| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics |
2Department of Biology, Coker Hall, University of North Carolina-Chapel Hill, North Carolina 27599 USA; 3Herbarium, Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3AB, UK; 4Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3DS, UK
Received for publication August 16, 2004. Accepted for publication May 6, 2005.
ABSTRACT
Parsimony and Bayesian analyses of plastid rbcL and trnL-F DNA sequence data of the pantropical family Euphorbiaceae sensu stricto (s.s.) are presented. Sampling includes representatives of all three subfamilies (Acalyphoideae, Crotonoideae, and Euphorbioideae), 35 of 37 tribes and 179 of the 247 genera of uniovulate Euphorbiaceae sensu lato (s.l.). Euphorbiaceae s.s. were recovered as a monophyletic group with no new adjustments in circumscription. Two clades containing taxa previously placed in Acalyphoideae are found to be successive sisters to all other Euphorbiaceae s.s. and are proposed here at subfamilial rank as Peroideae and Cheilosoideae. The remainder of the family fall into seven major lineages including Erismantheae and Acalyphoideae s.s. (parts of Acalyphoideae), Adenoclineae s.l., Gelonieae, articulated crotonoids and inaperturate crotonoids (parts of Crotonoideae), and Euphorbioideae. Potential synapomorphies and biogeographical trends are suggested for these clades. Acalyphoideae s.s., inaperturate crotonoids, and Euphorbioideae tribe Hippomaneae each have two major subclades that represent novel groupings without apparent morphological synapomorphies. Two subfamilies, 14 tribes, and 10 genera were found to be para- or polyphyletic. Noteworthy among these, Omphaleae are embedded in Adenoclineae, Hureae + Pachystromateae in Hippomaneae, Ditta in Tetrorchidium, and Sapium s.s. in Stillingia.
Key Words: Cheilosoideae • Euphorbiaceae • Malpighiales • molecular phylogenetics • morphology • Peroideae
The order Malpighiales sensu the Angiosperm Phylogeny Group (APG II, 2003
, but excluding Peridiscaceae following Davis and Chase [2004]
and including Rafflesiaceae sensu stricto [s.s.] [see Davis and Wurdack, 2004
]) contains 29 (to 37 depending on alternative sensu stricto circumscriptions) families of eudicots. Five of these families, Euphorbiaceae s.s., Pandaceae, Phyllanthaceae, Picrodendraceae, and Putranjivaceae (see Wurdack et al., 2004
, for family comparisons), have often been united as Euphorbiaceae sensu lato (s.l.). The circumscription of Euphorbiaceae s.l. has long been controversial (reviewed by Webster, 1987
), and the current APG II splitting of the group reflects long-known fundamental divisions that were in part used for previous subfamilial circumscriptions (i.e., Webster, 1975
, 1994b
; Radcliffe-Smith, 2001
). The origins of the APG II segregates include, with some adjustments, Euphorbiaceae s.s. derived from Acalyphoideae pro parte (excluding Dicoelieae and Galearieae), Crotonoideae, and Euphorbioideae; Pandaceae from Acalyphoideae–Galearieae, Phyllanthaceae from Phyllanthoideae pro parte (excluding Centroplaceae and Drypeteae, and including Croizatieae and Dicoelieae), Putranjivaceae from Phyllanthoideae–Drypeteae (excluding Lingelsheimia), and Picrodendraceae from Oldfieldioideae (excluding Croizatieae). The modern sensu stricto family delimitation was proposed by Corner (1976)
and Meeuse (1990)
to include members of Euphorbiaceae s.l. that possess a single ovule per locule (instead of two), excluding Pandaceae and Scagea (secondarily uniovulate Picrodendraceae). The remaining segregates (Phyllanthaceae, Picrodendraceae, and Putranjivaceae) possess two ovules per locule, although in many taxa only one of the paired ovules develops into a seed. Gynoecial evidence links Euphorbiaceae s.s. with Picrodendraceae and Phyllanthaceae, notably a shared and putatively unique combination of nucellar beak, epitropus ovules, and obturator (Sutter and Endress, 1995
). Putranjivaceae appear further removed based on embryological differences (Tokuoka and Tobe, 1999
).
Single-exemplar sampling of Euphorbiaceae s.s. (Euphorbia) employed in major angiosperm-wide molecular phylogenetic studies (Chase et al., 1993
; Soltis et al., 1997
, 2000
, 2003
; Savolainen et al., 2000a
; Hilu et al., 2003
) has failed to recover supported sister relationships for the family, and polyphyly of Euphorbiaceae s.l. has been indicated. Limited additional sampling using rbcL (Wurdack and Chase, 1996
, 1999
; Savolainen et al., 2000b
; Chase et al., 2002
; Wurdack, 2002
) and/or other genes (Spichiger et al., 1993
; Downie et al., 1996
; Savolainen et al., 1997
; Cho et al., 1998
; Wurdack, 2002
; Davis and Chase, 2004
; Davis and Wurdack, 2004
; Davis et al., 2005
) has revealed little about sister groups or core relationships in the family.
The focus of this paper is Euphorbiaceae s.s., which comprise c. 6300 species (estimated from Govaerts et al., 2000
) in 245 genera (Radcliffe-Smith, 2001
) or alternatively 219 genera in the previous classification adopting a broader generic concept (Webster, 1994b
). Two additional genera, Aubletiana and Colobocarpos, have been recently recognized for aberrant species of Conceveiba and Croton, respectively. Euphorbiaceae s.s. is the largest component of Euphorbiaceae s.l. and remains as one of the largest plant families even in this more restricted circumscription. Of all Malpighiales families, Euphorbiaceae s.s. are unsurpassed in species richness, morphological and phytochemical diversity, and economic importance. The notable economic products include cassava (Manihot esculenta), rubber (Hevea brasiliensis), castor (Ricinus communis) and tung oils (Vernicia spp.), candilla wax (Euphorbia spp.), and ornamental poinsettias (Euphorbia pulcherrima). Molecular-clock dating (Wikström et al., 2001
; Davis et al., 2005
) gives a Cretaceous estimate for divergence of the family. The fossil record is of poor quality with few reliable Euphorbiaceae s.s. macrofossils (see Webster, 1967
; Crepet and Daghlian, 1982
), although it does indicate considerable diversification by the Eocene (Muller, 1981
; Crepet and Daghlian, 1982
; Friis and Crepet, 1987
; Dilcher and Manchester, 1988
). Systematics of the family stands on the work of Webster (1975
, 1994b)
and is presently inferred to be largely coincident with that of the uniovulate subfamilies Acalyphoideae, Crotonoideae, and Euphorbioideae. Webster's 1994 system also placed Pandaceae as tribe Galearieae of Acalyphoideae, although previously he had maintained it as a family-level segregate within Euphorbiales (1975; see Webster [1987]
for a discussion of this change of opinion). Dicoelia had been referred to a monotypic tribe (Dicoelieae) either in Pandaceae (Webster, 1987
) or adjacent to Galearieae (Webster, 1994b
). The presence of two ovules per locule has long indicated misplacement of Dicoelia, and a position in Phyllanthaceae has recently been confirmed (Kathriarachchi et al., 2004
, 2005
). Relative to the classification of Webster (1994b)
, that of Radcliffe-Smith (2001)
is nearly identical in the composition of Acalyphoideae (20 tribes, 119 genera), Crotonoideae (12 tribes, 74 genera), and Euphorbioideae (5 tribes, 54 genera). His system recognizes a number of generic segregates that had been subsumed by Webster, contains a reevaluation and further generic splitting of Euphorbioideae–Hippomaneae, and changes the classification of seven other genera, including the creation of a new tribe (Acalyphoideae–Sphyranthereae) to accommodate Sphyranthera.
Systematic palynological studies have done much to guide or illuminate current classifications (Punt, 1962
; Nowicke, 1994
; Takahashi et al., 1995
, 2000
; Fernández-González and Lobreau-Callen, 1996
; Lobreau-Callen and Suárez-Cervera, 1997
; Nowicke et al., 1998
, 1999
, 2002
). The understanding of evolutionary trends in Euphorbiaceae s.l. is clouded by the belief that the uniovulate subfamilies are derived from biovulate lineages (i.e., from within a paraphyletic Phyllanthoideae [= Phyllanthaceae]). Similarly, many tribes and subfamilies Acalyphoideae and Crotonoideae are believed to be paraphyletic (Webster, 1987
). Evolutionary relationships among uniovulate Euphorbiaceae were first depicted in the intuitive phylograms of Pax (1924)
. A number of small morphological phylogenetic studies (Esser, 1994
; Welzen, 1995
, 1999
; Kruijt, 1996
; Esser et al., 1997
; Welzen and Stuppy, 1999
) have focused on specific tribes, but their sampling was constrained a priori to follow the circumscriptions of Webster (1994b)
and did not address the premise of monophyly. No published molecular phylogenetic study has focused solely on higher-level relationships within Euphorbiaceae s.s. Species-level work, however, is appearing with increasing frequency, including for Dalechampia (Armbruster and Baldwin, 1998
; Armbruster, 2002
), Euphorbia s.l. (Gielly and Taberlet, 1994
; Savolainen et al., 1997
; Molero et al., 2002
; Steinmann and Porter, 2002
; Ritz et al., 2003
; Haevermans et al., 2004
), Fontainea (Rossetto et al., 2001
), Hevea (Luo and Boutry, 1995
; Luo et al., 1995
), Macaranga (Blattner et al., 2001
; Davies et al., 2001
), Manihot (Bertram, 1993
; Olsen and Schaal, 1999
), and Mercurialis (Krähenbühl et al., 2002
) and is underway for Acalypha (Steinmann and Levin, 2003
) and Croton (Berry et al., 2002
). Mitochondrial gene-content surveys have shown potentially useful variation among core lineages (Adams et al., 2002
). Genomic work has been published on Ricinus (Loo et al., 1995
) and Euphorbia (Edqvist and Farbos, 2002
) and is underway for economically important Hevea and Manihot.
The phylogenetic analyses of Euphorbiaceae s.s. that we report here used DNA sequence data from the plastid regions rbcL, and the trnL-F spacer and trnL intron (hereafter referred to as "trnL-F"). The utility of rbcL for resolving intra- and interfamilial relationships is well known and has been demonstrated previously in Malpighiales (i.e., Cameron et al., 2001
; Chase et al., 2002
; Gustafsson et al., 2002
; Wurdack, 2002
; Wurdack and Chase, 2002
; Wurdack et al., 2004
). The lack of bootstrap support along the spine of these trees, even with the addition of genes of similar or slower evolutionary rates, and poor resolution among closely related taxa indicates that more rapidly evolving regions are required. Largely noncoding trnL-F is frequently used at the species level including in Euphorbiaceae s.s. (Gielly and Taberlet, 1994
; Rossetto et al., 2001
; Ritz et al., 2003
), but its broader utility has been demonstrated (e.g., Freeman and Scogin, 1999
; Richardson et al., 2000
; Davis et al., 2001
; Bremer et al., 2002
; Borsch et al., 2003
). Alignment issues with increasing sequence divergence can ultimately limit the higher-level usefulness of noncoding DNA (see Kelchner, 2000
; Simmons and Freudenstein, 2003
). Our study was undertaken to evaluate the circumscription of Euphorbiaceae s.s. and monophyly of suprageneric taxa recognized in recent classifications, and to elucidate patterns of infrafamilial relationships and character evolution. We have suggested modest classification changes but have refrained from proposing a full revision of the family classification because of limitations due to sampling and resolution.
MATERIALS AND METHODS
Taxon sampling
Taxa, voucher information, and GenBank numbers for all 213 rbcL and 221 trnL-F sequences used are listed in the Appendix. Unpublished sequences include 169 newly generated rbcL and 220 trnL-F, along with 36 published rbcL sequences that were originally generated for this study but used first elsewhere (Källersjö et al., 1998
; Savolainen et al., 2000b
; Chase et al., 2002
; Wurdack et al., 2004
; Davis et al., 2005
). Following Radcliffe-Smith (2001)
, sampling included 35 of 37 tribes and 179 of 247 genera attributed to the uniovulate subfamilies of Euphorbiaceae s.l. Generic names strictly follow Radcliffe-Smith (2001)
for ease of reference. Phylogenetically placed here but not tallied in our generic count are the unpublished new genus Brasiliocroton (Berry et al., in press
), as well as Vaupesia (data obtained during revisions and not available for inclusion in figured analyses). Not sampled from lack of adequate material were one monogeneric tribe (Sphyranthereae) and potentially significant genera (as suggested by Webster [1994b]
, Radcliffe-Smith [2001]
or other evidence) Anomalocalyx, Avellanita, Cheilosa, Chondrostylis, and Dendrocousinsia; the remaining unsampled tribe (Dicoelieae) has been excluded from Euphorbiaceae s.s. (see Introduction). Particular effort was made to sample more thoroughly the "basal" tribes of Acalyphoideae. An increased sampling of Hippomaneae was included to examine the generic delimitations and classification proposed by Esser (Esser et al., 1997
; Esser, 1999
; Radcliffe-Smith, 2001
, 352–398). Given the undisputed monophyly of Euphorbia s.l. and recent well-sampled molecular phylogenetic studies (i.e., Steinmann and Porter, 2002
), sampling in the cyathial clade was intentionally low relative to its species richness.
The choice of outgroup is somewhat arbitrary due to the lack of bootstrap-supported sister relationships to Euphorbiaceae s.s. in ordinal molecular phylogenetic analyses (see Chase et al., 2002
; Wurdack, 2002
; Davis and Chase, 2004
; Davis et al., 2005
). Humiriaceae were chosen based upon their low level of molecular divergence and concomitant ability to align with ingroup trnL-F sequences (see Davis et al., 2001
; Wurdack, 2002
). Pandaceae were included to demonstrate their monophyly and misplacement in the classifications of Webster (1994b)
and Radcliffe-Smith (2001)
, although interfamilial sequence divergence in trnL-F makes homology assessments more difficult. Additional available Malpighiales trnL-F sequences (mostly K. J. Wurdack, unpublished data) include those for Achariaceae, Malpighiaceae (from Davis et al., 2001
), and biovulate lineages (Phyllanthaceae, Picrodendraceae, and Putranjivaceae), but these proved unsatisfactory due to the introduction of additional ambiguously aligned regions.
Laboratory methods
DNA extraction, polymerase chain reaction (PCR) amplification, and sequencing methods followed those previously outlined (Wurdack, 2002
; Wurdack et al., 2004
). In some Hippomaneae, potent PCR inhibitors co-purified with Qiagen DNeasy Plant Mini kit extractions (Qiagen, Valencia, California, USA). Inhibition could usually be overcome by template dilution. Ex Taq DNA Polymerase (hot-start version; Takara Mirus Bio, Madison, Wisconsin, USA) was found to be especially sensitive for degraded or weak extractions. Manual radioactive methods were used for 19% of our rbcL sequences, and the rest were generated by automated fluorescent sequencing. The trnL-F sequences were generated entirely by automated sequencing using four primers (c, d, e, f; Taberlet et al., 1991
) after amplification as one piece (primers c + f) or two pieces (c + d, e + f) with some degraded templates. A newly designed trnL-intron internal primer (aligned positions 272–292, trnL-intF: 5'-GGTGCAGAGACTCAATGGAAG-3') was used as substitute for primer "c" with templates too degraded for amplification of the entire intron. Homopolymer regions affecting sequence quality were sporadic in both the trnL-F spacer and intron. Twenty-three sequences were amplified and sequenced as two pieces, in which case we treated the internal primer-binding region (primers d, e) as missing data. Data from Bernardia scabra, Conceveiba pleiostemona, Mareyopsis, Sampantaea, and Vaupesia could only be obtained for the trnL-F spacer because of the degraded nature of these herbarium DNAs. Micrococca, Strophioblachia (both Chase et al., 2002
), and Euphorbia epithymoides (as E. polychroma; Chase et al., 1993
) were resequenced for rbcL from the original DNAs as the published sequences were found to contain sequencing errors or missing data. Both strands were sequenced for all taxa.
Sequence assembly and data analysis
Sequences were assembled and edited in Sequencher 3.1.1 (Gene Codes, Ann Arbor, Michigan, USA) and trimmed of primer regions. Alignments of rbcL were easily managed by eye, whereas trnL-F sequences were aligned with CLUSTAL W (as implemented in CLUSTAL X v1.83; Thompson et al., 1997
), followed by extensive hand refinements in Se-Al v2.0a11 (Rambout, 1996–2002
) that considered homology assessments from the standpoint of both similarity and possible mechanism of molecular evolution (i.e., following Kelchner, 2000
; Borsch et al., 2003
). Indels were abundant in trnL-F. Although small groups of related taxa were easily alignable, ambiguous regions of high complexity were present in the complete matrix, especially in Euphorbia s.l. Secondary structure prediction in trnL-F used mfold v3.1 on a web server (http://www.bioinfo.rpi.edu/ applications/mfold; Zuker, 2003
) with default parameters to evaluate correspondence between structural motifs and some sequence changes (i.e., compensatory changes, conserved P-elements, and involvement of hairpins in inversions). Mononucleotide repeats were few (six "hotspots"), and much of the length variation involved short tandem repeats of 3–15 nucleotides. A similarity criterion was used to associate the least divergent copies of such repeats. Often the copies were identical, in which case incorrect homology assessment would have no phylogenetic impact under the data exclusion sets used. Patterns were evident of indel-poor regions interspersed with variable regions often containing multiple, independent (non-homologous) insertions. The data matrices are archived in TreeBASE (www.treebase.org/treebase/) and are available from the authors.
Maximum parsimony (MP) analyses were conducted using PAUP* 4.0b10 (Swofford, 2003
) and a two-stage search strategy with starting trees generated by 1000 random taxon addition sequence replicates (RAS), equal weights and unordered characters, tree-bisection-reconnection (TBR) swapping, holding 10 trees at each step, and MulTrees (holding more than one tree) off (Wurdack et al., 2004
). The resulting trees were used as starting trees with MulTrees on and a maximum tree limit of 1 x 106. We also used PAUPRat (Sikes and Lewis, 2001
); however, no shorter trees were found. The searches shown here were done on rbcL, trnL-F, and the combined data set. The combined data set included six composite terminals (i.e., each partition came from a different DNA source but representing the same species for Croton lucidus, Cr. lobatus, and Maprounea guianensis, or same genus for Bernardia, Dysopsis, and Tetraplandra). The individual data sets included 75 (34 rbcL and 41 trnL-F) terminals that were only present in one partition and not included in the combined analysis. Other analyses (not shown) included each separate partition of the combined sampling (see Table 1), combined searches with all data, and compartmentalized searches of large clades. Searches on subpartitions defined by the amplification primers used for degraded samples showed no bootstrap-supported incongruence that could otherwise indicate the assembly of chimeric sequences. The first 30 bp of rbcL were excluded to reduce missing data in the primer-binding region that was ragged due to different length primers used. Indels in trnL-F were treated as missing data, and for the final analyses described here, 1468 bp of that matrix were excluded. This reduced the missing data from 56.9% to 5.13% and eliminated most ambiguously aligned regions, one inversion hotspot (trnL-F spacer aligned positions 1776– 1813), and most insertions. Details of the excluded regions are embedded in the archived Nexus data files with accompanying exclusion sets. We acknowledge that this conservative approach might have caused the loss of some phylogenetic information (but not necessarily; Chase et al., 2000
) but believe this is better recovered (i.e., indel coding and fewer ambiguous regions to exclude) with analyses of subsets of the data from individual clades. Such compartmentalized analyses (noted earlier) were found to have little effect on local topologies despite being analytically superior in terms of reduced ambiguous regions and more thorough search implementation. Zero-length branches were collapsed and uninformative characters were included in analyses except, as noted, for the calculation of alternative tree statistics. Tree statistics included the consistency index (CI; Kluge and Farris, 1969
), retention index (RI; Farris, 1989
), and rescaled consistency index (RC; Farris, 1989
). Relative clade support was evaluated with 1000 bootstrap (Felsenstein, 1985
) replicates (split into 125 or 200 replicate blocks and treefiles subsequently combined using stored tree weights), each with 20 or 100 RAS using TBR swapping, holding 10 trees at each step, and saving no more than 10 trees (nchuck = 10, chuckscore = 1) per iteration. The number of RAS took into account the number of times the shortest tree was found during the first stage of the heuristic searches (see Kauff and Lutzoni, 2002
). Preliminary tests with 5–200 RAS also empirically verified the threshold needed to hit the same minimal length at least twice per bootstrap replicate. For rbcL and combined data sets, 100 RAS were used (shortest tree in heuristic search found an average of 1 per 100 RAS) and 20 RAS for trnL-F (shortest tree found an average of 4 per 10 RAS). Bootstrap percentages (BP) are described as high (85–100%), moderate (75–84%), or low (50–74%).
|
), and data were partitioned by gene in the combined analysis. The most appropriate model for each paritition was determined with Modeltest v3.06 (Posada and Crandall, 1998
) to be the general time reversible model (GTR + I + 
). Four Markov chain Monte Carlo (MCMC) chains (one cold and three heated with default heating parameter of 0.2, and each initiated with a random tree) were run simultaneously for 1 x 107 generations and sampled every 100 generations. We used default priors and a Dirichlet distribution for the base frequency parameters. Results (i.e., Bayesian trace files with sampled parameter values) were examined with Tracer v1.0.1 (Rambaut and Drummond, 2002–2003
) to determine when stationarity had been reached. The stationary phases of two independent replicate runs were pooled after discarding trees for each replicate from the burn-in period that included the first 1.0 x 106 generations (10 000 trees) for trnL-F and combined data sets and 1.5 x 106 generations (15 000 trees) for rbcL. Posterior probabilities (PP) 
0.95 were considered significant (Kauff and Lutzoni, 2002
). RESULTS
Data set characteristics and search statistics are presented in Table 1. The parsimony and Bayesian analyses exhibited a general correspondence and did not have supported incongruence; bootstrap-supported nodes mostly also had high PP. Bayesian results were better resolved than MP strict consensus trees when including nodes with PP > 0.50, but this difference and most of the slight topological differences disappeared when considering nodes with PP 0.95.
Euphorbiaceae s.s. and Pandaceae were recovered as strongly supported monophyletic groups. Of the three uniovulate subfamilies that comprise Euphorbiaceae s.s. in current classifications, only Euphorbioideae were here found to be monophyletic. Acalyphoideae are clearly paraphyletic by containing the acalyphoid grade separate from the core remainder of Acalyphoideae and by the exclusion of Omphalea, which was found embedded in a crotonoid lineage. Crotonoideae do not form a supported monophyletic group but relationships among the four main lineages are unclear. In total, nine major lineages (Acalyphoideae s.s., Adenoclineae s.l., articulated and inaperturate crotonoids, Cheilosoideae, Erismantheae, Euphorbioideae, Gelonieae, and Peroideae) of Euphorbiaceae s.s. were identified, although some were not recovered intact in the rbcL analysis.
Analyses of rbcL
The termination codon beginning at nucleotide 1426 was, as in most Malpighiales, predominant in Euphorbiaceae s.s., but RuBisCo length variants beyond the typical 475 amino acids occurred (fully sequenced until stop codon or inferred from lack of one at position 1426 where some sequences were truncated) in 17 taxa and were inferred to be from eight separate gains. The MP strict consensus is shown in Fig. 1. Euphorbiaceae s.s. are strongly supported, but the spine of the tree is poorly resolved; seven of the nine major lineages are recovered intact. Peroideae are not recovered as a monophyletic group, and Pogonophora is sister to the rest of Euphorbiaceae s.s., although this lacks support (BP < 50; PP 0.68). Euphorbioideae are not monophyletic due to the pulling away of Stomatocalyceae which group as the unsupported sister to the inaperturate crotonoids. The core acalyphoids and inaperturate crotonoid subclade C2 are also poorly resolved. Among all nodes with PP 0.50–1.0, 65 (36%) had PP 0.50–0.94 and fall below the significance criteria noted earlier (Materials and Methods, Sequence Assembly and Data Analysis).
|
). The MP strict consensus is shown in Fig. 2. Euphorbiaceae s.s. are strongly supported, and the spine of the tree is better resolved than with rbcL; all nine major lineages are recovered intact, but the core acalyphoids and inaperturate crotonoid subclade C2 are poorly resolved. There were 41 nodes (24% of nodes with PP 0.50) with PP 0.50–0.94. One notable topological difference was the placement of Manniophyton + Pausandra as sister to the rest of poorly resolved subclade C2; Trigonostemon occupies this position in the rbcL and combined analyses.
|
|
0.50) with PP 0.50–0.94, including two topological differences between the Bayesian and MP trees among the major lineages (i.e., Bayesian: Euphorbioideae + core acalyphoids; Erismantheae sister to articulated crotonoids + inaperturate crotonoids). The alternative placements (i.e., Erismantheae + Euphorbioideae) of these groups shown in the MP tree lack bootstrap support (BP < 50).
|
DISCUSSION
The phylogenetic analyses presented here show broad correspondence to recently proposed classifications that have not incorporated any information from DNA, the notable exceptions involving historically contentious or problematic groups. Until now, delimitation of Euphorbiaceae s.s. has largely relied on inferences from previous sensu lato family classifications. A robust circumscription of Euphorbiaceae s.s. is here established. Pandaceae are excluded here and are far removed from the position among the Acalyphoideae suggested by Webster (1994b)
. A sister relationship to Euphorbiaceae s.s. should not be inferred by close proximity in our trees because that position merely reflects our limited sampling of families in Malpighiales. Supported sister groups to both Pandaceae and Euphorbiaceae s.s. remain to be established, although larger ordinal phylogenetic studies beyond the present four-gene study (Davis et al., 2005
) are underway that may provide support for critical deep nodes (Wurdack and Davis, unpublished data). In our trees, no additional lineages were excluded from Euphorbiaceae s.s.; morphologically aberrant Dicoelia is now known to be a member of Phyllanthaceae (Kathriarachchi et al., 2004
, 2005
). Centroplacus, of uncertain familial affiliation (APG II, 2003
) but usually associated with Phyllanthaceae or Pandaceae, has also been excluded (not sampled here; see Wurdack et al., 2004
).
The spines of the trees in all analyses lack supported relationships (BP < 50) among major clades that relate to subfamilial delimitation sensu (Webster, 1994b
), especially among lineages of Crotonoideae. Crotonoid pollen and phorbol esters (co-carcinogenic diterpenes) present in part of the family have been used to link Euphorbiaceae with Thymelaeaceae (Schmidt, 1986
; Webster, 1987
; Nowicke, 1994
; Seigler, 1994
). These characters are homoplasious between the families (but informative within them), despite their remarkable co-distribution and absence elsewhere in angiosperms; Thymelaeaceae are firmly established in Malvales (Soltis et al., 2000
; APG II, 2003
). In Euphorbiaceae s.s., crotonoid pollen is confined to Crotonoideae and phorbol esters nearly exclusively (but see subclade A3) to Crotonoideae + Euphorbioideae, indicating that both are derived within the family rather than plesiomorphic.
The distributions of latex and associated laticifers have been used in the broad classification of Euphorbiaceae s.l., notably their presence in Crotonoideae s.l. and Euphorbioideae and absence in other lineages (i.e., Acalyphoideae, Phyllanthaceae, and Picrodendraceae). Great differences in latex properties, such as color (e.g., milky, yellowish, orange, red, or clear), chemical composition, and subcellular bodies (i.e., starch grains, rubber and Frey-Wyssling particles, and lutoids), indicate a potential source of characters that have been little studied on a broad scale. Milky latex (rarely yellowish) characterizes Euphorbioideae and articulated crotonoids; red latex is frequent in inaperturate crotonoids and Adenoclineae s.l. Recently documented observations show isolated occurrences of laticifers in Acalyphoideae (i.e., Dalechampia; Hayden and Hayden, 2000
) along with latex-like exudate in Macaranga. Laticifers have also been rarely observed in Phyllanthaceae (Phyllanthus, Balaji et al., 1996
; Dicoelia, Hayden and Hayden, 2000
). Older reports (reviewed by Rudall, 1987
) are difficult to evaluate and have often been dismissed due to lack of documentation and corroboration. They may reflect differences in interpretation (i.e., they actually are sclereids) and difficulty of observation when laticifers are few. Rudall (1994)
has suggested putative homology between laticifers and foliar sclereids. Foliar sclereids have been recently reported from Chaetocarpus (Peroideae) and the core acalyphoid Mallotus (Roth, 1984
; Rudall, 1994
), as well as in some inaperturate crotonoids (Rudall, 1994
). A preliminary survey (K. J. Wurdack, unpublished data) of c. 600 Euphorbiaceae s.l. cleared leaves deposited at the Smithsonian Institution (J. Wolfe collection, Division of Paleobotany) revealed a far more complex picture with wider sclereid distribution, intrageneric variation (i.e., present in some but not all species of Cleidion, Pera, and Trigonostemon), and intermediate states (perhaps weakly sclerified). Well-developed elongate sclereids were observed in acalyphoids (Adenophaedra, Agrostistachys, Angostylis, Argomuellera, Blumeodendron, Cleidion, Crotonogynopsis, Erismantheae, Haematostemon, Koilodepas, Pera, and Pseudagrostistachys) and crotonoids (Dodecastigma, Pausandra, and Trigonostemon). In addition to our poor knowledge of laticifer distribution (i.e., Gelonieae apparently remain to be investigated), the lack of support among major clades hinders uncovering the pattern(s) of laticifer evolution in the family. Likewise, we have found no support for the widely held hypothesis that articulated laticifers are derived from nonarticulated ones. The former are multicellular and originate in primary and secondary tissues, whereas nonarticulated types are unicellular (coenocytic) and nearly exclusively originate in primary tissues (Rudall, 1994
). Articulated laticifers occur in two clades (Jatropha and articulated crotonoids) and have been suggested to have two independent origins (Dehgan and Craig, 1978
). Neither occurence is in a supported position within the nonarticulated clades. The presence of articulated laticifers in "basal" lineages of Malpighiaceae has been used to suggest their presence in the common ancestor of Euphorbiaceae s.s. and Malpighiaceae (Vega et al., 2002
), although there is no support for a close relationship between these families. Elatinaceae, the apparent sister to Malpighiaceae, lacks laticifers, although some members possess a brownish resin (see Davis and Chase, 2004
).
Indument type and reproductive features including the presence of petals in pistillate and/or staminate flowers (mapped in Figs. 3, 4), pollen type (e.g., inaperturate crotonoid exines), and pollen nuclear number may be useful at higher levels. Trichome morphology varies greatly across the family and even within a genus (e.g., Croton s.l.; Webster et al., 1996
), but there appears to be broad utility in simple vs. stellate forms; specialized types such as malpighian and stinging hairs can be informative for some groups. Although pollen nuclear data are sparse, we estimate from comparisons of the data tabulated by Webster and Rupert (1973)
with our trees and for Euphorbioideae with those of Steinmann and Porter (2002)
that the trinucleate condition has been gained at least eight times. Cytological data show great variation but are sparse except for the larger genera (Hans, 1973
; Urbatsch et al., 1975
; Index to Plant Chromosome Numbers, Missouri Botanical Garden, St. Louis, Missouri, USA. Available at http://mobot.mobot.org/W3T/Search/ipcn.html). The articulated crotonoids appear to have chromosome numbers based on x = 9, whereas those of the core acalyphoids, Gelonieae, and most inaperturate crotonoids and Euphorbioideae are based on x = 11. Most Euphorbia s.l. are derived from dysploid series of x = 6–10. The alchorneoids and inaperturate crotonoid subclade C1 have lineages (Discoglypremna and Jatropha, respectively) based on x = 11.
A study of comparative seed and ovule anatomy has recently been completed across Euphorbiaceae s.s. (Tokuoka and Tobe, 1998
, 2002
, 2003
). The importance of these characters will depend on their evolutionary stability both within the family and across other lineages of Malpighiales as indicated in phylogenetic analyses such as ours; they need to be studied in Malpighiales generally. Vascular bundles in the outer integument (mapped in Figs. 3, 4) are clearly homoplasious within Euphorbiaceae s.s. and also when compared to other families (i.e., Phyllanthaceae; Tokuoka and Tobe, 2001
; Wurdack et al., 2004
). The presence of vascular bundles in the inner integument appears more conservative (absent in Phyllanthaceae; Tokuoka and Tobe, 2001
) and characteristic (a possible synapomorphy) of most crotonoids (mapped in Figs. 3, 4). It is unclear how much of their distribution reflects plesiomorphic absence, secondary losses, or possible multiple gains, especially when considering their isolated occurence in Anthosteminae (Euphorbioideae) and Klaineanthus (Adenoclineae). As in Phyllanthaceae and Picrodendraceae, typical explosive euphorbiaceous schizocarps are also predominant in Euphorbiaceae s.s., although most taxa in the acalyphoid grade have atypical forms. Indehiscent fruits are clearly homoplasious and presumably associated with shifts to zoochory (e.g., Crotonopsis, Elaeophorbia, Glycydendron, Mareyopsis, Ricinodendron, and Trewia; see Esser, 2003b
). Seed arils are of probable phylogenetic utility, although their homologies are poorly known, and they are notably scarce or absent in Acalyphoideae, Adenoclineae s.l., and Gelonieae (see Tokuoka and Tobe, 1998
, 2002
, 2003
, but underestimated compared with Radcliffe-Smith, 2001
). Most of these arils are confined to the micropylar region (caruncles); they have been usually considered to be involved in dispersal (e.g., Berg, 1975
; Narbona et al., 2005
), but they also effect seed physiology (Webster, 1994a
; Bianchini and Pacini, 1996
). Caruncles vary greatly in morphology and distribution across the family, even within some genera (e.g., Croton, Euphorbia s.l., and Stillingia) or individual plants (i.e., seeds from allomorphic vs. regular flowers in Acalypha). Peroideae are carunculate, although Tokuoka and Tobe (2003)
described Pogonophora as exarillate. Our observations of herbarium specimens indicate that Pogonophora has a fleshy and well-developed caruncle on mature, black seeds, in agreement with other reports (Secco, 1990
; Radcliffe-Smith, 2001
). Caruncles are widespread in Picrodendraceae but notably lacking in Podocalyx, which is sister to the rest of that family (K. J. Wurdack, unpublished data) and absent from Phyllanthacaeae except for Celianella (Stuppy, 1996
).
The acalyphoid grade
A series of mono- or digeneric tribes that were all previously assigned to Acalyphoideae fall into two clades that are subsequent sisters to the rest of Euphorbiaceae s.s.; the first clade containing Chaetocarpeae, Clutieae, Pereae, and Pogonophoreae (Peroideae; see conclusions) and the second with Cheiloseae (Cheilosoideae, see conclusion). These tribes were already considered primitive and placed first in Webster's (1994b)
phylogenetic classification that started with Clutieae and ended in Euphorbieae.
Peroideae is sister to all other Euphorbiaceae s.s. The fruit morphology (membranous, fragile septa without visible vascularization) of these taxa sets them apart from all other Euphorbiaceae s.l. (Esser, 2003a
). They also share distinctive smooth, shiny, black, carunculate seeds. Chaetocarpus, Clutia, and Pera, each from separate tribes, form a strongly supported subclade within Peroideae and have a unique exotestal seed coat (Tokuoka and Tobe, 2003
). The seed characters appear to be apomorphic in the family, and their strongly supported sister Pogonophora and the rest of Euphorbiaceae s.s. have exotegmic, palisadal seed coats (Tokuoka and Tobe, 2003
; Pogonophora reconfirmed here, K. J. Wurdack, personal observation). Pera has been considered isolated and classified in its own tribe or even family (see Radcliffe-Smith, 1987
) because of its unique pseudanthial inflorescences. A survey of Pera pollen noted high diversity, including an intectate form, despite the general morphological similarity of most species (Nowicke et al., 1998
). Specialized pollinator relationships are expected to be involved with floral novelties, but in the case of Pera they do not seem to be unusual; recorded visits are from small pollen-collecting bees (Gillespie and Armbruster, 1997
). The strongly supported relationship of Pera with Acalypha (Figs. 2, 4 in Davis and Chase, 2004
) is an artifact of sparse sampling relative to exceptionally high sequence divergence in PHYC (PHYC independently resequenced and verified, K. J. Wurdack, unpublished data). The putative association of Pera with Alchorneeae based on wood anatomy (Hayden and Hayden, 2000
) is not supported here. Pera is Neotropical, Clutia African (with greatest diversity in South Africa), and Pogonophora is vicariant (one Neotropical species and one West African). Chaetocarpus has a disjunct tropical distribution, including species in the Neotropics, Africa, and Asia.
The next node in the acalyphoid grade isolates Neoscortechinia (Cheilosoideae) from the rest, although this node is poorly supported (BP 72). Cheilosoideae comprise two small Southeast Asian genera that are palynologically unique in Euphorbiaceae s.s. due to their echinate pollen exines (Takahashi et al., 1995
; Nowicke et al., 1998
). Homoplasious echinate exines are present in other euphorbiaceous lineages, including Amanoa and Croizatia of Phyllanthaceae, and most Picrodendraceae. Ovule and seed characters also support the distinctiveness of Cheilosoideae (Tokuoka and Tobe, 2003
).
Core Euphorbiaceae s.s
The rest of the family forms a monophyletic group containing seven major lineages: (1) Erismantheae, (2) Acalyphoideae s.s., (3) Adenoclineae s.l. (including Omphalea but excluding Glycydendron), (4) Gelonieae, (5) articulated crotonoids, (6) inaperturate crotonoids, and (7) Euphorbioideae (including Stomatocalyceae).
Erismantheae
The next tribe after Cheiloseae in Webster's linear sequence is Erismantheae. This small Southeast Asian group of three genera, totaling only five species, is united by monoecy, opposite leaves, and interpetiolar stipules. The unusual stipular arrangement has been suggested to be the result of unequal internode growth (i.e., seemingly opposite leaves due to contracted internodes) and loss of one stipule per leaf (Welzen, 1995
; Radcliffe-Smith, 2001
). Opposite leaves have evolved independently multiple times across the family (e.g., Baloghia inophylla, Calycopeplus spp., Chamaesyce spp., Mallotus section Hancea, Mercurialis spp., Pera oppositifolia, Sebastiania hexaptera, and Stillingia oppositifolia), but interpetiolar stipules occur only in Erismantheae and, deeply embedded in the euphorbioid clade, Chamaesyce. The sampled Erismantheae, Moultonianthus (rbcL only) and Syndyophyllum form a monophyletic group in the rbcL analyses. The last, sampled for the combined analyses, is resolved among the lactiferous lineages, but this position lacks support (BP < 50). Laticifers have not been reported in the tribe (Hayden and Hayden, 2000
), but the leaves are sclerified (Herbert, 1897
; K. J. Wurdack, personal observation).
Acalyphoideae s.s
Acalyphoideae s.s. form a strongly supported clade uniting the remaining 11 tribes (excluding, as noted, Dicoelieae and Galearieae) of Acalyphoideae sensu Radcliffe-Smith (2001)
. The clade contains two strongly supported groupings with the alchorneoids sister to the "core acalyphoid" remainder. The core acalyphoids are poorly supported at the deepest nodes, but do contain some supported internal subclades (designated for discussion subclades A1–A7) that, in part, follow previous tribal or subtribal classifications.
Alchorneoids
One of the two constituent main groupings of Acalyphoideae s.s. is a novel association of Alchorneeae + Agrostistachydeae and several genera from three other tribes including Caryodendreae, which Webster (1994b)
suggested had links to Agrostistachydeae. Pseudagrostistachys is sister to the rest of this subclade in the combined analyses (but not rbcL), making Agrostistachydeae paraphyletic by not grouping with the other sampled tribal member, Cyttaranthus. Paranecepsia (rbcL only), a member of Bernardieae (Webster, 1994b
; Radcliffe-Smith, 2001
), is strongly supported as sister to Pseudagrostistachys in the rbcL analysis and this pair is in turn the unsupported (BP < 50) sister to the core acalyphoids. Alternative rbcL analyses (not shown) without Paranecepsia recover the placement of Pseudagrostistachys sister to the alchorneoids (BP 68) seen in the trnL-F and combined trees. The placement of Amyrea differs greatly between Webster (1994b
; Acalypheae-Claoxylinae) and Radcliffe-Smith (2001
; Bernardieae). Radcliffe-Smith (1998)
noted its morphological similarities with Agrostistachys + Pseudagrostistachys, and this affinity is supported here. Alchorneeae s.s. (sensu Webster, 1994b
) form a strongly supported clade within the alchorneoids and have potential synapomorphies in wood anatomy (lysigenous canals; Hayden and Hayden, 2000
) and pollen (stratified opercula; Nowicke and Takahashi, 2002
). The nested nature and supported nodes of sister taxa indicate that Alchorneeae s.s. might be expanded (i.e., Alchorneeae s.l.) to include some or all members of the alchorneoid clade. Outside Alchorneeae s.s. there is no great morphological discontinuity along this grade; however, Pseudagrostistachys and its presumed sister Agrostistachys (not sampled) have petals in flowers of both sexes, and Cyttaranthus has petals only in staminate flowers. Petals are absent in the remaining alchorneoids and Paranecepsia. The position in our analyses of Mareyopsis among the alchorneoids is far removed from Mareya (see core acalyphoid subclade A3) despite recent synonymization by Webster (1994b)
. Radcliffe-Smith (2001)
reinstated the genus as distinct from Mareya but placed both in a newly recognized subtribe Mareyinae within Acalypheae. The heterogeneous pollen observed (Nowicke and Takahashi, 2002
) in Mareyopsis is based on misidentification (Zenker "35" incorrect, K. J. Wurdack, personal observation). Trigeneric Caryodendreae dissolve with the placement of Discoglypremna and Alchorneopsis as separate alchorneoid lineages and Caryodendron in subclade A7. African Discoglypremna and Neotropical Alchorneopsis are distinct as indicated by seed coat data (Tokuoka and Tobe, 2003
) and not vicariants as suggested by Webster (1994b)
. Our data indicate that Conceveiba sensu Webster (1994b)
and Radcliffe-Smith (2001)
is not monophyletic and support the separation of Aubletiana, containing two African taxa originally described as Conceveiba species, from the Neotropical remainder (Murillo-A., 2000
). The position of Gavarretia is unresolved, but its inclusion along with Polyandra (not sampled) in Conceveiba s.l. has been proposed (Murillo-A., 2000
) and is supported in a better-sampled nuclear ITS ribosomal DNA analysis (K. J. Wurdack, unpublished data). Pantropical Alchornea is paraphyletic with the inclusion of Bocquillonia, a small genus endemic to New Caledonia. The spelling "Alchorneeae" is here adopted in accordance with the nomenclatural code (Greuter et al., 2000
) Art. 19.3. and 18.1. The spelling "Alchorneae" (Webster, 1994b
) or "Alchornieae" (Radcliffe-Smith, 2004) contravenes these rules. The same applies to "Ampereeae," not Ampereae (Webster, 1994b
; Radcliffe-Smith, 2001
). Both Alchornea and Amperea are based on the surname of a person.
Core acalyphoid subclade A1
Blumeodendron (Pycnocomeae–Blumeodendrinae) is strongly supported as sister to Macaranga, Mallotus, and Trewia (Acalypheae–Rottlerinae). Webster's (1994b)
doubts about the monophyly of Pycnocomeae are confirmed by our analyses, with all sampled members of Pycnocomeae–Pycnocominae (Argomuellera and Pycnocoma) far removed to subclade A3. Blumeodendron has a disc in flowers of both sexes; this is absent in Macaranga and Trewia but variably present in Mallotus (disc glands 0–
according to Radcliffe-Smith, 2001
). Blumeodendron shares the stellate indumentum with Macaranga, Mallotus, and Trewia, whereas Pycnocominae have simple hairs. Mallotus is sister to Trewia in our analyses, and this strongly supported clade is in turn sister to Macaranga. Mallotus japonicus (section Mallotus) and M. philippensis (section Rottlera; trnL-F only) in our sampling are embedded in the genus based on morphological analyses (Slik and Welzen, 2001
; although most relationships in their tree lack strong support and Trewia was not included there), indicating that Trewia may also be embedded within Mallotus. The only character used to distinguish Mallotus and Trewia is fruit type (dehiscent and indehiscent, respectively). Transitions between these have been shown to occur frequently within Euphorbiaceae s.l. (see Wurdack et al., 2004
) and in the case of Trewia may be adaptive for megafaunal dispersal (Dinerstein and Wemmer, 1988
). The granular laminar glands characteristic of Macaranga and Mallotus pro parte are also present on young leaves of Trewia (P. Hoffmann, personal observation; not noted in Radcliffe-Smith, 2001
). Webster (1994b)
considered the eight genera of Acalypheae–Rottlerinae (only two Mallotus species and monotypic Trewia sampled here) to be closely related, but further sampling within the subtribe and aberrant sections of Mallotus (i.e., Hancea and Oliganthae) is needed to determine how to treat Mallotus. Subclade A1 is nearly exclusively Old World, centered in tropical Asia. Exceptions are in Cleidion, the poorly supported sister to subclade A1, with three of c. 25 species from the Neotropics (including C. castaneifolium sampled here). Radcliffe-Smith (2001)
also included the poorly known Chilean endemic Avellanita (not sampled) in Rottlerinae, but an association with the members of principally American subclade A5 has also been suggested (see Webster, 1994b
).
Core acalyphoid subclade A2
Lobanilia (Acalypheae–Lobaniliinae) is embedded in Acalypheae–Claoxylinae in our analyses. The small Madagascan genus, formerly a section of Claoxylon, has been segregated as Acalypheae–Lobaniliinae because of its stellate indumentum and lack of reddish pigment (Radcliffe-Smith, 1989
), but is otherwise similar to the other taxa in this morphologically homogeneous group. Sister to this clade is Mercurialis, placed far from the two other members of Acalypheae–Mercurialinae (subclade A5). Mercurialinae is heterogeneous geographically (Eurasia vs. southern Africa) and with respect to breeding system, stamen number, as well as disc and seed morphology.
Core acalyphoid subclade A3
A clade uniting Spathiostemon and (rbcL only) Homonoia (Acalypheae–Lasiococcinae) is sister to two strongly supported groupings, one containing Acalypha and two new associates, Crotonogynopsis (rbcL only) and Mareya, and the other grouping with mostly Pycnocomeae–Pycnocominae. Acalypha, the third largest genus of Euphorbiaceae s.s. with 462 species (Govaerts et al., 2000
), is distinctive by virtue of a suite of wind pollination adaptations including small porate pollen, pendulous, flexuous to vermiform anther thecae, reduced perianth, dissected styles, and lack of floral nectar. As a consequence, it was placed in a monogeneric subtribe Acalyphinae. Our results show Acalypha as the strongly supported sister of Mareya. Mareya was placed in Acalypheae–Claoxylinae by Webster (1994b)
and in Acalypheae–Mareyinae by Radcliffe-Smith (2001)
, but the first species were described by Bentham in Acalypha (Hooker, 1849
; see also Bentham, 1878
). Both genera share dissected styles, divergent anther thecae, and similar pollen exines (Nowicke and Takahashi, 2002
). Mareya does not appear to have the full suite of anemophilous adaptations as it possesses interstaminal glandular disc segments and lacks the pollen reduction (i.e., smaller size and shorter colpi) of Acalypha. Crotonogynopsis is currently classified in tribe Adelieae but was placed next to Mareya by Pax and Hoffmann (1931)
. This ditypic African genus seems to be wind-pollinated with long pendulous, sometimes cauline staminate racemes and long stamens protruding from the calyx. The basally divergent anther thecae and laciniate styles provide another link with the derived features seen in Acalypha.
The last strongly supported grouping in subclade A3 includes Pycnocoma + Argomuellera (Pycnocomeae–Pycnocominae) and Sampantaea (trnL-F spacer only) + Wetria. The latter pair was considered closely related and perhaps congeneric (Webster, 1994b
), although they differ in pollen characters (Takahashi et al., 2000
). Palynologically, Sampantaea is close to Cleidion, but the latter is phylogenetically distant in our analyses (subclade A1). Wetria and Pycnocoma are anomalous among Acalyphoideae s.l. in having bioactivity for phorbol esters, but this was not detected in Sampantaea (Argomuellera not examined; Beutler et al., 1996
). Wood anatomy has been used to support an alignment of Wetria with Alchorneeae (i.e., anomalously large rays that might be pre-lysigenous; Hayden and Hayden, 2000
), but this disposition is also not supported palynologically nor by our results.
Core acalyphoid subclade A4
Subclade A4 is a convenient unit for discussion but poorly supported (BP 62) compared with the other acalyphoid subclades recognized here. The subclade is exclusively Old World and contains several groups recognized in previous classifications including Ampereeae, Epiprineae, and three of four subtribes of heterogeneous Chrozophoreae. Ricinus is here and previously has occupied an isolated position in the monotypic Acalypheae–Ricininae, although Webster (1994b)
suggested an affiliation with Adriana. The latter genus associates with Monotaxis (and Amperea for rbcL only) in our trees to form a strongly supported exclusively Australian grouping sister to the remainder of subclade A4. Our analyses uncover new associates for Ricinus including Speranskia, poorly supported as its sister. Both genera differ in many ways, notably in the numerous fascicled stamens of Ricinus vs. 10–15 free stamens in Speranskia. Fascicled stamens are homoplasious in Acalypheae–Lasiococcinae, including Homonoia (rbcL only) and Spathiostemon sampled here (subclade A3). In addition to Ricinus, many members of subclade A4 contain homologous genes for the proteinaceous seed-toxin ricin (K. J. Wurdack, unpublished data). Suregada (gelonin; Hosur et al., 1995
) and Jatropha (curcin; Lin et al., 2003
) produce functionally similar ribosome inactivating proteins (RIPs) but those taxa are unrelated in our analyses.
Core acalyphoid subclades A5, A6
An herbaceous habit is uncommon in Euphorbiaceae s.s., and four relatively delicate genera (Dysopsis, Leidesia, Mercurialis, and Seidelia, all sampled here) have been classified in adjacent Acalypheae subtribes Mercurialinae and Dysopsidinae. Southern African Leidesia and Seidelia form a supported monophyletic group (A5) as expected, given the suggestion by Webster (1994b)
to unite them. Tokuoka and Tobe (2003)
overemphasized the caruncle in Seidelia (absent in Leidesia) as indication of misplacement; the other two genera are not related (A2, Mercurialis; A6, Dysopsis). Members of the strongly supported subclade A6 include Adelieae as sister to Chrozophoreae–Ditaxinae plus Dysopsis (Acalypheae–Dysopsidinae). This clade is morphologically heterogeneous but many members have petals, and it is distributed nearly exclusively in the New World (except a few species of Caperonia). Relationships for Ditaxinae differ greatly from those recovered in a morphological cladistic analysis (Welzen, 1999
), but the latter results show weak support throughout as well as sampling constrained to follow Webster (1994b)
. Dysopsis, a genus of small Neotropical montane herbs, is sister to also herbaceous Caperonia. Dysopsis had been included in Acalypheae–Dysopsidinae, associated with other herbaceous taxa but is here recovered as sister to Caperonia which was classified in Chrozophoreae. Pollen evidence, in part from misinterpretation (see Takahashi et al., 2000
), has been used to suggest an isolated position for Dysopsis (Fernández-González et al., 1994
; Webster, 1994b
). Ditaxis and Chiropetalum, sisters in our analyses, have been considered questionable segregates of Argythamnia s.l., although florally and palynologically they are distinct (Ingram, 1979
). Argythamnia s.s. was not sampled. A grouping of Adelia, Lasiocroton, and Leucocroton is strongly supported as had been suggested by pollen evidence (Takahashi et al., 2000
) and shared tribal classification in Adelieae. The remaining two members of that tribe, Crotonogynopsis and Enriquebeltrania, belong elsewhere (subclade A3 and unresolved in core acalyphoids, respectively) based on rbcL. Enriquebeltrania, a monotypic southern Mexican endemic, had been considered by Webster (1994b)
to be questionably distinct from Adelia; its position in the core acalyphoids is unresolved among the major subclades but clearly far removed from Adelia.
Core acalyphoid subclades A7, A8
Subclade A7 containing Caryodendron and two of three sampled members of Bernardieae is strongly supported as sister to subclade A8, the monophyletic Plukenetieae. Preoccupation with the great diversity of pollen and stylar morphologies in Plukenetieae has obscured understanding of phylogenetic relationships within this group despite probable synapomorphies such as twining habit, pachychalazy (Tokuoka and Tobe, 2003
), and stinging hairs. Cnidoscolus (articulated crotonoids) has clearly independently evolved distinct stinging hairs (Solereder, 1908
; Webster, 1994a
). Among the three Plukenetieae subtribes, stinging hairs are present in Tr
