997American Journal of Botany 85(7): 997–1006. 1998.PARALLEL EVOLUTION OF GLUCOSINOLATEBIOSYNTHESIS INFERRED FROM CONGRUENT NUCLEARAND PLASTID GENE PHYLOGENIES1JAMESE. RODMAN,2,6PAMELAS. SOLTIS,3DOUGLASE. SOLTIS,3KENNETHJ. SYTSMA,4ANDKENNETHG. KAROL52Division of Environmental Biology, National Science Foundation, Arlington, Virginia 22230;3Department of Botany, Washington State University, Pullman, Washington 99164-4238;4Department of Botany, University of Wisconsin, Madison, Wisconsin 53706-1381; and5Laboratory of Molecular Systematics, Smithsonian Institution, Washington, DC 20560The phytochemical system of mustard-oil glucosides (glucosinolates) accompanied by the hydrolytic enzyme myrosinase(b-thioglucosidase), the latter usually compartmented in special myrosin cells, characterizes plants in 16 families of angio-sperms. Traditional classifications place these taxa in many separate orders and thus imply multiple convergences in theorigin of this chemical defense system. DNA sequencing of the chloroplast rbcL gene for representatives of all 16 familiesand several putative relatives, with phylogenetic analyses by parsimony and maximum likelihood methods, demonstratedinstead a single major clade of mustard-oil plants and one phylogenetic outlier. In a further independent test, DNA sequencingof the nuclear 18S ribosomal RNA gene for all these exemplars has yielded the same result, a major mustard-oil clade of15 families (Akaniaceae, Bataceae, Brassicaceae, Bretschneideraceae, Capparaceae, Caricaceae, Gyrostemonaceae, Koeber-liniaceae, Limnanthaceae, Moringaceae, Pentadiplandraceae, Resedaceae, Salvadoraceae, Tovariaceae, and Tropaeolaceae)and one outlier, the genus Drypetes, traditionally placed in Euphorbiaceae. Concatenating the two gene sequences (for atotal of 3254 nucleotides) in a data set for 33 taxa, we obtain robust support for this finding of parallel origins of glucosinolatebiosynthesis. From likely cyanogenic ancestors, the ‘‘mustard oil bomb’’ was invented twice.Key words: Capparales s.l.; DNA sequencing; glucosinolates; phylogeny; rDNA (18S).Mustard-oil glucosides (also named glucosinolates:Ettlinger and Kjaer, 1968) are oxime-derived sulfur-con-taining compounds whose breakdown products includethe familiar pungent principles of mustard, radish, andcapers (Fenwick, Heaney, and Mullin, 1983). The com-pounds usually are accompanied in the plant by a hydro-lytic enzyme, myrosinase (ab-thioglucoside glucohydro-lase, E.C. 3.2.3.1), which may be compartmented in spe-cial myrosin cells (Fig. 1). Hypothesized to deter herbi-vores or pathogens, this ‘‘mustard oil bomb’’ (Lu¨thy andMatile, 1984) is characteristic of all Brassicaceae, includ-ing the genomic model Arabidopsis, but occurs as wellin 15 other angiosperm families (Table 1). The phyto-chemical system of glucosinolates with accompanyingmyrosinase enzyme, the latter compartmented in myrosincells, is believed to characterize all the species in thesefamilies, except for Euphorbiaceae, where Drypetes (in-terpreted to include Guya and Putranjiva) is the only es-tablished source of mustard oils (Ettlinger and Kjaer,1Manuscript received 28 July 1997; revision accepted 19 November1997.The authors thank Ihsan Al-Shehbaz, Ray Cranfield, Robert Hirano,Hugh Iltis, Ching-I Peng, Robert Price, and Christopher Quinn for plantsamples, Kandis Elliot for graphics, and Peter Raven and George John-son for the idea behind our Fig. 1 cartoon. Support from the AndrewMellon Foundation (to P. S. Soltis and D. E. Soltis) and the NationalScience Foundation (DEB 9307000 to D. E. Soltis and DEB 9407270to K. J. Sytsma), advice from Elizabeth Zimmer, and the facilities ofthe Smithsonian Institution’s Laboratory of Molecular Systematics areappreciated. The authors also thank reviewers Neil Harriman and Kath-leen Kron for advice.6Author for correspondence (e-mail: [email protected]); agency ci-tation is solely for purposes of identification.1968; Rodman, 1981; Ettlinger, 1987). Traditional clas-sifications like Cronquist’s (1981, 1988) place these 16families in several widely separate taxonomic orders andthus imply multiple origins for the glucosinolates-with-myrosinase system. In turn, this multiple-origin view-point has fragmented the study of host fidelity and evo-lution by herbivores and pathogens adapted to glucosi-nolate-producing plants (Chew, 1988).In his early attempt to evaluate taxonomic relationshipsof angiosperms, Dahlgren (1975, 1977) challenged ortho-dox classifications of mustard-oil plants by expanding theorder Capparales to encompass nearly all families of glu-cosinolate taxa. Later he retreated from this position(Dahlgren, 1980, 1983; Dahlgren, Rosendal-Jensen, andNielsen, 1981), and emphasized the considerable mor-phological and habital diversity among these taxa. In acladistic analysis of 90 morphological and phytochemicalcharacters, Rodman (1991b) found weak support forDahlgren’s (1975) expanded Capparales. To test Dahl-gren’s radical reclassification of glucosinolate taxa withan independent molecular data set, representatives of all16 families were sequenced for the chloroplast rbcL geneand analyzed with parsimony and maximum likelihoodmethods (Gadek et al., 1992; Rodman et al., 1993, 1994,1996a). Comparisons were made with numerous putativerelatives, sequenced as part of a large collaborative anal-ysis (Chase et al., 1993; Price and Palmer, 1993). Withone exception, all mustard-oil taxa united into a singlemajor clade or lineage (Fig. 2), nested within a large‘‘rosid’’ plexus (Chase et al., 1993). The sole exceptionis the genus Drypetes, which is usually placed within thespurge family Euphorbiaceae and is affiliated with other998 [Vol. 85AMERICANJOURNAL OFBOTANYFig. 1. Biosynthesis of glucosinolates (mustard-oil glucosides) and of cyanogenic glycosides from amino acids via aldoximes (Du et al., 1995),and hydrolysis of glucosinolates to isothiocyanates (mustard oils; not shown are glucose, sulfate, other potential breakdown products) catalyzed bymyrosinases, a family of thioglucoside glucohydrolases (Xue et al., 1992, 1995) associated with cell membranes (Lu¨thy and Matile, 1984) andconcentrated in myrosin cells (stained cells in insert: Werker and Vaughan, 1974; Ho¨glund, Lenman, and Rask, 1992). Herbivore feeding ishypothesized to bring enzyme and substrate into contact, thereby releasing toxic mustard oils (Ehrlich and Raven, 1965), analogous to cyaniderelease from cyanogenic plants