Selection of dysfunctional alleles of bHLH1 and MYB1 has produced white grain in the tribe Triticeae

Abstract

Grain color is a key agronomic trait that greatly determines food quality. The molecular and evolutionary mechanisms that underlie grain-color regulation are also important questions in evolutionary biology and crop breeding. Here, we confirm that both bHLH and MYB genes have played a critical role in the evolution of grain color in Triticeae. Blue grain is the ancestral trait in Triticeae, whereas white grain caused by bHLH or MYB dysfunctions is the derived trait. HvbHLH1 and HvMYB1 have been the targets of selection in barley, and dysfunctions caused by deletion(s), insertion(s), and/or point mutation(s) in the vast majority of Triticeae species are accompanied by a change from blue grain to white grain. Wheat with white grains exhibits high seed vigor under stress. Artificial co-expression of ThbHLH1 and ThMYB1 in the wheat endosperm or aleurone layer can generate purple grains with health benefits and blue grains for use in a new hybrid breeding technology, respectively. Our study thus reveals that white grain may be a favorable derived trait retained through natural or artificial selection in Triticeae and that the ancient blue-grain trait could be regained and reused in molecular breeding of modern wheat.

This study provides evidence that both bHLH and MYB genes have played crucial roles in the evolution of grain color in Triticeae. Blue grain represents the ancestral trait, whereas white grain caused by the selection of dysfunctional alleles of bHLH1 or MYB1 genes is a derived trait in Triticeae. The bHLH–MYB-mediated blue-grain trait can be used in rational design for wheat molecular breeding.

Introduction

Wild wheat and other wild cereals in Triticeae have been used as food by humans for at least 10 000 years, even before the advent of cultivation (Harlan and Zohjary 1966; Harlan et al. 1973; Tanno and Willcox 2006; Purugganan and Fuller 2009; Liu et al. 2022). Modern wheat is currently one of the most important staple food crops worldwide, providing approximately 20% of the calories and 30% of the major food sources consumed by humans (Tilman et al. 2011). Other members of the tribe Triticeae, such as barley and rye, are also important for humans in their present-day life (Harlan and Zohjary 1966; Jayakodi et al. 2020; Kaur et al. 2021; Li et al. 2021; Zhang et al. 2023). However, little is known about the selection and molecular basis for the evolution of agronomic traits in Triticeae.

Grain color is a key agronomic trait and a typical target of natural and artificial selection in grain crops (Purugganan and Fuller 2009; Haas et al. 2019; Liu et al. 2022). The seed color of most crops is primarily white or yellow (e.g., rice, wheat, and maize), although some varieties have brown, black, red, purple, or blue seeds (Furukawa et al., 2007; Zhu, 2018). Differences in seed color are due mainly to the accumulation of anthocyanins, proanthocyanidins, or carotenoids in the endosperm, aleurone layer, or seed coat. Grain color, determined by the aleurone layer, can be blue or white in Triticeae (Zeven 1991; Zheng et al. 2009). Grains of species in tribe Triticeae (e.g., wheat, barley, and rye) are predominantly white, and humans have historically preferred to consume white grains of wheat (and other crops in tribe Triticeae), indicating that white grain is the dominant trait in Triticeae. Nonetheless, blue grains have been documented in tribe Triticeae for at least 130 years (Zeven 1991), and blue grain color has been used as a marker for crop breeding and gene mapping in Triticeae for dozens of years (Dubcovsky et al. 1996; Zheng et al. 2009). Although studies have shown that grain color in rice and maize is associated with domestication, little is known about the evolutionary history and molecular basis of grain color in tribe Triticeae (Haas et al. 2019; Levy and Feldman 2022).

Grain color also has a significant effect on food quality (Olsen and Wendel 2013; Zhang 2021). Colored grains contain anthocyanins, which have several beneficial effects on human health, including free radical scavenging activity; anti-cancer, anti-cardiovascular disease, and anti-inflammatory effects; effects on the hypoglycemic response; and even the ability to extend lifespan (Kähkönen and Heinonen 2003; Abdel-Aal et al., 2008; Butelli et al. 2008; Geleijnse and Hollman 2008; Bondonno et al. 2019; Zhang 2021); anthocyanins are also used as natural colorants in both food and nonfood applications (de Mejia et al. 2020). Therefore, the development of crop varieties with colored grains is a desirable goal for crop breeders.

Hybrid seed production technology (SPT) is a new generation of biotechnology for the commercial production of hybrid crop seeds (Perez-Prat and van Lookeren Campagne 2002; Albertsen et al. 2006). A seed-color marker gene, as well as a male sterility complementary fertility restoration gene and a pollen-killer gene, are required for construction of crop SPT; the seed-color marker gene is needed for obtaining transgenic maintainer lines by mechanical color sorting of transgenic maintainer and non-transgenic hybrid seeds (Perez-Prat and van Lookeren Campagne, 2002; Albertsen et al. 2006; Gupta et al. 2019; Cai et al. 2023). The red fluorescent protein (RFP) gene has been used successfully as a seed-color gene in maize (Wu et al. 2016; Zhang et al. 2018) and rice (Chang et al. 2016) SPT. However, RFP has not proved to be successful in wheat SPT. Therefore, a new seed-color marker gene is needed for the construction of propagation and commercial SPT in wheat.

In the present study, we demonstrate that blue grain is the ancestral trait in progenitors of the tribe Triticeae, whereas white grain is a derived trait caused by dysfunctionalization of bHLH and/or MYB genes. HvbHLH1 and HvMYB1 have been targets of selection in barley, and white grain exhibits high seed vigor in wheat, indicating that these loss-of-function mutations have experienced natural or artificial selection during the evolution of Triticeae. In addition, artificial expression of functional ThbHLH1 and ThMYB1 in the grain endosperm or aleurone layer gives rise to colored wheat grains with health benefits and to a seed sorting marker for wheat SPT, respectively. This work reveals the molecular basis for production of the white-grain trait and helps to explain how it became the predominant trait through natural mutation and natural and/or artificial selection during Triticeae evolution. This ancient trait whose frequency was reduced by natural selection can be regained and reused in modern technology for molecular breeding in modern wheat.

Results

The blue-grain trait is present in old lineages of the tribe Triticeae

The vast majority of taxa in the tribe Triticeae have white grains; blue grains are present in relatively few taxa of Hordeum, Thinopyrum, and Secale and in no or very few taxa of Aegilops and Triticum (Figure 1A). To examine the phylogenetic history of Triticeae and the evolution of grain color, we constructed a phylogenetic tree for representative Triticeae species using their chloroplast genome sequences (Supplemental Table 1). The Hordeum, Thinopyrum, and Secale families formed the older lineages in Triticeae and diverged from the ancestor of the Aegilops and Triticum families (Figure 1B). No or very few species with blue grains are observed in the Aegilops and Triticum taxa examined here, whereas species with blue grains are observed in the Hordeum, Thinopyrum, and Secale taxa (Figure 1), suggesting that the blue-grain trait was present in the progenitors of the tribe Triticeae. Phylogenetic profiling thus indicates that taxa with white grains may have been derived from an ancestor of Triticeae that had blue grains.

Figure 1.

Blue grain is a trait generated in the progenitors of the tribe Triticeae.

(A) Grain colors of Hordeum vulgare (WDM89 and WDM2972), Aegilops tauschii (AL8/78), T. urartu (G1812), T. turgidum (Landon), and T. aestivum (Fielder).

(B) Phylogenetic tree of species in the tribe Triticeae based on their chloroplast genomes. O. sativa and Brachypodium were used as outgroups. The yellow blocks indicate species or varieties with white grains, and the blue blocks indicate species or varieties with blue grains. The time in the tree indicates the divergence time between the two groups. Sources of chloroplast data can be found in Supplemental Table 1.

ThbHLH1 and ThMYBs are the genes responsible for blue grains in a bread wheat–Thinopyrum ponticum alien addition line

Cross breeding between species is common in Triticeae and provides opportunities to investigate the evolution of agricultural traits in crops and grasses. When we characterized the male-sterile mutant ms1 and cloned Ms1 in bread wheat (Wang et al. 2017), we accidentally observed that a bread wheat cultivar with blue grains was produced by crossing T. ponticum with bread wheat ms1 to make a wheat–T. ponticum chromosome alien addition line (Figure 2A), providing us with a valuable opportunity to investigate the genetic control of grain color in wheat. Genomic in situ hybridization (GISH) indicated that the wheat–T. ponticum alien addition line harbored the 4Ag chromosome from T. ponticum (Figure 2A and 2B). Further analysis confirmed that the alien addition line harbored a deletion of the long arm of 4Ag (Figure 2B), indicating that the gene(s) responsible for the blue-grain trait is located on the long arm of chromosome 4Ag. In addition, blue grains of the alien addition line that contained two copies of 4Ag were darker than those from the line that contained one copy (Figure 2B), indicating that the gene(s) controlling the blue-grain trait function in a dose-dependent manner. We observed that the blue-grain trait began to appear in seeds approximately 21 days after pollination, and the blue color was maintained through to seed maturity (Figure 2C). Analysis of seed cross-sections indicated that the blue color was specifically expressed in cells of the aleurone layer (Figure 2D).

Figure 2.

The grain from the ms1-4Ag wheat–T. ponticum alien addition line exhibits a blue color in the cells of the aleurone layer.

(A) Seeds of white-grained wheat (Fielder, NingChun4, and Chinese Spring) and blue-grained wheat (wheat–Thinopyrum 4Ag addition line).

(B) GISH analysis of chromosome 4Ag in Ms1, ms1-4Agʹ, ms1-4Agʹʹ, and ms1-4AgSʹʹ plants at mitotic metaphase. The red signal marks the 4Ag chromosome.

(C) Mature seeds of Ms1, ms1-4Agʹ, ms1-4Agʹʹ, and ms1-4AgSʹʹ.

(D) Seeds of Ms1 (top) and ms1-4Agʹʹ (bottom) at 15, 18, 21, 24, and 27 days post-anthesis (DPA).

(E) Seed cross-sections of Ms1 (top) and ms1-4Agʹʹ (bottom) at 15, 18, 21, 24, and 27 DPA.

To clone the gene(s) responsible for the blue-grain trait, we performed genome-wide RNA sequencing (RNA-seq) of cells from the aleurone layer of bread wheat and the ms1-4Ag alien addition line (Supplemental Figure 1A; Supplemental Table 2) and identified genes that were differentially expressed between the two materials. Several genes encoding enzymes involved in the anthocyanin biosynthetic pathway, including CHS, CHI, DFR, UFGT, and GST from the bread wheat genome, were upregulated in the aleurone layer of blue grains (Supplemental Figure 1B and 1C), indicating that the causal gene(s) function upstream of these genes in the pathway. Furthermore, two genes from T. ponticum, ThMYB1 and ThbHLH1, which encode an MYB and a bHLH transcription factor, respectively, were upregulated in the aleurone layer of blue grains (Supplemental Figure 1C). Because (1) several genes from the bread wheat genome that encode enzymes required for anthocyanin synthesis were upregulated in the blue grains (Supplemental Figure 1C); (2) the expression of these genes is reportedly activated by an MYB–bHLH complex (Xu et al. 2015); and (3) ThMYB1 and ThbHLH1 were identified from the T. ponticum genome, we hypothesized that ThMYB1 and ThbHLH1 might be the genes responsible for the blue grains observed in the wheat alien addition line.

An RNA-seq analysis revealed that there were two ThMYBs (ThMYB1 and ThMYB2) and two ThbHLHs (ThbHLH1 and ThbHLH2), all of which were expressed in the aleurone layer in the ms1-4Ag alien addition line (Figure 3A and 3B). To confirm that these genes were responsible for the blue grains of the alien addition line, we first expressed the four genes in bread wheat protoplasts and examined their ability to activate the expression of the aforementioned anthocyanin biosynthesis genes. Expression of any single gene had no effect on expression of CHS, CHI, DFR, or GST in the protoplasts, and co-expression of ThbHLH2 and ThMYB1 or ThbHLH2 and ThMYB2 also had no obvious effect. However, co-expression of ThbHLH1 and ThMYB1 or ThbHLH1 and ThMYB2 clearly activated the expression of CHS, CHI, DFR, and GST in the protoplasts (Figure 3C), indicating that both a ThbHLH gene and a ThMYB gene are required for expression of these anthocyanin biosynthesis genes. ThbHLH1, ThMYB1, and ThMYB2, but not ThbHLH2, were capable of activating expression of genes involved in anthocyanin biosynthesis in bread wheat protoplasts.

Figure 3.

The genes responsible for the blue-grain trait in the bread wheat–T. ponticum alien addition line are ThbHLH and ThMYB.

(A) and (B) Expression of ThMYB1, ThMYB2, ThbHLH1, and ThbHLH2 measured by RT–PCR (A) and quantitative RT–PCR (B) in the wheat–T. ponticum alien addition line (ms1-4AgSʹʹ). AL, aleurone layer; G, germ; L, leaf; R, root; P, pericarp; S, shoot; SE, starchy endosperm. ACTIN was used as an internal control. Error bars indicate the SD of three biological replicates.

(C) and (D) Effect of transient transformation of ThMYB1, ThMYB2, ThbHLH1, or ThbHLH2 driven by the ubiquitin promoter on the expression of CHS, CHI, DFP, and GST in bread wheat protoplasts (C) and on anthocyanin synthesis in bread wheat coleoptiles (D).

(E) Spikes (top) and seed cross-sections (bottom) of bread wheat var. Fielder and three representative pThbHLH1::ThbHLH1 + pThMYB1::ThMYB1 transgenic lines in the Fielder background. The red arrow indicates the aleurone layer. Among 93 independent transgenic lines, 41 produced blue grains.

This observation was confirmed by examination of anthocyanin synthesis in bread wheat coleoptiles transiently expressing ThMYBs and ThbHLHs. Transient co-expression of ThbHLH1 and ThMYB1 or ThbHLH1 and ThMYB2 induced anthocyanin biosynthesis in epidermal cells of the bread wheat coleoptiles, but no effect was observed upon expression of any single gene or upon co-expression of ThbHLH2 and ThMYB1 or ThbHLH2 and ThMYB2 (Figure 3D). Finally, genetic co-transformation of ThbHLH1 and ThMYB1 driven by their native promoters generated transgenic bread wheat with the blue-grain trait (Figure 3E). Observations of seed cross-sections revealed that anthocyanins accumulated specifically in cells of the aleurone layer in the transgenic bread wheat lines (Figure 3E), a pattern similar to that observed in the ms1-4Ag wheat–T. ponticum alien addition line (Figure 2D). Together, these results indicated that the genes responsible for blue-grain trait were ThbHLH1 and two ThMYBs.

THbHLH1 and ThMYBs physically interact to form a complex that triggers anthocyanin synthesis

To reveal how ThbHLH1 and ThMYBs participate in the synthesis and accumulation of anthocyanins in the aleurone layer, we investigated the physical interactions between ThbHLHs and ThMYBs as a requirement of a bHLH–MYB complex for the synthesis of anthocyanin. Pair-wise physical interactions between ThbHLH1/2 and ThMYB1/2 were observed in yeast (Supplemental Figure 2A). To determine why ThbHLH2 interacted with ThMYB1/2 in yeast but was not functional in bread wheat (Figure 3C and 3D; Supplemental Figure 2C), we examined the sub-cellular localization of ThbHLH1, ThbHLH2, ThMYB1, and ThMYB2. We observed that ThbHLH1, ThMYB1, and ThMYB2 were localized in the nuclei, but ThbHLH2 was localized in the cytosol (Supplemental Figure 2B). Further sequence analysis revealed that an insertion in the 3′ region of ThbHLH2 generated a premature stop codon predicted to result in a truncated protein (ThbHLH2) with a deletion of 113 amino acid residues in the C-terminal region compared with ThbHLH1, which contains a predicted nuclear localization sequence in the deleted region (Supplemental Figure 2C). This prediction was confirmed, as a form of truncated ThbHLH1 lacking the C-terminal region corresponding to the deleted C-terminal region in ThbHLH2 was localized in the cytosol, whereas fusion of the C-terminal domain of ThbHLH1 into ThbHLH2 caused the latter to exhibit nuclear localization (Supplemental Figure 2D). These results indicate that the mutation in ThbHLH2 leads to its dysfunction by disturbing its subcellular localization in T. ponticum. The pair-wise interaction between ThbHLH1 and ThMYB1 or ThMYB2 was further confirmed by fluorescence resonance energy transfer (FRET) and co-immunoprecipitation (Co-IP) assays in tobacco cells. A strong FRET signal was detected in the nuclei of tobacco cells co-expressing GFP-ThbHLH1 and ThMYB1-RFP or ThMYB2-RFP (Supplemental Figure 2E and 2F). In addition, specific Co-IP signals were typically observed when ThbHLH1 and ThMYB1 or ThMYB2 were co-transfected into tobacco leaves (Supplemental Figure 2G). These results indicate that both THbHLH1 and ThMYBs are required for the formation of the blue-grain trait in the alien addition bread wheat line, as both THbHLH1 and ThMYBs are components of a transcriptional complex required for the expression of anthocyanin synthesis genes.

bHLH and MYB have been dysfunctionalized by mutations in bread wheat and its progenitors

Southern blotting was performed to investigate homologs of THbHLH1 and ThMYBs in the genomes of bread wheat and its progenitors. Three copies of ThbHLH and three copies of ThMYB were detected in T. ponticum. A single copy of bHLH and a single copy of MYB were detected in the bread wheat genome, and they were from the D subgenome, as MYB and bHLH were detected only in Ae. tauschii (DD progenitor) but not in T. urartu (AA progenitor) or T. turgidum (AABB progenitor). Consistent with these results, three copies of the bHLH gene and three copies of the MYB gene were detected in the ms1-4Ag wheat–T. ponticum alien addition line (two from 4Ag and one from the D subgenome) (Supplemental Figure 3). Further genomic analysis suggested that the bread wheat MYB gene (named TaMYB1) and bHLH gene (named TabHLH1) were linked in the D subgenome, and subgenome collinearity analysis of the genomic region containing TabHLH1 and TaMYB1 indicated that both the A and B subgenomes had lost a small region containing TabHLH1, TaMYB1, and another gene in bread wheat (Figure 4A), consistent with the results of the Southern blot analysis (Supplemental Figure 3).

Figure 4.

Both TaMYB1 and TabHLH1 are dysfunctionalized by deletion(s), insertion(s), and nonsynonymous point mutation(s) in bread wheat and its progenitors.

(A) Chromosomal locations of ThbHLH1 and ThMYB1 homologs in bread wheat. CS4D02G224500 and CS4D02G224600 are the bread wheat homologs of ThMYB1 and ThbHLH1, respectively. The bread wheat 4D chromosome region from CS4D02G223900 to CS4D02G225100 and its homologous regions on the 4A and 4B chromosomes are shown. Homologous genes are marked with the same color and linked by straight lines. The box indicates the region specific to chromosome 4D.

(B) Expression of TaMYB1 and TabHLH1 measured by RT–PCR in bread wheat var. Chinese spring. AL, aleurone layer; G, germ; L, leaf; P, pericarp; R, root; S, shoot; SE, starchy endosperm. ACTIN was used as an internal control.

(C) Diagram of TaMYB1 and TabHLH1 structures. Exons are represented by dark blue arrows and introns by black lines. The loss of function mutations are marked on the diagrams.

(D) Interaction between TaMYB1 and TabHLH1 or their corresponding mutant proteins assessed by yeast two hybrid assays. The yeast was diluted to 10⁰, 10⁻¹, or 10⁻². DDO, double dropout supplements −Leu/−Trp; QDO, quadruple dropout supplements −Leu/−Trp−/−His/−Ade. Error bars indicate the SD of three biological replicates.

(E) Interaction between TaMYB1 and TabHLH1 or their corresponding mutant proteins assessed by FRET assays in N. benthamiana cells. Images of GFP and RFP channels before and after RFP bleaching are shown. The barcodes indicate the fluorescence intensity.

(F) FRET efficiency for the interaction between TaMYB1 and TabHLH1 or their mutant proteins.

(G) and (H) Effect of transient transformation of TaMYB1, TaMYB2, TabHLH1, or TabHLH2 or their mutant forms driven by the ubiquitin promoter on the expression of CHS, CHI, DFP, and GST in bread wheat protoplasts (G) and on anthocyanin synthesis in bread wheat coleoptiles (H).

Analysis of their expression patterns revealed that TaMYB1 was expressed in bread wheat in a pattern similar to that of ThMYB1 and ThMYB2 in the ms1-4Ag alien addition line, but TabHLH1 was not expressed in bread wheat (Figure 4B). There was an insertion in TaMYB1 and two insertions in TabHLH1 (Figure 4C), which led to a premature stop codon in TaMYB1 and inactivation of TabHLH1 expression in bread wheat (Figure 4B). Homologous alignment and analysis of protein interactions in yeast indicated that there was one more nonsynonymous point mutation in TaMYB1 and three more nonsynonymous point mutations in TabHLH1 that may affect their function in bread wheat (Figure 4C). The physical interaction between TaMYB1 and TabHLH1 was again absent, even after removal of the one insertion from TaMYB1 and the two insertions from TabHLH1 (Figure 4D; Supplemental Figure 4). The resulting TaMYB1 and TabHLH1 still failed to induce the expression of genes encoding anthocyanin synthesis-related enzymes in bread wheat protoplasts (Figure 4G) or the biosynthesis of anthocyanins in bread wheat coleoptiles (Figure 4H). The physical interaction between TaMYB1 and TabHLH1 was reconstructed, and the TaMYB1–TabHLH1 complex only regained function when we removed all insertions from the two genes and corrected all nonsynonymous point mutations in the two genes (Figure 4E and 4F; Supplemental Figure 4). The truncated TaMYB1 protein caused by the insertion in TaMYB1 showed a clear reduction in its interaction with TabHLH1 compared with that of full-length TaMYB1 (Figure 4D; Supplemental Figure 4). In addition, individual nonsynonymous point mutations in TabHLH1 contributed to its reduced ability to form the TabHLH–TaMYB1 complex (Figure 4D; Supplemental Figure 4). These results indicate that both TaMYB1 and TabHLH1 have been dysfunctionalized by deletion, insertion(s), and nonsynonymous point mutation(s) in bread wheat and its progenitors.

AebHLH and AeMYB have been dysfunctionalized by mutations in Aegilops species

Because both TaMYB1 and TabHLH1 are from the D subgenome and have been disrupted in bread wheat, we examined their homologs in Ae. tauschii, the progenitor of the bread wheat D subgenome, to determine whether AeMYB1 and AebHLH1 are intact in Ae. tauschii. We observed that the same insertion(s) and nonsynonymous point mutation(s) found in TaMYB1 and TabHLH1 of bread wheat were also present in AeMYB1 and AebHLH1 of Ae. tauschii (Figure 5A), indicating that the dysfunctionalization of AeMYB1 and AebHLH1 occurred before the hybridization that produced allohexaploid wheat. We further identified these insertion(s) and nonsynonymous point mutation(s) in AeMYB1 and AebHLH1 homologs across the genus Aegilops by sequencing these two genes from 102 representative Aegilops species. We observed that the deletion(s), insertion(s), and/or point mutation(s) were present in all of the tested species, with five indicated mutation types (Figure 5A; Supplemental Table 3). The number of mutations in these two genes decreased from five in AebHLH1 and two in AeMYB1 to one in each gene among members of the Aegilops family, indicating the processes of MYB1 and bHLH1 evolution in the Aegilops family. None of these genes were likely to be fully functional, on the basis of our previous analysis. Consistent with this observation, grains from all the representative Aegilops species are white (Supplemental Figure 5A).

Figure 5.

The majority of MYB1 and bHLH1 homologs in tribe Triticeae are dysfunctionalized by deletion(s), insertion(s) and nonsynonymous point mutation(s).

(A) The deletions, insertions, or point mutations in MYB1 and bHLH1 homologs from 102 representative Aegilops species. A dashed line indicates that the gene is not detectable and therefore may have been deleted from the genome.

(B) Circos plot showing the collinearity among assemblies of chromosome 4 or 7 from 13 representative Triticeae species. Red lines link the MYB1 and bHLH1 homologs.

(C) The collinearity of chromosome regions containing MYB1 and bHLH1 homologs from 13 representative Triticeae species, together with their corresponding grain colors.

(D) MYB1 and bHLH1 homologs and their mutation types that lead to dysfunctionalization of MYB1 and bHLH1 in 856 Triticeae species.

bHLH and MYB have been dysfunctionalized in the majority of Triticeae species

To further explore the evolution of grain color, we extended our examination to the tribe Triticeae. We first characterized TaMYB1 and TabHLH1 homologs in the tribe Triticeae. Chromosome-level genome assemblies of 13 Triticeae species were available, providing an opportunity to analyze the collinearity of their bHLH and MYB genes. We observed that both the TabHLH1 and TaMYB1 homologs were absent in the genomes of eight species, including five species from Aegilops, as well as Thinopyrum elongatum, Triticum urartu, and Triticum turgidum. Both the TabHLH1 and TaMYB1 homologs were present with various mutations on chromosome 4 or 7 in the genomes of another four species: two Hordeum species, one Aegilops species, and Triticum aestivum (one Hordeum species was missing the MYB1 homolog). Only Secale cereale appeared to have functional copies of both genes (Figures 5B and 5C; Supplemental Table 4).

We next examined 856 Triticeae species whose genome sequences were available; 700 of these species had white grains, 49 of them had blue grains (S. cereale and 48 Hordeum species), and 107 had no grain-color information (Supplemental Figure 5C; Supplemental Table 5). In addition, 644 of these species had the aforementioned deletions, insertions, and nonsynonymous point mutations in their TabHLH1 and TaMYB1 homologs (Figure 5D; Supplemental Table 6), and the grains from all of these 644 species were white (Supplemental Figure 5C). These mutations resulted in the loss of bHLH1 and MYB1 function due to deletion of TabHLH1 and TaMYB1 homologs, disruption of bHLH1 nuclear localization, generation of a non-functional, truncated bHLH1 protein, or impairment of the physical interaction between bHLH1 and MYB1 (Supplemental Figure 6). In another collection of Hordeum accessions, only 35 of 5413 accessions exhibited blue grains, and the others had white grains (Supplemental Figure 5C). The proportion of species with the blue-grain trait varied among different collections of the Hordeum genus (Supplemental Figure 5C), indicating that distribution of the grain-color trait may vary among different climatic regions. These results indicate that the blue-grain trait is less common in Triticeae and that the white-grain trait predominates due to dysfunctionalization of TabHLH1 and TaMYB1 homologs by deletions, insertions, and nonsynonymous point mutations.

Grain color is a target of selection in barley and bread wheat

As both white and blue grains are present in the Hordeum genus (Supplemental Figure 5C), we examined whether the grain-color trait has undergone selection during evolution in Hordeum. Among more than 5000 Hordeum accessions we examined, 131 had both a genome sequence and a known grain color, and we therefore analyzed these 131 accessions for selection on the grain-color trait in Hordeum. Forty-five of the accessions had blue grains, and 86 had white grains. None of the aforementioned mutations in HvbHLH1 or HvMYB1 were detected in the 45 accessions with blue grains, whereas these mutations in HvbHLH1 or HvMYB1 were present in 27 of the 86 white-grained accessions (Figure 6A). Our analysis of phylogenetic history and grain-color evolution in the 131 barley accessions based on their whole-genome coding sequences supported the hypothesis that the progenitors had the blue-grain trait and that white grain evolved independently and in parallel in the Hordeum family (Figure 6A). The evolution of grain color from blue to white began at least ∼0.53 million years ago (Mya) (Supplemental Figure 7). To determine whether HvbHLH1 and HvMYB1 are targets of selection, we analyzed genome-wide variations across the 86 white-grained and 45 blue-grained accessions by assessing three parameters of past selection. Both HvbHLH1 and HvMYB1 exhibited strong evidence of a past selective sweep, with a significant Pi ratio between white- and blue-grained accessions and highly divergent sites of the population differentiation statistic (F_ST) and cross-population composite likelihood ratios (XP-CLRs) of each SNP site between white- and blue-grain Hordeum accessions (Figure 6B). This result suggested that grain color has undergone natural selection or domestication and diversification in Hordeum species and that HvbHLH1 and HvMYB1 are the targets of convergent selection.

Figure 6.

Alleles of bHLH1 and MYB1 genes have been selected and contribute to grain-color determination in Triticeae.

(A) Rooted ML tree of white- and blue-grained accessions of H. vulgare.

(B) The Pi ratio (top), F_ST (middle), and XP-CLR (bottom) values estimated by comparing genome sequences from 131 white- and blue-grained accessions of H. vulgare. The genes required for anthocyanin synthesis that have undergone selection are indicated. The red dashed lines indicate the top 1% and top 5% for each value.

(C) Germination rate of the bread wheat variety Fielder and pThbHLH1::ThbHLH1 + pThMYB1::ThMYB1 transgenic lines under normal or stress conditions over 48 h of germination. Fielder, bread wheat variety and the recipient of transformation; #1 and #2, two independent transgenic lines. Error bars indicate the SD of three biological replicates. Statistical differences were assessed using a t-test. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Given that the vast majority of Triticeae species have undergone dysfunctionalization of these bHLH1 and MYB1 genes and that these allele genes were the targets of selection (Figures 5 and 6; Supplemental Figure 5C), the white-grain trait in Triticeae species is likely to have been an advantage under natural or artificial selection during evolution. To test this hypothesis, we compared the seed vigor of blue grains from transgenic bread wheat co-transformed with ThMYB1 and ThbHLH1 to that of white grains from the transgenic recipient parent bread wheat under various stress treatments. We observed that the seed vigor of the white grains was greater than that of the blue grains under stress treatments, including salt, osmotic stress, and low temperature (Figure 6C), indicating that blue-grained bread wheat expressing functional bHLH1 and MYB1 is more sensitive to stress.

The bHLH–MYB-mediated blue-grain trait can be used in the rational design of wheat molecular breeding

Understanding the molecular nature and evolution of the grain-color trait could enable the development of bread wheat with colored grain through rational design via transformation with functional bHLH1 and MYB1. To this end, we obtained bread wheat grains with purple endosperm by artificial expression of ThbHLH1 and ThMYB1 driven by an endosperm-specific promoter (Figure 7A), indicating that expression of ThbHLH1 and ThMYB1 in the endosperm leads to the synthesis of anthocyanins in the endosperm with more nutrients. Thus, an ancient trait with human nutritional benefits that has been lost through natural or artificial selection can be regained by trait engineering in modern crops.

Figure 7.

Rational molecular design for bread wheat breeding using functional ThbHLH1–ThMYB1.

(A) Production of bread wheat grains with purple endosperm by expression of ThbHLH1 and ThMYB1 driven by an endosperm-specific promoter. Among 82 independent transgenic lines, 54 exhibited grain with a purple endosperm.

(B) and (C) RFP-labeled and blue transgenic bread wheat grains viewed under fluorescence and white light, respectively.

(D) Sorting efficiencies for RFP-labeled and blue bread wheat grains. Error bars indicate the SD of three biological replicates. Statistical differences were assessed using a t-test. ∗∗∗p < 0.001.

bHLH–MYB-mediated blue grain is also useful in the new generation of hybrid SPT (Perez-Prat and van Lookeren Campagne, 2002; Albertsen et al. 2006). The seed-label gene is used to label and sort seeds containing the transformed male sterile restorer and pollen-killer genes, which act as the maintainer in SPT (Gupta et al. 2019; Cai et al. 2023). The blue grains generated by co-expression of ThbHLH1 and ThMYB1 driven by their native promoters exhibited almost 100% sorting efficiency, whereas the sorting efficiency of seeds labeled by RFP, which has been used as a seed-label marker for SPT in maize (Wu et al. 2016; Zhang et al. 2018) and rice (Chang et al. 2016), was only 70% (Figure 7B–7D). Given that the sorting efficiency must be almost 100% to avoid transgenic leak to the environment, these results suggest that the ThbHLH1–ThMYB-mediated blue-grain trait is a suitable sorting marker for SPT in bread wheat.

Discussion

It has been suggested that two types of adaptation are critical for the evolution and selection of cereal crops under human cultivation: successful germination of grain with increased soil disturbance and burial depth, and ease of harvesting (Westoby et al. 1996; Fuller et al. 2007; Purugganan and Fuller 2009). The present study reconstructed the evolutionary history of white grain color in the tribe Triticeae, which is caused by the dysfunction of ThbHLH1 and/or ThMYB1 homologs and has been maintained through natural and/or artificial selection. In addition, our preliminary results indicate that blue grains with functional bHLH1 and MYB1 are more sensitive to environmental stresses in terms of seed vigor (Figure 6C) and that white grains with mutant alleles of bHLH1 and MYB1 have been kept and enriched by natural or artificial selection in bread wheat (Figure 6B). Our results further suggest that germination vigor has undergone natural and/or artificial selection and that cultivars with higher seed vigor have become predominant during Triticeae evolution (Figure 6B), therefore providing evidence that germination vigor was critical for crop adaptation during evolution, even before human cultivation (Figure 6A; Supplemental Figure 7). Although we cannot rule out the possibility that blue grains have other advantage(s) or disadvantage(s) compared with white grains under specific conditions, our results still indicate that white grains have become the more common trait and that the blue-grain trait has gradually been lost due to natural or artificial selection under certain environmental conditions in Triticeae (Figure 8).

Figure 8.

Model of the molecular evolution of grain color in tribe Triticeae.

Functional bHLH1 and MYB1 produce blue grains by promoting the synthesis of anthocyanins in the aleurone. bHLH1, MYB1, or their downstream target genes in the anthocyanin pathway have undergone selection, and their mutations give rise to white grains, which exhibit higher seed vigor than blue grains under stress treatments. Therefore, natural mutations in bHLH1, MYB1, or their downstream target genes have been retained and enriched by natural and/or artificial selection in bread wheat and in tribe Triticeae. Triangles represent insertions, red dots represent point mutations, and solid and dashed boxes represent complete and partial gene deletions, respectively. Solid lines represent results based on experimental evidence, and dashed lines represent predicted results.

Nonetheless, there are some exceptions in which the blue-grain trait is absent even when none of the aforementioned mutations are observed in bHLH1 and MYB1 genes in tribe Triticeae (Figures 5D and 6A; Supplemental Figure 5C). There are two possibilities for these exceptional cases: (1) other unknown mutation(s), especially nonsynonymous point mutation(s), that lead to loss of gene function are present in the bHLH1 and MYB1 homologs; and (2) absence of the blue-grain trait is caused by mutations in downstream target gene(s) of the bHLH1–MYB1 complex in the anthocyanin pathway (Xu et al. 2023). In support of the latter possibility, we observed strong evidence for selection on several genes that are targets of the bHLH1–MYB1 complex and encode enzymes of anthocyanin biosynthesis in Hordeum species, e.g., ANS, ANR, F5′3′H, FLS, and GST (Figure 6B). It appears that mutations in downstream anthocyanin biosynthesis genes have occurred in Thinopytum but not in bread wheat, as transformation of T. ponticum ThbHLH1 and ThMYB1 into bread wheat leads to the blue-grain trait, whereas the grain of T. ponticum is white (Figure 3E). Because Hordeum and Thinopytum are ancient genera in tribe Triticeae (Figure 1B), this suggests that the evolution of grain color in ancient species reflects dysfunction of the bHLH1–MYB1 complex or the downstream anthocyanin-synthesis genes or both, whereas the grain color of derived species is inherited from their progenitors with dysfunctions in the bHLH1–MYB1 complex or evolved by selection on mutated alleles of bHLH1 and/or MYB1 in Triticeae (Figure 8). Nevertheless, all these results support a scenario in which white grain became the dominant trait through dysfunction of genes required for anthocyanin synthesis and was maintained through natural or artificial selection under specific environmental conditions in Triticeae (Figure 8).

The evolution of seed color shows diversity across different species in most crops, and the direction of seed color evolution tends to be from colored to colorless. However, the molecular mechanisms and selective pressures behind these changes are not identical. For example, the domestication process from the red wild variety to the white cultivated variety of rice is caused by loss of function in the Rc and Rd genes (Furukawa et al., 2007). The Rc gene is believed to be associated with seed dormancy, and humans therefore likely favored white-seeded varieties with mutations in the Rc gene during domestication (Sweeney et al., 2007). The direction of seed-color evolution in some other crops is the opposite of that in rice. For example, the Y1 gene regulates carotenoid synthesis, and an insertion in the Y1 promoter increases Y1 expression in maize seeds, significantly increasing their carotenoid content (Palaisa et al., 2003). Because animals cannot synthesize carotenoids and β-carotene is the precursor of vitamin A (Yeum and Russell, 2002), maize varieties with high carotenoid contents provide more precursors for vitamin A synthesis in animals. As a result, maize in the United States quickly shifted from predominantly white kernels to yellow kernels with a high carotenoid content (Doebley et al., 2006). In sorghum, mutations in the Tannin1 gene reduce anthocyanin content and increase fat content, making colorless sorghum varieties more attractive to birds (Xie et al., 2019).

Selection will cause a genetic bottleneck, which reduces the genetic diversity throughout crop genomes (Doebley et al. 2006; Meyer and Purugganan 2013; Huang et al. 2022); for example, approximately one-half of the resistance genes in wild soybean have been lost in landraces and improved soybean cultivars (Zhou et al. 2015). Artificial selection mainly focuses on yield (e.g., grain and fruit size) and taste, and it may therefore lead to a loss of food-quality traits; examples include the loss of cucurbitacin biosynthesis in cucumber and several other cucurbits (Shang et al. 2014), the reduced accumulation of anti-nutritional steroidal glycoalkaloids and vitamin C in tomato (Li et al. 2018; Zhu et al. 2018), and the loss of anthocyanin accumulation in wheat (Figures 1 and 3). However, when the molecular and evolutionary mechanisms underlying food-quality traits are understood, these traits can be regained by engineering of genes required for biosynthesis of the relevant metabolites in crops to improve food quality, as demonstrated for wheat (Figure 7), rice (Zhu et al. 2017; Zhang 2021), and tomato (Butelli et al. 2008; Li et al. 2018).

Materials and methods

Plant materials

Bread wheat (T. aestivum) and other species from tribe Triticeae, including T. urartu and T. turgidum, Aegilops species (Ae. tauschii, Ae. umbellulata, Ae. comosa, Ae. uniaristata, Ae. markgrafii, Ae. ovata), Th. ponticum, and H. vulgare species were used in this study. Aegilops species were provided by Professor Shoufen Dai (Wheat Research Institute, Sichuan Agricultural University), and H. vulgare species were provided by the Chinese Academy of Agricultural Sciences. The ms1-4Ag chromosomal addition lineage was made by crossing the ms1 male sterile mutant (as the female) with blue-grain wheat (as the male parent) (Zhou et al. 2006). All plants were grown in a greenhouse at 24°C with 16 h of light and 8 h of darkness.

GISH

Young plant root tips were collected, and cells at mitotic phase were used for the GISH assay. The GISH procedure was performed as described by Zheng et al. (2006). Genomic DNA probes were prepared using a DIG-Nick Translation Mix kit (Roche) with a 1:300 ratio of probe to blocking DNA, and hybridization signals were detected with an anti-digoxigenin–rhodamine kit (Roche). Chromosomes were stained and blocked using VECTASHIELD with DAPI (Vector Laboratories). Photographs were taken under a Zeiss Axio Imager M2 fluorescent microscope.

RT–PCR and RT–qPCR

Plant leaves, roots, stems, and seeds at different developmental stages were collected for total RNA extraction. Total RNA was extracted from leaves, roots, stems, pericarp, and germ using RNAiso Plus (Takara Bio Inc.) and from the starchy endosperm and aleurone layer as described by Pfeifer et al. (2014). Genomic DNA was removed using RQ1 RNase-Free DNase (Promega). Three micrograms of each RNA sample was used for reverse transcription with the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). RT–PCR was performed in a PCR instrument using LA Taq (Takara Bio Inc.), and amplification conditions were adjusted according to the properties of the target genes. ACTIN was used as an internal reference gene. RT–qPCR was performed in an ABI QuantStudio6 Q6 instrument, and TB Green Premix Ex Taq II (Takara Bio Inc.) was used for the amplification reaction. The reaction conditions were as follows: 95°C for 10 s, followed by 40 cycles of 95°C for 3 s, 60°C for 30 s, and 72°C for 34 s. ACTIN was used as the internal reference gene. Three biological replicates were performed, with three technical replicates for each replicate. The primers used for RT–PCR and RT–qPCR are provided in Supplemental Table 7.

RNA-seq analysis

Total RNA was extracted from the aleurone layer of 27-DPA grains of ms1-4Agʹʹ and Ms1 as described by Pfeifer et al. (2014). Two biological replicates for each sample from ms1-4Agʹʹ and Ms1 were performed, and RNA-seq was performed by BIOPIC. RNA-seq reads were mapped to the T. aestivum reference genome (IWGSC RefSeq v2.1) (Zhu et al. 2021) using HISAT2, and differentially expressed genes between the two samples were identified using the R package edgeR.

Transient gene expression in protoplasts and coleoptiles

The cDNAs of ThbHLH1, ThbHLH2, ThMYB1, ThMYB2, TabHLH1, TaMYB1, and their corresponding mutations were constructed using the pEASY-Blunt cloning vector (TransGen Biotech) driven by the ubiquitin promoter. Plasmids were obtained using cesium chloride density gradient centrifugation for transient gene expression in bread wheat protoplasts and coleoptiles.

Seven-day-old bread wheat seedlings were used to prepare protoplasts as described by Yoo et al. (2007). Fresh bread wheat leaves were cut into long strips of approximately 1 mm with a sharp blade, and the strips were transferred to enzyme solution containing 1.5% (w/v) Cellulase R-10 and 0.75% (w/v) Macerozyme R-10 (Yakult Pharmaceutical). After digestion, the enzymatic solution was removed, and the MMG solution was re-suspended to reach a concentration of 1 × 10⁶ protoplasts/mL. Plasmids were transferred into the protoplasts using 40% PEG–calcium solution, and the transformed protoplasts were cultured for 36 h. Protoplasts were collected, and protoplast total RNA was extracted using the RNeasy Plant Mini Kit (QIAGEN). The transcript levels of the corresponding genes were detected by RT–PCR.

Coleoptiles of bread wheat var. Chinese Spring were used for bombardment assays. Six coleoptiles were transformed simultaneously at one time, with three replicates for each transient transformation. The gold particle preparation and bombardment processes were performed as described by Vasil and Vasil (2006). Plasmids of pUBI::MYB and pUBI::bHLH were mixed with gold particles and used for bombardment with the Biolistic PDS-1000/He Particle Delivery System (Bio-Rad). The settings were 1100 psi gas pressure and a 9-cm distance between the termination screen and the coleoptiles. The bombarded coleoptiles were incubated at 22°C under long-day conditions (16-h light/8-h dark) for 2 d. Anthocyanin synthesis in coleoptile epidermal cells was photographed under a Zeiss Axio Imager M2 fluorescent microscope.

Genetic transformation of bread wheat lines

ThbHLH1 (2086 bp upstream of the ATG + cDNA + 510 bp downstream of the TAG) fusion DNA and ThMYB1 (1952 bp upstream of the ATG + the gene body + 441 bp downstream of the TAA) genomic DNA were used to generate transgenic bread wheat with blue grains. Both ThbHLH1 (cDNA + 510 bp downstream of the TAG) fusion DNA and ThMYB1 (gene body +441 bp downstream of the TAA) genomic DNA driven by the 1Dx5 endosperm-specific promoter (Halford et al. 1989) were used to produce transgenic bread wheat with purple grains. They were simultaneously constructed into the pCAMBIA1300 vector containing the BAR resistance gene. The plasmid was transformed into bread wheat var. Fielder via Agrobacterium tumefaciens. The primers used for construct preparation are provided in Supplemental Table 7.

Yeast two-hybrid assay

The full-length MYB genes (ThMYB1, ThMYB2, TaMYB1, or mutated forms of TaMYB1 or TaMYB1 homologs from other Triticeae species) were cloned into pGADT7 (AD). The full-length bHLH genes (ThbHLH1, ThbHLH2, TabHLH1, or mutated forms of TabHLH1 or TabHLH1 homologs from other Triticeae species) were cloned into pGBKT7 (BD). All constructs were confirmed by sequencing. AD-MYB and BD-bHLH plasmids were co-transformed into yeast strain AH109. The assays for transformation and yeast growth were performed as described in the Yeast Protocols Handbook (Takara Bio Inc.). The primers used for construct preparation are provided in Supplemental Table 7.

The protein interaction intensity was determined on the basis of yeast strain growth. The transformed yeast strains were inoculated into SD/-Leu-Trp and SD/-Leu-Trp-His-Ade liquid medium at 30°C and 220 rpm. Data were analyzed using IBM SPSS Statistics 26 software for non-linear regression according to the formula OD₆₀₀ = k/(1 + a × EXP ^{(−b × time)}) and the values of the parameters k, a, b, the protein interaction intensity was calculated according to the equation:

$$\frac{{\left(\text{time}\times {\ln }^{\left[a\right]}/b\right)}_{\left(\text{SD}/-\text{Leu}/-\text{Trp}/-\text{His}/-\text{Ade}\right)}}{\left(\text{time}\times {\ln }^{\left[a\right]}/b\right){}_{\left(\text{SD}/-\text{Leu}/-\text{Trp}\right)}}$$

FRET assay

Full-length MYB (ThMYB1, ThMYB2, TaMYB1, or mutated forms of TaMYB1) fused with RFP and full-length bHLH (ThbHLH1, ThbHLH2, TabHLH1, or mutated forms of TabHLH1) fused with GFP were cloned into the pCAMBIA1300-35S vector. We co-transferred 1300-35S::MYB-RFP and 1300-35S::GFP-bHLH into tobacco leaves by A. tumefaciens-mediated transformation as described by Li et al. (2016). The fluorescence signal was observed by confocal microscopy (Zeiss LSM 780) after 36 h of infiltration. The primers used for construct preparation are provided in Supplemental Table 7.

A FRET assay was used to examine the interaction between MYB and bHLH proteins and was performed as described by Li et al. (2016). The RFP fluorescence was bleached, and the pre- and post-bleach GFP fluorescence values were recorded. The FRET efficiency was then calculated according to the following equation: E_FRET = (F_[post] – F_[pre])/F_[post], where F refers to the GFP fluorescence values.

Southern blotting assay

T. aestivum L. (Ms1, AABBDD), the ms1-4Ag chromosomal addition line (AABBDD + 4Ag/4Ag), the T. turgidum L. accession Langdon (AABB), the T. urartu accession G1812 (AA), the Ae. tauschii accession AL8/78 (DD), and Th. ponticum (StStE^eE^bE^x) were used. Genomic DNA was extracted by the CTAB method. Forty micrograms of DNA from each sample was completely digested at 37°C with Hind III (Takara Bio Inc.). Agarose gel electrophoresis, membrane transfer, and hybridization were performed as described by Wang et al. (2017). The primers for the probe are listed in Supplemental Table 7. The membranes were probed and then analyzed using a chemiluminescence kit (RPN2106; GE Healthcare).

Co-IP assays

Full-length ThMYB1 or ThMYB2 fused with MYC and full-length ThbHLH1 fused with FLAG were cloned into the pCAMBIA1300-35S vector. We transferred 1300-35S::ThMYB1-MYC, 1300-35S::ThMYB2-MYC, and 1300-35S::ThbHLH1-FLAG either individually or together into tobacco leaves by A. tumefaciens-mediated transformation as described by Li et al. (2016). Total protein was extracted from 1.0 g of injected tobacco leaves using 500 μL of IP buffer (100 mM Tris–HCl [pH 7.5], 150 mM NaCl, 0.2% Triton X-100, 1 mM PMSF, and 1× protease inhibitor cocktail). The supernatant was collected and incubated with Anti-c-Myc Magnetic Beads (Cat# 88842, Thermo Fisher Scientific) for 3 h. The supernatant was then discarded, the beads were washed three times with IP buffer, 1× loading buffer was added, and the samples were boiled at 100°C for 10 min.

After denaturation, the protein samples were separated by 10% SDS–PAGE. ThMYB1-MYC and ThMYB2-MYC were detected using anti-MYC antibodies (SAB2103448; Sigma-Aldrich). ThbHLH1-FLAG was detected using anti-FLAG antibodies (DYKDDDDK-Tag(3B9) mAb, Abmart).

Seed germination assay

Seeds of pThbHLH1::ThbHLH1 + pThMYB1::ThMYB1 transgenic lines #1 and #2 and the recipient parent of the transgenic lines, Fielder, were used for the germination assay. One hundred seeds of each line were placed in medium with 10 mL 15% PEG6000 or 300 mM NaCl at 22°C for 5 days in a Petri dish (10-cm square), or seeds were placed at 22°C for 5 days in a Petri dish after being kept at −20°C, −15°C, or −10°C for 60 days for low-temperature treatment. Untreated seeds of each line were placed in medium at 22°C for 5 days in a Petri dish as a control. The embryo side of the seed was placed facing upward. Seeds were considered to have germinated when the radicle had broken through the seed coat. Germinated seeds were counted every 3 h during the germination process; three biological replicates were performed for each condition and line.

Analysis of genomic collinearity among Triticeae species

High-quality genomes and protein sequences from 13 Triticeae species were used: H. vulgare (Morex V3) (Mascher et al. 2021), H. muticum (H559) (Kuang et al. 2022), Th. elongatum (ASM1179987v1) (Wang et al. 2020), S. cereale (HAU_Weining_v1.0) (Li et al. 2021), Ae. searsii (cv. TE01), Ae. longissima (cv. TL05), Ae. sharonensis (cv. TH02), Ae. bicornis (cv. TB01), Ae. speltoides (cv. TS01) (Li et al. 2022b), Ae. tauschii (Aet v5.0) (Wang et al. 2021), T. urartu (Tu2.1) (Ling et al. 2018), T. turgidum (WEW v2.1) (Zhu et al. 2019), and T. aestivum (IWGSC RefSeq v2.1) (Zhu et al. 2021). The protein sequences of chromosome 4 were extracted for all species (except S. cereale, for which those of chromosome 7 were used), and collinearity analysis was performed using MCScanX (Wang et al. 2012) with blastp (v.2.11.0+).

Re-sequencing and SNP calling

This analysis used high-depth sequencing datasets from 756 Triticeae accessions, including 282 T. aestivum accessions (NCBI: PRJNA550304) (Li et al. 2022a), 116 T. turgidum accessions, 7 T. ispahanicum accessions, 28 Ae. tauschii accessions, 29 T. urartu accessions, 42 T. monococcum accessions (NCBI: PRJNA663409) (Zhou et al. 2020), and 240 H. vulgare accessions (NCBI: PRJEB36577) (Jayakodi et al., 2020). In addition, 10 more H. vulgare accessions (NCBI: PRJNA1012363) were re-sequenced by paired-end (150-bp) sequencing on the Illumina NovaSeq platform (Novogene). Raw reads were trimmed with fastp (v.0.20.1), and filtered reads were mapped to the reference genome using BWA–MEM (v.0.7.17) (Heng 2013). Sequences from all Triticum and Aegilops accessions were mapped to IWGSC RefSeq v2.1 (Zhu et al. 2021), and sequences from H. vulgare accessions were mapped to Morex V3 (Mascher et al. 2021). Duplicated reads were masked and variants were called using GATK (v.4.1.4.0) (https://software.broadinstitute.org/gatk/). SNPs called from the H. vulgare accessions were filtered using VCFtools (v.0.1.16) (Danecek et al. 2011) with a minor-allele frequency (MAF) of less than 0.01 and a missing rate of greater than 0.9 from approximately 116 million SNPs; 21 040 222 SNPs were selected for further analysis.

Chloroplast genome phylogeny

The chloroplast genome sequences of 32 Triticeae accessions were used to clarify the phylogenetic relationships among the Triticeae species: 3 H. vulgare accessions (NCBI: OX338658, OX338656, and OX338659), 3 Th. elongatum accessions (NCBI: MW888707, NC_043841, and MH331643) (Chen et al. 2021), 3 S. cereale accessions (NCBI: MZ507427, LC649171, and MW557517) (Du et al. 2022; Skuza et al. 2022), 3 Ae. searsii accessions (NCBI: KJ614413, NC_024815, and KJ614415), 2 Ae. longissima accessions (NCBI: MG958549 and NC_024830), 2 Ae. sharonensis accessions (NCBI: KJ614417 and NC_024816), 1 Ae. bicornis accession (NCBI: NC_024831), 3 Ae. speltoides accessions (NCBI: MG958553, KJ614406, and KJ614405) (Gornicki et al. 2014), 3 Ae. tauschii accessions (NCBI: MN223977, MN223976, and MN223975) (Su et al. 2020), 3 T. urartu accessions (NCBI: KJ614411, MG958555, and NC_021762) (Gornicki et al. 2014; Middleton et al. 2014), 3 T. turgidum accessions (NCBI: MG958552, KJ614402, and MG958550) (Gornicki et al. 2014), and 3 T. aestivum accessions (NCBI: CM022232, MW575926, and MW548259) (Zimin et al. 2017; Huaizhu and Xue 2021). The chloroplast genomes of 12 H. vulgare accessions in Supplemental Figure 7 were reassembled using GetOrganelle (version 1.7.7.0) (Jin et al. 2020); MCMCTree (Yang 2007) was used to predict species divergence times. The chloroplast genome sequences of Zea mays (NC_001666.2) (Maier et al. 1995), Sorghum bicolor (NC_008602.1) (Saski et al. 2007), Oryza sativa (GU592207) (Nock et al. 2011), and Brachypodium hybridum (LT558624) (Sancho et al. 2018) were used as outgroups. A maximum likelihood (ML) tree was constructed in IQ-TREE (version 1.6.12) (Nguyen et al. 2015), and iTOL version 6.8 (Letunic and Bork 2021) was used to visualize the trees.

Genomic phylogeny and genome-wide selection analysis

One hundred thirty-two H. vulgare accessions with known grain colors and genome sequences were used for this analysis. A phylogenetic tree of H. vulgare was constructed based on whole-genome sequences using genome-wide SNP data. The SNPs called from H. vulgare accessions were filtered using VCFtools (v.0.1.16) (Danecek et al. 2011) with a MAF of less than 0.01 and a missing rate of greater than 0.9 from approximately 116 million SNPs; 21 040 222 SNPs were selected, and the SNPs located on genes were filtered. An ML tree was constructed in IQ-TREE (version 1.6.12) (Nguyen et al. 2015) based on the coding regions of the whole genome sequences, and iTOL version 6.8 (Letunic and Bork 2021) was used to visualize the tree.

Genomic regions under selection were identified using the 21 040 222 SNPs among the 132 H. vulgare accessions. The nucleotide diversity (Pi) of white- and blue-grained accessions was calculated using VCFtools (v.0.1.16) (Danecek et al. 2011) with --window-pi 10000 and --window-pi 100000. The Pi ratio was calculated as Pi_{(white grain)}/Pi_{(blue grain)}, and regions with the top and bottom 10% of Pi ratios were identified as candidate selection regions. The F_ST values of white- and blue-grained accessions were calculated using VCFtools (v.0.1.16) (Danecek et al. 2011) with --fst-window-size 10000 and --fst-window-step 100000, and the regions with the top 10% of F_ST scores were identified as the candidate selection regions. The XP-CLR test was calculated using XP-CLR (v.1.0) (Chen et al. 2010) with window size 10000 and step 100000, and the regions with the top 10% of XP-CLR scores were identified as the candidate selection regions.

Seed sorting

A fluorescent sorter (with a red filter) and a visible light sorter were used to sort pink (RFP labeled) and blue seeds, respectively. One hundred pink or blue seeds were mixed into 200 000 white seeds. The sorting was repeated twice for each replicate, and four biological replicates were performed for each sorting experiment. The sorting efficiency was calculated based on the number of pink or blue seeds detected by the sorter from the 100 pink or blue seeds mixed into the white seeds.

Data and code availability

All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplemental Materials.

Funding

This research was supported by grants from the National Key Research and Development Program of China (2017YFD0101001 to L.M.) and the Beijing Municipal Government Science Foundation (IDHT20170513 to L.M.).

Acknowledgments

We thank S. Chen for analyzing RNA-seq data; Z. Ni., H. Yang, Y. Li, and X. Shang for suggestions on the manuscript; S. Dai for providing seeds of the Aegilops species; the Chinese Academy of Agricultural Sciences for providing seeds and seed-color information for the H. vulgare species; X. Sheng for help in performing the FRET assay; and H. Li for providing the bioinformatics analysis platform. The authors have a patent application.

Author contributions

L.M. conceived and conceptualized the study. L.M., J.P., Z.L., Z.W., Y.C., and X.W.D. designed the experiments. J.P. carried out most of the experiments and the bioinformatics analyses. Z.W., Y.H., and Z.C. characterized the ms1-4Ag wheat–T. ponticum alien addition line and transgenic lines, performed the Southern blot analysis, and prepared the constructs for genetic transformation. Z.W. and Z.C. performed RT–PCR and qRT–PCR analyses on the ms1-4Ag wheat–T. ponticum alien addition line. Q.X. analyzed chloroplast DNA sequences. L.J. performed the seed sorting assay. H.H. analyzed the RNA-seq data. K.W. and X.Y. performed the genetic transformation of wheat. Z.H. prepared DNA from Aegilops species. L.M., J.P., and Z.L. wrote the manuscript. Z.W. and X.W.D. provided critical editing of the manuscript.

Published: January 31, 2025