Genomic nucleotide data from 90 individuals of 21 populations across the species range from central Asia to North America were obtained from the previous study (Wang et al., 2018) (Supplementary Tables 1, 2). The reference genome of salt cress, Eutrema salsugineum, was reported (Yang et al., 2013).

Clean reads from each sample above were aligned to the Eutrema salsugineum nuclear genome sequence v1.0 (Salt cress) (Yang et al., 2013). Genome mapping was conducted using BWA software with “mem” option and default parameters (Li and Durbin, 2009). The Picard package (http://picard.sourceforge.net/) was subsequently used to check for PCR duplicates. The Genome Analysis Toolkit (GATK) (Mckenna et al., 2010) was used to perform local realignment of reads to enhance the alignments in the vicinity of putative indels.

After genome mapping, The SNP calling was done for all individuals using SAMtools v1.1 (mpileup and BCFtools). Only paired aligned reads were used for SNP calling. The genotype likelihoods for each individual per site were calculated, and allele frequencies were estimated. The “mpileup” command was used to identify SNPs with the parameters “-q 30 -C 50 -S -D -Q 30 -m 2 -F 0.002 -guf.” Low-coverage depth SNPs (summing all samples) were then filtered with the vcfutils.pl in BCFtools v1.1 (Li et al., 2009) with parameters “-d 135 -D 1800” and high-quality SNPs (RMS of mapping quality ≥10, the distance of adjacent SNPs in the vicinity of indel polymorphisms ≥5 bp, Hardy-Weinberg equilibrium (HWE) P <5e-3, SNP quality ≥ 30, 3.0 ≤ quality by depth (each individual) ≤ 30, SNPs with observed heterozygosity (Ho < 0.6) were further filtered by Perl scripts. We used the obtained about 1.76 million high-quality SNPs for subsequent analysis.

The software RAxML was then used to construct phylogenetic trees with the GTR-G model and 1,000 rapid bootstrapping replicates based on the Maximum Likelihood (ML) method (Stamatakis, 2015). The final Maximum Likelihood trees were viewed using FigTree (v1.4.0) (http://tree.bio.ed.ac.uk/software/figtree/).

To identify genomic regions that might have been subject to selection during stress tolerance, we applied genetic diversity tests to the entire data from all 90 individuals. Nucleotide diversity was calculated using the standard estimate of the scaled mutation rate: the average pairwise nucleotide diversity θ_π (Tajima, 1989). Tajima's D was also calculated by dividing the difference between the average pairwise nucleotide diversity and the proportion of segregating sites by the square root of its SE (Tajima, 1989). We scanned the genome for regions with the highest differences using a window size of 20 kb and a step size of 10 kb. Windows that shared the lowest 5% of θ_π and lowest 5% Tajima's D estimates of the entire data were identified as putatively selected regions in the salt cress. Genes located in these regions were considered putatively selected genes. The above analyses were conducted using Vcftools (Danecek et al., 2011) or PERL scripts.

To identify genomic regions that might have been associated with local adaptation, we also applied genetic diversity tests to subpopulation data: Y1 (from northern China) and Y4 (from North America-Russia). The fixation index (F_ST) (Weir and Cockerham, 1984), and nucleotide diversity (θ_π log-ratio Y1/Y4 and Y4/Y1) were also chosen as indicators of population differentiation. We scanned the genome for regions with the highest differences using a window size of 20 kb and a step size of 10 kb. Windows that shared the highest 5% of F_ST and highest 5% log-ratio estimates were recognized as positively selected regions in a given population. Genes located in these regions were considered putatively selected genes in local area.

We annotated functional categories of E. salsugineum genes based on the corresponding Arabidopsis thaliana orthologs. Functional enrichment analysis of Gene Ontology (GO) was performed using the KOBAS 2.0 web server (Wu et al., 2006). The chi-squared test was used to calculate the statistical significance of enrichment and only terms with a p-value <0.05 were considered significant.

We obtained re-sequenced data of the genomes of 90 E. salsugineum individuals spanning their worldwide geographic distributions from Wang et al. (2018). All of these data were mapped to an available reference genome (Yang et al., 2013). All high-quality single-nucleotide polymorphisms (SNPs) were further used for subsequent analysis.

To identify genomic regions that might have been subject to selection, we first combined the all populations into a single gene pool. Out of 24,039 windows of 20 kb in length sliding in 10 kb steps across the salt cress genome, 23,366 windows contain >10 SNPs and cover 97.9% of the genome (Figure 1; Supplementary Figure 1). These 23,366 windows were used to detect signatures of selective sweeps at the species level. To identify regions with selective sweep signals, we used an empirical procedure (Branca et al., 2011) and selected windows with both significantly low diversity (θ_π) (5% right tail, where θ_π is 0.00053) and an excess of low-frequency variants (low D_T) (5% right tail, where D_T is = −1.0197) when compared to the empirical distribution of these two statistics. This led to the identification of a total of 4.76 Mb genomic regions (1.97% of the genome, containing 262 genes) with strong selective sweep signals (Figure 1). Among these putatively selected regions, 227 genes were annotated and classified according to terms developed by the Gene Ontology Consortium (Berardini et al., 2004) and found in the TAIR database (Yon et al., 2003). Among them, 48% of genes were assigned to broad or unspecified GO categories: biological processes unknown, other cellular, other metabolic and other biological processes. We successfully classified the remaining genes to roles in transcription and signal transduction, protein, DNA or RNA metabolism, transport, development, response to abiotic or biotic stimulus and response to stress (Figure 2; Table 1; Supplementary Table 3). Of the categories with defined functions, the larger groups of biological processes were involved in response to abiotic or biotic stresses (10.5%) including stress (5.6%) and abiotic or biotic stimulus (4.9%), protein metabolism (8.5%), and transcription (8.5%) (Figure 2; Supplementary Table 3). Cell compartments clustered by GO terms showed that most of the predicted gene products were localized in the nucleus and endoplasmic reticulum (ER) (Figure 2). We identified 16 genes that were common to abiotic and biotic stresses with selection signals related to ion homeostasis, osmotic adjustment and growth regulation (Table 1). For example, seven of these (GLP9, NDPK1, ACO2, PTR3, PYK10, NUDT7, and MDAR2) are involved in response to osmotic adjustment and salt stress while some others (including ERF13, WRKY38, DAR4, and several disease resistance protein) are closely related to cell death and defense in plant growth regulations (Table 1).

We conducted phylogenetic analyses of all sampled individuals based on the nuclear genome SNPs. Consistenting with our earlier research (Wang et al., 2018), all samples of salt cress were clustered into four distinct lineages, Y1 to Y4 (Supplementary Figure 2; Supplementary Tables 1, 2). Y1 comprised all individuals from northern China (Y1), Y2 contained sampled individuals from western China (Xinjiang) while Y3 contained those from Altai. Y4 comprised all individuals from NE Russia and Canada.

To detect accurately the genomic footprints left by local selections between two main salt cress ecotypes defined by phylogenetic analyses (Wang et al., 2018), we applied genetic diversity tests to data from the two groups of populations, Y1 (northern China) and Y4 (North America-NE Russia). Putatively selected genes (PSGs) were identified by screening selected windows simultaneously with significantly high log₂ [Y1 ratio (θ_π, Y4/θ_π, Y1) and Y4 ratio (θ_π, Y1/θ_π, Y4)] (5% right tail, where log₂ (Y1 ratio) is 1.84 and log₂ (Y4 ratio) is 2.4043) and significantly high F_ST values (5% right tail, F_ST threshold: 0.9098) between them (Li et al., 2014; Qiu et al., 2015; Ma et al., 2018), which also exhibited significant differences (p-value <10⁻¹⁶, Mann-Whitney U-test) (Figure 3A). We identified 23 regions under selection for the population Y1 (northern China) with a total size of around 0.43 Mb (0.17% of the genome) and 17 regions for the population Y4 (North America-Russia) with a total size of around 0.33 Mb (0.13% of the genome). A total of 63 genes including 60 protein-coding genes and three hypothetical ones were annotated in the regions under selection of Y1 from northern China, while 42 protein-coding genes were annotated for those of Y4 from North America-Russia (Figure 3A; Supplementary Tables 4, 5). Functions of the genes from northern China are involved in DNA repair, plastid fission, meiosis, regulation of defense response to fungus, and regulation of sulfur metabolic process (Supplementary Table 4), while gene ontology (GO) enrichment analysis revealed that genes from North America-Russia are functionally related to response to various carbohydrates, transport, peptide biosynthetic process, and long-chain fatty acid biosynthesis process (Supplementary Table 5). Example of genes with strong selection sweep signals in northern China and North America-Russia are shown in Figure 3B.

Salt cress can tolerate salt concentrations up to 500 mmol/L NaCl (Zhu, 2001), allowing it to survive on soils with high salt content. Such habitats would be expected to exert a high selection pressure and reduce genetic diversity in genomic regions involved in salt resistance. Since salt cress is a selfing species in which a high level of linkage disequilibrum is expected, these selective effects could also extend well-beyond the genomic regions under selection. Indeed, LD in salt cress is higher than those of other species examined (Wang et al., 2018) and its genetic diversity is extremely low and similar to those of domesticated rice and soybean (Wang et al., 2018). Therefore, regions showing lower diversity appeared more likely to be under selection pressure (lower Tajima's D, Pearson correlation efficient = 0.516, p <2.2e-16), further suggesting that such regions resulted more from selection pressure than from genetic drift or other neutral demographic processes.

Abiotic stress tolerances of plants are likely to involve in several different physiological and developmental pathways, such as osmotic homeostasis, stress damage control and repair, and growth regulation (Zhu, 2001, 2002, 2003; Mahajan and Tuteja, 2005). Some genes among the 227 annotated genes for which selective sweeps were detected operate in pathways important in salt tolerance (Table 1; Supplementary Table 3). For example, ACO2 is a key member of the ethylene synthesis pathway, while ERF13 is an important ethylene-responsive transcription factor (Schellingen et al., 2014; Sogabe et al., 2014). Ethylene-mediated signaling pathways have been shown to be critically involved in enhanced salt tolerance in plants (Ryu and Cho, 2015). In addition, MDAR2 and NDPK1, are related to the removal of toxic H₂O₂ (Fukamatsu et al., 2003; Lisenbee et al., 2005), while NUDT7 plays a vital role in regulating redox homeostasis during salt stress/defense signaling and programmed cell death in plant disease resistance (Muthuramalingam et al., 2015). Finally, several annotated genes are involved in growth regulation: for instance, SCR, a GRAS family transcription factor, regulates stem cell fate of the immediately surrounding cells (Moreno-Risueno et al., 2015), while PYK10 encodes b-glucosidase in the sub-cellular compartments after wounding (Hara-Nishimura and Matsushima, 2003).

Local selections by the different habitats plays a fundamental role in the production and maintenance of genetic diversity (Savolainen et al., 2013). In the middle Pleistocene, the climate tended to become drier and cooler while desertification and salinization began to develop and expand in central Asia (Wang et al., 2015). These changes might have triggered origin of E. salsugineum and its divergence into northern China and NE Russia-North America via two long-distance dispersal ways (Wang et al., 2018), where may be important in local adaptation selected by their different habitats. Functional enrichment analysis of GO terms revealed a remarkable amount of divergence between salt cress populations in northern China and North America-Russia, suggesting genome-wide selection by these local habitats (Figure 3; Supplementary Tables 4, 5).

The acclimated freezing tolerance of salt cress was positively correlated with the average minimum habitat temperature (Yang et al., 2013), similar to those of Arabidopsis (Hannah et al., 2006; Zhen and Ungerer, 2008). The average minimum habitat temperatures during the coldest month of the growing season at collection sites in North America and Northeast Russia are lower 10 degrees below zero, while that at collection sites in northern China are above zero (Yang et al., 2013). The Yukon cress ecotype from North America-Russia can even tolerate temperatures as low as −19°C (Griffith et al., 2007). Low temperatures induce a number of alterations in cellular components, including the amount of unsaturated fatty acids (Mahajan and Tuteja, 2005) and changes in protein and carbohydrate composition (Lynch and Thompson, 1982). The accumulation of sucrose and other simple sugars that occurs with cold acclimation also contributes to the stabilization of membranes as these molecules protect membranes against freeze-damage (Mahajan and Tuteja, 2005). This was confirmed by the enrichment and classification of GO terms of genes exhibiting selection signals. Most of these annotated genes are involved in carbohydrate stimulus, peptide biosynthetic and unsaturated fatty acids metabolic process (Supplementary Table 5). For example, ADS2 is critical in the synthesis of unsaturated fatty acids that are an essential component for cold adaptation (Chen and Thelen, 2013) while the gene STH1 (Figure 2B), as the homolog of the Salt Tolerance protein (STO) in Arabidopsis, may similarly regulate photomorphogenesis in light signaling in response to low temperatures (Salazar, 2010). In addition, a homolog of two tandemly duplicated genes TKL1 (Figure 2B) in cucumber (CsTK) increases both photosynthetic rate and carboxylation efficiency under low temperature and light intensity (Bi et al., 2013). Both STH and TKL1 genes exhibited the higher expressions in Yukon cress ecotype from North America-Russia than in Shandong cress ecotype from northern China when they were grown in the low-temperature common garden or in cabinet (Lee et al., 2012). These findings suggest that salt cress variants found at high latitude in North America and NE Russia could be associated with adaptation to low-temperature habitats.

In northeast China, salt cress usually grows close to vast flood plains with high salinity, extremely wet air and high chemical pollution compared with other populations (Wang et al., 2015). Contaminated soil is a great threat for plants, which can even cause damage to DNA. As a predominantly selfing plant, rapid habitat range expansion could bring about a large amount of harmful mutations (Wang et al., 2018). It is likely that many of these selected DNA-repair associated genes could efficiently clear up the negative effects of exposure of harmful varients. Two of the over-represented gene ontology categories in northern Chinese populations were “telomere maintenance in response to DNA damage” and “DNA recombination” (Supplementary Table 4). Some of these genes, for example, IPT7, are involved in cytokinin (CK) biosynthetic process that promotes cell differentiation and regulates root length (Dello Ioio et al., 2012), while RAD54 is an important eukaryotic-specific recombination factor that plays a critical role in repairing damaged DNA due to radiation or heavy metal contamination (Sung, 1994; Raoul Tan et al., 2003). Moreover, only did it reach northern China, salt cress expanded and reached widespread distributions (Wang et al., 2018) might due to temperate climate conditions along the inland of the Yellow River and stronger resistance to pathogens (Yeo, 2014). Some other selected genes are related to defense against pathogens including “regulation of defense response to fungus” and “regulation of sulfur metabolic process.” Two genes, SPLAYED and EIL3 (Figure 3B), are known to play critical roles in enhancing defense against pathogens in biotic stress signaling networks (Van der Ent et al., 2008; Walley et al., 2008). In addition, the gene WES1 plays a key role in plant sulfur metabolism such as auxin and phytoalexin camalexin biosynthesis in pathogen stress response (Wang et al., 2012). Undoubtedly, genome-wide adaptive divergence further support that salt cress of northern China could display good resistance to biotic stress in their nature habitat (Pedras and Zheng, 2010; Yeo, 2014).

Due to its natural adaptations to various harsh climates and soil conditions, Eutrema salsugineum has long been used as an important model for deciphering mechanisms of salt and other abiotic stress in plants (Gong et al., 2005; Griffith et al., 2007; Lamdan et al., 2012). During adaptive evolution of the salt cress, genetic variation and natural selection are non-randomly fostering stress-related genes, gene interaction network in the whole genome, as well as prompting local adaptation, differentiation and diverse biological stresses between different accessions.