Different mechanisms participate in the R-dependent activity of the R2R3 MYB transcription factor C1.

The R2R3 MYB transcription factor C1 requires the basic helix-loop-helix factor R as an essential co-activator for the transcription of maize anthocyanin genes. In contrast, the R2R3 MYB protein P1 activates a subset of the C1-regulated genes independently of R. Substitution of six amino acids in P1 with the C1 amino acids results in P1(*), whose activity on C1-regulated and P1-regulated genes is R-dependent or R-enhanced, respectively. We have used P1(*) in combination with various promoters to uncover two mechanisms for R function. On synthetic promoters that contain only C1/P1 binding sites, R is an essential co-activator of C1. This function of R is unlikely to simply be the result of an increase in the C1 DNA-binding affinity, since transcriptional activity of a C1 mutant that binds DNA at a higher affinity, comparable with P1, remains R-dependent. The differential transcriptional activity of C1 fusions with the yeast Gal4 DNA-binding domain in yeast and maize cells suggests that part of the function of R is to relieve C1 from a plant-specific inhibitor. A second function of R requires cis-regulatory elements in addition to the C1/P1 DNA-binding sites for R-enhanced transcription of a1. We hypothesize that R functions in this mode by binding or recruiting additional factors to the anthocyanin regulatory element conserved in the promoters of several anthocyanin genes. Together, these findings suggest a model in which combinatorial interactions with co-activators enable R2R3 MYB factors with very similar DNA binding preferences to discriminate between target genes in vivo.

Flowering plants express a large number of proteins containing the conserved R2R3 MYB DNA-binding domain. About 125 R2R3 Myb genes are present in the Arabidopsis genome (1), and many more are predicted to be expressed in maize and related monocots (2,3). Similar to other transcription factor families, the R2R3 MYB factors show exquisite regulatory specificity in vivo, while recognizing very similar DNA sequences in vitro (4 -9). Thus, mechanisms other than discrimination between similar DNA-binding sites are at play in the control of specific sets of target genes by each R2R3 MYB transcription factor in vivo.
The regulation of flavonoid biosynthetic gene expression by the cooperation of R2R3 MYB and basic helix-loop-helix (bHLH) 1 transcription factors provides one of the best described examples of combinatorial gene regulation in plants (10,11). Anthocyanin accumulation in maize is controlled by two classes of regulatory proteins that act in concert: C1 or PL1, two closely related R2R3 MYB domain proteins (12), and R or B, which are members of the R/B family of bHLH domain proteins (13). Extensive genetic and molecular studies have shown that the C1 or Pl1 genes require a member of the bHLH-containing R or B gene family to activate transcription of the anthocyanin biosynthetic genes (10). The C1-and R/Bencoded proteins physically interact, and this interaction is mediated by the MYB domain of C1 and the N-terminal region of B (14) or R (15).
The maize P1 gene, encoding another R2R3 MYB transcription factor, controls the accumulation of 3-deoxyflavonoids and red phlobaphene pigments by activating a subset of the anthocyanin biosynthetic genes controlled by C1 and R, without a need for a known bHLH partner. P1 and C1 activate the expression of some common genes in the flavonoid pathway such as a1, and they interact with different affinities to the same cis-acting regulatory elements in the a1 gene promoter (4,6). The a1 promoter has a modular structure in which the proximal high affinity P1 binding sites ( ha PBS) and the distal low affinity P1-binding sites ( la PBS) are separated by the anthocyanin regulatory element (ARE) (16). In transient expression experiments, these three elements contribute to the regulation of a1 by P1 or by C1 ϩ R (4,6,17). In addition, transposon insertions and mutations in the ARE differentially affect the in vivo regulation of a1 by P1 or C1 ϩ R (18). Unlike P1, C1 activates the transcription of the a2, bz1, and bz2 genes, which are specific for the anthocyanin branch of the pathway (4,6,16). The a2, bz1, and bz2 gene promoters are also modular (16,19), containing ARE and C1-binding sites.
Although the MYB domains of P1 and C1 are over 70% identical (20) and they recognize very similar DNA sequences (4), only C1 has an absolute requirement for the bHLH factor R to activate transcription of the anthocyanin biosynthetic genes (R-dependent transcription). In contrast, P1 controls gene expression independently of R (R-independent transcription). The substitution of six residues in the MYB domain of P1 with the corresponding residues from C1 generates the P1* protein (Fig.  1A), which, unlike P1, is able to physically interact with R (15). Similar to P1, P1* activates a1 but not bz1 in the absence of R, indicating that the DNA-binding properties of P1* have not been altered. Interestingly, however, in the presence of R, P1* mediates a robust activation of the bz1 promoter (15).
The cooperation between bHLH and R2R3 MYB transcription factors is not limited to the regulation of flavonoid biosynthesis by C1-and R-like proteins. The GL1 R2R3 MYB protein interacts with the GL3 and EGL3 bHLH factors to regulate the accumulation of trichomes in Arabidopsis (21,22). Similarly, the bHLH rd22BP1 and R2R3 MYB AtMYB2 Arabidopsis proteins cooperate for drought-and abscisic acid-regulated gene expression (23). Whereas the Arabidopsis genome contains more than 120 genes encoding bHLH proteins (24 -27), the factors that cooperate with R2R3 MYB proteins belong to a small subgroup of bHLH proteins that share a common motif in their N termini. 2 This motif corresponds to the region in R that interacts with C1 (14,15). These findings suggest a general mechanism of cooperation between R2R3 MYB proteins and bHLH factors in transcriptional regulation. How this cooperation contributes to the regulatory specificity of R2R3 MYB proteins is the subject of this study.
Herein, we have investigated the cooperation between C1 and R for the regulation of flavonoid biosynthetic genes. Using P1*, we uncovered two components for this synergy. One component, manifested by the R-enhanced activity of C1 and P1* on the a1 promoter requires, in addition to the high affinity P1-binding sites, cis-regulatory sequences within the ARE. The second component is manifested by the R-dependent activity of C1 on promoters containing only the ha PBS. We generated a mutant of C1 (C1 SH ) that binds DNA with a higher affinity than C1 and comparable to P1. Using C1 SH , we demonstrate that the R-dependent activity of C1 is not solely due to the intrinsic low DNA-binding affinity of C1, since the C1 SH transcriptional activity continues to be R-dependent. The differential activity in yeast and maize cells of chimeras of C1 with the yeast Gal4 DNA-binding domain suggests that part of the function of R is to relieve C1 from an inhibitor. Together, our findings uncover two distinct and separable mechanisms by which R cooperates with C1 for transcriptional activity. In addition, our results provide a model to explain how R2R3 MYB factors with related DNA binding preferences achieve regulatory specificity through combinatorial interactions with accessory factors.

EXPERIMENTAL PROCEDURES
Plasmids Used in Yeast Experiments-The P1 and C1 cDNAs (20,28) were cloned under the control of the glyceraldehyde 3-phosphate dehydrogenase (GAPDH) promoter in the YEplac112 vector (29). A control vector was also constructed with the GAPDH promoter with no downstream gene. The fusions of the MYB domains of P1 and C1 to the Gal4 activation domain (Gal4 AD ) were constructed by synthesizing the MYB domains and the Gal4 activation domain by PCR and then ligating them at an engineered NotI site. The fusion was then cloned under the GAPDH promoter in YEplac112. The yeast expression constructs corresponding to the N-terminal region of R fused to the Gal4 DNA-binding domain (Gal4 DBD -R 1-252 ) was the previously described fusion (15), but with the yeast selection marker changed from TRP1 to LEU2 by replacing the Gal4 AD with the Gal4 DBD -R 1-252 fusion in pADGAL4 (Stratagene).
Plasmids Used in Transient Expression Experiments-As previously described, all P1 and C1 plant expression vectors include the CaMV 35S promoter, the TMV ⍀Ј leader, the first intron of maize Adh1-S in the 5Ј-untranslated region, and the potato proteinase II (pinII) termination signal. Plasmids described previously include p35SP1, p35SC1, p35SR, p35SP1* (corresponding to p35SP I77L,K80R,A83R,T84L,S94G,H95R (15)); pBz1Luc, containing 2.3 kb of the bz1 promoter and intron upstream of luciferase; and pA1Luc, containing 1.4 kb of the a1 promoter and the Adh1-S intron (4,6,15). The p35SC1 SH construct was generated by PCR mutagenesis of p35SC1 using the QuikChange® XL kit (Stratagene). p35SBAR (33) was used for normalizing the concentration of the 35S promoter delivered in each bombardment. The p35SGal4 DBD -C1 C-term construct was previously described (36). The pBz2Luc construct was also previously described (34) and was kindly provided by Dr. Bodeau. The pA2Luc construct was obtained from Dr. Lesnick (16). The construct containing three copies of the ha PBS present in the a1 gene upstream of the luciferase reporter (p( ha PBS) 3 Luc) corresponds to the 3ϫAPB1-35S construct previously described (6). The pGal4 BS Luc reporter construct consists of four CGGAGTACTGTCCTCCGAG motifs in tandem upstream of a minimal CaMV 35S promoter. The pA1 315L Luc construct is identical to pA1Luc, except for the 6-bp insertion present in the a1 315L allele, left as a consequence of the excision of Spm from the a1-m2-7991A::Spm-s allele (18). pUbiGUS (15) was used to normalize the efficiency of each bombardment.
Microprojectile Bombardment and Gene Expression Experiments-Bombardment conditions of maize black Mexican sweet (BMS) suspension cells and transient expression assays for luciferase and GUS were performed essentially as previously described (15). For each microprojectile preparation, the mass of DNA was adjusted to 10 g with p35SBAR (33) to equalize the amount of 35S promoter in each bombardment. One g of each of the regulators and 3 g of reporter plasmid were used in each bombardment. To normalize luciferase activity to GUS activity, 3 g of pUbiGUS was included in every bombardment. Each treatment was done in triplicate, and entire experiments were repeated at least twice. The assays for luciferase and GUS, and the normalization of the data were done as described (6). Data are expressed as the ratio of arbitrary luciferase light units to arbitrary GUS light units. -Fold activation is calculated as the ratio between the Luc/GUS units of the reporter construct with transcriptional activator divided by the Luc/GUS ratio without the activator. Luciferase for a typical bombardment with just pA1Luc (no activator) gives between 1,000 and 5,000 units, and GUS (from pUbiGUS) gives between 150,000 and 700,000 units.
Expression and Purification of Proteins Expressed in Bacteria-The plasmids for the expression of the MYB domains of P1, C1, and C1 SH in Escherichia coli were obtained by cloning into the pTYB2 vector (New England Biolabs). The MYB domains (residues 1-119) were synthesized by PCR, adding an NdeI restriction site at the N terminus and an XhoI restriction site at the C terminus, which adds a Cys residue at the N terminus of the protein splicing element intein from Saccharomyces cerevisiae. For expression, E. coli BL21 (DE3) PlyS cells bearing the corresponding plasmids were grown, induced, and purified essentially as described (35) with the following modifications. After induction of a 1-liter culture with 1 mM isopropyl-1-thio-␤-D-galactopyranoside, the cells were harvested by centrifugation and stored at Ϫ80°C until further use. The cells were resuspended in 40 ml of resuspension buffer (20 mM Tris, pH 8, 300 mM NaCl, 1 mM EDTA, 0.1% Triton X-100, and 1 mM phenylmethylsulfonyl fluoride) and passed twice through a French press. The cell lysate was centrifuged at 14,000 ϫ g for 20 min; the supernatant was filtered through two layers of Miracloth (Calbiochem). The chitin beads (New England Biolabs) were equilibrated in CB1 (20 mM Tris, pH 8, 300 mM NaCl, and 1 mM EDTA) and resuspended to give a 50% slurry. Ten ml of chitin bead slurry was added to the cell lysate supernatant and incubated for 2.5 h with rocking at 4°C. The beads were gently pelleted by centrifugation, and the pellet was then resuspended with 5 ml of CB2 (20 mM Tris, pH 8, 500 mM NaCl, 1 mM EDTA, and 0.1% Triton X-100), loaded onto a column, and washed twice with the same buffer. The column was washed three additional times with CB3 (20 mM Tris, pH 8, 1 M NaCl, 1 mM EDTA, and 0.1% Triton X-100). The column was incubated in CB3 plus 50 mM dithiothreitol overnight in order to allow the self-cleavage of the intein and therefore allow elution of the target protein. The elution was then dialyzed against A-0 buffer (10 mM Tris, pH 7.5, 50 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, 5% glycerol) at 4°C and stored at Ϫ80°C until further use. Each wash and elution fraction was collected and analyzed by SDS-PAGE, followed by staining with Coomassie Brillant Blue R-250. 2 E. L. Braun and E. Grotewold, unpublished observations.

Generation of C1 Monoclonal Antibodies and Western Analyses-
Monoclonal antibodies were generated using affinity-purified N 10 His-C1 protein by the Antibody Facility at Cold Spring Harbor Laboratory. For Western analyses, PJ69.4a yeast cultures were grown in 1.5 ml of SC-Leu-Trp medium at 30°C to an A 600 of 0.7. Cells were harvested by centrifugation at 14,000 rpm and washed once with 1.5 ml of distilled water. The pellet was then resuspended in 100 l of SDS loading buffer (0.06 M Tris-HCl, pH 6.8, 10% glycerol, 2% (w/v) SDS, 5% ␤-mercaptoethanol, 0.0025% (w/v) bromphenol blue) and heated to 95°C for 5 min. The suspension was then centrifuged at 14,000 rpm, and 15 l of extract was loaded and analyzed by 12% SDS-PAGE. The gels were transferred to polyvinylidene difluoride membrane by electrophoresis using the Bio-Rad Mini Trans-Blot transfer cell in 39 mM glycine, 48 mM Tris-HCl, and 20% methanol. The transfer was done at 4°C and 100 V for 1 h. Membranes were stained with Ponceau Red to verify transfer and then blocked in 1% nonfat dried milk dissolved in 1ϫ TBS (10 mM Tris-HCl, pH 8.0, 150 mM NaCl) with 0.05% Tween 20 overnight at 4°C. Following blocking, the membranes were incubated with the primary antibodies diluted in 1% nonfat dried milk dissolved in 1ϫ TBS with 0.05% Tween 20 for 2 h at room temperature. The C1 monoclonal antibody C1 40/1-249 was used at 1:150 dilution. For normalization, the blot was probed with an antibody that recognizes the yeast DED1p protein at a dilution of 1:4000. After washing the blots three times for 10 min in 1ϫ TBS, blots were incubated in anti-rabbit IgG, horseradish peroxidase secondary antibody (Amersham Biosciences) for DED1p and anti-mouse IgG, horseradish peroxidase (Amersham Biosciences) for C1 40/1-249 diluted 1:2000 in 1% nonfat dried milk dissolved in 1ϫ TBS with 0.05% Tween 20 for 2 h at room temperature. The blots were then rinsed again with 1ϫ TBS as before, and then they were visualized by using the Amersham ECL kit.
DNA binding assays were performed on ice for 30 min in a 25-l total volume in A-0 buffer with 0.8 g poly(dI/dC) and 1 mM dithiothreitol (unless otherwise indicated). After incubation on ice, ϳ10,000 cpm of end-labeled oligonucleotide probe was added and incubated on ice for an additional 30 min. Protein-DNA complexes were resolved on an 8% polyacrylamide gel (80:1, acrylamide/bis-acrylamide) with 0.25ϫ Tris borate-EDTA running buffer at 415 V for 55 min at 4°C. After electrophoresis, gels were dried onto Whatman paper and subjected to autoradiography at Ϫ70°C overnight, using x-ray film or a Kodak phosphor imager for 2 h and quantified using a Bio-Rad imaging system. The apparent dissociation constants were estimated by carrying out the DNA binding assays as described above with a fixed amount of protein (35 ng) and in the presence of varying amounts of cold doublestranded oligonucleotide (ranging from 0 to 5.240 nmol). After incubation on ice, ϳ10,000 cpm of end-labeled oligonucleotide probe was added and incubated on ice for an additional 30 min. Quantification of the free and bound APB1 oligonucleotide was estimated from the radioactivity present in each of the corresponding bands and calibrated against known standards. To compare our studies with others (4,35), data were assumed to fit to a linear relationship, although comparable RMS values are obtained using a hyperbolic fit (not shown). It should be noted that the APB1 probe is formed by two overlapping binding sites (6), and whereas only one protein can bind to APB1 at a time, thus ruling out a cooperative interaction (not shown), our studies cannot rule out the possibility that each binding site is recognized with different affinities by the different proteins.

RESULTS
The ARE Is Required for R-enhanced Activity-Previously, we showed that transferring the interaction with R from C1 to P1 resulted in the P1* protein with novel regulatory activities (15). In the absence of R, P1* activated the a1 promoter (pA1Luc; Fig. 1B) just like P1 ( Fig. 2A) (R-independent transcription; Table I) but not the bz1 promoter (Fig. 2B). In the presence of R, P1* activated bz1, similar to C1 (R-dependent transcription; Table I) (Fig. 2B), and induced anthocyanin accumulation (not shown). P1* displayed a 2-3-fold enhanced activity on the pA1Luc in the presence of R (Fig. 2A, compare P1* and P1* ϩ R). To ascertain whether this R-enhanced activity required cis-elements other than the MYB binding sites, we tested whether P1* displays R-enhanced activity on a promoter containing three copies of the ha PBS upstream of a minimal CaMV 35S promoter (p( ha PBS) 3 Luc; Fig. 1B). This syn- thetic promoter is activated by P1 and C1 ϩ R but not by C1 alone (Fig. 2C). On this promoter, however, the activity of P1* is not enhanced by R (Fig. 2C, compare P1* and P1* ϩ R). These results suggest that other cis-regulatory elements in the a1 promoter participate in the R-enhanced transcription of a1 by P1*.
The ARE element was identified as a conserved motif present in several flavonoid biosynthetic gene promoters (16). A 6-bp insertion in the a1 ARE generated by the excision of Spm from the a1-m2 allele dramatically reduced aleurone anthocyanin pigmentation controlled by C1 and R, without a significant effect on the pericarp phlobaphene pigmentation specified by P1 (18). To establish whether the ARE is responsible for the R-enhanced activity of P1*, we tested the promoter containing the 6-bp insertion (pA1 315L Luc; Fig. 1B) for activation by P1, P1*, P1* ϩ R, and C1 ϩ R (Fig. 2D). Supporting the significantly reduced anthocyanin pigmentation of the a1 315L allele in vivo (18), C1 ϩ R activate pA1 315L Luc significantly more weakly than pA1Luc (Fig.  2D), whereas P1 activates pA1Luc and pA1 315L Luc at similar levels (Fig. 2D, P1). P1* activates pA1Luc and pA1 315L Luc at similar levels, but in contrast to the activation of pA1Luc, the activation of pA1 315L Luc by P1* is not enhanced by R (Fig. 2D).
These results indicate that one function of R is to enhance the transcriptional activity of C1 and P1*. This R-enhanced activity requires the ARE cis-regulatory element.
Contrary to P1, C1 Does Not Activate Transcription in Yeast from the ha PBS-In addition to the ARE-mediated R-enhanced activity, R is essential for the activity of C1. To investigate what makes C1 activity R-dependent, we developed a yeast system in which two copies of the ha PBS were introduced upstream of a minimal yeast promoter driving the HIS3 selectable gene (Fig. 3). The p( ha PBS) 2ϫ HIS3 construct was integrated as a single copy in the yeast genome (linked to the URA3 marker) to generate the YEG102 yeast strain (see "Experimental Procedures"). The P1 and C1 proteins were expressed in YEG102 from the GAPDH promoter in a plasmid permitting selection in synthetic media lacking tryptophan (ϪTrp). When grown in ϪUra ϪLeu ϪTrp ϪHis media, a robust growth was observed for the yeast cells expressing P1 (Fig. 3A, P1). In contrast, cells expressing C1 verified by Western analysis using the C1 monoclonal antibody C1 40/1-249 (Fig. 3B) did not grow in the ϪUra ϪLeu ϪTrp ϪHis media (Fig. 3A, C1), even when in the presence of a version of the R paralog B, in which the bHLH domain was deleted (B ⌬HLH ) to permit the interaction with C1 (14). These results suggest that C1 is unable to activate transcription on its own, even in a heterologous, nonplant system. In contrast, if provided with a promoter containing high affinity binding sites, P1 activates transcription independently of other plant-specific factors.
C1 SH , a Mutant of C1 That Activates Transcription in Yeast-Previously, we showed that the second and third helices of R3 in C1 (Fig. 1A) participated in making the activity of C1 R-dependent (15). In addition, we established that the Gly 94 and Arg 95 residues in C1 (Fig. 1A) needed to be transferred to P1 to create the P1* protein (15). Thus, we targeted Gly 94 and Arg 95 as candidate residues that participate in making the C1 activity R-dependent. We generated the C1 SH protein containing the Gly 94 and Arg 95 amino acids replaced by S and H, respectively, as found in P1 (Fig. 1A). The MYB domains of P1, C1, and C1 SH were fused to the Gal4 activation domain (Gal4 AD ), and the resulting P1 MYB -Gal4 AD , C1 MYB -Gal4 AD , and C1 SHMYB -Gal4 AD proteins were expressed from the GAPDH promoter in the yeast strain YEG102. Cells expressing the fusion constructs were assayed for growth in media without histidine (ϪHis), which would be indicative of transcription of the p( ha PBS) 2ϫ HIS3 reporter. In these conditions, cells expressing P1 MYB -Gal4 AD and C1 SHMYB -Gal4 AD grew in the selective media (Fig. 4A, 1 and 5), whereas cells expressing C1 MYB -Gal4 AD did not (Fig. 4A, 3). The inability of C1 MYB -Gal4 AD to activate transcription of the p( ha PBS) 2ϫ HIS3 reporter is not complemented by the co-expression of either R or B ⌬HLH (not shown). The expression levels of C1 MYB -Gal4 AD and C1 SHMYB -Gal4 AD were comparable, as shown by Western analyses using the C1 monoclonal antibody C1 40/1-249 (Fig. 4C).  To investigate whether the G94S and R95H amino acid changes affected the interaction of C1 with R, we fused the N-terminal region of R (R 1-252 ), which is sufficient for the interaction with the MYB domain of C1 (15), to the Gal4 DNAbinding domain (Gal4 DBD ) to create R 1-252 -Gal4 DBD . We then co-expressed P1 MYB -Gal4 AD , C1 MYB -Gal4 AD , or C1 SHMYB -Gal4 AD with R 1-252 -Gal4 DBD in the PJ69.4a yeast strain and assayed the ability of cells co-expressing these proteins to grow in selective media (Fig. 4B). Because PJ69.4a has the HIS3and ADE2-selectable genes regulated by Gal4 binding sites (30), successful interaction would result in growth in ϪTrp ϪLeu ϪHis ϪAde media. C1 MYB -Gal4 AD and C1 SHMYB -Gal4 AD (Fig. 4B, 4 and 6), but not P1 MYB -Gal4 AD (Fig. 4B, 2), provide robust interactions with R 1-252 -Gal4 DBD . Thus, the G94S and R95H replacements in C1 do not interfere with the ability of C1 to interact with R, and they allow the corresponding MYB domain to activate transcription in yeast when fused to the Gal4 activation domain. FIG. 3. P1, but not C1, activates transcription in yeast. A, activation of transcription of the ( ha PBS) 2ϫ HIS3 construct (URA ϩ ) by P1 and C1 expressed from the constitutive GAPDH promoter (TRP ϩ ). Growth in media without histidine (ϪHis) is indicative of transcriptional activation of the His3 gene. B, Western blot analysis of yeast strains expressing empty vectors (Empty Vector) and C1 ϩ B ⌬HLH . The blot was probed with the C1 monoclonal antibody C1 40/ 1-249. As a loading control, an identical gel was stained with Coomassie Brilliant Blue (Coomassie).

FIG. 4. C1 SH activates transcription in yeast and interacts with R.
A, activation of transcription in yeast of the ( ha PBS) 2ϫ HIS3 construct (URA ϩ ) by fusions of the MYB domains of P1, C1, or C1 SH (C1 with G94S and R95H substitutions) to the Gal4 activation domain (P1 MYB -Gal4 AD , C1 MYB -Gal4 AD , and C1 SHMYB -Gal4 AD , respectively). Growth in a ϪUra ϪLeu ϪTrp ϪHis plate is indicative of activation. B, yeast two-hybrid experiment to verify interaction between P1 MYB -Gal4 AD , C1 MYB -Gal4 AD , or C1 SHMYB -Gal4 AD and the N-terminal region of R fused to the Gal4 DNA-binding domain (R 1-252 -Gal4 DBD ) in the yeast strain pJ69.4a (30) containing the HIS3 and ADE2 genes under the control of Gal4-binding sites. C, Western blot of yeast protein extracts from the YEG102 strain expressing C1 SHMYB -Gal4 AD (5, corresponding to 5 in Fig. 4A) or C1 MYB -Gal4 AD (3, corresponding to 3 in Fig. 4A). The blot was probed with the C1 monoclonal antibody C1 40/1-249 and with an antibody against the yeast cytoplasmic helicase DED1p protein (yDED1p) to normalize. C1 SH Binds DNA with High Affinity-The ability of C1 SHMYB -Gal4 AD to activate transcription from the ha PBS in yeast suggested that the G94S and R95H substitutions increased the normally low DNA-binding affinity of C1 (4). To determine whether C1 SH binds DNA better than C1, we expressed the corresponding MYB domains in E. coli utilizing the IMPACT™ (intein-mediated purification with an affinity chitin-binding tag) system to obtain 80 -90% pure P1 MYB , C1 MYB , and C1 SHMYB proteins (Fig. 5A) without any need for denaturation and only modified by the addition of a Cys residue at their C termini (see "Experimental Procedures"). Electrophoretic mobility shift assays showed that the C1 MYB protein (Fig. 5B, lane 4) binds the APB1 probe, representing the ha PBS present in the a1 promoter, much weaker than P1 MYB (Fig. 5B,  lane 2). Interestingly, C1 SHMYB bound DNA with a much higher affinity than C1 MYB (Fig. 5B, lane 3). Similar to P1 and C1, C1 SH did not bind to a mutant sequence in which the ha PBS have been mutated (APB5, not shown).
To quantitatively compare the DNA-binding affinity of C1 SHMYB with those of P1 and C1, we determined the apparent equilibrium dissociation constant (K d ) of the C1 SHMYB -APB1 complex. The results of Scatchard analyses (Fig. 5C) carried out in the conditions previously described for P1 and C1 (4,6,35) show that C1 SHMYB binds APB1 with an affinity comparable with that of P1 (Table II) or P1 MYB (Fig. 5C), significantly stronger than C1 (Table II) or C1 MYB (Fig. 5C).
High Affinity DNA Binding Is Not Sufficient for R-independent Transcription by C1-Having obtained a C1 mutant, C1 SH , capable of binding DNA with high affinity, we were poised to answer the important question of whether the need of C1 for R resides in the normal inability of C1 to bind DNA with high affinity. For this purpose, we assayed the activity of p35SC1 SH on the pA1Luc and pBz1Luc reporter constructs (Fig. 1B) in   5. C1 SH binds DNA with high affinity. A, SDS-PAGE analysis of the purified MYB domains of P1 MYB , C1 MYB , and C1 SHMYB used in the electrophoretic mobility shift assay experiments. B, comparison of the DNA binding activities of recombinant, purified C1 MYB , C1 SHMYB , and P1 MYB domains by electrophoretic mobility shift assay, using a fixed amount of protein and 32 P-labeled APB1 probe containing the ha PBS of the a1 promoter. C, Scatchard analysis of P1 MYB , C1 MYB , and C1 SHMYB binding to the a1 high affinity P1-binding sites ( ha PBS). Scatchard plots are shown with the x axis representing the bound/free APB1 ratio estimated from the amount of radioactivity in these two fractions and the y axis representing the amount of bound APB1, estimated from the amount of radioactivity in the bound fraction. For each experiment, the RMS and apparent equilibrium dissociation constants (K d ), deduced from the slopes assuming a linear relationship (see "Experimental Procedures"), are indicated. maize BMS cells by transient expression experiments. C1 SH does not efficiently activate transcription of either promoter on its own (Fig. 6, C1 SH ), yet in the presence of R, C1 SH efficiently activates pA1Luc and pBz1Luc (Fig. 6, C1 SH ϩ R). The inability of C1 SH to provide a strong activation of pA1Luc or pBz1Luc in the absence of R is not compensated by increasing the concentration of the protein in the transient expression experiments (not shown). These results show that a C1 mutant capable of binding DNA with high affinity remains dependent on R for function in maize cells. Thus, the function of R cannot be solely increasing the in vivo affinity of C1 for DNA.
The Fusion of C1 with the Gal4 DNA-binding Domain Makes Gal4-mediated Transcription Dependent on R-Previous studies have shown that the C-terminal activation domain of C1, when fused to the DNA-binding domain of the yeast Gal4 transcription factor (Gal4 DBD -C1 C-term ), was able to activate transcription independently of R (4,36,37), as shown in Fig.  7A. To investigate whether the full-length C1 protein remains R-dependent in the presence of another DNA-binding domain, we introduced the Gal4 DBD between the MYB domain of C1 and its C-terminal region to create C1 MYB -Gal4 DBD -C1 C-term . The activity of this protein was assayed on the pA1Luc reporter construct as well as on a reporter containing four tandem copies of the Gal4-binding sites upstream of a minimal CaMV 35S promoter driving luciferase (pGal4 BS Luc), in the presence or absence of R, in maize BMS cells by transient bombardment. The ability of C1 MYB -Gal4 DBD -C1 C-term to activate both pA1Luc and pGal4 BS Luc was dependent on R (Fig. 7A). These results suggest that the presence of the MYB domain of C1 or an interaction of the MYB domain with the C1 C-terminal region blocks the ability of the Gal4 DBD to direct transcription from the Gal4-binding sites in maize cells in the absence of R.
To determine why C1 MYB -Gal4 DBD -C1 C-term is unable to activate transcription using either reporter construct in the absence of R, we expressed C1 MYB -Gal4 DBD -C1 C-term in the PJ69.4a and YEG102 yeast strains. As shown in Fig. 7B (right panel), C1 MYB -Gal4 DBD -C1 C-term activates transcription from the Gal4-binding sites in PJ69.4a, manifested by a robust growth in ϪHis ϪAde media. In contrast, C1 MYB -Gal4 DBD -C1 C-term did not activate transcription from the ha PBS in the YEG102 strain (Fig. 7B, left  panel), similar to the results with C1 MYB -Gal4 AD (Fig. 4A). Together, these results demonstrate that the C1 MYB -Gal4 DBD -C1 C-term protein has the ability to bind to Gal4-binding sites and activate transcription in yeast but not in maize cells. Thus, it is likely that a plant-specific factor is involved in making the C1 MYB -Gal4 DBD -C1 C-term activity dependent on R.  Fig. 2. B, activation of transcription in the yeast strains YEG102 and PJ69.4a by the P1 and C1 MYB -Gal4 DBD -C1 C-term proteins expressed from the constitutive GAPDH promoter. Selection was done as described in Fig. 4A.

DISCUSSION
In this study, we investigated the mechanisms by which the bHLH factor R cooperates with the R2R3 MYB protein C1 to activate transcription of flavonoid biosynthetic genes. By comparing the regulatory activities of C1 and P1* on different promoters, we identified two mechanisms for R-dependent transcription. On synthetic promoters that contain only C1/P1 binding sites, R functions as an essential co-activator of C1. This function of R is unlikely to result solely in an increased C1 DNA-binding affinity, since a C1 mutant (C1 SH ) that binds DNA with comparable affinity as P1 remains R-dependent. A second function of R requires cis-regulatory elements in addition to the C1/P1 DNA-binding sites. We hypothesize that R functions in this mode by binding or recruiting additional factors that bind to the ARE element conserved in the promoters of several anthocyanin genes. Together, these findings suggest a model in which combinatorial interactions with co-activators enable MYB factors with very similar DNA binding preferences to discriminate between target genes in vivo.
Whereas P1 and P1* can activate transcription of a promoter containing the ha PBS independently of R (Fig. 2C), the activity of P1* is enhanced by R on the pA1Luc reporter but not on a promoter containing just the ha PBS (Fig. 2, A and C). The ARE, present in several maize flavonoid biosynthetic genes (16), provides a good candidate cis-regulatory element that participates in R-enhanced transcription. We show here that an insertion of 6 bp in the ARE, which dramatically affects a1 activation in vivo by C1 ϩ R but not by P1 (18) also abolishes R-enhanced transcription by P1* in transient expression experiments (Fig. 2D). Whereas this inhibitory effect of the insertion could be caused by altering the spacing between two cis-regulatory elements, three lines of evidence suggest that it is the ARE element itself and not the spacing between ARE-flanking elements that is important. First, a 4-bp insertion in the ARE does not result in a decrease of a1 activation in vivo (18) or in transient expression experiments (not shown); second, linkerscan analyses in which the sequence but not the distance of the ARE element was altered also showed an inhibitory effect on C1 ϩ R activity (16,17,19); and third, a conserved ARE element is necessary for the expression of a flavonoid biosynthetic gene in Gerbera hybrida (38). The participation of the ARE in R-enhanced transcription can be explained by proposing that R binds to the ARE or that the ARE is recognized by a DNA-binding R-interacting factor (RIF in Fig. 8). However, no factor(s) has yet been identified that may participate in the recruitment of R to DNA. The pac1 gene, discovered as a mutation that reduces R ϩ C1-specified anthocyanin pigmentation in maize aleurones (39,40), may encode a key player in this complex but is unlikely to recruit R to DNA because it contains no recognizable DNA-binding domains (39).
This R-mediated recruitment to the ARE would also participate in allowing P1 and C1 to activate different sets of target genes, thus contributing to their regulatory specificity, by providing R-dependent activation of promoters lacking DNA-binding sites for the R2R3 MYB factors. Our results suggest that, in the absence of P1-binding sites, R can recruit P1* to the bz1 promoter, explaining why P1* and C1, but not P1, activate this gene (Fig. 8, see activation of bz1 by P1*). This ability of R to recruit interacting R2R3 MYB factors to DNA was previously observed using a mutant a1 promoter lacking the ha PBS and la PBS. This mutant promoter can still be activated by C1 ϩ R, albeit at a reduced level (4). Similarly, a C1 protein defective in DNA binding also retains about 20% of the transcriptional activity of wild type C1 (4). In these cases, the role of the R2R3 MYB (P1* or C1) would be to provide the activation domains lacking in R.
Whereas the R-enhanced activity of P1* depends on the ARE, the R-dependent activity of C1 is observed on promoters containing just the ha PBS (Fig. 2C), suggesting a direct effect of R on C1. There are several possible mechanisms by which the activity of C1 could be dependent on R. R could increase the affinity of C1 for DNA, it could mediate C1 localization to the nucleus, relieve the effect of an inhibitory domain of C1 or an external C1 inhibitor, or it could stabilize the C1 protein.
Our results showing that the activity of C1 SH , a C1 mutant that binds DNA with an affinity comparable with that of P1 (Table  II), remains R-dependent strongly suggests that R is not required simply to increase the intrinsically low DNA-binding affinity of C1. The Gly 94 and Arg 95 residues are not required for the interaction with R (Fig. 7B), and given their positions in the MYB domain (41), they are unlikely to involve direct contacts with DNA. Thus, Gly 94 and Arg 95 could be involved in maintaining a specific MYB domain conformation that modulates DNA binding.
The C1 MYB -Gal4 DBD -C1 C-term chimeric protein is dependent on R for the activation of promoters containing C1-or Gal4binding sites in maize cells (Fig. 7A). However, C1 MYB -Gal4 DBD -C1 C-term activates transcription in yeast from the Gal4-binding sites in the absence of other plant factors. This result suggests that the inability of C1 MYB -Gal4 DBD -C1 C-term to activate transcription in maize cells is not due to the presence of an autoinhibitory domain, as found in ETS-1 and other transcription factors (42,43). Furthermore, the ability of C1 SH -Gal4 AD and C1 MYB -Gal4 DBD -C1 C-term to activate transcription in yeast also suggests that these proteins contain nuclear localization signals that function in yeast and therefore are likely to function in maize cells. Thus, it is unlikely that the R-dependent function of C1 resides solely in the R nuclear localization signals (44) driving the R-C1 complex to the nucleus. Together, these results suggest that there is a plant factor, absent in yeast cells, that either (i) inhibits C1 activity by masking the DNA-binding or activation domains, (ii) retains C1 in the cytoplasm, or (iii) destabilizes the C1 protein. The   FIG. 8. Mechanisms of cooperation between C1 or P1* and R. Model for the regulation of the a1 and bz1 genes by P1, C1 ϩ R, or P1* ϩ R. The cooperation between C1 and R has at least two components. The first component is dependent on C1 making contacts with the C1/P1 binding sites (PBS) and is probably involved in relieving C1 from the effect of an inhibitory plant cellular factor (shown as an orange circle). The second component of R is to make direct or indirect (through an as yet unidentified R-interacting factor (RIF)) contacts with the DNA through the ARE cis-regulatory elements. A similar mechanism of regulation by C1 and R on the a1 and bz1 promoters is suggested, the latter containing only C1-binding sites (CBS) to which P1 or P1* do not bind. P1* activates the bz1 promoter through its interaction with R and recruitment to DNA by the ARE. The thickness of the arrows is proportional to the levels of activation by the transcription factors.