Genome and transcriptome analyses reveal parallel altitude adaptation in Chenopodium

Abstract

Background

Elucidating how crops adapt to heterogeneous environments requires integrative analyses of genome-wide variation, regulatory architecture, and evolutionary processes. Chenopodium quinoa, a globally important pseudocereal, shows strong ecological differentiation between highland and lowland ecotypes, yet the genetic and regulatory bases of environmental adaptation across the genus remain incompletely understood.

Results

We generate a comprehensive genomic and transcriptomic resource consisting of whole-genome resequencing of 558 accessions from 20 Chenopodium species and transcriptomes from 295 accessions. Population genomic analyses reveal extensive genetic diversity, asymmetric evolution of the A and B subgenomes, and widespread interspecific introgression. Notably, Chenopodium berlandieri contributes adaptive variation to cultivated quinoa, particularly in genes related to stress response and immunity. Comparative analyses identify signatures of parallel adaptation to altitude in both quinoa and its wild relative Chenopodium berlandieri, including shared targets of selection such as PTR2, involved in nutrient transport, and CONSTANS, a key regulator of photoperiodic flowering. By genome-wide eQTL mapping, we identify 2,659 cis- and 407,628 trans-eQTLs regulating more than 11,000 genes. A major cis-eQTL controlling ELF3 expression is associated with large upstream deletions enriched in highland quinoa populations and correlated with reduced gene expression and elongated hypocotyls, implicating regulatory structural variation in altitude adaptation.

Conclusions

Our integrative analyses demonstrate how coding variation, regulatory divergence, and introgression jointly drive parallel environmental adaptation across wild and cultivated Chenopodium, providing insights into polyploid crop evolution and resources for breeding climate-resilient quinoa.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13059-026-04136-5.

Background

Understanding how crops adapt to diverse and often extreme environments is a central goal of plant evolutionary biology and a major challenge for sustainable agriculture. Natural and human-mediated selection shape both coding and regulatory variation, yet the genetic mechanisms underlying environmental adaptation—particularly in polyploid crops—remain incompletely understood [1, 2]. Comparative analyses that integrate wild relatives and domesticated forms across genomic and transcriptomic dimensions provide powerful opportunities to dissect these processes at scale [3, 4].

The genus Chenopodium L. (Amaranthaceae, subfamily Chenopodioideae) comprises approximately 150 species with a worldwide distribution and remarkable ecological adaptability [5, 6]. Members of this genus occupy a wide range of habitats, from saline coastal environments and arid lowlands to cold, high-elevation regions, reflecting exceptional tolerance to abiotic stress and environmental heterogeneity [5]. The genus exhibits extensive cytogenetic diversity, including diploid, tetraploid, and hexaploid taxa with multiple genome configurations shaped by recurrent hybridization and polyploidization events [7]. Both wild and domesticated Chenopodium species have played important roles as regional food crops and medicinal plants across ancient Eurasian and American civilizations, while others have evolved as highly competitive agricultural weeds [5, 6, 8]. This combination of ecological breadth, ploidy diversity, and complex evolutionary history makes Chenopodium an attractive system for studying plant adaptation, genome evolution, and crop domestication [9, 10].

Among Chenopodium species, quinoa (Chenopodium quinoa Willd.) has emerged as a globally important crop due to its exceptional nutritional value and broad environmental tolerance [11]. Quinoa is an allotetraploid (2n = 4x = 36) pseudocereal domesticated in the Andean region of South America, which boasts seeds rich in essential amino acids, free from gluten, and abundant in dietary fiber, minerals, and vitamins, factors contributing to its rapid spread beyond its native range [12]. Today, quinoa is grown in more than 100 countries [13], reflecting its adaptability to diverse agro-ecological conditions and its increasing importance for global food security.

During domestication and subsequent geographic expansion, quinoa diversified into distinct ecotypes that are strongly associated with environmental gradients, particularly altitude [14–16]. Highland quinoa populations are adapted to cold, frost-prone, and drought-stressed environments with intense ultraviolet radiation, whereas lowland and coastal populations exhibit traits suited to warmer, more humid, and often saline conditions [17]. Because altitude integrates multiple environmental variables—including temperature, photoperiod, radiation intensity, and nutrient availability—adaptation to elevational gradients represents a central axis of quinoa ecological diversification [9, 13]. Domestication further involved selection for agronomically important traits such as larger seed size, reduced saponin content, and enhanced environmental adaptability [18]. Although recent genomic studies have begun to identify candidate genes associated with flowering time [19, 20], saponin biosynthesis [21], seed size [22], panicle architecture [23], and abiotic stress responses [16, 24], the broader evolutionary and regulatory mechanisms underlying adaptation across ecological gradients remain poorly resolved.

Wild Chenopodium species represent a rich but underutilized source of genetic diversity and adaptive traits. Recent breeding efforts have increasingly leveraged wild relatives such as C. hircinum and C. berlandieri to introduce novel alleles for disease resistance and stress tolerance [25]. Notably, C. berlandieri—a semi-domesticated tetraploid species native to North America—shares the same AABB genome configuration as quinoa and is believed to have undergone an independent domestication trajectory [26]. While it retains wild traits such as seed shattering and high saponin levels, C. berlandieri also exhibits adaptive features similar to those of quinoa, including salt and drought tolerance [24, 27]. These genomic and ecological similarities raise important questions regarding the roles of shared ancestry, introgression, and parallel evolution in shaping adaptive traits across geographically separated Chenopodium lineages.

The rapid expansion of genomic resources in recent years has enabled the identification of both conserved and lineage-specific genome architectures across the Chenopodium genus [23, 26, 28–32]. Gene-based phylogenetic analyses across 12 Chenopodium species, encompassing eight genome variants (A–H), indicate that the subgenomes of quinoa and C. berlandieri cluster together, consistent with their shared allopolyploid origin. The B subgenomes of both species show close phylogenetic relationship to the diploid species C. ficifolium (BB), whereas the A subgenomes cluster with the diploid C. pallidicaule (AA), suggesting these species as potential diploid progenitors of the allotetraploid lineage. Moreover, several wild genome types (C–F) show closer genetic distance to the B subgenome, suggesting a shared evolutionary trajectory within the B-clade of Chenopodium [33].

Despite these advances, the genetic basis of local adaptation, ecotypic divergence, and trait innovation across wild and cultivated Chenopodium species remains largely unresolved. To address these gaps, we conducted comprehensive genome and transcriptome sequencing and analyses across 20 Chenopodium species representing multiple ploidy levels and geographic regions. Our analyses uncovered extensive genetic diversity and distinct evolution patterns in wild and domesticated Chenopodium lineages, and identified genomic regions as well as candidate genes under parallel selection associated with environmental adaptation. Overall, these findings advance our understanding of crop adaptation to diverse environments and highlight potential targets for breeding climate-resilient cultivars under increasing environmental challenges.

Results

A genome variation map of 558 Chenopodium accessions spanning 20 species

To capture a broad spectrum of genetic diversity within the Chenopodium genus, we resequenced 303 accessions representing 18 recognized species at an average depth of ~ 15 × (Additional file 1: Tables S1-S4). Additionally, we integrated publicly available data from a previous study [23], which included 255 nonredundant accessions from 5 species, selected based on documented records and phylogenetic relationships. The final dataset included 558 accessions representing 20 species, covering a broad geographic range and diverse ploidy levels. These comprised AA (9 species; 31 accessions), BB (2 species; 5 accessions), AABB (3 species; 483 accessions), BBCCDD (3 species; 33 accessions), and unknown ploidy (3 species; 6 accessions) (Fig. 1a; Additional file 1: Tables S2 and S3). The majority of C. quinoa accessions were collected from Peru (37.71%) and Bolivia (21.96%), regions recognized as the center of quinoa origin and domestication, whereas 90.32% of C. berlandieri accessions were sourced from the United States, a primary hub for their germplasm preservation (Fig. 1b).

Fig. 1.

Geographic distribution and taxonomic composition of Chenopodium accessions analyzed in this study. a Geographic distribution of Chenopodium accessions included in the analysis. b Pie charts showing the proportional representation of Chenopodium accessions by species across countries and regions. The number of accessions in each species group is indicated in parentheses below each pie chart

Approximately 160 million raw bi-allelic SNPs were identified across the 558 accessions (Additional file 1: Table S5). A higher SNP density was observed in the A subgenome (139 SNPs/kb on average) compared to the B subgenome (116 SNPs/kb), suggesting greater genetic diversity among sampled Chenopodium populations with AA genomes/subgenomes (Additional file 2: Fig. S1). After filtering out rare variants and excluding SNPs with a missing rate > 5%, a total of 8,576,501 high-quality SNPs were retained for downstream analyses. To validate SNP reliability, we first compared SNPs identified in AABB, AA, BB, and BBCCDD accessions with published reference genomes representing these groups [33]. For each high-confidence polymorphic site, a 400-bp sequence centered on the SNP was extracted and aligned to the corresponding reference genomes. Exact matches were observed for 97.84% of SNPs in C. berlandieri (AABB), 96.38% in C. watsonii (AA), 99.13% in C. ficifolium (BB), and 96.59% in C. formosanum (BBCCDD), indicating high consistency of SNP calls. In addition, PCR amplification and Sanger sequencing of 15 SNP loci across eight chromosomes in representative accessions from each genome group showed 94.67% concordance (71/75 sites) with inferred genotypes (Additional file 1: Table S6), further supporting the accuracy of SNP identification. Among the high-quality SNPs, 1,690,433 (19.71%) were located in coding regions, comprising 806,503 (9.40%) nonsynonymous substitutions, 875,386 (10.21%) synonymous substitutions, 816 start codon changes, and 18,489 stop codon changes. This comprehensive variant map offers a valuable genetic resource to advance quinoa research and breeding programs.

Phylogeny and population structure of Chenopodium

We inferred phylogenetic relationships within the genus Chenopodium using 1,554,832 SNPs at fourfold degenerate sites. The resulting phylogenetic tree delineated the accessions into distinct groups, closely reflecting their taxonomic classification (Fig. 2a). Among C. quinoa accessions, two major subclades were identified, largely consistent with differences in altitudinal distribution among accessions (t-test, P = 6.709e-05) (Fig. 2b). This phylogenetic clustering was further corroborated by t-SNE analysis (Fig. 2c).

Fig. 2.

Phylogeny and population structure of the Chenopodium diversity panel. a Phylogenetic tree of 558 Chenopodium accessions. Inset pie charts show the elevation (in meters) distribution of quinoa accessions. b Phylogenetic tree based on the p-distance matrix (top) and geographical distribution (bottom) of 419 quinoa accessions. c t-distributed stochastic neighbor embedding (t-SNE) analysis of 558 Chenopodium accessions (left), and 419 quinoa and 62 C. berlandieri accessions, both are tetraploid (right). d Population structure of 558 Chenopodium accessions. Each vertical bar represents an individual accession, with color proportions indicating the estimated genetic ancestry

Population structure analysis classified the accessions into six distinct groups, aligning well with their respective genome configurations at the optimal k value of 6 (Fig. 2d; Additional file 2: Fig. S2). Accessions with the AA genome, along with one accession of previously unrecognized genotype, formed a single cluster with a high average ancestry coefficient of 0.95 (Fig. 2d; Additional file 1: Table S7). BB accessions formed a separate cluster together with seven C. berlandieri (AABB) accessions, exhibiting the lowest ancestry coefficient (0.78). Population differentiation (F_ST) analysis further revealed that BB accessions were genetically more similar to C. berlandieri (F_ST = 0.24) (Additional file 1: Table S8). Moreover, quinoa showed substantial genetic differentiation from wild species (AA, BB, BBCCDD), with an average F_ST of 0.70. In contrast, these wild species exhibited a closer genetic relationship with C. berlandieri (avg. F_ST = 0.41), consistent with the topology of phylogenetic tree (Fig. 2a). Collectively, these results suggest that AA and BB accessions may have directly contributed to the A and B subgenomes of C. berlandieri, respectively.

For the two C. hircinum (AABB) accessions hypothesized to be direct ancestors of quinoa [16], we observed distinct clustering patterns. One accession grouped with highland quinoa and shared 61.91% and 37.58% of its genetic background with highland and lowland quinoa, respectively. The other accession clustered with lowland quinoa and exhibited an admixed genetic profile comprising 32.08% lowland quinoa, 20.55% highland quinoa, 28.46% BB, and 18.91% C. berlandieri genomic components (Fig. 2d; Additional file 1: Table S7). In addition, identity-by-descent (IBD) analysis among the 558 accessions further supported the inferred phylogenetic relationships within Chenopodium and revealed high levels of IBD sharing within groups but low or negligible levels between groups (Additional file 2: Fig. S3; Additional file 1: Table S9). These findings elucidate the genetic relatedness among Chenopodium species and confirm the independent origins of tetraploid quinoa and C. berlandieri.

Genetic diversity and deleterious mutations in Chenopodium

We estimated nucleotide diversity (π) separately for the A and B genomes in both diploid and polyploid populations. Overall, the A genomes exhibited higher genetic diversity (mean π = 4.10 × 10⁻⁴) than the B genomes (π = 1.84 × 10⁻⁴). Among B-genome-containing groups, the BBCCDD population showed notably higher diversity (π = 5.25 × 10⁻⁴) than the B subgenome of any other groups, including the wild BB species (Fig. 3a). This elevated diversity may reflect the broad geographic origins of BBCCDD accessions (Fig. 1a), potentially resulting from independent hybridizations involving genetically more divergent parental BB lineages. In AABB populations, the asymmetric distribution of genetic diversity between subgenomes aligned with the higher π observed in wild AA diploids compared to wild BB diploids (Fig. 3a; Additional file 1: Table S10), suggesting that much of the diploid genetic variation was retained during tetraploidization. Interestingly, the A subgenome of C. berlandieri showed significantly higher diversity than both wild AA diploids and all other groups (P < 2.2e-16), indicating possible introgression from additional AA sources after its emergence. Furthermore, among the tetraploid populations, both C. berlandieri (π = 3.53 × 10⁻⁴) and quinoa (highland, π = 7.34 × 10⁻⁵; lowland, π = 6.12 × 10⁻⁵) exhibited substantially lower nucleotide diversity than other crops such as rice (π = 2.4 × 10⁻³) [34], maize (π = 1.40 × 10⁻²) [35], wheat (π_subA = 1.7 × 10⁻³, π_subB = 2.5 × 10⁻³, π_subD = 2.0 × 10⁻⁴) [36], foxtail millet (landraces, π = 1.62 × 10⁻³; cultivars, π = 1.57 × 10⁻³) [37, 38], and soybean (landraces, π = 1.40 × 10⁻³; cultivars, π = 1.05 × 10⁻³) [39]. Notably, genetic diversity in quinoa was significantly lower than that in C. berlandieri (P < 2.2e-16). These findings suggest that tetraploid Chenopodium species have experienced strong genetic bottlenecks during their evolution and domestication, with particularly severe effects in quinoa. Additionally, we found that genetic diversity in highland quinoa was higher than in lowland quinoa, consistent with the hypothesis that quinoa was first domesticated in high-altitude regions [40].

Fig. 3.

Genetic diversity and demographic history of Chenopodium populations. a Nucleotide diversity (π) of different Chenopodium groups. Colors represent distinct species or ploidy groups. b Proportion of deleterious mutations in the genomes/subgenomes of the 558 accessions. Colors indicate different species or ploidy groups. c Demographic history of wild populations, including diploid AA and BB, and hexaploid BBCCDD taxa. d Demographic history of tetraploid species, including C. berlandieri, highland quinoa, and lowland quinoa

The accumulation of deleterious mutations is frequently associated with plant domestication and improvement [41–43]. We identified deleterious mutations across Chenopodium using 5,402,242 high-quality SNPs. Overall, the B genomes harbored a higher proportion (19.49%) of deleterious mutations than that of A genomes (17.87%) in both diploid and tetraploid populations (Fig. 3b; Additional file 1: Table S11). This pattern contrasted with that of nucleotide diversity, in which B genomes exhibited lower π values than A genomes (Fig. 3a). Furthermore, the AA accessions carried significantly fewer deleterious mutations than any tetraploid population, supporting the “cost of domestication” hypothesis. This finding is consistent with the reduced efficacy of purifying selection in removing deleterious mutations in polyploids [44]. However, the BBCCDD accessions appeared to be an exception, which carried a relatively lower burden of deleterious mutations. This suggests that additional evolutionary forces, such as hybridization and increased effective population size, may have mitigated the accumulation of deleterious variants in this hexaploidy lineage.

Demographic history

We inferred the demographic history of Chenopodium populations using the MSMC method [45]. Our analysis revealed a shared population bottleneck approximately 5,000–10,000 years ago across all Chenopodium lineages, followed by a gradual recovery in the BBCCDD lineage (Fig. 3c). Divergence time estimates indicated that quinoa and C. berlandieri split more than 10,000 years ago, well before the onset of quinoa domestication (Fig. 3d). Subsequent diversification within quinoa occurred ~ 2,000 years ago, giving rise to the highland and lowland ecotypes. Notably, highland quinoa experienced a population expansion starting ~ 500 years ago, temporally aligned with its widespread cultivation as a staple crop during the Inca Empire (1438–1533 CE). In contrast, lowland quinoa experienced a steady population decline over the past 500 years, likely due to its displacement by newly introduced crops in lowland areas.

Alien introgression in quinoa and C. berlandieri

Interspecific introgression, widely observed in crops [46–48], play an important role in plant domestication and adaptation. To detect potential introgression in domesticated/semi-domesticated Chenopodium (i.e., quinoa and C. berlandieri), we conducted chromosome-level principal component analysis (PCA) using SNPs, which revealed variable genetic relationships across populations and chromosomes (Additional file 2: Fig. S4). While most chromosomes exhibited clustering patterns aligning with taxonomic classifications, several showed divergent groupings. For example, chromosome 6 showed distinct wild versus cultivated quinoa clusters, consistent with phylogeny (Figs. 2a and 4a). In contrast, chromosome 11 revealed a subset of highland quinoa clustering more closely with C. berlandieri than with other highland accessions, and chromosome 8 grouped AA accessions with lowland quinoa, diverging from genome-wide patterns (Fig. 4a; Additional file 2: Fig. S4). These chromosome-specific clustering patterns suggest that introgression events have occurred among Chenopodium lineages.

Fig. 4.

Genetic introgression within the genus Chenopodium. a t-SNE analysis of 558 Chenopodium accessions based on SNPs from individual chromosomes (with the corresponding subgenomes indicated in parentheses). Black arrows indicate accessions showing evidence of introgression. b Gene flow analysis using the BBCCDD lineage as an outgroup. F_d statistics were used to assess the strength of gene flow from potential donor populations (AA, BB, and C. berlandieri) to recipient populations (highland quinoa, lowland quinoa and C. berlandieri). Asterisks denote cases where gene flow from a specific donor is significantly higher than from other donors (one-tailed t-test, P < 0.01). c Chromosomal distribution of f_d values for three recipient populations—highland quinoa (orange), lowland quinoa (green), and C. berlandieri (blue)—across chromosome 02 (with diploid AA as donor), chromosome 14 (diploid BB as donor), and chromosome 04 (C. berlandieri as donor)

ABBA-BABA analyses [49] using BBCCDD as the outgroup revealed asymmetric introgression patterns in tetraploid Chenopodium (Additional file 2: Fig. S5). The diploid AA lineage contributed more genetic material to C. berlandieri (Figs. 3a and 4b), which in turn exhibited significantly higher levels of introgression into both quinoa ecotypes than either the AA or BB donors (Fig. 4b). Chromosome-wide analysis of f_d statistics identified numerous introgressed regions (Additional file 2: Figs. S6-S8), including a 2.81 Mb AA-derived segment on chromosome 2 of C. berlandieri that encoded 77 genes. Several of these genes were homologs of known Arabidopsis immunity regulators, such as CRK10 [50], WRKY65 [51], CPR30 [52], RPP8 [53], and CAD1 [54], as well as the hypoxia-responsive aquaporin NIP2;1 [55]. Moreover, a 1.76 Mb BB-derived region on chromosome 14 in highland quinoa harbored 46 genes, including homologs of BBD2 potentially regulating drought response [56], RICESLEEPER 3 involved in flowering [57], and NEP1-interacting protein gene involved in light response [58]. Introgression from C. berlandieri was more frequent in highland than in lowland quinoa (Fig. 4b). However, a 1.11 Mb C. berlandieri-derived segment on chromosome 4 in lowland quinoa was identified. This region, harboring 99 genes, included homologs of BIG involved in auxin transport [59] and SEPALLATA 2 playing a role in reproduction [60] (Fig. 4c; Additional file 1: Table S12).

Parallel evolution in quinoa and C. berlandieri for environmental adaptation

Phylogeny and PCA analyses (Fig. 2a; Additional file 2: Fig. S4) revealed considerable genetic divergence within the C. berlandieri population, though to a lesser extent than in quinoa. This divergence was further supported by a phylogenetic tree constructed exclusively from C. berlandieri accessions, which delineated two distinct subgroups (Fig. 5a). These two subgroups differed significantly in elevation (P = 4.03e-09) and associated environmental factors such as annual mean temperature (P = 7.4e-05) (Fig. 5b; Additional file 2: Fig. S9; Additional file 1: Table S13). Geographic data revealed that lowland C. berlandieri populations were primarily distributed along the eastern coastal regions, while highland populations were mainly found on the western plateau regions (Fig. 5c).

Fig. 5.

Parallel selection for environmental adaptation in both quinoa and C. berlandieri. a Phylogenetic tree of the C. berlandieri population. b Boxplots showing elevation (left) and annual mean temperature (right) for highland and lowland C. berlandieri accessions. Statistical significance was assessed using two-tailed Student’s t-test. c Geographic distribution of highland and lowland C. berlandieri accessions. d Genome-wide distribution of nucleotide diversity across chromosomes 4 and 15, highlighting regions and overlapping genes (e.g., CO, ELF3, and PTR2) under parallel selection between highland and lowland populations of quinoa (top) and C. berlandieri (bottom). e AlphaFold-predicted structure of quinoa PTR2 (left), highlighting the position of a single amino acid substitution (valine vs. alanine) distinguishing lowland and highland genotypes. The right panel shows the frequency of the two amino acid variants in highland and lowland quinoa populations. f Maximum-likelihood phylogeny of CONSTANS (CO) orthologs from Arabidopsis, quinoa, rice, grape, and tomato. CqCO1 (AUR62023118) is highlighted in red. g Phenotypes of wild-type Arabidopsis and transgenic lines overexpressing CqCO1 (OE-5, OE-6, OE-7) or carrying the empty vector (EV). h Genotypic profiles of polymorphic sites within the gene body of CqCO1 in highland quinoa (HQ), lowland quinoa (LQ), highland C. berlandieri (HB), and lowland C. berlandieri (LB). Each row represents a single SNP, and each column represents an individual accession. 0/0: homozygous reference alleles; 1/1: homozygous alternative alleles; 0/1: heterozygous alleles. Gene structure is shown on the left. i Flowering time differences between highland and lowland quinoa under short-day and long-day conditions. j Expression levels of CqCO1 in highland and lowland populations of quinoa and C. berlandieri, measured as fragments per kilobase of transcript per million mapped fragments (FPKM). k CqCO1 expression in lowland accessions with either heterozygous or homozygous alleles. l PTR2 expression in highland and lowland populations of quinoa and C. berlandieri. Statistical significance in (i-l) was assessed using the Wilcoxon test

Exposure to similar environmental conditions can impose comparable selective pressures on different organisms, often leading to independent evolution of analogous adaptive traits driven by selection acting on homologous genetic loci [36, 61]. The existence of highland and lowland ecotypes in both C. berlandieri and quinoa provided a powerful system to investigate parallel selection associated with altitude adaptation. To identify genomic regions under parallel selection, we compared highland and lowland ecotypes of both species using F_ST and π ratios. Applying a top 5% threshold for both metrics, we identified selective sweeps spanning 45.71 Mb (2,747 genes) in C. berlandieri and 30.05 Mb (1,467 genes) in quinoa. Integrating both datasets yielded 3,393 unique genes (Additional file 2: Fig. S10; Additional file 1: Table S14). GO enrichment analysis revealed that these genes were significantly overrepresented in biological processes related to seed dormancy, floral development, UV-B response, and lignin and flavonoid biosynthesis (Additional file 2: Fig. S11), all of which are critical for plant survival and reproduction across elevational gradients.

While most elevation-adaptive genes appeared to be species-specific, we identified 87 genes under parallel selection in both quinoa and C. berlandieri, among which 26 contained highly differentiated SNPs (F_ST > 0.8) within coding and/or 5-kb upstream regions (Additional file 2: Fig. S10; Additional file 1: Table S14). These included orthologs of Arabidopsis PTR2 (AUR62017792) and CO (AUR62023118) (Fig. 5d), which were further supported by XP-CLR analysis (Additional file 2: Fig. S12). PTR2 encodes a high-affinity oligopeptide transporter, and its loss-of-function delays flowering and affects seed development and germination by modulating nitrogen availability and water uptake [62–64]. Heterologous expression of quinoa PTR2 in Arabidopsis accelerated flowering, suggesting a conserved functional role as a nutrient transporter (Additional file 2: Fig. S13a). A nonsynonymous SNP (Chr15: 19,296,867, C to T) in the PTR2 coding region was detected, resulting in an alanine-to-valine substitution at residue 413 (A413V), located within the cytosolic loop between two transmembrane domains—a region often linked to protein function variation [65] (Fig. 5e; Additional file 2: Fig. S13b). This variant (413 V) was highly prevalent in highland quinoa (85.1%) but much less common in lowland quinoa (13.5%), where it occurred redominantly in the heterozygous state (Fig. 5e; Additional file 1: Table S15). Comparative analysis of nine recently published quinoa genomes (five highland and four lowland) revealed that all highland lines carried the valine (V) allele, whereas all lowland lines harbored alanine (A) (Additional file 2: Fig. S13c). Notably, this was the only polymorphic site within the PTR2 protein sequence. Additionally, we identified five other NRT1/PTR family transporters associated with altitude adaptation, four of which showed evidence of selection specifically in highland populations (either quinoa or C. berlandieri; Additional file 1: Table S14), suggesting strong selection on nutrient transporters in nutrient-poor highland environments.

CONSTANS (CO), a central regulator of photoperiodic flowering, directly activates the expression of FLOWERING LOCUS T (FT) under long-day (LD) conditions, but not under short-day (SD) conditions [66]. The quinoa genome encodes two CO orthologs (Fig. 5f), among which CqCO1 (AUR62023118) exhibited signatures of selection in highland populations of both quinoa and C. berlandieri. The function of CO genes is often conserved across eudicots, and consistently, heterologous expression of CqCO1 in Arabidopsis markedly promoted flowering (Fig. 5g). Genotypic analysis revealed multiple segregating variants in the coding and upstream regulatory regions of CqCO1 between highland and lowland populations in both species, although the polymorphic sites occurred at different positions in quinoa and C. berlandieri (Fig. 5h). In C. berlandieri, highland and lowland accessions showed fixed but contrasting homozygous genotypes (1/1 vs. 0/0). In quinoa, two major haplotypes were identified, comprising 15 segregating SNPs (6 in the gene region and 9 in the promoter region). Haplotypes in highland quinoa primarily consisted of homozygous alternative alleles (1/1), while lowland quinoa exhibited two groups: one with homozygous reference alleles (0/0), and another (24% of lowland accessions) displaying heterozygous genotypes (0/1) (Fig. 5h). These patterns suggest parallel selection on CqCO1 in both species and indicate that lowland quinoa, descended from highland ancestors, may have acquired additional CqCO1 alleles through hybridization or introgression. To assess whether CqCO1 differentiation is associated with flowering time, we analyzed flowering time data for 260 quinoa accessions classified into highland and lowland groups [20]. Under LD conditions, highland quinoa (n = 166) flowered significantly later than lowland quinoa (n = 94), with an average delay of 9 days (P = 2.98e-14; Fig. 5i). Under SD conditions, however, both groups flowered earlier, and no significant difference in flowering time was observed between highland and lowland populations (P = 0.2598; Fig. 5i). Moreover, flowering time variation remained relatively stable across photoperiods in the lowland quinoa, but increased in highland accessions under LD conditions (Additional file 2: Fig. S14).

Given that many segregating variants were identified in the promoter regions and environmental adaptation is often linked to transcriptional regulation [67], we performed transcriptome sequencing on 295 leaf samples collected at the same developmental stage, including 169 quinoa and 61 C. berlandieri accessions (Additional file 1: Table S4). Our analysis revealed that CqCO1 expression was significantly lower in highland populations of both quinoa (P = 6.10e–14) and C. berlandieri (P = 0.04), with a more pronounced reduction in quinoa, consistent with delayed flowering observed in highland accessions under LD condition (Fig. 5j). Notably, CqCO1 expression also varied substantially within the lowland quinoa population. Although not statistically significant (P = 0.17), lowland accessions with heterozygous CqCO1 genotypes tended to exhibit higher expression of the gene than those with homozygous genotypes, suggesting that CqCO1 expression variation in lowland quinoa was in part contributed by genotypes (Fig. 5k). Given quinoa’s origin as a short-day species and its historical spread from low-latitude Andean highlands (with narrow photoperiod variation) to higher-latitude coastal lowland regions (with wide photoperiod variation) required adaptation to more variable photoperiods, our findings suggest that this adaptation to variable photoperiods may have been facilitated by selection for alleles that enhance CqCO1 transcription and/or mRNA stability in lowland Chenopodium, particularly in quinoa. Additionally, PTR2 expression was also significantly higher in highland quinoa than lowland accessions (P = 5.65e–14; Fig. 5l), suggesting a potential role in Chenopodium adaptation to nutrient-poor highland environments.

Genome-wide eQTL mapping

We performed expression quantitative trait loci (eQTLs) analysis using 7,985,852 high-quality SNPs and expression data from 32,887 genes across 230 accessions. This comprehensive analysis identified 410,287 significant eQTLs associated with the expression of 11,336 genes. Among these, 2,659 (0.65%) were classified as cis-eQTLs, spanning genomic intervals ranging from 9 bp to 85.89 kb, whereas 407,628 (99.35%) were trans-eQTLs, covering regions between 3 bp and 15.73 Mb in length (Fig. 6a, b; Additional file 1: Table S16). Of the trans-eQTLs, 321,572 (78.38%) were located within 5 Mb of their associated genes, while 88,715 (21.62%) extended beyond 5 Mb (Fig. 6b). Based on the genomic distribution of trans-eQTLs, we identified 3,991 trans-eQTL hotspots (P < 0.05).

Fig. 6.

Genome-wide mapping of eQTLs. a Genome-wide distribution of cis- and trans-eQTLs. b Length distribution of cis- (left) and trans-eQTLs (right). c Manhattan plot of eQTLs associated with ELF3 expression in quinoa (top). The horizontal dashed grey line indicates the Bonferroni-corrected significance threshold (α = 0.05). The middle panel shows the genomic location of ELF3, associated eQTLs, and structural variants. The bottom panel presents the corresponding LD heatmap. Arrow on ELF3 gene indicates the direction of its transcription. d Comparison of growth phenotypes among wild-type Arabidopsis, the Atelf3 mutant, and three independent transgenic lines overexpressing quinoa ELF3 (CqELF3). e Expression levels of ELF3 in highland and lowland populations of quinoa and C. berlandieri. f Comparison of total hypocotyl and root length (left) and plant height (right) between highland and lowland quinoa accessions. g Genotypic distribution of structural variants DEL1 and DEL2 in highland and lowland quinoa populations. Statistical significance in (e, f) was assessed using the Wilcoxon test

Among the identified eQTLs, the most significant one was a cis-eQTL located on chromosome 4 (Chr04:10,305,226–10,315,545), spanning 10.32 kb and strongly associated with ELF3 expression (P = 5.67e–47) (Fig. 6c). In Arabidopsis, ELF3 is a core component of the evening complex that represses circadian clock-related genes, thereby inhibiting hypocotyl elongation and delaying flowering [68]. Consistent with this role, Arabidopsis elf3 mutants exhibited accelerated growth and early flowering, while overexpression of quinoa ELF3 (CqELF3) in Arabidopsis delayed both growth and flowering (Fig. 6d), indicating functional conservation of ELF3 between the two species. We observed significantly lower expression of CqELF3 in highland quinoa accessions compared to lowland accessions, whereas no such difference was observed in C. berlandieri (Fig. 6e). To determine whether variation in CqELF3 expression contributes to morphological divergence between ecotypes, we analyzed phenotypic data from 133 quinoa accessions (68 highland and 65 lowland) [69]. In line with reduced CqELF3 expression, highland accessions displayed significantly longer hypocotyls and roots compared to lowland accessions (P = 0.0297; Fig. 6f), although mature lowland plants were significantly taller overall (P = 4.93e-06; Fig. 6f). Given that CqELF3 resides within a genomic region under selection in highland populations, these findings suggest that selection may have favored alleles associated with reduced CqELF3 expression, thereby contributing to longer hypocotyls in highland quinoa ecotypes.

Examination of the CqELF3 coding region revealed no segregating SNPs, indicating that the observed expression differences were likely driven by transcriptional regulation linked to the upstream cis-eQTL. This cis-eQTL was located 17.61 kb upstream of CqELF3, and within a 200-kb window surrounding the locus, we identified two structural variants: a 6,119-bp deletion relative to the lowland reference genome (DEL1) and a 11,409-bp deletion (DEL2). Linkage disequilibrium (LD) decay analysis revealed an extended LD span of 175.3 kb, while the LD heatmap showed strong genetic linkage across a 224-kb region encompassing CqELF3 and both deletions. Genotypic analysis further demonstrated that both DEL1 and DEL2 were predominantly found in highland quinoa populations (Fig. 6g). Computational prediction based on intact alleles from the lowland quinoa genome suggested that the DEL1 and DEL2 regions contain promoter-like sequences. To experimentally evaluate their regulatory potential, three candidate segments from each region, predicted to harbor promoter/enhancer elements, were tested using reporter-based assays. Both DEL1 and DEL2 regions exhibited measurable promoter activity, with varying strengths among segments, supporting the presence of functional cis-regulatory elements within the deleted intervals (Additional file 2: Fig. S15). Together, these findings suggest that DEL1 and DEL2 are closely associated with the regulation of CqELF3 expression, which may contribute to environmental adaptation through modulation of plant growth and development. Additionally, we identified two trans-eQTLs that showed the strongest associations with CqCO1 expression. These trans-eQTLs spanned 13,117 bp on chromosome 6 and 384,157 bp on chromosome 15 (Additional file 2: Fig. S16). Several segregating variants were detected within the coding and regulatory regions of these two loci. Although the causal variants underlying CqCO1 expression variation remain to be determined, the presence of multiple trans-eQTLs suggests that CO-related genes have been recurrently targeted during evolution to modulate photoperiodic responses across diverse developmental processes.

Discussion

Through extensive genome and transcriptome sequencing, this study provides a comprehensive genomic analysis of the genus Chenopodium, uncovering extensive genetic diversity across their genomes/subgenomes. Most importantly, we identified key genomic regions, functional genes, and regulatory variants that underly selection associated with adaptation to high- and low-altitude environments. These findings shed light on the molecular basis of altitude adaptation and highlight the role of parallel evolution in shaping the genetic architecture of crops across diverse environments.

Convergent and parallel selection occurs when different species or varieties independently evolve similar adaptive traits in response to comparable environmental pressures [61]. This phenomenon is frequently observed in crop domestication, where desirable traits repeatedly emerge across diverse species due to similar agricultural practices and environmental constraints [70–73]. Key traits commonly associated with convergent/parallel selection include resistance to biotic [74, 75] and abiotic stresses [71, 76], high-elevation adaptation [77, 78], and specialized metabolic pathways [79, 80]. Plant species, including many crops, have evolved complex mechanisms to adapt to altitudinal gradients. For example, changes in oxygen availability associated with elevation can alter plant sensitivity through ERFVII-mediated oxygen-sensing pathways [81]. In Heliosperma pusillum, parallel adaptation to lower elevations has been linked to enhanced plasticity of gene expression [82]. Among adaptive traits, adjustment of flowering time represents one of the most consistent mechanisms of altitude adaptation [77, 83]. Highland environments impose strong selection for rapid life-cycle completion and synchronization, making genes involved in photoperiod sensing and circadian regulation frequent targets of selection [84]. In rice, variation in CTB4a, COLD1, Hd1, Ghd7, and DTH8 contributes to adaptation across elevational and latitudinal gradients [85–88], while in maize, selection on ZmCCT, ZmPRR7, and ZmCOL9 facilitated adaptation of both highland teosintes and modern maize [89–91]. Similarly, adaptive variation in VRN and PPD loci underlies altitude and latitude adaptation in barley [92] and wheat [93]. Notably, many of these adaptive changes involve regulatory modifications rather than protein-coding alterations, enabling flexible developmental responses under heterogeneous environments [91, 93]. Adaptation to nutrient-poor highland environments further requires coordinated regulation of nutrient acquisition, utilization, and developmental timing. Key adaptive genes include nitrate and peptide transporters (NRT/PTR), phosphate transporters (PHT1), and central regulators such as PHR1 and SPX-domain proteins that control nutrient sensing and homeostasis [94, 95]. Genes regulating root architecture (e.g., ARF, LBD) enhance nutrient foraging capacity [96], while TOR and SnRK1 signaling pathways integrate nutrient availability with growth and metabolic regulation [97]. Consistent with these findings, our results show that genes identified under parallel selection associated with altitude adaptation include orthologs of Arabidopsis PTR2, involved in nutrient transport, and CONSTANS (CO), a key regulator of flowering time. Notably, genomic regions harboring these genes do not overlap with strong introgression signals, suggesting that these loci primarily reflect altitude-associated selection within each ecotype rather than introgression-mediated spread of adaptive haplotypes. The PTR2 gene homolog exhibited significant amino acid polymorphisms between highland and lowland quinoa, suggesting a potential role in quinoa adaptation to nutrient-poor high-altitude soils. Similarly, polymorphisms in both regulatory and coding regions of CO orthologs correlated with flowering time variation under different photoperiodic conditions. These observations suggest that Chenopodium species employ similar adaptive mechanisms, although not necessarily the same genes, to respond to altitudinal gradients, consistent with previous reports of convergent selection in plants adapting to comparable environmental conditions [70, 72].

Our transcriptome analysis identified significant cis- and trans-eQTLs regulating gene expression associated with environmental adaptation. The most significant cis-eQTL was linked to CqELF3, whose homolog in Arabidopsis is a core component of the evening complex known to regulate hypocotyl elongation and flowering [98]. In maize, ZmELF3.1 functions as a key regulator of photoperiod sensitivity and flowering time, contributing to the adaptation of maize from tropical to temperate environments [99]. Our data showed that CqELF3 expression was significantly reduced in highland quinoa compared to lowland accessions, correlating with elongated hypocotyls and roots, potentially reflecting adaptation to lower temperatures and reduced light at higher elevations. Consistent with these observations, two large deletions (DEL1 and DEL2) upstream of CqELF3 exhibited strong genetic linkage and were prevalent in highland quinoa populations, suggesting that these deletions may drive transcriptional changes that enhance adaptation to high-altitude conditions. Plants frequently adjust root morphology in response to harsh environmental conditions. For example, studies in Sorghum species have shown that root length and root radius increase with elevation in Sorghum nitidum, whereas S. arundinaceum maintains relatively constant root length while exhibiting reduced root radius with increasing elevation [100]. Such variation in root morphology between highland and lowland plants is likely influenced by differences in soil temperature, nutrient availability, and moisture conditions, which collectively shape root architecture to optimize resource acquisition under contrasting environments [101]. Additionally, we identified trans-eQTL hotspots linked to the regulation of CO genes, central regulators of photoperiodic flowering. These trans-eQTLs were particularly abundant, suggesting recurrent evolutionary recruitment of CO-related pathways for modulating developmental processes in response to altitude-specific photoperiodic variation. These findings are in agreement with prior research highlighting the evolutionary importance of CO in flowering time adaptation [19, 20].

Phylogenetic and population genetic analyses revealed distinct clustering patterns across Chenopodium species, reflecting genetic and ecological differentiation. Higher nucleotide diversity within the A subgenome was observed compared to the B subgenome, suggesting different evolutionary histories and selection pressures between subgenomes. Wild Chenopodium species, particularly hexaploid BBCCDD accessions, harbored high levels of genetic diversity, underscoring their potential as genetic reservoirs for breeding stress resilience quinoa varieties with improved yield stability [30, 31]. Furthermore, introgression between quinoa and other Chenopodium species were found to involve genomic regions encoding immunity, flowering, and stress response genes. Many of these genes were significantly differentially expressed between highland and lowland quinoa populations (Additional File 3), likely facilitating rapid adaptation and diversification by introducing novel genetic variation [46]. Demographic modeling also revealed insights into quinoa’s evolutionary history, with population expansion in highland quinoa historically coinciding with the Inca civilization, followed by a decline in lowland quinoa due to competition with newly introduced crops. The possible recent resurgence of lowland quinoa likely reflects renewed global interest and recognition of quinoa’s nutritional and agricultural value [9, 40].

Conclusions

This study uncovers extensive genetic diversity and distinct evolutionary trajectories across Chenopodium genomes, and identifies key genetic components and regulatory variants underlying quinoa’s adaptation to elevation. Although further work is required to determine whether and how these genes and their regulatory networks contribute to multiple aspects of altitude adaptation—including responses to variation in temperature, oxygen availability, light intensity, and humidity—and to identify additional genes and regulatory variants that may be revealed through long-read sequencing and expanded genomic analyses, the candidate genes and structural variants identified here provide valuable targets for marker-assisted breeding aimed at developing quinoa cultivars optimized for both high- and low-altitude cultivation.

Methods

Sample collection and sequencing

A total of 303 quinoa accessions were collected and grown in the greenhouse at Zhejiang A&F university. Mature leaves from healthy plants were used for genomic DNA extraction using the CTAB method. DNA concentrations were quantified with a Qubit Fluorometer (Thermo Fisher Scientific, USA), and DNA integrity was assessed via agarose gel electrophoresis and using an Agilent 2100 Bioanalyzer (Agilent Technologies, USA). High-quality DNA samples were then sheared into ~ 350 bp fragments using ultrasonic shearing (Covaris, Woburn, MA, USA). Fragmented DNA was end-repaired, A-tailed, and ligated to BGI-specific sequencing adapters. The ligated products were purified and amplified using PCR. The final libraries were size-selected using AMPure XP beads (Beckman Coulter, USA), assessed for quality using the Agilent 2100 Bioanalyzer, and quantified via qPCR. Libraries were sequenced on the MGISEQ-T7 platform.

For RNA sequencing, mature leaves were collected from plants at the same developmental stage. Samples were immediately flash-frozen in liquid nitrogen and stored at −80 °C until processing. Total RNA was extracted using the TRIzol reagent (Invitrogen, USA) following the manufacturer’s protocol. RNA quality and integrity were assessed using an Agilent 2100 Bioanalyzer and agarose gel electrophoresis. Samples with an RNA integrity number (RIN) > 8.0 were selected for library construction. mRNA was enriched using oligo(dT) magnetic beads, and the cDNA was fragmented and ligated to Illumina sequencing adapters. RNA-Seq libraries were sequenced on an Illumina NovaSeq 6000 platform.

Read mapping and variant calling

Raw sequencing reads were processed using fastp (v0.20.1) [102] to remove adapters and low-quality sequences. Cleaned reads were then mapped to the reference genome (CoGe ID: 33,827) [16] using BWA-MEM (0.7.17-r1188) [103] with default parameters. PCR duplicates were marked and removed using Picard (v2.26.10; https://broadinstitute.github.io/picard/). Variant calling was performed with GATK (v4.5.0.0) [104] following the best practices. The HaplotypeCaller module was used to generate raw variant calls, followed by joint genotyping across samples using the CombineGVCFs and GenotypeGVCFs modules. Variants were filtered using VariantFiltration, retaining only high-quality SNPs that met the following criteria: QUAL > 30, DP > 10, MQ > 40, QD > 2, FS < 60. The final high-confidence variant set was annotated using SnpEff (v5.0; https://pcingola.github.io/SnpEff/). Moreover, population-level large structure variations were identified using Smoove (https://github.com/brentp/smoove; v0.2.8) with short-read alignments and default parameters. Variants from all accessions, excluding those with unknown ploidy, were used for downstream analyses.

Phylogenetic and population genomic analyses

Bi-allelic SNPs with a missing data rate less than 5% and a minor allele count greater than three were retained for population genomic analyses. For phylogenetic analysis, only SNPs located at fourfold degenerate sites (1,554,832 SNPs) were used to construct a neighbor-joining tree based on the p-distance matrix. Principal component analysis was performed using Plink (v1.9; https://www.cog-genomics.org/plink/) with the full set of SNPs (8,576,501 SNPs). Population structure analysis was performed using ADMIXTURE [105]. The optimal K, which indicates the most likely number of clusters in the population, was calculated. Admixture analyses were run 20 times for each K value ranging from 2 to 15, using 10,000 randomly selected SNPs at fourfold degenerated sites.

Nucleotide diversity and deleterious mutations

Population diversity statistics, including π and F_ST, were computed using VCFtools (v0.1.16) [106]. A 50-kb sliding window with a 5-kb step size was used to estimate the π values across the genome. Pairwise F_ST values were calculated using the same window and step parameters to assess population differentiation between groups. LD values were measured within each population by calculating correlation coefficients (r²) of alleles using PopLDdecay (v3.41) [107].

To predict the functional effects of variants, we annotated the 8.58 million SNPs using SIFT 4G [108]. A quinoa-specific SIFT 4G database was constructed using the SIFT4G_Create_Genomic_DB tool, with UniRef90 (https://www.uniprot.org) serving as the reference protein set. Genome annotation for the quinoa reference genome V1 (CoGe ID: 33,827) was obtained from CoGe (https://genomevolution.org/coge/) and converted from GFF3 to Ensembl GTF format. SNPs with SIFT scores below 0.05 were identified as potentially deleterious nonsynonymous variants.

Identification of selective sweeps

To identify genomic regions under selection, the π ratio (π_group1/π_group2) and F_ST values were calculated in 50-kb sliding windows with a 5-kb step size across the quinoa genome. Genomic windows falling within the top 5% for both F_ST and π ratio were considered candidate regions under selection. Selection signals were further evaluated using the XP-CLR method [109], implemented with a high-density genetic map to improve the accuracy of genetic distance estimation [110].

Demographic history of Chenopodium species

We used the multiple sequentially Markovian coalescent (MSMC) approach [45] to infer historical changes of effective population size (N_e). Results of demographic modeling were scaled using a neutral mutation rate of 7e-9 mutations per base pair per generation, as estimated for higher plants [111], and a generation time of one year.

Introgression analysis with ABBA-BABA statistics

The f_d statistics [49] across the genome were calculated using the genomics_general Python scripts (https://github.com/simonhmartin/genomics_general), with 1 Mb windows (≥ 100 SNPs) and 100 kb steps. Each ancient population was tested as P3, BBCCDD was used as the outgroup (O), and each tetraploid population served as P2.

eQTL mapping

MatrixEQTL [112] was used to perform the eQTL analysis. To account for population stratification and reduce false-positive associations, principal components derived from genotype data using PLINK (https://www.cog-genomics.org/plink/) were included as covariates in the model. Expression data from 230 tetraploid accessions—selected from the 558 accessions with available SNP data—were used for the analysis. Genes with a median FPKM value of zero were excluded from downstream analysis. Gene expression values were further normalized using the quantile–quantile normalization (qqnorm) function in R. Significance threshold for associations between genetic variants and gene expression was determined using Bonferroni correction at α = 0.05, following established methods [113]. This corresponded to a Bonferroni-corrected P-value threshold of 1.90e–13 (− log₁₀(P) = 12.72). eQTLs located within 30 kb of a gene’s transcription start or stop site were classified as cis-eQTLs; all others were classified as trans-eQTLs.

LD decay was measured by calculating correlation coefficients (r²) for all pairs of SNPs within 1 Mb using PopLDdecay with parameters ‘-MaxDist 1000 -MAF 0.05 -Het 0.88 -Miss 0.999’. The stable r² value was considered as the background LD level, which was 0.20 in this study, corresponding to a physical distance of 175.30 kb. SNPs located in close physical proximity and associated with the expression of the same gene were grouped into eQTL blocks. Specifically, two adjacent SNPs were assigned to the same block if the physical distance between them was less than 175.30 kb, and blocks containing at least three significant SNPs were considered candidate eQTL blocks.

Creation of transgenic Arabidopsis plant

The Arabidopsis T-DNA mutant Atelf3 (At2g25930) was obtained from the Arabidopsis Resource Sharing Platform (https://www.arashare.cn). Arabidopsis wild-type Col-0 and transgenic plants were grown under long-day conditions (16 h light/8 h dark) at 23 °C/21 °C (day/night), 150 μmol photons m⁻² s⁻¹ light intensity, 70% relative humidity.

The coding sequences of CqCO1, CqPTR2, and CqELF3 were cloned into the pK7WG2D vector (CaMV 35S promoter, GFP marker, kanamycin resistance) via Gateway cloning (Invitrogen), using pDONR207 as the entry vector. BP and LR recombination reactions were performed following the manufacturer’s protocol. Sequence-verified constructs were introduced into Agrobacterium tumefaciens GV3101 by chemical transformation and subsequently into Col-0 plants via the floral dip method [114].

Seeds of transgenic plants were surface-sterilized in 10% bleach for 15 min, germinated on MS medium containing 50 μg mL⁻¹ kanamycin, and resistant seedlings were verified by PCR. Homozygous lines were selected for phenotypic analysis. Total RNA extraction, cDNA synthesis, and RT–qPCR was performed as described previously [115], using Arabidopsis actin encoding gene as the internal control. Reaction specificity was confirmed by negative controls and melting curve analysis. All transgenic lines were confirmed by gene expression and phenotypic analyses. Primers used here were listed in Additional file 1: Table S17.

Promoter activity assay

Target genomic sequences were cloned into the reporter vector pGreenII 0800-LUC. The resulting luciferase (LUC) reporter constructs were introduced into Agrobacterium tumefaciens strain GV3101 and transiently expressed in Nicotiana benthamiana leaves via agroinfiltration. Three days after infiltration, luciferase activity was imaged in vivo using a plant fluorescence imaging system following application of luciferin substrate to the abaxial leaf surface.

Authors’ contributions

C.J., X.S. and Z.F. conceived and supervised the project. J.H. and J.G. performed experiments. C.Z., X.L., Y.Z., X.S. and C.J. analyzed the data. C.L., B.X., T.L., W.W. and J.W. contributed to sample preparation and processing. X.L., X.S. and C.J. wrote the manuscript. Z.F. and H.L. revised the manuscript. All authors read and approved the final manuscript.

Funding

This research was supported by the National Natural Science Foundation of China (32470228), the Zhejiang Provincial Natural Science Foundation of China (LR23C150001; LR25C020001), and the ZJU-Cornell Joint Seed Fund awarded to C.J.

Data availability

The RNA sequencing and genome resequencing data generated in this study have been deposited in the Genome Sequence Archive of the National Genomics Data Center (CNCB-NGDC; https://ngdc.cncb.ac.cn) under accession numbers PRJCA038182 [116] and PRJCA044391 [117], respectively. The genome resequencing data are also available in the China National GeneBank Database (CNGBdb; https://db.cngb.org/) under accession no. CNP0007821 [118]. Previously published data used in this study are available in the NCBI Sequence Read Archive (SRA) under BioProject PRJNA673789 [119]. The code used for the bioinformatics analyses in this study is publicly available on GitHub at https://github.com/cj334new/Chenopodium_Pop_Gen under the MIT License [120]. The code has also been archived and is available on Zenodo (10.5281/zenodo.20371069) [121].

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

Zhangjun Fei is an Editorial Board Member for Genome Biology and the Guest Editor of the Genomics and pangenomics for understanding crop domestication collection but was not involved in the editorial process of this manuscript.