p53 : The Most Frequently Altered Gene in Human Cancers

It seems nearly impossible for a normal cell to become a cancer cell unless it inactivates the p53 network. Our current understanding of this network has come from diverse lines of investigation, many initiated by serendipitous findings that only later converged into a coherent picture. This story provides insights into the nature of cancer, as well as the nature of scientific progress.

In 1979, six groups of investigators, each working independently, reported the discovery of a 53 kDa protein that was present in human and mouse cells (DeLeo et al. 1979, Kress et al. 1979, Lane & Crawford 1979, Linzer & Levine 1979, Melero et al. 1979, Smith et al. 1979). In five of these studies, the protein was discovered because it bound to the large T-antigen in SV40 infected cells and was therefore co-immunoprecipitated with antibodies generated against the viral protein. The same protein was discovered serendipitously when an antiserum generated against a chemically-induced mouse sarcoma was found to react with a 53 kDa protein present in transformed but not normal mouse cells (Figure 1).

A variety of studies carried out with the protein, and later with the gene encoding p53, indicated that it was an oncogene (Eliyahu et al. 1984, Jenkins et al. 1984, Parada et al. 1984, Eliyahu et al. 1985). This interpretation reflected both the research climate of the time and apparently compelling experimental evidence. Oncogenes were thought to be the key to understanding cancer, and had been identified in both RNA and DNA tumor viruses. In contrast, the existence of tumor suppressor genes was entirely conjectural, and barely on the radar of most cancer researchers. The p53 protein was bound to the major oncogenic protein of SV40, strongly suggesting that it was a downstream effector of the large T-antigen pathway. This interpretation was consistent with the high levels of expression of p53 found in many cancers. And the piece de resistance was the discovery that the introduction of a "normal" p53 gene into a normal cell could transform it, converting it into a tumor cell. Though there were a few experimental observations that did not fit well with the idea that p53 was an oncogene, there was little reason to believe otherwise in the mid 1980s.

The turning point in p53 research occurred in 1989, as evident in the dramatic change in the slope of the timeline at right (Baker et al. 1989). As so often happens in research, this paradigm shift came from a seemingly unrelated field. In a search for a putative tumor suppressor gene on chromosome 17p, a small region that contained p53 came into focus. To formally exclude this oncogene as the searched-for tumor suppressor, the "two-hit" test was applied (Knudson 1971). Tumor suppressor genes act like the brakes in a car — they stop tumor growth. But because cells have two copies of every brake (one from the father, the other from the mother), it is necessary to alter both copies to get rid of the brake. In contrast, mutant oncogenes act like accelerators that are stuck to the floor of the car; only one accelerator needs to be stuck to make the car continue to go. This distinction allows a "two-hit" test to distinguish whether a mutant gene is an oncogene or a tumor suppressor gene: if both copies of the gene are altered, it is likely to be a tumor suppressor gene; if only one copy is altered, it is more likely to be an oncogene.

When this test was applied to p53, the results were entirely unanticipated. First, the majority of colorectal tumors were surprisingly found to have subtle mutations of p53, generally a single base substitution (such as C to T) resulting in a new amino acid (Baker et al. 1989). Mutations like this in p53 had never been observed before. Second, in virtually all cases, both copies of p53 were mutated. One copy was generally altered by a base substitution and the other copy was often completely deleted from the cell. This was the result expected for a tumor suppressor gene, not an oncogene. This "two-hit" test was then applied to many other tumor types and a similar result was found (Nigro et al. 1989). This result not only catapulted p53 into the center stage of human tumor research, but also provided cogent evidence that p53 was actually a tumor suppressor gene. This conclusion was confirmed by the subsequent findings that patients with inherited mutations of p53 were predisposed to diverse tumor types (Malkin et al. 1990, Srivastava et al. 1990) and that mice with engineered "knock-outs" of the p53 gene were also tumor prone (Donehower et al. 1992, Lowe et al. 1993)). Subsequent studies have demonstrated that p53 is more frequently mutated in human tumors than any other gene in the genome, and > 25,000 mutations have so far been reported (IARCTP53database).

The keys to understanding the mechanisms through which p53 suppresses tumorigenesis have largely come from the study of cultured cells. After initial confusion about the actual sequence of the wild type (wt) p53 gene was resolved, it became clear that the wt p53 gene, but not its mutant forms, could retard the growth of cultured cells when overexpressed (Eliyahu et al. 1989, Finlay et al. 1989). This growth retardation could in principle come from either a reduction in cell birth or an inhibition of cell death.

In a classic study in 1991, wt p53 expression in leukemia cells resulted in their death by suicide (apoptosis) (Yonish-Rouach et al. 1991, Shaw et al. 1992). This observation proved a major stimulant to the burgeoning field of apoptosis and showed that the regulation of cell death was as important to tumor suppressor gene action as the regulation of cell birth. It is now understood that wt p53 regulates both aspects of net cell growth, inhibiting the cell cycle in some circumstances, while promoting apoptosis in others (Figure 2).

In the early 1990s, two seminal observations were made: that p53 had a transactivation domain, and that it could bind to specific DNA sequences (Bargonetti et al. 1991, Kern et al. 1991). It was quickly demonstrated that the full length wt p53 protein functions as a sequence-specific transcriptional activator in vivo and in vitro (Farmer et al. 1992, Funk et al. 1992). Importantly, tumor derived mutants virtually always lost this critical function. Derivation of the first p53 consensus binding site revealed an unusually large and flexible sequence, with key invariant nucleotides (Bargonetti et al. 1991, el-Deiry et al. 1992). The crystal structure of the p53 core domain bound to DNA showed exactly which p53 residues made contact with DNA and revealed the way in which mutations abrogated this binding. Subsequent studies identified many genes that were transcriptionally activated by p53 through p53 binding sites in their regulatory regions. These included the cell cycle inhibitor CDKN128, and the pro-apoptotic proteins PUMA (Nakano & Vousden 2001, Yu et al. 2001) and NOXA (Oda et al. 2000), responsible in part for p53's control of cell growth.

From the early days of p53 research, scientists knew that, in normal unstressed cells, p53 protein is scant, and that its turnover is rapid. Mdm2 was discovered in 1992 to bind to, and negatively regulate, transactivation by p53, and was then itself found to be a transcriptional target of p53, defining a negative feedback loop (Momand et al. 1992, Picksley & Lane 1993). Accordingly, the embryonic lethality of Mdm2 knockout mice could be rescued by knockout of p53 (de Rozieres et al. 2000). Later studies revealed a similarly important role for MdmX, an Mdm2 homolog (Shvarts et al. 1996). Mdm2 proved to be an E3 ubiquitin ligase, stimulating p53 degradation (Haupt et al. 1997, Honda et al. 1997, Kubbutat et al. 1997). It is now recognized that the action of ubiquitinases and deubiquitinases determines the activity of the p53 network. These findings explain the relatively high levels of p53 in tumors, as p53 mutants are transcriptionally inert, disrupting the feedback loop.

The p53 protein is not active in the cell unless it is first modified by other proteins. In other words, the actual mass of p53 is not as important as the amount of activated p53, and only activated p53 can bind to DNA and stimulate the expression of its target genes. Although it was known since the 1980's that p53 levels increase after irradiation, only in 1992 was p53 shown to be regulated by ATM; a kinase orchestrating the DNA damage response (Banin et al. 1998, Canman et al. 1998). The p53 protein was subsequently demonstrated to be phosphorylated after DNA damage, and was the first non-histone protein shown to be acetylated by p300/CBP (Ionov et al. 2004). Biochemical studies showed DNA damage inducible kinases such as ATM (Westphal 1997) and Chk2 (Tominaga et al. 1999, Shieh et al. 2000) can phosphorylate key p53 residues that regulate its binding to Mdm2 and p300/CBP. The alternative reading frame (ARF) protein, known to be induced by a number of mitogens, was also shown to block the ability of Mdm2 to degrade p53, thereby linking p53 to key oncogenic pathways (Kamijo et al. 1998). The p53 protein has been shown to bind to dozens of other proteins, explaining its involvement in a wide range of physiologic processes.

Although there have been tens of thousands of publications on p53, much is still unknown (Figure 3). We still do not understand the microenvironmental conditions that favor the selection of cells with p53 mutations. Is the stimulus continuous or unrepairable DNA damage? Or is it perhaps reactive oxygen species, in association with alternating cycles of hypoxia and normoxia? In similar fashion, we don't yet understand why the expression of wt p53 results in apoptosis in some cells and cell cycle arrest in others, nor how the various post-translational modifications of p53 are related to this switch.

And perhaps most importantly, we don't yet know how to use the immense amount of knowledge so far gained about p53 for therapeutic purposes. Clever approaches to achieve this goal — small molecular weight compounds or peptides that reactivate mutant p53 or disrupt the interactions between MDM2 and wt p53, or viruses that only replicate in cells without a functional p53 network — have been developed and show great promise. However, the field is wide open to new, creative approaches that target p53, a protein that is inactivated in the majority of human cancers. History shows that the most novel ideas — the really bold and creative ones — often come from students.

The study of p53 has revealed many of the principles underlying human tumorigenesis. These include the critical differences between an oncogene and a tumor suppressor gene, the relationship between environmental exposures and cancer, the mechanisms through which cancer genes stimulate cell birth or inhibit cell death, and the striking networks that control the transcription, translation, and function of key cellular proteins. The many facets of these studies, coupled with the fact that p53 inactivation is essential for the formation of the majority of human tumors, has made p53 a uniquely valuable target for basic as well as applied research.