The p53 tumor suppressor protein has received a great deal of attention for its critical role in the prevention of neoplasias. It has been estimated that 80% or more of human cancers have defects in p53 signaling, while roughly half of all such cancers display overt structural alterations of one or both p53 alleles (Lozano and Elledge, 2000). In the last 10 years, enormous strides have been made in understanding the molecular mechanisms by which p53 functions in the cell (Levine, 1997; Giaccia and Kastan, 1998; Sherr, 1998; Prives and Hall, 1999; Vogelstein et al., 2000; Vousden, 2000). The overall picture that has emerged is that the p53 protein responds to a plethora of cell stresses in the dividing cell, integrating the stress signals and effecting one of multiple biological outcomes (Fig. 1).
Stressors that activate p53 include various physical and chemical DNA damaging agents (e.g., ultraviolet and ionizing radiation, chemotherapeutic drugs, reactive oxygen species), hypoxia, and aberrant growth signaling (e.g., activated oncogenes such as Ras or myc) (Giaccia and Kastan, 1998; Ljungman, 2000). These stressors can activate signaling pathways that directly modify p53 through post-translational mechanisms and result in increased stability and activity of the p53 protein, which is usually found at modest levels in a latent state in non-stressed cells (Ashcroft and Vousden, 1999; Jayaraman and Prives, 1999; Meek, 1999; Colman et al., 2000; Appella and Anderson, 2001). The increased stability of activated p53 is in large part due to prevention of its degradation by Mdm2, a protein that mediates p53 ubiquitination and proteolytic destruction (Lozano and Montes de Oca Luna, 1998; Prives, 1998; Momand et al., 2000). This blocking of Mdm2 function can be achieved in at least two different ways. For example, in cells stressed by ionizing radiation, the kinases Chk2 and ATM are activated and phosphorylate p53 at its amino terminus (Meek, 1999; Carr, 2000; Ljungman, 2000; Appella and Anderson, 2001). Since these phosphorylated sites on p53 lie within the region of p53-Mdm2 binding, phosphorylation here stabilizes p53 through prevention of Mdm2-mediated degradation. In cells with aberrant growth signaling due to oncogene overexpression, the p14ARF protein is activated and this protein sequesters Mdm2 and prevents its binding to p53, thus resulting in p53 stabilization (Sherr, 1998; Sherr and Weber, 2000; Sherr, 2001).
Once p53 is stabilized and activated, it binds to specific DNA sequences (a consensus p53 response element) and enhances the transcriptional activity of an array of genes containing such sequences (El-Deiry, 1998; Yu et al., 1999; Zhao et al., 2000). p53 target genes include important regulators of cell cycle inhibition, apoptosis, and genome integrity (El-Deiry, 1998). The induction of cell cycle arrest in G1 or G2 by p53 provides additional time for the cell to repair genomic damage before entering the critical phases of DNA replication and mitosis. This p53-induced cell cycle arrest is usually a transient state and the arrested cell will re-enter the cell cycle presumably when the stress-induced damage is repaired (Kastan et al., 1991). In other situations, perhaps where the stress may produce irrevocable damage or deleterious cell signaling alterations, p53 can induce apoptosis (Yonish-Rouach et al., 1991; Clarke et al., 1993; Lowe et al., 1993; Wu and Levine, 1994). Alternatively, it may mediate a terminal state of cell cycle arrest which is indistinguishable from the senescent state observed in long term passaged cells in culture (Di Leonardo et al., 1994; Linke et al., 1997). In any event, the p53-induced apoptotic or senescent cell cannot propagate and form a nascent tumor, thus protecting the organism from cancers.
CELLULAR REPLICATIVE SENESCENCE
Normal human cells cultured in vitro exhibit a finite number of cell divisions before they enter a non-dividing state referred to as replicative senescence (Hayflick, 1985; Smith and Pereira-Smith, 1996). Replicative senescence has been most extensively characterized in diploid human fibroblasts, though virtually all dividing human cell types exhibit it (Campisi, 1997). For example, human fibroblasts from fetal or newborn sources may undergo 50–80 divisions before they cease division and enter a terminal G1 arrest state insensitive to mitogens (Campisi, 1996; Smith and Pereira-Smith, 1996). Senescent cells can remain viable for years and are resistant to apoptosis (Smith and Pereira-Smith, 1996; Campisi, 1997).
The mechanism for cellular senescence in human cells after prolonged culture is related to telomere shortening. Telomeres, the repetitive DNA sequences that form the ends of chromosomes, are critical for chromosomal integrity and usually extend for 15–20 kb in human germ cells, which retain active telomerase, the enzyme that maintains the length of the telomeres (Stewart and Weinberg, 2000; Blackburn, 2001). Somatic cells lack telomerase, and thus with each division may lose 50–200 bp of telomere sequence (Harley et al., 1990; Stewart and Weinberg, 2000). Senescent human cultures have an average telomere length of about 4–7 kb, but it is possible that a single chromosome with a much shorter telomere length could induce the senescence phenotype (Harley et al., 1994; Hemann et al., 2001). The critical experiment supporting the telomere hypothesis was the demonstration that ectopic expression of telomerase in pre-senescent human cells leads to an immortal phenotype (Bodnar et al., 1998; Vaziri and Benchimol, 1998; Yang et al., 1999; Stewart and Weinberg, 2000). Thus, if telomere lengths are maintained, cells can divide indefinitely and retain an otherwise normal, non-transformed phenotype.
Despite the clear association of telomere length with senescence in human cells, it seems unlikely that telomere-based mechanisms are responsible for all senescence phenotypes. For example, mouse cells derived from inbred strains of mice undergo senescence in culture while retaining very long telomeres of 50 kb or longer (Kipling, 1997). One hypothesis is that senescence in mouse cells is a result of their higher sensitivity to the stresses of in vitro culture conditions (Sherr and DePinho, 2000). When agents that induce DNA damage or increased oxidative stress are added to the culture media, a senescent phenotype is often induced which is independent of telomere length (Toussaint et al., 2000; Chen et al., 2001). When activated oncogenes (e.g., Ras or Raf) are expressed in primary cells, a terminal senescent-like state is often observed (Serrano et al., 1997; Zhu et al., 1998).
Is in vitro replicative senescence relevant to organismal aging? Campisi has argued persuasively for a role of cellular senescence in in vivo aging, though the supporting data is mostly correlative in nature (Campisi, 1997, 2001a). For example, fibroblasts from older donors senesce after fewer divisions than fibroblasts from younger donors, suggesting that mitotic cells in the intact individual may gradually reduce their proliferative capacity with age (Goldstein, 1990; Campisi, 1997). Cells from Werner syndrome patients, who exhibit early aging phenotypes, reach senescence much earlier than cells derived from normal individuals (Martin et al., 1970; Norwood et al., 1979). Finally, alterations in levels of certain senescence-associated molecular markers are similar in both senescent fibroblasts and human skin (Dimri et al., 1995). Thus, there is some evidence that senescent cells exist in vivo and may contribute to some of the phenotypes associated with the aging organism.
Assuming that cellular senescence does occur in the aging organism, does it merely accompany the gradual degenerative processes associated with aging or does it have more significant roles? Sager, Campisi, and others have promoted the idea that cellular senescence is an important mechanism of tumor suppression (Sager, 1991; Smith and Pereira-Smith, 1996; Campisi, 2001b). Cells with a finite replicative span are much less likely to become tumorigenic (Campisi, 1997; Campisi, 2001b). Moreover, abrogation of programmed senescence seems to be a fundamental prerequisite for tumor formation (Hanahan and Weinberg, 2000).
If the hypothesis that senescence is a mechanism of tumor suppression is true, how is the increase in tumor incidence with age explained? As senescent cells accumulate, it might be expected that these would provide suppression of nascent tumors. Campisi (1997) has theorized that replicative senescence may fail to suppress tumors with age due in part to the microenvironmental disruptions produced by senescent cells. It has been shown that senescent fibroblasts secrete high levels of extracellular matrix degrading proteases, certain growth factors, and other molecules that may promote a pro-carcinogenic microenvironment (West et al., 1989; Millis et al., 1992; Campisi, 1997). Recently, Krtolica et al. (2001) have provided supporting evidence for this model by demonstrating that senescent human fibroblasts stimulate proliferation of premalignant and malignant epithelial cells in culture and promote tumor formation in mice, even when senescent cells were only 10% of the total fibroblast population. This stimulation was due to factors secreted by the senescent cells. It was concluded that although cellular senescence may suppress tumorigenesis during the early, reproductively active phase of an organism's life span, cancer may be promoted as the organism reaches the end of its life span.
p53 AND CELLULAR SENESCENCE
It is known that p53 can induce a transient cell cycle arrest (Kastan et al., 1991), but there is also considerable evidence that p53 may be responsible for the irreversible cell cycle arrest characteristic of senescent cells. Wahl and colleagues have shown that gamma irradiation can induce permanent arrest and senescent cell phenotypes in fibroblasts and epithelial cells (Di Leonardo et al., 1994; Linke et al., 1997). Moreover, the induction of senescence in human and rodent fibroblasts by overexpression of Ras is a p53-dependent phenomenon (Serrano et al., 1997). In the absence of p53, such cells become transformed, demonstrating directly that p53-induced senescence is a mechanism of transformation suppression. Ras-induced senescence may operate through upregulation of the tumor suppressor PML, which can also induce a p53-mediated cellular senescence when overexpressed (Ferbeyre et al., 2000; Pearson et al., 2000). A p53-dependent rapid growth arrest and senescence in human fibroblasts could also be effected by introduction of the cell cycle regulatory molecules E2F1 and p14ARF (Dimri et al., 2000). When p53 is introduced into or induced in cancer cells, it often causes apoptosis, but in some cases it has been shown to induce senescence (Sugrue et al., 1997; Chang et al., 1999; Goodwin and DiMaio, 2001). Thus, p53 can induce a senescence-like phenotype in cells in a number of different contexts.
Increased p53 activity correlates well with senescence. Human cells that undergo replicative senescence after prolonged culture show elevated p53 functional activity, though p53 protein levels are generally not increased (Atadja et al., 1995; Bond et al., 1996; Vaziri et al., 1997). It is likely that telomere attrition is associated with the augmented p53 response (Vaziri et al., 1997). Generally, the p53 in senescent cells shows increased DNA binding and transcriptional activation activity and is phosphorylated on some, though not all, of the amino acid residues which are phosphorylated following DNA damage (Webley et al., 2000). A p53 target gene, the p21WAF1/CIP1 cyclin-dependent kinase inhibitor, often shows elevated levels of expression in senescent cells (Noda et al., 1994), and senescence can be bypassed in human fibroblasts missing p21WAF1/CIP1 (Brown et al., 1997). In other instances, a p53-dependent induction of senescence could be generated in the absence of p21 (Pantoja and Serrano, 1999; Groth et al., 2000; Wei et al., 2001).
While senescence is associated with increased p53 activity, cells with reduced p53 can often delay or avoid senescence. Fibroblasts from Li-Fraumeni patients contain a mutant p53 allele and are more susceptible to spontaneous immortalization (Bischoff et al., 1990). Mouse embryo fibroblasts null for p53 do rapidly immortalize, in contrast to their normal counterparts (Harvey et al., 1993). Removal of p53 activity from pre-senescent cells through sequestration by viral oncoproteins extends the lifespan of human cells as many as 20 population doublings (though often in conjunction with Rb inactivation) (Shay et al., 1991; Harley et al., 1994). Introduction of dominant negative versions of mutant p53 also increases the lifespan (Bond et al., 1995). Microinjection of anti-p53 antibodies into senescent human fibroblasts reinitiates DNA synthesis and allows further population doublings (Gire and Wynford-Thomas, 1998).
Despite their extended lifespan, human cells with inactivated p53 terminally arrest in a state called cellular crisis, characterized by massive genomic instability and frequent cell death. In human cells with active p53, the terminal arrest state (M1) occurs when the telomeres are longer, 4–7 kb on average, versus the 2–4 kb length in the crisis state (M2) (Harley et al., 1994). In M1, the highly shortened telomeres may evoke a p53 arrest response because of their similarity to damaged DNA structures known to induce activated p53. p53 may sense these truncated telomeres through ATM, a known upstream kinase that directly phosphorylates p53 (Vaziri et al., 1997). The T-loop end structure formed by the telomere depends on the protein TRF2. Dominant negative TRF2 introduced into tumor cells evokes an ATM and p53-dependent apoptotic response, suggesting a mechanism by which shortened telomeres may be unable to form a stable t-loop or may be unable to bind sufficient TRF2 to prevent an ATM or p53 response (Karlseder et al., 1999; Stansel et al., 2001). However, the telomere would be sufficiently long to prevent the chromosomal instability and telomeric fusions that are prevalent in cells in M2 crisis (DePinho, 2000). Thus, the p53-induced senescent cell would be in a stable long term growth arrest situation and maintain chromosomal integrity, unlike the cell in crisis.
MODELS OF ORGANISMAL SENESCENCE
A number of organismal aging models have been developed to study the genetics and mechanisms relevant to longevity and aging (Jazwinski, 1996; Guarente and Kenyon, 2000). One apparent intervention that can prolong lifespan among many organisms is caloric restriction, where the animals are fed a nutritionally competent diet which is 30–50% reduced in calories from normal food intake (Masoro, 1996; Sohal and Weindruch, 1996; Weindruch, 1996). The mechanisms by which caloric restriction extend life span remain unclear, but a favored hypothesis is that it slows the metabolic rate and reduces production of reactive oxygen species, thus reducing cellular damage and allowing longer survival (Masoro, 1996; Sohal and Weindruch, 1996).
In the last decade, genetic interventions with effects on organismal longevity have received a large amount of attention. The most popular organisms for these types of studies are, not surprisingly, those species that are among the most widely studied and most genetically malleable, including Saccharomyces cerevisiae, Caenorhabditis elegans, Drosophila melanogaster, and inbred strains of Mus musculus (Jazwinski, 1996; Guarente and Kenyon, 2000). The use of such genetically tractable organisms by aging researchers implies a belief that there are single genes that can individually regulate the intrinsic aging process, despite evolutionary theories that argue that aging may represent a generalized deterioration during post-reproductive phases due to a lack of selection against many genes with late deleterious effects (Harrison and Roderick, 1997; Kirkwood and Austad, 2000; Vijg, 2000). Genes that affect longevity in multiple models might be expected to be the best candidates for longevity-associated genes and a number of these have been identified. For example, Sir2, a NAD-dependent histone deacetylase has been shown to enhance longevity in both yeast and worms when overexpressed (Guarente and Kenyon, 2000; Tissenbaum and Guarente, 2001). Removal of Sir2 from these organisms conversely causes a reduction in mean life span. Sir2 appears to function by gene silencing. In yeast, a reduction of silencing by Sir2 with age leads to an increase in the generation of extrachromosomal rDNA circles, which appear to shorten the life span (Guarente, 2000). In C. elegans, Tissenbaum and Guarente (2001) have shown that worms with additional copies of sir-2.1, which shows close similarilty to yeast Sir2, exhibit extended life span in part by altering the insulin-like signaling pathway, which has previously been shown to influence longevity in worms (Guarente and Kenyon, 2000).
Recently, two groups have shown that a mammalian ortholog of Sir2 directly binds and deacetylates human and mouse p53 protein, and attenuates the apoptotic and transcriptional regulatory activities of p53 in response to DNA damage and oxidative stress (Luo et al., 2001; Vaziri et al., 2001). Expression of catalytically inactive Sir2 molecules served to potentiate p53-associated apoptosis and radiosensitivity (Vaziri et al., 2001). Thus, the demonstration that Sir2 directly modulates p53 activity raises the intriguing possibility that it may influence longevity in mammals in part through p53-dependent effects.
MOUSE MODELS OF ORGANISMAL SENESCENCE
Inbred and mixed inbred strains of mice have a number of advantages as tools to model human aging (Sprott, 1997; Andersen, 2001; Anisimov, 2001). The obvious advantage is that they are mammals with a close evolutionary relationship to humans, yet their life span of 2–3 years is short enough to perform aging studies within a reasonable length of time. The development, biology, pathobiology, and aging of various inbred strains of mice have been well characterized. Moreover, the genetics of mice are in a very advanced state, with the draft sequence of the mouse genome available along with almost a century of work detailing a plethora of spontaneous and engineered mouse mutations. The ability to manipulate the mouse germline through transgenic and knockout methods is unsurpassed in mammalian systems. Added to this are major efforts to randomly introduce mutations across the entire mouse genome through random ENU mutagenesis protocols (Justice et al., 1999; Justice, 2000; Balling, 2001). Within the next decade, it is likely that mutations in virtually every mouse gene will be available to the scientific community. Unlike worms and flies, whose somatic cells are virtually all postmitotic as adults, mice have mitotic compartments of self-renewing tissue stem cells. These cells function to maintain organ homeostasis through replacement of senescent or dying cells. As will be discussed later, the adult stem cells of an organism may play an important role in mammalian aging.
Studies on the genetics of aging in mice are being carried out on inbred or even mixed inbred and wild mice, as well as mice with induced mutations (Harrison and Roderick, 1997; Miller et al., 1999; Takeda, 1999; Andersen, 2001; Anisimov, 2001; Klebanov et al., 2001). A recent trend in organismal aging studies has focused on the characterization of spontaneously arising or genetically engineered mutant mice with interesting aging-associated phenotypes (Andersen, 2001; Anisimov, 2001). Generally, a particular mutant line should exhibit enhanced or shortened longevity compared to their wild type counterparts to be considered potentially useful for aging studies. Many in the aging community believe that enhanced longevity is more likely to result from a bona fide alteration in aging or aging regulatory pathways, while shortened longevity must be evaluated with considerable caution (Harrison, 1997; Harrison and Roderick, 1997). This is justified because of the high likelihood that the shortened lifespan could be attributed to some non-aging related pathology (Harrison, 1994, 1997). For this reason, the burden of proof for a mutant model with shortened longevity is greater before it can be considered as a potential aging model. Such a model should show a relatively large constellation of aging-associated phenotypes and few or no pathologies that are not aging-related. While acknowledging the reservations inherent in the “strict” interpretation of mouse aging models, this review will take a “broad” interpretation and include those mice with shortened longevity that have numerous aging-associated phenotypes.
In mice, several single gene mutations have been shown to extend life span (Table 1). The Ames Dwarf mouse has a life span over 50% longer than that of normal mice due to an autosomal recessive mutation in the Prophet of Pit-1 (Prop-1) gene (Sornson et al., 1996; Bartke, 1998, 2000). This mutation results in the failure of the pituitary to produce growth hormone, prolactin, and thyroid stimulating hormone. These mice are infertile and appear resistant to oxidative stress. Their tumor incidence is not apparently different from that of normal mice. The Snell dwarf mouse, with extended longevity and phenotypes very similar to those of the Ames dwarf mouse, has a mutation in the Pit1 (pituitary-specific transcription factor 1) gene, which is downstream of the Prop-1 gene and also results in a growth hormone deficiency (Flurkey et al., 2001). In addition, mice with the little mutation are defective in the response to the hypothalmic peptide growth hormone releasing hormone (GHRH), have very low growth hormone levels, and also show significant extended longevity compared to strain-matched controls (Flurkey et al., 2001). Finally, growth hormone receptor knockout mice have been shown to recapitulate many of the phenotypes of the Ames and Snell dwarf mice, including extended longevity (Coschigano et al., 2000). Clearly, mutations in multiple genes affecting growth hormone production lead to increased longevity in mice, probably involving the growth hormone/IGF-1 axis (Flurkey et al., 2001).
Table 1. Mouse models with altered aging phenotypes
Indicates whether p53 signaling affected by targeted mutation or whether p53 deficiency can rescue aging phenotypes.
Prophet of Pit-1
Growth hormone receptor
The p66Shc−/− mouse is an example of a genetically engineered mutant mouse with an extended longevity phenotype. This mouse has been shown to have a 30% extended life span (Migliaccio et al., 1999). p66Shc is an adaptor protein important in the signaling response to reactive oxygen species (Luzi et al., 2000). Cells lacking this protein have enhanced resistance to apoptosis by UV light or hydrogen peroxide treatment and mice lacking this gene have an increased resistance to paraquat, a toxin that generates superoxide anions (Migliaccio et al., 1999). Interestingly, the p53 apoptotic response in these mice is impaired.
Some mutant strains of mice exhibit shortened longevity accompanied by phenotypes that are consistent with premature or accelerated aging (Table 1). Several lines will be considered as reasonable candidates for accelerated or premature aging models. In this instance premature aging is differentiated from accelerated aging by the observation of aging-like phenotypes prior of full maturity in the former and by observation of aging phenotypes after full adulthood in the latter (Hosokawa et al., 1997).
A good candidate for a premature aging model is the klotho mutant mouse. Initially discovered in mice homozygous for a transgene insertion, klotho mice displayed dramatic early aging-associated phenotypes and had a mean life span of 60.7 days (Kuro-o et al., 1997). Even before reaching adulthood, klotho mice displayed lordokyphosis (hunchbacked spine), thymic atrophy, hair loss, and osteoporosis. The klotho gene is not expressed in all tissues and appears to encode a secreted protein that Takahashi et al. (2000) have speculated might be an “anti-aging” hormone.
One of the earliest and best characterized mouse aging models is the senescence accelerated mouse (SAM-P) developed by Takeda and colleagues by propagation of recombinant inbred lines (Takeda et al., 1981; Takeda et al., 1997; Takeda, 1999). Nine senescence prone and three resistant lines have been characterized. The senescence prone lines have a significantly earlier appearance of a number of age-related disorders, including osteoporosis, lordokyphosis, senile amyloidosis, degenerative joint disease, loss of hair, and memory and learning dysfunction (Takeda et al., 1997; Takeda, 1999). The mean lifespan of the senescence prone lines was 9.7 months versus 18.9 months for the senescence resistant lines. Pathological analyses revealed earlier degenerative changes in the kidneys, brain, liver, bone, and other organs. The genetic loci responsible for these aging-like phenotypes appear to be unknown.
The Ku80 null mice developed by Hasty and colleagues are deficient in non-homologous end joining following DNA double strand breaks (Vogel et al., 1999). The Ku80−/− mice exhibited early age-specific changes that included osteopenia, skin atrophy, hepatocellular degeneration, hepatocellular lesions, and reduced longevity (Vogel et al., 1999). The embryo fibroblasts from these mice showed premature senescence in culture, which was rescued after introduction of p53 nullizygosity, indicating that the senescence phenotypes were p53-dependent (Lim et al., 2000).
One mutant mouse with accelerated aging-related phenotypes is the late generation telomerase deficient mouse (Lee et al., 1998; Rudolph et al., 1999). This mouse is null for the RNA component of telomerase. Interestingly, the first generation of telomerase null (mTR−/−) mice were phenotypically normal and had a normal lifespan (Blasco et al., 1997; Rudolph et al., 1999). This is probably due to the fact that telomeres of inbred mice are very long and would not erode enough in one life span to inhibit any cell or organismal functions. However, continued crossing of the mTR−/− mice revealed that later generation mice (e.g., sixth generation telomerase null mice) did have shortened life spans accompanied by aging-associated phenotypes, including decreased body weights, decreased stress responses, and an increased incidence of skin lesions (Rudolph et al., 1999). All of these phenotypes correlated with a decline in telomere length compared to normal control mice. Importantly, some of these aging phenotypes could be rescued by crossing the null p53 allele into the mTR null background (Chin et al., 1999). Mice deficient in both p53 and mTR (late generation) showed an attenuation of some of the organismal aging effects seen in the mTR−/− late generation mice. Moreover, late generation telomerase-deficient mice showed an induction of active p53 in some tissues, suggesting that p53 plays a role in the cellular response to shortened telomeres and may mediate some of the observed mTR−/− aging-like phenotypes (Chin et al., 1999; Artandi and DePinho, 2000).
Genetic lesions responsible for the early aging-associated phenotypes observed in human progeroid syndromes might be considered ideal candidates for modeling in mice. Werner's syndrome (WS) patients after a normal childhood exhibit a failed growth spurt at puberty, premature greying and loss of hair, skin atrophy, and early testicular atrophy, among other problems (Martin and Oshima, 2000). By the fourth decade of life, cataracts, osteoporosis, diabetes, atherosclerosis, cancers, and other age-associated symptoms arise, with a median age of death at 47–48 years. Two groups have generated targeted deletions in the critical helicase region of the WS gene (WRN) of mice (Lebel and Leder, 1998; Lombard et al., 2000). Surprisingly, the Wrn null mice show no obvious reduction in life span, though one of the mutant lines does show some evidence of early cardiac fibrosis and other pathologies, but classic age-related phenotypes appeared not to be in evidence (Lombard et al., 2000; Lebel et al., 2001). When p53 nullizygosity was introduced into the Wrn null background, the tumor incidence in the bideficient mice was accelerated compared to the singly deficient p53 null mice (Lombard et al., 2000; Lebel et al., 2001). The tumor spectrum of the doubly deficient mice was also more diversified, indicating an interaction between WRN and p53 in some tissues (Lebel et al., 2001). The likelihood of such an interaction is underscored by findings that WRN protein and p53 physically interact and that Werner syndrome cells have abrogated p53 apoptotic function (Blander et al., 1999; Brosh et al., 2001).
p53 AND ORGANISMAL AGING: THE p53+/m MOUSE
The role of p53 in sensing various stress signals and inducing senescence phenotypes in cultured cells is now well established. However, a major question is whether p53-induced effects play a significant role in organismal aging and longevity. What is the effect of p53-induced senescence in individual cells on the organism as a whole? It should be evident that p53 signaling may be affected in some of the mouse models discussed above (Table 1). The hint of p53 association in these aging models suggests that p53 itself may play an important role in regulating organismal aging. To test this hypothesis directly, levels of p53 could be modulated in the mouse. However, it has been shown in previous studies that p53-deficient mice develop early cancers and thus cannot be used for normal aging studies (Donehower et al., 1992; Harvey et al., 1993). Moreover, transgenic mice globally overexpressing wild type p53 appear to result in embryonic lethality (J. Butel and A. Bernstein, personal communications).
Despite these obstacles, my laboratory has recently generated a p53 mutant mouse line that appears to contain modestly overactive p53 (Tyner et al., 2002). This mutant mouse was generated accidentally following an attempt to introduce a p53 point mutation into the mouse germ line. While achieving this goal, the first six exons of the p53 gene were deleted and the last five exons of p53 were juxtaposed to an upstream promoter of unknown origin. This novel mutant p53 allele was termed the “m” allele and the mouse bearing this mutation was designated the “p53+/m” mouse. Most tissues of the p53+/m mice produced a truncated p53 message capable of synthesizing a C-terminal p53 fragment. The p53 response to ionizing radiation in tissues from the p53+/m mice was higher than that of p53+/− mice (also with one intact p53 allele) and appeared to approach or exceed the p53 response of p53+/+ mice in some tissues. Cell culture experiments indicated that the C-terminal p53 fragment encoded by the m allele could interact with wild type p53 and increase its transcriptional activation and growth suppression activities. This is consistent with earlier studies indicating that C-terminal fragments or peptides of p53 could augment wild type p53 activity for DNA binding and transactivation function (Muller-Tiemann et al., 1998; Selivanova et al., 1999). Thus, despite having only one intact wild type p53 allele, the p53+/m mice may possess a p53 response exceeding that of wild type mice.
Support for this interpretation arose from the cancer phenotypes observed in the p53+/m mice. Only 6% of p53+/m mice developed spontaneously arising tumors over the course of their life span. In contrast, virtually all of the p53+/− mice and approximately half of the p53+/+ mice of the same background developed overt tumors. Despite being almost cancer-free, however, the p53+/m mice had a 20% shortened median life span compared to their p53+/+ counterparts. This shortened longevity was accompanied by early appearance of a number of aging-associated phenotypes. These included osteoporosis, muscle atrophy, lymphoid atrophy in spleen and thymus, skin atrophy, lordokyhosis, reduced body weight, reduced cellularity of organs such as liver, kidney, and testis, reduced hair growth, as well as a poor response to stresses such as wound healing, and reduced anesthetic tolerance. The p53+/m animals appeared normal up to the age of 12 months (early middle age for a mouse), but by the age of 18 months many exhibited these early aging-associated phenotypes, while their wild type p53 counterparts remained robust in appearance. By 24 months of age, those differences became even more apparent.
While we believe the data is reasonably compelling that p53 is responsible for the observed cancer resistance and aging-associated phenotypes, two caveats should be mentioned. First, due to the relatively low expression level of the m allele in p53+/m tissues, we were unable to detect m-derived protein in these tissues, although m allele mRNA in vitro can be translated into a truncated C-terminal p53 protein. Second, at least one gene upstream of p53 has been deleted, raising the possibility that haplo insufficiency in genes upstream of p53 could contribute to the observed phenotypes. Therefore, we have attempted to recapitulate at least some of the aging phenotypes observed in other p53 mutant models. One mutant p53 transgenic line, pL53, developed by Alan Bernstein and colleagues (Lavigueur et al., 1989), contains approximately 20 copies of a temperature sensitive mutant allele of p53 that is globally expressed at high levels. At lower temperatures this p53 mutant protein behaves as a wild type molecule, while at high temperature it remains in a mutant non-functional configuration. While these mice have a modestly increased tumor incidence, some pL53 mice also show some of the aging-like phenotypes observed in the p53+/m mice, such as earlier osteoporosis, delayed wound healing, reduced hair growth, and reduced subcutaneous adipose. The skin and hair-associated aging-like phenotypes might be a result of an increased level of the wild type p53 conformation in the compartments of the animal exposed to the lower ambient temperatures.
One conclusion from these observations is that an augmented p53 response may induce an acceleration of the aging process. Of course, another possibility is that the many observed aging-associated phenotypes are merely pathologies which closely mimic the aging process. While the latter possibility cannot be definitively ruled out, we believe that the fairly extensive array of early aging-like phenotypes in the p53+/m mice is consistent with an accelerated aging process in some organs. Moreover, the involvement of p53 in in vitro senescence models and the aging phenotypes observed in other murine models with alterations in genes which affect p53 signaling support a role for p53 in regulating aging and longevity.
Despite the supporting evidence that enhanced p53 is associated with early aging-associated phenotypes, a more definitive proof would be the demonstration that reduction of p53 activity enhances longevity. Unfortunately, as indicated earlier, p53-deficient mice are susceptible to early cancers. It would be interesting to examine the longevity in p53-deficient mice in which tumors could somehow be prevented. A hint that reduced p53 could potentially increase longevity was our observation that two of 217 p53+/− mice in our colony that evaded tumors lived longer than any of the p53+/+ mice of similar background (L.D., unpublished data).
p53, STEM CELLS, AND ORGANISMAL AGING: A MODEL
My laboratory has generated mice that are deficient in p53 (p53−/− and p53+/−) or appear to have increased activity of p53 (p53+/m) (Donehower et al., 1992; Tyner et al., 2002). Consistent with the role of p53 as a tumor suppressor, mice with decreased p53 activity are highly susceptible to cancer while the mice with increased p53 appear highly resistant to cancer. The increased cancer resistance of the p53+/m mice was initially surprising, since one might assume that evolution would select for maximum p53 cancer prevention activities in normal mice. Earlier unsuccessful attempts to generate transgenic mice with extra copies of wild type p53 were supportive of this idea, since it suggested that more than wild type levels of p53 would cause embryonic lethality. Moreover, mice null for mdm2, a protein that degrades p53, show early embryonic lethality, presumably due to excess levels of p53 in the embryo (Jones et al., 1995; Montes de Oca Luna et al., 1995). The p53+/m mice may be viable because p53 activity appears only marginally greater than that of p53 in normal animals (Fig. 2A).
The increased cancer resistance seen in the p53+/m mice suggests that evolutionary forces haven't necessarily selected for maximum cancer fighting capability in normal mice. Instead, evolution may have selected for a balanced level of p53 activity, enough to efficiently prevent cancer early in life, yet perhaps not enough to accelerate the appearance of some deleterious aging-related phenotypes (Ferbeyre and Lowe, 2002) (Fig. 2B). Moreover, since most mice in the wild probably die by 1 year of age by predation, starvation, disease, or freezing (Hayflick, 2000), there may be little selection pressure for p53 activity which prevents cancer beyond this age. This is supported by the fact that we virtually never see cancers before 1 year in the normal C57BL/6 mice that we utilize in our studies (although there are other inbred lines that show early cancers, it is unlikely that mice in the wild live long enough to have a significant cancer incidence). Thus, p53 may represent an excellent example of “antagonistic pleiotropy,” an evolutionary theory of aging formulated by Williams (1957). In this model, pleiotropic genes have very positive early effects which are favored by selection, despite the fact that they may have deleterious effects later in life (where selection is minimal) (Williams, 1957; Kirkwood and Austad, 2000).
If p53 can play a role in regulating aging and longevity in mice, how might it perform these functions? One clue arises from the phenotypes of the p53+/m mice. As the p53+/m mice age, they appear to show an accelerated loss of mass and cellularity in many of their organs, particularly those that have a high cellular turnover rate. Atrophy of muscle, skin, and lymphoid organs, decreased hair growth, loss of bone density, reduced mass and cellularity of liver, kidney and testes all point to an earlier failure of the relevant tissues in maintaining homeostasis or normal cell numbers. Many of these aging-associated phenotypes are also observed in humans, particularly after the age of 60 (Arking, 1998). Moreover, the defects in re-epithelialization during wound healing, reduced white blood cell replenishment after hematopoietic ablation, and reduced hair growth suggest a reduction in stem cell functional capacity in the affected tissues. Adult tissue stem cells are critical for maintaining organ homeostasis by their ability to both self renew and differentiate into progeny cells that ultimately give rise to the mature cells that make up the functioning organ (Fuchs and Segre, 2000; Weissman, 2000). The age-associated reduction in stem cell functional capacity refers to an inability of the stem cells to replenish sufficient numbers of mature cells to a particular organ due to one or a combination of deficiencies. These could include a loss of stem cell number due to increased apoptosis or cell cycle arrest, a maintenance of stem cell number but a reduction in self renewal capacity, or an ability to maintain self renewal capacity but a reduction in the ability to provide more differentiated progeny cells. Loss of stem cell function could result not only from internal signaling defects, but also from insensitivity to external environmental cues to divide or differentiate. Alternatively, a systemic reduction in critical growth factors or hormones or alterations in stem cell environment could reduce the functional capacity of the stem cells.
Figure 3 illustrates one hypothetical model for the early aging phenotypes of the p53+/m mice. During the youth of both the p53+/+ and p53+/m mice, there is reserve functional capacity of the stem cells, so that organ homeostasis and cellularity can easily be maintained. However, with age this functional reserve may decline more rapidly in the p53+/m mice as their stem cells gradually fail to self renew or differentiate in response to environmental cues. The age-associated accumulation of genetic insults in the p53+/m stem cells may provoke augmented arrest or apoptosis responses which gradually result in fewer division-competent stem cells. Such an enhanced p53 response hypothesized for the p53+/m stem cells might have the added effect of preventing tumor formation in these animals. Nevertheless, the stem cells of the p53+/m mice may eventually reach a point in which their functional capacity is so reduced that sufficient numbers of mature cells cannot be provided to maintain organ homeostasis. The above hypothesis is primarily based on an intrinsic role for p53 activity in individual stem cells. Alternatively, a systemic role for p53 influence might result from stem cells in key endocrine organs being more affected by changes in p53 activity. Subsequent alterations in hormone levels secreted by these organs might then affect aging phenotypes. In either case, the resulting organismal phenotypes may include loss of organ mass, function, and tolerance for stress.
The wild type mice show some evidence of these organ atrophies in extreme old age (> 30 months), but these phenotypes are partially obscured by neoplasias and other pathologies. Multiple adult stem cell types in normal mice have shown age-related reductions in functional capacity (Bergman et al., 1996; de Haan and Van Zant, 1999a,b; Chen et al., 2000; Schlessinger and Van Zant, 2001), but presumably at a slower rate than in p53+/m mice. In fact, Van Zant and colleagues have shown that depletion of hematopoietic stem cell (HSC) function with aging varies according to strain (de Haan and Van Zant, 1999b; Geiger et al., 2001; Schlessinger and Van Zant, 2001). Interestingly, they have correlated an early reduction in HSC function with short lived strains and prolonged HSC function with long lived strains. This stem cell-based longevity correlation is not dependent on stem cell numbers, which increase with age, but with the overall “mobilizability” of the stem cells, which vary according to strain and age. This mobilizability is clearly dependent in part on the external environment of the stem cell, since quiescent stem cells in aged organisms can become quite active when transplanted into young animals (de Haan and Van Zant, 1999a,b). This suggests that stem cell functional depletion with age may result from a complex interaction of both intrinsic and extrinsic factors. As stated by Schlessinger and Van Zant (2001), “the underlying mechanisms and genetic factors determining the numbers, turnover rate, and mobilizability of stem cells remain to be elucidated, though they seem to be related to cell cycle checkpoint controls that may be more or less stringent in different strains, but always lead to increasing quiescence.”
The model in Figure 3 predicts that the reduced p53 activity in p53+/− mice might lead to an enhanced stem cell functional capacity. This is consistent with experimental findings that p53-deficient HSCs proliferate more rapidly and survive longer in culture than their p53 wild type counterparts (Palacios et al., 1996; Shounan et al., 1996). p53+/− mice could be expected to have a longer life span than p53+/+ mice if they were not susceptible to early tumors. Our finding of two long-lived p53+/− mice that avoided tumors is preliminary support for this, but more studies need to be done.
In conclusion, it is possible that p53, in addition to its role in protecting the organism against cancer, may affect components of aging. This idea is in keeping with previous aging hypotheses that cellular senescence is one mechanism of tumor suppression (Campisi, 2001b). At the very least, p53 is critical in allowing long-lived organisms their longevity by preventing cancers during the early reproductive years of life. Nevertheless, there is a price to be paid for this cancer-free early existence and that may be the gradual reduction in functionality of the stem cell compartment. p53 may in part be responsible for this aging-related reduction in stem cell function, leading to some aging-related phenotypes and influencing organismal longevity. Further experiments will be necessary to corroborate these hypotheses, using the current p53 mutant models and other new models.
I thank Gary Van Zant, Sundaresan Venkatachalam, Melissa Dumble, and Chao-Ling Wu for a critical reading of the manuscript. This work is supported by a grant from the National Cancer Institute and an Academic Award from the US Army Breast Cancer Research Program.