PHYLOGENETIC RELATIONSHIPS OF THE CHRYSOBALANACEAE INFERRED FROM CHLOROPLAST, NUCLEAR, AND MORPHOLOGICAL DATA1 Deepthi Yakandawala,2,5 Cynthia M. Morton,3 and Ghillean T. Prance4 ABSTRACT The Chrysobalanaceae, a pantropical family containing about 525 species, has often been nested within the Rosaceae despite evidence for recognizing it as a separate family. In 1963, Prance clearly placed Chrysobalanaceae as a distinct family containing 17 genera. However, the family has been linked with various other families and orders and recently has been placed within the order Malpighiales. Because of these discrepancies, a phylogenetic analysis for the family was launched to examine its monophyly and to investigate the relationships within the Chrysobalanaceae as well as its relationships to other groups. Comparative phylogenetic analyses were performed using morphological, rbcL, and ITS sequences. The data sets were analyzed independently and in combination. After exploration for hard incongruencies among the independent data sets, a simultaneous analysis of all the data was completed. The combined analysis resulted in a resolved, supported topology with several unambiguous morphological synapomorphies. The resulting topology indicated that the family is a well-defined monophyletic group that is sister to Euphronia Mart. & Zucc. (Euphroniaceae). The present tribal groupings, however, are paraphyletic. Key words: Chrysobalanaceae, Chrysobalaneae, Couepieae, Hirtelleae, ITS, morphology, Parinarieae, rbcL. The Chrysobalanaceae R. Br. includes 17 genera and about 525 species. Most of these species occur in the lowlands of the tropics and subtropics, and the family is especially well represented in the New World tropics. The present family circumscription places the 17 genera in four tribes: Chrysobalaneae, which includes Chryso- balanus L., Grangeria Comm. ex Juss., Licania Aubl., and Parastemon A. DC. Couepieae, which includes Acioa Aubl., Couepia Aubl., and Maranthes Blume Parinarieae, which includes Bafodeya Prance ex F. White, Exellodendron Prance, Hunga Prance, Neocarya (DC.) Prance ex F. White, and Parinari Aubl. and Hirtelleae, which includes Atuna Raf., Dactyladenia Welw., Hirtella L., Kostermanthus Prance, and Magni- stipula Engl. (Prance & White, 1988) (Table 1). All species of Chrysobalanaceae are woody, and most are trees or shrubs. The vegetative architecture of the plants is relatively uniform. The flower, by contrast, is comparatively diverse, although nearly every genus is characterized by an underlying uniformity of inflores- cence and floral structure. The flowers are bisexual, rarely unisexual, and markedly perigynous. The flower size and shape vary within wide limits, from minute patelliform flowers (Licania elaeosperma (Mildbr.) Prance & F. White) scarcely larger than a pinhead, to tubular flowers that are longer than 10 cm (Maranthes gabunensis (Engl.) Prance). Floral symmetry varies from almost completely actinomorphic to strongly zygomor- phic. In most genera, the entrance to the receptacle tube contains numerous long straight retrorse hairs. There are always five, completely free sepals that range from slightly to strongly imbricate. Petals, when present, are always five and inserted on the margin of the disc, and are mostly caducous however, they are absent in more than half of the species of Licania. The number of stamens also varies from two in Parastemon urophyllus (Wall. ex A. DC.) A. DC. to more than 300 in some species of Couepia. The ovary is fundamentally composed of three carpels, which are united only by the gynobasic style, but in most species only one carpel is functional. The fruit is basically a dry or fleshy drupe of varying size. The interiors of the fruit are often densely hairy, and the endocarp is variable, often containing a special mechanism for seedling escape (Prance, 1963 Prance & White, 1988). 1 The authors thank Walter Judd for helpful discussions. This research was supported by an Advanced Natural Environment Research Council (NERC) Research Fellowship granted to C.M. and a Commonwealth Scholarship granted to D.Y. 2 Centre of Plant Diversity and Systematics, School of Plant Sciences, University of Reading, Reading, United Kingdom. 3 Head of Section of Botany, Carnegie Museum of Natural History, 4400 Forbes Avenue, Pittsburgh, Pennsylvania 15213- 4080, U.S.A. Author for correspondence: mortoncm@verizon.net. 4 Herbarium, Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3AB, United Kingdom. 5 Department of Botany, University of Peradeniya, Peradeniya, Sri Lanka. doi: 10.3417/2007175 ANN. MISSOURI BOT. GARD. 97: 259���281. PUBLISHED ON 9 JULY 2010.

The family is locally important in the tropics for fruit, construction materials, fuel, charcoal, folk medicine, and shade trees. Several species are cultivated. Chrysobalanus icaco L. (coco plum), for example, is tinned and bottled as syrup in Colombia and Venezuela. Couepia rufa Ducke and C. bracteosa Benth. fruits are found in markets in Brazil. Parinari macrophylla Sabine (gingerbread plum) and P. curatellifolia Planch. ex Benth. (mobola plum) are both eaten in Africa. Parinari curatellifolia is used in beer making, and a red dye is extracted from its young leaves in West Africa. In Brazil, Licania tomentosa (Benth.) Fritsch is widely planted as an avenue tree providing shade and C. icaco is widely used as an ornamental in southern Florida. The taxonomic position of the Chrysobalanaceae has been particularly controversial. The majority of authors of general systems of classifications who treat Chrysobalanaceae as a family distinct from Rosaceae leave it in Rosales. Among other opinions, close affinities have been proposed for the Chrysobalana- ceae with a diversity of taxonomic groups: Dichape- talaceae and Trigoniaceae (Hallier, 1923), Gerania- ceae and Tropaeolaceae (Hauman, 1951), and Connaraceae (Gutzwiller, 1961). Many of these families��� and orders��� circumscriptions have changed in the last few years, and more recently the family has been placed in the rosid clade and sometimes in the order Malpighiales based solely on molecular data (Soltis et al., 2005). Cronquist (1981, 1988) placed Chrysobalanaceae in a much more widely conceived order of Rosales, including 24 families. He acknowledged that the Rosales formed an exceedingly diverse order. How- ever, he thought it was more useful to delimit the order broadly than to fragment it and lose sight of the interrelationships among its parts. In 1989, Dahlgren placed the family into Theales under the superorder Theanae, whereas Thorne, in 1999, placed the family in its own order, Chrysoba- lanales, under the superorder Rosanae. According to Prance and White (1988), the embry- ological differences are great among the Chrysobala- naceae, Rosales, and Theales. During 25 years of studying Chrysobalanaceae, Prance and White found no evidence pointing unequivocally to an evolutionary relationship between the Chrysobalanaceae and any other family. The preliminary attempts to apply mainly morpho- logical data to a cladistic analysis at the family level by Prance and White (1988) were thwarted because of the widespread occurrence of parallelism and because only a few of the many characters used at this time satisfied the requirements for inclusion in a cladistic analysis. Chappill (1992) carried out a cladistic analysis, using the 1988 monograph by Prance and White, to determine the level of parallelism in the family and to see whether or not the Prance and White system was supported. The results of the analysis support a monophyletic family however, many of the tribal groupings of Prance and White were not well supported. Phylogenetic work using molecular data within this family is in its infancy. A large molecular phyloge- netic study using rbcL sequence data (Chase et al., 1993) places the Chrysobalanaceae in the rosid clade, with Trigoniaceae as the sister group. A study using 18S ribosomal RNA (rRNA) sequence data (Soltis et al., 1997) also placed the family in the rosid clade. Both of these studies only used one species of Chrysobalanus and Licania. The combined analysis of rbcL, atpB, and 18S data sets (Soltis et al., 2000) and the analysis of the same three genes plus nad1B-C (Davis et al., 2005) housed the family in the Malpighiales within the eurosid I clade or, more recently, what has been called the Fabidae (Cantino et al., 2007) or the ������fabids������ (Judd & Olmstead, 2004). Other recent studies have examined one to three Chrysobalanaceae taxa with two to three genes and have also found the family positioned in the Malpighiales (Davis & Chase, 2004 Tokuoka & Tobe, 2006 Soltis et al., 2007) however, more taxa from different sources of DNA are needed to confirm this placement. Because of these discrepancies between and within the traditional taxonomy and the phyloge- netic analyses, further studies using phylogenetically Table 1. Genera and number of species of tribes of Chrysobalanaceae (from Prance & White, 1988). Tribe/genera No. of species Chrysobalaneae Chrysobalanus 2 Grangeria 2 Licania 192 Parastemon 2 Parinarieae Bafodeya 1 Exellodendron 5 Hunga 11 Neocarya 1 Parinari 44 Couepieae Acioa 4 Couepia 67 Maranthes 12 Hirtelleae Atuna 11 Dactyladenia 27 Hirtella 103 Kostermanthus 2 Magnistipula 11 260 Annals of the Missouri Botanical Garden

more informative characters and more taxa from this family are warranted. To test potentially competing hypotheses of family circumscription and generic relationships in the Chrysobalanaceae, two DNA regions that evolve at different rates were sequenced and a morphological data set was constructed. The first DNA region was the chloroplast gene rbcL, and the second was the nuclear region ITS. The present study carries out a phylogenetic analysis with the following purposes: (1) to test the monophyly of Chrysobalanaceae and their relationships with the Malpighiales (APG II, 2003 Davis & Chase, 2004 Davis et al., 2005 Tokuoka & Tobe, 2006 Soltis et al., 2007) using an expanded rbcL analysis, and (2) to investigate the internal generic relationships within the Chrysobalanaceae using rbcL, ITS sequencing, and morphological and anatomical studies. MATERIALS AND METHODS INGROUP SAMPLING The most recent systematic treatment of Prance and White (1988) has been followed as the basis for generic circumscriptions of the family. The genus Parastemon was not included in the analysis because of lack of material. OUTGROUP SELECTION Outgroups varied depending on the analysis per- formed and material available. For the larger rbcL analysis, the following species were used: a single species of Balanops Baill. (Balanopaceae), Cornus L. (Cornaceae), Connarus L. (Connaraceae), Euphronia Mart. & Zucc. (Euphroniaceae), Ochna L. (Ochnaceae), Comesperma Labill. (Polygalaceae), Spiraea L. (Rosa- ceae), Stylobasium Desf. (Stylobasiaceae), Cadellia F. Muell., Guilfoylia F. Muell., Suriana L. (Surianaceae), Tetracoccus Engelm. ex Parry, and Androstachys Prain (Picrodendraceae) two species of Trigonia Aubl. (Trigoniaceae) and Tapura Aubl. and three species of Dichapetalum Thouars (Dichapetalaceae). A subset of the most closely related taxa were used for a smaller rbcL analysis. For the ITS analysis, however, only Dichapetalum and Euphronia were used because of amplification problems. Dichapetalum and Tapura were chosen for the morphological analysis. These taxa were selected on the basis of the rbcL analysis of Chase et al. (1993) and Litt and Chase (1999), and various historical observations mentioned above. DNA EXTRACTIONS The total genomic DNA was extracted from herbarium, silica gel���dried, or air-dried leaf samples. Appendix 1 lists the voucher information, specimen numbers, type of material used, and GenBank numbers. Fresh (1���2.0 g) or dried (0.1���0.2 g) leaf material was ground into a fine powder and incubated according to the shortened 23 CTAB procedure of Doyle and Doyle (1987). Proteins were removed with SEVAG (24:1, chloroform:isoamyl alcohol), followed by an isopropanol precipitation. Purified DNA was stored at 280uC. RBCL GENE AMPLIFICATION The amplification of the rbcL gene was performed either as one complete piece using the forward primer that matched the first 20 base pairs (1F-ATGTCAC- CACAAACAGAAAC) of the exon and a reverse primer (1460R-TCCTTTTAGTAAAAGATTGGGCC- GAG) that matched a downstream control site (Olmstead et al., 1992) or as two overlapping pieces using three additional internal primers, 636F (GCGTTGGAGAGATCGTTTCT), 724R (TCGCATG- TACCYGCAGTTGC), and 1368R (CTTTCCAAATTT- CACAAGCA GCA). When primer mismatch occurred, primer combinations changed accordingly. The poly- merase chain reaction (PCR) was performed using standard protocols. The Thermal Cycler 480 (Perkin Elmer Inc., Waltham, Massachusetts, U.S.A.) was programmed to perform 25 cycles of denaturation at 94uC for 1 min., primer annealing at 50uC for 30 sec., and extension at 72uC for 1 min. Slight modifications to optimize the reaction conditions were found to be necessary for some taxa that is, the MgCl2 and bovine serum albumin (BSA) concentrations were varied. Products were purified using QIAGEN QIAquick PCR purification kit (QIAGEN Inc., Chatsworth, California, U.S.A.) following protocols provided by the manufac- turer. ITS GENE AMPLIFICATION The amplification of the ITS gene was performed using oligonucleotide primers 17SE (ACGAATT- CATGGTCCGGTGAAGTGTTCG) and 26SE (TA- GAATTCCCCGGTTCGCTCGCCGTTAC) as described by Sun et al. (1994). The PCR was performed using standard protocols however, the DNA template, depending on the concentration, was diluted 1:10 or 1:100. The thermal cycler was programmed to perform an initial one cycle of denaturation at 95uC for 2 min. followed by 24 cycles of 95uC for 30 sec., 55uC for 30 sec., and 72uC for 1 min. 30 sec. This was followed by 10 min. extension at 72uC. The same procedures described in the pre- ceding section were followed after completion of the PCR. Volume 97, Number 2 Yakandawala et al. 261 2010 Phylogenetic Relationships of Chrysobalanaceae