Synopsis of Trichosanthes (Cucurbitaceae) based on recent molecular phylogenetic data

Abstract The snake gourd genus, Trichosanthes, is the largest genus in the Cucurbitaceae family, with over 90 species. Recent molecular phylogenetic data have indicated that the genus Gymnopetalum is to be merged with Trichosanthes to maintain monophyly. A revised infrageneric classification of Trichosanthes including Gymnopetalum is proposed with two subgenera, (I) subg. Scotanthus comb. nov. and (II) subg. Trichosanthes, eleven sections, (i) sect. Asterospermae, (ii) sect. Cucumeroides, (iii) sect. Edulis, (iv) sect. Foliobracteola, (v) sect. Gymnopetalum, (vi) sect. Involucraria, (vii) sect. Pseudovariifera sect. nov., (viii) sect. Villosae stat. nov., (ix) sect. Trichosanthes, (x) sect. Tripodanthera, and (xi) sect. Truncata. A synopsis of Trichosanthes with the 91 species recognized here is presented, including four new combinations, Trichosanthes orientalis, Trichosanthes tubiflora, Trichosanthes scabra var. pectinata, Trichosanthes scabra var. penicaudii, and a clarified nomenclature of Trichosanthes costata and Trichosanthes scabra.


Background
Trichosanthes L. is the largest genus in the Cucurbitaceae family, with over 90 species. Th e genus has its center of diversity in Southeast Asia, but ranges from India throughout Asia east to Taiwan, the Philippines and Japan, and southeast to New Guinea, Australia, Fiji and Vanuatu (de Wilde and Duyfj es 2010). Th e snake gourd (Trichosanthes cucumerina L.) is a popular vegetable in South and Southeast Asian cuisine and cultivated in tropical and subtropical regions around the globe. Gymnopetalum Arn. includes four species (de Wilde and Duyfj es 2006) and ranges from India through China and Southeast Asia into the Malay archipelago, but does not occur in New Guinea and Australia (de Wilde and Duyfj es 2010).

Morphology and classification
Trichosanthes are mostly stout perennial climbers, 3-30 m long, dioecious, less frequently monoecious, with branched tendrils, distinctly fringed petals, and often eggsized brightly colored fruits. Dioecy, variation in vegetative morphology (esp. in juvenile plants), and incomplete herbarium collections, complicate taxonomical studies and have contributed to the description of nearly 300 taxa (de Wilde and Duyfj es 2010;IPNI 2011).
Infrageneric classifi cations of the genus Trichosanthes have been proposed by various authors (Yueh and Cheng 1974;Jeff rey 1980b;Chen 1985;Huang et al. 1997;Rugayah and de Wilde 1999;de Wilde and Duyfj es 2004 Rugayah and de Wilde (1999) and de Wilde and Duyfj es (2004; are reserved in their infrageneric classifi cations, and mention a need for further investigation. Pollen morphology has also been used for infrageneric classifi cation in the genus (Khunwasi 1998;Pruesapan and Van Der Ham 2005;Huang et al. 1997), but as a character is very variable in Cucurbitaceae, and its taxonomic value is not clear (Schaefer and Renner 2011). Palynological studies (Pruesapan and Van Der Ham 2005) have indicated that a variety of pollen types exist in Trichosanthes including 3(-4)-porate and 3(-4)-colporate pollen with psilate, perforate, verrucate, reticulate, and regulate ornamentation. Th eir study of pollen from 37 species distinguishes fi ve pollen types, two of which are further divided into subtypes, and categorized these using exine ornamentation patterns for the major types and ectoaperture characters for the subtypes.
Gymnopetalum strongly resembles certain Trichosanthes species, but lack threadlike fringes on the petals, and the overall shape of the folded petals in the mature bud is elongate (short and rounded in Trichosanthes) (de Wilde and Duyfj es 2006). A revision of the genus was published by de Wilde and Dufyjes (2006), with minor nomenclatural changes published later (de Wilde and Duyfj es 2008). Cogniaux (1881) divided Gymnopetalum into two sections, (i) sect. Gymnopetalum containing the type G. tubifl orum (Wight & Arn.) Cogn. from southern India and Sri Lanka, and (ii) sect. Tripodanthera (M.Roem.) Cogn. containing the Southeast Asian and Malesian species. Later authors did not follow this classifi cation (Jeff rey 1980a;Philcox 1997;de Wilde and Duyfj es 2006).

Recent molecular phylogenetic studies
Th e molecular phylogenetic study of Trichosanthes by de Boer et al. (Submitted) shows that Trichosanthes and Gymnopetalum are both non-monophyletic, but together form a clade with high support in the Bayesian tree and weak support in the ML tree (0.99/62). Th is indicates that Gymnopetalum should be merged with Trichosanthes, and that a revised infrageneric classifi cation is necessary. Some previously recognized sections in Trichosanthes and Gymnopetalum are well supported, but others need to be described or redefi ned. Within Trichosanthes the support for the two clades here defi ned as subgenera Trichosanthes (1.00/94) and Scotanthus (1.00/97) is high. However, splitting the genus into two genera corresponding to subg. Scotanthus and subg. Trichosanthes would not improve clarity, as both would consist of species with fringed and fringeless corollas. Maintaining a large Trichosanthes is in accordance with the recent taxonomic revisions of the genus (de Wilde and Duyfj es 2010; Chakravarty 1959;Lu et al. 2011;Duyfj es and Pruesapan 2004;Keraudren-Aymonin 1975;Cooper and de Boer 2011;Ohba 1984), and is the alternative that best provides taxonomic stability. Some proposed sections in Trichosanthes and Gymnopetalum are well supported: (i) sect. Cucumeroides (1.00/93), including subsect. Cucumeroides (0.99/-) and subsect. Tetragonosperma (0.99/69), (ii) sect. Edulis (1.00/75), and (iii) sect. Asterosperma (1.00/100). Th e subsections of sect. Cucumeroides are statistically supported, but subsect. Cucumeroides consists solely of accessions of Trichosanthes pilosa Lour. and species that have been reduced to its synonymy (Cooper and de Boer 2011). Sect. Involucraria is only weakly supported (0.83/65), primarily due to the low support for inclusion of its type, T. wallichiana (Ser.) Wight. Th e subsections of Involucraria are not supported, and taxa belonging to subsect. Pedatae are nested in diff erent locations within subsect. Bracteatae. Sect. Tripodanthera is not supported by the analysis, and could possibly form a grade at the base of sect. Edulis. However, morphological support for this section is strong as all taxa share characters of fl ower morphology, i.e. fringeless corollas. Sect. Foliobracteola Cheng & Yueh, which in its original sense included the species related to T. kirilowii Maxim. and T. villosa Blume (Yueh and Cheng 1974; (Yueh and Cheng 1980), and T. smilacifolia C.Y.Wu (Jeff rey 1980b), is not supported, as the three latter species all end up elsewhere in the phylogenetic tree.

Molecular phylogeny
Trichosanthes villosa was placed in sect. Involucraria by Yueh and Cheng (1974), and later in subsect. Involucraria by Jeff rey (1980b), but subsequently moved to sect. Foliobracteola by Rugayah and de Wilde (1999). In the protologue of T. phonsenae Duyfj es and Pruesapan (2004), the authors stated that the three species T. kerrii, T. phonsenae and T. villosa form a coherent, distinct group based on presence of white fruit pulp, male fl owers with the stamens inserted low in the receptacle tube, and a pseudo-ovary. Th e molecular evidence shows that all accessions of these species in this study form a well-supported monophyletic group, confi rming the observations by Duyfj es and Pruesapan (2004) and warranting placement of these taxa in a new section, sect. Pseudovariifera H.J.de Boer. Cooper and Ford (2010) and Cooper and de Boer (2011) placed Trichosanthes subvelutina F.Muell. ex Cogn. in section Foliobracteola as it has obovate seeds with an entire broad marginal band similar to those found in Malesian species of section Foliobracteola. However, Huang et al (1997) proposed to place it in section Foliobracteola subsect. Villosae based on it pollen morphology. Th e current phylogenetic data place the accessions of T. subvelutina as sister (1.00/91) to a well-supported clade (1.00/84) consisting of sections Gymnopetalum, Trichosanthes, and Cucumeroides, and the species is here placed in a separate section, sect. Villosae (Yueh & L.Q.Huang) H.J.de Boer.
Th e other genera in the tribe Sicyoeae have colpate-colporate pollen, similar to that in many other distantly related groups in Curcurbitaceae (Schaefer and Renner 2011). In the light of the molecular data and the phylogenetic analysis (de Boer et al. Submitted), a transition from colporate to porate apertures has taken place three times in the evolutionary history of Trichosanthes, in the common ancestors of: 1) sect. Involucraria; 2) sect. Edulis; and 3) sections Trichosanthes, Gymnopetalum and Cucumeroides.

Taxonomy and classification
A revision of the infrageneric classifi cations suggested by previous authors on the basis of morphological studies (Yueh and Cheng 1974;Jeff rey 1980b;Huang et al. 1997;Rugayah and de Wilde 1999;de Wilde and Duyfj es 2004) is here proposed on the basis of the molecular phylogenetic data (de Boer et al. Submitted). A synopsis is presented in which we attempt to assign all 91 species recognized here to sections using the clades recovered in the phylogenetic analysis as a framework, along with a plenitude of data from macromorphological studies of herbarium vouchers (Chakravarty 1959 Rugayah (1999) Reinwardtia 11: 265 T. longispicata Rugayah (1999) (1854)