Perspectives

Anecdotal, Historical and Critical Commentaries on Genetics Edited by Adam S. Wilkins

The Szilard Hypothesis on the Nature of Aging Revisited

Henrik Zetterberg,*,1 Magnus Ba˚th,†,‡ Madeleine Zetterberg,§,** Peter Bernhardt†,‡ and Ola Hammarsten†† *Institute of Neuroscience and Physiology, Department of Psychiatry and Neurochemistry, The Sahlgrenska Academy at University of Gothenburg, S-431 80 Mo¨lndal, Sweden, †Institute of Clinical Sciences, Department of Radiation Physics, The Sahlgrenska Academy at University of Gothenburg, S-413 45 Gothenburg, Sweden, ‡Department of Medical Physics and Biomedical Engineering, The Sahlgrenska University Hospital, S-413 45 Gothenburg, Sweden, §Institute of Biomedicine, Department of Medical Chemistry and Cell Biology, The Sahlgrenska Academy at University of Gothenburg, S-405 30 Gothenburg, Sweden, **Institute of Neuroscience and Physiology, Department of Clinical Neuroscience and Rehabilitation, The Sahlgrenska Academy at University of Gothenburg, S-431 80 Mo¨lndal, Sweden and ††Institute of Biomedicine, Department of Clinical Chemistry and Transfusion Medicine, The Sahlgrenska Academy at University of Gothenburg, S-413 45 Gothenburg, Sweden

ABSTRACT This year marks the 50th anniversary of a nearly forgotten hypothesis on aging by Leo Szilard, best known for his pioneering work in nuclear physics, his participation in the Manhattan Project during World War II, his opposition to the nuclear arms race in the postwar era, and his pioneering ideas in biology. Given a specific set of assumptions, Szilard hypothesized that the major reason for the phenomenon of aging was aging hits, e.g., by ionizing radiation, to the gene-bearing chromosomes and presented a mathematical target-hit model enabling the calculation of the average and maximum life span of a species, as well as the influence of increased exposure to DNA-damaging factors on life expectancy. While many new findings have cast doubt on the specific features of the model, this was the first serious effort to posit accumulated genetic damage as a cause of senescence. Here, we review Szilard’s assumptions in the light of current knowledge on aging and reassess his mathematical model in an attempt to reach a conclusion on the relevance of Szilard’s aging hypothesis today.

EO Szilard (born in Budapest, Hungary, 1898; died control of atomic energy out of the hands of the L in La Jolla, California, 1964) was one of the military and within a civilian department at a global leading contributors to the development of nuclear level. He joined the forces of a large number of fellow energy and the first atomic bomb. He understood the scientists and formed the Federation of Atomic Scien- concept of a nuclear chain reaction in 1933 and filed a tists, later Federation of American Scientists, which is patent for his idea the following year (Lanouette still active (http://www.fas.org/main/home.jsp). He also 1992). In 1942, with Enrico Fermi, he set up the first made efforts to encourage mutual disarmament and nuclear chain reaction. Realizing the power of what he the reduction of tensions between the United States had helped unleash, he became among the earliest and and the Soviet Union. To this end, he was active in the most active campaigners for nuclear arms control. In formation of the Pugwash Conferences on Science and spite of intense lobbying, he failed to prevent the World Affairs, together with Albert Einstein, Bertrand bombing of Hiroshima, which was a catastrophic event Russell, and others, a series of conferences concerned in his life that eventually made him leave nuclear phy- with reducing the danger of armed conflict and seeking sics (Lanouette 1992). In the postwar years, Szilard cooperative solutions for global problems (http://www. spent a major portion of his time working to place the pugwash.org/). In 1962, he helped found the Council for a Livable World, a Washington, D.C.-based lobby 1Corresponding author: Institute of Neuroscience and Physiology, De- organization for nuclear arms control, still active with the partment of Psychiatry and Neurochemistry, The Sahlgrenska Academy at University of Gothenburg, S-431 80 Mo¨lndal, Sweden. goal of providing the U.S. Congress with accurate E-mail: [email protected] technical and scientific information to facilitate bal-

Genetics 182: 3–9 (May 2009) 4 H. Zetterberg et al. anced decisions about weapons of mass destruction functioning when either both chromosomes in a chro- (http://www.clw.org/). mosome pair have suffered aging hits or one of the Leo Szilard lived a nomadic life and usually had no chromosomes carries an inherited fault and the other permanent job or address (Maas and Crow 2004). one suffers an aging hit (the possibility that both chro- However, he had an enormous capacity to grasp novel mosomes in a pair carry inherited faults is neglected). For fields and erect hypotheses for others to test experimen- the female, using a single-hit, multitarget (n ¼ 2) model tally and was often acknowledged in scientific communi- for chromosome pairs without inherited faults and a cations from research groups around the world (Klein single-hit, single-target model for chromosome pairs with 1990; Lanouette 1992). After World War II, he left an inherited fault, Szilard describes the surviving fraction, nuclear physics for good and turned himself to molecular f, of somatic cells as biology, a field that had developed rapidly in the United ÂÃ j 2 mr rj States under the lead of Salvador Luria, Max Delbru¨ck, f ¼ 1 ð1 e Þ e ; ð1Þ and others (Klein 1990). In 1946, he accepted an where m is the number of chromosome pairs, r is the appointment as professor of biophysics at the University number of inherited faults, and j is the average number of Chicago. One of his first accomplishments was the of aging hits that have been suffered by a somatic cell at development of the Chemostat, an instrument that a given age: allows for the maintenance of bacterial populations in growth phase over an indefinite period of time by age j ¼ : ð2Þ regulating dissolved oxygen, nutrient concentrations, 2mt and pH (Novick and Szilard 1950). Experiments in the Chemostat revealed, among other things, that the t induction of the lac operon was an all-or-none phenom- In Equation 2, is the ‘‘basictime interval of the aging enon (Novick and Weiner 1957), a result fundamental process,’’ i.e., the average time interval between two to the interpretation of the data by Monod and subsequent aging hits in a somatic cell. It is assumed that t colleagues, defining the concept of negative control of is a characteristic for the species and does not vary gene expression by repressors (Pardee et al. 1958). In appreciably from individual to individual. hisNobellectureof1965(http://nobelprize.org/nobel_ Szilard continues to assume that an individual will die prizes/medicine/laureates/1965/monod-lecture.pdf), within the year in which the surviving fraction has Monod acknowledged a significant part of his Nobel reached a certain critical value, f *. The maximum life Prize to discussions with Szilard, himself never a Nobel span of a species is therefore given by the age at which an laureate. Later in life, Szilard became interested in the individual without inherited faults reaches a surviving biology of aging (Szilard 1959) and higher brain fraction of f *. To include a variation in the model, a functions (Szilard 1964). genetic scattering (stemming from an individual varia- Fifty years ago, Szilard published an article describing tion in the number of inherited faults) and a non- a mathematical target-hit model for the aging process genetic scattering (estimated from the mean age (Szilard 1959). Szilard’s model enables the calculation difference at death between identical twins) are in- of the average and maximum life span of a species, as cluded. Szilard begins by using a Poisson distribution to well as the influence of increased exposure to DNA- describe the fluctuation in the number of inherited damaging factors on the aging process. Here, we re- faults, but to simplify the mathematics he motivates the assess this largely forgotten article, scrutinize the basic use of a Gaussian distribution. Using also a Gaussian assumptions and calculations being made, describe how distribution to describe the nongenetic scattering, the model influenced the thinking about accumulated comparing the result with the actually observed distri- genetic damage as a cause of aging, and update the bution of the ages of death of the U.S. population in the mid-20th century, and assuming that the critical surviv- model on some of its most outdated assumptions. 1 ing fraction of somatic cells (f *) is 6 , Szilard ends up with an average value of r ¼ 2.5 and a value of t ¼ 6 years for humankind. Thus, using m ¼ 23, all parameters in SZILARD’S MODEL Equation 1 are known. Szilard then continues to use this Szilard’s aging model is based on a number of result to discuss various subjects such as the ‘‘physiolog- assumptions, enabling a mathematical description of ical age,’’ the effect of changing the load of inherited the aging process. The elementary step in the process is faults, and the life-shortening effect of ionizing the ‘‘aging hit’’of a chromosome, which renders all genes radiation. carried by that chromosome inactive. These aging hits are random events that have a constant probability of SZILARD’S ASSUMPTIONS occurring per unit time throughout life for a chromosome. A chromosome may also carry an inherited fault, Szilard used the best estimates and assumptions making all its genes inactive from birth. A cell will cease from knowledge accumulated until 1958 to build his Perspectives 5 model of aging. Some of these assumptions are of a causal relation between DNA damage and aging are surprisingly close to present views, while others suffer the role of DNA as the primary informational bio- fromthefactthatlittleofthebiologyoftheDNAand molecule (Vijg 2000) and the fact that all other macro- protein machinery of the cell was known at that time. molecules that are subject to age-related damage In the following paragraphs, we review Szilard’s as- (mostly proteins and lipids) are renewable, whereas sumptions and describe some of the criticism his any acquired error in DNA may have irreversible model elicited. consequences (Garinis et al. 2008). Some of the first Assumption 1—aging is the accumulation of dam- experimental evidence for a causal relationship be- aged chromosomes or genes: The focus of Szilard’s tween somatic mutations and aging came from studies aging model on damage to chromosomes or genes as a on mouse liver cells showing an inverse correlation of major cause of aging was well in line with experimental induced (Curtis and Crowley 1963) and spontaneous data showing that late irradiation damage in animals, (Crowley and Curtis 1963) chromosome aberrations e.g., soft tissue atrophy, resembled premature aging with life expectancy. Results corroborating this view (Henshaw et al. 1947). Since radiation was known to have been seen more recently in experiments on cells induce mutations, it was proposed that aging could be from humans (Morley et al. 1982; Turner et al. 1985; the result of life-long, low-grade exposure to back- Ramsey et al. 1995; Tucker et al. 1999) and also in ground radiation and other agents that could damage computed models of aging (Kirkwood and Proctor DNA. As reviewed by Vijg, this view became the founda- 2003). As originally suggested by Burnet (1973), the tion of a number of somatic mutation theories of aging, somatic mutation theory of aging implies that impaired the first of which was Szilard’s model (Vijg 2000). fidelity of the DNA replication and repair system would Szilard’s article immediately elicited criticism. In a result in premature aging, a hypothesis supported by a letter to Nature, John Maynard Smith showed incon- wealth of recent data showing associations of genetic sistencies between predictions based on Szilard’s model defects in these systems with accelerated aging in both and observations on the life span of inbred compared mice and men (Garinis et al. 2008). with genetically variable wild populations of Drosophila Assumption 2—a certain species is characterized by a (Maynard Smith 1959). He later backed away from certain aging hit rate: The maximum life span is very part of that criticism when considering that, in contrast species specific at the group level (http://genomics. to mammals, there is almost no cell division in adult senescence.info/species/). Thus, it would be reason- insects, making them less suitable to study the role of able to hypothesize either that different genes deter- somatic mutations in aging (Holliday 2000). The main mine the aging process in different species or that criticism, however, was directed against the very unlikely members of a certain species experience different assumption that one aging hit would destroy the amounts of aging hits to the DNA. In general, the function of an entire chromosome (Maynard Smith accumulated data have revealed that similar genes affect 1959). An alternative to the inclusion of chromosome the aging process in different species but that they act in pairs (m) would be to incorporate information on the networks that get more complex and redundant when number of essential genes in a revised model; see moving up the evolutionary ladder (Kuningas et al. revision of the model and simulations below and 2008). It is well established that the amount of DNA calculations made in earlier studies (Holliday and damage per unit time correlates negatively with life span Kirkwood 1981; Kirkwood and Proctor 2003). across species (Adelman et al. 1988). In addition, the Szilard actually discusses this possibility but does not DNA repair capacity correlates with life expectancy in follow up on the issue in his calculations. He states that it different mammals (Hart and Setlow 1974; Burkle can be estimated that humankind has 15,000 genes, et al. 2005). Thus, Szilard may have been right in his 3000 of which are important for the functioning of the assumption that the rate of manifest aging hits may somatic cells of the adults. Now, this is a very accurate determine the maximum life span in different species. prediction for being made in 1958, 5 years after the Another related possibility, however, is that the life span unveiling of the structural basis for the genetic code might differ because the number of DNA-damaging hits (Watson and Crick 1953) and long before the needed to induce permanent cell-cycle arrest differs sequencing of any genome. The best guess today is between different species. 20,000–25,000 genes, while we still have no clear idea Assumption 3—aging hits are random events and do how many active genes are needed for a somatic cell to not vary interindividually: While the maximum lifespan be functional. Attempts to calculate the number of of an animal species is largely determined by genetic essential genes, defined as genes absolutely required for factors, the process of physiological aging varies widely survival, in mice reach numbers in the range of 5000– between individuals (Herndon et al. 2002; Passos et al. 10,000 (Hentges et al. 2007), not far from Szilard’s 2007). This dichotomy points to the key importance of estimate. stochastic factors in the aging process (Vijg 2000). On Is it really correct to emphasize DNA damage as the the basis of theoretical and experimental grounds, the key causal event in aging? Major arguments in support random character of many aging phenotypes has often 6 H. Zetterberg et al. been ascribed to random and destructive chemical mod- ifications of essential biomolecules, such as DNA, by reactive oxygen species (ROS) (Hekimi and Guarente 2003). Other possible shared effectors of aging include mitochondrial dysfunction and impaired protein ho- meostasis that leads to accumulation of aberrant proteins through an imbalance in protein biosynthesis, folding, translocation, assembly/disassembly, and/or clearance (Kirkwood and Kowald 1997; Morimoto 2008). However, biological aging is more than simply the occurrence of random changes in molecules. As discussed above, it is also balanced by the many repair systems found within cells, e.g., the elaborate DNA repair machinery and the ROS-detoxifying systems of Figure 1.—The number of observed female deaths in the all cells and the liver in particular. These biological Swedish population during the year 2007 as a function of age processes act in parallel, creating a network of inter- normalized to a total of 100,000 deaths (blue curve), the re- actions (Kirkwood and Kowald 1997). This opens the sult of simulations of Szilard’s original model based on chromosomes (f * ¼ 1 , t ¼ 6.35, n ¼ 1.4, green curve), and the way for interindividual variability, e.g., through genetic 6 result of simulations of the revised model based on genes polymorphisms, both in how the body prevents aging 1 (f * ¼ 6 , t ¼ 0.34, relative standard deviation of f * and t ¼ hits and in how it deals with damaged macromolecules. 0.083, red curve). Assumption 4—the surviving fraction of functional somatic cells determines the biological age of an gene as the target of the aging hit. Furthermore, instead individual: Szilard’s hypothesis postulates that when of including a nongenetic scattering based on the the surviving fraction of the somatic cells of an in- empirical mean age difference at death between female dividual approaches a critical value, the probability that identical twins, we have in our revised model added a that individual may die within 1 year becomes very high. fluctuation to the average time interval between two Szilard assumes that young mammalian organisms have subsequent aging hits in a somatic cell (the ‘‘basic time a large functional reserve and that the surviving fraction interval of the aging process,’’ t) as well as to the critical of somatic cells may fall substantially before the critical surviving fraction, f *. value of 10–30% is reached. The average number of The introduction of the essential genes instead of the functional cells needed for the body to survive is not chromosomes as the target for an aging hit does not known. However, for specific organs it is known that the alter Equation 1, although m and r now correspond to majority of cells can die before clinical signs of organ the number of essential gene pairs and the number of failure appear. For example, .50% of the contractile inherited faults on the gene level, respectively. The capacity of the heart may be lost before clinical signs of number of essential gene pairs and the average number heart failure appear (Aurigemma et al. 2001), .50% of of inherited faults were estimated to be 10,000 and 20, hippocampal neurons may be lost without any clinical respectively (McConkey 2003). Using Szilard’s original signs of Alzheimer’s disease (West et al. 2004), mice assumption of the critical value of the survival fraction, 1 survive an 80% reduction in renal mass without signs of f *, as 6 , we performed computer simulations to de- acute renal failure (Al-Awqati and Preisig 1999), and termine the number of deaths as a function of age for a by the time overt type I diabetes appears, almost all insulin- given population. In this way not being limited by producing cells have disappeared (Foster 1998). While mathematics, we used a Poisson distribution to describe it is clear that most organs work also after a significant the variation in the number of inherited faults (the loss of functional cells, Szilard’s model with an average genetic scattering), as originally suggested by Szilard. value of the critical surviving fraction for the whole The nongenetic scattering, which Szilard obtained from organism is a great oversimplification. Another criticism the mean age difference at death between female is the failure to recognize that a frequent consequence identical twins, was as described above included by of mutations is precancerous cell proliferation rather adding a fluctuation to the average time interval than cell senescence and death. between aging hits (t) and to the critical value of the survival fraction (f *). In Figure 1, the result of computer simulations of the REVISION OF THE MODEL AND SIMULATIONS revised model is compared with Swedish mortality data On the basis of current knowledge that it is inaccu- for females during the year 2007 (47,175 deaths), rate to expect that a single aging hit could inactivate a normalized to 100,000 deaths. The best fit-by-eye of whole chromosome and following calculations made by the revised model to the observed mortality data was Holliday and Kirkwood (1981), we have revised Szilard’s obtained using a value of the time between two aging model by replacing the chromosome with the essential hits (t) of 0.34 year and a relative standard deviation of Perspectives 7 t and f * of 8.3%. Since t now describes the average time ,70 generations (Hayflick 1965). It is now established interval between two subsequent aging hits in a somatic that this limit is caused by at least two interlinked cell at the gene level, it cannot be compared with the mechanisms: (i) the erosion of telomeres, which are value at the chromosome level, which was 6.0 years in regions at the ends of chromosomes that stabilize DNA, Szilard’s original model. Although there is a similarity and (ii) stress-induced expression of the p16 tumor between the model and the observed data, there are suppressor (Campisi 2005a). discrepancies. The distribution of the observed mortal- Telomerase adds specific DNA sequence repeats ity data is heavily skewed, with a large number of ob- (‘‘TTAGGG’’ in all vertebrates) to the 39 end of DNA served deaths at lower ages, whereas the revised model strands in the telomere regions at the ends of eukaryotic predicts a more symmetric distribution, thereby under- chromosomes. Without telomerase activity, the human estimating the number of deaths at lower ages and chromosomes become shorter during each replication. overestimating the number of deaths at higher ages. Eventually, the shortened telomeres loose their ability to The underestimation is larger than the overestimation, hide the chromosome ends from the DNA double- resulting also in a broader peak for the model than in strand break recognition system, resulting in a sustained the observed mortality data. DNA damage signal that stops the cell from continuous We also included simulations of Szilard’s original growth (Palm and De Lange 2008). In contrast to model based on chromosomes as the targets in the somatic cells, germ-line cells, some cells of the immune model. Since the current mortality data are different system, and most cancer cells express telomerase activity from those used by Szilard, an adjustment of some of the at a sufficient level to sustain continuous replication of parameters in Szilard’s original model was necessary to the chromosome ends without loss of genetic material. obtain a good fit-by-eye to the mortality data. As de- When telomere lengths are measured in somatic cell scribed above, Szilard obtained the values for the cultures, a correlation is observed between telomere average time interval between two subsequent aging length and the number of cell divisions preceding hits in a somatic cell (t) and the average number of senescence and death (Allsopp et al. 1992). However, inherited faults (average value of r) by fitting the model there is now a consensus that telomere length correlates to observed mortality data of the U.S. population in the to but does not cause aging in most individuals, with mid-20th century. It is therefore natural that slightly immunosenescence being the most tangible correlate different parameter values are needed to fit the model (Effros 2007). Exceptions may be a variety of rare to mortality data obtained half a century later in premature aging syndromes, such as dyskeratosis con- Sweden. The best fit to current mortality data was genita (Kirwan and Dokal 2008) and idiopathic obtained by using a value of 6.35 years for t and an pulmonary fibrosis (Blasco 2005), and some cases of average value of r of 1.4. For such small values of r, we are liver cirrhosis (Wiemann et al. 2002) and myelodysplastic not motivated to use a Gaussian distribution, as Szilard syndromes (Ohyashiki et al. 1994), where mutations in did, and we therefore used a Poisson distribution for the telomere maintenance genes have been identified. number of inherited faults. The nongenetic scattering Many other stimuli may induce a senescence re- was, however, included in the same way as Szilard’s by sponse. These include nontelomeric DNA damage, applying a Gaussian distribution with a standard de- particularly double-strand breaks that activate the p53 viation of 3.0 years to the results. It actually turns out pathway, and certain oncogenes that trigger the p16- (Figure 1) that Szilard’s original model better fits the pRB pathway (Campisi 2005b). In most human cell observed mortality data than the revised model. This is systems, only 10–40 double-strand breaks per cell are mainly because the Poisson distribution for an average required for permanent cell-cycle arrest. In his model, of 1.4 events, the average number of inherited faults in Szilard emphasizes ionizing radiation as a potent aging Szilard’s original model determined by fitting the factor. Indeed, ionizing radiation is the most powerful model to observed mortality data, is much more skewed inducer of double-strand breaks and it is well known than for an average of 20 events, the average number of that radiation therapy causes accelerated aging in inherited faults in the revised model determined from exposed tissues (Peters 1996). However, ionizing radi- empirical data (McConkey 2003). ation does not reduce the fraction of surviving cells by destruction of essential genes as postulated by Szilard. Instead, as stated above, the downstream effect that may WHAT IS NOT IN THE MODEL? be common to all aging phenomena seems to be p53- Two factors, not known at the time when Szilard induced terminal growth arrest and apoptosis. The presented his theory, but intimately linked to cellular central role of p53 in this process is corroborated by a senescence and the aging process, are the enzyme number of genetic observations. Transgenic mice ex- telomerase (Greider and Blackburn 1985) and the pressing overactive forms of p53 get less cancer but age transcription factor p53 (Rodier et al. 2007). Cellular faster than their wild-type controls (Tyner et al. 2002; senescence was first identified as a process that limits the Maier et al. 2004; Dumble et al. 2007). While p53 ability of normal human cells to proliferate in culture to knockout mice develop cancer at such a rate that it is 8 H. Zetterberg et al. impossible to examine any effects on aging phenotypes, Burkle, A., C. Brabeck,J.Diefenbach and S. Beneke, 2005 The the few p53/1 that do not get cancer age slower emerging role of poly(ADP-ribose) polymerase-1 in longevity. Int. J. Biochem. Cell. Biol. 37: 1043–1053. compared with p53 wild-type mice (Donehower et al. Burnet, F. M., 1973 A genetic interpretation of ageing. Lancet 2: 1992). In addition, humans homozygous for a p53 480–483. ampisi polymorphism that leads to a mild impairment of p53 C , J., 2005a Senescent cells, tumor suppression, and organis- mal aging: good citizens, bad neighbors. Cell 120: 513–522. function live longer but die 2.5-fold more often from Campisi, J., 2005b Suppressing cancer: the importance of being sen- cancer (van Heemst et al. 2005), suggesting that p53 escent. Science 309: 886–887. rowley urtis protects against cancer but at a cost of longevity. C , C., and H. J. C , 1963 The development of somatic mutations in mice with age. Proc. Natl. Acad. Sci. USA 49: 626–628. Unfortunately, the model suggested by Szilard does Curtis, H., and C. Crowley, 1963 Chromosome aberrations in not allow for analysis of these types of relatively liver cells in relation to the somatic mutation theory of aging. complicated aging events. To summarize, the model Radiat. Res. 19: 337–344. Donehower, L. A., M. Harvey,B.L.Slagle,M.J.McArthur,C.A. may be considered out of date for the following reasons: Montgomery,Jr. et al., 1992 Mice deficient for p53 are devel- (i) the assumption of chromosome inactivation is in- opmentally normal but susceptible to spontaneous tumours. Na- valid, (ii) the statistical assumption of a Gaussian ture 356: 215–221. Dumble, M., L. Moore,S.M.Chambers,H.Geiger,G.Van Zant distribution is flawed, (iii) the critical surviving fraction et al., 2007 The impact of altered p53 dosage on hematopoietic of cells is organ specific and cannot be approximated to stem cell dynamics during aging. Blood 109: 1736–1742. the whole organism, (iv) the role of cell proliferation in Effros, R. B., 2007 Telomerase induction in Tcells: A cure for aging and disease? Exp. Gerontol. 42: 416–420. replacing dead or senescent cells is ignored, and (v) Foster, D. W., 1998 Diabetes mellitus, pp. 2060–2081 in Harrison’s Prin- there is no recognition in the original model that a ciples of Internal Medicine,editedbyA.S.Fauci,E.Braunwald,K.J. frequent consequence of DNA damage is precancerous Isselbacher,J.D.Wilson,J.B.Martin et al. McGraw-Hill, New York. arinis van der orst ijg oeijmakers cell proliferation rather than cell senescence and apo- G , G. A., G. T. H ,J.V and J. H. H , 2008 DNA damage and ageing: new-age ideas for an age-old ptosis. Nevertheless, Szilard’s model was the first serious problem. Nat. Cell Biol. 10: 1241–1247. hypothesis to posit accumulated genetic damage as an Greider, C. W., and E. H. Blackburn, 1985 Identification of a spe- important cause of senescence, a view most researchers cific telomere terminal transferase activity in Tetrahymena ex- arinis tracts. Cell 43: 405–413. agree with today (G et al. 2008), and in spite of Hart, R. W., and R. B. Setlow, 1974 Correlation between deoxyri- being oversimplified and on some aspects wrong, it also bonucleic acid excision-repair and life-span in a number of mam- predicts population-based death rates in a quite accurate malian species. Proc. Natl. Acad. Sci. USA 71: 2169–2173. Hayflick, L., 1965 The limited in vitro lifetime of human diploid manner. In the literature, Szilard is often described as a cell strains. Exp. Cell Res. 37: 614–636. visionary man with a remarkable ability to make accurate Hekimi, S., and L. Guarente, 2003 Genetics and the specificity of predictions on grounds not always clear to his fellow the aging process. Science 299: 1351–1354. Henshaw, P. S., E. F. Riley and G. E. Stapleton, 1947 The biologic scientists. In a sense, his model of the aging process may effects of pile radiations. Radiology 49: 349–364. be another example of this extraordinary gift. Hentges, K. E., D. D. Pollock,B.Liu and M. J. Justice, 2007 Regional variation in the density of essential genes in The authors thank Georg Klein, Adam Wilkins, and two anonymous mice. PLoS Genet. 3: e72. reviewers for helpful comments on the manuscript. H.Z. was sup- Herndon,L.A.,P.J.Schmeissner,J.M.Dudaronek,P.A.Brown, ported by the Swedish Research Council grant 2006-6227; M.Z. was K. M. Listner et al., 2002 Stochastic and genetic factors influence supported by the Swedish Research Council grant 2008-3065; P.B. was tissue-specific decline in ageing C. elegans. Nature 419: 808–814. supported by the Swedish Cancer Society grant 4956, the Swedish Holliday, R., 2000 Interview: John Maynard Smith on ageing re- Research Council grant 2007-2801, the Swedish Radiation Safety search. Biogerontology 1: 185–189. olliday irkwood Authority, and King Gustav Jubilee Clinic Cancer Research Founda- H , R., and T. B. K , 1981 Predictions of the somatic tion; and O.H. was supported by the Swedish Cancer Society grant mutation and mortalization theories of cellular ageing are con- 070489 and the Swedish Pain Foundation grant B41-f 978/08. The trary to experimental observations. J. Theor. Biol. 93: 627–642. Kirkwood, T. B., and A. Kowald, 1997 Network theory of aging. authors declare no conflict of interest. Exp. Gerontol. 32: 395–399. Kirkwood, T. B., and C. J. Proctor, 2003 Somatic mutations and ageing in silico. Mech. Ageing Dev. 124: 85–92. Kirwan, M., and I. Dokal, 2008 Dyskeratosis congenita: a genetic LITERATURE CITED disorder of many faces. Clin. Genet. 73: 103–112. Klein, G., 1990 The Atheist and the Holy City. MIT Press, Cambridge, MA. Adelman, R., R. L. Saul and B. N. Ames, 1988 Oxidative damage to Kuningas, M., S. P. Mooijaart,D.van Heemst,B.J.Zwaan,P.E. DNA: relation to species metabolic rate and life span. Proc. Natl. Slagboom et al., 2008 Genes encoding longevity: from model Acad. Sci. USA 85: 2706–2708. organisms to humans. Aging Cell 7: 270–280. Al-Awqati, Q., and P. A. Preisig, 1999 Size does matter: Will knock- Lanouette, W., 1992 Genius in the Shadows. A Biography of Leo Szilard, out of p21(WAF1/CIP1) save the kidney by limiting compensa- the Man Behind the Bomb. University of Chicago Press, Chicago. tory renal growth? Proc. Natl. Acad. Sci. USA 96: 10551–10553. Maas, W., and J. F. Crow, 2004 Leo Szilard: a personal remem- Allsopp, R. C., H. Vaziri,C.Patterson,S.Goldstein,E.V. brance. Genetics 167: 555–558. Younglai et al., 1992 Telomere length predicts replicative ca- Maier, B., W. Gluba,B.Bernier,T.Turner,K.Mohammad et al., pacity of human fibroblasts. Proc. Natl. Acad. Sci. USA 89: 2004 Modulation of mammalian life span by the short isoform 10114–10118. of p53. Genes Dev. 18: 306–319. Aurigemma, G. P., J. S. Gottdiener,L.Shemanski,J.Gardin and D. Maynard Smith, J., 1959 A theory of ageing. Nature 184: 956–957. Kitzman, 2001 Predictive value of systolic and diastolic func- McConkey, E. H., 2003 Human Genetics: The Molecular Revolution. tion for incident congestive heart failure in the elderly: the car- Jones & Bartlett, Boston. diovascular health study. J. Am. Coll. Cardiol. 37: 1042–1048. Morimoto, R. I., 2008 Proteotoxic stress and inducible chaperone Blasco, M. A., 2005 Telomeres and human disease: ageing, cancer networks in neurodegenerative disease and aging. Genes Dev. 22: and beyond. Nat. Rev. Genet. 6: 611–622. 1427–1438. Perspectives 9

Morley, A. A., S. Cox and R. Holliday, 1982 Human lymphocytes Szilard, L., 1959 On the nature of the aging process. Proc. Natl. resistant to 6-thioguanine increase with age. Mech. Ageing Dev. Acad. Sci. USA 45: 30–45. 19: 21–26. Szilard, L., 1964 On memory and recall. Proc. Natl. Acad. Sci. USA Novick, A., and L. Szilard, 1950 Description of the chemostat. 51: 1092–1099. Science 112: 715–716. Tucker, J. D., M. D. Spruill,M.J.Ramsey,A.D.Director and Novick, A., and M. Weiner, 1957 Enzyme induction as an all-or- J. Nath, 1999 Frequency of spontaneous chromosome aberra- none phenomenon. Proc. Natl. Acad. Sci. USA 43: 553–566. tions in mice: effects of age. Mutat. Res. 425: 135–141. Ohyashiki, J. H., K. Ohyashiki,T.Fujimura,K.Kawakubo,T. Turner, D. R., A. A. Morley,M.Haliandros,R.Kutlaca and B. J. Shimamoto et al., 1994 Telomere shortening associated with Sanderson, 1985 In vivo somatic mutations in human lympho- disease evolution patterns in myelodysplastic syndromes. Cancer cytes frequently result from major gene alterations. Nature 315: Res. 54: 3557–3560. 343–345. Palm, W., and T. de Lange, 2008 How shelterin protects Mamma- Tyner, S. D., S. Venkatachalam,J.Choi,S.Jones,N.Ghebranious lian telomeres. Annu. Rev. Genet. 42: 301–334. et al., 2002 p53 mutant mice that display early ageing-associated Pardee, A., F. Jacob and J. Monod, 1958 The genetic control of cy- phenotypes. Nature 415: 45–53. toplasmic expression of ‘‘inducibility’’ in the synthesis of beta- van Heemst, D., S. P. Mooijaart,M.Beekman,J.Schreuder,A.J.de galactosidase by E. coli. J. Mol. Biol. 1: 165–178. Craen et al., 2005 Variation in the human TP53 gene affects old Passos, J. F., G. Saretzki,S.Ahmed,G.Nelson,T.Richter et al., age survival and cancer mortality. Exp. Gerontol. 40: 11–15. 2007 Mitochondrial dysfunction accounts for the stochastic Watson,J.D.,andF.H.Crick, 1953 Molecular structure of nucleic heterogeneity in telomere-dependent senescence. PLoS Biol. acids; a structure for deoxyribose nucleic acid. Nature 171: 737–738. 5: e110. West,M.J.,C.H.Kawas,W.F.Stewart,G.L.Rudow and Peters, L. J., 1996 Radiation therapy tolerance limits. For one or for J. C. Troncoso, 2004 Hippocampal neurons in pre-clinical all?–Janeway Lecture. Cancer 77: 2379–2385. Alzheimer’s disease. Neurobiol. Aging 25: 1205–1212. Ramsey, M. J., D. H. Moore,2nd,J.F.Briner,D.A.Lee,L.Olsen Wiemann, S. U., A. Satyanarayana,M.Tsahuridu,H.L.Tillmann, et al., 1995 The effects of age and lifestyle factors on the accu- L. Zender et al., 2002 Hepatocyte telomere shortening and se- mulation of cytogenetic damage as measured by chromosome nescence are general markers of human liver cirrhosis. FASEB J. painting. Mutat. Res. 338: 95–106. 16: 935–942. Rodier, F., J. Campisi and D. Bhaumik, 2007 Two faces of p53: ag- Vijg, J., 2000 Somatic mutations and aging: a re-evaluation. Mutat. ing and tumor suppression. Nucleic Acids Res. 35: 7475–7484. Res. 447: 117–135.