| 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; 4Department of Higher Plant Systematics and Evolution, Institute of Botany, University of Vienna, Rennweg 14, A-1030 Vienna, Austria; 5Section of Molecular Systematics, Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3DS, UK; 6Rand Afrikaans University, Auckland Park, Johannesburg, 2006, South Africa
Received for publication December 7, 2003. Accepted for publication July 9, 2004.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: Centroplacus • Euphorbiaceae • Malpighiales • molecular phytogenetics • morphology • Pandaceae • Phyllanthaceae • Phyllanthoideae • Putranjivaceae
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). The heterogeneity of Euphorbiaceae s.l. is reflected in a long history of attempts to dismember the family, and at least 20 segregates have been proposed (Webster, 1987
). Corner (1976)
and Huber (1991)
advocated a split of uniovulate Euphorbiaceae from the biovulate taxa based on seed coat characters. A notable, recent attempt by Meeuse (1990)
proposed the recognition of 10 segregate families. The current inclusive sensu lato circumscription of the family and infrafamilial classification was developed by Webster (1975
, 1994b)
and extended by Radcliffe-Smith (2001)
.
The most fundamental division in Euphorbiaceae s.l. is based on ovule number with a grouping of two biovulate subfamilies (Phyllanthoideae and Oldfieldioideae) and three uniovulate ones (Acalyphoideae, Crotonoideae, and Euphorbioideae). This binary division was conceptually provided by de Jussieu (1823
, 1824)
and considered the "first great advance in taxonomic insight" for the family (Webster, 1987
, p. 23). Separation of Phyllanthoideae, and with limited data Oldfieldioideae, from uniovulates is supported by seed-protein serological data (Vogel, 1986
; Jensen et al., 1994
), lack of phorbol ester bioactivity (Beutler et al., 1996
), predominantly ecarunculate seeds (present in most Oldfieldioideae), and derivation of cyanogenic glycosides from tyrosine (Hegnauer, 1977
; Seigler, 1994
). The biovulate subfamilies lack latex and laticifers (but see Balaji et al., 1996
) although those are by no means ubiquitous in the uniovulates.
Chase et al. (1993)
were the first to show the potential polyphyletic nature of Euphorbiaceae s.l. using molecular data that placed two taxa within the subclade of "rosid I" (= eurosid I sensu APG) later circumscribed as Malpighiales (APG, 1998
). Their sampling included only two widely divergent species, uniovulate Euphorbia polychroma A. Kern. (Euphorbiaceae sensu stricto [s.s.]) and biovulate Drypetes roxburghii (Wall.) Hurus. (= Putranjiva roxburghii Wall.; Putranjivaceae), which presented the prospect that the nonsister placement was either an evolutionary reality or a study artifact of inadequate sampling and/or suboptimal analysis of a large data set. Other studies (Conti et al., 1996
; Koontz and Soltis, 1999
; Setoguchi et al., 1999
; Schwarzbach and Ricklefs, 2000
) using these initial two sequences have shown concordance with the basic eurosid I topology.
Phylogenetic hypotheses about the family have been further advanced with additional taxon sampling (Wurdack and Chase, 1996
; Fay et al., 1997
; Litt and Chase, 1999
; Savolainen et al., 2000b
; Chase et al., 2002
) and additional genes (Soltis et al., 1997
, 2000
; Savolainen et al., 2000a
; Tokuoka and Tobe, 2002
; Wurdack, 2002
; Davis and Chase, 2004
). The original APG system (1998)
reclassified Euphorbiaceae s.l. into three lineages, Euphorbiaceae, Putranjivaceae (from Phyllanthoideae tribe Drypeteae), and Pandaceae (from Acalyphoideae tribe Galearieae). Savolainen et al. (2000b)
continued the dismemberment by removing the two remaining biovulate lineages, Phyllanthaceae (from Phyllanthoideae excluding tribe Drypeteae) and Picrodendraceae (as Pseudanthaceae; from Oldfieldioideae), and leaving Euphorbiaceae s.s. comprised of the remaining uniovulates (excluding tribe Galearieae = Pandaceae).
Following these changes, APG II (2003)
recognized three biovulate and two uniovulate families from Euphorbiaceae s.l. All five family-level monophyletic groups remain more or less closely related within Malpighiales. However, even with additional sampling of taxa and/or genes, there are notably few bootstrap supported associations (i.e., >50%) for the segregate families either among themselves or with any of the other families of Malpighiales (see Savolainen et al., 2000a
, b
; Soltis et al., 2000
, 2003
; Wurdack, 2002
; Davis and Chase, 2004
). The Neotropical ditypic genus Paradrypetes, previously considered a "basal" member of Oldfieldioideae (Levin and Simpson, 1994a
), is also a separate lineage (Wurdack and Chase, 1999
) allied with Rhizophoraceae (Wurdack, 2002
). It is aberrant in possessing raphides, colleters, and epipetiolar inflorescences.
The focus of this paper is Phyllanthaceae, the largest of the biovulate lineages. The systematics of the family are presently inferred to coincide with that of Phyllanthoideae (excluding Putranjivaceae). The recognition of Phyllanthoideae or its segregates at the family level has had a long, tumultuous history (Webster, 1987
) that well predates the recent molecular-based reevaluation of the question. Phyllanthoideae span the morphological and chemical diversity of Euphorbiaceae s.l. Their distribution is also pantropical, although they include fewer temperate taxa than Euphorbiaceae s.s. In contrast, they lack worldwide economic plants as present in Euphorbiaceae s.s. A number of taxa are regionally cultivated for their fleshy edible fruits [e.g., Phyllanthus acidus (L.) Skeels, P. emblica L., Baccaurea spp., Antidesma spp.], provide timber, or show medicinal promise (Rizk, 1987
; Calixto et al., 1998
).
Phyllanthoideae have a limited diversity of floral bauplan, notably lacking extreme floral reductions as, e.g., in uniovulate Euphorbia s.l., with the exception of Uapaca. They also lack specialized pollinator adaptations such as brightly colored flowers or zygomorphy, but Glochidion and Phyllanthus subgenus Gomphidium have an apparently species-specific obligate mutualistic relationship with their pollinators (Kato et al., 2003
; Kawakita and Kato, 2004
). Instead, they present innumerable small variations on reproductive organs, especially androecial and glandular elaborations. Their fossil record has been claimed to date back to the Upper Cretaceous based on wood (reviewed by Prakash et al., 1986
). Wheeler (1991
, p. 663) stated that the "phyllanthoid structural pattern is one of the earliest known for dicotyledonous woods," although it is generalized and not unique to any single extant family. Fruits compared to Phyllanthus are known from the Miocene (Nambudiri and Binda, 1989
; Mai, 1996
), and Phyllanthaceae appear well diversified by the Eocene from pollen evidence (Muller, 1981
; Gruas-Cavagnetto and Köhler, 1992
).
Phyllanthoideae have been considered the most primitive subfamily of Euphorbiaceae s.l., from which the others are derived or to which they are sister (Webster, 1994a
). Simpson and Levin (1994)
suggested the subfamily was paraphyletic and united by plesiomorphic characters. Webster (1994b)
classified Phyllanthoideae into eight tribes with 60 genera (including two, Centroplacus and Meborea, incertae sedis) and ca. 2200 species. He also suggested that Tacarcuna belongs in Phyllanthoideae, although placed incertae sedis for the entire family. Tacarcuna was described nearly concurrently with Webster's system (1994b)
; it was at the time poorly known and originally incorrectly described as uniovulate.
Radcliffe-Smith's (2001)
classification includes 10 tribes for 60 genera and closely follows that of Webster (1994b)
. It has detailed descriptions of genera and suprageneric taxa but provides few systematic innovations besides recognizing two monotypic tribes (Centroplaceae and Martretieae) for previously unplaced genera (Centroplacus and Martretia, respectively), submerging Meborea in Phyllanthus and placing Tacarcuna as incertae sedis for Phyllanthoideae. He failed to take notice of the exclusion of Phyllanoa as belonging to Violaceae (Hayden and Hayden, 1996b
). Two additional genera, Distichirhops and Nothobaccaurea (presumably related to Baccaurea), have been recently described (Haegens, 2000
). Dicoelia, previously aberrant in Euphorbiaceae s.s., or sometimes considered close to Pandaceae, and Lingelsheimia (= Aerisilvaea = Danguyodrypetes) that has been included in tribe Drypeteae (= Putranjivaceae sensu APG) have recently been found to be members of Phyllanthaceae (Katriarachchi et al., 2004a
).
Evolutionary relationships among Phyllanthoideae were first depicted in intuitive phylograms by Pax (1924)
. Webster (1984a)
appears to have been the first to conduct a cladistic analysis of any Phyllanthaceae (and any Euphorbiaceae) in his revision of Flueggea. Levin's (1986b)
work on Phyllanthoideae is the broadest published study to date and also groundbreaking in its demonstration of the utility of foliar morphology in resolving relationships. The few other morphological phylogenetic studies are on the Baccaurea alliance (Haegens, 2000
) and limited sampling of Phyllanthaceae as outgroups for Oldfieldioideae (Picrodendraceae) (Levin and Simpson, 1994a
) and Celastraceae (Simmons and Hedin, 1999
). Surveys of wood anatomy (Mennega, 1987
), pollen (Punt, 1962
; Köhler, 1965
), seed-coat structure (Stuppy, 1996
; Tokuoka and Tobe, 1999a
, 2001
), and additional work on leaf venation (Klucking, 1998
) have identified putatively significant systematic characters that have yet to be put in a phylogenetic framework.
No published molecular systematic study has yet focused solely on Phyllanthaceae. The sampling for explicit higher-level studies has included a total of six exemplars using rbcL (Fay et al., 1997
; Litt and Chase, 1999
; Savolainen et al., 2000b
; Chase et al., 2002
). No Phyllanthaceae are present in angiosperm-wide multi-gene phylogenetic studies (e.g., Savolainen et al., 2000a
; Soltis et al., 2000
, 2003
) or gene-content surveys (e.g., Adams et al., 2002
). This demonstrates limitations with exemplar sampling for poorly known polyphyletic groups. Källersjö et al. (1998)
used limited unpublished data generated early in this study, but only generalized trees were presented.
To evaluate relationships at the generic level, the commonly used rbcL gene is well suited for higher-level intrafamilial questions (i.e., Bremer et al., 1995
; Plunkett et al., 1996
; Cameron et al., 2001
). The large existing database of rbcL sequences makes this gene the locus of choice for evaluating taxa of uncertain affinity and the monophyly of families (i.e., Chase et al., 2002
). We acknowledge the preliminary nature of this study because of the limitations of sampling with only a single gene. This study was undertaken to evaluate the circumscription of Phyllanthaceae and monophyly of suprageneric taxa described in current Phyllanthoideae classification schemes, elucidate patterns of intrafamilial relationships, examine the evolution of selected morphological characters, and provide a broad framework for more detailed, future studies. We have refrained from proposing a formal phylogenetic classification of Phyllanthaceae based solely on the analysis of rbcL sequence data. We do, however, wish to highlight well-defined groups that could be incorporated in such a system and the putative morphological synapomorphies characters that appear to support them.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Savolainen et al., 2000b
; Chase et al., 2002
). Sampling included all 10 tribes and 51 of the 60 genera of Phyllanthoideae sensu Radcliffe-Smith (2001)
. A partial sequence of Celianella was obtained during manuscript revisions. Potentially the most significant taxa not sampled here as a result of lack of adequate material include Ashtonia, Dicoelia, Lingelsheimia, Protomegabaria, and Richeriella. Outgroups were chosen based on larger phylogenetic analyses (Chase et al., 1993
; Savolainen et al., 2000a
, b
; Soltis et al., 2000
; Wurdack, 2002
) on subsets of the 500+ members of Malpighiales for which rbcL sequences are presently available. Biovulate Euphorbiaceae are clearly polyphyletic (Fig. 1a in Chase et al., 2002
) and lack bootstrap-supported sister relationships, making the choice of malpighialean outgroups somewhat arbitrary on the basis of molecular evidence. Picrodendraceae are probably the best outgroup choice for Phyllanthaceae (see Discussion; Wurdack, 2002
; Fig. 4 in Davis and Chase, 2004
). Given the low bootstrap support for a sister relationship of Phyllanthaceae and Picrodendraceae, and the need to encompass all euphorbiaceous lineages, multiple outgroups were chosen. These included Euphorbiaceae s.s., Humiriaceae, Irvingiaceae, Lophopyxidaceae, Pandaceae, Picrodendraceae, and Putranjivaceae. Humiriaceae seem particularly useful for malpighialean lineages that lack supported sister groups due to their low levels of molecular divergence (see Fig. 1b in Chase et al., 2002
). An analysis including representatives of all biovulate lineages allows taxa to group with individual clades and readily reveals misclassified taxa either by exclusion or emerging phylogenetic patterns.
Laboratory methods
Samples and data were gathered over nine years and were subject to a diverse array of evolving methods and technologies. DNA extraction, polymerse chain reaction (PCR), and sequencing methods largely followed Chase et al. (2002)
or Wurdack (2002)
. Recent extractions involving about half of the accessions (and nearly all herbarium material) have been made using the DNeasy Plant Mini Kit (Qiagen, Valencia, California, USA) following tissue disruption of 0.5–1 cm2 of leaf tissue in a FastPrep FP-120 bead mill using lysing matrix "A" tubes containing one ceramic bead and garnet sand (Qbiogene, Carlsbad, California, USA). DNA extractions followed manufacturer's protocols with the modification of buffer AP1 lysis conditions by the addition of 0.4–0.7 mg of PCR-grade proteinase K (Roche, Indianapolis, Indiana, USA), 6.5% 2-mercaptoethanol or 15 mg dithiothreitol, and incubation at 42°C for 12–48 h on a rocking platform. This mini-prep method is primarily designed for herbarium material and minimizes sample mass and contamination risks. The addition of N-phenacylthiazolium bromide (PTB; Poinar et al., 1998
) or 4,5-phenacyldimethylthiazolium chloride appeared helpful for recalcitrant herbarium material. Many Phyllanthaceae (especially Flueggeinae and Leptopus) were mucilaginous and difficult to extract. Jablonskia contained a potent PCR inhibitor that was removed by centrifugal ultrafiltration with Ultrafree-MC (100 000 nominal molecular weight limit; Millipore, Billerica, Massachusetts, USA) following the manufacturer's directions.
The rbcL exon was amplified as one piece using primers 1F and 1368R/ 1460R (or variants) or as two overlapping fragments using combinations 1F-724R and 636F-1460R (For primer information, see Appendix 2 in Supplemental Data accompanying online version of this article.). Three noncontiguous fragments of Tacarcuna from a highly degraded sample were generated using 1F-367R, 636F-854R, and 1024F-1368R. PCR products were purified and directly sequenced for both strands. For 17 taxa, rbcL sequences were obtained by radioactive, manual-sequencing methods using eight primers (1F, 234F, 424F, 633F, 878F, 286R or 346R, 895R, 1460R). Autoradiograms were read by eye and recorded by hand.
The remaining sequences were more recently generated by fluorescent sequencing on ABI Prism 373A or 377XL automated sequencers (Applied Biosystems, Foster City, California, USA) with dye-terminator ABI Prism Ready Reaction mix and typically using four primers (or any additional amplification primers). Sequences were assembled and edited in Sequencher 3.1.1 (Gene Codes, Ann Arbor, Michigan, USA). Significant missing data are in Celianella, Maesobotrya sp., Tacarcuna, and Uapaca sp.
Data analysis
Sequences were aligned by eye (no indels present), and bases 1–30 that formed the primer-binding region most inclusive of all 1F primer variants used were excluded to reduce missing data. Maximum parsimony searches were conducted using PAUP* 4.0b10-Altivec (Swofford, 2003
) with 1000 replicates of random taxon addition, equal weights, and unordered characters (Fitch parsimony; Fitch, 1971
), and tree bisection-reconnection (TBR) branch swapping with 10 trees held at each step (MulTrees, saving multiple equally short trees, on) to save time swapping on large numbers of suboptimal trees. The resulting trees, including multiple minimal-length trees, were used as starting trees in another round of TBR and a maximum tree limit of 20 000. Branches with a minimum length of zero were collapsed. 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
). Partitioned analyses were conducted of 5' and 3' halves (or in the case of Tacarcuna, each of the three noncontiguous fragments was examined) of rbcL corresponding to primer pairs used for amplifying degraded samples. Incongruent positions between these separate searches could be evidence of chimeric sequences composed of data derived from different taxa (many taxa were amplified in two pieces using the internal primers described earlier, which could result in a contaminating sequence being preferentially amplified if the desired template DNA was highly degraded, as is often the case with herbarium DNA). Relative support for clades was evaluated using the bootstrap (Felsenstein, 1985
). For the bootstrap, 1000 replicates were performed using TBR swapping with each replicate consisting of five random taxon additions, holding 10 trees at each step and saving no more than 10 trees (nchuck = 10, chuckscore = 1) per iteration. Bootstrap percentages are described as high (85–100%), moderate (75–84%) or low (50–74%). Tree and character manipulations were carried out in MacClade 4.0 (Maddison and Maddison, 2000
). These included translations to amino acids to check for internal stop codons (none found).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
) were detected at the RuBisCo active site or other conserved residues noted by Kellogg and Juliano (1997)
. Putranjivaceae (including Drypetes, Putranjiva, and Sibangea) contains a suite of unusual synapomorphies (Y-33, W-248, Q-440, K-447, N-453), but these amino acid differences do not affect major structural functions and do not appear to be coordinated changes (see examples in Kellogg and Juliano, 1997
). The stop codon TGA (vs. TAA in all other Phyllanthaceae except as noted later) was present in Breynia, Phyllanthus flagelliformis Müll. Arg., P. fluitans Benth., P. liebmannianus Müll. Arg. subsp. platylepis (Small) G. L. Webster, and Reverchonia but otherwise absent in Phyllanthaceae and rare in Malpighiales. This is the result of a duplication within the 3' terminus of the gene where bases 1397–1428 have been shifted to 1422–1453 following a 25-base pair (bp) duplication (e.g., positions 1397–1421 and 1422–1446 form tandem paralogs, shifting the otherwise conserved TAA stop codon into the flanking noncoding region but resulting in a new stop codon that terminates the gene at the typical 1428 bases). One additional predicted amino acid beyond the usual malpighialean 475 is present in Phyllanthus nutans Sw. (I-476) and Spondianthus (K-476). Andrachne microphylla (Lam.) Baill. and Andrachne telephioides L. contain three additional amino acids (V-476,V-477, L-478). Four species, each represented by two sequences from separate accessions, have variation. Raw data were verified for Discocarpus (one base difference) and the Kenyan Heywoodia (AY663587). It is unknown if error or genuine polymorphism contributes to Breynia, Putranjiva roxburghii, and the other Heywoodia sequence differences. The two Poranthera huegelii Klotzsch sequences were identical as were Leptopus chinensis (Bunge) Pojark. and L. colchicus (Fisch. & C. A. Mey. ex Boiss.) Pojark.
The 109-sequence (76 Phyllanthaceae before the inclusion of Celianella) rbcL matrix contained 1398 characters (1.74% missing or ambiguous cells) of which 464 (365, Phyllanthaceae only) were variable and 359 (266, Phyllanthaceae only) potentially parsimony informative. Phylogenetic analysis yielded the upper limit of 20 000 most parsimonious trees (MPT) of length 1633 steps, CI = 0.39 (0.34 excluding uninformative characters), RC = 0.30 (0.27 excluding uninformative), and RI = 0.77. The strict consensus with bootstrap percentages (
50) is shown in Fig. 1 and a single most parsimonious tree in Fig. 2. A reduced sampling containing only Phyllanthaceae + Picrodendraceae (removing all unresolved outgroup taxa, especially Putranjivaceae) did not produce less than the tree limit of 20 000 MPT (results not shown). The partitioned analyses (i.e., the two halves of rbcL) did not detect any bootstrap-supported (BP > 50) incongruence.
|
|
are polyphyletic due to the inclusion of Centroplacus (although its sister position to Pandacaeae has BP < 50) and Drypetes, Putranjiva, and Sibangea (Putranjivaceae), and the exclusion of Croizatia (Oldfieldioideae). Phyllanthaceae are resolved into two well-supported sister clades (BP 91 and 98, respectively), hereby designated the fasciculate and tanniniferous clades. These names refer to different inflorescence types (axillary fascicles vs. elongate), and the absence and presence of tanniniferous epidermal cells. With the exception of bigeneric Hymenocardieae rbcL does not support the monophyly of any of the current tribes in Phyllanthoideae (Phyllanthaceae) (Webster, 1994b
).
The fasciculate clade (BP 91) contains Heywoodia and five supported subclades. Subclade F1 (BP 100) includes Phyllantheae subtribe Flueggeinae plus Savia bahamensis Britton (Wielandieae). Subclade F2 (BP 95) unites 10 genera from four phyllanthoid tribes (Amanoeae, Bridelieae, Phyllantheae, and Wielandieae), oldfieldioid Croizatia, and incertae sedis Tacarcuna. This subclade includes Savia section Savia, thereby making that genus biphyletic. Cleistanthus (Bridelieae) is paraphyletic in a well-supported subclade (BP 99) together with Bridelia (Bridelieae), Pentabrachion (Amanoeae), and Pseudolachnostylis (Phyllantheae). The strongly supported subclade F3 (BP 99) contains four genera from three subtribes of tribe Phyllantheae, Actephila (Wielandieae), and Poranthera, the only member of Antidesmeae to fall outside the tanniniferous clade. The anomalous position of "Poranthera sp." embedded in Euphorbiaceae s.s. in Chase et al. (2002)
is based on a misidentified collection. Chase 2162 is correctly Monotaxis megacarpa F. Muell. (Euphorbiaceae s.s.) and excluded in favor of new accessions of authentic Poranthera. All Wielandieae from the western Indian Ocean region (Malagassia) are found in subclade F4 (BP 100). Astrocasia and Chascotheca form a well-supported sister group (subclade F5; BP 100). Heywoodia (subclade F6) does not have supported relationships. The tanniniferous clade (BP 98) includes all members of Antidesmeae except Poranthera, as well as Bischofieae, Hymenocardieae, and Martretieae. The Antidesmeae subtribes Antidesminae (including Hymenocardieae and Martretieae) and Scepinae each form well-supported (BP 90+) subclades (T1 and T2, respectively). Bischofia, Jablonskia + Celianella (the latter not shown), Spondianthus, and Uapaca mostly lack supported sister groups, although they are resolved in the strict consensus (Fig. 1) and Spondianthus + Bischofia is weakly supported (BP 50).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
) recognized three segregate families (Phyllanthaceae, Picrodendraceae, and Putranjivaceae) of biovulate Euphorbiaceae s.l. Until now, delimitation of these families has relied on inferences from previous sensu lato classifications (e.g., Webster, 1994b
; Radcliffe-Smith, 2001
). Our analyses recover all three families for which circumscription is largely in accordance with that predicted from the morphologically based classifications. Due to the focus of this paper on Phyllanthaceae and the resulting limited outgroup sampling, a close relationship among these families should not be inferred by proximity in our trees. Additional sampling and analyses of more genes have so far only modestly improved support for deep branches along the spine of Malpighiales phylogenies. Three-gene analyses (atpB, rbcL, 18S rDNA; Wurdack, 2002
) recovered Phyllanthaceae + Picrodendraceae in the strict consensus, but this sister relationship received no support (BP < 50). These same analyses have low support for Phyllanthaceae but strong support for each of its two main clades. Davis and Chase (2004)
found weak support (BP 53) for Phyllanthaceae + Picrodendraceae using combined ndhF, rbcL, and PHYC (although BP < 50 in separate plastid and nuclear analyses) and their strong support for Phyllanthaceae (BP 98) reflects sampling only two members of the fasciculate clade that in our analyses also receives strong support (BP 91).
Picrodendraceae differ from Phyllanthaceae in pollen characters (Levin and Simpson, 1994a
), notably the echinate pollen exine first used by Köhler (1965)
to distinguish the group (as Oldfieldioideae) and more than three pollen apertures. Pollen with spinose sculpturing (e.g., in Amanoa and Securinega) or more than three apertures (e.g., in Phyllanthus) rarely occurs in Phyllanthaceae and appears to be the result of convergence (Levin and Simpson, 1994a
; Simpson and Levin, 1994
). Additionally, Phyllanthaceae are ecarunculate (but see Stuppy, 1996
, for Celianella) and frequently have petals, whereas Picrodendraceae are apetalous and usually have carunculate seeds. Compound leaves are common in Picrodendraceae but in Phyllanthaceae are restricted to Bischofia javanica Blume (Table 1).
|
; D. Sutter, University of Zurich, Zurich, Switzerland, personal communication). Ovule number must be used with care and in some cases, scrutinized from an anatomical-developmental standpoint. Many biovulate taxa appear uniovulate in fruit (e.g., Antidesma, Spondianthus, Tacarcuna, Wielandia) from the abortion of one of the paired ovules. Some taxa may have carried this further and secondarily lost an ovule although they have not been investigated for embryological vestiges at early stages. These include uniovulate Scagea, which was placed in the biovulate Oldfieldioideae. Pseudanthus ovalifolius F. Muell., from the same subfamily, is the only uniovulate species in an otherwise biovulate genus (Halford and Henderson, 2003
). This is also implicit in the proposed general derivation of uniovulates from ancestral biovulates noted earlier (Webster, 1994a
).
Sutter and Endress (1995)
noted a suite of gynoecial similarities (potential synapomorphies) between uni- and biovulate Euphorbiaceae s.l., including an extended nucellar beak, divided styles often with a ventral furrow, epitropous ovules, obturators present, carunculate seeds (usually lacking in Phyllanthaceae), unicellular-papillate stigmas, and often trimerous gynoecia. A nucellar beak, epitropous ovules, and obturator may in combination be unique to Euphorbiaceae s.s., Phyllanthaceae, and Picrodendraceae. The schizocarp with complex explosive dehiscence in these three lineages could also be synapomorphic and adaptively related to the embryology (Berg, 1975
; but see also Webster's [1994a]
critique). It is unclear whether these complex characters have evolved multiple times or whether they have been lost in some families that are related to and possibly interdigitated with Euphorbiaceae s.l. that formerly had them (i.e., become secondarily indehiscent).
Excluded taxa
The third biovulate lineage, Putranjivaceae, appears more distantly related to the other former members of Phyllanthoideae as suggested by embryological evidence (Tokuoka and Tobe, 1999b
), especially the lack of a nucellar beak. It has been primarily united with Phyllanthoideae based on characters that could be interpreted as plesiomorphic. The unusual sieve-element plastid type (PIcs) recorded for Drypetes and Bischofia but no other Phyllanthaceae (Behnke, 1981
) was erroneous (H.-D. Behnke, Heidelberg University, personal communication). Both Drypetes and all 31 examined species of Euphorbiaceae s.l. have S-type plastids (H.-D. Behnke, unpublished data), as have 16 other malpighialean families. An exception is Rhizophoraceae sensu APG II (including Erythroxylaceae), which have the PVc type. Despite the extremely limited phytochemical knowledge of the family, the presence of mustard oils (glucosinolates) in Putranjivaceae has been unduly emphasized for systematic comparisons, usually with Brassicales (Rodman et al., 1993
, 1996
; Seigler, 1994
). Putranjivaceae share n = 20 (Hans, 1973
, with probably erroneous counts; see Chattopadhyay and Sharma, 1988
), whereas Phyllanthaceae have predominantly x = 13 (see later).
Large-scale rbcL analyses (Savolainen et al., 2000a
, b
; Wurdack, 2002
) place Irvingiaceae and Putranjivaceae as sisters, whereas small-scale rbcL analyses (i.e., as shown here) and data from other genes (18S rDNA, atpB, and combined with rbcL; Wurdack, 2002
) support a relationship of Putranjivaceae to monotypic Lophopyxidaceae. All three families are isolated lineages, and Putranjivaceae is the most divergent member in this analysis (see Fig. 2). Lophopyxis maingayi Hook. f. has a scandent habit with tendrils and anomalous secondary growth, petals, and dry, winged fruits. Putranjivaceae, in contrast, are strictly shrubby or arborescent, apetalous, and drupaceous. Characters consistent with common ancestry include leaves with theoid teeth, dioecy, and many floral features such as pentamerous flowers with a disc, styles separate or nearly so, and two pendulous apical-axile anatropous ovules per locule (each with an obturator), one of which aborts during the development of the indehiscent fruit.
Putranjiva and Sibangea closely resemble the large pantropical genus Drypetes, and their generic distinctiveness has been questioned (Webster, 1994b
). In our analyses, relationships within Putranjivaceae are poorly resolved. It is indicated that Sibangea nests within a paraphyletic Drypetes and should be subsumed under that genus. This is concordant with the subtle generic differences given by Radcliffe-Smith (1978
, 2001)
and Webster (1994b)
. In Sibangea, the pistillate sepals are open in bud and persistent in fruit, whereas in Drypetes they are imbricate and deciduous. The pistillate sepals are narrow and therefore do not overlap as in regular Drypetes flowers. Pax and Hoffmann (1922)
included Sibangea in Drypetes section Hemicyclia. Sibangea arborescens Oliv. from Gabon belongs to a poorly supported clade also containing four Drypetes species from West Africa (two each of section Sphragidia and Stemonodiscus, respectively) which is in turn sister to two Australasian species, D. deplanchei (Brongn. & Gris) Merr. and D. macrostigma J. J. Sm. The three Neotropical species, D. lateriflora (Sw.) Urb., D. diversifolia Krug & Urb., and D. brownii Standl., do not resolve as a monophyletic group. Drypetes lateriflora is the only New World member of section Oligandrae, and the remaining Neotropical species belong to section Drypetes (Webster, 1967
), which also has Old World representatives. The sectional classification used in the latest revision of the entire genus (Pax and Hoffmann, 1922
) is not concordant with our results and is in need of reevaluation.
The small Asian genus Putranjiva is sister to Drypetes including Sibangea but this relationship lacks bootstrap support. Differential characters between Putranjiva and Drypetes lie in leaf venation, absence/presence of a floral disc, number of stamens, and shape of stigmas. The topology of Putranjivaceae presented here indicates a potentially complex biogeographic history as only c. 20 of c. 200 Drypetes species are Neotropical (G. Levin, Center for Biodiversity, Illinois Natural History Survey, personal communication). The inclusion of more species, especially southern Asian Drypetes, and the use of more rapidly evolving molecular markers is called for.
Lingelsheimia (not sampled here), a small poorly known Afromalagasy genus, has been placed in tribe Drypeteae (Webster, 1994b; Radcliffe-Smith, 2001
). It is anomalous in Putranjivaceae but not Phyllanthaceae based on seed coat data (Tokuoka and Tobe, 2001
) and leaf morphology (Levin, 1986b
). The genus has been confirmed as member of Phyllanthaceae (Katriarachchi et al., 2004a
).
The affinities of Centroplacus (reviewed by Radcliffe-Smith, 2001
) have been thought to be with Celastraceae, Flacourtiaceae, Pandaceae, or most recently Phyllanthaceae. The APG II (APG, 2003
) highlighted this wandering affiliation by considering it incertae sedis. Radcliffe-Smith (2001)
returned to the views of Pax and Hoffmann (1931)
by including Centroplacus in Phyllanthoideae, but his placement in a monotypic tribe Centroplaceae at the end of the subfamily did little to further clarify relationships. Pollen morphology (Punt, 1962
; Köhler, 1965
) excluded the genus from biovulate Euphorbiaceae and indicated a relationship with Acalyphoideae tribe Galearieae (= Pandaceae). Seed coat data have been interpreted controversially. Stuppy (1996)
excluded Centroplacus from biovulate Euphorbiaceae, whereas Tokuoka and Tobe (2001)
found no reason to separate it from Phyllanthaceae. Comparative seed coat studies with Pandaceae have not been published.
Our analyses place Centroplacus outside of Phyllanthaceae as the unsupported (BP < 50) sister to uniovulate Pandaceae. Airy Shaw (in Willis, 1966
) put Centroplacus in Pandaceae, an affinity recently accepted by Takhtajan (1997)
and Govaerts et al. (2000)
. Centroplacus differs from Pandaceae in, e.g., possessing two ovules per locule (vs. one), dehiscent fruits (vs. drupaceous), and carunculate seeds (vs. ecarunculate) (Table 1).
Relationships in Phyllanthaceae
The rbcL tree (Fig. 1) strongly supports relationships in Phyllanthaceae that are substantially different from those suggested in previous classifications. Seed coat data (Tokuoka and Tobe, 2001
) are partly congruent with these results, and exotegmens with ribbon-like cells appear plesiomorphic for the family. Wood anatomy (Aporosa vs. Glochidion types; Metcalfe and Chalk, 1950
; Mennega, 1987
) does not correlate well with the major clades recovered, although it has mainly been approached from a typological rather than a phylogenetic perspective and needs to be reconsidered. Vegetative anatomy (e.g., Gaucher, 1902
) indicates coherence within the tanniniferous (T1–T4) clade and within the Flueggeinae (F1) clade but provides no further characters uniting major groups recognized here. This corresponds with Pax and Hoffmann's (1931)
statement that anatomical data generally do not support delimitation of major groups but are often useful for defining smaller taxonomic units. The occurrence of tanniniferous epidermal cells in one of the major phyllanthoid clades recovered here, however, shows that a thorough examination of morphological/anatomical characters may still yield more significant indicators of phylogenetic relationships. In considering cytological data (reviewed by Hans, 1973
, but see critique by Webster, 1994a
, on the partly erroneous conclusions; also Urbatsch et al., 1975
; Tammaro and Pogliani, 1977
; Humphries et al., 1978
; Gill et al., 1981
; Podlech, 1986
; Rossignol et al., 1987
) in relation to our results, x = 13 appears to be the base chromosome number for the family.
Phyllanthaceae are resolved into two well-supported sister clades corresponding to the distribution of tanniniferous epidermal cells and to inflorescence types. The inflorescence type is used here as in the second couplet of Webster's tribal key (1994b
, p. 35): axillary clusters with "inflorescence axes usually less than 1 cm long" vs. "spicate or racemose" inflorescences with axes usually over 1 cm long. The clade with contracted inflorescence axes and lacking tanniferous cells is earlier designated as the fasciculate clade comprising subclades F1–F6, and its sister clade with predominantly elongated inflorescence axes is earlier designated as the tanniniferous clade comprising subclades T1–T4. These two clades could prove a logical subfamilial division, although they presently lack uncontradicted morphological distinctions.
The clustered vs. elongated appearance of the inflorescence was noted by Bentham and Hooker (1880)
and later adopted in major classifications to delimit the core of the tanniniferous clade (Pax and Hoffmann, 1922
, as Antidesmineae; Webster, 1994b
, as Antidesmeae + Hymenocardieae). Exceptions are in the tanniniferous clade subclade T3 with contracted inflorescences and in the fasciculate clade the rare occurrence of (possibly secondarily) elongated inflorescences (i.e., Amanoa strobilacea Müll. Arg., Sauropus racemosus Beille). Axillary fasciculate inflorescences are sometimes concentrated at the tips of the branches and subtended by much-reduced foliage leaves (e.g., in some species of Amanoa, Bridelia, Petalodiscus, and Poranthera). In these cases the subtending leaves resemble regular foliage leaves in all but size and have normal stipules. Bracts subtending flowers in elongated inflorescences of the tanniniferous clade are clearly differentiated from the foliage leaves and lack stipules. No systematic study (although Webster [1956–1958]
discussed variation and issues in Phyllanthus) of inflorescence morphology in Phyllanthaceae has yet been undertaken to establish homology of the inflorescence types.
The fasciculate clade
The first major clade of Phyllanthaceae includes members with axillary, fasciculate clusters of flowers and no tanniniferous epidermal cells in leaves. Tannin deposits, but not the enlarged idioblastic leaf epidermal cells characteristic of most of the tanniferous clade (see Levin, 1986a
), have been detected in the leaves of Discocarpus (Hayden and Hayden, 1996a; not reported by Levin, 1986a
) and the leaves and stems of some Phyllanthus species (Webster, 1956–1958
).
Subclade F1
Phyllantheae subtribe Flueggeinae + Savia bahamensis (Wielandieae) are strongly supported (BP 100) but include only few supported internal relationships. Members of this subclade contain securinine alkaloids that are synthesized by a unique pathway and not found in other plants (Seigler, 1994
and references therein). Savia is not monophyletic, and it would appear that Savia section Heterosavia belongs in Flueggeinae (F1) and Savia section Savia belongs to subclade F2 with other elements of Wielandieae. The morphological differences between the two sections of Savia are slight (P. Hoffmann, unpublished manuscript), but pollen of Savia section Heterosavia is aberrant in Wielandieae (Punt, 1962
; Köhler, 1965
). The generic identity of Flueggea neowawraea W. J. Hayden, a nearly extinct Hawaiian tree superficially resembling and at one time transferred to Drypetes, is in agreement with the findings of Hayden (1987)
.
Phyllanthus with over 800 species (Govaerts et al., 2000
) is the most species-rich genus of Phyllanthaceae. It has a diversity of growth forms (annual, arborescent, aquatic, pachycaulous, and phyllocladous), chromosome numbers, and pollen types rivalling that of any genus of flowering plants. Webster's ongoing treatment (1956–1958
, 1967
, 1970
, 1986
, 2001
, 2002
, 2003
; Webster and Airy Shaw, 1971
; Webster and Carpenter, 2002a
, b
) of Phyllanthus includes a broad circumscription (subsuming numerous segregate genera) and the creation of a more natural but complex infrageneric classification of 10 subgenera and over 30 sections based on branching patterns and pollen types in addition to floral characters. The 10 Phyllanthus species sampled include three species with spiral phyllotaxy (P. liebmannianus subsp. platylepis and P. calycinus Labill. from subgenus Isocladus; P. nutans from subgenus Xylophylla), two highly specialized phyllocladous species (P. epiphyllanthus L. from subgenus Xylophylla and P. flagelliformis from subgenus Phyllanthus), and the odd plagiotropic aquatic P. fluitans that was classified by Brunel (1987)
in subgenus Phyllanthus. The other four species belonging to subgenera Emblica (P. polyphyllus Willd.), Kirganelia (P. nummulariifolius Poir.), Phyllanthus (P. lokohensis Leandri), and Xylophylla (P. juglandifolius Willd.) have typical "phyllanthoid" branching (see Webster, 1956–1958
) with spirally arranged, deciduous, floriferous short-shoots, resembling compound leaves, borne on cataphyllous, indeterminate long-shoots.
The Phyllanthus species do not form a monophyletic group with rbcL data, but the sampling is poor considering the species richness of the genus. At face value, the strict consensus indicates complex patterns of habit evolution and biogeography in Phyllanthus. Morphological specialization has created classification difficulties, and it is not unusual for problems to occur when there is such diversity in form. Phyllanthus resembles the other large euphorbiaceous genera, Euphorbia s.l. (see Steinmann and Porter, 2002
) and Croton s.l. (also probably polyphyletic; Wurdack, 2002
; P. Berry et al., unpublished data), in which paraphyletic genera have been formed by recognition of specialized lineages, leaving a rump of more plesiomorphic species in the parent genus. It is clear that further work on Phyllanthus will not only require more rapidly evolving genes with greater numbers of variable positions but also more extensive sampling from Flueggeinae. As noted in the Results, an unusual and perhaps synapomorphic 3'-duplication unites Breynia, Phyllanthus flagelliformis, P. fluitans, P. liebmannianus subsp. platylepis, and Reverchonia. There is no support (BP < 50) for this grouping and alternative relationships (although partly united in two separate, poorly supported groups) are present in the strict consensus tree. The two phyllocladous species sampled are from two proposed independent origins of the growth form (Webster, 1956–1958
), as suggested by their classification in different subgenera and recent pollen investigations (Webster and Carpenter, 2002a
). Phyllanthus fluitans is the only fully aquatic Euphorbiaceae s.l. The plant is a free-floater resembling Salvinia, with which it sometimes grows, and is becoming popular in the commercial aquarium market. In the Adams consensus tree (not shown but see Fig. 2) it is sister to the phyllocladous P. flagelliformis.
According to our results, the sand dune annual Reverchonia, revised by Webster and Miller (1963)
, does not warrant generic status but rather represents a highly specialized Phyllanthus. The main generic characters of Reverchonia are a central staminate disc, otherwise unknown in Phyllanthaceae, and the narrow cotyledon shape. The latter is shared with Poranthera and unrelated taxa of ericoid habit in Euphorbiaceae s.s. and Picrodendraceae that were classified in a series "Stenolobeae" apart from all other Euphorbiaceae s.l. by Müller (1866)
. The distribution of this character (Fig. 2) suggests pleiotropic effects associated with extremely reduced leaves. The chromosome number in Reverchonia is 2n = 16. This is rare in Phyllanthaceae but corresponds with Phyllanthus section Isocladus (containing Phyllanthus liebmannianus), which, like Reverchonia, is characterized by the lack of phyllanthoid branching typical for most other Flueggeinae (Webster and Miller, 1963
). Pollen of Reverchonia has been reported from the Eocene of France (Gruas-Cavagnetto and Köhler, 1992
), suggesting its present limited North American distribution is relictual.
Subclade F2
The next subclade (F2 of Fig. 1) contains four genera of Wielandieae, as well as Amanoeae, Bridelieae, Croizatia (Oldfieldioideae), Tacarcuna (incertae sedis), and one member of Phyllantheae (Securinega). They share petaliferous flowers except for Pseudolachnostylis, Securinega, and perhaps Tacarcuna. In addition, the mitochondrial cox1 intron is absent in this clade, whereas all other Phyllanthaceae sampled possess it (Wurdack, 2002
). Wielandieae have been considered the "basal" tribe of Phyllanthoideae and of probable paraphyletic circumscription (Webster, 1994b
) and in our analysis fall in four of the five fasciculate-clade lineages.
Subclade F2 contains four supported groups including a weakly supported (BP 63) strictly New World subclade with Croizatia + Discocarpus + Tacarcuna and Gonatogyne + Savia section Savia. There are two strongly supported Old World groups, one comprising Bridelia, Cleistanthus, Pentabrachion, and Pseudolachnostylis, and the other, strictly African group, including Lachnostylis and Securinega. The fourth group consists of Amanoa, the only genus in this clade with a trans-Atlantic disjunction. Neotropical A. caribaea Krug & Urb. and African A. strobilacea Müll. Arg. are strongly supported (BP 100) as monophyletic. Amanoa shares sclerified walls of the leaf epidermis with Discocarpus (Rothdauscher, 1896
; Gaucher, 1902
; Hayden, 1980
; Levin, 1986a
; Hayden and Hayden, 1996a
), although this character was homoplasious in phylogenetic reconstructions using leaf morphological data (Levin, 1986b
).
Croizatia contains five species of Neotropical shrubs that closely resemble members of Wielandieae. The genus has been classified as the first branching lineage of Oldfieldioideae (Picrodendraceae), principally on shared echinate pollen even though it represented a discordant element in possessing petals, tri-aperturate pollen, and ecarunculate seeds lacking endosperm (Webster et al., 1987
; Levin and Simpson, 1994a
). Cladistic analyses of Oldfieldioideae by Levin and Simpson (1994a)
showed an unstable placement of Croizatia when using palynological characters alone vs. when combined with morphology, suggesting conflicting signal from homoplasious characters (e.g., foot-layer structure). Echinate pollen has been derived elsewhere in Phyllanthaceae (Amanoa, Securinega). In the case of Amanoa, the echinae appear clearly homoplasious and are not supratectal but derived from columellae of intectate pollen (Levin and Simpson, 1994a
).
Two species of Croizatia were sequenced (one nucleotide difference) for confirmation of the unexpected placement in Phyllanthaceae in a strongly supported heterogeneous clade containing Discocarpus and Tacarcuna. Croizatia is monophyletic in the Adams consensus tree (not shown). The lack of resolution in the strict consensus may be attributed to missing data in Tacarcuna and low levels of sequence divergence. Placement of Croizatia with Phyllanthaceae has been supported by analyses of other sequence data (18S, atpB, trnL-F, nad1; Wurdack, 2002
). Dorr (1999)
reviewed the disposition of monotypic Pseudosagotia and proposed a new combination, Croizatia brevipetiolata (Secco) Dorr, which is here supported based on sequence data. Species delimitation among the similar-looking species of Croizatia remains poorly understood and needs to be reevaluated in light of numerous new collections.
Tacarcuna contains three poorly known species of Neotropical trees. The genus resembles other Wielandieae, but this affinity was initially obscured by being incorrectly described (Huft, 1989
) as uniovulate. Subsequent observation has shown this to be by abortion and the undeveloped second ovule is sometimes even persistent on the columella of dehisced fruits (K. J. Wurdack, personal observation). Tacarcuna shares with Croizatia and Discocarpus (illustrated by Stuppy, 1996
) large embryos with thin, contorted cotyledons that fill the seed and lack (or nearly so) endosperm in the mature seed. It does not appear to have distinct petals that are usually present in members of subclade F2 but are highly reduced in Croizatia and Discocarpus. Huft (1989)
originally interpreted the flowers as containing a disc and a perianth of five sepals (T. gentryi Huft) or three sepals and three petals (T. amanoifolia Huft, T. tachirensis Huft). In the latter case, the perianth parts are poorly differentiated despite the positional distinctiveness of the two whorls and are persistent in fruit (K. J. Wurdack, personal observation). The previously unrecorded high stamen number in T. amanoifolia (14–19, among the highest in Phyllanthaceae) suggests this taxon is derived. Most members of subclade F2 (including T. gentryi) have only five stamens. Although we saw no evidence of chimerism or contamination, given the degraded nature of the DNA sample, additional data for Tacarcuna are desirable to confirm our results.
Lachnostylis (South Africa) and Discocarpus (northern South America) have been suggested to represent a vicariant pair (Bentham and Hooker, 1880
; Webster, 1994a
) and even treated as synonymous by Pax and Hoffmann (1922
, 1931)
. A close relationship is not supported by our data, although it has recently been reaffirmed morphologically, and an affiliation with the poorly known Chonocentrum dismissed (Hayden and Hayden, 1996a
, b
). Lachnostylis has a strongly supported (BP 94) relationship with the Madagascan Securinega capuronii Leandri despite little morphological resemblance between the genera. As with Savia and Andrachne (see later), Securinega has had considerable flux in generic delimitation. The broad circumscription of Pax and Hoffmann (1931)
has been reduced with the removal of Flueggea (Webster, 1984a
), Jablonskia (Webster, 1984b
), and Meineckia (Webster, 1965
), leaving a core Malagassian group characterized by distinctive spiny pollen and smooth seeds.
The Brazilian monotypic genus Gonatogyne has been included as a section of Savia by Pax and Hoffmann (1922
, 1931)
in a treatment adopting the widest circumscription of Savia, including Petalodiscus and species of Andrachne and Leptopus. The same authors reinstated Gonatogyne at generic rank after examining more material (Pax and Hoffmann, 1933
). In our analysis, Gonatogyne is sister to Savia section Savia (for Savia section Heterosavia see subclade F1). Morphological characters shared between Gonatogyne and Savia section Savia are the terete (vs. adaxially channeled in section Heterosavia) petiole, articulated pistillate pedicel, caducous perianth in fruit, and distally narrow columella. Differences between Gonatogyne and Savia lie in the shape of the floral disc, style division, fusion of the androecium, petiolar vascularization, and fruit dehiscence (P. Hoffmann, unpublished manuscript).
Due to their valvate sepals, Bridelia and Cleistanthus were united in tribe Bridelieae. This tribe was juxtaposed with tribe Phyllantheae containing all other phyllanthoid genera except Poranthera in most pre-Websterian classifications of Euphorbiaceae (Müller, 1866
; Jablonszky, 1915
; Pax and Hoffmann, 1922
, 1931
). The importance of this character was contested by Baillon (1873)
, who went so far as to include Bridelia, Cleistanthus, Gonatogyne, and Pentabrachion in Amanoa, which in turn, he considered to be closely related to Lachnostylis. Webster (1975
, 1994b)
took an intermediate view and maintained tribe Bridelieae in Phyllanthaceae using calyx aestivation as a key differential character. Our results show that this character has at most generic value in Phyllanthaceae.
Genera of subclade F2 are linked by pollen morphology. The Amanoa type of Köhler (1965)
united Amanoa, Bridelia, Cleistanthus, Pentabrachion, and Pseudolachnostylis and that of Punt (1962)
included Amanoa, Pentabrachion, and Pseudolachnostylis. Gaucher (1902)
noted the similarity of Pseudolachnostylis and Lachnostylis with regards to vegetative anatomy, calling the former a xeromorphic version of the latter. Many Cleistanthus spp., Bridelia, and monotypic Pentabrachion share conspicuous parallel (percurrent or scalariform) tertiary leaf venation, whereas other genera of this subclade have mostly reticulate tertiary venation. Vestured pits have been reported from Bridelia and Cleistanthus but not other Euphorbiaceae s.l. (reviewed by Jansen et al., 2001
).
The main distinction previously drawn between Bridelia and Cleistanthus lies in fruit morphology. Bridelia has one- to two-locular drupes or rarely capsules, whereas Cleistanthus has three-locular dehiscent capsules (Jablonszky, 1915
; Webster, 1994b
). We found this to be oversimplified as it ignores the trend towards drupaceous fruits in Cleistanthus section Chartacei, e.g., C. megacarpus C. B. Rob. with tardily dehiscent to nearly indehiscent fruits (P. Hoffmann, personal observation). The two Cleistanthus species sampled represent sections Chartacei Jabl. [C. oblongifolius (Roxb.) Müll. Arg.] and Cleistanthus (C. perrieri Leandri). Members of the latter section seem to have exclusively explosive schizocarps. Given the paraphyly of Cleistanthus in our phylogenetic study, more sampling and a careful study of reproductive characters such as sepal aestivation, locule number, and fruit type, as well as leaf venation (Levin, 1986a
), are needed to reevaluate generic boundaries.
Subclade F3
Actephila, Andrachne, Leptopus, Meineckia, Poranthera, and Zimmermannia form a strongly supported clade (BP 99) in this analysis, mixing two tribes and four subtribes. Poranthera was classified in series "Stenolobeae" with other unrelated ericoid Euphorbiacaeae s.l. by Müller (1866)
