Genome and systems biology of Melilotus albus provides insights into coumarins biosynthesis

Abstract

Melilotus species are used as green manure and rotation crops worldwide and contain abundant pharmacologically active coumarins. However, there is a paucity of information on its genome and coumarin production and function. Here, we reported a chromosome-scale assembly of Melilotus albus genome with 1.04 Gb in eight chromosomes, containing 71.42% repetitive elements. Long terminal repeat retrotransposon bursts coincided with declining of population sizes during the Quaternary glaciation. Resequencing of 94 accessions enabled insights into genetic diversity, population structure, and introgression. Melilotus officinalis had relatively larger genetic diversity than that of M. albus. The introgression existed between M. officinalis group and M. albus group, and gene flows was from M. albus to M. officinalis. Selection sweep analysis identified candidate genes associated with flower colour and coumarin biosynthesis. Combining genomics, BSA, transcriptomics, metabolomics, and biochemistry, we identified a β-glucosidase (BGLU) gene cluster contributing to coumarin biosynthesis. MaBGLU1 function was verified by overexpression in M. albus, heterologous expression in Escherichia coli, and substrate feeding, revealing its role in scopoletin (coumarin derivative) production and showing that nonsynonymous variation drives BGLU enzyme activity divergence in Melilotus. Our work will accelerate the understanding of biologically active coumarins and their biosynthetic pathways, and contribute to genomics-enabled Melilotus breeding.

Keywords: BGLU; coumarin biosynthesis; genome evolution; population genetics; sweet clover.

Figure 1

Phylogenetic relationships, comparative genomics, and evolutionary analyses. (a) Characteristics of the Melilotus albus genome. 1. Gene density. 2. Repeat sequences density. 3. LTR density. 4. Copia (red line) and Gypsy (blue line) density. 5. SNP (green line) and InDel (pink line) density. 6. Syntenic block. (b) Distributions of M. albus, Medicago truncatula, Cicer arietinum, Glycine max, and Vitis vinifera Ks values. (c) A phylogenetic tree of 13 plant species and comparison of gene families. The dark numerical value beside each node indicates the estimated divergence time of each node (Mya, million years ago), and the red and blue numerical values denote the numbers of expanded and contracted gene families, respectively. The stacked‐column plot represents the numbers of single‐copy, multicopy, unique, other, and unclustered genes in the 13 plant species.

Figure 2

Comparative analysis of repeat sequences in four legume species. (a) Phylogenetic analysis of Copia and Gypsy elements in Melilotus albus (red), Cicer arietinum (blue), Glycine max (green), Medicago truncatula (yellow), and Trifolium pratense (pink) using conserved protein domains. (b) Analysis of intact LTR numbers and insertion times in M. albus. (c) Distribution of sequence divergence among four types of TEs from M. albus, M. truncatula, C. arietinum, and G. max. The right y‐axis represents the genome percentage of M. albus TEs. (d) Comparison of intact LTR length among M. albus, M. truncatula, C. arietinum, and G. max. (e) Structure of LTRs and the corresponding genes (MaTLP and MaSTKc). The grey box represents a 5‐bp target site duplication, and red triangles represent dinucleotide palindromic motifs. (f) RNA abundance (FPKM) and relative expression levels of MaTLP and MaSTK in roots under drought stress. The patterns of FPKM values and relative expression levels under drought stress were similar. D‐0 h, D‐3 h, D‐24 h: roots at 0, 3, and 24 h of drought treatment, respectively.

Figure 3

Phylogeny and population structure of sweet clover in different categories. (a) NJ phylogenetic tree of 94 sweet clover accessions inferred from whole‐genome SNPs with 1000 nonparametric bootstrap replicates. Accessions of Melilotus albus, Melilotus officinalis, and 16 other species are indicated by blue, orange, and red letters, respectively. (b) Population structure of 94 Melilotus accessions. The Melilotus accessions were divided into two (K = 2) or three (K = 3) groups. (c) PCA score plot of the first two components for the 94 accessions. The colours of the symbols are coded the same as in (a). (d) Four‐taxon ABBA/BABA test of introgression based on D statistics. The upper plot shows the phylogenetic relationships among the four groups, and the lower plot shows the genealogies of the ABBA and BABA patterns. A and B denote derived alleles; populations P2 (M. officinalis accessions) and P3 (M. albus accessions) sharing derived alleles showed the ABBA pattern; and P1 (Other accessions) and P3 sharing derived alleles showed the BABA pattern. (e) Population splits and migrations among sweet clover accessions. G1–G8 represent eight groups. G2 and G3 mainly included M. albus accessions; G6, G7, and G8 mainly included M. officinalis accessions; and G1, G4, and G5 mainly included other species accessions. (f) Demographic histories of M. officinalis and M. albus. Estimates of the effective population size over time are shown for the M. officinalis and M. albus populations. (g) Genome‐wide linkage disequilibrium (LD) decay in the M. officinalis and M. albus groups

Figure 4

Selective sweep analysis of Melilotus traits in the natural population. (a) Manhattan plot of the Fst‐based detection of selective sweeps identified from the comparison of Melilotus albus and Melilotus officinalis accessions. Functionally characterized candidate genes associated with flower colour and development are highlighted. (b) Six tandem duplicated MaAGL80s gene structure and location on chromosome 7. (c) Manhattan plot of the π‐based detection of selective sweeps identified from the comparison of accessions with high and low coumarin contents. Functionally characterized candidate genes associated with coumarin biosynthesis are highlighted.

Figure 5

Coumarin biosynthesis in Melilotus albus. (a) Mass spectrum of coumarins from the leaves. Mass spectrum of coumaric acid glucoside (C₁₅H₁₈O₈) and its glucoside (C₆H₁₂O₅) (left), detected in negative ion mode; and coumarin (C₉H₆O₂) (right) detected in positive ion mode. (b) The intensity of the peaks for coumaric acid glucoside, its glucoside, and coumarin in Ma46, Ma49, and qc (qc, a mixed sample of Ma46 and Ma49, as a control). (c) Gene families involved in the coumarin biosynthesis pathway. Abbreviations for the enzymes involved in each catalytic step are shown in bold. The numbers under the enzyme abbreviations are (from left to right) the gene numbers per gene family in M. albus (red), Medicago truncatula, Glycine max and Lotus japonicus. PAL: phenylalanine ammonia‐lyase, BGLU: β‐glucosidase, UGT: UDP‐glycosyltransferase, C4H: cinnamic acid 4‐hydroxylase, 4CL: 4‐coumarate, HCT: hydroxy cinnamoyl transferase, C2’H: ρ‐coumaroyl CoA 2’‐hydroxylase, COSY: coumarin synthase, C3H: 4‐coumarate‐3‐hydroxylase, COMT: caffeic/5’‐hydroxyferulic acid O‐methyltransferase, CCoAOMT: caffeoyl CoA O‐methyltransferase, F6’H: feruloyl‐CoA 6’‐hydroxylase, S8H: scopoletin 8‐hydroxylase. (d) Coexpression network of coumarin biosynthesis pathway genes. (e) Manhattan plot of the two NILs based on BSA. The annotated genes were related to coumarin biosynthesis and identified with SNPs and InDels between NILs Ma46 and Ma49.

Figure 6

Evolutionary and functional analysis of the MaBGLU gene family. (a) Molecular structures and chromosomal locations of six MaBGLUs. (b) MaBGLU1 alleles with the variation sites identified through BSA. The bases showing variation are shown on the line. (c) The protein sequences of MaBGLU1 NIL‐Ma46 and NIL‐Ma49. A 2‐base insertion in MaBGLU1 of NIL‐Ma46 results in a frameshift and premature termination, as indicated by the *. (d) Phylogenetic analysis of the BGLU family in ten legume species and Arabidopsis. Upper right: phylogenetic analysis of the BGLU subfamily in Melilotus albus, Medicago truncatula, and Arabidopsis. (e) Relative expression of six MaBGLUs. (f) LC‐MS/MS analysis of scopolin and scopoletin in the presence of recombinant MaBGLU1‐Ma49 and the empty vector control. Recombinant MaBGLU1‐Ma49 was incubated with scopolin (bottom), and the empty vector control was incubated with scopolin and scopoletin (top). Product formation was analysed by LC‐MS/MS. (g) The relative expression level of MaBGLU1 in NIL‐Ma46 overexpressing MaBGLU1‐Ma49. (h) The relative expression levels of MaBGLU1 in NIL‐Ma49 overexpressing MaBGLU1‐Ma49.