Integrated metabolite profiling and transcriptomic analysis reveal mechanisms underlying quinoa inflorescence color

Abstract

Background

Quinoa exhibits striking inflorescence color variation, with red inflorescences (RF) characterized by higher anthocyanin accumulation than green inflorescences (GF). However, the molecular mechanisms underlying this phenotype remain insufficiently understood.

Results

Integrated metabolomic and transcriptomic analyses revealed substantial differences in the anthocyanin biosynthetic pathway between RF and GF accessions. RF plants showed marked upregulation of key structural genes, including ANS, 3GGT, and malonyltransferases (MaTs). Notably, MYB and bHLH transcription factors, including homologs of MdMYBA1, AtPAP1/2, and ZmbHLH125, were identified as master regulators, while the WD40 was significantly upregulated, supporting formation of the MYB-bHLH-WD40 (MBW) complex. Genes encoding transporters such as multidrug resistance-associated proteins (MRPs) and multidrug and toxic compound extrusion (MATE) proteins were also significantly upregulated, supporting the enhanced accumulation of malonylated cyanidin and petunidin derivatives. Regulatory analysis further revealed that abscisic acid (ABA) signaling appeared to indirectly influence anthocyanin biosynthesis by modulating transcription factor expression. Genomic analysis further showed that quinoa, as an allotetraploid species, possesses multiple paralogs of anthocyanin-related genes, potentially enhancing metabolic plasticity.

Conclusions

This study elucidates the metabolic, transcriptional, and genomic bases of inflorescence color formation in quinoa. The findings highlight that red panicle formation is driven by coordinated upregulation of biosynthetic genes (ANS, 3GGT, MaTs), transporters (MRP, MATE), and MBW transcription factors, with ABA signaling playing a key modulatory role. The findings provide valuable insight for improving pigment quality and advancing molecular breeding in this crop.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12870-026-08755-9.

Introduction

Quinoa (Chenopodium quinoa Willd.) is a nutrient-rich pseudocereal that has attracted substantial global interest due to its high-quality protein, balanced amino acid composition, and richness in bioactive compounds such as polyphenols, flavonoids, and saponins [1, 2]. Beyond its value as a grain crop, quinoa exhibits remarkable phenotypic diversity in inflorescence coloration, which not only contributes to its ornamental appeal but may also reflect different adaptation to environmental conditions [3, 4]. However, systematic studies on anthocyanin biosynthesis and its regulatory networks in quinoa remain scarce, particularly with respect to the metabolic and transcriptional differences among varieties exhibiting distinct panicle colors [5]. In parallel with growing consumer interest in functional foods, quinoa has received increasing attention due to its antioxidant, anti-inflammatory, and glycemic-regulation properties, further underscoring the importance of understanding pigment-related metabolic pathways [6]. Panicle color is one of the key phenotypic traits of quinoa, typically including white, yellow, red, and black hues, and is primarily determined by the accumulation of secondary metabolites such as anthocyanins, flavonoids in some varieties,, and betalains in others [7].

Anthocyanins are water-soluble pigments widely distributed in the plant kingdom, where they impart vibrant coloration to flowers, fruits, and vegetative tissues, while also contributing to growth, development, and responses to both biotic and abiotic stresses [8]. Current research indicates that anthocyanin production involves four major stages: the initiation and core reactions of the flavonoid pathway, anthocyanin formation, structural modification, and vacuolar transport [9]. Precursors such as 4-coumaroyl-CoA are synthesized via the phenylpropanoid pathway through PAL, C4H, and 4CL; CHS, CHI, and F3H drive the formation of dihydroflavonol intermediates. These compounds are then converted into stable anthocyanins through sequential catalysis of DFR, ANS, and UFGT [10]. Additional glycosylation and acylation steps, mediated by enzymes such as 3GGT and 5AT, generate diverse pigment structures [11]. Finally, the resulting anthocyanins are transported into the vacuole by GST, MRP, and MATE transporters [12].

The biosynthesis of anthocyanins in plants is regulated by both environmental cues and endogenous transcriptional networks. Transcriptional regulation primarily involves the coordinated activation of structural genes by MYB, bHLH, and WD40 transcription factors, which together form the conserved MYB–bHLH–WD40 (MBW) complex [13]. This complex functions as a central regulatory module that integrates developmental and environmental signals to activate key genes such as DFR and ANS, thereby orchestrating the spatial and temporal accumulation of anthocyanins [14].

In addition, plant hormones exert important modulatory effects on pigmentation. Auxin has been reported to exert concentration-dependent effects, enhancing anthocyanin accumulation at low levels while inhibiting it at high levels [15]. In contrast, abscisic acid [16], cytokinins [17], salicylic acid [18], and jasmonic acid [19] generally act as positive regulators. For instance, the rice SAUR gene, SAUR39, promotes expression of anthocyanin biosynthetic genes by suppressing auxin synthesis and transport, thereby functioning as a positive regulator independent of the auxin concentration effect [20, 21]. Similarly, upregulation of the ABA biosynthesis gene NCED3 enhances PAL transcription in rice [22]. Meanwhile, ABA induces MdABI5 in apple and ZmVP1 in maize, and these transcription factors further promote anthocyanin accumulation by regulating MYB1 and ZmC1, respectively [23].

Despite these advances, our understanding of quinoa panicle coloration remains constrained by two major knowledge gaps. First, the pigment profiles of quinoa varieties with different panicle colors - particularly red and green types, including the specific classes and quantities of anthocyanins and flavonoids - are still insufficiently characterized. Second, the transcriptional regulatory networks controlling pigment biosynthesis have yet to be fully elucidated. To address these limitations, this study integrates metabolomic and transcriptomic analyses to systematically dissect the metabolic basis and transcriptional regulatory networks underlying panicle color formation in red and green quinoa. The work not only elucidates the molecular mechanisms underlying panicle color differentiation but also provides critical molecular targets and a theoretical foundation for the molecular design breeding of quinoa varieties with enhanced nutritional attributes and tailored pigment profiles, thereby contributing to the industrial advancement of this high-value functional crop.

Materials and methods

Plant materials

Two independently bred advanced inbred lines of quinoa - the green panicle accession PI-614,888 (GF) and red panicle accession AUHQ-114 (RF) - obtained from the National Semi-arid Agricultural Engineering Technology Research Center (China) were grown in a phytotron at 25/18°C with a photoperiod of 16/8 h (light/dark). Seeds of GF and RF were planted in separate pots (15 cm × 20 cm) containing a crop-specific soil mix(Pindstrup^®, Denmark). Inflorescences were collected at 30 days after flowering, immediately frozen in liquid nitrogen, and stored at -80°C for subsequent pigment analysis and RNA sequencing. Three independent biological replicates were collected for each accession.

Identification of quinoa panicle color

Panicle samples from RF and GF were collected to quantify the color using a portable chromatic meter (CR-10plus, Konica Minolta, Japan) according to our previous study [24]. Color parameters (L, a, and b) were recorded, and total color difference (ΔE) values were calculated relative to the GF panicles using the formula ΔE=[(ΔL)²+(Δa)²+(Δb)²]^1/2 with GraphPad Prism 8 software. According to established criteria, ΔE values ≤ 2, indicate color differences imperceptible to the human eye; values > 2 to < 6 indicate perceptible but moderate differences, and values ≥ 6 indicate clearly distinguishable color differences.

Measuring total chlorophyll, carotenoid, anthocyanin and flavonoid contents

Total chlorophyll and carotenoid contents of quinoa inflorescences were measured following the procedure described in [25] with minor modifications. Briefly, 0.1 g of inflorescence sample was homogenized with 5 mL of extraction solution (ethanol: acetone = 1:2). The sample was then put in darkness until the quinoa inflorescences became colorless. Absorbance was measured at 663 nm and 645 nm using a spectrophotometer, and the contents were calculated according to the formulas cited in the literature.

Total anthocyanin content was quantified according to an established protocol [26] with minor modifications. Approximately 0.1 g of fresh inflorescence tissue was ground and mixed with 10 mL of 1% (v/v) HCl-methanol solution. The mixture was incubated at room temperature in darkness for 24 h. After centrifugation, the absorbance of the supernatant was measured at 530, 620, and 650 nm. The total anthocyanin content was expressed as cyanidin-3-glucoside equivalents (mg/g FW).

Total flavonoid content of quinoa inflorescence was determined by using a colorimetric assay method described in a previous study [27]. Approximately 0.1 g of fresh inflorescence tissue was ground in liquid nitrogen and extracted with 5 mL of 70% aqueous ethanol. The homogenate was centrifuged at 12,000 rpm for 10 min at 4 °C. A 0.5 mL aliquot of the supernatant was sequentially mixed with 2 mL of distilled water, 0.15 mL of 5% NaNO₂ solution, and after 6 min, 0.15 mL of 10% AlCl₃ solution. Following an additional 6 min, 2 mL of 4% NaOH solution was added, and the final volume was adjusted to 5 mL with distilled water. After incubation at room temperature for 15 min, the absorbance was measured at 510 nm. The total flavonoid content was calculated based on a standard curve of rutin and expressed as rutin equivalents (mg RE/g FW).

Metabolomic analysis

Freeze-dried quinoa panicle tissue (0.1 g) was ground into a powder using a ball mill at 30 Hz for 1.5 min. From each sample, 50 mg of powder was transferred to a microcentrifuge tube and extracted with 500 µL of extraction solvent consisting of 50% methanol–water supplemented with 0.1% hydrochloric acid. The suspension was vortex-mixed for 5 min, sonicated for 5 min, and centrifuged at 12,000 rpm for 3 min at 4 °C. The resulting supernatant was collected, and the extraction procedure was repeated once. Supernatants from both extractions were pooled and passed through a 0.22 μm membrane filter prior to analysis. Metabolite profiling was performed using an UPLC-ESI-MS/MS system (UPLC, ExionLC™ AD; MS, QTRAP^® 6500+). Chromatographic separation was achieved on an ACQUITY BEH C18 column (1.7 μm, 2.1 × 100 mm). The mobile phase was composed of solvent A, (water supplemented with 0.1% formic acid) and solvent B (acetonitrile supplemented with 0.1% formic acid). The gradient elution program was set as follows: 95:5 (A: B, v/v) at 0 min, 50:50 at 6.0 min, 5:95 at 12.0 min, and returned to 95:5 at 14.0 min. Samples were injected at a volume of 4 µL, with a constant flow rate of 0.40 mL/min and the column was operated at 40 °C.

The UPLC system was coupled to a triple quadrupole–linear ion trap mass spectrometer equipped with an electrospray ionization (ESI) source. In brief, the ESI source operated at 550 °C with an ion spray voltage of 5500 V in positive ion mode and a curtain gas pressure of 35 psi. In the Q-Trap 6500 + system, metabolites were detected using optimized declustering potential (DP) and collision energy (CE) parameters, and data acquisition was performed in multiple-reaction monitoring (MRM) mode. Differentially accumulated metabolites were identified based on a variable importance in projection (VIP) ≥ 1 and fold change (FC) ≥ 2 or ≤ 0.5.

Transcriptome analysis

Total RNA was isolated from quinoa panicle tissues using TRIzol reagent kit (Invitrogen, Carlsbad, CA, USA) in accordance with the manufacturer’s instructions. RNA integrity and quality were evaluated using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA) and further examined by RNase-free agarose gel electrophoresis. For library construction, polyadenylated RNA was isolated from total RNA using Oligo(dT) magnetic beads, and ribosomal RNA was depleted using the Ribo-Zero™ Magnetic Kit (Epicentre, Madison, WI, USA) to minimize rRNA contamination. The purified RNA was subsequently fragmented into short segments using fragmentation buffer and reverse-transcribed into first-strand cDNA with random primers. Second-strand cDNA synthesis was performed using DNA polymerase I, RNase H, dNTPs, and the supplied reaction buffer. The resulting double-stranded cDNA fragments were purified using the QIAquick PCR extraction kit (Qiagen, Venlo, The Netherlands), followed by end repair, poly(A) tail addition, and ligation of Illumina sequencing adapters. Adapter-ligated products were size-selected by agarose gel electrophoresis, amplified by PCR, and finally sequenced on an Illumina HiSeq 2500 platform at Gene Denovo Biotechnology Co. (Guangzhou, China). Raw sequencing reads were subjected to quality control using Trimmomatic (v0.39) to remove adapter sequences, reads containing more than 10% ambiguous bases (N), and reads in which more than 50% of bases had a quality score ≤ 20. Additional low quality read filtering was performed using FASTP (v0.18.0) with the parameters Q20, U50, N15, and L150 [28]. High-quality clean reads were then aligned to the quinoa reference genome (https://www.ncbi.nlm.nih.gov/genome/?Term=Chenopodium+quinoa+Willd) using HISAT2 software [29], which is based on the Bowtie2 alignment algorithm [30]. Gene annotation was subsequently performed based on the reference genome.

Differential expression analysis

Gene expression levels were estimated using the fragments per kilobase of transcript per million mapped reads (FPKM) metric [31]. Gene FPKM values were calculated using Cuffquant and Cuffnorm (version 2.2.1) [32]. DESeq2 [33] facilitated the identification of differentially expressed genes (DEGs) employing two distinct criteria: a fold change of ≥ 2 or ≤ 0.5, complemented by an adjusted p-value (q-value) of ≤ 0.01. Functional annotation was performed using WEGO software for Gene Ontology (GO) enrichment [34], while pathway analysis was conducted with the KEGG database to identify enriched biological pathways [35].

qRT-PCR validation

For cDNA synthesis, 1 µg of total RNA treated with DNase I was reverse-transcribed into first-strand cDNA using the PrimeScript RT reagent Kit with gDNA Eraser (TaKaRa, Japan) to remove genomic DNA contamination. The synthesized cDNA was diluted ten-fold with nuclease-free water and used as the template for qRT-PCR. All primers used for qRT-PCR (sequences listed in Table S1) were designed with Primer Premier 5.0 software according to the following criteria: amplicon length of 80–200 bp, annealing temperature of 58–62 °C, and GC content of 40–60%. The specificity of the amplified products was further verified by melt-curve analysis after qRT-PCR. Each 20 µL qRT‑PCR reaction contained 10 µL of 2× SYBR Premix Ex Taq II (Tli RNaseH Plus), 0.4 µL each of forward and reverse primers (final concentration 0.2 µM), 2 µL of diluted cDNA template, and 7.2 µL of nuclease‑free water. The thermal cycling protocol consisted of an initial denaturation at 95 °C for 30 s, followed by 40 cycles of 95 °C for 10 s, 60 °C for 20 s, and 72 °C for 15 s, with fluorescence acquisition at the annealing/extension step [24]. Three biological replicates, each with three technical replicates, were included for each sample. The quinoa actin gene served as the internal reference for normalization [24]. Relative gene expression levels were calculated using the 2^⁻ΔΔCt method, and statistical comparisons between the red- and green‑flowering panicle groups (RF and GF) were performed with GraphPad Prism 8 software [36].

Statistical analysis

All quantitative data in this study are presented as means ± standard deviation (SD). Comparisons between two groups (e.g., GF vs. RF accessions) were performed using a two-tailed Student’s t-test. The significance level was set at P < 0.05. All statistical analyses were performed using GraphPad Prism 8. The expression heatmap in this study was generated using TBtools [37].

Results

Analysis of inflorescence color phenotype and pigment content

Distinct color differences between the two quinoa varieties became evident at the seedling stage and became more pronounced after panicle initiation. At maturity, RF plants produced rose-pink panicles, whereas GF plants exhibited green panicles (Fig. 1A-C). RF fruits displayed pink coloration with deep purple perianths, while GF fruits remained light green with green perianths (Fig. 1D). Color contrast quantified by ∆E values showed that, using the GF panicle as baseline (∆E = 0), RF panicles reached a ∆E of 81.45, reflecting a striking chromatic divergence (Fig. 1I).

Fig. 1.

Panicle color phenotype and pigment content in quinoa. A–D Panicle phenotypes of the red-colored (RF) and green-colored (GF) quinoa varieties at maturity; E–H Quantification of total chlorophyll, carotenoids, anthocyanins, and flavonoids in RF and GF panicles. Different lowercase letters indicate statistically significant differences between RF and GF for each pigment (P < 0.05); I Color difference values (L, a, b, and ΔE) between RF and GF panicles

To identify the specific pigments responsible for the panicle color variation, we quantified total chlorophyll, carotenoids, anthocyanins, and flavonoids in the panicles of both RF and GF varieties. Total chlorophyll and carotenoids contents were significantly lower in RF than in GF (P < 0.05; Fig. 1E and F), whereas flavonoid content was significantly higher in RF (P < 0.05; Fig. 1H). Notably, anthocyanins were detected exclusively by the spectrophotometric assay in RF (Fig. 1G), suggesting its content in GF is below the detection limit of this method. These results indicate that anthocyanins are the primary determinants of the red panicle phenotype, with flavonoids likely serving as co-pigments that further modulate coloration.

Identification of differential metabolites

To characterize anthocyanin accumulation, 108 anthocyanin-related metabolites were profiled using LC–MS/MS in multiple reaction monitoring (MRM) mode. A total of 32 compounds were detected across the two varieties, encompassing petunidin, delphinidin, cyanidin, pelargonidin, malvidin, and peonidin derivatives as well as several flavonoids (Fig. 2A). Metabolites showing > 2-fold variation between RF and GF were defined as differentially accumulated.

Fig. 2.

Differential metabolite profiling in red (RF) and green (GF) panicle varieties of quinoa. A Heatmap of relative metabolite abundance. Hierarchical clustering of significantly differentially accumulated metabolites. Rows represent metabolites, and columns represent biological replicates of RF and GF. The color scale indicates relative abundance. B Venn diagram of differential metabolites. Shared and unique differentially accumulated metabolites between RF and GF, with 10 metabolites common to both groups. C Log2 fold change (FC) of major metabolite categories. Log2FC values of different metabolite categories between RF and GF. Red bars indicate higher accumulation in RF and green bars indicate higher accumulation in GF

Ten differentially accumulated metabolites were shared between RF and GF, and no RF-specific anthocyanin species were detected, suggesting that observed color variation arises from quantitative differences in common anthocyanin types rather than qualitative presence/absence (Fig. 2B).

Among the up-accumulated metabolites in RF, cyanidin-3-O-(6-O-malonyl-β-D-glucoside), petunidin-3-O-(6-O-malonyl-β-D-glucoside), naringenin, and afzelin exhibited fold changes ranging from approximately 1.03 to 2.03 (Fig. 2C). Conversely, delphinidin-3-O-rutinoside-5-O-glucoside and procyanidin B3 were significantly reduced in RF (down-regulated by 2.22-fold and 1.06-fold, respectively). These shifts highlight a re-allocation of metabolic flux toward specific anthocyanin derivatives in RF panicles.

Identification of differentially expressed genes

To further determine which gene expression changes led to the differences in anthocyanin content, we performed transcriptome sequencing on quinoa panicles of different colors. The complete list of all differentially expressed genes (DEGs) has been compiled into supplemental Table S2. Furthermore, details of the quality control for the sequencing data can be found in supplemental Table S3. We further screened the top 50 most significantly differentially expressed genes (Top 50 DEGs) and visualized their expression patterns in a heatmap Fig. S1. The results revealed substantial transcriptional divergence between RF and GF panicles, with 8,087 DEGs identified—5,076 up-regulated and 2,911 down-regulated in RF (Fig. 3A). Volcano plot analysis confirmed the strong enrichment of highly significant DEGs (Fig. 3B). GO and KEGG enrichment analyses further indicated that DEGs were predominantly associated with phenylpropanoid biosynthesis, cell wall organization, stress responses, and hormone signaling pathways. Hierarchical clustering showed clear separation of RF and GF transcriptomic profiles (Fig. 3C).

Fig. 3.

Identification and characterization of differentially expressed genes (DEGs) between RF and GF quinoa panicles. A DEG statistics. Number of up-regulated (red bars) and down-regulated (blue bars) DEGs in RF relative to GF. B Volcano plot of DEGs. Distribution of DEGs showing log2 fold change (log2FC) (x-axis) and -log10 adjusted p-value (FDR < 0.05) (y-axis). Yellow dots represent significantly up-regulated genes (FDR < 0.05), red dots represent significantly down-regulated genes (FDR < 0.05), and blue dots represent genes with no significant difference. C Hierarchical clustering heatmap of DEGs. Expression patterns of DEGs across RF and GF samples. Rows represent genes and columns represent biological replicates. D Gene Ontology (GO) classification. Distribution of DEGs across the three main GO categories (Biological Process, Molecular Function, Cellular Component). Yellow bars denote up-regulated genes and blue bars represent down-regulated genes. E GO enrichment analysis. Top 20 significantly enriched GO terms displayed by the rich factor (x-axis). Dot size corresponds to gene number, and color represents the enrichment significance (-log10(q-value)). F KEGG pathway enrichment bar plot. Top 20 significantly enriched KEGG metabolic pathways. The x-axis represents the number of differentially expressed genes in the pathway, and the y-axis represents the pathway names. G KEGG pathway enrichment analysis scatter plot. Top 20 KEGG pathways ranked by rich factor. Dot size indicates gene count and color represents the enrichment significance (-log10(q-value))

In the Biological Process (BP) category, “cell wall organization” (GO:0071554) and “metabolic process” (GO:0008152) were significantly enriched in RF (Fig. 3D and E), suggesting coordinated remodeling and secondary metabolite accumulation during red panicle formation. Molecular Function (MF) analysis revealed enrichment of genes related to “catalytic activity” (GO:0003824) and “binding” (GO:0005488) likely reflect activation of key phenylpropanoid enzymes (e.g., PAL, C4H, CHS). Enrichment of “cell membrane” (GO:0016020) and “cell wall” (GO:0005618) in the Cellular Component (CC) category pointed to the potential roles in metabolite transport and deposition.

KEGG pathway analysis highlighted phenylpropanoid biosynthesis and starch/sucrose metabolism as the most significantly enriched pathway in RF (Fig. 3F and G). Hormone-related pathways, particularly auxin and ABA signaling, also showed significant enrichment, implying hormonal modulation of anthocyanin biosynthesis in RF. Notably, altered expression of lipid metabolism genes further suggests additional contributions via effects on membrane dynamics and oxidative responses.

Functional analysis of structural genes in the anthocyanin biosynthesis and transport

To precisely identify the specific steps in the phenylpropanoid pathway that differ between RF and GF quinoa during anthocyanin synthesis, we analyzed the homologous structural genes across the pathway.

In the upstream stage, only one homolog was identified for PAL, C4H, and 4CL. In the second stage, four homologs of CHS, a key rate-limiting enzyme, were detected in quinoa (Fig. 4B). Phylogenetic analysis indicated that these CHS homologous genes clustered into two distinct clades, likely reflecting genome duplication in allotetraploid quinoa. Similarly, five CHI homologs were identified (Fig. 4C). By contrast, genes involved in the third stage– F3’H, F3’5’H, and F3H– did not undergo expansion, with only one homologous detected for each. During the fourth stages of anthocyanin synthesis, six DFR homologs (Fig. 4D), two LAR homologs, one ANS homolog, and five UFGT homologs (Fig. 4E) were identified. In addition, three homologous for 3GGT and two of anthocyanin malonyltransferases (MaTs) were identified, whereas only one homolog of 5GT were present.

Fig. 4.

Anthocyanin biosynthesis-related homologous genes and their expression patterns in quinoa. A Heatmap of key anthocyanin biosynthesis gene expression. Gene IDs in bold red font indicate that the expression level of the gene in RF is more than 2-fold higher than in GF. Gene IDs in green font indicate that the expression level of the gene in GF is higher than in RF. Furthermore, gene IDs in bold black font indicate that the gene exists as multiple homologous copies in the quinoa genome. B–E Phylogenetic trees of CHS, CHI, DFR, and UFGT gene families, respectively

Regarding anthocyanin transport, quinoa possessed homologous genes for GST, MRP, and MATE transporters, while no homolog of the BTL-like transporter was found. This suggests that anthocyanin sequestration in quinoa is predominantly mediated by GST- and MRP- dependent transport, as well as MATE-mediated transmembrane transport.

Expression profiling revealed moderate upregulation of upstream phenylpropanoid genes (PAL and C4H) in RF, with 4CL (XP_021755918.1) showing a 2.34-fold higher expression than in GF (Fig. 4A). In contrast, a CHS homolog (XP_021762432.1) exhibited markedly higher expression in GF, likely activating the flavonol synthesis branch. Coordinated upregulation of glycosyltransferase UFGT gene (XP_021725781.1) may enhance pigment stability through glycosylation modifications. Notably, the ANS homolog (XP_021769194.1) showed strongly elevated expression in RF, approximately 3.28-fold higher than in GF, identifying it as a central contributor to the red phenotype.

DFR homologs (e.g., XP_021759656.1) were scarcely expressed in both varieties, potentially directing metabolic flux away from colorless proanthocyanidins toward anthocyanin accumulation. The leucoanthocyanidin reductase (LAR) gene (XP_021761219.1) showed substantially higher expression in GF, suggesting a diversion of intermediates toward proanthocyanidin synthesis in green panicles, thereby reducing flux toward anthocyanin production.

All 3GGT homologous genes were more highly expressed in RF. with XP_021758620.1 showing a 2.73-fold increase. Similarly, both MaT homologous genes elevated expression in RF (2.01- and 1.64-fold higher), promoting the accumulation of glycosylated and malonylated anthocyanins characteristic of red coloration. All four MRP homologs were expressed at least 1.5-fold higher in RF, including XP_021763157.1, which showed a 3.26-fold increase. The MATE homologs exhibited a similar pattern, with 2.21-fold higher expression in RF. In contrast, GST gene expression was lower in RF, suggesting a distinct balance among transporter types.

Integrated KEGG analysis revealed synergistic enrichment of interconnected pathways—phenylpropanoid, flavonoid, and anthocyanin biosynthesis—indicating coordinated activation of the entire metabolic cascade from universal precursors to colored end products in RF panicles. Further expression profiling pinpointed a metabolic flux re-direction toward anthocyanin production, characterized by strong upregulation of ANS alongside relatively low expression of DFR and LAR, coupled with elevated expression of modification (3GGT, MaT) and transporter (MRP, MATE) genes, which collectively enable efficient synthesis, stabilization, and vacuolar sequestration of pigments.

Functional analysis of transcription factors in anthocyanin biosynthesis and their role in quinoa panicle color differentiation

A total of 17 MYB-type transcription factors, 5 bHLH-type transcription factors, and 1 WD40 protein -encoding gene homologous to known anthocyanin regulators were identified in quinoa. Expression analysis revealed that the homologs of positively MYB regulators – including quinoa homologs of MdMYBA1, RrMYB5, VvMYB6, SlMYB12, OsC1, ZmC1, AtPAP1, and AtPAP2– were consistently higher in RF than in GF. For example, the MdMYBA1 homolog (XP_021767009.1) showed a 2.34-fold increase in RF, and another key MYB homolog (XP_021750053.1) representing OsC1, ZmC1, AtPAP1, and AtPAP2, showed a 2.18-fold increase (Fig. 5A). In contrast, homologs of MYB repressors (e.g. quinoa homologs of MdMYB6, SmMYB35, FtMYB3, PpMYB20, PtrMYB93, and AmMYB308), showed similar or slightly higher expression in RF whereas the homolog of CmMYB012 (XP_021770126.1) showed drastically lower expression in RF (only about 3% of its expression level in GF) (Fig. 5A), indicating a reduction in repression. Collectively, these patterns suggest that MYB factors-, particularly homologs of MdMYBA1, OsC1, ZmC1, AtPAP1, AtPAP2, and CmMYB012, -play pivotal roles in promoting red panicle formation.

Fig. 5.

Expression patterns of transcription factors and hormone biosynthesis-related genes involved in anthocyanin biosynthesis in quinoa. A Heatmap of key transcription factors and hormone biosynthesis-related gene expression involved in anthocyanin biosynthesis. Gene IDs in red font indicate 1–2-fold higher expression in RF than in GF; gene IDs in bold red font indicate more than 2-fold higher expression in RF than in GF; gene IDs in green font indicate higher expression in GF than in RF. B Phylogenetic tree of MYB transcription factors. C Phylogenetic tree of bHLH transcription factors. D Phylogenetic tree of WD40 proteins

Among bHLH-type transcription factors, all the homologs of AtTT8 and ZmbHLH125 (XP_021748530.1, XP_021767015.1, XP_021761552.1, XP_021756333.1) were significantly upregulated in RF (Fig. 5A), supporting their role in activating key anthocyanin pathway genes. Similarly, the homolog of WD40 (XP_021773389.1) was expressed approximately 3.15-fold higher in RF (Fig. 5A), further highlighting the importance of the MYB–bHLH–WD40 regulatory complex.

Based on phylogenetic analysis, quinoa MYB genes cluster with known anthocyanin regulatory factors (e.g., AtPAP1/PAP2 in Arabidopsis, OsC1 in rice, ZmC1 in maize, MdMYB6 in apple) and are likely involved in regulating betalain/anthocyanin biosynthesis, with significant expansion of this gene family through gene duplication (Fig. 5B). The bHLH factors in quinoa are most closely related to ZmBHLH125 in maize, forming an evolutionary clade with AtTT8 in Arabidopsis and LhbHLH2 in tobacco, suggesting a recent common ancestor (Fig. 5C). The quinoa WD40 gene (XP_021773389.1) clusters closely with VvWDR2 in grape and forms a highly supported monophyletic branch with AtTTG1 in Arabidopsis and MdTTG1 in apple, belonging to a conserved subfamily in dicot plants (Fig. 5D). This analysis supports the evolutionary relationships of the three types of transcription factor genes in quinoa that may form the “MBW complex,” providing a basis for elucidating the regulatory network of pigment synthesis.

Roles of plant hormones in the color differentiation of quinoa inflorescences

We identified two homologs of OsNCED3 (a key enzyme in abscisic acid biosynthesis), two of ZmVP1, and three of MdABI5 in quinoa, all exhibiting higher expression in RF than in GF (Fig. 5A). The ZmVP1 homologs (XP_021763689.1 and XP_021765560.1) were upregulated 4.37-fold and 2.23-fold, respectively, in RF (Fig. 5A). The MdABI homolog, XP_021764816.1, showed a 2.79-fold increase (Fig. 5A). These results suggest that ABA signaling plays a major role in promoting anthocyanin biosynthesis in quinoa.

Two homologs of OsRR6– XP_021737823.1 and XP_021758238.1 were also identified, both showing nearly twofold higher expression in RF, indicating that cytokinin may enhance anthocyanin accumulation (Fig. 5A). Similarly, two homologs of OgUBC1– XP_021764938.1 and XP_021760361.1 which responds to salicylic acid, was significantly upregulated in RF, suggesting a possible role for SA in color formation.

Four homologs of OsABA8 ox1, involved in jasmonate-related pathways, showed markedly higher expression in RF, with one homolog (XP_021720820.1) exhibiting an 11.58-fold increase (Fig. 5A). These findings indicate that JA may further promote anthocyanin accumulation, with OsABA8ox1 serving as a key regulator.

Integrated metabolomic and transcriptomic analyses elucidate the anthocyanin biosynthesis, transport, and regulatory pathways

Integrated metabolomic and transcriptomic analysis (Fig. 6) revealed widespread differential expression across anthocyanin biosynthesis, transport, and regulatory pathways. Key structural genes - including 4CL, F3H, ANS, 3GGT, and MaTs - were strongly upregulated in RF, consistent with the elevated accumulation of anthocyanin derivatives such as cyanidin-3-O-(6-O-malonyl-beta-D-glucoside), petunidin-3-O-(6-O-malonyl-beta-D-glucoside), and cyanidin-3-O-sophoroside.

Fig. 6.

Proposed model of the anthocyanin biosynthesis, transport, and regulatory pathways in quinoa. () represents intermediates or end products in the anthocyanin biosynthesis and transport pathways. Arrows (→) indicates positive regulation or downstream conversion, and blunt-ended lines (⊥) indicate inhibition. Metabolites highlighted in red correspond to differentially accumulated metabolites (DAMs) between the RF and GF. For gene expression: gene IDs in red indicate 1–2-fold higher expression in RF; bold red indicates >2-fold higher expression in RF; black indicates similar or slightly higher expression in GF. Homologous gene families are indicated by bold black gene IDs

Genes involved in vacuolar transport (MRP and MATE), showed significantly higher expression in RF, suggesting enhanced transport efficiency and greater vacuolar accumulation of anthocyanins in red panicles. Several crucial transcription factors (e.g., quinoa homologs of ZmbHLH125, SlMYB012, AtPAP1/2, SlANT1) also showed more than twofold higher expression in RF. These regulators may enhance both biosynthetic gene expression and transporter activity, collectively promoting red coloration. Given the strong upregulation of transporter genes and related transcription factors in RF, we propose that anthocyanin transport may play a more decisive role than biosynthesis itself in determining quinoa panicle coloration.

Among hormone-responsive genes, ABA-responsive homologs showed the most consistent and pronounced upregulation in RF, suggesting that ABA likely the dominant hormonal regulator of anthocyanin accumulation in quinoa.

Validation of RNA-seq data by RT-qPCR assay

To verify the RNA-seq results, we conducted qRT-PCR on four structural genes involved in flavonoid biosynthesis, ANS (ncbi_110686439), CHI (ncbi_110721033), FHT (ncbi_110724781), CHS (ncbi_110702057) and four transcription factors related to anthocyanin regulation, MYB93 (ncbi_110730159), MYB12 (ncbi_110734325), bHLH125 (ncbi_110702384) and bHLH25 (ncbi_110706298). As shown in Figs. 7, qRT-PCR expression trends were consistent with transcriptome data, confirming the reliability of the RNA-seq-based differential expression patterns between RF and GF quinoa.

Fig. 7.

Validation of RNA-seq Data by RT-qPCR. The relative expression levels of each gene were determined by quantitative real-time PCR (qPCR, represented by grey bars) and RNA-seq (represented by black lines). The x-axis represents the two quinoa spikelet colors: RF and GF. The left y-axis indicates the relative expression level detected by qPCR, while the right y-axis shows the FPKM value obtained from RNA-seq

Discussion

Phenotypic basis of quinoa panicle color differentiation

The contrasting coloration of quinoa panicles—rose-pink in RF versus green in GF—results primarily from differential pigment accumulation. Quantitative colorimetry (ΔE = 81.45) and pigment assays showed that RF panicles accumulate substantial anthocyanins while exhibiting markedly reduced chlorophyll and carotenoid levels compared with GF (Fig. 1E-H). This pattern parallels pigment transitions observed in other species, such as grape (Vitis vinifera) berry skins, where anthocyanin accumulation accompanies chlorophyll decline during maturation [38]. Concurrent increases in flavonoid abundance in RF further suggest that co-pigmentation interactions contribute to pigment stability and hue enhancement, consistent with reports from Ranunculus, in which flavonols stabilize anthocyanins through molecular stacking and hydrogen bonding [39, 40]. In addition, our transcriptome analysis revealed enrichment of GO terms related to cell wall remodeling (GO:0071554) and membrane transport (GO:0016020), indicating that anthocyanin accumulation requires active cellular restructuring. This is consistent with pigment compartmentalization strategies described in petunia (Petunia hybrida), where vesicle-mediated transport and cell wall modifications are essential for petal pigmentation [41]. Together, these results demonstrate that quinoa panicle coloration likely adopts conserved molecular mechanisms integrating pigment biosynthesis, stabilization, and intracellular transport.

Differential metabolite profiles underlying red panicle color formation

Metabolomic profiling identified the accumulation of petunidin- and cyanidin-derived anthocyanins as the dominant contributors to the red phenotype in RF. Notably, cyanidin-3-O-(6-O-malonyl-β-D-glucoside) and petunidin-3-O-(6-O-malonyl-β-D-glucoside) were elevated by approximately 2.03-fold and 1.98-fold, respectively, compared with GF, indicating their dominant role in driving the red coloration. Additional increases in pelargonidin-3-O-galactoside and malvidin-3-O-(6-O-p-coumaroyl)-glucoside in RF further enriched the chemical complexity of RF pigmentation. These findings align with studies in blueberry, where developmental color transitions correlate with tissue-specific enrichment of cyanidin and delphinidin derivatives [42].

In contrast, flavonoids such as naringenin and afzelin showed only a slight increase in RF, suggesting their secondary roles — likely as co-pigments or metabolic intermediates. Naringenin, a key intermediate generated by CHI, did not exhibit a strong correlation with the red phenotype, supporting the view proposed by Motallebi et al. [43] that the major roles in plants are pharmacological rather than chromogenic. Meanwhile, the reduced accumulation of delphinidin-3-O-rutinoside-5-O-glucoside and procyanidin B3 (down-regulated by 2.22-fold and 1.06-fold in RF), suggests a metabolic shift away from colorless proanthocyanidins towards colored anthocyanins.

Integrated transcriptomic–metabolomic analysis revealed that elevated expression of key structural genes (such as ANS and 4CL) in RF supports enhanced anthocyanin synthesis, while suppression of DFR may redirect flux toward pigmented products. In summary, these patterns demonstrate that the red panicle phenotype predominantly arises from substantial enrichment of cyanidin- and petunidin- based anthocyanins, with flavonoids playing a relatively secondary role.

Differential gene regulation in anthocyanin biosynthesis and transport pathways

KEGG pathway enrichment analysis revealed that DEGs in RF were significantly overrepresented in pathways related to flavonoid, phenylpropanoid, and anthocyanin biosynthesis. This finding indicates that anthocyanin biosynthesis involves multiple enzymes encoded by early biosynthetic genes (EBGs) and anthocyanin-specific biosynthetic genes.

Phenylalanine ammonia-lyase (PAL), the rate-limiting enzyme initiating phenylpropanoid metabolism [44], showed higher expression in RF. Similar PAL-dependent enhancement of tissue pigmentation has been reported in peanut seed coats [45] and Fragaria chiloensis [46]. Elevated PAL expression in RF likely increases substrate flow toward downstream anthocyanin biosynthesis, supporting the red phenotype.

4-Coumarate: CoA ligase (4CL), another key metabolic regulator [47], was also significantly up-regulated in RF, consistent with findings in melon rind coloration [48]. Ho and Smith [49] indicated that low expression of CHI and F3H might enhance pigment deposition. In this study, 4CL expression in RF was significantly higher than in GF, indicating that high 4CL expression may positively regulate of red panicle color formation in quinoa, while its low expression in GF may be insufficient to drive anthocyanin accumulation, contributing to the green phenotype.

Chalcone isomerase (CHI) catalyzes the intramolecular cyclization of chalcone to produce flavanones, forming a critical branch point in anthocyanin biosynthesis [50]. Overexpression studies in tomato and tobacco have demonstrated strong positive correlations between CHI activity and flavonoid accumulation [51, 52]. These results consistently indicate a positive correlation between CHI activity and the amount of flavonoids synthesized. Interestingly, CHS, although essential in many species for directing flux into flavonoids [53], exhibited higher expression in GF. This deviation suggests quinoa may employ tissue-specific regulatory modes in which flux control downstream of CHS, rather than CHS abundance itself, determines anthocyanin accumulation.

Flavanone 3-hydroxylase (F3H), a key hydroxylase in flavonoid metabolism, catalyzes the conversion of flavanones to dihydroflavonols in the anthocyanin biosynthesis pathway, a core step in pigment formation [54]. Notably, studies in plants such as alfalfa and carnation suggest that downregulation of F3H expression might contribute to deeper pigmentation, possibly by redirecting metabolic flux toward more stable anthocyanin derivatives [55, 56]. These patterns suggest regulatory synergy among multiple enzymes rather than single-gene control.

In the third biosynthetic stage, elevated ANS expression in RF promotes the conversion of colorless leucoanthocyanidins into unstable anthocyanidins, whereas lower DFR and LAR expression in RF may limit competing flux into colorless products [57–59]. Downstream modification enzymes—including MaTs and 3GGT—were also more strongly expressed in RF, consistent with enhanced production of stable red/purple anthocyanin derivatives [60, 61].

Regarding transport, genes encoding MRP and MATE transporters—two major mechanisms of vacuolar anthocyanin sequestration [12]—were significantly up-regulated in RF. Four MRP homologs and one MATE homolog displayed high expression, suggesting that efficient intracellular transport contributes to the intense red pigmentation. Conversely, a GST homolog was expressed at higher levels in GF, implying that quinoa may utilize multiple transport pathways, possibly in a tissue- or pigment-specific manner.

The red panicle phenotype thus arises from a coordinated genetic program that concurrently enhances anthocyanin synthesis (via ANS upregulation), promotes pigment stabilization (through 3GGT- and MaT-mediated modifications), and ensures efficient vacuolar compartmentalization (via MRP/MATE transporters). This multi‑layer integration establishes a clear causal chain from transcriptional reprogramming and metabolic flux redirection to the visual manifestation of vivid red coloration.

Analysis of transcription factors and hormone-related genes involved in quinoa panicle coloration

The contrasting panicle colors of quinoa are regulated by a complex transcriptional network integrated with hormone signaling. In RF, high expression of MYB activators (e.g., quinoa homologs of MdMYBA1 and AtPAP1/2) and bHLH partners (e.g., quinoa homologs of ZmbHLH125) is consistent with the formation of a canonical MBW complex, suggesting that this complex likely functions as a central regulator of anthocyanin biosynthesis across species [62]. Simultaneously, repressor MYB homologs (e.g., CmMYB012-like) were expressed at only approximately 3% of GF levels, indicating derepression of the structural pathway [63]. Together, these expression dynamics are indicative of a strong transcriptional activation state in the RF panicle that is favorable for anthocyanin accumulation. The MBW complex regulatory mode proposed in this study reflects the conservation of transcription factor synergy in anthocyanin biosynthesis. Similarly, research in bananas has demonstrated that anthocyanin accumulation is controlled by a coordinated network involving multiple MYB transcription factors (e.g., MbaMA2, MusaMA8) and bHLH factors (TT8), where different MYB members exhibit specificity and hierarchy in regulating structural genes, thereby finely modulating metabolic flux [64].

Hormone -related expression differences further contribute to phenotype divergence. ABA signaling genes—including homologs of MdABI5 and ZmVP1—were significantly up-regulated in RF, which is consistent with the well-documented role of ABA in activating anthocyanin biosynthesis in multiple crops [65]. Additional up-regulated genes associated with cytokinin (e.g., quinoa homologs of OsRR6), salicylic acid (e.g., quinoa homologs of OgUBC1), and jasmonic acid (e.g., quinoa homologs of OsABA8ox1) indicate that multi-hormone coordination, rather than single-hormone dominance, shapes the pigmentation process. Such integration may enhance quinoa’s ability to fine-tune pigment synthesis under diverse environmental conditions. Notably, auxin-related regulatory components (e.g., quinoa homologs of OsSAUR39) were either absent or showed extremely low expression, suggesting that quinoa panicle coloration may be largely independent of canonical auxin signaling [66]. This divergence underscores the species-specific regulatory architecture governing pigmentation.

In summary, red panicle formation in quinoa is dominated by a dual regulatory system: (i) a transcriptional activation module centered on the MBW complex and the repression release of MYB inhibitors, and (ii) a multi-hormone signaling network dominated by ABA but involving CK, SA, and JA pathways. These layers jointly modulate structural genes in the anthocyanin synthesis pathway, ultimately driving pigment accumulation and the emergence of the red phenotype.

Multi‑level regulatory network drives the “green‑to‑red” panicle color transition in quinoa

This integrated analysis suggests that the striking “green-to-red” phenotypic shift in red-panicle quinoa is not determined by a single metabolic pathway or regulatory factor, but results from coordinated action of a series of molecular events operating at metabolism, transport, and transcriptional levels. At the metabolic level, the up-regulation of key structural genes in the anthocyanin biosynthesis pathway (e.g., PAL, 4CL, ANS, 3GGT, and MaTs) correlates with the specific, accumulation of cyanidin- and petunidin-derived anthocyanins, forming the chemical basis of the red coloration. Concurrently, the significant reduction in chlorophyll and carotenoid contents removes the competing green background pigmentation. At the transport level, enhanced expression of vacuolar transporter genes such as MRP and MATE ensures efficient sequestration and storage of newly synthesized anthocyanins in the vacuole, achieving high-concentration stable pigment accumulation at the cellular level. At the transcriptional level, activation of the MBW complex, coupled with strong suppression of key MYB repressors, creates a derepressed transcriptional state that provide a decisive “switch” for the entire anthocyanin biosynthesis pathway. Additionally, a multi-hormone regulatory network apparently dominated by ABA and involving coordinated CK, SA, and JA signaling integrate environmental and developmental cues to positively regulate pigment synthesis.

The MBW transcriptional regulatory module underlying red panicle color formation in quinoa demonstrates high evolutionary conservation. The expansion and functional diversification of its associated gene families may have enhanced the species’ regulatory potential for environmental adaptation. From an ecological adaptation perspective, this trait likely enables quinoa to cope with abiotic stresses such as intense ultraviolet radiation through anthocyanin accumulation. Moreover, the underlying multi-hormone synergistic regulatory network appears to tightly couple pigment biosynthesis with stress resistance responses, potentially conferring significant adaptive advantages.

In summary, red panicle formation represent the endpoint of a precisely orchestrated, multi-level regulatory program: at the transcriptional level, activation of the MBW complex and release of repression may initiate the genetic program for anthocyanin synthesis. At the metabolic level, this program drive the massive biosynthesis and modification of anthocyanins while suppressing competing chlorophyll/carotenoid pathways. At the cellular level, efficient vacuolar transport facilitates pigment compartmentalization and stable color manifestation. The concerted action of these events lead to a fundamental reprogramming of metabolic flux, shifting from a “green” metabolic state dominated by photosynthetic pigments to a “red” state characterized by the accumulation of specific anthocyanins. This reprogramming driving the macroscopically visible transition from green to rose-pink panicle coloration. This integrated model deepens our understanding of panicle color development in quinoa and provides a r theoretical framework and potential targets for molecular breeding to tailor coloration in quinoa and related crops.

However, the direct regulatory functions of specific transcription factors (e.g., MYB, bHLH) and structural genes (e.g., ANS, 3GGT) in quinoa panicle color formation remain to be validated through molecular biology approaches such as transgenic complementation assays, gene editing, or in vitro enzymatic activity analyses. Future work will focus on conducting functional studies of these candidate genes to establish causal relationships and confirm the findings of this study.

Conclusion

This study demonstrated that quinoa panicle color differentiation between red (RF) and green (GF) varieties is primarily driven by differences in anthocyanin accumulation and transcriptional regulation. RF panicles showed markedly higher levels of cyanidin- and petunidin-derived anthocyanins, supported by the up-regulation of key biosynthetic genes (ANS), modification enzymes (3GGT, MaTs), and transporters (MRP, MATE). These coordinated changes indicate that pigment synthesis, modification, and vacuolar sequestration act synergistically to determine the red phenotype. The findings provide a molecular basis for breeding quinoa varieties with enhanced pigment composition and potential nutritional value.

Authors’ contributions

X.S. conceived and designed the experiments. J.L., J.W., Z.T. conducted experiments. J.L., S.N., C.L., L.Z. analyzed the data. J.L., L.Z. wrote the manuscript. X.S., S.N., W.L., G.M. revised the manuscript. All authors read and approved the final manuscript.

Funding

This work was funded by the Fundamental Research Funds for Provincial Universities in Hebei Province (KY2022102); the National Natural Science Foundation of China (32471755); the Local Science and Technology Development Fund Projects Guided by the Central Government (236Z6302G); Key R&D Program of Hebei (21326305D) and Open Project of Yunnan Province Key Laboratory of Flower Breeding (Grant No. FKL-202305).

Data availability

The datasets presented in this study are deposited in the publicly accessible NCBI Sequence Read Archive database under accession number PRJNA1372918.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.