The genus Melothria L. (including Melancium Naudin, Cucumeropsis Naudin, and Posadaea Cogn.) includes 12–15 species, confined to arid plains, clearings and forest margins, grass- or woodlands from the southern United States through Central and South America down to northern Argentina (Schaefer and Renner 2011 a, b). One species, Melothria sphaerocarpa (Cogn.) H.Schaef. & S.S.Renner, is also found in tropical West Africa, where it might have arrived only recently through human-mediated transport (Schaefer and Renner 2010); and another species, Melothria pendula L., is locally invasive in tropical Asia (De Wilde and Duyfjes 2010). Melothria species are monoecious, small to medium-sized, herbaceous climbers with usually simple leaves, small yellow or white flowers, and the fruit a smooth, and often fleshy berry to 20 cm in Melothria sphaerocarpa. Each fruit contains numerous, strongly compressed seeds; and the testa tends to be smooth, white, and often covered by long appressed hairs (Schaefer and Renner 2011b).

In 1899, Alfred Cogniaux described a new species of Melothria based on herbarium material from the Caribbean island of Hispaniola (Haiti and the Dominican Republic), which he had received from the Krug and Urban collection (Urban 1899). The type material (L. Picarda 1503, BR) contains complete leaves and fruit plus seeds but no fully developed flowers. While the overall habit, leaf shape and fruit type are quite similar to Central American species of Melothria, the ovoid seeds would be unique in that genus and are very similar to another, very distantly related genus: Cayaponia Silva Manso in the tribe Cucurbiteae. And indeed, the Brazilian Cayaponia expert Vera Gomes-Klein annotated a specimen of Melothria domingensis Cogn. in the Berlin Herbarium (E. L. Ekman 1319, B) as “Cayaponia sp.” in July 1992; and two collections at New York Botanical Garden (P. Acevedo-Rodríguez et al. 13011 and 13274, NY) had originally been identified as Cayaponia racemosa (Mill.) Cogn. by M. T. Strong in 2010. The aim of our study here is to test this hypothesis using molecular and morphological data and to find out whether Melothria domingensis is indeed a misplaced Cayaponia or alternatively represents a unique type of seed morphology in Melothria.

Molecular analyses.Total genomic DNA was isolated from leaf samples of four herbarium specimens of Melothria domingensis (all collected in the Dominican Republic, Table 1) using a commercial plant DNA extraction kit (NucleoSpin, MACHEREY-NAGEL, Düren, Germany), and following the manufacturer’s manual. Polymerase chain reactions (PCR) following standard procedures were used to amplify the entire nuclear ribosomal ITS region plus the following five chloroplast regions: rbcL gene, trnL intron, and the three intergenic spacer regions trnL-trnF, rpl20-rps12, and trnH-psbA. PCR protocols and primers are given in Kocyan et al. (2007) and Schaefer et al. (2009). In addition, we used the primer pair trnH (5’- CGC GCA TGG TGG ATT CAC AAA TC) and psbA (5’- GTT ATG CAT GAA CGT AAT GCT C) designed by Sang et al. (1997) with an annealing temperature of 48°C to amplify the trnH-psbA spacer region. Crude PCR products were sent to Functional Biosciences, Inc. (Madison, WI, USA) for ExoSap cleaning and Sanger sequencing with the same primers used for PCR reactions. Seventeen sequences were newly generated for Melothria domingensis plus eleven ITS sequences for selected species of the genera Abobra Naudin, Calycophysum H. Karst & Triana, Cionosicys Griseb., Schizocarpum Schrad., and Tecunumania Standl. & Steyerm., all of the tribe Cucurbiteae, to obtain a more balanced sampling in our ITS matrix (for GenBank accession numbers and vouchers see table 1). Raw sequences were edited with Sequencher 4.9 (Gene Codes, Ann Arbor, Michigan, USA) and aligned by eye, using MacClade 4.08 (Maddison and Maddison 2003). We then added to those alignments sequences for all Cayaponia species available on Genbank (mainly from Kocyan et al. (2007), Schaefer et al. (2009), and Duchén and Renner (2010)) plus a set of Cucurbitaceae species representing all genera of the tribe Cucurbiteae, three representatives of Melothria and representatives of the genera containing Caribbean species based on Schaefer et al. (2008). Bayesian and maximum likelihood analyses were performed with a final dataset of 67 accessions representing 60 species.Data matrix and trees have been deposited in TreeBASE (http://www.treebase.org/ ) study number S13322.

Maximum likelihood (ML) analyses and non-parametric bootstrap searches (BS) with the fast-bootstrap algorithm were performed using RAxML-VI-HPC v. XX (Stamatakis et al. 2008). RAxML searches relied on the GTR + G + I model and model parameters were estimated over the duration of specified runs. Bayesian inference also used the GTR + G model (with the default four rate categories) and relied on MrBayes v. 3.2.1 x64 (Ronquist and Huelsenbeck 2003). We analyzed the combined dataset with two partitions (plastid and nuclear ITS), allowing partition models to vary by unlinking gamma shapes, transition matrices, and proportions of invariable sites. Markov chain Monte Carlo (MCMC) runs started from independent random trees, were repeated twice, and extended for ten million generations, with trees sampled every 1000th generation. Convergence was assessed using Tracer 1.5 (Rambaut and Drummond 2003). Trees saved prior to convergence were discarded as burn-in (2000 trees), and a consensus tree was built from the remaining trees.

Morphological analyses. We studied Melothria domingensis specimens from the following herbaria: BR, GH, NY, U, and US. For comparison, both authors also studied a large number of Cayaponia and other Neotropical Cucurbitaceae specimens from all major European and American herbaria over the past decade. All measurements given in the text are from dry herbarium specimens.

Analysis of pollination syndrome evolution and biogeography.We used the same approach as described in Duchén and Renner (2010), namely ancestral character state reconstruction under maximum likelihood in Mesquite v. 2.72 (Maddison and Maddison 2009) based on the Markov k-state one-parameter model. We added Melothria domingensis to the matrix of that previous study with the character states “bee pollination” (based on flower morphology) and “rainforest habitat” (based on herbarium label information).

Molecular and morphological evidence both show that Melothria domingensis is best placed in the Cayaponia clade and probably close to the North American Cayaponia quinqueloba. However, due to the poor resolution in parts of our Cayaponia phylogeny, we are unable to identify exact sister group relationships. Better taxon sampling and additional variable DNA markers are needed to solve this question.

The newly identified Cayaponia species from Hispaniola is a rain- or cloud forest inhabitant with small, probably diurnal flowers on short pedicels that match all characters for bee-pollination in Cayaponia discussed in Duchén and Renner (2010). It therefore does not fit to the pattern of pollinator shifts from bat- to bee-pollination as an adaptation to open savanna habitats hypothesized for Cayaponia by Duchén and Renner (2010) based on an analysis including 19 of the c. 60 species of Cayaponia. And indeed, our re-analysis of the pollination syndrome evolution in Cayaponia also questions the finding that bat pollination might be the ancestral condition in the clade. We find that based on the extended sampling, the most likely ancestral condition is bee pollination but with still 70% of the currently known Cayaponia species remaining to be sequenced, this pattern can easily change in the near future. We can confirm, however, at least one inferred shift from bat- to bee-pollination in Cayaponia espelina, which would be one of very few transitions in that direction (Duchén and Renner 2010; Van der Niet and Johnson 2012). Fieldwork on pollinators of Cayaponia espelina, a common species of the Brazilian Cerrado, would be needed to elucidate this interesting case further.

Cayaponia domingensis is one of only few species in the genus and in the entire family Cucurbitaceae that occur above 2000 m a.s.l. (the highest collections known are from 2100 m a.s.l., Liogier & Liogier 25690 (NY), A. Liogier 13134 (NY) and 2201 m a.s.l., P. Acevedo-Rodríguez et al. 13011 (NY)). The mechanisms that allow this species to thrive at those altitudes are unknown but might be interesting to study. Most Cucurbitaceae are highly frost-sensitive and survive cold periods only as seeds (e.g. North American Sicyos angulatus L.), rootstocks (e.g. Central European Bryonia dioica Jacq.), or tubers (e.g. Thladiantha dubia Bunge). The only perennial Cucurbitaceae that thrive at even higher altitudes are some Himalayan Thladiantha species that reach up to 3500 m a.s.l.

The mottled fruits characteristic for Cayaponia domingensis are very rare in the genus and might have evolved as an adaptation to a specialised seed dispersal agent on the island. Unfortunately, we lack field observations on dispersal agents and judging from fruit and seed size, a wide range of animals, including birds, lizards, and small mammals might be involved.

Recent transoceanic dispersal has been inferred for the ancestors of Cayaponia africana s.l. (Schaefer et al. 2009; Duchén and Renner 2010). While Duchén and Renner (2010) hypothesized a stepping-stone route via the Brazilian island Fernando de Noronha, our findings suggest that long-distance dispersal out of the Caribbean might be another option.

Our results highlight that even after two decades of Molecular Systematics, we still need more sequencing combined with morphological analyses to sort out taxonomic problems in Cucurbitaceae and other understudied families. We can now be confident about the phylogenetic position of this enigmatic Caribbean endemic. This is, however, only a first step and we now plan to do fieldwork on Hispaniola to obtain accurate distribution data, find out more about the ecology of this species, identify the threats to the remaining populations and ultimately develop a management plant to guarantee the survival of this unique endemic.