Pan-genome and multi-parental framework for high-resolution trait dissection in melon (Cucumis melo)

Abstract

Linking genotype with phenotype is a fundamental goal in biology and requires robust data for both. Recent advances in plant-genome sequencing have expedited comparisons among multiple-related individuals. The abundance of structural genomic within-species variation that has been discovered indicates that a single reference genome cannot represent the complete sequence diversity of a species, leading to the expansion of the pan-genome concept. For high-resolution forward genetics, this unprecedented access to genomic variation should be paralleled and integrated with phenotypic characterization of genetic diversity. We developed a multi-parental framework for trait dissection in melon (Cucumis melo), leveraging a novel pan-genome constructed for this highly variable cucurbit crop. A core subset of 25 diverse founders (MelonCore25), consisting of 24 accessions from the two widely cultivated subspecies of C. melo, encompassing 12 horticultural groups, and 1 feral accession was sequenced using a combination of short- and long-read technologies, and their genomes were assembled de novo. The construction of this melon pan-genome exposed substantial variation in genome size and structure, including detection of ~300 000 structural variants and ~9 million SNPs. A half-diallel derived set of 300 F₂ populations, representing all possible MelonCore25 parental combinations, was constructed as a framework for trait dissection through integration with the pan-genome. We demonstrate the potential of this unified framework for genetic analysis of various melon traits, including rind color intensity and pattern, fruit sugar content, and resistance to fungal diseases. We anticipate that utilization of this integrated resource will enhance genetic dissection of important traits and accelerate melon breeding.

Keywords: Cucumis melo; SNP; crop breeding; disease resistance; fruit-quality; genetic mapping; half-diallel; pan-genome; structural variation.

Figure 1

The MelonCore25 set and mapping platform development scheme. (a) Mature fruits of the 25 founders, ordered by subspecies and horticultural (cultivar‐) groups. (b) Schematic diagram for the components and development process of the multi‐parental framework.

Figure 2

Melon pan‐genome and distribution of structural variations (SVs) across HDA25. (a) Phylogenetic tree of the 25 founder lines presented with variation in genome size and examples of large inter‐chromosomal inversions on chromosomes 1 and 11 in ssp. agrestis. (b) The genomes of the ssp. agrestis accessions are significantly smaller than those of the ssp. melo accessions. (c) Size distribution of 108 011 large SVs (> 2 kb). The X‐axis is presented in log scale. (d) Distribution of allele frequency of 108 011 large SVs (> 2 kb). (e) Size distribution of 86 669 short (< 2 kb) bi‐allelic InDels across MelonCore25. The X‐axis is presented in log scale. (f) Distribution of allele frequency across 86 669 bi‐allelic sites (< 2 kb). (g) Pan InDels analysis. Horizontal black lines are the mean number of pan InDels under the corresponding number of genomes. Dark boxes are the 50th percentile and light boxes are the 99th percentile. (h) Correlation between parental distance (calculated using 24 000 genome‐wide GBS SNPs) and number of polymorphic short InDels across the 300 HDA25 crosses. (i) Number of polymorphic short InDels by chromosomes and bins, in the pan‐genome and two crosses. In parentheses is the total number of short Indels for each group.

Figure 3

Genetic analysis and mapping of mottled rind. (a) Representative mottled rind fruits from the GWAS180 collection. (b) Parallel frequencies of mottled‐ and non‐mottled rind accessions in the GWAS180 collection and the MelonCore25 set. (c) Projection of rind pattern (mottled or non‐mottled) on the genetic PCA plot of the MelonCore25 set. Dashed lines indicate crosses (and derived segregating populations) used for genetic analysis of the mottled‐rind trait. (d) PI414723 and DUL, the parents of the RILs population used for mapping the mottled‐rind trait. (e) Linkage mapping results of the MT‐2 locus to chromosome 2. (f) Zoom in on chromosome 2 mapping in the RILs of the cross PI414723 × DUL and comparison with a Manhattan plot derived from genome‐wide association (GWA) analysis of the GWAS180 collection. (g) Substitution‐mapping validation for the MT‐2 locus using recombinants from the RILs of the cross PI414723 × DUL. Green cells represent flanking markers for the interval. The yellow cell is the rind phenotype column. A = DUL allele, B = PI414723 allele. The genome‐wide association study (GWAS) peak interval is represented by the red horizontal line.

Figure 4

Multi‐parental characterization of the mottled rind trait. (a) Inheritance of rind pattern across the HDA20 set (190 half‐diallel F₁s and their 20 parents). Each cell represents the combination of parents from the horizontal and vertical axes. Squares arranged in the diagonal represent the phenotypes of the parents. (b) Contingency analysis for the segregation of rind pattern against the MT‐2 marker in the F₂ of the cross QME × TAD. Het = Heterozygote. (c) Contingency analysis for the segregation of rind pattern against the MT‐2 marker in the F₂ of the cross PI161375 × ESL. Het = heterozygote. Examples for young fruit segregants (mottled and non‐mottled) are displayed. (d) Contingency analysis for the segregation of rind color against the APRR2 marker in the F₂ of the cross PI161375 × ESL. Het = heterozygote. Examples for mature fruit segregants (light and dark rind) are displayed. (e) Contingency analysis for the segregation of rind pattern against the MT‐2 marker in the RILs from the cross PI414723 × DUL. (f) Schematic pivot table that presents the co‐segregation of the APRR2 and MT‐2 markers against rind color and pattern in the F₂ of the cross PI161375 × ESL. (g) Allelism test for the mottled‐rind trait in the F₂ of the cross DUD × PSR. (h) Allelism test for the mottled rind trait in the F₂ of the cross QME × PSR. (i) Allelism test for the mottled rind trait in the F₂ of the cross QME × PI414723.

Figure 5

Variation in fruit sugars across MelonCore25 and selected segregating populations for genetic mapping. (a) Distribution of total soluble solids (TSS) of mature fruits across MelonCore25. (b) Projection of accession TSS on the genetic PCA. Dashed lines represent crosses (and derived segregating populations) used for genetic dissection of fruit sugar accumulation. (c) Correlation between content of the monosaccharides, glucose and fructose, and disaccharide, sucrose, in mature fruits of MelonCore25. (d) Selection of tails segregants through the analysis of correlation between F₃ and F₄ lines of the cross TAD × QME (HDA008). High and low TSS tails are represented in red and blue, respectively. (e) Selection of tail segregants through the analysis of correlation between F₄ and F₅ lines of the cross TAD × PI164323 (HDA192). High and low TSS tails are represented in red and blue, respectively. (f) Selection of tail segregants through the analysis of correlation between replications in the F₄ population from the cross SAS × DOYA (HDA243). High and low TSS tails are represented in red and blue, respectively. (g) Correlation between TSS of parental means (mid‐parent) and TSS of their F₁ hybrids, across 190 F₁s (HDA20 population).

Figure 6

Variation in response to artificial inoculation with Fusarium oxysporum f. sp. melonis races 1 and 2 and haplotypic variation in the FOM‐2 gene across MelonCore25. (a) Projection of accession response to inoculation with Fusarium race 1 on the genetic PCA. (b) Projection of accession response to inoculation with Fusarium race 2 on the genetic PCA. (c) Haplotypic variation in the FOM‐2 gene (MELO3C021831) across MelonCore25. Vertical colored lines are SNPs. Gray histograms reflect short‐read depth. The dashed rectangle in haplotype 3 is the1100‐bp insertion discovered through Nanopore sequencing and de novo assemblies. Remarkably, a ~500‐bp deletion was initially indicated in this region based on absence of Illumina short‐read alignments. The actual insertion was discovered only based on the long‐read assemblies (Figure S5). Colored dots correspond to the Fusarium race 1 response as presented in (a). Part of the gene model is presented below the haplotypic view. Triangles below the gene model are PCR primers for FOM‐2 CAPs markers. Blue: marker from Oumouloud et al. (2015). Red: CAPs marker developed for the Newe Ya‘ar breeding program (Table S8). (d) Projection of accession FOM‐2 haplotype on the genetic PCA. (e) Gel image of polymerase chain reaction (PCR) validation of the 1100‐bp InDel.

Figure 7

Variation in response to Fusarium oxysporum f. sp. radicis‐cucumerinum (FORC) across MelonCore25 and QTL mapping of TAD × DUL RILs. (a) Experimental design and examples of disease response of susceptible (S) and resistant (R) accessions, 7–14 days post‐inoculation. (b) Projection of the responses of the 25 accessions to FORC on the genetic PCA. The dashed line marks a cross, HDA034, used for the quantitative trait loci (QTL) mapping. (c) Frequency distribution of FORC disease symptoms index (DSI) across 153 TAD × DUL RILs. Responses of parents are marked. (d) Manhattan plot for linkage mapping of response to FORC in the TAD × DUL RILs. (e) Allelic effect of the chromosome 7 QTL. For both respective genotypic groups, DUL and TAD, the two gray boxes represent the standard deviation and the blank spaces in‐between the allelic mean.