Introduction

Although most flowering plants are hermaphrodite (bisexual flowers develop on each plant), a significant number of species also produce unisexual flowers, leading to a diversity of sex phenotypes. The genetic and molecular studies performed on different sexual types of these species provided valuable information regarding sex determination mechanisms in plants (Westergaard 1958; Charlesworth 1991; Grant 1999; Ainsworth 2000; Filatov et al. 2000; Liu et al. 2004). Melon (Cucumis melo L., 2n=2x=24) is cultivated throughout the world. Breeding selections are generally based of fruit quality and resistance to pests and diseases (Whitaker and Davis 1962; Stepansky et al. 1999). Like other members of the Cucurbitaceae family, melon is remarkable for its diversity of sexual types, and sex expression is most challenging for melon breeders (Rosa 1928; Poole and Grimball 1939; Kenigsbuch and Cohen 1990). Although many traditional melon cultivars are andromonoecious (separate hermaphrodite and male flowers develop on each plant), they are being replaced by new hybrid varieties that tend to be monoecious (plants bearing female and male flowers). Monoecy is of agronomic interest both for breeding practices and for the seed industry due to three main reasons: (1) emasculation of the flowers is not required to avoid inbred lines; (2) the facility to produce hybrid seeds ensures confidentiality and preserves breeder rights; and (3) fruits produced by monoecious hybrids develop a smaller bottom scar (Fig. 1), resulting in a reduced risk of pathogen infections and better fruit quality (Périn et al. 2002b).

Fig. 1
figure 1

A The andromonoecious (a) gene has major effect on sexual type of melon. Monoecious plants (bearing the A allele) develop separate female (left) and male (center) flowers, while andromonoecious plants (homozygous aa) develop male and hermaphrodite (right) flowers. Moreover, the a gene has pleiotropic effects on the abscission zone of the flower which is smaller in hermaphrodite than in female flowers (arrows). B The scar bottom, which is the shutting kept on the fruit after the flower abscission, is also smaller in fruits developed from female (left) than from hermaphrodite (right) flowers. Scale represents 0.5 cm

Full size image

The genetic control of sex determination in melon depends on three genes, andromonoecious (a), gynomonoecious (g), and maleness (M), which combine to generate the different sexual types (Poole and Grimball 1939; Keningsbuch and Cohen 1990). Among these, monoecious (A-GG) and andromonoecious (aaGG) types are determined by the dominant and recessive allele of the a gene in combination with the gynoecious dominant allele (G-). Hermaphrodite type is homozygous for a and g recessive alleles (aagg), while mm genotype seems to stabilize gynoecy of AAgg plants (see Keningsbuch and Cohen 1990). For breeding purposes, monoecious and andromonoecious types are the ones commonly used as plant material, while other sexual types segregating for G and M genes are not usually considered by melon breeders. Recent mapping results published by Danin-Poleg et al. (2002), Périn et al. (2002a) and Silberstein et al. (2003) have located the a gene in two different linkage groups, although in all cases the monoecious US PI 414723 was a donor parent in the corresponding mapping populations. However, both linkage groups contain a resistance gene (Zym) and a microsatellite marker (CmGA36), indicating they correspond to the same genomic region. The partial map of Danin-Poleg et al. (2002) included an RAPD marker at 16.2 cM to the a gene, while Silberstein et al. (2003) described an RFLP marker linked at 7 cM to the same locus. The genetic map constructed by Périn et al. (2002a) placed the a gene on a 25.2-cM linkage region containing two QTLs for ovary and fruit shape. This linkage region is likely to contain genes controlling sex and fruit shape.

The combination of bulked segregant analysis [(BSA) Michelmore et al. 1991] and highly polymorphic PCR-based markers permits the identification and mapping of useful molecular markers for breeding purposes. Markers linked to sex-controlling genes were identified in several dioecious species, including asparagus (Asparagus offficinalis, Reamon-Bütter et al. 1998; Reamon-Bütter and Jung 2000), hop (Humulus lupulus, Polley et al. 1997), Pistacia vera (Hormaza et al. 1994), and papaya (Carica papaya, Urasaki et al. 2002; Deputy et al. 2002). In the Cucurbitaceae family, markers linked to sex expression were identified in cucumber (Cucumis sativus, Trebitsh et al. 1997; Kahana et al. 1999), in which sex expression is also under the control of three genes (see review of Robinson and Decker-Walters 1997). One of these genes, F, promotes female flower production and determines the gynoecious phenotype when combined with the MM genotype, whereas cucumber plants of ffMM genotype are monoecious. Moreover, genetic analyses performed on ethylene-related genes suggest that the F gene could affect ethylene biosynthesis and sex expression in cucumber plants (Trebitsh et al. 1997; Kahana et al. 1999). Association between sex-related and ethylene-related genes has not been reported in melon. The identification and mapping of molecular markers linked to sex controlling genes would be valuable for breeding and for elucidating the sex-determining mechanisms. In this paper, we report the development of a codominant SCAR marker linked to monoecious trait by using a BSA methodology (Michelmore et al. 1991) with a double-haploid (DH) population of Charentais type melon.

Material and methods

Plant material

Three homozygous breeding lines of melon (C. melo L.), all belonging to the Charentais varietal type, were used in this study. A monoecious line (CM) and an andromonoecius line (CA1) were the parents of a DH population composed of 38 individuals used in the BSA analysis. DH lines were generated by pollination of female flowers with Co60 gamma-irradiated pollen, followed by in vitro rescue of parthenogenic embryos. In vitro culture conditions and colchicine treatments for chromosome doubling were performed following the protocol described by Lotfi et al. (2003), with minor changes. The third line (CA2) is an andromonoecious line that was crossed with CM to generate a backcross (BC1) population [(CM×CA2)× CA2] composed of 530 individuals, which was used for linkage analysis. Both CA1 and CA2 were used as female parents and produced round fruits, whereas CM produced oblong fruits. In order to know the genotype and the stability of sexual type in every DH line, 12 offprint plants were grown in the greenhouse and their phenotype scored at three different developmental stages: at the initiation of flowering, in the middle of the reproductive cycle when the plants set the first two full-sized fruits, and at the end of the crop cycle. All plants were grown in a plastic greenhouse located in Almeria (Spain) under natural sunlight conditions.

DNA isolation

Young leaves either from 12 plants of every DH line or from individual BC1 plants were harvested, quickly frozen in liquid nitrogen, and stored at −80°C. DNA was extracted using the DNeasy Plant Mini kit (Qiagen), according to the recommendations of the supplier. The total DNA was dissolved in water, visualized after electrophoresis in 1% agarose gels 1× TBE (Tris-borate-EDTA), and quantified by comparison with DNA standards (GIBCO–BRL). Then it was diluted to 100 ng/μl and kept frozen at −20°C until used in PCR reactions.

Marker analysis

The AFLP analysis was carried out as described by Vos et al. (1995), using MseI (New England Biolabs) and EcoRI (Roche) as restriction enzymes. Digestion was performed with 5 U each enzyme in a final volume of 40 μl for 2 h at 37°C, and it was followed by the addition of 1 U T4-DNA Ligase (Roche) and incubation for 1 h at 37°C. Pre-amplification reaction were done in a volume of 20 μl, using A as the selective nucleotide for the EcoRI primer (EcoRI+A) and three different MseI primers (MseI+C, MseI+G, and MseI+T). The second PCR amplification was performed with primers that included three selective bases in their sequences being the EcoRI primer radioactively labeled with γ[33P]-ATP (3,000 Ci/mmol). AFLP products were separated on 6% denaturing polyacrylamide gels, using a Sequi-Gen Cell sequencing equipment (Bio-Rad). Electrophoresis was carried out for 2 h in 1× TBE; gels were then dried for 2 h at 80°C and exposed to radiographic Kodak film over 3 days at room temperature.

In order to increase the probability of identifying some marker linked to the a gene, we generated four bulks from the DH population, composed of eight plants each—two comprising monoecious plants and another two with andromonoecious plants. DNA from the bulks and the two parental lines was simultaneously screened with 79 AFLP primer combinations (Table 1). AFLP markers were named according to the standard list for AFLP primer nomenclature (Keygene, The Netherlands, http://wheat.pw.usda.gov/ggpages/keygene/AFLPs.html). Polymorphic bands among the contrasting DNA bulks, which were also displayed by the parents, were further analyzed both in DH and BC1 individual plants.

Table 1 AFLP primer combinations used to find markers linked to the a gene
Full size table

Conversion of AFLP markers

The DNA fragment corresponding to the AFLP marker closest to the a gene was excised from the gel and purified according to Qu et al. (1998). Briefly, the piece of gel containing the amplified fragment was hydrated and incubated at 37°C overnight; then it was centrifuged to pellet solid debris, and the upper phase was 100-fold-diluted. The DNA isolated was re-amplified by PCR, following the same AFLP selective amplification described above but using non-labeled primers. The resulting PCR products were purified by size-exclusion chromatography, using the GenElute PCR Clean-up kit (Sigma), visualized, and quantified by 1% agarose gel electrophoresis. The appropriate DNA fragment was cloned in the T/A-cloning vector pGEM-T Easy (Promega) and transformed into Escherichia coli strain DH5α, according to the manufacturer’s instructions. Clones of the expected insert size were selected and sequenced using the T7 and SP6 primers. Nucleotide sequences were obtained from an ABIPRISM 377 Sequencer (Applied Biosystem) and used to design SCAR primers by synthesizing two specific oligonucleotides from both ends of the AFLP fragment. Amplification of the SCAR marker was performed in 50 μl, using 10 ng total DNA, 50 ng each SCAR primer, 2.5 mM dNTP, 4 mM MgCl2, and 1 U RED-Taq DNA polymerase (Sigma) in 1x Taq buffer. Cycling condition were as follows: 94°C, 60 s; and then 35 cycle of 94°C, 30 s, 56°C 60 s, and 72°C, 60 s; and an extension of 5 min at 72°C. Products from the SCAR experiments were analyzed in 2% agarose gels in 1× TBE.

Analysis of data

A χ2 test was used to test the goodness-of-fit of the Mendelian segregation of the a gene classes. The recombination frequencies between the marker and the a gene were calculated according to the method of Allard (1956), whereas map distances were based on Kosambi’s mapping function (Kosambi 1944).

Results

F1 plants from crosses between a Charentais monoecious line, CM, and two Charentais andromonoecious lines, CA1 and CA2, were all monoecious, indicating that the development of unisexual flowers is a dominant character. To further characterize the genetic control of monoecy, 38 DH lines were obtained from the hybrid CM × CA1. A segregation analysis of the a gene was carried out in 12 descendants of each DH plant as well as in a BC1 population composed of 530 descendants from the backcross (CM × CA2) × CA2. Both segregating populations (DH and BC1) were generated from the same monoecious donor parent, and they were found to fit a segregation ratio 1:1 for either sexual type (monoecious:andromonoecious). χ2 test value was highly significant in the DH population (0.24, P=0.62), but showed a lower level of significance in the BC1 population (5.7, P=0.02). Nevertheless, all these results confirmed the monogenic inheritance of the monoecy/andromonoecy in melon as was previously described by Kenigsbuch and Cohen (1990).

Genetic linkage of AFLP markers to the andromonoecious gene

A total 79 AFLP primer combinations were used to screen DNA from the two parental lines (CM and CA1) and four DNA bulks constructed for the BSA analysis, two of them from monoecious and two from andromonoecious DH plants. PCR analyses resulted in the amplification of 5,119 AFLP fragments, of which 1,893 (37%) were polymorphic among the CM and CA1 parents. As expected, most of the primer combinations used in the BSA analysis did not detect any polymorphism among the four DNA bulks. However, one AFLP primer combination (E33M85) exhibited a polymorphic banding pattern between parents that was also maintained when compared DNA profiles from monoecious and andromonoecious bulks (Fig. 2). This primer combination produced a DNA fragment in the monoecious parental line, named M3A, which was also present in both monoecious DNA bulks but not in the andromonoecious plants, either parental CA1 line or bulks (Fig. 2). Interestingly, the same primer combination amplified another DNA fragment, M3a, with a complementary pattern to M3A. The M3a marker was shown by the andromonoecious plants (CA1 and bulks) but was absent in the monoecious parent and in one of the monoecious DNA bulks (Fig. 2). This result indicated that this primer combination amplified molecular markers putatively linked to the A/a gene. In order to confirm this hypothesis, both markers were mapped in the DH population (Fig. 2). This preliminary linkage analysis revealed a single recombination event between the marker and the gene, indicating a genetic distance of 3.3 cM between the marker M3 and the a gene (Kosambi mapping function). Polymorphism detected by this AFLP primer combination behaved as a codominant marker (M3) with two alleles, M3A and M3a, linked to the monoecious (A) and andromonoecious (a) alleles, respectively.

Fig. 2
figure 2

AFLP analysis performed in andromonoecious (CA1) and monoecious (CM) parents as well as in double-haploid (DH) lines, the latter grouped in two andromonoecious (bA) and two monoecious bulks (bM). Bulked segregant analysis is shown in the left panel, while the individual analysis performed in DH plants is shown in the right panel. The two M3 allelic markers, M3A and M3a (arrows), cosegregate with A (monoecious) and a (andromonoecious) alleles respectively, except for one monoecious DH plant (asterisk). Only a section of the AFLP gel is shown for a better quality of the figure

Full size image

Development of a codominant SCAR marker

M3 marker was converted to a SCAR marker, which could be useful for genotyping selection. With this objective, it was characterized by cloning and sequencing the two DNA fragments corresponding to M3A and M3a markers. The sequence of the monoecious allele M3A was 83-bp long and exhibited a 42-bp deletion when compared to the corresponding sequence from the andromonoecious allele M3a, which was 125-bp long (Fig. 3). This result proved that both fragments were alleles of the same locus. Database searches with both sequences did not show any homology in other genomes.

Fig. 3
figure 3

Alignment of DNA sequences from both alleles of the M3 marker, M3A and M3a, differs in a 42-bp insertion/deletion. Primer sequences designed to amplify the codominant SCAR marker (showed in Fig. 4) is in boldface. Note that sequence from M3a allele includes three tandem repetitions of a 21-bp motif (underlined)

Full size image

After the alignment of the nucleotide sequences from M3A and M3a, two specific PCR primers were designed from the ends of both allelic markers. This primer combination yielded products of the expected size in the parental lines CA1 and CM, and the corresponding polymorphism was readily observed in 2% agarose gels (Fig. 4). Additionally, a longer DNA fragment was also observed, with a variable intensity in the andromonoecius parental line. This slower-moving band most probably corresponds to a heteroduplex fragment formed by an almost-perfect tandem duplication present in the M3a sequence (Fig. 3). PCR amplification performed, using as the DNA template the originally cloned M3a DNA, confirmed this hypothesis (data not shown). The codominant marker was used to screen the BC1 population composed of 530 individuals generated from the cross CM × CA2 and using CA2 as the recurrent parent (Fig. 4). Previously, we have sequenced both parental lines, CA1 and CA2, to show they shared the same M3a allele of the marker. A total of 29 recombinant events were found between a gene and the M3 marker, indicating that the marker was linked to the a gene at a genetic distance of 5.5 cM (Kosambi function).

Fig. 4
figure 4

PCR amplification of the codominant SCAR marker M3 in the parental lines, CA1 and CM, the F1 progeny, and 16 BC1 plants. Among the BC1 plants, eight were andromonoecious and eight monoecious, and a recombinant individual is shown (asterisk). Molecular size of PCR products is indicated (left)

Full size image

Discussion

Sexual expression is currently an important objective in melon breeding programs. Breeders aim to introgress monoecy in different varietal types because of the agronomic advantages that this trait provides in terms of pollination control, seed production, and fruit quality (Kenigsbuch and Cohen 1990; Périn et al. 2002b). However, time-consuming selection based on phenotypic traits and the linkage of the a gene to fruit shape slowed down breeding progress. Taking into account the low genetic variability described in melon germplasm (Shattuck-Eidens et al. 1990; Neuhausen 1992; Staub et al. 2000), we have used AFLPs to identify molecular markers linked to a gene, as large numbers of polymorphic markers can be produced using this technique (Vos et al. 1995; Garcia-Mas et al. 2000). AFLPs have also proved to be extremely powerful for the identification of polymorphic markers in different plant species, since a high number of informative data points can be obtained per sequencing gel. Moreover, the results obtained in this work seem to confirm that one of the most straightforward applications of AFLPs in marker-assisted breeding includes the isolation of markers linked to specific genes. AFLP markers can also be detected as allelic markers, and their codominant nature can be demonstrated by analyzing heterozygous plants for a given polymorphism. In fact, the BSA analysis carried out among andromonoecious and monoecious plants revealed a polymorphic AFLP locus (M3) with two putative allelic fragments. These fragments segregated complementarily and differed in 42 bp, a greater molecular size than that suggested by Périn et al. (2002a), to consider two AFLP markers as allelic. Thus, there may not be a limit to the difference in size of two putative AFLP allelic fragments, although smaller-sized polymorphisms are obviously more likely to occur than greater chromosome mutations.

The BSA approach performed in this work consisted in the comparative DNA analysis of four bulks of plants constructed from a DH population. It was designed to find markers at tight genetic distances, and allowed us to identify an AFLP marker linked to the a gene, at a genetic distance of 3.3 cM. After cloning and sequencing, the AFLP fragment was converted into a codominant SCAR marker and was then mapped again in a BC1 population of 530 plants, revealing a genetic distance of 5.5 cM. Such difference in the genetic distance could be due to the different properties of the mapping populations (DH and BC1), particularly the number of individuals and the genetic background. In any case, the two alleles of the M3 marker enable us to distinguish between homozygous and heterozygous monoecious individuals, which in turn allow a faster and more reliable development of melon breeding lines and new hybrid varieties.

Some mapping approaches have located molecular markers in the vicinity of the a gene, but they are of limited use for genotyping selection since the genetic distances are greater than the one shown by the M3 marker developed here. A partial genetic map described by Danin-Poleg et al. (2002) located an RAPD marker at a distance of 16.2 cM from the a gene in linkage group IV, whereas Périn et al. (2002a) mapped this gene in a region covering 25.2 cM in linkage group II of a reference map. Although the latter is the most saturated map published to date for melon, with an the average genetic distance of 3.24 cM between markers, there is a poor density of molecular markers in the chromosome region flanking the a locus. More recently, the genetic map developed by Silberstein et al. (2003) describes an RFLP located 7 cM away from the a gene, which was the closest marker until the publication of the present paper. Nevertheless, the genetic distance and the dominant nature of the RAPD marker on the one hand, and the complexity of the RFLP technique on the other, make difficult the extensive and routine use of these markers for breeding purposes. Thus, the M3 SCAR developed may be considered the most reliable and closest molecular marker to the a gene identified in melon genome. The codominant nature of M3 and the late expression of the target trait (monoecious/andromonoecious type) make this marker a useful tool for a early genotyping selection of desirable sexual types in this species. Interestingly, M3 was identified from a DH population generated from a cross between two phylogenetically related Charentais lines. This suggests that in a BSA approach, not only the genetic distance between the parental lines but also the homozygosity level in chromosome regions flanking the locus of interest, may contribute to finding markers closely linked to this locus.

Searching analysis of the databases did not reveal homology of the M3 DNA fragment with any known sequence, indicating that M3 is most probably part of a non-coding region of the melon genome. The short length of the M3 allelic fragments does not refute the fact that they could be part of a coding sequence, which would not necessarily be related to sex determination given the genetic distance of M3 from the a gene. Whether this is the case, a more detailed characterization of the genomic region containing the M3 marker is needed to probe this hypothesis. Interestingly, a restriction fragment homologous to the ACC-synthase1 gene (CS-ACS1) has only been observed in gynoecious plants of cucumber (C. sativus), not in monoecious ones (Trebitsh et al. 1997). These results not only suggested a functional role of ethylene in sex determination but also that CS-ACS1, a gene involved in the biosynthesis of this hormone, could correspond to the Female (F) gene of cucumber. In melon, on the other hand, none of the few ethylene-related genes mapped to date are located in the same linkage group as the a gene (Périn et al. 2002a; Silberstein et al. 2003). Our results on the comparative sequencing analysis between the parental lines of the DH population regarding three ACC-oxidase genes and five ACC-synthase genes did not reveal any molecular polymorphism in the coding regions of these regulatory genes that can be associated with the sex type (data not shown). Exhaustive genetic and physiological studies will be required to determine the functions of ethylene-related genes in the sex differentiation process of plant species, for which melon and other cucurbits are excellent plant models.

Sex of melon flowers appears to be strongly associated to the fruit shape, as all monoecious DH plants produced more oblong fruits compared to the round ones produced by andromonoecious DH lines. These results agree with those obtained by Périn et al. (2002b), who demonstrated a pleiotropic effect of the a gene on fruit shape, using genetic and sex-reversion analyses. These results seem to confirm the influence of a gene on early developmental patterns of melon fruits. The same authors reported two tightly linked QTLs responsible for the variation of fruit shape located in the same 25.2-cM chromosome region as the a gene. Although fruit shape is a polygenic trait, the SCAR marker described in this work is located close to both the a gene and the two major QTLs that determine fruit shape. This could facilitate breeding selection not only of sexual types but also of fruit shape in melon.