| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Article |
Institut of Systematic Botany, University of Zürich, Zollikerstrasse 107, CH-8008 Zürich, Switzerland
Received for publication July 24, 2006. Accepted for publication November 13, 2006.
ABSTRACT
Recent molecular and morphological systematic investigations revealed that the cacti are most closely related to Anacampseroteae, Portulaca and Talinum of the family Portulacaceae (ACPT clade of suborder Portulacineae). A combined analysis of ndhF, matK, and nad1 sequence data from the chloroplast and the mitochondrial genomes indicates that the tribe Anacampseroteae is the sister group of the family Cactaceae. This clade, together with Portulaca, is well characterized by the presence of axillary hairs or scales. Relationships within Anacampseroteae are characterized by a grade of five species of Grahamia s.l. from North and South America, and Grahamia australiana is found to be sister to the genera Anacampseros and Avonia. A comparison of vegetative characteristics indicates an evolutionary transition from woody subshrubs to dwarf perennial and highly succulent herbs during the diversification of Anacampseroteae. Available evidence from the present investigation as well as from previously published studies suggests that a revised classification of Portulacineae on the basis of inferred phylogenetic relationships might consist of a superfamily that includes Cactaceae and the three genera Anacampseros s.l. (including Avonia and Grahamia s.l.), Portulaca, and Talinum (including Talinella), either referred to three monogeneric families or to a paraphyletic family Portulacaceae*.
Key Words: Anacampseroteae • Cactaceae • family classification • Grahamia • molecular phylogenetics • Portulacaceae
Recent molecular systematic studies using plastid DNA data indicate that the traditional families Basellaceae, Cactaceae, Didieraceae, and Portulacaceae form a distinct and well-supported clade within Caryophyllales (Manhart and Rettig, 1994
; Applequist and Wallace, 2001
; Cuénoud et al., 2002
). These four families were first grouped together by Thorne (1976)
as suborder Portulacineae; more recently the terms portulacaceous alliance (Hershkovitz, 1993
), portulacaceous cohort (Applequist and Wallace, 2001
), and higher core Caryophyllales II (Hilu et al., 2003
) were used to refer to this same clade. Evidence is available that Halophytaceae and Hectorellaceae are also part of Portulacineae (Bittrich, 1993
; Philipson, 1993
; Cuénoud, 2002
; Müller and Borsch, 2005
).
Overall, the suborder Portulacineae comprises some 2000 to 2200 species in 160 genera, of which about 80% are Cactaceae (estimates based on Kubitzki et al., 1993
). Except for a few taxa that have a cosmopolitan distribution [e.g., Basella alba L., Portulaca oleracea L., Talinum paniculatum (Jacq.) Gaertn.], most species of Portulacineae inhabit dry areas of North and South America, Africa, and, to a lesser extent, Australia. Common characteristics of this group of families are (1) the presence of a fleshy or succulent tissue in either stems, leaves, or underground parts, (2) Crassulacean acid metabolism (CAM) photosynthesis, (3) normal secondary growth (i.e., absence of internal phloem), (4) mucilage idioblasts in stems and leaves, and (5) calcium oxalate crystals in the stem epidermis (Stevens, 2001
).
Morphological and molecular studies that included several genera and species of Portulacaceae (Hershkovitz, 1993
; Hershkovitz and Zimmer, 1997
; Applequist and Wallace, 2001
; Cuénoud et al., 2002
) indicate that this family is paraphyletic, with Basellaceae, Cactaceae, and Didieraceae nested in it. Furthermore, evidence is accumulating that the suborder Portulacineae consists of three major subclades: (1) ACPT clade (from Anacampseroteae G. D. Rowley, Cactaceae, Portulaca L., Talinum Adans.), (2) CDP clade (from Ceraria H. Pearson & Stephens, Didieraceae, Portulacaria Jacq.), and (3) PAW clade (from Phemeranthus Raf., Australian Calandrinias, western American Portulacaceae; Hershkovitz, 1993
; Hershkovitz and Zimmer, 2000
). The relationships among these three major groups and between Basellaceae, Halophytaceae, and Hectorellaceae are not yet resolved.
The ACPT clade is moderately supported by published molecular analyses (Hershkovitz and Zimmer, 1997
; Applequist and Wallace, 2001
) and consists of four distinct subgroups: Anacampseroteae (Anacampseros group of Gerbaulet [1992]
), Cactaceae, Portulaca, and Talinum s.str. (excluding Phemeranthus; Hershkovitz, 1993
; Hershkovitz and Zimmer, 1997
, 2000
).
The tribe Anacampseroteae was originally proposed by Nyananyo (1990)
and later validated by Rowley (1994)
; it comprises the genera Anacampseros L., Avonia G. D. Rowley, and Grahamia Gill. ex Hook. s.l. (including Talinaria Brandegee, Talinopsis A. Gray, and Xenia Gerbaulet; Rowley, 1994
, 2002
). These closely related genera of the family Portulacaceae were first recognized as a distinct group by Carolin (1987)
on the basis of inflorescence, pollen, fruit, and seed characters. Grahamia s.l. is a small but diverse genus of six species with a disjunct distribution in America and Australia: Grahamia bracteata, G. kurtzii, and G. vulcanensis occur in central and northern Argentina (G. vulcanensis extends into adjacent Bolivia), G. coahuilensis occurs in Central Mexico, G. frutescens grows in northern Mexico and the southern United States, and G. australiana occurs in central and southern Australia. The two species of Grahamia subgen. Grahamia (i.e., G. bracteata, G. frutescens) are scrambling, sparsely branched subshrubs with stiff branches, and the terete leaves are separated by distinct internodes. In contrast, the four species of Grahamia subgen. Talinaria (Brandegee) G. D. Rowley (i.e., G. australiana, G. coahuilensis, G. kurtzii, and G. vulcanensis) are perennating, succulent herbs with weak, fleshy branches and distinctly succulent leaves that are aggregated near the tips of the branches. Previously, Gerbaulet (1992)
included G. australiana in the genus Anacampseros and recognized four monotypic genera for the remaining species (G. vulcanensis was not treated). Monotypic genera, however, are of limited scientific value because they do not carry information about species relationships and are to be avoided if not required to establish monophyly (e.g., Entwisle and Weston, 2005
). Therefore, the generic classification system of Rowley (1994
, 2002
) is used here for the tribe Anacampseroteae. In this classification system, the genus Anacampseros is more narrowly circumscribed and comprises about 15 species that grow in South Africa, Namibia, and Somalia. The plants of this genus are dwarf succulents with leaves arranged in dense rosettes and "stipules" consisting of axillary hairs. Avonia, with 12 species from southern, eastern, and northeastern Africa, is generally included either as a section or subgenus in a more broadly circumscribed genus Anacampseros. These species are well characterized by their papery, scale-like structures that completely cover the minute leaves.
The family Cactaceae is very well characterized by several distinctive synapomorphies (Anderson, 2001
; Nyffeler, 2002
). It includes about 1600 to 1850 species, which almost exclusively occur in North and South America. More than 98% of the species are leafless stem-succulents (except for minute, deciduous leaves in Opuntioideae and a very few taxa of Cactoideae), and only Maihuenia Phil., Pereskia Mill., Pereskiopsis Britton & Rose, and Quiabentia Britton & Rose include species with persistent, terete or flat leaves that are fleshy or distinctly succulent (Barthlott and Hunt, 1993
).
Portulaca is a very distinctive genus based on its circumscissile fruit capsules, but the number of species to be recognized is very controversial, ranging from 40 (Geesink, 1969
) to more than 100 (Legrand, 1953
). Some species have a worldwide distribution, but diversity is concentrated in the tropical and subtropical areas of North and South America. Portulaca, as circumscribed here, also includes species that have previously been referred to the segregate genus Sedopsis (Engler) Exell & Mendonça.
The circumscription of Talinum has recently been clarified by morphological and molecular studies (Carolin, 1987
; Hershkovitz, 1993
; Hershkovitz and Zimmer, 1997
). These investigations revealed that the herbaceous taxa with linear-terete leaves should be placed in a separate genus, Phemeranthus, with its closest relatives in the PAW clade (Hershkovitz and Zimmer, 2000
). Talinum s. str. includes about 10 species and is predominantly distributed in America and Africa (Eggli, 2002
). The plants are low shrubs, often with tuberous underground parts and occasionally with annually deciduous branches. The leaves are flat, alternate, and rather herbaceous. Furthermore, they often have paired, scarious emergences in the axils, which have a nonlateral position that is not stipular in nature but rather may represent prophylls (M. Ogburn, University of Missouri, personal communication). A unique characteristic of this genus is the papillate fruit epidermis (Carolin, 1987
, 1993
). Recent molecular studies (Hershkovitz and Zimmer, 1997
; Applequist and Wallace, 2001
) provided strong evidence that the genus Talinella Baill., which includes 12 species endemic to Madagascar (Applequist, 2005
), is very closely related to Talinum. Talinella has previously been treated as having uncertain relationships (Franz, 1908
; Nyananyo, 1986a
; Carolin, 1993
; Eggli, 1997
) because of its septate ovary and fleshy, mucilaginous fruits otherwise unknown in Portulacaceae.
In this study, I focus on the ACPT clade, trying (1) to resolve the relationships among the four major subclades of this monophyletic group to clarify the sister-group relationships of Cactaceae and (2) to investigate the relationships within the tribe Anacampseroteae. Previously published hypotheses regarding these two questions based on analyses of morphological and molecular data (Carolin, 1987
; Gerbaulet, 1992
; Hershkovitz, 1993
; Hershkovitz and Zimmer, 1997
; Applequist and Wallace, 2001
; see Figs. 1, 2) are evaluated in the context of the newly generated sequences of this study. Furthermore, potential morphological synapomorphies for the different clades are discussed and evaluated, and the different options for adjusting the familial and generic classification of the taxa of Portulacineae to the inferred phylogenetic relationships are outlined. Detailed and well-supported hypotheses about relationships in the present study group are critical for a better understanding of the evolution of succulence in leaves, stems, and underground parts within the ACPT clade.
|
|
; Olmstead and Sweere, 1994
; Baum et al., 2004
) and sequences of the intron 2 of the NADH dehydrogenase subunit 1 gene (nad1 b/c) from the mitochondrial genome (Demesure et al., 1995
; Bakker et al., 2004
) to investigate these questions. The two chloroplast markers have been used before to clarify relationships within Caryophyllales (Applequist and Wallace, 2001
; Cuénoud et al., 2002
; Nyffeler, 2002
; Müller and Borsch, 2005
). The ndhF marker is reported to have less variation than the matK marker; therefore, a combination of the two sequence parts should provide increased resolution on a broader taxonomic level (Soltis and Soltis, 1998
). Parts of the nad1 marker are used here for the first time for a systematic study of a subgroup of Caryophyllales. MATERIALS AND METHODS
Taxon sampling
In total, 36 exemplars representing the four subclades of ACPT and outgroups from within and outside of Portulacineae were included in the present analyses (Appendix). Within Anacampseroteae, all recognized species of Grahamia s.l., and seven species of Anacampseros and Avonia were sampled. Four representatives of Cactaceae, three representatives of Portulaca from three of the four infrageneric taxa recognized by Geesink (1969)
, and six representatives of Talinum s.str. and Talinella were added to the sampling as representatives of the ACPT clade. Altogether, 28 ingroup taxa were included in the present study. Outgroup taxa from the CDP and the PAW clades, Basellaceae, and the sister group of Portulacineae (i.e., Aizoaceae, Nyctaginaceae, Phytolaccaceae) were also included.
I generated as new sequences for this project 16 ndhF sequences, 21 matK sequences, and 28 nad1 sequences. The other sequences were downloaded from GenBank (http://www.ncbi.nlm.nih.gov) as published in previous investigations (ndhF: Applequist and Wallace, 2001
; Olmstead et al., 2000
; matK: Cuénoud et al., 2002
; Nyffeler, 2002
). For three terminals, sequences of two or three different species had to be pooled (Phytolacca americana for ndhF, P. dioica for matK; Lewisia cantelovii for matK, L. cotyledon for nad1, L. pygmaea for ndhF; P. eruca for matK and nad1, P. grandiflora for ndhF) because of the unavailability of sequences of the different markers for the same species.
DNA extraction, amplification, and sequencing
Details on DNA extraction, PCR protocol, and temperature profiles for amplification (i.e., initial denaturation at 94°C for 4 min, 34 cycles of 94°C for 30 s, 48°C for 60 s, and 72°C for 90 s), and cleaning of PCR products are given in Nyffeler (2002)
. The ndhF marker was amplified in two reactions using primers –52F/8B and 5C/2210R of Olmstead et al. (1993)
and Applequist and Wallace (2001)
. The nad1 b/c region was amplified as in Demesure et al. (1995)
. Cleaned double-stranded PCR products were directly sequenced using the Big Dye Terminator Cycle Sequencing Reaction Kit (Perkin-Elmer, Applied Biosystems, Rotkreuz, Switzerland). For matK, I used the amplification primers and internal primers as reported in Nyffeler (2002)
. For ndhF and nad1, six newly designed internal primers (Table 1) were used in addition to the external primers of the PCR reactions. The reactions were cleaned with Microspin G-50 (Amersham Pharmacia Biotech Europe, Dübendorf, Switzerland) using multiscreen plates to remove excess Big Dye Terminator before loading on the automated sequencer ABI PRISM 3100 Genetic Analyzer (Perkin-Elmer).
|
; GenBank GBAN-X04826), Beta vulgaris subsp. maritima (L.) Tell. for nad1 (L. Herman, Centre of Agricultural Research, Melle, Belgium, unpublished data; GenBank AJ428873), and Spinacia oleracea L. for ndhF (Schmitz-Linneweber et al., 2001
, GenBank AJ400848). All sequences were submitted to GenBank (accession nos. DQ855843-DQ855863 for matK, DQ855864-DQ855879 for ndhF, DQ855880-DQ855907 for nad1), and the combined data matrix is available from TreeBase (http://www.treebase.org).
After manual alignment of all three partitions, the matK and ndhF sequences were combined into a single chloroplast matrix (cpDNA data set). I used the partition-homogeneity test (incongruence length difference test; Farris et al., 1994
) as implemented in PAUP* version 4.0b10 (Swofford, 2002
) to check for incongruence between the two chloroplast markers (10 000 replicates with standard parameters). The nad1 partition from the mitochondrial genome (mtDNA data set) contained several informative indels, which were separately coded in binary form for the parsimony analysis.
Initially, I conducted separate parsimony and Bayesian phylogenetic analyses on the cpDNA and mtDNA data sets. The topologies were then compared to check for the presence of well-supported incongruences. In a second step, a partition-homogeneity test was run to assess the amount of conflict between the two different data sets for an indication of whether a combined analysis was feasable.
The maximum parsimony (MP) analyses were perfomed in PAUP* 4.0b10 (Swofford, 2002
) with 10 000 random taxon addition replicates (holding one tree at each step) and tree-bisection-reconnection (TBR) branch swapping. All characters (i.e., sequence data and coded indels) were equally weighted. To estimate clade support, I conducted 10 000 bootstrap replicates with simple taxon addition and TBR branch swapping (the parameter MAXTREE was kept at 1000).
Bayesian phylogenetic analyses were conducted on all three sequence data sets using the software MrBayes version 3.1 (Huelsenbeck and Ronquist, 2001
). In the case of the cpDNA data set, I defined six different partitions (i.e., first, second, and third coding positions of the two different genes) and allowed model parameters to vary independently. The nad1 intron sequence data (excluding coded gaps) was treated as a single partition, and for the combined cpDNA + mtDNA data set, seven different partitions were consequently recognized. I used MrModeltest version 2.2 (Nylander, 2004
) to determine the best available model of molecular evolution on the basis of hierarchical likelihood ratio tests with the help of the Akaike information criterion (Akaike, 1974
) Then, I conducted three independent Markov chain Monte Carlo (MCMC) analyses, each composed of four linked chains with a sequential heat of 0.2 that were run for 1 million generations (sampling every 100 generations). The burn-in period was set at a very conservative estimate of 100 000 generations based on findings from an earlier preliminary analysis. The topologies of the three majority rule consensus trees and the clade posterior probabilities were compared to check for good mixing during the MCMC analyses. In the case of a positive finding, the post-burn-in trees of the three independent runs were combined into a single majority rule consensus phylogram derived from 27 003 individual topologies with the help of the SUMT command. Furthermore, the SUMT command provided estimates of the branch lengths as the mean across the subset of trees having the branch in question.
RESULTS
Sequence characteristics
The aligned matK matrix has a length of 1546 bp and comprises 36 exemplars. All newly generated matK sequences consist of the entire coding part, except for Grahamia kurtzii and Portulaca bicolor. As a result of problems with a polyA-rich region near the 5'-end, the first 370 bp for these two exemplars are lacking. The 15 sequences downloaded from GenBank are lacking some 430 to 461 bp at the 5'-end and some 227 to 259 bp at the 3'-end. The alignment required the insertion of seven gaps, of which five are uninformative, one is informative for outgroup–ingroup relationships, and one concerns the polyA-rich region.
The sequencing of the ndhF marker of exemplars of Avonia and Grahamia subgen. Talinaria repeatedly failed, whereas those for Anacampseros contained a deletion in the aligned matrix of 710 bp (Applequist and Wallace, 2001
). Therefore, the aligned ndhF matrix comprises only 29 exemplars. It has a length of 2142 bp, but lacks 102 bp at the 5'-end and 50 bp at the 3'-end in comparision to the published total gene sequence of Spinacia oleracea (GenBank AJ400848). Furthermore, eight sequences are lacking an additional 1–23 bp at the 5'-end and 9–30 bp at the 3'-end. The alignment required the insertion of 12 gaps: five are uninformative, five concern relationships among outgroup exemplars, one falls into a polyT-rich region, and one supports a sister-group relationship between Talinum caffrum and T. polygaloides.
The partition–homogeneity test between the matK partition and the ndhF partition yielded a P value of 0.12 and hence failed to reject the null hypothesis of congruent data at a P value of 0.05. Therefore, the two partitions were treated as a single cpDNA data set.
The present nad1 sequences correspond to the total length of the intron 2 of the NADH dehydrogenase subunit 1 (nad1) gene as compared to Beta vulgaris subsp. maritima (GenBank AJ428873). The aligned nad1 matrix comprises 26 ingroup exemplars and two outgroups from the genera Cistanthe and Lewisia. More distantly related taxa proved to be difficult to align because of unique sequence motifs, and, therefore, they were not considered for the analysis. The nad1 matrix has a length of 1867 bp. However, the Grahamia frutescens sequence lacks 171 bp at the 5'-end as a result of sequencing problems. All 17 informative gaps in the nad1 matrix were coded in binary form and added as additional phylogenetic information to the mtDNA data set.
The partition-homogeneity test between the cpDNA and the mtDNA data sets yielded a P value of 0.043, indicating that the two organellar DNA data sets contain a significant proportion of incongruent phylogenetic information. Therefore, the topology and clade-support values derived from the combined analysis were carefully evaluated in comparison to those from the analyses of the two separate partitions (described later).
Phylogenetic analyses
The results from the three different parsimony analyses are summarized in Table 2, and the strict consensus trees, together with bootstrap support values, are shown in Figs. 3A, 4A, and 5A.
|
|
|
), including a parameter to allow for among-site variation following a discrete approximation to a gamma rate distribution and a parameter to estimate the proportion of invariant sites (GTR+G+I), is preferred for all three different data sets. All Bayesian analyses (i.e., cpDNA, mtDNA, combined data) yielded results that showed no differences in the topologies of the post-burn-in majority-rule consensus trees derived from the three independent runs of each analysis. Furthermore, only minimal variation (<2%) in clade posteriors was found among the different runs. This indicates good mixing and justifies combining the three individual pools into a single posterior distribution for each of the three analyses. The majority-rule consensus trees are shown in Figs. 3B, 4B, and 5B, each with clade posteriors for all distinct clades. By examining the topologies derived from the cpDNA (Fig. 3A, B) and the mtDNA (Fig. 4A, B) data sets, some distinct discordances in the inferred relationships become immediately obvious. However, when comparing only well-supported clades with bootstrap support values above 75% for parsimony analyses and posterior probabilities of 95% and more for Bayesian analyses, conflict is slight. In particular, the data from the nad1 intron provides reasonable statistical support for a very limited number of clades only. This observation might explain the marginally significant result of the partition-homogeneity test between the distinctly incongruent cpDNA and mtDNA data sets.
A further characteristic of the two individual data sets is the topological incongruence between the parsimony and the Bayesian analyses. In contrast, the two consensus topologies derived from the combined data analyses (Fig. 5A, B) are congruent, have the lowest number of polytomies and show almost always higher support values in comparison to the individual analyses. Furthermore, all groupings resulting from the combined analyses are congruent with those either present in the cpDNA or in the mtDNA analyses. This is taken as an indication that the "total evidence analysis" is the best estimate of the available phylogenetic signal (de Queiroz et al., 1995
).
|
The cpDNA and the combined data support the monophyly of the four exemplars of Anacampseros and the three exemplars of Avonia. These two clades are sister groups based on the combined data analyses and form a polytomy with Grahamia australiana based on the cpDNA data analyses. The remaining five species of Grahamia form a grade: G. frutescens takes the topologically most basal position, followed by G. bracteata, G. vulcanensis, G. kurtzii, and G. coahuilensis.
The mtDNA data supports partly different relationships. In particular, the paraphyly of Portulaca is remarkable: in the parsimony analysis, the representatives of Portulaca subgen. Portulaca (i.e., P. eruca and P. oleracea) are sister to Anacampseroteae + Cactaceae, but P. bicolor of subgen. Portulacella (F. Muell.) Legrand, which includes the taxa previously referred to the segregate genus Sedopsis, is sister to Cactaceae. In contrast, the Bayesian analysis resolves P. eruca and P. oleracea as a grade basal to Cactaceae and P. bicolor as sister to Anacampseroteae. Furthermore, Talinum forms an unresolved grade at the base of the ACPT clade. In addition, Grahamia forms a grade, with Anacampseros and Avonia together forming a clade at its tip. The differences in the consensus topologies of the parsimony and the Bayesian analyses (Fig. 4) stem from the fact that the former also includes the data of the coded gaps, which support, in part, quite different relationships among the species of Grahamia (data not shown). The parsimony analysis identifies a clade of Grahamia bracteata, G. frutescens, and G. vulcanensis, although only with marginally 27% bootstrap support. In contrast, in the Bayesian analysis, these three exemplars form a grade, with G. frutescens taking the topologically most basal position, followed by G. vulcanensis and G. bracteata. In the parsimony analysis, G. kurtzii is sister to a clade consisting of two subclades of G. australiana and G. coahuilensis and of Anacampseros and Avonia. In the Bayesian analysis, relationships are similar, except that G. bracteata forms a polytomy together with G. kurtzii and its sister group as identified by the parsimony analysis. Furthermore, Anacampseros s. str. is not monophyletic based on the mtDNA data, because Anacampseros telephiastrum is identfied as being sister to Avonia, with moderate statistical support.
DISCUSSION
Hypotheses on phylogenetic relationships in the ACPT clade and Portulacineae
The present study provides strong support for the ACPT clade and for its constituent subclades Anacampseroteae, Cactaceae, Portulaca, and Talinum + Talinella. Furthermore, this investigation presents us with a hypothesis for the topological interrelationships among these four subclades, although with only moderate statistical support. Relationships among Anacampseroteae, Cactaceae, Portulaca, and Talinum seem to resist straightforward enlightenment (e.g., Hershkovitz and Zimmer, 1997
; Applequist and Wallace, 2001
). Hence, the question about the sister-group relationships of Cactaceae can be reformulated as a four-taxon problem; there are 15 different bifurcating, rooted trees that describe the relationships among four terminal taxa (Felsenstein, 1978
). The present hypothesis is similar to the one offered by Hershkovitz (1993; see Fig. 1B) based on morphological data, and it disagrees concerning the sister-group relationships of Cactaceae with those provided by Hershkovitz and Zimmer (1997)
and Applequist and Wallace (2001)
(Fig. 1C and 1D, respectively). To make things more complicated, phytochrome C (phyC) sequence data from the nuclear genome provided evidence for a sister-group relationship of Cactaceae with Portulaca (Edwards et al., 2005
), but on the basis of a very limited taxon sampling of the ACPT clade other than Cactaceae.
The present analysis is based on information from sequences of the chloroplast and the mitochondrial genome; thereby the former contributes more than four times the number of informative characters compared to the latter (the cpDNA makers have about twice the length and about two times the substitution rate of the mtDNA marker). The mtDNA data, in contrast to the cpDNA sequences, do not support the monophyly of Portulaca and Talinum + Talinella: both genera are paraphyletic, although with moderate to very low statistical support. The mtDNA data are marked by partly extensive insertions and deletions of sequence motifs: a lack of informative sites for Talinum and Talinella (see branch lengths derived from the Bayesian analysis; Fig. 4B) as a result of deletions and potentially nonhomologous insertions, which are partly difficult to align, in the different accessions of Portulaca probably account for these unexpected findings. The mtDNA data set, despite its obvious limitations, contributes some phylogenetic signal to a combined organellar sequence matrix to further resolve relationships among the major clades in the ACPT clade as well as in the tribe Anacampseroteae—without disrupting the consistently strong signal from the cpDNA data set.
Characteristics of fruits (e.g., differentiation of the pericarp into two layers, caducous exocarp, and fruit epidermis with straight anticlinal cell walls), seeds (e.g., presence of a thin, paper-like structure formed by the outer testa, presence of a cutin layer, embryo shape), and chromosome base number may all be indicators of a close relationship among taxa such as Anacampseros, Avonia, Grahamia, Portulaca, and Talinum of the family Portulacaceae (Carolin, 1987
; Gerbaulet, 1992
; Hershkovitz, 1993
). However, so far, for many of these characters, no detailed comparative studies are available that also consider Cactaceae and Talinella as well as potential outgroups from the remaining Portulacineae.
The fruits of Anacampseroteae, Portulaca, and Talinum are capsular, either circumscissile or dehiscing by terminal valves, whereas those of Cactaceae are fleshy or more rarely dry and those of Talinella are juicy-mucilaginous and indehiscent (Barthlott and Hunt, 1993
; Carolin, 1993
; Eggli, 1997
). The unique floral architecture of Cactaceae, consisting of an inferior ovary enclosed by stem tissue with numerous nodes ("pericarpel"), however, makes the interpretation of fruit characters concerning homologous structures in the ACPT clade difficult.
The presence of axillary hairs, or large axillary scales in the case of Avonia, is unique to Anacampseroteae, Cactaceae, and most species of Portulaca. None of these distinctive structures are found in Talinum, Talinella, or other Portulacineae and are therefore identified here as a distinct synapomophy for the ACP subclade. In Cactaceae (i.e., in Pereskia, Pereskiopsis, and Quiabentia), the hairs are mostly uniseriate, whereas those of Anacampseroteae and Portulaca are reported to be bi- or oligoseriate (Stevens, 2001
). The nature of these structures have been interpreted differently in the past, but most authors refer to them as "stipules" (Schumann, 1899
; Pax and Hoffmann, 1934
; Rutishauser, 1981
).
Gerbaulet (1992)
argued for a sister-group relationship of Anacampseroteae with Portulaca based on the absence of fibers in the stem cortex and a base chromosome number of x = 8 or x = 9, whereas Talinum and other representatives of Portulacaceae are known to possess this distinctive stem anatomical feature and have a base chromosome number of x = 12. However, Turner (1994)
provided partly different base chromosome numbers for the taxa of this clade (Anacampseroteae: x = 9; Cactaceae: x = 11; Portulaca: x = 10, rarely x = 8 or x = 9; Talinum: x = 8) that question whether base chromosome number is an unambiguous synapomorphy for Anacampseroteae + Portulaca. There is even less evidence from structural data for a possible sister-group relationship between Portulaca and Cactaceae as suggested by the phyC sequence data used for investigations by Edwards et al. (2005)
. The evaluation of evidence from floral architecture and inflorescence structure fails because of the unavailability of detailed comparative information about these taxa. Thus, the third possibility of interrelationship between the three taxa of the ACP clade [((A+C)P), instead of ((A+P)C) or ((C+P)A) as discussed previously] is favored by the molecular data of the present study: a sister-group relationship of Anacampseroteae with Cactaceae. Micromorphological characters of the seed testa and embryo shape hold the potential to provide supporting evidence for a close relationship between these two taxa (Carolin, 1987
; Gerbaulet, 1992
; Hershkovitz, 1993
).
The focus of the present study is on the ACPT clade, and the other taxa of Portulacineae were primarily investigated as outgroups. Therefore, the sampling of the CDP clade, the PAW clade, and Basellaceae is very limited. Available evidence from other molecular systematic studies (Applequist and Wallace, 2000
, 2001
on CDP; Hershkovitz and Zimmer, 1997
, 2000
on PAW; Hershkovitz and Zimmer, 1997
on Basellaceae) indicate that these four clades represent the major groups of Portulacineae; their interrelationships, though, remain unclear to now. The present analysis provides some very limited evidence for a sister-group relationship of the ACPT clade with the CDP clade. These two clades, together with Basellaceae, represent the eastern American/African clade of Hershkovitz (1993)
, whereas the PAW clade is predominantly distributed in western America, and only Phemeranthus and Parakeelya are found in eastern America and Australia, respectively (Hershkovitz, 1998
; Hershkovitz and Zimmer, 2000
). This hypothesis of a sister-group relationship between the ACPT and CDP clades receives further support from stomata architecture: the species of these two clades mostly have parallelocytic stomata, whereas those of PAW and many other families of Caryophyllales have paracytic or some other type of stomata (Eggli, 1984
; Nyananyo, 1986b
; Kubitzki et al., 1993; M. Ogburn, University of Missouri, personal communication).
Relationships, diversification, and generic classification in Anacampseroteae
The combined data analyses (Fig. 5) also revealed relationships among the taxa of Anacampseroteae. In contrast to some of the hypotheses outlined by previous authors (Fig. 2; Carolin, 1987
; Gerbaulet, 1992
; Rowley, 1994
; Hershkovitz and Zimmer, 1997
), the current hypothesis favors a cladistically basal grade of Grahamia frutescens and G. bracteata, but not a clade of the two species that is sister to the other representatives of Anacampseroteae. Other than that, the current hypothesis is congruent with that offered by Gerbaulet (1992)
based on a morphological cladistic analysis. The topologically basal positions of G. frutescens and G. bracteata are also supported by evidence from characters related to habit (Table 3) and, in particular, by their lack of the caducous exocarp reported to be typical for the other species of Grahamia as well as for Anacampseros and Avonia (Gerbaulet, 1992
; Hershkovitz, 1993
).
|
and is known from a number of other taxa in different families (e.g., Cercidium/Parkinsonia: Haston et al., 2005
; Larrea: Lia et al., 2001
). Grahamia australiana, the sister group of the African Anacampseros and Avonia species, is the most recently emerging extant species of the "Grahamia grade" (Fig. 5), indicating that long-distance dispersal from America resulted in the diversification of the Anacampseroteae in southern and eastern Africa.
Recently, two different generic classification systems for Anacampseroteae have been proposed: Gerbaulet (1992)
recognized four monotypic genera for the species of the "Grahamia grade" (i.e., Grahamia australiana was placed in Anacampseros s.l. and G. kurtzii was not investiged in detail), whereas Rowley (1994
, 1995
, 2002
) placed them all in Grahamia s.l. to produce "greater uniformity" (Rowley, 1994
, p. 105) in the classification of Anacampseroteae. To avoid recognizing a distinctly paraphyletic Grahamia, either it has to be split into five (i.e., if G. australiana is included in Anacampseros as suggested by Gerbaulet [1992]
) or six monotypic genera or it needs to be combined into an expanded genus Anacampseros s.l. (including Avonia, Grahamia, Talinaria, Talinopsis, Xenia). Personally, I prefer the latter option to avoid recognizing a large number of monotypic genera for a taxonomic group counting overall about 30 to 35 species. This expanded genus Anacampseros is consistent with the tribe Anacampseroteae as circumscribed by Nyananyo (1990)
and Rowley (1994, 1995) and is characterized by the combination of (1) axillary hairs or scales, (2) capsular fruits dehiscing from the tip into 3 or 6 valves, and (3) a fruit wall consisting of a distinct exocarp and endocarp.
Family classification of Portulacineae
During the past two decades, evidence has accumulated that the family Portulacaceae as traditionally circumscribed is not monophyletic and that its subgroups are more closely related to either Basellaceae, Cactaceae, or Didiereaceae. Therefore, as further insight about the internal relationships within this alliance of four families is accumulating, it will have to be reclassified to better reflect phylogenetic history. Judd et al. (2002)
suggested that three families (in addition to Basellaceae, as traditionally circumscribed), corresponding to the ACPT clade, the CDP clade, and the PAW clade might ultimately be recognized. To date, no comprehensive systematic study of Portulacineae is available that is based on a thorough sample of the species diversity of this clade with adequate amounts of information from both molecular and structural data. However, based on recently published studies (Hershkovitz and Zimmer, 1997
, 2000
; Applequist and Wallace, 2001
, 2003
; Cuénoud et al., 2002
) and my present investigation, I propose here an outline for a revised familial classification and point out the potentially vexing problems arising from the close relationships between the well-characterized and well-established Cactaceae and Portulaca (i.e., the type genus of Portulacaceae).
The oldest available family name for the PAW clade (Hershkovitz and Zimmer, 1997
, 2000
), which comprises the herbaceous, rosulate taxa of Portulacaceae with a mainly western American distribution, the eastern American genus Phemeranthus, and the Australian genus Parakeelya (elsewhere erroneously referred to as Rumicastrum; Hershkovitz, 1998
), is Montiaceae (Hoogland and Reveal, 2005
). This family mainly includes taxa with rosettiform habits and clasping, nonconstricted leaf bases (Hershkovitz, 1993
). The circumscription of the family Didiereaceae has recently been revised based on molecular and structural evidence to include the three genera Calyptrotheca, Ceraria, and Portulacaria of Portulacaceae (Applequist and Wallace, 2000
, 2003
). The present investigation provides evidence that Talinum s.l. (including Talinella) is sister to a clade of Anacampseros s.l., Portulaca, and Cactaceae, the latter being well characterized by the presence of axillary hairs or scales. The relationships within this ACP clade, however, remain unclear because of contrasting hypotheses offered by other studies (Hershkovitz and Zimmer, 1997
; Applequist and Wallace, 2001
; Edwards et al., 2005
). Hence, we are left with the question of how to transform this insight about the relationships among the taxa of the ACPT clade into an updated classification system of flowering plants (APG, 2003
). Proponents of the three major views on the philosophy of contemporary classification and nomenclature (i.e., [1] monophyletic groups and PhyloCode [Cantino and de Queiroz, 2006
], [2] monophyletic groups and ICBN [Greuter et al., 2000
], and [3] monophyletic or paraphyletic groups and ICBN; Stevens, 2006
) will certainly offer different solutions; therefore, it does not come as a surprise that the present case has already made headlines (Brummitt, 2006
). The three most obvious variants are outlined in Table 4. There is currently no good evidence that a recircumscribed Portulacaceae, consisting only of Anacampseros s.l., Portulaca, and Talinum s.l., would form a monophyletic taxon if Cactaceae is recognized as a separate family; recognizing Portulacaceae in this revised circumscription would render this family paraphyletic. The nature of this taxon may be indicated with quotation marks (Judd et al., 2002
) or with an asterisk. The two family names Cactaceae and Portulacaceae were described by de Jussieu (1789)
in the same publication and at the same date, and both are conserved (Hoogland and Reveal, 2005
). Therfore, neither of them has priority over the other. Expanding Cactaceae to include the three most closely related genera of Portulacaceae (Table 4, variant 2) or recircumscribing Portulacaceae to include all genera of Cactaceae would lead to abandoning a very widely recognized and easily identified taxon at the family rank. Recognizing three monogeneric families in addition to Cactaceae for the ACPT clade would fulfill the requirement of recognizing only monophyletic families, but this solution does not carry any information on relationships but only satisfies formal classification. The option of expanding either Cactaceae or Portulacaceae to include all taxa of the suborder Portulacineae, and hence relegating the problem of reclassifying its diversity to the subfamily rank, is not further considered here. To offer a compromise that might be acceptable to a broad group of users, I suggest the introduction of a superfamily name for a taxon that includes Cactaceae and the three genera Anacampseros s.l., Portulaca, and Talinum s.l. The last three may either be referred to Portulacaceae* (Table 4, variant 1) or each of them placed in a separate family (Table 4, variant 3). Superfamilies are largely unknown in botanical classification systems (see Kress, 1990
), and different endings have been proposed for names at this rank. However, such a name for a taxonomic group not previously recognized carries the best promise for broad acceptance. A validating description for a name such as "Cactariae" for the ACPT clade will be provided in due time.
|
). Recent phylogenetic work has clarified the relationships among the major clades of cacti (Edwards et al., 2005
) and hence, has set the basis for developing a scenario of early cactus evolution through ancestral trait reconstruction (Edwards and Donoghue, 2006
). The basal and paraphyletic grades in Anacampseros s.l. and Pereskia (of Cactaceae) indicate that their common ancestors were slightly xeromorphic shrublets with fleshy leaves and swollen roots (see Giffith [2004] for an alternative view) from which arose, on the one hand, the diminutive leaf-succulent herbs of Anacampseros s.l. in southern Africa and, on the other hand, the huge diversity of stem-succulent cacti in America. Comparative studies on the morphology, physiology, and ecology will allow us in the near future to compile a detailed sketch of the evolutionary history of this prominent clade of succulent plants.
Available evidence suggests that a revised family classification of Portulacineae that reflects inferred phylogenetic relationships might consist of five major taxa: Basellaceae (as traditionally circumscribed), Didiereaceae (including Calyptrotheca, Ceraria, and Portulacaria [Applequist and Wallace, 2003
]), Halophytaceae (Cuénoud et al., 2002
; Müller and Borsch, 2005
), Montiaceae (corresponding to the PAW clade of Hershkovitz and Zimmer [2000
], including Hectorellaceae [Applequist et al., 2006
]), and Cactariae (a superfamily not yet formally described) to include Cactaceae, and the genera Anacampseros, Portulaca, and Talinum, which are either referred to their individual monogeneric families or are included in a paraphyletic family Portulacaceae*.
APPENDIX
Taxon; GenBank accession: ndhF, matK, nad1; Source or reference of original publication, Voucher specimen information.
Anacampseros karasmontana Dinter; DQ855872, DQ855859, DQ855900; Cult. ZSS 89 4014; D. Supthut 84155, Namibia. Anacampseros retusa Poelln.; DQ855873, DQ855860, DQ855901; Cult. ZSS 80 3518; Anonymous s.n., South Africa, Cape Prov. Anacampseros subnuda Poelln.; DQ855874, DQ855861, DQ855902; Cult. ZSS 90 1652; Anonymous s.n., South Africa, Cape Prov. Anacampseros telephiastrum DC.; DQ855875, DQ855862, DQ855903; Cult. ZSS 90 1523; J. Lavranos & Bleck s.n., South Africa, Cape Prov. Avonia albissima (Marloth) G. D. Rowley; —, DQ855856, DQ855897; Cult. ZSS 90 2269; J. Lavranos & Bleck s.n., South Africa, Cape Prov. Avonia papyracea (Fenzl) G. D. Rowley; —, DQ855857, DQ855898; Cult. ZSS 93 2223; J. Lavranos 29946, South Africa, Cape Prov. Avonia recurvata (Schönland) G. D. Rowley; —, DQ855858, DQ855899; ZSS T19507; Oliver 1151, South Africa, Cape Prov. Basella alba L.; AF194834 (Applequist & Wallace, 2001
), AY042553 (Cuénoud et al., 2002), —. Cistanthe grandiflora (Lindl.) Schltr.; AF194842 (Applequist & Wallace, 2001
), AY042568 (Cuénoud et al. 2002), DQ855881; Cult. ZSS 94 2055; D. Ford & Smith 359, Chile, Prov. Colchagua. Copiapoa bridgesii (Pfeiff.) Backeb.; DQ855879, AY015293 (Nyffeler, 2002
), DQ855907; ZSS 19863; K. Knize 1399, Chile, Prov. Chañaral. Decarya madagascariensis Choux; AF194844 (Applequist & Wallace, 2001
), AY042574 (Cuénoud et al. 2002), —. Delosperma cooperi L. Bolus; DQ855864, DQ855843, —; Cult. ZSS hort. org. Didierea trollii Capuron & Rauh; AF194845 (Applequist & Wallace, 2001
), AY042576 (Cuénoud et al. 2002), —. Grahamia australiana (J. M. Black) G. D. Rowley; —, DQ855855, DQ855896; P. I. Forster 1749, Australia, Queensland. Grahamia bracteata Gill. ex Hook.; AF194846 (Applequist & Wallace, 2001
), —, —. Grahamia bracteata Gill. ex Hook.; —, AY015273 (Nyffeler, 2002
), DQ855892; Cult. ZSS 94 1326; B. E. Leuenberger & U. Eggli 4184, Argentina. Grahamia coahuilensis (S. Watson) G. D. Rowley; —, DQ855854, DQ855895; Cult. ZSS 90 1259; Glass & Forster 1934, Mexico, Nuevo Leon. Grahamia frutescens (A. Gray) G. D. Rowley; DQ855871, DQ855851, DQ855891; Cult. ZSS 92 1739; D. J. Ferguson 1140, USA, New Mexico. Grahamia kurtzii (Bacigalupo) G. D. Rowley; —, DQ855853, DQ855894; Cult. ZSS 94 1057; B. E. Leuenberger & U. Eggli 4217, Argentina, La Rioja. Grahamia vulcanensis (Añón) G. D. Rowley; —, DQ855852, DQ855893; Cult. ZSS 90 4035; B. E. Leuenberger 3534, Argentina, Jujuy. Lewisia cantelovii J. T. Howell; —, AY042607 (Cuénoud et al., 2002), —. Lewisia cotyledon (S. Watson) B. L. Rob.; —, —, DQ855880; Cult. ZSS hort. org. Lewisia pygmaea (A. Gray) B. L. Rob.; AF194847 (Applequist & Wallace, 2001
), —, —. Maihuenia patagonica (Phil.) Britton & Rose; DQ855877, AY015281 (Nyffeler, 2002
), DQ855905; B 030-30-88–10; B. E. Leuenberger & S. Arroyo 3850, Argentina. Mirabilis jalapa L.; AF194826 (Applequist & Wallace, 2001
), AY042614 (Cuénoud et al., 2002), —. Montia parvifolia (DC.) Greene; AF194851 (Applequist & Wallace, 2001
), AY042616 (Cuénoud et al., 2002), —. Opuntia vestita Salm-Dyck; DQ855878, AY015278 (Nyffeler, 2002
), DQ855906; Cult. ZSS 85 1173; Swoboda 124, Bolivia. Pereskia aculeata Mill.; DQ855876, DQ855863, DQ855904; Cult. ZSS hort. org. Phytolacca americana L.; AF130229 (Olmstead et al., 2000
); —, —. Phytolacca dioida L.; —, AY042631 (Cuénoud et al. 2002), —. Portulacacf.bicolor F. Muell.; DQ855870, DQ855848, DQ855888; Cult. BRI; P. I. Forster 26799, Australia, Queensland. Portulaca eruca Hauman; —, DQ855849, DQ855889; Cult. ZSS 99 5251; B. O. Schlumpberger 66, Argentina, Córdoba. Portulaca grandiflora L.; AF194853 (Applequist & Wallace, 2001
), —, —. Portulaca oleracea L.; AY194867 (Applequist & Wallace, 2001
), —, —. Portulaca oleracea L.; —, DQ855850, DQ855890; Z ; R. Nyffeler s.n., Switzerland, Aarau. Portulacaria afra Jacq.; AF194857 (Applequist & Wallace, 2001
), AY042637 (Cuénoud et al., 2002), —. Talinella pachypoda Eggli; DQ855868, DQ855846, DQ855886; Cult. ZSS 95 2007; W. Röösli & Rechberger s.n., Madagascar, Antsiranana. Talinum caffrum (Thunb.) Eckl. & Zeyh.; AY194859 (Applequist & Wallace, 2001
), AY042662 (Cuénoud et al., 2002), —. Talinum caffrum (Thunb.) Eckl. & Zeyh.; —, —, DQ855882; Cult. B. 263-10-92-30; Anonymous s.n., South Africa, Transvaal. Talinum paniculatum (Jacq.) Gaertn.; DQ855866, AY015274 (Nyffeler, 2002
), DQ855884; Cult. ZSS 93 1952; S. Martinez & U. Eggli 203; Mexico, Puebla. Talinum polygaloides Arn.; DQ855867, DQ855845, DQ855885; Cult. ZSS 95 1725; D. J. Ferguson 263, Argentina, Salta. Talinum portulacifolium (Forssk.) Schweinf.; DQ855869, DQ855847, DQ855887; Cult. ZSS 90 2248; Anonymous s.n., South Africa. Talinum triangulare (Jacq.) Willd.; DQ855865, DQ855844, DQ855883; Cult. ZSS 95 1726; D. J. Ferguson 848.
|
1 The author thanks T. Bolliger and U. Eggli (Sukkulenten-Sammlung Zürich), P. Forster (Brisbane Botanical Garden), and B. E. Leuenberger (Botanischer Garten Berlin) for providing plant material and various information; A. Lendel and Ph. Reuge (University of Zürich) for help with lab work; the Linder lab (University of Zürich), P. Stevens (Missouri Botanical Garden, St. Louis), and two anonymous reviewers for helpful comments on earlier drafts of the manuscript. This work was supported by funding from the G. and A. Claraz-Schenkung of the University of Zürich. 
2 E-mail: rnyffeler{at}systbot.unizh.ch
LITERATURE CITED
Akaike H.. 1974. A new look at statistical model identification. IEEE Transactions on Automatic Control 19: 716-723.
