Functional evolution of the p53 regulatory network through its target response elements

Abstract

Transcriptional network evolution is central to the development of complex biological systems. Networks can evolve through variation of master regulators and/or by changes in regulation of genes within networks. To gain insight into meaningful evolutionary differences in large networks, it is essential to address the functional consequences of sequence differences in response elements (REs) targeted by transcription factors. Using a combination of custom bioinformatics and multispecies alignment of promoter regions, we investigated the functional evolution of REs in terms of responsiveness to the sequence-specific transcription factor p53, a tumor suppressor and master regulator of stress responses. We identified REs orthologous to known p53 targets in human and rodent cells or alternatively REs related to the established p53 consensus. The orthologous REs were assigned p53 transactivation capabilities based on rules determined from model systems, and a functional heat map was developed to visually summarize conservation of sequence and relative level of responsiveness to p53 for 47 REs in 14 species. Individual REs exhibited marked differences in transactivation potentials and widespread evolutionary turnover. Functional differences were often not predicted from consensus sequence evaluations. Of the established human p53 REs analyzed, 91% had sequence conservation in at least one nonprimate species compared with 67.5% for functional conservation. Surprisingly, there was almost no conservation of functional REs for genes involved in DNA metabolism or repair between humans and rodents, suggesting important differences in p53 stress responses and cancer development.

Gene expression largely depends on combinatorial interactions of transcription machineries with regulatory modules and is influenced by changes in chromatin structure (1–3). Sequence-specific transcription factors (TFs) can modulate expression of networks of target genes through dynamic interactions with unique cis-regulatory response elements (REs). Transcription network evolution is an important genetic component in phenotypic diversification between species. Transcriptional networks can transevolve through variation of master regulators or their targets or they can cis-evolve by changes in regulation of target genes. Cis-regulatory evolution, which can arise from point mutations, small deletions and additions, or large rearrangements of promoter regions, may be a major driver of species phenotypic differences (4). However, sequence-based analysis of phylogenetic conservation within regulatory regions has limitations because of an overall promoter organization that is less conserved compared with coding sequences. Although conservation of cis-regulatory sequence motifs may indicate preserved function within transcriptional networks, there are few comparative studies that address quantitatively the evolution of these regulatory interactions and examine the actual functionality of diverged REs in terms of ability to recruit TFs and mediate changes in transcription rates of associated genes (5–8). The commonly observed large variability in individual REs for a given TF within an organism, as exemplified by degenerate consensus sequences, challenges the ability to predict quantitative functional consequences of diverged REs on promoter outputs.

We have addressed evolution of REs that provide direct transcriptional control by the sequence-specific TF p53, a tumor suppressor gene and master regulator of cellular stress responses. The study focuses on functional evolution of p53 REs along with sequence evolution and draws on earlier investigations of p53 transactivation capacity and the effects of subtle variations in RE sequences, such as single nucleotide polymorphisms (SNPs) (9, 10). Limited to human and mouse species, the work also utilizes previous analyses of transactivation potentials by the orthologous p53 proteins (11). We combined a customized bioinformatics approach for phylogenetic footprinting and motif scanning with predictions of functionality (i.e., transactivation potential in terms of “on” or “off” and level of response) for p53 REs based on a set of described RE rules (11). This led to a new functional heat map representation of RE functionality and sequence variation within and across species. There is widespread evolutionary turnover of p53 TF binding sites (TFBS) that is not fully predicted by simple sequence analysis. Importantly, the p53 target genes involved in DNA-associated metabolic activities in primates appear to have been incorporated into the p53 network after branching the rodent and primate evolutionary paths.

Results

Functional and Sequence Conservation of p53 REs Between Species.

We addressed evolution of the p53 master regulatory network through an analysis of 47 previously validated human and mouse p53 target REs. Specifically, we investigated the extent to which functional variation tracks with sequence variation. In fact, the consensus p53 RE is degenerate, and our previous studies revealed that RE function was weakly predicted by position weight matrix (PWM) sequence evaluation (12). Functionality in the present study was evaluated from a combination of sequence inspection, RE rules [Methods and supporting information (SI) Table 1], and direct determination of p53-mediated transactivation capacity in yeast and mouse cells.

Conservation was examined in up to 14 species, depending on retrievable information. Borrowing microarray expression data representations, the functional and sequence conservation are presented together in a heat map format that presents levels of functionality (predicted and/or measured) of the REs, as described in Fig. 1. Sequence conservation of REs is indicated by symbols (see SI Tables 1 and 2 for supporting data). The reference/seed human or mouse sequences for cross-species comparisons are shown as filled triangles. Most of these REs have been measured directly for transactivation. Predictions of RE functionalities (“nonresponsive” to “high”) are based on previous findings (10–15) (J. Jordan, D.M., A.I., M. Nourredine, D.A. Bell, and M.R., unpublished data) and are described in SI Tables 1 and 2. Functionality (“relative transactivation”) is normalized to human p21 RE-5′, which mediates strong transactivation by human p53 and binding in vitro (12, 16). For example, in comparison with the p21 RE-5′ (red), transactivation at the REs of human GADD45A and PCNA is ≈50% (yellow) and ≈30% (green), respectively. Species are ordered by evolutionary tree, and conservation of the DNA sequence-specific binding domain of p53 proteins is shown.

Fig. 1.

Functional heat map depiction of p53 RE conservation, where relative function is conveyed in a heat map format, and sequence conservation is described by symbols. A modified microarray heat map graphic style was used to represent the predicted transactivation potentials of p53 REs. The estimated strength is relative to the strongly transactivated human p21 RE-5′, where red is comparable, and yellow, green, and gray correspond to decreasing transactivation capabilities. Using human or mouse (red gene symbols) p53 REs as reference sequences/seeds (filled triangle), the orthologous sequences from 13 other species (depending on retrievable sequence) were extracted and sequence conservation based on number of mismatches (MM) relative to seed RE was determined (see SI Table 2). * indicates extent of “sequence conservation relative to the reference sequence” as depicted by the size of the 0; X indicates no conservation. A black box indicates no sequence information was available in the corresponding region of the examined species. For each RE, functionality was assigned based on a set of rules derived from yeast-based assays summarized in the text and in SI Table 1 (# indicates reference sequences not experimentally tested in yeast). Species are ordered according to the evolutionary tree, and percentage of conservation of the DNA binding domain of p53 proteins is indicated (hedgehog p53 sequence information is currently not available currently). The color code values for functionality and symbol codes for sequence conservation are indicated in the lower part of the figure.

The 47 REs examined correspond to 38 genes and are grouped by biological processes [Gene Ontology annotations (17)]. They were used as seed/reference sequences for multispecies comparison and include 43 human REs (Fig. 1, black, uppercase) and 4 mouse REs (Fig. 1, red, lowercase). For each RE, the degree of sequence conservation relative to seed was determined, and transactivation capacity was predicted (SI Table 2). Sequence conservation is indicated by the size of the symbol “O” (“X” denotes lack of RE sequence) (Fig. 1 and SI Table 2).

Among the human REs, nearly 75% (32/43) showed high (≤2 mismatches relative to seed) to “moderate” (three to six mismatches) sequence conservation (Fig. 1 and SI Table 2) in at least one species beyond the primate lineage, with even greater representation (91%; 39 of 43) if sequence conservation included the “poor” conservation category. Within primates, there was nearly 100% high to moderate conservation.

Some level of functional conservation across at least one nonprimate species (i.e., a minimum of one gray box) was predicted for only 67.5% (27 of 40) of the human REs. Comparisons were not possible for three seed REs considered “nonresponsive to p53.” Conservation of RE functionality was clearly less than sequence conservation (high to poor range) (P = 0.013, Fisher's exact test). Within primates, a significant change in functional RE activity was found only for chimp EDN2, predicted to have “slight” transactivation (green), and rhesus PLK3, predicted to be nonresponsive (blue).

Surprisingly, there were no associations between functionality level of human RE and extent of RE sequence conservation. For example, p21 RE-5′, p21 RE-3′ BAX-C, and Mdm2 RE-A are human REs with predicted strong, moderate, weak, and poor levels of p53 responsiveness, respectively, and yet the sequences are highly conserved across species. Conversely, among REs only conserved in primates, DDB2 is an example of a highly responsive RE, whereas XPC and PCNA are weaker. There were examples of RE sequence conservation but species-specific changes in RE functionality. Higher activities relative to the human RE are predicted for PERP RE-C in mouse and TRAF4, PMAIP, CCNG1 in rodents. Among conserved RE sequences, much lower activities relative to the human RE are instead predicted for PLK2 (SNK) in dog and SFN in cow.

Comparison of REs, using corresponding orthologous human/rodent sequences, was possible for 46 REs (43 human and 3 mouse REs). The mouse Fas RE had no corresponding orthologous sequence in any species (SI Table 2). At least some RE sequence conservation (high to poor range in Fig. 1) between human and at least one rodent species was found for 30 of 46 REs (65%). Conservation of transactivation potential was somewhat less, with 23 (53%) being considered functional among 43 REs (3 of which were p53-nonresponsive).

Notable is the absence in rodents of p53 RE sequence conservation and predicted functionality for DNA metabolism genes. For XPC and MLH1 REs, there is no conservation in all nonprimate species examined, whereas GADD45A is conserved and predicted to be at most weakly active in rabbit, armadillo, elephant, hedgehog, and opossum. Species-specific conservation was observed for RRM2B (rabbit and hedgehog) and PCNA (dog) REs. The weak MSH2 RE is conserved in rabbit, cow, and dog (two base changes from human) but predicted to be nonresponsive to p53. Interestingly, this RE is not conserved in rat because of an 8-bp insertion; however, the resulting RE is predicted to have very limited p53 responsiveness, similar to the human RE.

p53 REs Confined to Orthologous Positions Within Large Promoter Regions.

We asked whether conservation of p53 control of target genes could occur through nonconserved and conserved RE sequences, providing compensatory p53 control. Analysis was expanded to broader promoter regions to identify additional putative p53 REs (either conserved or species-specific) not previously reported. We focused on a comparison between human and mouse genomes to address the presence of compensatory p53 binding-sites, particularly for p53 target genes in the DNA metabolism group, which lacked RE conservation. We chose a window of 10 kb centered on the transcription start site (TSS), because nearly all established p53 REs directly linked to changes in transcription rates of associated genes are found within this region (see Methods).

Putative hits were manually inspected in light of additional parameters, such as position of mismatches and nature of the CWWG sequence, based on our results reported in ref. 14 (SI Table 1) and consistent with recent structural data (20). We also took into account sequences comprising a 10-bp half-site and an adjacent consensus pentamer site (i.e., a p53 monomer binding site), because this type of ¾ RE structure exhibits moderate activity (Fig. 2 and J. Jordan, D.M., A.I., M. Nourredine; D.A., Bell, and M.A.R., unpublished data).

Fig. 2.

Transactivation potential of human and mouse REs by p53, p63, and p73. The transactivation capacities of human WT p53, p63-β, and p73-β toward p53 REs derived from human and mouse target genes were assessed in a set of isogenic yeast strains where there was a modest level of expressed proteins (cells grown in 0.008% galactose medium). The REs were placed upstream of a luciferase reporter (11). Included are the strongly active and evolutionary conserved p21-RE-5′, the diverged GADD45A REs, and the conserved MDM2 REs. The RE name, nucleotide sequence, and position of p53 monomer-binding sites (arrows) are shown. For each RE, nonconsensus bases are in red lowercase. Capitalized blue bases in the mouse REs denote changes compared with the orthologous human REs. For MDM2, the two lines of sequence correspond to the two REs present in the human and mouse MDM2 promoters, which are separated by 17 bases, although the second RE of sequence appears more like a ¾ site. Relative to the p53 consensus sequence the second p53 RE in MDM2 has one mismatch in the fourth position of the CWWG core, which is viewed as a critical component in p53 REs. Nevertheless, the remaining sequence of the first decamer matches the consensus and, when combined with the second half-site, contains only one mismatch in a noncritical position, which would result in a ¾ site. Bar graphs present average reporter light units and standard deviations of three independent replicates.

Twenty-four p53 target genes were examined, including the DNA metabolism group, except MLH1 and MSH2. As summarized in SI Table 3, our pattern search retrieved all known p53 REs from those promoter regions. No additional p53 REs, either conserved or species-specific, were identified except for a few partial RE sequences whose functional significance is uncertain. A functional mouse-specific, nonorthologous compensatory RE was only found in the Fas promoter, referred to as RE-3′.

Transactivation Potential of Human and Mouse p53 REs in Yeast.

To directly analyze our estimates of functionality between species, we examined a set of human and mouse RE sequences, using a luciferase-based assay in yeast that provides for variable expression of human p53. We also investigated transactivation by human p53 orthologs p63 and p73 (β isoform) because RE sequence divergence might be reflected by different relative contributions of p53 family members in modulating target genes, which could have selection pressure consequences on target RE sequences. p63 is considered the most ancient gene of the family and is more closely related to p73 than to p53. After the vertebrate transition in the phylum Chordata, gene duplications eventually led to a p53 family with three gene members (18). Results in knockout animal models and, in the case of p63 and p53, phenotypes associated with germ-line mutations in humans, provide compelling evidence for functional divergence within the p53 family. However, both p63 and p73 can participate in DNA damage responses and influence p53 function as a transcription factor and tumor suppressor gene, based on results in rodents and human cells. The orthologous human and mouse p63 and p73 proteins share ≈98% and 90% sequence similarity, respectively. Both human p63 and p73 proteins are ≈60% identical to p53 within their DNA binding domains and exhibit overlapping but nonidentical transactivation specificities toward p53 REs, although reports differ on optimal consensus RE for p63. A specific consensus RE for p73 has not been determined (19–22).

We chose the yeast-based assay because it can evaluate transactivation capacity of p53 toward defined REs integrated at a specific chromosomal locus upstream of a minimal promoter in isogenic strains (12). Using 16 REs (8 pairs of REs that differ by a SNP), we found that results in yeast are highly predictive of human cells (10). Functional analysis was performed on REs taken from the Gadd45a, Pcna, Apaf1, Mdm2, and Cdkn1A genes (Fig. 2 and SI Fig. 4). The conserved p21 RE-5′, used as a strong positive control, exhibited the highest activity with all three p53 proteins. Human GADD45A contains a moderately functional p53 RE, whereas the mouse RE has six nucleotide changes from the human RE and three changes from consensus (Fig. 2). Functional analysis confirmed the prediction that the mouse RE is inactive for p53-mediated transactivation. Similarly, p63-β and p73-β failed to stimulate transcription from the mouse RE, yet they are active on the human RE. The Mdm2 REs showed functional conservation with p53, p63-β, and p73-β. Both the human and mouse sites comprise two tetramer sequences with a 17-bp spacer (SI Table 2). Although the orthologous mouse site differs by 6 nt from human, those differences had little effect on functionality, as predicted.

We also tested human PCNA RE and three additional putative sites identified by pattern search in the human and mouse promoter region together with the partially conserved Apaf1 and the conserved Cdkn1A (p21–3′) REs (SI Fig. 4). Overall, our results indicate that the potential p53 transactivation capacity of the Pcna gene appears greatly reduced in mouse. For Apaf1, we observed functional conservation, as predicted. Finally, the p21 RE-3′ site should be considered a ¾ site with moderate activity, because a conserved monomer binding sequence adjacent to the 3′ half-site was found (SI Table 2).

Transactivation Potential of Homologous Genes and Orthologous Promoter Regions in Mouse Cells.

The transactivation assay in yeast of the 33 reference/seed sequences (except those marked by # in Fig. 1) generally confirmed our predictions of the presence or lack of functional conservation with defined p53 REs. However, p53 transcription can be influenced through cooperative interactions with other cis-acting TFs (23, 24). In the case of the FLT1 (VEGFR1) gene, we recently found that the p53 RE required for the interaction can even be noncanonical (9).

The finding that neither sequence or function of human p53 REs are conserved in the mouse genes Ddb2, Gadd45a, Rrm2b, Pcna, Xpc, Mlh1, and Msh2 (contains an 8-bp insertion in the RE) is surprising given the p53-mediated control of these genes in humans and the evolutionarily conserved function of their protein products (25–27). Also, there is conservation of the region surrounding the REs in primates and rodents for all but Pcna and Xpc (SI Fig. 5). Therefore, we summarized reported values of p53 dependent responsiveness of these target genes based on expression-microarray or real time PCR analyses (see SI Table 4). Of the seven DNA metabolism associated genes examined in mouse, expression results were available for all but Msh2. For five of the genes, there was no evidence of p53 dependent expression (only one of six reports demonstrated p53 dependent induction of Gadd45a), whereas, for Rrmb2, there was only a small level of induction.

To address the effect of sequence context and possible compensatory changes in other cis-regulatory sequences, the transactivation potential of large (≈1-kb) promoter fragments from human cells and corresponding mouse regions were compared. We analyzed GADD45A, a moderately functional RE in humans but proposed as nonfunctional in mouse, and RRM2B, a strong RE in human with no evidence of RE sequences in mouse. We also included Sema3B, a weak RE that is poorly conserved. Mouse embryo fibroblast (MEF) p53-null cells were cotransfected with an empty expression vector or a vector expressing mouse WT p53 in addition to the luciferase reporter plasmids with the cloned orthologous fragments. There were marked differences in p53-mediated transactivation potential of the GADD45A and RRM2B constructs, consistent with results for the isolated REs (12) (Fig. 3). The mouse-derived Gadd45a and Rrm2b sequences did not mediate p53-responsiveness, whereas the human sequences enabled strong induction. This observation also demonstrates that mouse p53 can efficiently transactivate from human RE sequences.

Fig. 3.

Transactivation potential of mouse and human p53 REs promoter fragments in MEF p53^−/− cells. MEF p53^−/− cells were transiently cotransfected with either an expression vector for mouse WT p53 or empty control vector along with mouse or human promoter fragments of the GADD45A, RRM2B, and SEMA3B genes cloned in the pGL3-promoter vector. The p53-responsive luciferase construct pG-13 provided a positive control. Transcriptional activity was measured 24 h later, using the luciferase reporter assay. Shown are average values for luciferase activity relative to empty pGL3 vector and standard deviations from at least three independent experiments.

The human SEMA3B fragment was inactive, but there was activity from the mouse orthologous fragment. The putative human RE sequence is actually a ¾ site, whereas the mouse is a full site (SI Table 2, lines 613 and 617). The p53 responsiveness of the SEMA3B gene in humans might be mediated primarily by another proposed weak RE (not present in the cloned region examined) that is ≈0.6 kb upstream of the TSS (28) and is not conserved in mouse (see RE alignment in SI Table 3).

Discussion

The combination of comparative genomics of REs in master regulatory networks in concert with functional analysis provides opportunities to address functional conservation of components within networks. We investigated 47 functionally validated p53 REs (35 human, 12 murine) shown to affect expression levels of the associated genes. We examined sequence conservation in 14 species spanning ≈500 million years of evolutionary separation and functional conservation, defined by transactivation capacity of the REs in gene reporter assays. Changes in the level of p53 responsiveness are not estimable simply by sequence inspection of homolog REs, because of the degenerate nature of the consensus p53 RE. Importantly, large differences in transactivation potential can result from introduction of even a single base change in an RE (10).

Our results demonstrate widespread turnover of p53 TFBSs and suggest independent gain and loss in different phylogenetic lineages. Of the established human p53 REs analyzed, 91% had sequence conservation in at least one other species beyond the primate lineage, compared with 67.5% for functional conservation. Of the REs that do show poor sequence conservation, some appeared to be primate specific (e.g., BAX-A, BAX-B, and P53AIP1), indicating recent acquisition or modulation of p53 responsiveness, whereas a few others were conserved in more distant species (e.g., SFN, RRM2B, GADD45A, and PCNA), indicating lineage-specific loss or changes in responsiveness. Notable was the lack of conservation in rodents of REs from all of the genes involved in DNA metabolism. For Rrm2b and Gadd45a, we also confirmed that a region of ≈1 kb that is homologous to the human p53 responsive regulatory region is not responsive to mouse p53. Because regulatory regions are known to occur further upstream or downstream of a gene, there remains a formal possibility of a distant compensatory site, even if there was an absence of REs in the 10 kb surrounding orthologous sites. Additionally, other than TFs, the involvement of chromatin structure and context in p53-dependent transactivation cannot be ruled out, especially when p53 REs are weak. In this study, the yeast-based assay reveals the potential for p53-mediated transactivation of defined REs placed in a constant chromatin environment (all our strains are isogenic and differ only in p53 RE sequence). Both the yeast and the mammalian cell-based assays can only determine the potential for p53 transactivation and do not factor in the possible influence of chromatin.

Of particular relevance to our rodent/human analysis of p53 REs is that mouse and human p53 proteins that are ≈86% identical in the DNA binding domain exhibit equal transactivation capacities in the yeast-based assay (11). However, differences in REs are not necessarily predictive of changes in associated target gene expression. Similar gene expression could be achieved through assemblage of analogously acting, although nonorthologous TFBSs. Alternatively, regulatory elements controlling the target genes could have evolved beyond recognition, through several coevolved changes. It is possible that TFBSs coevolve with their binding factors (29, 30), and it is their combined interaction rather than the primary structure of TFBSs that is preserved by selection over long periods of time, resulting in functional conservation of gene control in distantly related species. For example, examination of enhancer evolution in Drosophila species revealed sequences that can rapidly evolve (confounding traditional alignment approaches) while preserving their function across species (7, 31, 32). Comparison of human and mouse enhancer sequences also suggested similar widespread turnover of TFBSs (8, 33). However, comparative genomics-based analysis of p53-regulated genes and/or p53 REs in relation to biological function is largely absent.

The appearance of pathways in evolution may depend on selection of a new TFBS in response to a newly arising selection pressure, starting from a neutrally evolved initial state and progressing by point substitutions. Although it is formally possible that p53 modulates all corresponding target genes in orthologous species, our 10-kb promoter scans did not provide evidence for compensatory p53 REs in the mouse genome (SI Table 3). Furthermore, an analysis of the expression of DNA metabolism genes in mouse was consistent with the sequence bioinformatics evaluations in that there was little, if any, p53 responsiveness (SI Table 4). These results contrast with the apoptosis related genes. Of the 13 genes (19 total REs; 4 for BAX, 3 for PERP, and 2 for FAS) in Fig. 1, gene expression information was available for nine mouse genes, of which seven were clearly p53 responsive (SI Table 4; there was no information for Siva, Pten, Aip1, and Tp53I3). The present results are especially important, because rodents are often used as model organisms for cancer investigations and studies addressing the impact of DNA damaging agents. The findings were corroborated by our functional analysis in mouse cells of large promoter regions from mouse and human Gadd45a and Rrm2b (p53R2) genes. Unlike the human promoter region, the mouse was completely lacking in p53 responsiveness, consistent with the absence of a detectable compensatory site.

The lack of conservation among DNA metabolism p53 REs (Fig. 1) points to a recent emergence of primate-specific p53-regulated DNA metabolism pathways and suggests divergent positive selection. The cell cycle arrest responses mediated by the transcriptional activities of p53 are, in contrast, well conserved during the evolution. Although the sample size may not warrant a formal statistical analysis, the eight characterized cell cycle genes had five functionally conserved and six sequence-conserved p53 REs, whereas there was little if any conservation for the seven DNA metabolism/repair genes examined. For the case of DNA metabolism and repair, these capabilities must be on to some degree because of continuing DNA challenges. Although speculative, the additional stress responsiveness for the DNA metabolism and repair genes in the primate lineage may reflect increased necessity for coordinated events or greater repair activity possibly during DNA replication. For example, the Ddb2 gene was shown to be nonresponsive to p53 in mice, and it was suggested that because rodents are nocturnal animals and also have a fur shield, there may be reduced selection for systems that tune the DNA damage responses to the UV component of sunlight (34, 35).

Consistent with our findings, an independent study that used comparative genomics to examine sequence conservation of >80 p53 REs between human and rodents also observed pronounced RE sequence divergence (36). Those findings with p53 REs contrasted with the bona fide binding sites for the NRF2 and NFkB stress response TFs, which showed high interspecific conservation. Comparisons were made with other genomic coding and noncoding sequences. In that study, p53 REs in DNA repair/metabolism genes were found to have little sequence homology with rodent sequences (36). Similar to the present findings, where functionality has also been found to be conserved, there was conservation of p53 REs associated with the cell cycle genes. Also consistent is our observation that p53 target REs in many of the apoptotic genes described in Fig. 1 exhibit little conservation between rodents and humans. However, there are examples of species-specific functional REs (e.g., FAS) and cases where, among multiple human REs in a promoter, one is conserved in rodents (e.g., BAX and PERP) (Fig. 1). Interestingly, and as noted above, several mouse apoptotic genes lacking human p53 RE counterparts exhibit responsiveness to DNA damage stress. Along this line, apoptotic signaling machinery has been suggested to be evolutionarily conserved throughout vertebrates and most probably was established after the divergence between vertebrates and nonvertebrates (37).

Importantly, our combined analysis of p53 REs addresses conservation of level of functionality and facilitates comparisons between functionality and sequence conservation. Just as mutations in p53 can alter levels of transactivation from different REs in human cells (11, 13), changes in sequence can impact the level of transactivation. We found that some REs appear to display conservation of the potential for comparable levels of transactivation, whereas others show considerable variation (compare TRAF4 and BAX-C in Fig. 1; also see Results). There are, of course, other factors that could influence the RE-mediated levels of gene expression. For example, p53 proteins with altered functionalities could originate and be selected as part of the adaptation to specific environments, as proposed for the mole rat (38). Along this line (and analogizing p53 to a hand playing a piano), we have proposed p53 as a “master gene of diversity” based on change-of-spectrum p53 mutants and differences in transactivation toward a variety of REs (11).

Visual depiction that accurately and succinctly portrays functional and sequence homology is challenging. One of our goals was to summarize our findings of functional vs. sequence homology of the 47 p53 REs analyzed in a simple and concise manner. Heat maps provide easy mental access to a vast amount of detailed information, as has been well established for large bodies of microarray expression data. Our functional heat map approach may stimulate comparative genomics and/or bioinformatics researchers to explore new connections in evolving systems through more extensive experimental analyses addressing evolution of transactivation potentials and sequence divergence both at the level of transcription factor proteins and target REs. Last, note that p53 proteins that are quite diverged (e.g., armadillo or zebrafish) may exhibit different intrinsic transactivation specificities, limiting the predictions of orthologous RE functionality based on rules established with human or mouse p53. Finally, the combined sequence and functional conservation approach can be applied to other master regulatory networks provided that the relevant REs can be functionally characterized. In this regard, yeast-based model systems have also been developed to address sequence-specific transactivation capacity for NKX2.5, a TF required for human heart development (39) and for components of NFkB (A.I. and M.A.R., unpublished data) that are essential in many inflammatory responses.

Methods

Multispecies Comparative Analysis of p53 Target Genes.

To examine p53 RE sequence and functionality conservation, we conducted a phylogenetic analysis of 47 known p53 REs located in proximity to transcriptional start sites (TSS) of validated p53 target genes. This list included the 33 functional REs for which we had previously determined the transactivation potential plus14 REs identified based on DNA binding or transcription assays, including a recent genome-wide mapping of human p53 bound to DNA (40). We extracted corresponding sequences from other species, using the 14-species (human, chimp, rhesus monkey, dog, cow, mouse, rat, hedgehog, rabbit, elephant, armadillo, opossum, fugu, and zebra fish) multialignment [based on multiZ (41)] data from University of California, Santa Cruz (42). When downloading multialignment regions of p53 REs, we included 25-bp flanking regions (5′ and 3′), and the regions were manually examined for p53 REs, sequence conservation, and predicting transactivation potential based on the set of response element rules.

Computational Search for Putative Compensatory p53 Motifs in Established p53-Target Genes.

To address compensatory p53 control, we expanded our analysis to include broader promoter regions to identify possible additional putative p53 REs (either conserved or species-specific) not previously reported. We chose a window of 10 kb (5 kb upstream and downstream to TSS), because nearly all established p53 REs directly linked to changes in transcription rates of associated genes are found within this region. The human and mouse genomic sequences of 24 well established p53 target genes were downloaded from University of California, Santa Cruz, Golden Path (42). To identify putative p53 sites, we used p53 consensus sequence instead of p53 PUM from the TRANSFAC database (43), because it assumes that the spacer is always zero (which is not the case, because p53 REs have a varied spacer length of 0–17 bp). Additionally, because the consensus p53 RE is degenerate (two copies of 5′-RRRC(A/T)(T/A)GYYY-3′) and most of the established p53 REs contain mismatches, we adopted a two-step computational analysis where the initial search was done by using the 10-mer half-site (RRRCWWGYYY), and then the flanking sequences were searched for another half-site. We allowed for ≤2 mismatches in the 10-bp half-site, and the spacer between the two 10-mers could be ≤4 bp. These rules were determined from our earlier yeast-based functional assays (11, 12) and are consistent with our human cell findings (J. Jordan, D.M., A.I., M. Nourredine, D.A. Bell, and M.A.R., unpublished data).

Functional Analysis of REs or Promoter Regions, Using Luciferase Assays in Yeast and MEFs.

A group of conserved and diverged REs (relative to human) from the mouse genome was examined by using a yeast-based functional assay, as described in refs. 10 and 11 (see Fig. 1 for REs tested in yeast and SI Methods). Promoter regions of three genes were chosen for functional analysis in MEFs (SI Methods). We assessed transactivation potential of a promoter fragment (≈1 kb) containing a functional RE in the human genome after cloning the human sequence and the corresponding fragment from mouse into the luciferase reporter plasmid pGL3-p (Promega) as described in ref. 15.