| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics |
2Instituto de Botánica Darwinion, Labardén 200, Casilla de Correo 22, San Isidro B1642HYD, Buenos Aires, Argentina; 3Department of Biology, University of Missouri-St. Louis, 8001 Natural Bridge Rd., St. Louis, Missouri 63121 USA; and 4Department of Biology, University of Saskatchewan, Saskatoon, Saskatchewan S7N 5E2 Canada
Received for publication November 30, 2000. Accepted for publication April 24, 2001.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Our molecular phylogenies are not congruent with previous classifications of tribes or subtribes. Based on this sample of species, we infer that C4 photosynthesis has evolved independently several times, although a single origin with multiple reversals and several reacquisitions is only slightly less parsimonious. The phosphoenol pyruvate carboxykinase (PCK) subtype of C4 photosynthesis has evolved only once, as has the NAD-malic enzyme (ME) subtype; all other origins are NADP-ME. Inflorescence bristles are apparently homologous in the genera Setaria and Pennisetum, contrary to opinions of most previous authors. Some genera, such as Digitaria, Echinochloa, and Homolepis are supported as monophyletic. The large genus Paspalum is shown to be paraphyletic, with Thrasya derived from within it. As expected, Panicum is polyphyletic, with lineages derived from multiple ancestors across the tree. Panicum subg. Panicum is monophyletic. Panicum subg. Dichanthelium, subg. Agrostoides, and subg. Phanopyrum are unrelated to each other, and none is monophyletic. Only Panicum subg. Dichanthelium sect. Dichanthelium, represented by P. sabulorum and P. koolauense, is monophyletic. Panicum subg. Megathyrsus, a monotypic subgenus including only the species P. maximum, is better placed in Urochloa, as suggested by other authors.
Key Words: C4 photosynthesis • ndhF • Panicoideae • Poaceae
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
3300 species in 206 genera (following the Grass Phylogeny Working Group [GPWG], 2001
) and is larger than most angiosperm families. The subfamily is distributed on all continents except Antarctica, and its members are dominant in tropical and warm temperate regions. In addition, it includes some of the world's most important crop plants, such as maize (Zea mays), sorghum (Sorghum bicolor), sugar cane (Saccharum officinarum), common millet (Panicum miliaceum), pearl millet (Pennisetum glaucum), foxtail millet (Setaria italica), and Shama millet (Echinochloa colona).
Panicoideae form a monophyletic group, based on their paired florets, the lower of which is staminate or sterile (Brown, 1810, 1814
), and on distinctive simple starch grains (Tateoka, 1962
; Kellogg and Campbell, 1987
; GPWG, 2001
). Staminate flower development is apparently uniform throughout the subfamily and unique among the grasses. In staminate flowers, gynoecial development ceases after a clear ridge forms around the nucellus (LeRoux and Kellogg, 1999
). This correlates with loss of cytoplasm and nuclei in a small set of subepidermal cells, apparently the result of controlled cell death. This pattern matches that found in maize and Tripsacum, both panicoid grasses, in which the gynoecial cell death is known to be under the control of the product of the gene Tasselseed2 (TS2)(DeLong, Calderón-Urrea, and Dellaporta, 1993
). The cell death pathway is not active in Zizania aquatica, an oryzoid grass that has independently evolved male flowers (Zaitchik, LeRoux, and Kellogg, 2000
). Thus, TS2-induced cell death is apparently restricted to Panicoideae and may be the genetic basis of the subfamilial synapomorphy.
The monophyly of the subfamily (sensu GPWG, 2001
) has been supported by every molecular phylogenetic study to date, including phylogenies of both chloroplast and nuclear genes (reviewed in Kellogg, 1998
; Soreng and Davis, 1998
; support was relatively weak in Cummings, King, and Kellogg, 1994
). In addition, maps of the nuclear genomes of maize, sugar cane, sorghum, pearl millet, and foxtail millet show that they share a common genome arrangement, with linkage group 10 (corresponding to rice chromosome 10) inserted into group 3 and with linkage group 9 inserted into group 7 (Gale and Devos, 1998
; Kellogg, 1998
; Devos and Gale, 2000
). Recent molecular data place the Panicoideae as sister to Gynerium (Arundinoideae), with the panicoid/Gynerium clade sister to a clade consisting of Chasmanthium, Zeugites (both Centothecoideae), and Thysanolaena (Arundinoideae) (GPWG, 2000, 2001
). The genera Eriachne (tribe Eriachneae, now incertae sedis) and Danthoniopsis (formerly Arundinelleae, now Centothecoideae) are formally excluded from the subfamily (GPWG, 2001
; S. Aliscioni, Darwinion Institute, unpublished data).
Bentham (1881)
divided the Panicoideae into six tribes, although more divisions were proposed during the 20th century with the addition of new anatomical and cytological characters (Pilger, 1954
; Hsu, 1965
; Butzin, 1970
). At least seven tribes are commonly recognized, of which by far the largest are Paniceae (101 genera; cf. Clayton and Renvoize, 1986
) and Andropogoneae (85 genera). Other tribes include Arundinelleae (12 genera), Hubbardieae (1), Neurachneae (3), Isachneae (5), and Steyermarkochloeae (1); the last tribe is placed in Arundinoideae by Watson and Dallwitz (1992)
.
The tribe Paniceae, with almost half the genera and 60% of the species of panicoid grasses, is the central phylogenetic problem of the subfamily. There is no evidence that the tribe is monophyletic. It is morphologically diverse and lacks an obvious unifying morphological character. The lemmas are generally, but not always, indurated, and the glumes are often membranous. The inflorescences are commonly described as simple or compound, paniculate or racemose (Clayton and Renvoize, 1986
), although preliminary developmental data suggest that these standard descriptors are inaccurate and insufficient to describe the morphological diversity indicated by comparison of adult morphologies (L. G. LeRoux, A. N. Doust, and E. A. Kellogg, unpublished data). Chromosome number may be based on 9 or 10 or more rarely on 8 (Clayton and Renvoize, 1986
).
Over the years, the systematics of the Paniceae has involved several criteria. For instance, Brown (1977)
divided the Paniceae into four subtribes based on the different photosynthetic pathways among species (see below) and suggested that species with Kranz anatomy have evolved from non-Kranz taxa. Likewise, with the use of morphological features, Clayton and Renvoize (1986)
distinguished seven subtribes primarily differentiated by spikelet characters, such as presence of bristles and texture of the upper lemma.
Panicum, as circumscribed by Zuloaga (1987)
, is the largest genus in the Paniceae; it is highly variable and almost certainly polyphyletic. Zuloaga, Morrone, and Giussani (2000)
subdivided Panicum into groups with apparent synapomorphies and used these groups as terminal taxa in a morphological phylogenetic analysis of the tribe Paniceae; they concluded that Panicum is polyphyletic. In fact, the diversity of this genus encompasses almost all the variation in the tribe. Were it to be treated as a single terminal taxon for a morphological phylogenetic analysis, virtually all phylogenetically informative characters would be polymorphic. Clayton and Renvoize (1986)
depicted Panicum as a large amorphous blob, out of which most other genera in the tribe have arisen. It appears that for any phylogenetic study, the genus must be divided into putatively monophyletic subgenera or sections.
Based on available molecular sequence data, Andropogoneae, the other large tribe within subfamily Panicoideae, forms a monophyletic assemblage (Mason-Gamer, Weil, and Kellogg, 1998
; Spangler et al., 1999
; S. Mathews and E. A. Kellogg, unpublished data). However, its position within the subfamily is not resolved. These same investigations showed that the tribe Arundinelleae is polyphyletic (Spangler et al., 1999
; Kellogg, 2000
).
Diversification of photosynthesis is evident across the grass family, in which C3, C4, and C3/C4 intermediates occur. The C4 species differ anatomically and biochemically, but some combinations of anatomy and biochemistry recur frequently. This led Hattersley and Watson (1992)
to describe ten different structural-biochemical types within grasses, with the three most common combinations of characters being "classical NADP-ME," "classical NAD-ME," and "classical PCK" types. Although many panicoid species use the conventional C3 photosynthetic pathway, a large number exhibit the C4 photosynthetic pathway. The tribe Andropogoneae is entirely C4 "classical NADP-ME," whereas Paniceae includes eight different C4 types, as well as intermediate C3/C4 species (Hattersley and Watson, 1992
).
In C4 plants, enzymes associated with the C3 pathway are produced only in the bundle sheath, whereas enzymes such as phosophoenol pyruvate carboxylase (PEPC) are strongly up-regulated in mesophyll cells (Kanai and Edwards, 1999
; Leegood and Walker, 1999
); all are produced by nuclear-encoded genes. PEPC catalyzes the production of a four-carbon compound, oxaloacetate, from phosphoenol pyruvate (PEP) plus carbon dioxide (in the form of bicarbonate). The four carbon compound is then modified to malate or aspartate and shunted to the bundle sheath, where the CO2 is removed. The CO2 is then supplied directly to Rubisco, which accumulates only in the bundle sheath. The remaining three-carbon compound is transported back to the mesophyll, where it is phosphorylated to regenerate PEP. All C4 plants are alike in replacing Rubisco with PEPC in the mesophyll. They also have closely spaced leaf veins, with every bundle sheath cell in contact with a mesophyll cell. This is clearly necessary for the continual shuttling of substrates between the two different cell types. Species that translocate malate and decarboxylate with malic enzyme using NADP as a cofactor (NADP-ME) have only a single sheath around the vascular bundles. The single sheath is of procambial origin, has a suberized outer wall, and is positionally and structurally similar to the mestome sheath of C3 grasses. This combination of anatomy and biochemistry has been called "classical NADP-ME" by Hattersley and Watson (1992)
, in their typology of C4 variation. Conversely, species that translocate aspartate decarboxylate with either a malic enzyme or PEP carboxykinase (PCK) using NAD as a co-factor (NAD-ME) have a double bundle sheath, the inner part of which is a conventional mestome sheath, while the outer part is parenchymatous and thin walled. Rubisco is expressed in the outer sheath. In NAD-ME species, the outer walls of the bundle sheath cells generally form a regular outline, whereas in PCK species, the outer sheath cells are much less regular in size and shape. These constitute the "classical NAD-ME" and "classical PCK" types of Hattersley and Watson (1992)
. In a few members of Panicoideae, intermediate veins are reduced to lines of isolated bundle sheath cells, called "distinctive cells" by mystified earlier researchers. These were thought to characterize the Arundinelleae and were interpreted by some as intermediates on the way to full C4 anatomy ("Arundinelleae" type sensu Hattersley and Watson, 1992
). It is possible, however, that distinctive cells are actually a loss of intermediate veins, rather than a gain.
Diversity of photosynthetic pathways is as great within the genus Panicum (sensu Zuloaga, 1987
) as within the entire tribe. Some subgenera are photosynthetically homogeneous. For example, both subgen. Dichanthelium and subgen. Phanopyrum are entirely C3. Panicum subgen. Panicum is, anatomically and biochemically, a C4 "classical NAD-ME" type. Although some species within subgen. Panicum are anatomically similar to PCK type, biochemically they use the enzyme NAD-ME. Consequently, these species are included in the NAD-ME "classical PCK" type (Hattersley and Watson, 1992
). Subgenus Megathyrsus is a C4 "classical PCK" type (Zuloaga, 1987
), while Panicum subg. Agrostoides is C4 "classical NADP-ME" type. This photosynthetic type is found not only in all Andropogoneae but also in many genera of Paniceae, including Setaria, Pennisetum, and Echinochloa, to name only a few. Also, P. prionitis and P. petersonii, both in section Prionitia, are biochemically NADP-ME although the outer sheath is still present, which places them in the "Neurachneae type" (Hattersley and Watson, 1992
).
In this study, we have used sequences from the chloroplast gene ndhF to address whether the tribe Paniceae forms a monophyletic assemblage within the panicoid grasses. We also investigated the monophyly and phylogenetic relationships of major genera such as Panicum and Paspalum. We selected ndhF because of its relatively high rate of molecular evolution in the grasses (Clark, Zhang, and Wendel, 1995
). Virtually all the proteins involved in C4 photosynthesis are encoded by nuclear genes (Kanai and Edwards, 1999
; Leegood and Walker, 1999
); a chloroplast gene should thus provide an independent history from the genes selected for C4 photosynthesis. Our study provides new insight into the evolution of photosynthetic pathways and relationships of the major phylogenetic lineages of Panicoideae.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
, Murray and Thompson (1980)
, and Saghai-Maroof et al. (1984)
. Plant tissue was ground in liquid nitrogen and the protocols were scaled down for use with small amounts of either fresh and/or silica gel-dried plant material. In some cases, total DNA was purified with Geneclean III kit (BIO 101, Vista, California, USA). The ndhF gene was amplified via the polymerase chain reaction (PCR) using a Taq-mediated protocol (Promega, Madison, Wisconsin, USA) in several overlapping fragments: 5F/972R, 5F/1318R, and 972F/2110R. For difficult taxa the gene was amplified in smaller fragments, i.e., using primer pairs 5F/536R, 536F/1318R, 972F/1821R, and 1318F/2110R. In total, we used a battery of 10–14 sequencing primers described by Olmstead and Sweere (1994)
with the exception of primer 1821, which was designed by Clark, Zhang, and Wendel (1995)
. The PCR products were cleaned with the QIAquick PCR purification kit (QIAGEN, Valencia, California, USA), quantified by comparison to a low mass DNA ladder (pGEM 25 and 50 ng; Applied Biosystems, Foster City, California, USA) and then labelled with fluorescent dye terminators (Applied Biosystems) during cycle sequencing (10 µL reactions). Both forward and reverse strands were sequenced with a minimum overlap of 90% for every taxon on an ABI 377 automated sequencer using Long Ranger acrylamide gels (FMC Bioproducts, Rockland, Maine, USA). Assembly and editing of sequences used the software program Sequencher, version 3.1 (Gene Codes, Ann Arbor, Michigan, USA). We used the ndhF sequence of Oryza sativa (rice) as a reference for aligning our data. The rice gene is 2205 bp long, occupying coordinate numbers 103 637 (5' end) to 101 433 (3' end), (Hiratsuka et al., 1989
, corrected by Clark, Zhang, and Wendel, 1995
). Sequences were translated to check for stop codons and then manually aligned, preserving the reading frame. Gaps corresponding to indels were mapped onto the final trees to determine whether they were synapomorphies or homoplasies. Other gaps were added when contiguous blocks of sequence (contigs) could not be assembled after several attempts at amplification; these were treated as missing data. Panicum euprepes is incomplete between nucleotide positions 102 643 and 102 318; Chaetium bromoides between 102 294 and 102 203; and the following taxa were sequenced only between nucleotide position 103 580 and the position shown in parentheses: Tatianyx arnacites (102 295), Panicum pedersenii (102 197), Panicum piauiense (101 793), Panicum ovuliferum (101 704), Pennisetum alopecuroides (101 674), and Chasmanthium latifolium (101 643). The aligned data matrix has been submitted to TreeBASE (http://www.herbaria.harvard.edu/treebase) and has also been submitted as supplemental data to the American Journal of Botany website (http://ajbsupp.botany.org/).
Taxonomic sampling
In this study, subfamily Panicoideae was considered the ingroup, including sequences of the tribes Paniceae, Andropogoneae, and Arundinelleae. Delimitation of the subfamily follows GPWG (2001)
, and outgroup selection was based on the GPWG (2001)
phylogeny as well as the grass phylogeny proposed by Clark, Zhang, and Wendel (1995)
. Outgroups included members of the tribes Thysanolaeneae and Centotheceae (Centothecoideae) plus the formerly panicoid genus Danthoniopsis. Tribal classification follows the treatments proposed by Clayton and Renvoize (1986)
, Watson and Dallwitz (1992)
, and Zuloaga, Morrone, and Giussani (2000)
.
In all, 78 sequences of the chloroplast gene ndhF were generated, 76 of 78 within the tribe Paniceae. The remaining two sequences, for Danthoniopsis dinteri (Arundinelleae) and Chasmanthium latifolium (Centotheceae), were generated to verify sequences available in GenBank because we were concerned about possible misplacement of the species in preliminary trees. We used our own sequences in the analyses presented here (supplemental material, http://ajbsupp.botany.org/). Additional sequences from the ingroups, Paniceae (Panicum virgatum and Setaria viridis), Andropogoneae (22 species), and Arundinelleae (two species), and from the outgroup Centothecoideae (three species) were obtained from GenBank; accession numbers are also specified in the supplemental material (http://ajbsupp.botany.org)/.
Effort was made to encompass most of the morphological diversity of the Paniceae (represented by 35 genera), the tribe Andropogoneae (22 genera), and the tribe Arundinelleae (2 genera). Our sample included 19 species of Panicum, representing 14 sections in 5 subgenera; throughout this paper we follow the classification of Zuloaga (1987)
in discussing Panicum. We also included 9 species of Paspalum, representing 7 informal taxonomic groups; and 7 species of the large genus Setaria. One of the Panicum species, listed by Zuloaga (1987)
as P. maxima in subgenus Megathyrsus, is listed in Table 1 as Urochloa maxima, following Webster (1987)
.
|
) with all characters equally weighted and gaps scored as missing data. Overall, 3.5% of the data matrix cells were scored as gaps. Separate analyses using implied weights (Goloboff, 1993
) were run in Pee-Wee version 3.0 (Goloboff, 1997b
) using the same search strategies as in NONA. This reduces the number of trees by reducing the influence of homoplastic characters, which are downweighted in proportion to their number of extra steps (homoplasy). The weighting is based on a concave function, with six different concavities available in the program; 6 is the mildest and 1 the strongest weighting function (Goloboff, 1997b
). Searches were done using K = 1, K = 3, and K = 6. In weighted and unweighted analyses, uninformative characters were discarded using the "pack" command. All informative characters were considered unordered, and both the "amb-" (resolve clades only if they have unambiguous support) command, and "poly=" (polytomies allowed) command were used. Searches were performed using "mult*3000," which randomizes the order of taxa, creates a Wagner tree, and submits it to branch-swapping by tree-bisection reconnection (TBR). It stores up to 20 most-parsimonious trees in memory and repeats the process 3000 times. The shortest trees retained from the subsearches were then TBR swapped to completion with the "max*" command. To estimate the relative stability of individual clades and overall topology of the cladograms, strict consensus trees were generated from the most-parsimonious trees obtained from the NONA and Pee-Wee analyses. Data with equally weighted characters were also analyzed in PAUP*4.01b (Swofford, 1998
) with 1000 random addition sequences and no branch swapping; these found 208 islands of equally parsimonious trees. These trees were then used as starting trees for a heuristic search with TBR branch swapping; the memory limit was reached at 19 500 trees. The strict consensus of these 19 500 trees was then used as a negative constraint tree.
To assess the relative support for clades found in each analysis, bootstrap analyses (Felsenstein, 1985
) were performed with PAUP* version 4.01b for UNIX or for Macintosh Power PC (Swofford, 1998
) with 1000 replicates in a heuristic search using random taxon entry followed by TBR branch swapping (MULTREES). Constrained analyses were performed with NONA to calculate the number of additional steps it would take to make a monophyletic group. To perform these analyses we used a tree with a fixed monophyletic group as starting point (using the "force" command) and carried out a branch-swapping search on the initial tree ("max/" command) to look for trees with highest fit. To test for significant differences between constrained and unconstrained trees, we did Templeton tests (Templeton, 1983
), as implemented in PAUP*4.01b.
Haploid chromosome numbers, leaf structure, and physiological characters related to photosynthetic pathways were obtained from the literature (Table 1) at the generic and specific levels. Although these characters were not included in the analyses, they were added to the matrix and were unambiguously optimized on one of the most-parsimonious trees using Winclada Beta, version 0.9.9 (Nixon, 1999
) after the analyses. Optimization of these characters allowed us to look for evolutionary patterns and degree of relationships among the different lineages.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
|
The phylogenetic analysis with equally weighted characters (NONA) found 27 128 equally parsimonious trees of length (L) = 1472, consistency index (CI) = 0.43, and retention index (RI) = 0.77, excluding uninformative sites; the analysis reached completion after 35 h when running max*, mult*3000. One most parsimonious tree from the equally weighted (NONA) analysis, with branch lengths and bootstrap values, is shown in Fig. 1.
|
|
Panicoideae is divided into three well-resolved and strongly supported clades, corresponding largely to groups having the same basic chromosome number (Andropogoneae [x = 10], Paniceae [x = 10], and Paniceae [x = 9]). The relationship among these clades is uncertain. Trees with equally weighted characters place Andropogoneae sister to x = 10 Paniceae (Fig. 1), although there is no bootstrap support for this relationship. Constraining the trees from equally weighted characters to make Paniceae monophyletic added only one step; a Templeton test indicated that the difference was not significant (P < 0.706). Trees with characters under implied weight indicate that the Paniceae is monophyletic, with the x = 10 clade sister to the x = 9 clade (Fig. 2B). Thus, we cannot be certain whether the Paniceae is monophyletic or not.
The tribe Andropogoneae is strongly supported in all analyses with 17 nucleotide substitutions and 100% bootstrap value (Fig. 1). Analyses with equally weighted characters place Arundinella hirta (x = 7, 10, 12) (formerly Arundinelleae) as sister to the tribe, as found by previous authors (Mason-Gamer, Weil, and Kellogg, 1998
; Spangler et al., 1999
; Kellogg, 2000
), although not well supported by bootstrap analysis (57%). Analyses with implied weights, however, place A. hirta in the Andropogoneae and the clade Tripsacum-Elionurus-Zea appears as sister group to the tribe (Fig. 2B). Most members of Andropogoneae are x = 10, but some species, i.e., Coix and Sorghum are x = 5; Cleistachne and Coelorachis are x = 9.
The x = 9 Paniceae are clearly monophyletic (98% bootstrap, seven nucleotide substitutions; Fig. 1). Although most members of this clade have a chromosome number of x = 9, there are a few exceptions, such as Chaetium (x = 13), a few species of Setaria (x = 10), and a few species of Urochloa (x = 8) (Table 1). Within the x = 9 Clade, Digitaria is represented by three species and forms a robust group (100% bootstrap, branch length 29) (Fig. 1). The species investigated also have a unique 18 bp deletion (coordinate number 102 114, Fig. 1; Table 2). The position of the Digitaria clade is not stable; it is basal within the x = 9 clade in unweighted analyses, but results from implied weight analysis place Digitaria as sister taxa of the "Setaria/Urochloa/Panicum clade" (see below for description of this clade). This affects inferences about character evolution (see below).
Acroceras, Echinochloa, Lasiacis, Oplismenus, Panicum ovuliferum (subg. Dichanthelium, sect. Cordovensia), and Pseudechinolaena form a moderately well-supported clade (82% bootstrap, four nucleotide substitutions). Because all the taxa normally have lanceolate leaf blades and are associated with forest shade environment (Davidse, 1978
; Clayton and Renvoize, 1986
; Zuloaga, Morrone, and Saenz, 1987
), we call this the "Forest Shade Clade." Within this clade, Echinochloa, represented here by E. colona and E. frumentacea, forms a robust monophyletic unit (100% bootstrap, 25 mutations; Fig. 1). Pseudechinolaena, Lasiacis, and Oplismenus form a well-supported and constant group within the Forest Shade Clade (93%). The first two species are also linked in all trees, with 94% bootstrap value.
Panicum millegrana (subg. Phanopyrum, sect. Monticola) and Sacciolepis indica are sister taxa (84% bootstrap, four molecular synapomorphies; Fig. 1). Panicum koolauense and P. sabulorum (subg. Dichanthelium, sect. Dichanthelium) are also strongly supported as sisters (100% bootstrap, 18 molecular synapomorphies).
The remaining species of the x = 9 Clade fall into a single well-supported group (88% bootstrap, with 6 bp substitutions), here called the Setaria/Urochloa/Panicum clade. All species in this clade exhibit C4 photosynthesis. The clade is divided into three strongly supported subgroups, each of which represents a single C4 subtype (Fig. 2). All members of the clade containing Chaetium, Eriochloa, Panicum maximum (= Urochloa maxima), Melinis, and Urochloa (98% bootstrap, eight mutations) use PEP carboxykinase as a decarboxylating enzyme, whereas all members of Panicum subgenus Panicum (98% bootstrap, nine mutations) use NAD-malic enzyme. The clade, including species of Cenchrus, Panicum bulbosum (subg. Agrostoides, sect. Bulbosa), Paspalidium, Pennisetum, Setaria, and Stenotaphrum (100% bootstrap, and nine changes) uses the NADP-malic enzyme.
Several major genera (Urochloa, Setaria, and Pennisetum) are paraphyletic. Five species of Setaria (S. macrostachya, S. parviflora, S. palmifolia, S. sphacelata, and S. geniculata) form a monophyletic subclade (95% bootstrap; Fig. 1), but the remainder are placed together with Paspalidium, Stenotaphrum, and Panicum bulbosum, or are basal to the Setaria clade. Pennisetum is paraphyletic but forms a monophyletic assemblage with Cenchrus ciliaris. The latter species is treated as Pennisetum ciliare by Pohl (1980)
and Hitchcock (1951)
among others.
The x = 9 Clade includes members of five subgenera of Panicum (subg. Panicum [C4], Dichanthelium [C3], Agrostoides [C4], Phanopyrum [C3], and subg. Megathyrsus [C4], sensu Zuloaga [1987
]). Our results show that these subgenera are unrelated to each other. Subgenus Panicum is strongly supported as monophyletic, but subgenus Dichanthelium is polyphyletic. Subgenus Dichanthelium sect. Dichanthelium (represented by P. sabulorum and P. koolauense) is monophyletic and is unrelated to subgenus Dichanthelium sect. Cordovensia (represented by P. ovuliferum). Section Dichanthelium is sister to the large clade of C4 species, the Setaria/Urochloa/Panicum clade, or basal to the x = 9 Clade, depending on whether the data are unweighted or weighted, respectively. Panicum ovuliferum, on the other hand, appears more closely related to Echinochloa (Fig. 1). Subgenus Megathyrsus includes only Panicum (= Urochloa) maximum; this species falls in the Urochloa clade and is unrelated to other species of Panicum. Subgenus Agrostoides is represented by P. bulbosum, which falls in the Setaria clade, and subgenus Phanopyrum (represented by P. millegrana) is sister to Sacciolepis.
The x = 10 Paniceae form a robust clade (100% bootstrap, 12 nucleotide substitutions) that includes taxa with a base chromosome number of ten (Fig. 1). The only known exception is Streptostachys ramosa with x = 9. The x = 10 Paniceae share a 6-bp insertion (coordinate 101 735) that represents a synapomorphy of the clade. Thysanolaena contains a 6-bp insertion at the same position but the inserted nucleotides are completely different (Table 2).
Phylogenetic relationships among the terminal taxa within the x = 10 Paniceae are not fully resolved, but many small groups are well supported. A modest 72% bootstrap value and three base substitutions support the inclusion of Altoparadisium, Arthropogon, Homolepis, Mesosetum, Panicum euprepes (subg. Phanopyrum, sect. Lorea), P. prionitis (subg. Agrostoides, sect. Prionitia), Streptostachys ramosa, and Tatianyx in a single clade. Members of the clade have little in common in terms of morphology or ecology, so we refer to them as the "Ambiguous Clade." Several species pairs are strongly supported as sisters, but relationships among the pairs are unclear. In all cases, Homolepis (C3) is monophyletic, with two species represented (H. glutinosa and H. isocalycia) (99% bootstrap and nine common mutations). Arthropogon is paraphyletic; A. villosus and Altoparadisium are in a highly supported monophyletic clade (100% bootstrap, 18 nucleotide substitutions), while Arthropogon lanceolatus is together with Panicum euprepes and P. prionitis in a well-supported clade (87% bootstrap, 4 bp substitutions). Mesosetum and Tatianyx are sister taxa with ten mutations and 99% bootstrap support.
Panicum laxum (subg. Phanopyrum, sect. Laxa), Steinchisma hians, Plagiantha, Hymenachne, Otachyrium (all C3), and Leptocoryphium (C4), form a second well-supported clade (89% bootstrap, five mutations; Fig. 1) within the x = 10 Paniceae. Internal branches in this clade are strongly supported in bootstrap analyses, and Leptocoryphium is basal. Species of the Hymenachne to Steinchisma clade share a 9-bp deletion (coordinate 101 739), which removes 4 bp of the x = 10 inserted sequence plus an additional 5 bp.
The third clade within the x = 10 Paniceae clade includes Axonopus (C4) and Paspalum (C4), in addition to Anthaenantiopsis (C4), Echinolaena (C3), Ichnanthus (C3), Ophiochloa (C4), Panicum obtusum (C4) (subgenus Agrostoides, section Obtusa), P. piauiense (C3) (subg. Phanopyrum, sect. Stolonifera), Thrasya (C4), and Streptostachys asperifolia (C3) (80% bootstrap, four substitutions; Fig. 1). Ichnanthus is the sister of this clade in one of the two possible topologies. There are three major lineages common to all topologies: (1) Echinolaena and Panicum piauiense (89% bootstrap, four mutations, C3); (2) Axonopus, Ophiochloa, and Streptostachys (95% bootstrap, seven synapomorphies, C3 and C4). Within this clade, the two representatives of Axonopus, A. anceps and A. fissifolius, are placed with the monotypic genus Ophiochloa in a very strong clade (100% bootstrap, 18 mutations, C4); these three taxa share a 6-bp insertion (coordinate number 102 065, Fig. 1, Table 2). (3) Panicum obtusum, Anthaenantiopsis, plus all species of Paspalum and Thrasya (100% bootstrap, 12 mutations, C4). Paspalum forms a large paraphyletic group in which Thrasya is embedded (93% bootstrap, four mutations); however, Thrasya, represented by T. petrosa and T. glaziovii, is monophyletic (85% bootstrap, three mutations).
The Panicum species investigated fall in both the x = 9 (Panicum subg. Panicum, Panicum sect. Dichanthelium, P. millegrana, P. bulbosum, and P. ovuliferum) and x = 10 clades (P. euprepes, P. laxum, P. obtusum, P. prionitis, and P. piauiense). Forcing Panicum to be monophyletic in the traditional sense, including all the species under study, costs 116 extra steps relative to the most-parsimonious trees and 101 steps excluding Panicum (= Urochloa) maximum; both constraints are significantly different from the most-parsimonious topology (P < 0.0001).
Correlation with photosynthetic types
Optimization of photosynthetic pathway differs between the trees with equal weights and those with implied weights (Fig. 2A, B); the differences between the trees indicate our uncertainty about some aspects of photosynthetic evolution. Using the implied weights topology, the common ancestor of the panicoid clade appears as C3 (Fig. 2B); in this case there are eight or nine origins of the C4 pathway, with the exact number depending on the resolution within the Ambiguous Clade. Using the topology retrieved from the equally weighted analysis, the ancestral state is ambiguous (Fig. 2A). Multiple independent origins of the C4 pathway are possible, but it is equally parsimonious to postulate a single origin of the C4 pathway in the common ancestor of the subfamily. In the latter case, the single origin would be followed by multiple losses (reversals to C3) and then several reversals of the reversals to arrive at the current C4 condition. We prefer the hypothesis of multiple independent origins because it seems simpler than a hypothesis involving gain-loss-gain of C4.
When all C4 species of the subfamily Panicoideae were forced to be monophyletic, the shortest trees were 85 steps longer than the unconstrained trees with C3 species basal on the tree; if Danthoniopsis was excluded from the subfamily, and the remaining C4 species were constrained to form a clade, the trees were 73 steps longer. In both cases, the Templeton test (1983)
indicated that the difference was highly significant (P < 0.0001). Within the C4 constraint clade, species were grouped principally by basic chromosome number and several minor clades were still recognized. When the tree was constrained to match the photosynthetic classification of the Paniceae proposed by Brown (1977)
(i.e., C3, C4 NADP-me, C4 NAD-ME, and C4 PCK species in four different clades), 75 extra steps were necessary to keep this hypothesis, and 72 steps if Andropogoneae and Arundinella hirta were included within the C4 NADP-ME group, Subtribe 1. The molecular evidence thus argues strongly that the C4 pathway is highly homoplasious.
Assuming that the Panicoideae are ancestrally C3, C4 photosynthesis has originated once at the base of the Andropogoneae. There are at least four origins in the x =10 Paniceae clade (Fig. 2): (1) the clade including Anthaenantiopsis, Panicum obtusum, Paspalum, and Thrasya; (2) the Axonopus-Ophiochloa clade; (3) Leptocoryphium; and (4) the Ambiguous Clade. Optimization of C3/C4 characters on the basal branch is ambiguous. In the consensus tree from the equally weighted analysis, well-supported groups in the Ambiguous Clade collapse on the ancestral branch (Fig. 2A). Optimization over the two possible resolutions for this clade shows three independent origins for the C4 pathway. In analyses under implied weights, the ancestral state of this clade is also ambiguous, and two or three origins are possible (Fig. 2B). In addition, Steinchisma hians is a C3/C4 intermediate (Brown and Brown, 1975
; Morgan and Brown, 1979
; Morgan, Brown, and Reger, 1980
; Brown et al., 1985
) and here is clearly derived from C3 ancestors; there is no evidence that it represents a transition from C3 to C4.
Optimization of C4 in the x = 9 Paniceae depends on the position of the Digitaria clade, which is basal in the equally weighted analyses (Fig. 2A), and embedded in the clade in analyses under implied weights (Fig. 2B). In the former analyses, there are three independent C4 origins: (1) Digitaria; (2) Echinochloa; and (3) the large Setaria/Urochloa/Panicum clade. In analyses with implied weights, C4 originates only twice in the x = 9 species—once in Echinochloa and once in the large clade of Digitaria plus the Setaria/Urochloa/Panicum clade.
Most C4 Panicoideae use NADP-malic enzyme as a decarboxylating enzyme. Most NADP species have lost their outer bundle sheath and form agranal chloroplasts in the mestome sheath, located over the outer wall in centrifugal position; these characters thus appear in each origin of NADP-ME C4 (Fig. 2). Two exceptions are Panicum prionitis, which has retained an outer bundle sheath similar to that in C3 species (Zuloaga, Morrone, and Dubcovsky, 1989
), and Anthaenantiopsis rojasiana, with remnants of the outer sheath represented by one or two globose cells (Morrone et al., 1993
). A few NADP-ME panicoids, including Arundinella hirta and Altoparadisium chapadense sampled here, have isolated bundle sheath cells in the mesophyll instead of minor veins; this anatomy has thus originated multiple times independently.
The only panicoids in this analysis that use NAD malic enzyme are species of Panicum subg. Panicum, which is strongly supported as monophyletic. Similarly, all species using PEP carboxykinase (PCK-type) are in a single well-supported clade. Each of these C4 types thus originated only once. Both NAD-ME species and PCK species retain their outer bundle sheaths, a plesiomorphic character shared with their C3 ancestors. Centripetal chloroplasts in the outer bundle sheath is a derived character that appears three times on internal branches of subgenus Panicum. The order of evolution of the three C4 subtypes is uncertain and depends on the type of analysis done. However, if Digitaria is basal in the x = 9 Clade, then the ancestral decarboxylating enzyme is ambiguous (Fig. 2A). If Digitaria is sister to the Setaria/Urochloa/Panicum clade, a clade that includes the three C4 subtypes, NADP-ME is optimized as ancestral (Fig. 2B). In this case, NAD-ME and PCK are derived from ancestors with the NADP-ME subtype.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
and Gómez-Martínez and Culham (2000)
, based on the chloroplast gene trnL. Gómez-Martínez was the first to establish that Paniceae are divided into two clades corresponding to chromosome number; these she correlated with biogeography and called "the American Paniceae" (x = 10) and "the Pantropical Paniceae" (most of the x = 9 species). Because the taxonomic sample presented here is biased toward the New World, additional samples from Africa, Asia, and Australia will be necessary to determine whether the distributional hypothesis is supported.
The sample of species presented here is considerably larger than any molecular study to date, but is still lacking several small tribes. We did not have material of tribes Isachneae, Steyermarkochloeae, Hubbardieae (a total of 112 species, all C3), nor of Paniceae subtribes Neurachninae (10 species, C3 and C4) or Spinificinae (8 species, C4). Our sample of 59 genera represents slightly more than one-fourth of the total and our sample of species represents 3% of the total. The results presented here thus reflect whatever limitations are imposed by the set of sequences; data collected since these analyses, however, continue to support the results outlined here (J. Barber, A. Doust, L. Giussani, K. Hiser, and E. Kellogg, University of Missouri-St. Louis, unpublished data).
Zuloaga, Morrone, and Giussani (2000)
produced a morphological phylogeny of the Paniceae including >100 genera of the tribe. Their analysis could not test the monophyly of Paniceae because Andropogoneae were not included. The two basic chromosome numbers (x = 9 and x = 10) have several independent origins on the morphological tree rather than correlating with major clades, as in the molecular tree. (Note that chromosome number was not used in the phylogenetic analysis of Zuloaga, Morrone, and Giussani [2000]
due to the lack of information in many genera.) The morphological phylogeny also did not clearly correlate with biogeography. The morphological phylogeny places C3 species at the base of the tree and the C4 NADP-ME genera in a derived monophyletic clade (Zuloaga, Morrone, and Giussani, 2000
). This contrasts with the ndhF phylogeny, in which C4 NADP-ME appears in multiple independent clades.
Both molecular phylogenies (ndhF and trnL) and the morphological phylogeny of Zuloaga, Morrone, and Giussani (2000)
have several clades in common. All studies find that Panicum subg. Panicum is monophyletic, as is the group of genera grouped into the Setaria clade (Setaria, Cenchrus, Pennisetum, Panicum bulbosum, Paspalidium, and Stenotaphrum, the latter two genera not included in the trnL analysis). In the ndhF and the morphological phylogenies Acroceras and Lasiacis are part of the same clade, but are linked with Cyrtococcum and Microcalamus in the morphological study; these taxa were not included in the molecular data set. The clade consisting of Eriochloa, Urochloa (Brachiaria in Gómez-Martínez and Culham, 2000
), and Panicum (= Urochloa) maximum appears in the three phylogenies, although the molecular phylogenies also place Melinis and Chaetium (in the ndhF trees) within the group. Echinolaena and Panicum piauiense (subg. Phanopyrum sect. Stolonifera) are closely related in the studies, although the genera with which they are associated differ.
Both Zuloaga, Morrone, and Giussani (2000)
and the molecular data identify Panicum as a polyphyletic genus; within Panicum, subg. Agrostoides is also polyphyletic. Other genera, such as Streptostachys and Arthropogon, also appeared as polyphyletic in both studies (Zuloaga, Morrone, and Giussani, 2000
).
Our analyses confirm the polyphyly of the tribe Arundinelleae. The taxa included for study (Arundinella hirta, Danthoniopsis dinteri, and D. petiolata) appear in different and distantly related clades (Fig. 1), with Danthoniopsis clearly placed in a basal polytomy with the outgroups. Previous studies (Mason-Gamer, Weil, and Kellogg, 1998
; Spangler et al., 1999
) have also indicated a polyphyletic origin of the Arundinelleae and the need of further studies to recircumscribe this tribe. Likewise, relationships between the subfamilies Panicoideae and Centothecoideae (the latter including Thysanolaeneae) are not resolved by our analyses and need further investigation.
Morphological correlates of molecularly defined clades
Members of the Forest Shade Clade—Acroceras, Echinochloa, Lasiacis, Oplismenus, Panicum ovuliferum, and Pseudechinolaena—all have lanceolate leaf blades and are associated with the forest shade environment (Davidse, 1978
; Clayton and Renvoize, 1986
; Zuloaga, Morrone, and Saenz, 1987
). In addition, all taxa in this clade except Echinochloa use the C3 photosynthetic pathway. The ligules of these species are membranaceous, their glumes are herbaceous, and their upper paleas and lemmas are crustaceous, with the margins of the upper lemma tucked into the palea; their primary inflorescence branches are more or less racemose, with the spikelets borne close together on short pedicels. Acroceras and Echinochloa both have a protuberance on the apex of the palea (perhaps an homologous character for both genera); they are not sisters in our analyses, but the placement of Acroceras in the clade is ambiguous. Echinochloa is C4 NADP-ME subtype, with a single bundle sheath.
The Setaria clade includes species of Cenchrus, Paspalidium, Pennisetum, Setaria, and Stenotaphrum, all of which have setae or bristles in the inflorescences, the sole exception being Panicum bulbosum. These bristles vary in position and development, with some terminating branches, and others apparently representing modified pedicellate spikelets (Webster, 1988, 1992
). Previous authors (e.g., Clayton and Renvoize, 1986
) had thought that bristles were not homologous among members of this group, but our molecular data and that of Gómez-Martinez and Culham (2000)
suggest that bristles had a single evolutionary origin.
Setaria is one of the genera with the largest taxonomic sampling, in part because it includes >100 species distributed worldwide. Although the genus is easily recognized because of its bristles that persist on the rachis when the spikelet falls, our study indicates that this genus is paraphyletic. Of the species investigated, five of them (S. macrostachya, S. parviflora, S. palmifolia, S. sphacelata, and S. geniculata) form a monophyletic subclade. Setaria viridis and S. lachnea are sisters and together form a clade with Panicum bulbosum. The relationship between Pennisetum and Cenchrus is not surprising in that both have smooth cartilaginaceous to membranous paleas and lemmas, and the dispersal unit is the spikelet-bristle combination (i.e., bristles fall from the rachis along with the spikelets).
All taxa in the Urochloa clade (Chaetium, Eriochloa, Panicum (= Urochloa) maximum, Melinis and Urochloa) use the PCK subtype of the C4 photosynthetic pathway, and have the classical morphological correlates of the pathway (Brown, 1977
; Watson and Dallwitz, 1992
), although there are no other morphological characters that distinguish the clade. Although most species have a base chromosome number of x = 9, Chaetium bromoides has x = 13, an unusual chromosome number within the Paniceae. The paraphyletic Urochloa also appears to have undergone an active process of chromosome evolution, with the chromosome number of Panicum (= Urochloa) maximum varying from x = 9 to x = 8 or 7. Urochloa acuminata, which appears basal in the clade, also has x = 13 (Morrone et al., 1995
).
Panicum subg. Panicum is a robust monophyletic unit. (Here and elsewhere in this discussion, the classification of Panicum follows Zuloaga, 1987
). Physiologically, the group is characterized by the principal activity of NAD-ME enzyme during photosynthesis (see above under Correlation with photosynthetic pathways), and several species have centripetal chloroplasts in the outer part of the parenchymatous sheath. Morphologically, this subgenus includes caespitose plants with membranous-ciliate ligules, linear to linear-lanceolate blades, open and lax inflorescences, spikelets with crustaceous upper lemmas and paleas, and the apex of the palea with simple or compound papillae.
Panicum subg. Dichanthelium section Dichanthelium, represented by two species (P. sabulorum and P. koolauense), is sister to the large Setaria/Urochloa/Panicum clade in the equally weighted trees or basal to the x = 9 Paniceae clade in the implied weighted trees. The two species are C3, with foliar dimorphism usually present, and have upper lemmas and paleas covered with simple papillae.
The sister taxon relationship of Panicum millegrana (P. subg. Phanopyrum sect. Monticola) and Sacciolepis indica is one of several surprises in this study. Both species are C3, but this is the ancestral condition and does not indicate relationship; they differ by the presence of a spiciform panicle in Sacciolepis, with a gibbous upper glume and a smooth upper floret. Panicum millegrana bears open and lax panicles, nonswollen upper glumes, and upper florets that are transversely rugose. We have been unable to find any morphological characters shared by the two species.
Digitaria is clearly monophyletic in these analyses, but its position is uncertain, whether sister to all x = 9 Paniceae (in equally weighted trees) or sister to the Setaria/Urochloa/Panicum clade (in implied weighted analysis). Digitaria species share the C4 NADP-ME subtype and have only one bundle sheath. Digitaria is also characterized by dorsiventrally compressed spikelets with the lower glume reduced, the lower flower absent, the upper lemma and palea cartilaginous.
The clade that includes Altoparadisium, Arthropogon, Homolepis, Mesosetum, Panicum euprepes, P. prionitis, Streptostachys ramosa, and Tatianyx (the Ambiguous Clade), includes considerable morphological, anatomical, and physiological diversity. These taxa have not been placed together in previous classifications and few obvious morphological features unite them. Some are C3 with standard C3 vein spacing and double bundle sheaths, whereas others are C4 NADP-ME with close vein spacing and variously reduced outer bundle sheaths. Homolepis is monophyletic, with the two species investigated having glumes that are equal in length, cartilaginous upper lemmas and paleas covered by silica bodies, and linear hila that extend half the length of the caryopsis (Zuloaga and Soderstrom, 1985
). Two of the three species of the recently recircumscribed Arthropogon (Filguieras et al., 2001) are represented here and appear in different clades. Arthropogon lanceolatus is grouped with Panicum euprepes and P. prionitis, while A. villosus, a C4 NADP-ME species, is related to the recently established genus Altoparadisium (Filgueiras et al., 2001
). Arthropogon lanceolatus is a C3 species with fusoid cells, aristate lower glume, and cartilaginous upper lemma and palea; P. euprepes is also C3 but with stiff and sharp-pointed leaf blades, and open, lax panicles, while P. prionitis is C4 with flat leaves, keeled blades and open panicles.
Species in the clade including Anthaenantiopsis, Axonopus, Echinolaena, Ichnanthus, Ophiochloa, Panicum piauiense, P. obtusum, Paspalum, Streptostachys asperifolia, and Thrasya share an inflorescence pattern in which the spikelets are arranged in unilateral branches. Most genera included in this large clade are native to America. Streptostachys asperifolia, a C3 non-Kranz species, is sister to the Axonopus-Ophiochloa clade. Axonopus and the monotypic genus Ophiochloa are C4 "classical NADP-ME" type species. In some trees, Axonopus appears to be paraphyletic with Ophiochloa derived from within it. However, their relationships are not resolved in the consensus trees and Ophiochloa differs by nine mutations. Ophiochloa differs from Axonopus by its hyaline upper floret with the upper palea free at the apex, lower lemma free from the upper glume, and one raceme per inflorescence (Filgueiras, Davidse, and Zuloaga, 1993
).
Paspalum is, after Panicum, the second largest genus of the Paniceae, with 320 species. In our study, this genus is represented by nine species belonging to seven informal groups, which include a wide range of morphological features and geographic distribution. These species form a well-supported clade with representatives of Thrasya (Fig. 1). Paspalum is a paraphyletic assemblage, with two species of Thrasya embedded within it. Both genera have unilateral racemes with the lower glume absent or reduced. Additionally, several species of the Paspalum group Decumbentes are morphologically intermediate between the two genera, with the lower glume always or variably present and varying in length within the same inflorescence, the pedicels partly adnate to the rachis, and the lower lemma coriaceous and sulcate in the middle portion.
From the major clade composed by Hymenachne, Leptocoryphium, Otachyrium, Panicum laxum, Plagiantha, and Steinchisma hians, the last four genera form a well-supported subclade characterized by an expanded lower palea. Additionally, Panicum laxum, Plagiantha, and Steinchisma are linked in another subclade with all three entities having verrucose papillae all over the upper lemma and palea, although this character is not constant in all specimens of Panicum laxum.
Implications for classification
Brown (1977)
classified the Paniceae based on photosynthetic pathway, dividing it into four major groups. His subtribe 1 included all NADP-ME species, here shown to be polyphyletic. Brown's subtribe 2 included all C3 species, which he interpreted as basal, and the group from which all other members of the tribe originated. We agree with Brown's point of view, as it is the ancestral condition in the implied weighted trees, and we consider it the most probable state for the ancestor of the panicoid grasses (see below under Evolution of C4 photosynthetic pathways within panicoid grasses), although the ancestral state is ambiguous in the equally weighted trees. Brown's subtribes 3 and 4 corresponded to the Urochloa group (all with the PCK subtype of C4) and Panicum subg. Panicum (all with NAD-ME subtype); we have confirmed that these are monophyletic.
Clayton and Renvoize (1986)
recognized seven subtribes in Paniceae, based on several exomorphological and anatomical characters. Of the seven, only subtribe Cenchrinae is monophyletic on the molecular tree, although Cenchrus and Pennisetum were the only genera of the subtribe included. Their intuitive diagram of relationships among genera of the subtribe Setariinae showed the genera organized according to photosynthetic pathway; taxa are grouped largely without regard to chromosome numbers. Nonetheless, their groups of Otachyrium-Plagiantha-Steinchisma, Setaria-Paspalidium-Stenotaphrum, Paspalum-Thrasya, and Urochloa-Eriochloa correspond to clades in the molecular analyses. The latter group also includes Brachiaria s.s.; although we did not include any species of Brachiaria s.s. in this analysis, Urochloa mutica was treated as Brachiaria mutica by Clayton and Renvoize (1986)
. Clayton and Renvoize (1986)
suggested Panicum as a possible ancestor from which all Setariinae emerged. Our results, in contrast, show Panicum in many places over the tree, with many of these species associated with other genera of Paniceae (see discussion of Panicum).
Streptostachys ramosa appears in the x = 10 Paniceae clade, although the chromosome count (Davidse and Pohl, 1978
) shows that this species is x = 9. Streptostachys ramosa is separated from the type species of the genus, S. asperifolia, a species with a basic chromosome number of x = 10 (Morrone et al., 1995
), the latter in a well-supported clade together with Ophiochloa and Axonopus. Morrone and Zuloaga (1991)
have pointed out the differences between S. asperifolia and the other species of the genus, S. ramosa and S. macrantha, but they did not make any decision about the taxonomic position of these taxa. Our data suggest that the genus should be split into two, with S. ramosa placed in another genus.
Our data show that Paspalum is paraphyletic and includes the genus Thrasya. To maintain a monophyletic Paspalum, therefore, the clade must either be divided into smaller units or Thrasya must be merged with Paspalum. The latter possibility has already been suggested by morphological studies (Trillo and Rúa, 1999
; S. S. Aliscioni, unpublished data).
The ndhF data support segregation of Steinchisma from Panicum (Renvoize, 1988, 1998
; Zuloaga et al., 1998
) and its close relationship with Plagiantha (already established by Zuloaga et al., 1998
). The similarity between Panicum laxum and Steinchisma needs to be tested with addition of more species of subg. Phanopyrum sect. Laxa of Panicum. Species of sect. Laxa have been crossed with those of Steinchisma (Brown et al., 1985
), suggesting a close relationship between the two groups. It is also remarkable that Hymenachne and Sacciolepis, two genera usually cited as closely related in the tribe (Pohl and Lersten, 1975
), appeared in this analysis in two clearly distinguished clades.
Zuloaga and Soderstrom (1985)
removed two species from Panicum and included them in the genus Homolepis, a conclusion not followed by Clayton and Renvoize (1986)
and Renvoize (1998)
. In this study, one of these two species was included (H. glutinosa = Panicum glutinosum), and its position as sister with H. isocalycia supports its segregation as Homolepis.
Brown (1977)
placed Brachiaria, Urochloa, and Eriochloa in a natural group, supported by their PCK physiology. Many of the species of Brachiaria included by Brown were later transferred to Urochloa. Brown suggested that Panicum maximum, the only species of Panicum with PCK physiology, should be included within the Brachiaria group. Most genera of this C4 PCK-type group have rough, transversely rugose lemmas and primary inflorescence branches with spikelets on short pedicels, although U. maxima and the Fasciculata group of Urochloa do not fit this general description. Morrone and Zuloaga (1992, 1993)
, following Webster (1987, 1988)
transferred American species of Panicum with the PCK syndrome to Urochloa, a conclusion strongly supported by the molecular data.
Most genera currently accepted within Paniceae, such as Acroceras, Brachiaria, Digitaria, Homolepis, Ichnanthus, Otachyrium, and Urochloa, among others, have been gradually segregated from Panicum, starting with the pioneering papers of Chase (1906, 1908a, b, 1911)
, and the relationships of this genus with many other taxa are highlighted in the intuitive evolutionary diagrams of Clayton and Renvoize (1986)
. In spite of the continuous segregation of genera, Panicum is still polyphyletic. Panicum subg. Panicum and sect. Dichanthelium are the only taxa that are monophyletic. Panicum maximum, previously classified by Zuloaga (1987)
as P. subg. Megathyrsus, is found within the Urochloa clade, and therefore in this paper it is treated as U. maxima, as suggested by Webster (1987)
; this species is characterized by a transversely rugose upper lemma and palea. Species of Panicum subg. Dichanthelium are distributed in two different clades, with sect. Dichanthelium (represented here by P. sabulorum and P. koolauense) forming an isolated but well-supported clade, unrelated to P. ovuliferum (sect. Cordovensia), which is close to Acroceras and Echinochloa. Panicum bulbosum, which belongs to P. subg. Agrostoides sect. Bulbosa (Zuloaga, 1987
), is within the Setaria clade. The other species of subgenus Agrostoides, P. prionitis of sect. Prionitia, appears in the x = 10 Paniceae clade, related to P. euprepes (sect. Lorea, subg. Phanopyrum) and Arthropogon lanceolatus. In turn, Panicum piauiense (subg. Phanopyrum sect. Stolonifera) forms a clade with the genus Echinolaena. Panicum obtusum is distantly related to other NADP-ME taxa, such as Thrasya, Paspalum, and Anthaenantiopsis. Also, Panicum laxum, of subg. Phanopyrum section Laxa, is grouped with Steinchisma hians in a strongly supported clade related to Plagiantha, Hymenachne, and Otachyrium.
These results suggest that the name Panicum should be restricted to subgenus Panicum. Panicum maximum should be placed in Urochloa, Dichanthelium can be raised from the subgeneric to ge
