Haseltonia 14: 46–53. 2008 46 HASELTONIA 14, 2008 47

Grandifolia group Investigating and the earliest P. bahiensis 100 75 8 0.53 divergences in Cactaceae P. stenantha 69 0.97 88 P. grandifolia ssp. grandifolia Charles Butterworth 2 98 100 P. grandifolia ssp. violacea 1.00 Desert Botanical Garden, 1201 North Galvin Parkway, 16 Phoenix, Arizona 85008, USA; [email protected] P. sacharosa 67 98 98 0.92 4 P. nemorosa 1.00 Erika J Edwards Lychnidiflora group P. lychnidiflora Department of Ecology & Evolutionary Biology, Brown University, 77 2 100 Providence, Rhode Island 02912, USA; [email protected] 1.00 Zinniiflora group P. zinniiflora 99 97 4 P. portulacifolia 75 1.00 51 0.93 100 P. marcanoi 0.83 100 1.00 Abstract: The cacti are renowned desert of the New World. The are typically - 100 1.00 63 9 P. quisqueyana 100 1.00 less, spine-bearing stem succulents. of Pereskia possess broad, regular and have 95 1.00 Bleo group been viewed as being representative of ancestral cacti. A number of previous studies have at- 2 P. bleo 88 74 0.98 tempted to resolve phylogenetic relationships within Pereskia and between Pereskia and other Guamacho group P. aureiflora cacti. Here, we present the results of a joint analysis along with hypothesis testing using data- 74 78 97 sets previously published independently by each author. This study clearly shows a split 1 2 P. guamacho 1.00 85 1.00 in the cacti between a clade of Caribbean-basin-centered Pereskia species and all other cacti. Horrida group P. horrida ssp. rauhii Furthermore, hypothesis testing strongly supports the basal Pereskia hypothesis over an alter- 100 11 P. horrida ssp. horrida native hypothesis in which the form the basal split in the cacti (although the 100 latter hypothesis was not statistically rejected). 12 P. diaz-romeroana 100 100 77 1.00 0.98 Key words: 100 Pereskia, Cactaceae, cacti, , evolution 10 P. weberiana 0.98 100 11 1.00 Aculeata group P. aculeata 100 Opuntioideae 100 20 1.00 poeppigii Introduction morphies. The Pereskioideae has tradition- 100 Except for a single, epiphytic species, cacti are ally been comprised of two genera, Pereskia 1.00 naturally restricted to the New World, where and Maihuenia (Philippi ex F. A. C. Weber) Outgroup(s) they represent a spectacular arid land– Schumann, the latter being unusual due to radiation. In the majority of cacti, persistent its natural distribution in the extreme south ­Figure 1. Phylogenetic relationships in Pereskia and basal Cactaceae. The cladogram on the left (redrawn from leaves are replaced by spines, and the green suc- of South America and its stem and leaf fea- Butterworth and Wallace 2005) is the strict consensus of six most parsimonious for combined chloro- culent stem has become the primary photosyn- tures (for instance, distribution of mucilage- plast rpl16 intron and psbA-trnH intergenic spacer sequence data. Bootstrap support ≥50% are shown above thetic organ. A few genera do retain persistent bearing structures) unknown elsewhere in the branches, decay values are shown below the branches. The cladogram on the right (redrawn from Edwards leaves, however, and these have been treated as the family (Gibson 1977). The Opuntioideae and others 2005) is the Bayesian consensus for combined chloroplast trnK/matK, rbcL, and psbA-trnH IGS, the most ancestral-like members of the fam- Burnett includes prickly-pears, chollas, and mitochondrial cox3, and nuclear phyC sequence data. Maximum parsimony bootstrap support ≥50% is shown ily. Spines (developmentally modified leaves) other members that share distinct morpho- above branches, Bayesian posterior probabilities are shown below branches. Shaded boxes denote species groups defined by Leuenberger (1986). are typically borne from areoles (specialized logical and anatomical features, such as the axillary buds that may bear spines, , or presence of glochids (very short, fine, barbed ). structure is predominated by re- spines), and with a funicular envelope, A number of studies have attempted to genomes (chloroplast, mitochondrion, nucleus) ceptacular epigyny; however, in the leafy usually forming a bony, white aril (Gibson resolve subfamilial phylogenetic relation- for Pereskia and representatives from Opunti- Pereskia Miller there is variation in ovary po- and Nobel 1986). The Cactoideae are mor- ships in Cactaceae. Results of a cpDNA re- oideae, Maihuenioideae, and Cactoideae. The sition from superior to inferior ovaries. phologically variable, from tall, columnar spe- striction-site survey of members of Pereskia maternally inherited genomes (chloroplast and Traditionally most classifications cies to low-growing, globose cacti to epiphytic (Pereskioideae) were summarized by Wal- mitochondrion) did not yield strong support recognized three subfamilies (Barthlott and plants possessing flattened, almost leaf-like lace (1995). These data, supplemented by in their placements’ subfamilies. However, the Hunt 1993; Hunt and Taylor 1986, 1990). stems. There are no obvious morphological or chloroplast sequence data (Butterworth and nuclear component (phytochrome C) and the The Pereskioideae Schumann possess mor- anatomical synapomorphies uniting the Cac- Wallace 2005) suggest that Pereskioideae is maternal data set both yielded a monophy- phological features, such as persistent leaves toideae, except possibly stem succulence and not monophyletic for the following reasons: letic Opuntioideae, which formed the sister and superior to inferior ovaries, that are con- lack of persistent leaves and glochids, the lat- 1) sampled members of Cactoideae are em- group to a clade that included some Pereskia sidered primitive or relictual within the Cac- ter being a derived trait in the Opuntioideae. bedded within Pereskia, and 2) Maihuenia, species, Maihuenia, and Cactoideae. taceae (Boke 1964; Gibson and Nobel 1986). Furthermore, the chloroplast rpoC1 intron is Opuntioideae, and the Cactoideae form an Here we present a summary of the two pre- For these reasons, members of the Pereskioi- missing in all Cactoideae (Wallace and Cota unresolved trichotomy. vious molecular (DNA) phylogenetic studies of deae are considered to be united by shared 1996) with the exception of Blossfeldia (But- Edwards and others (2005) presented a phy- Pereskia (Butterworth and Wallace 2005; Ed- plesiomorphies rather than distinct synapo- terworth 2006). logeny based on sequence data from all three wards and others 2005). We then present, for 48 Butterworth and Edwards—Earliest Divergences in Cactaceae HASELTONIA 14, 2008 49

the first time, the analysis of combined DNA paraphyletic, although there were differ- phyC 1.00 1.00 Maternal sequence data used in those studies. ences regarding which other groups were 1.00 Pereskiopsis porteri 1.00 Butterworth CA and Wallace RS 2005. nested within the genus. Butterworth and Published in December, 2005, this study Wallace (2005) placed sampled members of 1.00 verticillata 1.00 0.75 added DNA sequence data from the chlo- the Opuntioideae and Maihuenioideae out- 1.00 Quiabentia zehntneri roplast rpl16 intron and psbA-trnH IGS to side the clade containing Pereskia. However, 1.00 Brasilopuntia brasilensis 1.00 the chloroplast restriction-site data of Wal- the basal polychotomy does not allow infer- 1.00 lace (1995). The strict consensus from max- ences regarding the “ancestral” lineage of the dilleni imum-parsimony analysis for the combined Cactaceae. Sampled species of the Cactoideae 1.00 subulata 0.79 data set (Fig 1) gave greater phylogenetic were placed, phylogenetically within Pereskia, articulatus P lych- resolution than that of Wallace (1995) and in a sister-group relationship with . 0.71 recovered several key Pereskia clades as out- nidiflora De Candolle. Edwards and oth- Echinocactus platyacanthus 0.77 lined by Leuenberger (1986), including the ers (2005) also demonstrated a paraphyletic 1.00 1.00 1.00 0.75 Cereus fernambucensis Horrida, Zinniiflora, Guamacho, and Gran- Pereskia, although their phylogeny indicated 1.00 1.00 diflora groups. This study also placed several the presence, within Pereskia, of a clade con- Calymmanthium substerile enigmatic taxa with good statistical support: taining members of the Cactoideae, Opun- 1.00 Blossfeldia liliputiana P. aculeata Miller was united with the Horrida tioideae, and Maihuenioideae. Furthermore, 1.00 Maihuenia patagonica group and P. bleo with the Zinniiflora group. according to Edwards and others, the fun- 1.00 Relationships among the various subfamilies damental split in the Cactaceae is between remained ambiguous, but importantly, there a lineage of northern Pereskia species and all Pereskia grandifolia var grandifolia 1.00 was no indication that Maihuenia should be other cacti termed the “Caulocacti” (in ref- 0.9 1.00 Pereskia stenantha 1.00 united with Pereskia, and there was moder- erence to the presence of stem stomata and 0.92 ately strong support for sampled Cactoideae delayed bark formation, two traits shared by Pereskia nemorosa 1.00 1.00 being nested within Pereskia. most members of this clade). The caulocacti Pereskia sacharosa 0.99 Edwards E, Nyffeler R, Donoghue MJ include the South American Pereskia species 2005. This study, published in August 2005, and the “core” cacti—Opuntioideae, Cactoi- 1.00 1.00 not only utilized chloroplast trnK/matK, psbA- deae, and Maihuenioideae. In general, statis- Pereskia horrida var horrida 1.00 trnH IGS, and rbcL, but also nuclear phyC tical support for this phylogenetic hypothe- 0.84 0.97 Pereskia marcanoi 1.00 and mitochondrial cox3 sequence data, that sis is the strongest of all previous studies of 0.91 is, DNA sequence data from all three DNA basal cactus phylogeny, but several nodes still Pereskia portulacifolia 0.96 1.00 0.98 1.00 compartments (chloroplast, mitochondrion, remain quite tentative. Furthermore, most of Pereskia quisqueyana 0.59 nucleus). The nuclear marker (phyC) pro- the support for this hypothesis comes strictly 1.00 Pereskia zinniiflora 1.00 vided the greatest statistical support to date from the phyC marker, as all sampled chlo- 1.00 for the deeper nodes within the phylogeny. roplast and mitochondrion markers have not 0.88 Pereskia bleo The Pereskia clades outlined by Leuenberger proved overly useful for resolving the basal 1.00 1.00 Pereskia aureiflora 1.00 (1986) and demonstrated by Butterworth and cactus problem. A logical next step in studies 1.00 Pereskia guamacho Wallace (2005) were also recovered in the Ed- of early cactus evolution will be to sequence 1.00 wards study, yet a number of differences in ar- additional gene regions, but we first thought 1.00 Pereskia lychnidiflora rangement of major clades (see Fig 1) can be it a useful exercise to perform a combined Portulaca oleracea 0.98 seen. While Pereskia species from northern analysis of data from both studies. Talinum paniculatum South America and the Caribbean (Guama- cho, Bleo, and Zinniiflora groups), and spe- Materials & Methods Grahamia bracteata cies from the region of South America Laboratory techniques (DNA extraction, PCR, Grahamia coahuilensis (Horrida and Aculeata groups) form single sequencing, etc) are detailed in Butterworth clades in both studies, topological differ- and Wallace (2005) and Edwards and oth- Anacampseros telephiastrum ences occur in phylogenetic placements of the ers (2005). DNA sequence data from both ­Figure 2. Bayesian consensus cladograms for the nuclear (phyC) and maternal datasets. Numbers above the Grandifolia and Lychnidiflora groups, which studies were combined using MacClade 4.08 branches are Bayesian posterior probabilities. The narrowly-dashed line for phyC represents the Maximum Like- form a clade with the Cactoideae in the But- (Maddison and Maddison 2005). Only taxa lihood topology. The widely-dashed line for the maternal dataset allows for direct comparison with the phyC terworth and Wallace phylogeny. Also worth for which data from all three genomes were cladogram without re-arranging taxon order. noting, both studies indicate a possible long- present were retained for analyses. The total distance dispersal event due to the phyloge- data matrix consisted of 7192 characters for 28 var2 (Nylander 2004) implemented in PAUP* distributed rates model (Yang 1993). Phylo- netic positioning of the Brazilian endemic taxa from the Cactaceae and five taxa from the (Swofford 2002). Selected models were F81 genetic reconstruction was undertaken using Pereskia aureiflora Ritter with P. guamacho Portulacaceae. The final data matrix had a total (Felsenstein 1981) for rpl16; HKY (Hasegawa MrBayes var 3.1.2 (Ronquist and Huelsenbeck Weber from . It is interesting that of 10.23% cells scored as missing data. These and others 1985) for rbcL and phyC; and GTR 2003). Incongruence Length Difference (ILD) both these species are unique within Pereskia cells included indels and regions from taxa for (Tavare 1986) for psbA-trnH, trnK/matK, and tests were implemented in PAUP* to test con- for their yellow-colored flowers, possibly re- which no sequence data was available. cox3. Additionally, all regions with the excep- gruence between the three genomes. With all lated to their major pollinators. Bayesian analyses. Models of sequence tion of cox3, which utilized the proportion of taxa included, the ILD test rejected the null In both studies Pereskia was shown to be evolution were estimated using MrModeltest invariable-sites model, also utilized a gamma- hypothesis of congruence with a p–value of 50 Butterworth and Edwards—Earliest Divergences in Cactaceae HASELTONIA 14, 2008 51

1.00 Pereskiopsis the MrBayes default prior parameters in Mar- P. sacharosa) in an unresolved polychotomy; and kov-chain Monte Carlo (MCMC) searches a final clade containing Maihuenia in a sister- 100 Pereskiopsis porteri 1.00 using four incrementally heated chains for group relationship to the Cactoideae. 100 1.00 Quiabentia verticillata 1,000,000 generations sampled every 100 Maternal data (chloroplast, mitochon- 0.57 100 Quiabentia zehntneri generations. Following the MCMC analyses, drion). With the exception of a single species posterior parameter distributions were viewed (P. lychnidiflora), the major early divergence 1.00 Brasilopuntia brasilensis using Tracer v1.4 (Rambaut and Drummond within the Cactaceae is between a clade con- 1.00 100 Opuntia dilleni 2007), which indicated that exclusion of the taining species of Pereskia and all other cacti 100 first 100 trees (10,000 generations) was more (Fig 2). Furthermore, P. lychnidiflora is not em- 1.00 Austrocylindropuntia subulata than adequate to ensure stationarity. bedded within the remaining cacti; rather, it ◊ 90 Hypothesis Testing. Different hypotheses forms the sister-group to the other Cactaceae. 1.00 Echinocactus platyacanthus of Cactaceae evolution were undertaken using Within the other cacti, there is an unresolved 0.59 constraint trees and Maximum Likelihood (ML) trichotomy involving Maihuenia, Opuntioi- 1.00 Cereus fernambucensis analyses using PAUP*. Suitable models of se- deae, and Cactoideae. 100 quence evolution were K81 + I + G (Kimura Combined data. The Bayesian consensus 1.00 Calymmanthium substerile 100 1981) for the complete dataset, TVM + I + G for cladogram (Fig 3) for the combined dataset is 1.00 Blossfeldia liliputiana the maternal dataset, and HKY + G (Hasegawa topologically highly congruent with the phyC 1.00 1.00 Maihuenia patagonica and others 1985) for nuclear dataset were esti- cladogram, differing only by being fully resolved. 100 Maihuenia poeppigii mated using ModelTest (Posada and Crandall Furthermore, all of the deep nodes within the 1998). These models were then used in maxi- cacti are supported with high Bayesian poste- 1.00 Pereskia grandifolia var grandifolia mum likelihood analyses on the unconstrained rior probabilities (that is, 99 or 100%). datasets. The same analyses and models were Hypothesis testing. With the exception 1.00 89 Pereskia stenantha then repeated using topological constraints for of the combined and nuclear datasets for the 100 1.00 Pereskia nemorosa the hypotheses being tested: 1. “basal” Pereskia; “basal” Maihuenia hypothesis, none of the hy- 0.99 99 Pereskia sacharosa 2. “basal” Opuntioideae; 3. “basal” Maihuenia. potheses of early cactus evolution were fully 62 The ML trees from each of these studies were rejected using the SH-Test (Table 1). A com- 1.00 Pereskia aculeata compared to their unconstrained counterparts parison of the p-values for “basal” Opuntioi- 1.00 100 Pereskia horrida var horrida using the Shimodaira-Hasegawa Test (Shimod- deae and “basal” Pereskia show that the latter 100 Pereskia marcanoi aira and Hasegawa 1999) as implemented in hypothesis has better statistical support than 0.74 PAUP* using full optimization and 1000 boot- the former. However, it is important to note ◊ 1.00 Pereskia portulacifolia strap replications. that both these hypotheses were not statisti- 87 cally rejected. 1.00 Pereskia zinniiflora Results Comparison between this and previous * 0.93 100 1.00 Pereskia quisqueyana The Bayesian consensus trees from all three da- studies. In essence, combination of the sequence 100 0.99 Pereskia bleo tasets (maternal, nuclear, combined) supported data from Butterworth and Wallace (2005) 75 Pereskia aureiflora monophyly of the Cactaceae as well as subfam- and Edwards and others (2005) is equivalent 1.00 ilies Opuntioideae and Cactoideae (Figs 2, 3). to adding chloroplast rpl16 data to the latter 1.00 1.00 96 Pereskia guamacho 95 Furthermore, all data analyses supported the dataset. In terms of topology, the combined Pereskia lychnidiflora hypothesis of Pereskia representing the most dataset in this paper (Fig 3) and Edwards and 1.00 ancestral group within the family. However, others (2005) are highly congruent. Portulaca oleracea there is incongruence between the maternally Discussion Talinum paniculatum inherited datasets and phyC, particularly with regard to the placement of the species from Given that the Bayesian consensus cladogram Grahamia bracteata and the Andes region of South America (Fig 3) in this study is topologically highly con- Grahamia coahuilensis (P. aculeata, P. horrida, P. grandifolia, P. ste- gruent with that in Edwards and others (2005) Anacampseros telephiastrum nantha, P. nemorosa, P. sacharosa). it is unnecessary to include a detailed discussion Nuclear Data. The earliest divergence within here. However, a number of issues pertinent to ­Figure 3. Bayesian consensus cladogram for the combined analysis. Bayesian posterior probabilities are shown the Cactaceae is between a clade containing early cactus evolution can be addressed based above the branches and maximum parsimony bootstrap above 50% are shown below the branches. Dashed line Pereskia species from the Caribbean (P. bleo, upon the results of the study presented here. represents branch collapsed in Maximum Likelihood cladogram. * represents change from superior to half infe- P. quisqueyana, P. zinniiflora, P. portulacifolia, “Basal” relationships in the Cactaceae. The rior ovaries; ◊ represents change from half-inferior to fully inferior ovaries. P. marcanoi) as well as P. lychnidiflora, P. aurei- phylogeny presented here, as well as results flora, and P. guamacho, and the remaining sam- from the hypothesis testing, allow a number of 0.01. However, maximum parsimony anal- test showed significant congruence between pled members of the Cactaceae (Fig 2). Within scenarios of cactus evolution to be scrutinized. yses of the individual genomes showed that all genomes (p = 0.08). Three dataset com- the remaining cacti, a divergence clearly reveals a The traditional point of view that the earliest the main topological incongruence was due binations were analyzed (all regions, chloro- monophyletic Opuntioideae and a second clade members of the Cactaceae were Pereskia-like to the placement of the Pereskia species from plast + cox3, and phyC) in partitioned anal- containing a trichotomy involving P. aculeata (Barthlott and Hunt 1993; Britton and Rose Brazil and the Andes region of South Amer- yses so that each region could be run under and P. horrida; the southern South American 1919; Cronquist 1981; Gibson and Nobel ica. When these taxa were removed, the ILD their specific model. Analyses were run using species (P. grandifolia, P. stenantha, P. nemorosa, 1986) seems to be borne out in this study. The 52 Butterworth and Edwards—Earliest Divergences in Cactaceae HASELTONIA 14, 2008 53

Table 1. Results from hypothesis testing using Maximum Likelihood and the Shimodaira- Literature cited Hasegawa Test. Maternal data incorporates sequence data from the chloroplast rbcL, rpl16 intron, psbA-trnH IGS, trnK/matK, and the mitochondrial cox3. Hypotheses rejected below Barthlott W, Hunt D. 1993. Cactaceae. pp 161– Kimura M. 1981. Estimation of the evolutionary the 5% value are marked with an asterisk (*). 197 In The families and genera of vascular plants distances between homologous nucleotide se- (K Kubitzki, ed) Springer-Verlag, Berlin. quences. Proceedings of the American Academy Basal Opuntioideae Basal Pereskia Basal Maihuenia Boke NH. 1964. The cactus gynoecium: a new of Arts and Sciences 78: 454–458. interpretation. American Journal of 51: Leuenberger BE. 1986. Pereskia (Cactaceae). Combined Data 0.067 1.00 0.047* 598–610. Memoirs of The New York Botanical Garden Maternal Data 0.335 1.00 0.418 Britton NL, Rose JN. 1919. The Cactaceae. Carn- 41: 1–141. egie Institution, Washington. Maddison D, Maddison W. 2005. MacClade, ver- Nuclear (phyC) 0.081 0.224 0.016* Butterworth CA. 2006. Resolving “Nyffeler’s sion 4.08. Sinauer. Puzzle”—the intriguing taxonomic position of Mauseth JD, Landrum JV. 1997. Relictual veg- Blossfeldia. Haseltonia 12: 3–10. etative anatomical characters in Cactaceae: Bayesian consensus cladogram (Fig 3) clearly tual within the family (Mauseth and Landrum Butterworth CA, Wallace RS. 2005. Molec- the genus Pereskia. Journal of Plant Research shows that Pereskia is paraphyletic, with Mai- 1997). Griffith (2004) points out that woody ular phylogenetics of the leafy cactus genus 110: 55–64. huenia, Opuntioideae, and Cactoideae embed- members of the usually possess Pereskia (Pereskioideae). Systematic Botany 30: Nyffeler R. 2006. The closest relatives of cacti: ded within. some form of anomalous secondary growth or 800–808. insights from phylogenetic analyses of chlo- Griffith (2004) discussed an alternative hy- stem structure (Cronquist 1981, 1988). Cronquist A. 1981. An integrated system of clas- roplast and mitochondrial sequences with sification of flowering plants. Columbia Univer- special emphasis on relationships in the tribe pothesis in which a leafless stem succulent of the The phylogenies presented in this paper sup- sity Press. Anacampseroteae. American Journal of Botany Opuntioideae represented the ancestral form of port the hypothesis of Pereskia representing the Cronquist A. 1988. The evolution and classifica- 94: 89–101. Cactaceae. However, the cladograms presented earliest members of the cacti. This conclusion tion of flowering plants. New York Botanical Nylander JA. 2004. MrModeltest, version 2. Evolu- in this study do not support a “basal” position supports earlier hypotheses of cactus evolution Garden, Bronx. tionary Biology Centre, Uppsala University. for the Opuntioideae (also demonstrated by a following a course from ancestral woody Edwards EJ, Nyffeler R, Donoghue MJ. 2005. Posada D, Crandall KA. 1998. Modeltest: test- parametric bootstrapping test in Edwards and with typical dicotyledonous leaves in Pereskia Basal cactus phylogeny: implications of Pereskia ing the model of DNA substitution. Bioinfor- others 2005). It is important to note, how- to leafless stem succulents encountered in the (Cactaceae) for the transition to the matics 14: 817–818. ever, that the results of the hypothesis testing Opuntioideae and Cactoideae (Barthlott and cactus life form. American Journal of Botany Rambaut A, Drummond AJ. 2007. Tracer. Uni- did not statistically reject the hypothesis of a Hunt 1993; Britton and Rose 1919; Cronquist 92: 1177–1188. versity of Edinburgh. Felsenstein J. 1981. Evolutionary trees from DNA Ronquist F, Huelsenbeck JP. 2003. MrBayes 3: “basal” Opuntioideae. Griffith presents a good 1981; Gibson and Nobel 1986). sequences: a maximum likelihood approach. Jour- Bayesian phylogenetic inference under mixed discussion of problems based on the concep- However, understanding early evolutionary nal of Molecular Evolution 17: 368–376. models. Bioinformatics 19: 1572–1574. tual model of the phylogeny of the Cactaceae events with certainty in the Cactaceae will be Gibson AC. 1977. Vegetative anatomy of Maihuenia Shimodaira H, Hasegawa M. 1999. Multiple (Griffith 2004) including ovary position, which difficult. The patterns of DNA sequence diver- (Cactaceae) with some theoretical discussions of comparisons of log-likelihoods with applica- he suggests is difficult to polarize in the Portu- gence in molecular data collected thus far suggest ontogenetic changes in xylem cell types. Bulletin tions to phylogenetic inference. Molecular Bi- lacaceae/Cactaceae group due to the presence rapid radiation in early cactus history. Uncov- of the Torrey Botanical Club 104: 35–48. ology and Evolution 16: 1114–1116. of half-inferior ovaries in Portulaca. However, ering the sequence of divergence events during Gibson AC, Nobel PS. 1986. The cactus primer. Swofford DC. 2002. PAUP*. Phylogenetic analy- based on the Bayesian cladogram (Fig 3), we this time (and thus, the relationships among the Harvard University Press, Cambridge. sis using parsimony (*and other methods), ver- can easily hypothesize an evolutionary transition major cactus clades) may require an extremely Griffith MP. 2004. What did the first cactus look sion 4. Sinauer Associates. like? An attempt to reconcile the morphological Tavare S. 1986. Some probabilistic and statistical from superior ovaried ancestors via the half-in- large amount of data. Equally challenging will and molecular evidence. Taxon 53: 493–499. problems on the analysis of DNA sequences. ferior state to the convergent acquisition of in- be accurate character reconstruction—as there Hasegawa M, Kishino H, Yano T. 1985. Dating Lectures in Mathematics and Life Sciences 17: ferior ovaries in the Pereskia bleo clade and the is no fossil record for cacti—and it is likely that the human-ape split by a molecular clock of mi- 57–86, 368–376. Opuntioideae/Cactoideae clade. many lineages that might have represented in- tochondrial DNA. Journal of Molecular Evolu- Wallace RS. 1995. Molecular systematic study of The presence of persistent broad leaves in termediate steps in the transition to the typical, tion 22: 160–174. the Cactaceae: Using chloroplast DNA varia- Pereskia also puzzled Griffith (2004). He argued leafless cactus have since gone extinct. In spite Hunt D, Taylor N. 1986. The genera of the Cac- tion to elucidate Cactus phylogeny. Bradleya that because Portulacaceous relatives of cacti of these obstacles, we have already gained much taceae: towards a new consensus. Bradleya 4: 13: 1–12. (namely Anacampseros and Talinum) possessed clarity on the subject in recent years using the 65–78. Wallace RS, Cota JH. 1996. An intron loss in Hunt D, Taylor N. 1990. The genera of Cac- the chloroplast gene rpoC1 supports a mono- small succulent leaves lacking strong venation, handful of extant taxa that are available, and taceae: progress towards consensus. Bradleya phyletic origin for the subfamily Cactoideae. the leaves of Pereskia could have evolved to su- we are certain that additional efforts will yield 8: 85–107. Current Genetics 29: 275–281. perficially resemble those of typical dicotyledon- further insight into the early evolution of this ous plants. However, until we have confirmed important plant lineage. an accurate placement of the Cactaceae within the Portulacaceae (see Nyffeler 2006; Nyffeler Acknowledgements and others, this volume pp 26–36) we cannot The authors would like to thank the follow- be sure of the polarity of leaf form in this group ing people and institutions for their assis- of plants. Furthermore, it is even plausible that tance and support: Wendy Applequist, Mir- Pereskia might have retained an ancestral leaf iam Diaz, Urs Eggli, Beat Leuenberger, the form while the occurrence of small succulent Desert Botanical Garden (Phoenix, Arizona, leaves happened in numerous independent lin- USA), the Berlin Botanical Garden (Berlin- eages in the Portulacaceae/Cactaceae as the re- Dahlem, Germany), the Sukkulenten-Sam- sult of convergent evolution. The same argu- mlung (Zürich, Switzerland), and the Jardín ment could also apply to wood characters of Botánico Nacional (Santo Domingo, Domin- Pereskia, which are considered to be highly relic- ican Republic).