Pieces of acyanic internal sheaths from seedlings of onion ‘California Early Red’ were cultured on 1/2 MS medium under 20–50 μmol m^-2 s^-1 light from Osram 36 W Grolux fluorescent tubes (16 h photoperiod) at 25°C. Under these conditions, anthocyanin production was sporadically initiated by 24 h and increased over 96 h. MYB1 was isolated from tissue at 48 h using 3′- and 5′-RACE (Invitrogen 5′-RACE system). Genomic regions corresponding to the MYB1 cDNA sequence were isolated using PCR with primers designed to the cDNA. A genomic region upstream of the coding sequence was obtained from three overlapping PCR products using the GenomeWalker^TMkit (Clonetech, Mountain View, CA, USA). Primer sequences are given in Supplementary Table 1.

SG4 (MYB4, MYB5) gene fragments were first isolated by 3′-RACE as described for MYB1, with full-length sequences subsequently identified from a transcriptome database generated using TGICL software as described in Crowhurst et al. (2008) from sequences from the following sources: GenBank NT division mRNA sequences, Onion ESTs (Kuhl et al., 2004), RNASEQ data from NCBI BioProjects PRJNA175446 and PRJNA175449 (Duangjit et al., 2013), and PRJNA238142 and PRJNA60277 (Baldwin et al., 2012). The SG7 (MYB29) sequence also was identified from the transcriptome database. Primer sequences are given in Supplementary Table 1. The onion MYB sequences were assigned numbers to reflect the order in which they were identified, as per the guidelines of Gray et al. (2009).

Maximum likelihood phylogenetic trees based on full-length sequences were generated using PhyML with 1000 bootstrap replicates (Guindon and Gascuel, 2003) upon MUSCLE (Edgar, 2004) amino acid alignments (with manual correction) within Geneious (v7). Analysis of the putative TF recognition sites in the promoter regions used the PLACE database (Higo et al., 1999). GenBank database accession numbers are Allium cepa MYB1 (KX785130), MYB4 (KX785131), MYB5 (KX785132), MYB29 (KX785133); Antirrhinum majus ROSEA1 (ABB83826), MIXTA (CAA55725), MYB308 (P81393); Aquilegia formosa MYB17 (ACQ82820); Arabidopsis thaliana, MYB4 (At4g38620), MYB12 (ABB03913), MYB20 (NP_176797), MYB24 (AF175987), MYB75/PAP1 (AAG42001), MYB79 (AEE83284), MYB103 (AED96722), MYB106 (AEE73615), MYB123/TT2 (CAC40021); Dahlia pinnata (syn. variabilis) MYB1 (AB601003); Gerbera hybrida MYB1 (CAD87007), GMYB10 (CAD87010); Glycine max MYBZ2 (LOC780553); Gynura bicolor MYB1 (BAJ17661); Ipomoea nil MYB1 (BAE94391); Lilium (hybrid division I) MYB6 (BAJ05399), MYB12 (BAJ05398); Lotus japonicus TT2a (AB300033); Malus x domestica MYB10 (ACQ45201), MYB16 (ADL36756); Medicago truncatula LAP1 (ACN79541); Oncidium Gower Ramsey MYB1 (ABS58501); Oryza sativa C1 (BAD04024), IF35 (BAB20661), Plw-OSB1 (BAB64301); Petunia x hybrida AN2 (AAF66727), PH4 (ADX33331); Phalaenopsis schilleriana UMyb6 (FJ039860); Picea mariana MBF1 (U39448); Vanda hybrida MYB6 (ADQ57817); Vitis vinifera MYBA1 (BAD18977), MYB4a (ABL61515), MYBPA1 (AM259485), MYBF1 (FJ948477); Zea mays C1 (AAA33482), P1 (AAC49394), MYB38 (P20025).

Quantitative RT-PCR (qPCR) was used to measure transcript abundance for MYB1, CHS-A, and DFR-A in white and red colored regions of onion seedlings. DFR-A (AY221249) is the only significantly active DFR gene in the onion bulb (Kim et al., 2004a) and CHS-A (AY221244) is one of the two significantly active CHS genes (Kim et al., 2005b). Outer leaves of red onion seedlings were removed to expose the first fully white sheath beneath. Each seedling was cut in half longitudinally and the white sheath tissue of one half was immediately flash frozen in liquid N₂. The other half was placed on moistened paper in a Petri dish and left to become pigmented. Pigmented sheath tissue was sampled after approximately 3 days. Two samples were analyzed, each consisting of sheath tissue from three seedlings. Total RNA was isolated using the RNeasy Plant Mini Kit (Qiagen). First strand cDNA was made from 500 ng DNAse I-treated RNA using the Transcriptor First Strand cDNA Synthesis Kit (Roche) and priming from the supplied anchored oligo dT₁₈. qPCR (on cDNA diluted 50-fold) was performed as described in Albert et al. (2011). Relative transcript abundance was determined by comparative quantification. Primers used are given in Supplementary Table 1.

Overexpression vectors for MYB1 (pKES22) and the maize anthocyanin bHLH factor Lc (pNASA3) were generated by cloning the respective cDNAs behind a CaMV 35S promoter (35Spro) in pART7. The RNAi vector for MYB1, pHMN3, was based on pDAH2 (Davies et al., 2012b). The construct, driven by 35Spro, had inverted repeats to give hairpin RNA. Primers used in making these constructs are given in Supplementary Table 1. Other constructs used were for a GFP internal transformation control (pPN93; 35Spro:GFP-ER; Shang et al., 2007), a snapdragon anthocyanin bHLH factor (pJAM1528; 2x35Spro:AmMutabilis in pJIT60; supplied by Cathie Martin) and a maize anthocyanin MYB factor (pLN44; 35Spro:ZmC1 in pART7; Albert et al., 2010). Constructs were transformed into dorsal petals of the snapdragon rosea^dorsea mutant line (Schwinn et al., 2006), seedling tissue of onion ‘California Red,’ and shoot tissue from bulbs of an unidentified commercial cultivar of garlic (Allium sativum). Particle bombardment using a helium-driven particle in flow gun was carried out generally as described in Davies et al. (2012b). Tissue was not surface sterilized before bombardment. Each experiment involved the bombardment of at least three pieces of tissue with each piece bombarded once (onion/garlic) or twice (snapdragon). Each experiment was replicated at least twice except for the experiment involving pJAM1528. Bombardment used a 30 ms burst of helium at a pressure of 200 kPa for the RNAi experiments, 300 kPa for the other experiments in onion and garlic, and 400 kPa for the experiments in snapdragon. DNA concentrations in the gold particle preparations were 0.4 μg/1 mg gold for the GFP construct, which was added to all preparations, and either 0.8 or 1.0 μg/1 mg gold for each of the other constructs added to a preparation. Post-bombardment onion tissue was placed on moistened filter paper in a Petri dish. Images were recorded using a Leica M205FA microscope with a Leica DFC550 digital camera or an Olympus SZX fluorescent microscope (Olympus Corp., Tokyo, Japan) with a Leica DC500 digital camera (Leica Camera AG, Solms, Germany). Any post image-capture adjustments of brightness or contrast for clarity of the GFP signal were applied equally to all images in a figure.

Transgenic garlic plants (‘Printanor’) were produced using the method of Kenel et al. (2010). Cassettes for both 35S:MYB1 and 35S:ZmLc were introduced into pART27H, a modified version of pART27 (Gleave, 1992) that has the NPTII selectable marker replaced with an HPTII (hygromycin) expression cassette. A schematic of the plasmid is shown in Supplementary Figure 2. Presence of the selection marker HPTII was detected using PCR on genomic DNA isolated from tissue culture leaf tissue.

Candidate R2R3-MYB genes for anthocyanin regulation were amplified using 3′-RACE PCR from cDNA derived from RNA purified from onion tissue that was induced to produce anthocyanins (see “Materials and Methods”). Degenerate oligonucleotides designed to the conserved MYB DNA-binding domain of R2R3-MYB genes were used as forward primers (Supplementary Table 1). Several MYB-related sequences were obtained, including one sequence with similarity to anthocyanin-related SG6 MYB genes. The partial cDNA sequence for this candidate gene was extended by 5′-RACE, followed by PCR amplification of a full cDNA coding sequence. This gene was named MYB1 (GenBank accession KX785130). The MYB1 cDNA is 1036 bp long with three possible ATG translation initiation codons, all of which are in frame from the first ATG. The second of these has the best fit to the Kozak consensus sequence, and on this basis the deduced polypeptide sequence consisted of 253 amino acids. The closest matches to MYB1 on a BLASTp search of the NCBI non-redundant protein sequence database were anthocyanin-related R2R3-MYB proteins, but from dicot rather than monocot species (data not shown).

Based on data from other species it was expected that a small multigene family of anthocyanin-related R2R3-MYB activators would be present in onion. A second 3′-RACE sequence was isolated that was identical to MYB1 except for possessing a 92 bp direct repeat in the coding sequence, and a truncated 3′-UTR (21 bp in length) containing one single nucleotide polymorphism. Compared with MYB1, the direct repeat altered the predicted polypeptide sequence, causing 10 sequential variant amino acids followed by a truncation of 75 amino acids in the C-terminus due to a premature stop codon (data not shown). It is not known if this cDNA represents an allele of MYB1. To look for further MYB1 related alleles or genes, a subsequently available in-house transcriptome database was searched (BLASTn and BLASTp) with the MYB1 sequence. The transcriptome examined contained approximately 296, 942 singleton and contig assemblies from sequencing 13 different onion mRNA derived libraries. No additional anthocyanin-related MYB genes were detected. However, four partial sequences (between 225 and 615 bp) were identified that differed from MYB1 by less than 10 nucleotides and which probably represent either allelic differences or sequencing errors (data not shown).

The genomic MYB1 sequence (from the ATG furthest upstream through the stop codon) was isolated from the same onion cultivar used to isolate the full cDNA. It was 4.915 kb in length and the gene contained three exons and two introns (intron 1 is 0.099 kb and intron 2 is 4.027 kb; Figure 1), a common structure in plant R2R3-MYB genes. In addition a total of 1.914 kb upstream of the distal start codon was isolated by genome walking. To understand how this gene may be regulated in response to light and other factors, the PLACE database for predicted DNA-binding recognition sites (Higo et al., 1999) was used to analyze the proximal 1.0 kb of this upstream region. A predicted TATA box was located at -25 bp. There was a cluster of predicted recognition motifs of interest within the -500 bp region, including candidate MYB and bHLH sites and a range of sites found in promoters of genes up-regulated by light, such as the T-Box, GT-1, and SORLIP1AT motifs (Figure 1; Supplementary Figure 1).

Flavonols are other key flavonoid metabolites in onion bulbs. Given that R2R3-MYB proteins of SG7 have been found to regulate the production of flavonols in a range of dicot species and in the monocot maize, we reasoned that SG7 MYB genes regulate flavonol production in onion. Several highly similar sequences were retrieved from a transcriptome database when searched (BLASTp) with the sequence for the flavonol regulator At-MYB12 from Arabidopsis, including a singleton sequence encoding a full-length candidate, MYB29 (GenBank accession KX785133). The MYB29 sequence was 1076 bp long, with a putative 5′- and 3′-UTR of 19 and 187 bp, respectively. The EST originated from a mixed tissue mRNA sample of the doubled-haploid red onion genotype 5225. Analysis of related sequences identified one additional SG7 gene family member in the data from the doubled-haploid line, represented by a partial 5′-EST (data not shown).

The deduced amino acid sequence of MYB29 was used to search (BLASTp) the NCBI non-redundant protein sequence database. The closest matches to MYB29 were At-MYB12 and another SG7 protein, Zm-P1, a regulator of flavone and phlobaphene production in maize. The motif identified as conserved among flavonol-related SG7 R2R3-MYBs ([K/R][R/x][R/K]xGRT[S/x][R/G]xx[M/x]K; Stracke et al., 2001) was fully conserved in MYB29 as KKRKGRTSRSAMK.

The initial 3′-RACE screen for anthocyanin MYBs predominantly isolated R2R3-MYB sequences belonging to SG4. Characterized SG4 sequences from other species encode proteins with a repressive action on genes of the phenylpropanoid pathway. Full-length cDNAs for two of the genes from the screen, MYB4 and MYB5 (GenBank accessions KX785131 and KX785132, respectively), were subsequently identified in the transcriptome database. The MYB4 sequence was 844 bp long with a deduced peptide of 213 amino acids, and matched a partial EST present on GenBank (CF441129). MYB5 was 1144 bp in length with a deduced peptide of 236 amino acids. The closest matches to MYB4 or MYB5 on a BLASTp search of the GenBank Swiss-Prot database were Am-MYB308 from snapdragon and At-MYB4 from Arabidopsis. As with other members of SG4, MYB4, and MYB5 contain in their C-terminals an ERF-associated Amphiphilic Repression (EAR) domain (C2 motif) and the lsrGIDPxT/NHR (C1) motif as defined by Kranz et al. (1998) (Supplementary Figure 2).

The relationships of MYB1, MYB4, MYB5, and MYB29 to a range of characterized, flavonoid-related R2R3-MYBs were examined by phylogenetic analysis (Figure 2). The candidate onion sequences grouped with the expected flavonoid regulators. Specifically, MYB29 grouped with SG7 (flavonol/flavone-related) factors, while MYB4 and MYB5 grouped with SG4 (repressors of flavonoid biosynthesis) factors, although in different subclades within SG4. MYB5 showed some amino acid residue changes within the conserved MYB domain as well as some variation in other C-terminal motifs that have been categorized by Cavallini et al. (2015) (Supplementary Figure 2).

There are two clades within the anthocyanin-related R2R3-MYBs, SG5, and S6, based on sequences that are currently available (Yamagishi et al., 2010; Feller et al., 2011; Davies et al., 2012a). These may be defined as whether sequences separate with Arabidopsis At-MYB75 (PAP1) or At-MYB123 (TT2), which regulate anthocyanin and proanthocyanidins (PAs or condensed tannins), respectively. MYB1 grouped with anthocyanin-related factors in SG6, including ones from another Asparagales taxon (Iris fulva; If-SG6) and a taxon in the monocot order Liliales (Lilium hybrida, Asiatic lily; Lh-MYB6, Lh-MYB12). This contrasts with the anthocyanin-regulating R2R3-MYBs from orchid taxa (Og-MYB1, Vh-MYB6, Ps-MYB6), which despite being members of the Asparagales, are located within SG5 along with representative sequences from the monocot order Poales.

Alignment of a region of 104 amino acid residues containing the MYB domain of SG5 and SG6 sequences (including MYB1) identified five conserved amino acid or motif differences among the clades (Supplementary Figure 3). These were [R] versus [N/H/K], [C/F] versus Y, R versus G, X versus N, and A versus D for the SG6/PAP1 versus SG5/C1-TT2 clade, respectively. These differences are strongly conserved within each clade and are a likely cause of the clade separation. Notably, the A versus D (position 90 in Supplementary Figure 3) was the start of a DNEI motif that is absolutely conserved in the SG5 sequences (and other subgroups like SG4 and SG7) but absent in the SG6 sequences, where the motif was ANDV in all but one example. Both SG5 and SG6 proteins (including MYB1) contain a conserved motif ([D/E]Lx₂[R/K]x₃Lx₆Lx₃R) within the MYB domain that is necessary to bind bHLH partners (Zimmermann et al., 2004), and is an essential component to the formation of MBW activation complexes.

Within the variable C-terminal region, a motif has been identified as defining the anthocyanin-related SG6 proteins. This was initially identified as KPRPR[S/T]F when SG assignments were made for Arabidopsis (Stracke et al., 2001), and later modified to [K/R]P[Q/R]P[Q/R] based on the Asiatic lily SG6 sequences (Yamagishi et al., 2010). This motif was reasonably conserved within MYB1, as KPQPxxx, but is very poorly conserved in the Orchidaceae and Poaceae sequences that form the C1-TT2 clade (Figure 3). A second motif, closer to the C-terminal, was conserved within the C1-TT2 clade, although it aligned less consistently than the previously defined KPRPR[S/T]F. This is VWAPKAVRCT in C1 and VIRTKAIRCS in TT2, and it was strongly conserved except in Vh-MYB6 (Figure 3). The V[V/I]RTKA sequence is strongly conserved in a range of R2R3-MYBs involved in activation of PA production (Liu et al., 2014).

The spatial and temporal expression patterns of the R2R3-MYB genes have been shown to be important in determining anthocyanin phenotypes in a range of species. Transcript abundance for MYB1 and two anthocyanin biosynthetic genes, chalcone synthase (CHS) and DFR, was compared between pooled samples of white sheaths and sheaths in which anthocyanin production was induced. Transcript amounts for MYB1, CHS, and DFR were much higher in the scales that had become pigmented (Figure 4).

The ability for MYB1 to activate anthocyanin synthesis was tested in both onion and garlic using biolistic transformation (Figure 5; Supplementary Figure 4). Introduction of 35S:MYB1 into acyanic sheath tissue of inner leaves or inner sprouts of red onion seedlings resulted in single cell red foci at low frequency that were visible by 48 h post-bombardment. However, co-bombardment of 35S:MYB1 with constructs for heterologous anthocyanin bHLH regulators (Zm-Lc and Am-Mutabilis) resulted in a greater frequency of red cells (Figure 5; Supplementary Figure 4) and these became visible earlier. Particularly in green sprout tissue, the red cells were visible within 12–24 h post-bombardment. The anthocyanin pigmentation co-localized with GFP fluorescence, further indicating that the pigmentation was due to introduction of the TFs and not from autonomous activation of pigmentation. Autonomous activation does occur, but its delayed timing allowed sufficient time to conduct the experiments. Introduction of the bHLH factor alone did not induce anthocyanin production, nor, interestingly, did the introduction of the SG5 anthocyanin regulator from maize, ZmC1, in conjunction with Zm-Lc (Supplementary Figure 4). Anthocyanin synthesis was also activated in garlic shoot tissue after co-bombardment with MYB1 and bHLH (Zm-Lc) constructs (Figure 5), but not when MYB1 alone was used.

To investigate further whether MYB1 was a key gene determining anthocyanin pigmentation in onion, transient RNAi was conducted using biolistic introduction of a hairpin construct for MYB1. The autonomous induction of anthocyanin production in seedling leaf sheath tissue described earlier allowed the effect of RNAi against MYB1 to be observed. Acyanic sheath tissue was bombarded with the 35S:MYB1-RNAi and 35S:GFP constructs after it was removed from red onion seedlings, and incubated to allow pigmentation to develop. In tissue bombarded with the MYB1 RNAi construct, zones with greatly reduced or absent pigmentation became visible by 48 h post-bombardment and by 72 h they were more prominent, as the pigmentation in the background of untransformed cells developed (Figure 6). As expected for RNAi, the inhibition was not cell autonomous, but rather spread to adjacent cells, as indicated by multi-cellular acyanic regions surrounding the central transformed cells (GFP positive cells). Bombardment with 35S:GFP alone had no visible effect on the anthocyanin phenotype (Figure 6).

Garlic rather than onion was chosen for this experiment because it is technically easier to transform. To maximize potential for anthocyanin production in the transgenic plants, garlic was transformed via Agrobacterium with both 35S:MYB1 and 35S:Zm-Lc transgenes. The transgenes were present on the same T-DNA, in a binary vector containing a hygromycin selectable marker (Supplementary Figure 5). Strong red pigmentation was present on the callus, and this was visible from early in the transformation/regeneration process (Figure 7). This phenotype has not been observed when using other gene constructs with the same tissues and methods (e.g., Kenel et al., 2010). The young regenerated transgenic plants had darker pigmentation of the leaves than control plants and red pigmentation at the base of the leaves, where the bulb would develop. Four independent transformation events were taken through to mature plants (1111-1A, 1111-3A, 1111-7A, 1111-8A). These showed variable intensity of novel pigmentation, including light-pink (1111-3A) and strong red (1111-8A). The plants were maintained through a season of bulb formation, plant dieback and dormancy, and bulb sprouting in the following spring. The newly formed foliage continued to show darker pigmentation of the leaves and red pigmentation at leaf bases (Figure 7).

Given that MYB1 is phylogenetically more similar to dicot anthocyanin regulators than those from orchid or grasses (i.e., from SG5), we tested whether it could substitute for a dicot regulator. Along with a GFP internal control, the 35S:MYB1 construct was biolistically introduced into acyanic petal tissue of the rosea^dorsea anthocyanin MYB mutant of snapdragon (Schwinn et al., 2006). MYB1 was able to complement the mutant, resulting in cell autonomous production of anthocyanin pigment, which co-localized with GFP fluorescence (Figure 8). Introduction of a reporter gene alone does not restore pigmentation in this mutant (Schwinn et al., 2006).

A candidate gene approach was taken to identify genes that regulate the production of anthocyanin pigments in onion. MYB1, a sequence sharing similarity to characterized anthocyanin-related MYB genes, was identified from red onion. In these onions, anthocyanin pigmentation is first present in the seedlings. It occurs in the swelling, predominantly achlorophyllous sheath tissue of the first few outer layers of the leaves. As the bulb develops and matures, pigmentation extends throughout the inner layers, comprised of the swollen sheaths and scales. MYB1 expression (along with expression of the investigated biosynthetic genes, CHS and DFR) was correlated with anthocyanin induction in sheath tissue of the seedlings (Figure 4). Furthermore transient overexpression and knockdown experiments induced and inhibited, respectively, anthocyanin production in the seedlings (Figures 5 and 6), and ectopic anthocyanin production was observed in garlic co-transformed with transgenes encoding MYB1 and a heterologous bHLH factor (ZmLc; Figure 7). These results demonstrate that MYB1 positively regulates anthocyanin pigmentation and we conclude that MYB1 is key for the red bulb phenotype of this crop. As such it is an excellent candidate for underlying a major Quantitative Trait Locus (QTL) controlling anthocyanin concentration and intensity of red bulb color. Duangjit et al. (2014) have identified two such QTLs that are different from the regions controlling red versus yellow bulbs that have been attributed to the L locus (which encodes the anthocyanin biosynthetic gene ANS) by Kim et al. (2006). They proposed that these newly characterized QTLs may be associated with factors that regulate the pigment pathway.

Vegetative tissues of plants often accumulate anthocyanins in response to light (Gavazzi et al., 1990; Dubos et al., 2008; Albert et al., 2009), particularly in juvenile tissues that may be more sensitive to excess irradiation (Gould, 2004). In onion seedlings only a few of the outer layers of leaves showed anthocyanin pigmentation in the sheaths and when these were peeled away, the exposed acyanic sheath tissue beneath rapidly became pigmented in our experimental conditions. This suggests that light is an important signal in anthocyanin production in onion, at least early in plant development. This signal is probably integrated through control of MYB1. The promoter region of MYB1 has recognition motifs that have been observed in light regulated genes (Supplementary Figure 1). In addition, the induced accumulation of anthocyanins in acyanic inner leaf sheaths was correlated with the expression of MYB1 (Figure 4). Juvenile shoots of Asiatic lily produce anthocyanins upon exposure to light because of the expression of the anthocyanin regulators Lh-MYB6, Lh-bHLH1, and Lh-bHLH2 (Nakatsuka et al., 2009; Yamagishi et al., 2010). Therefore, light-controlled anthocyanin synthesis is probably regulated by a similar mechanism in both lily and onion. Other signals are expected to be involved in the regulation of anthocyanin production in red onion since by the time the bulbs mature the inner swollen sheaths and scales, all substantially thickened tissues, have become pigmented without being directly exposed to light. It is possible that a light signal is being transduced to the inner leaves; however, pigment regulation during bulb development is likely to involve developmental signals as well.

A MBW complex regulates anthocyanin synthesis in all species studied to date. While most of the functional data for the MBW complex are from eudicots, the cooperative involvement of MYB, bHLH, and WDR proteins for regulating anthocyanin synthesis is well-established in the monocot model maize (Ramsay and Glover, 2005; Feller et al., 2011). The predicted amino acid sequence of MYB1 has the motif necessary to bind bHLH TF partners (Supplementary Figure 3), and the co-introduction of a bHLH factor with MYB1 in the transient assays allowed earlier development of anthocyanin pigmentation along with an increased frequency of pigmented cells (Figure 5). MYB1 was able also to substitute for the endogenous R2R3-MYB anthocyanin regulator in a dicot species that is a model for MBW regulation of anthocyanin pigmentation (Schwinn et al., 2006) (Figure 8). In the orchid family (also Asparagales), anthocyanin synthesis is also activated when R2R3-MYB and bHLH proteins are co-expressed (Ma et al., 2008; Albert et al., 2010), and endogenous R2R3-MYB genes that regulate anthocyanin production have been identified in the orchid genera Phalaenopsis (Ma et al., 2009) and Oncidium (Chiou and Yeh, 2008). This suggests that the MBW complex is conserved within Angiosperms, although some aspects of MBW activity have yet to be identified in monocots (Albert et al., 2014b). These include hierarchical activation of bHLH proteins, the involvement of gene-regulation networks, and the action of repressor proteins, which are conserved in eudicots (Albert et al., 2014a). In many species, the IN/AN1/MUT-clade bHLH is essential for regulating anthocyanin synthesis, even if LC/JAF13/DEL-clade bHLH proteins are expressed (Davies et al., 2012a). In Asiatic lilies, the anthocyanin-related R2R3-MYBs (Lh-MYB6 and Lh-MYB12) were able to activate the promoters of the Lh-CHSa and Lh-DFR genes only when Lh-bHLH2 (IN/AN1/MUT clade) was co-bombarded (Yamagishi et al., 2010), although Lh-bHLH1 (LC/JAF13/DEL) was not tested. In maize, the LC/JAF13/DEL clade bHLH factors regulate anthocyanin synthesis (Ludwig et al., 1989; Tonelli et al., 1991, 1994; Goff et al., 1992), but the IN/AN1/MUT-clade bHLH has lost the ability to activate transcription and acts as a competitive repressor (Burr et al., 1996). Similar patterns of anthocyanin induction were observed in the onion transient assays when bHLH factors of either clade (ZmLC of the LC/JAF13/DEL clade and AmMUT of the IN/AN1/MUT clade) were co-introduced with MYB1 (Figure 5; Supplementary Figure 4). This indicates either that MYB1 normally interacts with bHLH factors of both clades in its native environment or that it has retained the ability to function with either type. Identification of the bHLH factors important for pigmentation in Allium is required to investigate this further.

Flavonols accumulate to high concentrations in the bulbs of yellow and red onions, but are only present in trace amounts in white varieties (Patil et al., 1995). MYB29 was identified that has significant similarity to SG7 R2R3-MYBs, including the characteristic motif present in the C-terminus. While functional data are desirable, the phylogenetic support for Ac-MYB29 belonging to SG7 is very strong. In Arabidopsis, grape and tomato, the SG7 flavonol regulators coordinately regulate the early flavonoid genes CHS, CHI, flavanone 3-hydroxylase (F3H), and the flavonol-specific gene flavonol synthase (FLS; Mehrtens et al., 2005; Luo et al., 2008; Czemmel et al., 2009; Ballester et al., 2010; Stracke et al., 2010). Similarly in maize, P1 directly regulates the FLS genes (Ferreyra et al., 2010, 2012), although P1 also regulates flavone and phlobaphene synthesis in some tissues (Casati and Walbot, 2005; Grotewold, 2006; Morohashi et al., 2012). However, it is also worth noting that in maize, FLS is also regulated by the anthocyanin-related MYB and bHLH factors (Ferreyra et al., 2010), suggesting a degree of functional redundancy exists. Taking this into consideration may help to explain the nature of the C locus in onion, which is thought to encode a regulatory gene. C is required for the expression of two genes encoding CHS (Kim et al., 2005b), which is the first committed step toward the biosynthesis of flavonoids, including flavonols and anthocyanins. MYB29 and MYB1 are excellent candidates for regulating CHS, but it is not currently known if they co-localize to the C locus, which is on chromosome 6 (Khar et al., 2008).

Onion R2R3-MYB DNA sequences were also isolated that belong to the phylogenetic SG containing repressors of phenylpropanoid/flavonoid biosynthesis (SG4; MYB4, MYB5). The identification of MYB4 and -5 was based on phylogenetic (Figure 2) and amino acid sequence analysis, including the presence of the EAR repression motif in the C-terminus (Supplementary Figure 2). The plant-specific EAR domain occurs in other TF families besides the R2R3-MYBs (Kagale and Rozwadowski, 2010), but alignment of confirmed and candidate AtMYB4-like R2R3-MYB-EAR repressors for flavonoid biosynthesis from a range of species showed the (P/L)DLNL(E/D)L version of the motif to be strongly conserved, and the sequence PDLNL(E/D)L is present in MYB4, MYB5 and characterized phenylpropanoid repressors from Arabidopsis, maize, and petunia. There are dominant white bulb phenotypes in onion and a color-inhibiting factor encoded by the I locus, which is homozygous dominant and conditions white bulbs (El-Shafie and Davis, 1967). Therefore, it would be interesting to investigate repressive regulatory factors in relation to this.

Flavonoid regulatory genes have been previously characterized from only three monocot families, the Poaceae, Liliaceae, and Orchidaceae. The anthocyanin R2R3-MYBs activators from dicot species separate phylogenetically from those of the Poaceae and Orchidaceae (Figure 2; Supplementary Figure 5). The dicot sequences form the cluster termed SG6 (containing the anthocyanin regulator PAP1) based on the Arabidopsis R2R3-MYB gene family, while the Poaceae and Orchidaceae cluster with Arabidopsis SG5 (containing the PA regulator TT2). The absence of SG6 sequences from the Poaceae and Orchidaceae may suggest divergent evolution of the two MYB clades during the origins of dicots and monocots. However, the MYB genes from Asiatic lily (Liliaceae) and onion (Allioideae) cluster in the SG6/PAP1 clade. Thus, based on available sequence data it appears that the R2R3-MYBs that regulate anthocyanins have diverged within Asparagales evolution during formation of the Orchidaceae and Allioideae. Most phylogenetic analysis places the Orchidaceae as the sister group to the rest of the Asparagales, being one of the two initial phylogenetic branches (Angiosperm Phylogeny Group III, 2009), so divergence could well have occurred after this branching. The placement of the onion and Asiatic lily sequences with dicot rather than Orchidaceae/Poaceae sequences is not seen for the target genes of the flavonoid biosynthetic pathway. Rather, the biosynthetic gene sequences form phylogenetic trees close to the current suggested species taxonomic relationships (Supplementary Figure 6 shows a representative tree, for DFR). Additionally, the other flavonoid-related MYBs identified, for SG4 and SG7, group closely with the similar sequences from both dicots and the Poaceae (the only other monocot family for which sequences have been characterized).

SG5/C1-TT2 clade members were not identified in the approximately 250,000 onion Non-Redundant sequences. Nor have they been published for Asiatic lily, even though it may be expected that PA production is regulated by an R2R3-MYB in most monocots, as it is in the Poaceae (Himi and Noda, 2005; Himi et al., 2012) and dicot species (Nesi et al., 2001; Bogs et al., 2007; Hancock et al., 2012; Liu et al., 2014) studied to date. It may be that RNA sequencing for the Liliales and Asparagales species has not been conducted on samples from tissues rich in PAs. Interestingly, seed coats are typically rich in PAs but PAs have been functionally replaced by phytomelans in Allium and other taxa of Asparagales, but not in the Orchidaceae (Boesewinkel and Bouman, 1995; Fritsch and Friesen, 2002; Chase, 2004). Full genome sequences for species such as onion and Asiatic lily (∼130 and 900-fold larger, respectively, than that of Arabidopsis) are not yet available but sequencing of the gene rich regions is in progress for onion¹. Additional DNA sequence resources for monocot species other than the Poaceae may also help to address the lack of reported PAP1-like sequences in some monocot species.

The basis for the absence of identified PAP1/SG6 clade members in the Poaceae and Orchidaceae is unknown. While additional DNA sequencing may identify members of this clade in such species, no PAP1-like R2R3-MYBs have been identified in the rice genome based on analysis of the phylogenies in Katiyar et al. (2012) and nor are any apparent in current transcriptome databases for various orchid species (the authors’ unpublished analysis). It seems unlikely that PAP1-like R2R3-MYBs arose independently in dicots, the Liliales and the Asparagales – i.e., convergent evolution after the separation from the Orchidaceae. More likely is that there has been neofunctionalization of the C1-TT2 (SG5) clade in the Poaceae and Orchidaceae with the subsequent loss of the PAP1 clade. The presence of regulators of both anthocyanins and PAs in the same clade in grasses is consistent with this theory. Furthermore in this scenario, both clades would have been retained in the Liliaceae, Allioideae, and dicots, with separation of function developing for anthocyanin and PA regulation. Alternatively, the SG6 clade could have evolved from the SG5 MYBs to regulate anthocyanins, whilst SG5 genes were retained to regulate PA synthesis, reflected by the two subclades of genes associated with anthocyanins and PAs within SG5 in grasses (Figure 2). To address this question fully, further characterisation of the anthocyanin and PA regulators is required from diverse monocot and gymnosperm species.

Finally, it is worth noting that some phylogenetic differences in the anthocyanin regulators were translated into differential activities in the transient assays. In contrast to MYB1, the maize MYB regulator Zm-C1 was not able to activate pigmentation in onion even when its bHLH partner was co-introduced (Supplementary Figure 4). Indeed, the maize anthocyanin regulators have been observed to inhibit pigmentation when bombarded into onion (personal communication, Dr. David Collings). Also, Zm-C1 (-/+ its bHLH partner) was not able to complement the snapdragon anthocyanin MYB mutant (the authors’ unpublished observation). Whether this reflects evolutionary divergence or different lineages for the regulators awaits further analysis. In the first instance, it would be interesting test the anthocyanin regulators from other lineages within the Asparagales.