Analysis of Genetic Diversity in the Traditional Chinese Medicine Plant 'Kushen' (Sophora flavescens Ait.)

Abstract

Kushen root, from the woody legume Sophora flavescens, is a traditional Chinese medicine that is a key ingredient in several promising cancer treatments. This activity is attributed in part to two quinolizidine alkaloids (QAs), oxymatrine and matrine, that have a variety of therapeutic activities in vitro. Genetic selection is needed to adapt S. flavescens for cultivation and to improve productivity and product quality. Genetic diversity of S. flavescens was investigated using genotyping-by-sequencing (GBS) on 85 plants grown from seeds collected from 9 provinces of China. DArTSeq provided over 10,000 single nucleotide polymorphism (SNP) markers, 1636 of which were used in phylogenetic analysis to reveal clear regional differences for S. flavescens. One accession from each region was selected for PCR-sequencing to identify gene-specific SNPs in the first two QA pathway genes, lysine decarboxylase (LDC) and copper amine oxidase (CAO). To obtain SfCAO sequence for primer design we used a targeted transcript capture and assembly strategy using publicly available RNA sequencing data. Partial gene sequence analysis of SfCAO revealed two recently duplicated genes, SfCAO1 and SfCAO2, in contrast to the single gene found in the QA-producing legume Lupinus angustifolius. We demonstrate high efficiency converting SNPs to Kompetitive Allele Specific PCR (KASP) markers developing 27 new KASP markers, 17 from DArTSeq data, 7 for SfLDC, and 3 for SfCAO1. To complement this genetic diversity analysis a field trial site has been established in South Australia, providing access to diverse S. flavescens material for morphological, transcriptomic, and QA metabolite analysis. Analysis of dissected flower buds revealed that anthesis occurs before buds fully open suggesting a potential for S. flavescens to be an inbreeding species, however this is not supported by the relatively high level of heterozygosity observed. Two plants from the field trial site were analysed by quantitative real-time PCR and levels of matrine and oxymatrine were assessed in a variety of tissues. We are now in a strong position to select diverse plants for crosses to accelerate the process of genetic selection needed to adapt kushen to cultivation and improve productivity and product quality.

Keywords: DArT sequencing; copper amine oxidase; gene duplication; genetic diversity; lysine decarboxylase; quinolizidine alkaloids; single nucleotide polymorphisms; traditional Chinese medicine.

FIGURE 1

Proposed quinolizidine alkaloid biosynthesis pathway. The first two steps in the pathway have been confirmed. The gene for the first step, the conversion of L-lysine to cadaverine by a bifunctional lysine/ornithine decarboxylase (L/ODC), has been cloned from a variety of species including S. flavescens (Bunsupa et al., 2012). The conversion of cadaverine to 5-aminopentanal, requires a copper amino oxidase (CAO) and this intermediate is assumed to be spontaneously cyclised to Δ¹-piperideine. The CAO gene involved in QA biosynthesis has been cloned from L. angustifolius (Yang et al., 2017) but not from S. flavescens. Δ¹-piperideine is the precursor for different QAs in S. flavescens [including (+)-matrine and (+)-oxymatrine) and L. angustifolius (including (–)-lupinine and (+)-epilupinine] (Ohmiya et al., 1995; Sato et al., 2018). Dashed arrows indicate multiple reaction steps.

FIGURE 2

(A) Collection sites for S. flavescens seeds spanning the natural range of the species and representing nine different regions in China. A commercial seed sample was obtained from Internet retailer Alibaba. Regions 4–8 coincide with the silk road (Wang et al., 2019). (B) Phylogenetic analysis of SNP data generated using DArTSeq using a Bayesian inference method in MrBayes v3.2.6. A total of 89 samples were analysed, 85 samples grown from seeds collected from 9 provinces of China (10 accessions from most regions, except region 4, 6, and 8, with 8, 9, and 8 accessions, respectively), and 4 samples from commercial seed. Accessions from each population are labelled in a different colour, and are indicated by region number_sample number (e.g., 1_12). Samples in bold and indicated by an asterisk were used to sequence SfLDC and SfCAO gene fragments. A total of 1636 SNPs was used (those with less than 5% missing data and minor allele frequencies greater than 5%). Values reported on the nodes are Bayesian posterior probabilities and are 1.0 unless otherwise indicated. Scale bar represents the probability of nucleotide substitutions per site. Map image: https://upload.wikimedia.org/wikipedia/commons/d/d2/China_administrative_claimed_included.svg.

FIGURE 3

Anther dehiscence occurs before flowers fully open in S. flavescens. (A) Four discernible stages of flower development (1–4) prior to fully opened flowers. Longitudinal sections of representative buds at each stage (B–E) shows that by stage 4 (E) most of the pollen has dehisced, the keel petal has “opened,” but the other four petals are still together at the apex of the bud. Inset panel (E) shows a close up of the style, stigma, and one anther with released pollen. S. flavescens is in subfamily Papilionoideae and has a typical papilionoid-type flower with a forward facing petal, called a keel petal, and four other petals (Tucker, 2003). Scale bars, 1 mm.

FIGURE 4

Observed frequency of heterozygotes and minor allele frequency for each of 1636 SNPs detected by DArTseq genotyping by sequencing of 85 Sophora flavescens plants from nine regions in China. Solid black symbols represent markers for which observed genotype frequencies deviated significantly (p < 0.05) from Hardy–Weinberg equilibrium. Open black symbols represent markers for which observed genotype frequencies did not deviate significantly from Hardy–Weinberg equilibrium. The blue curve represents the expected frequency of heterozygotes given the minor allele frequency.

FIGURE 5

Sequence and analysis of deduced SfCAO protein sequence derived from the targeted capture and assembly of S. flavescens short read sequences matching LaCAO. (A) SfCAO was aligned with LaCAO (XP _019417507.1) and Cicar_XP_004501882.1 using Muscle (https://www.ebi.ac.uk/Tools/msa/muscle/) and the resulting alignment formatted in Geneious. The box indicates the conserved NYE/X motif and the three histidines (*) indicate the residues that interact with the catalytic copper ion (Yang et al., 2017). The horizontal line at the C-terminus indicates a predicted peroxisomal targeting signal peptide (Reumann et al., 2012). (B) Maximum likelihood tree (MEGA) of CAO genes from S. flavescens (Sf, green), L. angustifolius (La, blue), C. arietinum (Cicar, chickpea, brown), Glycine max (Glyma, black), and Arabidopsis thaliana (At, orange). Species known to produce QAs (Sf and La) are in bold. Nodes with ≥60% support (1000 bootstrap replicates) are indicated. The three sequences that are labelled LaCAO protein represent the CAO protein involved in QA biosynthesis from three different sources: ATY35659 [protein encoded by the cDNA sequence reported in Yang et al. (2017)], XP_019417507.1 (NCBI), and the protein predicted from the gene, Lup000530 as used to predict intron positions (legumeinfo.org, with some manual reannotation, see Supplementary Figure 8). One of 10 Arabidopsis thaliana CAO proteins (NP_181777.2 (= AtCuAO3), encoded by At2g42490) was selected based on data showing that the protein groups with other legume CAOs and the gene encoding this protein is the only Arabidopsis CAO gene that has 11 introns as found in lupin (Tavladoraki et al., 2016). The scale represents the number of amino acid substitutions per site.

FIGURE 6

Selected samples from regions 4 and 9 have consistent but different changes in SfCAO1. (A) Amplification of 3′ SfCAO gene fragments from nine individuals from Anshun Province (region 9) reveals diversity in SfCAO1 for the 181 bp deletion from intron 10 to intron 11 (bases 271–451, including exon 11). This deletion was originally identified by Sanger sequencing of sample 9_8 (*, blue arrow). (B) Amplification of 3′ SfCAO from six individuals from Hebei Province (region 4) reveals diversity for presence/absence of SfCAO1 with primers SfCAO_5p_Fd and SfCAO_e1_R. Sample 4_15 (*, pink arrow) was the sample selected for sequencing and thus only SfCAO2 could be sequenced from this sample. (C) Summary of presence or absence of: in samples from region 9, the deletion variant allele (containing a 181 bp deletion that codes for exon 11 and flanking intron sequence); or in samples from region 4, the SfCAO1 PCR product, based on the position of samples in the relevant clades (D) of the phylogenetic tree produced from 2600 DArTSeq SNPs. (E) A scaled down copy of the entire tree from Figure 2B (to provide context).

FIGURE 7

Expression of the first two genes in the quinolizidine alkaloid pathway as determined by quantitative real time PCR (qPCR) as described by Burton et al. (2011). (A) Lysine decarboxylase (SfLDC) and (B) cadaverine oxidase (SfCAO) normalised transcript levels (arbitrary units). Error indicates standard deviation of three replicate tissues samples from the same plant.

FIGURE 8

Quantification of oxymatrine. Oxymatrine (%) was determined by HPLC based on peak area compared to a standard curve produced by a commercial standard (Sigma). Oxymatrine was detected in root samples (both large and small root) (Supplementary Figure 1), immature seeds (Supplementary Figure 3), and the two control samples (kushen root from a TCM store, and dry commercial seed). Error bars indicate standard deviation of three replicate tissues samples from the same plant (except for sample AC1, small root, n = 2). Example chromatograms (concentrated samples) are shown in Supplementary Figure 12 and show that low levels of matrine were detected in the immature seeds (LC4), and the two control samples.