Large-scale whole-genome resequencing unravels the domestication history of Cannabis sativa

Abstract

Cannabis sativa has long been an important source of fiber extracted from hemp and both medicinal and recreational drugs based on cannabinoid compounds. Here, we investigated its poorly known domestication history using whole-genome resequencing of 110 accessions from worldwide origins. We show that C. sativa was first domesticated in early Neolithic times in East Asia and that all current hemp and drug cultivars diverged from an ancestral gene pool currently represented by feral plants and landraces in China. We identified candidate genes associated with traits differentiating hemp and drug cultivars, including branching pattern and cellulose/lignin biosynthesis. We also found evidence for loss of function of genes involved in the synthesis of the two major biochemically competing cannabinoids during selection for increased fiber production or psychoactive properties. Our results provide a unique global view of the domestication of C. sativa and offer valuable genomic resources for ongoing functional and molecular breeding research.

Fig. 1. Population structure of Cannabis accessions.

(A) Geographic distribution (i.e., sampling sites of feral plants or country of origin of landraces and cultivars) of the samples analyzed in this study. Color codes correspond to the four groups obtained in the phylogenetic analysis and shapes indicate domestication types. The two empty red squares symbolize drug-type cultivars obtained from commercial stores located in Europe and the United States. For sample codes, see table S1. (B) Maximum likelihood phylogenetic tree based on single-nucleotide polymorphisms (SNPs) at fourfold degenerate sites, using H. lupulus as outgroup. Bootstrap values for major clades are shown. (C) Bayesian model–based clustering analysis with different number of groups (K = 2 to 4). Each vertical bar represents one Cannabis accession, and the x axis shows the four groups. Each color represents one putative ancestral background, and the y axis quantifies ancestry membership. (D) Nucleotide diversity and population divergence across the four groups. Values in parentheses represent measures of nucleotide diversity (π) for the group, and values between pairs indicate population divergence (F_ST). (E) Principal component analysis (PCA) with the first two principal components, based on genome-wide SNP data. Colors correspond to the phylogenetic tree grouping.

Fig. 2. Demographic history of C. sativa and selection signatures identified from comparison between hemp- and drug-type cultivars.

(A) Demographic history inferred from the PSMC method (30). (B) Graphical summary of the best-fitting demographic model inferred by fastsimcoal2 (65). Widths show the relative effective population sizes (N_e). Arrows and figures at the arrows indicate the average number of migrants per generation among different groups. The point estimates and 95% confidence intervals of demographic parameters are shown in table S3. Examples of genes with selection sweep signals in hemp-type cultivars (C) and drug-type cultivars (D). Three independent sets of signals (F_ST, π ratio, and XP-CLR) are shown along the genomic regions covering the four genes. Dashed lines represent the top 5% of the corresponding values. Below the three plot schemes are the gene models in the genomic regions. Below each gene model are the SNP allele distributions along each of the four genes for the two groups (green, heterozygous site; orange, homozygous site of reference allele; blue, homozygous site of alternative allele; gray, missing data).

Fig. 3. Evolution of CBDAS and THCAS.

(A) Occurrence of CBDA-synthase gene (CBDAS), THCA-synthase gene (THCAS), and two CBDAS pseudogenes across 104 Cannabis accessions, based on mapping to a reference genome having both genes and many pseudogene copies of them [Jamaican Lion DASH (42)]. Cladogram on top and symbols are as in Fig. 1. For sample codes, see table S1. Below the cladogram is indicated for each gene whether reads from each sample mapped to the reference positions. The height of each gene box represents the length of the gene. The Jamaica Lion DASH genome sequence coordinates for the four genes are shown on the right. (B) Top left: Phytocannabinoids CBDA and THCA result from a biosynthetic reaction catalyzed respectively by the enzymes CBDA and THCA synthase from the common precursor CBGA. Bottom: The proportion of CBDAS and THCAS in each of the four groups. Top right: The proportion of CBDAS and THCAS in landraces versus cultivars within the Hemp-type group. Fisher’s exact test, *P < 0.05; ***P < 0.001. (C) Transcriptomic expression for the two genes and pseudogenes in different tissues and vegetative stages [data from (47)]. Wilcoxon rank-sum test, *P < 0.05.