Classification of Convolvulaceae: A Phylogenetic Approach

Because recent molecular studies, based on multiple data sets from all three plant genomes, have indicated mutually congruent, well-resolved, and well-supported relationships within Convolvulaceae (the morning-glory family), a formal reclassification of this family is presented here. Convolvulaceae, a large family of worldwide distribution, exhibiting a rich diversity of morphological characteristics and ecological habitats, are now circumscribed within twelve tribes. A key to these tribes of Convolvulaceae is offered. The group of spiny-pollen bearing Convolvulaceae (forming ‘‘Echinoconiae’’) and tribe Cuscuteae are retained essentially in their traditional sense, Cresseae are circumscribed with only minor modifications, Convolvuleae and Erycibeae are recognized in a restricted sense, while Dichondreae and Maripeae are expanded. Also, to produce a tribal taxonomy that better reflects phylogenetic relationships, the concept of Poraneae is abandoned as artificial, three new tribes are recognized (Aniseieae, Cardiochlamyeae, and Jacquemontieae), and a new tribal status is proposed for the Malagasy endemic Humbertia (Humbertieae). ‘‘Merremieae’’ are tentatively retained even though the monophyly of this tribe is not certain. In addition to the formal classification, we provide clade name definitions for the family as well as for most of the clades recognized presently as tribes. Also, five well-supported clades that are not assigned formal ranks are recognized and their names defined. The reevaluation of traditional taxonomic characters reveals that many homoplasious characters were emphasized in previous classifications, resulting in formal recognition of non-monophyletic groups. Putative morphological synapomorphies for many clades discovered through molecular cladistic analyses are discussed. However, the morphology of several clades that are well-supported by DNA evidence remains poorly understood, creating further challenges for future studies in Convolvulaceae. Recent advances in understanding of the phylogenetic relationships within Convolvulaceae, the morning-glory family, based on molecular data sets drawn from all three plant genomes, indicate the need for a revised classification of this family (Manos et al. 2001; Miller et al. 2002; Stefanovic et al. 2002; Stefanovic and Olmstead in press). The morning-glory family comprises approximately 1600–1700 species grouped in 55–60 genera (Mabberley 1987). The family is nearly cosmopolitan in distribution, but its members are primarily tropical plants, with many genera endemic to tropical zones of individual continents (Austin 1998). Uncertainties exist regarding both delimitation of the family and intrafamilial (tribal) circumscription. Currently available classifications within Convolvulaceae rely largely on the work of several earlier workers, including Choisy (1845), Bentham and Hooker (1873), Hallier (1893), Peter (1891, 1897), and Roberty (1952, 1964). An overview of these, including both points of conflict and congruence among them, is presented by Austin (1973; his Table 8). All of these pre-cladistic schemes are based on very few characters considered to be ‘‘the most important’’ by the author. Among the most influential are those of Peter (1891), based upon fruit type, and Hallier (1893), adopted by Peter (1897), based on pollen surface features as well as fruit and stylar characters. Hallier divided the family into two groups: ‘‘Echinoconiae’’, with spiny pollen surface, and ‘‘Psiloconiae’’, with smooth pollen. The major outlines of this classification were subsequently adopted by many authors (but see Roberty 1952, 1964), with modifications concerning mainly the taxonomic ranks. The most comprehensive traditional work at the family level was offered by Austin (1973) based mainly on chromosome numbers. Austin’s treatment (1973, modified 1998) is currently the most widely used traditional classification. Following Austin, the nonparasitic members of the family are circumscribed in nine tribes (Table 1). Cuscuta, the only parasitic taxon associated with Convolvulaceae, is treated as a separate monotypic family, Cuscutaceae. The major traditional synoptic works on flowering plants place Convolvulaceae in the order Solanales (sensibus Cronquist 1988; Dahlgren 1989; Thorne 1992) along with the Polemoniaceae, Hydrophyllaceae, and Solanaceae. Takhtajan (1997) placed this family in its own order, Convolvulales, due to a number of characteristics, such as presence of articulated latex canals and latex cells, intraxylary phloem position, distinct seed structure and pollen morphology, not shared with other Solanales. Results of broad molecular phylogenetic studies of angiosperms in general, and Asteridae in particular, based on limited molecular data for the Convolvulaceae suggested that the family is most closely related to the Solanaceae and Montiniaceae (e.g., Olmstead and Palmer 1992; Olmstead et al. 1992, 1993, 2000; Chase et al. 1993; Soltis et al. 1997, 2000; APG 1998; Savolainen et al. 2000). Some members of Convolvulaceae, especially Cus792 [Volume 28 SYSTEMATIC BOTANY TABLE 1. Synopsis of Convolvulaceae classification—a comparison of the most recent traditional (Austin 1973, 1998) and newly proposed phylogenetic classification. Asterisks (*) indicate genera not sampled in any molecular study of Convolvulaceae (see text), but retained in present classification in their putative respective tribes based on morphology. Pound signs (#) indicate genera found not to be monophyletic as circumscribed traditionally. Quotation marks (‘‘ ’’) indicate the tribe not confirmed to be monophyletic (see text). Austin (1973, modified 1998) Family Tribe Genus Phylogenetic classification (present study) Family Tribe Genus Convolvulaceae Argyreieae Argyreia Blinkworthia Rivea Convolvulaceae Ipomoeeae s.l. Argyreia (including Rivea) Astripomoea Ipomoeeae Ipomoea Astripomoea Lepistemon Lepistemonopsis Paralepistemon Stictocardia Turbina Blinkworthia* Ipomoea# Lepistemon Lepistemonopsis* Paralepistemon Stictocardia Turbina# ‘‘Merremieae’’ Merremieae Merremia Hewittia Decalobanthus Operculina Xenostegia Merremia# Hewittia Hyalocystis* Decalobanthus* Xenostegia Operculina Hyalocystis Aniseia Iseia Odonellia Tetralocularia Convolvuleae Convolvulus (including Calystegia) Polymeria Aniseieae Convolvuleae Calystegia Convolvulus Polymeria Jacquemontia Erycibeae Aniseia (including Iseia) Odonellia Tetralocularia Cuscuteae Cuscuta Maripa Dicranostyles Lysiostyles Erycibe Humbertia Hildebrandtieae Jacquemontieae Jacquemontia Maripeae Dicranostyles Maripa Lysiostyles* Hildebrandtia Sabaudiella Cladostigma Cresseae Seddera Evolvulus Cressa Bonamia Stylisma Wilsonia Cresseae s.l. Hildebrandtia (including Cladostigma and Sabaudiella) Seddera Evolvulus Cressa Bonamia# Stylisma Wilsonia Itzaea Neuropeltis Neuropeltopsis Dichondreae Itzaea Neuropeltis Neuropeltopsis* Dichondreae s.l. Dichondra Falkia Nephrophyllum Dichondra Falkia Nephrophyllum* Petrogenia TABLE 1. Continued. Austin (1973, modified 1998) Family Tribe Genus Phylogenetic classification (present study) Family Tribe Genus Poraneae Porana p.p. Porana Metaporana Calycobolus Dipteropeltis Metaporana Calycobolus# Dipteropeltis Rapona Rapona Cordisepalum Poranopsis Cardiochlamys Tridynamia Dinetus Erycibeae Erycibe Cardiochlamyeae Cordisepalum Poranopsis Cardiochlamys Tridynamia Cuscutaceae Cuscuteae Cuscuta Porana p.p. Dinetus Humbertieae Humbertia cuta and Ipomoea, have been used as model systems to address a wide variety of biological questions. Many species of the parasitic genus Cuscuta are recognized as pests on a large array of important agricultural crops. A substantial body of literature deals with the life history, ecology, and pest control of Cuscuta species (reviewed by Dawson et al. 1994, and references therein). These branch parasites are also frequently used to study haustorial initiation and formation (e.g., Dörr 1987; Heide-Jørgensen 1987; Lee and Lee 1989; Subramaniam and Mahadevan 1994). In addition, Cuscuta has been the subject of extensive molecular analyses. The diversity of photosynthetic ability among species prompted several physiological studies of photosynthetic enzymes and molecular evolution studies of the chloroplast genome (Machado and Zetsche 1990; Haberhausen et al. 1992; Haberhausen and Zetsche 1994; Bömmer et al. 1993; Freyer et al. 1995). Species attributed to the genus Ipomoea are exceptionally morphologically diverse, varying in habit and vegetative and reproductive morphology. Due to this diversity Ipomoea species have been the focus of a broad range of evolutionary and molecular genetic studies including maintenance of floral polymorphisms, mating system evolution, and evolution of flavonoid biosynthetic pathway (reviewed by Miller et al. 1999). Convolvulaceae have been the subject of only two family-wide molecular phylogenetic studies (Stefanovic et al. 2002; Stefanovic and Olmstead in press). The first study (Stefanovic et al. 2002) was based on sequences from four chloroplast loci: rbcL, atpB, psbE-J operon, and the trnL-F region. These sequences were obtained from a broad sample of taxa within the family, including 102 species from all nine traditionally recognized nonparasitic tribes (Austin 1973; 1998), seven Cuscuta species, as well as three outgroups. The 2003] 793 STEFANOVIC ET AL.: CLASSIFICATION OF CONVOLVULACEAE chloroplast DNA (cpDNA) results confirmed that Convolvulaceae are sister to Solanaceae, with 100% bootstrap support for each family and the clade comprising both families. Two groups, Cuscuta and tribe Dichondreae, that have been proposed as segregate families (Dumortier 1829), were found nested within the Convolvulaceae. The position of Cuscuta as sister to the rest of Convolvulaceae, which would be compatible with recognizing it as a separate family, was rejected. This result was further corroborated by the distribution of deletions in the atpB gene and trnL intron found in all Cuscuta species as well as in all other Convolvulaceae, with the exception of Humbertia, which is sister to the rest of the family (Stefanovic et al. 2002). The second study (Stefanovic and Olmstead in press) was designed to better address the question of the position of Cuscuta within Convolvulaceae. A new molecular data set consisting of mitochondrial (atpA) and nuclear (RPB2) genes was generated for a subset of species chosen to best represent the diversity of the family, and analyzed together with the existing chloroplast data matrix to which an additional chloroplast gene (rpl2) was added. That study confirmed the previously recovered polyphyly of Erycibeae, Poraneae, and Merremieae, the close relationships of tribes Hildebrandtieae with Cresseae, of Ipomoeeae with Argyreieae, and of Dichondreae with some members of Poraneae. At least two nonparasitic lineages were shown to diverge within the family before Cuscuta. However, the exact sister group of Cuscuta could not be ascertained, even though many alternatives were tested and rejected with confidence (Stefanovic and Olmstead in press). Halliers’s (1893) ‘‘Echinoconiae’’ circumscribe species with spiny pollen and include approximately onehalf of the species in Convolvulaceae. The results regarding relationships within this group inferred from broader sampling across Convolvulaceae, but more modest sampling of ‘‘Echinoconiae’’ (Stefanovic et al. 2002; Stefanovic and Olmstead in press) are in agreement with the more detailed studies focusing on this clade conducted by Manos et al. (2001) and Miller et al. (2002). Those studies sampled more taxa and used more rapidly evolving nuclear sequences (ITS and waxy), and offered a resolved and well-supported phylogeny for the Convolvulaceae with spiny pollen. The paraphyly of tribe Ipomoeeae and genus Argyreia as well as polyphyly of Ipomoea and Turbina are first reported and discussed in detail in those studies. The widespread congruence among independent data sets provides compelling support for the classification proposed here. The main goal of the present study is to develop a comprehensive, phylogeny-based classification, in conjunction with a reevaluation of traditional taxonomic characters. In a phylogenetic classification, only monophyletic groups are recognized and named (Wiley et al. 1991; de Queiroz and Gauthier 1994). Classification based exclusively on monophyletic groups provides a more efficient way to store, communicate, and retrieve information and has greater predictive value (Farris 1979; Donoghue and Cantino 1987; for different opinions see Cronquist 1987; Brummit 2002). Most importantly, phylogenetic classification provides an explicit basis for comparative biological studies (Funk and Brooks 1990). In this way the classification can communicate to the scientific community the currently available phylogenetic information on relationships within Convolvulaceae. MATERIALS AND METHODS At present, the best basis for intrafamilial classification of Convolvulaceae is provided by two family-wide molecular cladistic analyses (Stefanovic et al. 2002; Stefanovic and Olmstead in press) combined with two additional studies focusing on species-rich group of Convolvulaceae with spiny pollen (Manos et al. 2001; Miller et al. 2002). A series of goals for the research on Convolvulaceae has been established by Stefanovic et al. (2002) and the first three, to test the monophyly of Convolvulaceae, to circumscribe major lineages within the family, and to develop a wellsupported phylogenetic hypothesis of Convolvulaceae, have been addressed in that study. An additional goal, to narrow down the phylogenetic position of Cuscuta, the only parasitic genus of Convolvulaceae, was addressed in more detail in a separate study (Stefanovic and Olmstead in press). The questions regarding the monophyly of Ipomoea and the relatioships among its species and members of tribe Argyreieae were first addressed in depth by Manos et al. (2001) and Miller et al. (2002). A summary of the current evolutionary hypothesis for Convolvulaceae based on all currently available molecular analyses is depicted in Fig. 1. A synoptical overview of our proposed phylogenetic classification and comparison with most recent taxonomic treatment (Austin 1973, 1998) is given in Table 1. The PhyloCode is a rank-free system of classification based on phylogenetic principles (de Queiroz and Gauthier 1992; 1994) that has generated much discussion regarding its advantages (Cantino et al. 1997, 1999; Cantino 1998, 2000; Baum et al. 1998; de Queiroz and Cantino 2001; Lee 2001; Bryant and Cantino, 2002) and disadvantages (Brummitt 1997; Lindén and Oxelman 1996; Lindén et al. 1997; Benton 2000; Nixon and Carpenter 2000; Forey 2001, 2002) with respect to a traditional, or Linnaean, system. It is beyond the scope of this paper to partake in this debate. However, in addition to the formal Linnaean classification, as governed by the International Code of Botanical Nomenclature (ICBN), we provide phylogenetic definitions for the family and for most of the clades recognized as tribes according to the rankless phylogenetic nomenclature system (de Queiroz and Gauthier 1992; 1994) as governed by the set of rules published on-line in the draft PhyloCode (Cantino and de Queiroz http://www.ohio.edu/phylocode/). These definitions follow the discussions of individual groups. Node-based definitions are preferred whenever taxon sampling and/or confidence in given clades allows it. However, we offer stem-based tribe definitions in certain cases due to the uncertainties regarding the exact circumscription of some tribes. In this way the core members of a tribe can be joined eventually by other taxa of unsure position as additional information becomes available. Also, five well-supported clades that are not assigned formal ranks are recognized and defined in this manner. They are distinguished from the traditional taxa by a preceding slash mark (e.g., /Dicranostyloideae, Fig. 1), following Alverson et al. (1999). These additional groups are discussed following the formal taxonomic treatment. While deciding which clades are well enough supported to warrant naming, the priority was given to the molecular analyses, but 794 [Volume 28 SYSTEMATIC BOTANY FIG. 1. The summary evolutionary hypothesis for Convolvulaceae derived from sequence data from all three plant genomes and analyzed with a range of analytical methods (Manos et al. 2001; Miller et al. 2002; Stefanovic et al. 2002; Stefanovic and Olmstead in press). Numbers above branches represent bootstrap values for selected clades as reported by Stefanovic et al. (2002); numbers below branches are the Bayesian posterior probabilities for the same clades following Stefanovic and Olmstead (in press). Asterisks (*) indicate nodes for which Bayesian posterior probabilities were not estimated by Stefanovic and Olmstead (in press) due to the reduced taxon sampling. The areas of clades correspond approximately to the number of species found in those clades. Traditional deffinition of tribes (Austin, 1973, 1998) is indicated in the left column. The formal phylogenetic classification proposed in this paper is indicated in the right column. Compare with Table 1 for the generic circumscription of tribes according to both classifications. Letters a-b represent genera traditionally circumscribed in Cresseae (a 5 Itzaea, Neuropeltis, Neuropeltopsis, Wilsonia) and Poraneae (b 5 Dipteropeltis, Rapona, Calycobolus) for which molecular data did not ascertain more precise tribal affinities. Rankless names, indicated by a preceding slash mark, are provided for five additional clades (see text for further explanation). 2003] 795 STEFANOVIC ET AL.: CLASSIFICATION OF CONVOLVULACEAE the morphological distinctiveness as well as correspondence with previously named taxa also were taken into consideration. Each named clade was supported by the molecular analyses (Fig. 1). However, different degrees of support were deemed sufficient depending on whether the particular clade was corroborated by other lines of evidence or not. For example, the /Dicranostyloideae clade (Fig. 1) was found to be only moderately supported in some molecular analyses but this clade corresponds closely to Dicranostyleae sensu Hallier (1893) and is corroborated by morphology (divided style; see below). On the other hand, /Argyreiinae and /Astripomoeinae clades lack defining morphological features uniting all of their respective members and contradict traditional circumscription (see below) but are strongly supported by all molecular analyses (Fig. 1). Also, it is not necessary that all well supported clades be named; hence some clades, albeit well supported, were left unnamed at present due to the lack of known morphological synapomorphies and/or because their respective members do not correspond well to previously named groups.