A single residue substitution causes a switch from the dual DNA binding specificity of plant transcription factor MYB.Ph3 to the animal c-MYB specificity.

Transcription factor MYB.Ph3 from Petunia binds to two types of sequences, MBSI and MBSII, whereas murine c-MYB only binds to MBSI, and Am305 from Antirrhinum only binds to MBSII. DNA binding studies with hybrids of these proteins pointed to the N-terminal repeat (R2) as the most involved in determining binding to MBSI and/or MBSII, although some influence of the C-terminal repeat (R3) was also evident. Furthermore, a single residue substitution (Leu71 → Glu) in MYB.Ph3 changed its specificity to that of c-MYB, and c-MYB with the reciprocal substitution (Glu132 → Leu) essentially gained the MYB.Ph3 specificity. Molecular modeling and DNA binding studies with site-specific MYB.Ph3 mutants strongly supported the notion that the drastic changes in DNA binding specificity caused by the Leu → Glu substitution reflect the fact that certain residues influence this property both directly, through base contacts, and indirectly, through interactions with other base-contacting residues, and that a single residue may establish alternative base contacts in different targets. Additionally, differential effects of mutations at non-base-contacting residues in MYB.Ph3 and c-MYB were observed, reflecting the importance of protein context on DNA binding properties of MYB proteins.

One characteristic of most eukaryotic transcription factors is that they can be grouped into a small number of families, each including factors with sequence similarity over their DNAbinding domain (reviewed in Refs. [1][2][3]. In a given species, different members of the same family usually regulate unique, often partially overlapping, groups of target genes, at least in part due to distinct, although related, DNA binding specificities (4,5). The basis of distinctiveness/similarity in DNA binding specificity among members of each family of transcription factors is not yet fully understood, although some progress toward rationalizing this problem has already been made (Refs. 6 and 7 and references therein).
One of the families of transcription factors is that of MYB proteins, so named because the first member of the family to be discovered was the product of the avian myeloblastosis oncogene v-myb. Subsequently, members of this family, sharing the MYB DNA-binding domain, have been found in all eukaryotes investigated, from yeast to humans (reviewed in Refs. 8 -10).
Structurally, the best characterized member of the family is c-MYB, the cellular homologue of v-MYB, for which the solution structure of its DNA-binding domain has been solved, both in the free form and in complex with DNA (11)(12)(13)(14)(15). The c-MYB DNA-binding domain consists of three imperfect repeats (R1, R2, and R3), each of which folds into a variant of the homeodomain helix-turn-helix motif, similar to that of the prokaryotic LexA protein (7, 11-13, 15, 16). The third helix of the R2 repeat, the recognition helix, however, shows certain conformational flexibility in the free form, which is stabilized upon binding to DNA, and the same is true for the equivalent helix of B-MYB (14,15,17,18). MYB repeats are also characterized by the presence of three conserved tryptophan residues regularly spaced by 18 or 19 amino acids that play a relevant role in the folding of the hydrophobic core of the MYB domain (11,12,19,20). In their interaction with DNA, the recognition helices of both R2 and R3 lie on the major groove of the DNA and interact with each other, resulting in a cooperative binding to DNA sequences with the consensus pentanucleotide core CNGTT (12). The R1 repeat, which is missing in all plant MYB proteins (for examples, see Ref. 21) has no observable effect on DNA binding specificity, although it contributes to the stability of the protein⅐DNA complex (12,(22)(23)(24). The three key base contacts are established by residues Lys 128 (R2), Lys 182 (R3), and Asn 183 (R3), which are fully conserved in all known plant and animal MYB proteins (Refs. 12, 21, and 25 and references therein). However, whereas all known animal (R1, R2, and R3) MYB proteins recognize the same type of core sequence as c-MYB, in plants, there are at least some MYB proteins that show binding specificity differing from that of c-MYB (22,(25)(26)(27)(28)(29)(30)(31)(32)(33).
A striking case of this divergence in binding specificity is that of MYB.Ph3, which is a transcription factor predominantly found in epidermal cells of Petunia flowers. Like some other plant MYB proteins, such as the C1, Pl, and P proteins from maize, the Am305 protein from Antirrhinum, and others, MYB.Ph3 possibly regulates the flavonoid (phenylpropanoid) biosynthetic pathway (21,31,(33)(34)(35)(36)(37). MYB.Ph3 has been shown to bind to two types of site: MBSI 1 (A(a/D)(a/D)C(G/ C)GTTA, where a/D is A, G, or T, A being the preferred base), which conforms to the core consensus sequence CNGTT and is bound by c-MYB; and MBSII (AGTTAGTTA), which resembles the binding site of P and Am305 proteins and which is not bound by c-MYB (30,31,33). Our previous studies support the idea that binding of MYB.Ph3 to MBSI and MBSII does not involve alternative orientations of its two MYB repeats, despite of the resemblance of these sites to inverted and direct repeats of the GTTA motif, respectively (33,38).
Here we report on the analysis of the molecular determinants that enable MYB.Ph3 to recognize two different types of sequence. Remarkably, a single residue substitution in the R2 repeat of MYB.Ph3 (Leu 71 3 Glu) changes its specificity to that of c-MYB, and that the reciprocal substitution in c-MYB, Glu 132 3 Leu, essentially confers MYB.Ph3 specificity. We provide evidence that the ability of a single residue substitution to have such great effects on DNA binding specificity reflects the fact that certain residues influence this property directly, through base contacts, and also indirectly, through interactions with other base-contacting residues, and that some residues can establish alternative base contacts in different targets. In addition, we show that substitutions in (presumably) non-basecontacting residues can also affect the DNA binding properties of MYB proteins, and that their effect may be different in c-MYB and MYB.Ph3, thereby underlining the importance of protein context in determining DNA binding.

Plasmid Constructs and in Vitro Synthesis of Proteins-Constructs
coding for Petunia MYB.Ph3⌬C1, murine c-MYB⌬R1C1, and Antirrhinum Am305⌬C1 have been described previously (31,33). Constructs coding for the mutant derivatives of these proteins, used in this study, were obtained by PCR-mediated, site-directed mutagenesis of the corresponding cDNAs as described by Cormack (39). To prepare the constructs encoding MYB chimeric proteins, the cDNA fragments corresponding to the parts of the MYB proteins present in the chimeras were obtained by PCR amplification with one phosphorylated oligonucleotide, that corresponding to the internal part of the chimera. After ligation of the two fragments present in each chimera, a second PCR was performed with the oligonucleotides corresponding to the 5Ј and 3Ј ends of the chimeric cDNA. All cDNA fragments coding for mutant or chimeric proteins were cloned into the XbaI-BamHI sites (or XbaI-PstI for P2A3 and M2A3 chimeras) of the pBluescript vector. All PCR fragments used in the constructs were confirmed by sequencing.
RNAs, obtained by in vitro transcription of the corresponding constructs using T7 or T3 polymerase, were used for in vitro translation in the flexi-rabbit reticulocyte system (Promega) supplemented with magnesium acetate and potassium chloride to final concentrations of 2.05 and 75 mM, respectively, in the presence of [ 35 S]methionine, following the manufacturer's instructions. After in vitro translation, SDS-PAGE analysis of the reticulocyte extracts was performed to allow estimation of the amount of each translated protein by measurement of 35 S cpm in the corresponding protein band and correction for methionine content.
DNA Binding Reactions and EMSA-PCR labeling of MBSI, II, IG, and IIG oligonucleotides, DNA binding reactions, and electrophoretic mobility shift assays (EMSAs) were performed as described in Solano et al. (33). Each binding reaction (15 l) contained 4 ng of labeled DNA, 400 ng of poly(dI⅐dC), 150 ng of denatured salmon sperm DNA, and rabbit reticulocyte lysate consisting of a measured amount of the in vitro translated protein, supplemented with lysate incubated in the absence of external RNA to give a final volume of 2 l, so that all reactions had equimolar amounts of protein.
Molecular Modeling-The structures used for the analysis were the average NMR structure of c-MYB bound to DNA (GTCAGTTA), as deposited in Protein Data Bank (40) under code 1MSE by Ogata et al. (12), and the best 25 NMR solutions (Protein Data Bank code 1MSF). Modeling of MYB.Ph3 complexed with DNA (MSBI or MSBII) was carried out with the WHATIF package (41). The quality of the resulting structures was assessed by different standard structures based on normality of molecular contacts (42) and deviation from normal exposed hydrophobic surfaces (43). The analysis of alternative conformations for different residues corresponds to the WHATIF secondary structurespecific rotamer data base (version Feb. 1996).

Role of MYB Repeats in Sequence
Recognition-To investigate the molecular determinants that allow MYB.Ph3 protein to recognize two different types of sequence, we first analyzed the role of R2 and R3 repeats on DNA binding. For this purpose, we took advantage of the differential affinity of murine c-MYB and Antirrhinum Am305 proteins for both types of MYB.Ph3 consensus binding sites (28,31,33). As shown by EMSAs (electrophoretic mobility shift assays, Fig. 1), derivative MYB.Ph3⌬C1 binds both types of consensus sequence (MBSI and MBSII) with the same affinity, whereas derivative c-MYB⌬C2R1 only binds to MBSI but not to MBSII, and derivative Am305⌬C1 shows the opposite behavior. Protein Am305 also differs from MYB.Ph3 in that it prefers a variant of MBSII with a G in position ϩ2 (MBSIIG; Ref. 31; Fig. 1), whereas an additional difference between c-MYB and MYB.Ph3 is that a change of T for G at position ϩ2 in MBSI (MBSIG) still allows a certain binding by c-MYB and not by MYB.Ph3 (24).
We constructed hybrid proteins that combined R2 and R3 MYB repeats: P2 and P3 from MYB.Ph3; A2 and A3 from Am305; and M2 and M3 from c-MYB. These chimeric proteins, like their progenitors ( Fig. 1), also contained amino acid sequences beyond the strict R2 and R3 repeats, originating from the 5Ј or the 3Ј coding parts of the cDNAs, except for M2, which only included an additional methionine from an engineered initiation codon (33). However, previous work with c-MYB (12,22,28), as well as the studies with site-directed mutants described in the next sections, showed that the effect of these additional sequences on DNA binding specificity was negligible. As shown in Fig. 2, all chimeric proteins, except P2A3, recognized at least one of the four sequences, albeit generally with lower affinity than their parental proteins, particularly M2A3. The type of sequence recognized (i.e. I or II) was mainly dependent on the type of R2 repeat in the chimera. Thus, proteins with the R2 repeat of MYB.Ph3 (P2) were able to bind type I and type II sequences, as can MYB.Ph3, whereas proteins with the R2 repeat of c-MYB (M2) or Am305 (A2) showed a preference for type I or type II, as found for A2A3 or M2M3, respectively. Because type I and II sequences differ at their 5Ј halves, R2 should be mainly implicated in the interaction with the 5Ј half of the sites. On the other hand, M2P3 and M2A3, which share the same R2 repeat, showed differential affinity for I and IG sequences, respectively, thus implicating the R3 repeat in the interaction with the 3Ј part of the targets. The same conclusion could be drawn from a comparison of A2P3 with A2A3.
However, there must be some functional interdependence between repeats R2 and R3 in their interaction with the 5Ј and 3Ј halves of the sequence, respectively. For instance, the A2P3 protein bound to MBSIIG with higher affinity than P2P3 or M2P3, indicating some role of R2 (A2) in the interaction with the 3Ј half of the sequence. The same conclusion could be drawn by comparing A2M3 with P2M3 and M2M3, or P2P3 with M2P3. On the other hand, the higher binding affinity of A2P3 than A2A3 to MBSI can be taken as an example of the influence of the R3 repeat on the interaction with the 5Ј half of the sequence.
Residue Asn 125 in the R3 Repeat of MYB.Ph3 Determines Preferential Recognition of T at Position ϩ2-An analysis of R2-and R3-specific residues responsible for the differential binding specificity of MYB.Ph3 versus c-MYB and/or Am305 was then undertaken. The amino acid residue of the R3 repeat determining the preference for T rather than G at position ϩ2 was investigated. A comparison of the amino acid sequence of R3-recognition helices from MYB.Ph3 and Am305, which respectively prefer T and G at position ϩ2, revealed several differences (Fig. 3). Among these, residue Asn 125 of MYB.Ph3, which is substituted by an Arg residue in Am305, was selected for site-directed mutagenesis, based on previous evidence that implicated the equivalent residue from c-MYB in the interaction with ϩ2T (see Fig. 7; position Ϫ5T in Ref. 12). The MYB.Ph3 (Asn 125 3 Arg) mutant now preferred the MBSG to the MBS sequences (Fig. 3), thereby revealing the influence of residue Asn 125 from MYB.Ph3 (or Arg from Am305) in the specificity of ϩ2 contacts. Mutations of residue Asn 125 to Ser, Ile, or His decreased overall affinity without affecting specificity, indicating that this residue is not the only determinant of position ϩ2, in agreement with the studies using chimeric proteins described in the previous section (see also Fig. 7).
Major Role of Residue Leu 71 from the R2 Repeat in Dual Recognition by MYB.Ph3-Recognition determinants within the R2 repeat of MBSI and/or MBSII were investigated using chimeric proteins obtained by full or partial replacement of the recognition helix of the A2 repeat from protein A2M3 by its M2 counterpart, because repeats A2 and M2 determined the most extreme differences in binding to types I and II sequences (Fig.  2). As shown in Fig. 4, full substitution of the recognition helix (A2M3-3 protein) conferred the c-MYB specificity, whereas the partial substitution in A2M3-2 did not greatly alter the A2M3 specificity. This suggested that the N-terminal half of the R2 recognition helix was the major determinant of binding to MBSI or MBSII in the Am305/c-MYB context.
To confirm the importance of amino acid residues of the N-terminal part of the R2 recognition helix in the MYB.Ph3 context, we performed EMSA with mutant derivatives of the MYB.Ph3 and c-MYB proteins affecting the three nonconserved positions in these two proteins (Fig. 5). Each replacement had a different effect on the DNA binding properties of MYB.Ph3 and of c-MYB. Remarkably, a single residue substitution in MYB.Ph3 (Leu 71 3 Glu) conferred c-MYB specificity (with respect to the targets used), and the reciprocal change in   (dimers and others). b, EMSA using the four types of binding sites (Fig. 1) and the proteins shown schematically on the left. P2, P3, A2, A3 and M2, and M3 represent the R2 and R3 repeats of the MYB domain of the MYB.Ph3, Am305, and c-MYB proteins, respectively. The autoradiographies corresponding to P2M3 and M2A3 were 3-fold overexposed. c-MYB (Glu 132 3 Leu) showed the reverse behavior, because it rendered a c-MYB derivative able to bind to MBSII sequences, although showing a slight preference for MBSI sequences. An additional change in c-MYB (Gln-Glu 3 Ser-Leu) resulted in a preference of this mutant protein for MBSII. It is noteworthy that the MYB.Ph3 (Cys-Ser 3 Ile-Gln) also showed preference for MBSI sequences. In fact, its relative affinity to MBSI compared to MBSII was higher than that of the corresponding mutant of c-MYB, c-MYB (Glu 3 Leu), indicating that the effect of this residue is dependent on protein context. A similar conclusion can be drawn by comparing the DNA binding properties of MYB.Ph3 (Ser-Leu 3 Gln-Glu) and MYB.Ph3 (Cys-Ser-Leu 3 Ile-Gln-Glu) with those of c-MYB, or of MYB.Ph3 (Cys 3 Ile) with those of c-MYB (Gln-Glu 3 Ser-Leu).
Mutational Analysis of the Conserved Lys Residue in the R2 Repeat-A major difference between the two types of MYB.Ph3 binding site is the greater sequence constraints on MBSII compared to MBSI at positions Ϫ4 and Ϫ3 imposed by the presence of T instead of C at position Ϫ2 (33). Thus, for instance, exchanging T in MBSII with C has only a moderate effect on its binding by MYB.Ph3, whereas the reciprocal change in MBSI (i.e. C 3 T), results in a great impairment of binding ( Ref. 33; Fig. 6). Molecular modeling predicted that the highly conserved Lys residue from MYB.Ph3 (Lys 67 ) should interact with Ϫ2ЈG in MBSI, like the equivalent Lys of c-MYB, but with Ϫ4G (and perhaps with Ϫ3T) in the MBSII sequence (see "Discussion" and Fig. 7); thus, the Lys 67 residue may be responsible for these sequence constraints. To test this prediction, the effect of mutating the Lys 67 residue (to Ala or Ser) on DNA binding specificity was examined. As shown in Fig. 6, the two mutants had reduced DNA binding affinity but bound better to MBSI and MBSII than to MBSIG, MBSIA (AAAAG-GTTA), and MBSIIG, indicating that the mutations did not affect MYB.Ph3 specificity indiscriminately. However, in sharp contrast to wild-type MYB.Ph3, these mutant proteins bound similarly to MBSI and MBSIT (AAATGGTTA), in agreement with the prediction that the Lys 67 residue is responsible for sequence constraints at positions Ϫ3 and Ϫ4 in MBSII. In fact, when Lys 67 does not impose constraints on positions Ϫ3 and Ϫ4 (e.g. when MYB.Ph3 binds to MBSI, or when MYB.Ph3 (Lys 67 3 Ala/Ser) binds to MBSI or MBSII), it appears that the preferred base at these positions is an A, as in MBSI. It is also noteworthy that the Lys 3 Ala/Ser mutations showed higher The residues that were substituted in any of the constructs are highlighted (full boxes). b, SDS-PAGE analysis of the mutants and original proteins used. c, EMSA using the four types of binding sites (Fig. 1); the proteins are shown schematically on the left. Here, in each protein, the residue present at each of three positions where the mutants may differ from their wildtype protein is indicated: empty letter indicates wild type residue; boldface, mutant residue. overall affinity to MBSIT than MYB.Ph3. This could indicate that when the (large and charged) Lys residue of MYB.Ph3 does not establish a base-specific contact, it may perturb base contacts by other residues. DISCUSSION DNA binding studies with hybrid MYB proteins and with site-directed MYB mutants reported here indicate that the R3 repeat is the most responsible for differential binding to MBSI and MBSII compared to MBSIG and MBSIIG, whereas the R2 repeat was primarily involved in determining the MBSI/MBSII specificity (Figs. [2][3][4][5]. Additionally, these experiments indicated that both repeats influence each other's primary effect; for instance, the higher relative affinity for MBSIIG (versus MBSII) of the A2P3 chimera with respect to that of P2P3 reveals a role of R2 (A2) in the interaction with the 3Ј half of the sequence (Fig. 2).
The proposed primary roles of the R2 and R3 repeats and their functional interdependence are in good agreement with the available structural information on c-MYB. Indeed, the NMR solution structure of the complex between the c-MYB R2R3 domain and DNA shows that its repeats physically interact and bind to DNA in a partially overlapping way (12). In this context, it is not surprising that the R2 and R3 amino acid sequences of both Am305 and MYB.Ph3 fit well in the structure of the R2R3 domain from c-MYB (data not shown). This structural similarity is also manifest in the effects of particular residue substitutions, like that of Asn 125 3 Arg in MYB.Ph3, which resulted in a specificity change at position ϩ2 (Fig. 3), the position interacting with the equivalent residue from c-MYB, Asn 186 ( Ref. 12; Fig. 7).
The physical interactions between repeats of c-MYB bound to DNA result in (intramolecular) cooperativity (12). In this scenario, it is conceivable that for a MYB domain to be functional, its R2 and R3 repeats must adapt to each other. Our results with R2/R3 MYB chimeras are in line with this suggestion because most of them displayed reduced DNA binding affinity with respect to their progenitors (most notably P2A3). Thus, it appears that co-evolution may have placed constraints on the compatibility between repeats from different MYB proteins.
Leu 71 , a Key Residue for MYB.Ph3 Dual DNA Binding Specificity-In this study, we have found that a single residue substitution within the recognition helix of the R2 repeat of MYB.Ph3, Leu 71 3 Glu, switches the dual DNA binding specificity of MYB.Ph3 to the c-MYB specificity, and that the reciprocal (Glu 132 3 Leu) change in c-MYB essentially confers the MYB.Ph3 specificity. As discussed below, such drastic effects on specificity caused by the Leu 3 Glu substitution most likely indicate that there are key residues that influence binding specificity not only directly, through base contacts, but remarkably also indirectly, through interactions with other base-contacting residues, and that a single residue can establish alternative base contacts in different targets.
In the NMR average structure of the complex of the c-MYB minimal DNA-binding domain (R2R3) with its target DNA (GTCAGTTA; Ref. 12), Glu 132 interacts weakly with DNA (positions Ϫ2C and ϩ1ЈC in our nomenclature, see Figs. 1 and 7; Ref. 12). Additionally, Glu 132 establishes an electrostatic interaction with Lys 128 , a key base-contacting residue that interacts with Ϫ2ЈG. Using molecular modeling (see "Materials and Methods"), we predicted that the change of Glu for Leu would have two consequences: (i) Leu would be in a hydrophobic cavity adequate to allow interaction with C or T at position Ϫ2 (see also Ref. 6), the bases respectively present in MBSI and MBSII, whereas Glu would not specifically interact with T, and (ii) the electrostatic interaction between Lys 128 of c-MYB (Lys 67 in MYB.Ph3) and Glu does not occur, thereby facilitating the interaction of this Lys residue with an alternative position, Ϫ4 and to some extent with Ϫ3. This is particularly important in the binding to MBSII, which has A rather than G at position Ϫ2Ј. The possibility that a single residue can establish contacts at two alternative positions has been documented/invoked in several instances (44 -46).
Further evidence that Lys 128 (c-MYB coordinates) can establish alternative base contacts was obtained from the analysis of the 25 available NMR solution structures of the c-MYB(R2R3)⅐ DNA complex. Indeed, in two solutions Lys 128 was found to interact directly with Ϫ4G, whereas Glu 132 was located far away from the average structure and did not interact directly with the DNA (see alternate positions of Lys 128 and Glu 132 in Fig. 7). Hence, the preferential interaction of c-MYB residue Lys 128 with Ϫ2G could be due to the electrostatic attraction of Lys 128 toward Glu 132 (as close as 2.64 Å in some of the NMR These interpretations are also supported by our results with the Lys 67 3 Ala/Ser substitutions in MYB.Ph3, which broadened specificity at positions Ϫ4 and Ϫ3 (Fig. 6), and with missing nucleoside assays, which have shown that nucleoside at position Ϫ2Ј (A) in MBSII is fully dispensable in binding by MYB.Ph3 (33).The requirement for T at position Ϫ3 in MBSII (33) is not well understood, although it could reflect that the methyl group of T pushes Lys 67 to Ϫ4G, or alternatively that Lys 67 interacts with GT rather than only with G (see Ref. 6).
MYB DNA Binding Specificity Is Also Influenced by Protein Context-Base-contacting residues play a critical and direct role in determining the specificity of DNA-binding proteins. This is evident from the fact that specificity can be explained to a significant extent using simple rules: the base-contacting specificity of different residues and the (usually) fixed position of base-contacting residues within the DNA-binding domain in each protein family (6,7,47). For instance, in MYB.Ph3, the effect of the Asn 125 3 Arg substitution does conform to these rules. However, there is strong evidence that binding specificity in MYB proteins can also be indirectly modulated by nonbase and base-contacting residues. Thus, Am305 shares all putative base-contacting residues with MYB.Ph3 (Asn 125 3 Arg), but only the latter strongly binds to MBSI (Figs. 3, 4, and  7). Likewise, the maize P protein shares all the putative rec-ognition residues with MYB.Ph3 and/or c-MYB, but it binds to a different site (GGT(T/A)GGT(A/G); Refs. 30 and 35). Moreover, in addition to the indirect effects of the Leu/Glu substitutions discussed above, we have also shown that several substitutions in presumably non-base-contacting residues alter specificity and/or affinity (see Fig. 5), and in some instances (e.g. the Gln/Ser substitution) the degree of the effect on specificity was different in the MYB.Ph3 and in the c-MYB contexts.
One possible explanation for these indirect effects on binding specificity could be that residue substitutions affect conformational properties of the protein, thereby influencing the strength of possible contacts by recognition residues or imposing constraints in the structural properties of the DNA (3), because some MYB proteins induce bending/distortions upon binding to DNA (38,48). In this regard, note the presumed structural flexibility of the R2-recognition helix, a property expected to be very sensitive to mutations (14,15,17,18). Some of the specificity effects of substitutions of non-base-contacting residues could simply be mediated by side-chain interactions with base-contacting residues, such as that of Glu 132 with Lys 128 in c-MYB. For instance, residue Gln 129 (c-MYB coordinates) interacts in the average structure with the phosphate backbone of the DNA, but in some of the solutions, it interacts with Glu 132 or with Lys 128 . In solutions where Gln 129 interacts with Glu 132 or with Lys 128 , Lys 128 interacts very closely with Ϫ2ЈG (see Fig. 7). Hence, it seems that Gln 129 contributes to maintain Lys 128 in the conformation that favors the interaction with Ϫ2ЈG, and such effect could be accentuated when Glu 132 is missing (c-Myb (Glu 132 3 Leu) and MYB.Ph3 (Cys-Ser 3 Ile-Gln); Fig. 5).
The differential effects of some residue substitutions, such as Gln/Ser, in MYB.Ph3 and c-MYB further underline the importance of protein context in MYB DNA binding specificity, possibly involving interresidue interactions. Indeed, we noticed that, in c-MYB, Glu 132 and Gln 129 are part of a network with several residues (Asn 179 , Lys 182 , Asp 178 , Arg 131 , and His 135 ), phosphates, and bases. In MYB.Ph3, one of these residues, His 135 , is substituted by Ala (Fig. 7), and consequently, the effect of substitutions involving residues at positions 129 and 132 (c-MYB coordinates) cannot be the same in the two protein contexts.
The notion that DNA binding specificity is best viewed as the result of a network of interactions of residue side chains with the DNA backbone and bases, as well as with other residues, rather than the simple and independent contribution of basecontacting residues has also been highlighted for other protein families, such as bZIP, ribbon-helix-helix, homeodomain, prokaryotic helix-turn-helix, and others (for examples, see Refs. 44 -52). Obviously, the importance of interresidue interactions will be higher in proteins with physically interacting DNAbinding subdomains, such as the MYB and cut-homeodomain proteins (12,53).
The numbers of myb genes in plant species are large, in contrast to those in other types of eukaryotes; for instance, there are at least 20 -30 myb genes in Petunia (21), and Arabidopsis contains over 100 of these genes. 2 However, 6 of 8 putative recognition residues are fully conserved among all plant MYB proteins with known sequence (30 in data bases; data not shown), and the remaining 2 residues are conserved in at least 80% of the proteins. Therefore, mutations in non-basecontacting residues must have greatly contributed to the generation of functional diversity among the members of the plant MYB family.