| 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
Received for publication July 19, 2002. Accepted for publication December 12, 2002.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONTAXONOMIC TREATMENTLITERATURE CITED |
|---|
Key Words: codon position • Dichanthelium • Hymenachne • molecular phylogeny • ndhF • Paniceae • Panicoideae • Panicum • Phanopyrum • Poaceae
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONTAXONOMIC TREATMENTLITERATURE CITED |
|---|
; Watson and Dallwitz, 1992
). The subfamily comprises two main tribes: Paniceae, with approximately 101 genera, and Andropogoneae, with approximately 85 genera. Smaller tribes are Arundinelleae (11 genera), Isachneae (five genera), Neurachneae (three genera), and the monotypic tribes Hubbardieae and Steyermarkochloeae, the latter placed in Arundinoideae by Watson and Dallwitz (1992)
. Eriachneae, with two genera, is conventionally included in the subfamily, but was placed in Arundinoideae by Watson and Dallwitz (1992)
; recent molecular data have shown convincingly that Eriachneae does not belong in the Panicoideae (GPWG, 2001
).
Monophyly of the subfamily is supported by the presence of bifloral spikelets with the lower flower staminate or neuter (Brown, 1810
, 1814
), and distinctive, simple starch grains (Tateoka, 1962
; Kellogg and Campbell, 1987
), as well as by extensive molecular data from both chloroplast and nuclear genes (Gómez Martínez, 1998
; Kellogg, 1998
; Soreng and Davis, 1998
; Gómez-Martínez and Culham, 2000
; Giussani et al., 2001
; J. Barber, L. Giussani, F. Zuloaga, and E. Kellogg, unpublished data).
The tribe Paniceae contains almost half of the genera and 60% of the species of the Panicoideae with several genera having worldwide distribution; its greatest diversity is concentrated in the tropics. Members of the tribe differ extensively in morphological, physiological, anatomical, and karyological characters. This has resulted in different evolutionary schemes of the tribe and its genera.
Panicum, the type genus of Paniceae and Panicoideae, is a large and taxonomically difficult group of species. Taxonomists have held different opinions on the delimitation of this genus, which we will summarize below. For the purposes of this discussion, we will use the name Panicum sensu lato (s.l.) to refer to the classification proposed by Zuloaga (1987a)
and subsequent modifications (Zuloaga et al., 1989
, 1992
, 1993b
; Zuloaga and Morrone, 1996
; F. Zuloaga and O. Morrone, unpublished data; see details below and Table 1). Panicum sensu stricto (s.s.) refers to the more narrowly circumscribed group recognized herein. The genus was described by Linnaeus (1753)
, who recognized 20 species for Panicum. Fifteen of these species have been transferred to other genera (e.g., Dichanthelium, Digitaria, Echinochloa, Paspalum, Pennisetum, Stenotaphrum), and only three of Linnaeus's species are still recognized within Panicum subg. Panicum; the other two correspond to the same species, P. brevifolium L., which is in sect. Parvifolia subg. Phanopyrum.
|
). Panicum s.l. is fairly uniform in spikelet morphology but quite diverse in vegetative features (Crins, 1991
). The genus shows a wide range of variation, and no single character defines it. Brown (1977)
, in a study based on anatomical and physiological data, postulated that a C3 Panicum could be the ancestor of all remaining Paniceae. Later, Clayton and Renvoize (1986)
, in an intuitive classification of the tribe, placed Panicum as the core genus, from which all other genera were derived. The lack of an obvious synapomorphy, the broad range of morphological and physiological variation, and the suggestion of being a group from which all others are derived, indicate the likely para- or polyphyletic nature of Panicum.
Panicum s.l. includes, among its species, all photosynthetic types known to occur in the Poaceae. The genus has C3 and C4 photosynthetic systems (Brown, 1977
), and also some C3/C4 intermediate species (Brown and Smith, 1975
; Ku and Edwards, 1978
; Morgan and Brown, 1979
; Brown et al., 1985
; Hattersley et al., 1986
; Zuloaga et al., 1998
). The C4 species include representatives of the PEP-ck, NAD-me and NADP-me subtypes (Downton, 1975
; Hatch et al., 1975
). This variation is uncommon among genera of Poaceae; photosynthetic pathway is usually uniform among closely related species.
Basic chromosome numbers also vary within Panicum s.l. Multiples of 8, 9, 10, 11, and 15 are reported, suggesting up to five basic numbers for Panicum (Darlington and Wylie, 1955
; Gould, 1960
; Bolkhovskikh et al., 1969
; Christopher and Abraham, 1976
; Bouton et al., 1981
; Kumatsu and Suzuki, 1987
; Urbani, 1990
; Hamoud et al., 1994
). Of these, the most common basic chromosome numbers are 9 and 10, while other numbers, those not in multiples of 9 or 10, may have evolved by cytological aberrations (Jauhar and Joshi, 1965
; Tateoka, 1965
; Jauhar, 1969
). Of the proposed subgenera of Panicum s.l., Steinchisma, Panicum, and Dichanthelium are uniform regarding basic chromosome number, while Agrostoides and Phanopyrum are heterogeneous (Dubcovsky and Zuloaga, 1991
).
The taxonomy, anatomy, and physiology of Panicum have received considerable attention in recent years. A small sample of Panicum has been analyzed in recent morphological and molecular studies, using chloroplast and nuclear gene sequences (Gómez-Martínez and Culham, 2000
; Zuloaga et al., 2000
; Duvall et al., 2001
; Giussani et al., 2001
; J. Barber, L. Giussani, F. Zuloaga, and E. Kellogg, unpublished data); all these treatments concluded that the genus is polyphyletic and that more comprehensive studies, analyzing more taxa, are needed.
Several taxonomic studies have been conducted in the genus, but none has studied it in its entirety (Zuloaga and Soderstrom, 1985
). Hitchcock and Chase (1910
, 1915
) analyzed the North American species and placed them in informal groups or subgenera, suggesting natural relationships between the species (Table 1). Later, several of these groups were removed from Panicum, such as subg. Paurochaetium (transferred to Setaria), Fasciculata group (Urochloa), Geminata group (Paspalidium), and Maxima group (Urochloa). Pilger (1931
, 1940
) divided the genus into eight subgenera and 23 sections. Hsu (1965)
, who made a complete study of the tribe Paniceae considering new characters (e.g., lodicule morphology, epidermal characters), recognized five subgenera and 25 sections within Panicum (Table 1). Crins (1991)
, in a review of Paniceae for the southern United States, considered that several taxa currently placed in Panicum deserve generic status; according to him this goal can be achieved only when tribal, subtribal, and generic classification schemes are based on phylogenetic considerations and the infratribal groups are based on true apomorphies.
Zuloaga (1987a)
proposed an intuitive classification of the genus for America, based on exomorphological, anatomical, and karyological characters. In this classification and subsequent modifications, six subgenera and 24 sections were recognized. We based our study on Zuloaga's classification (Table 1) with the following minor changes. Zuloaga (1987a)
considered Repentia a section of subg. Panicum, with approximately 12 species. Afterwards, F. O. Zuloaga (unpublished data) considered that several species of this sect. should be included in section Virgata, and Panicum gouinii, P. repens, and P. pedersenii were assigned to section Dichotomiflora. Panicum validum was first considered by Zuloaga (1987a)
as an ungrouped species within subg. Agrostoides, but later, Zuloaga et al. (1989)
created section Valida based on this unique species. Panicum penicillatum, placed by Zuloaga (1987a)
within subg. Phanopyrum, was later classified in section Dichanthelium subgenus Dichanthelium (Zuloaga et al., 1993b
). Panicum stramineum was considered by Zuloaga (1987a)
as a synonym of P. hirticaule, but P. stramineum was revalidated by Zuloaga and Morrone (1996)
.
Because Zuloaga's classification is the starting point for the present study (Table 1), we summarize the six subgenera and their major characters below.
Subgenus Panicum, with five sections, includes Kranz species, which are characterized by having two sheaths around the vascular bundles, with the outer parenchymatous sheath usually having centripetal specialized chloroplasts (Fig. 1A). This subgenus was considered by W. V. Brown (1977)
as homogeneously C4, NAD-me subtype; nevertheless, some species (e.g., P. elephantipes) were described as having PEP-ck type anatomy by Zuloaga (1987a)
. The basic chromosome number of the subgenus is x = 9. Most of its species are cespitose with erect culms, usually growing in dry and open places. The species are distinguished by having pyramidal inflorescences, lax and diffuse, with ellipsoid to lanceolate spikelets (Fig. 2E–F), with the upper glume 7–9-nerved, and the upper anthecium smooth and shining with compound or simple papillae toward the apex of the palea (Fig. 1B).
|
|
to include sect. Dichanthelium, with species growing from Canada and the United States to southern America, and sect. Cordovensia, with five species ranging from Mexico and Central America to Argentina. Subgenus Dichanthelium includes non-Kranz, C3 species; all species have lax panicles with spreading or occasionally contracted branches, spikelets obovoid to ellipsoid, the upper glume and lower lemma usually 7–9-nerved, with the lower palea absent in species of sect. Cordovensia, and the upper anthecium apiculate, the lemma and palea covered by simple papillae regularly distributed (Fig. 1G). The basic chromosome number for subg. Dichanthelium is x = 9.
Hitchcock and Chase (1910)
considered Dichanthelium as a subgenus of Panicum; later, Hitchcock and Chase (1915)
recognized the Cordovensia group within subg. Dichanthelium, indicating that the morphology of this group was intermediate between Dichanthelium and Panicum (Table 1). Gould (1974)
raised Dichanthelium to the generic level. Later, Clark and Gould (1975)
, Brown and Smith (1975)
, Gould and Clark (1978)
, and Gould (1980)
, while studying North American species of the subgenus, confirmed this taxonomic decision on the basis of exomorphological characters, photosynthetic type, and upper anthecium ornamentation. In North America, Dichanthelium can be recognized easily by the presence of a basal rosette of leaves during the winter and a clear foliar dimorphism (Fig. 2H). This character is not seen in the Central and South American taxa, which led Lelong (1984)
to treat Dichanthelium as a subgenus within Panicum, as did Zuloaga et al. (1986
, 1993b
), Clayton and Renvoize (1986)
, Zuloaga (1987a)
, and Webster (1988)
.
Subgenus Steinchisma includes C3/C4 intermediate species (Fig. 1C), with a basic chromosome number of x = 10. The species grow in open and humid places. Spikelets are arranged in open or contracted panicles, the upper anthecium is covered with compound papillae over the entire surface (Fig. 1D), and the lower palea is conspicuously expanded at spikelet maturity (Fig. 2A). Steinchisma was established as a genus by Rafinesque (1830)
. Later, several authors included Steinchisma in sect. Laxa of Panicum (Hitchcock and Chase, 1910
, 1915
; Pilger, 1931
, 1940
; Hsu, 1965
; see Table 1). Brown (1977)
mentioned that habitat, ornamentation of the upper anthecium, lower palea features, and inflorescence type are strong characters to segregate Steinchisma from Panicum. Brown et al. (1985)
stressed that several artificial hybrids were produced between species of Steinchisma and species of Panicum sect. Laxa; these authors therefore concluded that Steinchisma should be retained within Panicum and not as an independent genus. More recently, Steinchisma has been considered either a subgenus of Panicum (Zuloaga, 1987a
; Webster, 1988
) or as a distinct genus (Clayton and Renvoize, 1986
; Renvoize, 1998
; Zuloaga et al., 1998
).
Subgenus Megathyrsus, as delimited by Zuloaga (1987a)
, contains a single species, Panicum maximum Jacq. Webster (1987)
, in a revision of Paniceae of Australia, transferred this species to the genus Urochloa. Panicum maximum is C4, of the PEP-ck photosynthetic subtype (Fig. 1E); exomorphologically it is characterized by its lax and pyramidal, many-flowered panicles, ellipsoid spikelets with the upper anthecium transversely rugose, with single papillae on the wrinkles (Fig. 1F). The basic chromosome number of this species is polymorphic x = 8 or 9.
When establishing subg. Agrostoides, Zuloaga (1987a)
included in it a few isolated species of Panicum, which are all C4, of the mestome sheath (MS) subtype, e.g., with a single mestome sheath with specialized chloroplasts around the vascular bundles (exceptionally with an outer parenchymatous bundle sheath in P. prionitis and P. petersonii, species of sect. Prionitia); additionally, most species have spikelets with the upper glume and lower lemma 3–5-nerved. Sections (mostly monotypic or oligotypic) assembled in this subgenus are American and each has a restricted distribution, with the exception of an Asian species, P. antidotale. Basic chromosome number can be x = 9 (in sects. Agrostoidea and Bulbosa and in P. antidotale) or x = 10 (in sects. Obtusa, Prionitia, Tenera, and Valida); no chromosome counts have been reported for sects. Tuerckheimiana and Discrepantia. Brown (1977)
previously placed these taxa in his "Miscellaneous Assemblage," indicating they were not closely related to each other. Later, Zuloaga et al. (1989
, p. 228) emphasized that most species placed in Agrostoides are diploids and may represent "an adaptative line differentiated from C4, PS species."
Subgenus Phanopyrum (Zuloaga, 1987a
) includes non-Kranz species of Panicum, with spikelets with the upper glume and lower lemma usually five-nerved (except in P. mertensii, of sect. Megista, with the upper glume and lower lemma 7–9-nerved). Some sections have spikelets arranged on unilateral branches (e.g., Phanopyrum, Stolonifera or Laxa), and other taxa have lax and open inflorescences. The upper anthecium may be covered with simple papillae and bicellular microhairs (e.g., Parvifolia, Monticola, Verrucosa), whereas in sects. Parvifolia and Lorea the upper anthecium is smooth, and in Monticola and Verrucosa it is slightly transversely rugose. Consistency of the upper anthecium ranges from hard and indurate to membranous, with the margins of the upper lemma not enclosing the upper palea completely. Chromosome counts also vary within this subgenus, with both x = 9 sections (Parvifolia, Monticola, Verrucosa, and Parviglumia), and x = 10 sections (Stolonifera, Laxa, Lorea, and Phanopyrum).
A broad molecular study of subfamily Panicoideae suggested the polyphyly of Panicum (Giussani et al., 2001
), although sampling of the genus was incomplete. In particular, it was shown that the six subgenera described above do not form a monophyletic group. Subgenus Panicum is monophyletic, but not related to the other five subgenera. Subgenus Dichanthelium is apparently polyphyletic, with sect. Dichanthelium falling in a different clade from sect. Cordovensia. Subgenus Steinchisma is monophyletic. Subgenus Megathyrsus (Panicum maximum) is indeed within the Urochloa group, supporting its transfer to that genus. The subgenera Agrostoidea and Phanopyrum were apparently polyphyletic, but too poorly sampled for definite conclusions.
The goal of this paper is to elucidate the phylogenetic placement of almost all sections of Panicum s.l. and to test their monophyly where possible. Our study is based on the chloroplast gene ndhF, building on the earlier study of Giussani et al. (2001)
. We also consider morphological characters of the various sections and their placement within Panicoideae. Molecular characters were matched with the diagnostic morphological characters, chromosome numbers, and with different photosynthetic pathways; similarly our results were contrasted with proposed classifications for Panicum. Validity of infrageneric taxa is discussed and new taxonomic arrangements are proposed wherever possible. Recent studies (Källersjö et al., 1999
, 2000
) have pointed out the relevance of mutations in the third codon position to resolve the phylogeny. In order to explore their role in the evolution of Panicoideae, we also analyzed the three positions independently and compared supported groups with bootstrap results of the whole data set.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONTAXONOMIC TREATMENTLITERATURE CITED |
|---|
, Murray and Thompson (1980)
, and Saghai-Maroof et al. (1984)
. The procedure to extract and amplify DNA was similar to that used by Giussani et al. (2001)
. A battery of 10 sequencing primers specified by Olmstead and Sweere (1994)
and Clark et al. (1995)
were used. Overlapping fragments were combined as follows: 5F/972R, 536F/1318R, 972F/1821R, and 1318F/2110R. For difficult taxa, including the herbarium material, smaller fragments were used: 5'F/536R or 1821F/2110'R. Additionally, two new internal primers were developed for polymerase chain reaction (PCR) amplification and sequencing of ndhF within the subfamily Panicoideae: 1660F, 5'CTTTTTACTTTGTTCATTGGAT3'; 1660R, 5'ATCCAATGAACAAAGTAAAAAG3'. Forward and reverse strands were sequenced with a minimum overlap of 90% for every taxon on an ABI 377 automated sequencer (Applied Biosystems, Foster City, California, USA) 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 Corporation, Ann Arbor, Michigan, USA). Sequences were translated to check for stop codons and then manually aligned, preserving the reading frame.
The aligned data matrix has been submitted to TreeBASE (http://www.treebase.org/treebase/).
Ingroup selection
Panicum, the largest and most problematic genus of Paniceae, was sampled to test its monophyly and infrageneric delimitation. Effort was made to encompass most of the morphological diversity and geographical distribution of the genus. Most morphological diversity appears in the Americas, but we have included selected Old World taxa as well. Panicum sampling includes 66 species representing all six subgenera and 22 of 24 sections of Panicum s.l. Sections Discrepantia, of subg. Agrostoides, and Parviglumia, of subg. Phanopyrum, were not sequenced because of lack of material. We also included two African species of sect. Clavelligera, an Asian species, Panicum antidotale of subg. Agrostoides, and three species from Oceania: P. faurieri, P. nephelophilum of sect. Panicum subg. Panicum, and P. koolaunse of sect. Dichanthelium subg. Dichanthelium, to extend the geographical sampling. Of the 66 sequences, 46 sequences were generated for this paper, and 20 were taken from Giussani et al. (2001)
or GenBank (Appendix 1; http://ajbsupp.botany.org/v90/).
Outgroup selection
Since Panicum appears as a polyphyletic genus within Paniceae (Gómez-Martínez and Culham, 2000
; Zuloaga et al., 2000
; Duvall et al., 2001
; Giussani et al., 2001
), any study of Panicum needs to include as many genera as possible of Paniceae and subfamily Panicoideae. Accordingly, we used the study of Giussani et al. (2001)
as a guide to taxon sampling. All genera of Paniceae already sequenced for ndhF were included in the analysis, and three new species, Brachiaria eruciformis, Oplismenopsis najada, and Pennisetum montanum, were also added. Any genus previously shown to be monophyletic (Digitaria, Echinochloa, Homolepis, and Thrasya) was represented here by one species; if the genus was shown by Giussani et al. (2001)
to be para- or polyphyletic, several species were included (e.g., Paspalum, Pennisetum, Setaria, and Urochloa).
The data matrix, including ingroup and outgroup taxa, has a total of 123 sequences, with 119 belonging to subfamily Panicoideae. Of these, 111 represent the tribe Paniceae (66 representing Panicum s.l.,), seven represent the tribe Andropogoneae, and one (Arundinella hirta) represents tribe Arundinelleae. Four other taxa, from subfamily Centothecoideae, were included in the analyses as close relatives of the subfamily Panicoideae (Clark et al., 1995
; GPWG, 2001
): Chasmanthium laxum subsp. sessiliflorum, Danthoniopsis dinteri, Thysanolaena maxima, and Zeugites pittieri. Danthoniopsis dinteri, formerly placed in Arundinelleae, is now seen as a member of Centothecoideae (GPWG, 2001
; J. G. Sanchez-Ken and L. G. Clark, Iowa State University, personal communication). Thysanolaena, previously placed in subfamily Arundinoideae as the tribe Thysanolaeneae, is now included in Centothecoideae (GPWG, 2001
). The number of representatives for tribe Andropogoneae and subfamily Centothecoideae was reduced from the number used by Giussani et al. (2001)
. The tribe Andropogoneae represents a rapid radiation with few mutations on internal branches of the tree (Giussani et al., 2001
) so that many equally parsimonious arrangements are possible. We were not interested in all these rearrangements in this study, so only placeholders were used.
To insure that the results are not biased by the selection of taxa, an additional analysis was performed using all available ndhF sequences of the tribe Paniceae, as taken from Giussani et al. (2001)
. The complete matrix contained 156 taxa. Appendix 1 (http://ajbsupp.botany.org/v90/) shows species used in this study and their accession numbers.
Phylogenetic analyses
Cladistic analyses based on maximum parsimony were performed using NONA version 2.0 (Goloboff, 1997a
). All characters were equally weighted with respect to codon position, were unordered, and gaps were scored as missing data. Parsimony uninformative characters were excluded. Indels were mapped onto the final tree to determine whether they were synapomorphies or homoplasies. In a separate analysis, we recoded indels as binary data following the "simple indel coding method" proposed by Simmons and Ochoterena (2000)
.
Heuristic searches were carried out with the options "amb-" (resolve clades only if they have unambiguous support) and "poly=" (polytomies allowed). Searches were performed using "h/2," "mult*3000," which randomizes the order of the taxa, creates a Wagner tree, and submits it to branch-swapping by tree-bisection reconnection (TBR), retaining a maximum of two trees for each initial Wagner tree; the procedure was repeated 3000 times. The shortest trees retained from the subsearches were then TBR swapped to completion with the "max*" command. Additionally, and only for the 156 taxon data set, we performed a parsimony ratchet analysis (Nixon, 1999
) using one of the most parsimonious trees found with NONA as starting point. A total of 10 000 replicates were done as implemented in Winclada version 1.0 (Nixon, 1999–2002
).
To evaluate relative support for individual clades, bootstrap analysis (Felsenstein, 1985
) was performed in NONA, using Winclada as an interface (Nixon, 1999–2002
). A total of 10 000 replicates was performed. Each replicate was analyzed using a Wagner tree as a starting point followed by TBR branch-swapping, saving only one tree per replicate. Supported clades are considered to be those with more than 50% bootstrap value.
Decay indices (d) or Bremer support values (Bremer, 1994
) were also calculated using PAUP version 4.0b10 (Swofford, 2002
). We used the consensus of all shortest trees as a negative constraint tree to look for trees one step longer. Constraint searches were done 70 000 times, performing TBR swapping on the tree and saving only one tree per replicate whenever it was the shortest suboptimal tree. To find clades that collapse in trees two steps longer, we used the consensus tree of the shortest plus the suboptimal trees already found as a reverse constraint tree and proceeded with a new series of 70 000 replicates. This procedure was repeated nine times.
To explore the monophyly of particular groups, constrained analyses were performed with NONA to determine the number of additional steps required to make a monophyletic group. We used a tree with a fixed monophyletic group as a 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 conducted Templeton tests (Templeton, 1983
), as implemented in PAUP version 4.0b10 (Swofford, 2002
).
An alternative analysis using implied weights (Goloboff, 1993
) was run in Pee-Wee version 3.0 (Goloboff, 1997b
) using the same search strategies as in NONA. Implied weight provides a way to resolve character conflict in favor of the characters that have less homoplasy on the trees and implies that the average weight for the characters is as high as possible (Goloboff, 1997b
). 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, 1998
). Searches were done using all six possible concavities.
Base-pair performance
Synonymous substitutions, principally due to variation in the third codon position, occur much faster than nonsynonymous substitutions (Li, 1997
; Page and Holmes, 1998
). However, Källersjö et al. (1999
, 2000
) showed that, for rbcL sequences of Poaceae, the third position contains most of the phylogenetic information in the data. Hence, to test the influence of the three base-pair codon positions on phylogenetic results, we treated them independently. We examined subsets of data that represent (1) first codon positions only, (2) second codon positions only, and (3) third codon positions only. Bootstrap analyses were performed on each data set (10 000 replicates, each starting with a Wagner tree and TBR-swapped), and results were independently compared with the bootstrap analysis performed on the entire data set (2061 bp data set).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONTAXONOMIC TREATMENTLITERATURE CITED |
|---|
|
, are listed in Table 2 and mapped on one of the most parsimonious trees (Fig. 3).
|
; sequences are A-T rich, with thymine 37.3%, adenine 26.8%, guanine 16.7%, and cytosine 15.8%. Overall, 3.4% of the data matrix cells were scored as missing data. Heuristic searches of the 123-taxa matrix with informative characters equally weighted (NONA) found 15 408 trees of length (L) = 1506; consistency index (CI) = 0.42; and retention index (RI) = 0.78. Bootstrap values above 50% are shown above the branches in Fig. 4. Figure 5 presents a consensus tree of the 15 408 most parsimonious trees with principal monophyletic groups or paraphyletic groups represented with triangles or thick lines, respectively. All these analyses treated gaps as missing data. Coding indels as binary characters produced identical trees with only a few additional steps (1518).
|
|
. The subfamily Panicoideae is monophyletic (bootstrap = 97%, decay (d) = 6), and is composed of three large clades, the Andropogoneae, Paniceae with a basic chromosome number of x = 10, and the remaining Paniceae with a basic chromosome number of x = 9. Relationships among these three clades are ambiguous. Analyses using implied weights were performed with Pee-Wee under six different concavities (K = 1–6) and excluding uninformative sites. All six concavities retrieve the panicoid clade and the same three major clades (trees not shown). The consensus trees derived from K = 4, 5, and 6 present similar topologies (L = 1515, fit = 3003.5, and rescaled index = 0.46). Searches with K = 3 resulted in trees of length = 1519 and fit = 3003.7. Major differences appear when using K = 1 and 2 (L1 = 1543, fit1 = 2996.2; L2 = 1528, fit2 = 3002, respectively), though they show few differences for Panicum resolution. The rescaled index was similar among all concavities.
To test the influence of the three codon positions on phylogenetic results, the distributions of the 431 informative characters in the first, second, and third codon position were analyzed separately. Third codon positions account for most variation (53.1%), followed by first (29.1%), while the second position represents only 17.8% of the variable characters. Separate bootstrap analyses for the first, second, and third codon positions were compared with bootstrap results based on all codon positions together (Fig. 4). Consistent with the variation found among codon positions, the topology was better resolved by the third codon positions. The third codon position recovered 66 well-supported clades of the 82 clades found by the complete data set. The first position data set recovered 36 clades, while the second codon position recovered only 14 clades.
Few clades are exclusively supported by characters of the first codon position, e.g., the x = 9 Paniceae clade and the Panicum/Urochloa/Setaria clade (Figs. 4 and 5). On the other hand, more than 30 clades are supported by bootstrap analysis and support derived almost exclusively from the third codon position, e.g., the Plagiantha-Otachyrium-Steinchisma clade, Setaria clade, Urochloa clade, Forest shade clade, sect. Cordovensia clade, sect. Panicum clade, and sect. Prionitia clade, among others (Figs. 4 and 5). No clade is uniquely characterized by the second codon position.
Panicum phylogeny
The phylogeny of Panicum is practically identical in both the implied weight and equally weighted analyses; therefore, we describe the results from equally weighted characters with indels as missing data. Bootstrap values refer to the results found using all codon positions, unless explicit references are made to particular clades. Results from constrained analyses are presented in Table 3. Comments on indels, implied weight, constrained results, decay, bootstrap analyses, and also results from the 156-taxa analyses are presented when necessary.
|
Subgenus Panicum is clearly monophyletic and completely included in the x = 9 Paniceae clade (Fig. 5). Subgenus Steinchisma is exclusively within the x = 10 Paniceae clade and its monophyly requires only one more step than the shortest tree (Table 3). The monotypic subgenus Megathyrsus (Panicum maximum) is placed within the Urochloa clade. Subgenera Agrostoides, Phanopyrum, and Dichanthelium are clearly polyphyletic. Subgenus Dichanthelium is entirely incorporated in the x = 9 Paniceae clade. Making the subgenus monophyletic requires 18 more steps than the most parsimonious tree (Table 3). Subgenera Agrostoides and Phanopyrum are spread over the x = 9 and x = 10 Paniceae clades. Constraining their monophyly requires 98 and 86 extra steps, respectively (Table 3).
Though subgenera Agrostoides and Phanopyrum are spread over x = 9 and x = 10 Paniceae clades, representatives of sections Laxa, Lorea, Megista, Phanopyrum, and Stolonifera (subg. Phanopyrum), and representatives of sections Agrostoidea, Obtusa, Prionitia, Tenera, Tuerckheimiana, and Valida (subg. Agrostoides) are entirely in the x = 10 Paniceae clade (Fig. 5). On the other hand, species of sections Monticola and Parvifolia (subg. Phanopyrum), P. antidotale (not included in any section), and P. bulbosum from sect. Bulbosa (both from subg. Agrostoides) are included in the x = 9 Paniceae clade. The African species of sect. Clavelligera are also placed in the x = 9 Paniceae clade.
Paniceae x = 10 is a strongly supported clade (99% bootstrap, d > 9), divided into three major lineages (Fig. 4). One clade (82%, d = 4) includes five species of Panicum sect. Laxa related to Hymenachne donacifolia, Plagiantha tenella, Otachyrium versicolor, and three species of Steinchisma; Leptocoryphium lanatum is sister to the whole group (Fig. 5). A deletion 9-bp long characterizes the Panicum hylaeicum to Steinchisma clade (73%, d = 2; see Fig. 3 and Table 3). Relationships among species of Panicum sect. Laxa show that the section is polyphyletic. Steinchisma would have to include Panicum laxum to be monophyletic (95%, d = 3; see results of constrained analyses in Table 3).
A second major clade within Paniceae x = 10 (80% bootstrap, d = 3) includes Panicum sect. Stolonifera (subg. Phanopyrum), three species of the polyphyletic subg. Agrostoides, and representatives of Anthaenantiopsis, Axonopus, Echinolaena, Ichnanthus, Ophiochloa, Paspalum, Streptostachys asperifolia, and Thrasya (Figs. 4 and 5). Species of sect. Stolonifera form a monophyletic group, though with low support (62%, d = 1). Echinolaena inflexa is its sister group, linked in a better supported clade (78%, d = 2). However, when a weight function is applied (Pee-Wee analyses), all six concavities resolve sect. Stolonifera as paraphyletic, including Echinolaena (results not shown). In all analyses, three species of subg. Agrostoides join Anthaenantiopsis, Paspalum, and Thrasya in a well-supported clade (99%, d > 9). Although subg. Agrostoides is not monophyletic, two species, one from sect. Tuerckheimiana and one from Valida are clearly related to Anthaenantiopsis (89%, due to informative characters of the third codon position; d = 3). Panicum obtusum, sect. Obtusa, is sister to the Paspalum-Thrasya clade with low bootstrap support (62%, d = 1) or sister to the whole group that includes species of sect. Agrostoidea, Anthaenantiopsis, and the Paspalum-Thrasya clade, in all concavities in the Pee-Wee analyses (not shown). The resolution shown within the Paspalum-Thrasya clade in the 123-taxa data set is not supported by analyses of the 156 taxon (Fig. 3); results similar to the latter were shown by Giussani et al. (2001)
.
A third major clade within the x = 10 Paniceae includes several species of subgenera Agrostoides and Phanopyrum, Oplismenopsis najada, and species of the Ambiguous clade previously defined by Giussani et al. (2001)
as a poorly supported group of morphologically dissimilar taxa: Altoparadisium, Arthropogon villosus, Canastra, Homolepis, Mesosetum, Streptostachys ramosa, Tatianyx, plus Panicum euprepes (subg. Phanopyrum sect. Lorea), and P. prionitis (subg. Agrostoides sect. Prionitia). The redefined Ambiguous clade (Figs. 4 and 5), still with very low bootstrap support (56%, d = 1), is also present in all concavities of the implied weight analyses, except for K = 1 where the basal branch of this clade collapses. In all analyses, two species of sect. Tenera (subg. Agrostoides) and the monotypic sect. Megista (subg. Phanopyrum) constitute a well-supported monophyletic unit (99% bootstrap, d > 9). The third species of sect. Tenera falls in a monophyletic clade with two species of sect. Agrostoidea (99%, d > 9). To make sect. Tenera monophyletic results in a tree 35 steps longer than the current one (Table 3). Section Prionitia (subgenus Agrostoides) is monophyletic (86%, d = 2) and sister to this latter clade (98% bootstrap, d = 5, mostly supported by third position bases; Fig. 4). Section Lorea, a monophyletic unit based on two sampled species (99%, d = 8, and a synapomorphic 15-bp deletion, Table 3), is linked with Canastra lanceolata to sect. Prionitia, sect. Agrostoidea, and P. tenerum with a bootstrap value = 72% and d = 2. Panicum gymnocarpon (from the monotypic sect. Phanopyrum) represents the sister taxon of that major clade. Phylogenetic relationships of the Arthropogon villosus-Altoparadisium clade, Homolepis, Oplismenopsis najada, Streptostachys ramosa, and Tatianyx-Mesosetum clade are still not clear within the Ambiguous clade.
Paniceae x = 9 is monophyletic with 98% bootstrap support and d = 6, due almost exclusively to informative characters of the first codon position (Fig. 4). Digitaria is the sister taxon of the x = 9 Paniceae under the equally weighted character analyses and also under all concavities of the Pee-Wee analyses. This relationship is not supported by Pee-Wee K = 1, and the 156-taxa analyses where it collapses (Figs. 3–5). Digitaria is well-differentiated by 29 changes including 14 autapomorphies in both the 123- and 156-taxa trees (Figs. 3 and 4). After Digitaria, four clear monophyletic groups are included in an unresolved polytomy (Figs. 3 and 5).
The Forest shade clade, including Acroceras, Echinochloa, Lasiacis, Oplismenus, Panicum ovuliferum (subg. Dichanthelium sect. Cordovensia), and Pseudechinolaena (Giussani et al., 2001
), is recovered in these analyses and also includes two other species of sect. Cordovensia, P. cordovense and P. missionum, as well as P. penicillatum (sect. Dichanthelium; Figs. 3–5). This clade constitutes a monophyletic unit with 80% bootstrap, d = 2, and is principally supported by mutations in the third codon position. Section Cordovensia plus P. penicillatum are characterized exclusively by synapomorphies of the third codon position (75%, d = 1; Fig. 4). Pseudechinolaena ovuliferum is sister to the other three species, which form a trichotomy (64%). Resolution within the Cordovensia clade is not seen in the consensus tree of the 156-taxa analysis, where the four taxa are collapsed (Fig. 3). Results for the Forest shade clade, from Pee-Wee analyses, considering all different concavities, show Echinochloa colona and Acroceras zizanioides as sister taxa of sect. Cordovensia (not shown). These relationships are not supported when equally weighted characters are used (Fig. 4). In the bootstrap results of the 156-taxa data set, Acroceras is sister to Oplismenus, Lasiacis, and Pseudechinolaena, with very low support (53%); this relationship is not resolved in the 123-taxa analyses nor in the consensus tree of the 156-taxa analysis (Figs. 3 and 4). The other five species included in the analysis from sect. Dichanthelium (subg. Dichanthelium) are assembled in a clade with sect. Clavelligera (99%, d > 9). Both sect. Clavelligera and sect. Dichanthelium are monophyletic (100%, d > 9 and 89%, d = 2, respectively; Figs. 4 and 5).
Species from subgenus Phanopyrum, representing sects. Monticola, Parvifolia, and Verrucosa, form a clade with Sacciolepis indica (bootstrap = 82%, d = 2; Figs. 4 and 5). The two sampled species of the three that constitute sect. Monticola form a monophyletic group (95%, d = 3). Panicum cyanescens, P. parvifolium, P. schwackeanum, and P. wettsteinii, of sect. Parvifolia, also form a highly supported clade (99%, d = 8) with Sacciolepis as its sister taxon; P. trichanthum is not included in this group, making sect. Parvifolia paraphyletic. However, these four species are included in the Parvifolia clade, sharing a 15-bp deletion (Figs. 3 and 5), a homoplastic change also present in Acroceras, P. sabulorum, and sect. Lorea (Table 3).
The fourth x = 9 Paniceae subclade includes species of the Panicum/Setaria/Urochloa clade (Fig. 5; see also Giussani et al., 2001
) with 87% bootstrap and d = 4. The monotypic subgenus Megathyrsus is included in the Urochloa clade (97%, d = 6), while Panicum antidotale (unassigned to section) and P. bulbosum (sect. Bulbosa), both from subg. Agrostoides, are grouped within the Setaria clade (99%, d > 9; Figs. 4 and 5). Although not supported by bootstrap analysis, Panicum antidotale and P. bulbosum are related to two species of Setaria, S. viridis and S. lachnea, in a clade that appears in the consensus tree with d = 1 (Figs. 3–5) and that is also present under all concavities of Pee-Wee analyses (not shown).
Subgenus Panicum is highly supported as monophyletic (97% bootstrap, d = 6; Fig. 4). Section Rudgeana (99%, d = 5) and sect. Panicum are each monophyletic, although the latter has a low bootstrap value (59%). Section Panicum also appears as monophyletic in the consensus tree from NONA (Fig. 5) and in the consensus trees from K = 3–6 of Pee-Wee analyses (not shown). Section Dichotomiflora is paraphyletic within the Panicum clade, while sects. Virgata and Urvilleana form a monophyletic unit supported by informative characters of the third codon position (91%, d = 5, and also present with K = 3–6, not shown).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONTAXONOMIC TREATMENTLITERATURE CITED |
|---|
, 2000
), even though changes at the third codon position occur at higher rates than the other codon positions, they should not be discarded or downweighted a priori. When we analyzed the three data sets the topology was recovered mostly by the third position followed by the first one. The second codon position was the most conservative, as expected. However, all codon positions were required to produce the well-resolved topology found with the complete gene data set. Variation in the third codon position characterizes some monophyletic clades also supported by morphological synapomorphies as in the Plagiantha-Otachyrium-Steinchisma clade, the sect. Cordovensia clade, the Setaria clade, the Urochloa clade, and the Forest shade clade, among others (Fig. 3).
Panicum s.s
Our results show clearly that the name Panicum must ultimately be restricted to subg. Panicum; all five sections, Panicum, Virgata, Dichotomiflora, Rudgeana, and Urvilleana form a monophyletic group. This clade is very well supported and our results are congruent with previous studies (Gómez-Martínez and Culham, 2000
; Duvall et al., 2001
; Giussani et al., 2001
; J. Barber, L. Giussani, F. Zuloaga, and E. Kellogg, unpublished data).
Relationships within Panicum s.s. are, on the other hand, not completely clear. Sections Panicum and Rudgeana are both monophyletic, while further studies are needed to understand relationships among Dichotomiflora, Virgata, and Urvilleana. Section Virgata (P. tricholaenoides and P. virgatum) and sect. Urvilleana (P. chloroleucum and P. racemosum) are together in an unresolved clade and related to another clade that includes P. olyroides and P. mystasipus, both ungrouped species in previous treatments (Zuloaga, 1987a
; Zuloaga et al., 1993a
); sect. Rudgeana is a sister taxon of this clade. Section Dichotomiflora (here represented by P. aquaticum, P. dichotomiflorum, P. elephantipes, P. gouinii, P. pedersenii, and P. repens) is paraphyletic, and affinities among its species are not resolved.
Morphologically, Panicum s.s. includes caespitose species, with membranous-ciliate ligules, open panicles with spikelets on long pedicels, the spikelets with upper glume and lower lemma 7–13-nerved (Fig. 2E–F), and upper anthecium indurate, with compound papillae at the apex of the upper palea (Fig. 1B), these papillae also present in some species at the apex of the upper lemma. All species are C4, of the NAD-me subtype, with specialized chloroplasts centrifugally or centripetally located in the outer parenchymatous sheath (Fig. 1A); the basic chromosome number for subgenus Panicum is x = 9.
Sections Panicum, Dichotomiflora, and Virgata are distributed worldwide, with species in America, Africa, Europe, Asia, and Oceania. Section Urvilleana is restricted to America (with one possibly related species present in northern Africa), while species of sect. Rudgeana grow from Central to South America (Zuloaga, 1987b
).
Steinchisma
Steinchisma clearly has to be considered independent from Panicum s.s., as previously stated by several authors (Brown, 1977
; Renvoize, 1998
; Zuloaga et al., 1998
). This subgenus is characterized by the presence of an expanded and large lower palea (Fig. 2A), five-nerved upper glume and lower lemma, and upper anthecium covered with compound papillae regularly distributed over the lemma and palea (Fig. 1D). All species have a basic chromosome number of x = 10. Some of these characters are shared by sister taxa in the cladogram, such as Plagiantha and Otachyrium. Steinchisma differs from its sister taxa in being a C3/C4 intermediate (Fig. 1C).
Panicum laxum, the type species of sect. Laxa, falls within the monophyletic Steinchisma clade (Fig. 5). Zuloaga et al. (1992)
noted that, in some specimens of P. laxum, the anatomy tends to the intermediate C3/C4 type and resembles Steinchisma in having specialized chloroplasts present in the outer bundle sheath cells. Some specimens of P. laxum share morphological features with Steinchisma, such as enlargement of the lower palea and upper anthecium with compound papillae (Zuloaga et al., 1992
). Later, Zuloaga et al. (1993a)
showed in a phenetic analysis that P. laxum appeared segregated from sect. Laxa and related to species of Steinchisma; nevertheless, these authors maintained P. laxum within sect. Laxa. This study provides additional data to support exclusion of P. laxum from section Laxa and its inclusion in Steinchisma (see taxonomic treatment).
This study also shows that Plagiantha, characterized by its enlarged lower palea and upper anthecium with compound papilla all over its surface, and Otachyrium, a genus with a conspicuous enlarged lower palea, are sister to the Steinchisma clade. This relationship was previously suggested based on morphological characters by Clayton and Renvoize (1986)
and Zuloaga et al. (2000)
.
Dichanthelium
As stated above, the systematic position of Dichanthelium, as an independent genus or a subgenus of Panicum, has been extensively discussed in the last three decades. Our analysis shows sect. Dichanthelium as a monophyletic group, with the African sect. Clavelligera as its sister taxon. The two sampled species are sisters.
Our results support recognition of Dichanthelium as an independent genus, excluding from it sect. Cordovensia, and placing P. penicillatum under the latter section. The taxonomic position of these taxa is discussed below. Dichanthelium is therefore considered as an American genus, with approximately 55 species ranging from Canada and the United States through Central America, reaching the center of Argentina and Chile in South America. The genus includes taxa usually with vegetative and floral dimorphism (Fig. 2H), ligules membranous-ciliate, spikelets with the upper glume and lower lemma 7–11(–15)-nerved, exceptionally five-nerved in two species, lower palea present (Fig. 2G), upper anthecium indurate, with simple papillae all over its surface, the lemma apiculate or crested at the apex (Fig. 1G), non-Kranz anatomy and a basic chromosome number of x = 9; most of the species grow in forest edges in humid places, from sea level to 3000 m a.s.l.
Section Clavelligera includes approximately 10 African species with distinctive and conspicuous clavellate hairs on the inflorescence branches, with overall morphology similar to species of Dichanthelium (except for the foliar and floral dimorphism, absent in species of Clavelligera). Although section Clavelligera should clearly be segregated from Panicum s.l., including it in Dichanthelium should wait until more species are scrutinized.
Cordovensia clade
Section Cordovensia was established by Parodi (1925)
based on the informal group Cordovensia and was subsequently included in subg. Dichanthelium (Zuloaga et al., 1986
). Section Cordovensia is distinguished from sect. Dichanthelium by the absence of foliar dimorphism, presence of a lower glume three-quarters the length of the spikelet, and absence of a palea in the lower anthecium. Our study indicates that P. penicillatum Nees ex Trin. and sect. Cordovensia were erroneously placed within Dichanthelium (Zuloaga et al., 1993b
). All these taxa are within the Forest shade clade, together with Acroceras, Echinochloa, Lasiacis, Oplismenus, and Pseudechinolaena. Section Cordovensia, now including P. penicillatum, is morphologically similar to the genus Acroceras. Both taxa include species of forest shade, with plants decumbent and leaning in the vegetation, ligules membranous, open panicles with spikelets with the lower glume one-half to four-fifths the length of the spikelet, upper glume and lower lemma five- (occasionally seven-) nerved, lower palea present or absent, and upper anthecium crested or
