The double haploid line, M62-113 (from Semillas Fito S.A.), belonging to the Piel de Sapo commercial type (Cucumis melo subsp melo var inodorus), was used as the starting cultivar. This is the parent of the melon genetic map and is one of the parents of the DHL line selected for the sequencing of the melon genome. Different batches of a total of ~12,000 M62-113 seeds were treated with the alkylant mutagen ethyl methanesulfonate (EMS). Different EMS concentrations were tested, and 1% was finally selected and tested on a larger batch of seeds. Seeds were first hydrated with tap water (16h), and were then soaked in tap water containing EMS concentrations (18 h). The treated seeds were thoroughly rinsed in tap water twice for 4 hours. After washing, seeds were placed on trays over wet paper, and were kept in a germination chamber overnight. ~5,000 mutant plants of this first (M1) generation were transplanted at the greenhouse. Plants were grown in two localities, in Barcelona (by Semillas Fito S.A. and IRTA) and in Valencia (by COMAV-UPV). Each M1 plant was selfed (1 fruit per plant) giving rise to the M2 seed.

Twenty seeds were sown per M2 family, and the germination percentage was scored. Finally, a total of 2,368 mutagenized M2 families, which produced at least 10 viable seedlings in at least two plantings, were sampled for DNA extraction. 1-cm discs of tissue from 10 individual plants per M2 were pooled, and total DNA was extracted, quantified, diluted and organized into twenty-five 96-well plates. Using equivalent amounts of DNA from individual M2 families, the samples were pooled four-fold and organized into a 96-well format. Seeds of all pooled M2 populations are maintained at the Genebank of the COMAV-UPV, coded as F (S. Fito), I (IRTA) or C (COMAV-UPV) plus the corresponding number.

Primer sets were designed from 3 genes that were selected based on their potential contribution to fruit quality and disease resistance: Cm-eIF4E (translation initiation factor 4E) and Cm-eIF(iso)4E (translation initiation factor E, Isoform) which are involved in resistance to viruses 31 , and Cm-PDS (Phytoene Desaturase), which is involved in carotenoid synthesis 32 . The CODDLe program (Codons Optimized to Deliver Deleterious Lesions) 33 was used to select the regions most likely to harbor deleterious changes induced by EMS (Table 1). These 3 genes were tilled in the Piel de Sapo population as well as in a set of 2,483 available lines of the cantalupensis TILLING population reported in 30 . Four additional genes involved in fruit ripening were tilled in the Piel de Sapo population: Cm-ACO1 (ACC oxidase 1), involved in the conversion of ACC into ethylene 34 ; Cm-NOR (non-ripening), a transcription factor related to ethylene-sensitive/insensitive phenotypes 35 ; Cm-DET1 (de-tiolated-1), a negative regulator of light-mediated responses that affects carotenoid and flavonoid pathways in tomato and other crops 36 37 ; and Cm-DHS (Deoxyhypusine Synthase), mutations of which delay fruit softening in tomato 38 . Primer sets used for these additional genes are the same as those used in 30 . The amplicons analyzed for each gene are indicated in Figures 1 and 2.

Nested PCR was used to improve the specificity of amplification. The PCR reactions were performed in a 25 μl volume consisting of dH₂O, 1× PCR buffer (Promega Corp, Madison, WI), 2.5 mM MgCl₂, 0.2 mM dNTPs, 1 U Taq polymerase, 0.4 mM forward and reverse primers, unlabeled in the first reaction, and 700 nm and 800 nm 5' labeled (MWG Biotech AG, Germany) in the second reaction, and 10 ng DNA. The thermocycling conditions were 95°C for two minutes for initial denaturing, followed by 35 cycles of 95°C for 20 seconds, 63-69°C (specific for each primer pair) for one minute, 72°C from 30 seconds to one minute and one cycle of 72°C for five minutes. PCR products (~0.2 μl) were separated in 1% agarose gels. The PCR products were heated and cooled in a thermocycler (a gradient starting at 94°C and decreasing 0.1°C per second to 4°C) to form the heteroduplexes. Once the heteroduplexes were formed, the products were treated with Endo I. Digestions were performed in 30μl of final volume with dH2O, 3μl digestion buffer, 150 ng of PCR product and 3μl of EndoI (diluted 1/5000). The products were incubated at 45°C for 20 minutes to digest mismatches in the heteroduplex. Digestion was stopped by adding 5μl of 0.15 M EDTA and placing the reaction on ice. After the digestion was completed, the products were filtered through a Millipore MultiScreen filter plate that was packed with hydrated Sephadex G-50 medium beads (Amersham Biosciences). Samples were concentrated in the Speed Vac at 65°C for 45 minutes. Loading dye was added and the PCR products were denatured and loaded onto a polyacrylamide gel attached to a LI-COR 4300 DNA Analyzer (LICOR, Lincoln, NE) for separation. A total of 903 pools were assayed (592 four-fold pools of the Piel the Sapo population and 311 eight-fold pools of the cantalupensis population). Once a mutation was revealed in a pool, the corresponding families were analyzed individually to select the mutant M2 family. Mutation was confirmed by sequencing in the pooled DNA, in the individual M2 families and in several individual plants used in the segregation studies. The effect of the mutations was analyzed with SIFT (Sorting Intolerant from Tolerant, http://blocks.fhcrc.org/sift/SIFT.html) 39 , which predicts whether an amino acid substitution affects protein function. Scores below 0.05 are predicted to affect protein function.

Seedling phenotypes were systematically evaluated in the M2 generation. Additionally, in order to increase the number of seeds in our mutagenized population, we obtained the M3 generation. M2 populations that are being reproduced are also being inspected for phenotypes that are distinct from the wild type in plant, flowering and fruit traits. To date, about 800 M2 plants have been reproduced by S. Fito. M3 plants are also being analyzed for ethylene-response mutants by growing seedlings in darkness with and without an air flux containing 10 microL L-1 ethylene, and analyzing the "triple response".

We selected 0.5%, 1 and 1.5% for generating a pilot population of ~600 lines, and studied the effect of EMS dosage on M2 seed viability and seedling vigor. Increased EMS concentration led to an increased percentage of M2 fruits in which seeds were non-viable or in which the number of viable seeds was too low to perform TILLING analysis (13.24%, 27.75% and 31.38%, respectively). Only a slight reduction in the average germination rate was observed in M2 families obtained from 0.5% M1 (93.51%), whereas higher reductions were obtained in 1% and 1.5% (82.73 and 75.94%, respectively). In addition, seedling growth was more restricted at the 1.5% concentrations. In contrast, the progeny from the 1%-treated M1 were robust. Based on these results, 1% EMS was used in the end to complete the mutagenized melon population. This EMS dose produces an acceptable level of M1 seed survival, fertility in M1 plants to set viable M2 seeds, and vigor in M2 seedlings. From a total of 3,140 M1 fruits finally obtained, 2,368 M2 families were sampled for DNA and used for TILLING purposes.

Ten plants per M2 line were scored for seedling phenotypes. A large number of the mutations affected cotyledons, with 5% of the M2 families showing variation in their number position and shape. About 4.3% of the M2 families segregated for "dwarf" or "semidwarf" plants. Albinism and chlorophyll deficiency occurred in 2.1% of the M2 families. Over 2% of the families showed alterations in leaves and shoot morphology. Gravitropic mutants and necrosis also appeared, but were less frequent. Figure 3 shows several seedling phenotypes. M3 families were characterized for plant flowering and fruit traits. Fruit traits of interest for melon breeding were identified: longer and rounder fruits than the standard Piel de Sapo fruits, with a lighter rind color, without the typical rind spots, and with different degrees of netting. Some lines have a subtle aroma and showed a color change and the formation of an abscission layer near maturity, traits that are more similar to climacteric cantalupensis types than to the standard non-climacteric inodorus Piel de Sapo. Differences in the dark-grown seedlings' response to ethylene were also observed. Wild type M62-113 showed the typical triple response: exaggeration of the apical hook, together with the inhibition of hypocotyl and root growth as well as the radial swelling of the shoot 40 , whereas some M3 families showed responses of varying intensities (Figure 3).

The TILLING approach allowed us to identify nucleotide changes in two of the three gene targets first screened in the Piel de Sapo population (Table 2). EMS-induced changes were mostly G:C to A:T transitions (67%) (Table 3), as was expected due to the frequent alkylation of guanine residues by EMS, thus forcing mispairing with T 41 . Four mutations were exonic, most of which were determined to be missense mutations, only one being silent. A graph of the target genes marking the location of the induced polymorphisms is shown in Figures 1 and 2.

Four new alleles were reported for Cm-eIF4E, defined by 4 mutations. Two (in exons 1 and 2) were predicted to affect protein function according to SIFT (p<0.05). The segregation of the missense mutations was analyzed in each mutant M2 family (Table 3) and homozygous individuals for each mutation are being produced for phenotypic analysis. Two new alleles were reported for Cm-PDS. Mutations were detected in the first intron and in the second exon of the amplicon (Figure 1). The exonic mutation was predicted to affect protein function.

The genotype/phenotype segregation was verified for the Cm-PDS mutations in 15 plants of each M2. The intronic mutation found in M2 family F1151 (mutation 20 in Figure 1) presented a Mendelian segregation of 4 wild type/7 heterozygous/4 homozygous mutant (fitting to a 1:2:1 ratio) (Table 3). No associated phenotype was observed, which agrees with the position of the mutation in an intronic region. The M2 family C384 carried a missense mutation in exon 2 (mutation 21 in Figure 1). The segregation of M2 was 9 heterozygous/6 wild type. In this family, we observed that some plants showed delayed growth and even death at the cotyledon stage, and therefore could not be sampled. M3 seeds from a heterozygous M2 plant were used to repeat segregation analysis. Fourteen plants germinated out of the 20 seeds sown. Again, plants with delayed growth were observed, but we also found a seedling with albino cotyledons and hypocotyl (Figure 4). This seedling was carefully grown until tissue could be sampled for DNA extraction. The albino was the only plant homozygous for the mutant allele; the remaining M3 plants were wild type or heterozygous. This mutant line showed a similar albino phenotype comparable to the previously reported phenotypes for PDS disruption 32 .

To further evaluate the mutation rate of the Piel de Sapo population, we extended the TILLING screen to 4 additional genes involved in fruit quality. Mutations were found for all genes, except for Cm-NOR, with Cm-DET1 being the most mutated gene. Six mutations were exonic, most determined to be missense, only one silent, but all were predicted to be tolerated (Figure 2, Table 3).

We used these results to calculate the mutation rate of the population. A total of 14 point mutations were detected out of the total 11,534 bp analyzed. Previous studies reported the difficulty in tracking mutations on the ends of the fragments (~100 bp) 42 . Therefore, 200 bp was subtracted off each amplicon, and we considered the length screened in this study to be 9,134 bp. The overall mutation density was then calculated by dividing the total base pairs screened, which includes the sum of the total length of the 12 amplicon sizes × the total number of individuals screened, by the total number of mutations revealed by TILLING ((9,134 × 2,368)/14 = 1,544,950). It was calculated to be ~1/1.5 Mb. The mutation frequency as determined by TILLING is coherent with the moderate level of phenotypic mutants found in our population.

In order to compare the inodorus and the cantalupensis populations, the same genes were screened in a set of 2,483 lines of the Charentais population 30 . The number of mutations was higher, and new alleles were found for the three genes (Table 2 Figure 1). The spectrum of observed changes was similar to that found in the Piel de Sapo population (Table 3). Only one mutation was found for Cm-eiF(iso)4E. Cm-eIF4E was also the most mutated gene with 14 mutations: seven intronic, two in the 3'-UTR, and five exonic. Of the exonic mutations, two, located in exons 1 and 2, were not tolerated (p<0.05). One intronic mutation was located in the first nucleotide of intron 3, and may affect a splicing site. TILLING analysis also revealed four mutations in Cm-PDS, one of which was not tolerated. The 19 point mutations discovered in the 3,871 bp analyzed in the Charentais population make a mutation rate of ~1/401 Kb ((3,071 × 2,483)/19 = 401,331). This is about 3 times higher than that found in the Piel de Sapo population. This high mutation rate in the cantalupensis population has also been observed by screening additional genes in the final version of the population, which consists of 4,023 lines 30 .