Incipient diploidization of the medicinal plant Perilla within 10,000 years

Abstract

Perilla is a young allotetraploid Lamiaceae species widely used in East Asia as herb and oil plant. Here, we report the high-quality, chromosome-scale genomes of the tetraploid (Perilla frutescens) and the AA diploid progenitor (Perilla citriodora). Comparative analyses suggest post Neolithic allotetraploidization within 10,000 years, and nucleotide mutation in tetraploid is 10% more than in diploid, both of which are dominated by G:C → A:T transitions. Incipient diploidization is characterized by balanced swaps of homeologous segments, and subsequent homeologous exchanges are enriched towards telomeres, with excess of replacements of AA genes by fractionated BB homeologs. Population analyses suggest that the crispa lines are close to the nascent tetraploid, and involvement of acyl-CoA: lysophosphatidylcholine acyltransferase gene for high α-linolenic acid content of seed oil is revealed by GWAS. These resources and findings provide insights into incipient diploidization and basis for breeding improvement of this medicinal plant.

Fig. 1. Genome of the allotetraploid P. frutescens.

a Images of mature plants of the allotetraploid PF40 and the diploid PC02 used for de novo assemblies. b Mapped features of the allotetraploid genome including (1) chromosomes arbitrarily numbered in descending order of their assembled lengths, (2) mapping depth distribution by PC02 in 10-kb windows, (3) distribution of 527 pairs of HE genes on PFA (as blue lines) and PFB (red lines) subgenomes, (4) density of predicted genes in 500-kb windows (with values 0–67), (5) density of predicted pseudogenes in 500-kb windows (0–67), (6) percentage of repeats in 500-kb windows (0.5–1.0), and (7) PFA-PFB synteny linked by red lines (n = 15,170). Ticks on the outer circumference represent 5-Mb units of chromosome length.

Fig. 2. Evolution of the allotetraploid Perilla.

a Distribution of synonymous nucleotide substitutions (dS) between the four perilla sequences. The dS = 0 signal between PFA-PC02 (n = 8939) was not displayed. b Chromosomal synteny between PF and PC genomes. Each dot represented syntenic gene relationship between PFA-PC02 (19,412 gene pairs, in red) or PFB-PC02 (15,422 gene pairs, in blue). Scattered segmental duplications not related to polyploidization were shown by magenta dots. PF chromosomes underlined were reversed for visual consistence. c Patterns and statistics of nucleotide mutational signatures of PFA and PC02 since polyploidization. The signatures are displayed according to the 96-substitution classification defined by substitution class and sequence context immediately 5′ and 3′ to the mutated base, and displayed alphabetically from ANA to TNT. d Subgenome expression dominance as calculated by log2 transformed TPM (Transcripts Per Million) ratio of PFA to PFB syntenic genes (n = 15,484). Solid lines represented RNA-seq data of PF40 from flower and leaf with three replicates each. For any paired TPM values of <1, a pseudo-count of 1 was added to both PFA and PFB values before log2 ratio calculation. Enlarged inset showed expression bias toward PFA with a minor peak around 0.2 (red downward arrowhead). Source data underlying Fig. 2c are provided as a Source data file.

Fig. 3. Patterns of subgenome exchanges in the allotetraploid.

a Sequencing depth distribution between syntenic subgenomes indicated homeologous exchange in PF185. Coverage by PC02 indicated that the chr1 block was of AA-origin. In PF185, a 397-kb segment of chr2 (deletion, shown in green) was replaced by 281-kb homeologous sequences of chr1, resulting in a seemingly duplication of the chr1 segment (in red). Red boxes marked the corresponding HE intervals. b Reciprocal HE swapped 180 kb of chr7 with its syntenic chr19 interval. This balanced exchange was inherited by all tetraploid accessions. c Replacement of chr13 segments by non-homeologous chr15 sequences resulted in gene dosage imbalance (1:3) in PF175.

Fig. 4. Population analysis of perilla germplasms.

a Neighbor-joining tree of 191 allotetraploid perilla accessions. Green, yellow, and purple branches indicated three perilla clades of south China (n = 120), north China (26), and crispa (45), respectively. Totally 4,789,738 filtered high-quality SNPs were used for tree construction. b PCA analysis of these samples with the same SNP dataset. c Population structure of these samples based on different numbers of clusters (K = 2–4). d Nucleotide diversity (π) and population divergence (F_ST) across the three clades. The value in each circle indicates calculated nucleotide diversity for that clade, and the value along each line represents population divergence between two neighboring clades. e Neighbor-joining tree constructed with 108,850 PFA-specific SNPs, and overlaid with leaf color information: dashed lines indicated green leaves, and solid lines represented red leaf phenotype. Color codes were the same in (a), (b), (d), and (e).

Fig. 5. GWAS analysis of perilla leaf color variation.

a Manhattan plot. The strongest associated SNP on chr8 was marked by a red arrow. Dashed horizontal red line indicated significant p-value threshold of 1.04e−8 (calculated as 0.05/n). Leaf (abaxial) phenotypes were shown as inset. b Exon structure of the causal Myb113 gene. Coding exons are shown as red boxes, and UTRs in gray. Positions of three coding variants are marked by vertical black lines. c Schematic representation of LTR insertion in diploid and subsequent 3′ end partial deletion in allotetraploid. Black box on chromosome (gray line) indicated the 5-nt target site that was duplicated upon LTR integration, and red box on chromosome represented the 5-nt micro-homologous sequences that initiated the 9967-bp segment deletion. Drawn not to scale. d A proposed scenario for evolution of perilla leaf color. Note that the 6-bp in-frame deletion in the 2nd exon of Myb113 (white asterisk) first emerged within the crispa clade.

Fig. 6. Analysis of perilla seed oil biosynthesis genes.

a Expression heatmap of TAG biosynthesis genes during perilla seed development. TPM values of TAG genes were extracted from seven transcriptomes corresponding to the seeds of 2, 6, 10, 14, 18, 22, and 26 days post anthesis (DPA) of the high oil content line PF40 (~56%), and displayed as log2(TPM + 1). Each row represented one predicted TAG-related genes in PFA-PFB alternating manner. The first column indicated functional categories of these genes. A detailed list of these 33-pairs of syntenic genes can be found in Supplementary Data 8. b Manhattan plot of GWAS analysis for ALA content of perilla seed oil. c Comparison of ALA content between two SV haplotypes in the GWAS population. Error bars, mean ± s.d. Source data underlying Fig. 6a are provided as a Source data file.