In plants, the vacuole occupies a large part (up to 90%) of the cell volume and is important for a variety of physiological processes, such as pH homeostasis, osmoregulation, ion transport, and storage of metabolites. Moreover, it plays an important role in cell growth, because the enlargement of a cell is mostly attributable to an increase in the volume of the vacuole rather than of the cytoplasm (reviewed in Taiz, 1992; Maeshima, 2001; Gaxiola et al., 2002).

The lumen of vacuoles is acidic compared with the cytoplasm, and in some cells (e.g., in lemon [Citrus limon] fruit) it can reach pH values as low as 1. Among the most abundant proteins on the vacuolar membrane are vacuolar ATPase (v-ATPase) and pyrophosphatase proton pumps (Szponarski et al., 2004) that transport protons from the cytoplasm into the vacuole, thereby contributing to the acidification of the vacuolar lumen. The resulting electrochemical gradient across the vacuolar membrane is the driving force for the transport of a variety of compounds (ions, sugars) via secondary symport and antiport transporters and channels (reviewed in Taiz, 1992; Maeshima, 2001; Gaxiola et al., 2002).

In most species, the coloration of flowers and fruits results from the accumulation of flavonoid pigments (anthocyanins) in the vacuoles of (sub)epidermal cells. Because the absorption spectrum of anthocyanins depends on the pH of their environment (de Vlaming et al., 1983), the color of a tissue depends in part on the pH of the vacuolar lumen, thus making flower color a convenient and reliable reporter to monitor alterations in vacuolar pH (Yoshida et al., 1995, 2003)

In morning glory (Ipomoea tricolor) petals, the vacuolar pH is relatively low when the flower bud opens, resulting in a red color, but upon further maturation, the vacuolar pH increases and the petals acquire a strong blue color (Yoshida et al., 1995). This color change and the increase of vacuolar pH require a putative Na⁺/H⁺ exchanger encoded by the PURPLE gene (Fukada-Tanaka et al., 2000). Most likely, PURPLE transports sodium ions into and protons out of the vacuole, resulting in a less acidic vacuole and a bluer color.

Petunia hybrida flowers normally have a lower pH than Ipomoea flowers, and the color of wild-type flowers stays on the reddish (low pH) side of the color spectrum. By genetic analyses, seven loci (named PH1 to PH7) have been identified that, when mutated, cause a more bluish flower color and an increase in the pH of crude petal extracts (Wiering, 1974; de Vlaming et al., 1983; van Houwelingen et al., 1998), suggesting that these genes are required for acidification of the vacuole (de Vlaming et al., 1983). Mutations in the genes ANTHOCYANIN1 (AN1), AN2, and AN11 cause, besides the loss of anthocyanin pigments, an increased pH of petal extracts. That this pH shift is at least in part attributable to an increased pH of the vacuolar lumen was evident from the bluish flower color specified by particular an1 alleles (formerly known as ph6) that lost the capacity to activate the vacuolar acidification function but could still drive anthocyanin synthesis (Spelt et al., 2002).

AN1 and AN11 are required for transcriptional activation of a subset of structural anthocyanin genes, encoding the enzymes of the pathway, in all pigmented tissues (Quattrocchio et al., 1993) and encode a basic-helix-loop-helix (BHLH) transcription factor and a WD40 protein, respectively (de Vetten et al., 1997; Spelt et al., 2000). AN2 encodes a MYB-type transcription factor whose function appears to be (partially) redundant, because it is expressed only in petals and not in other pigmented tissues (Quattrocchio et al., 1999). Moreover, even in the an2 null mutant, pigmentation of the petals is reduced, but not fully blocked, and the pH shift in an2 petal homogenates is smaller than that in an1 or an11 petals (Quattrocchio et al., 1993; Spelt et al., 2002). In addition, AN1 and AN11 play a role in the development of epidermal cells in the seed coat (Spelt et al., 2002).

The anthocyanin pathway has been shown to be activated by similar MYB, BHLH, and WD40 proteins in a wide variety of species, indicating that this function is well conserved (reviewed in Winkel-Shirley, 2001; Koes et al., 2005). Several studies revealed that these MYB, BHLH, and WD40 proteins could interact physically, indicating that they may operate in one transcription activation pathway and may activate their target genes as a (ternary) complex (Goff et al., 1992; Zhang et al., 2003; Baudry et al., 2004; Kroon, 2004; Zimmermann et al., 2004). Besides petunia, Arabidopsis thaliana is the only other species in which these activators are known to control multiple processes.

In Arabidopsis, the WD40 protein TRANSPARENT TESTA GLABRA1 (TTG1) (Walker et al., 1999) is required for the synthesis of anthocyanin and proanthocyanidin pigments, the production of seed mucilage, and the development of trichomes on stems and leaves (Koornneef, 1981), whereas in roots it suppresses the formation of root hairs in certain cells (Galway et al., 1994). During the regulation of anthocyanin synthesis, trichome development, and nonhair development in the root, TTG1 cooperates with two functionally equivalent BHLH proteins encoded by GLABRA3 (GL3) and ENHANCER OF GLABRA3 (EGL3) (Payne et al., 2000; Bernhardt et al., 2003; Ramsay et al., 2003; Zhang et al., 2003), whereas synthesis of proanthocyanidins in the seed coat depends on a distinct BHLH protein encoded by TRANSPARENT TESTA8 (TT8) (Nesi et al., 2000). TTG1 and GL3/EGL3 or TT8 activate these distinct processes by associating with distinct MYB partners. During trichome development, they interact with R2R3 MYBs encoded by GLABROUS1 (GL1) or MYB23 (Oppenheimer et al., 1991; Kirik et al., 2001, 2005), and for the development of nonhair cells in the root they interact with a functionally equivalent MYB encoded by WEREWOLF (WER) (Lee and Schiefelbein, 2001), whereas for anthocyanin synthesis, their MYB partner is probably PRODUCTION OF ANTHOCYANIN PIGMENT1 (PAP1) or PAP2 (Borevitz et al., 2000; Baudry et al., 2004; Zimmermann et al., 2004).

The involvement of anthocyanin regulators in trichome and root hair development is seen only in Arabidopsis and not in other species for which regulatory anthocyanin mutants have been isolated, such as Antirrhinum majus, maize (Zea mays), or petunia. Nevertheless, RED (R) (BHLH) and PALE ALEURONE COLOR (PAC1) (WD40) from maize can restore the hair defects in Arabidopsis ttg1 mutants (Lloyd et al., 1992; Carey et al., 2004), indicating that the functional diversification did not depend on alterations in these WD40 and BHLH proteins but on the divergence of their MYB partners and/or their downstream target genes. Whether these MYB, BHLH, and WD40 proteins also activate vacuolar acidification in species other than petunia is unclear.

To unravel the mechanisms and the biochemical pathways by which AN1, AN11, and AN2 control vacuolar pH, we set out to isolate the genetically defined PH loci by transposon-tagging strategies and the downstream structural genes by RNA profiling methods. Here, we describe the isolation and molecular characterization of PH4. We show that PH4 is a member of the MYB family of transcription factors that is expressed in the petal epidermis and that can interact physically with AN1 and JAF13, a functionally related BHLH protein that can also drive anthocyanin synthesis (Quattrocchio et al., 1998). Because PH4 plays no apparent role in anthocyanin synthesis, we propose that AN1 activates anthocyanin synthesis and vacuolar acidification through interactions with distinct MYB proteins.

Figure 1. Genetic Control of the Anthocyanin Pathway in Petunia Petals.

Figure 2. Phenotypic Analysis of Flower Pigmentation Mutants.

Among progeny of W138 and W137, we found several new mutations affecting flower color (van Houwelingen et al., 1998; Spelt et al., 2002). In one class of mutants, the color of the AN1 or AN11 revertant spots (in an an1-W138 or an11-W137 background) (Figure 2C) or of the whole corolla (in an AN1 or AN11 germinal revertant) had changed from red to purplish (Figures 2D and 2E). Subsequent complementation analyses showed that these mutations represented new alleles of PH2 (allele ph2-A2414), PH3 (allele ph3-V2068), and PH4 (alleles ph4-V2166, ph4-B3021, ph4-X2052, and ph4-V2153) (see Supplemental Table 1 online). Yet another unstable ph4 allele (ph4-X2377) was recovered from the Syngenta breeding program in a family segregating 3:1 for wild types with red petals and mutants with purplish petals with an occasional revertant red spot.

The alleles ph4-V2166, ph4-V2153, an1-W138, an1-W225, and an11-W134 all cause a similar increase in petal extract pH (Figure 2F) (Spelt et al., 2002). To determine at which stage AN1, AN11, PH3, and PH4 are active, we analyzed flowers of different developmental stages. Figure 2G shows that wild-type petals start to acidify around developmental stage 4, when the bud is about to open. In an1, ph3, and ph4 mutants, this acidification is reduced but not blocked completely. This finding suggests that multiple vacuolar acidification pathways operate in petals, some of which are independent of AN1, AN11, PH4, and PH3. Analysis of double mutant flowers showed that the pH of an1 ph4 or an11 ph4 petal homogenates is not significantly different from that of any of the single mutants, suggesting that these genes operate in the same vacuolar acidification pathway (Figure 2F).

To test whether alterations in pigment synthesis also contributed to the color change in ph4 petals, we analyzed anthocyanins in wild-type (R27) and ph4-V2153 petals of closed buds (stage 4) by HPLC. Figure 2H shows that both genotypes accumulate a nearly identical mixture of anthocyanins. R27 petals synthesize only small amounts of flavonol copigments (quercetin derivatives) as a result of its fl/fl genotype, and HPLC analysis did not reveal clear differences in the accumulation of these compounds in ph4 petals (see Supplemental Figure 1 online). Thus, the ph4 mutation has, at least in the R27 genetic background, little or no effect on the synthesis and modification of anthocyanins.

Early genetic work had shown that a mutation in PH4 or a closely linked gene triggers the complete fading of flower color and the disappearance of anthocyanins after opening of the flower bud, if combined with a dominant allele at the FADING locus (Wiering, 1974; de Vlaming et al., 1982). Anthocyanins with a 3RGac5G substitution pattern are particularly sensitive to fading, whereas the 3-glucosides (as in rt mutants like R27 and derived lines) and the 3-rutinosides show little or no fading (de Vlaming et al., 1982). When we crossed the unstable ph4-V2166 allele into a genetic background that allows the synthesis of 3RGac5G-substituted anthocyanins, the flowers displayed upon opening a blue-violet color and were dotted with red-violet spots and sectors resulting from reversions of ph4 (Figure 2H). In the next days, the color of the blue-violet (ph4) cells faded to nearly white, whereas the red-violet (PH4) revertant sectors retained their color. Control crosses with isogenic PH4 plants (lines R27 and W138) yielded only progeny with evenly red-colored, nonfading corollas, whereas crosses with a stable recessive ph4-V2153 parent gave only progeny with evenly colored blue-violet, fading corollas (ph4). This finding demonstrates that it is the mutation of PH4, and not that of a linked gene, that triggers fading.

The an1-G621 allele expresses a truncated AN1 protein that can drive anthocyanin synthesis but not vacuolar acidification (Spelt et al., 2002). When crossed into a background that allows the synthesis of 3RGac5G-substituted anthocyanins, the an1-G621 allele also triggered fading (Figure 2I), and similar results were obtained with the an1-B3196 allele in distinct crosses.

Strikingly, the unstable ph2-A2414 (Figure 2J) and ph5 (data not shown) alleles did not induce fading when crossed into an FA background synthesizing 3RGAac5G-substituted anthocyanins, suggesting that fading in an1 and ph4 petals is not triggered by the upregulation of vacuolar pH alone and may depend on some other vacuolar defect.

To identify the dTPH1 copy in ph4-B3021, we analyzed dTPH1 flanking sequences in mutant and wild-type plants by transposon display (van den Broek et al., 1998) and found a 98-bp fragment that was amplified from the three ph4-B3021 plants analyzed but not from the two wild-type plants homozygous for the parental PH4 allele (Figure 3A). Subsequent isolation, cloning, and sequencing of this fragment showed that it contained 66 bp of dTPH1 sequence and 32 bp of flanking sequence that was identical to a cDNA clone (MYBa) that had been isolated independently by yeast two-hybrid screen using an AN1 bait (see below).

Figure 3. Molecular Analysis of PH4.

PCR experiments with gene-specific primers showed that in ph4-B3021 and ph4-V2166 plants, MYBa was disrupted by a 284-bp dTPH1 insertion in the 5′ and 3′ ends of the protein-coding region, respectively (Figure 3B). Analysis of PH4 progeny plants that originated from germinal reversions of ph4-B3021 (Figure 3C) and ph4-V2166 (Figure 3D) showed that reversion of the ph4 phenotype correlated with excision of dTPH1 from MYBa.

DNA analyses of plants containing ph4-X2052 or ph4-C3540, either in homozygous or heterozygous condition, showed that these mutants also contained dTPH1 insertions in MYBa, whereas the stable recessive ph4-V2153 mutants contained an ~4-kb TPH6 insertion that is nearly identical to the TPH6 element found in the alleles an1-W17 and an1-W219 (Spelt et al., 2002) (Figure 3B). Plants harboring the weakly unstable ph4-X2377 allele contained a 177-bp insertion 29 bp downstream from the translation start site. The insertion is flanked by an 8-bp target site duplication and has a 12-bp terminal inverted repeat with similarity to the terminal inverted repeats of hAT family transposons, including dTPH1 and Activator from maize. Because the internal sequences of the insertion show no similarity to other transposons and are too short to encode a transposase, it apparently represents a new (sub)family of nonautonomous petunia transposons that we named dTPH7. The ph4 lines V64 and M60 both contain a dTPH1 insertion in MYBa in exactly the same position (Figure 3B).

In summary, these data show that mutations in the isolated gene (MYBa) fully correlate with the phenotype conferred by the ph4 mutant, implying that PH4 is identical to MYBa. Below, we refer to this gene as PH4.

Figure 4. Similarity of PH4 to Other MYB Proteins.

R2R3 MYB genes constitute a large family of ~125 genes in Arabidopsis and >85 genes in rice (Oryza sativa) and have been categorized into 42 subgroups based on similarities in the encoded proteins and intron–exon structures (Stracke et al., 2001; Jiang et al., 2004). Generally, sequence similarity between MYB proteins is restricted to the N-terminal MYB domain, but 19 subgroups of R2R3 MYBs share some conserved motifs in their C-terminal domains that may indicate similarities in function (Stracke et al., 2001; Jiang et al., 2004).

PH4 is most similar to the R2R3 MYB proteins BNLGHi233 from upland cotton (Gossypium hirsutum), MYBCS1 and MYB5 from grape (Vitis vinifera), MYB5 from Arabidopsis, and MYB4 from rice (Figure 4A). The clustering of R2R3 MYBs in Figure 4 is in good agreement with previous analyses based on a much larger set of MYBs (Stracke et al., 2001; Jiang et al., 2004) and extends the data at some points. For example, grape MYBA1 and tomato (Solanum lycopersicum) ANT1, recently identified regulators of the anthocyanin pathway in grape (Kobayashi et al., 2004) and tomato (Mathews et al., 2003), respectively, cluster with known members (AN2 of petunia, PAP1 and PAP2 of Arabidopsis) of subgroups 6 and N9 as defined by Stracke et al. (2001) and Jiang et al. (2004), respectively. Curiously, C1 and Pl, regulators of the anthocyanin pathway in maize, cluster in a distinct group (5/N8) together with TT2, a regulator of proanthocyanidin and anthocyanin synthesis in Arabidopsis (Shirley et al., 1995), again consistent with previous results.

ODO1, an activator of the synthesis of phenylpropanoid volatiles in petunia flowers, clusters with Arabidopsis MYB42 and MYB85, similar to previous results (Verdonk et al., 2005), but it does not contain the conserved motif found in the C termini of MYB42 and MYB85 (Jiang et al., 2004).

Figure 4B shows that PH4, Arabidopsis MYB5, grape MYBCS1 and MYB5, cotton BNLGHi233, and rice MYB4 share two conserved motifs in their C-terminal domains. Such motifs have been used as signatures for the classification of subgroups (Stracke et al., 2001; Jiang et al., 2004). Given that these C-terminal motifs in PH4 and related R2R3 MYBs are much better conserved than those in proteins of subgroup N9, we propose that these MYBs form a new subgroup that we tentatively named G20, in accordance with the numbering of Jiang et al. (2004). This classification is supported by the finding that both Arabidopsis MYB5 and PH4 contain only one intron (Li et al., 1996) (Figure 3), whereas the majority of R2R3 MYBs contain two (Jiang et al., 2004).

Yeast two-hybrid assays showed that Arabidopsis MYB5 (Zimmermann et al., 2004) and PH4 (see below) can interact physically with functionally similar BHLH proteins, consistent with the idea that they have similar functions. However, the expression pattern of MYB5 (Li et al., 1996) seems quite different from that of PH4 (see below). Rice MYB4 induces freezing and chilling tolerance when constitutively expressed in Arabidopsis (Vannini et al., 2004), and grape MYB5 induces the synthesis of anthocyanins and proanthocyanidins when overexpressed in tobacco (Nicotiana tabacum) (Deluc et al., 2006). Definite proof for functional equivalence and/or orthology will require the swapping of genes between species and/or the identification of (direct) target genes (see Discussion).

Figure 5. Expression Analysis of PH4.

Ovaries are the only other tissue besides petals in which we detected clear PH4 expression. This organ also expresses AN1 (Figure 5A) (Spelt et al., 2000), which directs the activation of the anthocyanin biosynthetic gene DFR, but for unknown reasons this does not result in the accumulation of anthocyanin pigments (Huits et al., 1994).

Anthers of V30 are pigmented by anthocyanins and express AN1 mRNA during early stages of development (stages 1 to 3). However, no PH4 transcripts were detected in this tissue. The same holds for the stigma and style. AN1 is weakly expressed in sepals, leaves, and stems of V30, which correlates with the synthesis of low amounts of anthocyanin; PH4, however, is not expressed in these tissues. Roots are normally not pigmented and do not express PH4 or AN1.

To examine to what extent the transposon insertions affected the expression of PH4, we analyzed ph4 mRNAs in petal limbs of different mutants by RNA gel blot analysis, RT-PCR (data not shown), and rapid amplification of 3′ cDNA ends (3′RACE). Figure 5B shows that the amount of PH4 transcripts in petal limbs homozygous for the ph4 alleles X2052, B3021, C3540, and V2166 is strongly reduced compared with that of the isogenic wild-type line R27. The same holds for PH4 transcripts expressed from the X2377 allele compared with an isogenic revertant (ph4-X2377R). Presumably, the dTPH1 insertions in these alleles result in highly unstable mRNAs that are rapidly degraded. The small amount of wild-type-size PH4 transcripts in these corollas presumably originates from cells in which dTPH1 was excised from PH4.

The insertions in ph4-V64 and ph4-V2153 cause the accumulation of short ph4 transcripts. Cloning and sequencing of these products revealed that they resulted from polyadenylation within the dTPH1 and TPH6 sequences, respectively, and encode truncated PH4 proteins that contain only part of the DNA binding and protein–protein interaction domains. Although truncations of transcription factors can have dominant-negative effects (Singh et al., 1998; Ferrario et al., 2004), this is apparently not the case here, as ph4-V64 and ph4-V2153 are recessive mutations.

In 23 of 5 × 10⁵ yeast transformants, the introduction of the cDNA plasmid resulted in reproducible activation of the GAL4-responsive HIS3 and ADE2 reporter genes, resulting in His and adenine auxotrophy. Sequence analysis showed that these 23 transformants contained cDNAs derived from four different genes, all encoding MYB proteins. Seventeen clones contained partial cDNAs from PH4, and three clones contained partial cDNAs of AN2. The three remaining clones encode two MYB domain proteins (named MYBb1 and MYBx) that had not been identified previously (Kroon, 2004). A detailed functional analysis of the latter will be published elsewhere.

To analyze the two-hybrid interaction between PH4 and AN1 in more detail, we made new constructs to express the full PH4 (PH4^1-291) and AN2 (AN2^1-255) proteins as fusions to either the activation domain of GAL4 (GAL4^AD) or GAL4^BD (Figure 6A). When expressed (alone) in yeast, PH4^1-191GAL4^BD and AN2^1-255GAL4^BD strongly activated the GAL4-activated HIS3, ADE, and LACz reporter genes, whereas GAL4^BD did not, suggesting that PH4 and AN2 both contain a strong transcription activation domain, as was shown previously for C1 from maize (Goff et al., 1991). Because of the autoactivation of PH4-GAL4^BD and AN2-GAL4^BD fusions, we used instead fusions to GAL4^AD to analyze two-hybrid interactions with AN1 and JAF13.

Figure 6. Interactions between PH4 and Regulators of the Anthocyanin Pathway.

Figure 6B shows that a fusion of full PH4 to GAL4^AD (PH4^1-291GAL4^AD) interacts in a two-hybrid assay with GAL4^BD fusions containing full AN1 or its conserved N-terminal domain (AN1^1-668GAL4^BD and AN1^1-238GAL4^BD), but not with GAL4^BD alone. Subsequent biochemical experiments showed that in vitro translated AN1 and PH4 could be coimmunoprecipitated with an anti-AN1 serum, confirming that the observed two-hybrid response resulted from a direct physical interaction between both proteins (see Supplemental Figure 3 online). The interaction between PH4 and AN1 appeared equally strong as that between AN2 and AN1, because both combinations activated the ADE2 and LACz genes to a similar extent. Furthermore, AN2 and PH4 interacted with similar efficiency with the N-terminal domain of JAF13, as they did with the same domain of AN1.

When tested separately, we found that the C termini of AN2 and PH4 (PH4^127-291 and AN2^115-255) did not interact at all with AN1 or JAF13, whereas the N-terminal parts containing the MYB domain (PH^1-134 and AN2^1-121) gave a weak two-hybrid response that could be unambiguously scored with the ADE reporter but was barely detectable with the LACz reporter (Figure 6B). The weak response of the separated AN2 and PH4 domains relative to the full proteins may result from (1) a reduced transcription activation potential of PH4^1-134 and AN2^1-121complexes relative to the full proteins, attributable to the removal of the strong transcription activation domain in their C termini, (2) aberrant folding of the isolated MYB domains, and/or (3) the MYB regions containing only part of the interaction domain.

A detailed characterization of maize C1 (Grotewold et al., 2000) and Arabidopsis PAP1 (Zimmermann et al., 2004) identified several amino acids in the first helix of the R3 domain that are critical for the interaction with their BHLH partner (R and EGL3, respectively) and/or the transcription activation of downstream genes. These residues are conserved in AN2 and PH4 (Figure 4B) and may play a critical role in the interaction with AN1 and JAF13 and the activation of downstream genes.

To assess whether PH4 and AN1 might be able to form complexes in vivo, we performed in situ hybridization analysis to examine whether PH4 and AN1 are coexpressed in the same petal cells. Figure 6C shows that DFR is expressed in the upper (adaxial) epidermis of the corolla and at a much lower level in the lower (abaxial) epidermis, consistent with the anthocyanin pigmentation pattern. AN1 mRNA was expressed in a very similar pattern: relatively high expression is seen in the upper epidermis, and lower expression (just above the detection limit) is seen in the lower epidermis. PH4 mRNA could be detected clearly in the upper epidermis. However, because of the low expression levels of PH4 mRNA, it was difficult to differentiate the weak colorigenic signals that were sometimes observed in the lower epidermis or the mesophyll from the background signal that results from nonspecific binding of the probe. Hence, we cannot exclude or confirm that PH4 is also weakly expressed in these cells. Together, these findings indicate that AN1 and PH4 may form complexes in the upper epidermis of the flower corolla, which are the same cells in which the color change is seen in an1 and ph4 mutants.

To test directly whether PH4 is involved in the expression of structural anthocyanin genes, we measured the amount of several of their mRNAs in wild-type (R27) and isogenic an1-W225 and ph4-V2153 petals. Figure 7A shows that in an1 petal limbs the structural genes CHS, CHI, and F3H are normally expressed, whereas the expression of DFR, AS (encoding anthocyanin synthase [Weiss et al., 1993]), AAT (encoding an anthocyanin-rutinoside acyltransferase; Brugliera and Koes, 2001; F. Brugliera and R. Koes, unpublished data), and AN9 (encoding a glutathione transferase–like protein; Alfenito et al., 1998) is strongly reduced. However, the ph4 mutation has no clear effect on any of these mRNAs. This finding indicates that PH4 plays no role in the regulation of anthocyanin genes, or that its role is redundant with that of another gene.

Figure 7. Effect of PH4 Function on Gene Expression.

To discriminate between these possibilities, we analyzed the activity of PH4 in a gain-of-function assay. We introduced a PH4 gene construct driven by the constitutive 35S promoter of Cauliflower mosaic virus (35S:PH4) into leaf cells by particle bombardment and measured its capacity to activate a LUCIFERASE (LUC) reporter driven by the promoter of DFR (DFR:LUC). In these experiments, we cobombarded a β-GLUCURONIDASE (GUS) reporter gene driven by the 35S promoter (35S:GUS) to correct for variations in transformation efficiency. Figure 7B shows that the DFR:LUC reporter showed little or no expression in leaf cells when introduced alone. Consistent with earlier results (Spelt et al., 2000, 2002), the expression of DFR:LUC was strongly induced (>35-fold) when it was cointroduced with 35S:AN1 and 35S:AN2. However, when 35S:AN2 was replaced by 35S:PH4, the DFR:LUC reporter was not induced, nor was it when we cointroduced 35S:JAF13 in addition. These findings show that for activation of the DFR promoter, AN2 cannot be replaced by PH4, suggesting that PH4 plays no role in the transcriptional activation of DFR.

To unravel the nature of the acidification pathway that is activated by PH4, we analyzed transcripts expressed in wild-type, an1, ph3, ph4, and ph5 petals by microarray and cDNA-AFLP analysis and identified nine mRNA fragments whose expression is reduced in an1 and ph4 petals (a full account of these experiments will be published elsewhere). One of the identified mRNAs, named CAC16.5 (for cDNA-AFLP clone 16.5), encodes a Cys proteinase–like protein (GenBank accession number AY371317) (see Supplemental Figure 4 online). Figure 7A shows that CAC16.5 mRNA is strongly reduced in an1 and ph4 petals compared with wild-type petals, indicating that CAC16.5 expression requires both AN1 and PH4.

To examine the role of AN2 in CAC16.5 expression, we analyzed CAC16.5 mRNAs in an2 petals and isogenic controls in which the an2 mutation was complemented by a 35S:AN2 transgene. Figure 7C shows that the restoration of flower pigmentation by 35S:AN2 was accompanied by strong (re)induction of DFR mRNA expression, consistent with previous results (Quattrocchio et al., 1998; Spelt et al., 2000). However, the expression of CAC16.5 and PH4 mRNA was similar in an2 and an2 35S:AN2 petals, indicating that AN2 has little or no effect on CAC16.5 or PH4 expression.

To examine the regulation of CAC16.5 by PH4 and AN1 in further detail, we analyzed transgenic plants containing 35S:PH4 and 35S:AN1. Figure 7D shows that CAC16.5 and DFR mRNAs are not expressed in wild-type leaves. Although the 35S:PH4 line expresses similar amounts of PH4 mRNA in leaves as the endogenous PH4 gene does in petals, this results only in a very weak activation of CAC16.5 in leaves, to a level that is barely visible in Figure 7D. Although the 35S:AN1 line expressed AN1 mRNA in leaves to a level that is twofold to threefold higher than that found in wild-type petals, no induction of CAC16.5 or DFR mRNA was observed. Thus, both AN1 and PH4 are on their own insufficient to induce DFR or CAC16.5 efficiently.

To obtain plants that ectopically express AN1 and PH4, we crossed the 35S:AN1 and 35S:PH4 lines. The 35S:PH4 parent used apparently contains multiple active copies of the transgene, because after crossing to other empty lines, all 84 progeny contained and expressed a 35S:PH4 transgene. The 35S:AN1 parent (Spelt et al., 2000), however, contained a single active transgene locus that segregated as a single Mendelian factor in crosses. After crossing the 35S:PH4 and 35S:AN1 parents, all 39 progeny contained 35S:PH4, whereas only 4 of them contained 35S:AN1 (plants 1 to 4). Plant 1 grew very poorly and died before it could be subjected to more detailed analyses. Plant 2 grew slowly, remained severely stunted, and bore leaves with a weak and patchy anthocyanin pigmentation pattern (see Supplemental Figure 5 online). Plants 3 and 4 developed normally. Moreover, leaf extracts of plant 2 were significantly more acidic than those of plants 3 and 4 or control plants that had not been transformed. RT-PCR analysis showed that plant 2 expressed the 35S:PH4 and 35S:AN1 gene in leaves to a similar extent as the parental lines did. In plants 2 and 3, however, the expression of 35S:AN1 was significantly reduced compared with the parental 35S:AN1 line, possibly as a result of (epigenetic) silencing of the transgene. Together, it appears that strong, ectopic coexpression of AN1 and PH4 has a deleterious effect, possibly attributable to ectopic activation of a vacuolar acidification pathway.

Despite the low expression of 35S:AN1 in plants 3 and 4, CAC16.5 RNA was induced in their leaves to levels that are close to CAC16.5 mRNA in petals, whereas DFR was hardly induced at all. In plant 2, however, the induction of CAC16.5 was even stronger than in leaves of plants 3 and 4; moreover, DFR was clearly activated, although the actual mRNA levels remained far lower than in petals. Given the aberrant phenotype of plant 2, we cannot distinguish whether the activation of DFR is a direct effect of AN1 and PH4 expression or an indirect (stress) effect caused, for example, by the physiological changes that are induced by AN1 and PH4 (see Discussion).

The BHLH and WD40 regulators AN1 and AN11 activate, in addition to anthocyanin synthesis, several other aspects of epidermal cell differentiation, such as growth and division of cells in the seed coat epidermis and vacuolar acidification in petals (Spelt et al., 2002). Here, we present evidence that AN1 activates these processes through interactions with two distinct MYB proteins encoded by AN2 and PH4.

In the R27 genetic background (rt/rt), the ph4 mutation does not affect the expression of structural anthocyanin genes or the accumulation of anthocyanin pigments, suggesting that the color change is attributable largely to the effect on vacuolar pH (Figure 2). However, in a genetic background containing RT (to allow the synthesis of 3RGac5G-substituted anthocyanins) and the dominant FADING allele (de Vlaming et al., 1982), ph4 or an1-G621 triggers the almost complete disappearance of anthocyanins in the petals and the fading of the flower color after opening of the bud (Figure 2).

The dependence of fading on a dominant FA allele and a 3RGac5G substitution pattern of the anthocyanins suggests that fading may involve an active (enzymatic) substrate-specific degradation process. However, because FA is not isolated and because the origin of the FA and fa alleles is unclear, it cannot be ruled out that FA is actually a dominant-negative allele and that the recessive fa allele encodes an active protein required for the stability of anthocyanins. Furthermore, it cannot be excluded that fading does not depend on the genes RT and GLYCOSYLATION AT FIVE (GF) and the resulting 3RGac5G anthocyanin substitution pattern but on distinct genes that are genetically linked to RT and GF. Consequently, the molecular role of PH4 in (preventing) fading is difficult to infer at this stage. Because ph2 and ph5 mutations increase vacuolar pH without inducing fading, it is possible that fading in ph4 and an1 is not attributable to a less acidic vacuolar lumen alone but depends on an additional defect in vacuolar physiology.

Given that PH4 is a MYB domain protein that can interact with the transcription factors AN1 and JAF13, it presumably activates vacuolar acidification indirectly, by regulating the transcription of downstream genes that encode proteins involved in proton metabolism. We consider it unlikely that PH4 activates the expression of one or more subunits of a v-ATPase proton pump, because mutations that inactivate v-ATPase were shown to cause severe developmental defects and often lethality in a variety of species (Davies et al., 1996; Ferea and Bowman, 1996; Inoue et al., 1999; Schumacher et al., 1999; Oka and Futai, 2000; Strompen et al., 2005). Moreover, immunoblot analysis showed that the v-ATPase A and B subunits as well as pyrophosphatase proteins are expressed at equal levels in PH4 and ph4 petals (N. Frange and E. Martinoia, unpublished data).

To determine the nature of the vacuolar pathway downstream of AN1 and PH4, we started the analysis of transcripts that are expressed at reduced levels in an1 and/or ph4 mutants. One of these is CAC16.5, which encodes a protein homologous with Cys proteinases that is expressed in petals but not in leaves. The CAC16.5 expression domain is largely determined by AN1 and PH4, as the gene is inactivated in an1 and ph4 petals. Moreover, forced expression of AN1 and PH4 is sufficient to activate CAC16.5 in leaves (Figure 7).

Cys proteases have regulatory functions in a variety of processes. In animals, Cys proteases play an important role in programmed cell death. In plants, Cys proteases have also been implicated in programmed cell death (Elbaz et al., 2002; Kuroyanagi et al., 2005) and various aspects of defense responses (Kruger et al., 2002; Hatsugai et al., 2004; Matarasso et al., 2005). Recent data suggest that (some) Cys proteases have DNA binding activity and can be involved directly in transcriptional activation (Matarasso et al., 2005). Currently, we are investigating the function of CAC16.5 and other AN1/PH4-controlled genes in vacuolar acidification by reverse genetic strategies.

First, screening of a petal cDNA library with the N terminus of AN1 revealed interactions with only four proteins (AN2, PH4, MYBb, and MYBx) (cf. Figure 6) but not with numerous other MYBs that are expressed in petals at much higher levels (Avila et al., 1993; Mur, 1995; van Houwelingen et al., 1998). This indicates that the observed yeast two-hybrid interactions are specific. Moreover, the four interacting MYBs all seem to have a function in pigmentation. AN2 is known to regulate structural anthocyanin genes together with AN1 (Quattrocchio et al., 1993, 1998; Spelt et al., 2000) (Figure 7B), whereas PH4 regulates, together with AN1, several aspects of vacuolar physiology (acidification and fading) (de Vlaming et al., 1982, 1983; Spelt et al., 2002; this study). MYBb1 and the closely related homolog MYBb2 encode proteins with high similarity to AN2 and are required for the activation of anthocyanin synthesis in distinct floral tissues, whereas MYBx appears to be an inhibitor of AN1 in anthocyanin synthesis and vacuolar acidification (Kroon, 2004).

Second, in vitro synthesized PH4 and AN1 proteins form an immunoprecipitable complex, indicating that the PH4–AN1 interaction is direct.

Third, AN1 and PH4 mRNAs are both expressed in the epidermis of petal cells; thus, AN1–PH4 complexes have the potential to be formed there (Figure 5D). Ovaries also express AN1 and PH4 mRNA, but because it is unknown whether both genes are coexpressed in the same cells, it is unclear whether AN1–PH4 complexes can also form in this organ.

Fourth, in petals, anthocyanin accumulation is limited to epidermal cells, and the underlying mesophyll cells are uncolored (Koes et al., 1990; Huits et al., 1994). Hence, the flower color phenotypes of an1, an2, and ph4 mutants result from alterations in the very same cells.

Fifth, mutations in an1 and ph4 affect the same target processes: vacuolar acidification, flower color fading, and the expression of CAC16.5. Moreover, AN1 and PH4 can induce ectopic expression of CAC16.5, but only when coexpressed. Because ph2 and ph5 mutations do not cause fading, it seems that fading is attributable to some other defect in vacuole structure or physiology occurring in both an1 and ph4.

Ectopic expression of the putative PH4 homologs from rice (MYB4) in heterologous systems can activate the phenylpropanoid gene PAL (encoding phenylalanine ammonia lyase) and other pathways that result in dwarfism and induced freezing/chilling tolerance (Vannini et al., 2004). The induction of PAL, however, is very moderate (less than twofold) (Vannini et al., 2004), possibly because the BHLH partner of rice MYB4 was missing or because the induction is indirect. Ectopic overexpression of the grape homolog MYB5 in tobacco enhanced the pigmentation in petals and stamens but not in other tissues (Deluc et al., 2006), possibly reflecting the dependence on the expression of an endogenous BHLH partner. Curiously, both early (CHS, CHI, and F3H) and late (DFR) anthocyanin genes were induced by MYB5 (Deluc et al., 2006). Whether this overexpression phenotype reflects the normal MYB5 function is difficult to assess from the available data.

Given that anthocyanin genes are normally expressed in ph4 flowers, these appear to be activated through interactions with a distinct R2R3 MYB protein. Most likely, the R2R3 MYB partner required for the activation of anthocyanin genes is AN2, because (1) AN2 is known to be required for the expression of structural anthocyanin genes (Quattrocchio et al., 1993); (2), it is, in addition to PH4, one of the few MYB proteins in petals that can bind to AN1 (Figure 6; see Supplemental Figure 3 online); (3) it can functionally replace C1 of maize (Quattrocchio et al., 1998), which is known to bind directly to the promoters of structural anthocyanin genes (Sainz et al., 1997); and (4) coexpression of AN2 with AN1 or JAF13 activates DFR fast enough to be detected within 24 h in transient expression assays, consistent with AN2 activating DFR directly, like AN1 (Quattrocchio et al., 1998, 1999; Spelt et al., 2000) (Figure 6). Moreover, using GLUCOCORTICOID RECEPTOR fusions, it was shown that TT2 of Arabidopsis activates DFR directly (Baudry et al., 2004). Together, these findings suggest a model in which AN1 activates vacuolar acidification or anthocyanin synthesis by binding to PH4 or AN2, respectively.

Work in Arabidopsis led to very similar models, which propose that the partially redundant BHLH proteins GL3 (Payne et al., 2000), EGL3 (Bernhardt et al., 2003; Zhang et al., 2003), and TT8 (Nesi et al., 2000) activate the synthesis of anthocyanin and proanthocyanidin pigments, the production of seed mucilage, the development of trichomes on leaves and stems, and the specification of nonhair fate (atrichoblast) in certain cells of the root epidermis through interactions with distinct R2R3 MYB proteins that are specific for each pathway (Zhang et al., 2003; Baudry et al., 2004; Pesch and Hulskamp, 2004; Zimmermann et al., 2004).

These simple models explain many of the observations, but not all of them, and therefore are likely to be incomplete and possibly even incorrect in some details.

First, these models predict that overexpression of PH4 might compete with AN2, resulting in an an2-like phenotype (reduced anthocyanin synthesis), whereas overexpression of AN2 might result in a ph4-like phenotype (blue flowers). However, such cross-inhibition phenotypes were not observed in any of the 35S:PH4 or 35S:AN2 lines (Quattrocchio et al., 1998; Spelt et al., 2000; this study). Also, for Arabidopsis lines overexpressing TT2, PAP1, GL1, WER, or Arabidopsis MYB23, no cross-inhibitory effects of anthocyanin regulators on hair development, or vice versa, were reported (Szymanski and Marks, 1998; Lee and Schiefelbein, 1999, 2002; Borevitz et al., 2000; Kirik et al., 2001; Nesi et al., 2001).

Second, these simple models do not explain the small but clear effect of AN2 on vacuolar acidification (Spelt et al., 2002). Gain-of-function experiments indicated that AN2 and TT2 might play an additional role as a higher order regulator in transcription activation of their BHLH partners, AN1 and TT8, respectively (Spelt et al., 2000; Nesi et al., 2001). This indirect effect of AN2 and TT2 on DFR expression is evident only in transgenic plants that constitutively express AN2 or TT2 (Spelt et al., 2000; Nesi et al., 2001), but it is apparently too slow to contribute to the induction of DFR within the <24 h of a transient expression assay (Quattrocchio et al., 1998; Spelt et al., 2000; Zimmermann et al., 2004). Thus, it is possible that AN2 affects vacuolar acidification indirectly through the activation of another regulatory gene that is distinct from AN1 or PH4, because these are normally expressed in an2 petal limbs (Figure 6) (Spelt et al., 2000).

Alternatively, it is possible that in vivo, larger complexes are formed that contain multiple R2R3 MYB and BHLH proteins, because yeast two-hybrid experiments indicated that TT2 can homodimerize (Baudry et al., 2004) and that EGL3 and GL3 as well as AN1 and JAF13 can form both homodimers and heterodimers (Zhang et al., 2003; Kroon, 2004). Thus, the in vivo interactions of AN2 and PH4 with AN1 need not be mutually exclusive. Clearly, the current genetic approaches need to be combined with biochemical analyses of protein complexes to solve this issue.

To introduce an1-G621 in a background synthesizing 3RGac5G-substituted anthocyanins, several progeny plants with an an1-W138, RT phenotype were selected from the backcross (W138 × V30) × W138 (de Vetten et al., 1999) and subsequently crossed to line R153 (an1-G621, hf1, rt). Detailed information about the structure and resulting phenotypes of an1-W138 and an1-G621 can be found elsewhere (Spelt et al., 2000, 2002).

The actual pH values measured for specific plants showed some variation in time, possibly as a result of variable environmental conditions in the greenhouse, but the differences between distinct genotypes were constant. Therefore, the absolute pH values can be reliably compared between samples/genotypes that were measured within one experiment (i.e., one figure panel) but much less between distinct experiments (i.e., distinct figure panels).

A full-size PH4 cDNA clone was isolated by screening a petal cDNA library made from the PH4 line R27; the corresponding genomic region was amplified from R27 with primers complementary to the cDNA ends.

cDNA-AFLP analysis of RNA isolated from wild-type, an1, ph3, ph4, and ph5 petal limbs was performed essentially as described (Bachem et al., 1996) using the restriction enzymes MseI and EcoRI and corresponding adapters and primers. CAC16.5 was identified as a 161-bp cDNA-AFLP fragment that was present in wild-type and ph5 petals but absent in an1, ph3, and ph4 petals. A 1169-bp CAC.16.5 cDNA containing part of the protein-coding sequence was isolated by screening a petal cDNA library with the 161-bp fragment.

RNA isolation and RT-PCR analysis were performed as described (de Vetten et al., 1997; Quattrocchio et al., 1998). cDNA products were amplified using primers specific for CHSa (primers 275 and 604), CHIa (311 and 312), AN3/F3H (112 and 113), DFR (97 and 98), AS (433 and 434), AAT (435 and 436), AN9 (229 and 230), CAC16.5 (1871 and 1769), PH4 (1233 and 1317), AN1 (123 and 126), or GAPDH (19 and 20) and a reduced number of PCR cycles (GAPDH and CAC16.5, 18 cycles; CHSa, CHIa, F3H/AN3, DFR, AS, AAT, and AN9, 20 cycles; AN1 and PH4, 24 cycles). Because the yield of PCR products was too low for detection by ethidium bromide, they were visualized by gel blot hybridization (using ³²P-labeled cDNA probes) and phosphor imaging, For each gene–primer combination, we determined whether the RT-PCR response was linear and quantitative (at different numbers of PCR cycles) by assaying dilution series of cDNA samples in which transcripts were highly abundant or that had been spiked with cDNA fragments.

The 3′RACE of PH4 was done as described previously (Frohman et al., 1988) using a PH4-specific primer (1136) and a primer complementary to the poly(A) tail.

RNA in situ hybridization was performed as described (Souer et al., 1996) using RNA probes obtained by in vitro transcription with T7 or SP6 RNA polymerase. The templates were generated by PCR on cDNAs cloned in plasmid vectors using a vector-specific primer that anneals just upstream of the T7 or SP6 promoter and one gene-specific primer. The PH4 probe spanned the region encoding the C-terminal domain (generated with primer 582 and a vector-specific primer) and lacked the conserved MYB domain region. For AN1 and DFR, a cDNA fragment spanning the coding sequence and the 3′ untranslated region was used. Because anthocyanins present in flowers made it difficult to detect the weak colorigenic signals for AN1 and PH4 mRNA unambiguously, we used petals from the an3 lines W62 and W37.

To construct the tree shown in Figure 4A, we aligned the 104–amino acid region spanning the MYB domain (see Supplemental Figure 2 online), as defined by Stracke et al. (2001), taken from the following GenBank accession numbers: Ph PH4, AY973324; Ph AN2, AAF66727; Ph ODO1, AAV98200; Ph MYBPh1, CAA78386; Ph MYBPh2, CAA78387; Ph MYBPh3, CAA78388; Gh BNLGHi233, AAK19611; Vv MYBCS1, AAS68190, Vv MYB5, AAX51291; Sl ANT1, AAQ55181; Vv MYBA2, BAD18978; Vv MYBA1, BAD18980; Vv MYBA3, BAD18179; Os MYB4, BAA23340; Zm Pl, AAA19821; Zm C1, AAK09327. For sequences of the Arabidopsis thaliana MYB proteins, we used the same accessions as Stracke et al. (2001).

Table 1. Description of Primers Used for PCR

For all experiments, we used the yeast strain PJ69 (James et al., 1996), which harbors HIS3, ADE2, and LACz reporter genes driven by distinct GAL4-responsive promoters.

A petunia cDNA library was generated in pAD-GAL4-2.1 from poly(A)⁺ RNA that had been isolated from petal limbs of line R27 using the HybriZap system (Stratagene), according to the instructions of the supplier. The original library (8 × 10⁷ primary recombinants) was amplified and used for mass excision of plasmid DNA according to the instructions. This cDNA library was screened by introducing 10-μg portions of library DNA into yeast PJ69 cells expressing the N-terminal 238 amino acids of AN1 fused to the GAL4 DNA binding domain (AN1^1-238GAL4^BD). Transformed cells were plated on dropout medium lacking Leu, Trp, and His. After ~1 week of growth at 30°C, the colonies that appeared were replicated on plates lacking Leu, Trp, His, and adenine and grown for another 3 d at 30°C. Plasmid DNA was isolated from positive (HIS, ADE) colonies, transformed in Escherichia coli, and sequenced. Plasmids isolated from E. coli were subsequently reintroduced in PJ69 yeast cells, either alone or together with the bait plasmid expressing AN1^1-238GAL4^BD. Only plasmids that specified a HIS, ADE phenotype when cotransformed with the bait vector, but not when cotransformed with the empty pAD-GAL4-2.1 vector or when transformed alone, were considered positives and analyzed further.

Transient expression assays were performed by particle bombardment of petunia W115 leaves (genotype AN1, an2, an4, AN11, PH4) as described previously (de Vetten et al., 1997; Quattrocchio et al., 1998). The an2 and an4 mutations in W115 do not affect the expression of anthocyanin genes in tissues other than petals and anthers (Quattrocchio et al., 1993).

Transgenic plants expressing 35S:PH4 were obtained by Agrobacterium tumefaciens–mediated leaf disc transformation of the F1 hybrid V26 × R162 (genotype ph2). Of four transformants, one expressed the 35S:PH4 transgene, resulting in partial rescue of the ph2 phenotype. This transformant was crossed to 35S:AN1 plants in a W242 (an1, PH2) background (Spelt et al., 2000) or, as a control, to nontransgenic W242 plants to obtain 35S:PH4 plants in a wild-type (AN1, AN2, AN11, PH4) and 35S:AN1 background. Segregation of the transgenes in these crosses was monitored by PCR, using primer combinations specific for each transgene.

Because for an2 null mutants no isogenic wild-type (AN2) lines are available, we transformed an an2 F1 hybrid (W115 × W59; genotype AN1, an2, an4, RT,AN11, PH2, PH4) with 35S:AN2 or, as a control, the empty T-DNA vector and used plants in which 35S:AN2 complemented the an2 mutation for comparison with an2 plants containing the empty vector (Quattrocchio et al., 1998).

For (co)immunoprecipitation, translation products were mixed in a 1:1 ratio, and 1 μL of a mouse anti-AN1 serum (Spelt et al., 2002) was added and incubated for 1 h at 4°C. Subsequently, 10 μL of protein G–agarose beads (Actigen) was added and incubation was continued for 16 h at 4°C in a rotation apparatus. The agarose beads were washed twice in 10 mM Na-phosphate, pH 7.2, 150 mM NaCl, and 0.05% Tween, resuspended in SDS sample buffer, boiled for 2 min, and size-separated on a 12.5% polyacrylamide-SDS protein gel. Gels were fixed and dried, and radioactivity was detected using a phosphor imager.

• Supplemental Figure 1. HPLC Analysis of Methanol-Extractable Anthocyanins and Phenylpropanoids in Petal Limbs of PH4 and ph4-V2153 Flowers in the R27 Genetic Background.
• Supplemental Figure 2. Alignment of MYB Domains of PH4 and Other R2R3 MYB Proteins.
• Supplemental Figure 3. Coimmunoprecipitation of AN1 and PH4 Synthesized in Vitro by Coupled Transcription and Translation.
• Supplemental Figure 4. Alignment of the Protein Encoded by a Partial CAC16.5 cDNA of Petunia to Cys/Thiol Proteases of Other Species.
• Supplemental Figure 5. Phenotype of 35S:AN1 35S:PH4 Double Transgenic Plants.
• Supplemental Table 1. Complementation Analysis of ph2, ph3, and ph4 Mutants.

We thank Pieter Hoogeveen, Martina Meesters, and Daisy Kloos for plant care and Riet Vooijs for help with the HPLC analyses. This research was supported by grants from the European Union, the Netherlands Technology Foundation, and the Netherlands Organization for Scientific Research.