Amino-terminal p53 mutations lead to expression of apoptosis proficient p47 and prognosticate better survival, but predispose to tumorigenesis

Significance

Mutations in the amino-terminal transactivation domain of the tumor-suppressor p53 are mostly insertions or deletions, and result in loss of full-length p53 expression. However, these changes concomitantly result in the expression of a truncated p47 isoform, which retains the ability to selectively transactivate some apoptotic target genes. The selectivity appears to be due to a default feature, stemming from the lack of acetylation on K382 at the carboxyl terminus, which requires the amino terminus. Consistently, expression of p47 could prognosticate better survival in sporadic cancer patients, corroborating with its ability to induce apoptosis. However, apoptosis proficiency appears to be insufficient for tumor suppression, because these amino-terminal mutations are found in the germ line, leading to the Li–Fraumeni syndrome.

Abstract

Whereas most mutations in p53 occur in the DNA-binding domain and lead to its functional inactivation, their relevance in the amino-terminal transactivation domain is unclear. We show here that amino-terminal p53 (ATp53) mutations often result in the abrogation of full-length p53 expression, but concomitantly lead to the expression of the amino-terminally truncated p47 isoform. Using genetically modified cancer cells that only express p47, we demonstrate it to be up-regulated in response to various stimuli, and to contribute to cell death, through its ability to selectively activate a group of apoptotic target genes. Target gene selectivity is influenced by K382 acetylation, which depends on the amino terminus, and is required for recruitment of selective cofactors. Consistently, cancers capable of expressing p47 had a better overall survival. Nonetheless, retention of the apoptotic function appears insufficient for tumor suppression, because these mutations are also found in the germ line and lead to Li–Fraumeni syndrome. These data from ATp53 mutations collectively demonstrate that p53’s apoptosis proficiency is dispensable for tumor suppression, but could prognosticate better survival.

Major efforts in cancer genome sequencing have confirmed that p53 is the most mutated gene in human malignancies (1), highlighting its crucial role in guarding against cellular transformation. Most mutations in p53 occur in the central DNA-binding domain (DBD), expectedly, due to p53’s major function as transcription factor that controls the expression of a plethora of genes that regulate apoptosis, senescence, cell-cycle arrest and DNA repair (2, 3). Mouse knock-in models that recapitulate human cancer-derived p53 mutations and mimic the Li–Fraumeni syndrome (LFS) have confirmed that these DBD mutations lead to loss-of function (LOF), or in certain cases, gain-of-novel oncogenic functions, which appears to be mutation-type specific (4–6). Moreover, mutant p53 has been shown to result in dominant-negative effect over the remaining wild-type allele, thereby inhibiting efficient transcriptional activation and, hence, therapeutic response (5, 7), collectively highlighting the importance of mutations in the DBD in contributing to carcinogenesis and affecting therapeutic outcome.

Mutations in other domains of p53 have also been noted, albeit to a lesser extent. For example, mutations in the carboxyl-terminal oligomerization domain and, in particular, the R337H residue, have been noted to be prevalent in the Brazilian LFS patients, giving rise to a variety of tumor types (8), and especially adrenocortical carcinomas in children (9). This particular mutation causes defects in tetramer formation leading to loss of function (10), thereby highlighting other possible avenues by which mutations can inactivate p53 functionally.

Similarly, mutations in the amino-terminal domain of p53, which contains the transactivation (TA) 1 and 2 domains within amino acid residues 1–40 and 41–61, respectively (11, 12), have also been noted. This region of p53 contains several regulatory elements, such as the MDM2 and p300 binding sites within the first 40 aa, which regulates p53 stability through ubiquitination and activation through acetylation, respectively (13, 14). Furthermore, it is to be noted that alternate translation initiation from the methionine in exon 4 (at amino acids 40 or 44 of human p53) leads to the production of the amino-terminal truncated form, termed as p47 (also referred to as p44, p53/p47, ΔΝp53, or Δ40p53), which lacks the TA1 (15–18). Thus, p47 was initially thought to lack the ability to transactivate targets genes and was indeed reported to lack the ability to induce apoptosis (15). However, subsequent data has suggested that it is capable of inducing expression of some p53 target genes (16, 19). Nonetheless, whether the presence of mutations in the amino terminus, especially in the region between the first two methionines (i.e., amino acids 1–40) (referred hereafter as ATp53 mutations), could affect the structure and functionality of p53 is unclear.

Whereas not much information is available on the functional role of ATp53 mutations found in humans, the role of the functional domains has been examined in mice. Mice with mutations that result in incapacitation of the TA1 alone (p53^25,26) or both TA1 and TA2 (p53^25,26,53,54) have been generated, and have demonstrated a distinct role for TA1 in regulating the transactivation of a subset of genes such as p21, but not genes such as Bax (20, 21). However, both TA1 and TA2 were found to be required together for complete transactivation of all p53 target genes and for tumor suppression (20, 21). Moreover, TA1 was dispensable for tumor suppression in mice, even in the absence of G₁ arrest or apoptosis in response to acute DNA damage (3), suggesting that selectivity in activation of target genes through the various TA domains can regulate tumorigenesis. In this context, several studies have identified various cofactors required for selective transactivation of target genes involved in G₁ arrest (e.g., p21) and apoptosis (e.g., Pig3). One such example is TAF1, which was shown to be recruited to the p21 promoter and, hence, critical for its transactivation by p53 (22). Similarly, BRD7 was shown to bind to p53 and p300 and to be recruited to target gene promoters, affecting histone acetylation, p53 acetylation, and promoter activity of a subset of senescence target genes (23). However, hCAS1 was shown to be recruited to apoptotic target promoters and necessary for their activation by p53 (24). Furthermore, selective posttranslational modifications have also been suggested to provide selectivity in target gene activation. In one case, phosphorylation on serine 46 has been shown to promote apoptotic target gene activation (25). Likewise, acetylation at K382 by p300, which binds to the amino terminus of p53, has been shown to regulate p21-like arrest gene activation (22). Thus, these data collectively indicate that modifications and cofactor binding on the amino terminus of p53 can have a profound effect on its functionality.

We have therefore focused on understanding the effects of ATp53 mutations. We report here that most cancer-derived insertions and deletions (indels) in the amino terminus within the first 40 aa result in abrogation of full-length p53 expression, but lead to the expression of the truncated p47 from the methionine in exon 4. Using genetically modified cancer cells that express p47 without full-length p53 (26), we show that p47 is capable of selectively activating a group of apoptotic targets genes and, hence, is required for apoptosis induction. Lack of efficient activation of p21-like genes by p47 is attributed to lack of acetylation on its carboxyl terminus, which depends on the presence of the amino terminus. Importantly, clinical data indicate that presence of ATp53 mutations in the germ line indeed predispose to LFS. However, patients with sporadic tumors capable of expressing high p47 expression and ATp53 mutations have a better overall prognosis, consistent with the capacity of p47 to induce apoptosis. These data therefore indicate that apoptosis proficiency is insufficient to prevent tumorigenesis, as exemplified by ATp53 mutants.

Results

Tumor-Derived ATp53 Indels Lead to Loss of Full-Length p53 Expression, but Gain of p47 expression.

Interrogation of the COSMIC database for mutations in the p53 gene analyzed in a large number of cancers revealed that whereas most of the mutations were found in the central DBD (1), a small but significant percentage of mutations were in the amino-terminal domain, especially within the first 40 aa (SI Appendix, Fig. S1A and Table S1), which contains the TA1 domain of p53. These alterations include frameshifting insertions and deletions (42%), substitutions (46%), and to a minor extent other large genomic mutations (9%) (SI Appendix, Fig. S1B). Whereas most of the substitutions are missense mutations scattered throughout the TA1 domain at a very low frequency, there are a few nonsense mutations that lead to a stop codon (e.g., p.Q5*; p.Q38*). Intriguingly, decoding several of the cancer-derived indel mutations indicated that many of these frame-shifts lead to a stop codon within the first 43 amino acids, especially within codons 41–43 (Fig. 1A and SI Appendix, Table S1). Interestingly, the subsequent methionines in codon 40 or 44 of exon 4 were unperturbed in all such cases. To determine whether such mutations would affect full-length p53 expression, we generated p53 cDNA constructs to mimic these mutations and expressed them in p53 null H1299 cells. Immunoblotting with the DO1 antibody that recognizes the amino terminus of p53 (amino acids 11–25) revealed the absence of any specific band in all of the cases tested (Fig. 1B). However, probing with the PAb1801 antibody that recognizes an epitope from amino acids 46–55 of p53 revealed the presence of a shorter p53 form, which corresponded to the p47 isoform of p53, in all cases, albeit being expressed at lower levels than the control p47 construct that was expressed due to the ablation of the first ATG of p53 (Fig. 1B). To determine whether the p47 that is expressed by these mutants is indeed initiated from the methionine in codon 40 and/or 44, we mutated these residues to alanines by using one of the ATp53 indel mutants, c.97delT. Indeed, mutating methionine 40 and 44 or 44 alone led to almost the complete abrogation of p47 expression (Fig. 1C).

Fig. 1.

ATp53 indels lead to loss of p53 expression but result in p47 expression. (A) Sequence alignment of cancer-derived alterations identified within amino acids 1–40. All indels lead to the generation of a stop codon (TGA/TAA) within the first 43 aa. The next methionine (ATG) after the stop signal is underlined in each case. (B and C) p53 null H1299 cells were transfected with selected ATp53 indels and analyzed by immunoblotting with an amino-terminal–specific p53 antibody (DO1) (recognizing amino acids 11–25), or an antibody that recognizes between residues 46–55 (Pab 1801) (B). One representative ATp53 indel, the 97delT, was mutated to substitute its methionines at amino acid 40 and/or 44 to alanine, and the plasmids were transfected similarly to analyze the expression of p47 (C). p53 and p47 (where the first ATG in p53 was mutated to a stop codon) plasmids were used as controls, and representative blots are shown. Arrowheads indicate position of p47 and p53. (D and E) Primary colorectal samples with either DBD-domain mutations (DBD), or amino-terminal mutations (AT) were used for immunoblot analyses with the DO1 antibody or the carboxyl-terminal–specific PAb421 antibody (recognizing amino acids 371–380) (D). RKO^+/+ and RKO^−/− cells treated with etoposide were used as positive controls for p53 and p47 expression, respectively (see Fig. 2 for details). Arrowheads indicate position of p47 and p53, and asterisk represents nonspecific bands. Lo and hi refer to low and higher exposures of the blots. Three of the samples, one representing a DBD mutation (3616: p.R175H), and two representing the amino-terminal mutations (3431: c.97_115del; 3624: frameshift) were used for immunohistochemical analysis with the pan-p53 (CM1) or amino-terminal specific (DO1) antibodies (E). Representative pictures from lower (Top) and higher (Bottom) magnification are shown.

To evaluate whether such mutations would lead to the expression of the p47 form in human tumor samples, we used colorectal samples from patients with either common DNA-binding domain (DBD) mutations, or with the amino-terminal mutations as indicated in Fig. 1A. Immunoblot analysis with the amino-terminal DO1 antibody indicated that whereas samples with DBD mutation expressed p53, the AT samples were hardly expressing p53 (Fig. 1D). However, detection with antibodies against both p53 and p47 (PAb 421) revealed the presence of the band corresponding to p47 in three of the four AT samples, but not in the DBD samples. To further evaluate whether p47 is expressed at the cellular level, we also performed immunohistochemical analysis by using these samples. Staining with CM1, a polycolonal p53 antibody resulted in the detection of strong signals in all cases, including both the DBD mutant samples (3616:p.R175H) and the AT samples (3431:c.97_115del and 3624:frameshift) (Fig. 1E). By contrast, and consistent with the immunoblot data, the DO1 antibody only detected p53 in the DBD-mutant sample, but not in both the AT samples. These data therefore collectively indicate that ATp53 indel mutations lead to loss of full-length p53 expression in human cancers, with a concomitant gain of expression of the alternate p47 isoform from the subsequent methionine in exon 4.

p47 Is Induced by Stress Signals and Is Required for Cell Death.

To understand the functional role of p47 in tumors with ATp53 indels, we used the gene-targeted RKO colorectal tumor cell line (RKOp53^−/−) in which the exon 2 containing the first ATG was removed for targeting the p53 allele, leaving behind the second ATG in exon 4 intact (26). This cell line does not express the full-length p53 protein and is incapable of inducing targets such as p21 (26). We therefore hypothesized that p47 would be expressed in this cell line, mimicking the ATp53 indels, and found that although the full-length transcript containing exon 2 was absent, a common transcript containing the exon 5–8 region was present (SI Appendix, Fig. S2A). Small hairpin RNA (shRNA) targeting a common region in the DBD domain of full-length p53/p47 led to a concomitant decrease in both the full-length and the exon 5–8 transcripts, confirming their specificity (SI Appendix, Fig. S2A). We therefore used mRNA from these cells to isolate the transcript and performed 5′ RACE (SI Appendix, Fig. S2B), which revealed that exon 2 was absent, and that the transcript contained the second ATG in exon 4 (SI Appendix, Fig. S2C).

Immunoblot analysis indicated that a p47 kDa protein was detected in RKOp53^−/− cells by using the pan-p53 CM1 polyclonal antibody (Fig. 2A). Basal p47 levels were higher in RKOp53^−/− cells compared with the counterpart RKOp53^+/+ cells that had no detectable p47 expression. To determine whether p47 would be induced by stress signals similar to full-length p53, we treated these cells with a variety of agents, including etoposide, cisplatin (CDDP), taxol, and IR, which revealed that p47 was similarly induced by these agents as p53 (Fig. 2A and SI Appendix, Fig. S3A). Moreover, treatment of these cells with agents inducing cell-cycle arrest at various stages indicated that serum starvation and hydroxyurea (HU) led to an increase of p47, although full-length p53 was only marginally induced by serum starvation in RKO^+/+ cells, as was the case with treatment with the thapsigargin (TG), a ER-stress inducer (Fig. 2A). Given that p47 does not contain the MDM2-binding domain that regulates full-length p53 stability, we examined whether protein stabilization has any contributory role in p47 induction. Evaluation of the half-life of p47 after etoposide treatment showed that p47’s stability was not prolonged by etoposide treatment (although being highly stable), unlike p53 (T_1/2 for RKO^+/+ vs. RKO^−/− for untreated and etoposide treated was the following: <2 h vs. 8 h and >16 h vs. >16 h, respectively) (Fig. 2B), suggesting that p47 increase in the RKO^−/− cells is likely due to other mechanisms but not due to protein stabilization, unlike in the case of p53. Moreover, consistent with the fact that p47 lacks the MDM2-binding domain and, hence, is less likely to be degraded, p47 expression was significantly noted mainly in the nucleus even in the absence of stress stimulation in RKOp53^−/− cells, by immunofluoresence analysis with the CM1 polyclonal antibody, but not with the amino-terminal–specific DO1 antibody (Fig. 2C). By contrast, endogenous p53 was detected mainly in the nucleus in RKOp53^+/+ cells only after treatment with etoposide and cisplatin by both antibodies (Fig. 2C and SI Appendix, Fig. S3C).

Fig. 2.

Endogenous p47 is induced by multiple stimuli. (A and B) Immunoblot analysis of p47 and p53 induction upon treatment with various stimuli was determined by immunoblotting using total cell lysates from the respective RKO^−/− and RKO^+/+ cells and a pan-p53 antibody (CM1) or the DO1 antibody (A). Half-life of p47 and p53 was determined in these cells with or without etoposide treatment for 24 h, followed by cycloheximide (CHX) treatment for the indicated time periods (B). Etop, etoposide; HU, hydroxyurea; -serum, serum starved for 24 h; TG, thapsigargin; UT, untreated. Arrowheads indicate position of p47 and p53. Numbers below the blots indicate quantification of p53/p47, relative to untreated or at time 0 h. (C) RKO^−/− and RKO^+/+ cells with or without etoposide treatment were used for immunofluorescence analysis using the DO1 antibody (Top), or the pan-p53 CM1antibody (Bottom). DAPI stain highlight nucleus.

To evaluate whether endogenous p47 observed in the nucleus is capable of transactivation, we first analyzed the expression of a series of p53 target genes upon etoposide treatment. Time course analysis after etoposide treatment showed that apoptotic (Dr5, Fas, Puma, Noxa), arrest/senescent (p21, Gadd45, 14-3-3, Btg2), and metabolic (Tigar) target genes of full-length p53 were induced to various extents in RKO^+/+ cells (Fig. 3A). By contrast, there was a distinct lack of activation of most of these genes in RKO^−/− cells, except for some apoptotic genes such as Dr5 and Fas (Fig. 3A), suggesting selectivity in the ability of p47 to induce some apoptotic genes, but not the arrest/senescence genes. Silencing p53/p47 using different shp53 or sip53 that target a common region of both p47 and full-length p53 led to the reduction of both full-length p53 (in RKOp53^+/+ cells) and p47 (in RKOp53^−/− cells) (Fig. 3B and SI Appendix, Fig. S4A), again confirming the validity of the p47 band. Analysis of target genes after p53/p47 silencing indicated the reliance of Fas, Pig3, and Aip1 expression on both p47 and p53 in RKO^−/− and RKO^+/+ cells, respectively, in contrast to p21, 14-3-3 and Noxa, which was only induced and subsequently reduced in RKO^+/+ cells (Fig. 3C and SI Appendix, Fig. S4B). We thus determined the relevance of endogenous p47 on cell death. Cell death after etoposide treatment in RKO^+/+ cells was reduced upon silencing p53 expression, to levels similar to scrambled siRNA-transfected RKOp53^−/− cells (Fig. 3D and SI Appendix, Fig. S4C), confirming an expected role of full-length p53 in the death process. Importantly, silencing of p47 expression in RKO^−/− cells lead to a further consistent reduction of death, demonstrating that p47 contributed to cell death. Similar results were obtained in HCT116 colon cancer cells in which p53 were similarly gene-targeted (26) (SI Appendix, Fig. S5). Together, these data demonstrate the propensity of endogenous p47 to be induced by a variety of stimuli, which is required and contributes to cell death.

Fig. 3.

p47 contributes to apoptotic target gene activation and cell death. (A) Various p53 target genes expression was analyzed by quantitative real-time PCR after treatment of RKO^−/− and RKO^+/+ cells with etoposide for the indicated time periods. Relative expression was normalized against gapdh expression. (B–D) Cells were transfected with control (pSuper) or Shp53 and treated with etoposide or cisplatin (CDDP) and used for immunoblot analysis (B), or for real-time PCR analysis of the indicated genes (C). Cell death was determined concurrently after 48 h after treatment (D). Arrowheads indicate position of p47 and p53 in B. All experiments are representative of at least three independent experiments, and data are represented as mean ± SDs.

Selective Induction of Some Apoptotic Target Genes by p47.

To further examine the role of p47 in inducing cell death, we transfected a CMV-driven p53 construct that expresses only p47 but not full-length p53 due to the ablation of the first ATG, into p53-deficient H1299 cells, which led to a significant increase in apoptotic cell death, similar to p53, as determined by Annexin V and propidium iodide staining (% cell death of vector vs. full-length p53 vs. p47: 4.2 vs. 22.4 vs. 16.6) (Fig. 4A and SI Appendix, Fig. S6A). To understand the potential mechanism by which p47 was able to induce cell death, we generated a mutant version of p47 that is incapable of binding to the DNA (p47R175H)—similar to the mutations found in human tumors (27)—to test its ability to inhibit cellular growth. Transfection of p47R175H did not lead to appreciable cell death, unlike full-length p53 and p47 (% cell death of vector vs. p53 vs. p47 vs. p47R175H: 4.2 vs. 22.4 vs. 16.6 vs. 6.0) (Fig. 4A, Left and SI Appendix, Fig. S6 A and B). Moreover, analysis of cell-cycle status upon expression of these constructs revealed a similar trend in the induction of sub-G₁ population, indicative of increased apoptosis by both p53 and p47 (Fig. 4B). However, there were no major differences in the G₁, S, or G₂/M populations upon expression of p53 or p47 in these cells (Fig. 4B). To confirm that p47 was capable of inhibiting cellular growth, we performed long-term colony formation assays (CFA), which again indicated that whereas p47 was able to inhibit cellular growth in CFA, similar to p53, p47R175H was completely devoid of this activity in various cell lines tested (Fig. 4C). Similarly, cancer-derived ATp53 indel mutants were also capable of inhibiting cellular growth to varying degrees (SI Appendix, Fig. S7). These data together indicate that p47 requires DNA-binding activity to inhibit growth and to induce cell death, suggesting potential involvement of transactivation in this process.

Fig. 4.

p47 is capable of inducing cell death and selectively activates apoptotic target genes. (A and B) The indicated constructs were transfected into H1299 cells, and cell death was analyzed 24 h later by flow cytometry, following propidium iodide/Annexin V staining (A), or by sub-G₁ analysis (B). Representative data are shown. (C) H1299 and Saos2 cells were transfected with the indicated plasmids and selected for 8–14 d, and cellular colonies were stained with crystal violet and visualized. Representative data from at least three independent repeats are shown. (D and E) The indicated plasmid constructs were transfected into p53-deficient H1299 cells together with the indicated p53-target promoters linked to the luciferase gene, and luciferase activity was determined (Left). Immunoblots indicate expression levels of the transfected proteins (Right) (D). Real-time quantitative PCR of endogenous target genes were analyzed 24 h after transfection of the indicated p53/p47 constructs in H1299 cells. Relative expression was normalized against gapdh expression (E). Experiments are representative of at least three independent repeats, and SDs are shown.

We therefore explored whether p47 was capable of transactivation of target genes. Analysis of several p53 target promoters driving the luciferase reporter gene showed that p47 was very much compromised in its capacity to activate p21, Mdm2, or Gadd45 promoters compared with full-length p53 (Fig. 4D). However, p47 was capable of inducing the proapoptotic Aip1 promoter. Importantly, the p47R175H mutant was totally defective in inducing Aip1, indicating that DNA-binding activity is required for transactivation of Aip1. To confirm the apparent selective ability of p47 to transactivate different sets of endogenous target genes, we evaluated the expression of a series of arrest and apoptotic target genes by quantitative real-time PCR. As noted with the reporter assays, expression of p21 and Mdm2 was not significantly induced by p47 (Fig. 4E). Similarly, Btg2, Tigar, Puma, and Noxa were also not significantly induced by p47 (Fig. 4E). By contrast, Pig3 and Aip1 expression were induced by p47, and again, the DNA-binding mutant p47R175H was incapable of activating them both (Fig. 4E), highlighting target gene specificity, consistent with the data obtained from the RKO^−/− cells. Altogether, these data demonstrate that p47 is capable of inducing selective activation of apoptotic target genes and cell death.

Defects in Activation of Selected Genes Corresponds to Lack of K382 Acetylation.

To understand the basis of lack of efficient activation of a subset of target genes including p21, Gadd45, and Mdm2 by p47, in contrast to the apoptotic genes such as Aip-1, Fas, and Dr5, we analyzed the status of posttranslational modifications on p47. Etoposide treatment led to an increase of S46 phosphorylation of both p47 and full-length p53 in the respective cells (Fig. 5A). However, and surprisingly, K382 aceylation was almost completely absent on p47 in the RKO^−/− cells, in contrast to full-length p53 in RKO^+/+ cells, although acetylation on K373 was not compromised on p47 (SI Appendix, Fig. S8A). Because the acetylase p300 has been demonstrated to play a role in K382 acetylation (22), we first evaluated its role by using a p300 defective cell line PACL4 (ref. 28 and SI Appendix, Fig. S8B), which showed significant reduction in p53 K382 acetylation upon etoposide treatment, compared with the RKO^+/+ cells (Fig. 5B). Thus, to determine whether acetylation on K382 is important for activation of target genes like p21, we generated full-length p53 mutants in which the amino-terminal sites capable of binding to p300 were mutated, i.e., L22Q and W23S mutants (14), as well as the K382R mutant that cannot be acetylated. These mutants were significantly compromised in K382 acetylation (Fig. 5C) and exhibited a reduction in their ability to activate p21 expression, although they were able to activate apoptotic targets Pig3 and Aip-1 efficiently (Fig. 5D and SI Appendix, Fig. S8C), suggesting a link between K382 acetylation and ability to efficiently activate p21 expression.

Fig. 5.

K382 acetylation is required for efficient p21 gene activation, but dispensable for apoptotic target gene activation. (A and B) Posttranslational modifications on endogenous p47 and p53 after cellular stimulation with indicated reagents were analyzed by immunoblotting using respective antibodies. PACL4 is a pancreatic cell line with p300 deletion and with a mutant p53 (B). Arrowheads indicate position of p47 and p53, and asterisks represents nonspecific bands. (C and D) H1299 cells were transfected with the indicated p53 mutant constructs, and cell lysates (C) or mRNA (D) were used for immunoblot or quantitative PCR analysis of target genes, respectively. (E) Chromatin immunoprecipitation was performed by using anti-TAF1, hCAS1, or IgG control antibodies, followed by PCR amplification of p21 (sites p21A and p21B) or Aip1 promoter regions, from RKO^−/− and RKO^+/+ cells with and without etoposide treatment. Input is from genomic DNA without chromatin immunoprecipitation. Asterisk represents primer dimers. (F and G) H1299 cells were transfected with the indicated p53 constructs in various combinations with the p47 plasmid, and K382 acetylation was determined by immunoblotting (F). Asterisk represents nonspecific band. mRNA from parallel cultures were used for quantitative PCR analysis of p21 (G). All experiments are representative of at least three independent experiments, and data are represented as mean ± SDs.

We thus explored the potential of p47 to bind to the prototype p21 promoter by chromatin immunoprecipitation (ChIP) assays and found p47 to be bound to a variety of target gene promoters including p21, Mdm2, and other apoptotic promoters even in the absence of any activation by etoposide in RKO^−/− cells (SI Appendix, Fig. S8D). By contrast, full-length p53 was generally bound more significantly onto these promoters only after etoposide treatment, with weak basal binding noted in naïve conditions in RKO^+/+ cells. Examination of cofactors recruited to the target promoters, such as TAF1 onto cell-cycle arrest promoters (22), or hCAS1 onto apoptotic promoters (24), revealed that while TAF1 was loaded efficiently upon etoposide treatment onto both the 5′ and 3′ p53-binding sites (referred to as p21A and p21B, respectively) on the p21 promoter in RKO^+/+ cells, it was not the case in RKO^−/− cells (Fig. 5E). Similar results were obtained upon UV treatment (SI Appendix, Fig. S8E). In addition, K382-acetylated p53 was also loaded onto the p21 promoter sites onlE in RKO^+/+ cells but not in RKO^−/− cells (SI Appendix, Fig. S8F). By contrast, hCAS1 was loaded equally onto the prototypic apoptotic Aip1 promoter upon etoposide treatment in both RKO^+/+ and RKO^−/− cells (Fig. 5E). Furthermore, we used both wild-type and K382R mutant p53 plasmids to determine the direct relevance of K382 acetylation on TAF1 recruitment to p21 promoter upon etoposide treatment, which revealed that unlike wild-type p53, the K382R mutant was unable to facilitate TAF1 binding to the p21 promoter sites (SI Appendix, Fig. S8G). These data therefore show that lack of K382 acetylation results in reduction of p21 expression due to failure of recruitment of TAF1 onto the p21 promoter in the p47-expressing RKO^−/− cells.

We therefore explored whether alternate means of K382 acetylation of p47 can lead to rescue of p21 activation. Because p47 lacks the amino terminus containing the p300-binding domain, we cotransfected p47 with either a wild-type or R175H mutant full-length p53, reasoning that coexpression would lead to oligomerization of both p47 and full-length p53 as reported (29), leading to recruitment of p300 by the latter’s amino terminus and, thus, promoting acetylation of p47 in trans. Confirming this hypothesis, coexpression of wild-type or R175H mutant full-length p53 resulted in acetylation of p47 on K382 (Fig. 5F). Concurrent analysis of p21 expression showed that the R175H mutant full-length p53 was inactive (Fig. 5G). However, coexpression of mutant full-length p53 led to a small but significant increase in p21 activation by p47. It is noteworthy that mutant full-length p53 was also found to inhibit p47’s transactivation potential due to dominant-negative effect, as noted with an apoptotic Pig3 expression (SI Appendix, Fig. S8H), probably explaining the lack of a massive activation of p21 when both are coexpressed. These results together demonstrate that the absence of the amino terminus in p47 is causal to lack of K382 acetylation and, thus, contributes to compromised p21 transactivation.

ATp53 Mutations in the Germ-Line Predispose to Li–Fraumeni Syndrome, but p47 Expression Prognosticates Better Survival in Sporadic Cancers.

Given that p47 expression from ATp53 mutations is capable of inducing cell death, we evaluated the role of these mutations in two clinical aspects: predisposition to cancer and overall survival in sporadic cohorts, the latter being reflective of response to treatment. To ascertain whether these ATp53 mutants that express p47 are predisposing to tumors, we first investigated wether such mutations are found in the germ line of LFS patients. Analysis of the International Agency for Research on Cancer (IARC) database indicated that whereas the frequency of such mutations within the first 44 aa is approximately 0.46% (129 of 27,595 tumors) in somatic cases, it is higher in the LFS cases, being 0.93% (12/1,295) (Fig. 6A) (30). This trend is generally similar to the frequency of mutations in the carboxyl terminus of p53 (amino acids 340–393) (somatic vs. germ line: 0.85% vs. 2.6%), which is another less frequently mutated region of p53. By contrast, the frequency of DBD mutations is reduced from 93.79% in somatic cases to 81.25% in the germ-line cases (Fig. 6A). To determine whether the somatic AT mutations are significant and not artifactual, we used the Li–Fraumeni rates as a population control because these mutations are passed on to the subsequent generations and confirmed in the progeny, and performed statistical analyses to determine the confidence interval (CI) for the population mutation rates based on the “exact” Clopper–Pearson method. The data are shown below in the SI Appendix, Table S2. Essentially, the 99% CI for the mutation rate in the amino acid 1–44 region of the somatic tumors was (0.37%, 0.58%), based on the Li–Fraumeni status. Similarly, the mutations rates in the other domains were also in the 99% CI range, thereby indicating that the mutation rates are statistically significant and are true mutations. We also further analyzed a large cohort of French LFS families collected over 20 y, which contained several families with ATp53 mutations (Fig. 6B). All these mutations were capable of leading to premature termination of full-length p53 expression (similar to Fig. 1B), and the patients carrying these mutations were predisposed to early tumors. Importantly, these mutations were passed on through the germ line to subsequent generations, who also succumbed to spontaneous tumors (Fig. 6C), similar to the classical DBD mutations, indicating that ATp53 mutations that retain apoptotic potential predispose carriers to spontaneous tumors.

Fig. 6.

ATp53 mutations lead to Li–Fraumeni syndrome. (A) Quantification of somatic or germ-line mutations within the various regions of the p53 gene in human cancers is shown. Data were analyzed from IARC database for p53 mutations. (B and C) Characteristics of French LFS families with ATp53 mutations leading to stop codon within amino acid 40 region is shown (B). Pedigree of family 36 is shown (C). Respective age of patients and cancer types are indicated.

Second, we analyzed the effects on overall survival. However, because the number of patients with such ATp53 is low, we took advantage of the fact that a p53 transcript that retains intron 2 (i.e., p53EII) has been demonstrated to produce significant amounts of p47 (29). To confirm the aforementioned, we generated these constructs, which showed that p53EII indeed led to the expression of both full-length p53 and p47 (Fig. 7A). Thus, we screened a panel of 70 early-stage colorectal samples (31), for p53EII expression, which revealed distinct clustering of groups of patients with high or low p53EII levels (p53EII^Lo or p53EII^Hi) (Fig. 7B). Analysis of their overall survival indicated that the p53EII^Hi patients with wild-type full-length p53 showed a trend of better survival compared with the p53EII^Lo patients (mean survival in months: p53EII^Lo vs. p53EII^Hi of 73.171 vs. 83.151) (log rank test, P = 0.06) (Fig. 7C). Importantly, we had identified four tumors in this cohort that had ATp53 indels that would lead to loss of p53 expression and gain of p47 expression (i.e., c.97delT; c.97insG; p.Q38*; and c.97_115del). The mean survival of these four patients was even higher, being 89.5 mo (Fig. 7C), altogether suggesting that p47 expression could lead to better overall patient survival. These data are therefore consistent with ATp53 mutants’ ability to regulate apoptosis, because p53EII^Hi-expressing tumors prognosticate better survival of patients.

Fig. 7.

Clinical effects of ATp53 alterations. (A) Schematic shows p53 and p53EII transcripts (Top). H1299 cells were transfected with the indicated constructs, with either the G or C polymorphism in codon 72, and analyzed by immunoblotting with pan-p53 polyclonal CM1 antibody. (B and C) p53EII mRNA levels were analyzed by quantitative real-time PCR in colorectal samples. Relative levels to gapdh are shown for a subset of samples, to highlight the p53EII^Lo (relative levels <1.5) and the p53EII^Hi (relative levels >1.5) groups (B). Kaplan–Meier plots of overall survival of patients from this cohort are shown. (C, Top). Table shows mean survival (in months) of patients with p53EII^Hi (relative levels >2.0), p53EII^Lo, or with ATp53 indels (C, Bottom). (D and E) Model for p47 generation by ATp53 indels and its role in regulation of cell death. ATp53 indels generate a stop signal (indicated by “xxxx”) and, thus, disrupt the expression of full-length p53. Translation from the next methionine therefore leads to the generation of the p47 isoform (D). Under naïve conditions, p47 is generally abundant and nuclear, and is found to be bound onto target promoters, although target genes are not activated. By contrast, p53 is a labile protein that is degraded, and is minimally bound to target promoter DNA (E, Top). Upon stimulation by various stress agents, p47 is activated, leading to accumulation without any increase in protein stability. However, p47 is not acetylated at K382 because of lack of the amino-terminal domain, and does not recruit TAF1 to targets such as p21, resulting in inefficient p21 expression. Nonetheless, hCAS1 is recruited to apoptotic target promoters such as Aip1, even in the absence of K382 acetylation, thereby leading to their activation. Cellular stimulation, however, leads to stabilization of p53, which is the primary mechanism for increasing its abundance (E, Bottom). Stabilized p53 is acetylated on K382 and is able to recruit TAF1 to arrest target promoters. Concomitantly, p53 is also able to bind to apoptotic target promoters, recruiting hCAS1 and result in their activation. Red dots represent acetylation on K382.

Discussion

The data presented here defines a subgroup of alterations in the amino terminus of p53 that while resulting in abrogation of full-length p53, leads to the expression of the truncated p47 form. Functional consequences of p47 expression in such cases indicate that it is capable of inducing apoptosis upon exposure to a variety of stress signals, albeit independent of protein stabilization, through the selective activation of some apoptotic genes. The inability to activate p21-like genes stems from lack of acetylation on K382 that requires the presence of the amino terminus. Clinically, the preservation of selective apoptotic function by p47 suggests better prognosis and survival (this report and ref. 32), although these findings require further large cohort validation. Nonetheless, the data indicate that cancer patients with ATp53 indels may have a higher possibility of being responsive to therapeutic treatments. Moreover, the presence of the apoptotic potential appears insufficient to prevent tumorigenesis.

The p53 amino terminus contains several regulatory regions, including the TA domains and binding sites for multitude of proteins such as MDM2 and p300. Interestingly, the ATp53 indels appear to affect most of these functions. For example, a noteworthy finding is that p47 is not acetylated on K382 upon stimulation, unlike full-length p53, suggesting a critical requirement for the amino-terminal domain that contains the p300-binding site in this process (14). Lack of p47 K382 acetylation correlates with lack of TAF1 recruitment to nonapoptotic target gene promoters like p21, but does not affect the activation of apoptotic target genes. Moreover, acetylation-defective full-length p53 mutants (L22Q, W23S, and K382R) were also partially defective in p21 activation, highlighting the requirement for K382 acetylation in this process. Consistent with these findings, acetylation on K382 of full-length p53 was shown to be required for binding to the bromodomain of TAF1, recruiting the latter to the p21 promoter (22). Furthermore, data from genetically modified mice in which the TA1 was mutated showed similar defects in arrest gene activation (i.e., p21) without much effect on apoptotic genes like Bax (20). Thus, it appears that the selectivity of p47 to induce apoptotic target genes is essentially by virtue of its inability to induce the other class of p21-like genes. Coexpression of mutant full-length p53 was found to be capable of inducing K382 acetylation on p47 in trans, and lead to p21 induction, further demonstrating the role of K382 acetylation for p21 activation. Nonetheless, apoptotic target genes are induced by efficient recruitment of the apoptotic cofactors such as hCAS1 to their promoters upon stress stimulation, both in the context of p47 or full-length p53 to similar extents. The preceding also explains the lack of activation of these apoptotic target genes in the absence of stimulation in RKO^−/− cells, even as p47 was generally found to be bound to target promoters and present in the nucleus.

Consistent with the fact that multiple stimuli are able to induce p47 expression, our data show that p47 is required and sufficient to induce cell death, even in the absence of full-length p53. p47’s role in cell death has been controversial, with the initial reports suggesting that it does not have the ability to induce death (15), but subsequent reports suggesting that it could promote cell death and apoptotic target gene activation (16, 19). Data presented here shows that p47 overexpression in various cell lines induces cell death to an extent as seen with full-length p53. Moreover, silencing p47 expression in RKO^−/− cells led to reduction of cell death, indicating a contributory role for p47 towards this process. Not surprisingly, p47 was found to be capable of efficiently inducing some of the p53 apoptotic targets genes, whose expression is reduced when p47 was silenced, demonstrating a link between p47 and apoptotic target gene expression. Supporting this observation, a DNA-binding p47 mutant is completely devoid of such activity and is incapable of inducing apoptosis, highlighting a crucial role for transactivation by p47 in this process.

It is noteworthy that whereas p47 is capable of inducing the expression of some apoptotic target genes like Fas, Dr5, Api1 and Pig3, it is defective in turning on the other classical apoptotic targets such as Puma and Noxa. Although the mechanistic basis and the consequence of this bifurcation is unclear, it is interesting to note that Fas and Dr5 regulate the extrinsic apoptotic pathway whereas Puma and Noxa and regulate the intrinsic apoptotic pathway (33). Aip1 and Pig3 are associated with dissipation of mitochondrial potential and related to oxidative stress, respectively (34, 35). Whether the p47 form is mainly associated with executing the extrinsic pathway of apoptosis is worth further investigation. Nonetheless, these data, together with the earlier reports that have highlighted that loss of Puma or Noxa (36), or loss of apoptotic and cell cycle/senescence activities (37) does not lead to abrogation of p53 tumor suppressor functions, suggest that the apoptotic potential of p53 is dispensable for its tumor-suppressive properties.

Finally, the observation that ATp53 alterations are also found in the germ line in LFS patients alludes to the fact that functional loss of the amino-terminal region is sufficient to predispose to tumorigenesis. Interestingly, mice in which the TA1 was incapacitated were not tumor prone, which led to the conclusion that targets genes regulated by TA1 were dispensable for tumor suppression. However, loss of both TA1 and TA2 led to spontaneous tumors in mice (21), indicating that transcriptional activity of full-length p53 is essential for tumor suppression. Nonetheless, given that acquisition of the ATp53 indels can predispose to cancers in humans indicate that the amino-terminal region, beyond its ability to transactivate target genes from TA1, may be required for suppression of tumorigenesis. Understanding the mechanistic basis of this observation, by generation ATp53 knock-in mice, will provide important clues in understanding whether full-length p53’s tumor suppressive properties are resident only within its transactivation capabilities or beyond.

In summary, the data presented here demonstrates a mechanism of p47 expression in human cancers with ATp53 alterations, through alternate start from the subsequent methionine (Fig. 7D). p47 is sufficient and is necessary for apoptosis, through the selective activation of some apoptotic target genes. The selectivity in target gene activation appears to be a default feature, occurring in the absence of K382 acetylation that depends on the amino-terminal domain, which is required for arrest related gene activation (Fig. 7E). Importantly, these data conclude that selective apoptosis proficiency of full-length p53 is insufficient to protect against tumorigenesis.

Methods

Standard cellular and molecular biology techniques of cloning, cellular transfections, cell death, immunoblotting, and PCR were used for the experiments. In essence, clinical samples obtained with SingHealth Centralized Institutional Review Board B ethics approval were used for p53EII transcript analyses and determination of p53 mutation status by standard sequencing and PCR, and also for immunoblot and immunohistochemical analyses, as described in detailed in SI Appendix. All cell culture and biochemical work, transfections, target gene analysis by quantitative real-time PCR, and chromatin immunoprecipitations were performed as described (28, 38, 39).