|Home | About | Journals | Submit | Contact Us | Français|
Little is known about the mechanisms that regulate differential transactivation by p53. We developed a system in the yeast Saccharomyces cerevisiae that addresses p53 transactivation capacity from 26 different p53 response elements (REs) under conditions where all other factors, such as chromatin, are kept constant. The system relies on a tightly regulated promoter (rheostatable) that can provide for a broad range of p53 expression. The p53 transactivation capacity toward each 20- to 22-bp-long RE could be ranked by using a simple phenotypic assay. Surprisingly, there was as much as a 1,000-fold difference in transactivation. There was no correlation between the functional rank and statistical predictions of binding energy of the REs. Instead we found that the central sequence element in an RE greatly affects p53 transactivation capacity, possibly because of DNA structural properties. Our results suggest that intrinsic DNA binding affinity and p53 protein levels are important contributors to p53-induced differential transactivation. These results are also relevant to understanding the regulation by other families of transcription factors that recognize several sequence-related response elements and/or have tightly regulated expression. We found that p53 had weak activity towards half the apoptotic REs. In addition, p53 alleles associated with familial breast cancer, previously classified as wild type, showed subtle differences in transactivation capacity towards several REs.
Complex modulation of gene expression is an integral component of specific cellular responses to environmental stresses. This is generally achieved through the activation and regulation of transcription factors by the integrated action of multiple signaling pathways (78). The p53 protein is a remarkable example of a biochemical signal integrator whose transactivation activity is critical in the prevention of cellular transformation (46). Its function may be altered in nearly all human cancers, and an average of 50% of tumors express mutated forms of the protein with single amino acid changes that in some cases retain native conformation and partial function (10, 48).
Wild-type p53 can modulate the transcription of a vast number of target genes that play a role in diverse processes, including cell cycle control, apoptosis, senescence, differentiation, and DNA repair as well as its own turnover and activity (77). Differential transactivation by p53, in terms of both selectivity and kinetics of gene activation and repression, is important in determining which of these various processes is elicited, particularly with respect to the choice between cell cycle arrest and apoptosis (41, 69). However, little is known about the mechanisms that determine the in vivo specificity of p53 responses (44, 78).
Specific DNA binding by wild-type p53 is essential to its role as a tumor suppressor. This view is supported by the presence of p53 response elements (REs) in p53-regulated genes, the strong selection for inactivation of transactivation function during tumorigenesis, and the spectrum of p53 mutations observed in cancer cells (43, 79). The consensus sequence of a p53 RE consists of two copies of a degenerate 10-bp motif, 5′-RRRCWWGYYY-3′, with a spacer of up to 13 bp (21) that provide for the binding of a p53 tetramer (50). Many potential p53 targets are being identified through genome-wide sequence analysis and expression profiling in response to expressed p53 (40, 80, 85, 88).
There are many factors that might contribute to differential transactivation of various genes by p53. These include the individual sequence differences between REs, number of REs, and location of an RE with respect to the transcription start site (20, 62). The amount of nuclear p53 protein and the regulation of its DNA binding capacity may also be factors in differential transactivation (15, 63, 88). In addition to p53 levels (1, 58), posttranslational modifications may directly influence p53 DNA binding specificity, providing a link between a particular stress signal, an upstream p53 modifier, and an induced response (2, 13, 19). Cell type and physiology could also modulate p53 activity through gene-specific or general coactivator proteins (27, 65).
The roles of the structural organization of the promoters containing p53 REs and locus-specific features of chromatin assembly and remodeling in regulating p53 transactivation have recently been investigated. p53 is latent (needs posttranslational modifications) for binding with an RE in naked DNA but is active for in vitro binding to the same RE when chromatin is present (23). Based on in vivo chromatin immunoprecipitation, p53 is an active DNA binding protein (39, 69). In addition, acetylation of chromatin is important for p53 transactivation (3), while acetylation of p53 does not influence its DNA binding activity. The p53 protein may bind to a RE and recruit coactivator or corepressor proteins whose chromatin-remodeling activities (e.g., histone acetylation and deacetylation) modulate transcription (23, 49, 81). Although the actual impact on gene expression is likely to be influenced by target locus-specific features, p53 protein availability and intrinsic relative affinity of binding to p53 REs can be anticipated to play an important role in achieving differential transactivation.
In this work we address the role that the RE sequence and protein level can have in the differential transactivation by p53 at different REs. To accomplish this, we developed an in vivo system based in the model eukaryote, the yeast Saccharomyces cerevisiae, that relies on (i) a wide range of p53 protein expression levels from a tightly regulated (rheostatable) promoter, (ii) the use of a p53-dependent reporter gene located in single copy at its native chromosomal locus, and (iii) comparison of the transactivation activities of 26 responsive promoters, each comprising a single p53 tetramer binding site RE in an otherwise isogenic context. The functional analysis system has been extended to four p53 alleles associated with familial breast cancer that appeared to be indistinguishable from wild-type p53 in previous mammalian and yeast assays (11, 67).
The S. cerevisiae yAFM-RE strains were constructed from strain yIG397 (MATa ade2-1 leu2-3,112 trp1-1 his3-11,15 can1-100 URA3::3xRGC::p-cyc1::ADE2::ura3-1) (24) as illustrated in Fig. Fig.1.1. First we removed the p53 reporter construct integrated at the URA3 locus by selecting Ura− cells on 5-fluoro-orotic acid plates (Toronto Research Chemicals Inc., Toronto, Ontario, Canada). The loss of the ADE2 cassette was confirmed by colony PCR. Genomic DNA from strain yIG397 was used as a template to amplify the cyc1::ADE2 promoter-reporter construct by using a pair of primers with tails containing restriction sites for cloning. The cyc1 promoter fragment corresponds to chromosome X positions 525782 to 526025. The resulting 2-kb fragment was then cloned into plasmid pUC18 to generate pUC-cyc1ADE2. This was used as a template to generate a PCR fragment targeting the yeast ADE2 locus and containing the cyc1 minimal promoter and the wild-type ADE2 open reading frame, but no p53 RE. This fragment was integrated by exploiting homologous recombination at chromosome XV between positions 564474 and 566541. Besides the ade2-1 open reading frame, 350 bp of the ADE2 promoter upstream of the transcription start site (position 566190) was replaced by the integration. This strain, named yAFM, was red on plates with low adenine levels, indicating that upstream activating sequences were required for effective ADE2 transcription. Next we used a recently developed in vivo site-specific mutagenesis system known as delitto perfetto (68) to create p53-responsive cyc1 promoters. The CORE cassette (CO, counterselectable, KLURA3; RE, reporter, KanMX4, G418 resistance gene) was amplified and integrated immediately upstream from the minimal cyc1 promoter. Finally, the CORE cassette in the strain now named yAFM-CORE was replaced by using a pair of oligonucleotides containing the 20-bp p53 RE flanked by 40-bp homology tails. In order to insert p53 REs longer than 20 bases (i.e., with spacers or more than one tetramer binding site), we used a pair of oligonucleotides with a 30-base overlap at their 3′ sites that were extended in vitro before transformation in yeast as described previously (68). CORE replacements were confirmed by colony PCR and automated DNA sequencing (Applied Biosystems, Foster City, Calif.). Each isogenic yAFM-RE strain is identified by the name of the specific p53 RE (Table (Table1).1). All yAFM-RE strains without p53 expression vector were red on plates with low adenine levels. The sequences of all primers and details on strain construction are available upon request.
The transactivation of the ADE2 gene, which is dependent on p53 activity, results in the color of colonies changing from red to pink and to white (24). Plasmid pLS89 (66), coding for the human wild-type p53 cDNA under the inducible GAL1-10 promoter, was introduced into the yeast strains by lithium acetate transformation (29). Transformants were purified onto selective glucose plates (lacking tryptophan) containing a high adenine level to overcome the growth defect caused by incomplete adenine biosynthesis. The transactivation assay was then performed on plates with a low adenine level (5 mg/level) to allow color detection, using three independent transformants of each yAFM-RE strain. In order to modulate the expression of p53 protein from the GAL promoter, we used a combination of different sugars in the plate assay. Two percent raffinose (Sigma, St. Louis, Mo.) was used as a basic carbon source to keep the promoter in a derepressed state; variable amounts of galactose (Pfanstiehl, Waukegan, Ill.) were added to induce different levels of transcription. A glucose plate was also used to achieve minimal p53 expression (the GAL promoter is actively repressed on glucose). Thus, each transformant was streaked to 12 different low-adenine plates allowing different amounts of transactivation (see Fig. Fig.3).3). The plates were incubated at 30°C for 3 days before the colony color was scored.
To verify that variable amounts of galactose led to a fine-tuning of p53 protein expression, yAFM transformants were grown overnight in selective glucose medium. Cells were washed and diluted to 106/ml in raffinose-galactose medium, using the same concentrations as in the plate assay (see Fig. Fig.22 for concentrations). A high concentration of adenine (200 mg/liter) was used to allow for equal growth of all strains irrespective of the extent of p53-induced ADE2 transactivation. Proteins were extracted measured and blotted as previously described (37). p53 was detected by using pAb1801 and DO-1 monoclonal antibodies (Santa Cruz, Santa Cruz, Calif.). Relative p53 protein induction was determined by film scan and densitometric quantification with the free National Institutes of Health image software Scion Image 1.62.
To quantitate p53-dependent transactivation, the ADE2 mRNA was measured from aliquots of the same cultures used to prepare the protein extracts. The ratio between ADE2 and ACT1 (actin) mRNAs was determined by using quantitative reverse transcription-PCR (RT-PCR) approaches. Total RNA from 4 ml of overnight cultures was prepared by using the RNAeasy kit and protocol (Qiagen, Valencia, Calif.). The RNA preparations were treated with DNase I to eliminate any possible DNA contamination. Total cDNA was prepared from 200 ng (final concentration, 20 ng/μl) of total RNA using random hexamers (Applied Biosystems, Foster City, Calif.) and MultiScribe reverse transcriptase (Applied Biosystems). Different amounts of cDNA (2 μl of 1:100, 1:10, and 1:1 dilutions) were used for the specific amplification of ADE2 and ACT1. The primer pairs were chosen by using the ABI PRISM 7700 sequence detection system software (Applied Biosystems). Quantification of the ADE2/ACT1 ratio was obtained by using both the SYBR Green kit (Applied Biosystems) and TAQMan chemistry (Applied Biosystems). Both protocols gave overlapping results. The first method quantifies the amount of amplification but cannot discriminate between specific and nonspecific PCR products. The second procedure exploits two fluorescent probes that are specific for the templates under examination, allowing for multiplex PCR. The amount of amplicons produced is measured in real time (cycle by cycle) in the ABI 7700 thermal cycler (Applied Biosystems), so that the linear fraction of the amplification reaction is used for quantification. The ADE2/ACT1 mRNA ratio was obtained by the comparative CT method (CT is the threshold cycle; it indicates the fractional cycle number at which the amount of amplified target reaches a fixed threshold) by using the CT values for the VIC and FAM probes calculated from the linear fraction of the amplification reaction, corrected by the spectral compensation for multiplex PCR (Applied Biosystems). The ΔCT values for the three different cDNA dilutions used were independently computed and averaged. The standard deviations presented in Fig. Fig.44 represent the variation observed between three independent RNA-cDNA preparations examined by quantitative PCR. The sequences of all primers and probes and details of the amplification reactions are available upon request.
We sought to develop an in vivo system to systematically evaluate the contributions of RE sequence and p53 protein level to differential transactivation. This required the ability to assess the capacities for transactivation by various amounts of p53 protein towards many REs inserted at the same position within a reporter gene. Yeast-based p53 functional assays have provided valuable tools for screening tumor samples, classifying mutant p53 alleles, and structure-function studies (8, 9, 14, 38). High-level, constitutive expression of wild-type p53, but not tumor mutants, could activate a promoter containing a p53 RE in yeast (66). Using an inducible p53 expression system, we had previously identified various functional classes of p53 mutant alleles (36). Moreover, at low expression levels, wild-type p53 was more active with RGC and P21 REs than with a BAX RE (37). However, because of differences in strain backgrounds and structural arrangements of the REs, these comparisons were qualitative and an evaluation of differential transactivation could not be addressed.
To explore differential transactivation by p53, we developed a new set of isogenic haploid yeast strains based upon the ADE2 red-white p53 reporter system initially described by Flaman et al. (24). Presented in Fig. Fig.11 is a general system for incorporating REs into a promoter. We utilized our recently developed two-step in vivo mutagenesis system, termed delitto perfetto (68), to rapidly generate a panel of yeast strains with modified p53-responsive CYC1 promoters. Briefly, the wild-type ADE2 open reading frame was integrated into the chromosome, replacing the ade2,1 allele in the yIG397 background (see Materials and Methods). The ADE2 promoter was also replaced with a minimal CYC1 promoter that lacked upstream activating sequences. Hence, ADE2 transcription is inactive, and the strain, named yAFM, is red in medium containing low levels of adenine due to the interruption of adenine biosynthesis at the step prior to the one catalyzed by Ade2. Applying the delitto perfetto approach, a CORE cassette (CO, counterselectable marker; RE, reporter gene) was inserted by homologous recombination upstream of the minimal CYC1 promoter, and then the cassette was replaced by using double-stranded oligonucleotides containing a specific p53 RE. Using this approach, we created 22 p53 REs from mammalian p53-regulated genes plus four artificial sequences (Table (Table1).1). The correct RE insertions were confirmed by DNA sequencing. At least two independent isolates for each RE, named yAFM-RE, were tested in the functional assay. With the exception of the c-fos RE (i.e., yAFM-cFOS), all of the RE sequences cloned upstream of the CYC1 promoter contained a single 20- to 22-bp p53 tetramer binding site. Most of the REs have no spacer between the dimer binding sites. Table Table11 describes the actual sequence of each element along with the mammalian gene name from which it was derived.
A single-copy centromere plasmid, pLS89, containing the human wild-type p53 cDNA under control of the GAL1-10 promoter was transformed into the yAFM-RE strains. The GAL promoter is repressed on glucose, while on raffinose a derepressed, higher basal transcriptional state is achieved (34). The p53 level on glucose is about 15 times lower than that on raffinose (data not shown). When galactose is added to the raffinose medium, the promoter can be induced to a high level. In order to establish that the system was rheostatable for p53 protein expression, we determined p53 levels after growth in medium containing raffinose plus low levels of galactose (Fig. (Fig.2).2). The amount of p53 was quantified by using densitometric scans of Western blots (Fig. (Fig.2A).2A). The levels of p53 protein were examined in five isogenic yAFM-RE strains, providing a mean relative p53 expression as a function of galactose concentration (Fig. (Fig.2B).2B). The induction of p53 expression was approximately linear over the range of galactose concentration from 0 to 0.12%. The p53 level after induction with 2% galactose (see also Fig. Fig.5B)5B) increased only two to three times relative to that with 0.12% galactose.
The relative abilities of p53 to transactivate the reporter gene at various REs were examined by using different levels of galactose inducer and ranking the amount of galactose inducer required for colonies to become pink or white. The various yAFM-RE transformants containing the p53 expression vector were streaked out for single colonies onto a series of 12 plates (which select for the vector) containing either glucose, raffinose, or raffinose plus different amounts of galactose (increasing by factors of 2, i.e., 0.001, 0.002, 0.004, 0.008, 0.016, 0.032, 0.12, 0.25, 0.5, and 2%). A small amount of adenine (5 mg/liter) was included to enable growth of the ade2 mutant cells to small red colonies if there was insufficient transactivation. Transactivation of ADE2 by p53 results in pink and white colonies, depending on the extent of induction. If the transactivation capacity with all p53 REs is the same, colony colors would be expected to change from red to pink to white at comparable levels of galactose. Instead, the responses often differed between REs based on level of galactose required to produce pink or white colonies. For example, the yAFM-P21-5′ strain was nearly white on glucose plates, as seen in Fig. Fig.3,3, indicating high activity by basal levels of p53 toward this element. Strains such as those containing m-cFOS or IGF-BP3 B remained pink or red at maximum p53 expression (2% galactose), as shown in Fig. Fig.44.
The 26 yAFM-REs could be ranked for ability to be induced by different amounts of p53 by using this phenotypic color assay. Since the relative p53 expression was measured under the same culture conditions and appeared to be linear over a broad range (r2 = 0.987) (Fig. (Fig.2B),2B), the relative amount of p53 protein required for a phenotypic change (pink or white) could be determined. This allowed us to rank p53 transactivation capacity towards the 26 REs as the minimum level of p53 required to produce changes in colony color. (As described below, color correlates well with level of ADE2 transcription.) Furthermore, since all strains were isogenic, the transactivation capacities directly reflect the combination of the DNA binding affinity and the ability to transactivate from individual REs by p53.
The functional ranking of p53 transactivation at the various REs of p53-regulated genes is presented in Fig. Fig.4;4; P21-5′is the strongest, and the IGF-BP3 box B is the weakest (the Con and Mut-RGC response elements are described below). In addition to P21-5′, five yAFM-REs (p53R2 to hFAS) showed some transactivation at the basal levels of expression on glucose-only plates. All other strains were red on both glucose and raffinose. For 15 of the yAFM-REs (p53R2 to MDM2/RE1), white colonies could be detected at some level of p53 induction. For 12 of the latter (p53R2 to cyclin G), colonies appeared white (or nearly white as in the case of hFAS, PUMA, PCNA, and cyclin G) when there was less than a 20-fold induction of p53 protein relative to the level on glucose. The yAFM-NOXA, -P21-3′ and -MDM2/RE1 elements enabled some p53 transactivation with a 20-fold relative increase in induction of p53, based on the appearance of pink colonies; much more induction was required to produce white colonies. The remaining nine elements (PA26 to IGF-BP3 box B) did not enable high levels of transactivation, since the colonies did not turn white even under conditions of maximum p53 expression. Except for the IGF-BP3 box B RE, which was completely inactive, the colonies exhibited a light to dark pink color on 2% galactose.
The red-pink-white phenotypic assay for assessing p53 gene-specific transactivation capacity and the impact of different levels of expression is simple and highly reproducible. However, the mechanism of red pigment accumulation and metabolism by the ADE2 gene product to result in white colonies is unclear and might be affected by processes that are independent of p53. We therefore addressed the relationship between colony color and ADE2 transcription. The ADE2 mRNA levels were measured relative to actin mRNA (ACT1) by using a real-time RT-PCR approach (see Materials and Methods). First, we checked that there were comparable levels of ADE2 expression from various REs in the absence of p53 protein. Two of the strongest (yAFM-P21-5′ and -p53R2) and the two weakest (BAX-A and IGF-BP3 box B) p53-responsive strains were compared. The red yIG397 strain with a wild-type ADE2 promoter in front of the ade2,1 allele was also included (Fig. (Fig.5A).5A). Total RNA was prepared from cultures grown in 2% raffinose medium for 24 h. Sufficient adenine was added to allow normal growth and prevent a selection for enhanced ADE2 expression. The uninduced ADE2 mRNA levels were comparable and reduced about 10 times for the p53-responsive promoters compared to the natural ADE2 promoter. The actin mRNA levels were similar for all cultures.
The levels of mRNA induced by p53 at various REs reflected colony color. p53-dependent ADE2 transactivation was measured with low (raffinose) and high (galactose) p53 expression. Total RNA was prepared from liquid raffinose cultures (i.e., basal level of induction) of yAFM-RE strains that produced red (RGC, PCNA, MDM2-RE1, BAX-A, p21-3′ cyclin G, and AIP1), pink (GADD45), and white (P21-5′ and p53-R2) colonies on this medium (Fig. (Fig.5B).5B). Consistent with the phenotypic assay, the ADE2/ACT1 ratio was about five times greater for the two white strains than for the red strains. However, the pink yAFM-GADD45 strain was not distinguishable from the red strains.
RNA was also examined following growth in 2% galactose for all but the RGC and p53R2 strains. As shown in Fig. Fig.5C,5C, the BAX-A strain, which is pink on 2% galactose plates (Fig. (Fig.4),4), showed the lowest ADE2 induction. All of the other strains were white on 2% galactose. However, there were differences in mRNA induction that ranged from 1.5 times higher for the yAFM-P21-3′ strain to 60 times higher for the P21-5′ strain compared to levels in the BAX-A strain. The other strains showed between 10 and 20 times higher levels of ADE2 mRNA. Since the p53 protein levels of cells grown on 2% galactose were comparable, the differences in expression must be due to differences in transactivation capacity (Fig. (Fig.5C5C).
Finally, we examined the relative ADE2 transcription from the P21-5′, GADD45, PCNA, MDM2-RE1, and BAX-A REs at different levels of p53 induction. Total RNA was prepared from the same cultures used for the Western blots presented in Fig. Fig.2A,2A, and the ADE2/ACT1 ratios are presented in Fig. Fig.5D.5D. The yAFM-BAX-A strain exhibited poor induction at all levels of p53. The P21-5′, GADD45, and PCNA strains had various degrees of strong induction of ADE2 at relatively low levels of p53 protein. Except for P21-5′, the induction of transactivation reached a plateau at high levels of p53 protein expression. The yAFM-MDM2-RE1 strain showed a modest stimulation of ADE2 transactivation up to a 100-fold increase in p53. This RE becomes highly responsive only with large amounts of p53 (i.e., 2% galactose). We conclude that there is a good correlation between the phenotypic assay and transactivation of the ADE2 gene.
We noticed that the 5′-CATG sequence at the junction between p53 monomer binding sites is found in both dimer sequences of the four strongest REs described in Table Table1:1: P21-5′, p53R2, m-FAS, and GADD45. This motif is present in only one dimer sequence in five other REs in Table Table1.1. To evaluate the contribution of this sequence element to p53 transactivation capacity, we examined 4 additional artificial sequences, described in Table Table11 as Con-A, Con-B, Con-C and Mut-RGC. The Con sequences correspond to the consensus RRRCWWGYYY. Interestingly, Con-A, which contains the CATG sequence in both dimer elements, showed the strongest activity of all of the REs, as shown in Fig. Fig.44 (the colonies were white on glucose plates). Con-B, which is identical to Con-A but has CTAG central elements, ranked sixth, with an approximate 20-fold reduction in transactivation capacity. There was less transactivation from Con-C (it ranked eighth), possibly because it lacks any CATG element. We also constructed a modified RGC RE (Mut-RGC), replacing the CTTG sequence element with CATG in both dimers. This RE showed a dramatic increase in activity.
The incidence of p53 mutations in familial breast cancer associated with germ line BRCA1/2 mutations is nearly 70%, compared to around 30% for sporadic breast cancer. Moreover, the spectrum of mutations appears to be different (30). Interestingly, a subset of BRCA1-associated p53 mutant alleles appeared wild type in mammalian functional assays (67), and four of the mutants, T150I, G199R, R202S, and S215C, were also wild type in a yeast transactivation assay at high p53 expression (11). We tested these mutants for the ability to transactivate from 10 of the REs in this study at various levels of p53 expression. All of the mutants were comparable to the wild type at high levels of p53 expression (data not shown); however, the matrix of responses differed from wild type at low levels. The difference in the minimal galactose amount needed to obtain white colonies (or light pink colonies with BAX-B) with mutant and wild-type alleles was used to estimate the relative transactivation capacity. Subtle differences in the transactivation pattern were detected for all four alleles (Table (Table2).2). Interestingly, the changes comprised both reduced and enhanced activity with specific REs and no effect with other REs relative to the wild type, suggesting that the amino acid changes do not simply affect protein stability.
What is the role of p53 intrinsic DNA binding capacity in differential transactivation? Since differential modulation of transcription of many genes by a common regulatory protein can provide for adaptive cellular responses, particularly in response to environmental stress, it is important to understand the underlying mechanisms. There are many factors that can influence the specificity, levels, and kinetics of transcriptional changes of various genes. As described in Fig. Fig.66 they include the amount and activity of the transactivation protein (e.g., p53), the availability of cofactors, the chromatin organization of the gene locus, and the actual sequence of the target promoter elements. A minimal functional unit that is required, but is not sufficient, for transcriptional activation is the transcription factor bound to DNA. The specific DNA binding activity of factors such as p53 may facilitate the recruitment of chromatin-remodeling systems that in turn modulate the efficacy of the assembly of the transcriptional machinery and of transcription initiation and elongation (17, 59).
A fundamental question that has not been adequately addressed with in vivo systems concerns the role of intrinsic DNA binding specificity in differential transactivation (5, 47, 61). A common feature of several families of response elements is variation in nucleotide sequence, reflected by a loose consensus within a family (51, 60). Hence, the relative binding affinity or transactivation potential of specific promoter elements by a common regulator might allow for flexibility in cellular responses. This is particularly relevant in the case of the p53 transcription factor, whose nuclear levels increase in response to cellular stresses and which can directly affect transcription of many genes with different functions. It is also relevant in light of the many mutant p53 alleles selected and expressed in tumor cells that appear to retain partial activity (11, 28, 67) (F. Storici et al., unpublished data).
Based with systems that utilize high constitutive p53 expression, there is an abundance of anecdotal evidence of differential transactivation capability in mammalian cells and in yeast by wild-type p53 or p53 mutants (18, 62, 72). Genome expression analyses have shown that there are large differences in transactivation (and repression) of the many p53 target genes in response to different levels of ectopically expressed wild-type p53 or in response to the induction of endogenous p53 by various sources of DNA damage (40, 41, 64, 88). However, the source and degree of differences have not been clear.
Most of the studies addressing p53 binding to REs and its regulation, which are based on in vitro experiments with naked DNA, are supportive of a general model whereby p53 in unstressed cells is a latent DNA binding protein that requires posttranslational modification for activation (35, 62, 71). However, these assays may not accurately mimic the nuclear environment. Recent in vivo studies that utilize chromatin immunoprecipitation have shown variation in promoter occupancy by a constitutively active p53 protein in unstressed cells at a small number of targets. Induction of p53 by genotoxic stress led to an increase in DNA binding that correlated with the increase in protein amount with no evidence of changes in affinity (39, 69). However, the correlation between level of promoter occupancy and strength of transactivation by p53 was not precise and is likely influenced by specific features of the target loci. With this in mind, we sought to develop a sensitive in vivo system that could address the importance of both p53 levels and RE sequence in differential transactivation. To do this, we kept constant all other potentially modifying factors (Fig. (Fig.6),6), such as number of REs, chromatin structure, coregulators, and p53 posttranslational modifications, by constructing a panel of isogenic yeast strains, each containing a p53-dependent reporter gene, that differ only in the sequence of the p53 RE.
Using our model system, we have established that both the RE sequence and the levels of p53 induction are important factors in determining which RE is activated and to what extent. The wide linear range of protein expression (Fig. (Fig.2)2) allowed us to distinguish both subtle and dramatic differences in relative p53 transactivation capacities among 26 p53 REs. The difference in p53 protein amount required to achieve the same transactivation with two REs provided a measure of relative transactivation capacity. This highly reproducible phenotypic assay for p53 transactivation function was validated by mRNA measurements. Since we used isogenic strains and conditions, the transactivation differences directly relate to DNA binding of p53 with its REs.
Our results support the dose-response model of p53 regulation in mammalian systems (44), in which various biological responses are dependent on the level of available nuclear p53. We could not correlate transactivation capacity with gene groupings in functional pathways. However, the weak transactivation group is enriched with REs from apoptotic genes, which is consistent with results in mammalian cells, where induction of the apopotic pathway appears to have a higher threshold of activation (15, 63, 78). Also, the weak transactivation potentials of MDM2-RE1 and -RE2 are consistent with the important role of this gene as a p53 repressor, and they parallel results obtained with mammalian cells (31, 45, 74).
The wide differences in transactivation capacity, as much as 1,000-fold, were unexpected. The patterns could be divided into three groups. The strongly binding group, where white colonies could be detected at low levels of galactose, includes the REs from Con-A to cyclin G in Table Table11 and Fig. Fig.44 (strongest to weakest). Presumably, p53 tetramers can efficiently form at these sites and lead to transcription. The next group, from NOXA to MDM2-RE2, showed modest induction over a broad range of p53 expression. For the yAFM-MDM2-RE1 strain, comparable levels of ADE2 mRNA were detected over a 100-fold induction of p53, and strong induction was observed only on 2% galactose. Six REs, from cFOS to IGF-BP3 box B, led to poor transactivation even at the highest levels of p53 (Fig. (Fig.4).4). This suggests that factors other than RE sequence, such as chromatin state, distance to or from the transcription start site, and availability of coactivators, might be important for transactivation from these promoters in mammalian cells (65). These REs may be ideal for identifying such factors in a yeast-based system. In addition, efficient transactivation of a subset of target genes by p53 may require the function of the p53-related p63 and p73 genes (26). Interestingly, in mouse embryo fibroblasts such cooperation appeared to be specifically required for the induction of a subset of p53 targets whose REs scored as intermediate or weak in our yeast-based functional rank (BAX and NOXA) (Fig. (Fig.4)4) (26). We are currently evaluating the transactivation capacities of p63 and p73 isoforms and the effect of their coexpression on p53 function by using the functional assay at low or variable expression levels (unpublished results).
To better compare the relative activities, we chose to focus this study on p53 REs containing a single tetramer binding site (with the exception of cFOS), even for those cases of human genes that have more than one RE. Perhaps more than one RE is required for transactivation of genes that contain a poorly responding RE. The presence of two closely spaced REs, as found, for example, in the human mdm2, bax, and IGF-BP3 genes, could result in a cooperative effect to make the promoter responsive to p53. (The results with cFOS do not contradict this view, since transcription from just one cFOS tetramer was not examined.) In a separate study (unpublished), we examined transactivation from the MDM2-P2 sequence (where the two REs are separated by 5 nucleotides) and from BAX-A+B, IGF-BP3 box A+B, and 2× RGC containing two contiguous copies of the RE without spacer. Unlike the situations with a single RE, these strains turned white with modest to high levels of p53 expression, supporting the idea that adjacent REs can have a cooperative effect on transactivation capacity. BAX-A+B showed a profile of activity similar to that of cyclin G (11th in the ranking), while the other three strains were similar to NOXA (data not shown). However, the transactivation capacities of these complex REs were still significantly lower than those of the top-ranked single REs such as P21-5′, p53-R2, and FAS (Fig. (Fig.44).
Overall, the yeast-based results are consistent with electrophoretic mobility shift assays (EMSA) and chromatin immunoprecipitation (ChIP) analysis (e.g., P21-5′, P21-3′ BAX-A, BAX-B, MDM2, NOXA, PUMA, and PIG3) (39, 62, 69, 71, 82). The yeast-based ranking of transactivation capacities is also consistent with the limited gene expression studies that are available for comparison (e.g., p21, Gadd45, Pig3, and bax) (25, 40, 69, 88). Comparisons between the yeast results and published DNA binding measurements for nine REs are presented in Table Table3.3. There is a good qualitative correlation between relative in vitro binding activity and transactivation capacity. The correlation is less apparent for comparisons with ChIP results. However, this in vivo assay is not well suited to evaluate relative DNA binding capacities, since each RE is analyzed within different promoters where individual differences in chromatin structure are expected to influence p53 occupancy. The large difference in transactivation capacity between the two p21 REs in yeast is also consistent with analysis of in vitro DNA binding to chromatin, although there was only a small difference in p53 binding to the two sequences (23). Similar to results from that chromatin study, we also observed that the carboxy-terminal deletion (Δ368-393) leads to a reduction, rather than stimulation, of p53 transactivation capacities with all REs, especially for the weaker REs that include P21-3′ (unpublished results).
We have also applied our analysis at low or variable p53 expression levels to the evaluation of the role of the two distinct transactivation (TA) domains located in the p53 amino terminus. Deletion of the first TA domain (Δ1-39) resulted in an approximately 50% reduction in transactivation capacity with all 12 REs that were tested (unpublished results). With strong REs like P21-5′, p53R2, and GADD45, white colonies were obtained with moderate p53 expression. At high p53 expression, a defect in transactivation was visible only with intermediate and weak REs such as P21-3′, MDM2-RE1, and PA26. Consistent with this observation, a previous analysis at high p53 expression of a double mutant in the first TA domain (L22E/W23S) showed almost no affect on transactivation function, pointing to differences in the interactions with the transcriptional machinery between yeast and mammalian cells (18). We also tested single amino acid changes (W53R and F54S) at key residues in the second TA domain (12) and observed a reduction in transactivation capacity similar to that seen with the Δ1-39 mutant. Deletion of both TA domains (Δ1-65) instead completely abolished p53 transactivation capacity. These combined observations suggest that the TA domains have overlapping functions in yeast and contribute comparably to total protein transactivation activity, a feature that could be appreciated only at low or variable p53 expression and that is consistent with observations for human cells (89).
The posttranslational modification status of p53 protein expressed in yeast has not been reported. Hence, we cannot exclude the possibility that the functional rank we described results from a particular state of p53. In mammalian cells, it is not clear whether p53 posttranslational modifications play any direct role in p53 DNA binding specificity (1, 6, 19). It is important to note that neither an MDM2 homologue nor a p53 homologue has been detected in budding yeast. The transactivation capacity of p53 alleles with amino acid changes at sites of posttranslational modifications when expressed alone or in combination with the MDM2 protein is currently under investigation.
We sought to explain how the small differences in 20 bp of the REs could account for the vast differences in transactivation capacity. The number of nonconsensus bases in the RE sequences (Table (Table1)1) clearly does not predict the wide variations. For example, both P21-5′and IGF-BP3 box A have two nonconsensus bases at the first position of monomer binding sites, and they display a dramatic difference in transactivation.
A statistical method based on nucleotide usage has been developed to predict the relative activity of each RE (4). We computed the nucleotide usage (Table (Table4)4) of a large panel of p53 REs previously identified by Tokino et al. (73) using yeast-based transactivation at high p53 expression (T values). A total of 170 p53 dimer REs were reported in that study. We also used the published nucleotide usage obtained with 20 REs identified by in vitro binding assays (B values) (21). The nucleotide usage value (percent) for each position was used to compute the heterology index (HI) of each RE with an equation derived from statistical mechanics theory (4) (Table (Table1).1). Although some of the more active REs have lower HIs, particularly when the T values of nucleotide usage are used, we found that nucleotide usage approaches have low predictive value. For example the most active P21-5′ RE has only the 9th and 10th best HIs, respectively, with the T and B nucleotide usage values. Moreover, RGC is predicted to have a similar or higher activity than P21-5′, despite about a 1,000-fold difference in transactivation capacity. The statistical approach is based on the assumption of equal contributions of each binding unit (i.e., each nucleotide in the RE) to the affinity or activity of the complete RE. No additional correction factor is introduced to weight the position of each nucleotide within a p53 monomer binding site. An alternative approach would involve weighting the contribution of nucleotides directly involved in protein-DNA interactions, i.e., positions 3 and 4, as predicted by the crystal structure (16). Even though the functional rank is partially consistent with a limited role of the nucleotide at position 1 of the p53 monomer binding site and a higher impact of deviations at position 3 (e.g., RGC and Mut-RGC) (Table (Table1),1), the overall result cannot be completely explained by the position of the nucleotide that deviates from the consensus. For example the RE used in the crystal structure study (Con-C) has a perfect consensus sequence but ranked eighth in our transactivation assay, and the PCNA RE with a mismatch at the invariant fourth position still retains strong transactivation capacity.
Another interesting algorithm that identifies putative p53 REs and computes their binding probabilities by using a novel filtering and weight matrix was recently reported (33). No correlation was found with results in the present study. The algorithm attributed the same binding probabilities to IGF-BP3 A and H-FAS (scored as 90.9 and 90.92 out of 100), while the yeast-based functional assays indicated almost a 1,000-fold difference in transactivation capacity. MDM2-RE1 had a much lower score (79.68), while the order in the yeast assays was H-FAS, MDM2-RE1, and IGF-BP3 A (Fig. (Fig.44).
We hypothesize that sequence-dependent structural features of the DNA greatly affect p53 activity in our in vivo system, as previously observed in vitro with purified p53 DNA binding domain and naked DNA RE (54). It is possible that upon interaction with DNA RE sequences the p53-DNA complex undergoes conformational changes that significantly contribute to binding affinity and activity. For example, axial bending of the DNA facilitated by the flexibility of the CATG sequence (55) may lead to tighter p53 tetramer binding by allowing the establishment of intersubunit interactions in addition to those at the tetramerization domain (50). These changes might also make available the amino-terminal transactivation domains for protein-protein interactions with components of the transcription machinery. Our results with the consensus sequences and the modified RGC RE support this interpretation. Modification of Con-A (CATG) sequences into Con-B (CTAG) (Table (Table1)1) led to a considerable reduction in activity. Furthermore, modification of the RGC RE (CTTG into CATG, corresponding to Mut-RGC) resulted in a dramatic improvement of p53 transactivation capacity. The observation that subtle changes in RE sequence can lead to significant differences in the responsiveness of a target gene to p53 also suggests that interspecies, interindividual, or somatically acquired variations in RE sequences may lead to diverse p53-dependent responses, which has implications for cancer susceptibility and evolution.
The yeast functional assay that we described provides a versatile tool to evaluate the impact of p53 mutations on transactivation potentials and specificity. One implication of the important role of p53 intrinsic DNA binding affinity in differential transactivation is that specific p53 mutations may lead to defined changes in the transactivation patterns that are compatible or advantageous during tumorigenesis and could be selected in particular cellular or genetic environments. The rheostatable system along with many REs enables the functionality of various p53 mutations to be addressed.
Previously we found several alleles that were toxic at high expression levels but exhibited increased transactivation toward the P21-5′ RE at basal-level expression (37). The functional profile of these alleles with the complete panel of REs is being determined. The system might provide the opportunity to select p53 alleles that would have altered functional patterns in mammalian cells, such as apoptosis (39, 64). For example, a V122A mutant shows a 4- to 8-fold higher transactivation capacity toward P21-5′, GADD45, and hFAS and about a 16-fold higher capacity toward BAX-A+B and IGF-BP3 box A (unpublished results). Its activity was equal to that of wild-type p53 with p53R2, while it was reduced four- to eightfold with 14-3-3 and cyclin G and was undetectable with P21-3′, MDM2-RE1, and RGC (unpublished results).
The p53 mutant alleles detected in familial BRCA1/2-associated breast cancer are an important example of how the opportunity to address transactivation capacity over a wide range of p53 levels can be useful in understanding functional alterations. We analyzed four alleles that had been characterized previously in both mammalian and yeast assays at high levels of expression. They appeared to be indistinguishable from wild-type p53 (11, 67) in a number of in vivo functional assays that were used. As shown in Table Table3,3, the yeast assays at low and variable p53 expression provide for functional discrimination. While the differences were small (about two- to fourfold) and a common pattern of functional changes was not evident, both reductions and increases in transactivation capacity with defined REs were observed. These results may implicate subtle changes as having an important impact on the etiology of tumor development.
These observations with wild-type p53 and mutant alleles have general implications for studies with other families of transcription factors whose nuclear levels are tightly controlled and which recognize response elements that have loose consensus sequences, as for the case of p53. The system of tight regulation of protein expression, rapid construction of responsive promoters by using in vivo delitto perfetto site-directed mutagenesis (68), and phenotypic and quantitative assessment of transcription levels provides a framework in which to investigate a variety of issues that relate to transactivation specificity. The approach provides the means to study altered selectivity and differential transactivation in vivo as well as the isolation and characterization of enhancer elements and transcription factors and the screening for external modifiers of transcription complex assembly.
Our thanks go to Richard Iggo for the generous gifts of the yIG397 yeast strain and p53 expression vectors; to Gilberto Fronza for the p53 mutant alleles; to Chris Miller, Dmitry Gordenin, Kirill Lobachev, Yong Hwang Jin, and Karthikeyan Gopalakrishnan for advice, helpful discussions, and comments on the manuscript; and to Trevor Archer and Alex Merrick for critical review.
This work was partially supported by DOD grant DAMD 17-011-0579 to M.A.R.