A phylogenetic analysis of morphological and molecular characters of Boraginaceae: evolutionary relationships, taxonomy, and patterns of character evolution

The angiosperm family Boraginaceae includes ca. 1600 species distributed among ca. 110 genera. Some floral features are constant within the family, but many vegetative, floral, pollen, and nutlet traits vary. Utilizing 224 species of Boraginaceae and related taxa, five matrices were constructed with various combinations of morphological characters, three chloroplast DNA regions, and one nuclear ribosomal DNA region. Phylogenetic analyses were conducted for these matrices, and patterns of character evolution were examined. Boraginaceae is resolved as monophyletic, with Wellstedia as its sister. Codon is sister to Boraginaceae + Wellstedia. Although most of the investigated morphological characters have a low consistency index, particular character states are synapomorphies for large clades in each of the tribes of the family. In Boraginaceae, the breeding system heterostyly evolved at least 12 times, which is the largest number of origins resolved in any family; therefore Boraginaceae can serve as a model for the evolution and development of heterostyly. Nutlet ornamentation is most diverse in Cynoglosseae and Trichodesmeae, while pollen and floral features are most variable in Boragineae and Lithospermeae. Phylogenetic relationships and patterns of character evolution identified in the present study set the stage for future work creating an updated taxonomic system of Boraginaceae. © The Willi Hennig Society 2013. Introduction The angiosperm family Boraginaceae includes ca. 1600 species distributed among ca. 110 genera. The family is characterized by a scorpioid cymose inflorescence (Buys and Hilger, 2003), a gynobasic style, and a two-part ovary that breaks into four nutlets. This circumscription is equivalent to, and has in the past been referred to as, Boraginaceae s.s. or Boraginoideae (Small, 1913; Gottschling et al., 2001; Diane et al., 2002). Boraginaceae has also been circumscribed in a broader context, which has been referred to as Boraginaceae s.l. or Boraginales. This broader circumscription has included four taxa treated as either subfamilies (Boraginoideae, Cordioideae, Ehretioideae, and Heliotropioideae) or families that are characterized by a scorpioid cyme and two-parted gynoecium (style position and fruit type vary) (Lawrence, 1937; Cronquist, 1981; Al-Shehbaz, 1991; Takhtajan, 1997). In the present study, the former circumscription is treated as Boraginaceae, while the latter is treated as Boraginales, which currently includes: the four traditionally recognized families (Boraginaceae, Cordiaceae, Ehretiaceae, and Heliotropiaceae); Hydrophyllaceae, which has been recognized as closely related to the aforementioned four taxa (Cronquist, 1981; Gottschling et al., 2001; Soltis et al., 2011); and three small families (Codonaceae, Lennoaceae, and Wellstediaceae) (Gottschling et al., 2001; Weigend and Hilger, 2010) that have yet to be critically studied in a phylogenetic context. Of the eight families in Boraginales, Boraginaceae is the most speciose, and although the inflorescence type, gynoecium position, and fruit type *Corresponding author: E-mail address: [email protected] Cladistics Cladistics (2013) 1–31 10.1111/cla.12036 © The Willi Hennig Society 2013 are consistent within the family, other vegetative, floral, pollen, and nutlet traits vary. The objective of the present study is two-fold: (i) to utilize morphological characters and DNA sequence data to reconstruct phylogenetic relationships within Boraginaceae; and (ii) to investigate patterns of morphological character evolution in the family. During the past 17 years, researchers have conducted several phylogenetic studies on Boraginaceae. Most have focused on relationships within a genus or among closely related genera (e.g. B€ ohle et al., 1996; Boyd, 2003; L angstr€ om and Oxelman, 2003; Hilger et al., 2004; Buys, 2006; Selvi et al., 2006; Cohen and Davis, 2009, 2012; Weigend et al., 2009; Khoshsokhan et al., 2010; Hasenstab-Lehman and Simpson, 2012; Trinh et al., 2012; Huang et al., in press), although some (L angstr€ om and Chase, 2002; Mansion et al., 2009; Weigend et al., 2010; Nazaire and Hufford, 2012) have addressed higher-level relationships. The lack of overlapping taxon samples across multiple studies has made it difficult to cobble together a phylogeny of Boraginaceae, and ca. 40% of the genera of the family have yet to be included in a phylogenetic analysis. Questions remain concerning the placement of the many small genera (< 5 species) in the family as well as the monophyly of large, geographically widespread genera (e.g. Anchusa L., Cynoglossum L., Myosotis L., and Onosma L.) and tribes. Moreover, Boraginaceae remains unplaced among the lamiids in the latest treatment of the Angiosperm Phylogeny Group (APG III, 2009). In Boraginaceae, tribes frequently have been recognized based on a combination of style division, stigma number, position of nutlet attachment, and nutlet ornamentation (e.g. Al-Shehbaz, 1991). This has led to the acceptance of between four (L angstr€ om and Chase, 2002) and 13 tribes (Popov, 1953), depending on the author, and has resulted in increased taxonomic complexity within the family. Recent phylogenetic analyses (L angstr€ om and Chase, 2002; Mansion et al., 2009; Weigend et al., 2010; Nazaire and Hufford, 2012) have led to the identification of four to five tribes—Boragineae, Cynoglosseae, Echiochileae Lithospermeae, and Trichodesmeae—that are congruent with the traditional taxonomic system of Boraginaceae. Phylogenetic relationships among tribes are becoming better resolved and better supported, but relationships within each tribe remain largely unresolved (e.g. Hilger et al., 2004), although the phylogeny of one tribe, Lithospermeae, has begun to be elucidated (B€ ohle et al., 1996; Buys, 2006; Thomas et al., 2008; Cecchi and Selvi, 2009; Cohen and Davis, 2009, 2012; Ferrero et al., 2009; Weigend et al., 2009). Phylogenetic analyses of genera of Boraginaceae suggest that morphological character evolution provides intriguing patterns (L angstr€ om and Oxelman, 2003; Buys, 2006; Ferrero et al., 2009; Cohen, 2011; Hasenstab-Lehman and Simpson, 2012; Huang et al., in press), but these patterns have yet to be explored throughout the entire family. This is unfortunate because Boraginaceae is well suited to serve as a model for the study of particular morphological features. For example, heterostyly, a complex and elegant breeding system that involves morphological and physiological components (Fig. 1g), is present in Boraginaceae in at least nine genera scattered among three tribes (Ganders, 1979; Naiki, 2012). Within these tribes, Thomas et al. (2008), Ferrero et al. (2009), Cohen (2010, 2011), and Hasenstab-Lehman and Simpson (2012) provide evidence for multiple origins of heterostyly, but patterns of this breeding system have yet to be studied critically throughout the family. Additionally, because Boraginaceae only produces one type of fruit—nutlets (Fig. 1a–c)—it is possible to focus investigations of fruit evolution on the modifications of one type of fruit rather than, as is the case in many taxa of comparable size (Clausing et al., 2000; Knapp, 2002), the origin of different types of fruit as well as modifications of each type of fruit. The present study provides a family-level phylogenetic investigation of Boraginaceae that includes both DNA sequence data and morphological characters, which allows for phylogenetic relationships to be elucidated and patterns of character evolution to be examined. Materials and methods Taxon sampling The present study includes 224 species (Appendix 1). Two hundred and six species from across 80 genera belong to the ingroup. This sampling comprises ca. 70% of the genera of Boraginaceae, and represents both the morphological and geographic range of variation in the family. The outgroup comprises 18 species from related families of Boraginales and Lamiidae (Gottschling et al., 2001; Luebert and Wen, 2008; Mansion et al., 2009; Soltis et al., 2011), including Codonaceae, Cordiaceae, Ehretiaceae, Heliotropiaceae, Hydrophyllaceae, Vahliaceae, and Wellstediaceae. No members of Lennoaceae were included in the present study, but this family has been resolved as nested within, or sister to, Ehretiaceae (Gottschling et al., 2001; Hilger et al., 2005). DNA sequence data Sequence data from four DNA regions were included in the present study: two protein-encoding plastid DNA (cpDNA) regions (matK and ndhF), one cpDNA intergenic spacer (trnL–trnF), and the nuclear 2 J. I. Cohen / Cladistics (2013) 1–31