Saxifragaceae sensu lato

Transcription

1 Proc. Nati. Acad. Sci. USA Vol. 87, pp , June 1990 Evolution rbcl sequence divergence and phylogenetic relationships in Saxifragaceae sensu lato (DNA sequencing/evolution/systematics) DOUGLAS E. SOLTISt, PAMELA S. SOLTISt, MICHAEL T. CLEGGt, AND MARY DURBINt tdepartment of Botany, Washington State University, Pullman, WA 99164; and tdepartment of Botany and Plant Sciences, University of California, Riverside, CA Communicated by R. W. Allard, March 19, 1990 (received for review January 29, 1990) ABSTRACT Phylogenetic relationships are often poorly understood at higher taxonomic levels (family and above) despite intensive morphological analysis. An excellent example is Saxifragaceae sensu lato, which represents one of the major phylogenetic problems in angiosperms at higher taxonomic levels. As originally defined, the family is a heterogeneous assemblage of herbaceous and woody taxa comprising 15 subfamilies. Although more recent classifications fundamentally modified this scheme, little agreement exists regarding the circumscription, taxonomic rank, or relationships of these subfamilies. The recurrent discrepancies in taxonomic treatments of the Saxifragaceae prompted an investigation of the power of chloroplast gene sequences to resolve phylogenetic relationships within this family and between the Saxifragaceae and other major plant lineages. Sequence data from the gene rbcl (ribulose-1,5-bisphosphate carboxylase, large subunit) reveal that (i) Saxifragaceae sensu lato is at least paraphyletic, and probably polyphyletic, (ii) the genera Parnassia and Brexia are only distantly related to other members of Saxifragaceae, and (iii) representatives of the Solanaceae (subclass Asteridae) appear more closely related to Saxifragaceae (subclass Rosidae) than traditionally maintained. These data illustrate the value of chloroplast gene sequence data in resolving genetic, and hence phylogenetic, relationships among members of the most taxonomically complex groups. In many groups of angiosperms phylogenetic relationships at higher taxonomic levels (family and above) have remained enigmatic despite detailed morphological analyses. Presumably, these difficulties reflect the more frequent occurrence of parallel and convergent character state change (homoplasy) at higher taxonomic ranks. Below the familial level, restriction site analysis of chloroplast DNA has been exceptionally valuable in resolving phylogenetic affinities (1, 2), even when extreme morphological divergence has obscured a close relationship (3). At higher taxonomic levels, however, restriction site analysis typically is inadequate for phylogenetic inference, owing to excessive homoplasy and length mutation; DNA sequencing appears to be the tool of choice in such instances (2, 4, 5). Sequencing of the slowly evolving ribosomal RNA genes, for example, has unequivocally demonstrated that chloroplasts are of endosymbiotic origin (33). In plants, DNA sequence comparisons of chloroplastencoded genes will be particularly useful in phylogenetic analyses at higher taxonomic levels (4, 5). Although few major studies of chloroplast gene sequences and phylogenetic relationships have been published (6, 7), the chloroplast gene rbcl, encoding the large subunit of ribulose-1,5-bisphosphate carboxylase, has emerged as the preferred gene for examining higher-level phylogenetic relationships (2, 4). The reasons for this preference include (i) rbcl has been widely se- The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C solely to indicate this fact. quenced and analyses to date indicate that it is reliable for phylogenetic analysis at higher taxonomic levels, (ii) rbcl is a large gene [>1400 base pairs (bp)] that provides numerous characters (bp) for phylogenetic studies, and (iii) the rate of evolution of rbcl is appropriate for addressing questions of angiosperm phylogeny at the familial level or higher. We used rbcl sequence data to analyze phylogenetic relationships in a particularly problematic group-engler's (8) broadly defined family Saxifragaceae (Saxifragaceae sensu lato). Based on morphological analyses, the group is almost impossible to distinguish or characterize clearly and represents one of the greatest taxonomic problems at higher levels in the angiosperms (9, 10). Following the traditional interpretation (8), the family is a large, morphologically diverse assemblage of annual, biennial, and perennial herbs, shrubs, trees, and vines comprising 15 subfamilies; this was later increased to 17 subfamilies (11). The morphological diversity encompassed by this group is so great that subsequent workers have provided substantially modified concepts of relationships (9, 12-16). However, these recent schemes often differ dramatically; little agreement exists regarding the circumscription, taxonomic rank, or relationships of Engler's subfamilies. The differences among taxonomic treatments are too substantial to review here (see reviews in refs. 10 and 12). We illustrate the magnitude of these discrepancies, however, in a simplified comparison of relationships (Fig. 1). Most recent workers have considered the Englerian concept of Saxifragaceae too broad. Takhtajan (15), for example, distributed the same taxa among more than 10 families and suggested that some members of Saxifragaceae sensu lato were only very distantly related. He placed most families in two orders (Cunoniales and Saxifragales, superorder Rosanae; subclass Rosidae) and considered these to be distantly related to Brexiaceae (Celastrales, superorder Celastranae, subclass Rosidae) and Hydrangeaceae (Hydrangeales, superorder Cornanae; subclass Asteridae). His Saxifragaceae are essentially limited to Engler's tribe Saxifrageae. In contrast, Cronquist (12) placed several of Engler's woody subfamilies in Grossulariaceae (Fig. 1); he maintained that woody taxa (his Grossulariaceae and Hydrangeaceae) were more usefully associated with other woody families of Rosidae. Cronquist's Saxifragaceae encompasses several of Engler's original herbaceous subfamilies. To understand completely the phylogenetic affinities of all members of Saxifragaceae sensu lato would require analysis of a large portion of Rosidae, as well as taxa in other subclasses. Here we concentrate on a suite of taxa that represent well the morphological diversity of the group and ask: can rbcl sequence data elucidate phylogenetic relationships in this enigmatic group? If so, this tool could be used to unravel the evolutionary history of other equally problematic groups of plants. 4640

2 Takhtajan (1987) ROSANAE Engler (1964) SAXIFRAGACEAE Cronquist (1981) ROSALES CUNONIALES BAUERACEAE Penthoroideae ---, SAXIFRAGACEAE SAXIFRAGALES / / Saxifragoideae PENTHORACEAE Ae Vahlioideae, SAXIFRAGACEAE Al Francooideae VAHUACEAE, / Eremosynoideae FRANCOACEAE 0 Lepuropetaloideae EREMOSYNACEAE Pamassioideae LEPUROPETALACEAE Baueroideae * CUNONIACEAE PARNASSIACEAE Ribesoideae GROSSULARIACEAE GROSSULARIACEAE Pterostemonoideae ITEACEAE * Hydrangeoideae HYDRANGEACEAE CELASTRANAE CELASTRALES Evolution: Soltis et al. BREXIACEAE CORNANAE HYDRANGEALES PFEROSTEMONACEAE HYDRANGEACEAE W TETRACARPACEAE ESCALLONIACEAE MONTINIACEAE DULONGIACEAE Tetracarpaeoideae gegoideae Iteoideae ~~Brexioideae Escallonioideae Montinioideae Phyllonomoideae FIG. 1. Schematic representation of relationships among taxa of Saxifragaceae sensu lato depicting some of the discrepancies that exist among the traditional (11) and more recent (12, 14) taxonomic treatments. MATERIALS AND METHODS We obtained rbcl sequence data for the following members of Saxifragaceae sensu lato (see Fig. 1): Astilbe taquetii (Saxifragoideae/Saxifrageae), Brexia madagascariensis (Brexioideae), Carpenteria californica (Hydrangeoideae), Francoa sonchifolia (Saxifragoideae/Francoeae), Heuchera micrantha (Saxifragoideae/Saxifrageae), Itea virginica (Iteoideae), Penthorum sedoides (Penthoroideae), and Parnassia fimbriata (Parnassioideae). Astilbe, Heuchera, Penthorum, Francoa, and Parnassia are herbaceous; Brexia, Itea, and Carpenteria are woody. Total DNAs were isolated (17, 18) and digested with EcoRI. Genomic libraries were constructed by using the Lambda ZAP II/EcoRI cloning kit (Stratagene) and Gigapack packaging extracts (Vector Cloning Systems). The genomic libraries (unamplified) were screened (19) with an rbcl probe from Cenchrus. Recombinant phage containing DNA inserts that hybridized to the rbcl probe were subsequently cloned into the plasmid pbluescript following the in vivo excision method provided by Stratagene. We used a double-stranded dideoxynucleotide sequencing protocol (20, 21) and used a series of synthetic oligonucleotide primers based on the rbcl sequence of Zea mays (obtained from G. Zurawski, DNAX). Sequence data were verified by sequencing both strands of the rbcl clones. rbcl sequences were compared by three methods: (i) unweighted pair group method of analysis (UPGMA; ref. 22), (ii) a maximum-likelihood procedure developed for nucleotide sequence data (23), (iii) Wagner parsimony. For UPGMA we used a bootstrap procedure with 500 replicates. Several investigators (24-26) have reviewed the methods of analyzing sequence data. We also performed relative rate tests (27) to determine whether rbcl evolves in a clock-like fashion among the taxa sequenced. We included in these analyses previously published rbcl sequences from taxa that show varying degrees of relatedness to Saxifragaceae (based on refs ): Z. mays (Poaceae) Proc. Nati. Acad. Sci. USA 87 (1990) 4641 (28), a monocotyledon, distantly related to the other taxa, which are all dicotyledons; Nicotiana otophora (Solanaceae, subclass Asteridae) (29); and Pisum sativum (Fabaceae, subclass Rosidae, but placed in a different order than Saxifragaceae) (30). These species served as outgroups in the parsimony analysis. UPGMA dendrograms were calculated from distance matrices based on the 3ST model (31). Maximum-likelihood and parsimony analyses used the DNAML and DNAPARS programs, respectively, provided in PHYLIP 3.2 (J. Felsenstein); the DNABOOT program (PHYLIP 3.1) was used to obtain confidence levels for the parsimony analysis. For the DNAML analysis, the transition/transversion ratio was set to 2.0 and the empirical frequencies of the bases were used (F option); runs were conducted with and without Z. mays and the categories option (C). Because of the length of time needed to run the program with Z. mays and the C option, these were not included when the jumble option (J) was used. RESULTS AND DISCUSSION rbcl in Saxifragaceae sensu lato is 1407 bp long, comparable to the size reported for other taxa (5, 7). Sequences for the eight taxa analyzed are given in Fig. 2. Due to EcoRI* activity, we did not obtain the full length of the rbcl gene from Francoa (our sequence for this taxon begins 471 bp from the 5' end of the gene). Thus, we have sequenced approximately two-thirds of rbcl from Francoa. Detailed sequence comparisons (4) indicate, however, that use of a partial gene sequence (particularly one this large) should not affect phylogenetic reconstruction. In fact, we obtained the same results when we used only the last two-thirds of the rbcl gene for all taxa in a parsimony analysis. A matrix of 3ST distances with standard errors for each codon position is provided in Table 1. As expected (4, 5), most of the base substitutions are in the third codon position. Relative rate tests provided no evidence for differential rates of evolution of rbcl in any of the lineages analyzed. These results are similar to those of earlier studies (6, 7). In 100% of the UPGMA analyses, (i) Nicotiana (Solanaceae) clustered more closely with Saxifragaceae than did Pisum (Fabaceae); (ii) Heuchera and Astilbe appeared as closest relatives; (iii) Penthorum, Francoa, Carpenteria, and Itea clustered with Astilbe and Heuchera; and (iv) Brexia and Parnassia formed a distinct cluster well separated from the cluster containing Penthorum, Francoa, Carpenteria, Itea, Astilbe, and Heuchera. In 45% of the analyses, Nicotiana (Solanaceae) actually appeared between the Brexia-Parnassia cluster and the Penthorum-Francoa-Carpenteria-Itea- Heuchera-Astilbe cluster. In 94% of the analyses, Itea appeared as the closest relative of Astilbe and Heuchera. The differences among the dendrograms mainly involved the alteration of relationships of Penthorum, Carpenteria, Francoa, and Itea to Heuchera and Astilbe. Parsimony analysis yielded two equally most-parsimonious trees that differ only in the placement of Francoa. One tree places Francoa in a clade with Itea, Penthorum, Astilbe, and Heuchera; the other places Francoa as the sister species of a large clade having two groups: Itea, Penthorum, Astilbe, and Heuchera in one and Carpenteria and Nicotiana in the other. We also analyzed the data by using 100 bootstrap replicates; the tree obtained (Fig. 3) provides confidence intervals for the nodes. The maximum-likelihood solution yielded a topology (Fig. 4) that differed from the parsimony tree only in the positions ofpenthorum and Itea relative to Heuchera and Astilbe. The maximum-likelihood solution is also very similar to the UPGMA dendrogram that appeared most frequently. Use of the jumble option (but with Zea omitted to shorten run times) yielded several different solutions with similar likelihood

3 4642 Evolution: Soltis et al. Proc. Natl. Acad. Sci. USA 87 (1990) NUCL HE AS PE PA BR CA IT FR 0015 C 8027 G C 8042 G G 0045 C 0048 C C 8057 T T T T T T T 0060 T T T 0068 A 0084 T T C C C C C 0087 T T T T T T T 0088 G G G G G 8095 C 0096 A A A A A A 0108 C C C C C C C 0111 A T 0124 A 0126 G 0138 A A 0141 G 0147 G G 0153 G G G G 0159 G G C 0162 A 0165 A A T G 0168 T 0177 T T T T T T 0186 C 0195 C 0207 C G G G G G G 0213 T 0219 G 0227 A 0228 T T 0231 G 0237 G 0243 G 0246 A A A A A A 0235 T T T T 0257 A A A A A A A A 0266 C C C C C C C 0267 C C C C C C 0270 G 0272 C C C C C C C 0280 G G G G G C G 0283 A A A A A A A 0284 G G 02S0 T T C 0293 C T T 0309 T T 0312 C C C A C 0324 C 0342 T T T T G T T 0345 T 0357 T 0369 T T T T T T T 0372 C C 0378 A A 0393 T T A 0405 A A 0408 G G C C G C G 0412 T T 0414 T 0423 C C 0424 G G G G A 0425 T T T 0433 T T T 0434 C C C 0444 T 0443 G 0450 G C G C C C C C 0456 G G 0439 T T T T T T T T 0462 C C T T C C 0468 G 0474 G G G G G G G 0483 G 0489 T 0498 A A G C 0501 C C 0504 T T 0307 A A A 0310 A A A A A 0513 C 0332 C 0537 C A A 0543 T T T T 0549 G G G G G G G 0552 T T 0555 T T 0564 A A A C A G G A 0579 C A A C C C C A 0582 T T T T T 0394 C C C 0600 C C C C C C C 0612 A A A A A 0628 A A 0624 C C C C C C C T 0648 C C C C C C C 0663 C 0666 A T A 0642 T T G T G 0673 A A A 0677 T 0684 T 0687 A A A A 0688 G G G C G G G 0696 G 070S T 0711 A A 0717 T G G G G G G C G G 0738 G 0751 C C 0753 G G G G NUCL HE AS 0757 PE PA C BR CA IT FR Heuchera scores. These solutions and always support Astilbe the following: are closely allied, (i) (it) Brexia and Par G G G G G G 0762 A A A nassia always appear together and are distantly related to 0763 A C A other Saxifragaceae sensu lato, (iii) Nicotiana is more similar 0771 C C C C C C 0778 C to Saxifragaceae than is Pisum. These solutions differ in their A 0783 placement of Carpenteria, Itea, Francoa, and Penthorum 0785 C relative to Heuchera and Astilbe C C 0789 T T A T T T C T The three methods of analysis are concordant in suggesting 0801 C C that (i) Saxifragaceae sensu lato is at least paraphyletic and 0804 T T 0813 A A C A A A probably polyphyletic, (ii) Brexia and Parnassia are well 0816 T A 0825 T T T T T T T T separated from the remaining taxa of Saxifragaceae analyzed, 0834 A T () the herbaceous genera Astilbe and Heuchera are closely 0840 A A A allied-both herbaceous (Penthorum) and woody (Itea) taxa 0843 C 084 T Tclslgera show a close relationship to these two genera, and (iv) 0845 T T T Solanaceae (Asteridae) are more closely related to Saxi T T T T T T fragaceae (Rosidae) than are members of Fabaceae (Ros T A A A A T T idae). The phylogenetic placement of saxifragaceous taxa 8097 AAAAAA A A paeet 3A~G0~~U 0915 A A closer to Solanaceae than to Fabaceae is not due to any 097 A A A A G A A A 0933 T peculiarities in the sequences of either Pisum or Nicotiana T T T T T T T 094 C C C C C C C C The seune sequences of these two w taxa are r extremely similar iia to 094 T T T G G T T T other representatives of the same two families (e.g., Medi- 094 G 0951 A C C C C cago sativa and Petunia hybrida, respectively). In fact, the C C C C same results were obtained when rbcl sequences for M T T Tanls. sativa and P. hybrida were included in a UPGMA analysis G The relationships suggested by rbcl sequence data have 1008 G T important evolutionary and systematic implications. For G G GG ATC TG G 1020 A T T G example, morphological parallelism and/or convergence 1023 C C C C C C T C may have prompted erroneous assessments of evolutionary A A A A A A A relationships. Furthermore, the woody vs. herbaceous habit 058 A C is not always a reliable indicator of relationship in this group C 106 A A A A A A A Aablt Importantly, our data also demonstrate the ability of rbcl 1866 A A A A A A A A sequence data to clarify relationships in a group notorious for 1071 C C C C C C its taxonomic difficulty. The implications of rbcl T T sequence data regarding the validity of recent classification schemes 1095 C 1098 G are discussed below ~~~~~~~C 1111 C C C C C C C Cronquist's (12) system in particular is called into question 1123 T T T by rbcl sequence data, as are some aspects ofthe systems of 1125 G G G G G G G G 1137 C Thorne (16) and Dahlgren (13). The system of Takhtajan (15) 1140 G G G C G G is largely supported. Not surprisingly, Heuchera and Astilbe 1167 C C are closely allied; they form the core of the Saxifragaceae 1168 T 1170 A according to most investigators. However, Cronquist in A A A T T A A cludes in Saxifragaceae not only Astilbe and Heuchera, but 1180 C 1194 T also Penthorum, Francoa, and Parnassia; he considered the 203 T A A A woody genus Itea to be distantly related to these herbaceous 1209 G G taxa and placed Brexia and Itea together in his Grossulari C C C C C C aceae. These concepts of relationship seem untenable in light A A A A A A A A of rbcl sequence data because (i) Parnassia is well separated 1243 C C C G G G T T G G G from the other members of his Saxifragaceae, (ii) Itea is just 1264 G G G G C G G G 1266 T T T as closely related to Astilbe and Heuchera as are Penthorum 1269 C and Francoa, and (iii) Brexia and Itea appear to be distantly 1209 G G G G G G 1302 G related T T T T T T T T Both Thorne (16) and Dahlgren (13) considered Parnassia 1320 G G G G G G to represent a distinct family; Dahlgren even placed Parnas C CC A 1338 T T T T T T T siaceae in a separate order. rbcl data support the view that 34C5 G A A A Parnassia is well differentiated from those taxa with which it 1346 C 1356 traditionally has been allied; rbcl u~. data as also suggest a rela T T T T T T T T tionship, heretofore unrecognized, of this genus to Brexia C C C C C 1389 G G G G G G G G Although both Thorne (16) and Dahlgren (13) included Fran A AAA cai '''~~~~ coa and Penthorum in 1397 A A A A A Saxifragaceae A A A ideae) (Thorne's and Saxifragoexcluded Itea, rbcl data fail to support this 1402 C C C G G C C G 1404 A A A A A A A A concept of relationships C C C C C C C FIG. 2. rbcl sequence differences between N. otophora (29) and Heuchera (HE), Astilbe (AS), Penthorum (PE), Parnassia (PA), Brexia (BR), Carpenteria (CA), Itea (IT), and Francoa (FR). The nucleotide positions (NUCL) at which differences exist between N. otophora and the taxa analyzed here are given followed by the nucleotides found in the members of the Saxifragaceae sensu lato at those positions. Minus symbols indicate portion of the gene for Francoa that was not available for sequence analysis.

4 Evolution: Soltis et al. Proc. Natl. Acad. Sci. USA 87 (1990) 4643 Table 1. Matrix of Kimura 3ST distance partitioned by codon position Codon Taxon position ZE PI NI HE AS PE PA BR CA IT FR ZE PI NI HE AS PE PA BR CA IT FR Standard errors appear below the diagonal. ZE, Z. mays; PI, P. sativum; NI, N. otophora; other abbreviations are the same as in Fig. 2. ZEA PISUM ASTILBE HEUCHERA ITEA FRANCOA PARNASSIA BREXIA FIG. 3. Parsimony tree obtained using DNABOOT for taxa of Saxifragaceae sensu lato and outgroup taxa. The branch lengths have been drawn to reflect the number of mutations that support each branch. The branch length for Francoa is an estimate for the entire gene based on the number of mutations detected in that portion of the gene actually sequenced. The percentages provide confidence levels based on 100 bootstrap replicates. rbcl data support Takhtajan's (15) narrowly defined Saxifragaceae, which he limits to only 30 genera. Astilbe and Heuchera represent well the morphological extremes of this group. This narrow circumscription of Saxifragaceae is also supported by chloroplast DNA restriction site data (D.E.S., A. Grable, P.S.S., and M. Edgerton, unpublished data). Furthermore, all taxa examined corresponding to Takhtajan's Saxifragaceae have lost the intron for the chloroplast gene rpl2 (S. Downie, R. G. Olmstead, G. Zurawski, D.E.S., P.S.S., J. C. Watson, and J. D. Palmer, unpublished data). This intron is present, however, in Parnassia, Francoa, Penthorum, Brexia, Itea, close relatives of Carpenteria, and other taxa representing many of the original subfamilies of Saxifragaceae sensu lato. Takhtajan considered Itea (but not Brexia) and Penthorum (his Iteaceae and Penthoraceae, respectively) to be close relatives of his narrowly defined Saxifragaceae; rbcl data are consistent with these hypotheses. Takhtajan (15) also placed Brexiaceae in a separate superorder; rbcl data support the distinctiveness of Brexiaceae. However, Parnassia (Parnassiaceae) and Francoa (Francoaceae) were considered close allies of Saxifragaceae by Takhtajan, and these views are not supported by sequence data. One point of disagreement among the three methods of data analysis involves the woody genus Carpenteria. UP- GMA indicates an association of Carpenteria with five other members of Saxifragaceae sensu lato. However, parsimony analysis and some maximum-likelihood solutions suggest a particularly close relationship of Carpenteria with Solanaceae (Figs. 3 and 4). As noted earlier, the relationship of woody and herbaceous members of Engler's Saxifragaceae

5 4644 Evolution: Soltis et al. ASTILBE.PENTHORUM FRANCOA PARNASSIA BREXIA PISUM NICOTIANA FIG. 4. The maximum-likelihood topology for the taxa analyzed by thejumble option based on 10 orderings of the data (log likelihood = ). Zea was not included in thejumble analysis because of the excessive computer run times involved. The relative distances between nodes and tips and between internodes are displayed. The placement of Francoa is based on the partial sequence data available for this taxon. All branch lengths are significant at the P = 0.01 level. has been controversial. The discrepancy involving Carpenteria is intriguing because some (13, 15) have suggested that Hydrangeaceae (which includes Carpenteria) are more closely related to some members of Asteridae (such as Solanaceae) than to Rosidae. rbcl sequence data indicate a closer relationship of Saxifragaceae (Rosidae) to Solanaceae (Asteridae) than to Fabaceae (Rosidae). Although most recent investigators have considered Asteridae to be derived from Rosidae, Saxifragaceae typically have been considered to occupy a basal position in Rosidae, well separated from Solanaceae. A few investigators (32) have, however, suggested that Solanaceae may be derived from Saxifragaceae. The suggestion from rbcl data that some Asteridae and Rosidae are more closely allied than traditionally maintained merits further investigation. In summary, rbcl sequence data have provided important phylogenetic information regarding Saxifragaceae sensu lato. In several instances, these data enabled us to discern among the competing phylogenetic hypotheses of Cronquist, Dahlgren, Thorne, and Takhtajan. In other instances,' rbcl sequence data provided insights into evolutionary relationships. This study also illustrates the importance of using several different methods of data analysis in the comparison of rbcl sequences. The ability of rbcl sequence data to elucidate phylogenetic relationships in a notoriously difficult group such as Saxifragaceae sensu lato indicates the tremendous potential of chloroplast gene sequence data for higherlevel phylogenetic analyses. Proc. Natl. Acad. Sci. USA 87 (1990) We thank Ed Golenberg, Jerry Learn, Amy MacRae, Joel Davis, and David Henderson for their assistance in the laboratory. This research was supported in part by National Science Foundation Grants BSR and BSR Palmer, J. D. (1987) Am. Nat. 130, S6-S Palmer, J. D., Jansen, R. K., Michaels, H. J., Chase, M. W. & Manhart, J. R. (1988) Ann. Mo. Bot. Gard. 75, Sytsma, K. J. & Gottlieb, L. D. (1986) Proc. Natl. Acad. Sci. USA 83, Ritland, K. & Clegg, M. T. (1987) Am. Nat. 130, S Zurawski, G. & Clegg, M. T. (1987) Annu. Rev. Plant Physiol. 38, Doebley, J., Durbin, M., Golenberg, E. M., Clegg, M. T. & Ma, D. P. (1990) Evolution, in press. 7. Giannasi, D. E., Zurawski, G. & Clegg, M. T. (1990) Syst. Bot., in press. 8. Engler, A. (1930) in Die Naturlichen Pflanzenfamilien, eds. Engler, A. & Prantl, K. (Engelmann, Leipzig), Vol. 18a, pp Dahlgren, R. M. T. (1980) J. Linn. Soc. Bot. 80, Spongberg, S. A. (1972) J. Arnold Arbor Harv. Univ. 53, Engler, A. (1964) in Syllabus der Pflanzenfamilien, eds. Melchior, H. & Werdermann, E. (Gebruder Borntraeger, Berlin), pp Cronquist, A. (1981) An Integrated System of Classification of Flowering Plants (Columbia Univ. Press, New York). 13. Dahlgren, R. M. T. (1983) Nord. J. Bot. 3, Takhtajan, A. (1980) Bot. Rev. 46, Takhtajan, A. (1987) System of Magnoliophyta (Academy of Sciences U.S.S.R., Leningrad). 16. Thorne, R. F. (1983) Nord. J. Bot. 3, Soltis, D. E., Soltis, P. S. & Bothel, K. D. (1990) Syst. Bot. 15, Doyle, J. J. & Doyle, J. L. (1987) Phytochem. Bull. 19, Maniatis, T., Fritsch, E. F. & Sambrook, J. (1982) Molecular Cloning:A Laboratory Manual (Cold Spring Harbor Lab., Cold Spring Harbor, NY). 20. Korneluk, R. G., Quan, F. & Gravel, R. A. (1985) Gene 40, Hattori, M. & Sakaki, Y. (1986) Anal. Biochem. 152, Sneath, P. H. A. & Sokal, R. R. (1973) Numerical Taxonomy (Freeman, San Francisco). 23. Felsenstein, J. (1981) J. Mol. Evol. 17, Felsenstein, J. (1983) in Statistical Analysis ofdna Sequence Data, ed. Weir, B. S. (Dekker, New York), pp Felsenstein, J. (1988) Annu. Rev. Genet. 22, Nei, M. (1987) Molecular Evolutionary Genetics (Columbia Univ. Press, New York). 27. Wu, C.-I. & Li, W.-H. (1985) Proc. Natl. Acad. Sci. USA 82, Zurawski, G., Clegg, M. T. & Brown, A. H. D. (1984) Genetics 106, Lin, C. M., Liu, Z. Q. & Kung, S. D. (1986) Plant Mol. Biol. 6, Zurawski, G., Whitfeld, P. R. & Bottomley, W. (1986) Nucleic Acids Res. 14, Kimura, M. (1981) Proc. Natl. Acad. Sci. USA 78, Hutchinson, J. (1959) The Families offlowering Plants (Oxford Univ. Press, New York), 2nd Ed. 33. Pace, N. R., Olsen, G. J. & Woese, C. R. (1986) Cell 45,