Oncogenic point mutations induce altered conformation, redox sensitivity, and DNA binding in the minimal DNA binding domain of avian myeloblastosis virus v-Myb.

c-Myb is the founder member of a class of transcription factors with tryptophan-rich repeats responsible for DNA binding. Activated oncogenic forms of Myb are encoded by the avian retroviruses, avian myeloblastosis virus (AMV) and E26. AMV v-Myb encodes a truncated protein with 11 point mutations relative to c-Myb. The mutations in the DNA binding domain (DBD) were reported to impose distinct phenotypes of differentiation on transformed myeloid cells (Introna, M., Golay, J., Frampton, J., Nakano, T., Ness, S. A., and Graf, T. (1990) Cell 63, 1287-1297). The molecular mechanism operating has remained elusive since no change in sequence specificity has been found. We introduced AMV-specific point mutations in the minimal DBD of chicken c-Myb and studied their effect on structure and function of the purified protein. Fluorescence emission spectra and fluorescence quenching experiments showed that the AMV-specific point mutations had a significant effect on the conformation of the DBD, giving rise to a more compact structure, a change that was accompanied by a reduced sensitivity toward cysteine-specific alkylation and oxidation. The DNA binding properties were also altered by the AMV-specific point mutations, leading to protein-DNA complexes with highly reduced stability. This reduction in stability was, however, more severe with certain subtypes of binding sequences than with others. This differential behavior was also observed in an in vivo model system where DBD-VP16 fusions were coexpressed with various reporters. These findings imply that different subsets of Myb-responsive promoters may react differentially toward the AMV-specific mutations, a phenomenon that could contribute to the altered patterns of gene expression induced by the AMV v-Myb relative to wild type c-Myb.

expressed in immature cells of all lineages, and the expression is strongly down-regulated during terminal differentiation. Aberrant overexpression of c-Myb inhibits differentiation of hematopoietic precursor cells (reviewed in Ref. 5), while antisense oligonucleotides directed at c-myb inhibit their proliferation (6 -8). Mice homozygous for a c-myb disruption die from multiple hematopoietic defects during embryonic development (9), and transgenic mice with T cell-specific expression of a dominant interfering allele of Myb display partially blocked thymopoiesis and diminished proliferation of mature T cells (10). The molecular mechanisms producing these phenotypes are poorly understood, in particular when it comes to Myb-regulated target genes with a role in proliferation and differentiation. Candidate target genes have recently been reviewed (2).
The c-myb gene encodes a 75-kDa sequence-specific DNAbinding transcription factor with at least three functional domains (2). The DNA binding domain (DBD) 1 located near the amino terminus is a highly conserved tryptophan-rich region composed of three imperfect repeats (R 1 , R 2 , and R 3 ), each related to the helix-turn-helix motif (11)(12)(13). Each repeat appears to have a distinct function. R 3 is a fully folded domain mainly responsible for the sequence-specific recognition of the AAC core in the binding site (13,14). R 2 is more flexible and seems to undergo a conformational change upon binding to DNA, possibly to allow the protein to adapt to a range of flanking sequences (15)(16)(17) (see also Carr et al. (18)). R 2 also contains a highly oxidizable cysteine implicated in redox control (16). An NMR-derived structure of a mouse R 2 R 3 -DNA complex was recently reported (14). The role attributed to R 1 has been a stabilization of the complex through electrostatic interactions (19 -21). An acidic transactivation domain is found centrally located in the protein, and a large carboxyl-terminal region appears to have a negative effect on Myb's transactivation and DNA binding functions. A recent report suggested that several subdomains may cooperate to form a functional transactivation domain, implying that the precise borders of this domain remain to be elucidated (22).
Distinct mechanisms of oncogenic activation seem to operate in the two isolated avian v-myb-containing retroviruses E26 and avian myeloblastosis virus (AMV) (reviewed in Refs. 1, 4, and 5). Both contain truncated v-myb oncogenes. E26 encodes an amino-and carboxyl-terminally truncated Myb protein that is fused to the v-Ets oncoprotein, while AMV encodes a truncated Myb protein that displays a number of specific amino acid substitutions relative to c-Myb. E26 is able to transform multipotent hematopoietic cells and causes erythroblastosis and a low level of concomitant myeloblastosis in chickens, while AMV transforms myelomonocytic cells and causes acute monoblastic leukemia.
The specific point mutations found in AMV v-Myb contribute to the establishment of the specific differentiation phenotype of AMV-transformed cells, probably by affecting sets of differentiation-specific genes. From experiments with hybrid viruses, the critical difference between the E26 and AMV viruses that causes the two distinct transformation phenotypes were mapped to the R 2 repeat of the DNA binding domain (23). The AMV-specific point mutations also contribute to the oncogenic potential of the virus. Dini et al. (24) found that AMV v-myb caused a 10-fold higher frequency of leukemic transformation of primary hematopoietic cells in culture than did a similarly truncated c-myb without the mutations. Mutations in both the DNA binding and the transactivation domains seemed to cooperate in attaining a high transformation frequency (24).
Precisely how the AMV-specific point mutations causes such prominent phenotypic effects is poorly understood. Since the DBD of Myb is involved, the most obvious hypothesis would be to expect an altered DNA binding specificity leading to activation of a distinct set of genes. However, binding-site amplification experiments in vitro have not revealed significant differences in sequence recognition between c-Myb and v-Myb (25). An alternative hypothesis suggested by some investigators has been that the mutations modify interactions between Myb and other regulatory proteins (see Thompson and Ramsay (2)). Indirect support for this hypothesis was the finding that all three amino acid replacements are exposed on the surface of the DNA binding domain as judged from its structure (14). However, so far no direct evidence for this assumption has been reported, and no interacting proteins have been identified. In the present work we have exploited the possibility of significant effects intrinsic to R 2 . We show that the AMV-specific mutations alter the conformation of the flexible R 2 -domain and that this change is accompanied by altered DNA binding properties, altered redox sensitivity, and altered transactivation in a model yeast effector-reporter system. We also demonstrate that the quantitative alterations in DNA binding is more severe with some recognition sequences than with others, suggesting that the altered DNA binding properties act differentially and could thus affect subsets of genes differently.

MATERIALS AND METHODS
Expression and Purification of Myb Proteins-The minimal DNA binding domain of the chicken c-Myb protein, R 2 R 3 , and two mutant derivatives were expressed in Escherichia coli using the T7 system (26), and proteins were purified as described previously (12).
Mutagenesis-Site-directed in vitro mutagenesis was performed as described elsewhere (12,27). Plasmids from positive clones, identified by DNA sequencing, were used directly to transform the E. coli BL21(DE3)LysS strain. The three AMV-specific mutations introduced in R 2 were I91N, L106H, and V117D as illustrated in Fig. 1A. The R 2 R 3 [AMV HD ] protein harbored the L106H and V117D mutations, while the R 2 R 3 [AMV NHD ] protein harbored all three alterations. Mutations were verified by amino acid sequencing of the purified proteins through 32 steps. Automated Edman degradation was performed on a 477A protein sequencer with an on-line 120A phenylthiohydantoin amino acid analyzer from Applied Biosystems (Foster City, CA).
Electrophoretic Mobility Shift Assay-DNA binding was monitored by the electrophoretic mobility shift assay (28), with the modifications described in Gabrielsen et al. (12). The basic duplex oligonucleotide probe used (Scheme 1) was based on the Myb recognition element (MRE) in the upstream region (site A) of the mim-1 gene (29). Derived variants with different configurations of Gs in positions 5 and 6 of the MRE consensus sequence were also used (Scheme 2) (30). Labeled duplex oligonucleotides were obtained by end-labeling of a small com-plementary oligonucleotide "MRE primer" (5Ј-GGCGCTAAA-3Ј) using polynucleotide kinase and [␥-32 P]ATP. After annealing and fill-in, all probes got identical specific activities and therefore directly comparable intensities in electrophoretic mobility shift assay. Labeled duplex oligos were purified by polyacrylamide gel electrophoresis.
N-Ethylmaleimide (NEM) and azodicarboxylic acid bis(dimethylamide) (both from Sigma, the latter abbreviated as "diamide") were made fresh from powder dissolved in water. H 2 O 2 was freshly diluted in water from a stock solution (30%, Aldrich). NEM alkylation reactions were stopped by addition of dithiothreitol to a final concentration of 10 mM. H 2 O 2 oxidations were terminated by addition of 0.5 l of 1 mg/ml catalase.
Protease Treatment of Myb R 2 R 3 -Chymotrypsin solutions were made fresh from powder dissolved in TE-buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0). Purified Myb proteins were diluted in TGE␤ 100 -buffer (20 mM Tris-HCl, 10% glycerol, 1 mM EDTA, 10 mM ␤-mercaptoethanol, 100 mM NaCl, pH 8.0) and incubated with protease for 15 min at 37°C. The reaction was stopped by addition of SDS loading buffer. The samples were heated at 95°C for 2 min and loaded onto a 10 -20% gradient polyacrylamide gel containing sodium dodecyl sulfate according to Laemmli (31).
Fluorescence Spectroscopy-Fluorescence experiments were performed essentially as described previously (16). A Perkin-Elmer LS-50B luminescence spectrometer and a Perkin-Elmer luminescence spectroscopy cell of 120 l were used for the fluorescence experiments. The exitation wavelength was 295 nm, and exitation slit 15 nm. Emission spectra were recorded between 310 and 400 nm with an emission slit of 5 nm, and a scanning speed of 500 nm/min. Each recording was made as an average of three accumulated scans. Samples were prepared from purified proteins in TE␤ buffer (10 mM Tris-HCl, 1 mM EDTA, 10 mM ␤-mercaptoethanol, pH 8.0) at a concentration of 2 M. Fluorescence values were corrected with respect to background fluorescence and the inner filter effect of acrylamide. For denaturation, proteins were incubated with 6 M guanidinium chloride for 15 min at room temperature before recording.
Transactivation Studies in Yeast with Myb-VP16 Fusion Proteins-The reporter plasmids used are described elsewhere (30). The c-Myb-R 2 R 3 [wt] and R 2 R 3 [AMV NHD ] effector plasmids were derived from the centromeric plasmid pDBD11 (32), a yeast centromeric plasmid designed to express DNA binding domains fused to the strong transcriptional activator from the herpes simplex virus VP16 gene under the control of the yeast GAL1 promoter. The insert encoding the R 2 R 3 domains were generated by polymerase chain reaction by which also proper restriction cleavage sites (HindIII and BglII) were generated.

RESULTS
The AMV-specific Mutations in R 2 Alter the Conformation of the DNA Binding Domain-To assess whether the three AMVspecific point mutations in R 2 had any effect on the structure of the minimal DNA binding domain of Myb (R 2 R 3 ), we first studied purified DBDs by fluorescence spectroscopy taking advantage of the high content of tryptophans in these domains conferring intrinsic fluorescent properties to the proteins. Three subdomains were expressed in E. coli and purified, the chicken wild type R 2 R 3 protein (designated R 2 R 3 [wt]) and two mutated R 2 R 3 -harboring mutations found in R 2 of the AMV v-Myb protein. The protein designated R 2 R 3 [AMV NHD ] harbored all three AMV-specific mutations in R 2 (I91N, L106H, and V117D), while R 2 R 3 [AMV HD ] harbored only the two latter (L106H and V117D). Fig. 1 shows the locations of the introduced mutations relative to secondary structure elements of Base numbering in MRE  123456789  GG:  5-GCATTATAACGGTCTTTTAGCGC-3Ј  TG:  5-GCATTATAACTGTCTTTTAGCGC-3Ј  GT:  5-GCATTATAACGTTCTTTTAGCGC-3Ј  TT: 5-GCATTATAACTTTCTTTTAGCGC-3Ј SCHEME 2 R 2 R 3 ( Fig. 1A) and relative to its three-dimensional structure (Fig. 1B). First we compared the fluorescence emission spectra of the two subdomains in three states, guanidinium chloridedenatured, native, or DNA-bound form, using the spectrum of denatured protein for normalization ( Fig. 2). As described previously (16), the emission maximum of R 2 R 3 [wt] was shifted toward a shorter wavelength (340 -344 nm) compared to that of the denatured protein (356 nm) due to the folding of the polypeptide. The folding was also accompanied by a large reduction in quantum yield. Addition of DNA lead to a further slight shift in emission maximum (337-339 nm) and also to a further reduction in quantum yield. We have previously presented evidence that a DNA-induced conformational change in R 2 contributes significantly to the decreased fluorescence of DNA-bound R 2 R 3 [wt] (16 could be more modest than in R 2 R 3 [wt]. To obtain more direct evidence for this hypothesis, we performed fluorescence quenching experiments to monitor if the AMV-specific mutations induced any alteration in the average exposure of tryptophans in the R 2 R 3 -protein. As seen from the Stern-Volmer plot in Fig. 3, both R 2 R 3 [AMV NHD ] and R 2 R 3 [AMV HD ] were quenched significantly less by the neutral quencher acrylamide than R 2 R 3 [wt], supporting our hypothesis of a more compact structure in the mutants relative to the wild type protein. After binding to DNA, all three forms showed a similar tryptophan exposure, indicating more similar conformations in the DNA-bound state. Since a purified R 3 domain was found to be less quenched than R 2 R 3 [wt] (results not shown), the tryptophans in R 3 must be less exposed to solvent than the average of all six tryptophans in R 2 R 3 [wt]. Hence, the three tryptophans in wild type R 2 must be significantly more exposed to solvent than their R 3 homologues in order for R 2 to make a major contribution to the increased slope of the Stern-Volmer plot of R 2 R 3 [wt]. This exposure is then reduced upon some critical mutations in R 2 , both the AMV-specific mutations studied here and the previously reported C130V mutation (designated C43V in Myrset et al. (16)). The L106H and V117D mutations made the major contributions to this conformational effect, since no significant differences were detected between the two proteins that differed with respect to the I91N mutation. For this reason, in some experiments below only R 2 R 3 [wt] and R 2 R 3 [AMV HD ] are compared.
If the difference in conformation between R 2 R 3 [AMV HD ] and R 2 R 3 [wt] were sufficiently large, the two proteins might be expected to display differences in their proteolytic sensitivity. The two purified proteins were therefore subjected to limited proteolysis by chymotrypsin. This protease was chosen since none of the amino acid replacements in R 2 R 3 [AMV HD ] should affect the specificity of the enzyme. The band pattern of proteolytic products revealed distinct differences between R 2 R 3 [AMV HD ] and R 2 R 3 [wt] as indicated by arrows in Fig. 4. Both the total number and the positions of specific bands differed. This result strongly suggests that the AMV-specific point mutations in R 2 indeed have a specific effect on the conformation of the DNA binding domain.
We also asked whether this conformational effect was specific to the AMV mutations or whether the conformation of the wild type protein was particularly sensitive to mutations in R 2 , such that any mutations in this domain would have a high probability of changing the conformation of R 2 . We analyzed a total of 15 different point mutations distributed over the entire R 2 domain by running emission spectra on purified recombinant proteins and by analyzing their DNA binding activity using the electrophoretic mobility shift assay (data not shown).
In the analysis, we found only two classes of mutations. The first class displayed unaltered DNA binding and no change in emission spectra, while the second displayed severely reduced DNA binding accompanied by a significant red shift in max as well as increased quantum yield in the emission spectra, suggesting a more open conformation and a defect in attaining the native active state. In no case did we observe spectra resembling the AMV emission spectra with significantly reduced quantum yield relative to the wild type protein. These results suggest that the AMV-derived mutations have a rather specific effect on the structure of the minimal DNA binding domain of Myb.
The AMV-specific Mutations in R 2 Alter Quantitative Aspects of Myb's DNA Binding Properties-To assess if the altered conformational state in the R 2 R 3 [AMV] proteins had any effects on their DNA binding properties, we performed a series of comparative quantitative DNA binding experiments with R 2 R 3 [AMV NHD ], R 2 R 3 [AMV HD ] and R 2 R 3 [wt] using a sequence derived from the strong A site upstream the mim-1 gene as a binding probe (29). A direct analysis of DNA binding using electrophoretic mobility shift assay did not reveal striking differences between the three proteins (Fig. 5A), consistent with  1, 4, and 7), 40 fmol (lanes 2, 5, and 8), and 100 fmol (lanes 3, 6, and 9) of purified Myb R 2 R 3 proteins were incubated with 10 fmol MRE(mim) probe at 25°C for 10 min and analyzed by the electrophoretic mobility shift assay as described under "Materials and Methods." Lanes 1-3 show the analysis of the Myb R 2 R 3 [wt] protein, lanes 4 -6 that of R 2 R 3 [AMV HD ], and lanes 7-9 that of the R 2 R 3 [AMV NHD ] protein. Panel B, time course of complex dissociation upon competition. Myb R 2 R 3 -DNA complexes were allowed to form at 25°C for 10 min before they were exposed to a 75-fold excess of unlabeled specific MRE(mim) probe for t ϭ 0, 2, 5, 10, and 20 min (lanes 1-5). DNA binding was monitored by the electrophoretic mobility shift assay as described above. For the R 2 R 3 [wt] protein, both free probe and complexes are shown; for the R 2 R 3 [AMV NHD ] and R 2 R 3 [AMV HD ] proteins, only the complexes are presented. Panel C, titration of protein-DNA complexes with poly(dI-dC). 30 fmol of purified Myb R 2 R 3 proteins were incubated with 10 fmol of MRE(mim) probe and increasing amounts of poly(dI-dC) and analyzed by the electrophoretic mobility shift assay as described above. The amounts poly(dI-dC) added was 0 g  6). For the Myb R 2 R 3 [wt] protein both free probe and complexes are shown; for the Myb R 2 R 3 [AMV HD ] protein, only the complexes are shown. After scanning, the ratio K s /K n was estimated as described previously (28). previous findings that the same consensus recognition sequence holds for both c-Myb and AMV v-Myb (25). However, a clear difference was observed when the stability of the three protein-DNA complexes were compared. Under the specific conditions used to compare the decays of the complexes, the R 2 R 3 [wt]-DNA complex dissociated slowly and a substantial fraction resisted 20 min of competition with excess unlabeled DNA. In contrast, both the R 2 R 3 [AMV]-DNA complexes dissociated much more rapidly. After 5 min of competition less remained of the two R 2 R 3 [AMV]-DNA complexes than of the R 2 R 3 [wt]-DNA complex after 20 min (Fig. 5B). These data indicated that the conformational change induced by the AMVspecific mutations in R 2 leads to a considerably reduced stability of the protein DNA complex.
To determine the specific to nonspecific binding constant ratios (K s /K n ) (28), specific DNA complexes were titrated with increasing amounts of poly(dI-dC). As seen in Fig. 5C, the R 2 R 3 [AMV HD ]-DNA complex was titrated at lower concentrations of poly(dI-dC) than the wild type protein. From the data shown a K s /K n ratio of 6 ϫ 10 3 was calculated for the R 2 R 3 [AMV HD ] protein compared to 6 ϫ 10 4 for the wild type protein. Assuming that the nonspecific DNA affinity is not affected, the AMV-specific mutations in R 2 causes a 10-fold reduction in the specific DNA binding constant.

The AMV-specific Mutations in R 2 Alter the Relative Stability of Complexes Formed with Subtypes of MREs-Previous in
vitro studies of preferred binding sequences found no evidence for any alteration of sequence specificity as a result of the AMV-specific mutations (25). We have recently shown that strong MRE sequences with different configurations of guanine bases in the binding site for R 2 gave remarkably different levels of transactivation in vivo at low expression levels of c-Myb (30). In vitro, corresponding differences between the binding sites variants could be seen only in analyses of protein-DNA complex stabilities. Since this phenomenon is related to the sequence of the half-site recognized by R 2 , we asked whether the AMVspecific point mutations in R 2 had any influence on the relative complex stabilities when bound to different MRE variants.
Using three different MRE variants (TAACGG, TAACTG, TAACGT) as binding probes, the stabilities of the complexes formed with R 2 R 3 [AMV HD ] and R 2 R 3 [wt] were compared after competition with an excess of unlabeled MRE (Fig. 6). For all three probes the R 2 R 3 [wt] protein formed more stable complexes than R 2 R 3 [AMV HD ]. A substantial fraction of all the R 2 R 3 [wt] complexes resisted 5 min of competition, but differences in stabilities allowed us to rank the complexes according to affinities in the order: TAACGG Ͼ TAACTG Ͼ TAACGT (Fig. 6). This order parallels their ability to mediate transactivation in vivo (see below). In contrast, the R 2 R 3 [AMV HD ] protein formed much less stable complexes. When bound to the TAACGG probe a visible fraction of the complexes resisted 5 min of competition (comparable to the R 2 R 3 [wt] protein in complex with the TAACGT probe). However, no visible complexes were seen even at the first time point (2 min of competition) when the R 2 R 3 [AMV HD ] protein was bound to the TA-ACGT or to the TAACTG probes. Thus, single-G MRE variants form highly unstable complexes with R 2 R 3 [AMV HD ], but are still reasonably stable when bound by R 2 R 3 [wt]. This suggests that the destabilization of the protein-DNA complexes caused by the AMV-specific mutations might be more severe with some MRE subclasses than with others.
The AMV-specific Mutations in R 2 Alter the Redox Sensitivity of the DNA Binding Domain-The R 2 repeat that is mutated in the AMV v-Myb also harbors a highly conserved redox-sensitive cysteine that was found to be essential for DNA binding, transformation, and transcriptional transactivation (36,37).
The same cysteine is located in a disordered flexible region of R 2 (15,18). We have previously proposed that that this cysteine could function as a molecular sensor for a redox regulatory mechanism turning specific DNA binding on or off by controlling a DNA-induced conformational change in R 2 (16). Having found that the AMV-specific mutations induced a more compact conformation in R 2 , we asked if this could influence the redox sensitivity of the conserved cysteine. First we titrated the two purified R 2 R 3 proteins with increasing concentrations of the SH-specific alkylation reagent NEM. Specific DNA binding was abolished for all three proteins at sufficiently high NEM concentrations, but the titration showed a clear difference between the three variants with respect to NEM sensitivity since R 2 R 3 [wt] was inactivated at significantly lower concentrations of NEM than both R 2 R 3 [AMV HD ] and R 2 R 3 [AMV NHD ] (Fig. 7A). A corresponding difference was observed when the time course of inactivation was measured (results not shown). In both experiments a C130V mutant was unaffected, demonstrating the specificity of the alkylation.
We next examined Myb inactivation by treating the proteins with the SH-specific oxidation reagent diamide. As shown in Fig. 7B, specific DNA binding of R 2 R 3 [wt] was lost at lower concentrations of the SH reagent than observed for R 2 R 3 [AMV HD ], suggesting a reduced redox reactivity as a result of the point mutations in R 2 . A time course experiment led to the same conclusion (results not shown). Finally, Myb inactivation by oxidation of the proteins with H 2 O 2 was monitored. Again, specific DNA binding of R 2 R 3 [wt] was lost at lower concentrations than observed for R 2 R 3 [AMV HD ] (Fig. 7C). Thus, we conclude that the conformational change induced by the AMV-specific point mutations in R 2 , changes the reactivity of the conserved cysteine in this repeat, making the DNA binding domain less susceptible to inactivation through modulation of the redox state of the critical cysteine.
The Effects of the AMV-specific Mutations in Vivo Measured in a Yeast Effector-Reporter System-To assess the effects of altered DNA binding properties in a model in vivo situation, fusions between the DNA binding domains and the strong VP16 transactivation domain were expressed in yeast. The fusion gene was under control of the galactose-inducible GAL1-10 promoter in a centromeric low copy number plasmid, the latter to better mimic expression levels in a physiological in vivo situation. All reporters were high copy number plasmids (2 m) containing the E. coli lacZ gene under control of the yeast FIG. 6. Time course of Myb R 2 R 3 -DNA complex dissociation upon competition with unlabeled probe. Myb R 2 R 3 -DNA complexes were formed and subjected to competition as described in the legend to Fig. 5B. Times of competition were t ϭ 0, 2, 5, 10, and 20 min (lanes 1-5 and 6 -10).  show the analysis of the Myb R 2 R 3 [wt] protein and lanes 6 -10 that of Myb R 2 R 3 [AMV HD ]. Three different labeled MRE variants were used as indicated. DNA binding was monitored by the electrophoretic mobility shift assay as described above. For the GG probe, both free probe and complexes are shown; for the rest, only the complexes are presented. CYC1 minimal promoter. Insertion of three upstream MREs made each of them Myb-responsive. The reference reporter pBP19 with 3 ϫ TAACGGAAC inserted has been described (34). Upon induction with galactose, the R 2 R 3 [wt]-VP16 and the R 2 R 3 [AMV NHD ]-VP16 fusion proteins were expressed to the same level as judged by Western blot analysis of representative yeast extracts using a Myb-specific polyclonal antibody (data not shown). Since the only difference between the two fusion proteins were their DNA binding domains, the levels of induced ␤-galactosidase activity were taken as estimates of their in vivo DNA binding. Both fusion proteins were found to bind the pBP19 reporter plasmid and activate lacZ transcription. Induction of R 2 R 3 [wt]-VP16 resulted in 979 Ϯ 160 ␤-galactosidase units, whereas only 395 Ϯ 65 ␤-galactosidase units were found with R 2 R 3 [AMV NHD ]-VP16 induced. These results suggest that the destabilization of the DNA binding caused by the AMVspecific mutations observed with the TAACGG binding site in vitro, correlated with a corresponding reduced DNA binding and transactivation in vivo.
The AMV-specific Mutations in R 2 Alter the Relative Response to Subtypes of Myb-responsive Promoters in Vivo-We next asked whether the AMV-specific point mutations in R 2 had any influence on the relative responses obtained with different subtypes of synthetic Myb-responsive promoters in vivo. For this purpose, we employed reporter constructs con-

FIG. 8. Myb R 2 R 3 [wt]-and R 2 R 3 [AMV NHD ]-dependent transactivation in a yeast effector-reporter system with different MRE variants.
The yeast effector-reporter system used to monitor the effectiveness of various MREs in vivo has been described elsewhere (30). A reporter plasmid expressing R 2 R 3 [wt]-VP16 or R 2 R 3 [AMV NHD ]-VP16 fusion protein was cotransformed with one of three lacZ reporters containing triple copies of either the GG, GT, or TG variants of the MRE. Transactivation was measured as ␤-galactosidase activity after induction by galactose of the R 2 R 3 -VP16 effectors. Panel A shows the absolute ␤-galactosidase units obtained with the different combinations. Error bars show standard deviations of the mean values from three measurements. Panel B shows the same data normalized by setting the activity of each GG reporter to 100%.  1-6) at 25°C for 5 min before incubation for 2 min with 10 fmol of MRE(mim) probe. As controls, cysteine-free purified Myb R 2 R 3 [C130V] proteins treated with 0 and 100 M NEM (lanes 8 and 9) were used. The complexes were analyzed by the electrophoretic mobility shift assay as described under "Materials and Methods." For the Myb R 2 R 3 [wt] protein, both free probe and complexes are shown; for the R 2 R 3 [AMV NHD ] and R 2 R 3 [AMV HD ] proteins, only the complexes are shown. Panel B, the same amount of purified Myb R 2 R 3 proteins was treated with 0, 30, 60, 90, 120, 150, and 180 M diamide (lanes 1-7) at 25°C for 10 min before incubation with the MRE(mim) probe as described above. Panel C, the same amount of purified Myb R 2 R 3 proteins was treated with 0, 1, 2, 4, 6, 8, and 10 mM H 2 O 2 (lanes 1-7) at 25°C for 20 min before incubation with the MRE(mim) probe. Myb R 2 R 3 [C130V] controls treated with 0 and 10 mM H 2 O 2 are shown in lanes 8 and 9. taining three copies of MRE sequences having different configurations of Gs in positions 5 and 6 of the MRE consensus sequence; that is either 3 ϫ TAACGG, 3 ϫ TAACGT, or 3 ϫ TAACTG (designated "3ϫGG," "3ϫGT," or "3ϫTG reporters," respectively). A reporter with 3 ϫ TAACTT was used as negative control. As illustrated in Fig. 8A, the 3ϫGG reporter gave the highest levels of transactivation for both fusion proteins. The transactivation obtained with the 3ϫGT and 3ϫTG reporters were generally much lower, but still at a significant level when activated by the R 2 R 3 [wt]-VP16 fusion protein, consistent with our previous report (30). In contrast, the R 2 R 3 -[AMV NHD ]-VP16 fusion protein resulted in very low levels of transactivation for both the 3ϫGT and 3ϫTG reporters. This suggests that the reduced DNA binding seen with the R 2 R 3 [AMV] proteins, relative to the wild type protein, is more severe for certain MRE sequences than for others as already seen in in vitro data in Fig. 6. This is illustrated in Fig. 8B where the transactivation data for each effector are normalized to the 3ϫGG reporter. If the reduction in DNA binding caused by the AMV-specific point mutations were proportional for the different MREs, the three synthetic promoters analyzed would have given pairwise equal levels of transactivation when normalized as in Fig. 8B, which is clearly not the case. In particular the 3ϫTG-reporter responded very poorly to the R 2 R 3 [AMV NHD ]-VP16 fusion protein, while it was reasonably activated by the R 2 R 3 [wt]-VP16 fusion protein.
Thus, we conclude that, even if the AMV-specific point mutations do not directly alter the sequence specificity of the protein as demonstrated previously (25), these mutations seem to weaken the interaction with subclasses of the recognition sequence to a different extent. It is therefore conceivable that different subsets of Myb-responsive promoters react differentially to these mutations and that these quantitative effects contribute to the altered patterns of gene activation induced by AMV v-Myb relative to wild type c-Myb. DISCUSSION In this work we have investigated how the properties of Myb are affected by the point mutations that have been selected for in the DNA binding domain of AMV v-Myb. Despite striking phenotypic effects in vivo (23,24), the mechanism of action of these mutations has remained elusive since no change in sequence specificity of the mutated protein was observed (25), and no alterations in protein-protein interactions with a hypothetical partner has been reported. To better understand which properties of the protein might have been altered, we performed a detailed biochemical analysis of the minimal DNA binding domain in mutated versus wild type forms.
Most of the properties analyzed were indeed altered. The AMV-specific mutations in the second repeat had significant effects on conformation, redox-sensitivity, and on quantitative aspects of the DNA binding properties of the protein. Even if the mutated protein binds the same range of sequences as the wild type, we found a sequence dependence in the magnitude of destabilization of the complexes in vitro and in DNA binding in vivo. These latter observations in particular offer a possible explanation of how the AMV-specific mutations can lead to a different spectrum of genes activated by AMV-Myb versus c-Myb.
The evidence for a conformational change comes primarily from fluorescence quenching experiments, which showed that the average exposure of the many tryptophans in the DNA binding domain is reduced in the two R 2 R 3 [AMV] mutants compared to that in R 2 R 3 [wt]. Since the introduced AMV-specific mutations are located in R 2 , it is reasonable that the conformation of this domain was most affected. The R 2 domain seems to be flexible and temperature-sensitive, and it changes conformation upon binding to DNA (15)(16)(17)38). This conformational flexibility might well be sensitive to mutations. Previously analysis of Cys mutations located in R 2 also changed the conformation of the domain. However, when we screened a series of 15 additional mutations distributed over the entire R 2 , none were found with the same properties as R 2 R 3 [AMV], suggesting that the AMV-specific mutations have a particular effect on the structure.
It is noteworthy that all three point mutations in R 2 represents changes from rather hydrophobic residues to more polar ones (I91N, L106H, and V117D). Since they are all located on the surface of the protein (14), one possibility is that they might stabilize the folded structure of the domain through solvation effects. In addition, inspection of the structure reported by Ogata et al. (14) revealed that two of the mutations (L106H and V117D) were close to each other in space (Fig. 1B). If both were ionized, the resulting electrostatic interaction between His ϩ and Asp Ϫ could also have a stabilizing effect on the AMV v-Myb DNA binding domain. In accordance with this hypothesis all differences observed in this work could be attributed to the L106H and V117D mutations, since the two proteins R 2 R 3 [AMV HD ] and R 2 R 3 [AMV NHD ] behaved equally in all comparative experiments. R 2 contains a single reactive cysteine that is exposed for modification in the free protein but protected in the tighter folded DNA-bound conformation (16). According to the structural model the cysteine is located in the interior of the protein in the DNA complex (14). The observation that the reactivity of this cysteine is reduced in the R 2 R 3 [AMV] proteins compared to the R 2 R 3 [wt] protein, suggests that the cysteine on average is in a less exposed conformation in R 2 R 3 [AMV]. It is possible that an unstable or nascent helix in the region around the cysteine (18) might be stabilized by a tighter conformation in other parts of R 2 . Alternatively, the reported cavity in R 2 might be affected (17). The cysteine in R 2 is highly conserved, but is not essential for DNA binding since hydrophobic substitutions may be introduced without reducing the DNA binding activity of the protein (16). Its high redox reactivity has led us and others to suggest that it might be conserved to keep the protein responsive toward a possible redox regulatory mechanism. The observation that the AMV-specific mutations make the protein less responsive toward Cys modification and oxidation suggests that the oncogenic version might remain partially active under conditions where the normal variant would be inactivated. This might contribute to its oncogenic potential under specific conditions.
Previous reports found no qualitative differences in DNA sequence recognition properties between AMV v-Myb and c-Myb (25). With respect to quantitative DNA binding properties AMV v-Myb subdomains were reported to bind stronger (19) or weaker (24) to MRE as judged by direct electrophoretic mobility shift assay using bacterial extracts or purified recombinant Myb. We therefore analyzed in more detail quantitative aspects of DNA binding using purified R 2 R 3 [wt], R 2 R 3 [AMV NHD ], and R 2 R 3 [AMV HD ]. Although the three minimal DNA binding domains seemed very similar when analyzed in a direct mobility shift assay, competition assays revealed striking differences in stabilities between the formed complexes. Our analysis clearly demonstrates that the AMV-specific mutations weakens the interaction with the mim-1-derived recognition sequence. This adds to our previous hypothesis that R 2 plays a role in modulation of complex stability. In keeping with that different variants of the second half-site bound by R 2 lead to different complex stabilities (20), we show here that the AMV mutant versions of R 2 also displays altered complex stability.
The most interesting effect of the AMV-specific mutations was the differential destabilization on certain MRE variants. Binding to a TAACTG variant of MRE was much more destabilized than a complex with a TAACGG variant. The consequence of this differential destabilization when analyzed in an in vivo yeast system was that R 2 R 3 [AMV NHD ]-VP16 still transactivated through a TAACGG variant of MRE, but gave very low level of transactivation through a TAACTG site. In contrast R 2 R 3 [wt]-VP16 transactivated both variants of synthetic promoters. Among the candidate target promoters for Myb reported in the literature, the TG variants of binding sites are found more often than the GG variants. It is therefore quite possible that certain subclasses of promoters activated by c-Myb will fall below a critical level of affinity and not be activated by AMV v-Myb while other subclasses are activated by both factors. This may result in a different spectrum of genes activated by AMV-Myb relative to c-Myb. Our analysis has revealed that several intrinsic properties of the minimal DNA binding domain of Myb are modified as a result of the AMV-specific mutations in R 2 . Even if we cannot directly link these changes to the observed phenotypic effects of the same mutations in transformed cells, it is quite probable that quantitative alterations in DNA binding properties will have important phenotypic effects. The prevailing alternative hypothesis has been to assume alterations in protein-protein interactions with a hypothetical partner. We cannot exclude that AMV v-Myb later will prove also to have modified interactions with such partner proteins, but our results show that we do not need to postulate such alterations to explain phenotypic effects of the AMV point mutations. Alterations in other subdomains of the protein also mutated in AMV v-Myb probably add to the changes analyzed in this work. It was recently reported that the transactivation properties of the protein was weakened as a result of some of these mutations (24,39). Although it is intriguing that a more potent oncogene encodes a transcription factor with weakened DNA binding and reduced transactivation properties, the finding that most of the mutations have a clear effect probably reflects the long time through which these mutations have been selected for.