Long-read bitter gourd (Momordica charantia) genome and the genomic architecture of nonclassic domestication

Abstract

The genetic architecture of quantitative traits is determined by both Mendelian and polygenic factors, yet classic examples of plant domestication focused on selective sweep of newly mutated Mendelian genes. Here we report the chromosome-level genome assembly and the genomic investigation of a nonclassic domestication example, bitter gourd (Momordica charantia), an important Asian vegetable and medicinal plant of the family Cucurbitaceae. Population resequencing revealed the divergence between wild and South Asian cultivars about 6,000 y ago, followed by the separation of the Southeast Asian cultivars about 800 y ago, with the latter exhibiting more extreme trait divergence from wild progenitors and stronger signs of selection on fruit traits. Unlike some crops where the largest phenotypic changes and traces of selection happened between wild and cultivar groups, in bitter gourd large differences exist between two regional cultivar groups, likely reflecting the distinct consumer preferences in different countries. Despite breeding efforts toward increasing female flower proportion, a gynoecy locus exhibits complex patterns of balanced polymorphism among haplogroups, with potential signs of selective sweep within haplogroups likely reflecting artificial selection and introgression from cultivars back to wild accessions. Our study highlights the importance to investigate such nonclassic example of domestication showing signs of balancing selection and polygenic trait architecture in addition to classic selective sweep in Mendelian factors.

Keywords: Momordica charantia; artificial selection; domestication; genome assembly; population genetics.

Fig. 1.

Features of the Goya v2 assembly. Shown are regions missed in the Dali-11 assembly, F_ST, reduction of diversity (ROD), composite likelihood ratio (CLR), and cross-population composite likelihood ratio (XP-CLR) test results.

Fig. 2.

Population structure and demographic history of M. charantia. Shown are the (A) population structure, (B) phylogenetic tree, (C) principal component analysis, and (D) demographic history of different wild (THAI and TAI) and cultivar (SA and SEA) groups.

Fig. 3.

Genetic architecture of selection in bitter gourd fruit traits. (A) Pictures (not to scale) showing the typical phenotypes of wild (Left, short, green, and spiny) and cultivar accessions. The typical weight of wild fruits is less than 30 g, whereas cultivar fruits could be more than 500 g. (B) Phenotypic distributions among genetic groups for fruit color, length, and presence of spines. (C–E) Relationship among top trait-associated SNPs and their mean (C) effect size, (D) F_ST between cultivar and wild groups, and (E) F_ST between the SA and SEA cultivar groups. The horizontal axes show SNPs with different magnitudes of association with traits (0.1, 1, 5, and 10% SNPs with the lowest P values). Solid lines represent mean test statistic for target SNPs, and shaded areas represent the 95% range of 1,000 resampled background SNP sets. Asterisks above solid lines represent values higher than the top 5% of background values. Blue represents WILD-SEA, red represents WILD-SA, and purple represents the SA-SEA comparisons.

Fig. 4.

The region with low differentiation between and high polymorphism within wild and cultivar groups in chromosome 1. (A) This region in chromosome 1 is labeled in shade, showing low F_ST, high PI, and the colocalization with the gynoecy QTL. The left half of chromosome 1 is repeat-rich centromeric regions, and the red asterisk labels the region enlarged for (B). (B) The high-polymorphism region in cultivars contains a candidate gene NUA likely affecting flower sex ratio. (C) Phylogenetic tree of the 1.77-Mb shaded area in A. The phylogeny is mostly consistent with genomewide phylogeny, except some wild accessions have obvious introgression from cultivars. (D) Phylogenetic tree of the NUA coding region, showing several balanced allelic groups with very low diversity within and high difference between groups. (E and F) Phylogenetic trees of 500-kb flanking regions upstream and downstream of the 1.77-Mb low-F_ST region. For C to D, tip label colors represent the population assigned based on genomewide SNPs. Red, wild accessions; gray, wild-cultivar admixed accessions; blue, South Asian cultivars; and cyan, Southeast Asian cultivars.