| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics |
2Cullman Program for Molecular Systematics, American Museum of Natural History, New York, New York 10024 USA; 3L. H. Bailey Hortorium and Department of Plant Biology, Cornell University, Ithaca, New York 14853 USA; 4Institute of Systematic Botany, New York Botanical Garden, Bronx, New York 10458 USA
Received for publication May 7, 2002. Accepted for publication July 19, 2002.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLiterature Cited |
|---|
Key Words: atpA • chloroplast genome inversions • Ecdeiocolea • Joinvillea • monocots • Poaceae • Poales • rbcL
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLiterature Cited |
|---|
as the Bromeliiflorae-Zingiberiflorae-Commeliniflorae (BZC) complex, plus the palms (Arecaceae) (see also Dahlgren and Rasmussen, 1983
). Dahlgren, Clifford, and Yeo (1985)
recognized five multifamily orders within the BZC complex, these being Zingiberales (eight families), Commelinales (five families), Cyperales (three families), Poales (seven families), and Typhales (two families), plus six additional orders, each of which included one family (Velloziales, Bromeliales, Philydrales, Haemodorales, Pontederiales, and Hydatellales). Several phylogenetic analyses published during the past 10 yr (e.g., Chase et al., 1993
, 2000
; Duvall et al., 1993
; Davis, 1995
; Linder and Kellogg, 1995
; Givnish et al., 1999
; Soltis et al., 2000
), and based on different taxonomic samples, have generally concurred in the addition of Dasypogonaceae and Hanguanaceae to this complex (both families were included in Asparagales by Dahlgren, Clifford, and Yeo, 1985
); in the removal of Velloziales (i.e., Velloziaceae) to a position in Pandanales; and in the placement of Arecaceae as the sister of this complex, which allows the recognition of Arecaceae either as the sister of the Commelinids or as a member of the complex. The Commelinoids of Chase et al. (2000)
thus comprise Arecaceae plus a group corresponding to a BZC complex that is slightly altered in membership from the complex recognized by Dahlgren, Clifford, and Yeo (1985)
. Because "Commelinids" and "Commelinoids" refer to the same general group, we will use the former name in this paper in the interest of uniformity.
With respect to the phylogenetic structure among the constituent groups of the Commelinids, the analyses cited above have generally confirmed the monophyly of some of the orders recognized by Dahlgren, Clifford, and Yeo (1985)
, such as Zingiberales and Typhales, while suggesting that others should be circumscribed differently. One of the most dramatic changes has been in the order Commelinales, to which Chase et al. (2000)
assigned five families (Commelinaceae, Haemodoraceae, Hanguanaceae, Philydraceae, and Pontederiaceae). Only one of these families (Commelinaceae, which provides the name for the order) was a member of the Commelinales of Dahlgren, Clifford, and Yeo (1985)
; the other four families in the order were Mayacaceae, Xyridaceae, Rapateaceae, and Eriocaulaceae, as had been suggested by Duvall et al. (1993)
and Linder and Kellogg (1995)
. In addition to this change in the recognized affinities of Commelinaceae, several of these studies also have supported the placement of this recircumscribed order Commelinales as the sister group of Zingiberales.
In the recent treatment by Chase et al. (2000)
, Arecaceae, Commelinales, and Zingiberales are regarded as monophyletic groups within the Commelinids, and all other families in this assemblage, except Dasypogonaceae, are placed in a fourth monophyletic group, which the authors designate the Poales. This assemblage (Poales sensu lato [s.l.]) includes the Poales of Dahlgren, Clifford, and Yeo (Poales sensu stricto [s.s.]) plus all the families that Dahlgren, Clifford, and Yeo had assigned to Cyperales and Typhales and various other families from the Commelinales (i.e., all members of that order except Commelinaceae) and other groups. Poales s.s. comprises seven widely recognized families (Flagellariaceae, Joinvilleaceae, Poaceae, Restionaceae, Centrolepidaceae, Anarthriaceae, and Ecdeiocoleaceae). Several analyses have placed Flagellariaceae as the sister of the rest of the group, while Joinvilleaceae has been placed as the sister of Poaceae, with Restionaceae, Centrolepidaceae, Anarthriaceae, and Ecdeiocoleaceae forming another clade, although relationships among this last group of families have been less clear.
In addition to the seven families of Poales s.s. mentioned above, two additional families, Hopkinsiaceae and Lyginiaceae, recently were proposed, each to accommodate one anomalous genus (Briggs and Johnson, 2000
). Two molecular analyses (Briggs et al., 2000
; Bremer, 2002
) concurred in resolving Hopkinsia and Lyginia as sister taxa and in placing the clade that includes them as the sister of Anarthria (the only genus of Anarthriaceae). Briggs et al. note that Lyginia and Hopkinsia are morphologically distinctive, relative to Anarthria and to each other, and suggest, for this reason, that these two genera would not be accommodated well within a more broadly defined Anarthriaceae. Hence, they recognize this group as a cluster of three monogeneric families. Although relationships among these taxa should be assessed by future studies, material of Lyginia and Hopkinsia was not available for the present study, and they are not included in the present analysis.
One group of characters that has been useful in the analysis of relationships within Poales s.s. is a set of three inversions in the plastid genome (Fig. 1), which were initially detected in comparisons between the plastid genome sequences of Oryza (Poaceae) and Nicotiana (Solanaceae) (Hiratsuka et al., 1989
; Shimada and Sugiura, 1991
). These have been designated the 28-kilobase (kb) inversion, the 6-kb inversion, and the trnT inversion. The absence of these inversions from the plastid genomes of most other monocot families, and from other lineages of embryophytes (see citations in Doyle et al., 1992
), indicated that the apomorphic state of each of the three inversions is the state that occurs in Oryza and other grasses. A survey of the taxonomic distribution of the three inversions, conducted by polymerase chain reaction (PCR) amplification with diagnostic combinations of primers situated near the ends of the inverted regions, detected all three inversions in all 15 grass species that yielded definitive results (Doyle et al., 1992
). This pattern was confirmed in another survey that included 30 grass species (Katayama and Ogihara, 1996
). Among the grasses that have been surveyed by these three studies are several species each of the BEP and PACCAD clades (cf. GPWG, 2001
), two large groups that have been resolved as sister taxa and that together include 9 of the 12 subfamilies within the grass family. All three inversions also have been observed in representatives of three additional lineages of grasses (Anomochloeae, Streptochaeteae, and Phareae) that are recognized as the earliest to diverge from the lineage that includes the BEP and PACCAD groups. These results support the conclusion that the last common ancestor of all extant grass species, like all grass species examined to date, had all three of these inversions. The trnT inversion has never been observed outside the grasses and therefore is interpreted as a synapomorphy of the family.
|
reported the presence of the 28-kb inversion in all sampled representatives of Restionaceae, Joinvilleaceae, and Ecdeiocoleaceae, and its absence in Flagellariaceae, Anarthriaceae, and all sampled taxa outside Poales s.s. (see Table 1). Katayama and Ogihara (1996)
confirmed the presence of the 28-kb inversion in grasses, Joinvilleaceae, and additional species of Restionaceae and its absence in Flagellariaceae and Anarthriaceae and all sampled taxa outside Poales s.s. They also reported its absence in Centrolepidaceae and in two species of Restionaceae (see Table 1).
|
; Doyle et al., 1992
) and therefore is inferred to have originated after the 28-kb inversion. From this it is predicted that the 6-kb inversion should occur in some or all of the taxa that have the 28-kb inversion, but not in any taxon that lacks the 28-kb inversion. This prediction has been borne out by previous reports of its universal occurrence, along with the 28-kb inversion, in Poaceae, its occurrence outside Poaceae only in Joinvilleaceae, which has the 28-kb inversion, and its absence in Flagellariaceae, Anarthriaceae, Centrolepidaceae, and Restionaceae, plus all taxa outside Poales s.s. (Doyle et al., 1992
; Katayama and Ogihara, 1996
). Doyle et al. included a representative of Ecdeiocoleaceae in their survey and detected the presence of the 28-kb inversion, as well as the absence of the trnT inversion, but obtained no amplification with primers diagnostic for either the presence or absence of the 6-kb inversion. The reported distribution of the 28-kb inversion supports the placement of Poaceae in a monophyletic assemblage (within Poales s.s.) that includes Joinvilleaceae, Ecdeiocoleaceae, and some elements of Restionaceae and that excludes Flagellariaceae, Anarthriaceae, Centrolepidaceae, and certain other elements of Restionaceae (Fig. 1). Within this assemblage, the distribution of the 6-kb inversion supports the placement of Poaceae in a smaller group that includes Joinvilleaceae, excludes all elements of Restionaceae, and may or may not include Ecdeiocoleaceae, for which there has been no observation, either of its presence or absence.
In the present paper we examine the phylogenetic structure of Poales s.l., with particular attention to the structure of Poales s.s. We extend the existing survey of the taxonomic distribution of the three inversions in the plastid genome and report, among other findings, the presence of the 6-kb inversion in Ecdeiocoleaceae. Data on the presence/absence of the three inversions are included in a structural character set that, in combination with two DNA sequence character sets, resolves a clade that consists of Joinvilleaceae, Ecdeiocoleaceae, and Poaceae, with the latter two families more closely related to each other than either is to Joinvilleaceae.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLiterature Cited |
|---|
were included in the study (see Appendix 1 at http://ajbsupp.botany.org/v90). The analysis also includes three families (Bromeliaceae, Mayacaceae, and Rapateaceae) that were included in the Commelinids in 1998, though not assigned to an order within that group, and that were included in Poales in the amended classification of Chase et al. (2000)
. We also include two of the remaining three Commelinid families that were not assigned to orders in 1998 (Dasypogonaceae, plus "Abolbodaceae," by virtue of our inclusion of Abolboda and Orectanthe, which are assigned to Abolbodaceae when it is segregated from Xyridaceae; see Kral, 1998
). We do not include the sixth of these unplaced families, Hanguanaceae, which is not believed to be closely related to the Poales (e.g., Rudall, Stevenson, and Linder, 1999
; Chase et al., 2000
); this family has been resolved repeatedly as a member of a clade that includes Zingiberales and Commelinales, and it is difficult to assign to an order only because, within this clade, its precise relationship to these orders is uncertain. Dasypogonaceae is unplaced because its position within the Commelinids is uncertain (e.g., Chase et al., 2000
, where it is variously placed as sister of Arecaceae or as sister of all of the Commelinids except Arecaceae), and it is included here as an outgroup (Nixon and Carpenter, 1994
). The matrix also includes Hydatellaceae, a family of uncertain affinity; it was not assigned to any suprafamilial group within the monocots in the APG classification (1998)
or in the classification of Chase et al. (2000)
. However, it was resolved among the Commelinids by the analyses of Chase et al. (1995)
and Stevenson et al. (2000)
. The data matrix includes a total of 40 terminal taxa.
Data matrix
The overall data set includes structural characters (morphological, anatomical, and chloroplast genome inversions) and DNA sequence data from two loci (atpA from the mitochondrial genome, rbcL from the plastid genome). The morphological portion of the matrix (Tables 2, 3) is a subset of a larger data set that includes all families of the monocotyledons (D. W. Stevenson et al., unpublished data), and it has been derived from previously published matrices (Chase et al., 1995
; Linder and Kellogg, 1995
; Stevenson and Loconte, 1995
; Rudall, Stevenson, and Linder, 1999
; Stevenson et al., 2000
). Because the coding in the most inclusive of these matrices was meant to represent entire families as terminals, rather than genera or individual species, the data have been revised to represent the particular genera in the present study, drawing especially from family-level analyses within the Poales (Soreng and Davis, 1998
; Linder, Briggs, and Johnson, 2000
; Linder and Caddick, 2001
) and our own new observations. Additionally, the character presence/absence of ligule was rescored under a more restrictive definition; we interpreted as ligules only those processes present on the adaxial surface of the leaf. Lateral flanges or those on the abaxial side of the leaves are not interpreted as ligules in this study, because it is not possible to establish whether or not these structures are homologous with those in the adaxial side of the leaves as present in most Poaceae.
|
|
, Linder and Rudall (1993)
, and Linder and Kellogg (1995)
. Linder and Rudall coded this character as binary, and we recognize it here as a multistate character (operculum absent, operculum present and covering the entire pore, pore covered by loose, coarse granules). Linder and Kellogg (1995)
coded this character similarly, but whereas they considered Restionaceae as having a solid operculum, we are coding the different genera for which there is data available as having loose coarse granules. In addition, new observations were made on the occurrence of three inversions in the large single-copy region of the chloroplast genome, which are present in Poaceae and closely related families (Doyle et al., 1992
; Katayama and Ogihara, 1996
), and these inversions were included as three characters in the structural matrix.
Chloroplast genome inversions
Genomic DNAs used to determine the presence/absence of three chloroplast genome inversions were isolated from silica-dried and herbarium material, using the method of Doyle and Doyle (1987)
or the Qiagen (Venlo, The Netherlands) DNeasy plant mini kit following the manufacturer's protocols. Alternatively, the DNA extraction method of Struwe et al. (1998)
was employed for silica-dried or cetyltrimethylammonium bromide (CTAB)-preserved leaf material (Rogstad, 1992
). The same methods were employed to isolated DNAs used to sequence the two genes detailed below.
Presence/absence of each of the three inversions in the chloroplast genome was determined using primer pairs specific for each condition. In each case, one priming site was outside the inversion and the other was in the adjacent region immediately inside the inversion, or alternatively (to detect the uninverted state), the second primer was in a similar position in the uninverted sequence (Fig. 1). Primer sequences were obtained from Doyle et al. (1992)
.
Double-stranded PCR amplifications were conducted with a Perkin-Elmer (Wellesley, Massachusetts, USA) GeneAmp 9600 thermal cycler. In all cases, 40-µL reactions were run for a total of 35 cycles of 94°C (30 s), 52–58°C (40 s), and 72°C (1 min). All reactions terminated with an additional 3 min at 72°C. The reactions contained Promega (Madison, Wisconsin, USA) 10x buffer, magnesium (1.5–2 mmol/L), deoxynucleotide triphosphates (1.25 mmol/L each), primers (0.5 µmol/L each), 1.5 units of Promega Taq polymerase, and 0.5–1.5 µL of template. Additionally, in selected cases, BSA at a final concentration of 0.1 µg/µL was used as an additive to the PCR mix. Annealing temperatures were, in general, lower than those reported by Doyle et al. (1992)
because this provided stronger products that were easier to detect, without the appearance of multiple bands. PCR products were separated by electrophoresis in 1% Tris borate/EDTA agarose gels containing ethidium bromide for later visualization under UV light.
DNA sequencing
Two loci were used in this study: atpA (from the mitochondrial genome) and rbcL (from the plastid genome). Sequence data were obtained from publicly available sources (Genbank, in which some atpA sequences are listed as atp1) or in a few cases provided by other investigators or generated for this study using published primers (atpA: Eyre-Walker and Gaut, 1997
; Davis et al., 1998
; rbcL: Chase et al., 1993
; Asmussen and Chase, 2001
) following standard PCR and automated cycle sequencing protocols.
The portion of rbcL included in this study comprises 1371 aligned bases, from positions 31 to 1401 of the coding region of Genbank accession D00207 (Oryza sativa). The portion of atpA comprises 1379 aligned bases, corresponding to 1256 sites (305–1560) of Genbank accession X51422, also Oryza sativa (Kadowaki, Kazama, and Suzuki, 1990
), plus inferred gaps totaling 123 sites. Length variation occurs in three general regions of the gene (J. I. Davis et al., unpublished data), and four aligned indels were included as characters in the atpA portion of the matrix (Simmons and Ochoterena, 2000
; Stevenson et al., 2000
), only one of which is cladistically informative among the 40 taxa in this analysis. Sequence data for atpA is missing for three taxa in the matrix (Trithuria, Aphelia, and Leptocarpus), for which repeated attempts yielded no sequence, except (in some cases) anomalous fragments with internal stop codons. One of the two representatives of Cyperaceae could not be identified definitively. The DNA for this taxon was isolated from a living collection at RBG Kew which was originally accessioned as Lagenocarpus sp. The plant died before flowering, but DNA from the same individual had been collected and used in an rbcL analysis of Cyperaceae by David Simpson. In his analyses this individual is resolved as sister of a species of Cladium (D. Simpson, personal communication). In analyses conducted for the present study (rbcL only, atpA only, and all combined analyses) this terminal always is placed as sister of Carex, the only other representative of Cyperaceae. Hence, we identify it as "aff. Cladium." Because of uncertainty regarding its precise identity, it was scored as a generalized member of Cyperaceae for the morphological matrix.
Voucher information and Genbank accession numbers for the taxa included in this study, including accessions used for the inversion survey but not in the cladistic analysis, can be found in the supplementary data at http://ajbsupp.botany.org/v90.
Phylogenetic analyses
Cladistic analyses of the combined data set were conducted with NONA (Goloboff, 1993
). Tree searches were conducted with 2000 random taxon-entry sequences, with ten trees held per iteration, followed by swapping of shortest trees to completion. The combined data matrix also was analyzed using the parsimony ratchet (Nixon, 1999a
) as implemented in Winclada (Nixon, 1999b
). Trees derived from four different ratchet runs of 400 iterations each were pooled and swapped to completion in NONA. Jackknife support values (Farris et al., 1996
) were calculated based on 2000 replicates. Each replicate was performed with 100 random taxon-entry sequences, with three trees held from each initiation, and the set of shortest trees then swapped to and through up to 1000 resulting trees, with the consensus tree retained in each case. This approach to jackknife computation counts a group as having been resolved by a given replicate only if the group occurs in the consensus tree retained from that replicate; thus, scores may be lower, but not higher, than those obtained by approaches that assign proportional scores to groups that occur in some but not all trees from a given replicate (for similar approaches using the bootstrap, see Linder, 1991
; Davis et al., 1998
). Two constrained analyses of the combined data set also were conducted to examine the implications of particular relationships that were not obtained by unconstrained analysis. The constrained analyses were conducted by adding to the matrix a heavily weighted dummy character scored 1 for taxa in the intended group and 0 for all other terminals. A third constrained analysis was conducted in a similar manner, to examine the implications of non-homoplasy in a character (the 28-kb inversion) that was suggested to be homoplasious by the results of the unconstrained analysis. In this case, the weight of this character was increased to enforce non-homoplasy in it, but the placements of taxa that were scored as unknown for this character were not constrained. Lengths of trees obtained by the constrained analyses are reported after subtracting steps in the dummy characters. Winclada also was used for morphological data entry, matrix editing, tree viewing, and calculation of the consistency index (CI; Kluge and Farris, 1969
) and retention index (RI; Farris, 1989
) for trees obtained in each search.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLiterature Cited |
|---|
|
; Katayama and Ogihara, 1996
). The distributions of these inversions among the seven families of Poales s.s., and a few additional families, also are depicted in Fig. 1. None of the three inversions has ever been observed in any taxon outside the Commelinids, and previously published observations for non-Commelinid taxa are not included in Table 1; similarly, every accession of Poaceae examined to date has all three inversions, and previously published observations for grasses also are not included in Table 1, though new observations for some of these taxa are included; new observations agree with previously published observations in all cases, except for a possible discrepancy in Elegia, as discussed below. In a few cases no amplification was observed for primer combinations diagnostic for either the presence or absence of an inversion in a particular taxon (e.g., the 6-kb inversion in Bambusa). Apart from these instances, in which information is lacking, the set of taxa observed to have the trnT inversion is a subset of those that have the 6-kb inversion, and the set of taxa with the 6-kb inversion is a subset of those that have the 28-kb inversion.
Inversion 1 (28-kb)
The taxonomic distribution of the 28-kb inversion was surveyed primarily by amplification at the rps14 end (Fig. 1). The PCR using primers flanking the trnG-UCC end often yielded multiple bands, so definitive observation of the presence of the first inversion at this end was not always possible. Similar results were obtained by Doyle et al. (1992)
. The 28-kb inversion was found to be present in all taxa surveyed for the present study from Restionaceae, Joinvilleaceae, Ecdeiocoleaceae, and Poaceae, but absent from all other taxa for which results were obtained, including representatives of Hydatellaceae, Mayacaceae, Rapataceae, Xyridaceae (including Abolbodaceae), and Cyperaceae. In addition to the two genera of Restionaceae surveyed by Doyle et al. (1992)
, Askidiosperma (as Chondropetalum) and Rhodocoma, we found this inversion to be present in Baloskion, Elegia, Leptocarpus, and Thamnochortus. However, no results were obtained, either for the presence or absence of this inversion, in Lepyrodia (Restionaceae) or Aphelia (Centrolepidaceae), despite our attempts with three different accessions of Aphelia. The 6-kb and trnT inversions, however, are absent in both of these genera (see below). In spite of repeated attempts, results also have not been obtained for any of the inversions in Sparganium (Sparganiaceae) or Prionium (Prioniaceae). These are the only families of Poales s.l. (except for Lyginiaceae and Hopkinsiaceae) that remain unscored for all three inversions.
Inversion 2 (6-kb)
The taxonomic distribution of the 6-kb inversion was surveyed primarily on the basis of the end that is adjacent to trnS (Fig. 1). The third inversion, which includes trnT, is situated beyond the other end of the second inversion, and because the available primers for that end of the second inversion are located inside the region affected by the trnT inversion, more than two primer pairs are needed to assess the occurrence of the 6-kb inversion at that end (i.e., the orientation of priming locations within trnT, and hence the ability to detect the 6-kb inversion at that end, are affected by the presence or absence of the trnT inversion). The 6-kb inversion was observed to occur only in Poaceae, Joinvilleaceae, and Ecdeiocoleaeceae and to occur in all sampled taxa from each of these families. Similar results for the first two of these families were reported previously by Doyle et al. (1992)
and by Katayama and Ogihara (1996)
. However, Doyle et al. never observed amplification of Ecdeiocolea with primers diagnostic for either the presence or the absence of the second inversion, and Katayama and Ogihara did not include a representative Ecdeiocoleaceae in their study, so there is no previous report for this family. We detected the second inversion in all three accessions of Ecdeiocolea that we surveyed.
Inversion 3 (trnT)
Occurrence of the trnT inversion was surveyed exclusively at the end adjacent to trnE, and it was found to occur in all genera of Poaceae that were examined; all other taxa examined yielded a positive PCR product with the trnE + trnT-r primer combination, which is diagnostic for the absence of this inversion.
Phylogenetic analyses
The combined data set included 661 cladistically informative characters, 563 from the DNA sequence data set (including one informative indel in atpA) and 98 from morphology and other structural characters (Tables 2, 3). Of the 3920 cells in the morphology matrix, 9% are scored as missing, 2% as inapplicable, and 1% as polymorphic, for a total of 12%. Of the 563 informative characters in the DNA sequence portion of the overall matrix, 351 are nucleotide sequence characters from rbcL, 211 are sequence characters from atpA, and one is an atpA insertion/deletion character (Appendix 2; http://ajbsupp.botany.org/v90).
Cladistic analysis of the morphological data set yielded 22 most-parsimonious trees (MPTs) of length 308, with CI = 0.39 and RI = 0.73, and with 21 clades resolved in the consensus tree (not shown) (tree length and CI calculated here and elsewhere with autapomorphies excluded; number of clades resolved calculated with the artifactual node between Baxteria and all other terminals collapsed). With the tree rooted between Baxteria and the remaining taxa, the taxa outside the Dasypogonaceae are resolved as four groups (Trithuria plus three multispecies clades) that arise from a polytomy. The first of the three clades consists of Sparganium and Typha; the second consists of the three representatives of Bromeliaceae as sister of a clade that includes the various representatives of Rapateaceae, Eriocaulaceae, Mayacaceae, and Xyridaceae; and the third, which includes all remaining taxa, has at its base a polytomy from which four subgroups diverge. These four groups are Thurnia; a clade consisting of the two species of Cyperaceae; a clade consisting of Prionium plus the two species of Juncaceae; and a clade consisting of all taxa of Poales s.s. Within the latter group there is a multichotomy from which seven groups diverge. Five of these are individual terminals (Flagellaria, Anarthria, Aphelia, Joinvillea, and Ecdeiocolea), and the two other groups are Restionaceae and Poaceae, each of which is resolved as monophyletic.
Analysis of the molecular data set resulted in six MPTs of length 1933, with CI = 0.42 and RI = 0.60, and with 31 clades resolved in the consensus tree (not shown). When the trees are rooted with Baxteria, a clade consisting of Bromeliacae, Sparganiaceae, and Typhaceae is sister to the remaining Poales s.l. Within the latter group, Rapateaceae are sister of a group in which there is a polytomy among four clades and Flagellaria. The first clade consists of Mayacaceae, Eriocaulaceae, Xyris, Hydatellaceae, Prioniaceae, Thurniaceae, Juncaceae, and Cyperaceae. The two other genera of Xyridaceae in the study (Abolboda and Orectanthe) constitute the second of the four clades. The third clade comprises Joinvillea, Ecdeiocolea, and Poaceae in an unresolved trichotomy, and the fourth clade consists of Anarthria, Aphelia, and Restionaceae. A review of the structures of the six most-parsimonious trees indicates that this large polytomy results from the alternative placements of the clade consisting of Abolboda and Orectanthe as sister of the Anarthria-Aphelia-Restionaceae clade or as sister of the clade that includes Eriocaulaceae, Cyperaceae, and other taxa.
Analysis of the combined data set yielded two MPTs of length 2273, with CI = 0.41 and RI = 0.62, and with 36 clades resolved in the consensus tree, a greater number than was resolved by the structural (21) or sequence (31) portions of the data set when analyzed separately. Figure 3 depicts the strict consensus of the two most-parsimonious trees resolved by the combined data set. In this tree, when rooted with Baxteria, the deepest point of divergence outside Dasypogonaceae is between a clade consisting of Rapateaceae, Xyridaceae, Eriocaulaceae, Mayacaceae, and Hydatellaceae, and another that includes all remaining taxa of Poales s.l. Among the remaining taxa, Bromeliaceae are sister of a clade that includes all other taxa, and within the latter group a clade consisting of Sparganium and Typha is sister of a group that consists of two clades, one including Prionium, Thurnia, Juncaceae, and Cyperaceae, and the other comprising Poales s.s. (i.e., Flagellariaceae, Anarthriaceae, Restionaceae, Centrolepidaceae, Joinvilleaceae, Ecdeiocoleaceae, and Poaceae). Within Poales s.s., Flagellaria is resolved as the sister of all other members of the group, which fall into two sister clades. In the first of these clades Anarthria is the sister of a clade in which Aphelia is sister of a clade that comprises all Restionaceae. In the second clade Joinvillea is sister of a clade in which Ecdeiocolea is sister of a group that includes all Poaceae.
|
The clade consisting of Joinvillea, Ecdeiocolea, and Poaceae, with a jackknife frequency of 99%, has three unambiguous structural synapomorphies, characters 18 (gain of ligule), 24 (differentiation of leaf epidermal cells into long and short cells), and 95 (presence of the 6-kb inversion in the plastid genome). Epidermis differentiated into short and long cells has been interpreted traditionally to be restricted to Poaceae and Joinvilleaceae. However, new observations (by D. W. Stevenson) showed that this character is also present in Ecdeiocolea but not in other taxa of Poales s.s. (i.e., Restionaceae, Anarthriaceae, etc.). Within this group, the sister group relationship of Ecdeiocolea with Poaceae has a jackknife frequency of 49% and two unambiguous synapomorphies, characters 49 (loss of scrobiculi) and 93 (gain of an operculum on the pollen pore).
Within the Joinvillea-Ecdeiocolea-Poaceae clade, jackknife frequencies from the unconstrained analysis of the combined data set, for two alternatives to the placement of Ecdeiocolea with Poaceae, are 14% for the placement of Joinvillea with Ecdeiocolea and 8% for the placement of Joinvillea with Poaceae. These alternative relationships also were examined in constrained analyses. Analysis of the combined data set, with Joinvillea constrained to be sister of Ecdeiocolea, yielded 14 trees of length 2277, four steps longer than those obtained by unconstrained analysis. Poaceae are sister of the forced Joinvillea-Ecdeiocolea clade in all of these trees, and no structural character is an unambiguous synapomorphy of the grouping of Joinvillea and Ecdeiocolea in any of these trees. Analysis of the combined data, with Joinvillea constrained to be sister of Poaceae, yielded two trees of length 2275, two steps longer than those obtained by unconstrained analysis. Ecdeiocolea is sister of the forced Joinvillea-Poaceae clade in both of these trees, and there is one unambiguous structural synapomorphy for the grouping of Joinvillea with Poaceae, character 19 (gain of pubescence).
The grouping of Anarthria, Aphelia, and Restionaceae is supported by four unambiguous structural synapomorphies, characters 0 (root hairs originating from any epidermal cell), 28 (floral dicliny, i.e., unisexual flowers), 34 (loss of the outer stamen whorl), and 38 (dorsifixed anthers). Within this group, the placement of Aphelia with Restionaceae is supported by three unambiguous synapomorphies, characters 12 (presence of compound starch grains in the embryo sac), 41 (one theca per anther), and 68 (presence of elongate cells in the nucellar epidermis).
An analysis of the combined data set also was conducted with the 28-kb inversion constrained to be non-homoplasious, by increasing the weight of this character. No scores were changed, so the positions of taxa that are scored as unknown for this character were not constrained. This analysis yielded 16 trees of length 2283 (after subtraction of extra steps attributable to the increased weighting), or seven steps longer than those obtained by unconstrained analysis. A clade consisting of all representatives of Centrolepidaceae, Restionaceae, Joinvilleaceae, Ecdeiocoleaceae, and Poaceae (i.e., all taxa scored as having this inversion, plus Aphelia, Elegia, and Lepyrodia) was resolved in all MPTs. The only unambiguous structural synapomorphy of this group is the 28-kb inversion, the character that was used to constrain the group. The jackknife support value of this group, in the combined data set, analyzed without constraint, is 4%.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLiterature Cited |
|---|
and the accompanying taxonomic treatment, which differs somewhat from the earlier treatment of the APG (1998)
, are useful benchmarks to which other analyses and taxonomic systems can be compared, including those of Dahlgren, Clifford, and Yeo (1985)
.
Chase et al. (2000)
recognize four orders among the Commelinids, plus one additional unplaced family within the Commelinids, Dasypogonaceae. They also recognize Hydatellaceae as one of three families that are included within the monocots, but not assigned to a suprafamilial grouping. This family, which earlier had been included in Poales s.l. by the APG (1998)
, after having been resolved as a member of that group by Chase et al. (1995)
, as sister of Typhaceae-Sparganiaceae (cf. their Fig. 4), and by Stevenson et al. (2000)
, as the sister of Xyridaceae, is thus a reasonable candidate for inclusion in the Commelinids and specifically in Poales s.l. Monophyly of each of the three other Commelinid orders (Arecales, Commelinales, and Zingiberales) was strongly supported in the analysis of Chase et al. (2000)
, and each of these groups was mutually exclusive in membership with Poales s.l., which included 17 families, or 19 families if Prioniaceae is segregated from Thurniaceae and Abolbodaceae from Xyridaceae. Thus, apart from the recently segregated Lyginiaceae and Hopkinsiaceae (Briggs and Johnson, 2000
), the group of candidate taxa for membership in Poales is this set of 17 (or 19) families of Poales s.l., plus Dasypogonaceae and Hydatellaceae, for a total of 19 (or 21) families. All of these families are included in the present analysis, including the two segregates, Prioniaceae and Abolbodaceae, plus representatives of the two families from which they often are segregated (Thurniaceae s.s. and Xyridaceae s.s., respectively).
Unconstrained cladistic analysis of the combined matrix of three character sets (atpA, rbcL, and morphology) has resulted in a well-resolved hypothesis of relationships within Poales s.l., and the results of this analysis are more fully resolved than those obtained with either the structural or molecular subset of the overall character set. With the exception of Xyridaceae s.l., all families that are sampled more than once in the combined analysis are resolved as monophyletic. If Xyridaceae is circumscribed narrowly, so as to exclude Abolbodaceae (represented here by Abolboda and Orectanthe), there is no test of monophyly of Xyridaceae s.s., for the family is represented just once (by Xyris), while Abolbodaceae, represented by two genera, are resolved as monophyletic. The combined analysis therefore is consistent with the segregation of a monophyletic Abolbodaceae from Xyridaceae, but it supports the monophyly of all other families represented by more than one terminal.
One noteworthy result of the present analysis is the placement of Trithuria (Hydatellaceae) within a clade that otherwise consists of representatives of Xyridaceae s.l., Eriocaulaceae, and Mayacaceae. Hydatellaceae were sampled, but only with morphological characters, in the combined analysis of Chase et al. (1995)
, and as noted above they were placed as sister of a Sparganiaceae-Typhaceae clade. The present analysis does not include representatives of all Commelinid lineages, or of any monocot groups outside the Commelinids, so it does not provide a rigorous test of the precise placement of Trithuria. However, the analysis of Stevenson et al. (2000)
, which included representatives of all major lineages of the monocots, placed Trithuria in the Commelinids, within a group that corresponds to Poales s.l., as sister of Xyris. Therefore, that analysis and the present one provide complementary evidence supporting this placement of Hydatellaceae (broader sampling of monocot lineages by Stevenson et al. vs. more intensive sampling of Poales in the present analysis); hence, we provisionally recognize Hydatellaceae as an element of Poales s.l.
With respect to other higher level relationships, the combined analysis provides strong jackknife support for several multifamily groupings within Poales s.l., while most relationships among these groups, though resolved in the consensus tree, have relatively weak support. Some of the major groups resolved here, possibly warranting recognition as orders, are: (1) Rapateaceae-Xyridaceae-Eriocaulaceae-Mayacaceae-Hydatellaceae; (2) Bromeliaceae; (3) Sparganiaceae-Typhaceae; (4) Prioniaceae-Thurniaceae-Juncaceae-Cyperaceae; and (5) Poales s.s. These groups correspond in many respects to those recognized by Dahlgren, Clifford, and Yeo (1985)
and resolved by Givnish et al. (1999)
, Chase et al. (2000)
, and Bremer (2002), but there are some inconsistencies. The focus of the present work is on Poales s.s., and we defer further discussion of higher level relationships in the Commelinids to a more comprehensive forthcoming work.
One of the major groups resolved within Poales s.l. is a clade of seven families corresponding in membership to Poales s.s. as delimited by Dahlgren, Clifford, and Yeo (1985)
. This group is resolved by the combined data set as well as by the structural and molecular character sets when analyzed separately. Among recent studies, the analyses of Givnish et al. (1999)
and Chase et al. (2000)
each included representatives of four of these seven families, and in both cases the four families were resolved as a monophyletic group. Bremer (2002) sampled all seven families of Poales s.s., plus Lyginia and Hopkinsia, and resolved the group as monophyletic. Groups corresponding to Poales s.s., based on various sets of taxa and characters, also have been resolved by Linder and Rudall (1993)
, Davis (1995)
, Linder and Kellogg (1995)
, Stevenson and Loconte (1995)
, Chase et al. (1995)
, Katayama and Ogihara (1996)
, Rudall, Stevenson, and Linder (1999)
, and Soltis et al. (2000)
. The present analysis thus agrees with these earlier studies in supporting the recognition of this group as monophyletic.
All three of the plastid genome inversions first observed in Oryza are limited in distribution to Poales s.s. and to only a subset of the families within this group. The various surveys that have been conducted of the distribution of these inversions (Doyle et al., 1992
; Katayama and Ogihara, 1996
; and the present paper) have covered all families of Poales s.l. except for Prioniaceae and Sparganiaceae (both of which have been placed outside Poales s.s. by several analyses) and the recently proposed families Lyginiaceae and Hopkinsiaceae (Briggs and Johnson, 2000
), which almost certainly belong within Poales s.s. Prioniaceae appear to be closely allied to Thurniaceae, which lacks all three inversions, just as Sparganiaceae are allied with Typhaceae, in which the 28-kb inversion and the trnT inversion are lacking, and in which unequivocal results for the 6-kb inversion have not been obtained. Hence, neither Prioniaceae nor Sparganiaceae are likely candidates for inclusion in Poales s.s. and are not expected to have any of the three inversions. The situation with Lyginiaceae and Hopkinsiaceae is not as clear; they are closely related to taxa that have at least one of the inversions, and sampling of these two groups is warranted.
The present survey of the taxonomic distribution of the three inversions (Table 1, Fig. 1) largely confirms the previous results of Doyle et al. (1992)
and Katayama and Ogihara (1996)
. The known occurrences of all three inversions are limited to four of the seven generally recognized families of Poales s.s., though some members of at least two of these families, as currently circumscribed, may be polymorphic. Available observations indicate absence of all three inversions from all sampled representatives of Flagellariaceae (of which Flagellaria is the sole genus), Anarthriaceae (of which Anarthria is the sole genus), and Centrolepidaceae (a family of three or four genera, two of which have been sampled, though unambiguous results have not been obtained for the 28-kb inversion in Aphelia). We have no reason to doubt the observations of Katayama and Ogihara. However, in the absence of confirmation that the 28-kb inversion is absent from Centrolepidaceae and in light of the placement of Aphelia with elements of Restionaceae that have this inversion (Fig. 3), further study of Centrolepidaceae appears to be in order, and we signify the implied phylogenetic placement of this family with a dashed line in Fig. 1. In contrast to the reported absence of all three of the inversions in Flagellariaceae, Anarthriaceae, and Centrolepidaceae, all unambiguous results obtained from a broad sampling of Poaceae indicate that all three inversions are present throughout this family. Of the three remaining families, Joinvilleaceae (of which Joinvillea is the sole genus) and Ecdeiocoleaceae (a family of two genera, of which only Ecdeiocolea has been sampled) have both the 28-kb and the 6-kb inversions and lack the trnT inversion.
Restionaceae, the last of the seven families of Poales s.s., is the only one for which there is direct evidence of polymorphism for any of the inversions, in the form of reports of its presence in some taxa and absence in others. Most representatives of Restionaceae that have been sampled are reported to have the 28-kb inversion, but two are reported to lack it (Table 1). As with Aphelia, the analysis conducted with this inversion forced to be non-homoplasious places all elements of Restionaceae within the clade that has the inversion, and we use a dashed line in Fig. 1 to signify the placement of representatives of this family that reportedly lack the 28-kb inversion.
The various observations of the three inversions, in summary (Table 1, Fig. 1), suggest that Poaceae is monophyletic (universal presence of the trnT inversion); that Ecdeiocoleaceae, Joinvilleaceae, and Poaceae constitute a more inclusive clade (presence of the 6-kb inversion); and that these three families, along with at least a subset of Restionaceae, form a clade that does not include Flagellariaceae, Anarthriaceae, or Centrolepidaceae and that also may not include some elements of Restionaceae.
Although the 6-kb and the trnT inversions are optimized as non-homoplasious in the unconstrained phylogenies obtained in this study and are interpreted as synapomorphies of Poaceae and the Joinvillea-Ecdeiocolea-Poaceae clade, respectively, the 28-kb inversion is optimized as homoplasious because it is present in the Joinvillea-Ecdeiocolea-Poaceae clade and in at least some Restionaceae (reportedly absent in other Restionaceae), unobserved in Aphelia (though reportedly absent in Centrolepis), yet absent from Anarthria, which is resolved as the sister of the Centrolepidaceae-Restionaceae clade. Alternatively, as discussed above, the constrained analysis with the 28-kb inversion forced to be non-homoplasious yields a different set of relationships. Hence, there is a conflict between the distribution of this inversion and the rest of the data set. In the unconstrained analysis of the complete data set, a clade that is consistent with non-homoplasy of the 28-kb inversion has a jackknife frequency of just 4%; moreover, when a clade consistent with the non-homoplasy of this character is constrained to be resolved, no other structural character in the matrix is found to be an unambiguous synapomorphy of this group. A key element of the conflict between the distribution of the 28-kb inversion and the overall structure of the data set is evident in the three unambiguous structural synapomorphies that link Aphelia with Restionaceae, and the four additional unambiguous structural synapomorphies that link this clade with Anarthria, rather than with other taxa that have the 28-kb inversion.
An additional complication in this overall pattern is presented by the evidence of polymorphism for the 28-kb inversion within Restionaceae. If confirmed, the presence of the 28-kb inversion in some elements of Restionaceae, and its absence in others, would suggest that the family is not monophyletic. Genera of Restionaceae reported to have the 28-kb inversion in at least one species are Baloskion, Askidiosperma, Elegia, Leptocarpus, Rhodocoma, and Thamnochortus, and genera reported to lack this inversion in at least one species are Desmocladus and Elegia. Elegia occurs on both lists, with the 28-kb inversion reportedly present in Elegia fenestrata (this paper) and absent in Elegia cuspidata (Katayama and Ogihara, 1996
). Information regarding this inversion is lacking for Lepyrodia, which is reported to lack the other two inversions, like all other members of the family. The only other genus for which there is a report is "Restio" (Katayama and Ogihara, 1996
), but the species in which the 28-kb inversion is reportedly present is Restio tetraphyllus, which has been transferred to the genus Baloskion, as B. tetraphyllum, a species for which we report presence of the inversion in this paper. The two reports for this species, though under different generic names, are in agreement with each other, and it should be noted that there is, as yet, no report for Restio s.s.
The present phylogenetic analysis is not the first to produce results that are inconsistent with the known distribution of the 28-kb inversion. This distribution also is inconsistent with the phylogenies resolved by Briggs et al. (2000)
, Linder (2000)
, Linder, Briggs, and Johnson (2000)
, and Bremer (2002). For example, Desmocladus, which reportedly lacks the inversion, is nested among several taxa that have the inversion (suggesting a gain followed by a loss) in the phylogenetic hypothesis of Briggs et al. (2000)
. Similarly, both presence and absence of the inversion are reported for taxa within two of the seven major clades resolved by Linder (2000)
(the Restio clade and the Coleocarya-Harperia clade), and in a like manner both presence and absence are observed within the Desmocladus-Centrolepidaceae-Loxocarya clade and the African clade of Linder, Briggs, and Johnson. Finally, the trees resolved by Bremer (2002) place all sampled genera of Restionaceae (with Centrolepidaceae nested among them) with Anarthria, Lyginia, and Hopkinsia, in a clade that is sister to one that includes Flagellariaceae, Joinvilleaceae, Ecdeiocoleaceae, and Poaceae.
Rather than speculate on the relative likelihood of multiple gains or losses of this inversion, as opposed to the various phylogenetic hypotheses that are inconsistent with the reported distribution of this character, we would again mention the need for verification of the various reports for Restionaceae and Centrolepidaceae and for broader sampling within these and related families, including Hopkinsiaceae and Lyginiaceae.
The occurrence of both the 28-kb and 6-kb inversions in Ecdeiocoleaceae supports the placement of this family, along with Joinvilleaceae, in a clade that otherwise includes only Poaceae, and a clade with this composition is, indeed, resolved by the combined and separate unconstrained analyses. These three families also were resolved as a monophyletic group by Bremer (2002). It has become conventional in recent years to regard Joinvillea as the sister of the grass family, but the present results suggest that this position may be occupied by Ecdeiocoleaceae or by a clade consisting of Joinvilleaceae and Ecdeiocoleaceae. One reason that this three-family grouping has not been observed frequently is that these three families have rarely been included in a common analysis. The analyses of Davis (1995)
, Katayama and Ogihara (1996)
, and Givnish et al. (1999)
included Poaceae and Joinvilleaceae, but not Ecdeiocoleaceae. This is also true of several analyses that have focused on grasses (e.g., Hilu, Alice, and Liang, 1999
; GPWG, 2001
). The analysis of Linder (2000)
included Joinvilleaceae and Ecdeiocoleaceae, but not Poaceae, while that of Briggs et al. (2000)
included Ecdeiocoleaceae and Poaceae, but not Joinvilleaceae. The latter study is notable, however, in placing Ecdeiocoleaceae (represented by Ecdeiocolea and Georgeantha) as the sister of Poaceae, while Bremer's analysis (2002) supported the same relationships among these three families as are reported in the present paper.
Some previous analyses, using morphological characters (and in some cases morphology plus DNA sequence data), have included all three of these families, but none has resolved a clade consisting of Joinvilleaceae, Ecdeiocoleaceae, and Poaceae. Linder and Rudall (1993)
resolved a clade consisting of Joinvilleaceae and Poaceae, but their analysis placed Ecdeiocoleaceae in another clade that included Restionaceae and Centrolepidaceae, and sometimes Anarthriaceae as well. Similarly, Chase et al. (1995)
, Kellogg and Linder (1995)
, Linder and Kellogg (1995)
, Stevenson and Loconte (1995)
, and Rudall, Stevenson, and Linder (1999)
obtained a variety of results, but never a clade comprising Poaceae, Joinvilleaceae, and Ecdeiocoleaceae. However, none of these analyses included DNA sequence data for Poaceae, Joinvilleaceae, and Ecdeiocoleaceae.
One limitation in many of these studies has been the lack of DNA sequence data for Ecdeiocoleaceae; hence, even in combined analyses (e.g., Chase et al., 1995
), the suite of morphological characters used in these analyses may have dominated in the placement of Ecdeiocoleaceae with Restionaceae. Another factor, and perhaps a more pernicious one, may have been the identities of the terminals themselves. As noted above, it is possible that Restionaceae are not monophyletic. In the present combined analysis we obtain a monophyletic Restionaceae s.s., as sister of Centrolepidaceae, but character polymorphisms observed in the past have led systematists to remove genera once assigned to Restionaceae to other families, including Anarthriaceae, Ecdeiocoleaceae, Hopkinsiaceae, and Lyginiaceae (Cutler and Airy Shaw, 1965
; Briggs and Johnson, 2000
), and as noted above, Restionaceae, as currently circumscribed, may be polymorphic for the 28-kb inversion. Moreover, it is not clear whether Centrolepidaceae are a close relative of Restionaceae s.s. (e.g., Briggs et al., 2000
) or nested among its members (Hamann, 1975
; Linder and Rudall, 1993
; Kellogg and Linder, 1995
; Linder, Briggs, and Johnson, 2000
). The use of "composite taxa" (presumptive monophyletic groups that are treated as solitary terminals in cladistic analyses and coded either as monomorphic or polymorphic for each character, the latter reflecting variation among constituent taxa) can cause severe distortions in phylogenetic analyses, even when the composite taxa actually are monophyletic (Nixon and Davis, 1991
; Prendini, 2001
; Simmons, 2001
). When a composite taxon is not monophyletic, there is an additional source of distortion, for scores present in the one line of data that represents this terminal (i.e., characters observed in particular subsets of the group) actually belong in different positions in the phylogeny. Hence, composite taxa should be used only as an expedient when alternative approaches are not feasible, and the alternative, exemplar sampling, should be employed whenever possible.
Within the Joinvillea-Ecdeiocolea-Poaceae clade the unconstrained analysis of the combined data set, like that of Bremer (2002), resolves Ecdeiocolea as sister of Poaceae, but there is some evidence within the data set for alternative phylogenetic structures. Jackknife frequencies, consistent with the consensus tree, indicate a higher degree of overall support for a sister group relationship between Ecdeiocolea and Poaceae (49%) than for either of the other two possible dichotomous resolutions (both alternatives less than 20%). In analyses in which the alternatives are forced to be resolved, no structural character is identified as an unambiguous synapomorphy of Ecdeiocolea and Joinvillea, while one is identified (gain of pubescence) for a constrained grouping of Joinvillea with Poaceae. In contrast, the unconstrained analysis identifies two synapomorphies for the grouping of Ecdeiocolea with Poaceae, both of them representing features of the pollen. One of these, the loss of scrobiculi (character 95), is a reversal of the gain of scrobiculi, which is a synapomorphy of Poales s.s. This character, if examined in isolation, could be interpreted differently, with gain of scrobiculi seen as a synapomorphy of all members of Poales s.s. except Ecdeiocoleaceae, Joinvilleaceae, and Poaceae. In addition to these unambiguous synapomorphies, one character of note is the occurrence of a nutlet in Ecdeiocolea. This is not a universal feature of Ecdeiocoleaceae, for the fruit of Georgeantha, the other genus of Ecdeiocoleaceae, is a capsule (Briggs and Johnson, 1998
). Although Georgeantha is not included in the present analysis, the character still cannot be optimized unambiguously in this region of the two most-parsimonious cladograms, because of the pattern of occurrence of various fruit types in nearby taxa. However, if the caryopsis of grasses is interpreted as a particular kind of nutlet, and scored as such (as it is in this analysis), one possible optimization of this character is as a synapomorphy of Ecdeiocolea and Poaceae. If the shared fruit indehiscence of these two taxa is homologous, Ecdeiocolea may be more closely related to Poaceae than either of them is to Georgeantha, in which case Ecdeiocoleaceae would not be monophyletic. However, this character should be interpreted with caution because it has been proven to be extremely variable within closely related taxa and might have evolved multiple times within the Restionaceae alone (Linder, 1992a
, b
).
Poales s.s., a set of seven families delimited as an order by Dahlgren, Clifford, and Yeo (1985)
, has been reconfirmed as a natural group in several subsequent studies, including the present one. Within this assemblage, the placement of Ecdeiocoleaceae in a clade that otherwise includes only Joinvilleaceae and Poaceae, consistent with our documentation of the presence in Ecdeiocoleaceae of the 6-kb inversion that was previously detected only in Joinvilleaceae and Poaceae, broadens the field of taxa that are implicated as being the closest relatives of the grass family. Our results also suggest that Restionaceae, and possibly Ecdeiocoleaceae, may not be monophyletic. It may also be the case that the three inversions in the plastid genomes of grasses have a more complicated history than previously has been believed. These results, in the context of an increasingly complicated picture of relationships among Restionaceae and their closest relatives, confirm the unity of Poales s.s., but suggest the need for additional study of relationships among the families of this group, with one possible result being the need to recircumscribe one or more of them.
|
|
FOOTNOTES |
|---|
), relative positions of the primers used in this study, and inferred phylogenetic relationships based on presence/absence of the three chloroplast inversions. For each of the four maps, genes above the line are transcribed from left to right and genes below the line are transcribed from right to left. Vertical lines between pairs of maps demarcate end points of the three inversions. Arrows with boldface numbers represent primer combinations used to identify presence or absence of each inversion (cf. Table 1 in Doyle et al., 1992
50%. Jackknife support for the Ecdeiocolea-Poaceae clade is also shown