View larger version (15K): [in this window] [in a new window] [Download PPT slide]   Figure 1. 12-yr survival and longevity of the general population by p53 Arg72Pro genotype. Death endpoints were collected from the Danish Civil Registration System, which is 100% complete. (A) For 12-yr survival, follow-up started at blood sampling and ended at death, emigration, or March 11, 2004, whichever came first. (B) For longevity using left-truncated age, follow-up started at blood sampling and ended at death, emigration, or March 11, 2004, whichever came first. (C) Comparison of the effect of p53 Arg72Pro genotype with that of smoking status on 12-yr survival and median survival in the same population. Arg/Arg homozygotes are in blue, Arg/Pro heterozygotes are in red, and Pro/Pro homozygotes are in green.

In accordance with our data, in the Leiden 85-plus study, Pro/Pro versus Arg/Arg homozygotes were associated with a 1.41-fold (95% CI: 1.03–1.92) increased survival among 1,226 participants 85 yr or older followed for 5–10 yr (10). A 1.41-fold–increased survival is equivalent to a 0.71-fold risk of mortality, where we observed a 0.82-fold risk of mortality in Pro/Pro versus Arg/Arg homozygotes (Table II). In contrast to our study, the former study (10) did not observe decreased mortality in Arg/Pro heterozygotes versus Arg/Arg homozygotes; however, that study had less statistical power than the present study.

Genetic variations in modifiers of p53 like ATM and MDM2 could have an impact on the effect of p53 Arg72Pro on mortality. However, the Ser49Cys and Ser707Pro polymorphisms of ATM and the –309 T/G polymorphism of the MDM2 promoter did not interact with the effect of the Arg72Pro polymorphism on mortality (Fig. S4, available at http://www.jem.org/cgi/content/full/jem.20062476/DC1).

Survival after a diagnosis of cancer
We next examined whether the increased longevity was explained by a better prognosis in Arg/Pro heterozygotes and Pro/Pro versus Arg/Arg homozygotes after the diagnosis of cancer, cardiovascular disease, or other life-threatening diseases. The cumulative 5-yr mortality after a cancer diagnosis was reduced in Arg/Pro heterozygotes and Pro/Pro versus Arg/Arg homozygotes by 9% (P = 0.003) and 13% (P = 0.03; Fig. 2). The reduced 5-yr mortality corresponds to a decreased gender- and age-adjusted hazard ratio for death after a diagnosis of cancer of 0.87 (95% CI: 0.77–0.98) in Arg/Pro heterozygotes and 0.74 (0.56–0.98) in Pro/Pro versus Arg/Arg homozygotes (Table II). The equivalent hazard ratios for death were 0.95 (0.85–1.06) and 0.85 (0.69–1.05) after a diagnosis of cardiovascular disease and 0.88 (0.81–0.97) and 0.81 (0.68–0.98) after a diagnosis of another life-threatening disease, respectively. Thus, the increased longevity associated with the Arg72Pro polymorphism may be due to increased survival after a diagnosis of cancer or other life-threatening diseases.

View larger version (16K): [in this window] [in a new window] [Download PPT slide]   Figure 2. 5-yr survival of the general population after a cancer diagnosis by p53 Arg72Pro genotype. Cancer diagnosis and death endpoints were collected from the Danish Cancer Registry and the Danish Civil Registration System, which are 98 and 100% complete. Only cancers diagnosed 1 yr before blood sampling and afterward were included. (A) For 5-yr survival after a cancer diagnosis, follow-up started at cancer diagnosis and ended at death, emigration, or March 11, 2004, whichever came first. (B) Effect of p53 Arg72Pro genotype on 5-yr survival after a cancer diagnosis. Arg/Arg homozygotes are in blue, Arg/Pro heterozygotes are in red, and Pro/Pro homozygotes are in green.

Genetic variations in modifiers of p53 (ATM and MDM2) did not interact with the effect of the Arg72Pro polymorphism on mortality after a diagnosis of cancer (Fig. S4). Inflammatory conditions and leukocyte activation release nitric oxide, a known activator of p53. Therefore, diseases with inflammatory elements could be particularly affected by the influence of the p53 Arg72Pro polymorphism. Interestingly, mortality after a diagnosis of respiratory diseases with a known inflammatory component was reduced in both Arg/Pro heterozygotes and Pro/Pro vs. Arg/Arg homozygotes (Table S1, available at http://www.jem.org/cgi/content/full/jem.20062476/DC1).

Different cancers have a different percentage of somatic p53 mutations. Therefore, the impact of the Arg72Pro polymorphism upon mortality after a cancer diagnosis could depend on cancer subtype with a different percentage of somatic p53 mutations. However, when we regrouped all types of cancers in three groups according to frequency of somatic p53 mutations as reported for each cancer type by the International Agency for Research on Cancer (low, middle, and high; references 11–14), the impact of the germline Arg72Pro polymorphism upon mortality was independent of frequency of somatic p53 mutations (Table S2, available at http://jem.org/cgi/content/full/jem.20062476/DC1; P = 0.98 on robust interaction test in Cox regression). Accordingly, for patients with tumors with high levels of somatic p53 mutations, where all the impact in the tumor of the germline Arg72Pro polymorphism might be expected to be eliminated, we speculate that the reduced mortality in Pro/Pro versus Arg/Arg homozygotes could be a characteristic of the entire person harboring the tumor rather than of the tumor itself.

Risk of cancer
Thereafter, we examined whether the increased longevity by TP53 Arg72Pro genotype was explained by decreased risk of developing cancer, cardiovascular disease, or other life-threatening disease (Table II). The gender- and age-adjusted hazard ratios in Arg/Pro heterozygotes and Pro/Pro versus Arg/Arg homozygotes were 1.10 (1.01–1.21) and 0.91 (0.76–1.10) for risk of cancer, 0.96 (0.89–1.04) and 0.99 (0.86–1.15) for risk of cardiovascular disease, and 0.99 (0.95–1.04) and 1.03 (0.94–1.12) for risk of other life-threatening diseases, respectively. Furthermore, there was no evidence for decreased risk of any cancer subtype in Arg/Pro heterozygotes and Pro/Pro versus Arg/Arg homozygotes (Table III). Thus, in accordance with a recent study (15), the Arg72Pro substitution did not associate with a decreased risk of cancer.

Surprisingly, the risk of hematologic cancer increased twofold in Arg/Pro heterozygotes versus Arg/Arg homozygotes (P < 0.001), with a similar trend in Pro/Pro homozygotes (Table III). This twofold-increased risk was significant even after correction for multiple testing and therefore might represent a real rather than a chance finding; however, this needs to be confirmed in another independent study.

The p53 pathway has elements of sexual dimorphism: female Li-Fraumeni patients with p53 mutations develop tumors earlier and with a higher frequency when adjusted for age (16), and the promoter of MDM2, a p53 interaction partner, contains a functional estrogen receptor signal in the DNA (17). Therefore, the effect of the p53 Arg72Pro polymorphism on risk of cancer in women could depend on menopausal status. However, we did not detect any statistically significant interaction between the Arg72Pro polymorphism and gender or menopausal status on risk of cancer (Fig. S5, available at http://www.jem.org/cgi/content/full/jem.20062476/DC1).

Potential limitations
For all the analyses described, multifactorial adjustment for other potential confounders, as shown in Table I, did not change the estimates (unpublished data). Furthermore, because of Mendelian randomization, our results are unlikely to be confounded by yet other environmental or lifestyle factors (18). Finally, because the p53 Arg72Pro polymorphism is well known to be functional at the cellular level, influencing apoptotic potential as well as cellular arrest in G1 of the cell cycle (5–7), the effects observed in the present study are most likely directly due to the Arg72Pro polymorphism rather than to another polymorphism nearby in linkage disequilibrium with this polymorphism.

Although performed in a large, well-characterized cohort of the general population with long-term follow-up (Fig. S1, available at http://www.jem.org/cgi/content/full/jem.20062476/DC1), our study has limitations. Because participants were only genotyped if they participated in the 1991–1994 examination of the Copenhagen City Heart Study, a selection bias might have occurred if death or morbidity prevented certain individuals from participating in this examination. However, we found Hardy-Weinberg equilibrium in the genotype distribution, making selection bias against any of the genotypes unlikely. Should selection bias have occurred in this study, it would tend to result in a conservative estimate concerning mortality, and therefore cannot explain our result. Misclassification of disease status, but not mortality, could also have occurred. However, such misclassification most likely is minimal because we have 100% follow-up of the participants and because systematic registration of diseases causing hospitalization as well as death are registered in the entire country.

Mechanism
p53 can act through several pathways when reacting to cellular stress. Apoptosis, cell cycle arrest at the G1 checkpoint, and cellular senescence are all mechanisms triggered by activated p53 (19). These mechanisms are all beneficial when the organism is young, but in older organisms, such effects probably reduce longevity and augment cancer risk (20), so-called antagonistic pleiotropy (19, 21). In accordance with this concept, p53 accelerates aging when responding to cellular stress (22). The Arg allele of the Arg72Pro polymorphism increases p53-induced apoptosis, whereas the Pro allele effectuates cell cycle arrest in the G1 phase (5–7). Our findings suggest that Arg/Pro heterozygotes and Pro/Pro versus Arg/Arg homozygotes have reduced mortality, which could result from a generally decreased aging process caused by decreased proapoptotic activity and increased cell cycle–arresting abilities of p53.

Conclusions
Although our results suggest that p53 Arg72Pro is not associated with risk of cancer or any other disease, this polymorphism has a profound gene dose–dependent beneficial effect on 5-yr survival after a diagnosis of cancer, on survival after other life-threatening diseases, and on longevity. Thus, the increased longevity may be due to a general increased robustness after a diagnosis of any life-threatening disease. We speculate that the decreased proapoptotic and increased cell cycle–arresting abilities of the Pro versus Arg allele (5–7) might be beneficial for a person experiencing any critical illness. This suggests that the TP53 Arg72Pro polymorphism is an important gain-of-function genetic variant affecting the entire person.

	MATERIALS AND METHODS

Top ABSTRACT RESULTS AND DISCUSSION MATERIALS AND METHODS REFERENCES

 
Using a prospective study of the Danish general population, we examined 9,219 participants ages 20–95 yr from the Copenhagen City Heart Study participating in the 1991–1994 examination (Fig. S1; references 23–25). Height and weight were measured, and participants were questioned about smoking habits, alcohol consumption, and physical activity. Blood was drawn for cholesterol measurement and DNA extraction. All participants gave written informed consent. Herlev University Hospital and the Danish ethical committee of Copenhagen and Frederiksberg approved the study (No. 100.2039/91), which was conducted according to the Helsinki Declaration.

Endpoints.
We collected information on morbidity and mortality from three different population registries: information on mortality (23) was obtained from the Danish Civil Registration System; information on morbidity was obtained from the Danish National Hospital Discharge Register (26) and subdivided according to the Global Burden of Disease classification (27); and information on diagnoses of invasive cancer was obtained from the Danish Cancer Registry (28).

Cancer diagnoses were classified according to criteria from the International Classification of Diseases, Seventh Edition (29), and divided into seven different subgroups: (a) gastrointestinal cancer, including oral, esophagus, stomach, small intestine, liver and biliary tract, pancreatic, colon, rectum, and anal cancer; (b) hematologic cancer, including non-Hodgkin's lymphoma, Hodgkin's disease, multiple myeloma, and leukemia; (c) respiratory cancer, including larynx and lung cancer; (d) urologic cancer, including kidney and bladder cancer; (e) female cancer, including breast, cervix uteri, corpus uteri, ovarian, and vaginal cancer; (f) male cancer, including testis and prostate cancer; and (g) other cancers, including melanoma, nonmelanoma skin cancer, sarcoma, brain, and other central nervous system cancers, thyroid and other endocrine cancers, metastasis with unknown primary tumors, and other tumors.

Genotyping.
We genotyped 9,219 participants for p53 Arg72Pro, an amino acid–changing polymorphism caused by a G215C substitution. By PCR, a 331-basepair fragment, corresponding to exon 4, was amplified from genomic DNA using an intronic forward and exonic reverse primer (forward, 5'-CATCTACAGTCCCCCTTGC-3'; reverse, 5-GCTTCCATGAGACTTCAATGCC-3'). After thermo cycling, genotypes were determined using the Nanogen system (30). ATM and MDM2 polymorphisms were genotyped using endpoint PCR (Taqman; Applied Biosystems). Probes and primers were designed by using Primer Express software (Applied Biosystems) and are available from the authors upon request.

Statistical analysis.
The statistical software STATA (version 8.2) was used. Two-tailed P < 0.05 was considered significant. We used Mann-Whitney U test and Pearson's ² test. Kaplan-Meier curves plotted overall survival and survival after a cancer diagnosis as a function of follow-up time as well as left-truncated age. Hazard ratios for morbidity and mortality were calculated using Cox proportional hazards regression analysis. We tested for proportionality of hazards over time to ensure that this assumption of the Cox proportional hazards regression model was fulfilled based on Schonefeld residuals. Individuals with endpoints developed before study entry were excluded from analysis.

Multifactorial adjustments for cardiovascular and other disease included gender, age, tobacco consumption, smoking habits, systolic blood pressure, alcohol consumption, total cholesterol, body mass index, and physical activity at the time of blood sampling. Multifactorial adjustment for risk of cancer included the aforementioned variables excluding total cholesterol and systolic blood pressure.

Online supplemental material.
Fig. S1 shows the design of the study. Fig. S2 shows p53 Arg72Pro genotype frequency as a function of age in the general population. Fig. S3 shows the frequency of the p53 Arg72Pro Pro allele and frequency of deaths per 5-yr age group as functions of age. Fig. S4 shows the overall hazard ratio of death and the hazard ratio of death after a cancer diagnosis by Arg72Pro genotype, stratified by two ATM genotypes and one MDM2 genotype. Fig. S5 shows risk of cancer and cancer subgroups by Arg72Pro genotype, stratified by gender and menopausal status. Table S1 shows mortality after other diseases and morbidity of other diseases according to p53 Arg72Pro genotype in the general population. Table S2 shows mortality after a diagnosis of cancer according to low, middle, and high p53 somatic mutation frequencies and p53 Arg72Pro germline genotype in the general population. Online supplemental material is available at http://www.jem.org/cgi/content/full/jem.20062476/DC1.