Sequence Selectivity of c-Myb in Vivo

We have investigated the basis for the striking difference between the broad DNA sequence selectivity of the c-Myb transcription factor minimal DNA-binding domain R2R3 in vitro and the more restricted preference of a R2R3VP16 protein for Myb-specific recognition elements (MREs) in a Saccharomyces cerevisiae transactivation system. We show that sequence discrimination in yeast is highly dependent on the expression level of Myb effector protein. Full-length c-Myb and a C-terminally truncated protein (residues 1–360) were also included in the study. All of the tested Myb proteins displayed very similar DNA binding properties in electrophoretic mobility shift assays. Only minor differences between full-length c-Myb and truncated c-Myb(1–360) were observed. In transactivation studies in CV-1 cells, the MRE selectivity was highest at low expression levels of Myb effector proteins. However, the discrimination between MRE variants was rapidly lost with high input levels of effector plasmid. In c-Myb-expressing K-562 cells, the high degree of MRE selectivity was retained, thereby confirming the relevance of the results obtained in the yeast system. These data suggest that the MRE selectivity of c-Myb is an intrinsic property of only the R2R3 domain itself and that the transactivation response of a specific MRE in vivo may be highly dependent on the expression level of the Myb protein in the cell.

The c-Myb protein is a transcription factor encoded by the c-myb proto-oncogene (reviewed in Refs. [1][2][3]. A short N terminus is followed by a highly conserved DNA-binding domain, a centrally located transactivation domain, and more C-terminally located negative regulatory domains. The DNA-binding domain is well characterized, comprising the three imperfect repeats, R 1 , R 2 , and R 3 . The R 2 and R 3 repeats alone are sufficient for sequence-specific DNA binding (4). This DNAbinding motif is highly conserved throughout evolution in both animal and plant kingdoms (5).
A large body of evidence suggests strongly that c-Myb is involved in regulating cell growth and differentiation in hematopoietic cells. In chickens, retroviral v-Myb proteins p48 v-myb (AMV-derived) and p135 gag-myb-ets (E26-derived) both elicit myeloid leukemia. In mice, the majority of retroviral insertions in the c-myb gene result in N-terminal truncations of the c-Myb protein and deregulated expression, leading to myeloid leuke-mia (6). The abrogation of fetal liver hematopoiesis in mice with a c-myb null mutation confirmed a vital role for c-Myb in the development of hematopoietic cells (7). A recent study showed that c-Myb is expressed in human primitive hematopoietic stem cells in the fetal aorta region, prior to colonization and definitive hematopoiesis in the fetal liver, confirming an important role for c-Myb in the development of hematopoietic cells (8).
Much effort has been put into identifying relevant target genes for c-Myb action, and binding sites for c-Myb have been identified in an increasing number of gene promoters, including mim-1, neutrophile elastase (ELA2), cdc-2, c-myc, bcl-2, T cell receptor ␦ and ␥ chains, CD4, ADA, and lck (2,6). However, it has been difficult to establish a direct link between candidate genes, growth control, and oncogenic transformation (9). A notable exception is that of v-Myb being able to transactivate the GBX2 gene promoter and up-regulate transcription of the transcription factor GBX2. GBX2 in turn regulates transcription of the chicken growth factor cMGF, thereby creating an autocrine regulatory loop (10). Other promising candidate cmyb target genes, such as tom-1 (11) and the adenosine 2B receptor (12), have been identified in differential screens, but promoter analysis data are not yet available.
Several approaches have been used to define the consensus core DNA-binding site for c-Myb, YAACNG (Myb recognition element (MRE)). 1 The Myb consensus sequence was first deduced by isolation of chicken genomic DNA fragments bound by v-Myb on filters (13) and from comparison of putative Myb binding sites within the SV40 enhancer (14). Polymerase chain reaction-based binding-site selection methods with Myb proteins resulted in minor extensions of the MRE consensus sequence (15,16). Mutational analysis confirmed by NMR structural data has revealed that the Myb MRE is bipartite. The first half-site (YAAC) has the majority of specific contacts to R 3 , and the less well defined second half-site has mainly specific contacts with the R 2 subdomain (17)(18)(19). The first half-site of the MRE is absolutely required for DNA-binding. Sequence substitutions in the second half-site mainly affected the halflife of the protein-DNA complex in vitro (17). A more detailed analysis of the second half-site revealed a flexible sequence requirement, possibly caused by a flexible structure in R 2 (20 -22). In particular, base changes were readily accommodated in the highly conserved G6 position of the MRE, as long as a G was present in position 5. Bacterially expressed R 2 R 3 protein bound the MRE variants TAACGG, TAACGT, and TAACTG with very similar binding affinities as measured by electrophoretic mobility shift assay (EMSA). In contrast, these sequences conferred strikingly different transactivation activities (GG Ͼ Ͼ TG Ͼ GT and nonfunctional TT) in a strictly Myb DBD-dependent effector/reporter system, using a R 2 R 3 VP16 fusion protein and MRE ␤-galactosidase reporter plasmids in Saccharomyces cerevisiae (23). This suggests that optimal MRE sequences defined by in vitro binding studies may not directly reflect their importance and/or activity in vivo. The present study investigates in detail the basis for this in vivo/in vitro paradox. We have compared the MRE selectivity of recombinant R 2 R 3 and R 2 R 3 VP16 in EMSAs and investigated the effect of Myb effector expression level on MRE preference in S. cerevisiae. The MRE selectivity of full-length and a C-terminally truncated Myb protein was evaluated both in vitro and in vivo in mammalian cells. We show that the MRE target sequence selectivity is an intrinsic property of the R 2 R 3 domain, independent of other regions in the protein. However, more importantly, the apparent MRE target selectivity in vivo is highly affected by the Myb protein expression level both in yeast and in mammalian cells.

MATERIALS AND METHODS
Cell Culture and Reagents-CV-1, COS-1, and K-562 cells were obtained from ATCC. COS-1 and CV-1 cells were maintained in Dulbecco's modified Eagle's medium. K-562 cells were maintained in RPMI 1640 medium. All media were supplemented with 10% fetal calf serum, 20 mM glutamine, 1% penicillin and streptomycin. Cells were grown in 5% CO 2 in a humidified atmosphere.
Electrophoretic Mobility Shift Assay-DNA binding was monitored by EMSA. The sequences of the double-stranded oligo for the mim-1 A site (27) and variant MRE GG, GT, TG, and TT oligos are as in Ref. 23, except that all variant oligos were extended at the 3Ј-end with nucleotides TGG in order to improve annealing efficiency (see Table I). Variant MRE oligos were labeled to the same specific activity and purified as described (23). The mim-1 A site oligo was labeled and purified as described (23). Cold double-stranded oligos for competition were similarly annealed, filled-in and PAGE-purified.
Myb proteins in whole-cell lysates were titrated by EMSA and standardized to the same amount of Myb DNA binding activity on the mim-1 A site oligo, due to effects of c-Myb on the transactivation level of the internal standard plasmid (see below on transactivation in mammalian cells). The variation of the DNA-binding activity (on mim-1 A) between extracts with the different Myb proteins within the same transfection series was within Ϯ50% in lysate volume; typically 1-3 l of COS-1 cell extract was used per binding reaction. COS-1 cell lysates with various Myb proteins were adjusted to equal volumes with Buffer F before incubation in Buffer H (20 mM HEPES, pH 8.0, 50 mM NaCl, 1 mM dithiothreitol, 0.1 mM EDTA, pH 8.0, 10% glycerol), with final NaCl concentration 60 or 75 mM. Where indicated, binding reactions with recombinant R 2 R 3 (25-50 fmol) were included. 5Ј-␥-32 P-Labeled DNA oligo probe (10 or 30 fmol) was added, and the binding mixture (total volume 20 l) was incubated for 10 min at 25°C before electrophoresis or competition with cold oligos. Binding reactions were run on 5% 0.5ϫ TBE PAGE gels and 0.5ϫ TBE buffer at 4°C and Ϸ15 V/cm. Autoradiography was performed by standard procedures, and relative amounts of protein-DNA complexes were quantified by densitometric scanning. For comparison of band intensities, one of the protein-DNA complexes per series was arbitrarily designated as 100%.
Transactivation Assays in Mammalian Cells-CV-1 cells were transfected by polyethyleneimine (28) or by CaPO 4 precipitation with precipitate parameters as described (29). Cells were seeded the day before at 1.5 ϫ 10 5 cells/plate in 60-mm plates. Cells were transfected with 4 g of pGL2/tk/3xMRE variant reporter plasmid and the amounts of expression plasmid indicated in the figures. In transfections with polyethyleneimine, the total amount of expression plasmid input was kept constant at 5 g by adding appropriate amounts of the pCIneo plasmid. Polyethyleneimine was added to 9 equivalents (ϭ 2430 nmol pr total 9 g DNA/plate). Cells were lysed and scraped in 400 l of reporter lysis buffer (Promega) or lysed in 400 l of passive lysis buffer (in cotransfections with pRL-CMV).
In preliminary experiments, 25 ng pRL-CMV (Renilla reniformis luciferase, Promega) or 0.5 g of pCMV-␤-galactosidase (Promega) per plate were included as internal controls of transfection efficiency. However, we consistently observed effects on the transactivation level of the internal control plasmid depending on the amount of input c-Myb expression plasmid (data not shown). All transactivation results are presented without standardization to internal control plasmids.
K-562 cells were washed once in RPMI 1640 medium without additions. Cells were resuspended at 2 ϫ 10 7 cells/ml. Aliquots (500 l) were added to 10 g of pGL2/tk reporter plasmid, transferred to 4-mm-gap electroporation cuvettes, and pulsed once at 310 V and 960 microfarads (optimized) in a Bio-Rad Gene Pulser giving a pulse decay of t1 ⁄2 Ϸ 20 ms. Cells were diluted in RPMI 1640 medium with additions. After 36 h, cells were washed once in phosphate-buffered saline, resuspended in reporter lysis buffer, snap-frozen in liquid N 2 , and thawed. Debris was pelleted by centrifugation.
All lysates were analyzed for luciferase activity using either a luciferase assay kit or dual luciferase assay kit (when pRL-CMV was cotransfected) as described by the manufacturer (Promega).
Transactivation Assays in Yeast-Transactivation assays in the yeast strain INVSC1 (Invitrogen) were performed in galactose-containing medium as described (23) with the control plasmids pDBD 11 or pDBD 11 HK and the c-Myb R 2 R 3 VP16 expression plasmids pDBD 11 R 2 R 3 and pDBD 11 HK/R 2 R 3 . The yeast reporter plasmids contained triple copy GG, GT, TG, and TT oligos (see above) inserted 5Ј of the reporter gene (see above). Induction of ␤-galactosidase activity was assayed with o-nitrophenyl ␤-D-galactopyranoside (Sigma) as substrate (30).
Western Blot-Cell lysates from COS-1-transfected cells were run on a 15% discontinuous SDS-PAGE gel and electroblotted onto a nitrocellulose filter. Filters were blocked in BLOTTO, and Myb proteins were detected by the murine Mab 5E11 (31), followed by horseradish peroxidase-conjugated donkey anti-mouse antibody (Amersham Pharmacia Biotech). Yeast strain INVSCI transformed with Myb expression plasmid and the 3xGG MRE reporter plasmid were grown in selective media. Equal amounts of cells were lysed with glass beads and 5% trichloroacetic acid (32), run on 15% SDS-PAGE gels, and blotted onto polyvinylidene difluoride membranes. R 2 R 3 VP16 fusion protein was detected by the murine monoclonal anti-VP16 antibody 2GV4 (33) and horseradish peroxidase-conjugated donkey anti-mouse antibody. Horseradish peroxidase activity was detected by ECL or ECL ϩ , as described by the manufacturer (Amersham Pharmacia Biotech).

The MRE Selectivity of R 2 R 3 VP16 in Vitro and in Vivo in
Mammalian Cells-We have previously demonstrated that recombinant c-Myb R 2 R 3 can bind equally well to MRE variants containing a guanine in positions 5 and/or 6 of the consensus core MRE TAACNG as analyzed by EMSA. In sharp contrast to this low selectivity in vitro, transactivation of the same MRE variants as 3xMRE reporter plasmids in S. cerevisiae by R 2 R 3 VP16 fusion protein was highly selective (23). We first asked whether this striking in vitro/in vivo difference in MRE discrimination could be attributed to the fusion of the VP16 transactivation domain to the c-Myb minimal DNA-binding domain in the in vivo experiment. A CMV-based expression plasmid for chicken c-Myb minimal DNA-binding domain R 2 R 3 fused to the VP16 transactivation domain, R 2 R 3 VP16, was transfected into COS-1 cells, and whole-cell lysates were used as the protein source for EMSA.
The DNA binding capacity of R 2 R 3 VP16 expressed in mammalian cells was compared with R 2 R 3 produced in E. coli by EMSA. The set of MRE variant oligos used in this study are listed in Table I with the nomenclature referring to the bases in nucleotide positions 5 and 6 in the MRE core consensus sequence. COS-1 cell lysate with R 2 R 3 VP16 or recombinant R 2 R 3 was bound to MRE oligo GG, GT, TG, or TT (Fig. 1). The oligos were labeled to equal specific activity, thus yielding directly comparable intensities of DNA-protein complexes. R 2 R 3 VP16 bound the GG, GT, and TG oligos (Fig 1, lanes 2-4) with very similar intensities (relative binding of the oligos was 100, 94, and 89%, respectively, by scanning densitometry). Recombinant R 2 R 3 also bound the GG, GT, and TG oligos with very similar intensities (100, 120, and 120% respectively) (Fig. 1, lanes [6][7][8]and Ref. 23). Neither R 2 R 3 VP16 lysate (Fig. 1, lane 5) nor recombinant R 2 R 3 (23) was able to bind to the TT MRE variant oligo. This comparison shows that the C-terminal fusion of the VP16 transactivation domain to c-Myb R 2 R 3 does not have any major effect on the MRE-selectivity. These data also exclude the possibility that the expression system for the Myb protein (i.e. E. coli versus mammalian cells) might affect the DNA binding properties of R 2 R 3 .
We next performed transactivation assays in mammalian cells with R 2 R 3 VP16 and 3xMRE reporter plasmids to evaluate whether the high selectivity in yeast also could be observed in mammalian cells or whether this selectivity was a property only of the yeast system. CV-1 cells were transfected with pGL2/tk/3xMRE reporter plasmids and increasing amounts of R 2 R 3 VP16 expression plasmid (Fig. 2). There was no significant difference in the ability of R 2 R 3 VP16 to discriminate between MRE variants GG, GT, and TG in mammalian cells, regardless of the input level of effector plasmid. This finding contrasts our earlier observations in S. cerevisiae (23). This poor selectivity of R 2 R 3 VP16 for MRE in CV-1 cells is in keeping with results from in vitro-based selection strategies for defining the Myb MRE (13)(14)(15)(16). One might therefore tentatively conclude that the high sequence selectivity of Myb R 2 R 3 VP16 in S. cerevisiae was an artifact of the yeast system. However, there are important biological differences between the two transactivation systems that might affect the sensitivity of the assay. In the mammalian CV-1 cell system, only about 20 -25% of the population was transfected (as judged by cells expressing green fluorescent protein, data not shown). In this type of transient transfection assay, there is no control over the number of plasmids taken up per cell, i.e. no control of gene dosage (effector) or percentage of cells harboring both effector and reporter plasmids. Also, the R 2 R 3 VP16 protein was expressed from the very strong CMV promoter, which presumably results in a very high expression level of Myb protein per transfected cell. In the yeast system, the cell population is homogeneous, with all cells containing both effector and re-FIG. 2. The transactivation of ability of R 2 R 3 VP16 on MRE variant recognition elements in CV-1 cells. CV-1 cells were transfected with expression construct for R 2 R 3 VP16 at the indicated concentrations (g) and pGL2/tk reporter plasmids (4 g) containing triple copies of variant Myb recognition elements (3xGG, 3x GT, 3xTG, and 3xTT). Fold relative transactivation was calculated relative to the pGL2/tk/3xGG reporter with empty expression plasmid (pCIneo). Results are expressed as mean fold induction of luciferase activity and S.E. (n ϭ 3).  1. DNA binding selectivity of R 2 R 3 VP16. COS-1 cell lysate with c-Myb R 2 R 3 VP16 (2 l, lanes 2-5) or recombinant R 2 R 3 (320 fmol, lanes 6 -8) was bound to oligo GG, GT, TG, or TT (10 fmol each for R 2 R 3 VP16 and 220 fmol each for recombinant R 2 R 3 ), as indicated. Probes were labeled to equal specific activity. Complexes were incubated for 10 -15 min at 25°C before analysis by EMSA. Protein-DNA complexes are marked with arrows.
porter plasmids. The Myb R 2 R 3 VP16 protein was expressed from the Gal1-10 promoter at a supposedly physiological level from a low copy number (CEN-ARS) plasmid. The gene dosage and expression level of effector protein is therefore well controlled in the yeast system. We hypothesized that the differences in sequence selectivity between yeast and mammalian cell system might in part be attributed to the expression level of the effector protein.
Effect of Copy Number on MRE Selectivity in Yeast-In S. cerevisiae, the expression level of a protein can easily be manipulated by using low (CEN-ARS) or high (2 ) copy number expression plasmids. We exploited this feature by examining the effects of the copy number of the Myb expression plasmid on the transactivation level of 3xMRE-HIS3-lacZ fusion reporter plasmids. At a low expression level of Myb R 2 R 3 VP16, a high degree of selectivity for MRE sequence in terms of transactivation level was evident (Fig. 3A). Transactivation of the 3xGG reporter was strong, whereas the 3xGT and 3xTG variants were transactivated only marginally above the background level, and the 3xTT variant was not transactivated at all. In contrast, when R 2 R 3 VP16 was expressed from a high copy plasmid, the high degree of selectivity between MRE sequence variants was lost, with the 3xGT and 3xTG variant reporter plasmids being transactivated only slightly less than the 3xGG variant (Fig. 3B). The expected difference in the expression levels of Myb R 2 R 3 VP16 from the low copy and high copy vectors (Fig. 3C) was verified by Western blotting and detection by an anti-VP16 transactivation domain antibody (2GV4) (33). Notably, the 3xGG MRE reporter plasmid was transactivated to the same level regardless of the copy number of the effector plasmid, implying that R 2 R 3 VP16 expressed at a low level was sufficient to saturate the MREs on this reporter plasmid. The less optimal GT and TG MRE variant reporters were maximally transactivated when R 2 R 3 VP16 was overexpressed. This suggested that overexpression of R 2 R 3 VP16 leads to occupation of suboptimal DNA-binding sites and high trans-activation, giving the impression that those DNA-binding sites were optimal for the R 2 R 3 VP16 protein. These data show clearly that the sequence selectivity of R 2 R 3 VP16 in S. cerevisiae is highly dependent on the protein expression level. These results also show that very subtle differences in DNA-binding affinity that are too small to be clearly detected in vitro may be amplified to give large differences in transcriptional output in vivo.
Analysis of the DNA Binding Properties of Full-length Myb Protein Expressed in COS Cells-The relevance of using the S. cerevisiae transactivation system for evaluating optimal DNAbinding sites for c-Myb requires that Myb proteins have very similar properties in yeast and in mammalian cells. The high selectivity for the GG MRE variant in yeast raised the question of whether R 2 R 3 VP16 (and R 2 R 3 ) accurately reflected the MRE selectivity of full-length c-Myb. We wished to characterize the selectivity of full-length c-Myb on a selected set of variant MRE sequences. Truncated c-Myb protein was also included, because there have been conflicting reports on whether the DNA binding activity of c-Myb is affected by C-terminal truncations (16, 34 -36). COS-7 cells have been successfully used for high expression levels of full-length murine c-Myb (24). We expressed c-Myb(FL) (full-length murine c-Myb) and c-Myb(1-360) (murine c-Myb residues 1-360) proteins in COS-1 cells using SV40dependent replicative plasmids with CMV-directed gene expression. Whole-cells lysates were made under mild denaturing conditions, and the content of Myb proteins was analyzed by Western blotting. As shown in Fig. 4, lysates from cells transfected with c-Myb(1-360) (lane 1) and c-Myb(FL) (lane 2) contained readily detectable amounts of Myb proteins with expected molecular masses of 45 and 75 kDa, respectively. We did not observe major breakdown fragments of c-Myb, as reported for other expression systems (34,35,37). Some weak extra bands representing possible proteolytic fragments were detected. Because the Myb proteins were detected by an anti-Myb DBD antibody, these fragments would be expected to have

FIG. 3. Effects of copy number on Myb transactivation in yeast.
Low copy (A) or high copy (B) expression plasmids for R 2 R 3 VP16 (pDBD 11 R 2 R 3 and pDBD 11 HK/R 2 R 3 , respectively) and yeast reporter plasmids containing the triple copies of the indicated variants of the MRE were cotransformed into S. cerevisiae INVSCI cells. Transactivation was measured as ␤-galactosidase activity. Background is the transactivation level on the 3xGG reporter plasmid with the empty expression plasmids pDBD 11 (A) or pDBD 11 HK (B). Data are presented as mean ␤-galactosidase units and S.E. (n ϭ 4). A Western blot (C) shows the expression level of R 2 R 3 VP16 as detected by the anti-VP16 antibody 2GV4 (33) in cells cotransformed with the 3xGG reporter and the following plasmids as indicated: N, pDBD 11 ; LC, pDBD 11 R 2 R 3 ; HC, pDBD 11 -HK/R 2 R 3 .
DNA binding activity. However, we did not observe additional shifted bands in the EMSA (see Fig. 5), suggesting that these minor bands would not interfere in our analysis.
In preliminary transactivation assays, we consistently observed that the input level of Myb expression vector affected the transactivation level of the internal control plasmid(s). We therefore omitted the use of an internal control plasmid as a method to normalize cell lysates with respect to transfection efficiencies. We found a good correspondence between the relative amounts of c-Myb(FL) and c-Myb(1-360) proteins present in the lysates and DNA binding activities. For example, the c-Myb(1-360) lysate in Fig. 4 contained twice as much Myb protein and twice as much DNA binding activity on the mim-1 A oligo (data not shown) as an equal volume of c-Myb(FL) lysate. Before EMSA, we therefore chose to standardized the relative amounts of Myb proteins present in the COS-1 lysates from the same transfection by equal DNA binding activity to a mim-1 A site oligo (see under "Materials and Methods").
We first investigated the ability of c-Myb(FL) and c-Myb(1-360) proteins to bind to selected MRE variant oligos. To allow direct comparison, standardized lysates (see above) containing c-Myb(1-360) or c-Myb(FL) were bound to GG, GT, TG, or TT oligos and analyzed by EMSA (Fig. 5). Both c-Myb(1-360) (Fig.  5, lanes 1-4) and c-Myb(FL) (lanes 5-8) bound with similar intensities to the GG, GT, and TG MRE variant oligos. The relative intensities for c-Myb(1-360) and c-Myb(FL) bound to GG, GT, and TG were 100, 91, and 78%, and 100, 77, and 70%, respectively, as measured by densitometric scanning (arbitrary units, see under "Materials and Methods"). A 20 -30% reduction in the binding of c-Myb(1-360) and c-Myb(FL) to the GT and TG oligos relative to GG was consistently observed. Also, c-Myb(FL) bound slightly more weakly to the GT and TG oligos compared with c- Myb(1-360). This reduced binding affinity was also observed in other experiments (data not shown) and in those shown in Figs. 6 and 7. For the R 2 R 3 VP16 protein, this slight reduction in binding to GT and TG oligos was approximately 10% (Fig. 1, lanes 2 and 3). We concluded that there was no major shift in the sequence selectivity of c-Myb(FL) or c-Myb(1-360) compared with R 2 R 3 VP16 for the chosen MRE oligo subset.
We next examined whether subtle differences between the DNA binding capacity of the three Myb proteins could be detected by allowing them to compete with each other for decreasing amounts of oligo. We hypothesized that a difference in the binding affinity between the three proteins would result in a difference in the ability to bind the MRE oligo under DNAlimiting conditions. Standardized lysates with each of the three Myb proteins were mixed and bound to serial dilutions of GG, GT, or TG oligo, and analyzed by EMSA (Fig. 6). There was no striking differences in ability of the three Myb proteins R 2 R 3 VP16, c-Myb(1-360), and c-Myb(FL) to compete with each other for limiting amounts of the same MRE oligo (Fig. 6). The amounts of protein-DNA complex for each protein and oligo combination were calculated relative to the amount of R 2 R 3 VP16/GG complex (Fig. 6B). The rate of decreasing protein-DNA complexes with decreasing oligo concentration was overall very similar (Fig. 6B). However, some differences were found. In general, all three Myb proteins formed less protein-DNA complexes with the GT (35-65%) and TG (45-65%) oligos relative to the GG. Also, c-Myb(FL) bound less well than c-Myb(1-360) to the GT and TG oligos (approximately 10 -20% less). This subtle difference between c-Myb(FL) and c-Myb(1-360) is also reflected in Fig. 5, as shown above. A binding reaction with the TT oligo (Fig. 6A, lane 15) was included as a negative control, demonstrating the absence of nonspecific protein-DNA complexes in the binding reactions. This experiment suggests that a subtle difference in binding affinity of c-Myb(FL) and also c-Myb(1-360) could be ranked in the order GG Ͼ TG Ն GT.
The stability and/or apparent affinity of the three different Myb proteins for the MRE variant oligos was also estimated by comparing the ability of these oligos to compete out a mim-1 A oligo from preformed protein-DNA complexes. Standardized COS-1 lysates with each of the three Myb proteins were mixed and bound to a radioactive mim-1 A site oligo, followed by competition with a 25-fold excess of nonlabeled oligos for the MRE variants GG, GT, TG, and TT. The MRE variant oligos showed a qualitative difference in the ability to compete out mim-1 A site oligo for all three forms of Myb protein (Fig. 7). The amounts of protein-DNA complex for each protein and oligo combination were calculated relative to the amount of R 2 R 3 VP16/mim-1 A (Fig. 7B). The exchange between radiolabeled oligo and competitor oligo appeared to be quite rapid for all three proteins, because there was no major difference between the 3 and 15 min time points. The GG oligo competed efficiently out the mim-1 A oligo bound to all three Myb proteins (complex intensities reduced by 40 -70%, relative to complex bound to mim-1 A oligo), as expected, whereas the GT and TG variants were less efficient (10 -40% reduction), and the TT was completely unable to compete out the bound mim-1 A oligo  1-4) or c-Myb(FL) (lanes 5-8) were bound to oligos GG, GT, TG, or TT as indicated. Probes were labeled to equal specific activity. Recombinant R 2 R 3 protein was included as an internal control (lane 9). Complexes were incubated for 10 -15 min at 25°C before analysis by EMSA. Protein-DNA complexes are marked with arrows.
(lanes [11][12][13], thus reflecting the results in Figs. 5 and 6. To summarize the EMSA experiments, we found only small differences between c-Myb(FL) and c-Myb(1-360) and R 2 R 3 VP16 proteins in the ability to bind to selected variants of the MRE, with an apparent DNA-binding affinity in the order GG Ͼ TG ϳ GT. All three Myb proteins expressed in COS-1 cells exhibited DNA binding properties similar to previously characterized recombinant R 2 R 3 (23). Because we were not able to quantitate the R 2 R 3 VP16 protein in the COS-1 lysates by anti-Myb antibody, we cannot formally prove that the DNA binding capacity of c-Myb(FL) and R 2 R 3 VP16 proteins are identical. However, the apparent properties of R 2 R 3 VP16 and c-Myb(FL) are very similar. More importantly, the DNA binding activity of c-Myb(FL) and the C-terminally truncated c-Myb(1-360) was very similar, in contrast to some reports (34,35,38).
Transactivation Ability of Myb Proteins on Variant MREs in Mammalian Cells-We next tested the sequence selectivity of c-Myb proteins in transactivation studies in mammalian cells. CV-1 cells were transiently transfected with increasing amounts of expression plasmids for c-Myb(FL) or c-Myb(1-360) and pGL2/tk/3xMRE reporter luciferase plasmids (Fig. 8). In general, the MRE selectivity of c-Myb(FL) and c-Myb(1-360) seemed to be very similar. For both c-Myb(FL) (Fig. 8A) and c-Myb(1-360) (Fig. 8B), some preference for MRE variant (GG Ͼ GT ϳ TG) was evident at a low input of expression construct (1 g). This result shows the same direction of selectivity as R 2 R 3 VP16 expressed at a low copy number in yeast (Fig. 3A), although less pronounced. However, at increasing input of effector plasmid, the preference of c-Myb(FL) and c-Myb(1-360) for MRE sequence was rapidly lost. Transactivation by c-Myb(1-360) was slightly higher than with c-Myb(FL), but the difference between the two is minor compared with other reports (39,40) Surprisingly, the c-Myb(FL) and c-Myb(1-360) proteins had marked effects on transactivation of the pGL2/tk/3xTT reporter plasmid. Transactivation of the pGL2/tk/3xTT reporter plasmid increased rapidly with a high input of Myb expression plasmids. This effect cannot be due to specific binding to the TT MRE variant, because the EMSA (Figs. 1 and 5-7) clearly showed that the TT oligo was not bound by any of the expressed Myb proteins. This result clearly differs from the transactivation data with R 2 R 3 VP16, where transactivation from the TT reporter is minimal over the same range of input effector vector (Fig. 2). However, these results may not be directly comparable, as Myb FL and 1-360 contains the native c-Myb transactivation domain, whereas R 2 R 3 VP16 carries the herpes simplex virus VP16 transactivation domain. This difference in transactivation levels may reflect the observation that the c-Myb transactivation domain but not the VP16 TA domain could have effect(s) on transcription/transactivation independent of specific DNA binding. These data emphasize the importance of the c-Myb expression level as a major determinant of target site selectivity. At low expression levels, MRE selectivity is better or comparable to in vitro results. At high expression levels, the selectivity is worse than the in vitro results: fine discrimination between MRE sites is lost, and the negative control reporter (pGL2/tk/3xTT) is misleadingly transactivated to a high level.
Finally, to assess the relevance of this conclusion, we evaluated the selectivity of c-Myb in naturally Myb-expressing cells. The c-Myb-expressing cell line K-562 was electroporated with pGL2/tk/3xMRE luciferase reporter plasmids or the basal level control pGL2/tk. In Fig. 9, a high selectivity for MRE variant sequences was observed with a strong preference for the GG sequence over the GT and TG MRE variants. The strong preference for the GG MRE oligo was evident in this cell type, very similar to the results obtained with a low copy number Myb effector plasmid in S. cerevisiae (Fig. 3A). This result further confirmed our hypothesis of the Myb expression level as a major determinant of Myb target site selectivity in vivo. We also included an AMV v-Myb expression construct in one experiment to test whether the transactivation response could be increased by a higher expression level of Myb protein or whether the transactivation levels with the 3xMRE reporter plasmids already were at a maximum. Addition of v-Myb increased the transactivation response further by 4-fold (Fig. 9, right column) compared with the endogenous transactivation level of pGL2/tk/3xGG, thus demonstrating that the observed transactivation level reflected the endogenous c-Myb protein level rather than the input of reporter plasmid DNA.
In conclusion, the transactivation data show that the target site selectivity of the Myb proteins c-Myb(FL) and c-Myb(1-360) is strongly influenced by the expression level of the Myb effector protein, both in S. cerevisiae and in mammalian cells.
Furthermore, the expression level of Myb protein may severely affect the outcome of the effector/reporter transactivation assay. DISCUSSION We have investigated the basis for the paradox between the broad MRE sequence selectivity of c-Myb R 2 R 3 in vitro and the highly restricted preference of R 2 R 3 VP16 protein for MRE target sequence in a S. cerevisiae transactivation system (23). We have shown that this striking difference could not be attributed to fusion of the VP16 transactivation domain to R 2 R 3 used in the in vivo studies. We found only subtle differences in the MRE preference of R 2 R 3 VP16 compared with recombinant R 2 R 3 in vitro, implying that the addition of the VP16 transactivation domain did not functionally affect the neighboring DBD. We also investigated the DNA binding properties of longer c-Myb proteins, and found that the in vitro MRE selec- tivity of c-Myb(FL) and C-terminally truncated c-Myb(1-360) was very similar to that of both R 2 R 3 VP16 and R 2 R 3 . In general, only minor differences were found between the in vitro DNA binding properties of c-Myb(FL) and C-terminally truncated c-Myb(1-360) proteins. However, we observed that c-Myb(FL) consistently bound slightly less well to the TG oligo (MBS-I-like) than the GG oligo (mim-1 A-like), compared with the other expressed Myb proteins. This difference was also reflected in the transactivation data, where a MRE TG variant reporter plasmid was transactivated more poorly than a MRE GG variant (see below).
Several studies have reported that a C-terminal truncation of c-Myb resulted in an increase in the DNA binding activity (34,35,38). It has been suggested that this increased DNA binding also could contribute to the transforming phenotype of v-Myb (34). Other studies have demonstrated that truncation of the C-terminal tail did not increase the DNA binding capacity of c-Myb tested in EMSAs on the mim-1 A binding site (GG-type) (24,36), an MRE sequence motif derived from DNA (TG-type) (41), or a mim-1 A-related sequence motif (GG-type) in the c-myc promoter (42). Our data with c-Myb proteins expressed in mammalian cells agree with the latter studies in that there is no major difference in the DNA binding activity between full-length and C-terminally truncated c-Myb. However, we do find subtle differences that are dependent on the MRE target site. Taken together, our results suggest that the ability to discriminate between MRE sequences is an intrinsic property of the c-Myb DNA-binding domain only, independent of other domains in the c-Myb protein.
A clue to resolution of the in vitro/in vivo DNA target specificity paradox was the observation that in the yeast system, the strong MRE sequence selectivity in vivo was highly dependent on the physiological expression level of Myb effector protein. The ability of c-Myb proteins to transactivate selected MRE target sites in mammalian cells was also found to be strongly dependent on the expression level of c-Myb effector protein, both in a effector/reporter system (CV-1) and in an intrinsic c-Myb-expressing cell line (K-562). At a low input of Myb effector in CV-1 cells and in c-Myb ϩ K-562 cells, transactivation of MRE reporter plasmids was dependent on the position 5 and 6 nucleotides (see Table I) in the order GG Ͼ GT ϳ TG, in accordance with results from the yeast system, with the c-Myb effector being expressed at a physiological level. At high expression levels of Myb effector proteins, this discrimination

FIG. 9. Discrimination between Myb recognition elements in
Myb-expressing cells. K-562 cells were electroporated with pGL2/tk alone or with pGL2/tk/3xMRE reporter plasmids (10 g) as indicated. Results are expressed as mean fold induction of luciferase activity over the pGL2/tk plasmid and S.E. (n ϭ 4). On the right, 5 g of expression plasmid for AMV v-Myb was cotransfected with the pGL2/tk/3xGG reporter plasmid (one experiment). between MRE sequences was lost, both in mammalian cells and in the S. cerevisiae system. At very high input levels of c-Myb proteins in CV-1 cells, the 3xTT MRE reporter plasmid was transactivated to a high level, even though this MRE was not bound by Myb proteins in EMSAs. This effect was particularly prominent for full-length c-Myb. Our findings suggest that activation of target genes by c-Myb may greatly depend on the expression level of Myb proteins in the cell, a notion also suggested by others (38). It has been difficult to define bona fide target genes for c-Myb in vivo. Several in vitro methods have therefore been employed to predict MRE target sequences. An important question is whether these in vitro methods are able to accurately predict in vivo Myb target sites. Several in vitro studies have demonstrated that the minimal Myb DNA-binding domain R 2 R 3 is able to accommodate a number of changes in the DNA MRE second half-site (14 -16, 18, 23). In binding site selection studies with random oligos, the MRE consensus sequence was determined as YAACKGHA (15) or YAACKGHH (16). In these studies, the (T/C)AACGG sequence was prevalent in selected single MRE-site oligos, 70% (15) and 38 -70% (16), depending on the Myb protein used. Efforts have also been made at thermodynamic measurements (43) and free energy matrixes (44) as aids in predicting MRE sites in vitro. However, these methods were based only on the MBS-I (TG-like) version of the MRE, which we find a suboptimal Myb binding site in vivo. In well characterized Myb target gene promoters (Table II), Myb binding sites with a G in both positions 5 and 6 is also strongly represented. In our study, the bona fide Myb MRE site TA-ACGG from the mim-1 promoter (27) was highly preferred in transactivation assays compared with the MBS-I-like sequence TAACTG, both in yeast and in mammalian cells with moderate expression of Myb effector protein. Taken together, there is good correlation between Myb MRE sites in Myb target genes and the preferred MRE sites in both mammalian cells (low effector input level) and in our model in vivo transactivation system in S. cerevisiae.
A general problem in the analysis of gene promoters is that putative transcription factor binding sites are present at a much higher frequencies than the expected number of the target genes for that particular transcription factor. This paradox is partially resolved by the concerted action of several transcription factors supplying the specificity of the particular gene promoter activity. Another possibility, illustrated in this study, is that the activity of transcription factor binding sites may be more stringent in vivo than suggested by in vitro analysis methods. A general weakness of transient transfection studies in mammalian cells is the lack of control of the test gene expression level on a per cell basis, and conditions may easily become nonphysiological, as demonstrated in this study. Our model transactivation system in S. cerevisiae has several promising features. First, our data demonstrate that c-Myb R 2 R 3 VP16 fusion protein expressed in yeast has DNA binding properties very similar to full-length c-Myb and is able to mimic the MRE sequence selectivity of full-length mammalian c-Myb in the yeast effector/reporter system. The R 2 R 3 VP16 fusion protein should therefore be a suitable model protein for measuring the DNA binding activity of c-Myb on putative Myb target DNA sequences independently of other domains in the c-Myb protein. Second, the system exploits features of yeast plasmids such as control of copy number (gene dosage), the use of metabolic markers to ensure the presence of effector and reporter plasmids in a homogenous yeast population (in contrast to cell-line-dependent variations and wide range of transfection efficiencies in mammalian cells), and the plasmids being in a physiological chromatin-packaged state (45). In summary, the advantage of the yeast system is that the expression level of the effector protein (transcription factor) is well controlled, thus avoiding overexpression and misleading transactivation of suboptimal target DNA-binding sites, as demonstrated in this study. Our yeast effector/reporter system may easily be applied as an in vivo tool for evaluating putative DNA-binding sites for other monomeric transcription factors.