View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Ghent University Academic Bibliography

American Journal of Botany 89(1): 132±144. 2002.

MOLECULAR PHYLOGENETICS OF BASED ON NUCLEAR 18S RDNA AND RBCL, ATPB, AND MATK DNA SEQUENCES1

PHILIPPE CUEÂ NOUD,2,5 VINCENT SAVOLAINEN,2 LARS W. C HATROU,3 MARTYN POWELL,2 RENEÂ E J. GRAYER,4 AND MARK W. C HASE2

2Molecular Systematics Section, Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3DS, UK; 3National Herbarium of the Netherlands, Utrecht University Branch, Heidelberglaan 2, 3584 CS, Utrecht, The Netherlands; and 4Biological Interaction Section, Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3DS, UK

To study the inter- and infrafamilial phylogenetic relationships in the order Caryophyllales sensu lato (s.l.), ϳ930 base pairs of the matK plastid gene have been sequenced and analyzed for 127 taxa. In addition, these sequences have been combined with the rbcL plastid gene for 53 taxa and with the rbcL and atpB plastid genes as well as the nuclear 18S rDNA for 26 taxa to provide increased support for deeper branches. The red pigments of Corbichonia, Lophiocarpus, and have been tested and shown to belong to the betacyanin class of compounds. Most taxa of the order are clearly grouped into two main clades (i.e., ``core'' and ``noncore'' Caryophyllales) which are, in turn, divided into well-de®ned subunits. and are polyphyletic, and Por- tulacaceae are paraphyletic, whereas Agdestidaceae, Barbeuiaceae, , and Sarcobataceae should be given familial recog- nition. Two additional lineages are potentially appropriate to be elevated to the family level in the future: the genera Lophiocarpus and Corbichonia form a well-supported clade on the basis of molecular and chemical evidence, and Limeum appears to be separated from other Molluginaceae based on both molecular and ultrastructural data.

Key words: betalain; Caryophyllales; DNA evolution; matK; phylogeny.

Caryophyllales sensu stricto (s.s), sometimes also called Downie and Palmer, 1994; Manhart and Rettig, 1994; Downie, Centrospermae, have long been identi®ed as a natural assem- Katz-Downie, and Cho, 1997; Fay et al., 1997; Lledo et al., blage of families (Braun, 1864; Eichler, 1876). Morphological 1998; Hoot, MagalloÂn, and Crane, 1999; Soltis, Soltis, and characters diagnostic of the group include free-central (some- Chase, 1999; Meimberg et al., 2000; Savolainen et al., 2000a, times basal) placentation, perisperm, and curved embryos (Bit- b; Soltis et al., 2000; Clement and Mabry, in press). As a trich, 1993a). Studies on pigment chemistry (reviewed in result, additional families have been shown to be related to Clement et al., 1994; see also Clement and Mabry, in press) Caryophyllales: , Drosophyllaceae, Nepenthaceae, have shown that all Caryophyllales families, except Cary- , and Polygonaceae (Albert, Williams, and ophyllaceae and Molluginaceae, produce betalain pigments in- Chase, 1992; Williams, Albert, and Chase, 1994), Asteropei- stead of anthocyanins (as is the case in the rest of ¯owering aceae and Physenaceae (Morton, Karol, and Chase, 1997), An- ). Studies of the ultrastructure of sieve-element cistrocladaceae, , Frankeniaceae, Rhabdod- have revealed that Caryophyllales share a unique type (P3; endraceae, Simmondsiaceae, and Tamaricaceae (Fay et al., Behnke, 1994), which is characterized by a peripheral ring of 1997). These increasing data led the Angiosperm Phylogeny proteinaceous ®laments, generally surrounding a protein crys- Group (APG, 1998) to rede®ne Caryophyllales to include the tal of either globular or angular shape. Cronquist and Thorne families listed above (``noncore Caryophyllales'') in addition (1994), summarizing the data available on Caryophyllales, list- to those recognized previously (``core Caryophyllales''), a to- ed 11 families included in this order: , Amarantha- tal of 26 families. The closest relatives of Caryophyllales are ceae, Basellaceae, Cactaceae, , Chenopodi- not yet known with con®dence, although and per- aceae, , Molluginaceae, , Phytolac- haps are the most likely candidates (Hoot, Magal- caceae, and . In addition, seven families were loÂn, and Crane, 1999; Soltis, Soltis, and Chase, 1999; Soltis excluded: Bataceae, , Plumbaginaceae, Poly- et al., 2000). gonaceae, Rhabdodendraceae, Theligonaceae, and Viviana- Despite the data available, uncertainties remain as to the ceae. delimitation of several families, their phylogenetic relation- Molecular systematic studies, starting with Giannasi et al. ships, and the placement of some enigmatic genera. Bittrich (1992), have substantially increased our knowledge of the phy- (1993a) listed six families that lack clear delimitation or valid logeny of the Caryophyllales (Albert, Williams, and Chase, synapomorphies: , Chenopodiaceae, Portula- 1992; Rettig, Wilson, and Manhart, 1992; Chase et al., 1993; caceae, Nyctaginaceae, Phytolaccaceae, and Molluginaceae. 1 Manuscript received 22 February 2001; revision accepted 27 July 2001. Generic composition of Phytolaccaceae, for example, has un- The authors thank James R. Manhart, who provided DNA of Phaulotham- dergone a continuous thinning, with the recognition of Steg- nus; Wendy L. Applequist, who provided most of the Portulacaceae samples, nospermataceae, (as in APG, 1998), and as well as tissue of Alluaudiopsis; Tom J. Mabry, who provided a copy sometimes also Petiveriaceae, Agdestidaceae, Gisekiaceae, and of his paper in press; and the late Patricia Geissler for her help in getting Barbeuiaceae (Nakai, 1942) as separate families. A morpho- funding for this study. This study was funded by the Royal Botanic Gardens, Kew, the Royal Society (London), the Swiss NSF (31-52378.97), and the logical phylogenetic analysis of the order by Rodman et al. Academic Society of Geneva. (1984) was criticized by Hershkovitz (1989), who wrote that 5 Author for reprint requests (e-mail: philcuenoud࿞[email protected]). because of several problems (mainly choice of outgroup, def- 132 January 2002] CUE NOUD ET AL.ÐMOLECULAR PHYLOGENETICS OF CARYOPHYLLALES 133 inition of taxonomic units, and polymorphic characters within taxon addition with equal weights and -bisection-reconnection (TBR) units) the results were less satisfactory than Cronquist's clas- branch swapping, with only 50 held at each replicate to reduce time si®cation (1981). spent in searching nonoptimal tree lengths. Using only the matK matrix, heu- In this paper, we investigate the phylogeny of the order at ristic searches did not swap to completion, therefore only 10 000 of the trees inter- and infrafamilial levels using partial sequences of the found in 200 independent replicates were retained for discussion. Internal matK plastid gene (sometimes referred to as ORF 542; Hira- support was assessed using 1000 bootstrap replicates with TBR swapping and tsuka et al., 1989). We also combined these matK sequences simple addition of taxa, with a limit of ten trees kept in each replicate. with those from plastid rbcL and atpB genes and nuclear 18S For calculation of substitution patterns, nucleotide changes for each of the rDNA to provide increased support for several deeper branch- four genes were optimized with ACCTRAN onto the four-gene tree (Delos- es of the order. In addition, the chemical nature of the red- perma being excluded for the plastid genes). Sequences of the three coding purple pigments in Corbichonia (Molluginaceae), Lophiocar- genes were translated into amino acids using MacClade 3.04 (Maddison and Maddison, 1992), and the amino acid changes were computed after optimi- pus (Phytolaccaceae), and Sarcobatus (Sarcobataceae) is re- zation onto the combined tree. ported here for the ®rst time. For chemical analysis of pigments, ϳ10±50 mg of dried stems of Sarcobatus vermiculatus Torr., Corbichonia decumbens (Forsk.) Exell, Adenogramma sp., MATERIALS AND METHODS and Lophiocarpus polystachyus Turcz. were cut into small pieces and extracted Sequences of the rbcL and atpB exons and 18S rDNA were taken from the for 5 min in boiling 50% aqueous methanol. The extracts were cooled, ®ltered, studies of Soltis, Soltis, and Chase (1999) and Soltis et al. (2000); all available and evaporated to dryness at 40ЊC under vacuum. The precipitates were dis- members of the Caryophyllales from the three-gene matrix were included. solved in two drops of distilled water, and 0.5 mL of 80% aqueous methanol Alignment of 18S sequences was as in Soltis et al. (2000). In a few cases, was added to this solution. Extracts of fresh red of Iresine sp. (Amar- we could not get all four sequences for the same (i.e., the18S rDNA anthaceae; containing betacyanins) and Papaver sp. (Papaveraceae; containing of is missing, and rbcL, atpB, and 18S rDNA of Stellaria and anthocyanins) were prepared in a similar way and used for comparison. The Mollugo were combined with matK of Moehringia and Pharnaceum, respec- extracts were analyzed for the presence of anthocyanins or betacyanins in two tively). After comparisons of the published rbcL sequence of Delosperma with ways. First, we used reverse-phase, high-performance liquid chromatography those from other Aizoaceae, it was found to be too divergent and doubtful (HPLC) with diode array detection (40 ␮L injections) to assess the polarity and and was therefore deleted from the analyses. In addition, rbcL and atpB were UV/visible spectrum of any pigment present. The HPLC equipment used (Waters sequenced for Trichodiadema and additional rbcL sequences for the rbcL/ Ltd., Watford, UK) consisted of a model LC600 pump coupled with a model matK analysis were retrieved from various sources (see http://ajbsupp. 996-photodiode array detector controlled by Millennium software (Column botany.org/). Merck Lichrospher 100RP-18, 250 ϫ 4.0 mm internal diameter; 5 ␮m particle Taxa, voucher information, and accession numbers of all DNA sequences size; temperature of the column was maintained at 30ЊC). Elution was performed have been archived at the Botanical Society of America website (http:// with a linear gradient of two solvents: 2% aqueous acetic acid (solvent A) and ajbsupp.botany.org/). Total DNA from fresh, silica gel-dried, or herbarium methanol : acetic acid : water (18:1:1; solvent B), starting with 25% B and chang- specimens was extracted using the 2 ϫ cetyltrimethylammonium bromide ing to 100% B over 20 min, followed by isocratic elution with 100% B. The (CTAB) method of Doyle and Doyle (1987) and subsequently puri®ed through diode-array detector was set to record UV/visible spectra of all eluting com- cesium chloride gradients (see Savolainen et al., 2000a, for further details). pounds between 200 and 600 nm. Second, we added two drops of 2 mol/L The middle part of matK was ampli®ed by polymerase chain reaction (PCR, potassium hydroxide (KOH) to the crude extract to see the color reaction; de- 26 cycles, 1-min denaturation at 94ЊC, 30-sec annealing at 48ЊC, 1-min ex- velopment of a blue color with KOH indicates presence of anthocyanins and of tension at 72ЊC, 7-min ®nal extension) using primers 390F (5Ј-CGATCTATT- a yellow color the presence of betacyanins (Harborne, 1998). CATTCAATATTTC-3Ј) and 1326R (5Ј-TCTAGCACACGAAAGTCGAAGT- 3Ј), which amplify ϳ930 base pairs between positions 429 and 1313 of the RESULTS matK sequence of Schmitz-Linneweber et al. (2001) for Spinacia oleracea L. (these primers appear to be nearly universal for angiosperms). For cases in matK sequencesÐWe obtained 127 matK sequences in Car- which ampli®cations were weak, a second round of PCR (26 cycles) was yophyllales (two of them from Meimberg et al., 2000; see performed, using 2 ␮L of the product of the ®rst PCR reaction as template. Table 1 for taxon coverage within families) plus 13 for mem- Ampli®cation products were then puri®ed using QIAquick columns (Qiagen, bers of the putative outgroups Dilleniaceae, Olacaceae, San- Crawley, West Sussex, UK). Cycle sequencing (26 cycles, 10-sec denaturation talaceae, and (only the ®rst half of this sequence at 96ЊC, 5-sec annealing at 50ЊC, 4-min extension at 60ЊC) with dye termi- could be obtained for Adenogramma, but this was enough to nators was performed in 10 ␮L volumes and the products were then puri®ed place this taxon with 100% bootstrap support). After align- by ethanol precipitation. The redissolved samples were run on a Applied Bios- ment, the matrix contained 1013 characters, of which 30 at ystems 377 automated DNA sequencer following the manufacturer's protocols the 5Ј end and 52 at the 3Ј end were trimmed to limit the (Applied Biosystems, Warrington, Cheshire, UK). Both strands were se- amount of missing data in the matrix. An additional stretch of quenced using the ampli®cation primers, and both sequences overlapped on 12 nucleotides was excluded, corresponding to two insertions average for ϳ30% of their overall length. Sequences were aligned manually in one outgroup and one ingroup taxon, because of dubious (there were no particular alignment problems), and the informative indels (12) homology. Twelve potentially informative indels were coded were coded at the end of the matrix (absent/present). Phylogenetic analysis at the end of the matrix if they shared same position and size was performed using PAUP*, version 4.0 (Swofford, 1998). Parsimony anal- (all indels were in sets of three). No stop codons were found ysis only was performed because previous studies (e.g., Hillis, 1996) have shown that maximum likelihood tends to agree with parsimony analysis, es- when the sequences were translated into amino acids. In three pecially for well-supported groups. Since partition homogeneity tests have belonging to Caryophyllaceae (i.e., Silene italica Pers., shown cases of rejection with no obvious evidence of incongruence (Soltis et Paronychia sp., and Arenaria tetraquetra L.), highly divergent al., 2000; Reeves et al., 2001), evidence of incongruence was ascertained by matK sequences were recovered; these clustered within the examining the levels of bootstrap percentages of the groups in the combined monocots (with low similarity) in parsimony analyses includ- analysis relative to the separate analyses. The speci®c issue of combining 18S ing a broad sampling of angiosperms (L. W. Chatrou et al., rDNA and plastid DNA gene sequences for analyses of these angiosperms unpublished data). The presence of these ``anomalous'' se- has been adequately covered previously in Soltis et al. (2000). quences is not lineage speci®c because another matK sequence Most-parsimonious trees were obtained using 1000 replicates of random was retrieved from another Silene species, S. rothmaleri Pinto 134 AMERICAN JOURNAL OF BOTANY [Vol. 89

TABLE 1. Number of genera per family in Caryophyllales (after Kubitzki, Rohwer, and Bittrich, 1993; Mabberley, 1997), and number of genera sampled in this study.

Family No. of genera No. of genera sampled Achatocarpaceae 2 2 Agdestidaceae 1 1 Aizoaceae 127 6 Amaranthaceae 166 21 Ancistrocladaceae 1 1 Asteropeiaceae 1 1 Barbeuiaceae 1 1 Basellaceae 4 2 Cactaceae 98 5 Caryophyllaceae 86 10 Didiereaceae 4 4 Dioncophyllaceae 3 3 Droseraceae 3 2 Drosophyllaceae 1 1 Frankeniaceae 2 1 Gisekiaceae 1 1 Halophytaceae 1 1 Molluginaceae 13 8 Nepenthaceae 1 1 Nyctaginaceae 31 8 Physenaceae 1 1 Phytolaccaceae (Rivinoideae) 9 6 Phytolaccaceae (Phytolaccoideae) 4 2 Phytolaccaceae (Microteae) 2 1 Plumbaginaceae 27 9 Polygonaceae 43 16 Portulacaceae 31 11 Rhabdodendraceae 1 1 Sarcobataceae 1 1 Simmondsiaceae 1 1 Stegnospermataceae 1 1 Tamaricaceae 3 1 da Silva, which matched the other sequences in Caryophyl- clade (85% BS). Limeum (Molluginaceae) stands as sister to lales. The ``anomalous'' sequences have not been included in the remaining taxa (50% BS) which are divided into two sister the matrix because they would have given only spurious re- clades, one comprising Aizoaceae, Phytolaccaceae, and Nyc- sults. taginaceae (100% BS) and the other one Cactaceae, Portula- caceae, and Mollugo (70% BS). Among noncore Caryophyl- Four-gene analysisÐThe combined matrix (rbcL, atpB, lales, Tamaricaceae are sister to Frankeniaceae (100% BS), matK, and 18S rDNA) contains 5554 characters (including 12 and Polygonaceae are sister to Plumbaginaceae (100% BS); indels in matK and 5 in 18S rDNA, 1880 variable characters, the four families are in turn grouped in a larger clade (61% and 1221 potentially informative characters) for 37 taxa of BS). A group consisting of Ancistrocladaceae, Dioncophylla- which 11 are outgroups. A single most-parsimonious tree was ceae, Droseraceae, and Nepenthaceae (72% BS) is a sister to found, length 5531 steps, consistency index (CI) 0.49, reten- the latter clade. tion index (RI) 0.55 (Fig. 1). Substitution patterns for each gene are given in Table 2. Table 3 presents the phylogenetic rbcL/matK analysisÐThe combined matrix for rbcL and contribution of each gene as measured by their respective matK contains 2356 characters (including 12 indels for matK, bootstrap percentages for all clades identi®ed in the combined 1287 variable characters, and 906 potentially informative char- analysis. The monophyly of Caryophyllales sensu lato (s.l.) acters) for 64 taxa of which 11 are outgroups (note that rbcL receives 100% bootstrap support (BS). The positions of Rhab- only was available for and ). Two short- dodendron (Rhabdodendraceae) as sister to the remaining Car- est trees were found, length ϭ 5102 steps, CI ϭ 0.40 and RI yophyllales and of Simmondsia (Simmondsiaceae) as sister to ϭ 0.61 (Fig. 2). The trees derived from these data are similar the core Caryophyllales (including , Asteropei- to the one derived from the four-gene analysis for taxa shared aceae) have low BS (50 and 73%, respectively). Core Cary- between analyses. (Gisekiaceae), , and ophyllales (plus Asteropeia) and a clade grouping all noncore (Phytolaccaceae, Rivinoideae) form a clade (71% BS) sister to Caryophyllales (excluding Rhabdodendron and Simmondsia) Nyctaginaceae (52% BS). (Agdestidaceae) and Sar- receive high BS (98 and 95%, respectively). Core Caryophyl- cobatus (Sarcobataceae) are grouped in a clade (77% BS), sis- lales receive 100% BS. Among the remainder of the taxa, ter to Nyctaginaceae/Gisekia/Rivina/Petiveria (but with BS Ͻ Amaranthaceae s.l. and Caryophyllaceae form a group sup- 50%). Corbichonia (Molluginaceae) is sister to the clade in- ported by 96% BS. As sister to this group, the representatives cluding the latter taxa plus Aizoaceae and Phytolaccaceae of Aizoaceae, Phytolaccaceae, Nyctaginaceae, Cactaceae, Por- (67% BS). In the clade sister to the latter, Alluaudia (Didi- tulacaceae, and Molluginaceae form a moderately supported ereaceae) and Basella (Basellaceae) are grouped with Cacta- January 2002] CUEÂ NOUD ET AL.ÐMOLECULAR PHYLOGENETICS OF CARYOPHYLLALES 135

Fig. 1. Single most-parsimonious tree from the combined analysis of 18S rDNA, rbcL, atpB, and matK. Branch lengths are indicated above the branches, bootstrap percentages above 50% are indicated below the branches. De®nition of the orders follows APG (1998) and Savolainen et al. (2000a). ceae and Portulacaceae (92% BS). Stegnosperma (Stegnosper- Large matK analysisÐThe matrix contains 928 characters mataceae) is sister to a clade comprising all the taxa mentioned (including 12 indels, 706 variable characters, and 594 poten- above plus Limeum (96% BS for the whole clade and 56% BS tially informative characters in total; and 689 variable, and 565 for the basal position of Stegnosperma). potentially informative characters among Caryophyllales) for (Achatocarpaceae) is associated with Caryophyllaceae and 140 taxa, of which 11 are outgroups. The length of the shortest Amaranthaceae (84% BS for the whole clade) and Physena trees derived from the matK matrix is 5102 steps, CI ϭ 0.31, (Physenaceae) is strongly associated with Asteropeia (100% RI ϭ 0.70 (Fig. 3a±c). BS). There is a major split in Caryophyllales between a clade

TABLE 2. Patterns of nucleotide substitution for the four genes included in the combined analysis.

Measurement 18S rDNA rbcL atpB matK Total no. of substitutions (steps) 614 1461 1357 2041 No. of amino acid changes N/A 336 431 1091 Variable sites (%) 13.9 36.6 34.2 65.4 Informative sites (%) 8.2 22.2 22.0 45.4 Mean no. of changes per site (total no.) 0.36 1.02 0.92 2.23 Mean no. of changes per variable site 2.57 2.79 2.70 3.41 Mean no. of changes per informative site 3.58 3.86 3.57 4.34 Consistency index 0.49 0.49 0.51 0.48 Retention index 0.54 0.57 0.54 0.52 136 AMERICAN JOURNAL OF BOTANY [Vol. 89

TABLE 3. Bootstrap support (%) based on single-gene analyses for the clades identi®ed in the four-gene matrix.

Clade 18S rDNA rbcL atpB matK Combined Dilleniaceae 100 98 100 100 100 Berberidopsidaceae/Aextoxicaceae 89 53 88 93 100 Santalaceae 99 99 99 100 100 Santalales 78 100 100 100 100 Caryophyllales 68 100 98 98 100 Caryophyllales except Rhabododendron Ͻ50 52 Ͻ50 Ͻ50 50 Noncore Caryophyllales Ͻ50 Ͻ50 64 Ͻ50 95 Ancistrocladaceae, Dioncophyllaceae Droseraceae, and Nepenthaceae Ͻ50 Ͻ50 64 50 72 and 60 56 96 99 100 Drosera and Ͻ50 Ͻ50 54 Ͻ50 59 and Tamarix 100 96 100 100 100 Polygonaceae 84 98 100 100 100 Plumbaginaceae 99 100 100 100 100 Polygonaceae and Plumbaginaceae 98 92 100 100 100 Polygonaceae, Plumbaginaceae, Tamarix, and Frankenia Ͻ50 Ͻ50 Ͻ50 67 61 Core Caryophyllales plus Simmondsia Ͻ50 Ͻ50 64 Ͻ50 73 Core Caryophyllales Ͻ50 99 78 66 98 Core Caryophyllales minus Asteropeia Ͻ50 94 99 97 100 Stellaria and Amaranthaceae 89 Ͻ50 Ͻ50 Ͻ50 96 Amaranthaceae 74 91 87 77 99 Pereskia and Portulaca 73 94 100 76 100 Pereskia, Portulaca, and Mollugo Ͻ50 Ͻ50 64 Ͻ50 70 Phytolaccaceae Ͻ50 89 Ͻ50 82 77 Nyctaginaceae Ͻ50 99 75 98 100 Aizoaceae Ð Ð Ͻ50 100 93 Phytolaccaceae and Nyctaginaceae Ͻ50 85 Ͻ50 79 65 Aizoaceae, Phytolaccaceae, and Nyctaginace Ͻ50 98 Ͻ50 87 100 Higher core Caryophyllales (as de®ned in Fig. 3c) Ͻ50 94 Ͻ50 Ͻ50 86 Number of clades with BS Ͼ 50% 13 20 20 21 28 including core Caryophyllales, Rhabdodendron, and Sim- Calyptrotheca is associated with Didiereaceae (68% BS), and mondsia (53% BS), and noncore Caryophyllales (63% BS). four genera or groups of genera attributed to Portulacaceae are The noncore Caryophyllales are, in turn, separated into two identi®ed. groups: the ®rst one (supported by 73% BS) includes the pairs Finally, there are two ``soft'' incongruences (Seelanan, Plumbaginaceae/Polygonaceae (100% BS) and Frankeniaceae/ Schnabel, and Wendel, 1997) between the matK and the com- Tamaricaceae (100% BS); and the second one includes Dro- bined analyses described above, i.e., the association of Rhab- seraceae, Nepenthaceae, Drosophyllaceae, Dioncophyllaceae, dodendron and Simmondsia as sister groups and the position and Ancistrocladaceae (78% BS). of Amaranthaceae (compare Figs. 1 and 3b), but none of these Within core Caryophyllales (75% BS), Asteropeia is sister placements receives BS Ͼ 50%. We therefore attribute these to the rest of the taxa (100% BS). Caryophyllaceae form a to sampling error, not true incongruence (Huelsenbeck, Bull, well-supported clade (100% BS), which is sister to the rest of and Cunningham, 1996). the core Caryophyllales (BS Ͻ 50%). A clade grouping Amar- anthaceae and Achatocarpaceae (57% BS) is sister to the rest Chemical analysis of pigmentsÐAddition of KOH to the of core Caryophyllales (62% BS). Another large clade (92% dark red Papaver extract, used as the anthocyanin source, BS) includes Nyctaginaceae and Aizoaceae (both recovered as turned its color to blue, as expected for anthocyanins. This monophyletic; 85 and 97% BS, respectively), as well as genera color slowly faded within a day. Addition of KOH to the dark attributed to Phytolaccaceae, Molluginaceae, and Sarcobatus, or pink-red Iresine, Sarcobatus, Corbichonia, and Lophiocar- Gisekia, (Barbeuiaceae), and Agdestis. Nyctagina- pus extracts turned them to bright yellow, indicating presence ceae are sister to Sarcobatus (with BS Ͻ 50%), as is the case of betacyanins. These colors did not fade even after a week. for Aizoaceae and Gisekia (52% BS). The members of Rivi- The HPLC of the Iresine, Corbichonia, and Lophiocarpus ex- noideae (Phytolaccaceae) form another monophyletic group tracts revealed presence of one or two compounds in each supported by 97% BS, also associated with Agdestis (59% extract that absorbed visible light at 536 nm in aqueous meth- BS). Phytolaccaceae s.s. ( and ) form anoth- anol/acetic acid. All these compounds had short retention er monophyletic group supported by 100% BS, which is sister times (2.1±2.2 min), indicating that they are polar. The extract to the Nyctaginaceae/Sarcobatus/Rivinoideae/Agdestis group of Sarcobatus contained three compounds absorbing visible (96% BS). Barbeuia is sister to this clade (59% BS) and the light at 536 nm in aqueous methanol/acetic acid, with retention pair Corbichonia, and Lophiocarpus is sister to Barbeuia and times of 2.2, 11.1, and 11.4 min, respectively. The compounds this clade (95% BS for the clade, 92% BS for its placement). with longer retention times not only absorbed visible light but The last large clade (88% BS) contains a subclade of genera also UV at ϳ299 and 326 nm. The Papaver extract contained attributed to Molluginaceae (89% BS) plus a clade composed ®ve compounds with absorbance in both the visible and UV of Cactaceae, Portulacaceae, Basellaceae, Didiereaceae, and range. Their retention times were 8.2, 10.4, 11.4, 15.3, and

Halophytum (97% BS). Cactaceae, Basellaceae, and Didierea- 16.2 min, and the ␭max (in nanometers) of their UV/visible spec- ceae are monophyletic (72, 99, and 90% BS, respectively). tra in aqueous methanol/acetic acid were 277±517, 283±517, January 2002] CUEÂ NOUD ET AL.ÐMOLECULAR PHYLOGENETICS OF CARYOPHYLLALES 137

Fig. 2. One of the two most-parsimonious trees from the combined analysis of rbcL and matK. The branch marked with an arrow collapses in the strict consensus tree.

283±523, 283±523, and 283±523, respectively. The Adeno- but ranging from 505 to 535 nm for orange-red to blue antho- gramma sample failed to produce pigments of either the be- cyanins). However, some species of Caryophyllales contain tacyanin or anthocyanin class. betacyanins with complex side chains containing phenolic ac- Because betacyanins are in general polar compounds, they ids, e.g., Christmas cactus (Schlumbergera ϫ buckleyi), and usually have short retention times using reversed phased these pigments have retention times like those of anthocyanins HPLC. Many betacyanins only absorb visible light (at ϳ536 and absorb UV as well as visible light (Kobayashi et al., 2000). nm) and not UV. Anthocyanins, on the other hand, are less However, their ␭max in visible light (537±549 nm) is always polar, and therefore have longer retention times; they absorb longer than that of anthocyanins. Using the above criteria, it both UV and visible light (Harborne, 1998). The visible ␭max is clear that Sarcobatus, Corbichonia, and Lophiocarpus all of red-colored anthocyanins is at a shorter wavelength than contain betacyanins rather than anthocyanins. Two of the be- that of betacyanins (e.g., ϳ523 nm for red cyanidin glycosides, tacyanins in Sarcobatus seem to have complex attachments 138 AMERICAN JOURNAL OF BOTANY [Vol. 89

Fig. 3. One of the most-parsimonious trees from the analysis of matK sequences. Because of its size, the tree has been split into three parts: (a) outgroups plus noncore Caryophyllales, (b) basal core Caryophyllales (plus Rhabdodendron and Simmondsia), and (c) higher core Caryophyllales. because they absorb UV and have relatively long retention 0.32 for rbcL and atpB, respectively). Although the number times. of changes at third codon position is similar across genes (Fig. 4), matK has a distinctively higher rate of change at the ®rst DISCUSSION and second positions, indicating that although all genes have a similar rate of neutral change, matK has an apparently higher Pattern of sequence evolutionÐIf measured by the absolute rate of amino acid change. Because matK is more variable than number of changes and the average number of change per total the other genes but has similar CI and RI values, we conclude number of sites, the four genes exhibit different levels of var- that it has a higher number of similarly informative characters. iation, with 18S rDNA being the slowest and matK the most For matK alone, 21 clades received BS Ͼ 50%, whereas there variable (about three- to six-fold increase, Table 2). However, was a total of only 28 for the four-gene combined matrix (Ta- when changes are averaged per variable site only, then rates ble 3). are roughly similar among genes. This indicates that most of the variation is due to the number of variable sites (about ®ve- fold difference in the proportion of variable sites) rather than Caryophyllales phylogenyÐThe subdivision of the order the rate of variation at these sites. This is corroborated by the into two subclades (core and noncore Caryophyllales with CIs and RIs, which are similar among genes (Table 2). Rhabdodendron and Simmondsia unplaced) is well supported Of the protein-coding genes, matK has a high proportion of by the combined data sets. Although the core Caryophyllales nonsynonymous change (the amino acid changes/nucleotide have long been identi®ed, the noncore Caryophyllales are a substitutions ratio is 0.53 for matK, compared with 0.23 and relatively recent discovery due mostly to molecular phyloge- January 2002] CUEÂ NOUD ET AL.ÐMOLECULAR PHYLOGENETICS OF CARYOPHYLLALES 139

Fig. 3. Continued. netics (Albert, Williams, and Chase, 1992; Chase et al., 1993; the betalain-producing families are not monophyletic. Our re- Nandi, Chase, and Endress, 1998). sults show two anthocyanin-producing taxa, Molluginaceae The positions of Simmondsia and especially Rhabdodendron (minus Corbichonia and Limeum) and Caryophyllaceae, are lack strong support (see Fig. 3b). Their respective placements nested within the betalain-producing clades. Limeum, which received weaker BS in the four-gene analysis (73 and 50%, appears as sister to a group of taxa producing both anthocy- respectively), compared to 18S rDNA/rbcL/atpB combined (79 anins and betalains, possesses none of these pigments (Clem- and 69%, respectively), indicating that matK does not bring ent et al., 1994). additional support for the placement of these taxa. The posi- Aizoaceae, Nyctaginaceae, and Phytolaccaceae are charac- tion of these two taxa as sister groups in the matK analysis is terized by the presence of raphide crystals (Judd et al., 1999). probably due to sampling error, given evidence derived from Aizoaceae appear to be monophyletic based on the matK anal- combined analyses; this association was also found in some ysis, with Ruschioidae (here represented by Delosperma, Phyl- of the shortest trees inferred from atpB sequences (Savolainen lobolus, , and Trichodiadema) forming a clade sister et al., 2000b), presumably for the same reason. to Aizooidae (Galenia and Plinthus). With Gisekia, the family is sister to a clade including principally Phytolaccaceae s.l. and Core CaryophyllalesÐThe combined data sets gave sub- Nyctaginaceae. This differs from several studies based on rbcL stantial support for the division of this group (minus Astero- peia) into two sister clades. The ®rst clade includes Nyctagi- alone; in Savolainen et al. (2000a), in which Trianthema (Se- naceae, Phytolaccaceae, Aizoaceae, Molluginaceae, Cactaceae, suvioidae) and Delosperma were represented, only the former Portulacaceae, Stegnospermataceae, and several rather isolated occupied a similar position to our analysis, whereas the latter genera. This clade has been shown by Clement and Mabry (in was nested within Phytolaccaceae s.s. (as speci®ed in the MA- press) to share the derived character of a globular crystal in TERIALS AND METHODS section, we did not include the sieve-element plastids for all taxa except Limeum and Steg- Delosperma sequence for rbcL in this paper because we nosperma, which have an angular inclusion. Unlike their re- judged it too divergent from other rbcL sequences of Aizo- sults, our data indicate that Limeum is sister to the rest of this aceae and hence doubtful). In Clement and Mabry's study (in clade (minus Stegnosperma), thus matching the distribution of press), in which all subfamilies were sampled (but Delosperma sieve-element plastids. Another observation made by Clement was not included), the family formed a grade at the base of and Mabry (in press) is that taxa possessing anthocyanins Phytolaccaceae/Nyctaginaceae. However, monophyly of Ai- (Molluginaceae and Caryophyllales) are present in both of the zoaceae has not generally been challenged in the literature two main clades within core Caryophyllales, thus showing that (e.g., Bittrich, 1993a). 140 AMERICAN JOURNAL OF BOTANY [Vol. 89

Fig. 3. Continued.

The position of Gisekia as sister to Aizoaceae (according to whereas, according to matK (in this paper), Sarcobatus is sister matK) is surprising because analyses using rbcL sequence in to Nyctaginaceae, and Agdestis is sister to Rivinoideae. The Manhart and Rettig (1994) indicated a strong af®nity with Ri- presence of betacyanins in Sarcobatus as reported here com- vina (Phytolaccaceae, Rivinoidae). Our combined rbcL/matK pares well with the phylogenetic pattern since this genus is analysis placed Gisekia within Rivinoideae, thereby indicating nested within betalain-producing families. that matK alone is possibly misleading with respect to this The three subfamilies of Phytolaccaceae (if one excludes taxon. Nyctaginaceae appear to be monophyletic, with the Agdestis and Barbeuia from this family) do not appear directly Nyctagineae (, , , , related to each other. The monocarpellate Rivinoideae (repre- Oxybaphus, Selinocarpus) forming a clade to the exclusion of sented here by , , Ledenbergia, Petiveria, Ri- (Bougainvilleae) and (Pisonieae). vina, and ) form a monophyletic group more closely Sarcobatus was separated from Chenopodiaceae by Behnke related to Nyctaginaceae (with which they share a monocar- (1997) on the basis of its sieve-element plastids. This genus pellate ) than to Phytolaccaceae s.s. Petiveria and has been shown by Downie, Katz-Downie, and Cho (1997) to Gallesia, which are sister in our analyses, share a strong garlic be associated with Nyctaginaceae/Phytolaccaceae s.l. based on smell (Rohwer, 1993). Phytolaccaceae s.s. (represented here the analysis of ORF2280 sequences. Clement and Mabry (in by Phytolacca and Ercilla) are sister to Nyctaginaceae/Rivi- press), using rbcL sequences, found that Sarcobatus and Ag- noideae/Sarcobatus/Agdestis, which is in agreement with the destis formed a clade sister to Nyctaginaceae-Rivinoideae, rbcL study of Clement and Mabry (in press). January 2002] CUEÂ NOUD ET AL.ÐMOLECULAR PHYLOGENETICS OF CARYOPHYLLALES 141

placed in Molluginaceae) and the South African genus Lo- phiocarpus (Phytolaccaceae, Microteoideae) was unpredicted. Rohwer (1993) emphasized the differences between the two genera from Microteoideae (Lophiocarpus and ) and the rest of Phytolaccaceae; however, he suggested af®nities with Chenopodiaceae (particularly Microtea, for which sieve- element plastids lack a central inclusion). Behnke (1994) showed that Lophiocarpus has the same type of plastid (i.e., P3 with a globular crystal) as some Molluginaceae; Corbi- chonia has not been studied in this respect. Clement and Ma- bry (in press) showed a relationship of Corbichonia with Ai- zoaceae/Nyctaginaceae/Phytolaccaceae in agreement with our study, and the rbcL/matK combined analysis gave a similar position to Corbichonia as with matK alone (no rbcL sequence was available for Lophiocarpus). The presence of betacyanins in Corbichonia, as reported in this paper, is an indication that this genus is probably misplaced among Molluginaceae. Be- cause we showed that Lophiocarpus also contains betacyanins, this might be considered as additional evidence for Corbi- chonia and Lophiocarpus to constitute a separate lineage that may ultimately be considered a new family. The clade formed by Cactaceae, Portulacaceae, Basellaceae, Didiereaceae, and Halophytum was named the ``succulent'' clade by Manhart and Rettig (1994). Non-DNA characters that de®ne it include the presence of an involucre (in Didiereaceae, Basellaceae, and Portulacaceae) and normal secondary growth (not always in Cactaceae). The paraphyly of Portulacaceae has long been suspected (see Carolin, 1987, 1993). However, the matK analysis suffers from a lack of support with respect to this group. Five portulacaceous genera (Calandrinia, Cistan- the, , Lewisia, and Montia) form a clade, all Didi- ereaceae genera are grouped with Calyptrotheca, Basella/An- redera are sometimes sister to , all Cactaceae sampled (Opuntia, Pereskia, Quiabentia, Rhipsalis, and Ta- cinga) are grouped together with Portulaca (but with BS Ͻ 50%), and Talinaria is sister to Anacampseros. Hence, the only supported evidence of paraphyly for Portulacaceae is the grouping of Calyptrotheca with Didiereaceae (68% BS), which corresponds to the ®ndings of Applequist and Wallace (2000), which were based on an rpl16/trnL-F/trnT-L combined analysis. The phylogenetic relationships of Didiereaceae as in- ferred from matK, with Alluaudiopsis as sister to the remaining three genera (which appear in an unresolved trichotomy), are identical to the results of Applequist and Wallace (2000). The grouping of the ®ve genera from Portulacaceae corre- sponds to what Hershkovitz (1993) identi®ed as the ``west American group'' on the basis of morphological characters (see also Hershkovitz and Zimmer, 2000). Moreover, the af- Fig. 4. Comparison of the total number of steps, CI (consistency index), ®nities of Talinaria and Anacampseros have been outlined by and RI (retention index) in the three codon positions for rbcL (black), atpB Hershkovitz (1993) and Hershkovitz and Zimmer (1997). (grey), and matK (hatched) based on the tree from the four-gene analysis Within Cactaceae, genera belonging to Opuntioideae (Opuntia, shown in Fig. 1. Tacinga, and Quiabentia) form a well-supported clade. The position of Halophytum in the portulacaceous clade is congru- ent with analyses of rbcL sequences that supported a sister Madagascan, monotypic Barbeuia was said by Rohwer group relationship of Halophytum with Basellaceae (Savolai- (1993) to be of ``completely obscure'' af®nities because of nen et al., 2000a). Such a relationship appears only in some divergent characters (i.e., semi-arillate seeds and capsular of the most parsimonious matK trees. Possible relatedness of fruits). Based on matK, Barbeuia is outside the Aizoaceae/ Halophytum to Basellaceae was suggested by Bittrich (1993b) Nytaginaceae/Phytolaccaceae clade, which indicates a rather because of their similar cuboidal grains and by Hun- isolated position for this taxon. More sequence data for Bar- ziker, Pozner, and Escobar (2000) because of their similar beuia would be of great interest, and, given its position, data chromosome numbers. about pigment nature would be of special relevance. According to our analysis, taxa formerly included in Chen- The sister group relationship of Corbichonia (previously opodiaceae (i.e., Atriplex, Axyris, , Blackiella, Chenopo- 142 AMERICAN JOURNAL OF BOTANY [Vol. 89 dium, Hablitzia, Kochia, Maireana, Rhagodia, Suaeda, and The west African endemic Triphyophyllum peltatum Airy Spinacia) form a monophyletic group in some of shortest trees Shaw (Dioncophyllaceae) provides a possible link between only, whereas genera formerly included in Amaranthaceae s.s. syndromes: it produces glandular, sticky hairs (in a similar (i.e., Achyranthes, Alternanthera, Amaranthus, Bosea, Celo- manner to Drosera and ) in its juvenile stage, sia, Froelichia, Gomphrena, Iresine, Psilotrichum, and Ptilo- whereas its adult stage, a with apical leaf-tendrils, is rem- tus) form a well-supported clade (97% BS). Using sequences iniscent of Nepenthes without pitchers (Albert, Williams, and of ORF 2280, Downie, Katz-Downie, and Cho (1997) found Chase, 1992). The evolution of carnivory in the Caryophyl- a lack of separation between the two families (although they lales group has been reviewed in detail in Meimberg et al. did not ®nd members of Amaranthaceae s.s. forming a clade), (2000). and for similar reasons APG (1998) included Chenopodiaceae Frankeniaceae are sister to Tamaricaceae, with which they within Amaranthaceae s.l. In addition, we found Beteae to be share many similarities; e.g., heath-like shrubby habit in some monophyletic (Beta and Hablitzia) and distinct from a clade species, capsular fruit, and straight embryo (Heywood, 1993), comprising various genera from Atripliceae/Chenopodieae but it is not possible to tell if one family is nested in the other (Atriplex, Axyris, Blackiella, Chenopodium, Rhagodia, and one with only the current information in which one genus was Spinacia). The association of Amaranthaceae s.l. with the Neo- sampled per family. The proximity of Plumbaginaceae to Poly- tropical family Achatocarpaceae ( and Phaulo- gonaceae is well supported, and the clade they form is the best thamnus) agrees with Manhart and Rettig's results (1994). resolved of all analyses. Within Plumbaginaceae, the dichot- Although Caryophyllaceae are placed as sister to the rest of omy between Plumbaginoideae (Dyerophyton, Plumbago, and the core Caryophyllales in the matK analysis, they are more Ceratostigma) and Staticoideae (Afrolimon, Armeria, Goniol- likely to be part of the Amaranthaceae/Achatocarpaceae clade, imon, Limoniastrum, Limonium, and Psylliostachys) is clearly as found in the combined analyses (Fig. 1). Based on the gen- established, as was already shown based on rbcL (Lledo et al., era we sampled, the separation into three subfamilies is not 1998). The situation is less clear in Polygonaceae, in which supported, with Paronychoideae (here represented by Tele- not all of the groupings we found correspond to previously phium, sometimes included in Molluginaceae) as sister to recognized entities. Triplareae (genera Triplaris and Ruprech- mixed members of Alsinoideae (Honckenya and Moehringia) tia) appear to be well supported, as do Rumiceae (Emex, Ru- and Caryophylloideae (Lychnis, Silene, Vaccaria, Saponaria, mex, Oxyria, and Rheum), whereas Persicarieae (Koenigia, and Gypsophila). Persicaria, Bistorta, and Fagopyrum), Coccolobeae (Brunni- Because Molluginaceae lack clear synapomorphies (Bittrich, chia, Coccoloba, and Muehlenbeckia), and the pentamerous- 1993a; Endress and Bittrich, 1993), Clement and Mabry (in ¯owered Polygoneae (Polygonum and Fallopia) are poly- or press) wrote that ``there is no obvious reason to expect the paraphyletic. family to appear as a clade.'' In the matK alone analysis, the Despite the fact that we used only a small portion of the six genera we sampled to represent this family appeared in matK exon, this DNA region contains enough variable sites to three different positions, showing no apparent congruence with resolve many infrafamilial relationships within Caryophylla- the classi®cation of Hofmann (1973). Adenogramma, Glis- les, its high variability corresponding to high phylogenetic chrothamnus, Glinus, Pharnaceum, and Suessenguthiella are content. However, although a great deal of progress has been sister to the ``succulent'' clade and may constitute the best made lately, the de®nition of all natural entities within Cary- candidate for a rede®ned, monophyletic Molluginaceae. Cor- ophyllales is not yet fully satisfactory. This phylogenetic study bichonia is sister to Lophiocarpus, and both together are sister presents evidence for the recognition of Agdestidaceae, Bar- to Aizoaceae/Phytolaccaceae/Nyctaginaceae. Limeum is sister beuiaceae, Petiveriaceae, and Sarcobataceae, and Limeum, and to the ``globular inclusion'' clade in the combined four-gene the pair Corbichonia/Lophiocarpus may be ultimately recog- analyses (Fig. 1), whereas its position is unresolved using nized as separate families. The genus Halophytum is de®nitely matK alone (Fig. 3c). Since Limeum possesses a different type not part of Amaranthaceae, but it is dif®cult to determine if it of sieve-element plastid inclusion, it represents most probably deserves familial status or should be included, for example, in a separate lineage from the remaining Molluginaceae and Basellaceae. The systematics of the ``succulent'' clade are es- could deserve familial status. The Australian genus Macarthu- pecially in need of revision, with perhaps the recognition of ria possesses the same type of sieve-element plastid (Behnke, an expanded family Cactaceae that could include portulaca- 1994) and might be sister to Limeum, but we have no DNA ceous genera such as Portulaca, Anacampseros,orTalinum sequence data for this genus yet. (Hershkowitz and Zimmer, 1997). Molluginaceae, excluding Limeum and Corbichonia, appear well de®ned, but some gen- Noncore CaryophyllalesÐThe monophyly of this group, al- era (e.g., and Polpoda) have not been sampled though not supported by some studies based on rbcL (e.g., yet for DNA work. Savolainen et al., 2000a), is well established by the combined Ehrendorfer (1976) provided an hypothesis to explain the analysis. Diagnostic characters for this clade include: secretory presence of betalains in Caryophyllales. According to his cells containing plumbagin, stalked, gland-headed hairs, basal ideas, ancestral taxa of the group may have evolved in dry placentation, and starchy endosperm (Judd et al., 1999), al- environments on mineral substrates with a tendency towards though some of these have been lost in one or more lineages. anemophily; having lost the need to attract pollinators, they All carnivorous families of Caryophyllales (Droseraceae, would have also lost the ability to synthesize anthocyanin pig- Drosophyllaceae, Nepenthaceae, Dioncophyllaceae), along ments. If this is true, betalain synthesis may have evolved with Ancistrocladaceae, form a monophyletic group based on during subsequent reversals to zoochory. Interestingly, Li- the matK analysis. All carnivorous syndromes except for the meum, which is unpigmented (Clement et al., 1994), comes as bladders found in Utricularia are present in this group (¯y- sister to the ``globular inclusion'' clade that produces both be- paper traps in Drosera, Drosophyllum, and Triphyophyllum, talain and anthocyanin, but the anthocyanin-producing taxa are traps in Aldrovanda and Dionaea, and pitchers in Nepenthes). nested among betalain-producing ones. Another interesting January 2002] CUE NOUD ET AL.ÐMOLECULAR PHYLOGENETICS OF CARYOPHYLLALES 143 case is that of Caryophyllaceae (producing anthocyanins) that DOYLE,J.J.,AND J. L. DOYLE. 1987. A rapid DNA isolation procedure for are associated with Amaranthaceae (producing betalains) and small quantities of fresh leaf tissue. Phytochemical Bulletin 19: 11±15. Achatocarpaceae, for which the nature of pigments is not EHRENDORFER, F. 1976. Closing remarks: systematics and evolution of cen- trospermous families. Systematics and Evolution 126: 99±106. known (Clement et al., 1994). Although we are far from fully EICHLER, A. W. 1876. Syllabus der Vorlesungen uber Phanerogamenkunde. understanding the evolution of all peculiarities in Caryophyl- Schwer'sche Buchhandlung, Kiel, Germany. lales, phylogenetic data are increasing rapidly and will lead to ENDRESS,M.E.,AND V. B ITTRICH. 1993. Molluginaceae. In K. Kubitzki, J. substantial progress in the near future. As Ehrendorfer (1976) G. Rohwer, and V. Bittrich [eds.], The families and genera of vascular put it, ``the evolutionary picture emerging for the Centrosper- plants, vol. II, Magnoliid, hamamelid, and caryophyllid families, 419± mae families differentiating and radiating in time and space is 426. Springer Verlag, Berlin, Germany. one of the most fascinating in the plant kingdom.'' FAY, M. F., K. M. CAMERON,G.T.PRANCE,M.D.LLEDO , AND M. W. CHASE. 1997. Familial relationships of Rhabdodendron (Rhabdodendraceae): plastid rbcL sequences indicate a caryophyllid placement. Kew Bulletin LITERATURE CITED 52: 923±932. GIANNASI, D. E., G. ZURAWSKI,G.LEARN, AND M. T. CLEGG. 1992. Evo- lutionary relationships of the Caryophyllidae based on comparative rbcL ALBERT, V. A., S. E. WILLIAMS, AND M. W. CHASE. 1992. Carnivorous sequences. Systematic Botany 17: 1±15. plants: phylogeny and structural evolution. Science 257: 1491±1495. HARBORNE, J. B. 1998. Phytochemical methods. Kluwer, Dordrecht, The APG (ANGIOSPERM PHYLOGENY GROUP). 1998. An ordinal classi®cation for Netherlands. the families of ¯owering plants. Annals of the Missouri Botanical Garden HERSHKOVITZ, M. A. 1989. Phylogenetic studies in Centrospermae: a brief 85: 531±553. appraisal. Taxon 38: 602±610. APPLEQUIST,W.L.,AND R. S. WALLACE. 2000. Phylogeny of the Madagas- HERSHKOVITZ, M. A. 1993. Revised circumscriptions and subgeneric taxon- can family Didiereaceae. Plant Systematics and Evolution 221: 157±166. omies of Calandrinia, and Montiopsis (Portulacaceae) with notes on phy- BAKKER, F. T., D. D. VASSILIADES,C.M.MORTON, AND V. S AVOLAINEN. logeny of the portulacaceous alliance. Annals of the Missouri Botanic 1998. Phylogenetic relationships of Biebersteinia Stephen (Geraniaceae) Garden 80: 333±365. inferred from rbcL and atpB sequence comparisons. Botanical Journal HERSHKOVITZ,M.A.,AND E. A. ZIMMER. 1997. On the evolutionary origin of the Linnean Society 127: 149±158. of the cacti. Taxon 46: 217±232. BEHNKE, H.-D. 1994. Sieve-element plastids: their signi®cance for the evo- HERSHKOVITZ,M.A.,AND E. A. ZIMMER. 2000. Ribosomal DNA evidence lution and systematics of the order. In H.-D. Behnke and T. J. Mabry and disjunctions of western American Portulacaceae. Molecular Phylo- [eds.], Caryophyllales: evolution and systematics, 87±121. Springer Ver- genetics and Evolution 15: 419±439. lag, Berlin, Germany. HEYWOOD, V. H. 1993. Flowering plants of the world. B. T. Batsford, Lon- BEHNKE, H.-D. 1997. SarcobataceaeÐa new family of Caryophyllales. Taxon don, UK. 46: 495±507. HILLIS, D. M. 1996. Inferring complex phylogenies. Nature 383: 130±131. BITTRICH, V. 1993a. Introduction to Centrospermae. In K. Kubitzki, J. G. Rohwer, and V. Bittrich [eds.], The families and genera of vascular HIRATSUKA, J., ET AL. 1989. The complete sequence of the rice (Oryza plants, vol. II, Magnoliid, hamamelid, and caryophyllid families, 13±19. sativa) chloroplast genome: intermolecular recombination between dis- Springer Verlag, Berlin, Germany. tinct tRNA genes accounts for a major plastid DNA inversion during the evolution of the cereals. Molecular Genes and Genetics 217: 185±194. BITTRICH, V. 1993b. Halophytaceae. In K. Kubitzki, J. G. Rohwer, and V. Bittrich [eds.], The families and genera of vascular plants, vol. II, Mag- HOFMANN, U. 1973. Morphologische Untersuchungen zur Umgrenzung und noliid, hamamelid, and caryophyllid families, 320±321. Springer Verlag, Gliederung der Aizoaceen. Botanische Jahrbucher fur Systematik, P¯an- Berlin, Germany. zengeschichte und Planzengeographie 93: 247±324. OOT AGALLO N AND RANE BRAUN, A. 1864. Uebersicht der naturlichen Systems nach der Anordnung H , S. B., S. M , P. R. C . 1999. Phylogeny of basal derselben. In P. Ascherson [ed.], Flora der Provinz Brandenburg, der based on three molecular data sets: atpB, rbcL, and 18S nuclear Altmark und des Herzogthums Magdeburg, vol. 1, 22±67. Hirschwald, ribosomal DNA sequences. Annals of the Missouri Botanic Garden 86: Berlin, Germany. 1±32. CAROLIN, R. 1987. A review of the family Portulacaceae. Australian Journal HUDSON, G. S., J. D. MAHON,P.A.ANDERSON,M.J.GIBBS,M.R.BADGER, of Botany 35: 383±412. T. J. ANDREWS, AND P. R. WHITFELD. 1990. Comparisons of rbcL genes CAROLIN, R. C. 1993. Portulacaceae. In K. Kubitzki, J. G. Rohwer, and V. for the large subunit of ribulose-bisphosphate carboxylase from closely Bittrich [eds.], The families and genera of vascular plants, vol. II, Mag- related C3 and C4 plant species. Journal of Biological Chemistry 265: noliid, hamamelid, and caryophyllid families, 544±555. Springer Verlag, 808±814. Berlin, Germany. HUELSENBECK, J. P., J. J. BULL, AND C. W. CUNNINGHAM. 1996. Combining CHASE,M.W.,ET AL. 1993. Phylogenetics of seed plants: an analysis of data in phylogenetic analysis. Trends in Ecology and Evolution 11: 152± nucleotide sequences from the plastid gene rbcL. Annals of the Missouri 158. Botanical Garden 80: 528±580. HUNZIKER, J. H., R. POZNER, AND A. ESCOBAR. 2000. Chromosome number CLEMENT, J.-S, AND T. J. MABRY. In press. rbcL sequence data and pigment in Halophytum ameghinoi (Halophytaceae). Plant Systematics and Evo- evolution in the Caryophyllales: non-monophyly of the betalain-produc- lution 221: 125±127. ing plants. American Journal of Botany. JUDD, W. S., C. S. CAMPBELL,E.A.KELLOGG, AND P. F. S TEVENS. 1999. CLEMENT, J.-S., T. J. MABRY,H.WYLER, AND A. S. DREIDING. 1994. Chem- Plant systematics: a phylogenetic approach. Sinauer, Sunderland, Mas- ical review and evolutionary signi®cance of betalains. In H.-D. Behnke sachusetts, USA. and T. J. Mabry [eds.], Caryophyllales: evolution and systematics, 247± KOBAYASHI, N., J. SCHMIDT,M.NIMTZ,V.WRAY, AND W. S CHLIEMANN. 261. Springer Verlag, Berlin, Germany. 2000. Betalains from Christmas cactus. Phytochemistry 54: 419±426. CRONQUIST, A. 1981. An integrated system of classi®cation of ¯owering KUBITZKI, K., J. G. ROHWER, AND V. B ITTRICH [EDS.]. 1993. The families plants. Columbia University Press, New York, New York, USA. and genera of vascular plants, vol. II, Magnoliid, hamamelid, and cary- CRONQUIST, A., AND R. F. THORNE. 1994. Nomenclatural and taxonomic ophyllid families. Springer Verlag, Berlin, Germany. history. In H.-D. Behnke and T. J. Mabry [eds.], Caryophyllales: evo- LLEDO , M. D., M. B. CRESPO,K.M.CAMERON,M.F.FAY, AND M. W. lution and systematics, 87±121. Springer Verlag, Berlin, Germany. CHASE. 1998. Systematics of Plumbaginaceae based upon cladistic anal- DOWNIE, S. R., D. S. KATZ-DOWNIE, AND K.-Y. CHO. 1997. Relationships in ysis of rbcL sequence data. Systematic Botany 23: 21±29. the Caryophyllales as suggested by phylogenetic analyses of partial chlo- LLEDO , M. D., M. B. CRESPO,A.V.COX,M.F.FAY, AND M. W. CHASE. roplast DNA ORF 2280 homolog sequences. American Journal of Botany 2000. Polyphyly of Limoniastrum (Plumbaginaceae): evidence from 84: 253±273. DNA sequences of plastid rbcL, trnL intron and trnL-F intergene spacer. DOWNIE,S.R.,AND J. D. PALMER. 1994. A chloroplast DNA phylogeny of Botanical Journal of the Linnean Society 132: 175±191. the Caryophyllales based on structural and inverted repeat restriction MABBERLEY, D. J. 1997. The plant-book, 2nd ed. Cambridge University sites variation. Systematic Botany 19: 236±252. Press, Cambridge, UK. 144 AMERICAN JOURNAL OF BOTANY [Vol. 89

MADDISON,W.P.,AND D. R. MADDISON. 1992. MacClade. Sinauer, Sunder- SAVOLAINEN, V., R. SPICHIGER, AND J.-F. MANEN. 1997. Polyphyletism of land, Massachusetts, USA. Celastrales deduced from a chloroplast noncoding DNA region. Molec- MANHART, J. R., AND J. H. RETTIG. 1994. Gene sequence data. In H.-D. ular Phylogenetics and Evolution 7: 145±157. Behnke and T. J. Mabry [eds.], Caryophyllales: evolution and system- SCHMITZ-LINNEWEBER, C., R. M. MAIER,J.P.ALCARAZ,A.COTTET,R.G. atics, 235±246. Springer Verlag, Berlin, Germany. HERRMANN, AND R. MACHE. 2001. The plastid chromosome of MEIMBERG, H., P. DITTRICH,G.BRINGMANN,J.SCHLAUER, AND G. HEUBL. (Spinacia oleracea): complete nucleotide sequence and gene organiza- 2000. Molecular phylogeny of Caryophyllidae s.l. based on matK se- tion. Plant Molecular Biology 45: 307±315. quences with special emphasis on carnivorous taxa. Plant Biology 2: SEELANAN, T., A. SCHNABEL, AND J. F. WENDEL. 1997. Congruence and 121±252. consensus in the cotton tribe. Systematic Botany 22: 259±290. MORGAN,D.R.,AND D. E. SOLTIS. 1993. Phylogenetic relationships among SOLTIS,D.E.,AND P. S. SOLTIS. 1997. Phylogenetic relationships in Saxifra- Saxifragaceae s. l. based on rbcL sequence data. Annals of the Missouri gaceae sensu lato: a comparison of topologies based on 18S rDNA and Botanical Garden 80: 631±660. rbcL sequences. American Journal of Botany 84: 504±522. MORTON, C. M., K. G. KAROL, AND M. W. CHASE. 1997. Taxonomic af®n- SOLTIS, P. S., D. E. SOLTIS, AND M. W. CHASE. 1999. Angiosperm phylogeny ities of Physena (Physenaceae) and Asteropeia (). Botanical Re- inferred from multiple genes as a tool for comparative biology. Nature view 63: 231±239. 402: 402±404. NAKAI, T. 1942. Notulae ad plantas Asiae orientalis. XVIII. Journal of Jap- SOLTIS,D.E.,ET AL. 1997. Angiosperm phylogeny inferred from 18S ri- anese Botany 18: 91±120. bosomal DNA sequences. Annals of the Missouri Botanical Garden 84: NANDI, O. I., M. W. CHASE, AND P. E. ENDRESS. 1998. A combined cladistic 1±49. analysis of angiosperms using rbcL and non-molecular data sets. Annals SOLTIS,D.E.,ET AL. 2000. Angiosperm phylogeny inferred from 18S rDNA, of the Missouri Botanic Garden 85: 137±212. rbcL, and atpB sequences. Botanical Journal of the Linnean Society 133: NICKRENT,D.L.,AND D. E. SOLTIS. 1995. A comparison of angiosperm 381±461. phylogenies from nuclear 18S rDNA and rbcL sequences. Annals of the SWOFFORD, D. L. 1998. PAUP*: phylogenetic analysis using parsimony (* Missouri Botanical Garden 82: 208±234. and other methods), version 4.0.b2±8. Sinauer, Sunderland, Massachu- REEVES, G., M. W. CHASE,P.GOLDBLATT,P.RUDALL,M.F.FAY,A.V.COX, setts, USA. B. LEJEUNE, AND T. S OUZA-CHIES. 2001. Molecular systematics of Iri- WILLIAMS, S. E., V. A. ALBERT, AND M. W. CHASE. 1994. Relationships of daceae: evidence from four plastid DNA regions. American Journal of Droseraceae: a cladistic analysis of rbcL sequences and morphological Botany 88: 2074±2087. data. American Journal of Botany 81: 1027±1037. RETTIG, J. H., H. D. WILSON, AND J. R. MANHART. 1992. Phylogeny of the XIANG, Q.-Y., D. E. SOLTIS,D.R.MORGAN, AND P. S. SOLTIS. 1993. Phy- CaryophyllalesÐgene sequence data. Taxon 41: 201±209. logenetic relationships of Cornus L. sensu lato and putative relatives RODMAN, J. E., M. K. OLIVER,R.R.NAKAMURA,J.U.J.MCCLAMMER, AND inferred from rbcL sequence data. Annals of the Missouri Botanical Gar- A. H. BLEDSOE. 1984. A taxonomic analysis and revised classi®cation den 80: 723±734. of Centrospermae. Systematic Botany 9: 297±323. YASUI,Y.,AND O. OHNISHI. 1996. Comparative study of rbcL gene sequences RODMAN, J. E., R. A. PRICE,K.G.KAROL,E.CONTI,K.J.SYTSMA, AND J. in Fagopyrum and related taxa. Genes and Genetic Systems 71: 219± D. PALMER. 1993. Nucleotide sequences of the rbcL gene indicate 224. monophyly of mustard oil plants. Annals of the Missouri Botanical Gar- YASUI,Y.,AND O. OHNISHI. 1998. Interspeci®c relationships in Fagopyrum den 80: 686±699. (Polygonaceae) revealed by the nucleotide sequences of the rbcL and ROHWER, J. G. 1993. Phytolaccaceae. In K. Kubitzki, J. G. Rohwer, and V. accD genes and their intergenic region. American Journal of Botany 85: Bittrich [eds.], The families and genera of vascular plants, vol. II, Mag- 1134±1142. noliid, hamamelid, and caryophyllid families, 506±515. Springer Verlag, ZURAWSKI, G. R., W. BOTTOMLEY, AND P. R. WHITFELD. 1982. Structures of Berlin, Germany. the genes for the beta and epsilon subunits of spinach chloroplast ATPase SAVOLAINEN,V.,ET AL. 2000a. Phylogeny of the eudicots: a nearly complete indicate a dicistronic mRNA and an overlapping translation stop/start familial analysis based on rbcL gene sequences. Kew Bulletin 55: 257± signal. Proceedings of the National Academy of Sciences, USA 79: 6260± 309. 6264. SAVOLAINEN, V., M. W. CHASE,S.B.HOOT,C.M.MORTON,D.E.SOLTIS, ZURAWSKI, G. R., B. PERROT,W.BOTTOMLEY, AND P. R. WHITFIELD. 1981. C. BAYER,M.M.FAY,A.Y.DE BRUIJN,S.SULLIVAN, AND Y.-L. QIU. The structure of the gene for the large subunit of ribulose-1,5-bisphos- 2000b. Phylogenetics of ¯owering plants based on combined analysis of phate-carboxylase from spinach chloroplast DNA. Nucleic Acids Re- plastid atpB and rbcL gene sequences. Systematic Biology 49: 306±362. search 9: 3251±3270.