| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics and Phytogeography |
Department of Biology, 3507 Laclede Avenue, Saint Louis University, St. Louis, Missouri, 63103 USA; Department of Systematic Botany, University of Uppsala, Norbyvägen 18D, SE-752 36 Uppsala, Sweden
Received for publication October 30, 2006. Accepted for publication May 29, 2007.
ABSTRACT
Cucumis (Cucurbitaceae) comprises 33 species of annuals and perennials with a major native center of diversity in tropical and southern Africa. The genus includes some economically important and widely grown vegetables such as cucumbers and melons. Monophyly of the genus has been disputed in previous studies, but with only limited sampling. Relationships within Cucumis are thus poorly understood; moreover, the validity of the closely related genera has not been thoroughly tested. The present study was undertaken to test the monophyly of Cucumis and several closely related genera, to test sectional circumscriptions within Cucumis, and to understand the biogeographical history of the genus. We sequenced the nuclear ITS and plastid trnS-trnG regions for 40 ingroup and three outgroup taxa, representing all recognized subgenera and sections. Parsimony, maximum likelihood, and Bayesian analyses found Cucumella, Oreosyce, Mukia, Myrmecosicyos, and Dicaelospermum nested within Cucumis. The clades recovered within the Cucumis complex in some instances represent the first phylogenetically derived hypothesis of relationships, whereas others correspond to previous subgeneric and sectional classifications. At least four introductions from Africa to Asia, as well as one reintroduction to Africa, are suggested within the Cucumis complex. Cucumis sativus (cucumber) is strongly supported as sister to the eastern Asian C. hystrix, whereas C. melo (melon) is strongly supported as sister to C. sagittatus in southern Africa.
Key Words: Cucumella • Cucumis • Cucurbitaceae • Dicaelospermum • ITS • Mukia • Myrmecosicyos • Oreosyce • trnS-trnG
The genus Cucumis L. is one of the economically most important genera of flowering plants and includes many commonly grown vegetables (e.g., cucumber, melon, West Indian gherkin, African horned cucumber) as well as ornamentals (hedgehog gourd, gooseberry gourd). Next to tomatoes and onions, cucumbers and melons are the third most widely cultivated vegetable crops in the world (Pitrat et al., 1999
). Almost all domesticated species of Cucumis are susceptible to numerous fungal, bacterial, viral, and insect diseases (Whitaker and Davis, 1962
). Currently, a number of biotechnological studies are underway to find beneficial genes from wild relatives that have better resistance to such diseases (Chung et al., 2006
). Elucidating the phylogenetic relationships of the genus Cucumis is of great importance because those relationships can provide valuable information useful for collection and utilization of genetic resources that can be employed for crop improvement.
Cucumis belongs to the Cucurbitaceae subfamily Cucurbitoideae and is currently placed in the tribe Benincaseae (Jeffrey, 2005
). Since Kirkbride (1993)
recognized 32 species in the latest revision of the genus, one further species has been described (Thulin and Al-Gifri, 1994
). Species of Cucumis, as the genus is currently circumscribed, are characterized by a trailing, climbing, or bushy growth habit; simple petiolate and mostly three- to five-lobed leaf blades; unisexual (rarely androgynous) inflorescences; triplicate sigmoid anther thecae; and fleshy often tuberculate or spiny fruits with many compressed whitish seeds (Kirkbride, 1993
; Rubatzky and Yamaguchi, 1997
).
Taxonomists over the years have differed on delimitation of Cucumis, which was first described by Linnaeus in 1753
. Linnaeus recognized seven species in the genus, all of which were cultivated or economically useful. Three of these have been transferred to other genera, one is synonymized, and only three are still maintained. The type of the genus is C. sativus L., the cucumber. Numerous taxonomic treatments of Cucumis have been proposed since the work of Linnaeus (Pangalo, 1950
; Jeffrey, 1962
, 1967
, 1980
, 1990
; Kirkbride, 1993
). Pangalo (1950)
analyzed domesticated species of Cucumis and suggested that members of C. melo, the melon, deserved generic status. No other authors have adopted this scheme for the melons, but Jobst et al. (1988) proposed that Cucumis should be restricted to C. sativus alone, whereas all other species should be placed in Melo. The most comprehensive treatment of Cucumis was that of Kirkbride (1993)
. Making use of chromosomal, morphological, crossing, and geographical data, Kirkbride (1993)
has contributed greatly to the systematics of Cucumis. Based on his investigations, two subgenera are recognized, differing in shape of leaf blade lobes, geographical distribution, and basic chromosome number. Subgenus Melo (31 spp., n = 12, leaf blades with triangular lobes) is centered in Africa and is partitioned into two sections (Melo and Aculeatosi), whereas subg. Cucumis (two spp., n = 7, leaf blades with elliptic, oblong, or ovate lobes) is widely distributed in Asia. This classification was challenged by the recent discovery of n = 12 in the Asian C. hystrix, the second species in subg. Cucumis (Chen et al., 1997
), which suggests a possible bridge between the two subgenera.
Kirkbride (1993)
organized subg. Melo into two sections, each with three series. The two sections were merely "convenient morphological groupings that may or may not have evolutionary significance" (op. cit., p. 22) and correspond to the primary divisions of Naudin (1859)
and Cogniaux (1881). Kirkbride (1993)
based his sections solely on the presence or absence of "aculei," outgrowths of the outer layer of the female hypanthium and fruit. He based his six series on various classes of data, including cucurbitacin and flavonoid chemistry, isozymes, DNA, cytology, crossability, and morphology, and considered the series to be "significant evolutionary groupings."
Several other genera have been associated with Cucumis. These include Cucumella Chiov., Oreosyce Hook. f., Myrmecosicyos C. Jeffrey, Mukia Arn., and Dicaelospermum C. B. Clarke (Jeffrey, 1962
, 1980
, 1990
, 2005
; Kirkbride, 1994
; Kocyan et al., 2007). Cucumella comprises 11 species mostly distributed in Africa but with two species in Asia (Kirkbride, 1994
). Jeffrey (1962)
resurrected the genus and transferred species into it from Cucumis and other related genera (e.g., Kedrostis Medik., Hymenosicyos Chiov., Oreosyce). All subsequent authors (e.g., Keraudren, 1966
, 1967
; Fernandes and Fernandes, 1970
; Jeffrey, 1967
, 1980
, 1990
, 1995
; Jeffrey and Thulin, 1993
; Kirkbride, 1993
, 1994
) have maintained Cucumella as a genus, but C. Jeffrey (personal communication, 2006) and one of the current authors (M. Thulin) have expressed the opinion that Cucumis and Cucumella may not be monophyletic and should perhaps be merged. According to De Wilde and Duyfjes (2006)
, Mukia consists of six described species, one of which (M. gracilis) they elevated from varietal status. The genus is restricted to Asia except for M. maderaspatana, which is also widespread in Africa. Oreosyce and Myrmecosicyos, both endemic to Africa, and Dicaelospermum, from India, are all monotypic (Kocyan et al., 2007). Jeffrey (1980
, 1990
) listed these genera according to his opinion of their relatedness to Cucumis, i.e., Cucumella as most closely related, followed by Oreosyce, Myrmecosicyos, Mukia, and Dicaelospermum. In a new treatment, Dicaelospermum has been sunk into Mukia (De Wilde and Duyfjes, 2006
).
A number of cytological, biochemical, and molecular investigations at various taxonomic levels have included species of Cucumis (Dane, 1976
; Dane et al., 1980
; Perl-Treves and Galun, 1985
; Puchalski and Robinson, 1990
; Staub et al., 1992
, 1997
; Jobst et al., 1998
; Garcia-Mas et al., 2000
, 2004
). These earlier studies were primarily focused on the phylogenetic relationships among the domesticated species and their origin. None of them included adequate taxon sampling to elucidate relationships among the component subgenera and sections. Furthermore, the phylogenetic relationships of Cucumis and the related genera have been tested using only narrow taxon sampling and/or molecular markers from a single genome. For example, Garcia-Mas et al. (2004)
included Oreosyce africana in their study and found it nested within Cucumis. Their sampling included 17 species of Cucumis, but no representatives of other potentially closely related genera. Kocyan et al. (2007), on the other hand, presented a multilocus chloroplast phylogeny for the Cucurbitaceae that included all putative close relatives of Cucumis but only six species of Cucumis. Their results support a paraphyletic Cucumis, with Cucumella, Dicaelospermum, Mukia, Myrmecosicyos, and Oreosyce nested among species of this genus. Whether Oreosyce and the other small genera represent distinct lineages, however, remains untested within a larger sample of Cucumis species.
The biogeograpical history and evolution of Cucumis are also uncertain because of the lack of a genus-wide phylogeny. Subgenus Cucumis was long considered primitive, with subg. Melo believed to be derived from it through chromosomal fragmentation (Whitaker, 1933
; Ayyangar, 1967
). In contrast, cytological investigations suggested that ancestral species of subg. Melo gave rise to those of subg. Cucumis via processes of fusion or nonhomologous translocation (Trivedi and Roy, 1970
; Singh, 1990
). Ramachandran and Seshadri (1986)
, however, maintain that the two subgenera are not closely related because of differences in number, size, organization, and behavior of their chromosomes, as well as their geographical distribution.
Here we present a molecular phylogeny of Cucumis and the related genera based on sequences from both nuclear and chloroplast genomes. This study uses a much more complete sampling of species within Cucumis than did previous studies and includes representatives of Cucumella, Dicaelospermum, Mukia, Myrmecosicyos, and Oreosyce. The goals of the present study were to (1) test the monophyly of Cucumis; (2) determine whether the morphologically based infrageneric groupings within Cucumis, including the relationships of cucumbers and melons, are supported by the molecular data; (3) examine the conflicting hypotheses regarding the biogeographical history of Cucumis; and (4) test the generic limits of related taxa.
MATERIALS AND METHODS
Taxon sampling
Forty ingroup taxa were sampled for this study (Appendix 1) including 31 species of Cucumis, four species of Cucumella, two of Mukia, and one species each of Myrmecosicyos, Oreosyce, and Dicaelospermum. All currently recognized infrageneric taxa were represented, and subspecies and varieties of individual species were sampled wherever possible, resulting in a total of 47 accessions. The only missing species of Cucumis were C. prolatior Kirkbride (Kenya) and C. jeffreyanus Thulin (Ethiopia and Somalia). Leaf material of 19 taxa was collected from plants grown in the Saint Louis University greenhouse from seeds obtained from USDA/ARS North Central Regional Plant Introduction Station in Ames, Iowa, USA. Leaf material of the remaining taxa was obtained from various sources identified in Appendix 1. Citrullus lanatus, Cucurbita pepo, and Muellerargia timorensis were used as outgroups based on their proximity to Cucumis in the recent classification of the family (Jeffrey, 2005
) and the broader phylogenetic analyses of Cucurbitaceae (Jobst et al., 1998
; Kocyan et al., 2007), as well as on previous systematic studies in Cucumis (Garcia-Mas et al., 2004
).
DNA extraction, amplification, and sequencing
Genomic DNA was extracted from fresh leaves of seed-grown plants, herbarium specimens, or silica-gel-dried samples of field-collected leaf material. DNA was extracted from fresh material following the CTAB method (Doyle and Doyle, 1987
) and purified via cesium chloride/ethidium bromide gradients (Sambrook et al., 1989
). For herbarium specimen and silica-gel-dried material, DNA was isolated using the DNeasy Plant Mini Kit (Qiagen, Valencia, California, USA) following manufacturer's specifications.
The internal transcribed spacer (ITS) region of nuclear ribosomal DNA (including the intervening 5.8S gene) was used as a nuclear marker. The ITS region is the most widely used nuclear marker for phylogenetic inference at generic and infrageneric levels in plants (Kim et al., 1996
; Alvarez and Wendel, 2003
). Furthermore, this region has been used successfully to investigate infrageneric relationships in Cucumis (Garcia-Mas et al., 2004
). Several noncoding regions in the chloroplast genome are also useful for resolving infrageneric relationships in plants. One such region is the trnS-trnG region (Hamilton, 1999
; Gaskin and Schaal, 2003
; Shaw et al., 2005
). Preliminary results showed that this region was variable in Cucumis; therefore the trnS-trnG region was sequenced for this study.
The entire ITS1–5.8S-ITS2 region was amplified via a polymerase chain reaction (PCR) using primers ITS-1A (a modification of ITS5; Downie and Katz-Downie, 1996
) and ITS4 (White et al., 1990
). PCR was carried out in 50-µL reactions containing 2 µL purified sample DNA, 5 µL 10x buffer, 2.5 mM MgCl2 (5 µL of 25 mM stock), 25 µM (0.125 µL of 10 mM stock) of each dNTP (0.5 µL total), 0.4 µM (1 µL of 20 µM stock) of each primer, and 1 µL Taq polymerase. In addition, 2.5 µL dimethyl sulfoxide was added to reduce the effects of secondary structure on amplification and sequencing. Amplification began with an initial 3-min denaturation step at 95°C, followed by 35 cycles of 30 s at 95°C, 30 s at 55°C, and 45 s at 72°C. This was followed by a final 7-min extension at 72°C. Amplification products were checked on 1% agarose gels and visualized under UV after ethidium bromide staining. PCR products were purified using the QIAquick PCR purification kit (Qiagen) following manufacturer's recommendations. Sequencing of PCR products was performed by Macrogen, Inc. (Seoul, South Korea) using the same primers as in the PCR reaction.
In preliminary analyses, four taxa were placed differently in the nuclear and chloroplast phylogenies. Therefore, we cloned ITS for Cucumis globosus, C. prophetarum, C. metuliferus, and C. rostratus to test for the presence of multiple ITS types that might potentially confound phylogenetic analyses. The purified ITS templates were ligated to the pGEM-T Easy vector (Promega, Madison, Wisconsin, USA) following the manufacturer's protocol and subsequently transformed into competent cells. After overnight culture at 37°C on LB-ampicillin/IPTG/X-gal selective plates, colonies carrying the ITS insert (white) were arbitrarily selected and used as template for PCR amplification using the protocol outlined earlier. Five clones from each species were sequenced.
The trnS-trnG region was amplified using primers trnG2G-F (5'-GCG GGT ATA GTT TAG TGG TAA AA-3') and trnG-R (5'-GTA GCG GGA ATC GAA CCC GCA TC-3') of Shaw et al. (2005)
. Reactions of 50 µL contained 2 µL purified sample DNA, 5 µL 10x buffer, 2.5 mM MgCl2 (5 µL of 25 mM stock), 25 µM (0.125 µL of 10 mM stock) of each dNTP (0.5 µL total), 0.4 µM (1 µL of 20µM stock) of each primer, 1 µL Taq polymerase, and 1 µL dimethyl sulfoxide. The thermal cycle protocol included an initial denaturing at 94°C for 4 min; 40 cycles at 94°C for 45 s, 52°C for 1 min, 72°C for 1 min; ending with an extension at 72°C for 7 min. PCR products were purified as described for ITS. The PCR products were sequenced using the same primers as for amplification.
Consensus sequences were assembled and edited with the program Sequencher (version 4.2; Gene Codes Corp., Ann Arbor, Michigan, USA) followed by manual adjustments in Se-Al (version 2.0a11; Rambaut, 1996–2002
). GenBank accession numbers for all sequences are given in Appendix 1. The data matrix and resulting trees are archived in TreeBase (SN 1797, M3281, M3282) at http://www.treebase.org/treebase/.
Phylogenetic analyses
Phylogenetic analyses were performed under maximum parsimony (MP), maximum likelihood (ML), and Bayesian inference. For the parsimony methods, we coded indels as characters, following the simple indel coding method of Simmons and Ochoterena (2000)
. Parsimony analyses were conducted in PAUP* 4.0b10 (Swofford, 2002
), excluding uninformative characters and with all characters equally weighted. Heuristic searches used the tree-bisection-reconnection (TBR) branch swapping option, with 1000 random addition sequence replicates. Bootstrap (BS) support was calculated from 10 000 pseudoreplicates (Felsenstein, 1985
) using TBR branch swapping and simple addition of taxa, but saving only one tree per replicate.
For ML searches, a best-fit model of nucleotide substitution and model parameters for the ITS, trnS-trnG, and combined data sets were determined using ModelTest 3.06 (Posada and Crandall, 1998
). A heuristic search with TBR branch swapping option under ML optimization was implemented using PAUP* with all characters included. Subsequently, 100 bootstrap replicate analyses were performed. Each data set was also analyzed under Bayesian inference using MrBayes (vers. 3.0b4; Ronquist and Huelsenbeck, 2003
), running four simultaneous chains for 106 generations and sampling trees every 100 generations. The number of generations needed to reach stationarity was determined by plotting likelihood scores against generation. The trees sampled in this burn-in stage were excluded, and the remaining trees were saved to a file. Three independent runs were performed for each data partition to ensure that analyses converged on the optimal set of trees. A majority consensus tree was generated from all optimal trees from the independent runs.
Statistical tests of incongruence
Data set incongruence was tested under parsimony using partition homogeneity tests (incongruence length difference, ILD; Farris et al., 1994
, 1995
) as implemented in PAUP*. The ILD test was conducted with 100 replicates, each with 10 random addition sequence replicates, TBR branch swapping, and the MulTrees option. Following Cunningham (1997)
, a significance level of P = 0.01 was adopted for this test. Competing tree topologies were also compared using a one-tailed Shimodaira–Hasegawa (SH) test (Shimodaira and Hasegawa, 1999
; Goldman et al., 2000
) as implemented in PAUP* (RELL method; 1000 bootstrap replicates). For the SH test, likelihood scores for the best tree from each Bayesian analysis were compared with 50 trees with the best likelihood scores from the rival data set, using a critical value of p = 0.05.
RESULTS
Phylogenetic analysis of ITS
For the four taxa with conflicting placements between ITS and trnS-trnG, five clones were sequenced and compared to the direct sequences obtained earlier. Sequence divergence between clones and direct sequences for each taxon was less than 1%, suggesting that it is unlikely that the clones represent different ITS types. Furthermore, all clones clustered with the associated direct sequence in parsimony analysis. We therefore included only the direct sequence in all subsequent analyses.
The length of the entire ITS1–5.8S-ITS2 region ranged from 747 to 784 bp. The resulting aligned data matrix was 813 characters long, of which 197 (24.2%) were potentially parsimony informative. Scoring indels as characters resulted in an additional 63 informative characters. Parsimony analysis yielded 16 100 equally parsimonious trees of 516 steps with consistency index (CI) of 0.516 and retention index (RI) of 0.748 (Table 1); the strict consensus of these trees is shown in Fig. 1A. Twenty-seven nodes in the parsimony analysis of the ITS data partition received 
50% bootstrap support (Table 1).
|
|
0.95 in the Bayesian analysis except one were also found in the corresponding parsimony analysis (Fig. 1A).
Phylogenetic analysis of trnS-trnG
The length of trnS-trnG sequences varied from 741 to 788 bp. The aligned matrix consisted of 818 characters, 110 (13.4%) of which were phylogenetically informative (Table 1). Scoring indels as characters resulted in an additional 18 informative characters. Parsimony analysis found 10 870 minimum length trees of 210 steps (CI = 0.610, RI = 0.821); the strict consensus of these trees is shown in Fig. 1B. A total of 20 nodes had support 50% in the MP bootstrap analysis (Table 1).
Modeltest 3.06 selected TIM+G as the best fit model for the trnS-trnG data. The ML search resulted in a single tree (not shown) that was topologically congruent with the strict consensus of parsimony trees (Fig. 1B). Nineteen of the 20 clades with 50% BS in the parsimony analysis were also recovered in the ML analysis. Bayesian analysis of the trnS-trnG data resulted in stabilization of the likelihood scores at ca. 22 500 generations; thus, 225 trees were discarded as burn-in. The consensus of trees remaining after discarding the burn-in was topologically congruent with that recovered by MP and ML analyses (Fig. 1B).
Congruence between data partitions
The ILD test found that the ITS and trnS-trnG data partitions were significantly incongruent (P = 0.001). Inspection of topological differences between individual analyses identified four taxa (C. globosus, C. prophetarum, C. metuliferus, and C. rostratus) that had different positions in the ITS and cpDNA phylogenies (Fig. 1). To investigate the impact of these differences in species placement, an ILD test was conducted with the four taxa removed. This second test yielded a nonsignificant P value (P = 0.03), suggesting that the differing placements of these four taxa contribute substantially to heterogeneity in the data set. Similarly, the SH test comparing trees of the individual data sets suggested incongruence between the two data partitions (P = 0.03). However, when the SH test was repeated with C. globosus, C. prophetarum, C. metuliferus, and C. rostratus removed, the test did not reject homogeneity of ITS vs. trnS-trnG (P = 0.08) and of trnS-trnG vs. ITS (P = 0.09).
In two of the four cases, the positions of the taxa that account for significant incongruence vary within the same major clade. In the ITS tree (Fig. 1A), for instance, C. prophetarum is sister to C. aculeatus, whereas C. globosus is basal in the clade containing the nondomesticated African taxa. In the trnS-trnG tree (Fig. 1B), however, C. globosus is sister to C. aculeatus and C. prophetarum is unresolved. In both trees, both C. globosus and C. prophetarum are part of a large clade comprising nondomesticated African species. In the third case, C. metuliferus is sister to C. rostratus in the ITS tree, but its alternative placement in the cpDNA tree is unresolved. Similarly, the alternative placement of C. rostratus in the cpDNA tree is unresolved within a major clade. Given the overall similarity of topology between trees inferred from the ITS and trnS-trnG data sets, as well as the insignificant level of incongruence when the four conflicting taxa are excluded, we chose to combine the two data sets. Several studies have used this approach in similar situations (Berry et al., 2004
; Levin et al., 2006
; Lohmann, 2006
). We performed analyses of the combined data partitions on (1) the complete set of taxa, and (2) the complete set of taxa minus C. globosus, C. prophetarum, C. metuliferus, and C. rostratus. Analyses of these two data matrices resulted in trees that are very similar in overall tree topology and support values. As a result, all final analyses used the combined data matrix.
Combined molecular analyses
The combined ITS and trnS-trnG data matrix contained 1631 characters with 307 (18.8%) potentially parsimony informative characters (Table 1). Scoring indels as characters resulted in an additional 80 informative characters. Parsimony analysis found 7684 minimum length trees of 763 steps (CI = 0.497, RI = 0.744); the strict consensus of these trees is shown in Fig. 2. Twenty-nine nodes in the MP analyses had bootstrap support 50% (Table 1).
|
DISCUSSION
Comparison of ITS, trnS-trnG, and combined data
The ITS and trnS-trnG data sets resulted in largely similar topologies. In terms of phylogenetic utility, the ITS region has a higher proportion of parsimony informative (PI) characters (Table 1). The trnS-trnG region was also phylogenetically informative but had a lower proportion of PI characters. The higher CI and RI values, and hence lower level of homoplasy, in the trnS-trnG than in the ITS region (Table 1), however, increased the utility of the trnS-trnG region for detecting interspecific variation in our study. These results are consistent with previous studies of the trnS-trnG region for the tribe Mesechiteae (Apocynaceae) by Simóes et al. (2004)
and for the genus Solanum (Solanaceae) by Levin et al. (2005)
.
The principal conflicts between the strict consensus trees for ITS and trnS-trnG data sets were in the placements of four African taxa (indicated by arrows, Fig. 1). The first of these is C. prophetarum. In the ITS tree, this taxon was sister to C. aculeatus, with very strong support (Fig. 1A), whereas in the trnS-trnG tree, it was unresolved but fell within the same major clade (Fig. 1B). The second conflicting placement is that of C. globosus. This taxon was resolved as sister to C. aculeatus with a strong support (Fig. 1B) in the trnS-trnG tree. In the ITS tree, C. globosus was moderately supported as sister to the largest clade in the tree (Fig. 1A). The third conflicting placement is the alternative position of C. rostratus as either sister to C. metuliferus, with a strong support, in the ITS analyses (Fig. 1A), or unresolved within the largest clade in the trnS-trnG analyses (Fig. 1B). The final conflicting placement is that of C. metuliferus. This species is weakly supported as sister to a clade comprising members of the C. melo group (Fig. 1B) in the trnS-trnG tree. In the ITS tree, however, C. metuliferus forms a strongly supported sister relationship with C. rostratus (Fig. 1A).
Incongruence between phylogenies based upon nuclear and chloroplast markers may result from a number of causes, including hybridization (Rieseberg et al., 1996
; Baldwin, 1997
; Barber et al., 2000
, 2007
; Levin, 2000
; Rieseberg et al., 2003
). Several previous evolutionary studies in Cucumis have suggested that hybridization has played a role in its evolution (Dane et al., 1980
; Singh and Roy, 1974
; Singh and Yadava, 1984
; Staub et al., 1992
; Kirkbride, 1993
; Staub et al., 1997
). In addition, hybrids among members of the nondomesticated African Cucumis have been documented (Dane et al., 1980
; Raamsdonk et al., 1989
). The four incidences of incongruence between the ITS and trnS-trnG trees in the present study all involve members of the wild African group. Therefore, incongruence may be due to chloroplast capture through hybridization. Additional studies with more extensive sampling, including multiple samples of these taxa, are necessary to confidently determine the cause of these incongruences.
The combined analysis of ITS and trnS-trnG data sets produced the most highly resolved and well-supported tree yet achieved for Cucumis s.s. and its related genera (Fig. 2). Parsimony, likelihood, and Bayesian analyses resulted in similar topologies and resolved 29, 30, and 28 nodes, respectively, with bootstrap support 50% and posterior probabilities 0.95. Given the overall similarity of the ITS and trnS-trnG topologies and better resolution of clades in the combined analyses, all further discussion of the results is based on the combined analyses.
Taxonomic implications
Results from parsimony, likelihood, and Bayesian analyses of the combined data sets show that Cucumis sensu Kirkbride (1993)
is paraphyletic (Fig. 2). All sampled representatives of Cucumella, Dicaelospermum, Mukia, Myrmecosicyos, and Oreosyce are resolved within Cucumis as presently delimited. A morphological synapomorphy that seems to unite the group is 3-colporate oblate-suboblate pollen grains (Marticorena, 1963
; Jeffrey, 1964
; De Wilde and Duyfjes, 2006
). Cucumis sensu Kirkbride is distinguished from the rest in having triplicate sigmoid anther thecae vs. straight or arcuate anther thecae in Cucumella, Dicaelospermum, Mukia, Myrmecosicyos, and Oreosyce (Jeffrey, 1962
; Kirkbride, 1994
). The disposition of taxa with straight anther thecae among several clades comprising primarily species with triplicate sigmoid anthers suggests that straight or arcuate anther thecae have evolved three times in the group.
Our results agree with those of Kocyan et al. (2007)
. Their study used cpDNA markers only and was focused on the entire family; thus, species-level sampling was sparser than in our study. Nevertheless, they also recovered a well-supported sister relationship between Cucumella and Oreosyce, as well as between Dicaelospermum and Mukia (the latter without support noted). With increased sampling among Cucumella and Mukia, our study found Oreosyce nested within Cucumella, and Dicaelospermum embedded within Mukia (Fig. 2). Kocyan et al. (2007) recovered the Madagascan-Malesian genus Muellerargia Cogn. as sister to the Cucumis complex, which is the reason we use it as outgroup. Morphologically, anther thecae in Muellerargia are straight; thus the presence of triplicate sigmoid anther thecae remains a synapomorphy for Cucumis in its new, wider sense, although with reversals in Clades IIB and V (Fig. 2), and a change to arcuate anther thecae in Myrmecosicyos in Clade IA.
Based on our analyses, taxonomic changes in the Cucumis complex appear to be inevitable. We propose that the genus Cucumis be expanded to also include the nested genera Cucumella, Dicaelospermum, Mukia, Myrmecosicyos, and Oreosyce. However, the formal nomenclatural changes will be made elsewhere.
Relationships within the Cucumis complex
Using ITS sequence data, Garcia-Mas et al. (2004)
split Cucumis into five groups: cucumbers, melons, C. metuliferus, a group containing 12 wild species of Cucumis and Oreosyce africana, and a fifth group comprising C. sagittatus and C. globosus. Our ITS results (with 82% increase in taxon sampling) largely support their findings; however, two well-supported conflicts were found. In their study, C. globosus is very strongly supported (MP = 99) as sister to C. sagittatus. In our ITS analyses, C. globosus is sister to all other species in Clade I with weak support, whereas C. sagittatus is strongly supported as sister to members of C. melo (Fig. 1A).
Our combined phylogenetic analyses do not support Kirkbride's (1993) subdivision of Cucumis into two subgenera based principally upon chromosome number and geographical distribution. Cucumis s.s. is paraphyletic in our analyses; moreover, the Asian Cucumis (subg. Cucumis of Kirkbride) form a well-supported clade (Clade IIA in Fig. 2) nested within African Cucumis s.s. Our combined tree identified six clades within the Cucumis complex, designated here as Clades I, II, III, IV, V, and VI (Fig. 2), which we discuss next.
Clade I
This clade comprises all nondomesticated African Cucumis s.s. after the exclusion of C. hirsutus, C. humifructus, C. metuliferus, C. rostratus, and C. sagittatus, and is moderately to strongly supported in our combined analyses (MP = 71; ML = 68; Bayes = 96; Fig. 2). The clade includes C. canoxyi from the Arabian Peninsula and the Afro-Asian C. prophetarum. This lineage was considered a natural group by Dane (1976
, 1983
), Esquinas-Alcazar (1977)
, Perl-Treves and Galun (1985)
, and Jeffrey (1980), and corresponds to Kirkbride's (1993) ser. Myriocarpi and ser. Angurioidei. Clade I is supported by a combination of characters: elliptic seeds, the presence of cucurbitacin D (Esquinas-Alcazar, 1977
), and the absence of chloroplast DNA restriction number 20 of restriction endonuclease pvuII (Perl-Treves and Galun, 1985
). Section Aculeatosi as recognized by Kirkbride (1993)
, including the members of both clade I and clade IV, is clearly paraphyletic in our analyses. The inclusion of Myrmecosicyos messorius within this clade suggests that the genus Myrmecosicyos is not distinct from Cucumis. Cucumis prolatior and C. jeffreyanus (unsampled in our study) were placed in ser. Angurioidei by Kirkbride (1993)
and would almost certainly fall into clade I.
Within clade I, a group comprising seven species—C. africanus, C. kalahariensis, C. heptadactylus, C. myriocarpus, C. quintanilhae, C. zeyheri, and Myrmecosicyos messorius (Clade IA in Fig. 2)—is very strongly supported as monophyletic (MP = 97; ML = 98; Bayes = 100). Members of this subclade are geographically restricted to southern Africa (except Myrmecosicyo, which is endemic to Kenya), and the group was recognized by Kirkbride (1993)
as ser. Myriocarpi, characterized by a lack of cucurbitacin A. However, he did not include C. zeyheri, which was placed in ser. Angurioidei. The paraphyletic ser. Angurioidei otherwise included the members of clade I except clade IA and was characterized by presence of cucurbitacin A (Kirkbride, 1993
). This characteristic, however, is also exhibited by some members of Clade IA (e.g., C. heptadactylus; Esquinas-Alcazar, 1977
).
Clade II
This clade received strong support (MP = 99; ML = 98; Bayes = 100; Fig. 2) in our combined analyses and consists almost exclusively of Asian taxa. Within this clade, two strongly supported subclades are resolved. The first of these (clade IIA) comprises the Asian species of Cucumis sensu Kirkbride. Kirkbride (1993)
placed C. hystrix in the Asian subg. Cucumis because of morphological similarities to C. sativus and its distribution in China, India, and Thailand. Recently, Chen et al. (1997)
reported that C. hystrix possesses the same chromosome number as species in subg. Melo (n = 12), challenging the existing classification. Nevertheless, our analyses strongly support both the placement of C. hystrix as sister to C. sativus and the monophyly of the Asian species (MP = 100; ML = 100; Bayes = 100; Fig. 2). This finding is consistent with previous suggestions by several workers (Jeffrey, 1980
; Kirkbride, 1993
; Garcia-Mas et al., 2004
, among others) that the Asian Cucumis form a natural group, but it should not be considered a subgenus within Cucumis. The position of C. sativus supports the hypothesis that the base chromosome number x = 7 is derived from x = 12 (Trivedi and Roy, 1970
; Raamsdonk et al., 1989
; Singh, 1990
; Kirkbride, 1993
), possibly by chromosomal fusion. A new synthetic artificial species with 2n = 38, formed from a chromosome-doubled hybrid between C. hystrix and C. sativus, was described as xC. hytivus by Chen and Kirkbride (2001)
.
The second subclade (clade IIB) comprises two representatives of Mukia (both Asian but M. maderaspatana is also widespread in Africa) and the monotypic Dicaelospermum (India) and received very strong support in our analyses (MP = 100; ML = 100; Bayes = 100; Fig. 2). Our data support the transfer of D. ritchiei to Mukia by De Wilde and Duyfjes (2006)
, because Mukia is paraphyletic in our analyses with respect to Dicaelospermum (Fig. 2). After it had been sunk into the genus Melothria by Cogniaux (1881)
, Jeffrey (1962)
reestablished the genus Mukia, characterized by the presence of tumid seeds and clustered female flowers. A position of Dicaelospermum close to Mukia was established by Singh (1965)
; together, they form a group that differs from clade IIA by having straight (not triplicate sigmoid) anther thecae.
Clade III
This group comprises C. melo and its varieties plus C. sagittatus and is very strongly supported in our analyses (MP = 98; ML = 99; Bayes = 100; Fig. 2). It corresponds to series Melo of Kirkbride (1993)
. Members of this clade are monoecious or andromonoecious plants with sessile male inflorescences and glabrous filaments. While all members of C. melo contain cucurbitacin B, C. sagittatus is anomalous in the group because of the presence of cucurbitacin F (Enslin and Rehm, 1957
). However, both species differ from the members of Clade IIA that contain cucurbitacin C.
Clade IV
This clade, composed of C. metuliferus and C. rostratus, corresponds to ser. Metuliferi of Kirkbride (1993)
but received only weak support in our combined analyses. Series Metuliferi is characterized by ovate seeds and an absence of cucurbitacin D. Previous investigations have been at odds as to the relationship of this clade within Cucumis. Perl-Treves and Galun (1985)
suggested that C. metuliferus had affinities with C. hirsutus and C. humifructus (our clade VI; Fig. 2). On the other hand, Kirkbride (1993)
placed members of clade II in sect. Aculeatosi along with members of clade I. In our combined molecular analyses, clade IV is weakly supported as sister to clades II and III together. Cucumis metuliferus is easily distinguishable from other Cucumis species because it is the only monoecious annual that bears large red spiny fruits (Jeffrey, 1980
). Cucumis rostratus is unique in the genus in that it is found in dense forest environments; all other members of the genus occupy drier open habitats (Kirkbride, 1993
). Furthermore, it is the only taxon in the genus with spindle-shaped fruits that taper strongly at both ends.
Clade V
Although only four of the 11 presently recognized species of Cucumella were sampled for this study, this strongly supported clade clearly suggests that Cucumella is monophyletic only if the monotypic Oreosyce is included in it. Further, both Cucumella and Oreosyce are nested within Cucumis and therefore do not deserve generic level recognition. Morphologically, members of Cucumis sensu Kirkbride are separated from those of Cucumella by only one character: species of Cucumis bear triplicate, sigmoid anther thecae, whereas species of Cucumella have straight anther thecae, with the apex sometimes bent over toward the anther connective. Oreosyce is similar to Cucumella in the shape of the anther thecae; however, it is distinguished from Cucumella, which has lenticular compressed seeds, by having wide-margined, depressed seeds.
Clade VI
This clade comprises C. hirsutus and C. humifructus and is very strongly supported in our analyses (MP = 100; ML = 100; Bayes = 100; Fig. 2). Furthermore, it is sister to the rest of Cucumis with strong support. Both species were treated by Kirkbride (1993)
as members of sect. Melo of subg. Melo. Cucumis hirsutus is anomalous in the genus in that it is the only dioecious taxon with pedunculate male inflorescences. Cucumis humifructus is the only Cucumis species having geocarpic fruit. Based on these morphological features, Kirkbride (1993)
placed these species in two separate series: ser. Hirsuti and ser. Humifructosi, respectively.
Geographic implications
Most members of the Cucumis complex are distributed in Africa, but species with an Asian or largely Asian distribution are found (or are predicted to be found) in four clades. In clade I (Fig. 2), the Afro-Asian C. prophetarum and C. canoxyi, endemic to the Arabian Peninsula, are nested among various African species, which suggests dispersal from Africa to Asia.
The monophyly of the group of Asian species in clade II (Fig. 2), and the nested position of the group among African taxa, suggests that this radiation has evolved from a single introduction to Asia from Africa. Furthermore, the nested position of Afro-Asian Mukia maderaspatana within this clade suggests dispersal of this species back to Africa. The position of Cucumis sativus within clade II supports the view that its aberrant chromosome number with n = 7 has been derived from n = 12, probably by fusion, because n = 12 is the most common number in Cucurbitaceae and the only other number known in the Cucumis complex (Trivedi and Roy, 1970
; Jeffrey, 1980
; Yadava et al., 1984
; Singh, 1990
; Kirkbride, 1993
). Furthermore, Esquinas-Alcazar (1997) suggested that fixation of some gene loci was found only in the Asian C. sativus, which suggests genetic drift or migration and establishment of populations giving rise to n = 7 (Kirkbride, 1993
).
In clade III (Fig. 2), C. melo is supported as sister to C. sagittatus. Wild forms of C. melo are widespread in Africa and Asia (Jeffrey, 1980
), whereas C. sagittatus is restricted to southern Africa (Kirkbride, 1993
). The position of C. melo as sister to C. sagittatus suggests an African origin of C. melo with dispersal to Asia.
Clade V (Fig. 2) consists of solely African species, but two closely related Indian species of Cucumella exist: C. ritchiei (Chakrav.) C. Jeffrey and C. silentvalleyi Manilal, T. Sabu & P. Mathew. They are known only from their types and thus could not be sampled, but judging from morphology, they would likely fall into clade V and constitute yet another example of Asian taxa with an African ancestry within the Cucumis complex.
Conclusions
Using maximum parsimony, maximum likelihood, and Bayesian analyses of sequence data from both the nuclear and chloroplast genomes, we have provided the first comprehensive phylogeny of Cucumis and the traditionally related genera. Our data clearly show that Cucumella, Dicaelospermum, Mukia, Myrmecosicyos, and Oreosyce are nested within Cucumis. Some clades recovered within the Cucumis complex represent the first phylogenetically derived hypotheses of relationships, whereas others either mirror or contradict previous morphologically based hypotheses of relationships. Finally, our results suggest at least four introductions from Africa to Asia, as well as one reintroduction to Africa, within the Cucumis complex.
APPENDIX
Taxa, collection locations, voucher numbers, and GenBank accession numbers for Cucumis, Cucumella, Dicaelospermum, Mukia, Myrmecosicyos, Oreosyce, and outgroup ITS and trnS-trnG sequences. Infrageneric circumscriptions are according to Kirkbride (1993)
. Voucher specimens are deposited in the following herbaria: Missouri Botanical Garden (MO); Namibian National Botanical Research Institute (WIND); Royal Forest Department, Bangkok, Thailand (BKF); University of Asmara-Eritrea, Asmara, Eritrea (UAE); Uppsala University, Uppsala, Sweden (UPS); National herbarium of the Netherlands-Leiden (L); Royal Botanic Gardens, Kew (K); East African Herbarium, Nairobi, Kenya (EA).
Taxon—Collection location, Voucher (herbarium); GenBank accession numbers: ITS; trnS-trnG.
Outgroup species
Citrullus lanatus (Thunb.) Matsum. & Nakai—Burkina Faso: Oudalan; Madsen s.n. (MO); EF595859, EF595909.
Cucurbita pepo L.—USA: USDA Torrey Pines Station, Lajolla, California; TCA s.n. (MO); EF595858, EF595908.
Muellerargia timorensis Cogn.—Papua New Guinea: Central Province (Port Moresby); Pullen 6848 (L); EF595860, EF595910.
Ingroup species
Cucumella Chiov.—Subg. Cucumella—Sect. Aculeatae J. H. Kirkbr—C. aspera (Cogn.) C. Jeffrey—Namibia: Omaruru; Giess et al. 6238 (MO); EF595861, EF595911.
Cucumella Chiov.—Subg. Cucumella—Sect. Cucumella.—C. kelleri (Cogn.) C. Jeffrey—Somalia: Eyl; Thulin et al. 10578 (UPS); EF595864, EF595914.
Cucumella Chiov. —Subg. Leptopericarpia J. H. Kirkbr.—C. bryoniifolia (Merxm.) C. Jeffrey—Zimbabwe: Nyamunyeche; Ryasiri 633 (MO); EF595862, EF595912. C. cinerea (Cogn.) C. Jeffrey—Namibia: Okahandja; Maggis et al. 1026 (WIND); EF595863, EF595913.
Cucumis L.—Subg. Melo (Mill.) C. Jeffrey—Sect. Aculeatosi J. H. Kirkbr.—Ser. Myriocarpi J. H. Kirkbr.—C. africanus L. f.—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1009 (MO); EF595866, EF595916. C. heptadactylus Naudin—South Africa; Degr (MO); EF595876, EF595926. C. kalahariensis A. Meeuse—Namibia: Okavango, Rundu; Maggis 1036 (WIND); EF595881, EF595931. C. myriocarpus Naudin—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1010 (MO); EF595888, EF595938. C. quintanilhae R. Fern. & A. Fern.—Botswana: Beitbridge; AGG 019 (UAE); EF595892, EF595942.
Cucumis L.—Subg. Melo (Mill.) C. Jeffrey—Sect. Aculeatosi J. H. Kirkbr.—Ser. Angurioidei J. H. Kirkbr.—C. aculeatus Cogn.—Tanzania: Saba Saba area; Kuchar 22468 (MO); EF595865, EF595915. C. anguria L.—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1012 (MO); EF595867, EF595917. C. anguria L. var. anguria—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1013 (MO); EF595868, EF595918. C. baladensis Thulin—Somalia: near Mogadishu; Thulin et al. 7464 (UPS); EF595869, EF595919. C. carolinus J. H. Kirkbr.—France: Institut de la Recherche Agronomique, Avignon; AGG 1017 (MO); EF595871, EF595921. C. canoxyi M. Thulin and A. N. Al-Gifri—Yemen: Hadramaut; Thulin et al. 9864 (UPS); EF595870, EF595920. C. dipsaceus Ehrenb. ex. Spach—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1021 (MO); EF595872, EF595922. C. ficifolius A. Rich.—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1019 (MO); EF595873, EF595923. C. globosus C. Jeffrey—Tanzania: Iringa, Mufindi district; Kayombo 681 (MO); EF595874, EF595924. C. hastatus Thulin—Somalia: near Gobyaxas; Kuchar 17327 (K); EF595875, EF595925. C. insignis C. Jeffrey—Ethiopia: Sidamo; Gilbert et al. 8061 (UPS); EF595880, EF595930. C. meeusei C. Jeffrey—USA: USDA Torrey Pines Station, Lajolla, California; AGG 1022 (MO); EF595882, EF595932. C. prophetarum L.—Yemen: Jabal al-Arays; Thulin et al. 8486 (UPS); EF595889, EF595939. C. pubituberculatus Thulin—Somalia: NE Mogadishu; Thulin 6321 (UPS); EF595890, EF595940. C. pustulatus Hook. f.—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1023 (MO); EF595891, EF595941. C. rigidus E. Meyer ex. Sond.—South Africa: Namaqualand; Pearson 4023 (MO); EF595893, EF595943. C. sacleuxii Paill. & Bois—Tanzania: Kidatu; Mhoro 727 (UPS); EF595895, EF595945. C. thulinianus J. H. Kirkbr.—Somalia: near Erigavo; Yohaness 3611 (UAE); EF595900, EF595950. C. zeyheri Sond.—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1027 (MO); EF595901, EF595951.
Cucumis L.—Subg. Melo (Mill.) C. Jeffrey—Sect. Aculeatosi J. H. Kirkbr.—Ser. Metuliferi J. H. Kirkbr.—C. metuliferus E. Mey. ex Naudin—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1024 (MO); EF595887, EF595937. C.rostratus J. H. Kirkbr.—Nigeria: SW Lagos; Babuker 8712 (MO); EF595894, EF595944.
Cucumis L.—Subg. Melo (Mill.) C. Jeffrey—Sect. Melo (Mill.) J. H. Kirkbr.—Ser. Hirsuti J. H. Kirkbr.—C. hirsutus Sond.—Congo: Orientale, Parc Nationale de la Garamba; De Saeger 1323 (MO); EF595877, EF595927.
Cucumis L.—Subg. Melo (Mill.) C. Jeffrey—Sect. Melo (Mill.) J. H. Kirkbr.—Ser. Humifructosi J. H. Kirkbr.—C. humifructus Stent.—Namibia: Okavango, Rundu; Strohbach 5630 (WIND); EF595878, EF595928.
Cucumis L.—Subg. Melo (Mill.) C. Jeffrey—Sect. Melo (Mill.) J. H. Kirkbride—Ser. Melo (Mill.) J. H. Kirkbride—C. melo L. subsp.agrestis (Naudin) Pangalo—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1028 (MO); EF595883, EF595933. C. melo L. var.cantalupensis Naudin—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1029 (MO); EF595884, EF595934. C. melo L. var. inodorus H. Jacq.—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1030 (MO); EF595885, EF595935. C. melo L. subsp.melo—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1031 (MO); EF595886, EF595936. C. sagittatus Peyr.—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1032 (MO); EF595896, EF595946.
Cucumis L.—Subg. Cucumis—C. hystrix Chakrav.—Thailand, Taeng nam, Chiang Mai; Phonesa et al. 3931 (BKF); EF595879, EF595929. C. sativus L.—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1033 (MO); EF595897, EF595947. C. sativus L. var.hardwickii (Royle) Gabaev—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1034 (MO); EF595898, EF595948. C. sativus L. var.xishuangbannanesis QiChunzhang & Yuan Zhenzhen, ined.—USA: Ames, Iowa, USDA, ARS Regional Plant Introduction Station; AGG 1035 (MO); EF595899, EF595949.
Dicaelospermum C. B. Clarke—D. ritchiei C. B. Clarke—India: Khaudala; Santapau 13946 (MO); EF595902, EF595952.
Mukia Arn.—M. maderaspatana (L.) M. Roem.—Ethiopia: Ilubabor region, Gambella; Hedrén 591 (UPS); EF595904, EF595954. M. javanica (Miq.) Jeffrey—China: S. Yunnan province; Tsi Zhanhuo 91–399 (MO); EF595903, EF595953.
Myrmecosicyos C. Jeffrey—M. messorius C. Jeffrey—Kenya: near Maralal; P. A. & W. R. Q. Luke 10782 (EA); EF595905, EF595955.
Oreosyce Hook. f.—O. africana Hook. f.—Kenya: Marsabit; Jonsell & Moberg 4535b (UPS); EF595906, EF595956. O. africana Hook. f.—Tanzania: near Rungwe Mt; Mwasumbi 16296 (UPS); EF595907, EF595957.
FOOTNOTES
1 The authors thank the Missouri Botanical Garden; the USDA/ARS North Central Regional Plant Introduction Station in Ames, Iowa; the University of Asmara, Eritrea; the Namibian National Plant Genetic Resource Center; the East African Herbarium in Nairobi, Kenya; and R. Perl-Treves, W. de Wilde, B. Duyfjes, and J. Garcia-Mas for providing leaf material and seeds. They are grateful to J. Kirkbride, S. Graham, J. Ray, J. Allen, A. Doust, and three anonymous reviewers for their constructive comments on the manuscript. Funding was provided by an American Society of Plant Taxonomists Graduate Student Research grant to AGG, a Beaumont Faculty Development grant to JCB, and financial support by the Swedish Research Council to MT. 
4 Author for correspondence (ghebreag{at}slu.edu
)
LITERATURE CITED
Alvarez I. Wendel J. F.. 2003. Ribosomal ITS sequences and plant phylogenetic inference. Molecular Phylogenetics and Evolution 29: 417-434.[CrossRef][ISI][Medline]
Ayyngar K. R.. 1967. Taxonomy of Cucurbitaceae. Bulletin of National Institute of Sciences of India 34: 380-396.
Baldwin B. G.. 1997. Adaptive radiation of the Hawaiian silversword alliance: congruence and conflict of phylogenetic evidence from molecular and non-molecular investigations. In T. G. Givnish and K. J. Sytsma [eds.], Molecular evolution and adaptive radiation 103-128 Cambridge University Press, Cambridge, UK.
Barber J. C. Finch C. C. Francisco-Ortega J. Santos-Guerra A. Jansen R. K.. 2007. Hybridization in Macaronesian Sideritis (Lamiaceae): evidence from incongruence of multiple independent nuclear and chloroplast sequence data sets. Taxon 56: 74-88.[ISI]
Barber J. C. Francisco-Ortega J. Santos-Guerra A. Marrero A. Jansen R. K.. 2000. Evolution of endemic Sideritis (Lamiaceae) in Macronesia: insights from a chloroplast DNA restriction site analysis. Systematic Botany 25: 633-647.[CrossRef][ISI]
Berry P. E. Hahn W. J. Sytsma K. J. Hall J. C. Mast A.. 2004. Phylogenetic relationships and biogeography of Fuchsia (Onagraceae) based on noncoding nuclear and chloroplast DNA data. American Journal of Botany 91: 601-614.
Chen J.-F. Kirkbride J. H.. 2001. A new synthetic species of Cucumis (Cucurbitaceae) from interspecific hybridization and chromosome doubling. Brittonia 52: 315-319.[CrossRef][ISI]
Chen J.-F. Staub J. E. Tashiro Y. Isshiki S. Miyazaki S.. 1997. Successful interspecific hybridization between Cucumis sativus L. and C. hystrix Chakr. Euphytica 96: 413-419.[CrossRef][ISI]
Chung S.-F. Staub J. E. Chen J.-F.. 2006. Molecular phylogeny of Cucumis species as revealed by consensus chloroplast SSR marker length and sequence variation. Genome 49: 219-229.[Medline]
Cogniaux A.. 1881. Cucurbitaceae. In A. L. P. P. de Candolle and A. C. P. de Candolle [eds.], Monographiae phanerogamarum vol 3: 325-951953-954 G. Mason, Paris, France.
Cunningham C. W.. 1997. Can three incongruence tests predict when data should be combined?. Molecular Biology and Evolution 14: 733-740.[Abstract]
Dane F.. 1976. Evolutionary studies in the genus Cucumis. Ph.D. dissertation Colorado State University, Fort Collins, Colorado, USA.
Dane F.. 1983. Cucurbits. In S. D. Tanksley and T. J. Orton [eds.], Isozymes in plant genetics and breeding, part B 369-390 Elsevier, Amsterdam, Netherlands.
Dane F. Denna D. W. Tsuchiya T.. 1980. Evolutionary studies of species of the genus Cucumis. Zeitschrift für Pflanzenzuchtung 85: 89-109.
De Wilde W. J. J. O. Duyfjes B. E. E.. 2006. Mukia Arn. (Cucurbitaceae) in Asia, in particular in Thailand. Thai Forest Bulletin (Botany) 34: 38-52.
Downie S. R. Katz-Downie D. S.. 1996. A molecular phylogeny of Apiaceae subfamily Apioideae: evidence from nuclear ribosomal DNA internal transcribed spacer sequences. American Journal of Botany 83: 234-251.[CrossRef][ISI]
Doyle J. J. Doyle J. L.. 1987. A rapid DNA isolation procedure for small quantities of fresh leaf tissue. Phytochemical Bulletin 19: 11-15.
Enslin P. R. Rehm S.. 1957. The distribution and biogenesis of the cucurbitacins in relation to the taxonomy of the Cucurbitaceae. Proceedings of the Linnean Society of London 168: 230-238.
Esquinas-Alcazar J. T.. 1977. Alloenzyme variation and relationships in the genus Cucumis. Ph.D. dissertation University of California, Davis, California, USA.
Farris J. S. Källersjö M. Kluge A. G. Bult C.. 1994. Testing significance of incongruence. Cladistics 10: 315-319.[CrossRef][ISI]
Farris J. S. Källersjö M. Kluge A. G. Bult C.. 1995. Constructing a significance test for incongruence. Systematic Biology 44: 570-572.[CrossRef][ISI]
Felsenstein J.. 1985. Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39: 783-791.[CrossRef][ISI]
Fernandes R. Fernandes A.. 1970. 88–Cucurbitaceae. In A. W. Exell, A. Fernandes, and E. J. Mendes [eds.], Conspectus florae angolensis, vol. 4 232-289 Junta de Investigações do Ultramar e Instituto de Investigações Científica de Angola, Lison, Angola.
Garcia-Mas J. Monforte A. J. Arus P.. 2004. Phylogenetic relationships among Cucumis species based on ribosomal internal transcribed spacer region. Plant Systematics and Evolution 248: 191-203.[ISI]
Garcia-Mas J. Oliver M. Gomez H. De Vicente M. C.. 2000. AFLP, RAPD and RFLP markers to measure genetic diversity in melon. Theoretical and Applied Genetics 101: 860-864.[CrossRef][ISI]
Gaskin J. F. Schaal B. A.. 2003. Molecular phylogenetic investigation of U.S. invasive Tamarix. Systematic Botany 28: 86-95.[ISI]
Goldman N. Anderson J. P. Rodrigo A. G.. 2000. Likelihood-based tests of topologies in phylogenies. Systematic Biology 49: 652-670.[CrossRef][ISI][Medline]
Hamilton M. B.. 1999. Four primer pairs for the amplification of chloroplast intergenic regions with intraspecific variation. Molecular Ecology 8: 521-523.[Medline]
