious and did not result in a robustly supported phylogeny for the subfamily (Barker et al. 2000). The most commonly used molecular markers in studies on the Danthonioideae have been ITS (Hsiao et al. 1998, Barker et al. 2000) and the grass-specific insert in the chloroplast rpoC2 gene (hereafter named ‘‘rpoC2’’; Barker et al. 1999, 2000, 2003). These studies have resulted in the delimitation of seven groups or clades of genera that have been given informal names (Barker et al. 2000). However, the support for these groupings has generally been weak, and the relationships among them insecure. These studies have also revealed numerous problems of generic delimitation. For example, Barker et al. (2003) showed that the genus Cortaderia is not monophyletic, being split along continental lines. The South American species, along with the closely related and possibly embedded genus Lamprothyrsus form one clade, while the species from New Zealand and nearby islands form a second clade, sister to the Australian genera Plinthanthesis and Notochloe. The relationships of Cortaderia archboldii from New Guinea (a species considered intermediate between Chionochloa and Cortaderia, not fitting traditional generic limits) have not been conclusively resolved by molecular data (Barker et al. 2003). The African genus Merxmuellera has also been shown to be grossly polyphyletic (Hsiao et al. 1998; Barker et al. 1999, 2000, 2003). An additional problem in this genus is the uncertain subfamilial affinities of Merxmuellera rangei and M. papposa, both rare grasses from arid regions; the former from Namibia, the latter from the Eastern Cape, South Africa. The position of the two species of Merxmuellera from Madagascar is unknown. Verboom et al. (2006) also show that the small genera Karroochloa, Tribolium and Schismus are completely intermingled. There are also persistent problems with delimitation of Rytidosperma, Austrodanthonia and Notodanthonia (Linder 2005, Linder and Verboom 1996). We thus have glimpses of the phylogenetic structure in the subfamily, but no confident pattern. In order to address the remaining taxonomic problems in the subfamily, there is a need to recognize well-supported monophyletic lineages. Once this is achieved, problems such as generic delimitation can be readily defined and addressed. The interpretations from current molecular studies have been hampered by incomplete sampling across different datasets. Consequently we followed a strategy of filling in the gaps in the data matrix. In addition, the number of genes sequenced has so far been too low to provide confident resolution to the phylogeny. Here we report on the use of two additional chloroplast data sets (the non-coding trnL intron and the coding rbcL gene) to aid in the reconstruction of evolutionary relationships in the Danthonioideae. Over and above the obvious benefits of increased taxon sampling size and data set size, material obtained for a number of unusual (taxonomically difficult and / or geographically isolated) taxa in the subfamily allows us to determine the relationships of these taxa based on at least one data set. For this reason, analyses of individual data sets are presented so as to allow for a discussion on the relationships of these unusual taxa. Materials and methods Taxon sampling. Ninety eight of the ca. 330 species in the subfamily have been included in at least one dataset (Table 1). This relatively dense sampling Genus and species Voucher Locality ITS Austrodanthonia auriculata (J.M. Black) H.P. Linder Austrodanthonia caespitosa (Gaudich.) H.P. Linder Austrodanthonia laevis (Vickery) H.P. Linder Centropodia glauca (Nees) T.A. Cope Linder 5569 (CANB) Australia, ACT, Belconnen Unknown AF367604 Linder 5633 (CANB) Linder 5410 (BOL) Barker 967 (BOL) Ward 10862 (NH) Australia, Wentworth Falls South Africa, N.Cape, Alexander Bay Namibia, Sossusvlei Angola, Mossamedes rbcL trnL intron DQ218157 AF019877 AF019875 U96313 DQ887097 DQ890450 U92265 U31100 DQ890449 DQ887125 DQ890492 AF019861 Barker 978 (PRE) South Africa, N.Cape, Kamiesberg DQ887158 U92483 Barker 1715 (GRA) South Africa, N.Cape, Alexander Bay AF367599 DQ887131 DQ887098 DQ890451 CHR 475418 (BOL) ex cult., New Zealand DQ887159 DQ887133 DQ887099 DQ890452 DQ887100 DQ890453 SWL Jacobs 935 DQ890484 AF019868 CHR 475278 (BOL) CHR 475279 (BOL) Linder 5710 (BOL) ex cult., New Zealand ex cult., New Zealand New Zealand AF367595 AF367596 AF367597 U92701 G7162 J. Marsden 115 (=MWC 8879; K) Lyle 1497 (HBG, BOL) Lyle & Carillo 920. (HBG) ex cult., New Zealand Indonesia, Irian Jaya AF367614 AF367620 AF355993 AF355998 Ecuador Venezuela, Edo. Merida AF367609 AF367612 AF355988 AF355991 DQ890485 DQ887101 DQ890454 137 Centropodia mossemadensis (Rendle) T.A. Cope Chaetobromus involucratus (Schrad.) Nees subsp. dregeanus (Nees) Verboom Chaetobromus involucratus (Schrad.) Nees subsp. involucratus Chionochloa flavescens Zotov Chionochloa frigida (Vickery) Conert Chionochloa macra Zotov Chionochloa pallens Zotov Chionochloa rigida (Raoul) Zotov Cortaderia araucana Stapf Cortaderia archboldii (Hitchc.) Connor & Edgar Cortaderia bifida Pilg. Cortaderia colombiana (Pilg.) Pilg. Hsiao et al. (1998) rpoC2 N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) Table 1. Voucher details for samples sequenced in this study and their associated GenBank numbers. GenBank numbers in bold were generated for this study, and have not been published previously. Voucher numbers preceded by ‘‘G:’’ are garden numbers for material from the Lincoln Botanic Gardens, New Zealand 138 Table 1. (Continued) Voucher Locality ITS rpoC2 rbcL trnL intron Cortaderia fulvida (Buchan.) Zotov G5088 AF367615 U93359 DQ887102 DQ890455 Cortaderia hapalotricha (Phil.) Conert Cortaderia jubata (Lem.) Stapf Cortaderia nitida (Kunth) Pilg. Cortaderia richardii (Endl.) Zotov Cortaderia rudiuscula Stapf Cortaderia selloana (Schult.) Asch. et Graebn. Lyle 1525 (HBG) New Zealand, Lincoln (ex cult., GDN No. 5088) Ecuador AF367610 AF355989 Lyle 1515 (HBG) Ecuador AF367608 AF355987 Lyle 1434 (HBG) Ecuador AF367611 AF355990 G3816 ex cult., New Zealand AF367618 AF355996 G11157 G11157 ex cult., New Zealand ex cult., New Zealand AF367613 AF367607 AF355992 U93360 Hsiao et al. (1998) Lyle 1128 (HBG) Ecuador AF019812 AF367606 AF355986 ex cult., New Zealand ex cult., New Zealand ex cult., New Zealand Indonesia, Irian Jaya AF367616 AF367619 AF367617 AF367620 AF355994 AF355997 AF355995 AF355998 Cortaderia sericantha (Steud.) Hitchc. Cortaderia splendens Connor Cortaderia toetoe Zotov Cortaderia turbaria Connor Danthonia archboldii Hitchc. Danthonia californica Bol. Danthonia compressa Austin Danthonia schneideri Pilg. Danthonia secundiflora J. & C. Presl Danthonia spicata Roem. & Schut. Danthonia subulata A.Rich. Joycea pallida (R. Br.) H.P. Linder Karoochloa tenella (Nees) Conert & Tuerpe DQ890483 G10872 G5042 G17358 J. Marsden 115 (=MWC 8879; K) Unknown. Guala s.n. Soreng 5638 (US) Lyle 1617 (HBG) USA, Pennsylvania China; Xizhang, Tibet La Azulita, Venezuela DQ887161 DQ887162 DQ887132 U93361 DQ887103 DQ887104 DQ890502 DQ890456 DQ890457 Kellogg, s.n. USA, Maine DQ887163 U93662 U31102 DQ890458 Linder 7669 (Z) Linder 5564 (CANB) Ethiopia, Bale Mtns Australia, ACT, Black Mountain South Africa, N.Cape, Botterkloof DQ887164 AF019880 DQ887129 DQ890497 U94394 AF019874 U94824 Linder 5360 (BOL) DQ890486 DQ887160 U31437 N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) Genus and species Lamprothyrsus peruvianus Hitchcl. Merxmuellera arundinacea (Berg.) Conert Merxmuellera cincta (Nees) Conert subsp. cincta Merxmuellera cincta (Nees) Conert subsp. sericea N.P. Barker Merxmuellera davyi (C.E. Hubb.) Conert Merxmuellera decora (Nees) Conert Merxmuellera disticha (Nees) Conert Merxmuellera drakensbergensis (Schweickerdt) Conert Merxmuellera dura (Stapf) Conert G11154 Barker 1017 (BOL) Barker 1160 (BOL) Barker1545 (GRA) AF367605 U94952 DQ887105 DQ890460 DQ887165 U94953 DQ887106 DQ890461 AF367593 U94954 AF367594 AF355985 DQ887107 DQ890462 Malawi AF367590 U94955 Barker 1168 (BOL) South Africa, W.Cape, Grootvadersbos South Africa, E.Cape, Nieu Bethesda Lesotho, Bokong Bog AF367592 AF355984 DQ887108 DQ890463 AF367600 U94956 DQ887109 DQ890464 DQ887166 DQ887134 DQ887110 DQ890465 DQ887167 U94957 DQ887111 DQ890466 Barker 1002 (BOL) Mafa 4 (BOL, GRA) Barker 983 (BOL) Barker 1009 (BOL) Linder 7004 (BOL) Mafa 1 (BOL, GRA) Barker 1008 (BOL) Merxmuellera papposa (Nees) Conert Merxmuellera rangei (Pilg.) Conert South Africa, W.Cape, Besemfontein South Africa, W.Cape, Silvermine South Africa, E.Cape, Rufanes River mouth DQ218161 Barker 942 (BOL) Linder 5421 (BOL) Merxmuellera guillarmodiae Conert Merxmuellera lupulina (Thunb.) Conert Merxmuellera macowanii (Stapf) Conert South Africa, W. Cape, Malgas Ex cult., New Zealand Barker & Mafa 1759 (GRA) Barker 960 (BOL) South Africa, N. Cape, Nieuwoudtville South Africa, N. Cape, Kamieskroon South Africa, E. Cape, Drakensberg South Africa, W.Cape, Wellington Lesotho, Mafika Lisiu Pass South Africa, E.Cape, Naudes Nek South Africa, E.Cape, Baviaanskloof Namibia, Aus DQ890493 DQ887168 U95075 DQ887112 DQ890467 DQ887169 AF355983 DQ887113 DQ890468 N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) Verboom 603 (BOL) DQ890496 AF019863 U95076 U31438 DQ890469 DQ887171 DQ887135 DQ887114x DQ890470 AF019862 U95077 DQ887115 DQ890471 139 140 Table 1. (Continued) Genus and species Voucher Locality ITS rpoC2 Merxmuellera rufa (Nees) Conert Merxmuellera setacea Barker & Ellis Merxmuellera stereophylla (J.G.Anderson) Conert Barker 1149 (BOL) South Africa, W.Cape, Bainskloof South Africa, W.Cape, Ceres Ha Lejone, Lesotho AF267591 U95078 AF019867 U95079 Merxmuellera stricta (Schrad.) Conert Barker 987 (BOL) Mafa 3 (BOL, GRA) Mafa 2 (BOL) Barker 1159 (BOL) Linder 5497 (BOL) Watson s.n. (CANB) Notodanthonia gracilis (Hook.f.) H.P. Linder Pentameris distichophylla Nees Pentameris glacialis N.P. Barker Pentameris macrocalycina (Steud.) Schweickerdt Pentameris oreophila N.P. Barker Pentameris swartbergensis N.P. Barker Pentameris thuarii Beauv. Linder 5683 (CANB) Pentaschistis airoides (Nees) Stapf ssp. airoides Pentaschistis aristidioides Stapf Pentaschistis aspera (Thunb.) Stapf Linder 5447 (BOL) Barker 1019 (BOL) Barker 988 (BOL) Verboom 223 (BOL) Linder 5490 (BOL) Linder 5456 (NBG) Linder 6971 (BOL) Barker 1158 (BOL) Barker 1164 (BOL) trnL intron DQ887116 DQ890472 DQ887127 AF019871 AF019869 U95126 AF367603 DQ887172 DQ887128 DQ890494 DQ890473 DQ887117 DQ890495 DQ887118 DQ890474 DQ887126 DQ218158 U95080 DQ887136 DQ890490 DQ887137 DQ887173 DQ887138 DQ887139 DQ887140 U95127 DQ887119 DQ887174 DQ890475 DQ890488 DQ887141 AF019865 U95128 DQ887120 DQ890476 N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) Notochloe microdon Domin Ha Lejone, Lesotho South Africa, W.Cape, Rhodes Memorial, Cape Town South Africa, W.Cape, Klein Swartberg Australia, ACT, Canberra Australia,Tasmania, Franklin River South Africa, W.Cape, Cedarberg South Africa, W.Cape, Klein Swartberg South Africa, W.Cape, Outeniqua Mts South Africa, W.Cape, Hex River Mts South Africa, W.Cape, Klein Swartberg South Africa, W.Cape, Tradouw’s Pass South Africa, N. Province, Kamiesberg South Africa, W.Cape, Muizenberg South Africa, W.Cape, Silvermine rbcL Pentaschistis borussica Pilg. Pentaschistis capensis Stapf Phillips 72 (K) Linder 5510 (BOL) Pentaschistis capillaris (Stapf) McClean Pentaschistis colorata Stapf Linder 5439 (BOL) Pentaschistis curvifolia (Schrad.) Stapf Barker 1165 (BOL) Pentaschistis densifolia Stapf Linder 5498 (BOL) Pentaschistis eriostoma Stapf Barker 1758 (GRA) Pentaschistis lima (Nees) Stapf Pentaschistis malouinensis (Steud.) Clayton Pentaschistis patula (Nees) Stapf Pentaschistis pyrophila H.P. Linder Linder 5422 (BOL) Pentaschistis rigidissima Pilg. ex H.P. Linder Pentaschistis triseta (Thunb.) Stapf Pentaschistis velutina H.P. Linder Plinthanthesis paradoxa (R. Br.) S.T. Blake Prionanthium dentatum (L.f.) Henrard Prionanthium ecklonii (Nees) Stapf Linder 5458 (BOL) Linder 5464 (BOL) Barker 1018 (BOL) Linder 5432 (BOL) Linder 5509 (BOL) Linder 5434 (BOL) Linder 5446 (BOL) Linder 5638 (CANB) Linder 5430 (BOL) Linder 5402 (BOL) South Africa, Free State, Golden Gate National Park Unknown South Africa, W.Cape, Klein Swartberg South Africa, W.Cape, Doornbosch South Africa, W.Cape, Garcias Pass South Africa, W.Cape, Silvermine, Cape Town South Africa, W.Cape, Klein Swartberg South Africa, E.Cape, Grahamstown South Africa, N.Cape, Kamiesberg South Africa, W.Cape, Klein Swartberg South Africa, N.Cape, Nieuwoudtville South Africa, W.Cape, Ladismith, Klein Swartberg South Africa, W.Cape, Motangu Pass South Africa, N.Cape, Nieuwoudtville South Africa, W.Cape, Cedarberg Australia, NSW, Wollongong South Africa, N.Cape, Nieuwoudtville South Africa, W.Cape, Clanwilliam DQ887175 DQ890489 DQ887142 DQ887143 DQ887144 DQ887145 DQ887176 U95129 DQ890491 DQ887146 DQ887147 DQ887148 DQ887149 DQ887150 DQ887151 DQ887152 DQ887153 DQ887154 DQ887177 U95361 U31440 DQ890477 DQ887178 DQ887155 DQ887121 DQ890478 AF019866 U95362 141 Galley 44 (Z) N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) Pentaschistis basutorum Stapf 142 Table 1. (Continued) Voucher Locality Pseudopentameris brachyphylla (Stapf) Conert Pseudopentameris caespitosa N.P. Barker Pseudopentameris macrantha (Schrad.) Conert Rytidosperma (=Pyrrhanthera) exigua Rytidosperma nudiflorum (Morris) Connor & Edgar Rytidosperma petrosum Connor & Edgar Rytidosperma aff. pumilum Rytidosperma pumilum (Kirk) Linder Rytidosperma setifolium (Hook. F.) Connor & Edgar Rytidosperma sp. Rytidosperma oreoboloides (F.Mueller) H.P. Linder Rytidosperma vestitum (Pilg.) Connor & Edgar Schismus barbartus (Loefl. ex L.) Thell. Barker 1669 (GRA) South Africa, W.Cape, Betty’s Bay DQ887156 Barker 1670 (GRA) South Africa, W.Cape, Betty’s Bay South Africa, W.Cape, De Hoop DQ887157 Linder 5470 (BOL) AF367598 U96307 GenBank GenBank Linder 5747 (CANB) Linder 5403 (BOL) Barker 1740 (GRA) Verboom 532 (BOL) trnL intron DQ887122 DQ890479 AY691638 Australia, Tasmania, Cradle Mt AF019876 U96314 U31441 DQ218159 AY752480 New Zealand, Mt Somers GenBank GenBank Sands 7265 (= MWC 8877; K) Marsden 132 (= MWC 8881; K) Linder 5359 (BOL) rbcL AY752484 AF019878 U96312 AY752482 Indonesia, Irian Jaya, AY752483 DQ887179 Indonesia, Irian Jaya DQ887180 South Africa, W.Cape, Botterkloof South Africa, N.Cape, Springbok South Africa, W.Cape, Clanwilliam South Africa, W.Cape, Gifberg South Africa, SWCape, Cape Peninsula AF019873 U96308 DQ887123 DQ218168 DQ887124 AF367602 DQ218178 N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) Linder 5693 (CANB, BOL) GenBank Verboom 572 (BOL) Tribolium hispidum (Thunb.) Desv. ITS rpoC2 Genus and species Verboom 530 (BOL) Verboom 558 (BOL) Barker 1163 (BOL) Tribolium uniolae (L.f.) Renvoize Tribolium pusillum (Nees) H.P. Linder & G. Davidse Linder 5402 (BOL) South Africa, W.Cape, Clanwilliam South Africa, W.Cape, Vanrhynsdorp South Africa, W.Cape, Silvermine South Africa, W.Cape, Cape Peninsula AF367601 U96311 U96310 DQ218182 DQ218174 N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) 143 represents all the genera and major clades previously identified in the subfamily (Barker et al. 2000). For the trnL and rbcL data sets, two samples each from four species were included as a test of species monophyly. Additional ITS data of species of Rytidosperma were obtained from GenBank. DNA extraction, amplification and sequencing. For the majority of sequences used here, DNA was extracted following the protocol outlined by Barker et al. (1999, 2003). The ITS regions were amplified and sequenced following Barker et al. (2003), the rpoC2 grass-specific insert as outlined by Barker et al. (1999), the trnL intron using the primers ‘‘c’’ and ‘‘d’’ (Taberlet et al. 1991) and the rbcL gene amplified in two overlapping segments using the primers rbcL Z1 (5¢-ATg TCA CCA CAA ACA gAA ACT AAA gCA AgT-3¢), rbcL-R3 (5¢-AgA CAA ACT AgT ATT TgC gg-3¢), rbcL-F2 (5¢-gTT ATg AgT gTC TAC gCg -3¢) and rbcL 1374-R (5¢-AAT TTg ATC TCC TTC CAT ATT TCg CA-3¢). One of us (AGV) provided some trnL intron data, obtained as described in Verboom et al. (2006). For some taxa, sequence data for different regions was obtained from two (and in one case three) different samples, a problem that is unavoidable when synthesising different sources of data. Sequence alignment. The sequences for each of the individual data sets were imported into MacClade v. 4.06 and aligned by eye. Gaps corresponding to insertion – deletion (indel) events were inserted to conserve the positional homologies of the nucleotides. As has been discussed previously, the rpoC2 region is unusual in its structure, being made up of heptameric (seven-amino acid long) repeat units (Barker et al. 1999, 2003). Because of this, a rule-driven method based on the amino acid sequences of this region was developed for the alignment of this region, described by Barker et al. (1999). This method was also followed here for this partition, and resulted in numerous gaps (usually 21 bp in length) being inserted in the alignment (data available from the first author upon request). Phylogenetic analyses. Parsimony and Bayesian Inference (BI) analyses were conducted for the individual data sets, as well as a combined plastid data set and a ‘‘combined DNA’’ data set (i.e. plastid + nuclear data). The plastid data set (comprising 34 taxa, including outgroup) and the combined data set include one ‘‘fictive taxon’’ (sensu Kellogg and Linder 1995), owing to the combination of rbcL data from Tribolium hispidum 144 N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) with the data for other regions from T. pusillum. Gaps were not encoded. Data sets are available from the senior author upon request, and all newly generated sequences have been deposited in GenBank. Parsimony analyses were conducted using PAUP v4.0b10 (Swofford 2000). One thousand random input analyses were conducted using the parsimony-informative characters, keeping one tree (TKEEP=1) from each of the analyses. A heuristic search was then conducted on the trees thus obtained, MAXTREES set 100,000, using the TBR branch swapping algorithm. FULL HEURISTIC Bootstrap analyses were conducted on each data set using 1000 replicates, with MAXTREES set to 1000. Bayesian inference (BI) analyses were carried out using MrBayes v.3.0b4 (Huelsenbeck and Ronquist 2001). The Metropolis-Coupled Markov Chain Monte Carlo (MCMCMC) analysis was applied with four chains, three cold and one hot, using the INVGAMMA model. Analyses were run for 1,000,000 generations, sampling every 100 generations. The likelihood scores from every 100 generations were plotted to evaluate when stationarity had been reached. From the plots, it appeared that the burn-in phase was complete by 50,000 generations. However, the first 1,000 trees were excluded as burn-in, this 10% exclusion being considered to be conservative. Posterior probabilities (pp) were calculated from the remaining 9,000 trees by means of a majority rule consensus tree produced using PAUP. Outgroup selection. There is considerable doubt about the sister group of the Danthonioideae. While the incorporation of a range of PACCAD taxa could be selected as outgroup, this approach is hindered by excessive sequence variability, especially of the ITS region, making the use of more distant taxa as outgroups difficult if not impossible. We selected Centropodia as the outgroup, as this genus (comprising two species) was previously thought to be part of the tribe Danthonieae (Clayton and Renvoize 1986, Watson and Dallwitz 1992), and is probably quite closely related. However, it lacks haustorial synergids, a synapomorphy for the Danthonioideae (Verboom et al. 1994), and is thus excluded from Danthonioideae by GPWG (2001), who consider it to be incertae sedis. There are indications that it is sister or basal to the Chloridoideae (Hilu et al. 1999; J.T. Columbus, pers. comm.) and the genus has a C4 leaf anatomy as do the Chloridoideae (Ellis 1984). Results The aligned size of each region, the numbers of variable and parsimony informative characters as well as the results from the parsimony analyses for each data set are presented in Table 2, as are the details of the parsimony analyses of each of the data sets. Where multiple samples of a taxon were included, these were all retrieved as monophyletic (Fig. 1a and c). The parsimony and Bayesian analyses usually produced similar topologies for all analyses, and the results from the parsimony analyses are presented here (Figs. 1 and 2), and annotated where necessary to show analysis-specific differences in topology. Table 2. Summary statistics of the parsimony analyses of the data sets Data set No. taxa No. aligned bases No. variable characters No. parsimony informative characters No. trees Tree length CI RI trnL rpoC2 rbcL CpDNA ITS Combined DNA 53 78 41 34 75 34 659 698 1338 2695 710 3405 117 191 134 374 310 640 74 (11.2%) 118 (16.9%) 90 (6.7%) 225 (8.3%) 203 (28.6%) 391 (11.5%) 66590 10,000+ 92 2 14040 11 131 221 201 417 743 1009 0.710 0.665 0.512 0.624 0.463 0.527 0.897 0.937 0.824 0.828 0.760 0.704 (17.7%) (27.4%) (10%) (13.8%) (43.6%) (18.8%) N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) Analysis of the individual data sets usually resolved the same suite of clades in the subfamily, albeit often with low support (irrespective of measure used), and there was no well supported conflict regarding species composition of the clades. The relationships between these clades were not always fully resolved, and there are some instances of moderate conflict between the data sets. Discussion Performance of the chloroplast markers. Previous studies in the grasses have found the trnL intron to be suitable for studies at generic level and above (Guo and Ge 2005, Catalán et al. 2004, Bess et al. 2005). This is confirmed here, as the trnL data presented little difficulty to align, and contained 74 (11.2%) informative characters. While the topology includes several polytomies, the nodes of the ‘‘backbone’’ of the tree are generally well-supported by both bootstrap and Bayesian posterior probability values (hereafter referred to as ‘‘pp’’). Owing to its unusual sequence structure, the rpoC2 region has not been widely used in grass phylogenetic studies, but its utility has been demonstrated by Barker et al. (1999), Duval et al. (2001) and Petersen et al. (2004). Our data set yielded 118 (16.9%) potentially parsimony informative sites. The consensus tree is not well resolved, as there are polytomies in many of the main clades, but the relationships of major clades generally receive moderate to good support (Fig. 1b), with only 14 of the 38 nodes receiving less than 80% bootstrap support, and 13 of these less than 70%. Two nodes in this tree receive conflicting BI pp and bootstrap support. Node A (in Fig. 1b) has <50% bootstrap support but a Bayesian posterior probability of 0.92. Nodes B and C are not retrieved in the BI tree, although the former has 100% and the latter a weak 63% bootstrap support. Conflict between the different measures of node support is not uncommon in situations where nodes have short branch lengths (Alfaro et al. 2003), which is the case in this instance (data not shown). 145 Despite including only 92 informative characters, rbcL sequences resulted in a well-resolved and well-supported tree. Only three nodes collapse in the parsimony consensus tree, and of the 28 nodes indicated in Fig. 1c, 14 receive less than 80% bootstrap support and 11 receive less than 70% support. Two nodes (marked A and B in Fig. 1c) have reasonably high posterior probabilities, but <50% bootstrap support. This partition has a low CI (0.510, Table 2), indicating extensive homoplasy. However, as noted by Källersjö et al. (1999) homoplasic characters (such as third codon positions in rbcL data) may contain phylogenetic structure. The rbcL gene has not been used extensively in grass systematics below the level of subfamily, but our results suggest that rbcL can be useful below the genus level. The topologies derived from each of the three individual plastid data sets are congruent (within the limits of resolution obtained using consensus trees). This suggests that the alignment procedure adopted for the rpoC2 data results in satisfactory assessments of positional homologies, despite the use of many indels in aligning the data, and out of all the chloroplast partitions used here, the rpoC2 region produces the greatest proportion of well-supported nodes. When the plastid markers are combined into a single data set of 34 taxa, only seven of the 29 nodes receive less than 80% bootstrap support, and four receive less than 70% (Fig. 2a). It would thus appear that for the purposes of elucidating relationships along the backbone of the phylogeny, it would not be worthwhile to sample more plastid genes. Performance of the nuclear marker. The ITS data required the use of numerous small indels in order to maintain positional homology. As previously noted by Barker et al. (2003) the sequences of taxa of the Pseudopentameris clade contain both a unique insertion and deletion in ITS1 and an insertion in ITS2. Alignment of the ITS1 sequences for the different species in this clade was problematic, and was further complicated by the fact that in N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) Tribolium hispidum 98 1.0 83 1.0 Danthonia schneideri 61 0.58 -0.61 Danthonia subulata 86 1.0 65 0.95 Merxmuellera dura NPB 983 100 Merxmuellera stricta NPB 1159 1.0 Merxmuellera stricta HPL 5497 88 1.0 85 1.0 Merxmuellera dura HPL 5421 55 0.93 99 1.0 Merxmuellera disticha 51 Chaetobromus involucratus ssp. dregeanus 0.86 Pseudopentameris macrantha 65 1.0 96 1.0 Cortaderia splendens Plinthananthesis paradoxa Danthonia secundiflora -0.89 Danthonia spicata Danthonia compressa Lamprothyrsus peruvianus Cortaderia colombiana Cortaderia rudiuscula 72 1.0 63 -0.92 99 Chionochloa macra -- Chionochloa rigida 76 Pentameris thuarii 0.82 Pentameris distichophylla 99 100 Pentaschistis airoides Prionanthium dentatum 0.77 99 1.0 99 1.0 B 64 0.63 87 1.0 Pentaschistis aspera -0.63 52 A -- 93 1.0 100 1.0 Psch 63 Chio 64 0.94 85 100 99 1.0 51 0.76 A 1.0 Chionochloa flavescens 1.0 88 1.0 98 1.0 0.67 C -- S. Am Cort Danth / A-NZ Cort Notochloe microdon Cortaderia fulvida 98 1.0 -- Pseud Chaetobromus involucratus ssp. involucratus 98 1.0 64 0.97 96 Pentaschistis curvifolia 1.0 Pentaschistis basutorum 97 Merxmuellera arundinacea 1.0 80 CMC Merxmuellera cincta ssp. sericea Merxmuellera decora Merxmuellera lupulina 0.99 100 1.0 Merxmuellera setacea 81 96 1.0 AMMC 0.99 Merxmuellera stereophylla Merxmuellera macowanii NPB 1008 63 0.98 Merxmuellera macowanii PM 1 89 62 0.98 98 1.0 Merxmuellera drakensbergensis 87 1.0 65 0.98 87 1.0 63 88 0.72 1.0 99 1.0 Merxmuellera papposa Merxmuellera rangei 0.96 Centropodia mossamedense b a -- Merxmuellera stricta 73 0.98 93 1.0 66 0.93 0.80 Merxmuellera dura 51 0.99 Danthonia scheideri Danthonia subulata Merxmuellera disticha 68 0.78 A 0.86 94 1.0 Tribolium hispidum 89 75 79 1.0 0.99 Rytido Karroochloa tenella 82 1.0 -- 0.99 72 72 -- Pyrrhanthera exigua 0.89 59 52 0.99 0.71 -- Austrodanthonia laevis 0.98 Schismus barbartus 1.0 Danthonia secundiflora Plinthanthesis paradoxa Lamprothyrsus peruvianus S. Am Cort Pseud 97 Chionochloa macra Chio 1.0 Chionochloa flavescens Chaetobromus involucratus ssp. involucratus Cortaderia colombiana 0.94 91 96 0.99 1.0 Pentaschistis aspera Prionanthium dentatum Pentameris thuarii 1.0 0.92 85 0.93 100 1.0 71 1.0 100 1.0 69 1.0 100 64 Merxmuellera decora 0.89 Merxmuellera lupulina 1.0 69 -97 1.0 90 1.0 78 0.95 99 1.0 76 Merxmuellera rangei 1.0 Merxmuellera papposa 100 1.0 Centropodia glauca Centropodia mossamedense c 63 100 1.0 Merxmuellera macowanii PM 1 99 78 0.91 64 0.91 97 Merxmuellera stereophylla PM 2 1.0 100 1.0 1.0 AMMC 0.99 Merxmuellera stereophylla PM 3 0.98 0.93 Merxmuellera macowanii NPB 1008 Merxmuellera drakensbergensis 1.0 1.0 Merxmuellera arundinacea 99 80 0.96 64 0.73 Merxmuellera setacea 1.0 82 0.99 100 100 CMC 63 0.89 100 1.0 77 1.0 Merxmuellera cincta ssp. cincta 0.85 62 99 Psch 61 0.95 99 -- Aust / NZ Cort Pseudopentameris macrantha 51 100 1.0 100 1.0 --- Cortaderia fulvida 0.96 B -- Danthonia spicata Notochloe microdon 62 0.88 89 1.0 100 1.0 -- -- Danth 54 -- 88 -- Merxmuellera guillarmodae 94 -- -- 0.95 1.0 Rytidosperma nudiflorum Notodanthonia gracilis -- 62 0.95 58 0.97 87 -- d 100 1.0 98 74 1.0 0.93 94 1.0 Centropodia glauca Austrodanthonia laevis Austrodanthonia ariculata Rytidosperma nudiflorum Tribolium pusillum Tribolium hispidum Danthonia vestita ITS1 Rytidosperma sp Danthonia oreoboloides ITS1 Notodanthonia gracilus Joycea pallida Rytidosperma pumilum Rytidosperma aff pumilum Rytidosperma petrosum Rytidosperma setifolium Austrodanthonia caespitosa Karroochloa tenella Danthonia schneideri Danthonia subulata ITS1 Merxmuellera disticha Merxmuellera dura Merxmuellera guillarmodae Merxmuellera stricta Schismus barbatus Chaetobromus involucratus ssp. involucratus Chaetobromus involucratus ssp. dregeanus Pseudopentameris macrantha Cortaderia fulvida Cortaderia richardii Cortaderia turbaria Cortaderia toetoe Cortaderia splendens Notochloe microdon Plinthanthesis paradoxa Danthonia secundiflora Danthonia spicata Danthonia californica Danthonia archboldii Cortaderia colombiana Cortaderia hapalotricha Lamprothyrsus peruvianus Cortaderia selloana GBank Cortaderia selloana Cortaderia jubata Cortaderia araucana Cortaderia rudiuscula Cortaderia nitida Cortaderia bifida Cortaderia sericantha Chionochloa flavescens Chionochloa macra Chionochloa rigida Chionochloa frigida Chionochloa pallens Merxmuellera arundinacea Merxmuellera setacea Merxmuellera cincta ssp. sericea Merxmuellera cincta ssp. cincta Merxmuellera decora Merxmuellera lupulina Merxmuellera rufa Pentameris macrocalycina Pentameris distichophylla Pentameris thuarii Pentaschistis aspera Pentaschistis airoides Pentaschistis curvifolia Prionanthium dentatum Prionanthium pholiuroides Pentaschistis basutorum Merxmuellera drakensbergensis Merxmuellera davyi Merxmuellera macowanii Merxmuellera papposa Merxmuellera rangei CMC 62 0.98 100 1.0 Schismus barbatus Chio Danth / A-NZ Cort Tribolium uniolae Rytido Rytido Tribolium pusilum 95 1.0 AMMC Pseud Karroochloa tenella 0.94 S.Am Cort 1.0 Psch 62 Rytido 67 0.86 60 0.94 93 1.0 100 Notodanthonia gracilis Pseud A / NZ Cort Austrodanthonia auriculata Danth Rytidosperma nudiflorum Centropodia glauca Centropodia mossamadense Merxmuellera papposa Merxmuellera rangei Merxmuellera davyi Merxmuellera macowani Merxmuellera drakensbergensis Merxmuellera rufa Merxmuellera lupulina Merxmuellera decora Merxmuellera arundinacea Merxmuellera cincta ssp. cincta Merxmuellera cincta ssp. sericea Merxmuellera setacea Chionochloa flavescens Chionochloa macra Danthonia secundiflora Danthonia spicata Notochloe microdon Plinthanthesis paradoxa Cortaderia fulvida Cortaderia splendens Cortaderia richardii Cortaderia toetoe Cortaderia turbaria Danthonia archboldii Merxmuellera stricta Merxmuellera dura Danthonia scheideri Merxmuellera disticha Merxmuellera guillarmodiae Tribolium pusillum Tribolium uniolae Joycea pallida Austrodanthonia laevis Rytidosperma pumila Rytidosperma nudiflora Schismus barbartus Karroochloa tenella Pseudopentameris brachyphylla Pseudopentameris caespitosa Pseudopentameris macrantha Chaetobromus involucratus ssp. dregeanus Chaetobromus involucratus ssp. involucratus Lamprothyrsus peruvianus Cortaderia araucana Cortaderia selloana Cortaderia jubata Cortaderia bifida Cortaderia hapalotricha Cortaderia colombiana Cortaderia nitida Cortaderia sericantha Cortaderia rudiuscula Pentameris glacialis Pentameris swartbergensis Pentameris distichophylla Pentameris thuarii Pentameris macrocalycina Pentameris oreophila Prionanthium dentata Prionanthium ecklonii Pentaschistis pyrophila Pentaschistis curvifolia Pentaschistis eriostoma Pentaschistis capensis Pentaschistis densifolia Pentaschistis aspera Pentaschistis triseta Pentaschistis patula Pentaschistis borussica Pentaschistis capillaris Pentaschistis malouenisis Pentaschistis aristidoides Pentaschistis velutina Pentaschistis rigidissima Pentaschistis lima Pentaschistis colorata S. Am Cort 99 1.0 Merxmuellera guillarmodae Chio Austrodanthonia laevis CMC Centropodia glauca Psch 146 AMMC N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) 147 b Fig. 1. Consensus trees of the analyses of the individual data sets. a Consensus of 5124 equally most parsimonious trees obtained from the trnL data set. The node annotated by ‘‘A’’ receives good support from only one of the support indices. b Consensus tree of 10,000 equally most parsimonious trees obtained from the rpoC2 data set. The nodes annotated by ‘‘A’’, ‘‘B’’ and ‘‘C’’ each receive good support from only one of the support indices. c Consensus tree of 92 equally most parsimonious trees obtained from the rbcL data set. d Consensus tree of 2340 equally most parsimonious trees obtained from the nuclear ITS data set. The star symbol indicates the node where the Danthonia clade is placed (also corresponding to nodes in trees shown in Fig. 2). Numbers above the branches are bootstrap values, numbers below the branches are posterior probabilities. Dashes above the branches indicate bootstrap support of <50% and dashes below the branch indicate topologies not retrieved or resolved in the BI tree. Each of the major clades is enclosed in a box, clade codes as follows: AMMC=African Mountain Merxmuellera clade, CMC = Cape Merxmuellera clade, Psch = Pentaschistis clade, Chio = Chionochloa clade, S. Am Cort = South American Cortaderia clade, Danth = Danthonia clade, A-NZ Cort = Australian – New Zealand Cortaderia clade, Pseud = Pseudopentameris clade, Rytido = Rytidosperma clade Chaetobromus involucratus subsp. involucratus, we observed multiple peaks in the sequence trace files, caused by a single base pair deletion in what is presumed to be one of two (or more) paralagous copies in this sample. This region was replaced in the data matrix by ‘‘?’’ characters. Analyses that excluded these regions made no difference to the consensus topology obtained (results not shown), and so the complete data were analysed. The BI analysis indicated that the branch lengths of the taxa in this clade are much longer than in other species, suggesting that these sequences may comprise pseudogenes (tree not shown). Razafimandimbison et al. (2004) suggest that pseudogenes could be characterised by a higher than expected substitution rate in the highly conserved 5.8S region. However, comparison of this region indicated that the 5.8S sequences of Chaetobromus and Pseudopentameris were identical to those of other species in the matrix, suggesting that the ITS sequences obtained here are indeed functional. A second problem encountered in the ITS data set was the presence of multiple or ambiguous bases in the sequence trace files, once again suggesting the presence of multiple paralogs. This phenomenon could be a consequence of nucleotide additivity due to hybridity (Yonemori et al. 2002) or paralogy. This problem was fortunately infrequent, being found predominantly in the sequence from Danthonia spicata. When such instances were encountered, IUPAC ambiguity codes were used. These sorts of problems are what lead Álvarez and Wendel (2003) and Bailey et al. (2003) to consider ITS a less than ideal region to use in phylogeny reconstruction. Cloning and subsequent sequencing of ITS PCR products would provide valuable information on the presence and nature of paralogues and pseudogenes, as well as possible ancestral parent taxa in cases of hybridization. This approach was successfully applied to the diploid and polyploid species of Hordeum L. by Blattner (2004). However, cloning and sequencing costs and time constraints have prevented us from following this strategy, and we adopt the use of ITS data here, heeding the caveats raised by Álvarez and Wendel (2003), Bailey et al. (2003) and Razafimandimbison et al. (2004). Of the 49 nodes shown in Fig. 1d, 25 receive less than 80% bootstrap support, 17 receive less than 70%, and the consensus tree comprises many polytomies. With approximately half of the nodes being well supported, there is clearly a need for another nuclear dataset, ideally based on a suitably variable single-copy nuclear marker. Although the ITS topology has several polytomies along the N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) Centropodia glauca 99 1.0 Austrodanthonia laevis 86 1.0 Karroochloa tenella 68 0.79 77 1.0 Centropodia glauca Karroochloa tenella 61 85 TRIBOLIUM 99 1.0 Austrodanthonia laevis 55 Rytidosperma nudiflorum Rytidosperma nudiflorum -- Tribolium pusillum Schismus barbartus Merxmuellera guillarmodiae 95 1.0 83 0.91 Danthonia schneideri -- -- Merxmuellera guillarmodae 100 Danthonia schneideri 100 1.0 96 1.0 Merxmuellera stricta 64 Merxmuellera dura Schismus barbatus Merxmuellera stricta Merxmuellera disticha -- Merxmuellera disticha 68 0.97 100 1.0 81 1.0 1.0 54 0.82 Pseudopentameris macrantha Cortaderia fulvida 75 Plinthanthesis paradoxa Plinthanthesis paradoxa Danthonia spicata Lamprothyrsus peruvianus 86 Cortaderia colombiana Danthonia secundiflora 100 Lamprothyrsus peruvianus Danthonia spicata Chionochloa flavescens Chionochloa flavescens 100 Chionochloa macra 100 1.0 Chionochloa macra -- 99 100 1.0 71 0.94 Merxmuellera arundinacea 69 Pentaschistis aspera 96 Prionanthium dentatum Merxmuellera setacea -- Merxmuellera cincta ssp. sericea Merxmuellera cincta ssp. sericea Merxmuellera decora 100 Merxmuellera decora 87 1.0 merxmuellera lupulina Merxmuellera lupulina Pentameris thuarii 99 Merxmuellera setacea 100 1.0 Pentaschistis aspera Merxmuellera arundinacea Prionanthium dentatum Merxmuellera drakensbergensis 100 1.0 Merxmuellera macowanii -- Merxmuellera papposa b cpDNA 98 1.0 90 1.0 99 1.0 58 0.86 -- -B 0.72 A 100 1.0 90 1.0 0.55 68 0.96 76 1.0 100 1.0 100 62 0.94 1.0 99 1.0 100 1.0 c 90 1.0 97 1.0 100 1.0 100 1.0 100 1.0 100 1.0 Merxmuellera papposa 100 Merxmuellera rangei 100 1.0 Merxmuellera drakensbergensis 71 Merxmuellera macowani 100 1.0 a Cortaderia colombiana 63 -- Notochloe microdon Pentameris thuarii 100 1.0 Notochloe microdon 77 -- Danthonia secundiflora 97 74 1.0 0.99 100 1.0 66 0.89 Chaetobromus involucratus ssp. involucratus 100 Cortaderia fulvida 90 1.0 83 1.0 -- Pseudopentameris macrantha 99 1.0 100 Merxmuellera dura Chaetobromus involucratus ssp. involucratus 51 76 0.74 0.99 100 1.0 Centropodia glauca Austrodanthonia laevis Rytidosperma nudiflorum Karroochloa tenella TRIBOLIUM Schismus barbatus Merxmuellera guillarmodae Danthonia schneideri Merxmuellera dura Merxmuellera stricta Merxmuellera disticha Merxmuellera rangei ITS BI topology 1.0 1.0 0.96 0.96 Rytido 148 1.0 1.0 1.0 0.99 0.78 Chaetobromus involucratus ssp. involucratus Pseud Pseudopentameris macrantha Cortaderia fulvida Aust / NZ Plinthanthesis paradoxa Cort Notochloe microdon Cortaderia colombiana S. Am Cort Lamprothyrsus peruvianus Danthonia secundiflora Danth Danthonia spicata Pentameris thuarii Psch Pentaschistis aspera Prionanthium dentatum Chionochloa flavescens Chio Chionochloa macra Merxmuellera arundinacea Merxmuellera setacea CMC Merxmuellera cincta ssp. sericea Merxmuellera decora Merxmuellera lupulina Merxmuellera drakensbergensis AMMC Merxmuellera macowanii Merxmuellera papposa Merxmuellera rangei Combined DNA N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) 149 b Fig. 2. a The consensus tree of two equally parsimonious trees (l= 417, ci = 0.624, ri = 0.828) obtained from the parsimony analysis of the plastid partition of the combined DNA data set component. b The consensus tree of two equally parsimonious trees (l= 539, CI = 0.505, RI = 0.616) obtained from the parsimony analysis of the ITS partition of the combined DNA data set component. c Consensus tree of 11 equally most parsimonious trees obtained from the parsimony analysis of the combined DNA data set. The taxon in upper case (Tribolium) is a fictive taxon comprising data from two different species (see text for details). The more resolved topology of the Rytidosperma clade as retrieved by the BI analysis is shown in the inset box to the right of this tree. In all trees the numbers above the branches are bootstrap values, numbers below the branches are posterior probabilities. Dashes above the branches indicate bootstrap support of <50% and dashes below the branch indicate topologies not retrieved or resolved in the BI pp tree. The nodes annotated by ‘‘A’’ and ‘‘B’’ receive support from only one of the support indices. The star symbol (also appearing in Fig. 1d) indicates the node where the Danthonia clade is placed. Each of the major clades is enclosed in a box in Fig. 2c, codes as explained in the caption to Fig. 1 backbone of the tree (Fig. 1d), the nuclear and plastid data sets generally retrieve the same major nodes (i.e. node composition is stable). The only exception to this is the conflicting relationships of the representatives of the genus Danthonia. The nuclear (ITS) analyses resolve four representative species of Danthonia as part of a moderately supported clade that includes all the South American Cortaderia species (69% bootstrap support, pp = 1; Fig. 1d, node marked with a star). In contrast, the combined plastid data places two Danthonia species in a clade with the Australian genera Plinthanthesis, Notochloe and the New Zealand Cortaderia species (81% bootstrap support and pp = 1 in Fig. 2a, node marked with a star). This result is not affected by data set size, as analyses of a reduced ITS data set of taxa, which has data for all regions (Fig. 2b), retrieves the same relationships of the major clades (but often with poor or no support) as the full ITS data. Combined data set. Owing to moderately supported conflict, the combination of the nuclear ITS and plastid data is potentially problematic, and results from the analyses of the combined DNA dataset are disappointing in terms of node support (Fig. 2c). Of the 25 nodes in the combined DNA tree, eight receive less than 80% bootstrap support, and six receive less than 70%. These values are lower than in the plastid analysis (Fig. 2a), but higher than in the ITS analysis. Many of the nodes in the combined DNA topology which lack good support are found within the Rytidosperma clade, which is also not well resolved. However, the BI analysis of the combined dataset does resolve the relationships of this clade, indicated in the inset in Fig. 2c. The results from the combined DNA dataset are thus disappointing and highlight the perils of combining nuclear (ITS) and plastid data. Two nodes (marked A and B in Fig. 2c) receive no bootstrap support and have low BI pp. The first and most obvious explanation for the difference between the nuclear and plastid results is that there is conflict between the data sets, involving the relationships of the Danthonia, the Chionochloa and the Pentaschistis clades, discussed below. While support for the respective positions of the latter two clades is not strong, the support for the conflict in the position and relationships of the Danthonia clade is moderate. This result could be explained by hybridization, chloroplast capture (which can be tested by means of a second nuclear data set) or possible problems with paralogy in ITS, noted above. Major clades of the Danthonioideae. Our expanded analyses indicate that there are nine major lineages in the subfamily, rather than the seven that a previous analysis suggested (Barker et al. 2000). The composition of these lineages is discussed below. 150 N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) The early diverging lineages. All analyses resolved Merxmuellera papposa (a rare endemic to the Eastern Cape of South Africa) sister to M. rangei (an endemic to Namibia) as the earliest diverging lineage in the ingroup. M. papposa is a new addition to the data set, but previous family level analyses of rpoC2 and rbcL data showed Merxmuellera rangei as sister to Centropodia, and these two taxa in turn are placed as sister to the Chloridoideae (Barker et al. 1995, 1999). Studies by Hilu et al. (1999) and Mason-Gamer et al. (1998) indicate that Centropodia glauca has a close affinity with Merxmuellera rangei. Furthermore, unpublished embryological data obtained by one of us (GAV) indicates that M. rangei does not possess haustorial synergids, supporting its exclusion from the Danthonioideae. It would thus probably be appropriate for these two species of Merxmuellera to be placed in a segregate genus, but morphological synapomorphies for these two species have not yet been identified. The GPWG (2001) considered Centropodia to be in the Chloridoideae, but placed as incertae cedis. We thus add Merxmuellera papposa and M. rangei to Centropodia in this category, pending an expanded molecular and morphological study of the PACCAD clade. The Basal Merxmuellera Assemblage: two lineages, or more? Earlier studies revealed a basal paraphyletic grade of species of Merxmuellera, which was called the Basal Merxmuellera Assemblage (BMA; Barker et al. 2000). The additional taxonomic sampling and expanded data sets used here allow us to more precisely elucidate the relationships of these species. The BMA is split into separate African mountain and Cape lineages. While the African mountain taxa form a monophyletic clade in all analyses, the Cape taxa are partly unresolved in the rpoC2 analysis (Fig. 1b) and form a partially resolved grade in the trnL analysis (Fig. 1a). The ITS (Fig. 1d) and rbcL (Fig. 1c) analyses show this clade to be monophyletic. These two clades shall be referred to here as the African Mountain Merxmuellera clade (AMMC) and Cape Merxmuellera Clade (CMC). The taxa of the AMMC (M. drakenbergensis, M. macowanii, M. davyi and M. stereophylla) probably form a monophyletic group, but as not all taxa are represented in all data sets and at least some analyses resolve this clade as paraphyletic, this is still not certain. Data for another Drakensberg species (M. aureocephala) and the two species from Madagascar are lacking. A suitable morphological marker for the clade may be the leaf-blades that break off just above the sheath, with the stub splitting and the parts recurved, but it is not clear whether this is found in all of the species. When fully resolved (as in the ITS and combined DNA analyses, Figs. 1d and 2c, respectively), the CMC comprises two subclades. One of these comprises the species of Merxmuellera in which the culm bases are modified into swollen, underground perennating organs that Linder and Ellis (1990a) consider to be adaptations to fire (M. rufa, M. lupulina and M. decora). The remaining taxa (M. cincta, M. setacea and M. arundinacea) form the second subclade. However, the position of M. arundinacea in particular is unsettled, as it is even placed as basal to the Pentaschistis clade by the trnL data (Node A in Fig. 1a). In the combined DNA analysis, the CMC only receives modest support (62% bootstrap support, pp=0.94; Fig. 2c). We have not been able to identify a morphological synapomorphy for this group as a whole. The Pentaschistis clade. This clade is well supported in all analyses, and is sister to an equally well-supported clade comprising the remaining taxa of the subfamily in the plastid analyses. The monophyly of Pentaschistis and Pentameris has previously been questioned (Linder and Ellis 1990b, Barker 1993, Barker et al. 1999) and this group is currently under intensive study by one of us (CG submitted). However, for data sets that include more than one species per genus (rpoC2 and ITS in particular; Fig. 1b and d), analyses indicate that Prionanthium is embedded within Pentaschistis, but that Pentameris is monophyletic and sister to a clade comprising Pentaschistis N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) and Prionanthium. This clade can be readily diagnosed morphologically: the spikelets are mostly two-flowered (with occasional reductions to a single floret), the palea-nerves do not continue to the tips of the palea, and the lemma setae are inserted in the sinus between the lemma lobes and the central lemma awn. The Chionochloa clade. This clade, including only the species of the Australian - New Zealand genus Chionochloa, is consistently retrieved with good support in all analyses. The five species sampled for the ITS data set had little or no sequence variation between them, as reflected in the polytomy of four of these species (Fig. 1d). The genus is rather distinctive in the Danthonioideae with several unusual features: tough sclerophyllous leaves, tufts of hair along the marginal flaps of the paleas, and ligular disarticulation of the leaf blades. These features also occur occasionally in other genera. However, the presence of long, overlapping microhairs along the bases of the adaxial leaf grooves appears to be unique to Chionochloa and this may constitute the most reliable synapomorphy for the genus. The genus appears to have radiated in the mountains of South Island, with a few species also on North Island, and one each on Lord Howe Island and Mt Kosziusko on the Australian mainland. The Danthonia clade. With a common core of D. spicata and D. secundiflora (North and South American respectively) that are included in all datasets, the individual analyses presented here suggest that the American species of Danthonia are monophyletic. In the trnL analysis, these two species are placed in a clade together with the North America D. compressa, although with poor support. D. californica is included in the ITS analysis, and once again, this species is placed in a clade with D. spicata and D. secundiflora (80% bootstrap support, pp = 0.96; Fig. 1d). Danthonia archboldii from New Guinea has long been a puzzle, as it is morphologically intermediate between Cortaderia and Chionochloa (Connor and Edgar 1974, Clayton and Renvoize 1986), and it has a reproductive 151 system reminiscent of these two genera (Connor 1970). However results based on ITS and rpoC2 data suggest that D. archboldii is a member of Danthonia s.str. Morphologically, Danthonia could be diagnosed by the presence of cleistogamous flowers among the basal leaves, but this character is absent from Danthonia archboldii. Two other species of Danthonia are excluded from this clade. Danthonia subulata (from Ethiopia) has been considered an anomalous species, and examination of the type specimen by one of us (HPL) suggested morphological similarities with Merxmuellera s.l. However, the trnL data indicates that this species is weakly related to another unusual taxon, Danthonia scheideri, from Tibet (61% bootstrap support, pp = 0.58; Fig. 1a). Similarly, the ITS data set (which includes only ITS1 data from D. subulata) also shows a weakly supported sister relationship with D. schneideri (59% bootstrap support, pp = 0.71; Fig. 1d). These two species are placed in the Rytidosperma clade (discussed below) and have affinities with Merxmuellera stricta and M. dura of the southwestern Cape of South Africa. This result is also retrieved by the rbcL data set, where D. subulata and D. schneideri are in a well-supported sister relationship (93% bootstrap and pp = 1) and placed in a clade with the same species of Merxmuellera (66% bootstrap support, pp = 0.93). Although D. subulata is not sampled in the rpoC2 data set, D. schneideri is retrieved as a member of the Rytidosperma clade. Clearly, the Ethiopian and Tibetan species of Danthonia have African relatives, rather than European or North American ones, and should together be considered as separate from Danthonia s.str. Thus, apart from D. archboldii, the genus Danthonia appears to be restricted to species from South and North America and Europe. The Pseudopentameris clade. This small clade comprises just two small African genera: Pseudopentameris and Chaetobromus. Barker (1994) noted similarities in caryopsis morphology that support this relationship. This clade receives good bootstrap support in all analy- 152 N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) ses, but its relationship to other clades is not clear. The combined DNA analysis places it as sister to the New Zealand Cortaderia clade, but with only 58% bootstrap support (pp = 0.86; Fig. 2c). Analyses of the other data sets place it in a (sometimes unresolved) clade with the Danthonia, Rytidosperma and Cortaderia clades. The South American Cortaderia clade. Barker et al. (2003) reported extensively on the polyphyly of Cortaderia, and noted that the genus was split along geographic lines. In this study, all South American species, as well as Lamprothyrsus, are resolved as a monophyletic group, except in analyses of the ITS data set (Fig. 1d), where the clade was broken into three moderate to well supported lineages, these being placed in a strongly supported relationship with the Danthonia clade. The Australia – New Zealand Cortaderia clade. Barker et al. (2003) found Plinthanthesis, Notochloe and the New Zealand species of Cortaderia to be a consistently retrieved clade with moderate to good bootstrap support. However, analyses of the data sets here present a less clear picture, with most analyses still retrieving this relationship (e.g. rpoC2; Fig. 1b and ITS; Fig. 1d). However the rbcL, plastid and combined data include the Danthonia clade with these taxa, although usually this arrangement receives less than 50% bootstrap support (but does have 81% bootstrap support in the plastid topology). In the combined DNA analysis only Cortaderia fulvida is placed between Plinthanthesis and Notochloe, leaving a confusing situation. It is therefore unclear whether the Danthonia clade is nested within the Australian – New Zealand Cortaderia clade as one lineage, or whether both clades should be recognised. The Rytidosperma clade. This clade is the largest in the subfamily in terms of number of genera and species. It is also inadequately sampled for the data sets analysed here, with Austrodanthonia, Notodanthonia (both Australasian) and Rytidosperma (Australasian and South American) notably undersampled. The Australasian genera have been revised by Linder and Verboom (1996), and the South American taxa by Baeza (1996). However, our analyses are completely lacking any South American taxa, an aspect now receiving attention in one of our labs (HPL). The ITS data set includes the largest sampling of Australasian species (owing to the inclusion of some data obtained from GenBank). The relationships of the species of Rytidosperma are largely unresolved by these data, and species of other Australasian genera such as Austrodanthonia, Joycea and Notodanthonia are placed in a terminal polytomy along with the southern African Tribolium, Schismus and Karroochloa (Fig. 1d). However, Verboom et al. (2006) question the monophyly of the latter three genera. Rytidosperma oreoboloides and R. vestita from Papua New Guinea and Indonesia are included in this clade. Although these taxa are only represented by partial ITS data they are included here since the affiliations of the species of ‘‘Danthonia sens. lat.’’ from this region have been the subject of considerable debate, as summarized by Linder and Verboom (1996). Despite the data limitations, these two species are firmly placed in the Rytidosperma clade, and are in no way related to D. archboldii from the same region, or other species of Danthonia. Relationships among the lineages. Our results emphasise the paraphyly of Merxmuellera. Apart from M. rangei and M. papposa (discussed above), the remaining basal clades of Merxmuellera species are difficult to discriminate morphologically, indicating the persistence of symplesiomorphic characters in these lineages. The increased resolution of relationships within what was termed the ‘‘Basal Merxmuellera Assemblage’’ by Barker et al. (2000) is a major step forward. The distinction of the AMMC and CMC (and clades within the CMC) reflects some interesting biogeographical and floristic patterns, and indicates that the subfamily may have had an afromontane grassland ancestry. The plastid (Fig. 2a) and nuclear (Figs. 1d, 2b) data disagree on the relative relationships N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) and positions of the Pentaschistis and Chionochloa clades. The full ITS dataset does not resolve the relationships of these clades 153 (Fig. 1d), as this node in the topology comprises a large polytomy. However, a parsimony analysis of the reduced ITS set retrieves Chionochloa Rytidosperma clade Caryopsis morphology Pseudopentameris clade Austral / NZ Cortaderia clade Complete haustorial synergid syndrome S. Am. Cortaderia clade Mostly with cleistogamous spikelets in the axils of the basal leaves Danthonia clade Setae from lemma sinusses, spikelets with 2 fls, palea with short veins, simplification of haustorial synergids Pentaschistis clade Overlapping microhairs, paleas with flanking hairs Chionochloa clade Haustorial synergids Swollen leaf bases (in part) Cape Merxmuellera clade Recurved leaf stubs ? African Merxmuellera clade M. rangei / papposa clade Outgroup (Centropodia) Fig. 3. Summary cladogram of relationships of the major clades in the subfamily Danthonioideae. Potential synapomorphies for the subfamily and some of the clades retrieved by our analyses are indicated 154 N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) as sister to a clade of Merxmuellera species (Fig. 2a) but with no bootstrap support. Thus, when compared to the plastid data, the analyses of the reduced ITS data set and the combined DNA data set retrieve differing and novel relationships for the relative positions of the Chionochloa and Pentaschistis clades, but support for these nodes in the combined analysis is weak. This suggests that, despite the use of three plastid regions, there is insufficient signal in the data to resolve relationships in this part of the tree, and the alternative scenario retrieved by ITS is not unlikely. This ambiguity could hint at a past series of rapid diversification events associated with these nodes, resulting in the absence of unequivocal phylogenetic signal in our data. The Maximum Likelihood topology (not shown) from this data set indicates that branch lengths in this region of the ‘‘spine’’ of the tree are very short, lending credence to this possibility. The plastid and nuclear data also disagree on the relationships of the Danthonia clade, which is summarized as a polytomy with the Cortaderia, Danthonia, Pseudopentameris and Rytidosperma clades (Fig. 3). The combined DNA analysis follows the ITS data set in placing Danthonia secundiflora and D. spicata as sister to the South American Cortaderia clade, with good BI pp support (96%; Fig. 2c, marked with a star) but only 68% bootstrap support. The Pseudopentameris clade is also associated with this conflict, being unresolved in the plastid topology (Fig. 2a) but placed sister to the New Zealand Cortaderia clade in both the reduced ITS and Combined DNA topologies, although support in these instances is weak. This conflict could have several causes, some biological (e.g. ancestral hybridisation, polyploidy, chloroplast capture) others simply a function of the choice of nuclear region (ITS paralogy, saturation, and other issues as elucidated by Álvarez and Wendel, 2003). Polyploidy is widely documented in the grasses (De Wet 1987), and this subfamily is no exception (e.g. Barker et al. 2003). As noted by Blattner (2004), polploidy can have a marked effect on phylogeny reconstruction, and thus additional data from low or single copy nuclear markers would be of considerable value in testing our ITS phylogeny. Despite these conflicts, the inclusion of two additional data sets (trnL and rbcL) and the expansion of existing ITS and rpoC2 data has resulted in substantial progress being made in disentangling the relationships among the major clades in Danthonioideae. We can now recognize nine major lineages, the composition of which appear to be stable. Furthermore, we have identified morphological synapomorphies for at least some of these lineages, and these features can be used to predict the affinities of species for which we have no molecular data. The relationships between the nine lineages are partially resolved, and the necessity of additional nuclear data sets is essential if further progress is to be made in this regard. NPB thanks the National Research Foundation of South Africa for funding though Grant Unique Number 2053645, and also gratefully thanks the Institute for Systematic Botany, University of Zurich, for hosting him during a sabbatical visit in 2002–2003, during which time the two additional data sets were generated. HPL and CG would like to acknowledge support from the University of Zurich, and CG thanks the Swiss Science Foundation for financial support. We thank all those who have collected material for us, sometimes from remote places, and A.H.A. Haller-Barker for proofreading the manuscript. References Alfaro M. E., Zoller S., Lutzoni F. (2003) Bayes or bootstrap? A simulation study comparing the performance of Bayesian Markov Chain Monte Carlo sampling and bootstrapping in assessing phylogenetic confidence. Molec. Biol. Evol. 20: 255–266. Álvarez I., Wendel J. F. (2003) Ribosomal ITS sequences and plant phylogenetic inference. Molec. Phylogenet. Evol. 29: 417–434. Baeza P. C. M. (1996) Los géneros Danthonia D.C. y Rytidosperma Steud. (Poaceae) en América una revisión. Sendtnera 3: 11–93. Bailey C. D., Carr T. G., Harris S. A., Hughes C. F. (2003) Characterization of angiosperm nrDNA N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) polymorphism, paralogy, and pseudogenes. Molec. Phylogenet. Evol. 29: 435–455. Barker N. P. (1993) A biosystematic study of Pentameris (Arundineae, Poaceae). Bothalia 23: 25–47. Barker N. P. (1994) External fruit morphology of southern African Arundineae (Arundinoideae: Poaceae). Bothalia 24: 55–66. Barker N. P., Linder H. P., Harley E. H. (1995) Polyphyly of Arundinoideae (Poaceae): Evidence from rbcL sequence data. Syst. Bot. 20: 423–435. Barker N. P., Linder H. P., Harley E. H. (1999) Sequences of the grass-specific insert in the chloroplast rpoC2 gene elucidate generic relationships of the Arundinoideae (Poaceae). Syst. Bot. 23: 327–350. Barker N. P., Morton C. M., Linder H. P. (2000) The Danthonieae: generic composition and relationships. In: Jacobs S. W. L., Everett J. (eds.) Grasses: Systematics and evolution. CSIRO, Melbourne, pp. 221–230. Barker N. P., Linder H. P., Morton C. M., Lyle M. (2003) The paraphyly of Cortaderia (Danthonioideae; Poaceae): Evidence from morphology, chloroplast and nuclear DNA sequence data. Ann. Missouri Bot. Gard. 90: 1–24. Bess E. C., Doust A. N., Kellogg E. A. (2005) A naked grass in the ‘‘bristle clade’’: A phylogenetic and developmental study of Panicum section Bulbosa (Paniceae: Poaceae). Int. J. Pl. Sci. 166: 371–381. Blattner F. R. (2004) Phylogenetic analysis of Hordeum (Poaceae) as inferred by nuclear rDNA ITS sequences. Molec. Phylogenet. Evol. 33: 289–299. Catalán P., Torrecilla P., Rodrı́guez J. A. L., Olmstead R. G. (2004) Phylogeny of the festucoid grasses of subtribe Loliinae and allies (Poeae, Pooideae) inferred from ITS and trnL-F sequences. Molec. Phylogenet. Evol. 31: 517–541. Clayton D. W., Renvoize S. A. (1986) Genera Graminum. Her Majesty’s Stationary Office, London. Connor H. E. (1970) Gynodioecism in Danthonia archboldii. Austral. J. Bot. 18: 233–236. Connor H. E., Edgar E. (1974) Names and types in Cortaderia Stapf (Gramineae). Taxon 23: 595– 605. De Wet J. M. J. (1987) Hybridisatoin and polyploidy in the Poaceae. In: Soderstrom T. R., Hilu K. W., Campbell C. S., Barkworth M. E. (eds.), Grass systematics and evolution. 155 Smithsonian Institution Press, Washington D.C., pp. 188–194. Duval M. R., Noll J., Minn A. H. (2001) Phylogenetics of Paniceae (Poaceae). Amer. J. Bot. 88: 1988–1992. Ellis R. P. (1984) Leaf anatomy of South African Danthonieae (Poaceae). IX. Asthenatherum glaucum. Bothalia 15: 153–159. Galley C., Linder H.P. (2006) The phylogeny of Pentaschistis (Danthonioideae, Poaceae) based on cpDNA, and the evolution and loss of complex characters. Evolution (in press). GPWG (The Grass Phylogeny Working Group: Barker N. P., Clark L., Davis J. I., Duval M. R., Guala G., Hsiao C., Kellogg E. A., Linder H. P., Mason-Gamer R., Mathews S., Simmons M., Soreng R., Spangler R.) (2001) Phylogeny and subfamilial classification of the grasses (Poaceae). Ann. Missouri Bot. Gard. 88: 373–457. Guo Y.-L., Ge S. (2005) Molecular phylogeny of Oryzeae (Poaceae) based on DNA sequences from chloroplast, mitochondrial, and nuclear genomes. Amer. J. Bot. 92: 1548–1558. Hilu K. W., Alice L. A., Liang H. (1999) Phylogeny of Poaceae inferred from matK sequences. Ann. Missouri Bot. Gard. 86: 835–851. Hsiao C., Jacobs S. W. L., Barker N. P., Chatterton N. J. (1998) A molecular phylogeny of the subfamily Arundinoideae (Poaceae) based on sequences of rDNA (ITS). Austral. Syst. Bot. 11: 41–52. Huelsenbeck J., Ronquist F. (2001) MrBayes: Bayesian Inference of phylogenetic trees. Bioinformatics 17: 754–755. Källersjö M., Albert V. A., Farris J. S. (1999) Homoplasy increases phylogenetic structure. Cladistics 15: 91–93. Kellogg E. A., Linder H. P. (1995) Phylogeny of Poales. In: Rudall P. J., Cribb P. J., Cutler D. F., Humphries C. J. (eds.) Monocotyledons: Systematics and evolution. Royal Botanic Gardens, Kew, pp. 511–542. Linder H. P. (2005) Danthonieae. In: Mallett W. W. (ed.) Flora of Australia 44B: 19–70. Linder H. P., Barker N. P. (2000) Biogeography of the Danthonieae. In: Jacobs S. W. L., Everett J. (eds.) Grasses: Systematics and evolution. CSIRO, Melbourne, pp. 231–238. Linder H. P., Barker N. P. (2005) From Nees to now – changing questions in the systematics of 156 N. P. Barker et al.: Phylogeny of the Danthonioideae (Poaceae) the grass subfamily Danthonioideae. Nova Acta Leop. 92, Nr 342: 29–44. Linder H. P., Ellis R. P. (1990a) Vegetative morphology and interfire survival strategies in the Cape fynbos grasses. Bothalia 20: 91–103. Linder H. P., Ellis R. P. (1990b) A revision of Pentaschistis (Arundineae: Poaceae). Contrib. Bol. Herb. 12: 1–124. Linder H. P., Verboom G. A. (1996) Generic limits in the Rytidosperma (Danthonieae, Poaceae) complex. Telopea 6: 597–628. Mason-Gamer R. J., Weil C. F., Kellogg E. A. (1998) Granule-Bound Starch Synthase: Structure, function, and phylogenetic utility. Molec. Biol. Evol. 15: 1658–1673. Petersen G., Seberg O., Baden C. (2004) A phylogenetic analysis of the genus Psathyrostachys (Poaceae) based on one nuclear gene, three plastid genes, and morphology. Pl. Syst. Evol. 249: 99– 110. Razafimandimbison S. G., Kellogg E. A., Bremer B. (2004) Recent origin and phylogenetic utility of divergent ITS putative pseudogenes: a case study from the Naucleeae (Rubiaceae). Syst. Biol. 53: 177–192. Swofford D. L. (2000) PAUP: Phylogenetic Analysis Using Parsimony. Version 4.0b10. Smithsonian Institution, Washington, D.C. Taberlet P., Gielly L., Pautou G., Bouvet J. (1991) Universal primers for amplification of three non-coding regions of chloroplast DNA. Pl. Molec. Biol. 17: 1105–1109. Verboom G. A., Linder H. P., Barker N. P. (1994) Haustorial synergids: an important character in the systematics of Danthonioid grasses (Arundinoideae: Poaceae). Amer. J. Bot. 81: 1601–1610. Verboom G. A., Ntsohi R., Barker N. P. (2006) Molecular phylogeny of African Rytidospermaaffiliated danthonioid grasses reveals generic polyphyly and convergent evolution in spikelet morphology. Taxon 55: 337–348. Watson L., Dallwitz M. J. (1992) The grass genera of the world. C.A.B. International, Cambridge. Yonemori K., Honsho C., Kanzaki S., Eiadthong W., Sugiura A. (2002) Phylogenetic relationships of Mangifera species revealed by ITS sequences of nuclear ribosomal DNA and a possibility of their hybrid origin. Pl. Syst. Evol. 231: 59–75. Addresses of the authors: Nigel. P. Barker (e-mail: nbarker@ru.ac.za) and Matthew Gilbert, Molecular Ecology and Systematics Group, Dept. Botany, Rhodes University, Grahamstown, 6140, South Africa. Chloé Galley and H. Peter Linder, Institute for Systematic Botany, Zollikerstrasse 107, 8008 Zurich, Switzerland. Anthony G. Verboom and Paseka Mafa, Dept. Botany, University of Cape Town, P. Bag, Rondebosch, Cape Town, South Africa.