Cladistic Parsimony Analysis of Internal Transcribed Spacer Region (nrDNA) Sequences of Bouteloua and Relatives (Gramineae: Chloridoideae)

The primary goal of the study was to estimate the phylogeny of Boute/oua and relatives (Gramineae: Chloridoideae) employing cladistic parsimony analysis of nuclear ribosomal internal transcribed spacer region (ITS I + 5.SS + ITS2) DNA sequences . Included were Aegopogon (2 of 4 species), Boute/oua (34 of 42) , Buchloe (I of I), Buch/omimus (I of I), Cathestecum (2 of 4), Cyclostachya (I of I) , Griffithsoch/oa ( I of I) , Hilaria (I of 7), Opizia (2 of 2), Pentarrhaphis (2 of 3), Pleuraphis (2 of 3) , Pringleochloa ( I of I), Soderstromia (I of I), and fi ve outgroup genera/species for a total of IS genera, 56 species, and ten varieties . In all , the ITS region of 72 plants was sequenced and analyzed utilizing PAUP. Aegopogon, the Hilaria-Pleuraphis clade, and Tragus (an outgroup representative) formed a tetratomy with a clade containing the remaining ingroup taxa. Neither Bouteloua nor its two subgenera, Bouteloua and Chondrosium, were found to be monophyletic. Bouteloua chondrosioides was sister to Opizia. Bouteloua rigidiseta fo rmed a clade with Buchlomimus and Pringleoch/oa. Boute /oua e/udens formed a clade with Buchloe, Cathestecum, Griffithsochloa, Pentarrhaphis, and Soderstromia. Bouteloua annua and B. aristidoides (subg. Boute/oua) formed a clade with B. eriopoda, B. eriostachya, B. hirsuta, and B. pectinata (subg. Chondrosium). Bouteloua j uncea, which has been included in the B. curtipendula complex, was not a member of that clade. No new circumscriptions were proposed, although recognition of Boute/oua in the broad sense, with Chondrosium reduced to synonymy, was advocated. The findings suggested homoplasy in morphological, anatomical, and breeding system traits.

The principal differences between the subgenera are shown diagramatically in Fig. 1 (see Gould [1980] and Clayton and Renvoize [1986] for additional characters). Species in subg. Bouteloua usually have 7-80 deciduous branches per inflorescence, each 0.8-2 cm long and bearing 1-10(-20) appressed spikelets, whereas Chondrosium species generally possess fewer branches (1-6) that are persistent (spikelets disarticulating at the base of the fertile [proximal] floret), longer (2-5 cm), and bear more (20-100), spreading (pectinate) spikelets (Gould 1980; Clayton and Renvoize 1986). Not all species, however, conform fully to this general characterization. For example, Bouteloua chondrosioides ("like Chondrosium") has been placed by most authors in subg. Bouteloua in spite of its relatively few inflorescence branches (usually 3-8) and " clearly" (Reeder and Reeder 1963b) or " moderately" (Gould 1980) pectinate spikelets. In subg. Chondrosium, B. eriopoda and B. eriostachya are exceptional in having relatively few (8-18) appressed or ascending spikelets. Other species in possession of atypical characteristics were noted by Gould (1980) and Clayton and Renvoize (1986). These authors, however, were not dissuaded by these exceptions, concluding, "the species of Bouteloua comprise two well-defined subgenera" and " divergence within each of the groups has resulted in a slight overlap of characteristics in a few taxa" (Gould 1980), and "though closely related, and with some overlap of individual characters, the two [groups] seem distinct enough" (Clayton and Renvoize 1986). Clayton and Renvoize (1986) placed Bouteloua s.s. and Chondrosium in subtribe Boutelouinae of tribe Cynodonteae. Shown in Fig. 2 are the other genera comprising Boutelouinae, their relationships as suggested by these authors, and numbers of species. The species now treated in the monotypic genera Buchlomimus Reeder, C. Reeder, & Rzed., Cyclostachya Reeder & c. Reeder, and Neobouteloua Gould were transferred out of Bouteloua in the 1960s. Buchlomimus and Cyclostachya were described upon discovery that their constituent species are dioecious and sexually dimorphic (Reeder and Reeder 1963a; Reeder et al. 1965). Reeder and Reeder (1966) also reported "dioecy (or gynodioecy)" in some Bouteloua chondrosioides, but refrained from erecting a new genus because correlated characters were lacking. Five other taxa having unisexual flowers distributed in separate inflorescences are Buchloe Engelm. (nom. cons.), Cathestecum brevifolium (Pierce 1979), Opizia J. Presl, Pringleochloa Scribn., and Soderstromia C. V. Morton. Monoecy (obligate in Opizia bracteata, predominate in Pringleochloa) and dioecy are known to be expressed in these taxa. Markedly dimorphic are the carpellate (pistillate) and staminate inflorescences of Buchloe, Opizia, and Pringleochloa. Interestingly, the staminate inflorescences of these three genera and Buchlomimus and Cyclostachya closely resemble one another and inflorescences of species in Bouteloua subg. Chondrosium (Fig. 1, right), prompting Clayton and Renvoize (1986) to suggest the staminate inflo- rescences "seem to have been undisturbed by the evolutionary pressures which have shaped the female [= carpellate] plants. " In Aegopogon Humb. & Bonpl. ex Willd., Cathestecum J. Presl, Griffithsochloa G. J. Pierce, Hilaria Kunth, and Pleuraphis Torr. (often treated as a subgenus of Hilaria, e .g., Hitchcock 1951; Sohns 1956; Clayton and Renvoize 1986), each inflorescence branch bears three spikelets, the terminal or central spikelet containing a hermaphroditic or carpellate floret (with or without staminate, neuter, and/or rarely hermaphroditic florets above) while the two lateral spikelets (one often not developed in A. bryophilus Doll) usually contain staminate florets (sometimes neuter, rarely hermaphroditic) . . Aegopogon and Pleuraphis, with hermaphroditic central spikelets, are andromonoecious, and Hilaria, possessing a carpellate central spikelet, is monoecious. As described by Pierce (1978Pierce ( , 1979, sexuality in Cathestecum and Griffithsochloa is quite labile; both andromonoecy and monoecy are known in Griffithsochloa and two species of Cathestecum, while C. brevifolium may be monoecious or dioecious and C. varium may be andromonoecious, monoecious, or trimonoecious (staminate, carpellate, and hermaphroditic flowers all present; Cruden and Lloyd 1995). Other than sexuality, the lateral spikelets of Aegopogon, Cathestecum, and Griffithsochloa primarily differ from the central in being smaller and possessing fewer florets (in Cathestecum the first glumes also differ), whereas the spikelets of Hilaria and Pleuraphis are quite dimorphic. Clayton and Renvoize (1986) grouped these genera plus the monoecious/dioecious Soderstromia together in their diagram of relationships (Fig. 2) because all possess three spikelets per inflorescence branch (in Soder-stromia each lateral spikelet is represented by a sterile bract).
Apart from Melanocenchris, distributed from Chad to India and Sri Lanka, and a single collection of Aegopogon from an isolated mountain top in Papua New Guinea (Veldkamp 1985), the genera in subtribe Boutelouinae have a natural distribution in the New World. Ten of the 16 genera and 58 of the 75 species are restricted to North America and the West Indies. With 14 genera (five endemic) and 64 species (27 endemic), Mexico is the center of diversity. Only Aegopogon (two species, one endemic), Bouteloua (nine species representing both subgenera, one endemic), Neobouteloua (one endemic species), and Pentarrhaphis (two species) are represented in South America. The most widespread species is Bouteloua curtipendula, extending from southern Canada to Argentina and Uruguay.
Bouteloua and relatives occur in areas, including deserts, characterized by relatively high temperatures and low precipitation. They are grassland associates or grow in openings in other vegetation types such as scrub and woodland. Bouteloua curtipendula, B. gracilis, B. hirsuta, and Buchloe are common prairie elements on the North American Great Plains; Bouteloua gracilis and Buchloe dominate the western, shortgrass region (Sims 1988). These and many of the other taxa are associated with and sometimes dominate semidesert grasslands, which extend in North America from the southwestern U.S.A. to southern Mexico (Rzedowski 1975;Brown 1982).
In his taxonomic revision "The grama grasses: Bouteloua and related genera," Griffiths (1912) wrote, "it is doubtful whether there is another group of native pasture grasses which is of as much economic impor-tance as this, when both quality and quantity are considered." Stubbendieck et al. (1992) included nine species of Bouteloua and relatives among their 75 most important North American native range grasses. OBJECTIVES In their classification of the world's grass genera, Clayton and Renvoize (1986) explained that "the diagrams are intended to give a visual impression of phenetic relationships, progressing from simple to complex structures; they obviously have phylogenetic implications, but no attempt has been made to treat these rigorously." Nonetheless, Fig. 2 represents the only explicit hypothesis of relationship that has been published for Bouteloua and relatives. In addition, insights into possible intrageneric relationships have been largely confined to remarks concerning a few species. The principal goal of this study, therefore, was to improve upon our limited understanding of the evolutionary relationships among Bouteloua and relatives. Some specific objectives were to ascertain the monophyly of the genera, intrageneric taxa, and informal groups (i.e., Bouteloua curtipendula complex [Gould and Kapadia 1964], B. repens complex [Gould 1969]) and to determine the phylogenetic distribution of the various sexual phenotypes (andromonoecy, monoecy, dioecy, etc.). The method chosen for this investigation was cladistic parsimony analysis of internal transcribed spacer (ITS) region sequences of nuclear ribosomal DNA (Baldwin et al. 1995).

Taxa and Collections
The taxa and collections used in this study are listed in Table 1. Included were 72 collections representing 18 genera (five outgroup), 56 species (five outgroup), and ten varieties. All genera in subtribe Boutelouinae were sampled except Melanocenchris, Neobouteloua, and Schaffnerella. Living material of these three genera was not obtained and attempts to amplify the ITS region from herbarium specimens were unsuccessful. Most species of Bouteloua were sampled; those not sampled included six species that are undisputed members, based on morphology and leaf blade anatomy (Columbus 1996a), of the Bouteloua curtipendula complex of 12 species (Gould and Kapadia 1964), B. quiriegoensis Beetle (scarcely distinct from B. hirsuta), and the South American endemic B. megapotamica (Spreng.) Kuntze. Gould (1980) and Clayton and Renvoize (1986, their Fig. 16) suggested that Bouteloua s.l. is closely related to Chloris; C. virgata, therefore, was chosen as an outgroup species. Other outgroup species, likewise members of subfamily Chloridoideae, included: Cy- Table I. Taxa, collections/vouchers, and origin of collections utilized in cladistic parsimony analysis of ITS region sequences. Asterisks denote those taxa not included in Boutelouinae by Clayton and Renvoize (1986)   the pulverized tissues total genomic DNA was extracted using a 2X CTAB buffer protocol (Doyle and Doyle 1987) with the following modifications: after addition of isopropanol the sample was kept at -20 C overnight to enhance precipitation, centrifuged, and the resultant pellet was washed for 10 min with 5 ml of 76% ethanol containing 10 mM ammonium acetate. Pellets were dried in a vacuum oven and resuspended in 0.2-1.0 ml of 10 mM Tris-HCI and 1 mM EDTA at pH 8.0. For samples taken from herbarium specimens, DNA extraction followed a 2X CTAB microprep protocol (Cullings 1992). All samples were diluted with sterile deionized water to a final concentration of 10 ng/ILI. Most plants were grown in a controlled environment chamber, screenhouse, and/or greenhouse at Rancho Santa Ana Botanic Garden (RSABG) from seed (caryopses) or transplants. Frank Axelrod kindly sent live material of Bouteloua juncea from Puerto Rico. Bouteloua americana was grown from caryopses removed from a specimen recently accessioned at RSA. For some plants, at least 1 g of healthy living leaves and young shoots was removed while in the field, wrapped in aluminum foil, and immediately plunged into a dewar of liquid nitrogen; these samples were transported to RSABG and transferred to a -80 C freezer. Tissues removed from plants growing at RSABG were immediately placed in a -80 C freezer or used directly for DNA extraction. Samples of Bouteloua chasei and two collections of Cyclostachya stolonifera (Columbus 2206 and 2601) were obtained from herbarium specimens. Determinations of all collections were made by the first author and, other than the B. juncea material, vouchers were deposited at RSA (Table 1).

DNA Extraction
About 1 g from each sample was ground to a powder in liquid nitrogen with a mortar and pestle. From The ITS region (ITSI + 5.8S + ITS2) was amplified using the polymerase chain reaction (PCR). Equal proportions of primers ITS4 and ITS5 (White et al. 1990) and 4-40 ng of total genomic DNA were included in each 100 ILl reaction; for most taxa 20 ng of DNA resulted in the best amplification. A PTC-100@) (MJ Research) or RobocyclerC® 96 (Stratagene) was used to carry out PCR: an initial denaturing step of I min at 97 C was followed by 40 cycles of 1 min at 97 C, I min at 48 C, and 2 min at 72 C, and concluded with a final extension step of 7 min at 72 C. The double-stranded PCR product was electophoresed on a 1.5% agarose gel to verify amplification. Purification of the PCR product was accomplished via filtration through Millipore Ultrafree-MC@ filters or a polyethylene glycol precipitation protocol (Morgan and Soltis 1993), followed by resuspension in sterile deionized water. The template DNA was then cyclesequenced using the PRISM@ DyeDeoxy@Terminator Kit (Perkin Elmer) following the manufacturer's recommendations. The primers used for sequencing were ITS2, ITS3, ITS4i, and ITS5i (White et al. 1990; Porter 1997). Sequencing products were read by an Applied Biosystems 373A automated DNA sequencer using a Sequagel-6 polyacrylamide gel (National Diagnostics).
The four sequences obtained per sample were assembled, edited, and a consensus sequence constructed using Sequenche~ version 3.0 (Gene Codes Corporation). The bounds of ITS1, 5.8S, and ITS2 were determined by comparison with sequences in Hsiao et al. (1994) and Buckler and Holtsford (1996; sequences in GenBank). The latter authors included four more bases on the 3' end of ITS2, a decision we followed.

Cladistic Analysis
The consensus sequences were aligned visually and analyzed utilizing Phylogenetic Analysis Using Parsimony (PAUP) version 3.1.1 (Swofford 1993). The entire ITS region, including 5.8S, was analyzed. Characters (nucleotide sites) were treated as unordered, weighted equally, and optimized via accelerated transformation. Gaps were treated as missing data. For a particular taxon, a site having multiple nucleotides was interpreted as a polymorphism. The heuristic option was used to search for all most parsimonious (minimum-length) trees. Starting trees were obtained via random stepwise addition, with one tree held at each step. Tree bisection-reconnection was employed as the branch-swapping algorithm. The steepest descent option was not in effect. Zero-length branches were collapsed. One hundred replicates were performed.
Also computed using PAUP were the character status, pairwise distance matrix, strict consensus tree, consistency index (CI), rescaled consistency index (RC), and retention index (RI). The trees were drawn by PAUP.
To assess support for clades, a bootstrap analysis of 100 replicates was performed employing the same settings as above except that a closest addition sequence was used. In addition, decay indices (Bremer 1988; Donoghue et al. 1992) were calculated following the method of Baum et al. (1994) using a PAUP block appended to the file to automate the procedure (Leigh Johnson pers. comm.).

RESULTS
The aligned sequences are provided in Appendix 1. Alignment necessitated insertion of gaps, resulting in a length of 705 base pairs (bp), 94 (Pleuraphis mutica) to 121 (Soderstromia) bp longer than the unaligned sequences. In some regions alignment was not straightforward; in these cases, an alignment was arrived at that minimized variation. Of the 705 characters, 387 (55%) vary and 324 (46%) are potentially phylogenetic ally informative. Most variable is ITS2 (182 of 266 characters vary, or 68%), followed by ITS 1 (179/ 274,65%) and 5.8S (26/165, 16%). The percentage of variable characters that are potentially informative is nearly the same, ca. 84%, in each of ITS1, 5.8S, and ITS2. The pairwise distance matrix is shown in Appendix 2. Mean distances (proportions of divergent nucleotide sites to total sites, excluding gaps and polymorphic sites) range from 0 between several conspecific samples to 28% (162 sites) between Bouteloua simplex and Hilaria ciliata.
The heuristic search located 4747 most parsimonious trees of 1817 steps (nucleotide substitutions). The trees have a CI of 0.43, RC of 0.31, and RI of 0.73. The strict consensus of these trees is shown in Fig. 3, including bootstrap percentages and decay indices. Figure 4 is one of the most parsimonious trees drawn as a phylogram with branch lengths (numbers of nucleotide substitutions) indicated. As seen in the strict consensus tree (Fig. 3), a tetratomy is formed among Tragus racemosus (an outgroup representative), the Aegopogon clade, the Hilaria-Pleuraphis clade, and a poorly resolved but well-supported clade comprising the majority of the ingroup taxa. With the aim of improving the resolution of this last clade, a second analysis was carried out after removing Aegopogon, Hilaria, Pleuraphis, and the five designated outgroup species. This search found 376 most parsimonious trees of 1293 steps, each tree with a CI of 0.50, RC of 0.39, and RI of 0.77. Figure 5 is the strict consensus tree and Fig. 6 is one of the most parsimonious trees, upon which are mapped the insertions/deletions (indels) identified and enumerated in Appendix 1. Only unambiguous indels shared by two or more taxa are identified; these can be regarded as another measure of support for clades. Rooting of these trees was accomplished by designating Bouleloua kayi and B. trifida as the outgroup; common to these species and Aegopogon, Hilaria, Pleuraphis, and the five outgroup species in the more inclusive analysis are a single base (indel 11) and a string of six bases (indel 25) lacking in the remaining taxa (Appendix 1; Fig. 6).

Aegopogon
Circumscription of Aegopogon has never been disputed and monophyly of the genus is strongly sup-    ported in this study (Fig. 3). Aegopogon is one of six genera, grouped together by Clayton and Ren .. voize (1986), having spikelets arranged in triads. In the ITS region phylogeny, Aegopogon does not form a monophyletic group with any of these genera (Cathestecum, Griffithsochloa, Hilaria, Pleuraphis, Soderstromia). Columbus (1996a) found that Aego .. pogon is unlike these and the other genera in sub .. tribe Boutelouinae in microscopic features of the ab .. axial epidermis of the lemma and in leaf blade tran .. sectional structure. In the strict consensus tree (Fig .  3), Aegopogon, along with the Hilaria .. Pleuraphis

Hilaria-Pleuraphis Clade
Not surprisingly, Hilaria and Pleuraphis form a strongly supported clade (Fig. 3). A close relationship between these taxa, often treated as congeneric, has never been questioned. Clayton and Renvoize (1986) remarked that Hilaria (including Pleuraphis) is "rather isolated but its "spikelet triads strongly suggest a link with the Cathestecum group of genera." Such a link is not evident from the ITS region phylogeny (Fig.  3), save possibly with Aegopogon. Among the most divergent sequences are those of Hilaria-Pleuraphis (Appendix 2) and the branch leading to these genera is by far the longest in the entire phylogeny (Fig. 4), mirroring the numerous morphological synapomorphies that set them apart from the other members of subtribe Boutelouinae. Hilaria and Pleuraphis also differ in basic chromosome number from the other genera in the subtribe (Avdulov 1931;Nielsen 1939;Brown 1950;Gould 1958Gould , 1960Gould , 1965Gould , 1966Gould , 1968Gould , 1980 Reeder 1966Reeder , 1967Reeder , 1968Reeder , 1971Reeder , 1977Reeder , 1984 A significant amount of sequence divergence has also occurred between the sole representative of Hilaria in this study and the two Pleuraphis species ( Fig.  4; Appendix 2), paralleling morphological and anatomical divergence (Columbus 1996a: lemma micromorphology, leaf blade anatomy). In contrast, the Pleuraphis sequences differ relatively little from one another.

Bouteloua
Monophyly of Bouteloua, either s.s. or s.l., is unsupported by the findings of this study ( Fig. 3-6). Bouteloua subg. Chondrosium is also not monophyletic. What follows is discussion of those clades comprised entirely of Bouteloua species. Bouteloua chondrosioides, B. eludens, and B. rigidiseta will be discussed later in conjunction with the genera they form clades with.
Bouteloua kayi-B. trifida clade.-As indicated in the results section, Bouteloua kayi and B. trifida are the only species in the genus that share indels with Aegopogon, Hilaria, Pleuraphis, and the five outgroup species in the more inclusive analysis. The two species constitute a strongly supported clade (Fig. 5, 6). The two plants of B. trifida used in this study, collected in Arizona and Texas, were substantially less divergent in sequence than either one was from B. kayi ( Fig. 6; Appendix 2), a narrow endemic known only from Brewster Co., Texas, lending no support to Correll and Johnston's (1970) assertion that B. kayi is "probably only a form" of B. trifida. These species are atypical in subg. Chondrosium because they possess relatively few spikelets per branch (6-24[-32], Gould 1980), the spikelets are appressed or ascending rather than spreading, and the base of the sterile (distal) floret lacks a tuft of hairs.
Bouteloua gracilis-B. simplex clade.-Bouteloua chasei, B. gracilis, B. scorpioides, and B. simplex, all members of subg. Chondrosium, form a strongly supported clade (Fig. 5). Bouteloua simplex, an annual distributed in both North and South America, is sister to the perennial B. scorpio ides, endemic to central Mexico. A close relationship between these two species is also suggested by morphology and leaf blade anatomy (Columbus 1996a). Bouteloua scorpioides and B. simplex are unique in Bouteloua in consistently developing only one branch per inflorescence.
A single polymorphic site (529) in ITS2 is the only difference between the sequences of the two Bouteloua gracilis plants sampled, from Arizona and Durango (Appendix 2; Fig. 6). The species is distributed from southern Canada to central Mexico. Its sister species is B. chasei, restricted to gypsum soils in northeastern Mexico. Support for the B. chasei-B. gracilis clade is not as strong as that for the B. scorpioides-B. simplex clade (Fig. 5, 6). The infolded, cylindrical leaf blades of B. chasei and B. scorpio ides differ little and share a unique combination of anatomical features (Columbus 1996a).
Bouteloua barbata-B. breviseta clade.-Bouteloua barbata, B. elata, B. parryi, B. breviseta, and B. ramosa are members of subg. Chondrosium and constitute a well-supported and well-resolved clade (Fig. 5). The first three species form a clade sister to a clade formed by the last two species, each clade also strongly supported (Fig. 5, 6).
Bouteloua barbata and B. parryi in turn comprise a strongly supported clade that is sister to B. elata (Fig.  5, 6). Close relationships among these species have been suggested previously. Watson  (Fig. 5, 6). The first four species make up a strongly supported and wellresolved clade that in turn forms a tritomy with B. hirsuta and B. pectinata in the strict consensus tree (Fig. 5). In 78% of the most parsimonious trees, including the tree in Fig. 6, B. hirsuta and B. pectinata form a clade that is sister to the B. aristidoides-B. eriopoda clade.
Bouteloua annua, B. aristidoides, B. eriopoda, and B. eriostachya (originally described as a variety of B. eriopoda) are similar in spikelet orientation and form, including lodicules and the abaxial epidermis of the fertile lemma, and in transectional structure of the leaf blade (Columbus 1993(Columbus , 1996a(Columbus , 1999. The two most widely distributed species, B. aristidoides (southwestern U .S.A., Mexico, South America) and B. eriopoda (southwestern U.S.A., northern Mexico), differ considerably in general appearance. Bouteloua eriopoda usually has 2-6 branches per inflorescence, each 2-5 cm long and persistent (characteristics of subg. Chondro-sium), while B. aristidoides typically has more numerous (7-20), shorter (1-3 cm), deciduous branches (characteristics of subg . Bouteloua). In addition , branches of B. eriopoda usually bear 8-18 spikelets (exceptionally few for subg. Chondrosium) and are usually distichous and ascending or spreading (see Griffiths 1912, Plate 74B), whereas branches of B. aristidoides usually have fewer spikelets (2-10) and, in var. a ristido ides, are frequently pendulous along one side of an ascending or arching main axis (rachis), rendering the inflorescence secund (see Griffiths 1912, Plate 77 A) . The differences between the inflorescences, along with the fact that B. eriopoda is a stoloniferous perennial with pubescent internodes while B. aristidoides is a nonstoloniferous annual with glabrous internodes, have long masked their close relationship and led to the two species being placed in separate subgenera. In the ITS region phylogeny ( . Lemma micromorphology and leaf blade anatomy also suggest a relationship more distant than believed by the above authors (Columbus 1996a). A conspicuous feature of B. hirsuta and B. pectinata, often employed in keys to distinguish these taxa from B. gracilis, is that the inflorescence branch axis is prolonged beyond the terminal spikelet node. The branch axis is also prolonged in the four species comprising the B. aristidoides-B. eriopoda clade. In fact, B. eriopoda, B. eriostachya, B. hirsuta, B. pectinata, and B. quiriegoensis (not included in this study, but unequivocally closely related to B. hirsuta based on morphology and leaf blade anatomy [Columbus 1996aJ) are the only members of subg. Chondrosium that exhibit this feature.
No author has hypothesized that Bouteloua hirsuta or B. pectinata is closely related to any of the species in the B. aristidoides-B. eriopoda clade as is suggested in the ITS region phylogeny. This is for good reason. Other than the prolonged axis of the inflorescence III branch, obvious morphological similarities are lacking (see figures and plates in Griffiths 1912). The two groups also differ markedly in leaf blade transectional structure (Columbus 1996a) .
With regard to the two varieties of Bouteloua hirsuta, the plants used in this study came from the same population. Their sequences were found to be identical except for a polymorphism at a single site (469) in ITS2 of var. hirsuta (Appendix 1). The varieties, recognized by Gould (1980), are distinguished simply by the presence (var. glandulosa) or absence (var. hirsuta) of pubescence on culm internodes. Expressing doubt and citing a mixed collection, Griffiths (1912) nonetheless recognized B. hirticulmis Scribn. (= B. hirsuta var. glandulosa) as a distinct species. We assert that if these taxa are to be recognized at all, it should be at the rank of form. Bouteloua chihuahuana-B . johnstonii clade.-A sister relationship between these species in subg. Bouteloua is strongly supported (Fig. 5, 6). These little-collected species are known only from Chihuahua and Coahuila, respectively. Bouteloua chihuahuana grows on calcareous substrates while B. johnstonii is an obligate gypsophile. The close relationship is also suggested by morphology and leaf blade anatomy (Columbus 1996a).
Bouteloua americana-B. repens clade.-This strongly supported clade is a group of species recognized by Griffiths (1912) and Gould (1969) plus Bouteloua ala-mosana, which Gould (1980) eventually discovered also belongs to this group. The species are members of subg. Bouteloua. Although there is considerable sequence divergence among the species (Fig. 6; Appendix 2), the relationships are unresolved (Fig. 5). In 79% of the most parsimonious trees, including the tree in Fig. 6, this clade is sister to the B. curtipendula clade (discussed below). Columbus (1996a) found the groups to be very similar in microscopic features of the abaxial surface of the fertile lemma.
Bouteloua curtipendula clade.-Gould and Kapadia (1964) recognized 12 species and five varieties in the Bouteloua curtipendula complex (subg. Bouteloua) . Included in our study were six species, four varieties, and a possible interspecific hybrid inferred from morphology. All but B. juncea (discussed below) comprise a strongly supported and well-resolved clade (Fig. 5,  6). Incongruent with morphology, the two varieties of B. uniflora are quite divergent in sequence (Appendix 2) and the species is polyphyletic (Fig. 5). The sequence of B. uniflora var. coahuilensis scarcely differs from that of B. warnockii ( Fig. 6; Appendix 1, 2), although the latter species is morphologically more similar to B. curtipendula. Sequences from the two B. Bouteloua juncea.-This species in subg. Bouteloua, endemic to the West Indies, was suspected by Columbus (1996a), based on morphology (including lemma micromorphology) and leaf blade anatomy, to be misplaced in the B. curtipendula complex (Gould and Kapadia 1964; discussed above). The sequence divergences (Appendix 2) and phylogeny (Fig. 5) lend support to this hypothesis.

Cyclostachya
Clayton and Renvoize (1986; Fig. 2) suggested that the mono typic and dioecious Cyclostachya, endemic to central Mexico, is closely related to the dioecious Buchlomimus and monoecious/dioecious Pringleochloa. The ITS region phylogeny (Fig. 5), however, supports Reeder et al. (1965) who contended, based on morphology of bicellular microhairs on the abaxial surface of the leaf, that Cyclostachya does not appear to be closely related to either of these genera. In the phylogeny, Cyclostachya is situated among members of Bouteloua subg. Chondrosium (all having hermaphroditic flowers), but note its numerous autapomorphies (Fig. 6). Although the staminate and carpellate inflorescences of Cyclostachya consist of a single Chondrosium-like branch, the branches are deciduous, unlike most species of Chondrosium. Also noteworthy is the lack of sequence divergence among the three plants originating from widely separated sites throughout the range of the species (Fig. 6; Appendix 2).

Bouteloua chondrosioides-Opizia Clade
The monophyly of Opizia is strongly supported by the ITS region phylogeny (Fig. 5, 6) Bouteloua), a relationship well supported by a bootstrap percentage of 83, decay index of 4, and a three-base insertion in ITS2 of all three species (Fig. 5, 6; Appendix 1). A close relationship between B. chondrosioides (Arizona, Texas, Mexico, Central America) and Opizia has never been hypothesized. Interestingly, B. chondrosioides is the only species of Bouteloua known to be dioecious (or gynodioecious), although the condition is facultative and without associated dimorphism (Reeder and Reeder 1966). Note also the long branches in this clade.

Bouteloua rigidiseta-Buchlomimus-Pringleochloa Clade
Based on morphology, including copossession of linear bicellular rnicrohairs on the abaxial surface of the leaf, Reeder et al. (1965) suggested that the dioecious Buchlomimus and monoecious/dioecious Pringleochloa are closely related, a hypothesis supported by the ITS region phylogeny (Fig. 5). These two stoloniferous monotypic genera have small distributional ranges in central Mexico, the former known only from the states of Hidalgo and Mexico and the latter from Puebla. The other member of this strongly supported clade is the caespitose, perfect-flowered Bouteloua rigidiseta, a member of subg. Bouteloua and distributed in Oklahoma, Texas, Coahuila, and Tamaulipas. A close relation of B. rigidiseta to Buchlomimus and Pringleochloa has not previously been suggested.

Bouteloua eludens-Buchloe-Cathestecum-Griffithsochloa-Pentarrh aphis-Soderstromia Clade
This strongly supported clade, though backing monophyly of both Cathestecum and Pentarrhaphis, is poorly resolved (Fig. 5). In 1978 Pierce segregated the monotypic and andromonoecious/monoecious Griffithsochloa from the andromonoecious/monoecious/trimonoecious/dioecious Cathestecum. These genera plus the monotypic and monoecious/dioecious Soderstromia, along with Aegopogon, Hilaria, and Pleuraphis (discussed above), constitute a group of genera that Clayton and Renvoize (1986) considered to be related (Fig. 2). These authors also situated Pentarrhaphis (perfect-flowered, although the second floret may be staminate) next to the Old World genus Melanocenchris (not included in our study) and apart from the above genera (Fig. 2), although earlier Clayton and Richardson (1973) wrote that Aegopogon, Cathestecum, and Soderstromia "are clearly related to Melanocenchris Nees and Pentarrhaphis H. B. K., and they to Bouteloua Lag." Bouteloua eludens, a member of subg. Bouteloua having hermaphroditic flowers, has never been postulated to be closely related to any of the other members of this clade. Gould (1980) and Reeder and Reeder (1990) considered B. eludens to be closely related to B. chondrosioides (discussed further below).
Of the members of this clade, the monotypic, monoecious/dioecious, and markedly dimorphic Buchloe has been compared only to Soderstromia because each possesses unisexual flowers (Reeder and Reeder 1963a; Reeder et al. 1965). Inflorescence morphology, however, does not indicate a close relationship. In fact, the staminate and carpellate inflorescences of Buchloe, which are Chondrosium-like and burrlike, respectively, not only differ markedly from each other but from the inflorescences of all the other taxa in this clade.

CONCLUSIONS
Although monophyly of Bouteloua s.l. or s.s. and Chondrosium is not supported by cladistic parsimony analysis of ITS region sequences, no new circumscriptions are proposed herein. This study represents one line of evidence in a larger systematic investigation, and additional phylogenetic estimates, particularly from the nonrecombining chloroplast genome, are needed before taxonomic changes, if any, are proposed. Columbus (1999), however, presents morphological and anatomical evidence corroborating the close relationship between Bouteloua aristidoides and B. (Chondrosium) eriopoda (Fig. 5). He recommends that for now Bouteloua be treated in the broad sense, with Chondrosium reduced to synonymy and no subgeneric divisions, a position we also advocate.
If the ITS region phylogeny provides an accurate estimate of the organismal phylogeny, then some intriguing cases of parallel or convergent morphological, anatomical, and breeding system evolution have taken place, although retention of pleisiomorphic traits is a less parsimonious alternative. For instance, inflorescence branches of Aegopogon, Cathestecum, Griffithsochloa, Hilaria, Pleuraphis, and Soderstromia each bear a triad of spikelets, the central spikelet differing in some manner from the laterals, but Aegopogon, Hilaria, and Pleuraphis appear distantly related to the other genera (Fig. 3). Also, inflorescence branches of Bouteloua hirsuta, B. karwinskii, B. pectinata, species in the B. barbata-B. breviseta and B. gracilis-B. simplex clades, Buchlomimus, Cyclostachya, and the staminate branches of Buchloe, Opizia, and Pring leochloa all bear numerous spreading spikelets (Fig. 1,  right), but these taxa are scattered throughout the phylogeny (Fig. 5). While a close relationship among Bouteloua chihuahuana, B. chondrosioides, B. eludens, and B. johnstonii has been advanced by Swallen (1943), Gould (1980), Reeder and Reeder (1990), and Columbus (1996b), the ITS region phylogeny supports a close relationship only between B. chihuahuana and B. johnstonii (Fig. 5). Bouteloua eludens commonly grows sympatrically with B. chondrosioides and their inflorescences are so similar that the two species can be difficult to distinguish in the field. In addition, the leaf blades of these two species and B. chihuahuana are virtually identical in transectional structure to one another and to the blades of B. hirsuta, B. pectinata, B. quiriegoensis, and Pentarrhaphis scabra (Columbus 1996a). Also similar are the blades of B. johnston ii, Griffithsochloa, Opizia, the two other species of Pentarrhaphis, and Soderstromia. Examination of the ITS region phylogeny (Fig. 5) reveals that these taxa are distributed in four separate clades, suggesting homoplasy at the anatomical level.
Taxa possessing unisexual spike lets are distributed throughout the ITS region phylogeny. The andromonoecious Aegopogon and Pleuraphis, monoecious Hilaria, and dioecious Cyclostachya are more distantly related to the other taxa, which are distributed in three separate well-supported clades (Fig. 3, 5). However, the relationship among these three clades is unclear as evidenced by the weak bootstrap and decay index support for the branches linking the clades in the strict consensus tree (Fig. 5). Interestingly, hermaphroditism, monoecy, and dioecy are expressed in each of these clades. These data suggest, therefore, that spikelet unisexuality and its various manifestations are homoplastic, and that Bouteloua and relatives appear to be predisposed to this condition.