Population genomic evidence for radiative divergence of four Orychophragmus (Brassicaceae) species in eastern Asia

(1)

Botanical Journal of the Linnean Society, 2019, 191, 18–29. With 3 figures.

Population genomic evidence for radiative divergence of four Orychophragmus (Brassicaceae) species in eastern Asia

LINLING ZHONG

^†

, HUANHUAN LIU

^†

, DAFU RU, HUAN HU and QUANJUN HU*

^, Key Laboratory for Bio-Resource and Eco-Environment of Ministry and Education, College of Life Sciences, Sichuan University, Chengdu 610065, PR China

Received 29 January 2019; revised 18 March 2019; accepted for publication 9 June 2019

Radiation rather than bifurcating divergence has been inferred through a number of phylogenetic analyses using different DNA fragments. However, such inferences have rarely been tested by examining alternative hypotheses based on population genomic data. In this study, we sequenced the transcriptomes of 32 individuals from 13 populations of four Orychophragmus spp. (Brassicaceae) to investigate their divergence history. Cluster and population structure analyses recovered four distinct genetic clusters without any genetic mixture. Most orthologous genes produced unresolved bifurcating interspecific relationships with a star phylogeny. The resolved gene trees were highly inconsistent with each another in reconstructing interspecific relationships. Population genomic analyses suggested unexpectedly high genetic divergence and a lack of gene flow between the four species. We examined radiation vs. bifurcating divergence between these four species based on coalescent modelling tests of population genomic data. Our statistical tests supported a radiation of these species from a common ancestor at almost the same time, rejecting stepwise bifurcating interspecific divergence with time. This nearly simultaneous radiation was dated to the Quaternary, during which climate changes are suggested to have promoted species diversity in eastern Asia.

Our results highlight the importance of population genomic data and statistical tests in deciphering interspecific relationships and tracing the divergence histories of closely related species.

ADDITIONAL KEYWORDS: non-bifurcating divergence – population genomic data – radiation – transcriptome.

INTRODUCTION

Stepwise bifurcating divergence over time is assumed to be the basic rule that governs the tree of life and it is therefore widely used in phylogenetic analyses (Fitch & Margoliash, 1967). Under this hypothesis, one common ancestor diverges into two descendant clades, each of which further bifurcates into descendant subclades over time (Woese, 2000; Yang & Rannala, 2012). Gene trees are generally consistent with such bifurcating species trees, with each bifurcating clade or subclade receiving a high level of statistical support in phylogenetic analyses of sequence variations in different genes (Fitch & Margoliash, 1967; Pamilo &

Nei, 1988). However, numerous phylogenetic studies have shown low support values for specific clades or subclades with multiple terminal species units, and

highly inconsistent gene trees between species or subclades have sometimes been obtained based on different DNA squences (Maddison, 1997; Nichols, 2001; Richardson et al., 2001; Degnan & Rosenberg, 2006). This has usually been explained by two, not mutually exclusive, factors. First, these phylogenetic complexities may have arisen from hybridization and gene flow during/after speciation. Introgression events result in inconsistent gene trees and low support for bifurcating clades or subclades (Novikova et al., 2016;

Feng et al., 2019). Second, no bifurcating relationship may apply to multiple (≥3) descendant lineages, for example when they split simultaneously from a common ancestor that has a widespread distribution (Simões et al., 2016). Geographical isolations trigged by rapid geological or climatic change may have played important roles in such radiative divergences (Liu et al., 2006; Glor, 2010; Simões et al., 2016; Sun et al., 2018). These radiative lineages may have accumulated their own mutations in the course of genetic divergence Keywords=Keywords=Keywords_First=Keywords

HeadA=HeadB=HeadA=HeadB/HeadA HeadB=HeadC=HeadB=HeadC/HeadB HeadC=HeadD=HeadC=HeadD/HeadC Extract3=HeadA=Extract1=HeadA

REV_HeadA=REV_HeadB=REV_HeadA=REV_HeadB/HeadA REV_HeadB=REV_HeadC=REV_HeadB=REV_HeadC/HeadB REV_HeadC=REV_HeadD=REV_HeadC=REV_HeadD/HeadC REV_Extract3=REV_HeadA=REV_Extract1=REV_HeadA BOR_HeadA=BOR_HeadB=BOR_HeadA=BOR_HeadB/HeadA BOR_HeadB=BOR_HeadC=BOR_HeadB=BOR_HeadC/HeadB BOR_HeadC=BOR_HeadD=BOR_HeadC=BOR_HeadD/HeadC BOR_Extract3=BOR_HeadA=BOR_Extract1=BOR_HeadA EDI_HeadA=EDI_HeadB=EDI_HeadA=EDI_HeadB/HeadA EDI_HeadB=EDI_HeadC=EDI_HeadB=EDI_HeadC/HeadB EDI_HeadC=EDI_HeadD=EDI_HeadC=EDI_HeadD/HeadC EDI_Extract3=EDI_HeadA=EDI_Extract1=EDI_HeadA

CORI_HeadA=CORI_HeadB=CORI_HeadA=CORI_HeadB/HeadA CORI_HeadB=CORI_HeadC=CORI_HeadB=CORI_HeadC/HeadB CORI_HeadC=CORI_HeadD=CORI_HeadC=CORI_HeadD/HeadC CORI_Extract3=CORI_HeadA=CORI_Extract1=CORI_HeadA ERR_HeadA=ERR_HeadB=ERR_HeadA=ERR_HeadB/HeadA ERR_HeadB=ERR_HeadC=ERR_HeadB=ERR_HeadC/HeadB ERR_HeadC=ERR_HeadD=ERR_HeadC=ERR_HeadD/HeadC ERR_Extract3=ERR_HeadA=ERR_Extract1=ERR_HeadA

INRE_HeadA=INRE_HeadB=INRE_HeadA=INRE_HeadB/HeadA INRE_HeadB=INRE_HeadC=INRE_HeadB=INRE_HeadC/HeadB INRE_HeadC=INRE_HeadD=INRE_HeadC=INRE_HeadD/HeadC INRE_Extract3=INRE_HeadA=INRE_Extract1=INRE_HeadA App_Head=App_HeadA=App_Head=App_HeadA/App_Head BList1=SubBList1=BList1=SubBList

BList1=SubBList3=BList1=SubBList2

SubBList1=SubSubBList3=SubBList1=SubSubBList2 SubSubBList3=SubBList=SubSubBList=SubBList SubSubBList2=SubBList=SubSubBList=SubBList SubBList2=BList=SubBList=BList

*Corresponding author. E-mail: huquanjun@gmail.com

†These authors contributed equally to this work

Downloaded from https://academic.oup.com/botlinnean/article/191/1/18/5549094 by guest on 03 November 2022

(2)

but lack the common mutations that are necessary to demonstrate phylogenetically bifurcating clusters, and their morphological traits and ecological adaptations will have diversified independently (Osborn, 1902; Glor et al., 2010; Givnish et al., 2014). In addition, numerous ancestral polymorphisms may have been retained across radiative lineages. However, it is difficult to distinguish examples of hybridizing reticulate evolution from true radiative divergence by means of phylogenetic analyses, although numerous species- rich groups, especially some occurring in regions that have experienced recent extensive geological or climatic changes, have been claimed to have resulted from radiative diversification (Richardson et al., 2001;

Liu et al., 2006, 2012, 2014; Rundell & Price, 2009;

Givnish, 2015; Chaves et al., 2016).

Modelling tests of alternative hypotheses combined with phylogenetic description based on population genomic data provide a robust alternative approach to tracing in depth the history of species divergence and examining gene flow during and after speciation (Ru et al., 2018; Sun et al., 2018; Wang et al., 2019).

In this study, we examine the histories of divergence among four Orychophragmus spp. (Brassicaceae):

O. violaceus (L.) O.E.Schulz, O. diffusus Z.M.Tan &

J.M.Xu, O. longisiliquus Huan Hu, J.Quan Liu &

Al-Shehbaz and O. zhongtiaoshanus Huan Hu, J.Quan Liu & Al-Shehbaz, based on population genomic data and modelling tests. All of these species occur in wet forest habitats or at forest edges, mainly in China;

only O. violaceus extends into Korea (Supporting Information, Table S1). (Hu et al., 2018). Two of them have been classified as separate from O. violaceus because they represent independently evolving lineages that exhibit morphological distinctions (Fig.

1) and reproductive isolation (Hu et al., 2018). All four species are diploid with 2n = 22 (O. diffusus) or 24 (the other three) (Zhou et al., 2001; Hu et al., 2018). However, interspecific relationships among them have been found to be highly inconsistent in previous studies and these relationships therefore remain unresolved (Hu et al., 2016). Phylogenetic analyses of nuclear ITS sequence variations or plastomes have suggested inconsistent relationships between these four species (Hu et al., 2015, 2016).

Two, not mutually exclusive, factors have been used to explain such phylogenetic inconsistencies, namely introgressive hybridization after stepwise bifurcating divergence and incomplete lineage sorting of ancestral polymorphisms because of radiative divergence (Hu et al., 2016). It remains necessary to discern which of these factors makes a greater contribution to these evolutionary complexities, using an approach based on additional genomic evidence and modelling tests.

The Chinese mainland was extensively affected by significant Quaternary climate fluctuations, and

palaeovegetation studies have suggested that dry, cold climates occurred there repeatedly and that forest and steppe vegetation alternated (Yu et al., 2000; Harrison et al., 2001). These climate changes not only resulted in range retreats and recolonizations by numerous species (Liu et al., 2012), but also promoted rapid speciation through geographical isolation (Qian & Ricklefs, 2000;

Qian et al., 2003). It is also likely that these Quaternary climate fluctuations may have triggered interspecific introgression among the four Orychophragmus spp.

or radiation as a result of geographical isolation.

To clarify these unresolved issues, we sequenced transcriptomes of 32 individuals from 13 populations of the four species. We aimed to address the following questions based on population genomic data. (1) Are the four species well differentiated, as suggested by reproductive isolation and morphological differences?

(2) What are their interspecific relationships? How did they diverge and is their genetic differentiation consistent with a bifurcating or radiative pattern?

Did gene flow occur during and/or after speciation? (3) Can their divergences be correlated with Quaternary climatic changes?

MATERIAL AND METHODS

RNA extRActioNANdtRANscRiptomesequeNciNg To obtain genes expressed under the same environmental conditions, we used materials grown together in the same common garden. First, seeds from different populations of four species in the field were collected. Based on previous studies, the distributions of the four species are allopatric (Hu et al., 2015, 2018).

To represent the distributions of each of the four species, we selected at least three populations from each species for the present study. In total, 32 individuals were collected from 13 populations (Supporting Information, Table S1, Fig. 1). For each individual, seeds were germinated and the subsequent seedlings were used to extract RNA. Voucher specimens were deposited in the Sichuan University Herbarium (Hu et al., 2018). One plant of Sinalliaria limprichtiana (Pax) X.F.Jin, Y.Y.Zhou & H.W.Zhang from the sister genus (Hu et al., 2015, 2016; Zhang et al., 2018) was also sampled as an outgroup. Sequencing was conducted on an Illumina HiSeq platform, and 2× 150-bp paired-end reads were generated.

RefeReNcetRANscRiptomeAssemblyANd filteRiNg

The quality of raw reads was assessed using FastQC v.0.11.3 (http://www.bioinformatics.babraham.

ac.uk/projects/fastqc/) and reads were filtered using Trimmomatic v.0.36 (Bolger, Lohse & Usadel, 2014)

(3)

based on the following parameters: SLIDINGWINDOW:

4:15; LEADING: 3; TRAILING: 3; MINLENGTH:

36; and adaptor trimming. Using Trinity v.2.1.1 with default parameters, one O. zhongtiaoshanus individual was de novo assembled as a reference (Haas et al., 2013). Redundant sequences were filtered from

this transcriptome assembly by using CD-HIT v.4.6 with a cutoff value of 0.9 (Fu et al., 2012). From all the sequences with similarity >90%, the longest one was retained. We used TransDecoder v.3.0.0 (http://

transdecoder.github.io) to predict the coding sequence (CDS) of each transcript and used RSEM (Li &

Figure 1. Locations of the 32 Orychophragmus individuals sampled and an outgroup sample. The species are indicated in yellow (O. violaceus), blue (O. zhongtiaoshanus), purple (O. diffusus), green (O. longisiliquus) and black (Sinalliaria limprichtiana).

(4)

Dewey, 2011) to justify the use of FPKM (fragments per kilobase of transcript per million mapped reads) for further quality filtering based on a cutoff value of 0.5. Finally, possible contaminating fungal transcripts were removed by using blastn searches against a custom database of 18S rRNA sequences (Supporting Information, Table S2), because fungal infections are common in Orychophragmus plants (Delhomme et al., 2015).

sequeNceAligNmeNtANd sNp cAlliNg SAM files were generated by mapping reads from the other individuals to the O. zhongtiaoshanus reference transcriptome using the bwa-mem algorithm (maximal exact matches) from BWA v.0.7.10 (Li

& Durbin, 2009). The SAM files were converted to BAM files using SAMTOOLS v.1.2 (Li et al., 2009).

We then removed duplicates using PICARD TOOLS v.1.92 (https://github.com/broadinstitute/picard) and realigned insertion/deletion (INDEL) regions to give a minimum error rate using Genome Analysis Toolkit v.3.4-46 (McKenna et al., 2010). We called single mucleotode polymorphisms (SNPs) in all samples using SAMTOOLS. We calculated genotype likelihoods from reads and estimated the allele frequencies at each location for each sample. Using the ‘MPILEUP’

command with the parameters ‘-q 30 -Q 30 -t DP -t DP4 -t SP’ to identify SNPs, SNP calling errors were corrected. To reduce the false discovery rate, we excluded sites with depth (DP) < 10 per sample and indels that were within five bases of a splice site.

Genotypes were treated as being missing when Phred- scaled genotype likelihoods below 30. Finally, sites with a missing rate <50% of the data were retained for further analysis (Ru et al., 2016).

iNteRspecificdiffeReNtiAtioNANdphylogeNetic RelAtioNships

To cluster all sampled individuals, principal component analysis (PCA) was performed using the package EIGENSOFT v.5.0.1 (Patterson, Price &

Reich, 2006) after pruning the dataset to reduce the linkage disequilibrium effect. Eigenvectors were calculated from the covariance matrix using the R function ‘reigen’. Tracy–Widom tests were performed, and two eigenvectors together were used to distinguish different clusters. The maximum-likelihood (ML) STRUCTURE approach was implemented by the program ADMIXTURE v.1.23 to place each individual in a genetic cluster (Alexander & Lange, 2011). The input data file was converted using VCFTOOLS v.0.1.14 and PLINK v.1.07 (Purcell et al., 2007; Danecek et al., 2011). ADMIXTURE with cross-validation was

employed for values of genetic clusters K from 1 to 7 to explore the convergence of individuals, using default methods and settings. Based on the genomic data, we further compared the genetic diversity of each species and examined genetic differentiation (F_ST) among species and populations within species (Hudson, Boos

& Kaplan, 1992). Tajima’s D (Tajima, 1989) and π (nucleotide diversity) (Nei & Li, 1979) were estimated for each locus in each species using VCFtools (Pavlidis, Laurent & Stephan, 2010; Danecek et al., 2011).

Sinalliaria limprichtiana was used as an outgroup to reconstruct genetic relationships between the four Orychophragmus spp. based on genome-scale data of all the samples. We performed neighbour-joining (NJ) and ML analyses with the population-scale SNP data.

These analyses were performed by TREEBEST v.1.9.2 (http://treesoft.sourceforge.net/treebest.shtml) (Vilella et al., 2009), or RAXML v.8.2.9 under the GTRGAMMA model for heuristic tree searches (Stamatakis, 2014).

Then using Trinity, four transcriptomes were de novo assembled from each species and the outgroup, which were the best-quality raw reads (Supporting Information, Table S3). Transcriptomes were filtered with CD-HIT and TransDecoder. We defined gene sequences that were orthologous across the four species and the outgroup based on protein sequences (derived from CDS) by CD-HIT, using orthoMCL v.2.0.9. We filtered out CDS with length <300 bp (Li, Stoeckert

& Roos, 2003) and obtained 1038 gene sequences orthologous between the five species. ML trees were reconstructed for each orthologous gene using RAXML.

Gene trees with bootstrap value <60% at any branch were excluded (Qiu, Yoon & Bhattacharya, 2013).

Highly supported gene trees were used to reconstruct the ‘species tree’ via the program MP-EST v.1.5 (Liu &

Yu, 2010; Liu, Xi & Davis, 2015). We used DensiTree v.2.01 to combine all gene trees into one consensus tree (Bouckaert, 2010).

geNeflowANdtestofdiveRgeNcehypotheses Genotypes shared between the four species were examined based on identity-by-descent (IBD) block analyses using BEAGLE v.4.1 (Browning &

Browning, 2013). We set the following parameters:

‘window = 1000; overlap = 100; ibdtrim = 10; ibdlod = 7’.

Only a limited number of IBD blocks were detected, and there was an almost complete lack of such blocks shared between O. zhongtiaoshanus and the other three species. Therefore, ABBA–BABA analyses was used to examine potential gene flow between the four species with a script based on four-fold degenerate sites extracted from all individuals of each species (Excoffier et al., 2013; Martin, Davey & Jiggins, 2015).

We randomly assumed the bifurcating relationships

(5)

of the four species and examined gene flow from two focused species. D-statistics and corresponding test values (Z-scores) were calculated using all individuals of each species and S. limprichtiana as an outgroup (Supporting Information, Table S1). Z-scores were calculated through block-jackknifing (Svardal et al., 2017). When Z-scores had an absolute value >3, gene flow was considered significant (Excoffier et al., 2013;

Martin et al., 2015). These previous studies suggested that such a threshold could exclude incomplete lineage sorting to alternatively explain shared alleles.

Finally, we used fastsimcoal2 v.2.5.2.21 to test the possible models of species divergence and gene flow that were indicated by ABBA–BABA analyses (Excoffier & Foll, 2011). Our modelling tests were performed based on the neutral joint site frequency spectrum data from four-fold synonymous sites in all individuals. Thirty-three alternative models of species divergence were designed based on hypotheses of radiative divergence, different orders of bifurcation, divergence timescale, gene flow and species population expansion. The times of divergence among the four species in each model were automatically estimated based on joint site frequency and mutation rates. All four of the species are usually referred to as ‘biennial’, but their generation time is only one year. They usually germinate between September and December and start to flower and set fruit between February and May.

We therefore assumed a mutation rate of 7 × 10^–9 per base pair per generation and a generation time of one year for these four species, as was used for Arabidopsis Heynh., a member of the same family (Douglas et al., 2015). Gene flow between each pair of species was assumed to be either absent or bi-directional. Rates of gene flow were estimated based on joint site frequency.

Akaike’s information criterion (AIC) was used to select the best-fitting model (Burnham & Anderson, 2002).

RESULTS

stRuctuReANAlysis, geNeticdiffeReNtiAtioN ANdgeNeticdiveRsity

A de novo transcriptome was assembled from O. zhongtiaoshanus as reference, which is 26 270 000 bp long with 29 678 transcripts, having an average length of 885 bp and N50 of 1152 bp (Supporting Information, Table S2), after quality filtering and sequence trimming. To obtain SNPs, reads from the four species and outgroup were mapped to the reference transcriptome. After quality control, for each sample, 65.64% of reads mapped to 95.00% of the reference genome assembly with an average depth of 117.81-fold (Table S4). In total, 24 500 000 SNPs were identified from all individuals, and 3 706 000 of

these SNPs were retained after filtering and quality control (Fig. S1).

We used PCAs of all sampled individuals of the four Orychophragmus spp. using the pruned SNPs.

The first and second eigenvectors distinguished four clear groups, corresponding to the morphological classifications of the species (Fig. 2A). Consistent with these PCAs, Structure ADMIXTURE analyses similarly revealed that when the number of clusters (K) was 4 (with the lowest CV error = 0.4308 as the best model), the sampled individuals clustered into four distinct groups (Fig. 2B). When K was set to 3 (CV error = 0.4320), the sampled individuals of O. diffusus and O. longisiliquus clustered into a single group, and those of O. zhongtiaoshanus and O. violaceus were retained as two separate groups. However, when K = 2 (CV error = 0.4534), all individuals of O. diffusus and O. zhongtiaoshanus clustered into one group, and those of O. longisiliquus and O. violaceus comprised the other (Supporting Information, Table S5).

The SNPs from all samples were used to examine interspecific differentiation and determine indexes of genetic diversity for each species. We found that the observed F_ST values between pairs of species varied from 0.27 between O. violaceus and O. longisiliquus to 0.33 between O. zhongtiaoshanus and O. diffusus (Supporting Information, Table S6, Fig. S2). Genetic diversity expressed as π values for each species varied from 0.0047 (O. diffusus) to 0.0069 (O. zhongtiaoshanus) (Table S7). Tajima’s D was calculated from the four species with three positive values suggesting either a recent population bottleneck or some form of balancing selection (Table S7, Fig. S3).

iNteRspecificRelAtioNshipsbAsedoN phylogeNeticANAlyses

We used S. limprichtiana as the outgroup to group all individuals based on phylogenetic analyses of SNPs.

NJ and ML analyses produced a consistent topological relationship (Supporting Information, Fig. S4) in which all 32 individuals initially clustered into four well- delimited groups. Orychophragmus longisiliquus was the first to diverge, with O. zhongtiaoshanus being sister to the subclade comprising O. diffusus and O. violaceus.

We then assembled a de novo transcriptome for one individual of each of three Orychophragmus spp. and the outgroup. In total, 65 923, 57 240, 79 589 and 66 292 contigs were obtained with N50 values of 1158, 1155, 978 and 1080, respectively, from O. longisiliquus, O. diffusus, O. violaceus and S. limprichtiana (Table S3). Together with the O. zhongtiaoshanus reference transcriptome, orthologous genes across all of the five assembled species were identified by OrthoMCL. After deleting single-copy orthologues of lengths <300 bp, a

(6)

total of 1038 orthologous gene groups across the five species were identified. ML trees were constructed using RAXML for each orthologous gene group with S. limprichtiana as an outgroup. We found that c.

56% of the orthologous gene groups (581) produced a star-phylogeny of the four species without statistical support for bifurcating interspecific relationships. We obtained 457 gene trees with statistical support for each node at a bootstrap cutoff ≥60%. We detected 15 different interspecific relationships between the four

species among these gene trees, with the number of gene trees for each relationship varying from 14 to 79 (Fig. 2D). The interspecific relationship with the highest number of gene trees (79) indicated that O. diffusus and O. violaceus comprised a subclade, sister to the other which comprised O. longisiliquus and O. zhongtiaoshanus. All 457 gene trees were combined by DensiTree and the results clearly indicated reticulate relationships between the four species (Fig. 2C). Further, integrating all 457 gene Figure 2. Delimitation and phylogenetic relationships of the four Orychophragmus spp. A, principal component analysis (PCA) plots. B, bar plots indicative of assignment probabilities from ADMIXTURE analysis. When the number of clusters (K) is 4, four species can be clearly distinguished. C, overlapped maxiumum-likelihood (ML) trees for orthologous genes constructed using DensiTree. (D, orthologous gene relationships revealed by ML analysis. The number of gene groups is presented for each tree type.

(7)

trees using MP-EST, the resulting integrated tree had a topology that was consistent with those of the ML and NJ trees (Fig. S4).

shARedhAplotypes, geNeflowANdtestof AlteRNAtivediveRgeNceRelAtioNship BEAGLE was used to detect IBD haplotypes among the four species (Fig. 3A). Only a few haplotypes were shared between O. zhongtiaoshanus and O. longisiliquus or between O. zhongtiaoshanus and O. violaceus. The paucity

of IBD shared between pairs of species suggested a low rate of gene flow. We further examined gene flow among the four species of all possible bifurcating relationships, similarly using S. limprichtiana as an outgroup, by the ABBA–BABA approach (Fig. 3B, C). Z values were considered as a test of significance. An absolute value of Z > 3 is often used as a critical threshold for gene flow (Excoffier et al., 2013; Martin et al., 2015). Absolute Z values for most of our comparisons were <3 (Fig. 3C) with only two comparisons >3 but <5, indicating that gene flow among the four species was extremely low.

Figure 3. Shared haplotypes, D statistic (ABBA–BABA test) and a radiation demographic history. A, haplotypes shared among the four Orychophragmus spp. Heat map colours represent the total length of identity-by-descent (IBD) blocks for each pairwise comparison. B, the distribution D statistic values in different topologies. Topologies are shown in different permutations from O. violaceus (yellow), O. zhongtiaoshanus (blue), O. diffusus (purple), O. longisiliquus (green) and Sinalliaria limprichtiana (black). C, Z value of ABBA–BABA estimations used to determine the significance of the test. An absolute value of the Z score of >3 is often regarded as significant. D, demographic parameter estimates with confidence intervals for the best-fit model of speciation in Orychophragmus. Estimates of effective population sizes (N_e) for O. violaceus, O. zhongtiaoshanus, O. diffusus and O. longisiliquus are given as single values, and estimates of the timing of the origin of ancestral groups (T1) are given in years before present. Confidence intervals are given in parentheses.

(8)

Thirty-three alternative models of the interspecific relationships from the four species were designed (Fig. 3D; Supporting Information, Fig. S5, Table S8), which included all bifurcating relationships recovered previously from different gene datasets, radiation, and relationships with or without gene flow. We compared all models and found the best-fit one with the lowest AIC value to be radiation without gene flow (Fig. 3D).

Estimated parameters for this mode suggested that the four species diverged simultaneously from a common ancestor 33 000 (21 000–58 000) years ago. The current effective population sizes (N_e) of O. diffusus, O. violaceus, O. longisiliquus and O. zhongtiaoshanus were estimated to be 66 708, 121 756, 102 167 and 90 731, respectively. Under the expansion radiation scenario, the AIC value is better than for the remaining scenarios, except for model 2. We speculated that the ancestral population experienced expansion and then diverged into the different populations.

DISCUSSION

In the study presented here, we used genome-scale SNPs to examine interspecific differentiation, genetic diversity and divergence history of four closely related Orychophragmus spp. Our results suggested that the four species were well delimited based on both PCA and STRUCTURE analyses. In addition, we found a high level of interspecific genetic differentiation.

However, interspecific relationships among the four species were highly inconsistent across different genes.

Further coalescent tests based on population genomic data supported almost simultaneous radiation of the four species during the late Quaternary, rather than the stepwise bifurcating divergence that is generally assumed to give rise to speciation.

fouRwell-delimitedspecieswithAhighlevel ofgeNeticdiffeReNtiAtioN

Our analyses provided genetic evidence from the nuclear genome that the four Orychophragmus spp.

were lineages that had evolved independently of one another, thus supporting previous inferences based on nuclear ITS and plastome sequence variation (Hu et al., 2015, 2016). In addition, clear morphological distinctions and post-pollination reproductive isolation in terms of fruit set have been identified among the species (Hu et al., 2018). Our PCAs of all nuclear- genome SNPs recovered four groups (Fig. 2A). Using STRUCTURE applied to all nuclear SNPs based on Bayesian clustering analyses similarly distinguished the four species when K = 4, each group corresponding to a different species with high probability (Fig. 2B).

Note that genetic differentiation values between

any two of the four species, measured by F_ST values based on the transcriptome-derived SNPs, were high, between 0.27 and 0.33. These values are close to those (0.2–0.4) observed for closely related Arabidopsis spp. of the same family based on total genomic SNPs (Novikova et al., 2016). All of our analyses provided further support for the species status of each of the four lineages.

lAckofobviousgeNeflowANdRAdiAtioN Gene flow is common between closely related species, and it is apparent in the divergence histories of numerous plant groups (Arnold & Martin, 2009;

Abbott et al., 2013). For example, in comparisons between closely related Arabidopsis spp., numerous IBD haplotypes were shared, and all ABBA–BABA analyses suggested the occurrence of widespread interspecific gene flow (Novikova et al., 2016).

However, among the four Orychophragmus spp. in this study, only a few IBD haplotypes were detected (Fig. 3A); infrequent sharing of haplotypes between species may result from ancestral polymorphisms and incomplete lineage sorting (Durand et al., 2011).

Therefore, we used ABBA–BABA analyses and the outgroup S. limprichtiana to examine gene flow between the four species. We used a Z-score with an absolute value >3 to exclude incomplete lineage sorting (Durand et al., 2011). Most such tests, which have different underlying assumptions, indicate a lack of gene flow although some with Z-scores >3, but <5, suggest a low level of gene flow; this can be ignored compared with obvious gene flow. For example, five comparisons between vervet monkeys (Chlorocebus) were not highly significant with a Z-score <5 (Svardal et al., 2017). In addition, incomplete lineage sorting might be present due to rapid evolution.

The phylogenetic relationships among the four species obtained from different datasets in the present and previous studies are highly inconsistent. In addition to the conflicting relationships constructed based on ITS and plastome sequence variation (Hu et al., 2015, 2016), we found that such inconsistencies were widespread on the basis of different orthologous genes (Fig. 2D), which revealed extremely reticulate relationships (Fig. 2C). In addition, c. 56% of the orthologous gene groups identified here could not discern the bifurcating relationships of the four species, indicating that the star phylogeny possibly resulted from radiation (Richardson et al., 2001; Liu et al., 2006; Glor, 2010; Arakaki et al., 2011).

Alternative hypotheses for the divergence histories of the four species were tested. The hypothesized models included all probable bifurcating relationships of the four species and radiative divergences with or without gene flow (Fig. 3D; Supporting Information, Fig. S5). We found that radiation without gene flow gave a better fit

(9)

than any of the stepwise bifurcating divergence models or those radiative models with gene flow, having the lowest AIC value (Table S8). All of these tests support that the hypothesis that the four species experienced radiation from a common ancestor without obvious gene flow. The inconsistent phylogenetic relationships between the four species inferred from different molecular datasets probably resulted from radiation and incomplete lineage sorting of the ancestral polymorphisms (Suh, Smeds & Ellegren, 2015). Thus, our case study suggests that population genomic data combined with coalescent modelling tests provide a robust means of examining potential radiation that has previously been proposed on the basis of phylogenetic analyses of sequence variation at a limited number of DNA loci (e.g. Richardson et al., 2001; Liu et al., 2006).

On the basis of genome-scale SNP joint site frequency, we dated the radiation of the four species as occurring c. 33 kya in the late Pleistocene, when dramatic climate fluctuations probably resulted in the explosive species diversification of some groups in other parts of the world (e.g. Johnson et al., 1996). Similarly, during this period, in northern and eastern China the climate started to become colder, and cold-adapted vegetation replaced the former warm-climate trees and/or shrubs in most regions as these species retreated to separate refugia (Yu et al., 2000). The climate then shifted from the cold, arid Last Glacial Maximum to the warm Holocene interglacial (Harrison et al., 2001). Such climatic oscillations have been found to cause range retreat and subsequent recolonization by numerous plant species in China (Liu et al., 2012). It is therefore likely that a previously widely distributed ancestor of the four Orychophragmus spp.

became isolated in different refugia, initiating radiation.

They then evolved independently because of strong geographical isolation, although nucleotide diversity varies among each species (Supporting Information, Table S7) possibly due to habitat differences. Because of their short generation times, these four species have developed clear post-pollination reproductive isolation (Hu et al., 2018). Note that after the initial isolation and divergence, the population size of each species seems not to have expanded distinctly in response to post-glacial climate warming given that the radiation–expansion model did not receive a better fit (Fig. S5). This is also consistent with the fact that the current distributions and population sizes of the four species remain small (Hu et al., 2018).

These results together suggest that Orychophragmus may have experienced radiation due to geological or climatic changes, as suggested for other genera (Richardson et al., 2001; Liu et al., 2006; Rundell &

Price, 2009; Zhang et al., 2014; Givnish, 2015; Chaves et al., 2016), which led to non-bifurcating interspecififc relationships (Sun et al., 2018). However, this differs from the recently suggested non-bifurcating divergence

caused mainly by gene flow (Novikova et al., 2016; Feng et al., 2019). Although eastern Asia lacked extensive ice cover, the climate nevertheless oscillated greatly during the Quaternary (Hewitt, 2000), which probably led to fragmentation of the continuous distribution of the ancestral species into isolated populations and initialized allopatric radiation. In addition, such a radiation mediated by geographical isolation caused by climate change during the Pleistocene, and the near lack of gene flow, provides support for a previously proposed hypothesis for the high species diversity found in eastern Asia (Qian & Ricklefs, 2000; Qian et al., 2003).

CONCLUSIONS

Our analyses of population genomic data show that the four Orychophragmus spp. were well delimited as independently evolving lineages. Rather than having experienced stepwise bifurcation, as is commonly assumed to underlie the fundamental tree of life, radiation from a common ancestor produced the four species almost simultaneously with a near lack of gene flow. This radiation was dated to the late Pleistocene, triggered by geographical isolation in different refugia, thus providing a further example of allopatric speciation brought about by climate change during the Pleistocene in eastern Asia, as has also been hypothesized in other cases.

ACKNOWLEDGEMENTS

We thank Prof. Jianquan Liu for his help in preparing the paper. We are grateful to Qianlong Liang for help with preparing Figure 1. This work was supported by grants from the National Natural Science Foundation of China (41771055, 31700323), National Key Research and Development Program (2017YFC0505203), Science & Technology Basic Resources Investigation Program of China (2017FY100100) and the Fundamental Research Funds for the central Universities (YJ201714).

DATA ARCHIVING

The data have been uploaded to NCBI. SRA:

SRR6655828–SRR6655859, SRR6441722.

CONFLICT OF INTEREST

The authors declare that they have no conflicts of interest.

(10)

REFERENCES

Abbott R, Albach D, Ansell S, Arntzen JW, Baird SJ, Bierne N, Boughman J, Brelsford A, Buerkle CA, Buggs R, Butlin RK, Dieckmann U, Eroukhmanoff F, Grill A, Cahan SH, Hermansen JS, Hewitt G, Hudson AG, Jiggins C, Jones J, Keller B, Marczewski T, Mallet J, Martinez-Rodriguez P, Most M, Mullen S, Nichols R, Nolte AW, Parisod C, Pfennig K, Rice AM, Ritchie MG, Seifert B, Smadja CM, Stelkens R, Szymura JM, Vainola R, Wolf JB, Zinner D. 2013. Hybridization and speciation. Journal of Evolutionary Biology 26: 229–246.

Alexander DH, Lange K. 2011. Enhancements to the ADMIXTURE algorithm for individual ancestry estimation.

BMC Bioinformatics 12: 246.

Arakaki M, Christin P-A, Nyffeler R, Lendel A, Eggli U, Ogburn RM, Spriggs E, Moore MJ, Edwards EJ. 2011.

Contemporaneous and recent radiations of the world’s major succulent plant lineages. Proceedings of the National Academy of Sciences of the United States of America 108:

8379–8384.

Arnold ML, Martin NH. 2009. Adaptation by introgression.

Journal of Biology 8: 82.

Bolger AM, Lohse M, Usadel B. 2014. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30: 2114–2120.

Bouckaert RR. 2010. DensiTree: making sense of sets of phylogenetic trees. Bioinformatics 26: 1372–1373.

Browning BL, Browning SR. 2013. Improving the accuracy and efficiency of identity-by-descent detection in population data. Genetics 194: 459–471.

Burnham KP, Anderson DR. 2002. Model selection and multimodel inference: a practical information-theoretic approach, 2nd edn. Ecological modelling, Vol. 172. Dordrecht:

Springer.

Chaves JA, Cooper EA, Hendry AP, Podos J, De Leon LF, Raeymaekers JA, MacMillan WO, Uy JA. 2016. Genomic variation at the tips of the adaptive radiation of Darwin’s finches. Molecular Ecology 25: 5282–5295.

Danecek P, Auton A, Abecasis G, Albers CA, Banks E, DePristo MA, Handsaker RE, Lunter G, Marth GT, Sherry ST, McVean G, Durbin R; Genomes Project Analysis Group. 2011. The variant call format and VCFtools. Bioinformatics 27: 2156–2158.

Degnan JH, Rosenberg NA. 2006. Discordance of species trees with their most likely gene trees. PLoS Genetics 2:

e68.

Delhomme N, Sundstrom G, Zamani N, Lantz H, Lin YC, Hvidsten TR, Hoppner MP, Jern P, Van de Peer Y, Lundeberg J, Grabherr MG, Street NR. 2015.

Serendipitous meta-transcriptomics: the fungal community of Norway spruce (Picea abies). PLoS One 10: 1–9.

Douglas GM, Gos G, Steige KA, Salcedo A, Holm K, Josephs EB, Arunkumar R, Ågren JA, Hazzouri KM, Wang W. 2015. Hybrid origins and the earliest stages of diploidization in the highly successful recent polyploid Capsella bursa-pastoris. Proceedings of the National Academy of Sciences of the United States of America 112:

2806–2811.

Durand EY, Patterson N, Reich D, Slatkin M. 2011. Testing for ancient admixture between closely related populations.

Molecular Biology & Evolution 28: 2239–2252.

Excoffier L, Dupanloup I, Huerta-Sánchez E, Sousa VC, Foll M. 2013. Robust demographic inference from genomic and SNP data. PLoS Genetics 9: e1003905.

Excoffier L, Foll M. 2011. Fastsimcoal: a continuous-time coalescent simulator of genomic diversity under arbitrarily complex evolutionary scenarios. Bioinformatics 27:

1332–1334.

Feng S, Ru D, Sun Y, Mao K, Milne R, Liu J. 2019. Trans- lineage polymorphism and nonbifurcating diversification of the genus Picea. New Phytologist 222: 576–587.

Fitch WM, Margoliash E. 1967. Construction of phylogenetic trees. Science 155: 279–284.

Fu L, Niu B, Zhu Z, Wu S, Li W. 2012. CD-HIT: accelerated for clustering the next-generation sequencing data.

Bioinformatics 28: 3150–3152.

Givnish TJ. 2015. Adaptive radiation versus “radiation” and

“explosive diversification”: why conceptual distinctions are fundamental to understanding evolution. New Phytologist 207: 297–303.

Givnish TJ, Barfuss MH, Van Ee B, Riina R, Schulte K, Horres R, Gonsiska PA, Jabaily RS, Crayn DM, Smith JA, Winter K, Brown GK, Evans TM, Holst BK, Luther H, Till W, Zizka G, Berry PE, Sytsma KJ. 2014.

Adaptive radiation, correlated and contingent evolution, and net species diversification in Bromeliaceae. Molecular Phylogenetics and Evolution 71: 55–78.

Glor RE. 2010. Phylogenetic insights on adaptive radiation.

Annual Review of Ecology Evolution & Systematics 41:

251–270.

Haas BJ, Papanicolaou A, Yassour M, Grabherr M, Blood PD, Bowden J, Couger MB, Eccles D, Li B, Lieber M, MacManes MD, Ott M, Orvis J, Pochet N, Strozzi F, Weeks N, Westerman R, William T, Dewey CN, Henschel R, LeDuc RD, Friedman N, Regev A. 2013. De novo transcript sequence reconstruction from RNA-seq using the Trinity platform for reference generation and analysis.

Nature Protocols 8: 1494–1512.

Harrison SP, Yu G, Takahara H, Prentice IC. 2001.

Palaeovegetation (communications arising): diversity of temperate plants in East Asia. Nature 413: 129–130.

Hewitt GM. 2000. The genetic legacy of the Quaternary ice ages. Nature 405: 907–913.

Hu H, Al-Shehbaz IA, Sun Y, Hao G, Wang Q, Liu J. 2015.

Species delimitation in Orychophragmus (Brassicaceae) based on chloroplast and nuclear DNA barcodes. Taxon 64:

714–726.

Hu H, Hu Q, Al-Shehbaz IA, Luo X, Zeng T, Guo X, Liu J.

2016. Species delimitation and interspecific relationships of the genus Orychophragmus (Brassicaceae) inferred from whole chloroplast genomes. Frontiers in Plant Science 7: 1826.

Hu H, Zeng T, Wang Z, Al-Shehbaz IA, Liu J. 2018. Species delimitation of the Orychophragmus violaceus species complex (Brassicaceae) based on morphological distinction and reproductive isolation. Botanical Journal of the Linnean Society 188: 257–268.

(11)

Hudson RR, Boos DD, Kaplan NL. 1992. A statistical test for detecting geographic subdivision. Molecular Biology and Evolution 9: 138–151.

Johnson TC, Scholz CA, Talbot MR, Kelts K, Ricketts RD, Ngobi G, Beuning K, Ssemmanda I, Mcgill JW. 1996.

Late Pleistocene desiccation of Lake Victoria and rapid evolution of cichlid fishes. Science 273: 1091–1093.

Li B, Dewey CN. 2011. RSEM: accurate transcript quantification from RNA-seq data with or without a reference genome. BMC Bioinformatics 12: 323.

Li H, Durbin R. 2009. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25:

1754–1760.

Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G, Abecasis G, Durbin R; Genome Project Data Processing Subgroup. 2009. The sequence alignment/map format and SAMtools. Bioinformatics 25:

2078–2079.

Li L, Stoeckert CJJ, Roos DS. 2003. OrthoMCL: identification of ortholog groups for eukaryotic genomes. Genome Research 13: 2178–2189.

Liu JQ, Duan YW, Hao G, Ge XJ, Sun H. 2014. Evolutionary history and underlying adaptation of alpine plants on the Qinghai-Tibet Plateau. Journal of Systematics and Evoltuion 52: 241–249.

Liu JQ, Sun YS, Ge XJ, Gao LM, Qiu YX. 2012.

Phylogeographic studies of plants in China: advances in the past and directions in the future. Journal of Systematics and Evolution 50: 267–275.

Liu JQ, Wang YJ, Wang AL, Hideaki O, Abbott RJ.

2006. Radiation and diversification within the Ligularia–

Cremanthodium–Parasenecio complex (Asteraceae) triggered by uplift of the Qinghai-Tibetan Plateau. Molecular Phylogenetics and Evolution 38: 31–49.

Liu L, Xi Z, Davis CC. 2015. Coalescent methods are robust to the simultaneous effects of long branches and incomplete lineage sorting. Molecular Biology and Evolution 32:

791–805.

Liu L, Yu L. 2010. Phybase: an R package for species tree analysis. Bioinformatics 26: 962–963.

Maddison WP. 1997. Gene trees in species trees. Systematic Biology 46: 523–536.

Martin SH, Davey JW, Jiggins CD. 2015. Evaluating the use of ABBA-BABA statistics to locate introgressed loci.

Molecular Biology and Evolution 32: 244–257.

McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A, Garimella K, Altshuler D, Gabriel S, Daly M. 2010. The genome analysis toolkit: a mapreduce framework for analyzing next-generation DNA sequencing data. Genome Research 20: 1297–1303.

Nei M, Li WH. 1979. Mathematical model for studying genetic variation in terms of restriction endonucleases. Proceedings of the National Academy of Sciences of the United States of America 76: 5269–73.

Nichols R. 2001. Gene trees and species trees are not the same. Trends in Ecology & Evolution 16: 358–364.

Novikova PY, Hohmann N, Nizhynska V, Tsuchimatsu T, Ali J, Muir G, Guggisberg A, Paape T, Schmid K,

Fedorenko OM, Holm S, Säll T, Schlötterer C, Marhold K, Widmer A, Sese J, Shimizu KK, Weigel D, Krämer U, Koch MA, Nordborg M. 2016. Sequencing of the genus Arabidopsis identifies a complex history of nonbifurcating speciation and abundant trans-specific polymorphism. Nature Genetics 48: 1077–1082.

Osborn HF. 1902. The law of adaptive radiation. The American Naturalist 36: 353–363.

Pamilo P, Nei M. 1988. Relationships between gene trees and species trees. Molecular Biology and Evolution 5: 568–583.

Patterson N, Price AL, Reich D. 2006. Population structure and eigenanalysis. PLoS Genetics 2: 2074–2093.

Pavlidis P, Laurent S, Stephan W. 2010. MsABC: a modification of Hudson’s ms to facilitate multi-locus ABC analysis. Molecular Ecology Resources 10: 723–727.

Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, Maller J, Sklar P, de Bakker PI, Daly MJ, Sham PC. 2007. PLINK: a tool set for whole- genome association and population-based linkage analyses.

American Journal of Human Genetics 81: 559–575.

Qian H, Ricklefs RE. 2000. Large-scale processes and the Asian bias in species diversity of temperate plants. Nature 407: 180–182.

Qian H, Song JS, Krestov P, Guo Q, Wu Z, Shen X, Guo X.

2003. Large-scale phytogeographical patterns in east Asia in relation to latitudinal and climatic gradients. Journal of Biogeography 30: 129–141.

Qiu H, Yoon HS, Bhattacharya D. 2013. Algal endosymbionts as vectors of horizontal gene transfer in photosynthetic eukaryotes. Frontiers in Plant Science 4: 366.

Richardson JE, Pennington RT, Pennington TD, Hollingsworth PM. 2001. Rapid diversification of a species-rich genus of Neotropical rain forest trees. Science 293: 2242–2245.

Ru D, Mao K, Zhang L, Wang X, Lu Z, Sun Y. 2016. Genomic evidence for polyphyletic origins and interlineage gene flow within complex taxa: a case study of Picea brachytyla in the Qinghai-Tibet Plateau. Molecular Ecology 25:

2373–2386.

Ru DF, Sun YS, Wang DL, Chen Y, Wang TJ, Hu QJ, Abbott RJ, Liu JQ. 2018. Population genomic analysis reveals that homoploid hybrid speciation can be a lengthy process. Molecular Ecology 27: 4875–4887.

Rundell RJ, Price TD. 2009. Adaptive radiation, nonadaptive radiation, ecological speciation and nonecological speciation.

Trends in Ecology and Evolution 24: 394–399.

Simões M, Breitkreuz L, Alvarado M, Baca S, Cooper JC, Heins L, Herzog K, Lieberman BS. 2016. The evolving theory of evolutionary radiations. Trends in Ecology and Evolution 31: 27–34.

Stamatakis A. 2014. RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies.

Bioinformatics 30: 1312–1313.

Suh A, Smeds L, Ellegren H. 2015. The dynamics of incomplete lineage sorting across the ancient adaptive radiation of neoavian birds. PLoS Biology 13: 1–18.

Sun YS, Abbott RJ, Lu ZQ, Mao KS, Zhang L, Wang XJ, Ru DF, Liu JQ. 2018. Reticulate evolution within a spruce

(12)

(Picea) species complex revealed by population genomic analysis. Evolution 72: 2669–2681.

Svardal H, Jasinska AJ, Apetrei C, Coppola G, Huang Y, Schmitt CA, Jacquelin B, Ramensky V, Muller-Trutwin M, Antonio M, Weinstock G, Grobler JP, Dewar K, Wilson RK, Turner TR, Warren WC, Freimer NB, Nordborg M. 2017. Ancient hybridization and strong adaptation to viruses across African vervet monkey populations. Nature Genetics 49:

1705–1713.

Tajima F. 1989. Statistical method for testing the neutral mutation hypothesis by DNA polymorphism. Genetics 123:

585–595.

Vilella AJ, Severin J, Ureta-Vidal A, Heng L, Durbin R, Birney E. 2009. EnsemblCompara GeneTrees: complete, duplication-aware phylogenetic trees in vertebrates. Genome Research 19: 327–335.

Wang TJ, Ru DF, Zhang D, Hu QJ. 2019. Analyses of genome-scale variation real divergence of two Sinalliaria species (Brassicaceae) with continuous but limited gene flow.

Journal of Systematics and Evolution 57: 268–277.

Woese CR. 2000. Interpreting the universal phylogenetic tree. Proceedings of the National Academy of Sciences of the United States of America 97: 8392–8396.

Yang Z, Rannala B. 2012. Molecular phylogenetics: principles and practice. Nature Reviews Genetics 13: 303.

Yu G, Chen X, Ni J, Cheddadi R, Guiot J, Han H, Harrison SP, Huang C, Ke M, Kong Z. 2000.

Palaeovegetation of China: a pollen data-based synthesis for the mid-Holocene and last glacial maximum. Journal of Biogeography 27: 635–664.

Zhang JQ, Meng SY, Allen GA, Wen J, Rao GY. 2014. Rapid radiation and dispersal out of the Qinghai-Tibetan Plateau of an alpine plant lineage Rhodiola (Crassulaceae). Molecular Phylogenetics and Evolution 77: 147–158.

Zhang L, Zeng T, Hu H, Fan L, Zheng H, Hu QJ. 2018.

Interspecific divergence of two Sinalliaria (Brassicaceae) species in eastern China. Frontiers in Plant Science 9: 77–89.

Zhou TY, Lu LL, Yang G, Al-Shehbaz IA. 2001.

Orychophragmus Bunge. In: Wu ZY, Raven PH, eds. Flora of China, Vol. 8. Beijing: Science Press; St. Louis: Missouri Botanical Garden Press.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this article at the publisher’s website.

Figure S1. Venn diagram showing the number of identical SNPs or unique SNPs in four Orychophragmus spp.

Figure S2. The distributions of estimates of population genetic differentiation (F_ST). Fixation index (F_ST) values between four Orychophragmus spp. were calculated in all divergent genes.

Figure S3. Box of the summary statistic for Tajima’s D. Species are illustrated by yellow (Orychophragmus violaceus), blue (O. zhongtiaoshanus), purple (O. diffusus) and green (O. longisiliquus).

Figure S4. Trees inferred using the ML method (left) and the NJ method (right).

Figure S5. The 33 models hypothesized for simulations. Model 1: radiative divergence with gene flow; Model 2:

radiative divergence without gene flow (see Fig. 3D); Model 3: expansion demographic after radiative divergence without gene flow; Models 4–33: all possible bifurcating relationships detected in Figure 2D with or without gene flow. Arrows indicate the migration rate and T1, T2 and T3 represent divergence times (years). NA1, NA2 and NA3 indicate the effective population size of the ancestral clade. Numbers in parentheses are bootstrap confidence intervals.

Table S1. Locations of the 32 Orychophragmus sampled and the outgroup.

Table S2. Statistics for de novo assembled transcriptomes.

Table S3. Statistical transcriptome information.

Table S4. Details of sample mapping depth and coverage information.

Table S5. Cross-validation (CV) error scores for ADMIXTURE models.

Table S6. Mean (±SD) values of_. the fixation index (F_ST) between the four species.

Table S7. Mean (±SD) values of π and Tajima’s D.

Table S8. The 33 models hypothesized for simulations of species divergence.