This Article Abstract Full Text (PDF) Supplemental Data All Versions of this Article: 18/4/831    most recent tpc.105.039255v1 Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (19) Citing Articles via Google Scholar Google Scholar Articles by Schwinn, K. Articles by Martin, C. Search for Related Content PubMed PubMed Citation Articles by Schwinn, K. Articles by Martin, C. Agricola Articles by Schwinn, K. Articles by Martin, C.

A Small Family of MYB-Regulatory Genes Controls Floral Pigmentation Intensity and Patterning in the Genus Antirrhinum^[W]

^a New Zealand Institute for Crop and Food Research Limited, Private Bag 11600, Palmerston North, New Zealand
^b Department of Cell and Developmental Biology, John Innes Centre, Norwich Research Park, Colney, Norwich NR4 7UH, United Kingdom
^c Institute of Molecular Biosciences, Massey University, Private Bag 11222, Palmerston North, New Zealand

³ To whom correspondence should be addressed. E-mail cathie.martin{at}bbsrc.ac.uk; fax 44-1603-450045.

	ABSTRACT

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

 
The Rosea1, Rosea2, and Venosa genes encode MYB-related transcription factors active in the flowers of Antirrhinum majus. Analysis of mutant phenotypes shows that these genes control the intensity and pattern of magenta anthocyanin pigmentation in flowers. Despite the structural similarity of these regulatory proteins, they influence the expression of target genes encoding the enzymes of anthocyanin biosynthesis with different specificities. Consequently, they are not equivalent biochemically in their activities. Different species of the genus Antirrhinum, native to Spain and Portugal, show striking differences in their patterns and intensities of floral pigmentation. Differences in anthocyanin pigmentation between at least six species are attributable to variations in the activity of the Rosea and Venosa loci. Set in the context of our understanding of the regulation of anthocyanin production in other genera, the activity of MYB-related genes is probably a primary cause of natural variation in anthocyanin pigmentation in plants.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

 
Some of the most striking features of insect-pollinated flowering plants are their highly distinctive patterns of floral pigmentation. Many color patterns provide honey guides that direct pollinators toward the reproductive organs and the source of nectar within the flowers (Waser and Price, 1983). In Antirrhinum majus, color patterning results from the localized production of yellow aurone pigment in the hinge region created by the ventral and lateral petals of the corolla. The yellow is surrounded by magenta anthocyanin, which is produced more or less evenly in both inner and outer epidermal layers of the corolla lobes and at somewhat lower levels in the epidermal layers of the corolla tube (Figures 1A and 1J ) (Jackson et al., 1992). This arrangement of yellow (aurone) and magenta (cyanidin) pigments provides a visual target for prospective pollinating bumblebees and probably guides them to the mouth of the fused corolla (Harbourne and Smith, 1978; Penny, 1983; Lunau et al., 1996). Some natural isolates of A. majus have additional patterning, in the form of increased pigmentation in regions of the epidermis overlying the vascular strands of the petal (Baur, 1910a, 1910b; Kuckuck and Schick, 1930; Stubbe, 1966). This venal pattern of pigmentation predominates on the inner (adaxial) epidermis of the dorsal petal lobes but continues into the inner epidermis of the corolla tube and, in contrast with the UV light–absorbing flavonoids, probably provides additional visual guides for the bees once they enter the corolla tubes (Thompson et al., 1972; Lunau et al., 1996).

View larger version (133K): [in this window] [in a new window]   Figure 1. Floral Phenotypes of Regulatory Mutants of A. majus. (A) Wild type. (B) roscol (grown outside). (C) rosdor (grown outside). (D) to (F) Phenotypes of individuals in the F2 population of roscol x rosdor. (D) roscol homozygote. (E) roscol rosdor heterozygote. (F) rosdor homozygote. (G) rosdor Ve+. (H) Surface view and transverse section of a dorsal petal from a rosdor Ve+ individual. Arrows indicate the anthocyanin in the epidermal layer in regions overlying the vascular tissue. (I) rosdor grown in the greenhouse. (J) to (L) Phenotypes of individuals in the F2 population of eluta (wild type) x Eluta. (J) eluta (wild type) homozygote. (K) Eluta eluta heterozygote. (L) Eluta homozygote.

A number of loci control anthocyanin production in flowers of the model species, A. majus. Mutations that affect the activity of structural genes encoding the enzymes of anthocyanin biosynthesis are well characterized and include mutations in the genes encoding chalcone synthase (nivea; CHS) (Sommer and Saedler, 1986), flavanone 3-hydroxylase (incolorata; F3H) (Martin et al., 1991), dihydroflavonol 4-reductase (pallida; DFR) (Martin et al., 1985), anthocyanidin synthase (candica; ANS) (Martin et al., 1991), and flavonoid 3'-hydroxylase (eosinea; F3'H) (Stickland and Harrison, 1974). Generally, knockout mutations at these loci give rise to acyanic flowers or, in the case of eosinea, flowers that produce alternative types of anthocyanin. Some mutations of these genes can give rise to patterned alleles (Coen et al., 1986; Martin et al., 1987; Coen and Carpenter, 1988; Martin and Lister, 1989), but patterned phenotypes invariably result from alleles in which the regulation of the expression of the structural gene has been affected, either through changes to the regulatory motifs of the promoters of the structural genes or through likely production of small interfering RNAs (Coen et al., 1986; Martin et al., 1987; Coen and Carpenter, 1988; Almeida et al., 1989; Martin and Lister, 1989; Robbins et al., 1989; Bollmann et al.,1991; Della Vedova et al., 2005). A number of mutations that affect the activity of regulatory genes that control the expression of the structural genes of floral pigment production have also been described, including delila (del), Eluta (El), rosea (ros), Venosa (Ve), and mutabilis (mut) (Wheldale, 1907; Baur, 1910a, 1910b; Kuckuck and Schick, 1930; Stubbe, 1966). Mutations in the regulatory genes do not abolish pigmentation but change the pattern of pigmentation within the flowers: Delila affects pigmentation in the corolla tube; Mut affects pigmentation in the corolla lobes; Rosea affects the pattern and intensity of pigmentation in both lobes and tubes; and Venosa affects pigmentation of the epidermis overlying the veins in both lobes and tubes. Of the regulatory genes, only Delila has been characterized molecularly and shown to encode a basic helix-loop-helix (bHLH) transcription factor that is required for the activation of expression of the late biosynthetic genes (including F3H, DFR, AS, and UDP-glucose 3-O-flavonoid transferase [UFGT]) in the corolla tube (Martin et al., 1991; Goodrich et al., 1992; Martin and Gerats, 1993). The other regulatory loci also influence the levels of transcripts of the biosynthetic genes active late in the anthocyanin biosynthetic pathway for their control of pigment patterning (Martin et al., 1991; Schwinn, 1999).

In maize (Zea mays), it has been shown that two types of transcription factor, a MYB-related protein and a bHLH-containing protein, interact to activate genes in the anthocyanin biosynthetic pathway (Cone et al., 1986, 1993; Paz-Ares et al., 1986, 1987; Chandler et al., 1989; Ludwig et al., 1989; Consonni et al., 1992, 1993; Goff et al., 1992; Sainz et al., 1997b; Zimmermann et al., 2004). In different accessions of maize, the genes encoding these proteins have been variously amplified, such that genes encoding functionally equivalent proteins are active in different tissues of the vegetative and reproductive phases of growth (Chandler et al., 1989; Cone et al., 1993; Pilu et al., 2003). Some of the resultant variation in plant pigment patterning may be attributable to human selection for more exotic pigmentation forms, because such lines were highly prized by the indigenous peoples of Central America (Lonnig and Saedler, 1997). In maize, an additional gene, pac1, encodes a WD repeat protein that affects the levels of anthocyanin production in kernels (Carey et al., 2004). By analogy to the activity of the homologous protein TRANSPARENT TESTA GLABRA1 (TTG1) in Arabidopsis thaliana (Walker et al., 1999), it seems likely that WD repeat proteins stabilize the interaction between MYB and bHLH proteins and so promote the transcriptional activation of the structural genes of the anthocyanin biosynthetic pathway (Serna, 2004; Zimmermann et al., 2004; Broun, 2005; Ramsay and Glover, 2005; Lepiniec et al., 2006).

In A. majus, the molecular characterization of delila demonstrated that gene products similar to those in maize regulate anthocyanin production in dicotyledonous species (Goodrich et al., 1992). This has also been shown to be the case for flowers of other dicotyledonous species such as morning glory (Ipomoea purpurea, Ipomoea tricolor) (Park et al., 2004; Chang et al., 2005), in vegetative tissues of Perilla frutescens (Gong et al., 1999; Sompornpailin et al., 2002; Springob et al., 2003), and in Petunia, in which a MYB-related gene, AN2, is required for anthocyanin production in flowers (Quattrocchio, 1994; Quattrocchio et al., 1999) and a bHLH protein, AN1, interacts with AN2 to activate anthocyanin biosynthesis (Spelt et al., 2000). A second gene encodes another bHLH protein, JAF13, which also interacts with AN2, but JAF13 is not able to complement an1 mutants and so is unlikely to be involved directly in regulating the transcription of anthocyanin biosynthetic genes (Quattrocchio et al., 1998; Spelt et al., 2000). It has been suggested that AN2 activates the expression of AN1 (Spelt et al., 2000). The AN2 MYB protein is thought to interact with AN1 to activate anthocyanin biosynthetic gene expression in conjunction with AN11, a WD repeat protein structurally similar to pac1 and TTG1 (de Vetten et al., 1997). Mutants of AN2 still synthesize anthocyanins in the corolla tubes and the limb retains pale pigmentation, which has led to the suggestion that other genes encoding R2R3 MYB proteins structurally related to AN2 also contribute to the regulation of anthocyanin biosynthesis in Petunia flowers. One possible candidate is the AN4 gene (Spelt et al., 2000; Koes et al., 2005).

We have characterized the activity of three genes encoding MYB-related transcription factors active in the flowers of A. majus. Analysis of mutant phenotypes showed that these genes contribute to the intensity and pattern of anthocyanin production in flowers. Despite the close structural similarity of these regulatory proteins, analysis of mutants showed that they influence the expression of target genes encoding the enzymes of anthocyanin biosynthesis with different specificities. Consequently, they are not precisely equivalent, biochemically, in their activities. Different species of the genus Antirrhinum, native to Spain and Portugal, show striking differences in their patterns and intensities of floral pigmentation. The differences in anthocyanin production between the six species investigated are attributable to variations in the activity of the Rosea and Venosa loci encoding these MYB-related regulatory proteins. Set in the context of our understanding of the regulation of anthocyanin production in other genera, the activity of MYB-related genes is probably a primary cause of natural variation in this trait in plants.

	RESULTS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

 
Phenotypes of the rosea and venosa Mutants Affecting Floral Pigmentation Intensity and Patterning
There are two extant mutant alleles of the rosea locus of A. majus called rosea^colorata (ros^col) (Figure 1B) and rosea^dorsea (ros^dor) (Figure 1C) (Baur, 1910a; Stubbe, 1966). Wild-type (Ros⁺) flowers are almost completely self-colored, with high levels of red/magenta anthocyanin produced in both lobes and tubes of the corolla. Only the central region of the fused ventral and lateral petals is colored yellow, as a result of the accumulation of aurone in the absence of anthocyanin (Figure 1A). The vegetative parts of the plant also accumulate anthocyanin, including leaves (especially in the cells of the ventral [abaxial] epidermis) and stems.

Both rosea mutants have paler floral pigmentation. ros^col has weak anthocyanin production restricted to the inner epidermis of the petals and a ring of pigment at the base of the corolla tube (Figure 1B). There is no anthocyanin pigment produced in the stems of the plant, although the abaxial surfaces of the leaves accumulate anthocyanin in field-grown plants. ros^dor has flowers that have a weak ring of anthocyanin production toward the base of the tube and anthocyanin accumulation on the outer epidermis of the dorsal lobes (Figure 1C). The floral pigmentation in ros^dor is very dependent on environmental conditions. At the higher temperatures (25°C) and lower light levels typical of greenhouses in the United Kingdom, no pigment is produced in the lobes (Figure 1I). Under the higher light but lower temperatures outside, field-grown ros^dor plants have prominent pigmentation on their dorsal lobes (Figure 1C). The vegetative parts of ros^dor plants are darkly pigmented, including stems and leaves. The mutations are fully recessive to the wild-type Ros⁺ allele. Furthermore, crosses between ros^col and ros^dor do not complement (Figures 1D to 1F). F1 plants have flowers with an intermediate phenotype between the two alleles, indicating that they carry mutations at the same locus.

The phenotype conferred by Venosa (Ve⁺) in A. majus involves the production of magenta anthocyanin pigment in tissue over the veins of the corolla (Figure 1G) (Baur, 1910a; Kuckuck and Schick, 1930; Stubbe, 1966). Pigment production is limited to the inner epidermis of the petal lobes and to the inner epidermis of the corolla tube (Figure 1H). The phenotype of Ve⁺ is not apparent in wild-type A. majus because the strong self-color masks the venal patterning. However, the phenotype is clear in the ros^dor and ros^col backgrounds (i.e., Ros⁺ is epistatic to Ve⁺). Unfortunately, the dominant Venosa⁺ (Ve⁺) accession, described by Stubbe (1966), has been lost from the Gatersleben germplasm collection. However, after outcrossing of another Gatersleben accession, decipiens, to stocks from the collection at the John Innes Centre, including ros^col and ros^dor, segregation of the Ve⁺ phenotype was apparent. Wild type accessions (John Innes Centre stocks 7 and 522) carry the recessive ve^– allele, as do the ros^dor and ros^col stock accessions. The Ve⁺ allele in the decipiens accession is very closely linked to the recessive decipiens allele, and we have recovered no recombinants in examination of >300 F2 progeny. The Ve⁺ phenotype segregates independently of Rosea; therefore, these two loci are unlinked.

Isolation of MYB-Related Genes Controlling Anthocyanin Biosynthesis from A. majus
To identify genes encoding MYB-related proteins controlling anthocyanin pigmentation in A. majus, oligonucleotide primers to the most conserved regions of MYB genes C1 (from maize) and AN2 (from Petunia), which are the regions encoding the recognition helices of the DNA binding domain (Figure 2A ), were used for 3' rapid amplification of cDNA ends (RACE) PCR amplification of cDNA from flowers of the wild type and rosea, Eluta, and mut mutants. cDNA fragments were amplified from wild-type flowers with an oligonucleotide encoding part of the first recognition helix and with one encoding part of the second recognition helix. One of the fragments amplified by the oligonucleotide from the first recognition helix in R2 (G1709) was not present when cDNA from flowers of ros^dor or ros^col were used (Figure 2B). This fragment was amplified from cDNA from flowers of the other mutants.

View larger version (16K): [in this window] [in a new window]   Figure 2. Structure of the Rosea Locus in Wild-Type A. majus and in roscol and rosdor Mutants. (A) Scheme of the gene structure of members of the R2R3 MYB subgroup 6 family (Kranz et al., 1998). The intron positions are conserved between monocots and dicots. Below the line, a scheme of the encoded R2R3 MYB proteins is shown aligned to the exons (E1 to E3). R2 and R3 indicate the MYB repeats most similar to the second and third repeats of the prototypic MYB protein, c-Myb, respectively. C indicates the C-terminal domain. The position of the sequence in oligonucleotide G1709 is indicated by an arrow. (B) Ethidium bromide–stained agarose gel showing PCR products of 3' RACE PCR amplification using oligonucleotide G1709 and the adaptor oligonucleotide (Frohmann et al., 1988) on first-strand cDNA from flowers of wild-type, roscol, and rosdor individuals. The arrow indicates the band present in the wild type and missing from rosea cDNAs. (C) Maps of independent EMBL4 clones identified as containing the Ros1 and/or Ros2 genes. (D) Map of the Rosea locus from wild-type A. majus. The raised blocks indicate exon sequence. The initiating ATG codons for Ros1 and Ros2 are shown. Restriction enzyme sites are as follows: E, EcoRI; X, XbaI (only relevant sites are shown). The 2.4-kb XbaI fragment mentioned in the text is also indicated. (E) Map of the Rosea locus from roscol. The raised blocks indicate exon sequence. The initiating ATG codons for Ros1 and Ros2 are shown. Changes from the wild-type sequence are indicated by red lines. (F) Map of the Rosea locus from rosdor. The raised blocks indicate exon sequence. The initiating ATG codon for Ros1 is shown. Changes from the wild-type sequence are indicated by red lines.

Genomic clones encoding the Ros1 gene were isolated by screening a genomic library from A. majus in EMBL4 (a gift from Hans Sommer) with the Ros1 cDNA. Five positive clones were identified and mapped (Figure 2C). Two of the independent genomic clones contained the Ros1 coding sequence within three adjacent EcoRI fragments. They also contained a second region of homology with the Ros1 cDNA, situated 4.3 kb downstream of Ros1 within a 2.4-kb XbaI fragment (Figure 2D). The three other clones contained only the second region of homology defined by the 2.4-kb XbaI fragment (Figures 2C and 2D). The sequences of the three EcoRI fragments containing the genomic sequence for Ros1 were determined. The ORF was interrupted by two introns situated at equivalent positions to the introns in the C1 gene of maize (Paz-Ares et al., 1987) and AN2 from Petunia (Quattrocchio et al., 1999), the first of 231 bp and the second of 1498 bp (Figure 2D). There was a sequence, TATTT, located 115 bp upstream of the initiating ATG in Ros1 that approximates a plant TATA binding site (Molina and Grotewold, 2005). Farther upstream at –489 and –587 were two additional motifs (TATAA) that approximated better to the TATA binding consensus.

The second region of homology with the Ros1 cDNA (contained within the 2.4-kb XbaI fragment) in all five EMBL4 clones was also sequenced (Figure 2C). This contained a sequence encoding the N-terminal region of a second MYB-related protein, distinct but structurally very similar to the N terminus of Ros1. This gene fragment was termed Rosea2 (Ros2). All five clones ended within the second intron of Ros2, and we were unable to identify any overlapping clones that contained the downstream ORF of Ros2. To determine whether this second gene was expressed and functional, we examined the expression of Ros2 in the wild type and ros mutant lines.

The Ros2 cDNA was amplified by 3' RACE PCR (Frohmann et al., 1988) using an oligonucleotide primer to the sequence encoding the initiating ATG codon of the Ros2 protein in the genomic clone and cDNA made from flowers of ros^col. A cDNA fragment was amplified that contained the entire Ros2 ORF, 3' to the priming oligonucleotide (Figure 3B ). Comparison with the partial genomic sequence of Ros2 confirmed that it contained at least two introns, the first two being in identical positions to those in Ros1, An2 from Petunia, and C1 and Pl from maize. The first intron in Ros2 was 111 bp, whereas the size of the second intron remains unknown but it is >9 kb, as deduced from the genomic sequences in the EMBL4 clones available. A potential TATA box was identified 109 bp upstream of the first ATG in the Ros2 ORF (Figure 2D).

View larger version (44K): [in this window] [in a new window]   Figure 3. Expression of Rosea Genes in Flowers of Wild-Type A. majus and rosea Mutants. (A) RNA gel blot of poly(A)+ RNA from different organs of wild-type A. majus (R, root; ML, mature leaf; YL, young leaf; P, petal) and from petals of roscol, rosdor, and roscol Ve+ lines probed with the Ros1 cDNA. Below, the expression of Ros1 is shown at different stages of petal development. Maximum expression was observed when the flowers had just opened. (B) Ethidium bromide–stained agarose gel of 3' RACE PCR products produced using oligonucleotide K17 and the adaptor oligonucleotide (Frohmann et al., 1988) on first-strand cDNA from flowers of wild-type, roscol, and rosdor individuals. Below is a blot of the same gel probed with a fragment of the Ros2 gene. The arrow indicates cDNA that was amplified from roscol only. M, size markers of DNA digested with HindIII; C, control amplification without cDNA.

Genotypic Constitution of Wild-Type, ros^col, and ros^dor Lines at the Rosea Locus
Because Ros1 cDNA was not amplified using the G1709 oligonucleotide primer from cDNA derived from RNA from flowers of either ros^dor or ros^col, the possibility that the phenotypes conferred by rosea resulted from mutations in the Ros1 or Ros 2 gene was investigated at the molecular level. First, the linkage of the two genes to the ros mutations was analyzed in crosses between the wild type and ros^col and the wild type and ros^dor. The Ros1 gene was highly polymorphic between wild-type, ros^col, and ros^dor lines. Using EcoRI digests, 126 progeny were examined from F2 crosses between the wild type and ros^col and 96 progeny were examined from crosses between the wild type and ros^dor. The ros^col and ros^dor plants that segregated (scored phenotypically) in F2 were all homozygous for the parental versions of the Ros locus identified in the stock ros^col and ros^dor lines.

The expression of Ros1 in the wild type, ros^col, and ros^dor was examined by RNA gel blot analysis. A high level of transcript was detected in maturing petals of wild-type plants (Figure 3A). A considerably reduced level of transcript of a slightly smaller size accumulated in petals of the same age from ros^col (Figure 3A). This transcript was amplified by RT-PCR and sequenced. It contained several single-nucleotide substitutions compared with wild-type Ros1, but the most significant difference was a 64-bp deletion that included the region encoding the first recognition helix (in R2) of the DNA binding domain (Figure 2E). This mutation introduced a stop codon 36 bp downstream of the deletion, indicating that a nonfunctional Ros1 product was produced, because the encoded protein would lack more than half of its DNA binding domain and the entirety of its C terminus. The 64-bp deletion included sequence encoding the recognition helix of R2, which explains why the original oligonucleotide, G1709, failed to amplify a cDNA product from ros^col RNA by 3' RACE PCR.

The ros^dor line produced no Ros1 transcript detectable on RNA gel blots (Figure 3A). However, RT-PCR did amplify a Ros1 cDNA that was cloned and sequenced. This showed 11 differences from the sequence of Ros1 from wild-type lines, of which 8 affected the identity of the encoded amino acids (codon Glu-27 to Lys, Gly-116 to Arg, Leu-131 to Gln, Lys-174 to Thr, Thr-176 to Lys, Ala-183 to Val, Glu-190 to Asp, and Asp-191 to Glu). However, comparison with the sequences of Ros2 and Ve suggested that it was unlikely that any of these changes would contribute to the phenotype, implying that the primary change in ros^dor is a change in the expression pattern of Ros1. To check this, we cloned the genomic DNA encoding the promoter and N terminus (up to the second intron) of Ros1 from the ros^dor background. Within the 910-bp region of the Ros1 promoter (compared between the wild type and ros^dor), there were nine single-nucleotide substitutions, two single-nucleotide additions, one single-nucleotide loss, and an insertion of 12 bp at 440 bp upstream of the initiating ATG codon. The biggest difference, however, was a deletion of 187 bp in the promoter (from –140 to –326), which is likely to have a significant impact on the transcription of Ros1.

The expression of the full-length Ros2 transcript was analyzed using RT-PCR. Using a 5' oligonucleotide covering the initiating ATG of Ros2 and a reverse primer representing sequences in the 3' untranslated region, a full-length transcript was detected in ros^col flowers but not in wild-type or ros^dor flowers (Figure 3B). These data showed that ros^col expresses a full-length transcript of Ros2 at a very low level, but there is no expression of full-length Ros2 in the wild type or ros^dor. Therefore, the wild type and ros^dor were deemed to be ros2^–.

These data suggested that both ros^col and ros^dor are determined by mutations in the complex Rosea locus. ros^col carries a loss-of-function allele of Ros1 but expresses Ros2. The Ros2 gene contributes the weak pigmentation restricted to the inner epidermis of the corolla lobes and the base of the tube seen in ros^col, but it is not expressed in wild-type lines. ros^dor has no Ros2 expression and so no pigment on the inner epidermis of the lobes. Ros1 expression is modified in ros^dor, most likely as a result of changes to the promoter of Ros1, such that it induces pigmentation only in the outer epidermis of the dorsal petals and in a ring at the base of the tube under specific environmental conditions of high light and relatively low temperature. Its contribution to vegetative pigmentation compared with the wild type is relatively normal.

To confirm that Ros1 and Ros2 both control anthocyanin production, and that Ros1 and Ros2 could complement rosea mutants, we used particle bombardment of ros^dor lines grown in the greenhouse, so that the corolla lobes were acyanic. Each cDNA was cloned between the cauliflower mosaic virus (CaMV) 35S promoter and the octopine synthase terminator and was bombarded into the inner epidermis of the dorsal petals of ros^dor flowers, as well as a control plasmid containing ß-glucuronidase (GUS) driven by the 35S promoter. In tissue bombarded with the control plasmid alone, patches of gold were visible after 2 d (Figure 4A ). The gold particles were surrounded by tissue that had browned slightly as a result of physical damage from bombardment. No spots of magenta anthocyanin pigment were ever observed with the control plasmid (Figure 4A). However, histochemical GUS staining revealed that the control plasmid was active in the bombarded tissue (Figure 4B). Bombardment with the Ros1 cDNA resulted in the development of single magenta cells within 48 h of bombardment. On average, for Ros1 >100 spots were observed per replicate piece of petal (Figures 4C and 4D). Bombardment with Ros2 also gave rise to magenta spots, but at a lower frequency than in Ros1 (Figure 4D).

View larger version (103K): [in this window] [in a new window]   Figure 4. Complementation of the Mutant Phenotype Conferred by rosdor by Particle Bombardment of Petal Lobe Tissue with the Rosea Genes. (A) Petal tissue bombarded with a construct carrying the GUS gene under the control of the CaMV 35S promoter. Only patches of gold are visible. (B) Tissue bombarded with a construct carrying the GUS gene under the control of the CaMV 35S promoter, stained for GUS activity. (C) Petal tissue bombarded with a construct with the Ros1 gene expressed under the control of the 35S promoter. (D) Magnification of tissue shown in (C), demonstrating magenta anthocyanin being produced in single petal epidermal cells. (E) Petal tissue bombarded with a construct with the Ros2 gene expressed under the control of the 35S promoter.

A Third Unlinked Gene Encoding a MYB-Related Protein Controls the Phenotype Conferred by Venosa
Because the Ve⁺ allele confers strong magenta pigmentation to the epidermal cells overlying the petal veins in ros mutants, but other structural and regulatory mutations are epistatic to Ve⁺, we reasoned that Venosa most likely encodes a MYB-related transcription factor that was functionally similar to Ros1 and Ros2. To identify such a gene, we digested genomic DNA from Ve⁺/ve^– heterozygotes and ve^– homozygotes with a range of restriction enzymes, separated the DNA by gel electrophoresis, and blotted it onto nitrocellulose filters. The DNA gel blots were probed with a fragment of cDNA encoding the Ros1 MYB DNA binding domain and were subsequently washed at low stringency. The EcoRI digest revealed a number of hybridizing bands, one of which (2.9 kb) segregated clearly with the phenotype conferred by Ve⁺ (Figure 5A ). This fragment of genomic DNA was cloned from DNA from a Ve⁺/ve^– heterozygote in NM1149 and subsequently was subcloned in pBluescript SK+ and sequenced. The genomic DNA fragment contained the sequence encoding the N-terminal region of a MYB-related transcription factor. This sequence was used to design a gene-specific primer to the sequence, including the initiating ATG. This oligonucleotide was used for 3' RACE PCR on cDNA prepared from Ve⁺ and ve^– corolla tissue. A transcript was amplified only from Ve⁺ cDNA. This was cloned and sequenced. The Ve cDNA encoded an R2R3 MYB protein very similar to Ros1 and Ros2 (see Supplemental Figure 1 online). By comparison with the genomic DNA of the active allele, the gene was deduced to have at least two introns, the first of 480 bp and the second of 1450 bp. The 5' upstream region contained a motif, TATAT, 75 bp upstream of the initiating ATG, and another, TTATTT, 101 bp upstream. Either of these may constitute the TATA box. The cDNA clone was used to map the Ve gene relative to the phenotype conferred by Ve⁺ in 179 segregating F2 individuals. No recombinants were found between the phenotype conferred by Ve⁺ and the 2.9-kb EcoRI fragment (Figure 5B).

View larger version (46K): [in this window] [in a new window]   Figure 5. Identification of the Venosa Gene and Phylogenic Analysis of the MYB Proteins Encoded by Ros1, Ros2, and Ve. (A) DNA gel blot of genomic DNA from individuals with different Ve phenotypes digested with EcoRI and probed with the region of the Ros1 cDNA encoding the DNA binding domain. Washing was at low stringency. The arrow indicates a fragment of DNA present only in lines carrying the Ve+ allele. Lane 1, rosdor; lane 2, rosdor Ve+; lanes 3, 4, 6, and 7, roscol; lanes 5 and 8, roscol Ve+. (B) Mapping of the 2.9-kb EcoRI fragment identified in (A). Genomic DNA from individuals with the phenotype indicated above the line was digested with EcoRI and probed with the 2.9-kb EcoRI fragment from the Venosa locus. Arrow a indicates a ve– allele, and arrow b indicates a Ve+ allele. (C) RNA gel blot of poly(A)+ RNA from petal tissue of roscol Ve+ and roscol ve– lines probed with the Ve cDNA. The arrow indicates the Ve transcript. L, lobes; T, tubes of flowers. (D) Top, map of the Ve locus from wild-type A. majus. The raised blocks indicate exon sequence. The initiating ATG codon is shown. Bottom, map of the Ve locus from ve–. Differences from the wild-type sequence are indicated by red lines. (E) Left, petal tissue bombarded with an empty vector construct; right, petal tissue bombarded with a construct with the Venosa gene expressed under the control of the 35S promoter. (F) Phylogenetic tree comparing the amino acid sequences of the DNA binding domains of the R2R3 MYB subgroup 6 members. The regions used for the alignments are indicated in Supplemental Figure 1 online. The numbers indicate bootstrap values.

To confirm that Ve could regulate anthocyanin biosynthetic gene expression, the cDNA was cloned between the double CaMV 35S promoter and the octopine synthase terminator and was bombarded into ros^dor petals (greenhouse-grown). A large number (>100 per replicate) of magenta single-cell spots were observed within 48 h, whereas the control plasmid gave no magenta spots (Figure 5E).

Ve, Ros1, and Ros2 Appear to Have Arisen by Gene Duplication in Antirrhinum
The deduced amino acid sequences of the Ros1, Ros2, and Ve peptides were compared with those of other MYB proteins identified as regulators of flavonoid biosynthesis: AN2 from Petunia (Quattrocchio et al., 1999), Vv MYBA1 and Vv MYBA2 from grape (Vitis vinifera) (Kobayashi et al., 2002), TT2 (At MYB123), PAP1 (At MYB75), PAP2 (At MYB90), At MYB113, and At MYB114 from Arabidopsis, and C1 and Pl from maize. GLABROUS1 (GL1), WERWOLF (WER), and At MYB23 from Arabidopsis (Stracke et al., 2001) were included as outgroups (Figure 5F; see Supplemental Figure 1 online). All of the anthocyanin-regulatory MYB proteins were very similar over their DNA binding domains, with 59 of 100 amino acids in the R2R3 MYB DNA binding domain conserved in all proteins. In R2R3 MYBs, the DNA binding domain forms two repeats of helix-helix-turn-helix structure. The third helix in each repeat is believed to interact with DNA to bind it in a sequence-specific manner. The greatest conservation of sequence between these proteins was in the third region of helix in R2 and in the second region of helix in R3 (see Supplemental Figure 1A online). Surprisingly, there were quite a number of amino acid substitutions within the third helix (the recognition helix) of R3 between Ros1, Ros2, and Ve, suggesting that the different proteins might have somewhat different binding site preferences (see Supplemental Figure 1 online).

Several amino acids within helix 1 and helix 2 of R3 have been identified by Grotewold et al. (2000) as important for interaction with bHLH proteins (Leu-76, Arg-79, Arg-82, Leu-83, Gly-94, and Arg-95; C1 numbering). All are conserved in Ros1, Ros2, and Ve. The extended interaction signature identified by Zimmermann et al. (2004) as [DE]Lx₂[RK]x₃Lx₆Lx₃R is also conserved and identical in the R3 regions of Ros1, Ros2, and Ve (DLivRlhkLlgnkwsLiagR; where uppercase letters indicate amino acids that are completely conserved, lowercase letters indicate alternative amino acids, and x indicates any amino acid). This implies that all three Antirrhinum proteins interact with bHLH proteins in their activation of anthocyanin biosynthesis, as has been shown for C1, Pl, and AN2.

The C-terminal domains of the Antirrhinum MYB regulators were much less well conserved, but one patch of conserved sequence was obvious: a more acidic region toward the C terminus of each protein that surrounds a core sequence of higher conservation (in dicot MYB proteins, it was Ww/lx+/–LL, where the annotation is as before and +/– indicates a charged amino acid). This motif is part of a longer motif identified by Kranz et al. (1998) from Arabidopsis MYB proteins defining R2R3 MYB subgroup 6. This short motif shows similarity to sequence in the KIX domain of animal c-MYB, which lies within the activation domain of that protein and interacts with the transcriptional coactivator, CREB. In C1 and Pl, there are also acidic regions at the C termini of the proteins structured around the sequence WLRC+T, which might represent the equivalent domain in the monocot proteins. One residue, Asn, which lies immediately N terminal to the WLRC+T motif, has been shown to contribute significantly to the ability of the C-terminal domain of C1 to activate transcription (Sainz et al., 1997a). The C-terminal domains of C1 and AN2 have also been shown to activate transcription in yeast one-hybrid assays (Goff et al., 1992; Sainz et al., 1997a; Quattrocchio et al., 1999).

Phylogenetic analysis of the anthocyanin MYB regulators Ros1, Ros2, and Ve was performed. A well-supported phylogeny was recovered that placed the Antirrhinum proteins as more closely related to one another than to their orthologs from other species (Figure 5F; see Supplemental Figure 1 online). This suggested that the three Antirrhinum genes were derived from relatively recent duplications of a common ancestral gene. The first duplication created the unlinked genes, the progenitor of Venosa and the progenitor of Rosea. The second duplication gave rise to Rosea1 and Rosea2 and occurred intrachromosomally. Similar duplication events appear to have also occurred recently in Arabidopsis to create MYB75, MYB90, MYB113, and MYB114, which are clustered on chromosome 1 (Stracke et al., 2001), in grape to create MYBA1 and MYBA2, which are closely linked (Kobayashi et al., 2002, 2004, 2005), in tomato (Lycopersicon esculentum) to create ANT1 and a very closely linked MYB-like gene (Mathews et al., 2003; De Jong et al., 2004), and in potato (Solanum tuberosum) to create the linked F locus controlling floral pigmentation and the D(I) locus controlling tuber color (De Jong et al., 2004); a duplication of the whole genome may have created C1 and Pl in maize (Cone et al., 1993).

Comparison of the Effects of Ros1, Ros2, and Ve on Target Gene Expression Reveals the Proteins to Have Similar but Distinct Biochemical Specificities
Although Ros1, Ros2, and Ve can all activate anthocyanin biosynthesis, we were interested to know whether they are truly functionally identical proteins or whether they have evolved distinct specificities in their regulatory properties. To this end, we examined the target genes of Ros1, Ros2, and Ve by comparing the levels of transcripts encoding the enzymes of anthocyanin biosynthesis in plants, both wild type and mutant, for the expression of each regulatory gene. To analyze the effects of Ros1, we compared the expression of the biosynthetic genes in flowers of the wild type and ros^dor (Figure 6 ). No change in transcript levels was observed for CHS. A small decrease in the level of chalcone isomerase (CHI) transcript was observed in ros^dor petals compared with wild-type petals. Much less transcript was observed in ros^dor than in the wild type for F3H, although greater differences were observed for F3'H, DFR, and UFGT. Smaller but still significant reductions in transcript levels in ros^dor compared with the wild type were observed for flavonol synthase (FLS), ANS, and AT, an anthocyanin permease from A. majus (GenBank accession number AJ796511) homologous with a tomato MATE transporter induced by a MYB transcription factor regulating anthocyanin production in tomato (Mathews et al., 2003; Bey et al., 2004). These data showed that Ros1 increases the transcript levels of the late biosynthetic genes (F3H, DFR, ANS, and UFGT) as well as F3'H, FLS, and AT, although not all of these target genes are equally dependent on Ros1 activity. F3'H, F3H, DFR, and UFGT were particularly dependent on Ros1 activity, because virtually no transcripts for these genes were detected in ros^dor flowers. By contrast, ANS and AT were much less dependent on Ros1. Ros1 was not required for the accumulation of transcripts of the early biosynthetic gene CHS and had only a very minor influence on CHI. This is similar to the effect of Delila on the regulation of the anthocyanin biosynthetic pathway in Antirrhinum (Martin et al., 1991), supporting the view that Ros1 may interact with Delila to activate gene transcription.

View larger version (50K): [in this window] [in a new window]   Figure 6. RNA Gel Blots of Poly(A)+ RNA from Petal Tissue of Wild-Type, roscol, rosdor, and rosdor Ve+ Plants Probed with cDNA Fragments of the Genes Encoding the Enzymes of the Anthocyanin Biosynthesis Pathway. The composition of active MYB proteins in each genotype is shown below the blots. R1, Rosea1; R2, Rosea2; Ve, Venosa. Genes assayed were as follows: CHS, gene encoding chalcone synthase; CHI, chalcone isomerase; F3H, flavanone 3-hydroxylase; F3'H, flavonoid 3'-hydroxylase; FLS, flavonol synthase; DFR, dihydroflavonol 4-reductase; ANS, anthocyanidin synthase; UFGT, UDP-glucose 3-O-flavonoid transferase; AT, anthocyanin transporter.

The influence of Ve on anthocyanin biosynthetic gene expression was assayed by comparing transcript levels in Ve⁺ and ve^– lines in the ros^dor background (Figure 6). No difference was detected in the levels of CHS transcripts. The presence of an active Ve⁺ allele did increase the transcript levels of CHI, F3H, F3'H, FLS, ANS, UFGT, and AT. Surprisingly, no difference could be detected in the levels of DFR transcript between Ve⁺ and ve^– lines; virtually no transcript was detected in either line. These data showed that Ve does increase the transcript levels of most of the biosynthetic genes and is not required for the expression of the early gene, CHS. However, they also indicate that Ve does not activate DFR expression nearly as effectively as Ros1. These data imply that there are subtle but reproducible differences in the abilities of Ve, Ros1, and Ros2 to activate the transcription of different target genes. This could reflect differences in target site specificity affecting the affinity with which each activator binds to the target promoters of the different biosynthetic genes. Such specificity might also involve differences in the efficacy of the complexes that each MYB protein makes with the bHLH and WD repeat proteins.

Ros1 and Ve Show Differences from Ros2 in the Specificity of Their Interaction with the bHLH Regulator Mut
We examined the interactions of the different MYB transcription factors with bHLH proteins through genetic analyses. The Delila gene product is a bHLH protein that is required for anthocyanin biosynthesis in the corolla tubes. It is also expressed in the lobes, as demonstrated by in situ hybridization, and it is active in the lobes (Almeida et al., 1989; Goodrich et al., 1992). A second gene, Mut, is required for anthocyanin biosynthesis in the lobes if Delila is nonfunctional. This encodes another bHLH protein (P. Piazza, C. Tonelli, and C. Martin, unpublished data). The specificity of Ros1, Ros2, and Ve interactions with these two bHLH proteins was tested by making double and triple mutants.

The line mutant for mut and del is acyanic. Ve⁺ is not visible in this background, suggesting that Ros1, Ros2, and Ve must interact with either Del or Mut to activate anthocyanin biosynthesis. When ros^dor is combined with del, the tubes are unpigmented, but there is little effect on pigmentation in the lobes, indicating that Del interacts with Ros1 in the tubes but that in the lobes Ros1 interacts with Mut to direct anthocyanin biosynthesis. This confirms the phenotype of del alone, in which the tubes are acyanic but the lobes (principally the product of Ros1 and Mut interaction) have wild-type levels of pigmentation.

When ros^col (in which only Ros2 of the MYB regulators is expressed) was combined with del, the double mutant was completely acyanic, despite Ros2 being active in the lobes (cf. Figures 7A and 7B ). This indicates that in the lobes, Ros2 can interact with Del but not with Mut. This could be because Ros1 activates the expression of Mut, because in Petunia the AN2 MYB protein has been reported to activate the expression of AN1 (encoding a bHLH protein) in leaves (although not in flowers) (Spelt et al., 2000). However, although when ros^col Ve was combined with del the background pigmentation in the corolla lobes of these plants was missing (acyanic), the venal pattern, determined by Ve, remained distinct (cf. Figures 7C and 7D). These results indicated that although Ros2 interacts effectively only with Del in the lobes, Ve (like Ros1) can interact effectively with Mut. These interactions are summarized in Figure 7E. Because our analysis of target gene activation showed that the three MYB proteins, Ros1, Ros2, and Ve, had differential specificities for each of the target genes, the ineffectual interaction between Ros2 and Mut may reflect the loss of activity of these proteins on a single target promoter, or their lack of activity on all target promoters. These possibilities cannot be resolved through genetic analysis, which measures only the accumulation of the end product, anthocyanin, and not the activation of the individual biosynthetic genes.

View larger version (100K): [in this window] [in a new window]   Figure 7. Interaction of MYB Proteins with bHLH Proteins Revealed by Mutant Analysis. (A) roscol. (B) roscol del. (C) roscol Ve+. (D) roscol Ve+ del. (E) Diagram summarizing the interactions of the regulatory proteins in the different regions of the flower of A. majus. Red diagonal stripes sloping right indicate the area of Mut expression; red diagonal stripes sloping left indicate the area of Del expression. Ros1 and Ve can interact with both Mut and Del in the lobes, whereas Ros2 can interact only with Del.

View larger version (32K): [in this window] [in a new window]   Figure 8. Floral Phenotypes of Accessions of Different Species within the Genus Antirrhinum. (A) A. majus subsp majus var majus (Toulouse) without venal pigmentation. (B) A. majus subsp majus var majus (Toulouse) with venal pigmentation. (C) A. majus subsp majus var majus (Barcelona) without venal pigmentation. (D) A. majus subsp majus var majus (Barcelona) with venal pigmentation. (E) A. australe. (F) A. barrelieri. (G) A. latifolium (Pyrea). (H) A. latifolium (Marseilles). (I) A. graniticum. (J) A. meonanthemum. (K) A. molle. (L) A. mollissimum.

Crosses of A. meonanthemum and A. latifolium to the mut del double mutant or to wild-type A. majus gave darker pigmentation to the lobes of flowers of the F1 progeny than that in A. meonanthemum and A. latifolium, but pigmentation was paler than in wild-type A. majus, especially over the outer edges of the corolla lobes (see Supplemental Figures 2B and 2C online). This phenotype matched exactly that of the Eluta mutant of A. majus (Baur, 1910a; Stubbe, 1966) (Figures 1J to 1L), which is a semidominant diluter of anthocyanin pigment in flowers. In the F2 of these species crossed to wild-type A. majus, pale pigmentation, the diluted pigmentation phenotype, and full-red pigmentation segregated 1:2:1, respectively, indicating that the pale pigmentation in these species was attributable to a semidominant allele at a single locus. The fact that in F2 populations of A. meonanthemum and A. latifolium (Marseilles) crossed to ros^dor, no full-red plants segregated (30 and 96 progeny examined, respectively) indicated that both A. meonanthemum and A. latifolium carry semidominant alleles of the Rosea locus that dictate their pale floral pigmentation. These alleles are likely to be very similar in their activity to the Eluta mutant of A. majus. Indeed, Eluta has been reported to be very closely linked to rosea by Stubbe (1966). In F2 populations of Eluta crossed to ros^dor or ros^col, we have never observed full-red recombinants, supporting the idea that Eluta is a semidominant allele of the Rosea locus. We have indicated the semidominance of the pigment-diluting Rosea alleles in these species with the term Ros^El.

Crosses of the species accessions to ros^dor suggested that A. majus subsp majus from both Barcelona and Toulouse was polymorphic for an active Ve⁺ allele (Figures 8A to 8D). Crosses between these accessions and ros^dor Ve revealed that the venal patterning of pigmentation could be attributed to the activity of the Ve locus, because no unpatterned progeny segregated in the F2 population (see Supplemental Table 1 online). Similar analyses with the species showed that all A. molle, A. mollissimum, A. meonanthemum, and A. latifolium individuals were homozygous for active Ve⁺ alleles. The deduced genotypes of the species analyzed for the loci affecting floral pigmentation are listed in Table 1.

These data revealed that the loci controlling anthocyanin pigmentation patterning in different Antirrhinum species principally encoded MYB-related transcription factors. DNA gel blots revealed that all three genes, Ros1, Ros2, and Ve, are present in the genomes of each of the accessions tested. The differences in background pigmentation patterns between the different species (most readily observed in the F2 segregants from crosses to wild-type A. majus, in which genetic background is homogenized, allowing for easier comparisons) suggest that differences in the activity of the Rosea locus is the single most important determinant of pigmentation intensity in the species tested. In addition, the other MYB-related gene, Venosa, contributes significantly to pigmentation in A. latifolium, A. molle, A. meonanthemum, and A. mollissimum and is also significant and variable in wild accessions of A. majus (see Supplemental Table 1 online).

	DISCUSSION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

 
In Antirrhinum majus, a small family of MYB-related proteins controls the pattern and intensity of floral pigmentation. Members of this family are closely related structurally to MYB proteins known to regulate anthocyanin production in other plant species, including maize (Paz-Ares et al., 1987; Cone et al., 1993) Petunia (Quattrocchio et al., 1999), grape (Kobayashi et al., 2002, 2004, 2005), pepper (Capsicum annuum) (Borovsky et al., 2004), morning glory (Chang et al., 2005), tomato (Mathews et al., 2003; De Jong et al., 2004), potato (De Jong et al., 2004), and Arabidopsis (Borevitz et al., 2000). Like other species, these different genes appear to have been derived by gene duplication and subfunctionalization, although our data from the investigation of mutant alleles suggests that the Ros1, Ros2, and Ve proteins are not functionally equivalent. These analyses were complicated by the different expression levels and patterns of the Ros1, Ros2, and Ve genes, but analysis of the transcript levels of potential target genes in lines defective in Ros1, Ros2, and Ve activity showed F3H, F3'H, FLS, DFR, and UFGT to be highly dependent on Ros1 for induction, whereas ANS and AT were less dependent and CHI was very much less dependent on Ros1 for induction. The only gene that was significantly induced by Ros2 was F3'H, although CHI transcript levels were also increased to a small extent by this protein. Ve induced the expression of CHI, F3H, F3'H, FLS, ANS, UFGT, and AT, but its induction of F3H, F3'H, and UFGT was stronger than that for ANS and AT and much stronger than that for CHI. Remarkably, Ve did not detectably activate DFR expression/transcript levels. These data suggest that each of these MYB proteins has distinct biochemical specificity in terms of its ability to activate transcription from different target promoters. This might reflect differences in their DNA binding affinities. There are a number of amino acid differences in the recognition helices of these three MYB proteins that could influence their DNA binding affinities and sequence motif recognition (see Supplemental Figure 1 online). Alternatively, differences in specificity might be attributable to differences in target promoter architecture, which could make binding or activation easier by one MYB protein relative to another, or to differences in the ability of the MYB regulators to interact with target promoters with different chromatin structures. However, although understanding of this specificity must await comparative biochemical characterization of the MYB proteins, we conclude that these regulatory proteins have diverged functionally as well as in their expression patterns, unlike the situation in maize, in which C1 and Pl appear to be functionally equivalent (Cone et al., 1993).

Given the observed specificity for target gene activation by the three MYB proteins, it was surprising that all could complement the ros^dor mutation and produce pigmented cells after particle bombardment of petals (Figures 4C, 4E, and 5D). In addition, Ve can complement the loss of Ros1 activity in the ros^col line to give regions of highly pigmented epidermal tissue (Figure 7C), and it can complement the loss of both Ros1 and Ros2 activity in ros^dor to give highly pigmented epidermal tissue in the regions overlying the veins on the inner epidermis of the petal lobes (Figure 1G). An explanation for this is that our complementation assays are based on the ability of cells to synthesize anthocyanins, an ability that is determined principally by the degree of activation of the biosynthetic step limiting the flux to anthocyanin accumulation in flowers. Although DFR was not detectably induced by Ve on RNA gel blots, 40 cycles of RT-PCR amplification of cDNA revealed low levels of DFR transcript in both ros^dor and ros^col flowers (see Supplement