p53: Multiple Facets of a Rubik’s Cube

Abstract

The p53 tumor suppressor has been studied for decades, and still there are many questions left unanswered. In this review, we first describe the current understanding of the wild-type p53 functions that determine cell survival or death, and regulation of the protein, with a particular focus on the negative regulators, the murine double minute family of proteins. We also summarize tissue-, stress-, and age-specific p53 activities and the potential underlying mechanisms. Among all p53 gene alterations identified in human cancers, p53 missense mutations predominate, suggesting an inherent biological advantage. Numerous gain-of-function activities of mutant p53 in different model systems and contexts have been identified. The emerging theme is that mutant p53, which retains a potent transcriptional activation domain, also retains the ability to modify gene transcription, albeit indirectly. Lastly, because mutant p53 stability is necessary for its gain of function, we summarize the mechanisms through which mutant p53 is specifically stabilized. A deeper understanding of the multiple pathways that impinge upon wild-type and mutant p53 activities and how these, in turn, regulate cell behavior will help identify vulnerabilities and therapeutic opportunities.

INTRODUCTION

The complexity of the p53 tumor suppressor pathway is astounding. Although p53 is under the radar most of the time, in response to DNA damage or cellular stress, such as metabolic, oxidative, or ribosomal stress, p53 functionally activates cellular processes that include cell cycle arrest, senescence, and cell death. In this scenario, p53 is responsible for maintaining homeostasis by repairing or eliminating cells with a damaged genome (Aylon & Oren 2016). Thus, it is the gene most often mutated in human cancers. However, in contrast to other tumor suppressors that are often deleted in cancers, the most common p53 mutations are missense mutations that result not only in loss of transcriptional activity but also in gain-of-function (GOF) activities that lead to aggressive tumor cell behaviors. Additionally, p53 has a crucial function in cell survival by promoting autophagy and metabolic restructuring in the face of starvation. The signaling pathways through which p53 functions are cell type– and context-specific, although the processes by which this occurs are still murky. This review focuses on current concepts that have evolved from in vivo studies of the p53 pathway.

p53: A STRESS RESPONSE PROTEIN

The major activity of p53 is as a transcriptional activator of a large repertoire of genes through its recognition and binding to a specific DNA sequence (Vousden & Prives 2009). p53 is present at barely detectable levels in normal cells, but in response to stimuli such as DNA damage, ribosomal stress, metabolic stress, and other abnormalities, p53 is posttranslationally modified, leading to rapid stabilization and subsequent transactivation of hundreds of genes with multiple cellular functions (Vousden & Prives 2009).

p53 transcriptionally activates multiple genes that ultimately determine whether a cell lives or dies (Figure 1) (Kenzelmann Broz et al. 2013). The first identified target of p53 was the cell cycle inhibitor CDKN1A, commonly referred to as p21, with important roles in cell cycle arrest and senescence (Brown et al. 1997, Brugarolas et al. 1995, Bunz et al. 1998, Chang et al. 1999, Deng et al. 1995, El-Deiry et al. 1993, Harper et al. 1993, Waldman et al. 1995). p21 remains one of the p53 targets most robustly activated in response to stress and, thus, serves as a great reporter. However, some p21-null mice develop tumors at a late stage (Martin-Caballero et al. 2001), and p21 is infrequently deleted in human cancers (Shiohara et al. 1994), indicating that it is not a potent tumor suppressor on its own. Perhaps the activation of p21 occurs to slow down the cell cycle enough so that cell fate can be determined. A mouse model recently developed with luciferase knocked into the p21 promoter will be useful for monitoring the p53 response in vivo (Tinkum et al. 2011). Senescence has also been examined in mouse tumor models. Restoration of p53 activity in hepatocellular carcinomas and in some soft tissue sarcomas that develop in mice in the absence of p53 resulted in tumor clearance via senescent mechanisms (Ventura et al. 2007, Xue et al. 2007). In addition, senescence occurs in response to doxorubicin treatment in a mouse model of breast cancer that retains wild-type p53 ( Jackson et al. 2012). Breast tumors with mutant p53 treated with doxorubicin did not senesce, which allowed tumor cells to traverse the cell cycle with double-strand breaks, leading to a mitotic catastrophe and improved outcomes. Senescence remains a poorly understudied p53 response that has important implications for the management of human cancers.

Figure 1.

p53 is activated in response to various stresses. In different contexts, the activation of p53 can either keep the cells alive (the survival pathway) or kill the cells (the death pathway).

In addition to cell cycle arrest and senescence, p53 can initiate other prosurvival mechanisms under certain conditions, such as autophagy in cells with limited nutrient availability, which destroys unnecessary or dysfunctional intracellular components (White 2016). Autophagy, in turn, downmodulates p53 functions to prevent cell damage and tissue degeneration (White 2016). In response to catastrophic changes in metabolism, p53 also remodels metabolic pathways for survival (Berkers et al. 2013). For example, p53 activates genes such as AMPKβ, TSC2, and PTEN, which inhibit mTOR (a sensor of nutrient supply) signaling, and TIGAR, among others, which temper aerobic glycolysis and promote oxidative phosphorylation. Finally, the ability of p53 to signal its own destruction via the activation of a potent inhibitor, Mdm2 (murine double minute 2), referred to as a feedback loop, restores normality. This ability to regulate pathways that sense and manage nutrient availability for cell survival is an important p53 response that ensures homeostasis.

p53 is also a potent inducer of different forms of cell death. In response to DNA-damaging ionizing radiation, p53 induces apoptosis in the thymus and developing central nervous system in vivo, a phenotype absent in p53-null mice (Lee et al. 2001, Lowe et al. 1993). Unrestrained p53 activity in the developing embryo through Mdm2 deletion induces apoptosis, resulting in embryonic lethality (Chavez-Reyes et al. 2003, Jones et al. 1995, Montes de Oca Luna et al. 1995). Another p53-induced cell death mechanism recently discovered, ferroptosis (an iron-mediated nonapoptotic cell death), has been implicated in tumor suppression ( Jiang et al. 2015). In addition, cells expressing a p53 polymorphism (P47S) are partially impaired for transcriptional activation and are resistant to ferroptosis ( Jennis et al. 2016). Thus, p53 functions in a cell similar to a manager of an assembly plant whose primary role is to stay vigilant, monitor the workflow, and slow down various processes to maintain a functioning facility (or cell). Importantly though, p53 as the manager has the ability to shut down everything. Further studies of the molecular factors that determine which p53-dependent response is activated and when are warranted.

Mouse models have probed the importance of cell cycle arrest, senescence, and apoptosis as mechanisms of tumor suppression. A p53 missense mutant, p53R175P (R172P in the mouse), observed in a few human cancers that eliminates the apoptotic function of p53 but retains a partial cell cycle arrest phenotype showed delayed tumor development, implicating cell cycle arrest as a tumor suppressive mechanism (Liu et al. 2004). A p53 transactivation domain mutant, p53^25,26, was compromised for both G1 arrest and apoptosis, but induced senescence and delayed tumor development in an HrasV12 lung model (Brady et al. 2011), implicating senescence as a tumor-suppressive mechanism. Moreover, other mouse mutants indicated that depletion of all these phenotypes also delayed tumor development. Mice with deletion of p21 (a cell cycle arrest and senescent target of p53) and Puma and Noxa (apoptotic targets of p53) showed the absence of a tumor phenotype at a time point where all p53-null mice have died, implicating other unknown mechanisms in tumor suppression (Valente et al. 2013). 3KR mice defective in lysine acetylation of three amino acids also lacked p53-dependent senescence and apoptosis and showed delayed tumor phenotypes (Li et al. 2012). These mice retained the ability to transcriptionally activate some p53-dependent metabolic genes, implicating the regulation of metabolism as a tumor-suppressive mechanism.

The activities mentioned above make p53 a potent tumor suppressor, and p53 is the most frequently mutated gene in human cancers (Kandoth et al. 2013). With all of these tumor-suppressive properties, the discovery that p53-null mice were viable was remarkable (Donehower et al. 1992, Jacks et al. 1994). This observation suggests that development has built-in redundancies, so p53 loss results in normal embryos for the most part (some p53-null fetuses succumb to exencephaly, which is a rare malformation of the neural tube with massive protruding brain tissue). Thus, perhaps the normal developing embryo is somehow protected by other players such that the damage that most cells receive is not a lethal event. Or, perhaps during embryogenesis, there are more cells than required so that damaged cells can be eliminated. The developing embryo also has a readily available nutrient supply, but neonates are extremely dependent on autophagy through p53 (White 2016).

REGULATION OF WILD-TYPE p53

p53 is regulated at different levels by numerous players. Among all of the regulators, the Mdm family of proteins has key roles and is the most well studied in terms of p53 regulation. Mdm2 is an E3 ubiquitin ligase that ubiquitinates and targets the p53 protein for proteasomal degradation (Pant & Lozano 2014). Mdm2 also inhibits p53 activity by directly binding to the p53 transcriptional activation domain 1 (TAD1) (Kussie et al. 1996) and domain 2 (TAD2) as indicated by a recent study (Shan et al. 2012). The importance of this relationship is clear because Mdm2 loss in mice leads to embryonic lethality due to induction of apoptosis at preimplantation, a phenotype that is rescued by the concomitant deletion of p53 (Chavez-Reyes et al. 2003, Jones et al. 1995, Montes de Oca Luna et al. 1995). Mdm4 (christened MdmX) is closely related to Mdm2 and binds to and inhibits the activity of the p53 transactivation domain (Shvarts et al. 1997). However, unlike Mdm2, Mdm4 does not have an enzymatically active RING domain for ubiquitination. Instead, the RING domain of Mdm4 binds Mdm2 and enhances the E3 ligase activity of Mdm2 (Sharp et al. 1999, Tanimura et al. 1999). Similar to Mdm2 loss in mice, Mdm4-null mice exhibit p53-dependent embryonic lethal phenotypes (Finch et al. 2002, Migliorini et al. 2002, Parant et al. 2001). In addition, mice with the deletion of, or a point mutation in, the RING domain of Mdm4, which disrupts interactions with Mdm2, also have embryonic lethality due to increased p53 activity, demonstrating the importance of the Mdm2–Mdm4 interaction during development (Huang et al. 2011, Pant et al. 2011). Lastly, a mouse in which only the E3 ligase activity of Mdm2 is disrupted but the interaction with Mdm4 (Mdm2^Y487A/Y487A) is retained is viable, suggesting that the ability of Mdm2 and Mdm4 to mask the TAD of p53 is sufficient to keep p53 levels low (Tollini et al. 2014). However, DNA damage caused mortality of the Mdm2^Y487A/Y487A mouse, which is p53 dependent. These models, among others, show that slight decreases in Mdm2 and Mdm4 activity maintain homeostasis and are compatible with life, but this activity is insufficient to prevent lethality upon DNA damage (Eischen & Lozano 2014, Gannon & Jones 2012). These in vivo studies emphasize important roles for the Mdm proteins in inhibiting p53 during development and in responding to DNA damage.

Mouse tumor models also indicate that Mdm2 and Mdm4 are drivers of tumorigenesis. A transgenic mouse model that exhibits a two- to fourfold increase in levels of Mdm2 is tumor prone, and tumor incidence increases upon haploinsufficiency of p53 ( Jones et al. 1998). Similar to the case with Mdm2, the pathological consequences of Mdm4 overexpression have been modeled using transgenic mice. Two independent Mdm4 transgenic lines developed spontaneous tumors, with sarcomas being the most frequent tumor type (Xiong et al. 2010).

As indicated above, numerous tumors either delete or mutate the p53 gene, driving tumor initiation and progression. As MDM2 and MDM4 are potent p53 inhibitors, they are frequently overexpressed or amplified in a myriad of tumors. In fact, alterations of p53 and increased levels of MDM2 or MDM4 are mutually exclusive events in many cancers, underscoring the importance of the p53 pathway in tumor suppression (Wasylishen & Lozano 2016). However, there is some tissue specificity in terms of alterations to the p53 pathway in human cancers. Alterations of the p53 gene occur in nearly 100% of ovarian serous cystadenocarcinomas and in 90% of small-cell lung cancers, but p53 mutations are absent or rare in liposarcomas and retinoblastomas. Instead, MDM2 is amplified in 80% of liposarcomas and MDM4 is amplified or overexpressed in 65% of retinoblastomas (Wasylishen & Lozano 2016). The reasons for the differences in how the p53 pathway is inactivated in tumor development are not well understood, but may reflect altered accessibility to genes resulting from chromatin modifications. Nonetheless, the p53 pathway is nonfunctional in the development of most human cancers. Differences in p53 pathway inactivation also mean that treatment options will vary. For example, tumor cells with increased MDM2 or MDM4 levels retain wild-type p53, and MDM2 inhibitors that release p53 from its shackles, such as Nutlin, have been developed and are being tested. Some antitumor activity has been observed. However, identifying the optimal dose schedule for the inhibitors is needed to achieve robust p53 activation and, at the same time, to minimize toxicity (Cheok & Lane 2016, Wang et al. 2016).

TISSUE- AND STRESS-SPECIFIC p53 ACTIVITY

Numerous p53 target genes have been identified, many of which are cell type– and tissue-specific. A p53-mediated, tissue-specific response correlates with the differential induction of p53 target genes. For example, in a study in which p53 was stabilized through global somatic deletion of Mdm2, p21 was elevated almost 200-fold in the kidney compared with 7- and 12-fold in liver and spleen, respectively (Zhang et al. 2014). And Puma, another p53 transcriptional target, was elevated 18-fold in the kidney, but was not elevated in the liver. In another study, upon γ-radiation, four apoptotic targets of p53, DR5, Bid, Puma, and Noxa, were induced in the jejunum and ileum, the two tissues most sensitive to radiation (Fei et al. 2002). In addition, differences in p53 target gene activation between the transverse and descending colon have been observed. Strikingly, even within the same tissue, different compartments exhibit distinct regulation of p53 activity. In response to γ-radiation, Puma was induced in the splenic white pulp, whereas Noxa and Bid were induced in the red pulp (Fei et al. 2002). The induction of these genes depended on p53 and correlated with the activation of caspase 3 in both compartments. Finally, in the liver, where radiation did not lead to caspase 3 activation, p21 appears to be the major p53-dependent target gene activated (Fei et al. 2002). These data indicate that cell- and tissue-specific differences exist with regard to the activation of p53 targets (Figure 2), which may determine the specific fate of different cells or tissues in response to stress.

Figure 2.

Mdm2 and Mdm4 are potent inhibitors of the p53 tumor suppressor. Loss of either or disruption of the interaction between the two results in embryo lethal phenotypes (red). Mutation of the RING domain of Mdm2 alone is viable. In response to Mdm2 loss or radiation in the adult mouse, p53 activates numerous genes (see the section titled Tissue- and Stress-Specific p53 Activity). Major genes with important tissue-specific roles are listed (labeled in green when activated by radiation), as identified in studies discussed in this review.

In addition, p53 transcriptionally activates the leukemia inhibitory factor (LIF) gene, which encodes a cytokine critical for embryo implantation (Hu et al. 2007). p53-null females exhibit implantation defects that are alleviated by administration of exogenous LIF. Although other tissues express LIF upon γ-radiation in a p53-dependent manner, the only defect observed in LIF-null females is failure to implant blastocysts (Hu et al. 2007, Stewart et al. 1992). These studies highlighted the importance of a single p53 target for a specific tissue: the uterus. Hypomorphic Mdm2 mice (that retain Neo in intron 1 in feedback loop–defective mice) with decreased Mdm2 levels also exhibit unique p53-dependent pathologies, such as increased skin pigmentation of the epidermis, reduction in seminiferous tubules, decreased spermatocytes in the testes, and a decreased number of ovarian follicles (Pant et al. 2016). None of the defects are rescued by deletion of p21, but the reproductive phenotypes are rescued by Puma loss, again highlighting that there are dose- and tissue-specific differences in the p53 response. Importantly, the deletion of a single, downstream target of p53, Puma, reverted a p53-dependent reproductive phenotype, whereas the skin phenotype appears to be driven by Kit ligand signaling. The Kit ligand is another p53 target expressed in melanocytes, and it is linked to their migration and accumulation (Zeron-Medina et al. 2013). These studies showed, for the first time, that the depletion of distinct p53 targets has specific outcomes in vivo. Combined, the data suggest a cell context–dependent differential induction of p53 targets, which may not be limited to the genes being examined. In fact, in another study, human HCT116 colorectal cancer cells and diploid lung fibroblast IMR90 cells showed distinct p53 genome-wide binding profiles (Botcheva & McCorkle 2014). A deeper knowledge of the p53 response in vivo may identify downstream effectors that could be used therapeutically to treat specific cancers.

Tissue- or cell-specific cofactors, which regulate p53 activity through protein–protein interactions, may also contribute to the tissue- or cell-specificity of p53 activity. A recent study showed that the carboxy terminus (C terminus) of p53 regulates gene expression in a target- and tissue-specific manner. Hamard et al. (2013) generated a murine p53 allele lacking the C-terminal 24 amino acids (p53ΔCTD). In bone marrow and thymus, p53ΔCTD is hyperactive. It increases senescence and p21 transcription in bone marrow, and it induces apoptosis and the transcription of Puma and Noxa in the thymus, suggesting that the C terminus dampens p53 activity in those tissues. In liver, however, p53ΔCTD exhibits impaired activity. In the spleen, the p53ΔCTD protein is overexpressed, suggesting that the C terminus has a role in controlling p53 protein levels in this tissue. Surprisingly, the overexpressed p53ΔCTD protein does not lead to any senescence or apoptosis phenotypes in the spleen. One explanation for the above intriguing results is that the p53 C terminus binds to cell- or tissue-specific cofactors, endowing specific p53 activity and leading to tissue-specific phenotypes. Alternatively, the p53 C terminus, which is essential for maintaining the protein structure of p53, might indirectly regulate the interaction between non-C-terminal domains of p53 and cell-specific binding partners. New in silico experiments have provided structural information about p53 in its tetrameric form and in interaction with a consensus DNA-binding site, indicating that the C terminus molds the DNA-binding domain and allows a better fit between p53 and DNA (D’Abramo et al. 2016). It will be interesting to examine whether this induced-fit mechanism is context dependent and how it is regulated, which may help understand why the p53ΔCTD protein induces the above tissue-specific phenotypes.

In addition, different DNA damage signals seem to deploy distinct mechanisms to induce functionally active p53. Early studies showed that UV, but not γ-radiation, was able to induce phosphorylation at the C-terminal serine 392 of human p53 (serine 389 in mouse p53), which significantly increased its DNA-binding activity (Kapoor & Lozano 1998, Lu et al. 1998). Phosphorylation of p53 has been implicated in regulating protein–protein interactions. As UV and γ-radiation cause different kinds of DNA lesions (bulky adducts versus strand breaks, respectively) and trigger different repair machinery (nucleotide excision repair versus base excision repair, respectively), it is possible that different site-specific modifications of p53 allow differential interaction with proteins involved in a specific DNA repair pathway. In addition, although both UV and γ-radiation stabilize p53, ubiquitin–p53 conjugates are detected only in γ-irradiated RKO and U2OS cells (Maki & Howley 1997), suggesting that the UV, but not γ-radiation–induced stabilization of p53 results from a loss of p53 ubiquitination.

p53 ACTIVITY AS A FUNCTION OF AGE

The role of p53 in aging remains controversial. Mice carrying an additional copy of the whole p53 genomic region were remarkably resistant to tumor development but aged normally (Garcia-Cao et al. 2002). In contrast, mice possessing increased constitutive activity of p53, through the expression of an artificial (p53m) or naturally occurring (p44) amino-terminal truncated form of p53 in the presence of wild-type p53, display early aging-associated phenotypes and have a shortened life span (Maier et al. 2004, Tyner et al. 2002). In contrast, mice carrying C-terminal truncated p53, p53ΔCTD (Hamard et al. 2013) or p53Δ31 (Simeonova et al. 2013), do not have an aging phenotype, even though they exhibit increased p53 activity in a lot of tissues. Similarly, mice with mutated p53 response elements in the Mdm2 promoter display an elevated p53 DNA damage response compared with normal mice, but age normally (Pant et al. 2013). These discrepancies indicate there are special roles in aging for the short forms of p53 that lack the amino terminus. Indeed, Maier et al. (2004) showed that activation by p44—a p53 protein that lacks the amino terminus owing to initiation at an internal methionine (codon 41)—altered p53-dependent gene expression, including increasing IGF-1 signaling. Persistent IGF-1 signaling caused sustained ERK activity, resulting in senescence. In addition, in mice, p44 promoted the phosphorylation of tau, the microtubule-binding protein (Pehar et al. 2010, 2014). Noteworthy, the hyperphosphorylation of tau plays critical parts in the development of neurodegenerative disease (Wang & Mandelkow 2016). Because the p53 amino terminus provides interacting sites for regulatory proteins, such as Mdm2 and Mdm4 (Chillemi et al. 2016), it is possible that disrupting these interactions plays an important part in aging.

On the other hand, emerging evidence suggests there is an age-dependent decrease in p53 activity. Mice treated with γ-radiation exhibited an age-associated decline in p53 functionality (Feng et al. 2007). In the same study, the function of the ataxia telangiectasia mutated (ATM) kinase was found to decline significantly with age as well, which may explain the decline of the p53 response to radiation because ATM phosphorylates and activates p53 (Feng et al. 2007). Thus, it has been suggested that the age-associated decline in ATM and p53 signaling functions in mice provides a potential mechanism for age-associated increases in cancer rates. In another study, in which Mdm2 was deleted globally in adult mice, the loss of Mdm2 caused less severe symptoms in old adult mice compared with those in young mice (Zhang et al. 2014). These pathologies resulting from Mdm2 loss were p53-dependent. At a molecular level, the transcripts of Bbc3 (Puma), Cdkn1a ( p21), and all of the senescence markers examined were much less efficiently activated in old mice. Interestingly, chromatin immunoprecipitation (ChIP) assays showed that in old mice p53 had poorer access to its binding sites on target promoters, which could, at least partially, account for the decreased sensitivity of p53 (Zhang et al. 2014). Epigenetic changes in nucleosomes occurring as a function of age are a potential reason for the decreased accessibility of p53 to its target promoters in old mice. In fact, various studies have demonstrated that the deregulation of DNA methylation is a common feature of aging. During the adult life span, methylation tends to decrease in repetitive elements (Bollati et al. 2009, Talens et al. 2012), but increase in the CpG islands of many key developmental genes (Bell et al. 2012, Maegawa et al. 2010, Rakyan et al. 2010, Teschendorff et al. 2010). The changes in p53 regulators occurring with age offer other potential reasons for the age-dependent decline of p53 activity.

GAIN-OF-FUNCTION ACTIVITIES OF MUTANT p53

Missense mutations account for approximately 75% of all p53 mutations in human cancers (Olivier et al. 2010, Soussi & Lozano 2005). A single base change can alter an amino acid, thus eliminating or severely compromising wild-type p53 transcriptional activity. Modification of any of five arginine residues that are hot spots for p53 mutations in human cancers disables p53 sequence-specific DNA binding (Cho et al. 1994). Moreover, signature mutations in the DNA-binding domain (some of which occur at these same hot spots) can arise through exposure to common insults, such as UV light, smoking, and exposure to aflatoxin or other chemical carcinogens (Olivier et al. 2010, Vogelstein & Kinzler 1992).

The fact that most p53 alterations in tumors are missense mutations suggests that cells expressing mutant p53 have an advantage over cells lacking p53 (Brosh & Rotter 2009). Numerous experiments have tested this hypothesis. Early experiments indicated that immortalized cells lacking p53 gained additional transforming properties upon overexpression of mutant p53 proteins (Dittmer et al. 1993). Cells expressing the most common p53 mutants, in contrast to cells lacking p53, also showed increased metastatic potential and invasiveness (Crook & Vousden 1992, Hsiao et al. 1994). In Li–Fraumeni syndrome (LFS), individuals with p53 missense mutations show a higher cancer incidence and an earlier age of tumor onset (9–15 years earlier, depending on the study) than individuals with other kinds of mutations (Birch et al. 1998, Bougeard et al. 2008, Zerdoumi et al. 2013).

The generation of p53 knockin alleles in mice provided direct in vivo evidence for the GOF activities of mutant p53. Knockin mouse models heterozygous for p53^R172H and p53^R270H mutations, which mimic hot spot alterations in human cancers that correspond to, respectively, amino acids 175 and 273, developed highly metastatic and invasive tumors that are not seen in p53^+/− mice (Lang et al. 2004, Liu et al. 2000, Olive et al. 2004). These studies produced two other interesting observations. First, even though p53^R172H heterozygous mice developed highly metastatic tumors, the survival curves were identical to those of p53^+/− mice. Second, analyses of tumors from p53^R172H heterozygous and homozygous mice showed that mutant p53 protein was not necessarily stable, implying that tumor-specific alterations stabilized mutant p53. In fact, analyses of normal tissue from p53^R172H homozygous mice indicated that mutant p53 was not stable in normal cells (Terzian et al. 2008). To understand the implications of these data, the first question asked was whether Mdm2, an E3 ubiquitin ligase that targets wild-type p53 for proteasomal degradation, could also regulate mutant p53. In p53^R172H/R172H; Mdm2^−/− mice, the loss of Mdm2 stabilized mutant p53R172H. This led to decreased tumor latency and increased metastases as compared with p53/Mdm2 double null mice (Terzian et al. 2008). Thus, in this example, the stabilization of mutant p53 led to more aggressive phenotypes than those found in mice lacking p53. Furthermore, mutant p53^R172H mice treated with ionizing radiation or crossed with mice containing an activated K-Ras oncogene also showed more aggressive tumor phenotypes as compared with mice with p53-null alleles (Suh et al. 2011). Together, these studies suggest that stable mutant p53 proteins have additional activities that fuel tumor cell proliferation and invasiveness and support the GOF phenotype of mutant p53 in tumorigenesis.

MECHANISMS OF MUTANT p53 GAIN OF FUNCTION

Several mechanisms have been identified that contribute to mutant p53 GOF activities. The first mechanism discovered showed that mutant p53 proteins suppressed the tumor-suppressive activities of p53 family members p63 and p73 (Gaiddon et al. 2001, Lang et al. 2004, Olive et al. 2004). In addition, TGF-β and EGFR/integrin signaling pathways stabilized mutant p53 (p53R175H and p53R273H introduced into p53-null H1299 cells) and inhibited the function of p63, properties that were essential for the invasive nature of these cells (Adorno et al. 2009, Muller et al. 2009). These studies strengthened the evidence that mutant p53 proteins bind and disrupt p63 activities. However, p63 expression is limited to epithelial cells, and therefore, inhibition of p63 cannot explain mutant p53 GOF in tumors of mesenchymal origin. Moreover, mutant p53 was found to regulate gene expression independently of p63 and p73 in some tumors (Scian et al. 2004, Weisz et al. 2004, Xiong et al. 2014, Zalcenstein et al. 2003).

ChIP-on-chip experiments and expression arrays using SKBR3 breast cancer cells with the p53R175H mutation identified mutant p53 complexes with the vitamin D receptor (VDR), which augmented expression of survival genes and dampened expression of proapoptotic genes (Stambolsky et al. 2010). Importantly, in these experiments, an intact TAD1 was required. Using expression arrays of MDA-MB-468 (p53R273H) breast cancer cells, Freed-Pastor et al. (2012) identified increased expression of genes encoding several enzymes of the mevalonate pathway. Mutant p53-bound SREBP proteins disrupted the acinar architecture of breast epithelial cells when grown as spheroids. In these experiments, both TAD1 and 2 were required, emphasizing the impact of mutant p53 on transcriptional programs. Transcription profiling in murine pancreatic cancer cell lines with K-Ras and p53R172H mutations identified PDGFRb as an important contributor to invasion and migration properties in culture (Weissmueller et al. 2014). Mechanistically, mutant p53 inhibited the p73–NF-Y complex, which represses PDGFRb expression. In other studies, comparisons of primary osteosarcomas that had metastasized from p53^R172H/+ mice to p53^+/− tumors that lacked metastases identified a unique set of transcriptional changes (Xiong et al. 2014). In this system, mutant p53 bound the transcription factor Ets2 and enhanced the expression of a phospholipase, Pla2g16, the protein product of which induced migration and invasion in culture (Xiong et al. 2014). Similarly, diverse p53 mutants, not the wild type, mediated through Ets2, were found to transcriptionally upregulate nucleotide metabolism genes, leading to the proliferation and invasion of various cancer cell lines (Kollareddy et al. 2015). The suppression of one of the mutant p53 target genes, GMPS, which catalyzes the final step of de novo synthesis of GMP, abrogated the metastatic activity of a breast cancer cell line (Kollareddy et al. 2015). Lastly, ChIP sequencing experiments using LFS fibroblasts with the p53R248W mutation identified numerous promoters enriched with mutant p53 (Do et al. 2012).

Emerging evidence implicates the contribution of chromatin remodeling to the GOF phenotype of mutant p53. The SWI–SNF chromatin remodeling complex (CRC) interacts with mutant p53 proteins. The SWI–SNF CRC–mutant p53 interaction is required for mutant p53–mediated VEGFR2 gene promoter remodeling and subsequent optimal VEGFR2 expression (Pfister et al. 2015). Another chromatin-remodeling protein, Pontin, was found to preferentially bind to mutant p53 compared with wild-type p53, as determined by co-immunoprecipitation assays (Zhao et al. 2015). The interaction of Pontin with mutant p53 is crucial for the transcriptional regulation of a group of mutant p53 target genes that are associated with tumorigenesis and metastasis. However, mutant p53 can also modify chromatin by inducing the expression of chromatin regulatory genes. Zhu et al. (2015) showed that p53 mutants, not wild-type p53, bind to and upregulate chromatin regulatory genes, including the genes encoding the methyltransferases MLL1 and MLL2 and the MOZ acetyltransferase, leading to genome-wide increases of histone methylation and acetylation. Genetic ablation or pharmacological inhibition of MLL1 significantly slowed cell proliferation. Furthermore, upregulation of MLL1, MLL2, and MOZ was found in human tumors with p53 mutants, but not in wild-type p53 or p53-null tumors.

In summary, these data suggest that multiple pathways contribute to the GOF phenotypes of cells with mutant p53. The emerging themes by which mutant p53 exhibits its GOF are (a) through forming mutant p53 complexes with other proteins that modify their activities (p63 and p73, for example), (b) by mutant p53 interacting with other transcription factors (VDR, SREBP, and Ets2, for example) that bring a potent transcriptional activation domain to promoters not normally regulated by wild-type p53, and (c) by remodeling chromatin through interactions with chromatin remodeling proteins (SWI–SNF CRC and Pontin, for example) and inducing chromatin regulatory gene expression (MLL1, MLL2, and MOZ, for example). These mechanisms are not necessarily mutually exclusive in the genesis of different cancers and may be context dependent ( Jackson et al. 2011, Oren & Rotter 2010).

In the experiments described above, different systems were used to identify the various mechanisms of mutant p53 GOF. A breast cancer cell line, MDA-MB-468, that expresses p53R273H implicated the mevalonate pathway (Freed-Pastor et al. 2012). A pancreatic tumor cell line from a mouse tumor that expresses p53^R172H and K-Ras^G12D identified PDGFR-β as an important contributor to metastasis (Weissmueller et al. 2014). Two studies, one in an LFS fibroblast cell line that contains the p53R248W mutation and one in an osteosarcoma cell line with the p53R172H mutation (equivalent to p53R175H in humans), implicated Ets2 as a driver of invasion and metastases (Do et al. 2012, Xiong et al. 2014). The SWI–SNF CRC interactions with mutant p53 proteins were identified in multiple breast cancer cell lines with different p53 mutations. Interactions of mutant p53 with Pontin were observed in lung cancer cell lines with ectopic expression of different p53 hot spot mutations and in one breast cancer cell line with the p53R175H mutation. The MLL1, MLL2, and MOZ interactions with mutant p53 were also observed in breast cancer cell lines containing different p53 mutants. Understanding is lacking as to whether these mechanisms are specific to the type of tumor cell being studied or the kind of p53 mutation examined. Other important questions that need to be addressed are which mechanisms drive the development of invasive tumors in vivo, and at what step of the invasion–metastasis cascade does the p53 mutant effect its activities?

In addition to the observations that mutant p53 proteins exhibit GOF activities, a growing body of evidence suggests that cells may be addicted to mutant p53 expression. Experiments using small interfering RNA knockdown of mutant p53 in cancer cell lines showed there was a higher apoptotic response to drug treatment (Maddocks et al. 2013, Oren & Rotter 2010). Mutant p53 depletion experiments showed decreases in cell growth rate, viability, replication, and clonogenicity. The constitutive inhibition of mutant p53 reduced tumor growth in nude mice and reduced stromal invasion and angiogenesis (Berkers et al. 2013). Recently, Prives and colleagues (Freed-Pastor et al. 2012) showed that mutant p53 depletion in breast cancer cells (MDA-MB-231 cells with p53R280K and MDA-MB-468 with p53R273H) in three-dimensional culture led to phenotypic reversion to more normal, differentiated structures with hollow lumens. An in vivo knockin model expressing a conditional, humanized p53R248Q mutation developed tumors, many of which regressed upon Cre-mediated removal of mutant p53 (Allen et al. 2014). In summary, these data suggest that cells with mutant p53 may be addicted to the GOF activities of mutant p53. As mutant p53 proteins have, thus far, proved undruggable (Alana et al. 2014, Basu & Murphy 2016), identifying the mechanisms driving mutant p53 GOF activities is likely to provide options for new, targeted therapies to undermine mutant p53 activities. A deeper understanding of how different p53 mutations contribute to GOF activities is important for developing new therapeutic targets and personalized treatment for cancer patients.

STABILIZATION OF MUTANT p53

Despite having different roles, wild-type and mutant p53 share the majority of regulatory factors (Terzian & Lozano 2012, Vijayakumaran et al. 2015). In this section, we focus on regulation mediated through the chaperone machinery, which in many different studies appears to be specific for mutant p53. Protein stability is the major level at which p53 is regulated. In fact, the stabilization of mutant p53 is a prerequisite for its oncogenic GOF phenotype (Suh et al. 2011, Terzian et al. 2008). The hyperstability of mutant p53 in cancer cells had been previously explained by the lack of a negative feedback loop between mutant p53 and Mdm2. However, mice engineered to express mutant p53 proteins, either with or without the wild-type allele, show high levels of mutant p53 protein expression only in tumor tissue and not in normal tissues (Lang et al. 2004, Olive et al. 2004, Terzian et al. 2008). In addition, in mice with mutated p53 response elements in the Mdm2 P2 promoter, p53 was still degraded, suggesting that the Mdm2–p53 negative feedback loop is dispensable for p53 stability (Pant et al. 2013). Therefore, in malignant cells, there must be an additional mechanism (or mechanisms) to stabilize mutant p53.

Studies have highlighted the roles of chaperone heat-shock proteins (HSPs) in stabilizing mutant p53 in cancer cells. The molecular chaperone machinery, including HSP70 and HSP90, has long been known to interact with mutant p53 (Finlay et al. 1988, Whitesell et al. 1998). Two early studies (Nagata et al. 1999, Peng et al. 2001) demonstrated that HSP90 was involved in the stabilization of mutant p53. More recently, Li et al. (2011b) showed that in contrast to wildtype p53, the mutant p53R172H proteins are engaged in complexes with the HSP90 chaperone machinery in human cancer cells, which blocks endogenous MDM2 and CHIP (C terminus of Hsp70–interacting protein) E3 ligase activities. Importantly, mutant p53 exists in a positive forward loop with HSP90; it facilitates the recruitment of HSF1, the transcriptional activator of HSP90, to its specific DNA-binding elements and stimulates transcription of HSPs, including HSP90, which further stabilize mutant p53 (Li et al. 2014). Indeed, interfering with HSP90 by downmodulating HSF1 disrupts the HSP90 chaperone machinery–mutant p53 complex and liberates mutant p53, allowing endogenous MDM2 and CHIP to degrade mutant p53 (Li et al. 2011b). In another study, HSP70 was found to partially inhibit Mdm2-mediated degradation of mutant p53 in mouse embryonic fibroblasts (Wiech et al. 2012).

Recently, Bcl2-associated athanogene family member 2 (BAG2) was also identified as a binding partner of the p53R172H mutant (Yue et al. 2015). BAG2 contains the BAG domain that interacts with another molecular chaperone, the heat-shock cognate protein 70 (Hsc70), and, therefore, can be considered to be a cochaperone. BAG2–mutant p53 interaction inhibited mutant p53 degradation mediated by Mdm2, and thereby led to mutant p53 accumulation in cancer cells. In addition, knockdown of BAG2 greatly abolished metastatic activity and the proliferation of cancer cells harboring various p53 mutants. Notably, BAG2 expression was found to be elevated in many types of human tumors, including colorectal cancers, lung cancers, breast cancers, and sarcomas (Yue et al. 2015). High levels of BAG2 are associated with poor prognosis in cancer patients and mutant p53 protein accumulation in human tumors.

The identification of mechanisms that protect mutant p53, as shown by the chaperone HSPs, has therapeutic significance as it implicates novel approaches to exposing mutant p53 to its negative regulators and driving its destruction. Indeed, pharmacological inhibition of HSP90 activity with 17AAG (17-allylamino-17-demethoxy-geldanamycin) induces a stronger cell death phenotype in mutant p53 compared with wild-type p53 cancer cells (Li et al. 2011b). In addition, SAHA (suberoylanilide hydroxamic acid)-mediated inhibition of HDAC6, which is an essential positive regulator of HSP90, exhibits preferential cytotoxicity in mutant p53 human cancer cells rather than wild-type and null (Li et al. 2011a). Strikingly, long-term HSP90 inhibition with 17DMAG (17-dimethylaminoethylamino-17-demethoxygeldanamycin) plus SAHA or the potent HSP90 inhibitor ganetespib significantly extends the survival of mice with mutant p53 but not p53-null mice (Alexandrova et al. 2015). Further identification of the mechanisms stabilizing mutant p53 in tumors will provide more opportunities for therapeutic intervention.

CONCLUSIONS

The p53 tumor suppressor is a multifaceted protein whose primary function is to ensure cell homeostasis. It is a potent transcriptional activator that has the ability to either keep cells alive— through cell cycle arrest, senescence, autophagy, or metabolic reprogramming—or kill cells— through apoptosis and ferroptosis (Figure 1). Often, p53 activates transcription in a tissue-specific fashion, and this activity is dampened with age. Tissue specificity needs to be explored further to determine whether any of the downstream targets are linchpins in manifesting p53 activities. However, untethered p53 activity is lethal. How can we harness this activity for tumor destruction in those tumors with a wild-type p53 gene without triggering the lethal side effects that affect normal cells? Mutant p53–expressing cells may be addicted to mutant activities, and the elimination of mutant p53 may also be beneficial for tumor destruction. In this case too, mutant p53 has numerous GOF activities and their physiological significance needs to be explored. Ultimately, targeting some of these activities may have therapeutic potential. It is time to think outside the box and expand beyond p53 to identify other critical components of the p53 tumor suppressor pathway.