The genus Hylocereus (Cactaceae) of the vine cacti, known as dragon fruit or pitayas, comprises 15 perennial species (Bauer, 2003). These Hylocereus species bear medium to large (200–800 g) edible fruits with broad scales, various peel and flesh colors, and numerous small soft seeds (Tel-Zur et al., 2004a, 2011b). Among these species, those cultivated worldwide include: the diploids H. undatus and H. monacanthus, which have white, purple or red flesh and a red peel; the tetraploid H. megalanthus, with white flesh and a yellow spiny peel (Lichtenzveig et al., 2000); and the spontaneous (natural) or induced (artificial) interspecific hybrids of these three species (Tel-Zur et al., 2012a; Pagliaccia et al., 2015). Interest in these species has increased markedly in recent years, since they are regarded as exotic fruit crops with high economic potential and as a rich source of natural micronutrients (Wu et al., 2006) and phytochemicals with therapeutic activities (Song et al., 2016). This rise in interest is indeed reflected in the registration of many new cultivars in recent years (Lobo et al., 2016). An important plant characteristic promoting the expansion of areas under cultivation with these species is their crassulacean acid metabolism (CAM), which confers exceptional tolerance to extreme drought: such plants show four- to six-fold higher water use efficiencies than other fruit crops grown under similar conditions (Mizrahi et al., 2007). These cactus species are thus a valuable commodity for dryland farmers in regions with increasing water scarcity and land degradation.

Despite the growing interest in these fruit crops, they remain underutilized minor perennial crops, and there is a need for effective breeding programs to facilitate the expansion of areas under cultivation. To this end, a long-term breeding program was begun some three decades ago at Ben-Gurion University of the Negev (BGU) to develop improved hybrids with superior fruit quality, good yields, and enhanced tolerance to extreme temperatures (Tel-Zur, 2013). However, this breeding program—like many breeding programs for other minor crops—has been hampered by a variety of factors, including difficulties in applying molecular and other updated breeding approaches, the limited genomic information currently available, and the plants' large genome size and high heterozygosity (Cisneros and Tel-Zur, 2013; Plume et al., 2013; Pagliaccia et al., 2015). In addition to all these factors, 6–10 generations of self-pollination are needed to obtain phenotypically stable homozygote lines. Since Hylocereus species are perennial species with a prolonged juvenile stage and a strong self-incompatibility system in the diploid species, the use of both traditional and advanced breeding approaches faced severe limitations. There was thus a need to develop the accurate platform research required to breed these crops.

To meet this need, the first step in the BGU breeding program was an investigation of the taxonomy and ploidy of the Hylocereus species. The original classification of H. megalanthus was performed by Britton and Rose (1920) on the basis of its unique morphology: since the species has a triangular stem, like that of all Hylocereus species, and spiny fruits, like those of Selenicereus, they classified it into a separate genus, which they named Mediocactus, thereby implying both an intermediate morphology and an intermediate taxonomic status. Nonetheless, Bauer has placed this species in the Hylocereus genus, reflecting the close affinity between this tetraploid species and the other species of this genus, which are all diploids (Bauer, 2003; Tel-Zur et al., 2011b). The fact that a tetraploid taxon shares morphological features with two genera (Hylocereus and Selenicereus), which are also cross compatible (Lichtenzveig et al., 2000; Tel-Zur et al., 2012b), might imply an allopolyploid origin. Indeed, early studies suggested an intermediate taxonomic status for H. megalanthus (Tel-Zur et al., 2004a,b), but no evidence of an allopolyploid origin was found in later work (Plume et al., 2013). Thus, the origin of H. megalanthus remains an open question—one that can perhaps be resolved taking an approach that differs from the more conventional methodologies described above, namely, by applying the idea of “gametoclonal variation” put forward by Morrison and Evans (1987) to describe the variation observed among plants regenerated from gamete cells in culture. The regeneration of H. megalanthus plants with half a set of chromosomes (two instead of four) via anther or ovule culture may offer an additional tool to explore the origin of this tetraploid species.

In both H. megalanthus and H. monacanthus, successful haploid induction via androgenesis or gynogenesis (Garcia et al., 2009a,b) has paved the way for breeding programs and genetic studies, since gamete-derived progenies have, theoretically, a lower level of heterozygosity than that of the parental lines (Verdoodt et al., 1998). Moreover, favorable outcomes may be expected from haploid induction, since polymorphism, at the molecular and phenotypic levels, between the parental line and the gamete-derived progenies will lead to the creation of novel phenotypes due to homozygous recessive genes and gametoclonal variation (Morrison and Evans, 1987). These outcomes will facilitate a search for molecular markers related to agronomic traits for plant improvement (Germanà, 2011; Ravi et al., 2014).

Here, we report a comprehensive evaluation of the gamete-derived progenies from H. monacanthus and H. megalanthus that were previously generated by Garcia et al. (2009a,b). On the basis of a previously reported correlation between ploidy level and stomatal density for a number of species (Beaulieu et al., 2008) and also for Hylocereus (Tel-Zur et al., 2011a,b), we performed a comparative stomatal study between the donor plants and the putative gamete-derived lines to evaluate the utility of applying this tool as a preliminary screen for ploidy. We then confirmed the ploidy level by using flow cytometry. In parallel, we proved the gamete origin of these lines by using simple sequence repeat (SSR) markers at the sequencing level. We also evaluated the gamete-derived progenies in terms of morphological traits, pollen viability, ovule development, fruit weight and reproductive potential in comparison with the donor species. In addition to reporting these findings, in this article we also address gametoclonal variation in the regenerated di-haploid H. megalanthus lines, thereby shedding light on the possible origin of this tetraploid species. Finally, we discuss the value of these lines for further genomic analysis and breeding programs.

In ovule or anther culture, plants can regenerate not only from gamete cells but also from somatic cells. Several plant species produce both reduced (n) and unreduced (2n) gametes with the somatic chromosome number (Bretagnolle and Thompson, 1995). Unreduced (2n) gametes (pollen or ovules) result from meiotic dysfunction, mainly as a result of two processes, namely, first division restitution (FDR) and second division restitution (SDR), which occur during abnormal first and second meiotic divisions, respectively (Bretagnolle and Thompson, 1995 and references therein). FDR gametes have two non-sister chromatids and exhibit levels of heterozygosity similar to those of the parent (maintaining all the parental genes except cross-over fragments), while SDR type gametes have two sister chromatids and exhibit lower levels of heterozygosity than those the parent (Bretagnolle and Thompson, 1995; Younis et al., 2014). Thus, a diploid status and a lower level of allelic diversity can be attributed either to an unreduced (2n) male gamete from SDR (the sister chromatids do not separate during the second meiotic division) or to spontaneous chromosome doubling in a haploid (n) male gamete, which restores the diploid status. However, for the diploid H. monacanthus there is evidence only of normal reduced gamete formation (Tel-Zur et al., 2003) and regular chromosome disjunction at anaphase I (Lichtenzveig et al., 2000), suggesting only a negligible occurrence of unreduced gametes in this species. Ploidy estimation (using flow cytometry analysis) for these five lines showed the same ploidy as the donor (Table 1). Nonetheless, these lines exhibited a lower level of allelic diversity in comparison with the donor plant (Figure 1A), and therefore they were identified as true androgenic double-haploid lines. Thus, we can assume that spontaneous chromosome doubling occurred during early plant development in all five androgenic H. monacanthus lines.

The tetraploid H. megalanthus is known to produce both normal reduced (n) gametes and unreduced (2n) gametes (Tel-Zur et al., 2003). Ploidy estimation for the putative gamete-derived H. megalanthus lines showed 46 diploids and 24 tetraploids (Table 2), where the diploids are indeed true gamete-derived lines (“di-haploid lines”). Among the 24 lines estimated as tetraploids, 21 lines showed the same five-band pattern as the donor line. This finding could indicate a somatic rather than a gamete-derived origin, but unreduced gametes of the FDR type possess levels of heterozygosity that are identical or close to identical with those of the parent. Furthermore, the origin of the three double di-haploid (432, 1347, and 1736) lines could derive either from a normal reduced gamete that had subsequently undergone spontaneous chromosome doubling or from unreduced gametes of the SDR type. Additional work using cytological, morphological and/or phenotypic markers—alongside molecular tools—is required to reveal the origin of these lines and to further examine their potential value for breeding.

On the basis of the general correlation between genome size (or ploidy) and stomatal parameters, extensive attempts have been made to develop methodologies for rapid pre-screening and ploidy estimation in plant species (see Germanà, 2011 and references within). In our hands, we found that stomatal density was a good proxy for the preliminary estimation of ploidy in gamete-derived H. megalanthus lines, since only few values were outliers or overlapped in the 2x and 4x results (Figure 4A). In contrast, considerable overlapping ranges in stomatal length were found between di-haploid and tetraploid regenerated lines (Figure 4B).

The five double-haploid H. monacanthus lines showed low vigor and developed abnormal flowers with malformed ovules and aborted pollen grains. Thus, the main obstacle to using double-haploid H. monacanthus lines in breeding programs is the severe male and female infertility of these lines, in keeping with the low fertility in double-haploid lines that has also been reported in other plant species. Joyce et al. (2003) suggested that under in vitro culture conditions a high level of mutations gives rise to deleterious characteristics. In a study on Solanum, low fertility was attributed to an aneuploid status, with 2C values deviating from those of the normal euploid (Valkonen et al., 1994). In robusta coffee (Coffea canephora), only half the double-haploid lines that had been planted out survived field conditions, and those lines showed low vigor and reduced fertility, suggesting a strong negative effect of homozygosity on vigor and reproductive characteristics (Lashermes et al., 1994). In apple (Malus × domestica), double-haploid lines also showed low vigor, together with compressed tree architecture and abnormal flower morphology (Höfer and Flachowsky, 2015). In double-haploid cocoa lines (Theobroma cacao), fertility varied from levels similar to those of the diploid donor to very low levels, which were not improved upon grafting (Lanaud, 1987). The reduced fertility in these cocoa lines, which was manifested in the abnormal development of female and male organs (as expressed by both pollen abortion and the lack of normal differentiation of embryo sacs), was attributed to the phenomenon known as “homozygote depression” that is forced in cross-fertilized species (Lanaud, 1987). Similar findings were obtained for our double-haploid H. monacanthus lines, i.e., low vigor and abnormal flower development.

The di-haploid and double di-haploid H. megalanthus lines were female fertile to different degrees (Table 2). Among the 46 di-haploid lines, 14 set fruit, with a wide range of average fruit weights between lines and a large standard deviation at the line level (Table 2); performances therefore ranged from “similar to donor lines” to “very low female fertility.” Reduced fruit weight was found in all the lines of gamete origin (di-haploids and double di-haploids) in comparison with the tetraploids of somatic origin (Figure S5). Some of these di-haploid and double di-haploid lines were controlled crossed with pollen donors from other Hylocereus species (interspecific crosses); fruits of these interspecific crosses reached maturity and contained viable seeds (data not shown). These new interspecific hybrids will be used for continuing the breeding program and further genetic studies. Although another 13 di-haploid lines did bloom (and pollen viability was measured), the developing fruits abscised before maturation (or set less than three fruits). Similar to the reduced fertility discussed above for H. monacanthus, reduced fertility in H. megalanthus might be linked to the abnormal ovules produced in the di-haploid and double di-haploid lines (Figure 3F). Thus, it is likely that for this tetraploid species, too, reduced heterozygosity drastically affected viability, supporting once more the theory of “homozygote depression.”

Male fertility (measured as the percentage of pollen grain viability) was low for all the di-haploid and double di-haploid H. megalanthus lines, i.e., from very low levels to totally sterile lines (Table 2). In haploid lines, chromosome doubling is needed to restore fertility (Germanà, 2011). However, in our hands, the three double di-haploid H. megalanthus lines and the above-described double-haploid H. monacanthus lines also showed very low level of pollen viability, pointing to the simple fact that doubling the set of chromosomes was not sufficient to restore fertility in gamete-derived H. monacanthus or H. megalanthus lines.

Knowledge about the origin of the species in any crop improvement program is fundamental to the success of the project. In this regard, our above-described findings support Bauer's taxonomy that placed the tetraploid H. megalanthus in the same taxon as the diploid Hylocereus species (Bauer, 2003). Further support comes from the study of Plume et al. (2013), which showed a low level of polymorphism within or among H. megalanthus accessions, meaning an autopolyploid rather than an allopolyploid origin. Their findings, in turn, are corroborated by the almost insignificant gametoclonal variation (namely, variation in plant morphology) observed in our study in the di-haploid H. megalanthus lines in comparison with the donor plant: the only difference lay in the size of the floral organs (bud length and width) of the di-haploid lines, which was smaller than that of the tetraploids (Figures S3, S4) These findings contrast with the wide variability in plant morphology reported for a di-haploid potato by Rousselle-Bourgeois and Rousselle (1992).

Among the different facets of our long-term breeding and research program for expanding the cultivation of Hylocereus—as a new emerging fruit crop—an attempt was made to improve breeding efficiency by the development of gamete-derived lines in two Hylocereus species. The current study shows that the use of double-haploid H. monacanthus lines is severely limited by the sterility of these lines. Nonetheless, hybrid plants produced from the single fruit that was obtained from the interspecific cross of the double haploid H. monacanthus line 011 as the female parent are currently growing in our greenhouse, and additional efforts to obtain more progeny will be made in the near future. Hybrids from some of the di-haploid and double di-haploid H. megalanthus lines are also currently growing in our greenhouse. For these lines, the pollen grains were almost totally sterile, but the ovules showed different levels of fertility. Thus, in practice, these lines are male sterile (require cross pollination) but they are female fertile to different degrees and can therefore be used to develop new hybrids. Although, the number of produced hybrids was limited and probably lower than that desirable for intensive breeding, this new plant material offers unique possibilities for genetic research and breeding.

DL and NT-Z conceived the experimental design, DL conducted the cytological and molecular assessments, MA performed the morphological evaluation, RS conducted the histological work, and DL and NT-Z wrote the manuscript. All authors read and approved the final manuscript.