Phased genomics reveals hidden somatic mutations and provides insight into fruit development in sweet orange

Abstract

Although revisiting the discoveries and implications of genetic variations using phased genomics is critical, such efforts are still lacking. Somatic mutations represent a crucial source of genetic diversity for breeding and are especially remarkable in heterozygous perennial and asexual crops. In this study, we focused on a diploid sweet orange (Citrus sinensis) and constructed a haplotype-resolved genome using high fidelity (HiFi) reads, which revealed 10.6% new sequences. Based on the phased genome, we elucidate significant genetic admixtures and haplotype differences. We developed a somatic detection strategy that reveals hidden somatic mutations overlooked in a single reference genome. We generated a phased somatic variation map by combining high-depth whole-genome sequencing (WGS) data from 87 sweet orange somatic varieties. Notably, we found twice as many somatic mutations relative to a single reference genome. Using these hidden somatic mutations, we separated sweet oranges into seven major clades and provide insight into unprecedented genetic mosaicism and strong positive selection. Furthermore, these phased genomics data indicate that genomic heterozygous variations contribute to allele-specific expression during fruit development. By integrating allelic expression differences and somatic mutations, we identified a somatic mutation that induces increases in fruit size. Applications of phased genomics will lead to powerful approaches for discovering genetic variations and uncovering their effects in highly heterozygous plants. Our data provide insight into the hidden somatic mutation landscape in the sweet orange genome, which will facilitate citrus breeding.

Figure 1

Pronounced haplotype differences in sweet orange revealed by a diploid assembly. (A) Genetic admixtures of pummelo and mandarin in the diploid sweet orange genome detected using 50-kb non-overlapping windows. The percentages from pummelo and mandarin are indicated on the y-axis and highlighted with different colors. (B) Comparison of unique kmers between two haplotypes (A_B) and between mandarin and pummelo (M_P). k = 21 and 61. (C) A 180-kb haplotype-specific region on chromosome 1 revealed by syntenic analysis.

Figure 2

Detection and characterization of somatic mutations in sweet orange. (A) Evaluation of somatic mutations using a single-reference genome and haplotype-based mapping methods. The percentage of detected mutations are indicated on the y-axis. The simulated reads with different amounts of coverage are indicated on the x-axis. (B) Statistics for somatic variations in the sweet orange genome calculated using a single-reference genome and haplotype-based mapping methods. (C) Correlations between genome size and somatic variations in two haplotypes. (D) Network phylogeny analysis for somatic variations in the 87 accessions from the sweet orange population. (E) Annotation of somatic variations with high allele frequencies (>60 accessions). (F) Normalized dN/dS in a somatic population of sweet orange. The distribution of neutral somatic mutations was estimated in simulations.

Figure 3

Allele-specific expression during fruit development in sweet orange. (A) Distribution of fold changes in expression. The fold change in expression for the allele expressed at relatively high levels compared to the allele expressed at relatively low levels is indicated on the x-axis. The threshold for extremely different fold change (EASE) genes is highlighted. (B) Correlation analysis between allele specific gene expression and heterozygous variations. The proportion was calculated based on the number of differentially expressed alleles and the corresponding number of genes in each 500-kb window. (C) Three expression patterns (clusters I, II, and III) and module-trait relationships from a WGCNA. The seven modules are indicated on the y-axis. The asterisks indicate corrected P values <0.05.

Figure 4

Integrative analysis of candidate genetic factors affecting fruit size in sweet orange. (A) Phenotypic differences between BT_2 and BT_5 fruit. Scale bar, 2 cm. (B) Cross sectional diameters (mm) of BT_2 and BT_5 fruit at 11 developmental stages. *P value <0.05; **P value <0.01 (Student’s t test). (C) Principal component analysis (PCA) of transcriptomes from BT_2 and BT_5 fruit at 70, 120, and 170 DAB. (D) Heatmap of expression levels for 17 differentially expressed fruit size and cell expansion-related genes. Genes were expressed at significantly different levels during at least one developmental period. (E) Expression of different EXP10.2 (HA6g01160) alleles. Counts for the expressed reads from different alleles are indicated on the y-axis. (F) Somatic variations located in the gene region or in the 3-kb upstream and downstream regions. (G) Validation of somatic mutations in two alleles of FS8.1 using Sanger sequencing. (H) Expression of somatic mutation related alleles from the FS8.1 gene (haplotype B). Counts for allelic expressed reads are indicated on the y-axis.