| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics |
Department of Organismic and Evolutionary Biology, Harvard University Herbaria, 22 Divinity Avenue, Cambridge, Massachusetts 02138 USA
Received for publication February 22, 2001. Accepted for publication July 31, 2001.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: biogeography • Cactaceae • matK • phylogeny • trnK intron • trnL-trnF
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). Centers of diversity are the arid regions of North and South America, notably the southwestern United States and Mexico, East Brazil, and the eastern and western slopes of the South American Andes. Only a single epiphytic species, Rhipsalis baccifera (J. S. Muell.) Stearn, has a distribution range that naturally extends beyond the New World to southern Africa, Madagascar, and Sri Lanka (Barthlott, 1983
; Barthlott and Taylor, 1995
).
The cactus family is remarkable for its great diversity in growth forms. Cacti form xerophytic trees or shrubs with conspicuous persistent leaves (Pereskia Mill.) or most often branched or unbranched, columnar to globular stem succulents. Cacti can be scandent, epiphytic, or epilithic and have either slender, terete stems or flattened, leaflike cladodes. The unusual vegetative morphology is the result of the following major modifications of the general structure of a perennial dicotyledonous flowering plant (Goebel, 1889
; Rauh, 1979
): (1) the leaves are highly reduced or lost, (2) the stems remain green and photosynthetically active for several years with retarded bark formation, (3) cortex and pith are transformed into a succulent water-storage tissue, (4) short side-branches are modified into clusters of spines called areoles, and (5) branching is often highly reduced.
Cacti have fascinated botanists and plant enthusiasts for centuries (Rowley, 1997
), and many are grown today as pot plants for their unusual habits and large, showy flowers. The family comprises 
1500–1800 species in slightly >100 genera (Barthlott and Hunt, 1993
). The current classification scheme is based on the recent work of an ad hoc Working Party under the auspices of the International Organization for Succulent Plant Study (Hunt and Taylor, 1986, 1990
). This group of cactus taxonomists was charged with proposing a consensus on the generic classification of Cactaceae that would provide a compromise between widely divergent views in numbers of genera to be recognized, ranging from 42 (Mottram, 1990
) to 233 (Backeberg, 1966
).
The cactus family is characterized by the following unique morphological features: (1) short shoots that are modified into areoles, (2) shoot apical meristems that are organized into four distinct zones, and (3) ovaries that are "sunken" in the receptacles, which in turn are covered with bracts and areoles (Boke, 1941
; Gibson and Nobel, 1986
; Leuenberger, 1986
). Due to their highly modified vegetative and floral morphology, taxonomists generally regarded the cacti as a very distinct group and placed it in its own order, Cactales (Opuntiales according to Engler, 1892
; e.g., Hutchinson, 1973
; Benson, 1979
). There was disagreement about the closest relatives of the cacti until studies of embryology (e.g., Schnarf, 1931
), plant pigment chemistry (e.g., Mabry, Taylor, and Turner, 1963
), and sieve-element plastids (e.g., Behnke, 1972
) suggested a close relationship of the family Cactaceae to the core Caryophyllales. Molecular studies have confirmed this inference and have identified a distinct clade consisting of Portulacaceae, Basellaceae, Cactaceae, and Didieraceae (e.g., Manhart and Rettig, 1994
). Additionally, recent studies based on increased taxon sampling have suggested that the three latter families are in fact nested in paraphyletic Portulacaceae (Hershkovitz and Zimmer, 1997
; Applequist and Wallace, 1999
).
The family Cactaceae is generally classified into three subfamilies: Pereskioideae, Opuntioideae, and Cactoideae (Schumann, 1899a
; Barthlott and Hunt, 1993
). Recently, however, it was suggested that Maihuenia (F. A. C. Weber) K. Schum., traditionally placed with Pereskia in the subfamily Pereskioideae, should be considered as a subfamily of its own (Wallace, 1995a
; Anderson, 2001
). Indeed, Maihuenia and Pereskia have been placed together essentially because they lack distinct synapomorphies. The subfamily Opuntioideae is characterized by the bony aril of the seeds and the presence of glochids (barbed hairs) in the areoles, while the subfamily Cactoideae is distinct in its lack of leaves (with a few exceptions, e.g., Corryocactus brevistylus (K. Schum.) Britton & Rose), the characteristic hilum-micropylar region of the seeds (Barthlott and Voit, 1979
), and an intron loss in the chloroplast gene rpoC1 (Wallace and Cota, 1995
).
Cactoideae is by far the most diverse and species-rich subfamily, comprising >80% of the cactus species. The present tribal classification of Cactoideae (Barthlott and Hunt, 1993
) is largely based on the work by Franz Buxbaum (Buxbaum, 1958
; Endler and Buxbaum, 1974
) and has received only minor modifications through the reshuffling of a few problematic genera and the lumping together of some tribes (check Table 6 for a summary of four recent tribal classifications for the taxa considered in this study). Recently, Anderson (2001)
moved a few genera to new or different tribes (i.e., Calymmanthium F. Ritter to Calymmantheae; Uebelmannia Buining to Cereeae; Acanthocereus (A. Berger) Britton & Rose, Corryocactus Britton & Rose, Echinocereus Engelm., and Leptocereus (A. Berger) Britton & Rose to Pachycereeae; Harrisia Britton to Trichocereeae) based on results from recent systematic studies (Taylor and Zappi, 1989
; Wallace 1995a, b, 1998
), but otherwise adhered to the general framework of Buxbaum's tribal classification of Cactoideae.
|
; see also Buchheim, 1964
; Buxbaum, 1980
) proposed that there is a consistent pattern of increased specialization and morphological reduction in the evolution of various characters of the stems, flowers, fruits, and seeds. This idea of directionality, or trend, in cactus evolution influenced many subsequent ideas about phylogenetic relationships (e.g., Buxbaum, 1967
; Barthlott and Hunt, 1993
). However, these generalizations about the evolutionary history in cacti can only be rigorously tested with the help of a detailed phylogenetic hypothesis. For instance, Porter, Kinney, and Heil (2000)
investigated the homoplasy of morphological characters and the evolution of paedomorphosis in Sclerocactus Britton & Rose based on an explicit phylogenetic framework using sequence data. The present study aims at providing a starting point for a phylogeny of the major relationships within Cactaceae using molecular sequences from the chloroplast genome for later comparative analyses of growth-form evolution.
The trnK intron (referred to here as trnK/matK) consists of the matK gene and two flanking introns. In the Cactaceae this region comprises almost 2600 base pairs (bp). While the matK gene, which encodes a protein structurally related to the maturases (Neuhaus and Link, 1987
), has been used in a number of infrafamilial studies (e.g., Johnson and Soltis, 1994
; Plunkett, Soltis, and Soltis, 1996
; Kron, 1997
), the adjacent noncoding introns of 1050 bp have only recently been used for phylogenetic analyses (e.g., Hiroshi, Thien, and Kawano, 1999
; Hu et al., 2000
; Lavin et al., 2000
). In addition, sequences of the trnL intron and the trnL-trnF intergenic spacer (together referred to here as trnL-trnF) were generated and added to a combined analysis in order to improve resolution in three large subclades of Cactoideae. The trnL-trnF sequences are rich in indels and have been used previously to resolve relationships among closely related genera and tribes (e.g., Böhle et al., 1994
; Bayer and Starr, 1998
).
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
: fig. 34). Exemplars were chosen in order to evenly represent cactus diversity. Later, additional taxa were added based on questions arising from a preliminary analysis. A list of the 72 taxa included in the present study has been archived on the Botanical Society of America website (http://ajbsupp.botany.org/). While trnK/matK was sequenced for all 72 exemplars, trnL-trnF sequences were added to a combined analysis for 41 taxa to further resolve relationships within three large clades of Cactoideae. Species of Grahamia Gill. and Talinum Adans. (Portulacaceae) were used as outgroup taxa (Hershkovitz and Zimmer, 1997
).
Sequencing methods
Genomic DNA was isolated from 10–20 mg dried stem cortex tissue using the DNeasy Plant Mini Kit (QIAGEN, Valencia, California, USA). In order to overcome problems with excessive mucilage, the manufacturer's extraction protocol was modified as follows: (1) spinning for 1 min at 628 rad/sec (6000 rpm) after step 3 and only using supernatant for step 4, and (2) spinning for 5 min at 1256 rad/sec (12 000 rpm) after step 4 and only using supernatant for step 5. The two chloroplast markers were amplified using primers trnK-3914F and trnK-2R for trnK/matK (Johnson and Soltis, 1994
) and primers trnL-c and trnF-f for trnL-trnF (Taberlet et al., 1991
). The polymerase chain reaction (PCR) amplifications were performed in 30-µL reactions containing 10 µL PCR buffer, 2.5 mmol/L MgCl2, 1 mmol/L dNTPs, 0.5 µmol/L of each primer, and 2.5 units of Taq DNA polymerase (QIAGEN). Amplifications were carried out with an initial denaturation step at 94°C for 4 min, followed by 34 cycles of 94°C for 30 sec, 48°C for 60 sec, and 72°C for 90 sec, and finished with a final elongation step at 72°C of 7 min. For trnL-trnF, the annealing temperature was kept at 52°–54°C. The PCR products were run out on 1% Tris/Boric acid/EDTA (TBE) agarose minigels, and bands were cut out and cleaned using the QIAquick Gel Extraction Kit (QIAGEN). Double-stranded PCR products were directly sequenced using the ABI Prism Big Dye Terminator Cycle Sequencing Reaction Kit (Perkin-Elmer Applied Biosystems, Foster City, California, USA) and run on an ABI 377 automated sequencer using Long Ranger gels (FMC Bioproducts, Rockland, Maine, USA). Sequencing of both strands was accomplished for trnK/matK using amplification primers, primer matK-4F (Manos and Steele, 1997
), and five newly designed internal primers (Table 1). For trnL-trnF, both strands were sequenced using external primers, internal primer trnL-d (Taberlet et al., 1991
), and a newly designed primer trnL-edF (Table 1).
|
; GenBank GBAN-X04826) and of Arabidopsis thaliana (L.) Heynh. for trnL-trnF (Sato et al., 1999
; GenBank GBAN-AP000423). (The prefix GBAN- has been added to link the online version of American Journal of Botany to GenBank, but is not part of the actual GenBank accession number.) Sequences of trnK/matK were easily aligned manually with the insertion of distinct gaps. However, the alignment of the trnL-trnF sequences proved to be cumbersome; an initial automated alignment with ClustalW 1.74 (Thompson, Higgins, and Gibson, 1994
) using standard parameters was then manually adjusted. Informative gaps were coded as additional binary or multistate characters (Baum, Sytsma, and Hoch, 1994
). Gaps of variable lengths due to runs of just one type of nucleotides were not considered (Schwarzbach and Ricklefs, 1998
). For the combined data set, question marks were used as placeholders for those exemplars that lacked the trnL-trnF sequence. All sequences were submitted to GenBank (accession numbers GBAN-AY015273–GBAN-AY015344 for trnK/matK and GBAN-AY015345–GBAN-AY015426 for trnL-trnF) and the aligned data matrix and the consensus trees are available from TreeBase (http://www.herbaria.harvard.edu/treebase).
All analyses were conducted with PAUP* 4.0 (Swofford, 2000)
on a Celeron 300 MHz with 128 MB RAM. Adjusted pairwise sequence differences were calculated for the trnK/matK data set. Six data sets were analyzed with unweighted maximum parsimony: (1) trnK/matK sequences including coded gaps, (2) trnK/matK sequences excluding coded gaps, (3) matK gene sequences including coded gaps, (4) trnK noncoding intron sequences including coded gaps, (5) combined trnK/matK and trnL-trnF sequences including coded gaps, and (6) combined trnK/matK and trnL-trnF sequences excluding coded gaps. Due to the high number of trees yielded from these analyses, an inverse constraint search approach (Olmstead and Palmer, 1994
; Catalán, Kellogg, and Olmstead, 1997
) was used with two search strategies: (1) 500 random-taxon-addition replicates with tree bisection-reconnection (TBR) branch swapping with each replicate restricted to 60 sec of swapping, and (2) a simple taxon addition with TBR branch swapping allowed to proceed for 24 h. The strict consensus trees yielded by these analyses were then applied to inverse constraint searches to check for trees equal or shorter in length that contradict the consensus topology. If no such trees were found, it was assumed that the strict consensus tree adequately summarized the available evidence of the analyzed data.
Support for individual branches was determined with parsimony bootstrapping (Felsenstein, 1985
) and decay analyses (Bremer, 1988
; Donoghue et al., 1992
) for the following four different data sets, all including coded gaps: (1) trnK/matK, (2) matK gene only, (3) trnK noncoding intron only, and (4) combined trnK/matK and trnL-trnF. For the bootstrap analysis each data set was analyzed with 500 replicates and TBR branch swapping restricted to 100 trees per replicate. The inverse constraint search strategy was used to determine the decay index for each clade of the four strict consensus trees with branch swapping restricted to 30 min.
A number of constraint parsimony searches based on 500 random-taxon-addition replicates (search strategy 1) were conducted to determine the cost of rearranging the phylogeny to fit a number of alternative hypotheses as listed in Table 5. The rationale for these constrained analyses is explained in detail in the Discussion part. Random samples of 100 most-parsimonious trees yielded by these constraint searches were compared to a similar number of unconstrained trees using a Wilcoxon signed-rank test (Templeton, 1983
; henceforth referred to as "Templeton test") as implemented in PAUP* 4.0 (Swofford, 2000
). The highest P value of the 10 000 comparisons was determined, and statistical significance of the tests is reported after sequential Bonferroni adjustment (Rice, 1989
).
|
; Sanderson et al., 2000
). The placement of Blossfeldia Werderm. as sister to the rest of subfamily Cactoideae was a highly unexpected finding warranting detailed investigation. In order to investigate the possible cause of this puzzling result, the following three different analyses were conducted (Huelsenbeck, 1997
): (1) a parsimony analysis based on the trnK/matK data set with coded gaps included but excluding the two exemplars of Blossfeldia, (2) a maximum likelihood analysis (i.e., a phylogenetic inference method putatively less prone to "long-branch attraction"; e.g., Kuhner and Felsenstein, 1994
; Huelsenbeck, 1995
), and (3) a Monte-Carlo simulation analysis with parametric bootstrapping (Huelsenbeck, Hillis, and Jones, 1996
).
A maximum likelihood analysis of the trnK/matK data set (excluding coded gaps) was conducted using the HKY85 model of molecular evolution (Hasegawa, Kishino, and Yano, 1985
). A successive approximation approach (e.g., Sullivan, Holsinger, and Simon, 1996
) was used to accommodate for the computationally intensive estimation of model parameters and tree topology. First, two different starting trees were generated: (1) a maximum likelihood tree using the HKY85 model with standard values for the model parameters (base frequencies was empirical; transition/transversion ratio was 2; among-site rate variation was equal) and (2) a minimum evolution tree based on this same model but allowing for substitution rate heterogeneity following a discrete approximation of a gamma (
) distribution. For the resulting two different topologies, parameters for the HKY85 + model were estimated. Then heuristic maximum likelihood searches with simple taxon addition and TBR branch swapping were conducted based on these estimated parameters. This successive approximation process of alternately estimating model parameters and tree topologies was repeated until both analyses converged to the same single result and parameter estimates ceased to change.
The single tree obtained from the maximum likelihood analysis was constrained to place Blossfeldia liliputana Werderm. as sister to Parodia microsperma (F. A. C. Weber) Speg. (see below for justification of this hypothesis), and branch lengths and model parameters were estimated using the HKY85 + model. Then 500 data sets of similar size and substitution rates were simulated based on the estimated parameters using Seq-Gen 1.1 (Rambaut and Grassly, 1997
) and analyzed for the difference in tree length between unconstrained and constrained searches (maximum parsimony, simple taxon addition, TBR branch swapping limited to 100 trees) that forced Blossfeldia liliputana to be sister to Parodia microsperma. The distribution of tree length differences from the simulated data sets was compared with the tree length difference estimated from the analysis of the real data using a significance level of P < 0.05.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
|
Complete sequences of the trnL intron and the trnL-trnF intergenic spacer were added to a combined trnK/matK and trnL-trnF data matrix for 41 exemplars of three major clades of Cactoideae (see below; no sequences were available for Browningia chlorocarpa (Kunth) W. T. Marshall, Hylocereus peruvianus Backeb., Matucana intertexta F. Ritter, Pfeiffera miyagawae Barthlott & Rauh, and P. monacantha (Griseb.) P. V. Heath). The aligned trnL-trnF matrix comprised 1123 sites, of which 903 were constant and 96 (8.5%) were informative. Length variation in trnL-trnF sequences between different taxa was remarkable (gaps ranging 2–57 bp) and required the insertion of 12 informative indels, which were coded as binary or multistate characters (Table 2). The addition of trnL-trnF sequences to a combined data set increased the ratio of terminal taxa to informative characters from 1:2.41 for the trnK/matK data set to 1:5.05 for the combined data set (comparison based on 41 exemplars for which trnK/matK and trnL-trnF were available). Further sequence information of the different partitions of the trnK/matK and trnL-trnF data sets are given in Table 3.
|
). Descriptive information about the six different parsimony analyses are listed in Table 4. The strict consensus of 61 354 most-parsimonious (mp) trees from the trnK/matK analysis, including coded gaps, yielded from search strategy 1 (500 random-taxon-addition replicates) is given in Fig. 1. This analysis resolved 39 clades, with bootstrap support values ranging from 47 to 100% (Fig. 1; values above the branches). The three clades that were not supported by the analysis with coded gaps excluded are marked with asterisks.
|
|
The combined analysis of trnK/matK and trnL-trnF sequences (search strategy 1) yielded 62 850 mp trees, and the strict consensus tree of this analysis is given in Fig. 2. The tree topology is largely congruent with the one derived from the trnK/matK analysis. The only taxa with incongruent positions in the two analyses were Harrisia pomanensis (F. A. C. Weber) Britton & Rose (sister to Echinopsis pentlandii (Hooker) A. Dietrich based on the trnK/matK data; sister to Echinopsis chiloensis (Colla) Friedrich & G. D. Rowley based on the combined data) and Parodia microsperma (sister to P. maassii based on the trnK/matK data; sister to P. ottonis based on the combined data). The addition of trnL-trnF data resulted in improved resolution, increasing the number of clades on the strict consensus tree from 39 to 51. In addition, bootstrap support values and decay indices increased consistently for the different subclades of these three groups.
|
Maximum likelihood analysis
The successive approximation approach used for the maximum likelihood analysis of the trnK/matK data set (excluding coded gaps) based on the HKY85 + model converged from the two suboptimal starting trees, the initial maximum likelihood tree (–log L = 9876.224) and the minimum evolution tree (–log L = 9891.006), to the same optimal tree and the same estimated parameters within two successive approximation cycles. The resulting phylogram is given in Fig. 3 and the estimated parameters are –log L = 9875.797, transition/transversion ratio = 0.646543, nucleotide parts are A = 0.30996, C = 0.16795, G = 0.17639, T = 0.34570 and gamma shape parameter alpha = 0.37489. The maximum likelihood tree, when compared to the strict consensus tree of the maximum parsimony analysis, favored slightly different relationships in the tribe Cacteae and Echinopsis Zucc., while providing no indication for distinct clades formed by Maihuenia + Opuntioideae and Calymmanthium + Copiapoa Britton & Rose (see below for more details).
|
Parametric bootstrapping
The tree lengths yielded from the 500 simulated data sets with Blossfeldia liliputana forced to be sister to Parodia microsperma ranged from 842 to 1045 steps (mean = 944.35; SD = 39.24). The constrained analyses of these simulated data sets generally resulted in mp trees of the same length or trees just one to three steps longer (number of simulated data sets with corresponding tree length difference in steps: no difference in 398 data sets, one step difference in 85 data sets, two steps difference in 14 data sets, three steps difference in 3 data sets) compared to unconstrained analyses of the same data sets, which showed the two tested taxa nested together in a small clade. The difference in tree lengths yielded from the real data set, however, comprised 34 steps. This large discrepancy in tree lengths derived from simulated data sets compared to those from the real data set made it safe to reject the hypothesis that branch length heterogeneity is a reasonable explanation for the unexpected result of Blossfeldia being sister to all other Cactoideae (Huelsenbeck, Hillis, and Jones, 1996
).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Givnish and Sytsma, 1997
). However, the large number of exemplars, compared to the number of informative characters available (ratio about 1 : 4) and the uneven allocation of this information among the clades (see branch length differences in Fig. 3) are likely responsible for the large number of trees yielded from the parsimony analyses. For this reason, it was not possible to conduct TBR branch swapping to completion, and an inverse constraint search approach was chosen to assess the validity of the strict consensus trees based on the incomplete parsimony analyses. In order to improve resolution within three larger clades of Cactoideae, I added sequences of the trnL intron and the trnL-trnF intergenic spacer to a combined analysis of chloroplast markers. The additional information helped to resolve relationships among rather closely related taxa within three large clades of Cactoideae.
Researchers have recently raised a number of concerns about potential sources for misleading results in molecular systematic analyses (e.g., Doyle, 1992
; Sanderson and Doyle, 1992
; Swofford et al., 1996
; Wendel and Doyle, 1998
). For the present study, I verified a number of unexpected findings, such as the position of Blossfeldia, Browningia Britton & Rose, Copiapoa, Frailea Britton & Rose, and Pfeiffera, with sequencing of additional congeneric exemplars in order to decrease the possibility of contamination and mistakes. Furthermore, the parsimony and the maximum likelihood analyses yielded very similar topologies, which indicated that inconsistencies of one or the other algorithm in reconstructing the phylogenetic relationships are not an obvious explanation for some of the unexpected results. This was, in addition, confirmed for the special case of Blossfeldia with a Monte-Carlo simulation analysis as described by Huelsenbeck, Hillis, and Jones (1996)
.
Relationships among the major clades in Cactaceae
The monophyly of the family Cactaceae, as indicated by a number of unique morphological features (e.g., Gibson and Nobel, 1986
) and molecular data (Wallace, 1995b
; Wallace and Cota, 1995
; Hershkovitz and Zimmer, 1997
), is also strongly supported by the present trnK/matK data set (bootstrap [BS] = 100%, decay index [DI] = 10). The comparison of pairwise sequence distances in the trnK/matK data set confirmed previous observations (Hershkovitz and Zimmer, 1997
) that the cacti are remarkable for their small genetic differences, even among morphologically very different taxa, such as Pereskiopsis diguetti and Parodia microsperma, when compared to a small sample of taxa from the sister group in Portulacaceae.
Relationships as inferred by the combined analysis of trnK/matK and trnL-trnF data (Fig. 2) are summarized in a cladistic classification scheme for the genera investigated in this study (Table 6). A number of clade names, derived from tribal and generic names, are listed together with four more recent traditional tribal classifications for the Cactoideae. All clades, with the exception of the tribe Cereeae, recognized for this classification scheme received bootstrap support values of 70% or higher.
Well-supported major clades within Cactaceae are the genus Maihuenia (trnK/matK data: BS = 100%, DI = 18; Fig. 1), and the subfamilies Opuntioideae (trnK/matK data: BS = 100%, DI = 22) and Cactoideae (trnK/matK data: BS = 97%, DI = 7), while the three exemplars of Pereskia do not form a monophyletic group, but a basal grade, in the present analysis. Bootstrap support for a paraphyletic genus Pereskia is low (trnK/matK data: BS = 63%, DI = 1) and is only indicated by the trnK noncoding intron partition. Furthermore, a constraint analysis forcing the exemplars of Pereskia to form a distinct clade yielded mp trees just one step longer. Hence, neither paraphyly nor monophyly of Pereskia can be ruled out based on the present data. The genus Maihuenia, traditionally placed in subfamily Pereskioideae because it lacks the synapomorphies of Opuntioideae and Cactoideae, is ecologically specialized (Leuenberger, 1997
; Mauseth, 1999
) and has a restricted distribution in southern Argentina and Chile. Recently, the genus was placed in a monotypic subfamily Maihuenioideae (Wallace, 1995b
; Anderson, 2001
). Opuntioideae and Cactoideae, in contrast, are very well-characterized by a number of structural and molecular synapomorphies. Different taxa of both subfamilies occur over the entire distribution area of the cacti and often are found at the same localities; however, Cactoideae comprises about seven times the number of species of Opuntioideae (estimate based on Barthlott and Hunt, 1993
).
The present study supported a clade consisting of Maihuenia, Opuntioideae, and Cactoideae (trnK/matK data: BS = 84%, DI = 1; signal provided by the matK partition only) sister to Pereskia. In addition, Maihuenia and Opuntioideae form a rather weakly supported monophyletic group in parsimony analyses (trnK/matK data: BS = 70%, DI = 1; signal provided by the matK partition only). However, there is indication that this finding is affected by including Blossfeldia to the parsimony analysis. The strict consensus tree of the trnK/matK analysis excluding Blossfeldia showed a polytomy consisting of Maihuenia, Opuntioideae, and Cactoideae. Furthermore, the maximum likelihood analysis did not provide support for this sister group relationship. Summary trees of other molecular analyses (Wallace, 1995b
; Martin and Wallace, 2000
) identified the same major clades of Cactaceae, though these trees suggested different relationships among them.
Relationships in Cactoideae
The relationships in the subfamily Cactoideae, by far the most diverse clade of Cactaceae, are the focal point of the present study. The findings revealed by this molecular analysis conflict in a number of cases with traditional, well-established ideas about relationships within this clade, but they also provide strong evidence to resolve various old debates about the placement of certain enigmatic taxa.
Blossfeldia
Certainly the most controversial result of the present study is the placement of Blossfeldia liliputana. This monotypic genus of tiny globular cacti from the eastern slopes of the Andes of northern Argentina and southern Bolivia (Leuenberger and Eggli, 1999
) forms the sister group to the rest of the subfamily Cactoideae (trnK/matK data: BS = 99%, DI = 7; Fig. 1). The same relationship is also favored by the maximum likelihood analysis (Fig. 3). Blossfeldia occurs in crevices of more or less vertical cliffs and is ecologically and morphologically specialized. The plants lack any xeromorphic structures that might prevent them from desiccation. Under water stress, the plants may lose up to 80% of their initial mass and then recover within a few weeks to their original vital condition (Barthlott and Porembski, 1996
). The tiny seeds with prominent strophiola are very similar to those of Parodia microsperma, wherefore a close relationship between these two genera was widely accepted (Buxbaum, 1967
; Taylor, 1989
). However, this alternative hypothesis was strongly opposed by the results of a Monte-Carlo simulation analysis and a Templeton test (Table 5), as was a less stringent hypothesis of a clade consisting of Blossfeldia, Frailea, and Parodia Speg. (Templeton test, P = 0.0018).
Several DNA extractions from different samples were prepared in order to overcome the problem of contamination or mistake. However, all the different sequences yielded the same result as reported here. Further studies are needed to investigate whether the present chloroplast gene tree might significantly differ from the organismal phylogeny or whether in fact this result reflects the correct relationships of Blossfeldia in the cactus phylogeny.
Cacteae
The circumscription of the tribe Cacteae has never been challenged in recent classification schemes (Table 6), and it emerges as the only suprageneric group undisputed by the present molecular analysis with high support values (trnK/matK data: BS = 100%, DI = 10; unique 4 bp indel in 3' trnK partition). This clade of mostly short cylindric to globular or globular-caespitose cacti comprises 20 genera and 500 species (estimate based on Barthlott and Hunt, 1993
) and is predominantly found in Mexico and the southwestern United States, with a number of species of Mammillaria Haw. extending into Central America, northern South America, and the Caribbean. This tribe is well circumscribed by its seeds, which have hilum and micropyle disjunct and only rarely conjunct (Barthlott and Voit, 1979
). The tribe Cacteae is sister to a large unresolved clade, here referred to as core Cactoideae (trnK/matK data: BS = 87%, DI = 3).
Core Cactoideae
The core of Cactoideae consists of the ACHLP clade, the RNBCT clade and three orphan genera (Fig. 2). Calymmanthium (columnar, densely branched cacti from northern Peru; 1 sp.) forms a very weakly supported clade (trnK/matK data: BS = 47%, DI = 1; Fig. 1) with Copiapoa (globular or subcolumnar cacti from the Atacama Desert, Chile; 20 sp.) when coded gaps are included in the analysis. A similar tendency of these two morphologically very different genera to form a clade was also found in preliminary analyses of ITS sequences (R. Nyffeler, unpublished data). On the other hand, this clade collapsed when coded gaps were excluded from the analysis, and a constraint analysis (Table 5) showed that the present data do not conflict with the traditional conception of a close relationship of Copiapoa with other globular Notocacteae (Templeton test, P = 0.49). Frailea is a genus of tiny globular cacti from southeastern South America. It has often been thought to be closely related to either Blossfeldia (Barthlott, 1988
) or Parodia (Eggli and Nyffeler, 1998
; R. Nyffeler, unpublished data). However, the present data conflicts with various alternative hypotheses, i.e., a sister group relationship with Blossfeldia, Parodia sensu stricto (s.s.), or with any other globular Notocacteae (Table 5).
The ACHLP clade, HLP clade, Hylocereeae, and Pachycereeae
The ACHLP clade (name derived from the initials of the genera Austrocactus and Corryocactus and the tribes Hylocereeae, Leptocereeae, and Pachycereeae) comprises 25 genera and 250 species (estimate based on Barthlott and Hunt, 1993
) with predominantly columnar or scandent habit and two distinct groups of epiphytes (Pfeiffera and the tribe Hylocereeae). The combined analysis with additional trnL-trnF sequences (Fig. 2) helped to further resolve relationships within this large clade. There are four distinct subclades in the ACHLP clade. Austrocactus + Eulychnia and Corryocactus are columnar cacti from the western slopes of southern South America (Chile, Peru, Bolivia; a few species of Austrocactus also occur in southern Argentina) and were previously thought to be "basal" members of the tribe Notocacteae (Buxbaum, 1967, 1969
). Pfeiffera comprises a handful of epiphytic and epilithic species from the eastern Andes of Bolivia and northwestern Argentina. This genus was traditionally included in the tribe Rhipsalideae and was regarded as "transitional" between Corryocactus and the other members of Rhipsalideae (e.g., Gibson and Nobel, 1986
). The fourth clade, here called HLP clade, consists of taxa previously referred to either the tribes Browningieae, Echinocereeae, Hylocereeae, Leptocereeae, or Pachycereeae (Table 6). Relationships within this group are not well resolved, with the exception of the two clades comprising the members of the tribe Pachycereeae + Echinocereus (trnK/matK data: BS = 88%, DI = 2 [Fig. 1]; combined data: BS = 99%, DI = 6 [Fig. 2]) and those of the tribe Hylocereeae + Acanthocereus (combined data: BS = 84%, DI = 2). The close relationship of Acanthocereus with the Hylocereeae is favored by the trnL-trnF data, while the maximum likelihood analysis (Fig. 3) of the trnK/matK data indicates a closer relationship of this genus with Pachycereeae. In addition, the maximum likelihood analysis also revealed a distinct clade consisting of Acanthocereus, Echinocereus, and the two tribes Hylocereeae and Pachycereeae. Both clades occur to a large extent in Central America and Mexico, though they are generally found in different habitats. The tribe Pachycereeae comprises large columnar cacti (only Bergerocactus Britton & Rose and in particular Echinocereus form densely branched, low-growing shrubs) and occur in rather arid scrub and desert habitats. In contrast, the tribe Hylocereeae includes scandent or mostly epiphytic cacti, often with very large showy flowers, and is generally found in tropical forests. Acanthocereus, though recently excluded from the tribe Hylocereeae (Barthlott, 1988
; Barthlott and Hunt, 1993
; Anderson, 2001
), mainly differs from that group by forming terrestrial shrubs with slender branches, but is otherwise very similar. Armatocereus Backeb., Neoraimondia Britton & Rose, and Castellanosia Cárdenas are large columnar cacti from South America (Bolivia, Colombia, Ecuador, Paraguay, and Peru), the former two genera largely from the western Andes of South America. Castellanosia was generally thought to be very closely related to Browningia s.s. and was recently included in the latter genus (Barthlott and Hunt, 1993
; Hunt, 1999
), although this is not supported by the present analysis. The genus Leptocereus is found in the Caribbean and forms small trees or shrubs and grows in dry forests.
The RNBCT clade
The other large clade of the core Cactoideae consists of the core Rhipsalideae (R), the core Notocacteae (N), and the Browningieae-Cereeae-Trichocereeae (BCT) clade. This clade is weakly supported (trnK/matK data: BS = 76%, DI = 2 [Fig. 1]), and information is largely contributed by the matK partition. Similar relationships in this RNBCT clade are also identified by the maximum likelihood analysis. Core Rhipsalideae is sister to the latter two clades. The clade consisting of core Notocacteae and the BCT group received high support values (trnK/matK data: BS = 97%, DI = 3 [Fig. 1]; combined data: BS = 98%, DI = 4 [Fig. 2]) from both trnK/matK partitions as well as the trnL-trnF data.
Core Rhipsalideae
The epiphytic cacti of the tribe Rhipsalideae, with the exception of Pfeiffera and some species of Acanthorhipsalis Kimnach, which are associated in the present analysis with taxa of the ACHLP clade, form a well-supported clade (trnK/matK data: BS = 100%, DI = 9; signal provided by both partitions). However, a monophyletic tribe Rhipsalideae in the traditional circumscription (including the genus Pfeiffera sensu lato [s.l.]) cannot be rejected (Templeton test, P = 0.57; Table 5). In any case, the present analysis revealed that sinking the genus Pfeiffera in Lepismium Pfeiff., as recently proposed by Barthlott and Taylor (1995)
, leaves the latter at best paraphyletic. A monophyletic genus Lepismium in the broad sense is opposed by the present trnK/matK data (Templeton test, P = 0.0205; Table 5).
Core Notocacteae
The circumscription of the tribe Notocacteae has always been rather unclear (Table 6). Based on the present analysis, the core group of Notocacteae consists of the genera Eriosyce (including Neoporteria Britton & Rose sensu Barthlott and Hunt, 1993
), Neowerdermannia Fric, and Parodia. This clade of 100 species is very strongly supported (trnK/matK data: BS = 98%, DI = 6; combined data: BS = 100%, DI = 11; with a unique 6 bp indel in the matK partition), with a large share of information contributed by the trnK partition. Core Notocacteae comprises mostly globular, or in a few cases subcolumnar, often unbranched cacti from southern South America. Eriosyce is predominantly West Andean (Chile, Peru, Argentina), Neowerdermannia occurs on the Altiplano, and Parodia is exclusively East Andean (Argentina, Bolivia, Brazil, Paraguay, Uruguay). Various other genera, like Austrocactus, Blossfeldia, Copiapoa, Corryocactus, Eulychnia, and Frailea, previously always included in the tribe Notocacteae, seem to be more closely related to other groups of Cactoideae, and a clade including these genera is rejected by the present data (Templeton test, P = 0.0042; Table 5).
The BCT clade, Cereeae, and Trichocereeae
The BCT clade (name derived from the initials of the tribes Browningieae, Cereeae, and Trichocereeae) is sister to the core Notocacteae and comprises 30 genera and 400 species (estimate based on Barthlott and Hunt, 1993
) of mostly South American columnar and globular cacti. Besides the tribes Cereeae and Trichocereeae, Browningia s.s. (excluding Castellanosia) of the tribe Browningieae, Harrisia of the tribe Echinocereeae, and Uebelmannia of the tribe Notocacteae fall into this strongly supported clade (trnK/matK data: BS = 99%, DI = 8 [Fig. 1]; combined data: BS = 100%, DI = 12 [Fig. 2]; with two indels of 11 and 6 bp in the 5' trnK and matK partitions, respectively). The combined analysis indicates that this BCT clade consists of a basal grade formed by Stetsonia Britton & Rose and Uebelmannia and a polytomy comprising Gymnocalycium Mittler, Cereeae + Browningia, and Trichocereeae. It remains unclear whether the three taxa of the Cereeae included in this study, i.e., Cereus Mill., Coleocephalocereus Backeb., and Micranthocereus Backeb., in fact form a clade. As judged from the branch lengths yielded by the maximum likelihood analysis, there is virtually no information present in the trnK/matK data set that would help resolving relationships among these taxa. The combined analysis, with the addition of the trnL-trnF sequences, resolved a monophyletic clade Cereeae as sister to Browningia s.s., but with a very low support values (combined data: BS = 42%, DI = 1 [Fig. 2]). The support for Cereeae + Browningia s.s. is even lower (combined data: BS = 26%, DI = 1). In contrast, the tribe Trichocereeae, including Harrisia, Rauhocereus Backeb., and Samaipaticereus Cárdenas, receives considerably stronger support (trnK/matK data: BS = 64%, DI = 1; combined data: BS = 76%, DI = 3). However, information for this relationship is only contributed by the matK partition of the trnK/matK data set. Harrisia has traditionally never been associated with the taxa of the tribe Trichocereeae although similarities in pollen characters led Leuenberger (1976)
to propose a possible relationship of Harrisia with Trichocereus Riccob. (alias Echinopsis), and this was later confirmed by a molecular study (Wallace, 1995b
). Buxbaum (1967)
argued that Rauhocereus and Samaipaticereus are basal in the lineage of Browningia s.l., while others suggested a closer relationship between those two genera and the tribe Trichocereeae (Rauh, 1979
; Ritter, 1980, 1981
). The parsimony analysis of the combined data and the maximum likelihood analysis both support a close relationship of Rauhocereus with the genera Haageocereus Backeb., Matucana Britton & Rose, and Oreocereus (A. Berger) Riccob. These taxa either occur on the West Andean side of Chile and Peru or in the Altiplano. The relationships of Samaipaticereus, a monotypic genus from East Andean Bolivia, within the tribe Trichocereeae remain unresolved.
Biogeography, origin, and age of the cacti
All major lineages of the cacti, i.e., Pereskia, Maihuenia, Opuntioideae, and Cactoideae, occur mostly or exclusively in South America. Furthermore, the closest relatives of the cacti from the "portulacaceous cohort" (Applequist and Wallace, 1999
) have their highest diversity on continents of the former Gondwana landmass (Hershkovitz and Zimmer, 1997
). This is generally taken as circumstantial evidence that the family Cactaceae originated in South America (e.g., Schumann, 1899b
; Buxbaum, 1969
; Mauseth, 1990
). Hence, various groups of Pereskia, Opuntioideae, and Cactoideae invaded Central and North America and the Caribbean from their postulated northwestern South American center of origin (Leuenberger, 1986
). The present molecular analysis suggests that three major clades of Cactoideae contributed to the cactus flora of Central and North America (including the Caribbean). The tribe Cacteae is exclusively North American, with some species of the supposedly derived genus Mammillaria occurring in Venezuela and Colombia. The tribes Hylocereeae and Pachycereeae of the HLP clade are widely distributed in Central and North America. The present data are ambiguous about whether these two tribes, including Acanthocereus and Echinocereus, in fact represent a single monophyletic group or whether they form two independent lineages within the HLP clade. The BCT clade, whose taxa most often are found in South America south of the equator, contributed only a few species of the genera Harrisia, Melocactus Link & Otto, and Pilosocereus Byles & G. D. Rowley (the latter two genera were not sampled for this study) to the flora of Central and North America. Finally, there is one widely distributed species of Rhipsalis Gaertn., which extends its distribution to Central and North America.
It is interesting to note that the South American taxa of the ACHLP clade, whose bulk of species diversity is contributed by the Central and North American Hylocereeae and Pachycereeae (180 species, or 70% of the total diversity), mostly occur in West Andean Colombia, Ecuador, Peru, and Chile. Prominent exceptions are Castellanosia from East Andean Bolivia and Pfeiffera from eastern Bolivia and adjacent northwestern Argentina and eastern Peru. In contrast, the RNBCT clade is almost exclusively South American, with different groups either occurring East or West of the Andes.
The presence of a Rhipsalis species in tropical Africa, Madagascar, and Sri Lanka (Barthlott, 1983
; Barthlott and Taylor, 1995
) led some authors to propose that this distribution indicates an old vicariance between South America and Africa (e.g., Backeberg, 1942
) or even an origin of the cacti in the Old World (Croizat, 1952
). This would imply that the cacti originated before the split of the two continents during the late Cretaceous and that all other cacti that might have naturally occurred in Africa got extinct. More recently, however, this distribution pattern of Rhipsalis baccifera, which is characterized by having very sticky seeds comparable to those of the mistletoe (Barthlott, 1983
), has been explained as the result of relatively recent long-distance dispersal by birds (Gibson and Nobel, 1986
; Barthlott and Hunt, 1993
).
While there is growing consensus concerning the spatial origin of cacti in northern South America, there is a disagreement about the temporal aspect of cactus origin. Traditionally, a late Cretaceous origin of cacti, perhaps 65–90 million years ago (mya) immediately following the breakup of the western part of the Gondwana supercontinent, has been favored (e.g., Gibson and Nobel, 1986
; Mauseth, 1990
). This time frame would allow explanation of the absence of endemic cacti in the Old World, while maximizing the time for the evolution of the various distinctive morphological features of extant cacti (Hershkovitz and Zimmer, 1997
). Based on molecular investigations of ITS sequences, Hershkovitz and Zimmer (1997)
proposed a much more recent origin of cacti in mid-Tertiary 30 mya. The present molecular investigation, though not specifically analyzed for this aspect, adds support to the latter hypothesis. The small amount of sequence divergence found in the present data set of chloroplast markers is indicative of a fairly recent origin of the major radiations in Cactaceae. Unfortunately, the large differences in sequence divergence among different groups of cacti are not conducive to a straightforward molecular clock analysis of the present data set.
Conclusions
The present molecular study helped to resolve a number of old disputes about the relationships of some enigmatic taxa. Gymnocalycium, Stetsonia, and Uebelmannia take a basal position in a large clade of South American cacti consisting of the tribes Browningieae, Cereeae, and Trichocereeae (BCT clade). Neowerdermannia is closely related to Eriosyce and Parodia of the core Notocacteae. Harrisia, Rauhocereus, and Samaipaticereus are members of the tribe Trichocereeae, while Acanthocereus shares affinities to the tribe Hylocereeae and Echinocereus is nested in the tribe Pachycereeae. Furthermore, this study provides for the first time a detailed phylogenetic hypothesis of the major relationships within the subfamily Cactoideae, though the basal position of Blossfeldia, as sister to the rest of Cactoideae, is difficult to understand based on structural data. The tribe Cacteae, composed of Central and North American globular cacti, is sister to a few orphan genera (Calymmanthium, Copiapoa, Frailea) and two major clades of basically South American cacti. The predominantly Central and North American columnar cacti of the two tribes Hylocereeae and Pachycereeae share close relationships to largely West Andean South American taxa of the former tribes Browningieae and Leptocereeae. However, the present study also unearthed a number of new problems for cactus systematics, e.g., the question about the closest relatives of Blossfeldia, Castellanosia, Copiapoa, Frailea, and Pfeiffera. More comparative sequencing of various other molecular markers, in particular from the nuclear genome, is needed to test relationships as proposed by the present study of trnK/matK and trnL-trnF data.
|
FOOTNOTES |
|---|
2 Current address: Gysulaweg 6, CH-5034 Suhr, Switzerland (rnyffeler{at}hotmail.com
).
|
LITERATURE CITED |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Anderson E. F. 2001 The cactus family. Timber Press, Portland, Oregon, USA
Backeberg C. 1942 Zur Geschichte der Kakteen im Verlauf der Entwicklung des amerikanischen Kontinentalbildes. Cactaceae—Jahrbuch der deutschen Kakteen-Gesellschaft 1942: 4-72
Backeberg C. 1966 Das Kakteenlexikon. Enumeratio diagnostica Cactacearum. VEB Gustav Fischer Verlag, Jena, Germany
Barthlott W. 1983 Biogeography and evolution in neo- and palaeotropical Rhipsalinae. In K. Kubitzki [ed.], Dispersal and distribution, 241–248. Verlag Paul Parey, Hamburg, Germany
Barthlott W. 1988 Über die systematischen Gliederungen der Cactaceae. Beiträge zur Biologie der Pflanzen 63: 17-40
Barthlott W. D. R. Hunt 1993 Cactaceae. In K. Kubitzki, J. G. Rohwer, and V. Bittrich [eds.], The families and genera of vascular plants, vol. 2, 161–197. Springer Verlag, Berlin, Germany
Barthlott W. S. Porembski 1996 Ecology and morphology of Blossfeldia liliputana (Cactaceae): a poikilohydric and almost astomate succulent. Botanica Acta 109: 161-166[ISI]
Barthlott W. N. P. Taylor 1995 Notes towards a monograph of Rhipsalideae (Cactaceae). Bradleya 13: 43-79
Barthlott W. G. Voit 1979 Mikromorphologie der Samenschalen und Taxonomie der Cactaceae: ein raster-elektronenmikroskopischer Überblick. Plant Systematics and Evolution 132: 205-229[CrossRef][ISI]
Baum D. A. K. J. Sytsma P. C. Hoch 1994 A phylogenetic analysis of Epilobium (Onagraceae) based on nuclear ribosomal DNA sequences. Systematic Botany 19: 363-388[CrossRef][ISI]
Bayer R. J. J. R. Starr 1998 Tribal phylogeny of the Asteraceae based on two n
