Phylogeny and Subfamilial Classification of the Grasses (Poaceae) Author(s): Grass Phylogeny Working Group, Nigel P. Barker, Lynn G. Clark, Jerrold I. Davis, Melvin R. Duvall, Gerald F. Guala, Catherine Hsiao, Elizabeth A. Kellogg, H. Peter Linder Source: Annals of the Missouri Botanical Garden, Vol. 88, No. 3 (Summer, 2001), pp. 373-457 Published by: Missouri Botanical Garden Press Stable URL: http://www.jstor.org/stable/3298585 Accessed: 10/11/2009 02:35 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=mobot. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. Missouri Botanical Garden Press is collaborating with JSTOR to digitize, preserve and extend access to Annals of the Missouri Botanical Garden. http://www.jstor.org Volume Number 88 3 2001 Annals of the Missouri Botanical Garden PHYLOGENY AND Grass Phylogeny Working Group2,3 SUBFAMILIAL CLASSIFICATION OF THE GRASSES (POACEAE)' ABSTRACT A large collaborative effort has yiel(led a comprehensive study of the phylogeny and a new suhfanilial classification on an integratedlandlrepresentativeset of 62 grasses of the grass family (Poaceae/Graminieae).The stu(ly was (con(luc(ted (0.6% of the species and ca. 8% of the genera) plus four outgroup taxa using six molecular sequence (lata sets ({ndhFl, rbcL, rpoC2, phyB, ITS2, and (;BSSI or waxy), chloroplast restriction site (lata, and(morphologicalidata.A parsimony analysis using 2143 informativecharacters (the comblinedanalysis) resulted in a single most parsimonioustree of 8752 steps with an RI of 0.556 and bootstrap support of > 90% for more than half of the internal no(les. Significant relationships that appear consistently in all analyses of all (lata sets and are strongly supported by the combined analysis include the following: Joinvilleaceae are sister to a monophyletic Poaceae; the earliest (liverging lineages of the Poaceae are Anomochlooideae, Pharoideae, and Puelioideae, respectively; and(all remaining grasses form a clade. Multiple monophyletic clades were recovere(, including Bambusoideae s. str., Ehrhartoideae,Pooideae s.l., Aristidoideae, l)anthonioideae, Chloridoideaes. str., Chloridoideaes.l., Panicoideae, Parianeae, Olyreae s. str., Oryzeae,Stipeae, Meliceae, Lygeum + Nardus, and Molinia + Phragmites. 'The PACCAI)Clade is monophyletic, containing Aristidoideae, Danthonioideae, Arundinoideae s. str., Chloridoideae s.l., Centothecoideae, Panicoideae, Eriachne, Micraira,and Gynerium.Based on the phylogeny, a classification of 11 previously published subfamilies (Anomochlooideae, Pharoideae, Puelioideae, Bambusoideae, Ehrhartoideae,Pooideae, Aristidoideae, Arundinoideae, Chloridoideae, Centothecoideae, and Panicoideae) and 1 new subfamily (Danthonioideae) is proposed. Several changes in the circumscription of traditionally recognized subfamilies are included. Previous phylogenetic work and classifications are reviewed in relation to this classification and circumscription, and major characteristics of each subfamily are discussed and described. The matrix, trees, and updated data matrix are available at (http://www.virtualherbarium.org/grass/gpwg/ default.htm). Key words: cereals, classification, DNA sequence data, evolution, grass, phylogeny, Poaceae. I Work presented here was supportedin part by NSF grants DEB-9806584 and DEB-9806877 to LGC,DEB-9727000 to JID, DEB-9419748 and DEB-9815392 to EAK, and BIR-9508467 to SYM. Miwa Kojima prepared the line illustrations of leaf anatomy and spikelets. We thank T. Cope, J. Everett, S. W. L. Jacobs, S. Phillips, S. A. Renvoize, and P. F. Stevens for helpful comments on the manuscript. 2 This paper is to be cited as authored by the Grass Phylogeny WorkingGroup, or GPWG. The group includes the following members, listed here in alphabetical order; there is no senior author. Nigel P. Barker, Departmentof Botany, Rhodes University, P.O. Box 94, Grahamstown,6140, South Africa; Lynn G. Clark, Department of Botany, Iowa State University, Ames, Iowa 50011-1020, U.S.A.; Jerrold I. Davis, L. H. Bailey Hortorium,Cornell University, 462 Mann Library, Ithaca, New York 14853, U.S.A.; Melvin R. Duvall, Department of Biological Sciences, Northern Illinois University, DeKalb, Illinois 60115-2861, U.S.A.; Gerald F. Guala, Fairchild Tropical Garden, 11935 Old Cutler Road, Miami, Florida 33156, U.S.A.; Catherine Hsiao, 6005 CrossmontCourt, San Jose, California95120, U.S.A.; Elizabeth A. Kellogg, Department of Biology, University of Missouri-St. Louis, 8001 Natural Bridge Road, St. Louis, Missouri 63121, U.S.A.; H. Peter Linder, Institut fur Systematische Botanik, Zollikerstrasse 107, CH-8008, Zurich, Switzerland; ANN. MISSOURI BOT. GARD.88: 373-457. 2001. 374 Annals of the Missouri Botanical Garden The economic and ecological significance of the grasses (Poaceae) has led to widespread interest in their evolution and classification. The cereals, sugarcane, bamboos, and forage and weedy grasses are of pre-eminent importance in human economy. Grasses, which occur in virtually every terrestrial habitat, cover as much as one-fifth of the Earth's land surface (Shantz, 1954). Long recognized as a "natural" group, the grass family includes approximately 10,000 species in over 700 genera (Dahlgren et al., 1985; Tzvelev, 1989; Watson & Dallwitz, 1992; Renvoize & Clayton, 1992). Efforts to produce a comprehensive, formal taxonomic structure of the family began over 200 years ago, while serious study of grass evolution began late in the 19th century. The Grass Phylogeny Working Group (GPWG) was established in 1996 to (1) combine a series of existing data sets to produce a comprehensive phylogeny for the grass family; (2) focus taxon sampling in the development of existing and future data sets; and (3) reevaluate the subfamilial classification of the grass family based on the results of the phylogenetic analyses. We combined and performed cladistic analyses on eight data sets (one structural, four plastome, and three nuclear) from 62 grasses and 4 outgroup taxa. The phylogenetic results and a revised subfamilial classification of the grass family are presented in this paper. and are somewhat more distantly related to Flagellariaceae (Dahlgren et al., 1985; Campbell & Kellogg, 1987; Linder, 1987; Doyle et al., 1992; Kellogg & Linder, 1995; Briggs et al., 2000); this group forms the graminoid clade, a subset of the order Poales (APG, 1998). A sister-group relationship between Poaceae and Joinvilleaceae consistently has been supported (Campbell & Kellogg, 1987; Doyle et al., 1992; Clark et al., 1995; Soreng & Davis, 1998), although rbcL sequence data suggest that + Ecdeiocoleaceae is the sister Joinvilleaceae clade to Poaceae (Briggs et al., 2000). The grass family was recognized as distinctive and coherent long before the term monophyly was ever applied. The fruit (caryopsis) is unique to the family, with the outer integument developmentally fused to the inner wall of the ovary. The embryo is lateral and, unlike most monocot embryos, is highly differentiated, with clear shoot and root meristems, leaves and vascular system. The pollen, as is typical for the whole order Poales, has only one aperture, but in grasses the pollen wall lacks scrobiculi. In all but the earliest-diverging lineage, the grass spikelet consists of a set of distichous bracts, the basal two empty (glumes) with a series of one to many bracts (lemmas) above, each lemma subtending a contracted floral axis on which is borne a presumed prophyll (palea), two or three reduced perianth parts (lodicules), the androecium, and the gynoecium (see discussion under Spikelet). Although subdivision of the grasses into groups we today call tribes began in the 18th century (see reviews in Calder6n & Soderstrom, 1980; Gould & Shaw, 1983; Pohl, 1987), works by Brown (1810, 1814) represent the earliest attempt to define groups of tribes, or what we now call subfamilies. Brown (1814) divided the grasses into the tribe Paniceae (roughly equivalent to the modern Panicoideae) and the tribe Poaceae (roughly equivalent to the Festucoideae of Hitchcock & Chase, 1950) based on spikelet compression, articulation, and floret number. Brown is credited with describing grass spikelets in detail and recognizing them as branched structures, as well as noting the tendency for the panicoids to grow in warm climates and the pooids in cooler climates (Gould & Shaw, 1983; Pohl, 1987). Brown's division of the family into two major groups was formalized by Bentham (1878), was retained by Bentham and Hooker (1883) and Hackel (1887), and persisted well into the 20th 1935; Hitchcock & century (e.g., Hitchcock, REVIEWOF GItASS PHYLOGENYAND CLASSIFICATI)N Historically, the Poaceae were thought to be related to Cyperaceae (Engler, 1892; Cronquist, 1981) based on floral reduction and chemical characters, but evidence accumulated during the past 15 years unequivocally shows that the similarities are convergent. Phylogenetic studies based on morphological and molecular characters show that the grasses are most closely related to Joinvilleaceae, Restionaceae, Anarthriaceae, and Ecdeiocoleaceae, Roberta J. Mason-Gamer,Department of Biological Sciences, University of Idaho, Moscow,Idaho 83844, U.S.A.; Sarah Y. Mathews, Division of Biological Sciences, University of Missouri-Columbia,226 Tucker Hall, Columbia, Missouri 56211, U.S.A.; MarkP. Simmons, The Ohio State University Herbarium, Ohio State University, 1315 Kinnear Road, Columbus, Ohio 43212, U.S.A.; Robert J. Soreng, Department of Botany, Natural History Museum, Smithsonian Institution, Washington, D.C. 20560-0166, U.S.A.; Russell E. Spangler, Departmentof Ecology, Evolution, and Behavior, Universityof Minnesota, 1987 Upper Buford Circle, St. Paul, Minnesota 55108, U.S.A. 3Author for correspondence: Elizabeth A. Kellogg, tkellogg@umsl.edu. Chase, 1950). Several classifications for the grasses based on spikelet and inflorescence morphology were pro- Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae posed in the 19th century (see reviews in Calder6n & Soderstrom, 1980; Gould & Shaw, 1983; Campbell, 1985; Pohl, 1987), with usually nine or ten tribes recognized. Some tribes, for example Paniceae, Andropogoneae, and Bambuseae, contain largely the same genera now as nearly 200 years ago. Others, such as the various "pooid"tribes, included disparate elements and are now seen as artificial. Whether explicit or not, a different perspective on the evolution of grasses and relationships within the family began to emerge by the end of the 19th century. Workerssuch as Celakovsky (1889), Goebel (1895), and Schuster (1910) carefully analyzed spikelet structure and proposed that Streptochaeta, or something very much like it, was representative of the most primitive grasses. With the development of leaf anatomical (Duval-Jouve, 1875; Prat, 1932), embryological (van Tieghem, 1897), and cytological (Avdulov, 1931) data, a profound reassessment of evolutionary relationships among grasses began. Additional data on embryo anatomy(Reeder, 1957, 1961, 1962), starch grains (Tateoka, 1962), lodicules (Jirisek & Jozifova, 1968; Guedes & Dupuy, 1976), and leaf anatomy (Brown, 1958; Metcalfe, 1960) accumulated and were also incorporatedinto evolutionary and classification schemes. Several classification systems were published in the 20th century (e.g., Roshevits, 1937, 1946; Tateoka, 1957; Prat, 1960; Stebbins & Crampton, 1961; Jacques-Felix, 1962; Caro, 1982; Clayton & Renvoize, 1986; Tzvelev, 1989; Renvoize & Clayton, 1992; Watson & Dallwitz, 1992); major ones that are global in scope are compared in Table 1. The number of subfamilies recognized ranged from 2 (Tzvelev, 1989) to 13 (Caro, 1982). All but the Watson and Dallwitz (1992) classification, which is avowedly phenetic, were based on presumed evolutionary relationships. The major change was the subdivision of the old Festucoideae (or Pooideae) into several subfamilies; Panicoideae were retained almost without modification. Other differences among the major classification systems primarily relate to the treatment of Arundinoideae and Bambusoideae. Clayton and Renvoize (1986) in particular published a number of diagrams depicting relationships based on their synthesis of knowledge at that time. These diagrams have served as a starting point for much subsequent work. Phenetic analyses of the grass family generally found groups consistent with the five or six subfamilies commonly recognized by the mid 1980s. Hilu and Wright(1982), in a cluster analysis of morphological and anatomical data, found eight major groups with strong support. Except for the cluster of Diarrhena, Nardus, and Lygeum, the remaining seven clusters corresponded to the subfamilies Festucoideae, Oryzoideae, Arundinoideae, Centothecoideae, Panicoideae, Eragrostoideae, and Bambusoideae. Watson et al. (1985) used the DELTA system to conduct comprehensive phenetic analyses of the family, and their character list continues to be developed. Watson and Dallwitz (1992) initially recognized five subfamilies and subsequently updated their classification to include seven (Watson & Dallwitz, 1999; http://biodiversity.uno.edu/ these are Stipoideae, delta/grass/www/class.htm); Pooideae, Bambusoideae, Centothecoideae, Arundinoideae, Chloridoideae, and Panicoideae. Subsequent phenetic analyses of immunological data (Esen & Hilu, 1989) and plastid DNA reassociation (Hilu & Johnson, 1991) were limited in sampling but in each case produced four major groups. Only within the past 15 years have cladistic methods been applied to questions of grass phylogeny and evolution. The first attempt to produce an explicit hypothesis of relationships was the morphological phylogeny of Kellogg and Campbell (1987), who analyzed 33 characters scored for virtually all grass genera. The pooids (including Stipeae), Panicoideae, Chloridoideae, and Bambusoideae were consistently in their monophyletic analyses, but Arundinoideae were polyphyletic, and the pooid clade formed the basal lineage in the family. Bambusoideae s.l. (including herbaceous tribes such as Anomochloeae, Phareae, Streptochaeteae, and Streptogyneae) were interpreted as monophyletic based on the presence of arm and fusoid cells; several tribes often included in the tra(litional Bambusoideae were placed in other clades (e.g., Brachyelytreae, Diarrheneae, and Phaenospermatidae in the pooid clade). Hamby and Zimmer (1988) and Doebley et al. (1990) published the first molecular phylogenies for the family, based respectively on ribosomal RNA and plastid gene rbcL (ribulose 1,5-bisphosphate 375 carboxylase/oxygenase, large subunit) sequence data. Relatively few taxa were sampled in both studies, but both supported the core Pooideae as well as the group that came to be known as the PACC clade (Davis & Soreng, 1993), containing subfamilies Panicoideae, Arundinoideae, Centothecoideae, and Chloridoideae. The first extensive application of molecular data to grass phylogeny was undertaken by Davis and Soreng (1993), using plastid DNA restriction site variation for 31 taxa representing the six subfamilies of Clayton and Renvoize (1986). This study marked the beginning of wider sampling in the traditional Bambusoideae (= Bambusoideae s.l.), long Table 1. Comparisonof the major 20th century classification systems of the Poaceae. Wa ( Roshevits (1946) Tateoka (1957) Prat (1960) Caro (1982) Clavton & Renvoize Tzvelev (1989) (1986) Bambusoideae Pharoideae Bambusoideae Bambusoideae Bambusoideae Bambusoideae Bam Bambuseae Bambuseae Bambuseae s.l. Bambuseae Bambuseae Bambuseae Arundinarieae Shibataeeae Dendrocalameae Melocanneae Oxytenanthereae Bam Oryzoideae Olyreae Pariana Olyroideae Olyreae Olyreae Parianeae In Olyreae Olyreae Parianeae Buergersiochloeae 77 77 Olyreae Parianeae In Olyreae Olyreae Parianeae ? 9 ? O Oly Anomochlooideae Anomochloeae Anomochloeae Incertae Sedis Anomochloeae Anomochloeae Anomochloeae An Streptochaeteae Streptochaeteae Streptochaeta Streptochaetoideae Streptochaeteae Streptochaeteae Streptochaeteae Str Phareae Leptaspideae Pha In Bambuseae In Bambuseae Atractocarpeae In Bambuseae Pue Gu Streptogyneae Streptogyneae Pooideae Str Oryzeae Diarrheneae Brachyelytreae Oryzeae Diarrheneae Brachyelytreae Or Dia Bra Phareae In Bambuseae In Bambuseae In Olyreae 9 9 9 In Bambuseae In Bambuseae Pooideae See Arundoideae Oryzeae Oryzeae ? See Arundoideae Incertae Sedis Oryzeae 9 Oryzoideae Oryzeae 9 9 Table 1. Continued. Phenospermeae See Arundoideae Phyllorachieae In Oryzeae See Arundoideae Centotheceae Hordeae Brachypodieae Bromeae Festuceae Aveneae Agrostideae Phalarideae Phaenospermatae Phyllorachideae Phaenospermatae Phyllorachideae Pha Phy Ehrhartoideae Ehrharteae Ehrharteae Ehrharteae Eh Centhostecoideae Centothecoideae Centosteceae Cen 9 Incertae Sedis See Arundoideae Micraireae Incertae Sedis Centhosteceae Centotheceae Pooideae Festucoideae Festucoideae Pooideae Triticeae Hordeae Festuceae Agrosteae Festuceae Aveneae Agrostideae Phalarideae Triticeae Bromeae Poeae Aveneae In Poeae Meliceae See Arundoideae In Festuceae Stipeae Nardeae Lygeae Monermeae See Arundoideae See Arundoideae See Arundoideae Monermeae Stipeae Incertae Sedis Incertae Sedis Meliceae Brylkinieae Hainardieae Stipeae Nardeae Lygeae Poo Triticeae Brachypodieae Bromeae Poeae In Phleeae Ampelodesmeae Phleeae Meliceae Brylkinieae 9 Stipeae Nardeae Lygeeae Tri Bra Bro Poe Av In Ses Me In See See See Beckmannia Arundoideae Stipeae Nardeae Lygeae Phragmitiformes Phragmnitoideae Arundinoideae Aru Sti Ste Na Lyg Table 1. Continued. Arundineae In Aveneae Arundineae Danthonieae Arundineae Arundineae Danthonieae ,, Arundineae ,, Aru Da Spa Cy Eri Micrairoideae See Pharoideae Incertae Sedis Micraireae Micraireae Mi Aristideae Aristideae Thvsanolaeneae Aristideae Thysanolaeneae Ari In Eragrostoideae Chloridoideae Micraireae Aristidoideae Aristideae Thysanolaeneae Aristideae Incertae Sedis Coleantheae Brachyelytreae Meliceae Glycerieae Centotheceae Streptogyneae Ehrharteae Phaenospermeae Garnotieae Arundinelleae Arundinelleae Eragrostoideae Chloridoideae Pappophoreae Pappophoreae Pappophoreae Chlorideae Zoysieae Chlorideae Lappagineae Spartineae In Chlorideae Chlorideae Zoysieae See Panicoideae Unioleae Panicoideae Thysanolaeneae Eragrosteae Eragrosteae In Eragrostideae Pappophoreae Orcuttieae Cynodonteae Eragrostideae Ch In Cynodonteae Cynodonteae ?? Tri Pap Or Ch Table 1. Continued. Sporoboleae In Chlorideae Sporoboleae Leptureae various isolated genera Panicoideae In Eragrosteae In Festuceae Paniceae Boivinelleae Melinideae In Paniceae In Paniceae In Zoysieae In Zoysieae In Melinidae Arundinelleae Andropogoneae Maydeae 9 ? Isachneae Paniceae Bovinelleae Melinideae Anthephoreae Lecomtelleae Trachyeae 9 9 See Arundoideae Andropogoneae Maydeae Panicoideae 9 9 Isachneae Paniceae Boivinelleae Melinideae Panicoideae Panicoideae Steyermarkochloeae Eriachneae Hubbardieae Isachneae Paniceae ,, Pan Steyermarkochloeae See In Arundineae See 9 In Isachneae Isa Paniceae Pan ,, ,, 9 9 ,? 9 ,, ,, Arthropogoneae See Phragmitiformes Andropogoneae Maydeae various isolated genera including Phaenosperma Incertae Sedis Anomochloa Centotheca Ehrharta Lygeum,Nardus Micraira Streptogyna Thysanolaena and other genera Arundinelleae Andropogoneae I,, Arundinelleae Andropogoneae ,, Neu In Aru An Ma 380 Annals of the Missouri Botanical Garden presumed to include the most ancestral elements of the grass family. Davis and Soreng's (1993) results supported an expanded pooid clade, the PACC (now PACCAD) clade, and suggested that the traditional Bambusoideae were not monophyletic. Nadot et al. (1994) analyzed sequences of the plastid gene rps4 (ribosomal plastid small subunit, protein 4) for 26 genera of grasses. Their sampling was heavily weighted towardthe pooid grasses, but they did include three genera of woody bamboos and Zizania and Oryza of the ehrhartoids (oryzoids). They recovered both a monophyletic pooid clade (including Stipa) and the PACC clade. The bambusoid/oryzoid taxa were paraphyletic and formed a polychotomy with the PACC clade. Cummings et al. (1994), using sequence data from the plastid rpoC2 (RNA polymerase II, [" subunit) gene, sampled only 13 genera, but did derive a monophyletic PACC clade and a monophyletic pooid clade. The rbcL sequence analysis of Barker et al. (1995) focused on the subfamily Arundinoideae. Both the PACCand pooid clades were shown to be monophyletic, although the traditional Arundinoideae appeared as polyphyletic. Bambusoideae, represente(dby a woody bamboo (Bambusa) and Ehrhartoideae (Zizania and Oryza), were paraphyletic to the rest of the family. Clark et al. (1995) were the first to include a broad sample of bambusoid and ehrhartoid taxa. Using ndhF (NADH dehydrogenase, subunit F) sequence data, they confirmed the polyphyly of the traditional Bambusoideae and demonstrated that Anomochloa and Streptochaeta, two broad-leaved Neotropical forest genera, formed the earliest diverging branch of the family, with Pharus, another broad-leaved tropical forest genus, constituting the next most basal branch. Their results also confirmed strong support for monophyly of the PACC clade, an expanded pooid clade (including Stipeae, Phaenospermatideae, Brachyelytreae, and Diarrheneae), a derived, monophyletic core bambusoid clade (Olyreae + Bambuseae), and the polyphyly of the traditional Arundinoideae. They also recovered a weakly supported clade including the core bambusoids, ehrhartoids, and pooids, which they named the BOP clade (here updated to the BEP Clade based on nomenclatural priority of Ehrhartoideae over Oryzoideae). They concluded that many features previously used to define the traditional Bambusoideae, including the presence of arm and fusoid cells and pseudopetiolate leaf blades among others, were probable synapomorphies for the family. The rbcLstudy of Duvall and Morton(1996) confirmed the basal placement of Anomochloa, as well as the monophyly of the core Bambusoideae, in addition to supporting the PACC and pooid clades. The topology recovered by Liang and Hilu (1996) from analysis of matK (maturase K) sequence data was similar to the rbcL topologies, with the PACC and pooid clades sister to each other, and Oryza sister to that clade and a woody bamboo sister to the whole family. By this time, reassessment of subfamilial classification was necessary; Clark and Judziewicz (1996) resurrected Anomochlooideae and Pharoideae to accommodate the basal lineages of the family, which could not be retained in a monophyletic Bambusoideae. Soreng and Davis (1998) combined a structural data set (including morphological, anatomical, chromosomal, and biochemical characters as well as structural features of the chloroplast genome) and an expanded chloroplast restriction site data set to analyze phylogenetic relationships within the grass family. They confirmed the basal positions of Anomochlooideae and Pharoideae, monophyly of an expanded Pooideae, monophyly of Panicoideae, Centothecoideae, and Chloridoideae, and polyphyly of the traditional Arundinoideae. The core Bambusoideae, supported as monophyletic in other analyses along with the ehrhartoid grasses, appeared as a set of clades paraphyletic to the [Brachyleytrum + (Pooideae + PACC)] clade. Soreng and Davis (1998) also identified structural synapomorphies for major clades, including, for example, loss of the epiblast and gain of an elongated mesocotyl internode in the PACC clade. Barker et al. (1999) used sequences of the grassspecific insert in the chloroplast gene rpoC2 (hereafter referred to only as rpoC2) to study relationships among a broad sample of "arundinoid" taxa. They were the first to include molecular data on such traditionally arundinoid genera as Centropodia, Merxmuellera, Notochloe, Tribolium, Monachather, Pentaschistis, Prionanthium, Cortaderia, and Spartochloa. Because Arundinoideae were known to be polyphyletic, previous classifications were not helpful in placing these genera. rpoC2 sequences of Anomochloa and Streptochaeta could not be aligned with those of other grasses, so their basal position could not be tested. Relationships among the Bambusoideae, Pooideae s.l., and the PACC clade varied depending on the analytical method and inclusion of phylogenetically informative insertion/deletion characters. Consistent with previous studies, they identified Panicoideae and Chloridoideae as monophyletic. They showed clearly that a large clade corresponding to Danthonioideae is monophyletic and that this corresponds at least in part to the clade with haustorial synergids identi- Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae fled by Verboom et al. (1994). They also showed that the genus Merxmuellera is polyphyletic, with one species, M. rangei, most closely related to Centropodia and the chloridoids. Hilu et al. (1999) sequenced the chloroplast gene matK for 62 species of Poaceae and produced a tree that was quite similar to those found in previous studies. Streptochaeta and Anomochloa were the earliest diverging lineages, although paraphyletic instead of monophyletic. The matK data supported a PACC clade and a clade including Pooiand Bambusoideae. deae ( Oryzoideae Ehrhartoideae) was sister to the PACC clade, rather than the pooid/bambusoid clade, but this was not sistent with certain broad phylogenetic groups (Kellogg, 1998). Unique combinations of linkage groups are synapomorphic for subfamilies Pooideae (Moore et al., 1995; Gale & Devos, 1998), Panicoideae (Moore et al., 1995; Gale & Devos, 1998; Wilson et al., 1999), and Ehrhartoideae (Kennard et al., 1999). In addition, unique linkages support monophyly of Triticeae (Devos et al., 1993) and Andropogoneae (Wilson et al., 1999). Phylogenetic analyses of individual molecular data sets within the last decade have converged on a set of well-supported relationships within Poaceae. Changes in the circumscriptions of subfamilies, and in the number of subfamilies recognized, clearly are necessary. The GPWG analyses presented here provide robust support for the major clades within the grass family, and provide the basis for the first family-wide subfamilial classification based on an explicit phylogenetic hypothesis. strongly supported. Zhang (2000) used the intron in the chloroplast gene rpll6 (ribosomal protein 16) to construct a phylogeny of the grasses and confirmed (again) the early divergence of Anomochloa and Streptochaeta, although his data did not support the monophyly of the pair. The next branch was Pharus. The rpll6 data supported a PACC clade and a BEP Clade; Puelia olyriformis was sister to the BEP Clade, with modest bootstrap support. Analyses of nuclear sequence data have provided results complementary to those obtained for molecular plastid data sets. Mathews and Sharrock (1996) and Mathews et al. (2000) sequenced loci in the phytochrome gene famlily antl resolved a topology similar to that (lerived from IIndtlhsequence data, although the phytochrome data provided significantly stronger support for the B3EP Clade than did the ndhF data. Additionally, the basal positions of Anomochloa, Pharus, and Puelia (as noted in Zhang, 1996) were confirmed. Hsiao et al. (1999) within the inferred phylogenetic relationships tranon of the internal based sequences grasses scribed spacer (ITS) region of nuclear ribosomal DNA. As in the other studies, Streptochaeta and Pharus were resolved as the basal lineages, and monophyly of the PACC clade and monophyly of the Pooideae were strongly supported. Unlike previous studies, however, some of Hsiao et al.'s (1999) analyses found that the traditional Arundinoideae were monophyletic. Combined analysis of sequence data from two chloroplast genes (ndhF, rbcL) and one nuclear gene (phytochrome B) provided strong support for the placement of Puelia + Guaduella as the next most basal lineage of the family after Anomochlooideae and Pharoideae (Clark et al., 2000). These results necessitated the description of a new subfamily, the Puelioideae. Mapping studies of the nuclear genome are in their infancy, but genome rearrangements are con- 381 MATERIALSANI) METHODS OF THlE CPWC O()RGANIZATION The Grass Phylogeny Working Group was formed explicitly to combine available (lata on the phylogeny of the grass family an(l to use these (lata to )rop)ose a new classification. Most contributors had already published papers on grass phylogeny an(d were invite(l to contri)ute their (lata, both pul)lished and unpublished. Each contributor retaine(l control over his or her data an(l was free to publish at any time, but the group agreed that the entire data set woul(l be published as a single paper. Most of the collaboration has been conducted via e-mail, an(d the entire group has never met in a single place. This may serve as a model for future collaborations in plant systematics. A list of taxa was drawn up in 1995 by LGC, JID, and EAK to improve parallel sampling for all data sets (Table 2; Appendix I). This list was chosen to include as many of the major lineages in the family as possible, based on our knowledge from previous studies. Although sampling of taxa is still not perfectly parallel, many sequences were generated for this particular set of taxa. DNA was exchanged as necessary among members of the group. The list was expanded slightly based on results acquired during the study. All sequences available at the end of 1997 were assembled by EAK into a single large data set in NEXUS format. The data set was then distributed via e-mail to all participants, who had the opportunity to comment on it. A "final" version of the data set was then distributed. Based on the results of the first round of analyses (GPWG, 2000), the Annalsof the MissouriBotanicalGarden 382 Table 2. Summary of genes and taxa included in the combined analysis. Taxa are listed approximately in the order in which they appear in Figure 1. cp rs = chloroplast restriction sites; GBS = GBSSI; struc. = structural data; * = composite taxon, represented by sequences from several genera; # = composite taxon, represented by sequences of different species within the same genus (as in Appendix I). For details of species, authorities, origina1 publications, and GenBank accession numbers, see Appendix I. Merxmuellera m. = Merxmuellera macowanii, Merxmuellera r. Merxmuellerarangei. Genus Flagellaria Elegia# Baloskion Joinvillea# Anomochloa Streptochaeta# Pharus# Guaduella Puelia Eremitis Pariana Lithachne# Olyra# Buergersiochloa Pseudosasa* Chusquea# Streptogynca Ehrharta# Oryza Leersia# Phaenosperma Brachyelytrum Lygeum Nardus Anisopogon Ampelodesmos Stipa# Nassella# Piptatherum# Brachypodium# Melica# Glyceria# Diarrhena# Avena* Bromus# Triticum* Aristida# Stipagrostis Amphipogon# Arundo Molinia* Phragmites Merxmuelleram. Karroochloa Danthonia# Austrodanthonia Merxmuellerar. Centropodia Eragrostis# Uniola Zoysia# cp rs ndhF phyB rbcL x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x rpoC GBS x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x ITS X X x x x x x x x x x x x x x x x x x x X x x x x x x x x x x x x x X X X X X x x x x x x x X x X X X X X X X X X X X X X X X X X X X struc. x x x x x x x x x x x x x x x x x x x x x x x x x x x X X X X X X X X X X X X X X X X X X X X X X x x Volume 88, Number 3 2001 Table 2. Grass Phylogeny Working Group Phylogeny and Classification of Poaceae 383 Continued. Genus cp rs ndhF Distichlis Pappophorum* Spartina# Sporobolus# Eriachne# Micraira# Thysanolaena Gynerium Chasmanthium# Zeugites Danthoniopsis# Panicum# Pennisetum# Miscanthus* Zea X X X X X X X phyB rbcL rpoC X X X GBS ITS X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X taxon list was expanded to include several more danthonioid taxa, and the matrix was recompiled by JID. Although all participants in the GPWG were invited to undertake data analyses and comment on the final version of matrices, abstracts, and text, this was not required. Thus the analyses and text of this paper reflect largely the work of LGC, JID, and EAK, with input from several members of the group. The GPWG website was created and is maintained by GFG. TAXA The taxa used in this analysis include four genera representing the families Flagellariaceae, Restionaceae (two genera), and Joinvilleaceae as out- X X X X X X X X X X X X X X X X X X X X X X struc. X X X X X X X X X X X X X X X groups. This choice was based on previous work (summarized in Kellogg & Linder, 1995) indicating that these represent the closest relatives of the grasses. The results of Briggs et al. (2000) suggest that the Ecdeiocoleaceae should be included in outgroup comparisons for the grasses in the future. Within the grass family, 62 exemplar species were chosen to represent the commonly recognized subfamilies Anomochlooideae, Arundinoideae, Bambusoideae, Centothecoideae, Chloridoideae, EhrPhahartoideae Panicoideae, (=Oryzoideae), roideae, and Pooideae, as well as species from several genera whose placement was uncertain (Amphipogon, siochloa, Anisopogon, Brachyelytrum, Danthonia, Danthoniopsis, BuergerEriachne, Table 3. Tree statistics for subsets of the (lata matrix. The percent missing data is the number of cells that are missing for the particularblock when included in the total data matrix, and thus is equivalent to the number of missing taxa times the total number of informativecharacters plus missing data within sequences for scored taxa. Chloroplast r.s. = Chloroplastrestriction sites. Data set Morphological Chloroplastr.s. ndhF phytochromeB rbcL rpoC2 GBSSI ITS cp sequence data All cp data Nuclear All molecular Combined data # Taxa 66 45 65 40 37 34 19 47 66 66 57 66 66 % Missing Total # # Inf. characters characters data 53 364 2210 1182 1344 777 773 322 4331 4695 2277 6972 7025 50 293 680 417 213 150 213 127 1043 1336 757 2093 2143 16.0 42.2 4.7 45.5 44.8 49.9 71.2 28.8 25.2 26.5 51.9 33.6 33.2 Length CI RI # Trees 227 939 2894 1997 651 374 720 745 3952 4903 3513 8488 8752 0.300 0.312 0.379 0.369 0.448 0.503 0.479 0.349 0.399 0.381 0.382 0.378 0.375 0.690 0.569 0.582 0.522 0.660 0.648 0.504 0.541 0.597 0.589 0.512 0.554 0.557 38,000+ 7 16 1 1 33 1 24 8 3 8 6 1 384 Annals of the Missouri Botanical Garden Table 4. List of structuralcharacters and states. The first number in parentheses indicates the number of the same character in Soreng and Davis (1998), and the second number indicates the number of the same character in Kellogg and Campbell (1987); a "-" indicates that the characterwas not used in one or both of those analyses. Culm 1 (1;-). Perennating woody culms: 0 = absent; 1 = present. 2 (2;-). Culm internodes: 0 = solid; 1 = hollow. Leaf 3 (3;-). Leaf sheath margins:0 = free; 1 = fused for at least 1/4 of length. 4 (4;-). Adaxial ligule type: 0 = membrane (with or without fringe of hairs); 1 = fringe of hairs only. 5 (5;-). Abaxial (contra-) ligule: 0 = absent; 1 = present. Leaf blade: 0 = absent; 1 = present. 6 (-;-). 7 (6;-). Pseudopetiole: 0 = absent; 1 = present. Spikelet 8 (-;-). Floret with a structureidentifiable as a palea, this recognized as present when a flowerarises on a contracted axis above an enshrouding prophyll (or something like it), in the axil of a lemma: 0 = absent; 1 = present. 9 (-;-). Spikelet pairs: 0 = absent; 1 = present. 10 (7;-). Pedicel of spikelet: 0 = absent; 1 = present. 11 (8; 3, 4). Proximal female-sterile florets in female-fertile spikelets: 0 = absent; 1 = present. 12 (9; 6). Number of female-fertile florets per female-fertile spikelet: 0 = two or more; 1 = one. 13 (10;-). Awn or mucro on fertile or sterile lemma: 0 = absent; 1 = present. 1 awn; 2 = 13-23 awns (unique to Pappophoruzm); 3 3 awns. 14 (-;-). Number of awns: 1 15 (11;-). Awn attachment:0 = terminal / subterminal; 1 = from a sinus; 2 = dorsal. 16 (12; 1). Disarticulation above glumes: 0 = absent; 1 = present. 17 (13;-). Germinationflap in lemma: 0 = absent; 1 = present. Flower 18 (14;-). Lodicules: 0 = absent; 1 = present. 19 (15; 7). Lodicule number:2 = two; 3 = three. 20 (16;-). Fusion of anteriorpair of lodicules: 0 = free; 1 = fused. 21 (17; 8). Distally membranousportion of lodicule: 0 = absent; 1 = present. 22 (18; 9). Lodicule vascularization:0 = very faint to absent; 1 = prominent. 23 (19; 1(). Inner whorl, posterior stamen: 0 = absent; 1 = present. 24 (19; 10). Inner whorl, anterior stamen pair: 0 = absent; 1 = present. 25 (19; 10). Outer whorl, anteriorstamen: 0 = absent; 1 = present. 26 (19; 10). Outer whorl, posterior stamen pair: 0 = absent; 1 = present. 27 (-;-). Anthers tetrasporangiate,dithecal = 0; anthers bisporangiate, monothecal = 1. 28 (20;-). Styles fused at least at base: 0 = absent; 1 = present. 29 (21;-). Number of stigmas: 1 = one; 2 = two; 3 = three; 4 = four. 30 (22;-). Highest order of stigmatic branching present: 1 = simple (unbranched,or with branches comiposedof single elongate papillate receptive cells, or with very short branches composed of a few papillate receptive cells, but in the latter case the stigmas linear in outline); 2 = primary (branches well developed, composed of series of dispersed 3= papillate receptive cells, with secondary branches absent or minimally developed, stigmas lanceolate or [broader); secondary (secondary to tertiarybranches well developed, branches composed of series of dispersed papillate receptive cells). 31 (23;-). Number of locules and ovules per pistil (all three families have one ovule per locule): 1 = oile; 2 = two; 3 = three. Embryogeny 32 (-;-). Haustorial synergids: 0 = absent; 1 = present. Fruit and Embryo 33 (24; 11). Hilum: 0 = long-linear, > 1/3 length of grain; 1 = nonlinear, < 1/3 length of grain, elliptical or broader to punctiform. 34 (25;-). Embryo position and structure:0 = embedded, simple; 1 = lateral, grass-type. 35 (26; 15). Embryo epiblast: 0 = absent; 1 = present. 36 (27; 16). Embryo scutellar tail: 0 = absent; 1 = present. 37 (28; 17). Embryo mesocotyl internode: 0 = negligible; 1 = elongate. 38 (29; 18). Embryonic leaf margins:0 = meeting; 1 = overlapping. 39 (30;-). Endosperm lipid: 0 = absent; 1 = present. 40 (31;-). Endosperm starch grain syndromes: 0 = Triticum-type(simple grains only, dimorphic in size, round or lenticular, free); 1 = Festuca-type (highly compound grains present, with or without simple grains also present); 2 = Andropogon-type(simple and compound grains both present, the latter consisting of few granules); 3 = Panicum-type (simple grains only, uniform in size, small to medium, angular or sometimes smooth walled, densely packed); 4 = Brachyelytrum-type(simple only, large). Seedling 41 (32; 20). Lamina of first seedling leaf: 0 = absent; 1 = present. Volume 88, Number 3 2001 Table 4. Grass Phylogeny Working Group Phylogeny and Classification of Poaceae 385 Continued. Vegetative Anatomy 42 (-;-). Differentiation of leaf epidermal cells into long and short (cork) cells: 0 = absent (i.e., cells ? undifferentiated); 1 = present (Campbell & Kellogg, 1987). 43 (34; 21). Multicellular microhairs:0 = absent; 1 = present. 44 (35; 22). Occurrence in multicellular microhairs of a broad, short terminal cell, often with a longer basal cell, the walls of the terminal and basal cells similar in thickness: 0 = absent; 1 = present. 45 (36; 31). Arm cells: 0 = absent; 1 = present. 46 (37;-). Fusoid cells: 0 = absent; 1 = present. Chromosomes Base chromosomenumber is same as state number except that 0 = 10; 1 = 11; 2 = 12; 3 = 13; 4 = 18; 47 (-;-). 5 = 19. Biochemistry 48 (38; 30). Carbonfixation pathway:0 = C,; 1 = C4 NADP-ME classical-type; 2 = C4 NADP-ME Aristida-type;3 = C4 NAD-ME;4 = C4 NADP-ME Arundinelleae-type;5 = C4 NADP-ME Eriachne-type. 49 (39; 30). Carbon fixation PCK: 0 = absent; 1 = present. Deletion in Phytochrome B 50 (-;-). 3-bp DNA deletion in phytochromeB: 0 = 3-bp DNA present (i.e., non-deleted state); 1 = DNA absent (i.e., deleted state; the deleted codon is at position 402 in the alignment of Mathews et al., 1995). Chloroplast Genome Structure 51 (40;-). 6.4 kb inversion in the large single-copy region of the chloroplast genome, relative to the gene arrangement in Nicotiana: 0 = absent; 1 = present. 52 (41;-). trnT inversion in the large single-copy region of the chloroplast genome, relative to the gene arrangement in Nicotiana: 0 = absent; 1 = present. 53 (42;-). 15 bp insertion in ndhF at position 101951 of the chloroplast genome of Oryza sativa: 0 = absent; 1 present. Lygeum, Micralira, Nard(is, Pari(t(a, Phaenosperma, Puelia, Streptogyna, 7Thysanolaelna). For 31 of the terminal taxa in the matrix, all molecular data were taken from a single species; for an additional 27, data were from two or more species of the same genus (noted l)y # in Table 2; Appendix I). In eight cases, however, data from several genera were combined to create a "conglomerate" taxon (asterisks in Table 2). For example, although one listed representative of the Andropogoneae is Miscanthus, there is no rbcL sequence available for that genus. There is, however, a sequence for Sorghum. Thus the Sorghum sequence for rbcL was combined with the Miscanthus sequences for ndhF, creating a fictive taxon, an approach used previously by Kellogg and Linder (1995). This assumes that both genera are part of a monophyletic higher-level group (in this case, Andropogoneae, which are certainly monophyletic; Spangler et al., 1999). The results of such combinations are potentially misleading, in that they assume certain combinations of characters that may not ever actually occur in a single plant. We feel that the number of characters involved, however, is small, and the addition of phylogenetically informative characters by including the line of data outweighs the risk of misleading results. Any subsequent studies, particularly those for which there are more than two representatives of taxa combined here, should break up each conglomerate taxon into real species (i.e., exempflar taxa). 'he numb)er of taxa was (dictated by the numb)ers of available sequences in the largest of the original data sets (rdh F andt chloroplast restriction sites). Recent work on large phylogenies suggests that phylogenetic accuracy is improve(d by a very dense sample of taxa (e.g., Hillis, 1996, 1998; Graybeal, 1998). Producing a large data set with perfectly parallel sampling, however, would have required either a centralized effort in a single lab, or a formal, coordinated, and separately funded effort among multiple labs, rather than the decentralized approach used here. CHARACTE RS The data matrix included 7025 characters assembled from the following sources: 1. NADH dehydrogenase, subunit F (ndhF)Clark et al. (1995, 2000); Davis et al. (this paper); Spangler et al. (1999). 2. Ribulose 1,5-bisphosphate carboxylase/oxygenase, large subunit (rbcL)-Barker et al. (1995); Barker (1997); Doebley et al. (1990); Duvall and Morton (1996). 3. RNA polymerase II, [3" subunit (rpoC2)-Cummings et al. (1994); Barker et al. (1999). 386 Annals of the Missouri Botanical Garden 4. Chloroplast restriction sites-Davis and Soreng (1993); Soreng and Davis (1998). 5. Phytochrome B (phyB)-Mathews and Sharrock (1996); Mathews et al. (2000). 6. Internal transcribed spacer of the nuclear ribosomal RNA (ITS)-Hsiao et al. (1998, 1999). 7. Granule bound starch synthase I (GBSSI, or et al. (1998). waxy)-Mason-Gamer 8. Morphology-Soreng and Davis (1998, and additional members of the GPWG, this paper). Information on numbers of characters and taxa for each matrix is in Table 3, and the structural character list is in Table 4. The morphological (structural) matrix is in Appendix II. The first four data sets represent the chloroplast genome, and the next three the nuclear genome. Five data sets, ndhF, rbcL, rpoC2, phyB, and GBSSI, are all protein coding sequences; introns of GBSSI were not included in the alignments. The full data matrix included 66 taxa and 7025 characters, for a total of 463,650 cells. The amount of missing data for the total data set is 33.2% and varies among genes and taxa (Table 3). The full data matrix can be obtained from LGC, JID, EAK, or HPL, or from the GPWG website, or at Tree BASE (http://herbaria.harvard. different ways the two programs count resolutions of polytomies. PAUP* analyses used 10 random addition sequences, MULPARS on, TBR branch swapping, and MAXTREES set to automatically increase by 100. Bootstrapanalyses (bts) used the full heuristic option, 500 or 1000 replicates. Bremer support(abbreviated here as brs; Bremer, 1988; Kallersji et al., 1992; also called decay index, cf. Donoghue et al., 1992) was also calculated. For tree lengths up to 11 steps longer than the shortest tree (up to 8763 steps), all trees were saved and the strict consensus computed. Because of memory limitations the method of negative constraints (Baum et al., 1994) was used to compute higher Bremersupportvalues. The search for optimal trees was found to be quite inefficient with this method and often led to inflated support values. To minimize this problem, each search was done with 10,000 random addition sequences. Even so, the search frequentlyfound trees shorter than the negative constrainttree, indicating that the previous searches had missed some trees. Computing Bremer support thus took almost two weeks of computer time on a G3. For the tree presented here, we arbitrarily chose a cut-off of 34 steps, so brs values above that are simply reported as "> 34." To assess robustness of the results to choice of markers, each data set was analyzed by itself. The morphological data set was omitted from one analysis, and the chloroplast data were analyzed separately, as were the nuclear data. For analyses of individual data sets, PAUP* was set to perform heuristic searches using maximumparsimony,gaps were coded as missing data, multistate taxa were coded as uncertain, and starting trees were obtained by ten random addition sequences, holding one tree at each step; branch swapping used tree bisection and regrafting (TBR), steepest descent was not in effect, and MULPARS was in effect. Bootstrap analyses of individual data sets were done to facilitate comparisons with combined analyses. All bootstrapsof individual data sets included 500 bootstrap replicates; MAXTREES was set to 500 to minimize times for searches. Analyses conducted with Nona ver. 1.6 (Goloboff, 1993) used the default settings amb- (clades resolved only if they have unambiguous support) and poly= (polytomies allowed). Tree searches involved 1000 Wagner tree initiations using random taxon entry sequences, followed by tree bisection reconnection (tbr) swapping with up to 20 mostparsimonious trees retained in each search (hold/ 20, mult*1000); shortest trees retained from the subsearches were then tbr-swapped to completion, edu/treebase/index.html). Alignments were provided by the contributors except for the ITS data, for which the alignment was constructed by EAK, beginning with an initial alignment in ClustalW (Thompson et al., 1994) and then continuing by eye. It became apparent that ITS1 could not be aligned reliably across the family, so it was omitted from the data set. Later after extensive data exploration, several regions of ambiguous alignment were also omitted from ITS2. Gaps were treated as missing data. A few indels, identified as phylogenetically informative in analyses of individual data sets, were coded as binary characters and included in the structural data matrix (Appendix II). DATA ANALYSIS Data were analyzed by parsimony algorithms, as implemented in PAUP*4.0 d64 (Swofford, 1998) on a Power Macintosh G3, and Nona (Goloboff, 1993) on an Intel-chip-based workstation running Windows NT. Data sets were analyzed individually by JID, LGC, EAK, and HPL to be sure that e-mail transmittal of such a large file had not introduced any errors (for which we suggest the term "networkinduced homoplasy"). Numbers of informative characters and tree lengths were the same for the two programs, although in some cases the number of equally parsimonious trees differed because of 387 Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae with up to 10,000 trees held in memory (holO000, max*). Structural character autapomorphies of terminals, and synapomorphies of clades, were determined by optimizing the morphological data on most-parsimonious trees obtained by the various analyses, using Winclada ver. 0.9.99m6.1 (Nixon, 2000). Strict-consensus bootstrap frequencies for just the total evidence analysis (see Soreng & Davis, 1998) were computed with Clados ver. 1.9.95 (Nixon, 1993) running Nona (Goloboff, 1993) as a daughter process for the tree searches, using a copy of the data set from which uninformative characters had been removed (with the "mop" function of Winclada). One thousand bootstrap replicates were conducted, using the same ambiguity and polytomy settings as in the basic analyses. Each replicate consisted of 10 random taxon entry sequences followed by tbr swapping with up to 10 trees retained from each subsearch (ho/10, mult* 10), and with further tbr swapping then conducted on the resulting trees from the 10 subsearches, with 101 trees held (ho 101, max*). Uninformative characters were excluded for all analyses, so all tree statistics reported in this paper (consistency index [CI] and retention index [RI]) reflect only potentially phylogenetically informative characters. Chloroplast and nuclear trees were compared using the incongruence length difference test (random partition test of Farris et al., 1994), as implemented in PAUP*. They were also compared using simple inspection, as recommended by Mason-Gamer and Kellogg (1996). To compare tree topologies, constraint trees were constructed as necessary in MacClade (Maddison & Maddison, 1993); these were then loaded, the constraint enforced, and a heuristic search undertaken using the same parameters as in unconstrained searches. The combined data were constrained to fit topologies suggested in previous studies by loading a constraint tree in PAUP* and then searching for the most parsimonious tree compatible with that constraint tree. Constrained and unconstrained trees were compared using the Wilcoxon signed ranks test (WSR) as suggested by Templeton (1983) and implemented by Mason-Gamer and Kellogg (1996). Significance values were determined using a two-tailed test. The entire data set could not be analyzed with or maximum likelihood algoneighbor-joining rithms. The inclusion of morphological and restriction site data with sequences made it nonsensical to specify a single model of evolution. While a model could in principle be hypothesized for morphological or restriction site data, it would have to be different from the model specified for sequence data. Calculation of base frequencies and transition/transversionratios would be meaningless. Several neighbor-joininganalyses were done with morphological and restriction site data omitted, but this also required omitting several taxa for which distances were then undefined because of missing data. By the time data sets and taxa were omitted, the results were difficult to compare to those of parsimony algorithms. Several maximum likelihood analyses were also undertaken on sequence data alone. These did not reach completion even after three to five days of analysis time. As with the neighbor-joining analyses, missing data and different models of evolution for the different genes made the results of questionable validity. RESULTS Consensus trees for analyses of the individual data sets are presented in Appendix III-A to H and tree statistics in Tables 3 and 5. Note that the taxa included are generally selected from more comprehensive analyses that have been published elsewhere, as described in Materials and Methods. Many of the trees differed in topology, but in no case was a strongly supported group in one tree contradicted by a strongly supported group in another tree. We interpreted this as lack of significant conflict. Nonetheless, the ILD test indicated significant differences between the nuclear and chloroplast data sets, between nuclear protein-coding and chloroplast, between ndhF and phyB, and between ndhF and rbcL.These differences persisted in most cases even when taxa with conflicting placements were removed. In the only exception to this observation, ndhF and phyB were not significantly different if the PACCAD Clade was reduced to Panicoideae, Chloridoideae, and the clade of Molinia plus Phragmites. This provides weak evidence that differences in resolution of the PACCAD Clade (Panicoideae, Arundinoideae s. str., Chloridoideae s.l., Centothecoideae, Aristidoideae, Danthonioideae) are partly responsible for the significant differences. Differences between ndhF and rbcL, however, are puzzling because both are part of the same linkage group. Because of the ambiguity of the results, we did not attempt to do all possible pairwise comparisons of trees. Despite the differences in the data sets, we chose to combine the data in a single analysis. Different histories for the various genes remain a formal possibility. However, in other investigations we have seen that the ILD test may return significant differences if there is extensive missing data (as we have in some data sets here) Table 5. Bootstrap support values for subsets of the total data matrix. Numbers of nodes at particular support values are given as fraction as decimals. Poly = polyphyletic; para = paraphyletic. Anom. = Anomochlooideae; Phar. = Pharoideae; Puel. = Puelioideae. *Panicoid excluded then support values are much higher. Total data # Nodes 100 # Nodes 90-99 # Nodes 70-89 Fraction nodes > 70 Poaceae Spikelet Clade Bistigmatic Clade BEP + PACCAD Clade Bambusoideae Ehrhartoideae Pooideae BEP Aristidoideae Chloridoideae Panicoideae* Danthonioideae PACC Arundinoideae s. str. Centothecoideae 27/64 (0.42) 9/64 (0.14) 11/64 (0.17) 0.73 100 100 100 100 97 100 100 71 100 86 65 98 100 77 Para Molec. data 25/64 (0.39) 14/64 (0.22) 8/64 (0.13) 0.73 100 99 100 100 98 100 100 90 100 83 Poly 97 100 78 Poly ChloroNuclear Structural ndhF plast 22/64 (0.34) 14/64 (0.22) 7/64 (0.11) 0.67 100 98 100 99 97 100 93 62 100 86 Para 98 99 <50 Para 2/55 (0.04) 12/55 (0.22) 5/55 (0.09) 0.35 97 87 97 81 Para 92 95 50 84 51 63 Poly 77 Poly Para 0 1/64 (0.02) 4/64 (0.06) 0.08 Para Para <50 Poly Poly Para Para Para 84 Para Para Para <50 Para Poly 23/63 (0.36) 11/63 (0.17) 4/63 (0.06) 0.59 100 94 100 100 100 98 88 53 100 <50 Poly 99 100 <50 Para cprs rbcL rpoC2 2/43 (0.05) 8/43 (0.19) 9/43 (0.21) 0.45 98 81 <50 Not tested Para 72 <50 Para Not tested 98 Para Not tested 95 Not resolved Not tested 8/35 (0.23) 7/35 (0.20) 6/35 (0.17) 0.60 99 98 Not tested Para Para 100 70 Para 98 54 95 83 73 23 64 2/32 (0.06) 6/32 (0.19) 4/32 (0.12) 0.37 Not tested Not tested Not tested Not tested Para 51 88 56 67 63 93 70 73 Poly Para (0 1 (0 1 (0 0 9 5 7 7 9 9 9 8 N 9 9 N 9 5 9 Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae or if a single terminal taxon differs in its placement (Z. Magombo, pers. comm.). Because of the lack of obvious points of conflict between the data sets, and because of the clear congruence at the deep nodes with which we are concerned, we interpret the significant ILD tests as misleading. Analyses of the complete data set were faster than analyses of many of the individual data sets, as has been found in studies of other large data sets (Soltis et al., 1998). For example, a heuristic search of the complete data set in PAUP* on a Macintosh G3 with 10 random addition sequences took 19.6 seconds. With all data combined, there were 2143 parsimony informative characters, which produced a single tree of 8752 steps, consistency index (CI) of 0.375, and retention index (RI) of 0.557 (Figs. 1 and 2). Bootstrap analyses (1000 replicates) indicated that 27 branches were supported in 100% of the bootstrap replicates, 9 branches in 90-99%, and 11 branches in 70-89% (Table 5). Put another way, of 64 internal nodes, slightly more than half (36) have bootstrap values over 90% and a clear majority (47) have values over 70%. Bootstrap values were virtually identical whether done using strict consensus bootstrap in Nona (Goloboff, 1993) or the frequency-within-replicates bootstrap in PAUP*4.0 (Swofford, 1998); for comparison with individual analyses, we report the values from PAUP*4.0. The analysis of the combined data confirms many results of previous studies and clarifies some relationships that were previously ambiguous. The two species of Restionaceae form a clade. Joinvillea is sister to a monophyletic Poaceae. The three earliest diverging lineages are the Anomochlooideae, Pharoideae, and Puelioideae, in that order, together accounting for 30 species of grasses. The vast majority of extant grasses fall into two distinct lineages. One of these is the PACC clade (Davis & Soreng, 1993), here called the PACCAD Clade (Panicoideae, Arundinoideae s. str., Chloridoideae s.l., Centothecoideae, Aristidoideae, Danthonioideae) to reflect the inclusion of two additional subfamilies within the clade. Within this clade, Panicoideae s. str. (excluding Danthoniopsis) are monophyletic (bts 94; brs 10), as are the core Paniceae sampled here (bts 100; brs 25) and Andropogoneae (bts 100; brs 32). Other strongly supported groups in the PACCAD Clade correspond to Aristidoideae (bts 100; brs 25) and Danthonioideae (bts 98; brs 15). The traditional Chloridoideae are supported at bts 99 (brs 16), and the clade including the chloridoids plus Centropodia glauca and Merxmuellera rangei (Chloridoideae s.l.) is also reasonably well sup- ported at bts 86 (brs 8). A clade corresponding to Arundinoideae s. str.-Arundo, Amphipogon, Molmodest support inia, and Phragmites-receives from this analysis (bts 77, brs 6), but the sister relationship of Molinia and Phragmites is well supported (bts 100; brs 16). The other major clade (the BEP Clade) is less well supported (bts 71; brs 8) and includes Bambusoideae s. str., Ehrhartoideae (= Oryzoideae), and Pooideae. Bambusoideae are monophyletic (bts 97; brs 15), as is the clade including the herbaceous bamboos (bts 100; brs 18). Likewise Ehrhartoideae are monophyletic (bts 100; brs 24), as are Oryzeae (bts 100; brs > 34). Pooideae include Brachyelytrum (bts 100; brs 15), and most nodes within the pooid clade are strongly sup- 389 ported. Despite the strong phylogenetic pattern shown by the combined analysis, placement of some taxa remains ambiguous. The major uncertainty remains the monophyly of the BEP Clade. As noted earlier (GPWG, 2000), it is almost equally parsimonious to place Pooideae as sister to the PACCAD Clade, and this makes evolution of particular morphological characters more parsimonious. The Pooideae plus PACCAD group appears in analyses of rbcL (Appendix III-C), chloroplast restriction sites (Appendix III-A), morphology (Appendix III-H), and ITS (Appendix III-F), whereas the BEP Cla(le is retrieve(l by analyses of nidhF (Appendix III-B), rpoC2 (Appendix Ill-I)), and phyB (Appendix IIIE). GBSSI (Appendix III-G) forms a novel topology, in which neither the PACCAD nor the BEP Cla(les is monophyletic. An analysis combining rbcL, chloroplast restriction sites, ITS, and morphology retrieves, not surprisingly, a clade that links the Pooideae with the PACCAD Clade. Bootstrap analysis, however, finds that the Pooideae + PACCAD clade occurs in only 23% of the replicates, although it appears in 40% if Streptogyna is considered part of the clade. The BEP Clade was not found in any of the bootstrap partitions. Constraining the entire data set to place Pooideae sister to the PACCAD Clade resulted in a single tree eight steps longer than the most parsimonious tree. The net change of eight steps, however, was produced by changes of one or two steps in 107 characters from throughout the data set. A Wilcoxon signed rank test (Templeton, 1983; MasonGamer & Kellogg, 1996) resulted in a test statistic of 2654; for n = 107, this corresponds to p < 0.406 (two-tailed test). This means that we cannot rule out the possibility that Pooideae are indeed sister to the PACCAD Clade. This is true even if the morphological characters are excluded (z = 1.146; P < 0.254). Annals of the Missouri Botanical Garden 390 118 Flagellaria 115 Elegia 136 136 I 113 113 aloskion 94 Joinvillea 106 134 140 Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae 100 >34 100 >34 I 100 15 d00 100 32 1100 >34 73 100 4 97 18 1064 36 15 628 2 140 | l 2 24 36 17 2 10 0 0 71 100 >34 8 88 100 15 41 3 7J 76 33 2 86 3 8 10 ?8 4 1 87 8 868 16 100 - 7 4 99 20 oO{ >34r >34 9 0 100 85 4 68 3 68 3 Flagellaria Elegia Baloskion Joinvillea Anomochloade StreptochaetaJ Anomochlooideae Pharus -Pharoideae Guaduela Pueliodeae Puelioideae Puelia Eremitis Pariana Lithachne Olyra Bambusoideae Buergersiochloa Pseudosasa Chusquea Streptogyna - Incertae sedis Ehrharta Ehrhartoideae Oryza 7 Leersia Phaenosperma Anisopogon Ampelodesmos Stipa Nassella Piptatherum Brachypodium Avena Pooideae Bromus Triticum Diarrhena Melica Glyceria Lygeum Nardus Brachyelytrumn 100 Arstida Aristidoideae A tidodeae 25 Stipagrostis m. a IMerxmuellera ~8 98Q Karroochloa 100 Dant honioideae 158 Austrodanthonia 1 5 >001 27 Danthonia 79 -Amphipogon 7L' Arundo Arundinoideae Ar LQ Molinia 99 8 111 32 27 25 2 1 34 1 I 53 1 28 I-00 12 60 29977 3 2 1 93 o~94i 882 96 9 no ^ 10 7 >34 88 100 >34 E 100 >34 61 100 20 100 -4 391 7 196 7 ' 65 3 L- 65 100 2 194 25 10 O|--00 32 -Zea _Phragmites Merxmuellerar. Centropodia Eragrostis Uniola(P ChlorUnold Chloridoideae Pappophorum Zoysia Spartina Sporobolus Distichlis - Incertae sedis Eriachne Thysanolaena Zeugites Centothecoideae Chasmanthiumn_ Incertae sedis Gynerium Danthoniopsis Panicoideae Panicum Pennisetum Miscanthus Micraira - ncertae sedis Figure 2. Same tree as Figure 1, showing percent of bootstrap replicates above lines and Bremer support below. Brackets indicate the revised classification for the Poaceae. Figure 1. Single most parsimonious tree for the grasses and relatives, based on eight sets of data. Length = 8752 steps, CI = 0.375, RI = 0.557. Numbers above branches are numbers of unambiguous changes. Branches are drawn proportionalto length. 392 Annals of the Missouri Botanical Garden The combined analysis places Streptogyna as sister to Ehrhartoideae, but this result is not strongly supported (bts 40; brs 2). This partly reflects missing data, in that only ndhF and phyB sequences are available for Streptogyna, in addition to morphological data. ndhF places S. americana sister to Ehrhartoideae, whereas phyB places it as sister to the entire BEP Clade, and morphological data fail to resolve its position. The combined data place the woody bamboos, Pseudosasa and Chusquea, in a clade (bts 68; brs 2), as would be expected from previous studies (Zhang, 2000; Zhang & Clark, 2000). The pair appears monophyletic in chloroplast restriction site, morphological, and phyB trees. The two are paraphyletic, however, in trees using rbcL and ndhF, although this result is not well supported in these trees. "Pseudosasa" is a composite taxon, made up of data from several different genera, and this may also affect its placement in the combined tree. Phaenosperma and Anisopogon are clearly members of the expanded pooid clade, where they are placed by all data sets, either singly or in combination. Their position within the clade, however, remains uncertain. They are sister taxa when all data are combined, but this result is not strongly supported (bts 53; brs 1); together they are sister to the Stipeae, also a poorly supported result (bts 28; brs 1). In ndhF, chloroplast restriction site, and phyB trees, Anisopogon is placed on a branch that diverges after the Lygeum + Nardus clade, but before the rest of the Pooideae (i.e., Stipeae, Meliceae, Diarrhena, Brachypodium, Aveneae, and Poeae). ITS places it sister to Aveneae/Poeae, and rpoC2 places it sister to Stipa. In no case is the placement strongly supported. The position of Phaenosperma is based only on ndhF and morphological data, and the latter are largely uninformative about its position. Meliceae are monophyletic in all gene trees and in the combined tree. Their position, however, varies among the individual gene trees. The combined tree provides good evidence that Meliceae diverged after Lygeum + Nardus (bts 82; brs 7), but evidence is weak that it was the next diverging branch (bts 29; brs 1). Other possible placements include sister to Stipeae (ndhF, phyB), sister to Diarrhena + Brachypodium + Aveneae/Poeae (cp restriction sites), sister to a clade of Brachypodium + Brachyelytrum + (Lygeum + Nardus) (ITS), or paraphyletic at the base of the Pooideae (GBSSI). The ambiguity cannot be ascribed to missing data, although additional sampling among early-diverging Pooideae might be warranted. The positions of Eriachne and Micraira are not firmly resolved by the combined data set, although both are clearly members of the PACCAD Clade. This almost certainly reflects missing data. In addition to morphological data, Eriachne is represented only by rbcL and ITS sequences, and Micraira by ndhF, rpoC2, and ITS. They are both isolated taxa, and in individual analyses fall at the base of other well-supported clades. The position of Micraira as sister to the entire PACCAD Clade appears only in the combined analysis, and likewise the position of Eriachne as sister to the Arundinoideae s. str. + Chloridoideae + Aristidoideae + Danthonioideae clade is both novel and poorly supported. Bootstrapanalysis of the combined data set placed Eriachne and Micraira as sisters in 51% of the 1000 replicates, a position not supported by the most parsimonious tree. Aristidoideae and Danthonioideae are both clearly monophyletic, and each is strongly supported by both bootstrap and decay analyses (bts 100 and 98; brs 25 and 15, respectively). In the combined tree they appear as sister taxa. The aristidoid/danthonioid clade is not stronlglysupported, however (bts 61; brs 8), and is reflected only in the phyB tree. rbcL places Aristidoideae sister to Chloridoideae, whereas ndhF and cllloroplast restriction sites put Aristidoideae sister to the rest of the PACCAD Clade, and ITS places the subfamily sister to Amphipogon + Chloridoideae. rpoC2 indicates that Aristidoideae is derived from within Arundinoideae. ndhF places Danthonioideae sister to Panicoideae + Centothecoideae, whereas chloroplast restriction sites do not resolve the position of Danthonia. GBSSI retrieves a novel arrangement in which Danthonioideae are polyphyletic, but this result is not strongly supported and is likely affected by skewed taxon sampling in the GBSSI data set. rpoC2 suggests that Danthonioideae are sister to Amphipogon. The relationships of Zeugites, Thysanolaena, Chasmanthium, Danthoniopsis, and (ynerium to each other and to the Panicoideae are not resolved by this analysis. The entire group is well supported as monophyletic (bts 87; brs 8), but other relationships are less clear. Only morphologicaland ndhF data are available for Zeugites, so its placement may be affected by missing data. The morphological data have little effect on the analysis. When they are omitted, 6 trees are found in two islands (length 8488, CI = 0.378, RI = 0.554). The topology of the strict consensus (Appendix III-K) is similar to that of the combined tree except for the position of Zeugites, which is sister to Danthoniopsis, and Gynerium,which is sister to Panicoideae. In the consensus of the six trees, the Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae relationship of Pseudosasa and Chusquea is unresolved, as is the relationship of Phaenosperma and Anisopogon, and the position of Meliceae in the pooid clade. These are areas that were poorly supported even in the combined tree, and thus already known to be ambiguous. The most notable difference is the increased support for the BEP Clade, which is supported at a bootstrap value of 90%. The number of nodes with support greater than 90% (Table 5) is somewhat greater without the structural data, but the overall consistency index is not changed appreciably (Table 3). The results for the entire data set largely reflect the results for the chloroplast data alone. The chloroplast data contribute 1336 potentially phylogenetically informative characters, or about 62% of the total. Analysis of these data alone produces 2 trees (length = 4903, CI = 0.381, RI = 0.589) that differ only in the relative positions of Phaenosperma and Anisopogon (Appendix III-I). The numbers of strongly supported nodes are about the same as for the entire data set (Table 5), and differences between the chloroplast tree and the entire data set are all in poorly supported areas of the tree Differences appear only in the placement of the Meliceae, Eriachne, Micraira, Zeugites, and Danthoniopsis, all poorly supported areas of the trees. (see below). The chloroplast tree is only slightly affected by mixing sequence data with restriction site data. If the restriction site data are exclulded so that the data set consists only of ndh F, rbcI, and rpoC2 data, the tree is virtually identical to the chloroplast tree except that Pseiidosasa plus Chusquea, an(l Phaenosperma plus Anisopogon form monophyletic pairs rather than being paraphyletic (not shown). Piptaltherum and Nassella are paraphyletic rather than sisters, and the Meliceae are sister to the core Pooideae rather than to the Stipeae. Analysis of only the three nuclear genes (phyB, GBSSI, and ITS) required elimination of nine taxa for which nuclear data were not available. The analysis thus included 57 taxa and 757 characters and found eight trees (length = 3513, CI = 0.382, RI = 0.512) on one island (Appendix III-J). The nuclear trees were not as well supported as the chloroplast tree or the tree for the entire data set, which presumably reflects extensive missing data for GBSSI, and a generally smaller number of informative characters. Only two nodes were supported in 100% of the 1000 bootstrap replicates, and 11 had values between 90 and 99%. Analysis of chloroplast data plus the data from the two protein-coding nuclear genes (that is, excluding morphological and ITS data) has little effect on either topology or support for the tree, perhaps because omitting morphology and ITS only eliminates 178 characters, or about 8% of the total. 393 DISCUSSION WELL-SUPPORTED CLADES Some relationships appear consistently in all analyses of all data sets and are strongly supported by the combined analysis. Among these are the following (in order from the bottom of the tree): 1. Joinvilleaceae are sister to Poaceae. 2. Poaceae are monophyletic. 3. The earliest diverging lineage of Poaceae is Anomochlooideae (even if Anomochloa and Streptochaeta prove to be two separate lineages, they would still be the two earliest-diverging lineages in the family). 4. The next diverging lineage is Pharoideae. 5. The next diverging lineage is Puelioideae. 6. All remaining grasses form a clade, which appears to have diversified well after the origin of the family. 7. Bambusoideae s. str., Ehrhartoideae, Pooideae s.l., Aristidoideae, Danthonioideae, Arundinoideae s. str., Chloridoideae s. str., Chloridoideae s.l., and Panic oideae are all monophyletic. 8. Bambuseae, Parianeae, (lyreae s. str., Oryzeae, Stipeae, Meliceae, and LygeiLm + Nardus, and Molinia + Phragmnites are all monophyletic. 9. The PACCAI) Clade- now including Panicoideae, Arundinoideae s. str., Chloridoideae s.l., Centothecoideae, Aristidoideae, Danthonioideae, Eriachne, Micraira, and Gynerium-is monophyletic. As noted in the introduction, all of these relationships have been supported by previous studies and none is unique to the combined analysis. Previous studies, however, were limited because they were based on a single gene, a modest number of morphological characters, and/or a restricted sample of taxa. Because of the strong support for the relationships found in the present study, we propose a revised subfamilial classification (see Taxonomic Treatment). The revisions primarily reflect changes in circumscriptions of the Bambusoideae and Arundinoideae and involve only a small fraction of the species in the family. Over three quarters of the species are included in the subfamilies Pooideae, Chloridoideae, and Panicoideae, the circumscriptions of which are changed only slightly by the revisions. MOLECULAR CHARACTERS Virtually all of the phylogenetic signal in this analysis comes from the molecular data (Appendix 394 Annals of the Missouri Botanical Garden III-K), as expected. The combinable component consensus (Bremer, 1990) of the molecular trees (Kellogg, 1998) is remarkably well resolved; all nodes found in this consensus are strongly supported in the combined analysis presented here. When the molecular data are analyzed alone, all strongly supported nodes from the combined analysis are recovered, and support for the BEP Clade is increased. Previous theoretical (Graybeal, 1998) and empirical (Soltis et al., 1998) studies have indicated that large numbers of characters may be necessary to resolve phylogenetic patterns, a conclusion only partially supported by this study. The molecular data alone and the entire data matrix included 2093 and 2143 phylogenetically informative characters, respectively (Table 3). These data sets found the largest percentage of nodes with bootstrap values above 70% (0.73 in both cases) but did not have the highest consistency or retention indices. The highest CI was produced by the rpoC2 data alone (150 informative characters), although this could be in part an artifact of alignment (see below), and the highest RI by the rbcL data alone (213 informative characters). The fraction of nodes with bootstrap values over 70% was almost as high for the phytochrome B data alone (with 417 informative characters) as for all data combined. We conclude that, while large numbers of informative characters may provide increased reliability, small numbers are not necessarily misleading or inaccurate samples of the whole. Other studies have shown that the number of taxa included may affect phylogenetic accuracy was difficult and confirmed our suspicion that it may not be useful at this level of divergence. ITS1 and parts of ITS2 had to be eliminated because of difficulty in assessment of sequence similarity.The rpoC2 sequences used here code for repeated amino acid motifs inserted into the protein. The insertion appears only in the grasses and thus constitutes a synapomorphyfor the family (Cummingset al., 1994), although we did not code it as such in this analysis. The repeats are similar but not identical to each other in sequence, making alignment problematical, a point discussed at length by Barker (1995) and Barker et al. (1999). Efforts to improve alignments necessarily reduce apparent homoplasy; this may result in the high CI mentioned above. Although phylogenetic results from these molecules are similar to those from the other genes, by themselves the two sets of sequences do not permit confident assessments of relationships among subfamilies. Trees from ndhF and phyB, individlually,are particularly well resolved and well su,pported.Their congruence in early-divergingbranches contributes to the strength of the overall topology of the combined data. In particular,phyB provi(les considerable support for the BEP Clade, a to,,ology that is only weakly supported by ndhF, an(l not at all by several other data sets. The two data sets do appear to conflict in relationships among members of the PACCADClade, and this may be an area for future (e.g., Hillis, 1996, 1998; Graybeal, 1998), although this is not necessarily the case (Poe & Swofford, 1999). Certainly future studies should include more taxa than just the set of exemplars used here. However, the results numbered 1 to 9 above have been found in analyses of virtually every individual data set, as well as in the combined tree, and we would be surprised if they were overturned by inclusion of more taxa. Most of the molecular data come from chloroplast genes, so it is not surprising that the tree from the chloroplast alone closely matches the tree for the entire data set. The ndhF data set is missing the least data (Table 3) and has the most informative characters of the molecular data sets, presumably because it is the longest molecule. Our results confirm the utility of this molecule for resolving relationships among grass genera (Clark et al., 1995; Giussani et al., in press). Alignment is a particular problem for the rpoC2 and ITS data used here. Alignment of the ITS data investigations. As noted in Methods, sequences for a given genus were in some cases taken from (lifferent, congeneric species. This procedure assumes that the genus is monophyletic, an assumption that is almost certainly correct in many cases (e.g., Joinvillea, Streptochaeta),and perhaps not as likely in others. For example, the three species of Stipa sampled here have been placed in the genera Achnatherum, Stipa, and Jarava (Barkworth & Fverett, 1987; Barkworth, 1993; Jacobs & Everett, 1997), which are distinct and possibly not a monophyletic group within Stipeae (Jacobs et al., 2000). While this problem is not likely to compromise our conclusions regarding subfamily relationships, it means that relationships among species of the Stipeae (or other tribes or genera where composite terminal taxa were used) cannot be addressed by this analysis. STRUCTURAL CHARACTERS The structural characters, comprising the morphological data set (Table 4), were optimized on the phylogeny (Fig. 3). Our results suggest that some Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae 395 3 3147 Flagellaria 3053 31 Joinvillea 2 242938 A4 n 2 28424345464751 hAnomochloa 354547 11111141 0-Streptochaeta 2 9 43 1 1 1112 'Puelia 1EP 221 020000 2 122135374050 1o-AooCoPACCAD 0000131 Figure 3 (pp. 395-397). Same tree as Figure 1, with structuralcharactersmapped on using ACCTRANoptimization. Characternumber is above the branch, and the state to which the character changes is below. Filled circles represent unique occurrences of character states; open circles represent homoplasies. of the morphological characters may be useful for delimiting groups within tribes or subfamilies, but are too variable to be useful in delimiting subfamilies; these include 2 (culms hollow or not), 3 (leaf sheath margins fused or free), 10 (pedicel present or not), 11 (presence or absence of proximal reduced flowers), 12 (number of flowers per spikelet), 13 (presence or absence of awns), 15 (attachment of awns), 16 (disarticulation above or below glumes), 20 (lodicule fusion), 28 (style fusion), 29 Annals of the Missouri Botanical Garden 396 5 1215253045 Streptogyna 1 0 1 0 35 /r />-~Ehrharta 111923242846 0 17 1 2 1 1 0 0 41 Oryza OO 23242530 194546 Leersia o0003 pOO- 13 1 1 12232429 1 5 40 Pseudosasa 0113 2 1147 13 I0 Chusquea 1 -0 r-Buergersiochloa \_ 13 "13 0-- gLithachne 17 1 0 A (-_ I Olyra 2529 Eremitis 1630 0 1 01Y 2324 01 Pariana 2 4047 0- Brachyelytrum 2 1344 10182943 / 01 -. 00 ....j28364347 /OOO-O-( Lygeum 7123047 Nardus 0011 13 0000 1 0000 \ 122238 o 1 1 3 I 31647 3 13202130 -Melica 00 1 0/1 Glyceria 13 L, Diarrhena r~-Brachypodium 404,7 153940 07 21 Avena 3 35 Bromus Triticum 0 7 131640 14153638 Anisopogon 1 21 0947 12 Phaenosperma 3 2 12 o Ampelodesmos 25 _ Piptatherum 1940 20 Figure 3. Continued. Stipa -Nassella Grass Phylogeny WorkingGroup Phylogeny and Classificationof Poaceae Volume 88, Number3 2001 397 182547 o0o-- Micraira 2835 21 1 1126 xO-Chasmanthium 1 0 1 5 2845 712223740 2 33 Thysanolaena 1 11 0 01 Zeugites 1 2 212545 2 122135374050 000000 0000131 ,OOO0-- Gynerium 4 133340 {?? 1121748 Danthoniopsis 2 134047 Miscanthus 9 1747 00 1 1111 2830 Zea 16 48 ? 28 Panicum 15 ^ 0 ~ Pennisetum 48 [*-Eria rchne Aristida 12174748 48 1 --2 1421 4\t~~ 31 4 1340 Stipagrostis Merxmuellera macowanii 1532 47 21(2 4Danthonia 1 47 0 ~-Karroochloa 1\33 111 1 Austrodanthonia 47 33 o 0 olinia (2V 112845 2 Phragmites 1 1 1 ~ 1217222844 Amphipogon / A3 Arundo 10 33 13I 4348 Merxmuellera 1547 16 rangei Centropodia 13 1- 35748 Pappophorum 14 103 20 3 Eragrostis 112835 1344 Uniola 1JJ '110 0 1 4 40 . / 0 0,3,4 Distichlis 13 _ 1012161849 01001 / 1 o--Zoysia 2 28 o-Spartina 101647 119 11 Figure 3. Continued. 9 Sporobolus 398 Annals of the Missouri Botanical Garden (number of stigmas), 30 (highest order of stigmatic branching), 39 (lipid in the endosperm), 40 (endosperm starch grain syndrome), and 47 (base chromosome number). For other structural characters, our assessment of character states is inadequate for use in phylogenetic analysis. For example, the woodiness of bamboo culms (char. 1) is due to numerous, closely spaced vascular bundles around the periphery of the culm, each bundle with massive sclerenchyma fiber caps on both sides, combined with heavily sclerified ground tissue, but isolated fiber strands may also occur (Soderstrom,1981; Liese, 1998). It is not clear if the hardness of culms in Arundo, Thysanolaena, and Gyneriumand a number of other flower and spikelet affects characters 8, 18, and 19. C4 photosynthesis (chars. 48 and 49) is known to be a set of characters that do not co-vary;its origin is still poorly understood.These charactersare discussed in more detail below. SPIKELETS AND FLOWERS Spikelet. The flowers of most grasses are arranged in bracteate units known as spikelets. These usually consist of two bract-like appendages (glumes) at the base of a central axis (rachilla) on which are borne one or more florets, all in a distichous pattern. According to Clayton (1965: viii), the "grass spikelet never ceases to fascinate, for the "woody" taxa of the PACCAD Clade (Watson & simplicity of its theme is matched by the elegance Dallwitz, 1992) is derived in the same way; these of its variations." The apparent simplicity of the taxa need more careful anatomical study. Similarly, grass spikelet notwithstanding, its origin and hothe fringed membranous ligule of some Pooideae mologies, as well as those of the grass flower,have (char. 4) may not be the same as the ciliate mem- been much debated (see reviews in Clifford, 1987, brane found in many members of the PACCAD and Soreng & Davis, 1998). The phylogeny highlights the difficulty of disClade and requires developmental investigation. Spikelet pairing (char. 9) appears in various pani- cussing the evolution of the grass spikelet. The coid grasses and also in Pharus, but the pattern of standard spikelet (or some modificationof it) is predevelopment of the pairs in Pharus is unknown. sent in all members of the Pooideae and the PACMembranouslodicules (char. 21) are scored as be- CAD Clade (Figs. 4 and 5). 'The pattern of bracts ing the same wherever they occur, but development and flowers, however, is variable among Pharoof these structureshas never been compared.Haus- ideae, Puelioideae, Bambusoideae, and Ehrhartotorial synergids are apparently uniquely derived ideae (Fig. 4). Pharoideae bear single flowers,each within the Danthonioideae, but many taxa have not with a lemma, a palea, and a pair of glumes, but been investigated for this character. Embryo char- no rachilla extension (Fig. 4B; Judziewicz, 1987; acters (35, 36, 37, 38) are often phylogenetically Soderstrom et al., 1987). Puelioideae have multiinformative, but lack of observation of critical taxa flowered spikelets, in which each flowerhas a lemmakes their use difficult in some cases. Starch ma and palea, and the whole unit has a pair of grains (char. 40) are classified according to their glumes (Fig. 4A). Proximal incomplete florets ocapparent structure when viewed with the light mi- cur, but distal reduction is seen only in Guaduella. croscope. With much recent work done on the bio- In Bambusoideae, the unisexual, one-flowered chemistry and molecular genetics of starch granule spikelets of Olyreae are standard, but the bracteate, formation (e.g., Whistler et al., 1984; Frazier et al., rebranching spikelets (pseudospikelets) of many 1997), this character could and should be recir- Bambuseae are difficult to interpret (Fig. 4F; Judcumscribed, although then much scoring will need ziewicz et al., 1999). The multifloweredspikelet of to be redone. Arm cells and fusoid cells (chars. 45 Streptogyna presents no difficulties of interpretaand 46) are now seen to be ancestral in the family, tion, but in the Ehrhartoideae,extra proximal,steryet their development and ultrastructureare poorly ile bracts (usually called sterile lemmas) are comstudied, and their physiological function is un- mon (Fig. 4G), and extreme reduction of glumes is known. Comparisonsof chromosomebase numbers known (Oryzeae). It is not clear whether the prox(char. 47) will almost certainly become more pre- imal sterile bracts of Ehrhartoideae are truly hocise because of recent studies of nuclear genome mologous to the proximalincomplete floretsthat occur in Puelioideae and some bambusoids, or if they arrangement(e.g., Gale & Devos, 1998). Finally, some of the structural characters are are phylogenetically (and possibly developmentally genuine morphological puzzles, ones for which and genetically) distinct. The uncertain position of strict comparison is difficult. These include the flo- Streptogyna makes the homology assessment even ral bracts (glumes, lemmas, paleas, lodicules), more ambiguous. Most characters of the spikelet and the floret which occur only in the grasses, and may not even occur in Anomochlooideae. Homology of the grass (chars. 9-12 and 14-17) are treated as inapplicable Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae 399 1 mm 1 cm E F I 1 mm 1 C B Figure 4. Spikelets and spikelet equivalents of early-diverginglineages and the BEP clade. -A. Puelia schumanniana, Puelioideae (Letouzey12930, US). -B. Pharus mezii, Pharoideae (Hinton 16059, US; redrawnfrom Judziewicz, 1987). -C. Streptochaetaspicata, Anomochlooideae (Bailey & Bailey 723, US; redrawnfrom Judziewicz & Soderstrom, 1989, from originals by A. Tangerini at US). -D. Anomochloa marantoidea, Anomochlooideae (Calderon2046, US; redrawn from Judziewicz & Soderstrom, 1989, from originals by A. Tangerini at US). -E. Stipa comata, Pooideae (Pearson s.n., ISC). -F. Guadua chacoensis, Bambusoideae (Nee 35467, ISC). -G. Ehrharta bulbosa, Ehrhartoideae (Barker1119, ISC). -H. Festuca idahoensis, Pooideae (Pohl 15642, ISC). Annals of the Missouri Botanical Garden 400 1mmi I1 -0? ,?i E 1 mm F G 2mm 1 mm immi C Figure 5. Spikelets of the PACCADclade. -A. Arundo donax, Arundinoideae (Bradley & Sears 3558, ISC). -B. Dichanthelium oligosanthes, Panicoideae (Lelong 2063, ISC). -C. Aristida arizonica, Aristidoideae (Griffiths7373, ISC). -D. Tridensflavus, Chloridoideae (Thorne 18302, ISC). -E. Andropogon gerardii, Panicoideae (Clark s.n., teaching collection, ISC). -F. Danthonia californica, Danthonioideae (Pohl 9459, ISC). -G. Centothecalappacea, Centothecoideae (Liang 66250, ISC). -H. Chloris cucullata, Chloridoideae (Malacara & Gutierrez33, ISC). Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae when grass-type spikelets and florets are absent, as in the non-grass outgroups (see char. 8). Although pseudospikelets occur in some genera of Bambuseae, the two in the present study are regarded as having true spikelets and florets, and thus scorable for features of these structures. Neither Anomochloa nor Streptochaeta(Fig. 4C and D) has structures clearly homologous with glumes, lemmas, or paleas, and thus neither can be described as having grass-type spikelets or florets. We therefore follow Clark and Judziewicz (1996) in using the term "spikelet equivalent" to refer to the flowering units of the inflorescences in the Anomochlooideae to emphasize this lack of recognizable homology. Characters of the spikelet and the floret are scored as inapplicable or ambiguous in these two genera except for char. 13 (see Appendix IV). The grass spikelet may have originated either before or after the divergence of Anomochlooideae. If before, then the spikelet was extensively modified in the long history of Anomochlooideae. The origin certainly must have occurred before divergence of Pharoideae. We refer to the clade of all grasses except Anomochlooideae as the Spikelet Clade. Bracts outside of the spikelets subtend inflorescence axes and often have a blade. Short to elongate prophylls are usually present on the branches subtended by these bracts. Such bracts occur )primarily in Bambuseae and Andropogoneae,but they are not necessarily homologous between the two groups (see Renvoize & Clayton, 1992). Well-developed sul)tending bracts are usually absent in other memlers of the Spikelet Clade, although there may be a ridge or scar which presumablyrepresents the subtending bract at the base of the inflorescence branch. the structure still could be derived as a prophyll if the axis rotated through 180? (Clifford, 1987). The gemmiparousbracts of many Bambuseae are essentially glumes with a bud in the axil. If the bud develops it becomes a second- or higher order pseudospikelet (Judziewicz et al., 1999). Glumes may be highly reduced or lost, as in Oryzeae. A glume-like prophyll at the base of the pseudospikelet is observed in many Bambuseae (Fig. 4F), although the axis bearing the prophyll is not elongated (McClure, 1966). 401 Numbersofflorets. The rachilla may or may not extend beyond the most distal floret, and reduced or modified florets may be present below or above (or both below and above) the fertile ones. For this analysis, we have assumed that the grass flower is terminal to the axis on which it is borne. In the Anomochlooideae, there is no identifiable palea and thus the flower appears to be truly terminal to the main sympodial axes of the inflorescence in Anomochloa, as discussed in Soreng and Davis (1998); the same is true under Soderstrom's(1981) interpretation of the spikelet equivalent of Streptochaeta.Within the Spikelet Clade, however, the flowers are borne on lateral branches, as indicated by the presence of a prophyll (i.e., the palea) in the proximal, a(laxial position on the branch. This pattern is clear in those taxa with multiflowered spikelets or spikelets with one floret ancda rachilla extension. There are a number of taxa with a single floret and no rachilla extension, including the Pharoideae, in which a well-developed palea is found in the floret. It is simple enough to imagine the reduction of the branch apex to the point where no evidence of a rachilla extension can be observed, but, as Soreng and Davis (1998) noted, the presence of a single-flowered floret (or equivalent) appears to be plesiomorphic for the family. This implies that either the rachilla extension and additional fertile florets evolved subsequently, or that multiflowered spikelets evolved before the divergence of the Pharoideae, and reduction to a single fertile floret occurred in that lineage (Soreng & Davis, 1998). Clearly, single-flowered spikelets evolved a number of times in various lineages in the BEP + PACCAD Clade. Glumes. We have assumed here that glumes are homologous across the Spikelet Clade. Glumes are typically defined as the two sterile bracts at the base of the spikelet, but additional sterile bracts (usually called sterile lemmas or sterile florets even if there is no evidence of any, even vestigial, floral axis) may occur between the glumes and the flowerbearing lemmas (e.g., Ehrhartoideae, Chusquea, many Centothecoideae). In general, the first (lower) glume is abaxial and the second (upper) glume is adaxial (E. A. Kellogg, pers. obs.; Clifford, 1987), Lemma. Each floral axis is subtended by a lembut the first glume may be adaxial in position, as ma (Fig. 6A and B), a structure that appears to be in a number of Paniceae (Clifford, 1987). Grassl universally present across the Spikelet Clade. The (1956) and Stapleton (1997) argued that the first lemma apparently is formed wholly by the spikelet glume is actually a prophyllfor Andropogoneaeand meristem, and thus is a bract on the rachilla (ClifBambuseae, respectively, and that the prophyll was ford, 1987). Lemma morphology is extremely varidisplaced upward to assume the position and func- able, but the number of nerves is consistently odd, tion of a glume. Even if the first glume is abaxial, varying from 1 to 15 (Clifford, 1987). Some taxa Annals of the Missouri Botanical Garden 402 ]E Ii L I! N e P e K: 0 I st' I \\ | ;.fv i E E lo . LJ i II ii rI I! i: C : E E E r CD 0 ijj A Figure 6. Grass flowers, fruits, and embryos. -A. Floral diagram of a grass with three lodicules and six stamens. -B. Floral diagram of a grass with two lodicules and three stamens. -C. Flower of Yushania (Bambuseae, original by D. Friedrick). -D. Lodicules of Pooideae (Poa, redrawn from Jirasek, 1968). -E. Lodicules of Chloridoideae (Muhlenbergia,redrawnfrom Soderstrom, 1967). -F. Lodicules of Panicoideae (Setaria, redrawnfrom Jirasek, 1968). -G. Lodicules of Bambusoideae (Chusquea),showing the anterior pair (lower two) and the posterior one (upper). H. Generalized ovule section (Danthonioideae) showing haustorial synergids (stippled) (redrawnfrom Verboomet al., 1994). -I. Longitudinal section of a panicoid embryo showing presence of a scutellar tail (st) and the elongated mesocotyl internode (mi). J. Longitudinalsection of a pooid embryo showing presence of an epiblast (ep). -K. Cross section of a panicoid embryo apex showing overlapping embryonic leaf margins. -L. Cross section of a pooid embryo apex showing embryonic leaf margins that meet. M, N. Caryopsis of Eustachys (Chloridoideae). -M. Hilum side, showing a punctiformhilum (h). -N. Embryo side, showing the large embryo (e). O, P. Caryopsis of Chusquea(Bambusoideae) (redrawnfrom McClure, 1973). -0. Hilum side, showing a linear hilum (h). -P. Embryo side, showing the small embryo (e). c-coleoptile; co-coleorhiza; e-embryo; ep-epiblast; es-embryo sac; h-hilum; i-inner integument; 1-lemma; lo-lodicule; mi-mesocotyl internode; o-outer integment; p-palea; pl-placenta; r-rachilla; s-scutellum; st-scutellar tail; t-vascular trace to placenta; w-ovary wall. 403 Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae have more than two glumes at the base of the spikelet (see Glumes), but the sterile lemma of the panicoid spikelet does appear to represent an evolutionary reduction from a fertile floret. cules continue to be controversial. The grass flower has been interpreted variously as dichlamydeous (with two perianth whorls, the palea representing two fused sepals and the lodicules the petals), monochlamydeous (with only one perianth whorl, the lodicules, present, and the palea homologous to a prophyll), achlamydeous (with no perianth, the lodicules representing modified bracts or stipules), or pseudanthial (with the flower representing a highly reduced branch system, and the lodicules derived from leaves or branches) (Clifford, 1987). The monochlamydeous interpretation of the grass flower is the most widely accepted. Recent molecular genetic studies provide support for a petaloid homology of the lodicules (Irish, 1998; Schmidt & Ambrose, 1998; Ambrose et al., 2000), thus rejecting the achlamydeous and pseudanthial hypotheses. Ambrose et al. (2000) also provided genetic evidence that the palea and possibly also the lemma have characteristics in common with an outer perianth whorl, thus suggesting that the dichlamydeous interpretation might be revived. Bossinger (1990) and Pozzi et al. (2000) described mutations in barley in which the lemma is converted to a leaf. This may suggest that the lemma is more leaf-like than sepal-like. Kellogg (2000a) suggested that these two interpretations are not mutually exclusive, and that the complex structure of the grass spikelet may reflect simultaneous expression of both leaf and floral developmental "programs." If the latter interpretation is correct, then it may be meaningless to discuss whether the lemma is "really" a leaf or a sepal. Palea. The palea (Fig. 6A and B) is widely interpreted as a prophyll (Linder, 1987; Stapleton, 1997; Clayton, 1990; Soreng & Davis, 1998; Judziewicz et al., 1999), and prophylls do occur in the inflorescences of many Bambuseae and Andropogoneae (Judziewicz et al., 1999; E. A. Kellogg, unpublished obs.) relative to subtending bracts in precisely the same way a prophyll is related to a subtending leaf. Although the "palea as prophyll" explanation may be more parsimonious, the derivation of the palea from the fusion of two sepals also has been supported (Schuster, 1910; Stebbins, 1974; Irish, 1998; Schmidt & Ambrose, 1998). Nufiez (1968) and Clifford (1987) have suggested that the odd-nerved, often 1-keeled paleas in the Oryzeae represent a separate origin of a palea-like structure, but this interpretation is inconsistent with the phylogeny. There is no reason to suppose paleas have been replaced with something different in Oryzeae. Single-flowered spikelets lacking a palea (e.g., Alopecurus)probably represent loss of the rachilla extension and the palea on the floral axis. Vegetative branching in the grasses and other monocots typically involves a leaf with a bud or branch in its axil, although bud displacement is observed in some palms and grasses (Dahlgren et al., 1985; Serebryakova,1971; Fisher & Dransfield, 1977). The first appendage of the branching axis is an adaxial two-keeled bract (= prophyll).This prophyll encloses the bud, but often persists once the branch develops. The origin of the vegetative prophyll is not clear, but whether it is a single structure or the result of fusion of two bracts, ultimately it is foliar in nature (Stebbins, 1974). The relationship between the subtending leaf and the prophyll is generally constant, and therefore a prophyll marks the presence of a lateral branch. If the vegetative branching pattern is reiterated in the inflorescence, then the palea ought to be homologous to a prophyll, and prophylls should occur in fully bracteate inflorescences as the adaxial first appendage of branches in the axils of subtending bracts. Flower. Grass flowers are made up of a gynoecium, androecium, and two or three flap-like structures (lodicules) that force the floret open at maturity (Fig. 6A-G). The literature on the anatomy and development of the grass flower should probably be reinterpreted in the light of the present phylogeny. The origin and homology of the lodi- Androecium. All six stamens (arranged in two alternating whorls of three stamens each, Fig. 6A; see Clifford, 1987) are plesiomorphically present within the study group and at the point of origin of the grass family, but the entire outer whorl is lost in the Restionaceae. Within the grass family, all six stamens (both whorls) are maintained in the three earliest-diverging subfamilies (except for the loss of the inner anterior pair in Anomochloa, as noted above). The outer whorl is maintained throughout most of the grass family, except for the various aunoted tapomorphic losses and polymorphisms above. Loss of the entire inner whorl (e.g., Fig. 6B) is interpreted as a synapomorphy of the BEP + PACCAD Clade, but in one subclade within this large group (the bambusoid/ehrhartoid alliance) there are three or more independent reversions to presence of this whorl, and possibly a secondary loss in Leersia. An alternative interpretation of the outer stamen whorl, in the context of the present reconstruction of phylogenetic relationships, and involving no secondary origins, would have the in- 404 Annals of the Missouri Botanical Garden ner whorl lost independently in Pooideae and the PACCAD Clade, as well as in a series of small lineages within the bambusoid/ehrhartoid alliance. Alternatively, if a PACCAD + Pooideae lineage is considered, loss of the inner stamen whorl could be interpreted as a synapomorphy of that clade; the presence of this whorl in some genera of Ehrhartoideae and Bambusoideae might then be interpreted as retention of a plesiomorphy, while the absence in others would be interpreted as having arisen independently of the loss in the PACCAD + Pooideae. are additional differences in expression of the enzymes involved in photosynthesis (Sinha & Kellogg, 1996). For example, all Chloridoideae except for the C3 Eragrostis walteri and Merxmuellerarangei form aspartate, use the NAD-malic enzyme, and have double bundle sheaths, with the inner one of thick-walled cells. (Centropodiahas not been biochemically typed, but is anatomically the same as other NAD-ME taxa.) Stipagrostis,in Aristidoideae, is similar to the chloridoids in having two bundle sheaths, the outer of which appears to be carbonreducing (Sinha & Kellogg, 1996). Aristida itself forms malate, using the NADP-malic enzyme; unlike other NADP taxa, it has two bundle sheaths. The ultrastructureof the outer sheath is similar to that of NADP species, but the inner sheath is more like an NAD or PCK plant; the extent to which these sheaths are developed varies throughoutthe genus (Brown, 1977; Carolinet al., 1973). Eriachne also uses NADP-ME and has a double bundle sheath, but unlike Aristida the inner sheath has thick-walled cells. The C4 Panicoideae also vary biochemically and histologically, although large groups are uniform. For example, the entire tribe Andropogoneae (ca. 100 genera and 1000 species) uses NADP-ME and has a single bundle sheath. Both the phylogeny and structural/biochemical data indicate that the C4 pathwayis not homologous wherever it occurs. The close relationship of the C4 lineages, however, suggests that there were a set of changes at the base of the entire PACCAD Clade that made the pathway easier to evolve. If this were true, then those changes, whatever they were, would be homologous, even though the final manifestations of the pathway are not. BIOCHEMISTRY The phylogeny suggests that the C4 photosynthetic pathway has evolved multiple times within the PACCAD Clade. In the C4 pathway, the CalvinBenson cycle, and hence Rubisco, is relegated to the bundle sheath cells surrounding the veins (Hattersley & Watson, 1992; Sinha & Kellogg, 1996). Phosphoenolpyruvate carboxylase then catalyzes CO2 reduction in the mesophyll to produce the fourcarbon compound oxaloacetate. This compound is then reduced to malate or aspartate and transported to the bundle sheath, where the newly fixed CO2 is immediately removed and taken up by Rubisco. This keeps CO2 concentration high at the active site of Rubisco, preventing competition by O2. Consistent with the constant flow of materials between the bundle sheath and mesophyll, C4 species (Fig. 7C) have closer vein spacing than C3 species (Fig. 7B). Other anatomical manifestations of C4 photosynthesis include enlarged bundle sheath cells (Kranz anatomy), closely packed chlorenchyma cells, and in some, radiate chlorenchyma (Fig. 7C). Despite the foregoing generalizations, C4 photosynthesis actually represents a suite of characters, rather than a single genetic and phylogenetic change. Only down-regulation of Rubisco in the mesophyll, up-regulation of PEP carboxylase, and closer vein spacing are common to all C4 grass lineages. The four-carbon compound that transports the carbon to the bundle sheath may be malate or aspartate, the decarboxylating enzyme may be a malic enzyme using NAD as a co-factor (NAD-ME), or using NADP (NADP-ME). If the former, additional decarboxylation activity may be provided by PEP carboxykinase, so PCK activity is dependent upon the presence of NAD-ME (Kanai & Edwards, 1999). Some C4 grasses have only one bundle sheath, whereas others have two. In those with two bundle sheaths the inner sheath may be made up of thick-walled cells, forming a conventional mestome sheath, or it may be parenchymatous. There IMPLICATIONSFOR MORPHOLOGICALEVOLUTION The phylogeny provides a unique and powerful tool for description of evolutionarypattern(Kellogg, 2000a). Major clades and evolutionary transitions are summarized in Figure 8. Additional detail can be found at http://www.virtualherbarium.org/grass/ gpwg/default.htm. Sister relationship of Poaceae and Joinvilleaceae. The presence of long and short cells in the leaf epidermis (char. 42), with at least some of the short cells containing silica bodies, unambiguously supports the sister relationship of Joinvilleaceae and Poaceae. This arrangement is apparently unique among angiosperms (Campbell & Kellogg, 1987; Kellogg & Linder, 1995). The presence of the 6.4 kb inversion in the chloroplast genome (char. 51) is also an unambiguous synapomorphy supportingthis sister relationship, although it is not Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae A.k. 405 A Figure 7. Leaf anatomy.-A. Dinochloa maclelandii (Soderstrom2607). -B. Poa sp. (CarolinaBiological Supply Co.). -C. Bouteloua sp. (CarolinaBiological Supply Co.). ac-arm cell; bc-bulliform cell; ch-chlorenchyma; fcfusoid cell; is-intercellular space; ms-mestome sheath; p-phloem; ps-outer parenchyma sheath; rch-radiate chlorenchyma;sg-sclerenchyma girder; st-stomatal apparatus;x-xylem. known whether Ecdeiocoleaceae possess version (R. J. Soreng, pers. comm.). Monophyly analysis, the (Fig. 61 and and the trnT this in- Two characters in this of Poaceae. differentiated grass embryo highly J) and its lateral position (char. 34) inversion in the large single-copy re- gion of the chloroplast genome (char. 52), unambiguously support the monophyly of Poaceae. Two other features unique to and characteristic of Poaceae but not included in this analysis are the caryopsis and the presence of intraexinous channels in the pollen wall (Linder & Ferguson, 1985; Campbell & Kellogg, 1987; Kellogg & Linder, 1995). Panicoidea Centotheco Aristidoide Danthonioi Arundinoid Chloridoide Eriachneae Micraireae Pooideae Ehrhartoid Streptogyn Bambusoid Puelioideae Pharoideae Anomochloo Joinvilleace Other Po Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae The caryopsis is a single-seeded, usually dry, indehiscent fruit with the pericarp fused to the seed coat in the hilar region and otherwise closely adnate (Sendulsky et al., 1987). The caryopsis develops from a unilocular ovary containing a single ovule (Fig. 6H). Within Poaceae, the basic caryopsis has been modified to the fleshy (baccoid) or the achene-like (nucoid) caryopses of some woody bamboos (Sendulsky et al., 1987) or the follicoid or cistoid caryopses of some chloridoids in which the seed is free from the pericarp or separates from it when moistened. The highly differentiated grass embryo and its lateral position at the base of the caryopsis (char. 34; Fig. 61 and J) are synapomorphiesfor the family (Campbell & Kellogg, 1987; Kellogg & Linder, 1995). In the grasses, the embryo has leaves, vascular tissue, and clearly localized shoot and root meristems before the fruit is dispersed, and thus looks much more like a seedling than the embryos of non-grass relatives (Fig. 61 and J; Reeder, 1957; Sendulsky et al., 1987). Constant features of the grass embryo are the scutellum, coleoptile, and coleorrhiza. A scutellar cleft (Fig. 6I) may or may not separate the scutellum from the coleorrhiza,and the epiblast, an extra flap of tissue opposite the scutellum, may be present (Fig. 6J) or absent (Fig. 61). Whether or not the embryonic leaf margins meet or overlap varies throughoutthe family (Fig. 6K and L). Changes in the inflorescence occurred between the time that Joinvillea divergetdand the time of divergence of Anomochlooideae. Relative to most of their poalean sister families (excluding Restionaceae, Ecdeiocoleaceae, and Centrolepidaceae),the grasses, including Anomochlooideae, have well-developed bracts (what are normally called subtending bracts, prophylls, glumes, lemmas, and paleas) subtending and enclosing contracted inflorescence branches and flowers (Figs. 4 and 5). This appears to be a derived character marking the origin of the family, although this was not used explicitly as a characterin this analysis. Under this interpretation, the characteristic grass spikelet (found in all grass- es except Anomochlooideae) is then likely the result of a cumulative series of changes that occurred during the early history of the family. Clayton (1990), however, pointed out that there are welldeveloped bracts in the spikelets of Restionaceae, and what could be interpreted as subtending bracts are present, although not necessarily well developed, in Joinvilleaceae (Dahlgren et al., 1985). Clayton (1990) also noted that if the palea in grasses is interpreted as a prophyll, then there is no homolog for it among other Poales; however, prophylls are occasionally found at the base of spikelets in Restionaceae (H. P. Linder, pers. obs.). The results of this analysis and prior studies (Clark et al., 1995; Soreng & Davis, 1998) support the following as plesiomorphic within the grass family: an herbaceous, perennial, rhizomatoushabit; pseudopetiolate and relatively broad leaf blades bearing multicellular microhairs and anatomically with commissural veins, fusoid cells (Fig. 7A), and alternating long and short cells on the epidermis with at least some of the short cells including silica bodies; leaves with an adaxial ligule and open sheaths; a highly bracteate inflorescence; one-flowered spikelets or spikelet equivalents; six stamens in two whorls with tetrasporangiate, dithecal anthers; monoporate pollen with intraexinous channels in the wall; a uniloculate, uniovulate gynoecium with three stigmas and one order of stigmatic branching; a basic caryopsis with a linear hilum; a highly differentiated, laterally positioned embryo with a scutellum, coleoptile, coleorhiza, and a negligible mesocotyl internode; Festuca-type starch 407 grains in the endosperm; and the C3 photosynthetic pathway.The grass spikelet and lodicules may have evolved before the divergence of the Anomochlooideae and the rest of the family, and were lost in the Anomochlooideae (cf. Soreng & Davis, 1998), but these features are not plesiomorphic under the present optimization. Bisexual flowers are probably also plesiomorphic, but unisexuality evolved early and in a number of different lineages within the family. Figure 8. Summaryphylogeny of the grasses indicating significant morphological,ecological, and molecular(cpDNA = chloroplast DNA) events in the evolution of the family. Infrequentlosses, parallel gains, and reversals are not shown for these characters. The 12 subfamilies recognized by the GPWG appear in boldface. Poales sensu APG include Cyperaceae. Markedtaxa: (star)At least some included species have unisexual flowers/florets;(?) At least some included species have a C4 carbon fixation pathway,Kranz anatomy,or both. Dark circles indicate nodes strongly supported by all data combined (bootstrap> 99; Bremersupport > 16). Subfamilies with common names: Aristidoideae (wiregrasses, etc.), Arundinoideae (reeds, etc.), Bambusoideae (bamboos), Chloridoideae (lovegrasses, tef, etc.), Danthonioideae (oatgrasses, pampas grass, etc.), Ehrhartoideae(rice, wild-rice, etc.), Panicoideae (maize, panic grasses, millets, sorghum, sugar cane, etc.), and Pooideae (barley,brome grasses, oats, rye, wheat, etc.). 408 Annals of the Missouri Botanical Garden The basal divergence beAnomochlooideae. tween Anomochlooideae and the rest of the Poaceae (the Spikelet Clade) is well supported based on molecular evidence. Monophyly of Anomochlooideae, however, is supported morphologically only by the unreversed presence of the adaxial ligule as a fringe of hairs, a character that appears elsewhere in the family. The pulvinus at the summit of the pseudopetiole is a possible synapomorphy for this clade but requires further study to determine similarities with the structure in other grasses. Molecular support for this clade may be due at least in part to long-branch attraction (see Unresolved let equivalent of Anomochloa is indeed homologous to a standard grass lemma, a palea is lacking and the flower is terminal, whereas in a true grass floret the flower is borne on a lateral axis as indicated by the presence of the palea if it is interpreted as a prophyll. Some authors interpret the distal three bracts of the spikelet equivalent of Streptochaeta as lodicules and the next proximal two bracts as a bifid palea (e.g., Clayton, 1990), but there is no compelling evidence for this. The spikelet equivalent of Streptochaeta might represent a condensed branching system (Soderstrom, 1981), but it is not a pseudospikelet as found in the Bambuseae. In any case, lack of a palea in Streptochaeta also implies that the flower is terminal. questions). Anomochlooideae have been recognized as a separate family (Nakai, 1943), a point of view that is completely consistent with the phylogeny. We have chosen here to retain Anomochlooideae within Poaceae because of the strong synapomorphies linking them (notably the caryopsis and the highly differentiated embryo, see Fig. 8). Retention of Anomochlooideae in Poaceae is also taxonomically conservative, in line with all previous studies of the family, and consistent with the efforts of the APG to limit monotypic or small families (Chase et al., 2000a, b). Anomochlooideae are variable with respect to embryonic leaf margins and the epiblast. Embryonic leaf margins meet and the epiblast is present in Anomochloa, whereas the embryonic leaf margins overlap and the epiblast is absent in Streptochaeta (Judziewicz & Soderstrom, 1989). Both genera have an inconspicuous scutellar cleft. The coleoptile is usually represented as a more or less conical "cap" protecting the embryonic leaves and shoot apex, but the coleoptile margins are entirely free and overlapping in Streptochaeta, whereas in Anomochloa the margins at the base of the coleoptile are fused but free toward the apex, as is also seen in Pharoideae (Reeder, 1953; Judziewicz & Soderstrom, 1989). The inflorescences of both Anomochloa and Streptochaeta are bracteate, but the lack of clear homology of these bracts with those of the standard grass spikelet has been noted. The spikelet equivalents in both Anomochloa and Streptochaeta are one-flowered and bisexual. The upper bract in Anomochloa exhibits a laminar anatomical structure, with the transversely elongated cell layer subjacent to the abaxial epidermis of the bract (Judziewicz & Soderstrom, 1989). This laminar structure is not found in bracts of Streptochaeta. (Similar but not identical laminar anatomy characterizes female lemmas in Pharoideae.) As Soreng and Davis (1998) pointed out, if the upper bract of the spike- The Spikelet Clade (Pharoideae + [Puelioideae + {BEP + PACCAD]]). This clade, which includes all of the grasses except for Anomochlooideae, is defined by the unambiguous presence of true grass spikelets, florets (char. 8), and lodicules (char. 18). The plesiomorphic condition of the spikelet is clearly the presence of a pedicel (char. 10), two glumes, and a well-developed lemma and palea in the floret. The single-flowered spikelet may be synapomorphic for the family above the point of divergence of Anomochlooideae, with a transformation to multiflowered spikelets above Pharoideae and then numerous reversals, but the first true spikelets in grasses may have been multiflowered (see discussion under Spikelet). The plesiomorphic condition for lodicules is clearly three (char. 19), unfused (char. 20), and with a distally membranous portion (char. 21). Presence of unisexual flowers may be synapomorphic for this clade, with numerous reversals to bisexual florets, but it may be more likely that unisexuality arose multiple times. A base chromosome number of x = 12 (char. 47) was established before the divergence of Pharoideae. Monophyly of this clade is strongly supported by the presence of resupinate leaf blades, oblique lateral veins in the leaf blades, and uncinate hairs wholly or partially covering the female lemmas (Judziewicz, 1987). The female lemmas exhibit a laminar anatomical structure similar to that of Anomochlooideae, but the transversely elongated cell layer is subjacent to the adaxial epiand dermis. In Pharoideae, Anomochlooideae, some Bambusoideae (Ghopal & Ram, 1985), the coleoptile margins are free for at least a portion of their length, but the distribution of this feature in the rest of the BEP + PACCAD Clade is not well documented. Pharoideae embryos have an epiblast and a small scutellar cleft, and embryonic leaf margins meet (Judziewicz, 1987). The Pharoideae uniPharoideae. Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae 409 formly exhibit one-flowered, unisexual, paired phological synapomorphies support its monophyly: (char. 9) spikelets (Fig. 4B). When present in the loss of the pseudopetiole (char. 7), reduction to two male spikelets, the three lodicules are small, which lodicules (char. 19), loss of the inner whorl of stamay be plesiomorphic or may represent a reduc- mens (chars. 23 and 24), and loss of arm and fusoid tion. Disarticulation is variable, in that the whole cells (chars. 45 and 46). The pseudopetiole is reinflorescence may disarticulate as in Scrotochloa, gained in Bambusoideae, as well as in a few memor the whole inflorescence or branches usually dis- bers of the PACCAD Clade. Arm and fusoid cells articulate in Pharus, or female spikelets disarticu- are also regained in Bambusoideae. The inner late above the glumes in Leptaspis, and perhaps whorl of stamens is interpreted as having been realso in the other genera (Soderstromet al., 1987). gained three or four times within the bambusoid/ These, along with the uncinate hairs, appear to be ehrhartoid clade. Within the BEP + PACCAD adaptations to epizoochorous dispersal. Multicel- Clade, the lamina on the first seedling leaf is lost lular microhairs (char. 43) are lost in Pharoideae. only in Bambusoideae and Oryzeae. Unisexual floThe Bistigmatic Clade (Puelioideae + [BEP + PACCAD]). This clade is marked by three morphological synapomorphies: transformation from three to two stigmas (char. 29), transformationfrom one to two orders of stigmatic branching (char. 30), and presence of the 15 bp ndhF insertion (char. 53). Multiple florets per spikelet (char. 12; Fig. 5) may have arisen within Puelioideae as shown in Figure 3, or in the common ancestor of the Puelioideae + (BEP + PACCAD)clade. Regardless of which scenario is correct, reversals to one floret occurred numerous times in the BEP + PACCAD Clade. Disarticulation above the glumes (char. 16) clearly is established before the divergence of Puelioideae. rets have evolved in most lineages of this clade, e.g., Olyreae (Bambusoideae), Zizania (Ehrhartoideae), Lamarckia (Pooideae), several genera of Chloridoideae and Centothecoideae, and very commonly in Panicoideae. Most lineages include taxa with one floret per spikelet and taxa with multiple florets per spikelet. The presence or absence of an epiblast is variable, as is the presence or absence of the scutellar cleft, although in the PACCAD Clade the scutellar cleft is generally present (Reed- er, 1957). The large number of reversals hypothesized in this part of the tree raises a number of intriguing questions regarding morphological evolution. We do not know anything about the development or underlying genetics of the characters, so we are forced Puelioideae. The forest habitat and broad, into the agnostic assumptions that gains and losses are equally likely, and that pseudopetioles, arm pseudopetiolate leaf blades of the Anomochlooideae and Pharoideae are retained in this subfamily, cells, fusoid cells, epiblasts, and unisexual flowers but no unique morphological synapomorphies for are all developmentally and genetically the same Puelioideae have been identified. The culms ap- wherever they occur. The changes that we interpret as reversals could actually represent retained primparently do not produce aerial branches, nor basal itive characters if loss of these characters is more in as Anomochlooideae and Pharoideae. tillering The presence of proximal female-sterile florets in likely than their regain. Equally possible, the the spikelet (char. 11) is an unreversed synapo- changes interpreted as reversals could represent morphyfor Puelioideae in this analysis, but is also the origin of novel characters that look superficially a synapomorphy for Paniceae and autapomorphic similar to ancient ones. We have some evidence for for multiple other taxa on the tree. In Puelioideae, the latter (see below) in that arm cells in the bamthe pattern of sexuality in the spikelets is somewhat busoids are actually morphologically different from more complex than in many other subfamilies, but those in Anomochlooideae and Pharoideae (Zhang within a spikelet, at least some proximal florets are & Clark, 2000). Finally, the character optimizations male. In Guaduella, the 1 to 3 proximal florets are reflect the hypothesis that the BEP Clade is monomale, and additional florets are bisexual with the phyletic. If, as we outline below, the Pooideae are distalmost one or few reduced, but in Puelia the actually sister to the PACCAD Clade-a hypothesis proximal 3 to 6 florets are male or neuter, with the that is neither favored nor excluded by the datasingle apical floret unisexual and female. Multicel- then the pattern of morphological evolution is diflular microhairs (char. 43) are lost in Puelia, and ferent. a reversion to three stigmas (char. 29) occurs in some species of Puelia. The BEP + PACCAD Clade. This clade includes the vast majorityof grass species. Six mor- The BEP Clade (= BOP clade of Clark et al., This clade is supported by molecular se1995). quence data, particularly from ndhF, rpoC2, and phyB (see Results), but other data sets support a 410 Annals of the Missouri Botanical Garden Pooideae + PACCAD clade (Soreng & Davis, 1998). In this analysis, constraining Pooideae + PACCAD as monophyletic was only slightly less parsimonious than BEP + PACCAD (see Results). In addition, no morphological synapomorphies supporting the BEP Clade have been identified. Loss of the lemma awn is optimized to this node, but awns are regained in many taxa within the BEP Clade. The lack of sequence data for Streptogyna contributes to the uncertainty of its position within the BEP Clade and may also affect assessment of the monophyly of the clade. Streptogyna appears as sister to Ehrhartoideae (Fig. 1), but in other analyses of these data it appears as sister to the rest of the BEP Clade. Pooideae. Monophyly of the pooid clade is strongly supported by molecular data including cpDNA restriction site data (Soreng et al., 1990; Davis & Soreng, 1993; Nadot et al., 1994; Soreng & Davis, 1998, 2000). Parallel-sided subsidiary cells, lack of microhairs, nonvascularizedlodicules (Fig. 6D), and the presence of an epiblast and lack of a scutellar cleft in the embryo (Fig. 6J) are characteristic of a majorityof the subfamily but do not constitute unequivocal synapomorphies. In this analysis, the loss of stylar fusion (char. 28) is an unreversed synapomorphyfor the Pooideae. Loss of the scutellar tail (char. 36) is widespread in the clade, but polymorphismsprevent its unambiguous optimization. An unreversed transformationto faint or absent vascularizationof the lodicules (char. 22) occurs within the Pooideae after the divergence of Brachyelytrum.A transformationto the embryonic leaf margins meeting (as opposed to overlapping; Fig. 6L) also occurs after the divergence of Brachyelytrum but is reversed in Phaenosperma (or Phaenosperma + Anisopogon).Multicellularmicrohairs (char. 43) are known only in Lygeum + Nardus; although this character is scored only for the abaxial leaf surface, it appears that Pooideae, at least above this divergence, are the only group of grasses to lose completely the ability to make multicellular microhairsanywhere on the plant (except possibly the lodicules). Chromosomalevolution in Pooideae is complex (see char. 47), but x = 12 is apparently plesiomorphic in the BEP Clade, so numbers such as x = 10 and x = 11 in the earlierdiverging lineages of Pooideae may well be derived from this condition. The presence of x = 12 in Phaenosperma, Ampelodesmos, and some Stipeae may be a retention;x = 7 is clearly a synapomorphy of the core Pooideae (here represented by Brachypodium,Avena, Bromus, and Triticum).Twolodicules (char. 19) are found at the base of Pooideae, but a reversal to three occurs in Stipeae (in which another transformation,to two, occurs in Nassella); this is undoubtedly an oversimplificationof the pattern in the Stipeae in which lodicule numbervaries considerably (Vickery et al., 1986). Loss of the distally membranousportion of the lodicule (char. 21) is a synapomorphyfor Meliceae. The earliest-diverging lineages of the pooid clade have one floret per spikelet (char. 12) (although a rachilla extension is present in Brachyelytrum),multiple florets appear in Meliceae, single florets characterize the (Phaenosperma + Anisopogon) + Stipeae clade, multiple florets are found at the base of the core pooids, and many taxa within the core pooids have one floret per spikelet. Multiple independent origins of multiple florets per spikelet can be hypoth- Bambusoideae. Monophyly of the true bamboos (i.e., olyroid + woody bamboos) is supported by molecular data in this and other analyses (Clark et al., 1995; Zhang, 1996; Zhang & Clark, 2000). Morphologically, secondary gain of the pseudopetiole (char. 7) and secondary loss of the lamina of the first seedling leaf (char. 41) are synapomorphies. Although only presence or absence of arm cells was scored in this analysis, Zhang and Clark (2000) found that the presence of strongly asymmetrically invaginated arm cells (Fig. 7A) is a potential synapomorphy for this clade. Fusoid cells are characteristic of the Bambusoideae (Fig. 7A), but it is not known whether their presence represents retention of the plesiomorphic condition or reversal after loss of fusoid cells at the base of the BEP Clade. Bambuseae are here supported by the presence of perennating woody culms (char. 1), abaxial ligules (char. 5), and Panicum-type starch grains (char. 40). A secondary gain of the inner stamen whorl (chars. 23 and 24) occurred at least once but possibly several times. Olyreae have a synapomorphic base chromosome number of x = 11 (char. 47), but the tribe is also characterized by unisexual spikelets. Ehrhartoideae. This lineage is strongly supported by molecular data, and is characterized by the presence of one female-fertile floret per spikelet, often with one or two proximal female-sterile florets (char. 11). This character is coded as ambiguous in Oryza and Leersia, but if the vestigial structures at the base of the spikelets in these genera are interpreted as highly reduced glumes, then the presence of proximal female-sterile florets is an unambiguous synapomorphy. Two lodicules (char. 19) are found in this clade; in addition, the inner whorl of stamens (chars. 23 and 24) is regained, styles are not fused (char. 28), and fusoid cells (char. 46) are lost. 411 Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae esized, but subsequent reduction to one floret per spikelet has clearly occurred in several groups. Patterns of divergence within this clade are complex and still are being evaluated, so some inferences regarding character evolution are likely to change. lished for the BEP + PACCAD Clade, but no transformations to any other number occur in the PACCAD Clade. The positions of Micraira and Eriachne in the phylogeny are not well resolved, presumably due to a lack of sequence data for both genera (see Results and also Unresolved Questions). The two species of Eriachne are from quite distinct parts of the genus, based on the informal classification of Lazarides (1995). This undoubtedly affects the interpretation of character state transformations within the entire clade. Early in the evolution of the PACCAD Clade, some lineages developed the capacity for C4 photosynthesis, apparently as an adaptation to high light/high temperature conditions and perhaps also to falling levels of atmospheric CO2 (Sage & Monson, 1999). Most members of the Panicoideae, all but two Chloridoideae, the Aristidoideae (except for Sartidia), and the Eriachneae are C4. The poor resolution of the phylogeny at the base of the PACCAD Clade makes it impossible to determine precisely how many origins of C4 photosynthesis there were, but certainly there were at least two, and possibly more. The data are also consistent with a polymorphism at the base of the PACCAD Clade. The PACCAD Clade. Over half the species of the grass family are included in this clade. Even as early as the 1930s (Avdulov, 1931; Prat, 1932, 1936; Roshevits, 1937), taxa of this clade have been grouped together. Hilu and Wright (1982) were the first to retrieve this clade in a formal analysis, and subsequently support for the monophyly of the clade is found in all molecular analyses to date with sufficient sampling (Hilu & Esen, 1988; Hilu & Johnson, 1991; Davis & Soreng, 1993; Nadot et al., 1994; Barker et al., 1995; Clark et al., 1995; Duvall & Morton,1996; Liang & Hilu, 1996; Mathews& Sharrock, 1996; Soreng & Davis, 1998; Hsiao et al., 1999; Mathewset al., 2000) except for Cummingset al. (1994), in which an oryzoid clade nested within the PACC clade. Davis and Soreng (1993) named this the PACC clade based on the four subfamilies that were then recognized as comprising the clade, but we here modify the name to reflect the recognition of two additional subfamilies, the Aristidoideae and the Danthonioideae. The PACCAD Clade is robustly supported based on molecular data and additionally is supported by the presence of an elongated mesocotyl internode (char. 37) and the loss of the epiblast (char. 35; Fig. 61). The latter character reverses in the clade, so that secondary gain of the epiblast is an apparent synapomorphy for Centhothecoideae. Two characters (chars. 21 and 50) are possible synapomorphies for the PACCADClade, but because of a lack of data or lack of a structurein Micraira,placement of these transitions is ambiguous. The lack of lodicules in Micraira prevents unambiguous placement of the loss of the distally membranousportion of the lodicule (char. 21), and Micrairaremains unsampled for the presence or absence of the 3 bp deletion in phytochrome B (char. 50). Solid culm internodes (char. 2) are shown here as synapomorphic, although hollow ones reappear in other members of the clade. Non-linear hila (char. 33; Fig. 6M) are widespread in the PACCADClade, but the point of origin is ambiguous.The Panicum-type starch grain syndrome (char. 40) may be a synapomorphyfor the PACCAD Clade, with a reversal to the Festuca-type in the clade containing Eriachne, Aristidoideae, Danthonioideae, Arundinoideae, and Chloridoideae (the Ligule of Hairs Clade, as defined below), but other optimizations are possible. Two lodicules (char. 19) are estab- + Centothecoideae Clade. The Panicoideae This clade was recovered in virtually all subanalyses, and had reasonable support (bts 85, brs 8) in the combined analysis. The presence of non-linear hila (char. 33; Fig. 6M) is a potential synapomorphy for this clade. Although support for the monophyly of Panicoideae (excluding Gynerium and Danthoniopsis) was strong (see Results), relationships of the centothecoid taxa, Gynerium, and Danthoniopsis to the Panicoideae and to each other remain unresolved, but the placement of Gynerium as sister to traditional Panicoideae is a novel result. The presence of proximal femalePanicoideae. sterile florets (char. 11) and the transformation to the classical NADP-ME C4 subtype (char. 48) are unambiguous synapomorphies for Danthoniopsis + Panicoideae. Some reversions to the C: pathway occur within the Paniceae among unsampled taxa, and at least one secondary transformation to the NAD-ME C4 subtype occurs in Panicum. This clade is also supported by the presence of one female-fertile floret (char. 12) as a reversal and the gain of a germination flap (char. 17), but the placement of this latter transformation is ambiguous. The loss of disarticulation above the glumes (char. 16) is a synapomorphy for Panicoideae excluding Danthoniopsis. The presence of paired spikelets (char. 9) is a synapomorphy of Andropogoneae in this 412 Annals of the Missouri Botanical Garden relationship of these two clades is relatively modest. The gain of the NAD-ME C4subtype (char. 48) is a possible synapomorphy for the entire clade, Centothecoideae. Monophyly of this subfamily however, and if so it would then revert to C3 in M. as currently constituted is not strongly supportedin rangei. The gain of chloridoid-type microhairs this analysis. The secondary gain of an epiblast (char. 44) is a synapomorphy for the traditional (char. 35) is a possible synapomorphy(but is un- Chloridoideae, although the character does occur known for Thysanolaena), as is fusion of the styles elsewhere in the PACCADClade, and many genera (char. 28). of chloridoids also include species with panicoidThe Ligule of Hairs Clade (Eriachne + [[Aristi- type microhairs (Jacobs, 1987; Van den Borre, doideae + Danthonioideae] + [Arundinoideae + 1994; Van den Borre & Watson, 1994). The periChloridoideae}]). The adaxial ligule as a fringe of carp is often free or loose, but this feature is not hairs (char. 4), awned lemmas (char. 13), and com- uniform and is also found in non-chloridoid grasspound starch grains (char. 40) are synapomorphies es. of this clade, but characters 4 and 13 reverse multiple times, and character 40 once, within this UNRESOLVED QUESTIONS clade, as well as elsewhere on the tree (Fig. 3). The Monophyly of Anomochlooideae. Anomochloa recovery of this clade is a novel finding, but further and Streptochaetaare each distinctive genera but investigation is warranted, given the lack of seto have little in common. We have not yet appear quence data for Eriachne. The transformationto found a uniquely derived morphological character embryonic leaf margins meeting (char. 38; Fig. 6L) is an unreversed synapomorphyof the four subfam- that unites them as members of a single clade. Although the analyses presented here indicate that ilies above Eriachne. the two form a monophyletic group, analyses of sinThe Aristidoideae + Danthonioideae Clade. gle data sets sometimes show them to be paraphyAlthough each of these subfamilies is well sup- letic or unresolved (Mathews et al., 2000; Hilu et ported as monophyletic, their sister relationship is al., 1999; Zhang, 2000). Because both genera ocanother novel result, and one that is only moder- cupy long branches in gene trees, they may form a ately supported. Nonetheless, the presence of a ba- clade only because of long-branch attraction (Felsic pattern of three awns (char. 14; Fig. 5C and F) senstein, 1978). Molecular studies of other species is an unreversed synapomorphyfor this clade. Re- of Streptochaeta would help break up the long appearance of the distal membranousportion of the branch to S. angustifolia and might affect the lodicules (char. 21) also may be a synapomorphy, monophyly of the clade. Resolution of Anomochloa and Streptochaeta as two separate basal lineages although this reverses within the Danthonioideae. would affect interpretationsof character Aristidoideae. Gain of a germinationflap (char. obviously evolution within the family. 17) and transformationto a base chromosomenumber of x = 11 (char. 47) are unambiguous synaPosition of Streptogyna. As noted in Results, the position of Streptogyna is ambiguous, apparpomorphies for the clade. caused by lack of data. There are two species Danthonioideae. The presence of haustorial ently in the genus, one in the New Worldtropics and the synergids (char. 32) is interpretedas an unreversed other in Africa. Neither has been collected fresynapomorphy,but wider sampling within the clade quently, and we do not know of any plants in culis needed. tivation. Morphologically,the genus would fit comArundinoideae. No unambiguous morphologi- fortably within the Bambusoideae, but molecular cal supportfor the monophylyof this subfamily was data suggest that it is an early-diverging member found, although a reversal to hollow culms (char.2) of the BEP Clade or the Ehrhartoideae.The characters it shares with Bambusoideae are thus preoccurs in this clade. sumably ancestral, not indicative of relationship. Chloridoideae. Chloridoideae, including Cen- Accurate placement of Streptogynais necessary for tropodia and Merxmuellera rangei, are supported interpretation of character evolution in the earlybased on molecular data, although no clearcut mordiverging members of Bambusoideae, Ehrhartophological synapormorphies have been identified. ideae, and Pooideae. Monophylyof Centropodia+ M. rangei is well supEarly-divergingPooideae. The combined analported as is that of the traditional Chloridoideae (i.e., Chloridoideaes. str.), but supportfor the sister ysis confirmsthe position of Brachyelytrumas sister analysis, but paired spikelets do occur within Paniceae (e.g., Brachiaria, Digitaria, Paspalum). Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae to the rest of the Pooideae, and Lygeum + Nardus as the next-diverging lineage; both these results are well supported. The next diverging lineages include supported as monophyletic by our data. The list of genera included in and excluded from each subfamily, however, is based on a ratherlimited sample of species and genera, combined with inferences from classical morphological studies. In particular, the exact circumscriptions of Danthonioideae, Arundinoideae, and Centothecoideae are not precisely determined by this study. A comprehensive effort by multiple systematists is needed to improve understanding of the many poorly known species and genera. Phaenosperma, Anisopogon, Stipeae, Ampelodesmos, Meliceae, and Diarrheneae, but the order of divergence is not resolved by any data collected to date. In the case of Phaenosperma, Anisopogon, Diarrhena, and Ampelodesmos, the problem may be ascribed to insufficient sequence data in this analysis. For our sample of Stipeae and Meliceae, however, appreciable sequence data are available, yet the relative positions of the two lineages remain unclear. If the phylogenetic problem is indeed soluble with molecular data, the sample of genera and species in each tribe may have to be increased substantially. A combined analysis of cpDNA restriction sites and morphology (Soreng & Davis, 2000) represents the broadest taxon sample for Pooideae among studies to date. The order of divergence of these lineages affects interpretation of the evolution of such characters as parallel-sided subsidiary cells, loss of microhairs, and trends in reduction of chromosome number (Kellogg, 1998). The latter may correlate with a marked increase in genome size (Bennetzen & Kellogg, 1997) and may suggest possible mechanisms of genome evolution. The PACCAD Clade. Relationships among the major lineages in the PACCAD Clade are not resolved by this or any other phylogenetic analysis to date. In the combined analysis, the branches at the base of the clade are short, marked by relatively few mutations each (11, 41, and 16 steps; Fig. 1). This suggests that the PACCAD radiation may have occurred relatively rapidly. If this is true, then relationships may remain difficult to resolve with certainty. The clade also contains a number of taxa of uncertain placement, many of which have received little or no attention in phylogenetic studies. The tribe Eriachneae, which includes the Australian genera Eriachne and Pheidochloa, is represented here only by an rbcL sequence of Eriachne triodioides and an ITS sequence of E. triseta. The genus Micraira, the only member of the Australian tribe Micraireae, is represented only by an ndhF sequence of M. lazaridis and by an ITS sequence of M. subulifolia. Such genera as Cyperochloa, Steyermarkochloa, and the Crinipes group were not included in this combined analysis. An rbcL seit with the quence of Cyperochloa places centothecoids, whereas a sequence of the crinipoid genus Styppeiochloa places it sister to Arundineae s. str. (Barker, 1997; Linder et al., 1997). The subfamilies recognized within the PACCAD Clade are, except for Centothecoideae, strongly 413 Centothecoideae. Of the groups recognized here as subfamilies, Centothecoideae are the only one not strongly supported as monophyletic by the combined analysis. We have retained the subfamilial name and expanded the circumscription to include Thysanolaena, formerly a member of the Arundinoideae. As with the remainder of the PACCAD Clade, a clear picture of the limits of the centothecoid clade depends on much more data, particularly on the remaining centothecoid genera, but a study is under way (J. G. Sanchez-Ken, pers. comm.). Circumscriptionof tribes. This paper does not address tribal circumscription. This will require far more extensive sampling, particularly in Pooideae, Panicoideae, Chloridoideae, and Bambusoideae, which constitute the four largest subfamilies. Choice of outgroups for such studies is now clear, however. Biogeography. Present-day distributions do not indicate much about where the grasses originated. Restionaceae are clearly a Gondwanangroup, with representatives in Africa and Australia. Joinvilleaceae, however, are insular, occurring on Borneo, New Caledonia, and Pacific Islands. The basal lineages of the grasses are found in the tropical regions of South America, Africa, and Asia; the Anomochlooideae are restricted to South and Central America (Judziewicz & Soderstrom, 1989), the Pharoideae are pantropical (Soderstrom et al., 1987), and the Puelioideae are restricted to tropical Africa (Soderstrom & Ellis, 1987; Clark et al., 2000). Due to the absence of an early fossil record, it is not clear how this distributionwas established, whether by long-distance dispersal across the Atlantic and Indian Oceans, or whether across a continuous Gondwanan equatorial forest. Either way, the continent of origination cannot be determined with current data. Timing and causes of diversification. The earliest unequivocal grass fossils are pollen grains from the Paleocene of South America and Africa, 414 Annals of the Missouri Botanical Garden deposited approximately 60 to 55 million years (my) ago (Jacobs et al., 1999), although some grains of Monoporitesfrom the upper Maastrichtian(Cretaceous) may also represent remains of grasses (Linder, 1987). The earliest known grass macrofossil appears in an early Eocene formation (ca. 55 mya) in North America (Crepet & Feldman, 1991). Based on the fossil record therefore, the family is at least 55 my and possibly as much as 70 my old. Establishment of all major lineages had occurred by the mid-Miocene (Jacobs et al., 1999), which is about the time that grass-dominatedecosystems appeared. Attempts to date particular nodes on the cladogram using molecular clocks are confounded by non-clocklike behavior of several of the genes (Gaut et al., 1996, 1997; Mathewset al., 2000; Kellogg & Russo, unpublished obs.). Using sequences of GBSSI, which has been shown to exhibit clocklike mutation, Gaut and Doebley (1997) placed the divergence of maize and Pennisetum at 25 mya, whereas Kellogg and Russo (unpublished) place the divergence of Danthoniopsisdinteri from the rest of the panicoids at ca. 16 mya. The two dates conflict with each other, but do suggest that the PACCAD Clade originated in the early Miocene or late Oligocene. All C4 lineages are included in the PACCAD Clade, so paleontological evidence for C4 photosynthesis can establish a minimumage for the common ancestor of the clade. The earliest known C4 grass macrofossil is dated at 12.5 mya (Nambudiriet al., 1978), and the earliest isotopic evidence for C4 is ca. 15 mya (Kingston et al., 1994; Latorre et al., 1997). This suggests that the origin of the PACCAD Clade occurred no later than 15 mya and possibly as early as 25 mya. Both fossil data and molecular clock estimates seem at odds with the apparent Gondwanandistribution of many grass taxa (see for example Simon & Jacobs, 1990). The Gondwanan distribution of such derived groups as the subfamily Danthonioideae might suggest that the PACCAD Clade originated sometime before the breakup of Gondwana, which would then place the origin of the family long before the earliest known fossils were deposited. This cannot be ruled out, of course, because it is an assumption based on negative evidence. If, however, we assume that groups within the PACCAD Clade originated before the breakup of Gondwana (a process hard to date precisely but perhaps 10070 mya), then we would have to assume that the family originated more than 200-140 mya, before the time of the first appearance of angiosperms in the fossil record. It seems more likely, and more consistent with available data, that the grasses achieved their Gondwanan distribution by dispersal (Soreng & Davis, 1998), as has been suggested for other taxa with an apparent Gondwanan distribution (e.g., Adansonia, Baum et al., 1998; Atherospermataceae, Renner et al., 2000). The combination of fossil data and molecular clock evidence suggests that the major diversification of the grasses occurred between 15 and 25 mya, long after the origin of the family at 55 to 70 mya. This is consistent with the observed branch lengths on the phylogeny in Figure 1. There may have been many more representatives of the Anomochlooideae, Pharoideae, and Puelioideae (or even additional lineages) that are now extinct, but grasses are generally rare in the fossil record until the Miocene (Jacobs et al., 1999). The simplest explanation is that the family diversified long after its origin. The novel characters that arose after the divergence of Joinvillea-the caryopsis, differentiated embryo, reduction in perianth-therefore did not lead immediately or directly to the current dominance of the family. Other characteristics acquired later in the evolution of the family may have been more important in its diversification and ecological success. Possibilities include such characters as formation of intercalary meristems or the acquisition of mechanisms for drought tolerance. We do not know the phylogenetic distribution of intercalary meristems, however, and it is possible that intercalary meristems of the leaves evolved after such meristems in the stems. This character needs to be investigated further. Acquisition of drought and heat tolerance would also be worth investigating, but would require a precise definition of what is meant by each term. The cellular components of such physiological responses are being identified and could perhaps be studied across a range of taxa. CONCLUSIONS We present here a resolved and strongly supported phylogeny of the grass family. It can be used to understand the diversification of morphology, genes, and genomes, to interpret comparative studies of cereal crops and forage grasses, and to develop hypotheses of adaptation to past and future environments. Some phylogenetic questions remain unresolved, and these affect inferences about such important characters as C4 photosynthesis. Nonetheless, this phylogeny is one of the most comprehensive and robust available for any family of plants, making the grasses an excellent clade for studies of evolutionary pattern and process. Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae TREATMENT TAXONOMIC Thysanolaeneae placed in the Centothecoideae and Gynerium as Incertae Sedis. Centropodia and Merxmuellera rangei are placed in Chloridoideae. Pooideae have grown by inclusion of Brachyelytreae, Lygeeae, Nardeae, Phaenospermatideae, Diarrheneae, Stipeae, and Ampelodesmeae, all formerly classified within either Bambusoideae or Arundinoideae by some authors; note, however, that Clayton and Renvoize (1986) placed Lygeeae, Nardeae, and Stipeae in Pooideae in agreement with the classification proposed here. A detailed comparison of the GPWG classification with the major grass classification systems of the 20th century is presented in Table 1. Primary sources for suprageneric names were the STAR Database (http://matrix.nal.usda.gov:8080/ the Catalog of New star/supragenericname.html), World Grasses (http://mobot.mobot.org/W3T/ and Clayton and Renvoize Search/nwgc.html), (1986). Diagnoses of the subfamilies were extracted from various sources including Clayton and Renvoize (1986) and Watson and Dallwitz (1992). Tribes in Chloridoideae and Panicoideae (except for the exclusion of Eriachneae) follow the treatment of Clayton and Renvoize (1986); tribes listed for the other subfamilies generally are treated according to more recent studies and/or consultation with specialists in those groups. Twelve subfamilies are recognized formally in this classification system (Table 1). A description is provided for each subfamily, and where appropriate, synonymy is indicated. To permit easy comparison with previous work, we have listed for each subfamily which of the tribes recognized by Clayton and Renvoize (1986) are to be included. In some cases (e.g., Pharoideae or Danthonioideae), the new circumscription of subfamilies makes tribal recognition largely unnecessary. For example, the subfamily Pharoideae includes three genera in a single tribe; the tribe is effectively redundant and serves no useful function in the subfamilial classification. Nonetheless we list the names for comparison. Our sample of taxa was explicitly designed to explore relationships among major clades that can be recognized at the subfamilial level, but it is not dense enough to evaluate tribal limits. We have in many cases combined molecular data from several species to represent a genus (as is also commonly done for morphological analyses), and in a few cases have combined data from several genera that represent a putatively monophyletic group. Such combinations assume, rather than test, monophyly. We therefore refrain from formal discussion of tribal limits, which cannot be addressed by our (lata; these limits will have to be re-evaluated by future studies. Three tribes and two genera are placed as Incertae Sedis at the end of the classification, although the genera may be provisionally placed as noted below. This classification reflects our attempt to use the phylogeny as the basis for recognizing subfamilies while remaining nomenclaturally conservative. Except for Centothecoideae, all subfamilies recognized are well supported as monophyletic in our analyses. While we could create an unranked classification for the grasses using our phylogeny, we feel that the practical interests of the potential users of this classification currently are best served by retaining the Linnaean hierarchy. Nonetheless we have applied informal names to several of the well-supported clades (see above). The most significant changes in our proposed subfamily classification are the breakup of the traditional Bambusoideae and Arundinoideae and the expansion of Pooideae. The diversity encompassed by the traditional Bambusoideae (or Bambusoideae s.l.) is now recognized as Anomochlooideae, Pharoideae, Puelioideae, Bambusoideae s. str., and Ehrhartoideae. Elements of the traditional Arundinoas Aristidoideae, ideae are now recognized Danthonioideae, and Arundinoideae s. str., with 415 Poaceae (R. Br.) Barnh., Bull. Torrey Bot. Club 22: 7. 1895. (Nom. alt. Gramineae Juss., Gen. P1.: 28. 1789.) A monophyletic family, recognizable by the following synapomorphic morphological characters: Inflorescence highly bracteate. Perianth reduced or lacking. Pollen lacking scrobiculi, but with intraexinous channels. Seed coat fused to inner ovary wall at maturity, forming a caryopsis. Embryo highly differentiated with obvious leaves, shoot and root meristems, and lateral in position. I. Anomochlooideae Pilg. ex Potztal, in Willdenowia 1: 772. 1957. TYPE: Anomochloa Brongn. Figure 4C and D. Syn.: Streptochaetoideae (Nakai) Butzin, Neue Unters. Blute Gram.: 148. 1965. Plants perennial, rhizomatous, herbaceous, of shaded tropical forest understories. Culms hollow or solid. Leaves with phyllotaxis either distichous or spiral; abaxial ligule absent; adaxial ligule a short fringe of cilia or absent, not membranous; blades usually relatively broad, venation parallel, with pseudopetioles short to very long, these with dark, turgid swellings (pulvini) at both ends (An- 416 Annals of the Missouri Botanical Garden omochloa) or only at the summit (Streptochaeta); sheaths non-auriculate. Inflorescences spicate, with complicated branching patterns, bracts outside of the spikelet equivalents present, large and with a blade or small and bladeless, or absent. Ultimate structures of the inflorescence (spikelet equivalents) of uncertain homology with typical grass spikelets but one-flowered and bisexual; bracts within the spikelet equivalents with phyllotaxis distichous or spiral, lacking uncinate macrohairs, sometimes awned but if so, the awns single; lodicules absent, or, in Anomochloa, their position occupied by a ring of short brownish cilia borne on a low membranousring; stamens 4 or 6; ovary glabrous, apical appendage absent, haustorial synergids presumed absent, style 1, stigma(s) 1 or 3. Caryopsis with the hilum linear, shallow and inconspicuous; endosperm hard, containing compound starch grains; embryolarge, epiblast present or not, scutellar cleft present but shallow, mesocotyl internode absent, embryonic leaf margins overlapping or not. Basic chromosome numbers: x = 11 or 18 (note: Clark & Judziewicz, 1996, erroneously cited these as 12 or 18). II. Pharoideae (Stapf) L. G. Clark & Judz., Taxon 45: 643. 1996. TYPE: Pharus P. Browne. Figure 4B. Foliar anatomy. Mesophyll nonradiate, an adaxial palisade layer absent, with fusoid cells very large and well developed, arm cells only weakly developed; Kranz anatomyabsent; midrib complex; adaxial bulliform cells present. Foliar micromorphology. Stomata with low dome-shaped and triangular subsidiary cells; bicellular microhairsvery large (0.075-0.15 mm), the pointed apical cell usually one and a half times as long as the basally constricted basal cell; papillae absent. Photosyntheticpathway. Presumed C3. INCLUDED TRIBES: Syn.: Leptaspidoideae (Tzvelev) C. 0. Morales, Sendtnera 5: 244. 1998. Nom. superfl. Plants perennial, rhizomatous, monoecious, herbaceous, of shaded tropical to warm temperate forest understories. Culms hollow or solid. Leaves distichous; abaxial ligule absent; adaxial ligule a fringed membrane; blades resupinate, relatively broad, with pseudopetioles prominent and twisted, with lateral nerves diverging obliquely from midnerve and running straight to margins; sheaths nonauriculate. Inflorescences paniculate, the main axis and branches disarticulating or not, covered with uncinate macrohairs, bracts outside of the spikelets absent. Spikelets unisexual, one-flowered, mostly in male-female pairs on short branchlets, or some female spikelets solitary. Female spikelets large, short-stalked; glumes 2, shorter than the floret; lemma tubular or inflated, covered wholly or in part by uncinate macrohairs, awnless; palea well developed; lodicules absent; ovary glabrous, apical appendage absent, haustorial synergids ptresumed absent, style 1, stigmas 3. Caryopsis with the hilum linear, extending the full length; end(osperm hard, without lipid; embryo small, epiblast present, scutellar cleft present but shallow, mesocotyl internode absent, embryonic leaf margins overlapping. Male spikelets small, short- to long-stalked, membranous; glumes 2, shorter than the floret; lodicules 3 or 0, if present then minute, elliptic, glabrous, and nerveless; stamens 6. Basic chromosome number: x= 12. Foliar anatomy. Mesophyll nonradiate, an adaxial palisade layer absent, fusoid cells large and well developed, arm cells weakly to moderately well developed; Kranz anatomy absent; midrib complex; inflated adaxial interstomatal cells present, bulliform cells poorly developed or absent. Anomochloeae C. E. Hubb., in Hutchinson, Fam. Foliar micromorphology. Stomata with parallelFl. P1. 2: 219. 1934. TYPE: Anomochloa sided to dome-shaped subsidiary cells; bicellular Brong. Streptochaeteae C. E. Hubb., in Hutchinson, Fam. microhairs and papillae absent. Fl. PI. 2: 205. 1934. TYPE: Streptochaeta Presumed C3. Photosynthetic pathway. Schrad. ex Nees. Notes. There is no unique morphological synapomorphyfor this subfamily, but both tribes lack lodicules and they apparently also lack grass-type spikelets. As noted above (Unresolved Questions), this lineage may not be monophyletic, in which case two subfamilies would need to be recognized. The subfamily includes 4 species. INCLUDED TRIBE (NOW IDENTICAL TO SUBFAMILY AND THUS REDUNDANT): Phareae Stapf, in Thiselton-Dyer, Fl. Cap. 7: 319. 1898. TYPE: Pharus P. Browne. Notes. In his original description of the tribe, Stapf specifically included Olyra (based on its uni- 417 Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae sexual spikelets), but did not explicitly list Pharus or Leptaspis, although his choice of the name Phareae implicitly recognized the membership of Pharus in the tribe and automatically placed Pharus as its type, according to Article 10.6 of the Code (Greuter et al., 2000). As long as Olyra was retained in the same tribe as Pharus, Phareae was a superfluous name for the Olyreae. When Pharus and Leptaspis are segregated into their own tribe, and Olyra is excluded, then Phareae becomes the valid, correct name for the tribe. Clark and Judziewicz (1996) based the name of the subfamily on this tribal name. Tzvelev (1989) argued that the name Phareae was illegitimate because the type of the previously described tribe Olyreae was included in it, and provided the name Leptaspideae for this tribe. Morales (1998) agreed with Tzvelev and rejected the name Pharoideae for this subfamily, according to Article 52.1 of the Code (Greuter et al., 2000), replacing it with Leptaspidoideae. Under Article 52.3, however, "A name that was nomenclaturally superfluous when published is not illegitimate ... if it is based on the stem of a legitimate generic name." We therefore accept the name Pharoideae for this subfamily, as Pharus is a legitimate generic name. The subfamily includes 12 species. axial palisade layer absent, fusoid cells well developed, arm cells only weakly developed; Kranz anatomy absent; midrib complex or less commonly simple; adaxial bulliform cells present. L. G. Clark, M. Kobay., S. Mathews, Spangler & E. A. Kellogg, Syst. Bot. 25: 181-187. 2000. TYPE: Puelia Franch. Figure 4A. III. Puelioideae Plants perennial, rhizomatous, herbaceous, of Culms hollow. shaded rainforest understories. Leaves distichous; abaxial ligule absent (Guaduella) or present (Puelia); adaxial ligule a fringed membrane; blades relatively broad, pseudopetiolate, venation parallel; sheaths non-auriculate. Inflorescences racemose or paniculate, bracts outside of the spikelets sometimes present. Spikelets with two glumes and several florets, the 1 to 3 proximal florets male, the next several florets female-fertile, with distal incomplete florets (Guaduella), or the proximal 3 to 6 florets male or neuter with the single distal floret female (Puelia), disarticulating above the glumes and between the florets (Guaduella) or not (Puelia); lemmas lacking uncinate macrohairs, awnless; palea well developed, sometimes tubular; lodicules 3, membranous, ciliate; stamens 6; ovary glabrous or hairy, an apical appendage present or not, haustorial synergids presumed absent, styles 2 or 3, the bases close, stigmas 2 or 3. Caryopsis with a long-linear hilum; embryo small. Basic chromosome number: x = 12. Foliar anatomy. Mesophyll nonradiate, an ad- Foliar micromorphology. Stomata with domeshaped to triangular subsidiary cells; microhairs absent (Puelia) or multicellular, uniseriate microhairs present (Guaduella);papillae present or more commonly absent. Photosynthetic pathway. INCLUDED Presumed C3. TRIBES: Guaduelleae Soderstr. & R. P. Ellis, in Soderstrom et al. (editors), Grass Syst. Evol.: 238. 1987. TYPE: Guaduella Franch. Puelieae Soderstr. & R. P. Ellis, in Soderstrom et al. (editors), Grass Syst. Evol.: 238. 1987. TYPE: Puelia Franch. Notes. This subfamily, which comprises approximately 14 species, is poorly known, and morphological, anatomical, cytological, and ecological studies are needed. IV. Bambusoideae Luerss., Grundz. Bot., ed. 5: 451. 1893. TYPE: Bambusa Schreb. Figures 4F, 6C, G, O, P, 7A. Syn.: ()lyroideae Pilger, Nat. Pfl.-Fam. ed. 2, 14(1: 168. 1956. Parianoideae (Nakai) Butzin, Neue Unters. lIiite (ram.: 148. 1965. Plants perennial (rarely annual), rhizomatous, herbaceous or woody, of temperate and tropical forests, tropical high montane grasslands, riverbanks, and sometimes savannas. Culms hollow or solid. Leaves distichous; abaxial ligule absent (Olyreae) or present (Bambuseae);adaxial ligule membranous or chartaceous, fringed or unfringed;blades usually relatively broad, pseudopetiolate, venation parallel; sheaths often auriculate. Inflorescences spicate, racemose or paniculate, completing development of all spikelets in one period of growthand subtending bracts and prophylls usually absent, or pseudospikelets with basal bud-bearing bracts producing two or more orders of spikelets with different phases of maturityand subtending bracts and prophylls usually present. Spikelets (or spikelets proper of the pseudospikelets) bisexual (Bambuseae) or unisexual (Olyreae), consisting of 0, 1, 2 or several glumes, 1 to many florets; lemma lacking uncinate macrohairs, if awned, the awns single; palea well developed; lodicules usually 3 (rarely 0 to 6 or many), membranous, vascularized, often ciliate; stamens usually 2, 3, or 6 (10 to 40 in Pariana, 6 418 Annals of the Missouri Botanical Garden to 120 in Ochlandra); ovary glabrous or hairy, sometimes with an apical appendage, haustorial synergids absent, styles 2 or 3, sometimes very short but close, stigmas 2 or 3. Caryopsis with hilum linear (or rarely punctate), extending its full length (or rarely less than full length); endosperm hard, without lipid, containing compound starch grains; embryo small, epiblast present, scutellar cleft present, mesocotyl internode absent, embryonic leaf margins overlapping. Basic chromosome numbers: x = 7, 9, 10, 11, and 12. ittate (Phyllorachideae), somewhat broad to usually narrow, sometimes pseudopetiolate, venation parallel; sheaths sometimes bearing auricles. Inflorescences paniculate or racemose, bracts outside of the spikelets rarely present (Humbertochloa). Spikelets bisexual or unisexual, with glumes 2 (absent in some Oryzeae), sterile florets 0 to 2, and female-fertile floret 1, disarticulating above the glumes or infrequently primary branches disarticulating as units; lemma lacking uncinate macrohairs, if awned, the awn single; palea well developed; lodicules 2, membranous or rarely fleshy, heavily vascularized; stamens usually 3 or 6 (sometimes 1, 2, or 4); ovary glabrous, apical appendage absent, haustorial synergids absent, styles 2, free, fused basally or for their full length (Zizaniopsis), close, stigmas 2. Caryopsis with the hilum longlinear; endosperm hard, without lipid, containing compound starch grains (rarely simple); embryo small, epiblast usually present (absent in Ehrharta), scutellar cleft usually present (absent in Leersia and Potamophila), mesocotyl internode absent (present but short in Microlaena), embryonic leaf usually with overlapping margins (meeting in Potamophila). Basic chromosome numbers: x = 12 (10 in Microlaena; 15 in Zizania). Foliar anatomy. Mesophyll nonradiate, an adaxial palisade layer absent, fusoid cells large and well developed, arm cells usually well developed and strongly invaginated; Kranz anatomy absent; midrib complex or simple; adaxial bulliform cells present. Foliar micromorphology. Stomata with domeshaped, triangular, or parallel-sided subsidiary cells; bicellular microhairs present, panicoid-type; papillae common and abundant. Photosynthetic pathway. C3. INCIUDED TRI:LS: Bambuseae Dumort., Anal. Fam. Pi.: 63. 1829. TYPE: Bamuttsa Schreb. Olyreae Kunth ex Spenn., Fl. Friburg. 1: 172. 1825. TYPE: Olyra L. (Including Buergersiochloeae Blake, Blumea, Suppl. 3: 62. 1946; Parianeae C. E. Hubbard, in Hutch., Fam. Fl. P1. 2: 219. 1934.) Notes. The current circumscription of this subfamily is much narrower than the traditional view. In their recent analysis, Zhang and Clark (2000) recovered two robustly supported clades, the olyroid bamboos and the woody bamboos, which they recognized as tribes Olyreae and Bambuseae, respectively. Following Zhang and Clark (2000), Buergersiochloeae and Parianeae are included in Olyreae. This subfamily includes approximately 1200 species. V. Ehrhartoideae Link, Hort. Berol. 1: 233. 1827. TYPE: Ehrharta Thunb. Figure 4G. Syn.: Oryzoideae Kunth ex Beilschm., Flora 16(2): 52, 109. 1833. Plants annual or perennial (rhizomatous or stoloniferous), herbaceous to suffrutescent, of forests, open hillsides, or aquatic habitats. Culms hollow or solid. Leaves distichous; abaxial ligule absent; adaxial ligule a fringed or unfringed membrane, or a fringe of hairs; blades rarely basally cordate or sag- Foliar anatomy. Mesophyll nonradiate, an adaxial palisade layer usually absent, fusoid cells absent or sometimes present (Zizania and Zizaniopsis), arm cells absent or present; Kranz anatomy absent; midrib simple or complex; adaxial bulliform cells present. Foliar micromorphology. Stomata with domeshaped or triangular subsidiary cells; bicellular microhairs present, panicoid-type; papillae often present in Oryzeae, otherwise absent. Photosynthetic pathway. C3. INCLUDED TRIBES: Ehrharteae Nevski, Trudy Bot. Inst. Akad. Nauk SSSR 4: 227. 1937. TYPE: Ehrharta Thunb. Oryzeae Dumort., Observ. Gramin. Belg.: 83. 1824. TYPE: Oryza L. Phyllorachideae C. E. Hubb., in Hook. Ic. P1. 34: t. 3386, p. 5. 1939. TYPE: Phyllorachis Trimen. Notes. Although we did not sample Phyllorachideae, we place it here based on morphological similarity. Nonetheless, any future studies of this clade should include this tribe to test its relationship to Ehrharteae and Oryzeae. Under the present circumscription, this subfamily includes approximately 120 species. Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae VI. Pooideae Benth., Fl. Hongk. 407. 1861. TYPE: Poa L. Figures 4E, H, 6D, J, L, 7B. or circular and less than 1/3 the length of the fruit; endosperm hard or sometimes soft or liquid (some Poeae), with or without lipids (some Poeae), containing compound starch grains, or simple starch grains (Brachyelytreae, Bromeae, Triticeae, some Stipeae); embryo small, epiblast present (rarely absent), scutellar cleft absent (rarely present, but not deeply incised), mesocotyl internode absent (rarely short, Brachyelytrum), embryonic leaf margins meeting (infrequently margins overlapping). Basic chromosome numbers: x = 7 (Bromeae, Triticeae, Poeae generally, few Brachypodieae), 2, 4, 5, 6, 8, 9, 10, 11, 12, 13 represented in a few Poeae and the other tribes, generally medium or large. Syn.: Avenoideae Link, Hort. Berol. 1: 108. 1827. Festucoideae Link, Hort. Berol. 1: 137. 1827. Glycerioideae Link, Hort. Berol. 1: 160. 1827. Echinarioideae Link, Hort. Berol. 1: 197. 1827. CynosuroideaeLink, Hort. Berol. 1: 198. 1827. AnthoxanthoideaeLink, Hort. Berol. 1: 232, 271. 1827. Agrostidoideae Kunth ex Beilschm., Flora (Beib.) 16(2): 52, 104. 1833. Stipoideae Burmeist., Handb. Naturgesch. 199. 1837. Hordeoideae Burmeist., Handb. Naturgesch.202. 1837. Phalaroideae Burmeist., Handb. Naturgesch. 208. 1837. Secaloideae Rouy, Fl. France 14: 2, 298. 1913. Plants annual or perennial (rhizomatous, stoloniferous, or neither), herbaceous, of cool temperate and boreal regions, extending across the tropics in the high mountains. Culms hollow (rarely solid). Leaves distichous; abaxial ligule absent; adaxial ligule scarious or membranous, the margin not or infrequently short ciliate fringed (rarely long ciliate, Anisopogon); blades somewhat broad to usually narrow, rarely pseudopetiolate (Phaenosperma), venation parallel; sheaths sometimes auriculate. Inflorescences spicate, racemose, or paniculate, bracts outside of the spikelets absent or rarely present (e.g., Sesleria, Echinaria, Ammochloa). Spikelets bisexual, infrequently unisexual or mixed, usually with two glumes (rarely without glumes, Lygeum, or the first absent, Hainardia, Lolium, Nardus, except on terminal spikelets), 1 to many female-fertile florets with apical or infrequently basal reductions, compressed laterally, infrequently not or dorsally compressed, disarticulating above the glumes (infrequently below the glumes, some Poeae, or at the nodes of the inflorescence, various genera); lemma lacking uncinate macrohairs, if awned, the awn single; palea usually present and well developed, but variable and sometimes very reduced or absent; lodicules 2 (rarely 3, Anisopogon, Ampelodesmeae, many Stipeae and few Poeae; fused, Meliceae; rarely absent, Lygeum, Nardus, and few Poeae), usually lanceolate, broadly membranous apically (fleshy, truncate, Meliceae), often lobed (Triticeae, Poeae), obscurely few-nerved, or infrequently + distinctly few-nerved, not or conspicuously ciliate on the margins; stamens usually 3 (infrequently 1 or 2); ovary glabrous or pubescent, rarely with an apical appendage (Bromus, Diarrhena) or rostellum (e.g., Brachyelytrum, Rostraria), haustorial synergids absent, styles usually 2, close, stigmas 2 (rarely 1, Lygeum, Nardus, and a few others, or 3, scattered genera). Caryopsis with the hilum linear and up to as long as the fruit, or subbasal and punctiform, linear, ellipsoidal, ovate, 419 Foliar anatomy. Mesophyll nonradiate, an adaxial palisade layer absent, fusoid cells absent, arm cells absent; Kranz anatomy absent; midrib simple; adaxial bulliform cells present. Foliar micromorphology. Stomata with parallelsided subsidiary cells; bicellular microhairs absent (rarely present, Lygeum, where chloridoid, Nardus, where panicoid), unicellular microhairs absent (rarely present, few Stipeae); papillae usually absent, when present rarely more than one per long cell. Photosynthetic pathway. C3. INCI,LUDI)E! TRIEKS: Ampelodesmeae (Conert) Tutin, Bot. J. Linn. Soc. 76: 369. 1978. TYPE: Ampelodesmos Link. Brachyelytreae Ohwi, Bot. Mag. Tokyo 55: 361. 1941. TYPE: Brachyelytrum P. Beauv. Brachypodieae (Hack.) Hayek, Oesterr. Bot. Z. 74(10): 253. 1925. TYPE: Brachypodium P. Beauv. Bromeae Dumort., Observ. Gramin. Belg.: 83. 1824. TYPE: Bromus L. Brylkinieae Tateoka, Canad. J. Bot. 38: 962. 1960. TYPE: Brylkinia F. Schmidt. Diarrheneae (Ohwi) C. S. Campb., J. Arnold Arbor. 66: 188. 1985. TYPE: Diarrhena P. Beauv. Lygeeae J. Presl, Wsobecny Rostl. 2: 1708, 1753. 1846. TYPE: Lygeum Loefl. ex L. Meliceae Link ex Endl., Fl. Poson.: 116. 1830. [as "Melicaceae"] TYPE: Melica L. Nardeae W. D. J. Koch, Syn. Fl. Germ. Helv.: 830. 1837. TYPE: Nardus L. Renvoize & Clayton, Kew Phaenospermatideae Bull. 40: 478. 1985. TYPE: Phaenosperma Munro ex Benth. Poeae R. Br., Voy. Terra Austral. 2: 582. 1814. TYPE: Poa L. (Including Aveneae Dumort., 420 Annals of the Missouri Botanical Garden Observ. Gramin. Belg.: 82. 1824; Agrostideae Dumort., Observ. Gramin. Belg.: 83. 1824.) Stipeae Dumort., Observ. Gramin. Belg.: 83. 1824 [as "Stipaceae"]. TYPE: Stipa L. Triticeae Dumort., Observ. Gramin. Belg.: 82, 84, 91. 1824. TYPE: Triticum L. fusoid cells absent, arm cells absent; Kranz anatomy absent (Sartidia) or present (Stipagrostis, Aristida), when present with one (Stipagrostis) or two (Aristida) parenchyma sheaths, although both not equally well developed throughout the genus; midrib simple; adaxial bulliform cells present. Relationships among some of the major the core Pooideae clade remain unreof lineages solved, and conflicts between molecular data and morphologically based tribal classifications exist (e.g., Poeae vs. Aveneae; see Soreng & Davis, 2000). This is one of several reasons that we do not offer a formal classification of tribes at this point. Relationships among the earlier diverging lineages of the whole pooid clade are only weakly supported, and also require further investigation. The tribal classification presented here is almost certain to change as additional data accumulate, and thus should be taken only as an indication of the taxa included within the subfamily. The subfamily includes approximately 3300 species. Foliar micromorphology. Stomata dome-shaped or triangular; bicellular microhairs present, panicoid-type; papillae absent. Notes. VII. Aristidoideae Caro, Dominguezia 4: 16. 1982. TYPE: Aristida L. Figure 5C. Plants annual or perennial, caespitose, herbaceous, xerophytic or less commonly mesophytic, of temperate, subtropical and tropical zones, often in open habitats. Culms solid or hollow. Leaves distichous; abaxial ligule absent or present as a line of hairs; adaxial ligule a fringed membrane or a fringe of hairs; blades relatively narrow, without pseudopetioles, venation parallel; sheaths non-auriculate. Inflorescences paniculate, bracts outside of the spikelets absent. Spikelets with bisexual florets, glumes 2, female-fertile floret 1, and no rachilla extension, cylindrical or laterally compressed, disarticulating above the glumes; lemma with three awns, the awns separate from each other, or fused below into a twisted column; palea short, less than half the lemma length; lodicules present or rarely absent, when present 2, free, membranous, glabrous, heavily vascularized; stamens 1 to 3; ovary glabrous, apical appendage absent, haustorial synergids absent, styles 2, free, close, stigmas 2. Caryopsis with the hilum short or long-linear; endosperm hard, without lipid, containing compound starch grains; embryo small (Sartidia) or large (Aristida, Stipagrostis), epiblast absent, scutellar cleft present or absent (Sartidia), mesocotyl internode elongated, embryonic leaf margins meeting. Basic chromosome numbers: x = 11, 12. Foliar anatomy. Mesophyll radiate or nonradiate (Sartidia), an adaxial palisade layer absent, C3 (Sartidia); C4 (ArPhotosynthetic pathway. istida, NADP-ME; Stipagrostis, not biochemically typed, but anatomically NAD-ME; Hattersley & Watson, 1992). INCLUDED TRIBE (NOW IDENTICAL TO THE SUBFAMILY AND THUS REDUNDANT): Aristideae C. E. Hubbard, in Bor, Grasses Burma, Ceylon, India & Pakistan: 685. 1960. TYPE: Aristida L. Notes. The presence of a basal column of the awn is a potential morphological synapomorphy for this clade. Sartidia diverges from Stipagrostis and Aristida in other respects, and should be sampled in future analyses. The subfamily includes approximately 350 species. VIII. Arundinoideae Burmeist., Handb. Naturgesch.: 204. 1837. TYPE: Arundo L. Figure 5A. Syn.: PhragmitoideaeParodi ex Caro, Dominguezia4: 13. 1982. Plants perennial (rarely annual), rhizomatous, stoloniferous, or caespitose, herbaceous to somewhat woody, of temperate and tropical areas, mesophytic or xerophytic, the reeds found in marshy habitats. Culms hollow or less commonly solid. Leaves distichous; abaxial ligule absent or rarely present as a line of hairs (Hakonechloa); adaxial ligule a fringed or unfringed membrane or a fringe of hairs; blades relatively broad to narrow, without pseudopetioles, venation parallel; sheaths usually non-auriculate. Inflorescences usually paniculate, rarely spicate or racemose, bracts outside of the spikelets absent. Spikelets with bisexual florets, glumes 2, a sterile lemma sometimes present, female-fertile florets 1 to several, apical reduction usually present, usually laterally compressed, disarticulating above the glumes; lemma lacking uncinate macrohairs, if awned, awn usually single, sometimes awns three, but then lacking a basal column; palea usually well developed; lodicules 2, free (rarely joined at the base), fleshy, glabrous or 421 Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae infrequently ciliate, not or scarcely vascularized to heavily vascularized; stamens (1 to)3; ovary glabrous, apical appendage absent, haustorial synergids absent, styles 2, usually free, close, stigmas 2. Caryopsiswith the hilum short or long-linear (Molinia); endosperm hard, without lipid, containing compound starch grains; embryo large or small (Amphipogon),epiblast absent, scutellar cleft present, mesocotyl internode elongated, embryonicleaf margins meeting or overlapping (Hakonechloa).Basic chromosome numbers: x = 6, 9, 12. monophyletic arundinoid clade, although Linder et al. (1997) linked Arundo, Phragmites, and Molinia by the presence of hollow culm internodes, a punctiform hilum, and convex adaxial rib sides in the leaf blade. This subfamily clearly requires further study. The subfamily includes 33 to 38 species, counting the crinipoids. Foliar anatomy. Mesophyll nonradiateor rarely radiate (Arundo,Amphipogon),without an adaxial palisade layer, without fusoid cells, arm cells absent or present (Phragmites); Kranz anatomy absent; midrib simple; adaxial bulliformcells present. IX. Danthonioideae Barker & H. P. Linder, subfam. nov. TYPE: Danthonia DC. Fl. Franc. 3: 32. 1805. Figure 5F. Haec subfamiliaab aliis subfamiliisPoacearumsynergidis haustorialibus, ligula ciliata,embryonemesocotyledone praedito,spicula plurifloravel si uni- vel biflora nunc rhachillain extensionemdesinente,stylorumbasibus plerumquedistantibusatqueanatomia"Kranz" et micropilischloridoideiscarentibusbene distincta. Distinctfromthe othersubfamiliesof the grassesby the Foliar micromorphology. Stomata with low haustorialsynergids,and by the conjunctionof a ciliate dome-shaped or triangular subsidiary cells; bicel- ligule, the presenceof an embryomesocotyl,a severallular microhairs present or less commonly absent, floweredspikelet,which,if 1- or2-flowered,has a rachilla when present of panicoid-type except in Amphipo- extension,usuallydistinctlyseparatedstylebases,the absence of Kranzanatomy,and the absence of chloridoid gon, which has unique microhair morphology;pa- microhairs. pillae absent except in Amphipogon. Plants perennial (caespitose, rhizomatousor stoPhotosyntheticpathway. C,. loniferous) or less commonly annual, herbaceous or rarely suffrutescent, of mesic to xeric open habitats INCLUDEI) TRIBE (NOW II)FNTICA, T() SU3BFAMIL,Y in grasslands, heathlands, and open woodlands. AND THUS REDUNI)ANT): Culms solid or very rarely hollow. Leaves distiArundineae Dumort., Obs. Gram. Belg.: 82. 1824. chous; abaxial ligule usually absent (sometimes present in Cortaderia, Karroochloa, and PentasTYPE: Arundo L. chistis); adaxial ligule a fringe of hairs or a fringed Notes. The traditionalArundinoideaewere well membrane; blades relatively narrow, without a known as a dustbin group (e.g., Clayton & Renvo- pseudopetiole, venation parallel; sheaths not auricize, 1986; Kellogg & Campbell, 1987). A number ulate except in Pentameris thuarii. Inflorescences of studies indicated that this subfamily as tradi- paniculate or less commonly racemose or spicate, tionally circumscribed was polyphyletic (e.g., Bark- bracts outside of the spikelets absent (but the suber et al., 1995; Clark et al., 1995), although some tending leaf + spatheate and disarticulating with support for a monophyletic Arundinoideae (includ- the inflorescence in Triboliumpusillum). Spikelets ing Arundinoideae s. str., Danthonioideae, Aristi- bisexual (but sometimes without bisexual florets in doideae, Micraira, and Eriachne) was found by Cortaderia) or unisexual (Cortaderia, LamprothyrHsiao et al. (1999). The results of the combined sus), glumes 2 and usually equal, female-fertile floanalysis presented here suggest that a monophyletic rets 1 to 6(to 20), with apical reduction and a rachcore arundinoid group does exist, even though in- illa extension usually present, laterally compressed, dividual data sets do not stronglysupportthe group. disarticulating above the glumes and between the The exact generic membership of the subfamily re- florets, less commonly below the glumes; lemma mains to be determined; however, we include the lacking uncinate macrohairs, awn single and from following genera: Amphipogon, Arundo, Dregeo- a sinus; palea well developed, sometimes relatively chloa, Hakonechloa, Molinia (and Moliniopsis if short; lodicules 2, free (rarely joined), fleshy or recognized), and Phragmites. We provisionally rarely with an apical membranousflap, glabrous or place the crinipoid group (Crinipes, Dichaetaria, ciliate, often with microhairs, sometimes heavily Elytrophorus,Leptagrostis, Nematopoa, Piptophyl- vascularized; stamens 3; ovary glabrous or rarely lum, Styppeiochloa, and Zenkeria) here as well, with apical hairs (Pentameris), apical appendage based on molecular evidence from Linder et al. absent, haustorial synergids present, only weakly (1997) and Barker (1997). No morphological syn- developed in a few taxa, styles 2, the bases usually apomorphies have been identified to support the widely separated, stigmas 2. Caryopsis with the hi- 422 Annals of the Missouri Botanical Garden perate woodlands and tropical forests. Culms solid or hollow. Leaves distichous; abaxial ligule absent or present as a line of hairs (Calderonella,Thysanolaena); adaxial ligule membranous or ciliate, or membranouswith ciliate margins;blades relatively broad to narrow, often pseudopetiolate, venation parallel; sheaths sometimes auriculate. Inflorescences racemose or paniculate, bracts outside of Foliar anatomy. Mesophyll nonradiate, an adthe spikelets absent. Spikelets bisexual or unisexaxial palisade layer absent, fusoid cells absent, arm to with reduction either cells absent; Kranz anatomy absent; midrib simple, ual, (1 to)2- many-flowered above or below the fertile florets, often compressed usually with one bundle, an arc of bundles in Corlaterally; lemma lacking uncinate macrohairs, if taderia; adaxial bulliform cells present or absent. awned, the awn single; palea usually well develFoliar micromorphology. Stomata with dome- oped, sometimes relatively short;lodicules 2 or abshaped or parallel-sided subsidiary cells (rarely sent, + cuneate, many-nervedor less commonlynot high dome-shaped or slightly triangular);bicellular or scarcely vascularized; stamens (1 to)2 or 3; ovary microhairs present, panicoid-type, sometimes ab- glabrous, apical appendage absent, haustorial synsent; papillae usually absent but often present in ergids presumed absent, styles 2, free or fused, Chionochloa and Merxmuellera. close, stigmas 2. Caryopsis with the hilum basal, endosperm hard, withoutlipid, containpunctiform; Photosyntheticpathway. C3. ing simple or compound starch grains; embryo small or large, the epiblast present, scutellar cleft INCLUDED TRIBE (NOW IDENTICAL TO SUBFAMILY mesocotyl internode present, embryonic present, AND THUS REDUNDANT): leaf margins overlapping. Basic chromosome numDanthonieae Zotov, New Zealand J. Bot. 1 (1): 86. ber: x = 12 (x = 11 or 12? in Thysanolaena). 1963. (Including CortaderieaeZotov,New ZeaFoliar anatomy. Mesophyll nonradiate, often land J. Bot. 1 (1): 83. 1963.) TYPE: Danthonia with an adaxial palisade layer, fusoid-like cells freDC. quently present as extensions of the outer parenNotes. The presence of haustorial synergids in chyma bundle sheath, arm cells absent; Kranzanatthe ovule and distant styles support the monophyly omy absent; midrib simple; adaxial bulliform cells of this clade (Verboomet al., 1994). Bilobed pro- large. phylls also may be a synapomorphy,but this feature Foliar micromorphology. Stomata with domehas not been investigated sufficiently in the rest of the family. The results of this study indicate robust shaped and/or triangularsubsidiary cells; bicellular molecular support for the monophyly of this clade microhairs present, panicoid-type; papillae absent. (excluding Centropodiaand Merxmuellerarangei), Photosyntheticpathway. C3. but its placement within the larger PACCADClade is equivocal. Pending further studies of the diver- INCLUDED TRIBES: sity of the danthonioid grasses, we recognize only one tribe, which includes the following genera (sen- Centotheceae Ridl., Mat. Fl. Malay Pen. 3: 122. 1907. TYPE: CentothecaP. Beauv. su Barker et al., 2000): Austrodanthonia, Chaetobromus, Chionochloa, Cortaderia,Danthonia, Joy- Thysanolaeneae C. E. Hubb., in Hutch., Fam. Fl. P1. 2: 222. 1934. TYPE: Thysanolaena Nees. cea, Karroochloa, Lamprothyrsus, Merxmuellera (minus M. rangei), Notochloe, Notodanthonia, PenNotes. Support for the monophyly of this subtameris, Pentaschistis, Plinthanthesis, Prionanfamily as recognized here is moderate, and no morthium, Pseudopentameris,Rytidosperma,Schismus, have been identified. and Tribolium. The subfamily includes approxi- phological synapomorphies The sister relationship between the centothecoid mately 250 species. and panicoid clades, however, is relatively robust. The positions of Gyneriumand Danthoniopsis are X. Centothecoideae Soderstr. [as "Centostecoiunstable. A majority of the Centotheceae are chardeae"], Taxon 30: 615. 1981. TYPE: Cento- acterized by unusual leaf anatomy, including the theca Desv. Figure 5G. of presence palisade mesophyll and laterally exPlants annual or perennial (rhizomatous or sto- tended bundle sheath cells. Additional study of this loniferous), herbaceous or reedlike, of warm tem- clade is under way (J. G. Sanchez-Ken, pers. lum short or long-linear; endosperm hard, containing compound starch grains (simple in Prionanthium); embryo large or small, epiblast absent, scutellar cleft present, mesocotyl internode elongated, embryonic leaf marginsmeeting (overlapping in Danthonia decumbens).Basic chromosomenumbers: x = 6, 7, 9. Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae comm.). The subfamily includes approximately 45 species. lar or dome-shaped subsidiary cells; bicellular microhairs present, panicoid-type, rarely absent; papillae absent or present (mostly in the Andropogoneae). XI. Panicoideae Link, Hort. Berol. 1: 202. 1827. TYPE: Panicum L. Figures 5B, E, 6F, I, K. Syn.: Andropogonoideae Burmeist., Handb. Naturgesch.: 201. 1837. Rottboellioideae Burmeist., Handb. Naturgesch.: 202. 1837. Saccharoideae (Rchb.) Horan., Char. Ess. Fam.: 34. 1847. Plants annual or perennial (rhizomatous, stoloniferous, caespitose or decumbent), primarily herbaceous, of the tropics and subtropics, but also diverse in the temperate zone. Culms solid or less commonly hollow. Leaves distichous; abaxial ligule usually absent, occasionally present as a line of hairs; adaxial ligule a fringed or unfringed membrane, or a fringe of hairs, or sometimes absent; blades relatively broad to narrow, sometimes pseudopetiolate, venation parallel; sheaths usually nonauriculate. Inflorescences panicles, racemes, or spikes, or complex combinations of these, bracts outside of the spikelets present (Andropogoneae) or absent (Paniceae). Spikelets bisexual or unisexual (if the latter plants dioecious or monoecious), frequently paired in combinations with long and short pedicels, usually with glumes 2, sterile lemma 1, and female-fertile floret 1, dorsally compressed or less commonly not compressed or laterally compressed, disarticulating below the glumes (above the glumes in Arundinelleae) or the inflorescence axes breaking apart; lemma lacking uncinate macrohairs, if awned, the awn single; palea well developed (Paniceae) or reduced to absent (Andro2 or sometimes absent, pogoneae); lodicules cuneate, free, fleshy, usually glabrous; stamens 3; ovary usually glabrous, apical appendage absent, haustorial synergids absent, styles 2, free or fused, close, stigmas 2 (rarely 1 or 3). Caryopsis with the hilum usually short; endosperm hard, without lipid, containing simple or less commonly compound starch grains; embryo usually large, epiblast absent or rarely present, scutellar cleft present, mesocotyl internode elongated, embryonic leaf margins overlapping or rarely meeting. Basic chromosome numbers: x = 5, (7), 9, 10, (12), (14). Foliar anatomy. Mesophyll radiate or nonraan adaxial diate, palisade layer absent, fusoid cells absent except in Homolepis and Streptostachys, arm cells usually absent; Kranz anatomy present or absent; midrib simple or rarely complex; adaxial bulliform cells present. Foliar micromorphology. Stomata with triangu- 423 C3, C4 (PCK, NAD-ME Photosynthetic pathway. and NADP-ME), and some C3/C4 intermediates. INCLUDED TRIBES: AndropogoneaeDumort.[as "Andropogineae"],Observ. Gramin. Belg.: 84. 1824. TYPE: Andropogon L. Arundinelleae Stapf, Fl. Cap. 7: 314. 1898. TYPE: Arundinella Raddi. Hubbardieae C. E. Hubb., in Bor, Grasses India Burma Ceylon Pakistan: 685. 1960. TYPE: Hubbardia Bor. Isachneae Benth., J. Linn. Soc. Bot. 19: 30. 1881. TYPE: Isachne R. Br. Paniceae R. Br., Voy. Terra Austr. 2: 582. 1814. TYPE: Panicum L. Steyermarkochloeae Davidse & R. P. Ellis, Ann. Missouri Bot. Gard. 71: 994. 1984. TYPE: SteyermarkochloaDavidse & R. P. Ellis. Notes. While supportfor the panicoid/centothecoid clade is high, relationships within the clade remain unclear. No robust phylogeny for the Panicoideae is yet available, although work is in progress (Giussani et al., in press; Duvall et al., in press). Preliminary results indicate that the Paniceae as currently circumscribed may not be monophyletic, and that the large genus Panicum is polyphyletic (Zuloaga et al., 2000; G6mez-Martinez& Culham, 2000). Andropogoneae + Arundinella appear to be monophyletic (Spangler et al., 1999); other genera of the Arundinelleae are likely to be distributed among the Andropogoneae, Paniceae, and perhaps even the Centothecoideae (Kellogg, 2000b). This subfamily includes approximately 3270 species. XII. Chloridoideae Kunth ex Beilschm., Flora 16(2): 52, 105. 1833. TYPE: Chloris Sw. Figures 5D, H, 6E, M, N, 7C. Burmeist.,Handb. Naturgesch. Syn.: Pappophoroideae 205. 1837. Eragrostoideae Pilger,Nat. Pfl.-Fam.ed. 2, 14d: 167. 1956. Plants annual or perennial (rhizomatous,stoloniferous, caespitose or decumbent), herbaceous (rarely woody), of dry climates, especially in the tropics and subtropics, also found in the temperate zone. Culms solid or hollow. Leaves distichous; abaxial ligule usually absent, rarely present as a line of hairs; adaxial ligule a fringed or less commonly 424 Annals of the Missouri Botanical Garden unfringed membrane; blades relatively narrow, without pseudopetioles, venation parallel; sheaths usually non-auriculate. Inflorescences paniculate, paniculate with spicate branches, racemose, or spicate, bracts outside of the spikelets absent. Spikelets bisexual or sometimes unisexual (if so the plants dioecious or monoecious), with glumes 2, rarely a sterile lemma, and female-fertile florets 1 to many, apical reduction usually present, usually laterally compressed, sometimes dorsally compressed, usually disarticulating above the glumes (below in a few Eragrostis species); lemma lacking uncinate macrohairs, if awned, the awns single or if multiple, lacking a basal column; palea well developed; lodicules 2 or absent, fleshy, glabrous;stamens 1 to 3; ovary glabrous, apical appendage absent, haustorial synergids absent, styles 2, free, close, stigmas 2. Caryopsis with the pericarp often free or loose; hilum short; endosperm hard, without lipid, containing simple or compound starch grains; embryo large or rarely small, epiblast present or rarely absent, scutellar cleft present, mesocotyl internode elongated, embryonic leaf margins meeting or rarely overlapping. Basic chromosomenumbers: x =(7), (8), 9, 10. INCERTAE SEDIS: Foliar anatomy. Mesophyll usually radiate, without an adaxial palisade layer, fusoid cells absent, arm cells absent; Kranz anatomy present; midrib simple; adaxial bulliform cells present. CentropodiaReichenb., Merxmuellerarangei (Pilg.) Conert Notes. Reduction in the number of veins in the lemma is a general trend within the subfamily but is clearly not a synapomorphy.Except for the C3 Eragrostis walteri and Merxmuellera rangei, the Chloridoideaeare uniformlyC4with both the NADME and PCK subtypes. The current tribal classification for this subfamily conflicts with molecular data and is likely to be modified (Hilu et al., 1999). This subfamily includes approximately 1400 species. XIII. Incertae Sedis Eriachneae (Ohwi) Eck-Borsb., Blumea 26: 128. 1980. Micraireae Pilger, Nat. Pfl.-Fam. Ed. 2, 14d: 167. 1956. Streptogyneae C. Calder6n & Soderstr., Smithsonian Contr. Bot. 44: 18. 1980. Cyperochloa Lazarides & L. Watson, Brunonia 9: 216. 1987. GyneriumWilld. ex P. Beauv., Ess. Agrostogr.138, 153, t. 24. 1812. Notes. These five taxa are left Incertae Sedis because the data presented here do not firmly support their inclusion in any of the 12 subfamilies. This approach has also been taken by the APG for taxa of uncertain placement. Some posFoliar micromorphology. Stomata with dome- (1998) sible placements of the five taxa above will require shaped or triangularsubsidiary cells; bicellular miof new names, and we feel stronglythat crohairs present, usually chloridoid-type; papillae publication nomenclatural changes should not be made until absent or present. appreciable data support the conclusion. That said, recent unpublished data (J. G. Sanchez-Ken, pers. Photosyntheticpathway. C3 (Eragrostiswalteri, that Gyneriumcan be placed as its Merxmuellera rangei), otherwise C4 (PCK, NAD- comm.) suggest own tribe in Panicoideae, and the tribal name may ME, but reported as NADP-ME in Pappophorum, be available by the time this paper is published by Hattersley & Watson, 1992; the latter may be (Sanchez-Ken & Clark, 2001). It is likely that Cyan error). perochloa will be placed in Centothecoideae, but this is based on its morphological similarities to INCLUDED TRIBES: Spartochloa and not on any data on Cyperochloa itself. Streptogyneae will probably fall within EhrCynodonteae Dumort., Observ. Gramin. Belg.: 83. hartoideae, but limitations of our data and lack of 1824. TYPE: Cynodon Rich. support in our trees make us cautious about placing Eragrostideae Stapf, Fl. Cap. 7: 316. 1898. TYPE: it there unequivocally; there may be an argument for recognition of the tribe as its own subfamily. Eragrostis Wolf. Leptureae Dumort., Observ. Gramin. Belg.: 83. The name Micrairoideae has been published (Pil1824. TYPE: LepturusR. Br. ger, 1956). Our data are too limited and the placeOrcuttieae Reeder, Madrofio18: 20. 1965. TYPE: ment of the group too uncertain to add it as a thirOrcuttia Vasey. teenth subfamily, although flora writers may choose Pappophoreae Kunth, Rev. Gramin. 1: 82. 1829. to do so. Our data on Eriachne are weak, and show TYPE: PappophorumSchreb. only that the genus does not fall within the Pani- Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae coideae, where it has been placed traditionally. Its placement near the base of the PACCAD Clade is based on a single-stranded rbcL sequence from one species, and an ITS sequence from a second. The two species represent two sections of the genus, one of which has actually been recognized as its own genus. We therefore feel that Incertae Sedis best reflects what we know of the position of the tribeits position is uncertain. a one-way ticket to genomic obesity? P1. Cell 9: 15091514. Bentham, G. 1878. Flora Australiensis 7: 449-670. & J. D. Hooker. 1883. Gramineae. Pp. 10741215 in Genera Plantarum,vol. 3, pt. 2. L. Reeve, London. Borre, A. Van den. 1994. Taxonomyof the Chloridoideae (Poaceae), with Special Reference to the Genus Eragrostis. Unpublished Ph.D. Dissertation, Australian National University, Canberra. & L. Watson.1994. The infragenericclassification of Eragrostis (Poaceae). Taxon 43: 383-422. Bossinger, G. 1990. Klassifizierung von Entwicklungsmutanten der Gerste anhand einer Interpretation des Pflanzenaufbausder Poaceae aus Phytomeren. Inaugural-Dissertation, Rheinischen Friedrich-WilhelmsUniversitat zu Bonn. Bremer, K. 1988. The limits of amino acid sequence data in angiosperm phylogenetic reconstruction. Evolution 42: 795-803. 1990. Combinablecomponent consensus. Cladistics 6: 369-372. Briggs, B. G., A. D. Marchant,S. Gilmore & C. L. Porter. 2000. A molecular phylogeny of Restionaceae and allies. Pp. 661-671 in K. L. Wilson & D. A. Morrison (editors), Monocots: Systematics and Evolution. CSIRO Press, Sydney. Brown, R. 1810. Prodromus florae Novae Hollandiae et insulae Van-Diemen, vol. 1. J. Johnson, London. 1814. General remarks, geographical and systematical, on the botany of Terra Australis, Appendix 3. Pp. 580-583 in M. Flinders (editor), A Voyageto Terra Australis; Undertaken for the Purpose of Completing the Discovery of that Vast Country,and Prosecuted in the Years 1801, 1802, and 1802. W. Bulmer, l,ondon. Brown, W. V. 1958. Leaf anatomy in grass systematics. Bot. Gaz. (Crawfordsville)119: 170-178. 1977. The Kranz syndrome and its subtypes in grass systematics. Mem. 'rorreyBot. Club 23: 1-91. Calderon, C. E. & T. R. Soderstrom. 1980. The genera of Bambusoideae (Poaceae) of the American continent: Key and comments. Smithsonian Contr. Bot. 44: 1-27. Campbell, C. S. 1985. The subfamilies and tribes of Gramineae (Poaceae) in the southeastern United States. J. Arnold Arbor. 66: 123-199. & E. A. Kellogg. 1987. Sister group relationships of the Poaceae. Pp. 217-224 in T. R. Soderstrom, K. W. Hilu, C. S. Campbell & M. E. Barkworth(editors), Grass Systematics and Evolution. Smithsonian Institution Press, Washington,D.C. , P. E. Garwood& L. P. Specht. 1986. Bambusoid affinities of the North American temperate genus Brachyelytrum (Gramineae). Bull. Torrey Bot. Club 113: 135-141. Caro, J. A. 1982. Sinopsis taxonomica de las gramineas argentinas. Dominguezia 4: 1-51. Carolin, R. C., S. W. L. Jacobs & M. Vesk. 1973. The structure of the cells of the mesophyll and parenchymatous bundle sheath of the Gramineae. Bot. J. Linn. Soc, London 66: 269-273. Celakovsky, L. 1889. Uber den Archenbau der Brasilianischen GrasgattungStreptochaetaSchrader.Sitzungsber. Konigl. Bohm. Ges. Wiss. Prag, Math.-Naturwiss.C1.3: 14-42, table 2. [Unpublished English translation by Mrs. G. Saad, 1974, for the Smithsonian Institution and the National Science Foundation, Washington,D.C.; in LiteratureCited Ambrose, B. A., D. R. Lerner,P. Ciceri, C. M. Padilla, M. F. Yanofsky& R. J. Schmidt. 2000. Molecular and genetic analyses of the silky] gene reveal conservation in floral organ specification between eudicots and monocots. Molec. Cell 5: 569-579. APG (Angiosperm Phylogeny Group). 1998. An ordinal classification for the families of flowering plants. Ann. Missouri Bot. Gard. 85: 531-553. Avdulov, N. P. 1931. Kario-sistematicheskoyeissledovaniye semeystva zlakov. Trudy Prikl. Bot. Prilozheniye 44: 1-352. [Karyosystematicstudies in the grass family. Supplement 44 to The Bull. Appl. Bot. Genet. P1.Breed., Leningrad. Russian text, pp. 1-352; German summary, pp. 353-425; index pp. 426-428. English translation(mislabeled as supplement 43) published by the Smithsonian Institution and the National Scientific Documentation Centre, New Delhi. 1975. TT 70- 53085.] Barker, N. P. 1995. A Molecular Phylogeny of the Suhfamily Arundinoideae. Ph.D. Thesis, University of Cape Town. 1997. The relationships of Amphipogon,Elytrophorus, and Cyperochloa(lPoaceae)as suggested by rbc, sequence data. Telopea 7: 205-213. , H. P. Linder & E. H. Harley. 1995. Polyphyly of Arundinoideae(Poaceae): Evidence fromrbcl, sequence data. Syst. Bot. 20: 423-435. & . 1999. Sequences of the grass, specific insert in the chloroplast rpoC2 gene elucidate generic relationships of the Arundinoideae (Poaceae). Syst. Bot. 23: 327-350. , C. M. Morton & H. P. Linder. 2000. The Danthonieae: Generic composition and relationships. Pp. 221-229 in S. W. L. Jacobs & J. E. Everett (editors), Grasses: Systematics and Evolution. CSIROPublishing, Collingwood,Victoria. Barkworth,M. E. 1993. North American Stipeae: Taxonomic notes and other comments. Phytologia 74: 1-25. & J. Everett. 1987. Evolution in the Stipeae: Identificationand relationships of its monophyletictaxa. Pp. 251-264 in T. R. Soderstrom, K. W. Hilu, C. S. Campbell & M. E. Barkworth(editors), Grass Systematics and Evolution. Smithsonian Institution Press, Washington,D.C. Baum, D. A., K. J. Sytsma & P. C. Hoch. 1994. A phylogenetic analysis of Epilobium (Onagraceae)based on nuclear ribosomal DNA sequences. Syst. Bot. 19: 363388. ,R. L. Small & J. F. Wendel. 1998. Biogeography and floral evolution of baobabs (Adansonia, Bombacaceae) as inferred from multiple data sets. Syst. Biol. 47: 181-207. Bennetzen, J. L. & E. A. Kellogg. 1997. Do plants have 425 426 Annals of the Missouri Botanical Garden the files of the McClure Bamboo Library, Department of Botany, Smithsonian Institution, Washington,D.C.] Chase, M. W., M. F. Fay & V. Savolainen. 2000a. Higherlevel classification in the angiosperms: New insights from the perspective of DNA sequence data. Taxon49: 685-704. , D. E. Soltis, P. S. Soltis, P. J. Rudall, M. F. Fay, W. J. Hahn, S. Sullivan, J. Joseph, M. Molvray,P. J. Kores, T. J. Givnish, K. J. Sytsma & J. C. Pires. 2000b. Higher-level systematics of the monocotyledons:An assessment of currentknowledge and a new classification. Pp. 3-16 in K. L. Wilson & D. A. Morrison (editors), Monocots: Systematics and Evolution. CSIRO Press, Sydney. , R. G. Olmstead, D. Morgan,D. H. Les, B. D. Mishler, M. R. Duvall, R. A. Price, H. G. Hills, Y.-L. Qiu, K. A. Kron, J. H. Rettig, E. Conti, J. D. Palmer, J. R. Manhart,K. J. Sytsma, H. J. Michaels, W. J. Kress, K. G. Karol, W. D. Clark, M. Hedren, B. S. Gaut, R. K. Jansen, K.-J. Kim, C. F. Wimpee, J. F. Smith, G. R. Furnier,S. H. Strauss, Q.-Y. Xiang, G. M. Plunkett, P. S. Soltis, S. M. Swensen, S. E. Williams, P. A. Gadek, C. J. Quinn, L. E. Eguiarte, E. Golenberg, G. H. Learn, Jr., S. W. Graham, S. C. H. Barrett, S. Dayanandan & V. A. Albert. 1993. Phylogenetics of seed plants: An analysis of nucleotide sequences from the plastid gene rbcL. Ann. Missouri Bot. Gard. 80: 528-580. Chatterton, N. J., C. Hsiao, K. H. Asay, K. B. Jensen & R. R. Wang. 1992. Nucleotide sequence of the internal transcribedspacer region rDNA in barley,Hordeumvulgare L. (Gramineae).P1. Molec. Biol. 20: 165-166. Chen, Z., S. Muthukrishnan,G. H. Liang, K. F. Schertz & G. E. Hart. 1993. A chloroplast DNA deletion located in RNA polymerase gene rpoC2 in CMS lines of sorghum. Molec. Gen. Genet. 236: 251-259. Clark, L. G. 1991. The function of fusoid cells in bamboo: An hypothesis. Amer.J. Bot. 78 (suppl.): 22. [Abstract.] & E. J. Judziewicz. 1996. The grass subfamilies Anomochlooideae and Pharoideae (Poaceae). Taxon45: 641-645. W. Zhang & J. F. Wendel. 1995. A phylogeny of the grass family (Poaceae) based on ndhF sequence data. Syst. Bot. 20: 436-460. , M. Kobayashi, S. Mathews, R. E. Spangler & E. A. Kellogg. 2000. The Puelioideae, a new subfamily of Poaceae. Syst. Bot. 25: 181-187. Clayton, W. D. 1965. Introductionto Arber's "The Gramineae." Pp. i-xxxii in The Gramineae by A. Arber, 1965 reprint. J. Cramer,New York. .1990. The spikelet. Pp. 32-51 in G. P. Chapman (editor), Reproductive Versatility in the Grasses. Cambridge Univ. Press, Cambridge,U.K. & S. A. Renvoize. 1986. Genera Graminum, Grasses of the World. Her Majesty's Stationery Office, London. Clifford, H. T. 1987. Spikelet and floral morphology.Pp. 21-30 in T. R. Soderstrom,K. W. Hilu, C. S. Campbell & M. E. Barkworth (editors), Grass Systematics and Evolution. Smithsonian Institution Press, Washington, D.C. Crepet, W. L. & G. D. Feldman. 1991. The earliest remains of grasses in the fossil record. Amer. J. Bot. 78: 1010-1014. Cronquist,A. 1981. An IntegratedSystem of Classification of Flowering Plants. Columbia Univ. Press, New York. Cummings, M. P., L. M. King & E. A. Kellogg. 1994. Slipped-strand mispairing in a plastid gene: rpoC2 in grasses (Poaceae). Molec. Biol. Evol. 11: 1-8. Dahlgren, R. M. T., H. T. Clifford& P. F. Yeo. 1985. The Families of the Monocotyledons. Springer-Verlag,Berlin, Heidelberg, New York. Davis, J. I. & R. J. Soreng. 1993. Phylogenetic structure in the grass family (Poaceae) as inferred from chloroplast DNA restriction site variation. Amer. J. Bot. 80: 1444-1454. Dehesh, K., J. Tepperman, A. H. Christensen & P. H. Quail. 1991. phyB is evolutionarilyconserved and constitutively expressed in rice seedling shoots. Molec. Gen. Genet. 225: 305-313. Devos, K. M., M. D. Atkinson, C. N. Chinoy, H. A. Francis, R. L. Harcourt, R. M. D. Koebner, C. J. Liu, P. Masoj'c, D. X. Xie & M. D. Gale. 1993. Chromosomal rearrangementsin the rye genome relative to that of wheat. Theor. Appl. Genet. 85: 673-680. Doebley, J., M. Durbin, E. M. Golenberg, M. T. Clegg & D. P. Ma. 1990. Evolutionaryanalysis of the large subunit of carboxylase (rbcL) nucleotide sequence data among the grasses (Poaceae). Evolution44: 1097-1108. Donoghue, M. J., R. G. Olmstead, J. F. Smith & J. D. Palmer. 1992. Phylogenetic relationships of Dipsacales based on rbcLsequences. Ann. Missouri Bot. Gard. 79: 249-265. Doyle, J. J., J. I. Davis, R. J. Soreng, D. Garvin & M. J. Anderson. 1992. Chloroplast DNA inversions and the origin of the grass family (Poaceae). Proc. Natl. Acad. Sci. U.S.A. 89: 7722-7726. Duval-Jouve, M. J. 1875. Histotaxie des feuilles des Graminees. Ann. Sci. Nat. Bot. 8: 227-346. Duvall, M. R. & B. R. Morton. 1996. Molecular phylogenetics of Poaceae: An expanded analysis of rbcL sequence data. Molec. Phylogenet. Evol. 5: 353-358. , M. T. Clegg, M. W. Chase, W. D. Clark, W. J. Kress, H. G. Hills, L. E. Eguiarte, J. F. Smith, B. S. Gaut, E. A. Zimmer & G. H. Learn, Jr. 1993. Phylogenetic hypotheses for the monocotyledonsconstructed from rbcL sequence data. Ann. Missouri Bot. Gard. 80: 607-619. , J. D. Noll & A. H. Minn. In press. Phylogenetics of Paniceae (Poaceae). Amer. J. Bot. Ellis, R. P. 1981. Leaf anatomyof the South African Danthonieae (Poaceae). V. Merxmuelleramacowanii, M. davyi and M. aureocephala. Bothalia 13: 493-500. Engler, A. 1892. Syllabus der Vorlesungeniiber specielle und Medicinisch-pharmaceutisch Botanik. Gebrtider Borntrager,Berlin. Esen, A. & K. W. Hilu. 1989. Immunological affinities among subfamilies of the Poaceae. Amer. J. Bot. 76: 196-203. Farris, J. S., M. Kallersjo, A. G. Kluge & C. Bult. 1994. Testing significance of incongruence. Cladistics 10: 315-319. Felsenstein, J. 1978. Cases in which parsimonyand compatibility methods will be positively misleading. Syst. Zool. 27: 401-410. Fisher, J. B. & J. Dransfield. 1977. Comparativemorphology and development of inflorescence adnation in rattan palms. Bot. J. Linn. Soc. 75: 119-140. Frazier, P. J., P. Richmond & A. M. Donald (editors). 1997. Starch: Structure and Functionality.Royal Society of Chemistry,InformationServices, Cambridge,U.K. Gale, M. D. & K. M. Devos. 1998. Comparativegenetics in the grasses. Proc. Natl. Acad. Sci. U.S.A. 95: 19711974. Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae Gaut, B. S. & J. F. Doebley. 1997. DNA sequence evidence for the segmental allotetraploid origin of maize. Proc. Natl. Acad. Sci. U.S.A. 94: 6809-6814. ,S. V. Muse, W. Clark & M. T. Clegg. 1992. Relative rates of nucleotide substitution at the rbcL locus in monocotyledonous plants. J. Molec. Evol. 35: 292303. , B. R. Morton, B. McCaig & M. T. Clegg. 1996. Substitution rate comparisons between grasses and palms: Synonymousrate differences at the nuclear gene Adh parallel rate differences at the plastid gene rbcL. Proc. Natl. Acad. Sci. U.S.A. 93: 10274-10279. , L. G. Clark, J. F. Wendel & S. V. Muse. 1997. Comparisons of the molecular evolutionary process at rbcL and ndhF in the grass family (Poaceae). Molec. Biol. Evol. 14: 769-777. Ghopal, B. H. & H. Y. Mohan Ram. 1985. Systematic significance of matureembryoof bamboo.P1. Syst. Evol. 148: 239-246. Giussani, L. M., J. H. Cota-Sanchez, F. 0. Zuloaga & E. A. Kellogg. In press. A molecular phylogeny of the grass subfamily Panicoideae (Poaceae) shows multiple origins of C4 photosynthesis. Amer. J. Bot. Goebel, K. 1895. Ein Beitrag zur Morphologieder Graser. Flora 81: 17-19. Goloboff,P. 1993. Nona, version 1.16. Buenos Aires: Distributed by the author. Gomez-Martinez,R. & A. Culham.2000. Phylogeny of the subfamily Panicoideae with emphasis on the tribe Paniceae: Evidence from the chloroplasttrnL-F cpDNA region. Pp. 136-140 in S. W. I. Jacobs & J. E. Everett (editors), Grasses: Systematics and Evolution. CSIRO Publishing, Collingwood, Victoria. Gould, F. W. & R. B. Shaw. 1983. Grass Systematics, 2nd ed. Texas A&M Univ. Press, College Station. GPWG (Grass Phylogeny WorkingGroup). 2000. A phylogeny of the grass family (Poaceae), as inferred from eight character sets. Pp. 3-7 in S. W . . Jacobs & J. E. Everett (editors), Grasses: Systematics and Evolution. CSIRO Publishing, Collingwood, Victoria. Grassl, C. 0. 1956. The morphologyof the grass spikelet with special reference to Saccharum. Proc. Int. Soc. Sugar Technologists 1: 764-780. Graybeal, A. 1998. Is it better to add taxa or characters to a difficult phylogenetic problem? Syst. Biol. 47: 917. Greuter, W., J. McNeill, F. R. Barrie, H. M. Burdet, V. Demoulin, T. S. Filgueiras, D. H. Nicolson, P. C. Silva, J. E. Skog, P. Trehane, N. J. Turland & D. L. Hawksworth (editors). 2000. InternationalCode of Botanical Nomenclature (Saint Louis Code). Regnum Veg. 138. Gu6eds, M. & P. Dupuy. 1976. Comparativemorphology of lodicules in grasses. Bot. J. Linn. Soc. 73: 317-331. Hackel, E. 1887. Gramineae.Pp. 1-97 in A. Engler & K. Prantl (editors), Die natiirlichen Pflanzenfamilien.Engelmann, Leipzig. Hamby, R. K. & E. A. Zimmer. 1988. Ribosomal RNA sequences for inferring phylogeny within the grass family (Poaceae). P1. Syst. Evol. 160: 29-37. Hattersley, P. W. & L. Watson. 1992. Diversification of photosynthesis. Pp. 38-116 in G. P. Chapman (editor), Grass Domestication and Evolution. Cambridge Univ. Press, Cambridge,U.K. Hillis, D. M. 1996. Inferringcomplex phylogenies. Nature 383: 130. . 1998. Taxonomic sampling, phylogenetic accuracy, and investigator bias. Syst. Biol. 47: 3-8. Hilu, K. W. & A. Esen. 1988. Prolamin size diversity in the Poaceae. Biochem. Syst. Ecol. 16: 457-465. & J. L. Johnson. 1991. Chloroplast DNA reassociation and grass phylogeny. P1. Syst. Evol. 176: 2133. & K. Wright. 1982. Systematics of Gramineae:A cluster analysis study. Taxon 31: 9-36. ,L. A. Alice & H. Liang. 1999. Phylogeny of Poaceae inferred from matK sequences. Ann. Missouri Bot. Gard. 86: 835-851. Hiratsuka, J., H. Shimada, R. Whittier, T. Ishibashi, M. Sakamoto, M. Mori, C. Kondo, Y. Honji, C. R. Sun, B. Y. Meng, Y. Q. Li, A. Kanno, Y. Nishizawa, A. Hirai, K. Shinozaki & M. Sugiura. 1989. The complete sequence of the rice (Oryza sativa) chloroplast genome: Intermolecular recombination between distinct tRNA genes accounts for a major plastid DNA inversion during the evolution of the cereals. Molec. Gen. Genet. 217: 185-194. Hitchcock, A. 1935. Manual of Grasses of the United States. U.S.D.A. Misc. Publ. 200. & A. Chase. 1950. Manual of Grasses of the United States, 2nd ed. U.S.D.A. Misc. Publ. 200. Hsiao, C., S. W. L. Jacobs, N. P. Barker& N. J. Chatterton. 1998. A molecular phylogeny of the subfamily Arundinoideae (Poaceae) based on sequences of rDNA. Austral. Syst. Bot. 11: 41-52. , N. J. Chatterton& K. H. Asay. 1999. A molecular phylogeny of the grass family (Poaceae) based on the sequences of nuclear ribosomal DNA (ITS). Austral. Syst. Bot. 11: 667-688. Igloi, G. L., A. Meinke, I. Dory & H. Kossel. 1990. Nucleotide sequence of the maize chloroplast rpo B/C1/C2 operon: Comparison between the (lerived protein primary structures fiom various organisms with respect to functional domains. Molec. Gen. Genet. 221: 379-394. Irish, E. E. 1998. Grass spikelets: A thorny problem. BioEssays 20: 789-793. Jacobs, B. F., J. D. Kingston & L. I. Jacobs. 1999. The origin of grass-dominated ecosystems. Ann. Missouri Bot. Gard. 86: 590-643. Jacobs, S. W. L. 1987. Systematics of the chloridoid grasses. Pp. 277-286 in T. R. Soderstrom,K. W. Hilu, C. S. Campbell & M. E. Barkworth(editors), Grass Systematics and Evolution. Smithsonian Institution Press, Washington,D.C. & J. Everett. 1997. Jarava plumosa (Gramineae), a new combination for the species formerly known as Stipa papposa. Telopea 7: 301-302. , M. E. Barkworth& C. Hsiao. 2000. Relationships within the stipoid grasses (Gramineae).Pp. 75-82 in S. W. L. Jacobs & J. E. Everett (editors), Grasses: Systematics and Evolution. CSIRO, Melbourne. Jacques-F6lix, H. 1962. Les graminees (Poaceae) d'Afrique tropicale, 1: G6n6ralit6s, classification, description de genres. Bull. Agron. 246: 1-4. Jirasek, V. 1968. Morphologic der Schuppchen (Lodiculae) von Grasern und ihre Terminologie, ein weiterer Beitrag zur Kenntnis des Baues der Lodiculae. Acta Univ. Carol., Biol. 1968: 321-344. & M. Jozifova. 1968. Morphology of lodicules, their variability and importance in the taxonomy of the Poaceae family. Bol. Soc. Argent. Bot. 12: 324-349. Johnston, C. R. & L. Watson. 1976. Microhairs: A universal characteristic of non-festucoid grass genera? Phytomorphology26: 297-301. 427 428 Annals of the Missouri Botanical Garden Judziewicz, E. J. 1987. Taxonomyand Morphologyof the Tribe Phareae (Poaceae: Bambusoideae). Unpublished Ph.D. Dissertation, University of Wisconsin, Madison. & T. R. Soderstrom. 1989. Morphological, anatomical, and taxonomic studies in Anomochloa and Streptochaeta (Poaceae: Bambusoideae). Smithsonian Contr. Bot. 68. , L. G. Clark, X. Londono & M. Stern. 1999. American Bamboos. Smithsonian Institution Press, Washington,D.C. Kallersjo, M., S. J. Farris, A. G. Kluge & C. Bult. 1992. Skewness and permutation.Cladistics 8: 275-287. Kanai, R. & G. E. Edwards. 1999. The biochemistry of C4 photosynthesis. Pp. 49-87 in R. F. Sage & R. K. Monson (editors), C4 Plant Biology. Academic Press, San Diego. Katayama,H. & Y. Ogihara. 1996. Phylogenetic affinities of the grasses to other monocots as revealed by molecular analysis of chloroplastDNA. Curr.Genet. 29: 572581. Kellogg, E. A. 1998. Relationships of cereal crops and other grasses. Proc. Natl. Acad. Sci. U.S.A. 95: 20052010. 2000a. The grasses: A case study in macroevolution. Ann. Rev. Ecol. Syst. 31: 217-238. 2000b. Molecularand morphologicalevolution in Andropogoneae. Pp. 149-158 in S. W. L. Jacobs & J. E. Everett (editors),Grasses: Systematics and Evolution. CSIRO, Melbourne. & C. S. Campbell. 1987. Phylogenetic analyses of the Gramineae. Pp. 310-322 in T. R. Soderstrom, K. W. Hilu, C. S. Campbell & M. E. Barkworth(editors), Grass Systematics and Evolution. Smithsonian Institution Press, Washington,D.C. & H. P. Linder. 1995. Phylogeny of Poales. Pp. 511-542 in P. J. Rudall, P. J. Cribb, D. F. Cutler & C. J. Humphries (editors), Monocotyledons: Systematics and Evolution. Royal Botanic Gardens, Kew, England. Kennard, W, R. Phillips, R. Porter & A. Grombacher. 1999. A comparativemap of wild rice (Zizaniapalustris L. 2n = 2x = 30). Theor. Appl. Genet. 99: 793-799. Kingston, J. D., B. D. Marino & A. Hill. 1994. Isotopic evidence for Neogene hominid paleoenvironments in the Kenya Rift Valley. Science 264: 955-959. Klijsgen, R. B., A. Gierl, Z. S. Schwarz-Sommer& H. Saedler. 1986. Molecular analysis of the waxy locus of Zea mays. Molec. Gen. Genet. 203: 237-244. Latorre, C., J. Quade & W C. McIntosh. 1997. The expansion of C4 grasses and global change in the late Miocene: Stable isotope evidence from the Americas. Earth Planet. Sci. Lett. 146: 83-96. Lazarides, M. 1995. The genus Eriachne (Eriachneae,Poaceae). Austral. Syst. Bot. 8: 355-452. LeRoux, L. G. & E. A. Kellogg. 1999. Floral development and the formation of unisexual spikelets in the Andropogoneae (Poaceae). Amer. J. Bot. 86: 354-366. Liang, H. & K. W. Hilu. 1996. Application of the matK gene sequences to grass systematics. Canad. J. Bot. 74: 125-134. Liese, W. 1998. The Anatomy of Bamboo Culms. Technical Report 18. InternationalNetworkfor Bamboo and Rattan, Beijing. Linder, H. P. 1987. The evolutionaryhistory of the Poales/ Restionales-A hypothesis. Kew Bull. 42: 297-318. . 1992a. The gynoecia of Australian Restionaceae: Morphology,anatomyand systematic implications. Austral. Syst. Bot. 5: 227-245. 1992b. The structure and evolution of the female flower of the African Restionaceae. Bot. J. Linn. Soc. 109: 401-425. & G. Davidse. 1997. The systematics of Tribolium Desv. (Danthonieae, Poaceae). Bot. Jarhb. Syst. 119: 445-507. & I. K. Ferguson. 1985. On the pollen morphology and phylogeny of the Restionales and Poales. Grana 24: 65-76. & G. A. Verboom. 1996. Generic limits in the Rytidosperma(Danthonieae, Poaceae) complex. Telopea 6: 597-627. & N. P. Barker. 1997. Phylogeny and evolution in the Crinipes group of grasses (Arundinoideae: Poaceae). Kew Bull. 52: 91-110. Lou, H., B. Zhang & D.-F. Qi. 1989. M13 cloning and nucleotide sequence of the sorghum chloroplast gene for the large subunit of ribulose-l,5-bisphosphate carboxylase. Acta Biochim. Biophys. Sin. 21: 179-188. Maddison, W. P. & D. R. Maddison. 1993. MacClade,version 3. Analysis of Phylogeny and CharacterEvolution. Sinauer, Sunderland, Massachusetts. Mason-Gamer,R. J. & E. A. Kellogg. 1996. Testing for phylogenetic conflict among molecular data sets in the Triticeae. Syst. Biol. 45: 524-545. , C. F. Weil & E. A. Kellogg. 1998. Granule-bound starch synthase: Structure, function, andtphylogenetic utility. Molec. Biol. Evol. 15: 1658-1673. Mathews, S. & R. A. Sharrock. 1996. The phytochrome gene family in grasses (Poaceae): A phylogeny and evidence that grasses have a subset of the loci found in dicot angiosperms. Molec. Biol. Evol. 13: 1141-1150. , M. Lavin & R. A. Sharrock. 1995. Evolution of the phytochrome gene family and its utility for phylogenetic analyses of angiosperms. Ann. Missouri Bot. Gard. 82: 296-321. ,R. C. Tsai & E. A. Kellogg. 2000. Phylogenetic structure in the grass family (Poaceae): Evidence from the nuclear gene phytochromeB. Amer. J. Bot. 87: 96107. McClure, F. A. 1966. The Bamboos:A Fresh Perspective. HarvardUniv. Press, Cambridge,Massachusetts. . 1973. Genera of bamboos native to the New World (Gramineae:Bambusoideae). Smithsonian Contr. Bot. 9: 1-148. Metcalfe, C. R. 1960. Anatomy of the Monocotyledons.I. Gramineae.Clarendon Press, Oxford. Moore, G., K. M. Devos, Z. Wang & M. D. Gale. 1995. Grasses, line up and form a circle. Curr.Biol. 5: 737739. Morales, C. 0. 1998. NomenklatorischeProbleme in Bezug auf die Namen Phareae und Leptaspideae (Gramineae). Sendtnera 5: 243. Nadot, S., R. Bajon & B. Lejeune. 1994. The chloroplast gene rps4 as a tool for the study of Poaceae phylogeny. P1. Syst. Evol. 191: 27-38. Nakai, T. 1943. Anomochloaceae. In Ordines, Familiae, Tribi, Genera, a Prof. Nakai, Edita 222. Nambudiri, E. M. V., W. D. Tidwell, B. N. Smith & N. P. Hebbert. 1978. A C4 plant from the Pliocene. Nature 276: 816-817. Nishizawa, Y. & A. Hirai. 1987. Nucleotide sequence and expression of the gene for the large subunit of rice ribulose 1, 5-bisphosphate carboxylase.Jap. J. Genet. 62: 389-395. Nixon, K. C. 1993. Clados, version 1.4.98 (computersoftware and manual). Distributed by the author. Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae . 2000. Winclada, version 0.9.99m6.1 (computer softwareand manual). Distributed by the author. Nfiez, 0. 1968. El problemade la palea de OryzaL. Bol. Soc. Argent. Bot. 12: 57-97. Pilger, R. 1956. Gramineae II, Unterfam. Micrairoideae. P. 167 in A. Engler & K. Prantl, Die naturlichenPflanzenfamilien (ed. 2) Band 14d. Duncker & Humblot, Berlin. Poe, S. & D. L. Swofford.1999. Taxon sampling revisited. Nature 398: 299-300. Pohl, R. W. 1987. Man and the grasses: A history. Pp. 355-358 in T. R. Soderstrom,K. W. Hilu, C. S. Campbell & M. E. Barkworth(editors), Grass Systematics and Evolution. Smithsonian Institution Press, Washington, D.C. Pozzi, C., P. Faccioli, V. Terzi, A. M. Stanca, S. Cerioli, P. Castiglioni, R. Fink, R. Capone, K. J. Miiller, G. Bossinger, W. Rohde & F. Salamini. 2000. Genetics of mutations affecting the development of a barley floral bract. Genetics 154: 1335-1346. Prat, H. 1932. L'epiderme des gramin6es: etude anatomique et systematique. Ann. Sci. Nat. Bot. 14: 117324. 1936. La syst6matiquedes Graminees. Ann. Sci. Nat. Bot., ser. 10, 18: 165-258. 1960. Vers une classification naturelle des graminees. Bull. Soc. Bot. France 107: 32-79. Reeder, J. R. 1953. The embryo of Streptochaetaand its bearing on the homologyof the coleoptile. Amer.J. Bot. 49: 77-80. 1957. The embryo in grass systematics. Amer.J. Bot. 44: 756-769. . 1961. The grass embryo in systematics. Pp. 9196 in Recent Advances in Botany, Vol. 1. Univ. Toronto Press, Toronto. 1962. The Bambusoid embryo: A reappraisal. Amer. J. Bot. 49: 639-641. Renner, S. S., D. B. Foreman& I). Murray.2()()().Timing transantarctic disjunctions in the Atherospermataceae (Iaurales): Evidence from coding and noncoding chloroplast sequences. Syst. Biol. 49: 579-591. Renvoize, S. A. & W. D. Clayton. 1992. Classificationand evolution of the grasses. Pp. 3-37 in (. P. Chapman (editor), Grass Evolution and Domestication.Cambridge Univ. Press, Cambridge,U.K. Rohde, W., D. Becker & F. Salamini. 1988. Structural analysis of the waxy locus from Hordeumvulgare. Nucl. Acids Res. 16: 7185-7186. Rosengurtt,B., A. Laguardia& B. R. Arrillagade Maffei. 1972. El caracter lipido del endosperma central en especies de Gramineas. Bol. Fac. Agron. Univ. Montevideo 124: 1-43. Roshevits, R. Yu. 1937. Zlaki: Vvedenie v Izuchenie Kormovykh Zlakov. [Grasses: An introductionto the study of fodder and cereal grasses.] Sel'khozgiz, Gosudarstvennoe Izdatel'stvo Kolkhoznoi I Sovkhoznoi Literatury. Agricultural Publishing House, Moscow, Leningrad. [English translationpublished for the Smithsonian Institution and the National Science Foundation,Washington, D.C., by the Indian National Scientific Documentation Centre, New Delhi. 1980. TT 72-51033.] . 1946. Sistema zlakov v svyazi s ikh evolyutsiei. KamarovBotanical Institute, U.S.S.R. Academy of Sciences, Leningrad. [Portuguesetranslationby T. Sendulsky. 1969. Evolucao e sistematica das gramineas. Bol. Inst. Bot. (Sao Paulo) 5: 1-20.] Sage, R. F. & R. K. Monson. 1999. C4 Plant Biology. Academic Press, San Diego. Sanchez-Ken, J. G. & L. G. Clark. 2001. Gynerieae, a new Neotropical tribe of grasses (Poaceae). Novon 11: 350-352. Schmidt, R. J. & B. A. Ambrose. 1998. The blooming of grass flower development. Curr.Opinion P1. Biol. 1: 6067. Schuster, J. 1910. tber die Morphologie der Grasbliite. Flora 100: 213-266, plates 2-5. Seberg, O. & I. Linde-Laursen. 1996. Eremium, a new genus of the Triticeae (Poaceae) from Argentina. Syst. Bot. 21: 3-15. Sendulsky, T., T. S. Filgueiras & A. G. Burman. 1987. Fruits, embryos, and seedlings. Pp. 31-36 in T. R. Soderstrom, K. W. Hilu, C. S. Campbell & M. E. Barkworth (editors), Grass Systematics and Evolution. Smithsonian Institution Press, Washington,D.C. Serebryakova,T. I. 1971. Morfogenezpobegov i evoluiutsiia zhiznennykh form zlakov [Shoot Morphogenesisand Evolution of Life Forms in Grasses]. Izdatel'stvo"Nauka," Moskva. Shantz, H. L. 1954. The place of grasslands in the earth's cover of vegetation. Ecology 35: 143-145. Simon, B. K. & S. W. L. Jacobs. 1990. Gondwanangrasses in the Australian flora. Austrobaileya 3: 239-260. Sinha, N. R. & E. A. Kellogg. 1996. Parallelism and diversity in multiple origins of C4 photosynthesis in the grass family. Amer. J. Bot. 83: 1458-1470. Soderstrom,T R. 1967. Taxonomicstudy of subgenus Podosemum and section Epicampesof Muhlenbergia(Gramineae). Contr. U.S. Natl. Herb. 34: 75-189 + 14 pl. 1981. The grass subfamily Centostecoideae. Taxon 30: 6144-16. & R. . llis. 1987. The position of bamboo genera and allies in a system of grass classification. Pp. 225-238 in rr.R. Soderstrom, K. W. Hilu, C. S. Campbell & M. E. Barkworth(editors), Grass Systematics and Evolution. Smithsonian Institution Press, Washington, D.C. & E. J. Judziewicz. 1987. The Phareae and Streptogyneae of Sri I,anka: A morphological-anatomical study. Smithsonian Contr. Bot. 65. Soltis, D. E., P. S. Soltis, M. E. Mort, M. W. Chase, V. Savolainen, S. B. Hoot & C. M. Morton.1998. Inferring complex phylogenies using parsimony:An empirical approach using three large DNA data sets for angiosperms. Syst. Biol. 47: 32-42. Soreng, R. J. & J. I. Davis. 1998. Phylogenetics and character evolution in the grass family (Poaceae): Simultaneous analysis of morphological and chloroplast DNA restriction site character sets. Bot. Rev. 64: 1-85. . 2000. Phylogenetic structure in Po& aceae subfamily Pooideae as inferred from molecular and morphological characters: Misclassification vs. reticulation. Pp. 61-74 in S. W. L. Jacobs & J. E. Everett (editors), Grasses: Systematics and Evolution. CSIRO Publishing, Collingwood,Victoria. & J. J. Doyle. 1990. A phylogenetic analysis of chloroplast DNA restriction site variationin Poaceae subfam. Pooideae. P1. Syst. Evol. 172: 83-97. Spangler, R., B. Zaitchik, E. Russo & E. A. Kellogg. 1999. Andropogoneae evolution and generic limits in Sorghum (Poaceae) using ndhF sequences. Syst. Bot. 24: 267-281. Stapleton, C. M. A. 1997. The morphologyof woody bam- 429 430 Annals of the Missouri Botanical Garden boos. Pp. 251-267 in G. P. Chapman(editor),The Bamboos, Academic Press, London. Stebbins, G. L. 1974. Flowering Plants: Evolution above the Species Level. Belknap Press of Harvard Univ. Press, Cambridge,Massachusetts. & B. Crampton.1961. A suggested revision of the grass genera of temperate North America. Pp. 133-145 in Recent Advances in Botany, Vol. 1. Univ. Toronto Press, Toronto. Sugiura, M. 1989. Oryzasativa chloroplast DNA 134,525 bp. Nagoya University, Center for Gene Research, Nagoya, Japan. Swofford, D. 1998. PAUP*. Phylogenetic Analysis Using Parsimony (*and other methods). Version 4. Sinauer, Sunderland, Massachusetts. Takaiwa, F., K. Oono, Y. Iida & M. Sugiura. 1985. The complete nucleotide sequence of a rice 25S rRNA gene. Gene 37: 255-259. Tateoka, T 1957. Miscellaneous papers on the phylogeny of the Poaceae (10). Proposition of a new phylogenetic system of Poaceae. J. Jap. Bot. 32: 275-287. 1962. Starch grains of endosperm in grass systematics. Bot. Mag. (Tokyo)75: 336-343. , S. Inoue & S. Kawano. 1959. Notes on some grasses. IX. Systematic significance of bicellular microhairs of leaf epidermis. Bot. Gaz. 121: 80-91. Templeton, A. R. 1983. Phylogenetic inference from restriction endonuclease cleavage site maps with particular reference to the evolution of humans and the apes. Evolution 37: 221-244. Terrell, E. E. 1971. Survey of occurrences of liquid or soft endosperm in grass genera. Bull. TorreyBot. Club 98: 264-268. Thompson, J. D., D. G. Higgins & T. J. Gibson. 1994. CLUSTALW: Improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucl. Acids Res. 22: 4673-4680. Tieghem, Ph. van. 1897. Morphologiede l'embryon et de la plantule chez les Gramin6eset les Cyperacees. Ann. Sci. Nat. Bot. 3: 259-309. Tzvelev, N. N. 1989. The system of grasses (Poaceae) and their evolution. Bot. Rev. 55: 141-203. Verboom,G. A., H. P. Linder & N. P. Barker.1994. Haustorial synergids: An importantcharacter in the systematics of danthonioid grasses (Arundinoideae:Poaceae). Amer. J. Bot. 81: 1601-1610. Vickery, J. W., S. W. L. Jacobs & J. Everett. 1986. Taxonomic studies in Stipa (Poaceae) in Australia. Telopea 3: 1-132. Wang, Z. Y., F. Q. Zheng, J. P. Gao, X. Q. Wang, M. Wu, J. L. Zhang & M. M. Hong. 1994. Identificationof two transposon-like elements in rice Wx gene. Sci. in China Ser. B-Chem. Life. Sci. & Earth Sci. 37: 437-447. Watson, L. & M. J. Dallwitz. 1992. The Grass Genera of the World. CAB International,Wallingford,U.K. & . 1999. Grass Genera of the World:Descriptions, Illustrations, Identification, and Information Retrieval; Including Synonyms, Morphology,Anatomy, Physiology, Phytochemistry, Cytology, Classification, Pathogens, World and Local Distribution, and References. http://biodiversityunoedu/delta/ Version: 18th August 1999. ,H. T. Clifford& M. J. Dallwitz. 1985. The classification of the Poaceae: Subfamilies and supertribes. Austral. J. Bot. 33: 433-484. Whistler, R. L., J. N. BeMiller & E. F. Paschall (editors). 1984. Starch: Chemistry and Technology. Academic Press, Orlando. Wilson, W. A., S. E. Harrington,W. L. Woodman,M. Lee, M. E. Sorrells & S. R. McCouch. 1999. Inferences on the genome structure of progenitor maize throughcomparative analysis of rice, maize and the domesticated panicoids. Genetics 153: 453-473. Zhang, W. 1996. Phylogeny and Classificationof the Bamboos (Poaceae: Bambusoideae). Ph.D. Thesis, Iowa State University, Ames, Iowa. . 2000. Phylogeny of the grass family (Poaceae) from rp/16 intron sequence data. Molec. Phylogenet. Evol. 15: 135-146. & L. G. Clark. 2000. Phylogeny and classification of the Bambusoideae (Poaceae). Pp. 35-42 in S. W. L. Jacobs & J. E. Everett (editors), Grasses: Systematics and Evolution. CSIRO Publishing, Collingwood,Victoria. Zuloaga, F. O., O. Morrone& L. Giussani. 2000. A cladistic analysis of the tribe Paniceae: A preliminaryapproach. Pp. 123-135 in S. W. L. Jacobs & J. E. Everett (editors), Grasses: Systematics and Evolution. CSIRO Publishing, Collingwood,Victoria. Zurawski, G., M. T. Clegg & A. H. D. Brown. 1984. The nature of nucleotide sequence divergence between barley and maize chloroplast DNA. Genetics 106: 735749. Volume 88, Number3 2001 Grass PhylogenyWorkingGroup Phylogeny and Classificationof Poaceae 431 Appendix I. Taxa included. For each data set, species name, voucher, and reference are listed, as well as GenBank accession numbers for gene sequences. EAK = Elizabeth Kellogg; HPL = Peter Linder;JID = Jerrold Davis; LGC = Lynn Clark;NPB = Nigel Barker;PMP = Paul Peterson; RJS = Robert Soreng; SJ = Surrey Jacobs; WZ = Weiping Zhang; XL = Ximena Londono; BBG = Berlin Botanic Garden; BHC = L. H. Bailey HortoriumConservatory;FTG = Fairchild Tropical Garden; NTBG = National Tropical Botanical Garden (Hawaii); PI = USDA Plant Introduction Station (Pullman, Washington)as source of seed. Genus Species Voucher Reference GenBank # ndhF Anomochloa Streptochaeta Pharus Guaduella indica L. stipularis Mast. tetraphyllum(Labill.) B. G. Briggs & L. A. S. Johnson ascendens Gaudich. ex Brongn. & Gris. marantoidea Brongn. angustifolia Soderstr. latifolius L. marantifolia Franch. Puelia ciliata Franch. Flagellaria Elegia Baloskion Joinvillea Eremitis Pariana Lithachne Olyra Buergersiochloa Arundinaria Chusqlue Streptogyna Ehrharta Oryza Leersia Phaenosperma Brachyelytrum sp. nov. radicifora Sagot ex Doll hurmilis Soderstr. latifolia L,. bambusoidesPilg. gigantea (Walter)Muhl. latifolia 1. (;. Clark americana C. fE.Hubb). calycina Sin. sativa L. virginica Willd. globosum Munroex Benth. erectum(Schreb.) P. Beauv. spartum L. stricta L. LGC & WZ 1305 (ISC) Eldenas 2 (BOL) Kew-6565-1977 (BH) Clark et al. (1995) This paper This paper U22007 AF251443 AF251444 NTBG-800379 (living) Clark et al. (1995) U21973 LGC 1299 (ISC) LGC 1304 (ISC) LGC 1302 (ISC) Kobayashi et al. 1539 (ISC) Kobayashi et al. 1541 Clark et Clark et Clark et Clark et (1995) (1995) (1995) (2000) U21991 U21982 U21992 AF164777 Clark et al. (2000) AF164779 (ISC) LGC & WZ 1343 (ISC) LGC & WZ 1344 (ISC) LGC 1298 (ISC) XL & LGC911 (1SC) Iransfield 1382 (K) WZ 8400703 (ISC) LGC& XL 4/7 (ISC) lohl & Davidse 12310 Zhang & Clark (2000) Zhang & Clark (2000) Clark et al. (1995) Clark et al. 1995 Zhang & Clark (2000) Clark et al. (1995) Clark et al. (1995) Clark el al. (1995) AF182353 AF182354 U21977 U21971 AF182341 U21846 U21989 U21965 (ISC) NPB s.n. (BOL) Sugiura (1989) LGC 1316 (ISC) LGC 1292 (ISC) Clark et Clark et Clark et Clark et (1995) (1995) (1995) (1995) U21995 X159()1 U21974 U22005 LGC 1330 (ISC) Clark et al. (1995) U22004 This paper This paper AF251445 AF251446 This paper This paper AF251447 AF251448 This paper This paper Clark et al. (1995) AF251449 AF251450 U21924 This paper This paper This paper Clark et al. (1995) AF251451 AF251452 AF251453 U21998 Clark et al. (1995) U22000 This paper AF251454 RJS 3698 (BH) BBG: Royl & Schiers s.n. 1988 (B) HPL 5590 (BOL) avenaceus R. Br. Anisopogon mauritanica (Poir.)T. BBG: Royl & Schiers Ampelodesmos Durand & Schinz s.n. 1988 (B) barbata Desf. PI-229468 (BH) Stipa Nassella viridula (Trin.)Barkworth PI-387938 (BH) Oryzopsis(=Piptath- racemosa (Sm.) Ricker ex LGC & WZ 1288 (ISC) Hitchc. erum) Brachypodium distachyon (L.) P. Beauv. PI-422452 (BH) Melica altisssima L. PI-325418 (BH) striata (Lam.) Hitchc. JID & RJS s.n. (BH) Glyceria Diarrhena obovata (Gleason) Bran- LGC & WZ 1216 (ISC) denburg sativa L. Avena material from R. Wise (ISU) Bromus inermis Leyss. PI-314071 (BH) Lygeum Nardus al. al. al. al. al. al. al. al. 432 Annals of the MissouriBotanicalGarden Appendix I. Continued. Genus Hordeum Aristida Stipagrostis Amphipogon Arundo Molinia Phragmites Merxmuellera Karroochloa Danthonia Austrodanthonia Merxmuellera Centropodia Eragrostis Uniola Zoysia Distichlis Pappophorum Spartina Sporobolus Micraira Thysanolaena Gynerium Chasmanthium Zeugites Danthoniopsis Panicum Pennisetum Miscanthus Zea Species Voucher material from R. Wise (ISU) purpureaNutt. var. longi- Gabel 2700 (ISC) seta (Steud.) Vasey ex Rothr. zeyheri (Nees) DeWinter NPB 1133 (BOL) HPL 5634 (BOL) strictus R. Br. donax L. LGCs.n. (ISC) LGC 1165 (ISC) caerulea (L.) Moench LGC 1294 (ISC) australis (Cav.)Trin. ex Steud. macowanii (Stapf) Conert NPB 1008 (BOL) purpurea (L.f.) Conert & HPL 5360 (BOL) Tuirpe PI-232247 (BH) californica Bolander HPL 5633 (BOL) laevis (Vickery) H. P. Linder NPB 960 (GRA) rangei (Pilg.) Conert NPB 967 (BOL) glauca (Nees) Copt curvula (Schrad.) Nees LGC1303 (ISC) paniculata L. JID s.n. (BH) matrella (L.) Merr. LGC 1174 (ISC) Allred s.n. (BH) spicata (L.) E. Green subsp. stricta (Torr.)R. F. Thorne bicolor E. Fourn. Pohl 12464 (ISC) LaDuke s.n. (BH) pectinata Link indicus (L.) R. Br. LGC 1293 (ISC) lazaridis L. G. Clark, LGC 1157 (ISC) Wendel & Craven maxima (Roxb.) Kuntze FTG (living) LGC & P Asimbaya sagittatum (Aubl.) P. 1472 (ISC) Beauv. D. Lewis s.n. (ISC) laxum (L.) H. O. Yates LGC 1171 (ISC) pittieri Hack. LGC 1173 (ISC) petiolata (J. B. Phipps) Clayton LGC 1164 (ISC) virgatum L. alopecuroides(L.) Spreng. RJS s.n. (BH) Arnold Arboretum301 japonicus Andersson 80c (living) Materialfrom M. Lee mays L. cv. 'B73' (ISU) vulgare L. Reference GenBank # Clark et al. (1995) U22003 Clark et al. (1995) U21966 This paper This paper Clark et al. (1995) Clark et al. (1995) Clark et al. (1995) AF251455 AF251456 U21997 U21994 U21996 This paper This paper AF251457 AF251458 This paper This paper AF251459 AF251460 This paper This paper Clark et al. (1995) This paper Clark et al. (1995) This paper AF251461 AF251462 U21988 AF251463 U21975 AF251464 This paper This paper Clark et al. (1995) Clark et al. (1995) AF352581 AF251465 U21983 U21972 Clark et al. (1995) This paper U21984 AF251466 Clark et al. (1995) Clark et al. (1995) Clark et al. (1995) U27296 U21987 U22008 Clark et al. (1995) This paper Spangler et al. (1999) U21986 AF251467 AF117417 Clark et al. (1995) U21985 Chase et al. (1993) Duvall & Morton(1996) L12678 L12675 Katiyama& Ogihara (1996) Duvall & Morton(1996) D38296 L01471 Duvall & Morton(1996) Clark et al. (2000) AF021875 AF164778 rbcL Flagellaria Elegia Baloskion Joinvillea Anomochloa Guaduella indica Chase 206 (NCU) Schel209 (NCU) Chase f.) capensis (Burm. pe No voucher tetraphyllum plicata (Hook. f.) Newell & B. C. Stone marantoidea marantifolia Thren84 (NO) LGC1299 (ISC) Kobayashi et al. 1539 (ISC) Volume 88, Number3 2001 Appendix I. Grass Phylogeny WorkingGroup Phylogeny and Classificationof Poaceae Continued. Genus Species Voucher Puelia ciliata Lithachne Bambusa humilis multiplex (Lour.) Raeusch. ex Schult. & Schult. f. circinata Soderstr.& C. Quail Botanic Garden Calderon (living) No voucher sativa Chusquea Oryza Leersia Stipa Avena Bromus Hordeum Aristida Stipagrostis Amphipogon Arundo Moliniopsis Phragmites Merxmuellera Karroochloa Danthonia Centropodia Eragrostis Enneapogon Eriachne Thysanolaena Gynerium Chasmanthium Pennisetum Sorghum Zea rpoC2 Joinvillea Olyra Bambusa Ehrharta Oryza Lygeum Nardus Anisopogon Stipa Briza Bromus 433 Kobayashi et al. 1541 (ISC) LGCs.n. (ISC) Sanders 62-616 (UCR) oryzoides(L.) Sw. dregeana Steud. var. dregeana sativa inermis Leyss. LGCs.n. (ISC) McDowell s.n. (BOL) vulgare congesta Roem. & Schult. zeyheri strictus donax japonica (Hack.) Hayata australis macowani purpurea spicata (I.) P. Beauv. ex Roem. & Schult. No voucher No voucher vulgaris Schrad. ex J. C. Wendl. dura Nees ex Trin. sativa spartum stricta avenaceus dregeana maxima L. tectorumL. Clark et al. (2000) AF164780 Duvall & Morton(1996) Duvall & Morton(1996) U13231 M91626 Duvall & Morton(1996) U13227 Nishizawa & Hirai (1987) Duvall & Morton(1996) Barker et al. (1995) D00207 U13228 L15300 Z49836 No voucher NPB 1130 (BOL) NPB 1133 (BOL) HPL 5634 (BOL) NPB 1131 (BOL) Kobayashi 1253 NPB 1132 (BOL) NPB 1008 (BOL) HPL 5360 (BOL) EAK V1 (GH) Barker et al. (1995) Barker (1997) Barker et al. (1995) Barker et al. (1995) Barker et al. (1995) Barker et al. (1995) Barker et al. (1995) Barker et al. (1995) U31378 U88403 U31360 U31439 U29900 U31438 U31437 U31102 Barker et al. Barker et al. Barker et al. This paper Barker et al. U31100 U31104 U31103 AF352580 U31380 HPL 5410 (BOL) NPB 1 135 (BOL) NPB 1023 (BOL) EAK s.n. (GH) Kew 1979-3225 Warr (living) Kew 1991-1276 Kall sagittatum (living) latifolium (Michx.) H. O. Snow 5944 Yates No voucher glaucum (L.) R. Br. bicolor (L.) Moench No voucher No voucher mays latifolia GenBank # Duvall et al. (1993) Seberg & Linde-Laursen (1996) Zurawski et al. (1984) Barker et al. (1995) glauca capensis (Thunl.) Trin. scaber Lehm. triodioides Domin maxima plicata Reference No voucher HPL 5742 (BOL) Durban Botanic Garden (living) NPB 1118 (BOL) No voucher Kew (living) Kew (living) HPL 5590 (BOL) McDowell s.n. (BOL) EAK s.n. (GH) EAK s.n. (GH) (1995) (1995) (1995) (1995) X00630 U31359 Barker et al. (1995) U31105 Barker et al. (1995) U31101 Doebley et al. (1990) Lou et al. (1989) Gaut et al. (1992) L14623 1515164A Z11973 Barker et al. (1999) Barker et al. (1999) Barker et al. (1999) AF001864 U90825 U90824 Barker et al. (1999) Hiratsuka et al. (1989) Cummings et al. (1994) Cummings et al. (1994) Barker et al. (1999) Barker et al. (1999) Cummings et al. (1994) Cummings et al. (1994) AF064761 X15901 L25381 L25382 U92263 U90826 L25376 L25377 Annals of the Missouri Botanical Garden 434 Appendix I. Continued. Genus Species Voucher Stipagrostis Amphipogon Arundo Moliniopsis Phragmites Merxmuellera Karroochloa Danthonia Austrodanthonia Merxmuellera Centropodia Eragrostis Enneapogon Spartina Micraira Thysanolaena congesta Roem. & Schult. subsp. barbicollis (Trin. & Rupr.) DeWinter zeyheri subsp. zeyheri strictus donax japonica australis macowanii purpurea spicata laevis rangei glauca capensis scaber alterniflora lazaridis maxima Gynerium sagittatum Chasmanthiuml Panicum Pennisetum Sorghum Zea latifolium maximum Jacq. sp. bicolor mays NPB 1133 (BOL) HPL 5634 (BOL) NPB 1131 (BOL) Kobayashi 1253 NPB 1132 (BOL) NPB 1008 (BOL) HPL 5360 (BOL) EAK V1O(GH) HPL 5633 (BOL) NPB 960 (GRA) HPL 5410 (BOL) NPB 1135 (BOL) NPB 1023 (BOL) EAK s.n. (GH) LGC 1157 (ISC) Kew, 1979-3225 Warr (living) Kew, 1991-1276 Kall (living) Snow 5944 NPB 1125 (BOL) No voucher No voucher No voucher indica RJS 77 394 (BH) Joinvillea ascendens Moore 10438 (NY) Anomochloa Streptochaeta Pharus Puelia marantoidea angustifolia lappulaceus Aubl. Aristida PhytochromeB Flagellaria Eremitis Pariana Lithachne Olyra Buergersiochlo( Pseudosasa Chusquea Streptogyna Ehrharta Oryza Lygeum Nardus Anisopogon Nassella Brachypodium Melica Reference GenBank # NPB 1130 (BOL) Barker et al. (1999) U90827 Barker et al. (1999) Barker et al. (1999) Barker et al. (1999) Barker et al. (1999) Barker et al. (1999) Barker et al. (1999) Barker et al. (1999) Barker et al. (1999) Barker et al. (1999) Barker et al. (1999) Barker et al. (1999) Barker et al. (1999) Barker et al. (1999) Cummingset al. (1994) Barker et al. (1999) Barker et al. (1999) U90828 U94335 U92264 U95081 U95130 U95076 U94824 U93362 U96313 U95077 U92265 U96317 U96319 L25386 U96318 U96315 Barker et al. (1999) U94392 Barker et al. (1999) Barker et al. (1999) Cummingset al. (1994) Chen et al. (1993) Igloi et al. (1990) U94334 AF000021 L25383 Z14983 X86563 Mathews& Sharrock (1996) Mathews & Sharrock (1996) Mathewset al. (2000) Mathewset al. (2000) Mathewset al. (2000) Mathewset al. (2000) U61203 AF137291 AF137328 AF137321 AF137324 Mathewset Mathewset Mathewset Mathewset Mathewset Mathewset (2000) (2000) (2000) (2000) (2000) (2000) AF137304 AF137317 AF137307 AF137315 AF137295 AF137323 Mathewset al. (2000) Mathewset al. (2000) Mathewset al. (2000) Dehesh et al. (1991) Mathewset al. (2000) Mathewset al. (2000) Mathewset al. (2000) Mathewset al. (2000) Mathewset al. (2000) Mathewset al. (2000) AF137298 AF137329 AF137302 X57563 AF137309 AF137313 AF137290 AF137314 AF137294 AF137310 LGC 1299 (ISC) LGC 1304 (ISC) LGC 1,329 (ISC) ciliata Kobayashi et al. 1541 (ISC) LGC & WZ 1343 (ISC) sp. nov. LGC & WZ 1344 (ISC) radiciflora pauciflora (Sw.) P. Beauv. LGC 1297 (ISC) XL & LGC911 (ISC) latifolia bambusoides Dransfield 1382 (K) EAK V6 (A) japonica (Sieb. & Zucc. ex Steud.) Makino ex Nakai oxylepis (Hack.) Ekman LGC 1069 (ISC) americana Johnston 433 EAK V44 (GH) erecta Lam. no voucher sativa RJS 3698 (BH) spartum BBG: Royl & Schiers s.n. stricta HPL 5590 (BOL) avenaceus Lavin s.n. (MONT) viridula pinnatum (L.) P. Beauv. PI-440176 (GH) PI-383702 (BH) cupanii Guss. al. al. al. al. al. al. U61205 Volume 88, Number3 2001 Appendix I. Grass PhylogenyWorkingGroup Phylogeny and Classificationof Poaceae Continued. Genus Glyceria Diarrhena Phalaris Bromus Triticum Aristida Molinia Phragmites Danthonia Eragrostis Sporobolus Thysanolaena Chasmanthium Danthoniopsis Panicum Pennisetum Miscanthus Zea Species Anomochloa Streptochaeta Pharus Eremitis Lithachne Olyra Pseudosasa Chusquea Ehrharta Oryza Leersia Brachyelytrum Lygeum Nardus Anisopogon Ampelodesmos Stipa Piptatherum Brachypodium Melica Glyceria Diarrhena Voucher grandis S. Watson obovata arundinacea L. JID & RJS s.n. (BH) LGC & WZ 1216 (ISC) RJS 3427 (BH) inermis Lavin s.n. (MONT) aestivum L. Mason-Gamers.n. (GH) purpurea subsp. longiseta Lavin s.n. (MONT) caerulea RJS 3305 (BH) australis Keller s.n. (GH) EAK V1O(GH) spicata cilianensis (All.) Vignolo Lavin s.n. (MONT) ex Janch. PMP 10008 (US) giganteus Nash maxima Farnsworths.n. (GH) EAK V13 (A) latifolium dinteri (Pilg.) C. E. PI-207548 (GH) Hubb. Lavin s.n. (MONT) capillare L. EAK s.n. (A) alopecuroides Arnold Arboretum301japonicus 80C (living) Lavin s.n. (MONT) mays Chloroplastrestriction site polymorphisms indica Flagellaria Baloskion tetraphyllum ascendens Joinvillea Nassella 435 BHC-77394 Kew-6565-1977 (BH) NTBG-800379 (H. Moore 10438) LGC 1299 (ISC) marantoid(ea PMP 9525 (US) sodiroanaL Hack. IHC from USZ latifolius USNHG-153, Soderstrom sp. 2182 (US) or USNHG286 (US) humilis BHC from U. S. National Zoological Gardens PMP 7311 (US) latifolia BHC-71467 japonica aff. subulata L. G. Clark PMP 9499 (US) PI-208983 (BH) calycina no voucher sativa RJS 3399 (BH) virginica erectum RJS 3427 (BH) RJS 3698 (BH) spartum BBG: seed from Royl & stricta Schiers s.n. 1988, Hempel s.n. 1987 (B) HPL 5590 (BOL) avenaceus mauritanica BBG: Royl & Schiers s.n. 1988 (B) barbata PI-229468 (BH) viridula PI-387938 (BH) PI-284145 (BH) miliaceum (L.) Coss. PI-440170 (BH) pinnatum PI-325418 (BH) altissima striata JID & RJS s.n. (BH) obovata Seed from Tiedye 5186 (DAO) Reference GenBank # Mathews et Mathews et Mathews et Mathews et Mathews et Mathews et Mathews et Mathews et Mathews et Mathews et al. al. al. al. al. al. al. al. al. al. (2000) (2000) (2000) (2000) (2000) (2000) (2000) (2000) (2000) (2000) AF137305 AF137301 AF137320 U61193 AF137331 AF137292 AF137312 AF137322 AF137299 U61200 Mathews et Mathews et Mathews et Mathews et al. al. al. al. (2000) (2000) (2000) (2000) AF137327 AF137330 AF137297 AF137300 Mathews et al. (2000) Mathews et al. (2000) Mathews et al. (2000) AF137316 AF137318 AF137311 Mathews et al. (2000) AF137332 Soreng & Davis (1998) Soreng & Davis (1998) Davis & Soreng (1993) Soreng & D)avis(1998) Soreng & Davis (1998) I)avis & Soreng (1993) Soreng & Davis (1998) Davis & Soreng (1993) Soreng & Davis (1998) Davis & Soreng (1993) Soreng & Davis (1998) Soreng & Davis (1998) Hiratsuka et al. (1989) Davis & Soreng (1993) Davis & Soreng (1993) Soreng & Davis (1998) Davis & Soreng (1993) Soreng & Davis (1998) Soreng & Davis (1998) Davis & Soreng (1993) Soreng & Davis (1998) Davis & Soreng (1993) Davis & Soreng (1993) Davis & Soreng (1993) Davis & Soreng (1993) Davis & Soreng (1993) Annals of the MissouriBotanicalGarden 436 Appendix I. Continued. Genus Avena Bromus barbata Pott ex Link inermis Triticum aestivum L. cv. 'Susquehanna' purpurea strictus Aristida Amphipogon Arundo Molinia Phragmites Danthonia Eragrostis Uniola Zoysia Distichlis Spartina Sporobolus Chasmanthium Panicum Pennisetum Miscanthus Voucher Species donax caerulea australis californica curvula paniculata sp. spicata subsp. stricta pectinata giganteus latifolium virgatum alopecuroides sinensis Andersson var. gracillimus Hitchc. Reference GenBank # No voucher RJS 3428 (BH), PI314071 (BH) RJS s.n. (BH) Soreng & Davis (1998) Davis & Soreng (1993) Allred s.n. (BH) HPL 5634 (BOL) FTG-83-130 (BH) RJS 3305 (BH) RJS 3884 (BH) PI-232247 (BH) PI-365034 (BH) JID s.n. (BH) JID s.n. (BH) Allred s.n. (BH) LaDuke s.n. (BH) PMP 10008 (US) Cornell University gardens (living) USDA 421520 (BH) RJS s.n. (BH) RJS s.n. (BH) Soreng & Davis (1998) Soreng & Davis (1998) Soreng & Davis (1998) Soreng & Davis (1998) Davis & Soreng (1993) Davis & Soreng (1993) Davis & Soreng (1993) Soreng & Davis (1998) Soreng & Davis (1998) Soreng & Davis (1998) Soreng & Davis (1998) Soreng & Davis (1998) Davis & Soreng (1993) Wilson 7126 PMP & Annable 9525 (US) PMP & Annable 6944 (US) Utah State University s.n. (living) LGC & XL 417 (ISC) Kew 1973-15875 (living) No voucher Jacobs 7782 IntermountainHerbarium 1669 Catalan 1593 IntermountainHerbarium 203443 Dalby 94/01 Kew 150-90.00982 (living) Renvoize& Flores 5301 (K) Houck s.n. Hsiao et al. (1999) Hsiao et al. (1999) AF019784 AF019785 Hsiao et al. (1999) AF019786 Hsiao et al. (1999) AF019787 Hsiao et al. (1999) Hsiao et al. (1999) Takaiwaet al. (1985) Hsiao et al. (1999) Hsiao et al. (1999) AF019788 AF019791 AF019793 AF019794 Hsiao et al. (1999) Hsiao et al. (1999) AF019797 AF019796 Hsiao et al. (1999) Hsiao et al. (1999) AF019800 AF019799 Hsiao et al. (1999) AF019803 Hsiao et al. (1999) L36520 Hsiao 199 Hsiao et al. (1999) AF019802 University of Leicester Botanic Gardens 347 Curto 719 Curto 826 IntermountainHerbarium 218465 Hsiao et al. (1999) AF019805 Hsiao et al. (1999) Hsiao et al. (1999) Hsiao et al. (1999) L36518 L36516 AF019798 Soreng & Davis (1998) Soreng & Davis (1998) Davis & Soreng (1993) Davis & Soreng (1993) ITS Streptochaeta plicata sodiroana Pharus latifolius Lithachne humilis Chusquea Microlaena Brachyelytrum latifolia stipoides (Labill.) R. Br. sativa hexandra Sw. erectum Lygeum Nardus spartum stricta Anisopogon Ampelodesmos avenaceus mauritanica Stipa ichu (Ruiz & Pavon) Kunth leucotricha (Trin.& Rupr.) R. W. Pohl songaricum (Trin.& Rupr.) Roshev. ex Nikitina mexicanum (Roem. & Schult.) Link californica Scribn. striata americana P. Beauv. Joinvillea Oryza Leersia Nassella Piptatherum Brachypodium Melica Glyceria Diarrhena Volume88, Number3 2001 Appendix I. Grass PhylogenyWorkingGroup Phylogenyand Classificationof Poaceae 437 Continued. Genus Avena Bromus Hordeum Aristida longiglumis Durieu mnermis vulgare purpurea Stipagrostis Amphipogon Arundo Molintia zeyheri subsp. zeyheri caricinus F. Muell. Phragmites Merxmuellera Karroochloa Danthonia australis Rytidosperma donax caerulea macowant'i purpurea califoricn pumilum (Kirk) H. P. Linder Merxmuellera rangei Centropodia glauca Eragrostis dielsii Pulg. cx Diels & Pritz. gracilis Trin. Spartina Sporobolus Eriachne Micratir(L Thysanolaena Gynerium Chasmanthiutm PatnicuLm Pennisetum Miscantlhus Zea Voucher Species Reference Hsiao et al. (1999) Hsiao et al., (1994) Chattertonet al. (1992) IntermountainHerbarium Hsiao et al., 1999 209381 NPB 1133 Hsiao et al. (1999) Hsiao et al. (1998) Macfarlane 2155 Hsiao 196, Evans s.n. Hsiao et at. (1999) Kewl973 10386 Hsiao et al. (1999) Chatterton s. a. Hsiao et at. (1999) Kew 142-83.01715 Hsiao et al. (1998) HPL 5360 Hsiao et al. (1998) Curto 974 Hsiao et at. (1999) HPL 5747 Hsiao et at. (1998) Fritz, CN Hsiao 103 Hsiao 200 NPB 960 (GRA) NPB 967 Jacobs 7195 Hsiao et al. (1998) Hsiao et al. (1998) Hsiao et al. (1999) IntermountainHerbarium Hsiao et 194828 Curto s. n. Hsiao et airoides (Torr.) Torr. triseta Nees cx Steud. Hsiao et Jacobs 7184 Clarkson 1 0300 Hsiao et subulifolia F. Mucll. maxima Kew1979-3225 Hsiao et Kewl991-1276Kall Hsiao et sagtttaturn ltiterl)iountainHerbarium Hsia(aet latij6lium 216008 rThunth. )isalcatl rntr Hsitao160, 1)-19486 Hsiao et setaceum (Forssk.) Chiov. Cutrto 976 Hsiao et smiersuis 1va(ns s.n. Hsiao ct Hsiao et mnaysL. sub)sp. mexicaLnaL Hsiao 197 GenBank # Z11758 L11579 Z11759 AF019807 AF019845 AF019849 AF019809 AF109857 AF019810 AF019863 AF019874 AF019813 AF019878 AF019862 AF019861 AF019834 al. (1999) AF019844 al. al. al. al. al. al. (1999) (1999) (1999) (1999) (1999) (1999) AF019842 AF019818 AF019859 AF019854 AFO1 9853 A 0O 19815 al. (1999) at. (1999) al. (1999) al. (1999) AF0l 9829 AFO19833 AVO19822 AFO19817 (Schrad.) Ittis GBSSI marantoiden LGC 1299 (ISC) laippulaceus LGC 1.329 (ISC) Eremitis sp. nov. LGC & WZ 1343 (ISC) Pariana radicflora LGC & WZ 1344 (ISC) Chusquea exasperata L. G. Clark LGC et al. 1003 (ISC) Aniomochloai Oryza sativa Lygeum spartum No voucher RJS 3698 Melica cupanii PI-383702 (A) Glycerin grandis JID & RJS s.n. Hordeum vulgare Merxmuellera macowanal Karroochloa purpurea laevis Austrodanthonia No voucher NPB 1008 (BOL) HPL 5360 (BOL) HPL 5633 (BOL) Mason-Gameret al. (1998) Mason-Gameret al. (1998) Mason-Gameret al. (1998) Mason-Gameret al. (1998) Mason-Gameret al. (1998) Wang et at. (1994) Mason-Gameret al. (1998) Mason-Gameret al. (1998) Mason-Gameret al. (1998) Rohde et al. (1988) This paper This paper This paper AF079290 AF079298 AF079295 AF079297 AF079293 X65183 AF079289 AF079296 AF079291 X07932 AF353520 AF353519 AF353517 Annals of the Missouri Botanical Garden 438 Appendix I. Continued. Genus Reference Voucher Species Merxmuellera Centropodia Danthoniopsis rangei glauca dinteri NPB 960 (GRA) NPB 967 (BOL) PI-207548 (A) Pennisetum alopecuroides Park Seed 3650 (A) Sorghum bicolor PI-156549 (A) Zea mays No voucher GenBank # This paper This paper Mason-Gameret al. (1998) Mason-Gameret al. (1998) Mason-Gameret al. (1998) Klosgen et al. (1986) AF353521 AF353518 AF079251 AF079288 AF079258 X03935 Appendix II. Matrix of structural characters, as assembled for analysis in NONA (Goloboff, 1993). Taxa in the matrix appear in groupings according to what was known about the phylogeny at the time the taxon samlplinglist was prepared. Thus, the four outgroupscome first, followed by the early-divergingtaxa, then bambusoids, rices, pooids, etc. Abbreviations of taxon names and associated underlines are required by the program.Charactersand c haracterstates are described in Table 4, and are optimized on the cladogramin Figure 3. Codes used for polymorphisns (presence of two or more states) and subset ambiguities (when one or more states are not present, but the observed attributecannot be assigned to any of the states not eliminated) are as follows: ? = unobserved; - = inapplicable; \ = intermediate/ uncertain homology/unassignableto defined states; A = [01]; C = [03]; D = [12]; E = [13]; F = [231; H = [034]; J = [234]; K = [01341; L = [14]; N = [29]; Q = [07]; R = [012]; S = [57]; T = [127]. Characternumbers Taxon Flagellaria Elegia Baloskion Joinvillea Anomochloa Streptochaeta Pharus Guaduella Puelia Eremitis Pariana Lithachne Olyra Buergersiochloa Pseudosasa Chusquea Streptogyna Ehrharta Oryza Leersia Phaenosperma Brachyelytrum Lygeum Nardus Anisopogon Ampelodesmos Stipa Nassella Piptatherum Brachypodium_d Melica_a 00000000011111111112222222222333333333344444444445555 789016789012346789012345678901 234567890123456789023 \01-01AO- _-- 000--0-0- ----- _- -0\o o-0\ 070--0-0- \1111003130-0----??00?-0050- -\000 ,1100103310-0----??1-0---?0- -???1 ,1100102D2?-0----011-0---TO- -?0?0 01000110- - ----0\ \1111OA3130-0----???110\140- -1100 0001011\- \0---\0\ \10110\111?011100?101101140- -0110 01010110-\\\\---\0\\\\111101311?010101?10?100110- -0110 0000011111010--A0130??111101311?01110101110-1120- -0110 0100011101100--1013011111100221?01???????110A1?O- -???1 0100111101110--1013011111101F21?01???????10-11200100011101010--001301100010\111?011101?1011011\0- -0??1 -01?1 010001110A010--0\1301111110A211?011101??011011DO- -0??1 01000111A1010--11130110011012F1?0111010?01101110- -0111 0100011101010--11130110011012F1?0111010101101110010001110101110101301100A1012?1?01???????11011?0110011110100A1010130111111013210011101?301101120100011110111A--1013011001101221?0111010?01101100- -01?1 010011A101001111013011000101F11?0 -0??1 -01?1 -01?1 11101? 00120-0?? -0110 010OOO01A101\lA10\11201111110A221?0110010101101020- -?1?0 0100010101\lA10\0120A1000100231?011001010110A020OA00011101010--0013010011002D?01110104 0020- -???1 0000010101011101012011001100221?011A01?4110-0010- -?111 000001010\000--1?0----00110\121?01100001111100000100010100011101?0----00110\111?0110000111100030- -0111 0100010101011311013010001100F21001?????1?10-00-0- -0111 00000101010011A1013010001100221?0110000?110-0020- -?1?1 -?1?1 -01?1 -?1?1 -01?1 -01?1 01000101010111A101FO10001100J2100110000Kll0-0ORO01000101010111A1012010001100221?0110000Cl10-00\0010001010101A101013010000100221?01????01110-00200100010101001101?1201?001100221?01AO0000110-0OSO- 01100101010AO--0012100001100231?011000011AO-0090- -01?? Volume88, Number3 2001 Appendix II. Grass Phylogeny WorkingGroup Phylogeny and Classificationof Poaceae Continued. Characternumbers Taxon 00000000011111111112222222222333333333344444444445555 123456789012 345678901 23456789013458901234567890123 Glyceria-s Diarrhena Avena Bromus Triticum Aristida Stipagrostis Amphipogon Arundo Molinia Phragmites M_merxmuellera-m_ Karroochloa Danthonia Austrodanthonia R-merxmuellera-r_ Centropodia Eragrostis Uniola Zoysia Distichlis Pappophorum Spartina Sporobolus Eriachne Micraira Thysanolaena Gynerium Chasmanthiunml Zeugites I)anthoniopsis Panicum Pennisetum Miscanthus Zea 0110010101000--1012A00000100231?011000011AO-0000-01? 01000101010AO--101201000A100221?011AOA01110-0000-01? 0100010101001E21012010001100221?011AO\1llAO-0070-01?1 011001010100A1R1012010001100221?010000001AO-0070-0171 010001010000A101012010001100221?01100\001AO-0070-01?1 00OA010101011E011120110011002210010110011A100012-11?1 OA010101010113011A2011001100221?010110?D111000130???1 01010101010113A1112000001101221?110110?1?11100-0-?1?1 110001010100AEA1012001001100221?110110?1?A100020-?111 010101010100A101?12A01001100221?0101100111100090-1111 \10101010110A1010120010011012210\101100111101020-1111 0001010101001E11012011001100221101???????11000?0-???1 0101010101001E110120010011002211110110???1100060-???1 010101010100131101200100110022110101100D11100040-11?1 01010101010013110120010011002211110110???1100020-???1 0?01010101001E11012001001100221??????????10-0060-???1 0001010101001111012001001100221011?????L?11000630???1 0\01010101000--1012001001100221?11111001111A000301111 0\01010101100--1012001001101221?110110??11110003\?1?1 000A1OOA0001110000 ----0A1OA221?1111100D111100031?1?1 0000010101000--1012001001100221?1111100H?11100030?1?1 0?010101010A120101200\001100221?\111100??111000E-???l 0101010100010--000----001101221?1111100111A100Q31?1?1 00OAO10101010--O1A200100AA002D1?1111100E11110093Al1?1 OA01A101010AA1A101200100A100221?010111??111000?5-???? 00OAO1A101AAO--100----000101221?0101???Hl1100000-???l 1100111101010--101200000A100221?11?????????11010DO-1111 100AA1\101000--101201\000100221?117??????11010\0-???1 0100010101100--1012001001001221?1111110311100020-1111 010001110101A1OA01200\001101221?111101?1?11000\0-???1 OAO1A1A1AllllE11112001001100221?01?????4111000NL-1??i 0100010101110--01120A100110022101101110E111000\3011?1 010A01010111A1001A200100110A221?1101110311100091-11?1 0000010111111110012001001100221?110111?211100051-11?1 010001011A10O--0?12001001101211?\101110311100001-11?1 439 Annals of the MissouriBotanicalGarden 440 Appendix III. Consensus trees for individual data sets and combinations of data sets. Numbers above branches indicate percent of 500 bootstrap replicates, except for K (all molecular data), for which 1000 replicates were done. Tree statistics are listed in Table 3. The GPWG classification is overlain on each tree for comparison with Figure 2. -A. Chloroplastrestriction sites; strict consensus of seven trees. -B. ndhF; strict consensus of 16 trees. -C. rbcL; single most parsimonioustree. -D. rpoC2;strict consensus of 33 trees. -E. PhytochromeB; single most parsimonious tree. -F. ITS; strict consensus of 24 trees. -G. GBSSI, single most parsimonious tree. -H. Structuraldata; strict consensus of 38,000 trees. -I. Chloroplastdata; strict consensus of two trees. -J. Nuclear data; strict consensus of eight trees. -K. All molecular data; strict consensus of six trees. Flagellaria Baloskion Joinvillea Anomochloa CI Streptochaet7 91 100 A m h Anomochlo Pharus Eremitis Lithachne 67 Phar. Bambus. Olyra 98 9 8 "-~~~~9 Brachyelytrum 9 Lygeum Nardus Anisopogon Ampelodesmos Stipa Nassella Piptatherum Brachypodium Avena Bromus Triticum Diarrhena 70 96 -8 88 81 _ 51 62 I 68 ~ - 59 13 100 X Melica Glyceria Aristida 95 Aristid. Arundin. Danthoni. Uniola 9.8 84 97 74 ,- 98t 58j' Zoysia Chlorid. Spartina Sporobolus Distichlis Chasmanthium Centothec. Pennisetum Miscanthus Panic. Panicum Pseudosasa Chusquea 72 - Amphipogon Arundo rMolinia Molinia Phragmites Danthonia Eragrostis 80 Po. I Ehrharta Oryza Leersia Appendix III-Figure A, cp restriction sites Bambus. Ehrhart. Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae Flagellaria 1 00 Elegia Baloskion Joinvillea Anomochloa Anomochlo StreptochaetaAnmcl . 96 100 .1 ~0~~0n 10 0 Pharus 1Guaduella 10 00 ~1 81 63 1 00 94 7i:--I 100- 69 98 I 100- 1 00 95 99 6 98 1 00 72 9 6 ~96- 67 o88 10 00 441 00 1 1d00uI ul-- 1 100 I I• 1 00 ' 00 - 53 74 70 100 00-- 99 0 Appendix III-Figure B, ndhF Incertae sedis Ehrhart. Po. Arist id. Arundin. Arundo Molinia Phragmites Merxmuelleram. Karroochloa Dant honi. Austrodanthonia Danthonia Pan i c. Danthoniopsis ChasmanthiumMerxmuellerar. Centropodia Eragrostis Centothec. Zoysia Chlorid. Uniola 0 1 000 Q_--l62j-97 Bambus. Zea I 1 00 ] Phar. Pue I. Centothec. ZeugiThysanolaena Zeugites Incertae sedis Gynerium Panicum Pennisetum Panic. Miscanthus 67 69 Eremitis Pariana Lithachne Olyra Buergersiochloa Pseudosasa Chusquea Streptogyna Ehrharta Oryza Leersia Phaenosperma Anisopogon Ampelodesmos Stipa Piptatherum Nassella Melica Glyceria Brachypodium Avena Bromus Triticum Diarrhena Lygeum ~~ ^~Nardus BrachyelytrunAristida Stipagrostis 1 001I9 9 ] Spartina - Sporobolus Distichlis Pappophorum Micraira Incertae sedis Annals of the Missouri Botanical Garden 442 - Anomochlo. ] Pueli. Po. Arist id. Chlorid. Incertae sedis Arundin. 64 Thysanolaena Chasmanthium 87 95 1 1 00[-- Gynerium Pennisetum Miscanthus J Centothec. - Incertae Panic. Zea Lithachne 100 Appendix III-Figure C, rbcL Pseudosasa Bambus. Chusquea Oryza Leersia Ehrhart. sedis Volume 88, Number 3 2001 443 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae Joinvillea Ehrharta 51 I 61 EtOryza Oryza Olyra Pseudosasa 56 99 88 ~~~~~88 I 82 1 I Ehrhart. Bambus. Babus. Lygeum Nardus Anisopogon Stipa 63 ] Po. Avena Bromus Aristida 67 Stipagrostis Molinia 100 Phragmites Arundo Arundin. Amphipogon 70 g97 99 73 I | -,~~~~~~~~73 I 51 I 63 1000 ' 99-I ~I ~ Merxmuelleram. Karroochloa Austrodanthonia Dant honi. Danthonia Merxmuellerar. Centropodia Chlorid. Eragrostis Pappophorum Spartina Micraira sedis -Incertae Cent ot hec. Thysanolaena Chasmanthium Gynerium Panicum 80 I I Pennisetum 98 I Miscanthus I Zea 93 Appendix III-Figure D, rpoC2 -Incertae an Panic. sedis 444 Annals of the Missouri Botanical Garden Flagellaria Joinvillea Anomochloa Anomochlo. Phar. Streptochaeta Pharus Puelia - Pueli. Eremitis Pariana Buergersiochloa Bambus. Lithachne Olyra Pseudosasa Chusquea Ehrharta jEhrhart. EOryza Lygeum Nardus Anisopogon Nassella Melica Po. Glyceria Brachypodium Diarrhena Avena Bromus Triticum Incertae sedis A id. Arist Streptogyna Aristida Danthonia Molinia - Danthoni. Phragmites Eragrostis J Sporobolus i Thysanolaena ChasmanthiumDanthoniopsis Panicum Pennisetum Miscanthus Zea Appendix III-Figure E, phy B Arundin. Chlorid. Cent ot hec. Panic. Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae 83 g9~ Q ~8 18 Joinvillea Anomochlo. Streptochaeta Pharus Phar. Lithachne Z5-7_ Bambus. Chusquea Ehrharta Ehrhart. R6Q9|Oryza L-- eersia ,83 _-4-Brachyelytrum -Lygeum 90 Nardus Brachypodium Melica 67 Glyceria Anisopogon o 99 Stipa 1Nassella Piptatherum Avena 64 1 Bromus 90 Triticum Ampelodesmos Diarrhena 76 isti Aristida. I--~idStipagrostis I__| - Arundin. Amphipogon 62 1 Spartina 1-00 Chlorid. Chorid. Sporobolus 93 Karroochloa Danthoni. Dant honi. *--Austrodanthonia Arundo Arundin. Merxmuellera m. - Dant honi. Merxmuellera r. - Ch o r id. Danthonia Dant honi. Molinia Arundin Arundn. Phragmites Centropodia Chlorid. Eragrostis ~J Pac Panicum 79 Pa IPennisetum __=__ Miscanthus Thysanolaena Centothec. Chasmanthium Panic. Zea Gynerium sedis Incertae Eriachne Micraira L 19 _ 445 Appendix III-Figure F, ITS Annals of the Missouri Botanical Garden 446 100 | Anomochloa - Anomochlo. Pharus - Phar. Eremitis Pariana Oryza 98 j 77 764 7 Ehrhart. Lygeum Triticum 64 Barmbus. Po. Glyceria Melica Danthoniopsis 61 Pennisetum 63 93 | Panic. Miscanthus Zea 67 100 Karroochloa Danthoni. Austrodanthonia 96 Merxmuellerar. 95 95 Chorid. Chlorid. Centropodia 1 00 Appendix III-Figure G, GBSSI Merxmuelleram. - Danthoni. Chusquea Bambus. Grass Phylogeny WorkingGroup Phylogeny and Classificationof Poaceae Volume88, Number3 2001 97 = Flagellaria Elegia Baloskion Joinvillea Streptochaeta Anomochloa Pharus Pariana Eremitis Chusquea Puelia -Guaduella - -Pseudosasa 80 LI LI 86 447 _ Anomochlo. - Phar. Bambus. Pue I Puel_ -Lithachne Olyra Buergersiochloa Zeugites Streptogyna Oryza Leersia Ehrharta Phaenosperma Brachyelytrum Lygeum Nardus Diarrhena Melica Glycerias Ampelodesmos Stipa _. . .. eru Plitarnerum Bambus Bam b us. Centothec. Incertae sedis Ehrhart. Po. Anisopogon Nassella Brachvpodium Triticumo Triticum QA A Ps:',&#io<a Stipagrostis Phragmites Molinia Amphipogon Aristid. A nd Arundo L 79 . Appendix III-Figure H, structural othec. Thysanolaena -Cent Incertae sedis Gynerium Eragrostis Zoysia Spartina Distichlis Chlorid. Pappophorum Sporobolus Uniola Merxmuellerar. Centropodia - ncertae sedis Micraira Chasmanthium - Cent ot hec. Panicum Pennisetum Miscanthus Panic. Zea Danthoniopsis Merxmuelleram. Karroochloa Danthoni. Dant Damnto/a honi. Danthonia Austrodanthonia - Incertae sedis Eriachne 448 Annals of the Missouri Botanical Garden 1 00 I 99 100 =- I, 1 00 98 00 11 00 87 O0 1-78,- 1 97 100 93 , 99 7 ,90 -n 7 6 00 93 7 Anomochlo. Streptochaeta Phar. Pharus Guaduella 7 Pueli Puelia Eremitis Pariana Lithachne Bambus. Olyra Buergersiochloa Pseudosasa Chusquea Streptogyna - Incertae sedis Ehrharta Ehrhart. Oryza Leersia Phaenosperma Anisopogon Ampelodesmos Stipa Nassella Piptatherum Melica Glyceria Po. Brachypodium Avena Bromus Triticum Diarrhena Lygeum Nardus Brachyelytrum Aristida Aristid A id. Stipagrostis Amphipogon Arundo Arundin. Molinia Phragmites Danthoniopsis - Panic. Thysanolaena nZeugites sa Centothec. Chasmanthiumj Incertae sedis Gynerium Panicum Pennisetum Panic. Miscanthus Zea Merxmuelleram. Karroochloa Danthoni. ___Anomochloa 100 62 Flagellaria Elegia Baloskion Joinvillea 100 99 100 51 P60 51 1 s^l5 -- -a 99 94 73 ~5 ~ ~~0 5 0uiAustrodanthonia 81 -_ 0 1J00 98Q ,I 8100- 100 99 ! 100 86 i78 100 7L -1 00 86 98 Appendix III--Figure I, all chloroplast Danthonia Merxmuellerar. Centropodia Eragrostis Uniola C . Chlorid. Pappophorum Zoysia Spartina Sporobolus Distichlis Micrairaae seis sedis -lncertae Eriachne Grass Phylogeny Working Group Phylogeny and Classification of Poaceae Volume 88, Number 3 2001 I 64 I Flagellaria Joinvillea Anomochloa Anm hlo. Anomochlo. Streptochaeta Pharus Puelia 97 99 Eremitis Pariana 99 Lithachne Olyra -- Phar. Pueli. Bambus. Buergersiochloa 63 - 87 92 Pseudosasa Chusquea Ehrharta Ehrhart. Oryza Leersia Brachyelytrum Lygeum Nardus Anisopogon Ampelodesmos Piptatherum Diarrhena Avena Bromus Triticum Brachypodium Stipa 5 9Q|- 97 50 97 95 98 Glyceria Streptogyna - Incertae sedis Aristida Arist id. Aist Stipagrostis Chlorid. Centropodia Arundo -Arundin. Merxmuellerar. Merxmuelleram. Danthoni. Danthonia Amphipogon - A ru n d in. Spartina Sporobolus Chlorid. 8481 { 8~7~~~~---Eragrostis '~~~7 93 62 r[ r! Karroochloa Dant Austrodanthonia Danthoni. Molinia Phragmites Eriachne Micraira I o6"3 76 Po Nassella Melica 93 57 449 Gynerium-Chasmanthium Thysanolaena Danthoniopsis Panicum Pennisetum Miscanthus 83 I --- Zea Appendix III-Figure J, all nuclear J' A ru ndin. Incertae sedis Centothec. Centhc. Panic. Annals of the Missouri Botanical Garden 450 Flagellaria Elegia Baloskion Joinvillea --Anomochloa Anomochlo. i Streptochaeta Phar. Pharus f C PueiGuad ueli. - Puelia = ' 100 Eremitis -66 z Pariana 1 00 I 100 100 100 100 100 Lithachne | 98 99 Bambus. Olyra Buergersiochloa Pseudosasa Chusquea Incertae sedis Streptogyna |Ehrharta Ehrhart. Oryza Leersia Phaenosperma Anisopogon Ampelodesmos 100 53 70 J 100 1 100 1 1 11 0000 90 87 91 100 1 00 5100 Stipa Nassella Piptatherum 100 Melica Glyceria Brachypodium 93 94 Avena 9f8 =Bromus 98 --|-Triticum 1 n00 Diarrhena 100 Lygeum Nardus Brachyelytrum 100 Aristida 53 ,97i 7. - 100 1 00 79 78 98 Arst id A Stipagrostis Merxmuelleram. Karroochloa Dant honi. Austrodanthonia Danthonia Amphipogon Arundo Arundin. Molinia Phragmites Merxmuellera r. Centropodia [ 00 1FH Eragrostis 3 7 Z 8Uniola Pappophorum 78 83 Po. 9 8 9-"99 84 - Zoysia -98 ' 100 I Spartina Sporobolus Distichlis Eriachne Thysanolaena '-Chasmanthium 90 550 100 Z96 _ 100 Appendix III-Figure K, all molecular Chlorid. Incertae sedis Centothec. Zeugites Panic. Danthoniopsis Gynerium Panicum Pennisetum Panic. Miscanthus Zea -- Incertae sedis Micraira Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae Appendix IV. Notes on morphologicalcharacters. In this section the structuralcharactersare defined and/ or discussed, numbered as in Table 4, and their distributionon the most parsimonioustree is outlined. The behavior of each character on the most parsimonioustree is signified by a series of three numbers (numberof steps, CI, and RI); p/a refers to presence/absence of a character. developed blade, but these are clearly interpretable as losses. 7 (4 steps, CI = 0.25, RI = 0.80). Pseudopetiole p/a: The pseudopetiole is a constriction at the base of the leaf blade. Both states occur in the grasses and in the outgroups. Loss of the pseudopetiole is a synapomorphyof the clade that includes all grasses except Anomochlooideae, Pharoideae, and Puelioideae (i.e., the BEP + PACCAD clade). In the present cladogram, the pseudopetiole is interpreted as secondarily gained in Bambusoideae, Thysanolaena + Zeugites, and Phaenosperma. CULM 1 (4 steps, CI = 0.25, RI = 0.25). Perennating woody culms p/a: Highly lignified, perennial culms are absent among outgroups (except for the score of "uncertain"in Flagellaria) and in most grasses. While most, if not all, grasses produce some lignin in their culms, the distinction between "woody" and "herbaceous" is usually easy to draw, and our scoring was based on this qualitative criterion. Of the taxa in this analysis, presence is an unambiguous and unreversed synapomorphyof Bambuseae, and an autapomorphyof Arundo, Thysanolaena, and Gynerium;Phragmites is scored as intermediate. 2 (14 steps, CI = 0.07, RI = 0.18). Hollow culms p/ a: This character is variable in the grasses (14 steps), and polymorphic in at least five of the sampled genera, and many additional ones. Occurrence of a small pore was scored as intermediate. Solid culms are uncommon and scattered in occurrence in the early-diverginglineages and the BEP Clade, but are frequent in the PACCAD Clade, where they are often associated with C1 photosynthesis. There are no unambiguous synapomorphiesin this analysis and the character is highly homoplasious globally, but, as is well known among grass systematists, for many small groups of genera and species one state or the other of this character likely is a synapomorphy. If the leaf blade is absent, characters 45 and(46 are considered inapplicable, but characters 3, 4, an(l 5 are scored. 3 (3 steps, CI = 0.33, RI = 0.33). ,eaf sheath margins free/fused: Fused margins are an unreversed synapomorphy of Meliceae, and autapomorphiesof Flagellaria and Bromus.Fused sheaths are frequent in Poeae and Aveneae and may provide a tribal or more local synapomorphy,or may be plesiomorphic. Sampling outside the grasses would help establish the point of origin of the free leaf sheath. 4 (5 steps, CI = 0.20, RI = 0.76). Adaxial ligule type: The membranousligule is the most common state in the sample. Transformationto a fringe of hairs is an unreversed synapomorphyof Anomochlooideae, and a synapomorphy of the clade of Eriachne plus its sister group (a set of four subfamilies, Aristidoideae, Danthonioideae, Arundinoideae, and Chloridoideae), although the character reverses multiple times within this group. The ligule as a fringe of hairs also is an autapomorphyof Danthoniopsis. 5 (4 steps, CI = 0.25, RI = 0.25). Abaxial ligule p/a: Most grasses lack an abaxial ligule. Presence is an unreversed synapomorphy of Bambuseae, and an autapomorphy of Puelia, Streptogyna,and Thysanolaena. Abaxial ligules occur sporadically in the PACCAD Clade, and are known in a few Pooideae. 6 (1 step, CI = 1.0, RI = 1.0). Leaf blade p/a: All sampled species have a leaf blade, except in Restionaceae, where loss of the blade is a synapomorphy.A few species in some grass genera, such as Ehrharta, lack a 451 SPIKELET 8 (1 step, CI = 1.0, RI = 1.0). Floret p/a: The floret was defined for the morphological matrix as a unit of the grass inflorescence consisting of a subtending bract ( lemma) enclosing a short axillary axis bearing a flower, the first appendage of which is an adaxial, usually twokeeled bract (= palea). The floret is present only in grasses, but not in all grasses. Of the taxa included in this analysis, the floret is regarded as absent in Streptochaeta and of undetermined status in Anomochloa, based on the uncertain homologies of their floral bracts and the lack of an identifiable palea (Judziewicz & Soderstrom,1989; Soreng & Davis, 1998). These two genera have flowers subtended by well-developed bracts, but not in any configuration that can be compared directly to the above definition. Gain of the floret is interpreted as a synapomorphy of the clade of all grasses except Anomochlooideae. Within the spikelet clade, the palea is absent in a number of taxa, including some species of Agrostis, An(ropogoneae, and, in this analysis, Zoysia. All of these taxa, however, have an identifiable lemma and other congeners have paleas. Following a strict definition of the floret,Zoysia was scored as polymorphicfor this character, although the phylogenetic context shows that a coimplete floret was present ancestrally. 9 (2 steps, Cl = 0.50, RIl = 0.50). Spikelet pairs: Spikelet pairs are infrequent in the grasses, and their origin is a synapomorphyof Andropogoneaein this analysis (but note that Danthoniopsis is regarded as polymorphic), although spikelet pairs are also found in some Paniceae. Their presence may also be an autapomorphyof Pharus, but since Anomochlooideae and non-grasses are not scored for this character, the placement of this transformation is ambiguous (i.e., paired spikelets could be interpreted as plesiomorphic anlong the floret-bearinggrasses). Developmentally, spikelet pairing appears to occur in the same way wherever it appears in the Panicoideae (LeRoux & Kellogg, unpublished obs.), but developmental studies have not been done on Pharus, so spikelet pairing is an inference based on adult morphologyalone. 10 (4 steps, CI = 0.25, RI = 0.0). Pedicel p/a: The pedicel is present in the earliest-diverging grass lineages that have spikelets. Multiple losses occur, but only the autapomorphicloss in Triticumcan be placed unambiguously. Loss of the pedicel may be a synapomorphyof Lygeum + Nardus, but this is ambiguous because this character is scored as ambiguous for Lygeum. 11 (7 steps, CI = 0.14, RI = 0.45). Proximal femalesterile florets: Presence of proximal female-sterile florets is interpreted as an unreversed synapomorphyof Puelioideae and of Panicoideae (including Danthoniopsis, excluding Gynerium),and is a potential synapomorphyfor Ehrhartoideaeexcept that this is coded as ambiguous for Leersia and Oryza. Multiple origins occur elsewhere as 452 Annals of the Missouri Botanical Garden autapomorphies, e.g., in the traditional Aveneae (including Phalarideae), Phragmites, Chasmanthium,Uniola, and Chusquea,and there are no unambiguouslosses once such florets are gained. As noted in the discussion on spikelets, some proximal female-sterile florets may be homologous to glumes, as in Ehrhartoideae or some Bambuseae, whereas others, as in Panicoideae, are clearly derived from reduction of fertile florets. 12 (13 steps, CI = 0.07, RI = 0.47). Numberof femalefertile florets per female-fertile spikelet: The plesiomorphic state among the grasses is one. Increase in the number of female-fertile florets has occurred multiple times, but the placement of these changes is ambiguous, in part because five genera are scored as polymorphic. The only unambiguous synapomorphicreduction from multiple female-fertile florets to one is for the the clade of Zoysia, Spartina, and Sporobolus,although this may not hold up when sampling density is increased. Among unsampled taxa, there are numerous additional transformationsto one floret. 13 (12 steps, CI = 0.08, RI = 0.52). Awn or mucro p/ a: There are multiple origins of awns and mucros on the lemma, mostly of ambiguous placement. Approximately half the genera of the family have awns, so this is another example of a locally useful but globally highly homoplasious character. In Streptochaeta, eleven of the twelve bracts lack awns, whereas one (bractVI) has an awn (Judziewicz & Soderstrom,1989). These bracts have been variously interpreted (Soderstrom,1981), but the ones most closely associated with the flower lack awns, so interpretation of the single awned bract as a lemma is doubtful, and Streptochaetaconsequently is scored as ambiguousfor this character. 14 (4 steps, CI = 0.50, RI = 0.50). Number of awns present: This character is scored for taxa that are polymorphic for character 13, but it is inapplicable when character 13 is scored as state 0 or of questionable homology. Danthonia and Austrodanthoniaare scored as state 3 to reflect the basic pattern in these taxa, in which the lemmas have nine veins, with the central vein forming a median, usually articulated or cork-screwed, awn, and the third vein out from the median on each side forms a hairlike awn or seta at the apex of its respective lateral lemma lobe. This seta varies from an acute lobe to a long hairlike extension. Occasionally the lateral lobes are fused with each other, and presumablywith the base of the central awn. From the basic pattern all kinds of deviations occur, including fusion of the lateral lobes, loss of the setae, loss of the median awn, and fusion of the lateral lobes with the median awn still present as a stout mucro. In this analysis, presence of three awns is an unreversed synapomorphyof the clade of Aristidoideae + Danthonioideae, but this state also occurs in Amphipogonand Anisopogon and in other genera not sampled here such as Plectrachne, Triodia, and Pentaschistis, among others. Presence of numerous awns is an autapomorphyof Pappophorum. 15 (7 steps, CI = 0.28, RI = 0.44). Awn attachment: This character is scored for taxa that are polymorphicfor character 13, but it is inapplicable when character 13 is scored as 0 or of questionable homology.Awn attachment at the apex of the lemma is the most common and widespread state of this character.Attachment in a sinus is an unreversed synapomorphyof Danthonioideae (although it appears to be attached at the apex of the lemma in some instances because of fusion of the lateral lobes with the base of the awn or loss of the lateral lobes; Linder & Verboom, 1996; Linder & Davidse, 1997) and of Merxmuellera rangei + Centropodia.This state also occurs in Streptogyna,Anisopogon, Bromus, and some Panicoideae. Awn attachmenton the back of the lemma is widespread in Aveneae; here it is an autapomorphyof Avena, the only taxon from the tribe in this analysis. Among unsampled grasses, dorsal attachment is known from one genus of Meliceae and a few genera of the PACCADClade including Arthraxon. 16 (6 steps, CI = 0.16, RI = 0.44). Disarticulation above glumes: Glumes are considered to be the two empty bracts subtending the spikelet, and glumes across the spikelet clade were assumed to be homologous even though there is disagreement on this point (see Discussion). This character is not scored in Anomochlooideaeor in the non-grass outgroups, which lack spikelets and therefore glumes, but the presence of this type of disarticulation in at least some members of the Pharoideae (e.g., Leptaspisand Pharus; Soderstromet al., 1987), and in both genera of the Puelioideae, argue that this type of disarticulation is plesiomorphic in the Spikelet Clade. Synapomorphicloss of disarticulation above the glumes occurs in Pariana + Eremitis and in the clade within Panicoideae that consists of Paniceae + Andropogoneae. Additional losses, all autapomorphicor potentially so, occur in Zoysia, Spartina, Phaenosperma,and Melica. 17 (6 steps, CI = 0.16, RI = 0.37). Germinationflap: The germinationflap, a small flap of tissue at the base of the lemma through which the germinatingembryo grows, is derived independently within the PACCAD and BEP Clades. There are unambiguous independent synapomorphic gains of the germinationflap in Aristidoideaeand Olyreae; germinationflaps also are present in some Panicoideae (but placement of the transformationis ambiguous although no Andropogoneae have germinationflaps), and the character is also autapomorphicin Oryza and in Amphipogon. FLOWER Characters 19-22 are scored as inapplicable for taxa that lack lodicules. 18 (4 steps, CI = 0.25, RI = 0.70). Lodicule p/a: Lodicules, as organs that become turgid at anthesis and force open the flower,occur in most grasses and are not present outside the grasses. The evidence that lodicules are modified tepals is not universally accepted. Anomochloa, Streptochaeta,and non-grasses were scored "uncertainhomology"by Soreng and Davis (1998), but these two genera are scored as lacking lodicules in the present analysis. In the Pharoideae,lodicules are present or absent in the male spikelets and lacking in female spikelets (Clark & Judziewicz, 1996). Pharus is scored here as having lodicules. Thus, the first unambiguous occurrence of lodicules is in the clade that consists of all grasses except Anomochlooideae. There is an unambiguoussynapomorphicloss in Lygeum + Nardus and another in Zoysia + Spartina + Sporobolus, plus an autapomorphic loss in Micraira. Among unsampled grasses, 84 taxa lack lodicules; among these, 60 belong to the PACCAD Clade, and 13 to the Pooideae (Watson& Dallwitz, 1992). Lodicules (whatever their origin) might be plesiomorphic for the family, and then lost in the Anomochlooideae. 19 (5 steps, CI = 0.20, RI = 0.71). Number of lodicules: Three lodicules are present at the point of first unambiguous occurrence of lodicules (see char. 18), and this plesiomorphy is retained in Pharoideae and Puelioideae. Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae Bambusoideae and Streptogynaalso have three lodicules, while Ehrhartoideaehave two. Transformationto two lodicules may be a synapomorphyof Ehrhartoideae,but cannot be inferred unambiguously because the earliest-diverging lineages of Pooideae also have two lodicules, as do all taxa of the PACCAD Clade that have been scored. Thus two lodicules may be a synapomorphyof the BEP + PACCAD group, with a reversal to three lodicules in the Bambusoideae/Ehrhartoideaegroup, and re-reversalto two lodicules in the Ehrhartoideae.Alternatively,two lodicules might have been retained in Ehrhartoideae(following the transformationfrom three to two in the origin of the BEP + PACCAD group) while Streptogynaand Bambusoideae independently experienced reversals to three; still other transformation sequences also are possible. Within Pooideae, an unambiguoustransformationfromtwo lodicules to three is a synapomorphy of the clade that includes Anisopogon, Phaenosperma,Ampelodesmos,and Stipeae, but within this group there is yet another unambiguous transformationback to two lodicules in Nassella (while Stipa is polymorphic). 20 (uninformative).Fusion of anteriorpair of lodicules: Lodicules are unfused at the first point at which they are unambiguously present (see char. 18), and in almost all grasses. The anteriortwo are fused in Melica, but the character is polymorphic in Glyceria, so the transformation may be a synapomorphyof this pair of genera, but the precise placement is ambiguous. Elsewhere, Molinia also is polymorphic. 21 (5 steps, CI = 0.20, RI = 0.82). Distally membranous portion of lodicule: The earliest lodicules apparently had a distally membranousportion (see char. 18), and this state is retained in early-diverging lineages. Within the 13EPClade, loss of this membranousportion is an unreversed synapomorphy of Meliceae. Most elements of the PACCADClade lack a distally membranousportionof the lodicule, including various early-diverging lineages, but Micraira, because it lacks lodicules, is not scored for this character.Thus, transformationto this state may be a synapomorphyof the entire PACCADClade, or of the subset that includes all members except Micraira. Within the PACCAD Clade, a distally membranouslodicule in Aristidoideae, Gynerium, and Merxmuellera macowanii implies at least three additional steps in this character, including either three independent transformationsto this lodicule type, or two independent gains including one in the common ancestry of Aristidoideae and Danthonioideae, followed by a loss in the ancestor of Danthonia, Karroochloa,and Austrodanthonia.Polymorphismin Panicum and Leersia implies two additional transformations, a gain of the membranous portion of the lodicule within the former,and a loss within the latter. 22 (3 steps, CI = 0.33, RI = 0.83). Lodicule vascularization: Lodicules were originally vascularized (see char. 18). In the BEP Clade, a single unreversed loss of vascularization is inferred within Pooideae, after divergence of Brachyelytrumfrom the rest of the subfamily,but absence of lodicules in the Lygeum + Nardus group means that this transformationcould have occurredbefore or after divergence of this group from the rest of the Pooideae. In the PACCAD Clade, an independent transformation to faint vascularizationis an autapomorphyof Amphipogon. The state also occurs in Thysanolaena, but the association of the latter with Zeugites, which is not scored for this character, prevents an unambiguousplacement of that transformation. 23 (5 steps, CI = 0.20, RI = 0.66). Posterior stamen of the inner whorl:The posteriorstamen of the inner whorl is present in all outgroups and among the earliest-diverging lineages in the grasses, so its presence is a plesiomorphy for the grasses. Loss of this stamen is a synapomorphy of the BEP + PACCAD clade, but it is regained at least three (possibly four) times in Ehrharta, Oryza(but not Leersia),Pseudosasa, and Pariana. The absence of this stamen in Leersia, coupled with its presence in Oryzaand Ehrharta, is equally consistent with independent gains in the latter two, or a gain in the ancestor of Ehrhartoideae followed by a secondary loss in Leersia. This stamen is not gained elsewhere in the family. 24 (6 steps, CI = 0.16, RI = 0.54). Anterior stamen pair of the inner whorl: Except for the loss of this pair in Anomochloa, while the posterior stamen is retained, the distributions of states of these two characters (23 and 24) are identical. Thus, as with the posterior stamen of the inner whorl, this stamen pair is unambiguouslyinterpreted as plesiomorphically present at the point of origin of the grasses, lost as a synapomorphyof the BEP + PACCAD clade, and regained three or four times in the Bambusoideae/Ehrhartoideae, possibly with a secondary loss in Leersia. 25 (8 steps, CI = 0.12, RI = 0.12). Anterior stamen of outer whorl: This stamen, though absent in Restionaceae, is present in Flagellaria, Joinvillea, and all earlydiverging grass lineages, and thus is plesiomorphically present within the grasses, and lost independently in Restionaceae and various grass lineages. There are seven autapomorphic losses within the grasses (in Streptogyna, Leersia, Eremitis,Glyceria,Piptatherum,Micraira, and Gynerium), plus polymorphisms in Biuergersiochloa,Diarrhena, Eriachne, Zoysia, Sporobolus, and Thysanolaena. The loss in Restionaceae is the only unambiguous synapornorphicloss in the taxon sample. 26 (2 steps, CI = 0.50, RI = (.50). Posterior stamien pair of outer whorl: This stamen pair, like the anterior stamen of the outer whorl, is present in all early-diverging grass lineages arndin all outgroups except Restionaceae. Thus, like the anterior stanmenand like all three stamens of the inner whorl, this pair is plesiomorphically present in the study sample as well as in grasses, and as with the anterior stamen of the outer whorl, all absences are interpreted as losses. Except for a few unscored taxa within the grasses, plus an autapomorphicloss in Chasmanthium and a polymorphism in Sporobolus,presence of this pair of stamens is constant within the grasses. 27 (1 step, CI = 1.0, RI = 1.0). Anthers tetrasporangiate, dithecal, vs. bisporangiate, monothecal: The presence of tetrasporangiate,dithecal anthers is interpretedas a plesiomorphy of the entire taxon sample, and of the grasses. The only transformationto bisporangiate,monothecal anthers is an unambiguoussynapomorphyof Restion- 453 aceae. 28 (12 steps, CI = 0.08, RI = 0.31). Fusion of styles: The presence of one style (see char. 29) can be interpreted as either a fusion or reduction in number, so taxa scored with state 1 of character 29 are scored as ambiguous for character 28. Unfused styles in Flagellaria and both representatives of Restionaceae, and fused styles in Streptochaeta, Pharus, and Puelia together suggest that transformation to the fused state occurred near the origin of the grasses, but polymorphismin Joinvillea and the unknown state in Anomochloa prevent unambiguous placement of this transformation.Reversal to unfused styles is interpreted as an autapomorphyin Guaduella, and as unreversed synapomorphies(i.e., lacking secondary reversion to 454 Annals of the Missouri Botanical Garden fused styles) of Pooideae and Ehrhartoideae (except for polymorphism in Oryza). Within the PACCAD Clade, fused styles appear to be plesiomorphic, with one or more transformationsto the nonfused state, and multiple reversals to the fused state (e.g., in Phragmites, Uniola, and Spartina). 29 (6 steps, CI = 0.33, RI = 0.50). Numberof stigmas: The plesiomorphic state for the grasses, as well as for the entire taxon set, is three, with autapomorphictransformations to two in Baloskion and, among earliest-diverging grass lineages, to one in Anomochloa. Transformationto two stigmas is a synapomorphyof the clade that includes all grasses except Anomochlooideae and Pharoideae (i.e., the Bistigmatic Clade), and there are additional transformations to one in Eremitis and in Lygeum+ Nardus, and to three in Pseudosasa. Another transformationto one appears to occur in Eremitis, but this is the result of a miscoding, as there are actually two stigmas in Eremitis (V. Hollowell, pers. comm.). Puelia is actually polymorphic for this character,having two or three stigmas (the species in this analysis has two). Other polymorphisms(see data table) signify additional transformations.All members of the PACCAD Clade have two stigmas. 30 (8 steps, CI = 0.25, RI = 0.50). Highest order of stigmatic branching: Only one order of stigmatic branching occurs in Flagellaria, Joinvillea, Anomochlooideae, and Pharoideae, and this state therefore appears to be plesiomorphicfor the grasses and for the entire taxon sample, with the various states in Restionaceae interpretable as apomorphic within that family. Transformationto two orders of branching appears to be a synapomorphyfor the clade that includes all grasses except Anomochlooideae and Pharoideae (the Bistigmatic Clade), and this state, once established, is constant in Puelioideae and nearly so in the PACCAD Clade (exceptions being the presence of state one in Zea, and polymorphismin Sporobolus).Within the Bambusoideae/Ehrhartoideae,transformationto state one is an autapomorphyof Streptogynaand a synapomorphy of Pariana + Eremitis,while a transformationto state 3 is an autapomorphyof Leersia,and polymorphismsoccur in unsampled Olyreae. Within the Pooideae, there is an autapomorphic transformationto state one in Nardus, a synapomorphictransformationto state three in Meliceae, and polymorphismin Phaenosperma. 31 (3 steps, CI = 0.66, RI = 0.0). Number of locules: All grasses have one locule with one ovule, while three locules, each with one ovule, are found in Flagellaria and Joinvillea, although there is a strongtendency to abortone or two of the ovules in both of these genera. A transformation from three locules to one may be a synapomorphy of the grasses. This transition has also occurred in Restionaceae, where the basal condition is three locules with numerous reductions to a single functioning locule, as in Elegia in our sample (Linder, 1992a, b). Although not sampled in this study, Anarthriaceae have three locules each with one ovule, but the fruit is 1-seeded, and Ecdeiocoleaceae have two locules, each with one ovule, and the fruit is 1-2-seeded. Thus, reduction in both number of locules and number of ovules that develop in the fruit are common in the Poales; Centrolepidaceae, however, have uniloculate, uniovulate ovaries that are apparently monocarpellary(Dahlgren et al., 1985). synergids are observed to be present only in the four sampled taxa of the Danthonioideae, and absent in all other taxa that have been examined (Verboom et al., 1994). Thus, presence of haustorial synergids is an unambiguous and unreversed synapomorphy of Danthonioideae, but continued investigation is warranted. EMBRYOGENY 32 (1 step, CI = 1.0, RI = 1.0). Haustorial synergids p/a: Data are unavailable for many taxa, but haustorial FRUITANDEMBRYO Characters35-38 describe features of the typical grasstype embryo. They are inapplicable for non-grass genera, which lack the grass-type embryo (i.e., state 0 of char. 34). Data sources include those listed by Soreng and Davis (1998) plus Klak (unpublished). These characterscan be difficult to score, and the literature contains conflicting reports for some taxa. 33 (5 steps, CI = 0.20, RI = 0.76). Hilum shape: This character is recognized as a feature of the caryopsis, and thus is inapplicable for non-grasses. Among grasses, taxa with a short hilum less than one-third the length of the grain are scored as ambiguous. All groups except the PACCAD Clade have a long hilum that is greater than one-third the length of the grain in our sample. This may thus be a synapomorphyof grasses (i.e., the plesiomorphic state for the caryopsis), but because the characteris treated as inapplicable outside the grasses there is no observed transformationat the origin of the family. Nonlinear hila are reported from the Olyreae and the Poeae/Aveneae in the BEP Clade. There are five character transformations within the PACCAD Clade, but lthe placements of two of these are ambiguous. All Centothecoideae and most Panicoideae have a nonlinear hilum that is less than 1/3 the length of the grain, but the hilum is long-linearand greater than 1/3 the length of the grain in Danthoniopsis,and this is interpretedas one origin of this state in the common ancestor of the clade, followed by a reversal in Danthoniopsis. Elsewhere in the PACCAI) Clade, the only unambiguous transformationof this character is as a synapomorphy of Karroochloa + Austrodanthonia,a subset of Danthonioideae.This state also occurs in all taxa of Arundinoideae and Chloridoideae for which there are observations, except in Molinia. Thus, it is interpretableeither as having arisen twice (once in Amphipogon + Arundo, and once in Chloridoideae), or as a synapomorphyof Arundinoideae + Chloridoideae, with reversionto a longlinear hilum in Molinia or in Molinia + Phragmites (not scored for Phragmites). 34 (1 step, CI = 1.0, RI = 1.0). Grass-typeembryop/ a: The grass-type embryo is lateral, peripheral to the endosperm, and differentiated in fruit (Reeder, 1957; Cronquist, 1981; Sendulsky et al., 1987). This embryotype is absent outside the grass family and present in all grasses for which observationsare available (unobservedin Merxmuellera rangei), and thus interpreted as an unambiguous and unreversed synapomorphyof the grasses. 35 (7 steps, CI = 0.14, RI = 0.66). Epiblast p/a: Because this character is inapplicable for non-grasses, its origin is not unambiguously fixed, but available data for the earliest-diverging lineages within the grasses (except Streptochaeta)indicate presence of an epiblast, and thus presence is interpretable as a plesiomorphy within the family. Loss of the epiblast is an unambiguous synapomorphy of the PACCAD Clade, but there are also three autapomorphic losses outside of the PACCAD Clade in Streptochaeta, Ehrharta, and Bromus, plus a polymorphism in Brachypodium.Within the PACCAD Clade the epiblast is secondarily gained in the Centothecoideae(but Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae there is no observation for Thysanolaena) and is a synapomorphyfor either the Chloridoideae or a subset of that clade. The precise point of origin within the Chloridoideae is ambiguous because there is no observation for Merxmuellera rangei or Centropodia.Within the Chloridoideae the epiblast is lost in Uniola. This character is highly homoplasious within the family. 36 (3 steps, CI = 0.33, RI = 0.81). Embryo scutellar tail p/a: The scutellar tail is present in Anomochlooideae and Pharoideae, throughoutthe PACCADClade (wherever observations are available) and in most taxa of the Bambusoid/Ehrhartoidclade, absent in most Pooideae, and unobserved in Puelioideae. Despite the variation just described, plus the inapplicability of this character outside the grasses and the absence of data for some critical taxa, the balance of evidence suggests that the scutellar tail is a plesiomorphyof the grasses. Within the Bambusoid/Ehrhartoid clade, loss of the scutellar tail is unambiguously interpretedas an unreversed synapomorphyof Oryzeae.A more complex pattern is present in Pooideae, where the scutellar tail is usually absent, but is present in Phaenosperma, and there are polymorphismsor conflicting reports for Brachyelytrum,Diarrhena, and Avena. These polymorphisms in the matrixprevent unambiguousoptimizationof the character in Pooideae. Loss of the scutellar tail may be a synapomorphyof Pooideae or of all Pooideae except Brachyelytrum;presence of the scutellar tail in Phaenosperma may be either a unique reversal or a synapomorphy for Phaenospermaand Anisopogon(which is not scored). 37 (2 steps, CI = 0.50, RI = 0.95). Embryomesocotyl internode, negligible vs. elongated: The embryo mesocotyl internode is negligible in length in all early-diverging grass lineages for which scores are available (e.g., Anornochlooideae, Pharoideae),so although this character is inapplicable for the non-grass outgroups the internode is likely to have been negligible at the origin of the grasstype embr-yo.All taxa that have been scored in the PACCAD Clade have an elongated internode. 'Transformation from a negligible to an elongated interrno(e is either a synapomorphyof the entire PACCAD Clade or of the entire clade except Micraira, for which no observation is available. 38 (4 steps, CI = 0.25, RI = 0.85). Embryonic leaf margins meeting vs. overlapping. The margins of the embryonic leaf meet in Anomochloa but overlap in Streptochaeta, Pharus (unobserved in Puelioideae), and earlydiverging lineages of both the BEP and PACCADClades. Thus, although the character is inapplicable in the nongrass outgroups, overlapping leaf margins are plesiomorphic at the origin of the grass-type embryo, with an autapomorphic transformation to margins meeting in Anomochloa, and with parallel synapomorphic transformations in both major lineages. Margins are overlapping in all observed taxa in the bambusoid/ehrhartoidalliance, as well as in Brachyelytrumand Phaenosperma of the Pooideae (also, Diarrhena is polymorphic).Given this distribution, transformationto the margins meeting is interpreted as a synapomorphy of all Pooideae except Brachyelytrum,with a reversal either in Phaenospermaor in the ancestor of Phaenospermaand Anisopogon(there is no observation for the latter). All observations for Panicoideae and Centothecoideae, plus Eriachne, are of leaf margins overlapping, while all observations for the other four subfamilies of the PACCAD Clade are of leaf margins meeting. Thus, transformationto the latter state is an unambiguous and unreversed synapomorphy of the sister group of Eriachne, the clade that includes Aristidoideae, Danthonioideae, Arundinoideae, and Chloridoideae. 39 (uninformative).Endosperm lipid p/a: Observations are unavailable for several taxa. Of those taxa that are scored, including Baloskion, only Avena has lipid in the endosperm. Thus, absence of lipid in the endosperm is plesiomorphic for the grass family and for the taxon set as a whole, with the presence of lipid in Avena an autapomorphy. Among unsampled grasses, all reports of endosperm lipid are from the Poeae-Aveneae. Liquid and semi-liquid endosperm are indicative of the presence of lipid, but "semi-solid" and solid states do not imply absence of lipid (Terrell, 1971; Rosengurtt et al., 1972). 40 (12 steps, CI = 0.33, RI = 0.20). Starch grain syndromes: Scoring here follows Tateoka's (1962) classification with one exception. Tateoka scored Brachyelytrumas having simple Panicum-type grains, but emphasized a major size difference, and we recognize Brachyelytrum-type as a separate state (see also Campbell et al., 1986). Watson and Dallwitz (1992) distinguished between starch grains "simple only" (coded here as [034]) or "compound" (coded here as [12]). Polymorphisms, ambiguity of state delimitation, and lack of observations together preclude unambiguous optimizations of many character-statetransformations in this multistate character, but some patterns are evident. First, the Festuca-type grain (state 1) is present in Baloskion (the only non-grass that is scored) and is widespread in early-diverging grass lineages, among which other types are not observed. This pattern suggests that this starch grain syndrome is plesiomorphic for the grass family and for the taxon set as a whole. The Triticum-type syndromne(state 0) occurs in most "core" Pooid(eae(represented here by Brachypodiunm, Arena, Bromus, and TriticuLm) that are collectively the sister group of Diarrhena, and state ( may lbe either a synapomorphyof this group (reversed, however, in Arena) or a parallelism that arises separately in Brachypodiumand in the ancestor of Bromus + Triticum.The Panicum-type syndrome (state 3) may be a synapomlorphyof Bambuseae, of Stipa + Nassella, and of all Panicoideae except Danthoniopsis, or of the entire PACCADClade. If the latter is true, then there is a reversal to the Festuca-type in the clade of Eriachle plus the set of four subfamilies that is its sister; the Festuca-type is also a potential synapomorphyof Thysanolaena + Zeugites, or an autapomorphyof Zeugites. The Brachyelytrum-type syndrome occurs in Phaenosperma and Brachyelytrum.Available information suggests that Stipa may also have this state (see Soreng & Davis, 1998). 455 SEEDIING 41 (4 steps, CI = 0.25, RI = 0.70). Lamina of first seedling leaf p/a: The lamina of the first seedling leaf is absent in Flagellaria and Anomochlooideae, and present in Restionaceae and Pharoideae, while the character is unobserved in other non-grass taxa and other early-diverging lineages within the grasses. Consequently, optimization of this character is ambiguous in this region of the tree. However, presence of the lamina is unambiguously established by the point of divergence of Pharoideae from the lineage that includes most other grasses; it is present in the PACCADClade and in Pooideae, but is lost twice within the BEP Clade, as a synapomorphyof Oryzeae and as a synapomorphyof Bambusoideae. VEGETATIVEANATOMY 42 (uninformative). Differentiation of leaf epidermal cells into long and short cells: Differentiationis absent in 456 Annals of the MissouriBotanicalGarden Flagellaria, inapplicable in Restionaceae (blades absent), and present in Joinvillea and all grasses (except for a few that are polymorphic).This differentiationis thereforeestablished by the time of divergence of Joinvillea from the grasses, but the point of origin is ambiguous. 43 (6 steps, CI = 0.16, RI = 0.72). Multicellular microhairs p/a: Presence of multicellular microhairs on the abaxial surface of the leaf blades in Joinvillea, Anomochlooideae, and Guaduella, and their absence in Restionaceae, Pharoideae, and Puelia, make it difficult to place the origin (or origins) of multicellular microhairs,but they appear to have been present in the common ancestor of Joinvillea and the grasses, and are inferred to have been lost independently (among the aforementioned taxa) in Pharoideae and Puelia. Multicellular microhairs are universally present within the PACCADClade in our sample, except for an autapomorphicloss in Merxmuellerarangei and a polymorphismin Spartina. The score for M. macowanii was inferred from reports for the rest of the genus; Ellis (1981) reported the absence of microhairs on the abaxial epidermis but did not investigate the adaxial epidermis, where they are most likely to occur. We note, however, that multicellular microhairshave not been detected on the abaxial leaf surface in some species of 40 PACCAD genera, but many of these genera are polymorphicfor this character (Watson& Dallwitz, 1992). In contrastwith their widespread occurrence throughoutmost of the family,multicellular microhairsoccur in only two genera of the Pooideae, Lygeumand Nardus. Either microhairswere lost in the common ancestor of the subfamily,followed by a secondary gain in the common ancestor of Lygeumand Nardus, or were lost twice, once in Brachyelytrum,the other time in the ancestor of all pooids except Brachyelytrum, Lygeum, and Nardus. Some taxa reported to lack multicellular microhairs on the abaxial surface of the leaf blades may have such hairs on the adaxial leaf surface or elsewhere on the plant, particularlylemmas or lodicules, and more detailed examination should be undertaken to verify this. Only the clade within Pooideae that is the sister of Lygeum + Nardus lacks multicellular microhairs entirely, but even in this clade, unicellular microhairs have been reported in several genera of Stipeae (Watson & Dallwitz, 1992). 44 (3 steps, CI = 0.33, RI = 0.71). "Chloridoid-type" microhairs p/a: Tateokaet al. (1959) distinguished microhairs with short and wide apical cells as "Chloridoidtype," as they are mainly restricted to subfamily Chloridoideae. This type of hair is contrasted with "panicoid-type" microhairs, which have relatively longer and thin-walled terminal cells and are widespread among non-Chloridoid grasses. This distinction has been recognized in subsequent studies (Johnston & Watson, 1976; Clayton & Renvoize, 1986; Watson& Dallwitz, 1992). Our scoring of presence vs. absence of chloridoid-typemicrohairs implies that the two types are clearly distinguishable, but in fact there is a continuum of variationbetween them (Van den Borre, 1994; Van den Borre & Watson, 1994; E. A. Kellogg, unpublished data). In future analyses this character should be considered very carefully, as many taxa may actually be intermediate or polymorphic. This character is inapplicable for taxa scored 0 for character 43. Absence of chloridoid microhairs is plesiomorphic for the family. This type of hair is gained in Lygeum, in Amphipogon, and as a synapomorphy in Chloridoideae as traditionallycircumscribed (i.e., excluding Centropodiaand Merxmuellerarangei). This gain in the Chloridoideae is unreversed, except for a polymor- phism in Eragrostis (Van den Borre, 1994; Van den Borre & Watson, 1994). 45 (8 steps, CI = 0.12, RI = 0.46). Arm cells p/a: Arm cells are invaginated chlorenchyma cells that, when present, reach their maximum development in the layer of chlorenchyma beneath the upper surface of the leaf. Variation in arm cells in the Poaceae is known but has not been investigated, and only presence/absence was scored for this analysis. Presence in Anomochloa, Pharoideae, and Puelioideae (including a polymorphismin Guaduella), and absence in Streptochaeta,combined with absence outside the Poaceae, together suggest a first occurrence of arm cells near the origin of the family, but optimizationof the transformationto arm cells at the point of origin of the family is ambiguous. Placement of additional gains and losses is complicated by a widespread occurrence in the bambusoid and ehrhartoidclades, polymorphywithin Ehrharta and Leersia, total absence in Pooideae, and sporadic occurrence in the PACCAD Clade (in Thysanolaena, Gynerium, and Phragmites). This overall distributionis consistent with a variety of optimizations that imply multiple origins and losses. Some taxa with arm cells lack fusoid cells, and vice versa, but the occurrence of both cell types is correlated with broad leaf blades and the forest habitat. 46 (4 steps, CI = 0.25, RI = 0.76). Fusoid cells p/a: Fusoid cells are large, clear, cigar-shaped empty cells that flank each vascular bundle and can occupy up to 30% of the leaf blade volume. Although their function is still unknown, they appear to form gas spaces ratherthan liquid spaces and may play some role in plotosynthesis (Clark, 1991). Fusoid cells are absent outsi(e the Poales and, among the families sampled in this analysis, are also absent in Flagellariaceae. They are present in Joinvillea and all three of the early-diverginglineages within the grasses. Thus, presence of fusoid cells may be a synapomorphyof Joinvillea and Poaceae, but the precise point of origin of fusoid cells is ambiguous in the present analysis, in part because the character is scored as inapplicable in Restionaceae. Elsewhere in the grasses, fusoid cells occur only in Bambusoideae and Streptogyn.a.Fusoid cells are absent in Ehrhartoideae, Pooideae, and the PACCAD Clade, and this pattern is interpretable either as a synapomorphic loss in BEP + PACCAD (followed by secondary gain in Streptogyna and Bambusoideae), or parallel losses in Ehrhartoideae, Pooideae, and the PACCAD Clade. Fusoid-like cells are known in some Paniceae, but these appear to be derived from laterally extended bundle sheath cells and thus are not homologous to fusoid cells. CHROMOSOMES 47 (23 steps, CI = 0.34, RI = 0.44). Base chromosome number: This multistate character varies among taxa, is scored as ambiguous in some, and is polymorphicin others, so the positions of most transformationscannot be reconstructed unambiguously.However, the base number x = 12 in Pharoideae, Puelioideae, the BEP Clade, and the PACCAD Clade, suggests early establishment of this state. If the base numbers 11 (in Streptochaeta)and 18 (in Joinvillea and Anomochloa)are derived from 12, then x = 12 may predate the origin of the grass family, but other reasonable interpretationsalso are possible. Nevertheless, our data support the interpretationthat x = 12 was established prior to divergence of Pharoideae and Puelioideae from the BEP + PACCAD lineage, and that this base number was maintained in some sublineages of both the BEP and PACCADClades. This numberis main- Volume 88, Number 3 2001 Grass Phylogeny Working Group Phylogeny and Classification of Poaceae tained (or re-evolved) in Streptogyna,Ehrhartoideae,and some Bambusoideae, and transformationsto other base numbers occur within the Bambusoideae (e.g., x = 11 in Olyreae). Within Pooideae, the occurrence of x = 10 in Lygeum,Diarrhena, and some Meliceae suggests that this number was established early in the history of this subfamily (with x = 11 in Brachyelytrumpossibly derived from x = 10 as well, but also possibly derived directly from x = 12 in the common ancestor of Pooideae and the bambusoid/ehrhartoidlineage). x = 12 in Phaenosperma, Ampelodesmos,and some Stipeae is interpreted here as a secondary transformationfrom x = 10, but placement of these lineages as early-diverging groups within Pooideae, and the occurrence of x = 12 in other early-diverging lineages, suggests that this state may be a plesiomorphy retained from its original establishment near or before the point of origin of the family. Transformationto x = 7 is a synapomorphyof the clade that includes Brachypodium, Avena, Bromus,and Triticum,which suggests that all other numbers within the tribes represented by this sample of genera (including x = 5 in some Brachypodium,and various numbers in Aveneae and Poeae) are derived from x = 7. Poeae are not sampled in this study, but they would undoubtedly be placed within the x = 7 clade. In the PACCAD Clade, the base number x = 12 occurs in several disparate taxa and is interpretedas the plesiomorphic state of this clade. Among the many other base numbers in the PACCAD Clade, there are a few unambiguous autapomorphictransformations(e.g., to x = 9 in Molinia and Sporobolus), but only one unambiguous synapomorphic transformation,to x = 11, in Aristidoideae. Arundinella, some species of Microstegium, Arthraxon, and in some species of Danthoniopsis, which is here scored as polymorphic. 49 (1 step, CI = 1.0, RI = 1.0). PCK-type carbon fixation, p/a: This syndrome arises once, as an unambiguous transformationfrom the "normal"NAD-ME C4 syndrome in a subclade of the Chloridoideae. Zoysia and Spartina exhibit PCK-type carbon fixation, and Sporobolus is polymorphic for presence/absence of this character. 457 INDELIN PHYTOCHROME B 50 (2 steps, CI = 0.50, RI = 0.92). 3-bp deletion in phytochrome, p/a: The deleted genotype (state 1) occurs in all sampled taxa of the PACCAD Clade, while the undeleted genotype (state 0) occurs in all other sampled grasses. Joinvillea has a 3-bp deletion and Flagellaria a 12-bp deletion in this region. Sesleria (not included in the present taxon sample) also has a 3-bp deletion (Mathews et al., 1995). The deletion occurs in a region of exon I characterized by extensive length and nucleotide variability, and it seems likely that grasses outside the PACCAD Clade have a synapomorphic insertion and that the deletion in Sesleria is apomorphic. Under this reconstruction, the deleted genotype in the PACCAD Clade would be synapomorphic, but the absence of data for Micraira precludes unambiguousoptimizationof the transformation, which is either a synapomorphy of the entire PACCAD Clade (including Micraira)or of the subclade that consists of all taxa except Micraira. CIHIOROPLAST GEN)MESTRUCTURE = 1.0, RI = 1.0). 6.4 kb inversion in 51 CI (1 step, BIOCHEMISTRY the chloroplast genome, p/a: Absence of the inversion in 48 (7 steps, Cl = 0.57, RI = 0.75). Carbon fixation Flagellaria and Baloskioli (no infolrmationfor Elegia), pathways: All of the non-glass outgroups andl all grasses couple(l with presence in Joinvillea and all grasses that outside the PACCAD Clade share the C3 photosynthetic have been sampled, allows unambiguous optimization of pathway.Taxa with all five C, subtypes, as well as addi- this character as a synapomorphy of Joinvillea plus Potional C3 taxa, occur within the PACCAI) Clade. These aceae. 52 (1 step, CI = 1.0, RI = 1.0). trnT inversion in the are intermixed to such an extent that this characterwould l)e homoplasious even if all subtypes of C, carbon fixation chloroplast genome, p/a: Absence of this inversion in Flahad been scored as a single state. One unambliguouspoint gellaria and Joinvillea (no informationfor either genus of of origin of C, carbon fixation is the transformationto Restionaceae in the present study), coupled with presence NADP-ME decarboxylation (state 1) as a synapomorphy in all grasses that have been sampled, allows unambiguof Panicoideae (although recent studies on the phylogeny ous optimization of this character as a synapomorphyof of Panicoideae show that even this is ambiguous(Giussani Poaceae. Although this character has not been scored for et al., in press)). The occurrence of NAD-ME decarbox- Elegia or Baloskion, it has been scored in another genus ylation (state 3) in Panicum is interpretedas a secondary of Restionaceae, Chondropetalum(Doyle et al., 1992), and transformationfrom NADP-ME. Outside of the Panicoi- absence of the inversion in that genus is consistent with deae and the Chloridoideae (discussed below), the occur- the interpretationthat this inversion is a synapomorphyof rence of two different types of C4decarboxylatingenzymes Poaceae. in taxa with and without mestome sheaths (states 2, 3, and 53 (3 steps, CI = 0.33, RI = 0.71). 15 bp ndhF in5, in Aristida, Stipagrostis, and Eriachne, respectively) re- sertion, p/a: This insertion was previously reportedas present in all sampled grasses except Anomochlooideae, quires three additional transformations,but various sequences of transformationamong states can explain the Pharoideae, and Oryzeae (though present in Ehrharteae), variation observed in these three genera. Finally, the pre- and absent outside the Poaceae (Clark et al., 1995). New dominant occurrence of the NAD-ME type of decarbox- sequences reported here confirm this distribution except ylation in Chloridoideae, including Centropodiabut not for the presence of the insertion in Elegia. Thus, the inMerxmuellerarangei (which has C, photosynthesis), sug- sertion arises independently as an autapomorphyof Elegia gests a synapomorphic gain of this syndrome in the an- and as a synapomorphyof the Bistigmatic Clade (all grasscestor of Chloridoideae, followed by a return to C3 pho- es except Anomochlooideae and Pharoideae);it is secondtosynthesis in Merxmuellerarangei, but multiple gains of arily lost (i.e., deleted) in the ancestor of (or within) the the NAD-ME subtype also are possible. Reversal from C4 Oryzeae. Examination of this character in other Restionto C3 also occurs in Eragrostis walteri (van den Borre, aceae and related families such as Anarthriaceae,Ecdeio1994). State 4, the NADP-MEArundinella-type,occurs in coleaeae, and Centrolepidaceae is warranted.