Redefining the p53 response element

Identification of a Bona Fide Repressive p53RE.

To compare and contrast the functional properties of p53RE sequences, the p21 gene was selected as a model for p53 activation, and for the repression counterpart a recently identified Lasp1 gene was used (24). Because previous studies on p53-repressed genes could not always identify a p53RE, we took additional steps to verify that Lasp1 p53RE is indeed repressive. To test whether p53 functionally represses Lasp1 gene expression, pCMV-p53 or pCMV-p53-R175H expression constructs were transfected into HCT116 (p53^−/−) cells. Fig. 1A illustrates that the overexpression of p53 induced opposite effects on p21 and Lasp1 transcription in HCT116 (p53^−/−) cells. Wild-type p53 but not p53-R175H mutant repressed Lasp1 protein expression, whereas p21 transcription was activated by wild-type p53 but not p53-R175H mutant. The use of p53-specific siRNA in HCT116 (p53^+/+) cells also elicited the same divergent pattern of p53 transcriptional behavior, where Lasp1 mRNA was up-regulated to 154%, whereas p21 mRNA was repressed by p53 siRNA by 63% relative to that of control (P < 0.05; Fig. 1B).

To confirm the presence of a p53RE, luciferase assay using a pGL3-luciferase reporter construct L(−790) containing the (−790 ≈ +74 bp) fragment of Lasp1 gene was performed, and the results showed that wild-type p53 but not p53-R175H mutant inhibited the Lasp1 promoter activity in a dose-dependent manner (Fig. 1C). Luciferase assay using internal deletion construct LΔRE confirmed that removing the putative p53RE site (−59 ≈ −36 bp, AGGCCAGTTCcccaGCTCCAGCCG) significantly abrogated p53-mediated transcriptional repression (P < 0.01). As controls, 2 additional deletion constructs targeting 24 bp either upstream (LΔups-RE) or downstream (LΔdns-RE) of the RE remained repressed by p53 (Fig. 1D). To verify whether the identified Lasp1 p53RE is associated with p53 interaction in vivo, we examined the region (−651 to +762 bp) of the Lasp1 gene by using ChIP-quantitative PCR (qPCR) in HCT116 cells. The results confirmed p53 binding with the highest level of occupancy coinciding with the region containing the Lasp1 p53RE (−59 to −36bp) and further supported by serial deletion analysis (Fig. 1E). Taken together, the above data show that the (AGGCCAGTTCcccaGCTCCAGCCG) sequence functions to mediate repression by p53.

A Regulatory Dinucleotide Core Is Located Within the p53RE.

To understand the basis behind the differences between p53-mediated activation and repression, we compared the repressing p53RE of Lasp1 with the well-known activating p53RE of p21. In addition to the nucleotide sequence differences, the Lasp1 p53RE has a 4-bp spacer separating the 2 decamer motifs. To examine whether this spacer has any impact on p53RE function, successive nucleotide deletion of the 4-bp spacer was performed. The results showed that deleting part or all of the spacer did not abrogate p53-dependent repression (Fig. 2A).

Next, we proceeded to test the properties of the 2 half-sites of p53RE by performing domain swaps. Three chimera constructs (LΔLasp1-p21, LΔp21-Lasp1, and LΔp21-p21) were generated, in which one or both half-sites of the Lasp1 p53RE were replaced with that of p21 p53RE. Surprisingly, replacement of the second decamer half-site resulted in conversion from repression to activation and the effect was increased further when both half-sites were replaced (Fig. 2B). Because changing the half-sites alone resulted in such functional conversion, we asked the question of whether part or all of the decamer is required for this function. Interestingly, the sequence difference between these 2 p53REs is essentially a dinucleotide core within the conserved CWWG motif, where it is CA in Lasp1 and AT in p21. Fig. 2C shows that when the CA dinucleotide core alone was substituted by AT in both half-sites (LΔAT-AT), it resulted in a dramatic conversion from repression to activation. This effect is mediated principally by the dinucleotide core as a p21 promoter region (2.4 kb) containing p21 p53RE when compared with the (−790 bp) Lasp1 promoter containing the p21 p53RE (LΔp21-p21) insert showed comparable levels of activation in response to p53 (Fig. 2D). These results suggest that the promoter sequence outside of the p53RE has relatively minor or no effect on the function of the RE itself.

To further probe the properties of the dinucleotide core, all combinatorial permutations were generated and functionally tested against both repressing and activating p53RE half-site backgrounds (Fig. 3 A and B). The results revealed that p53RE is strongly activating if its dinucleotide core is either AT, AA, or TT, whereas CG, GG, TG, CC, GC, or CA are repressive. The remaining dinucleotides, TA, AC, AG, CT, GT, TC, and GA, can be activating or repressing depending on the p53RE half-site background where it is located.

To investigate whether there is a physical basis to account for the differential function of p53 binding, oligonucleotides of p21 and Lasp1 p53RE were coated onto ELISA microtiter plates and tested for their ability to bind p53 protein (25). Fig. 3C shows that p53 had a distinctly higher binding affinity to p21 p53RE oligonucleotides compared with Lasp1 p53RE. Moreover, replacement of the AT dinucleotide core of p21 RE with CA (p21-CA) resulted in a significant decrease in p53 binding, whereas converse replacement of CA with AT in Lasp1 RE (Lasp1-AT) resulted in an increased level of p53 binding (Fig. 3D). These results demonstrate that the dinucleotide core is important in stabilizing the binding of p53 to its RE.

Triplet Flanking Sequences Modulate p53 Transcriptional Behavior.

A series of chimera half-site constructs were generated to functionally evaluate the role of the triplet RRR and YYY flanking sequences. Lasp1 p53RE was first replaced by either duplicated half-site 1 or duplicated half-site 2. Although these replacements did not change the overall p53 repressive function, slight loss of repression was observed when Lasp1 p53RE half-site 1 was duplicated, whereas the duplicated half-site 2 gave rise to an even stronger repression effect (Fig. 4A). Because the triplet flanking sequences in the half-site 2 of Lasp1 p53RE is RYYCCAGYYR, which differs from the YRRCATGYYR pattern in p21 p53RE half-site 2, we generated 3 additional constructs based on the Lasp1 p53RE background to replace only the flanking triplet of Lasp1 p53RE half-site 2 with that of p21 p53RE. Such replacement, however, did not result in any change in p53 repression (Fig. 4B).

Next, either one or both of the triplet flanking sequences in the half-site 2 of p21 p53RE was replaced by that of Lasp1 p53RE. Although the CAACATGTTG sequence was changed to CAACATGCCG, they both conform to the same YRRCATGYYR pattern, and thus there was only a slight loss of activation observed. In contrast, the change of triplet sequence pattern from YRRCATGYYR (CAACATGTTG) to RYYCATGYYR (GCTCATGTTG or GCTCATGCCG) resulted in a significant reduction in p53-mediated activation (Fig. 4C). Although the p21 RE is well studied, its half-site 1 possesses the canonical RRRCWWGYYY motif whereas half-site 2 is YRRCWWGYYR. Using this natural variation in half-site 2, the positional effects of R and Y nucleotides were evaluated by creating different triplet combinations (Fig. 4D). The results show that the activating property of the dinucleotide core AT can be abrogated by changing the nucleotide adjacent to the CWWG motif (YRRCWWGYYR to RRYCWWGRYY). Hence the data indicate that the positional weight of individual nucleotides within the triplet flanking sequence is stronger when closer to the CWWG motif and decreases when further away from it.

Arising from this observation, a comprehensive set of synthetic triplet flanking motifs was constructed to assess the combinatorial possibilities on p53RE function (Fig. 4E). As expected, the RRR-YYY:RRR-YYY combination, which conforms to the classical canonical motif, is most favorable for activation. Alterations to this motif and in particular the YYY-RRR:YYY-RRR combination will disrupt activation even when the dinucleotide motif is of the activating type. Taken together, these results reveal a set of general principles or rules governing the roles of nucleotides within the p53RE: (i) the dinucleotide core within CWWG motif is the major factor determining activation or repression function, (ii) the triplet flanking sequences can exert a modulating effect, and (iii) the nucleotide adjacent to the CWWG motif has the strongest positional effect. In repressive p53REs, the flanking triplet can only modulate the degree of repression but not reverse it; whereas in activating p53REs selected triplet combinations can convert activation to repression.

Reassignment of 20 Discrepant p53REs.

To test the robustness of these rules, a list of published p53REs were compiled and grouped into activating or repressing types. When the above rules were applied to published data, 142/162 (87.7%) were found to be concordant (Table S1), whereas 20 REs (16 activating, 2 repressing, 2 unknown) (26–28) were discrepant. To experimentally verify these discrepancies, all 20 p53REs were synthesized and individually cloned into pGL3-Lasp1–790 luciferase reporter plasmid. The validation assays showed that the assigned roles of these published p53REs turned out to be erroneous (Table S2). For instance, IER3/IEX-1 gene has a reported p53RE that functions as a repressor (29), however, the sequence CCACATGCCT:CGACATGTGC contains the classical activating dinucleotide core motif of AT in both decamers. Functional assay performed on this p53RE confirmed that the sequence is indeed activating rather than repressing. Another example is the second CTSD/IRDD p53 binding motif (AAGCTGGGCC:GGGCTGACCC), where it was reported to be activating (30), but the functional assay proved it to be of a repressing type. This result agrees with the repressing dinucleotide core combination (Fig. 3 A and B), which is TG in both decamers. The role of triplet flanking nucleotides in modulating the p53RE function can be seen in the p53RE of BCL2L14/BCL-G (31), where both dinucleotide cores are of the activating type, but the RE was functionally repressing. This observation can be explained by the positional effect of the nucleotides adjacent to the CWWG motif as shown in Fig. 4 D and E where instead of the canonical RRRCWWGYYY:RRRCWWGYYY, the BCL2L14/BCL-G p53RE (AGCCAAGGCT:GGTCTTGAAC) is RRYCWWGRYY:RRYCWWGRRY. Similarly, a second example can be seen in the p53RE of PLK2/SNK gene, GGTCATGATT:TAACTTGCCT, i.e., RRYCWWGRYY:YRRCWWGYYY (32), in which the dinucleotide core is also activating type, but the p53 function was modulated by the nucleotides adjacent to CWWG motif in half-site 1 to become repressive.

Activating and Repressing p53REs Have Characteristic Nucleotide Distribution Patterns.

To capture the general principles and rules of p53RE behavior, we analyzed the distribution frequency of the dinucleotide core among the 123 activating p53REs and 39 repressing p53REs. To do this, dinucleotide cores were grouped into 3 categories. Group 1 dinucleotides include AT, AA, and TT, which had consensus activating properties in both repressing (Fig. 3A) and activating (Fig. 3B) p53RE backgrounds. Similarly, group 3 dinucleotides (TC/GA/TG/CA/GC/GG/CC/CG) showed consensus repressing functions in both backgrounds, whereas group 2 (TA/AC/AG/CT/GT) are those that are relatively neutral in activity. It was found that the 123 activating p53REs have predominantly AT/AA/TT in each half-site (Fig. 5A), and when taken together as a complete RE, only group 1 and/or group 2 dinucleotide combinations were observed (Fig. 5B). The 39 repressing p53REs, however, have very different dinucleotide core composition with a high proportion of TG/TC/GC/CA/GG/CG motifs that are rarely found in activating p53REs (Fig. 5C). As a complete RE, >60% of the repressing p53REs contain at least 1 group 3 dinucleotide core combination (Fig. 5D).

Turning our attention to the nucleotides adjacent to the CWWG motif, Fig. 6A shows that in 123 activating p53REs the frequencies of these nucleotides agree well with the canonical sequence. In contrast, the repressing p53REs have a much higher frequency of noncanonical nucleotides in the position immediately adjacent to the CWWG motif (Fig. 6B). Similarly, at the complete RE level, these nucleotides in activating p53REs conform well to the canonical RCWWGY-RCWWGY pattern (Fig. 6C), whereas among the 39 repressing p53REs, this pattern is seldom intact (Fig. 6D). Therefore, the repressing p53REs were found to differ from the activating p53REs in 2 aspects. First, the dinucleotide core is likely to be of the repressing type, and second, the flanking nucleotides adjacent to the CWWG motif are frequently found to depart from the canonical p53RE motif.

Cell Lines.

The human colon cancer cell line HCT116, its derived p53^−/− cells, the wild-type p53 (pCMV-p53), and mutant p53 (pCMV-p53-R175H) expression vectors were kindly provided by Bert Vogelstein (The Johns Hopkins University, Baltimore).

Luciferase Reporter Assay.

DNA fragments containing 5′ flanking region (−790 to +74 bp) of Lasp1 promoter were first amplified by PCR and subsequently subcloned into pGL3-Basic vector. Different luciferase constructs used in this study were then generated based on L(−790) template. Different promoter luciferase reporter constructs (2 μg/mL) were cotransfected with pCMV-p53, pCMV-p53-R175H expression vectors, or pcDNA3.1 control plasmid (unless specified, 200 ng/mL of the expression vectors was used), together with 20 ng/mL of Renilla vector PRL-null into HCT116 (p53^−/−) cells plated in 96-well plates. The luciferase reporter activity was measured by using the Dual-Luciferase Reporter Assay System (Promega). For all luciferase assays, the normalized luciferase activity of each construct (Firefly/Renilla ratio) in the pCMV-p53 or pCMV-R175H cotransfected cells was then compared with that of the pcDNA3.1 control cotransfected cells.

ChIP-qPCR.

ChIP assays were carried out with HCT116 (p53^+/+) cells treated with 5-fluorouracil (5-FU) (375 μM) for 24 h as described (23).

ELISA.

To analyze binding of p53 protein to DNA sequences from different p53REs, a DNA-protein binding ELISA was performed as described (25) with modifications. Briefly, 25 ng of purified wild-type p53 protein (Aviva Systems Biology) was incubated with anti-p53 antibody DO-1 and then mixed with different dilutions of annealed double-stranded biotin labeled DNA probes (2, 10, or 50 ng) before plated into streptavidin-coated 96-well plates. After incubation with goat-anti-mouse HRP-conjugated IgG, measurement of color resulting from the enzymatic reaction (TMB substrate) was carried out in a Tecan plate reader at 450/570-nm wavelength.

For additional information see SI Text.