Evidence of Natural Hybridization and Introgression between Vasconcellea Species (Caricaceae) from Southern Ecuador Revealed by Chloroplast, Mitochondrial and Nuclear DNA Markers

Abstract

• Background and Aims Vasconcellea × heilbornii is believed to be of natural hybrid origin between V. cundinamarcensis and V. stipulata, and is often difficult to discriminate from V. stipulata on morphological grounds. The aim of this paper is to examine individuals of these three taxa and of individuals from the closely related species V. parviflora and V. weberbaueri, which all inhabit a hybrid zone in southern Ecuador.

• Methods Molecular data from mitochondrial, chloroplast and nuclear DNA from 61 individuals were analysed.

• Key Results Molecular analysis confirmed occasional contemporary hybridization between V. stipulata, V. cundinamarcensis and V. × heilbornii and suggested the possible involvement of V. weberbaueri in the origin of V. × heilbornii. In addition, the molecular data indicated unidirectional introgression of the V. cundinamarcensis nuclear genome into that of V. stipulata. Several of the individuals examined with morphology similar to that of V. stipulata had genetic traces of hybridization with V. cundinamarcensis, which only seems to act as pollen donor in interspecific hybridization events. Molecular analyses also strongly suggested that most of the V. × heilbornii individuals are not F₁ hybrids but instead are progeny of repeated backcrosses with V. stipulata.

• Conclusions The results of the present study point to the need for re-evaluation of natural populations of V. stipulata and V. × heilbornii. In general, this analysis demonstrates the complex patterns of genetic and morphological diversity found in natural plant hybrid zones.

INTRODUCTION

Natural hybridization occurs widely in plants and has an important role in their evolution (Arnold, 1997; Rieseberg and Carney, 1998). Plant hybridization zones have recently been studied because they are perceived to be the dynamic centres of ecological and evolutionary processes for plants and their associated communities (Whitham et al., 1999). Introgressive hybridization is particularly interesting for plant evolutionary studies because it produces considerable numbers of new genotypes, thereby increasing genetic diversity, which may lead to the establishment of novel strains, ecotypes or even sexual species that are adapted to particular environments (Arnold and Hodges, 1995; Arnold, 1997; Seehausen, 2004). Other genetic and ecological consequences that have been proposed for introgression of genetic material between otherwise divergent plant populations are the breakdown or reinforcement of isolating barriers and the promotion of dispersion and colonization (Anderson, 1948; Potts and Reid, 1988; Rieseberg and Wendel, 1993). However, hybridization and introgression between plant populations can have an opposite effect: Levin et al. (1996) showed how hybridization might contribute to the extinction of rare plant species through demographic swamping and genetic assimilation by an abundant congener.

This study presents the results of analyses of molecular data from taxa of the genus Vasconcellea (Caricaceae), i.e. V. cundinamarcensis Badillo and V. stipulata (Badillo) Badillo and their putative natural hybrid, i.e. V. × heilbornii (Badillo) Badillo, and the closely related species V. weberbaueri (Harms) Badillo and V. parviflora De Candolle.

Vasconcellea cundinamarcensis is widespread in temperate zones of the Andes (1500–3000 m) ranging from Colombia and Venezuela to Bolivia. It is cultivated on a small scale in Chile and introduced to similar climates where it has become naturalized (Badillo, 1971, 1993). In contrast to the large geographical range colonized by V. cundinamarcensis, V. stipulata is only known from the central-southern part of Ecuador (Loja and Azuay Provinces) and the northern part of Peru where it grows at an altitude between 1600 and 2450 m (Badillo, 1971, 1993). Consequently this area of distribution overlaps with a part of the distribution area of V. cundinamarcensis. Natural interspecific hybridization between V. cundinamarcensis and V. stipulata has been reported to take place in this area where the species are sympatric (Badillo, 1967; Horovitz and Jiménez, 1967). Furthermore, at the morphological level, Horovitz and Jiménez (1967) observed signs of introgression of V. cundinamarcensis into V. stipulata in Loja Province, Ecuador. Besides natural hybridization, Horovitz and Jiménez (1967) and De Zerpa (1980) also investigated artificial hybridization between V. cundinamarcensis and V. stipulata and the reciprocal cross. Both crosses resulted in viable F₁ plants and these plants were succesfully backcrossed using V. stipulata as the recurrent parent (De Zerpa, 1980). Backcrossing with V. cundinamarcensis was not attempted. Several other Vasconcellea species have also been demonstrated to be intercompatible (Jiménez and Horovitz, 1957; Horovitz and Jiménez, 1967; Mekako and Nakasone, 1975; De Zerpa, 1980). Based on the attempted artificial hybridizations six out of eight taxa appeared to be intercompatible (V. microcarpa, V. cundinamarcensis, V. stipulata, V. cauliflora, V. monoica and V. horovitziana), although reciprocal crosses were not always possible. Vasconcellea goudotiana and V. parviflora are intercompatible but can only be crossed with a few of the above-mentioned species. As far as is known, V. weberbaueri has never been tested for possible intercompatibility with other Vasconcellea spp.

Although V. cundinamarcensis and V. stipulata are relatively similar in their vegetative appearance, Badillo (1993) noted clear differences that should enable discrimination between both species. Usually V. cundinamarcensis is recognized by the pubescence on flowers and leaves and its greenish flowers, whereas V. stipulata is characterized by glabrous mature leaves, hard and pointed stipular spines on each side of the petiole and its orange flowers (Badillo, 1971, 1993, 1997). In contrast to the strictly dioecious V. stipulata (as most other species in this genus), monoecious forms of V. cundinamarcensis have been observed (Badillo, 1993).

Badillo (1967, 2000) first named the supposed hybrid between V. stipulata and V. cundinamarcensis as Carica × heilbornii Badillo, later on as V. × heilbornii (Badillo) Badillo and described it as a dioecious, monoecious or polygamous plant generally showing some degree of parthenocarpy (up to complete absence of seeds), somewhat pubescent to completely glabrous and mostly with stipular spines. The flowers are orange, green or greenish-yellow and calyx lobes alternate with corolla lobes. Probably because some selected plants of V. × heilbornii are vegetatively propagated by local farmers for fruit production, the male flowers are sometimes unknown. Different varieties of V. × heilbornii have been described. Actually two varieties (var. chrysopetala and var. fructifragrans) and one cultivar ('Babaco') are withheld in the latest Caricaceae revision (Badillo, 2000). However, besides the varieties described and the cultivar, Badillo (1993) states that several other undescribed varieties of V. × heilbornii can be found in Ecuador. This was recently confirmed by an ethnobotanical study of Vasconcellea spp. from southern Ecuador (Scheldeman, 2002). The author observed an extensive variability of morphological characters among undescribed V. × heilbornii genotypes and reported difficulties in distinguishing V. × heilbornii from V. stipulata (Scheldeman, 2002). Such morphological variations among the hybrids suggest repeated interspecies hybridization events.

Up to now, isozymes and RAPD (Jobin-Décor et al., 1997), AFLP (Van Droogenbroeck et al., 2002; Kyndt et al., 2005a) and chloroplast markers (Aradhya et al., 1999) have only been used in biodiversity and phylogenetic studies of Caricaceae, but not for hybrid identification within the genus Vasconcellea. Van Droogenbroeck et al. (2002) were the first to examine a diverse set of V. × heilbornii genotypes using nuclear AFLP markers. The specific clustering of these genotypes together with either one of the putative parent species, provided support for the involvement of V. cundinamarcensis and V. stipulata in the origin of V. × heilbornii and suggested ongoing bi-directional introgression events between the parent species (Van Droogenbroeck et al., 2002). AFLP data of Kyndt et al. (2005a) emphasized the close genetic relationship between V. stipulata and V. × heilbornii on the nuclear level. A close association between V. stipulata, V. × heilbornii, and two other species, V. parviflora and V. weberbaueri, growing in the central-southern part of Ecuador, has been observed throughout previous molecular studies (Van Droogenbroeck et al., 2004; Kyndt et al., 2005a, b). These species are morphologically easy to distinguish from other taxa: V. parviflora has pinkish flowers and small orange fruit with ten grooves, whereas V. weberbaueri is the only species of Vasconcellea with serrate leaves. A PCR-RFLP biodiversity study (Van Droogenbroeck et al., 2004) surprisingly revealed a close relationship between V. × heilbornii and V. weberbaueri, as they shared the same chloroplast and mitochondrial haplotype. Kyndt et al. (2005a) also found the typical haplotype of V. stipulata in two V. × heilbornii individuals.

Results of sequence analysis of the nuclear ITS region (Kyndt et al., 2005b) confirmed a close genetic relationship of V. stipulata with V. × heilbornii, but also with V. weberbaueri and V. parviflora. Vasconcellea cundinamarcensis was only distantly related to V. × heilbornii. Moreover, significant ITS sequence heterogeneity was found within the genome of the analysed specimens of V. stipulata, V. cundinamarcensis and V. × heilbornii by cloning the different ITS variants within each individual plant. Because of the fact that so-called ‘concerted evolution’ homogenizes variation across the multiple nuclear rDNA gene copies (Baldwin et al., 1995), the observed heterogeneity strongly suggests a hybrid origin for the specimens of these taxa investigated. Sequencing of the chloroplast DNA regions matK and trnL-trnF (Kyndt et al., 2005b) confirmed a very close phylogenetic relationship between V. × heilbornii and V. weberbaueri, showing a very low amount of divergence. Incongruence between nuclear ITS and chloroplast data sets, which can be attributed to hybridization or introgression, was shown to be caused by several taxa of Vasconcellea, including V. stipulata, V. cundinamarcensis and V. × heilbornii (Kyndt et al., 2005b). However, a combination of additional data derived from both nuclear and cytoplasmic markers from a larger set of individuals was needed to increase the possibility of an accurate detection of hybridization and introgression phenomena between these taxa.

The main objectives of the present study were (a) to investigate a more extensive set of V. × heilbornii individuals from a described Vasconcellea hybrid zone, i.e. Loja and Azuay Provinces in southern Ecuador, with respect to their hybrid genealogies and (b) to demonstrate the potential of hybridization and introgression between Vasconcellea spp. within the zone. Therefore, both nuclear and cytoplasmic molecular markers were used to analyse individuals of V. cundinamarcensis, V. stipulata, V. × heilbornii and the related species V. weberbaueri and V. parviflora. The AFLP technique was chosen to analyse the nuclear genomes because of its high multiplex ratio and reproducibility. In addition, PCR-RFLP markers of chloroplast DNA (cpDNA) and mitochondrial DNA (mtDNA) were analysed in this study. Maternal inheritance of these markers can be assumed based on intergeneric crosses between Carica papaya and Vasconcellea spp. (Van Droogenbroeck et al., 2005).

MATERIALS AND METHODS

Study site, plant collections and DNA extraction

Most of the sampling sites were located in southern Ecuador (Loja and Azuay Provinces), which is recognized by several authors as a natural hybrid zone for V. cundinamarcensis and V. stipulata (Badillo, 1967; Horovitz and Jiménez, 1967). However, a few V. × heilbornii specimens were collected in the South of Napo and Pichincha Provinces. Individuals were identified based on the most recent key (Badillo, 1993) and named according to Badillo (2001). A unique chronological number was assigned to all specimens collected (Table 1).

In total, 61 individuals including V. stipulata (eight individuals), V. cundinamarcensis (eight individuals), V. × heilbornii (35 individuals), and the related species V. weberbaueri (five individuals) and V. parviflora (four individuals) were investigated (Table 1). Within the taxon V. × heilbornii,Badillo (1993) described two varieties, V. × heilbornii var. chrysopetala (four individuals) and V. × heilbornii var. fructifragrans (three individuals), and one cultivar V. × heilbornii ‘Babaco’ (three individuals). In addition, 25 different genotypes of this taxon could not be identified as one of the described varieties or cultivar, and were therefore further referred to as ‘undescribed variety’ (‘var?’). Finally, one genotype resulting from an artificial hybridization between V. × heilbornii var. chrysopetala and V. cundinamarcensis was added to the sample set. Unfortunately, plant material from the parental genotypes of this artificial hybrid was not available. Among the specimens of V. stipulata collected, four contained all diagnostic morphological features as described by Badillo (1993), while four others revealed some slightly aberrant morphology in one of these characteristics. Therefore the latter specimens were referred to as ‘V. stipulata?’.

For DNA analysis, young leaves were collected in the field and dried with silica gel. DNA extraction was performed on dried leaf tissue ground in liquid nitrogen. Total genomic DNA was extracted using the Qiagen Dneasy Plant Mini-kit (Qiagen GmbH, Hilden, Germany).

AFLP analysis

AFLP analysis was performed as reported by Van Droogenbroeck et al. (2002). In a preliminary AFLP assay, multiple EcoRI+2/MseI+4 primer combinations were pre-screened. Based on the number of fragments amplified and the polymorphism rate observed, nine primer pairs were selected. From earlier studies on Vasconcellea spp. (Van Droogenbroeck et al., 2002, 2004; Kyndt et al., 2005a, b) it could be concluded that besides V. stipulata and V. cundinamarcensis only V. weberbaueri could possibly have acted as progenitor of V. × heilbornii. Therefore, special attention was given to the detection of species-specific AFLP fragments of V. cundinamarcensis, V. stipulata and V. weberbaueri in the V. × heilbornii individuals. Species-specific markers were identified for each putative parent species based on the results of AFLP studies already published (Van Droogenbroeck et al., 2002; Kyndt et al., 2005a). Markers were considered as species-specific when present in at least 95 % of the individuals of one putative parent species, and absent in at least 95 % of the other putative parent species. Vasconcellea stipulata? specimens were excluded from this identification.

In a second step, the same nine EcoRI+2/MseI+4 primer combinations were used to fingerprint the complete sample set. For each individual, the DNA fingerprints were scored by visual inspection for presence (1) or absence (0) of specific AFLP bands. Only distinct, major bands were scored. Cluster analysis was performed using PAUP^* 4·0b10 (Swofford, 2002). Genetic similarities between all pairs of individuals were calculated using the Nei and Li coefficient (Nei and Li, 1979). Genetic similarity/diversity estimates between or within groups of individuals were obtained by calculating the average of all the values representing the genetic similarity or diversity within or between the corresponding group(s). The dissimilarity matrix was analysed by UPGMA cluster analysis. Reliability of the dendrogram was tested by bootstrap analysis with 1000 replications. Additionally, a principal co-ordinate analysis (PCO) based on the genetic similarity matrix was performed using the DCENTER and EIGEN algorithms of the NTSYS-pc software package Version 2·10L (Rohlf, 2000).

CpDNA and mtDNA haplotype determination using PCR-RFLP

PCR-RFLP analysis was performed as reported in Van Droogenbroeck et al. (2004). The eight PCR-fragment/enzyme combinations selected in that study were used to fingerprint the complete set of individuals: K1K2/EcoRV, K1K2/ScaI, K1K2/AfaI; ML/PstI, ML/MseI for the cpDNA regions and nad4/1–2/HinfI, nad4/1–2/BstOI, nad4/1–2/DdeI for the mtDNA region. Haplotypes were defined as a set of specific combinations of the variants observed for all mutations detected. Species-specificity of the haplotypes was checked and confirmed against earlier PCR-RFLP fingerprinting results (Van Droogenbroeck et al., 2004; Kyndt et al., 2005a).

RESULTS

AFLP analysis

AFLP analysis of the complete dataset (61 individuals) with the nine selected primer combinations yielded 234 fragments that could be scored easily. Of these, 208 were polymorphic (89 %) and 26 monomorphic (11 %). Using these nine primer pairs, it was possible to detect 24 fragments that were specific for V. stipulata, 47 for V. cundinamarcensis and 21 for V. weberbaueri (Table 2). In a first step, these diagnostic markers were scored for all V. × heilbornii individuals and the V. stipulata and V. cundinamarcensis individuals. The results are shown in Table 3. The four V. stipulata? specimens missed some of the V. stipulata-specific markers, and in one of these individuals (stip?312), a single V. cundinamarcensis-specific marker was detected. Among the individuals identified as a described variety or cultivar of V. × heilbornii, no markers specific for V. cundinamarcensis could be detected. The average number of V. stipulata-specific fragments detected in the babaco-individuals was much lower (8–12; Table 3) in comparison with the individuals belonging to the varieties chrysopetala and fructifragrans (15–21; Table 3). On the other hand, 12 of the 26 individuals of V. × heilbornii that could not be identified as one of the described varieties or cultivar, contained at least one or more V. cundinamarcensis markers in combination with different numbers of V. stipulata markers. Four out of these 12 had all 47 V. cundinamarcensis markers, while a fifth individual (var?317) missed only one V. cundinamarcensis marker. Among these five, two individuals (var?247 and var?308) also displayed nearly all the V. stipulata markers and therefore may represent F₁ hybrids between V. stipulata and V. cundinamarcensis. Finally, a very low number of V. weberbaueri-specific fragments (one or two recurrent markers; Table 3) could be detected in some of the V. × heilbornii individuals.

Secondly, the group structure and molecular diversity within the complete set of individuals was studied by UPGMA and PCO analysis using the AFLP data obtained from scoring all markers. Figure 1 shows the result of an UPGMA analysis based on the Nei and Li coefficient of genetic similarity observed between all pairs of individuals. Vasconcellea stipulata and V. × heilbornii did not group into discrete, species-specific subclusters. Within the first subgroup of cluster 1, the individuals identified as a described variety (chrysopetala or fructifragrans) or as the cultivar ‘Babaco’ were grouped together in discrete subclusters, both moderately supported by bootstrap analysis (Fig. 1, cluster 1A). This group further holds some individuals that could not be identified as one of the described forms of V. × heilbornii together with four individuals identified as V. stipulata? (Fig. 1, cluster 1A). Next to cluster 1A, subgroup 1B consists of a mixture of V. stipulata and V. × heilbornii individuals (Fig. 1, cluster 1B). A single V. × heilbornii individual (var?202) clustered separately from these two subgroups and completed cluster 1. The remaining V. × heilbornii individuals, all unidentified specimens of V. × heilbornii, clustered together with the artificial hybrid in a well-supported subgroup (cluster 4B; bootstrap value 100 %), linked to the species-specific V. cundinamarcensis cluster (cluster 4A). The related species V. parviflora and V. weberbaueri were grouped in two highly supported clusters (clusters 2 and 3). These two clusters appeared to share more common AFLP markers with the group including V. stipulata (cluster 1) than with the group containing V. cundinamarcensis (cluster 4).

As mentioned above, within cluster 1A, the varieties fructifragrans and chrysopetala formed a moderately supported subgroup (bootstrap support of 68 %). The close genetic relationship between these two varieties was also confirmed by the observation of one AFLP marker that was present in these individuals, but absent in all others. Similarly, one specific AFLP marker was detected among the babaco-individuals (bootstrap support of 64 %) (data not shown). Apart from V. × heilbornii, all species displayed low intraspecific diversity, ranging from 0·07 for V. stipulata to 0·01 for V. weberbaueri. Low levels of intraspecific diversity were observed among the babaco-individuals (0·04) and the varieties chrysopetala (0·01) and fructifragrans (0·01), in contrast to the higher genetic diversity observed among the unidentified individuals of V. × heilbornii (0·21). This results in a higher overall intraspecific diversity of the species V. × heilbornii (0·19).

Figure 2 shows the results of a PCO analysis based on the pairwise genetic similarity between all individuals calculated using the Nei and Li coefficient. The first two principal co-ordinates accounted for 61·4 % of the variation and PCO separations support the results obtained with the cluster analysis. The four major groups identified with the cluster analysis (Fig. 1) could easily be recognized in the PCO analysis (Fig. 2). The first principal co-ordinate (45·7 %) separated the V. cundinamarcensis individuals from all the others, while the second principal co-ordinate (15·7 %) differentiated the species V. stipulata, V. parviflora and V. weberbaueri. Five V. × heilbornii hybrid individuals which could neither be identified as one of the described varieties nor as the cultivar, together with the artificial hybrid were placed between V. cundinamarcensis and V. stipulata. This positioning was likely due to the additivity of species-specific AFLP markers (Fig. 2 and Table 3). All the other V. × heilbornii individuals, including the varieties described and the cultivar, were positioned closer to V. stipulata.

Chloroplast and mitochondrial haplotypes

The cpDNA regions K1K2 and ML of the 61 individuals listed in Table 1 were amplified and restricted with, respectively, three (EcoRV, ScaI and AfaI) and two (MseI and PstI) restriction enzymes. For K1K2 this resulted in the detection of five polymorphisms among the individuals investigated. Two point mutations were specific for V. cundinamarcensis. The analysis of cpDNA region ML produced similar results. Among the four point mutations detected within this region, two were again specific for V. cundinamarcensis. In none of the cpDNA regions was an insertion/deletion found. Combining the nine point mutations detected in the cpDNA regions, K1K2 and ML allowed for the identification of four chloroplast haplotypes (Table 3). Two haplotypes were specific for a single species, i.e. that of V. cundinamarcensis (C haplotype; Table 3) and V. parviflora (P haplotype; Table 3). The third haplotype was shared by the four V. stipulata individuals and ten of the unidentified V. × heilbornii individuals. For reasons discussed below this haplotype henceforth will be considered as the V. stipulata haplotype (S haplotype; Table 3). Surprisingly, all other V. × heilbornii individuals, including the individuals belonging to the varieties described and the cultivar, together with the artificial hybrid and four individuals identified as V. stipulata?, shared all mutations detected in the PCR-RFLP analysis with the V. weberbaueri individuals. This fourth haplotype is therefore designated as the V. × heilbornii–V. weberbaueri haplotype (H-W haplotype; Table 3). An example of the differences between these haplotypes, as revealed by restricting K1K2 with AfaI, is shown in Fig. 3A.

PCR-RFLP analysis of the mtDNA region nad4/1–2 of all individuals with three restriction enzymes (HinfI, BstOI and DdeI) permitted the detection of two point mutations that were present only in the V. cundinamarcensis individuals, while a third point mutation was present in all other individuals but absent in the V. cundinamarcensis individuals. No insertion/deletions were found. Two mitochondrial haplotypes could thus be determined easily: (1) a haplotype specific for V. cundinamarcensis and (2) a haplotype common to all other individuals (Table 3). An example of the difference between these haplotypes, as revealed by restricting nad4/1–2 with HinfI, is shown in Fig. 3B.

All cpDNA and mtDNA haplotypes defined in this study were in agreement and confirmed the results obtained in previous PCR-RFLP biodiversity studies with smaller sample sizes (Van Droogenbroeck et al., 2004; Kyndt et al., 2005a).

Finally, for both cpDNA regions and the mtDNA region investigated, the artificial hybrid displayed restriction profiles identical to that specific for its maternal progenitor, i.e. V. × heilbornii var. chrysopetala, thereby suggesting that the chloroplast and mitochondrial genomes are maternally inherited in Vasconcellea, as is the case in most other angiosperms (Mogensen, 1996) and as already demonstrated for intergeneric hybrids between Carica papaya and Vasconcellea spp. (Van Droogenbroeck et al., 2005).

Combination of molecular data

Based on the data obtained from both AFLP and PCR-RFLP analyses, individuals could be grouped into seven classes. Vasconcellea cundinamarcensis, V. weberbaueri and V. parviflora could unambiguously be identified by the molecular analyses and are placed in classes 5, 6 and 7, respectively. However, the remaining individuals identified with the key of Badillo as either V. stipulata or V. × heilbornii, were distributed intermingled among classes 1–4 (Table 3). In general, this classification based on the combined molecular evidence corresponds well with the AFLP-based dendrogram and PCO analysis (Figs 1 and 2).

All individuals that shared the S haplotype carried at least 22 of the 24 AFLP markers specific for V. stipulata (Table 3; classes 1 and 2). Therefore this haplotype was designated as the V. stipulata haplotype (S haplotype; Table 3) and these individuals are likely to represent the most ‘pure’ V. stipulata individuals. The only exception to this rule was individual var?247, displaying the S haplotype but having only 19 of the 24 V. stipulata-specific AFLP markers. In contrast to the individuals from class 1 lacking V. cundinamarcensis-specific AFLP markers, a few (class 2a) or all (class 2b) V. cundinamarcensis-specific AFLP markers were found in three unidentified V. × heilbornii individuals with the S haplotype.

The H-W haplotype was present in all individuals belonging to classes 3, 4 and 6 (Table 3). Again the presence of one or more V. cundinamarcensis-specific AFLP markers was used to distinguish between classes 3 and 4. The majority of V. × heilbornii individuals belonged to class 3, including the cultivar ‘Babaco’ displaying a lower number of V. stipulata markers (class 3a) in comparison to the varieties (chrysopetala and fructifragrans) described, six unidentified V. × heilbornii individuals and three individuals identified as V. stipulata? (class 3b). As explained above, all individuals within this class combined the H-W haplotype with a variable number of V. stipulata-specific markers but lacked V. cundinamarcensis-specific AFLP markers. Morphologically, these specimens resembled V. stipulata in having orange flowers and spiny stipules (Romeijn-Peeters, 2004).

Finally, class 4 consisted of nine undefined V. × heilbornii individuals, one individual identified as V. stipulata and the artificial hybrid. In addition to the genetic characteristics of the individuals of class 3, class 4 individuals displayed a few (class 4a) or (nearly) all (class 4b) V. cundinamarcensis-specific AFLP markers. This genetic relationship with V. cundinamarcensis was also reflected in some morphological features as these specimens were observed to reveal pubescence on some parts of the plants, no stipules and yellow or greenish yellow flowers (Romeijn-Peeters, 2004).

DISCUSSION

Vasconcellea stipulata or V. × heilbornii?

The molecular data are congruent with the prior morphological classification based on the most recent key of Badillo (1993) for V. cundinamarcensis and the related species V. weberbaueri and V. parviflora (Table 3). In contrast, molecular data revealed that the morphological description of Badillo (1993) was in some cases too vague to identify V. stipulata or V × heilbornii and its cultivar/varieties accurately. As addressed in the Results, the S haplotype that was recorded very likely represents the true cpDNA haplotype of V. stipulata. Given the fact that this haplotype was only found in individuals carrying at least 22 of the V. stipulata-specific AFLP fragments and the conservative nature of cpDNA (Palmer, 1987), the annotation of the S haplotype is justified. A possible explanation for the observation that these ‘pure’ V. stipulata individuals did not carry all AFLP fragments specific for this species is that these perhaps were not really ‘species specific’. After all, if the species displays high levels of genetic diversity, it is possible that the set of V. stipulata plants analysed might not have been completely representative of the levels of genetic diversity present in this species. In the four V. stipulata? individuals that were morphologically identified as V. stipulata using the key of Badillo (1993), but for which some aberrant morphology was observed, the present study revealed molecular markers that distinguished them from the ‘pure’ V. stipulata plants. In addition, seven individuals that could not be morphologically identified as one of the varieties or cultivar of V. × heilbornii, can be considered as V. stipulata based on the results of the molecular analysis (Table 3). This ambiguity is reflected in the morphology, as V. stipulata is not as clearly morphologically defined as it was proposed by Badillo (1993). Vasconcellea stipulata is restricted by Badillo to wild plants with firm spiny stipules and orange flowers, definitively excluding any plant of garden origin. Due to the absence of flowers in many specimens, the development of the spiny stipules was the only omnipresent feature proposed by Badillo to distinguish V. stipulata from other taxa. However, the intermingling of V. stipulata with V. × heilbornii in the dendrogram derived from the AFLP analysis, together with the observation of spiny stipules on specimens initially identified as either V. stipulata or V. × heilbornii, clearly indicate that the development of the spiny stipules might not be as reliable as currently assumed: large spiny stipules are no guarantee for a correct identification of V. stipulata specimens. Nevertheless, these results indicate the necessity of adding new vegetative features to make identification of V. stipulata more reliable.

Repeated hybridization between V. stipulata, V. cundinamarcensis and V. × heilbornii

Intra-individual ITS sequence heterogeneity reported by Kyndt et al. (2005b) already indicated a hybrid or introgressed origin for some individuals of V. stipulata, V. cundinamarcensis and V. × heilbornii. In the present study, markers specific to V. stipulata and V. cundinamarcensis were identified in the nuclear genome using the AFLP technique (Table 2). In addition to the diagnostic AFLP markers, species-specific chloroplast haplotypes were determined for V. stipulata and V. cundinamarcensis using PCR-RFLP. Similar analyses of the mtDNA region nad4/1-nad4/2 could only distinguish V. cundinamarcensis from the four other species, but this was expected since mtDNA has a lower mutation rate than cpDNA in plants (Wolfe et al., 1987).

According to Badillo, many Vasconcellea spp. are intercompatible and can produce hybrids with various degrees of fertility. The occasional spontaneous occurrence of hybrids in sympatric regions (Badillo, 1971) was already illustrated by molecular and morphological data by Kyndt et al. (2005a). Based on morphological observations, V. × heilbornii is considered as a natural hybrid derived from crosses between V. cundinamarcensis and V. stipulata (Badillo, 1967; Horovitz and Jimenez, 1967). However, in this study only 12 out of the 35 individuals identified as V. × heilbornii displayed a combination of V. stipulata and V. cundinamarcensis AFLP markers. Of these 12 individuals, only three possessed the cpDNA haplotype of V. stipulata, while the other nine individuals surprisingly shared their cpDNA haplotype with V. weberbaueri. So, none of the individuals displaying a combination of V. stipulata and V. cundinamarcensis AFLP markers possessed the haplotype of V. cundinamarcensis. Consequently these results suggest that V. cundinamarcensis appears to act only as pollen donor in interspecific hybridizations and that V. stipulata occasionally can serve as the ovule/cytoplasm donor in interspecific hybridization events.

Based on the criterion of additivity of species-specific AFLP markers, var?247 is the only individual possessing the S haplotype that comes close to the genetic constitution of a true V. stipulata × V. cundinamarcensis F₁ hybrid. After all, a true F₁ should possess all 71 markers that are constant in one parent and absent in the other (24 from V. stipulata and 47 from V. cundinamarcensis). But only 66 constant markers were detected, suggesting that this individual may be a later-generation hybrid. However, Caraway et al. (2001) studied an F₁ artificial hybrid with the RAPD technique and concluded that the assumption of parental markers being fixed if polymorphism was not detected was violated; only 38 of the 40 diagnostic markers were present in the artificial hybrid. This taken into account, it remains possible that var?247 indeed is a F₁ hybrid. It is not surprising that none or possibly only a single F₁ is found within this hybrid zone. Rieseberg and Ellstrand (1993) emphasized in their review on interspecific hybridization in plants that hybrid zones are seldom limited to F₁s. Furthermore, substantial evidence that chromosomal and genetic sterility barriers are responsible for strong selection against F₁ hybrids has been presented (reviewed by Rieseberg and Carney, 1998).

The two other individuals of class 2, clustering with V. stipulata in the UPGMA and PCO analyses and also possessing the S haplotype, combined all or most of the V. stipulata-specific markers with a much lower number of V. cundinamarcensis-specific markers (Table 3). The observed combination of molecular markers in these individuals implies that they are products of backcrossing with V. stipulata beyond the BC₃. An alternative explanation for the low number of V. cundinamarcensis-specific AFLP markers present in these plants is that they are the products of an F₂ or later generation hybrid, with an already lower number of V. cundinamarcensis-specific markers, backcrossed to V. stipulata. Since no V. stipulata-specific AFLP markers could be detected in any of the V. cundinamarcensis individuals investigated, these results suggest the occurrence of unilateral introgressive hybridization at the nuclear level in the direction of V. stipulata, which as a whole demonstrates greater amounts of variation (an assumed consequence of introgression) in the characters examined. This conclusion does not agree with the results from a previous AFLP analysis (Van Droogenbroeck et al., 2002), where bi-directional introgression between V. stipulata and V. cundinamarcensis was suggested. The latter conclusion was based solely on the UPGMA phenogram obtained. In the present study, a comprehensive analysis of diagnostic AFLP markers was combined with cpDNA and mtDNA PCR-RFLP fingerprinting. Therefore the conclusions derived here demonstrate a higher degree of confidence. In addition, the molecular results obtained in this study agree with the morphological observations of Horovitz and Jiménez (1967) who reported signs of introgression of V. cundinamarcensis into V. stipulata in Loja province.

All other individuals displaying a combination of V. stipulata and V. cundinamarcensis diagnostic AFLP markers possessed the H-W cpDNA haplotype. Inspecting the AFLP data for these individuals, two subgroups were identified. In the first subgroup, individuals displayed all or most of V. cundinamarcensis-specific markers. This must be the direct result of hybridization between V. cundinamarcensis, again as pollen donor, and V. × heilbornii as seed parent. As expected, their artificial hybrid was attributed to this group. Subsequent pollination of the progeny of such hybridization events with V. stipulata pollen could have given rise to the other individuals of class 4. Such backcrossing events will eventually lead to individuals having a varying number of V. stipulata-specific AFLP markers in combination with a lower number of V. cundinamarcensis-specific AFLP markers.

The individuals belonging to class 3 (Table 3), including the described varieties and cultivar of V. × heilbornii, display none of the V. cundinamarcensis-specific AFLP markers and lack typical V. cundinamarcensis morphological features. These specimens may be the products of further backcrossing of individuals belonging to classes 2 and 4 with V. stipulata, ultimately leading to the loss of all V. cundinamarcensis-specific AFLP markers. Alternatively, they also can be the result of other backcross events such as an F₁ or F₂ crossing with an introgressed individual.

Theoretically, if an individual is a backcross progeny, the constant markers of the non-recurrent parent should decrease by one-half with each successive backcross. Thus it is expected that a hybrid backcrossing once (BC₁) to V. stipulata would lose approximately half (23 or 24) of the V. cundinamarcensis markers. A second backcross (BC₂) would result in a further loss of approximately half of the remaining V. cundinamarcensis markers and so on. However, the genealogical categorization of individuals, as hypothesized in some cases above, should be considered with caution pending analysis of controlled crosses among the species included in this study. After all, Rieseberg and Linder (1999) reported that there might be a deviation from neutral random assortment. They found that genetic markers from Helianthus petiolaris were disproportionately lost when H. annuus × H. petiolaris hybrids were successively backcrossed to H. annuus; the BC₂ generation included individuals that appeared to be later generation backcrosses (BC₃ to BC₇). But, it is the actual genetic constitution of hybrids, not the pedigree that is most predictive of their characteristics or behaviour. Rieseberg and Linder (1999) attributed this lack of introgressive neutrality to selection against H. petiolaris as a consequence of chromosomal translocations and inversions that distinguish the two species. The loss of V. cundinamarcensis-specific markers among the backcross progeny (classes 2–4) may be skewed in a manner similar to that found in Helianthus.

In conclusion, strong evidence has been found for contemporary hybridization between V. stipulata, V. cundinamarcensis and V. × heilbornii in southern Ecuador.

Hypothesis on the origin of V. × heilbornii

Based on combined information from nuclear, chloroplast and mitochondrial DNA, different scenarios for the origin of V. × heilbornii can be proposed. PCR-RFLP analysis of the chloroplast DNA regions trnK1-trnK2 and trnM-rbcL and the mitochondrial region nad4/1-nad4/2 identified V. weberbaueri or an ancestor of this species as parental progenitor of V. × heilbornii. As shown before in smaller sample sets (Van Droogenbroeck et al., 2004; Kyndt et al., 2005a) most V. × heilbornii shared all cpDNA and mtDNA restriction sites with V. weberbaueri. This was unexpected on the basis of morphological observations (Romeijn-Peeters, 2004) and nuclear DNA data. Sequence analysis of the chloroplast regions matK and trnL-trnF (Kyndt et al., 2005b) of one or two specimens of each taxon revealed a very limited genetic divergence between the haplotype of V. weberbaueri and V. × heilbornii. Since cpDNA and mtDNA is usually maternally inherited in angiosperms (Reboud and Zeyl, 1994; Birky, 1995) and intergeneric Carica-Vasconcellea hybrids also revealed maternal cytoplasm inheritance (Van Droogenbroeck et al., 2005), the maternal ancestor of V. weberbaueri is possibly the maternal progenitor of V. × heilbornii, or was at least involved as cytoplasm donor in the origin of this species. These findings are substantiated by biogeographical data as both species have a restricted distribution in southern Ecuador and northern Peru (Badillo, 1993). In contrast to the cpDNA data, a possible contribution of V. weberbaueri cannot be observed in morphological features (Romeijn-Peeters, 2004). In addition, as only very few V. weberbaueri-specific AFLP fragments were detected in only some of the V. × heilbornii individuals in this study, clear nuclear support for the involvement of V. weberbaueri or its maternal ancestor was not found in the genesis of V. × heilbornii.

Wendel and Doyle (1998) reviewed the possible causes of incongruence between nuclear and chloroplast-based phylogenetic relationships, including organism-level processes of convergence, rapid diversification, hybridization and introgression, lineage sorting and horizontal gene transfer. The effects of hybridization, both in terms of origin of new taxa and gene flow, are considered to be the major cause of incongruence between nuclear and cpDNA-based plant phylogenies (Rieseberg et al., 1996; Wendel and Doyle, 1998). In the case of Vasconcellea, there is additional evidence that hybridization may affect patterns of relationships. As already mentioned, according to Badillo (1971), many Vasconcellea spp. are intercompatible and can produce hybrids with various degrees of fertility in nature, and a few probable interspecific hybrid populations have already been observed by Kyndt et al. (2005a). In addition, several artificial interspecific hybridizations among Vasconcellea species were reported to result in viable hybrids (Jiménez and Horovitz, 1958; Mekako and Nakasone, 1975). Hybridization and introgression therefore seem the most likely explanation for the observed incongruency between the two data sets.

Based on the results of this study and the phylogenetic data of Kyndt et al. (2005b), two hypotheses for the origin of V. × heilbornii can be proposed. The first hypothesis of evolution that might explain the origin of V. × heilbornii is founded on the provided evidence for occasional hybridization between V. stipulata and V. cundinamarcensis. It is possible that these two species also hybridized at an earlier phase during evolution, which resulted in the first V. × heilbornii-like individuals. Subsequently, a certain amount of interfertility between one of these first V. × heilbornii-like individuals and V. weberbaueri at that moment, possibly resulted in hybridization. Perhaps these first V. × heilbornii-like individuals have captured the ancient V. weberbaueri cytoplasm through these hybridization events, but they were able to retain their nuclear genome via extensive backcrossing (‘chloroplast capture’).

A second hypothesis is the direct hybridization between the maternal progenitor of V. weberbaueri, as seed parent, and V. stipulata, as pollen donor, followed by backcrossing of the hybrids to V. stipulata and/or subsequent molecular divergence. These first V. × heilbornii-like individuals, all possessing the ancient V. weberbaueri cytoplasm, further crossed with V. stipulata. As discussed in the previous section, backcrossing of these individuals with V. stipulata seems possible. Notice that V. cundinamarcensis is not involved in this hypothesis. This is in contrast to what has been suggested in literature so far, but also seems likely given the large group of V. × heilbornii-individuals lacking any of the V. cundinamarcensis-specific markers. Cytoplasmic gene flow is frequently observed without evidence of nuclear introgression (Rieseberg et al., 1996; Wendel and Doyle, 1998). This could explain the detection of only a limited number of nuclear markers of V. weberbaueri in V. × heilbornii. It is not known if hybridization between V. weberbaueri and V. stipulata or V. × heilbornii is possible, but in the light of the present results this would certainly be worthwile to investigate.

Based on the current knowledge, it is difficult to exclude any of the hypotheses. Further detailed analyses, both at the nuclear and cytoplasmic DNA level, of larger sample sets of V. stipulata and V. weberbaueri and the study of faster evolving chloroplast markers (e.g. chloroplast SSRs) are therefore highly recommended. From both morphological observations and molecular analyses there was no evidence for V. parviflora being involved in the genesis of V. × heilbornii.

Genetic diversity of V. × heilbornii

In this study and previous reports (Van Droogenbroeck et al., 2002, 2004; Kyndt et al., 2005a) a high degree of genetic diversity was observed in V. × heilbornii. This could be explained in two ways: (1) multiple origins, possibly the result of ongoing hybridization events; and (2) segregation following sexual reproduction. First, our molecular analysis of the plants collected in Loja and Azuay Province (Ecuador) showed that V. cundinamarcensis, V. stipulata and V. × heilbornii occasionally hybridize and confirmed that multiple origins are indeed one of the causes of a high level of variation within the hybrid. Secondly, although Badillo (1971, 1993) stated that V. × heilbornii individuals show some degree of parthenocarpy (up to complete absence of seeds) and male flowers are very rare in V. × heilbornii, morphological observations in the field (Romeijn-Peeters, 2004) show the presence of male specimens. Moreover, Scheldeman (2002) reported that fruits of V. × heilbornii contained 0–40 seeds/100 g fruit, revealing germination levels of at least 10 % (Scheldeman, 2002). Cursory pollen staining and germination tests revealed that at least some of the male V. × heilbornii plants in the authors' living collection produced viable pollen that germinated well (data not shown). In conclusion, considering that some of the V. × heilbornii individuals indeed retain at least partial fertility, gene flow via both pollen and seed is likely to be occurring at low frequency. These individuals can be involved in further hybridization events, either as seed parent or pollen donor, resulting in a set of genetically and morphologically diverse individuals, as observed in the present study. Both multiple origins and segregation after sexual reproduction have probably contributed to the pattern of high genetic diversity observed among the V. × heilbornii individuals.

This research project was funded by the Flemish Fund for Scientic Research (FWO-Vlaanderen Project nr. 3G005100) and a grant to Tina Kyndt from the ‘Institute for the Promotion of Innovation through Science and Technology in Flanders (IWT-Vlaanderen)’. The authors thank Dr Ir. Xavier Scheldeman for his continued support during their stay in Ecuador, for valuable advice and useful discussions.

LITERATURE CITED

Author notes

1Department of Molecular Biotechnology, Faculty of Bioscience Engineering, Ghent University (UGent), Coupure links 653, B-9000 Ghent, Belgium, 2Department of Biology, Faculty of Sciences, Ghent University (UGent), K.L. Ledeganckstraat 35, B-9000 Ghent, Belgium and 3Naturaleza & Cultura Internacional, Mercadillo 18-10 y José Maria Peña, Loja, Ecuador