An Essay on "Genes for Aging"
Miller lab home page


A Position Paper on Longevity Genes

Richard A. Miller

November, 2001


Originally prepared for SAGE-KE.


Like all too many ill-used terms in biogerontology, the notion of "longevity genes" has meant different things to different researchers. As a result, sorting out these nuances is a key first step in organizing a discussion on this slippery topic. I think we need to consider six kinds of "longevity genes."

1. Genes that cause aging and do so for a living. This category (an empty set, I think), would, if it existed, include genetic alleles that evolved to bring about the aging process--that is, genetic variants molded by selective pressures to transform healthy young adults into senescent, vulnerable old organisms. I mention this notion only to dispose of it: There are many scientists, including some gerontologists, who are confident that genes that cause aging can be positively selected because they create old individuals whose prompt death will promote the success of their offspring. That may be true for salmon and other semelparous species whose convenient death provides quick nutrition for needy fry; but for other kinds of organisms, an allele that promotes rapid aging and early death can only diminish the number of healthy offspring produced by its unlucky owner.

2. Genes that alter longevity because they increase the risk of a specific illness early in life. Genes that cause phycomelia, blindness, congenital cardiac abnormalities, or juvenile diabetes will (in most environments) lead to a dramatic diminution of life-span. In my view, these genes, however much they might tell us about development, physiology, and pathology, are unlikely to provide important insights into aging. Although these mutations affect life expectancy, they do so by different means other than aging. These mutations affect life expectancy, and so does aging, but by different means.

In some situations, particularly experiments that use genetically identical rodents, it may be hard to distinguish genes of this kind from (hypothetical) genes that actually do alter aging per se. Consider a population of mice, for example, in which nearly all the mice die, at about 12 months, from an autoimmune syndrome for which a specific recessive allele is a key risk factor. In this situation, other alleles at the susceptibility locus are likely to have a major positive effect on longevity simply by preventing the autoimmune syndrome.

The literature is filled with publications whose titles announce genes for "accelerated aging" or "senescence-accelerated mice," based largely on evidence that the genes shorten life-span. I endorse David Harrison's pithy discussion of such claims (1) and believe that authors of such papers should be required to present very strong evidence that their allele of interest speeds up multiple aspects of the aging process. These effects should be observed at the cellular, extracellular, and intracellular levels and in multiple organs and tissues before one accepts the radical claim that the shortened life-span is really the result of accelerated aging.

2a. Genes that alter longevity because they increase the risk of a specific illness early in life whose features resemble, to some extent, some of the consequences of aging. Werner syndrome and Hutchinson-Guilford progeria are the poster children for this category, as Klotho (2) is the poster rodent. In some cases, including Werner syndrome, the list of traits seen in both patients and in normal old people is impressive, but so is the oft-omitted list of traits that don't overlap. I think those who wish to work out the pathogenesis of these "premature aging" syndromes and want to determine whether the same pathways contribute to pathogenesis in normal aging are working on a problem of real interest. And there is a reasonably good chance that these studies will provide important clues about aging. But the notion that these loci play a major role in timing aging per se seems to me much less likely, if only because it is so easy to tell the difference between a Werner patient and an old person.

3. Genes that affect what kind of old person you are. There are a lot of these: genes that "cause" (more strictly, genes that influence the risk of) Alzheimer's disease, breast cancer, macular degeneration, adult-onset diabetes, hair loss, sarcopenia, immune senescence, and a raft of other complaints. Estimates of the number of genes that "influence aging"--bruited in the famous essay by George Martin (3) on "segmental progerioid genes"--are based on the following kind of criterion: A locus where a mutation produces one or more of the changes seen in old people can be scored as "age-related." Such calculations lead to the conclusion that thousands or tens of thousands of genes "influence aging." It's important, however, to realize that this calculation does not provide an estimate of the number of genes that time the aging process: that is, the (hypothetical) genes that determine how long it takes for all these bad things to start becoming worrisome. Evolutionary theory provides useful insights into why the genome accumulates so many polymorphic variants whose late-life consequences are unfortunate: Mutations that lead to deleterious late-life changes are subject to little or no negative selection pressure, because very few breeding individuals survive to the ages at which such changes become apparent. Alleles at such loci may well alter life expectancy. Indeed, alleles that promote or postpone adult-onset diabetes or atherosclerosis could well lead to important differences in the likelihood of living to age 85 or 90 (or beyond, in countries that boast a high standard of living and an excellent health care system), just as alleles that slow or speed the loss of immune and muscle function might have had similar effects among 40-year-old Cro-Magnon matrons. Some of these genes--those of largest effect--may deserve to be counted as "longevity genes." Whether they are genes that alter the rate of aging is another matter.

4. Low-fitness genes that extend maximum life-span, probably by slowing down aging. There are now dozens of such discoveries in flies and worms and a handful in mice, cataloged as of early 2001 in a recent review (4). In very few cases does the evidence go beyond documentation of extended longevity to include analyses of other age-related traits. In such circumstances, the inference that the alleles alter aging (rather than, for example, a single common cause of death) rests on the plausible intuition that so many nasty things happen in old age that abolition of any one of them would have little measurable net effect on adult life expectancy. This intuition stands on firm demographic data in humans (5, 6), although for mice and particularly for squishier creatures, such as flies or worms, a certain amount of hand-waving must be carefully applied. In a few cases, data are just beginning to establish that long-lived mutant mice, such as the Snell dwarf dw/dw (Fig. 1), show other signs of slowed aging, including slower senescence of immune, joint, and connective tissues, in addition to their exceptional longevity (7).

I think it is a good bet that many of these alleles are not only "longevity genes" but also anti-aging genes. That is, they have their observed effect on longevity because they slow progression of a wide range of age-related changes in biochemical and physiological pathways, although I'll be more confident of this inference once additional relevant data are acquired.

How such discoveries illuminate questions in biogerontology requires careful consideration. The specific alleles in question are almost certainly not involved in regulating differences in life expectancy among individuals in normal populations, because in many cases these alleles have other phenotypic consequences that diminish Darwinian fitness except in well-defined laboratory settings. A hampered ability of some long-lived mutant worms to compete for food under conditions designed to mimic natural fluctuations of food availability has been demonstrated directly (8). In addition, the observation that long-lived mutations do not spontaneously arise in continuous laboratory cultures suggests that such alleles do not provide any selective advantage under ordinary conditions. Dwarf mice, similarly, make poor spouses, because the females cannot carry litters to term unless they receive hormone injections, whereas the males have mechanical difficulties in romantic engagements, even if they have not previously been eaten by their normal-sized brothers and uncles.

No, the value of such loser alleles is that they provide revealing glimpses of the control panel by which aging can be regulated. The continuing triumphs (9) of the worm, fly, and maybe someday mouse biologists as they use long-lived mutants to develop and test mechanistic hypotheses about aging are beyond the scope of this short article. The value of these alleles is, in a curious sense, more physiological than genetic: Although the loci themselves are unlikely to play a role in longevity determination in nature (and the specific alleles are even less likely to do this), the specific pathways--biochemical and cellular--altered by the mutations may well be those that nature uses to produce long-lived species or individuals.

5. Alleles and allele combinations that alter life-span because they affect aging. Heritability of life expectancy in most studies of flies, rodents, and people is approximately 15 to 25%. For the non-geneticists in the audience, heritability is a well-defined quantitative measure used throughout modern genetics. It is, roughly speaking, the proportion of the variance in a trait--life-span in this case--that can be accounted for by genetic factors in a specific set of genotypes and environments. Given the 15 to 25% estimate, it is thus safe to conclude that life expectancy is influenced by inherited genetic variants, although to a degree that many nonspecialists feel is surprisingly low. Some of this variation is undoubtedly due to genes that we've already discussed as category 3: genes that influence the risk of specific late-life illnesses and thus influence overall life expectancy but do not do so by an effect on the aging process itself. This leaves an unknown proportion of variance (not more than about 25% and possibly a good deal less) to be blamed on hypothetical polymorphisms that act by changing the aging rate (that is, by changing in parallel the timing of a wide range of mid-life and late-life alterations, some of them diseases). Loci with these properties can be termed "quantitative trait loci" (QTL). If there are indeed any polymorphic genetic loci that influence aging rate within a species, then in principle these may include a large number with tiny invisible effects, a small number with detectably large effects, or both.

There is now a small amount of evidence, from flies (10), mice (11), and worms (12, 13), that QTL with detectably large effects on life-span are indeed present in segregating populations of laboratory stocks. Our own laboratory, for example, has reasonably strong evidence for at least five such loci segregating in a four-way cross derived from four stocks (BALB/c, C57BL/6, DBA/2, and C3H/He) of mice. Such data provide the starting point for asking a number of newly available questions about the genetics of aging (as opposed to the genetics of longevity!) in mice:

Do these alleles alter life-span by altering aging? So far the data are razor-thin, but not quite invisible. In our own study, each of the polymorphisms that has a significant impact on all-cause mortality seems to influence life-span in two groups of mice: those dying of neoplasia and those dying of something else. These data thus suggest that the alleles in question may be influencing longevity by means of a general effect on multiple forms of late-life vulnerability. More specific tests are clearly needed (and are now in progress) to see whether mice that differ systematically in the alleles of interest show altered trajectories of age-dependent change in multiple cells, tissues, and organ systems.

Do these alleles have similar effects in other environments? Studies in flies (10) have suggested that the effects of specific alleles may depend to a great extent on other, "background" alleles, on gender, on environmental variables such as population density, and on complex interactions among these independent factors. Our own mouse data show that some of our QTL alleles have longevity effects only in males, or only in females, but rarely in both sexes. In addition, our unpublished work suggests that QTL that alter longevity in virgin females may be less influential in mated females.

Are the polymorphic alleles available for study in common laboratory stocks a reliable sampling of loci that influence aging in naturally occurring populations? There are excellent reasons (14, 15) for thinking that adaptation to laboratory environments may systematically eliminate alleles that are associated, in the wild population, with variation in the timing of life history events, potentially including regulation of aging. There is some evidence (scant, but accumulating) that populations of both flies and mice derived from recently caught wild populations may indeed live substantially longer than laboratory-adapted stocks of comparable genetic heterogeneity (for example, see "Wild Thing").

How do these QTL affect life-span: What specific biochemical, hormonal, or cellular changes do these QTL induce that lead to differences in longevity and risk of multiple late-life diseases.

What do the QTL encode? QTL studies produce evidence for the existence and rough position of loci but do not easily yield specific cloned candidate loci for follow-up studies. Although it is possible that fine-scale mapping followed by pursuit of attractive candidates may eventually yield cloned genes with effects on life-span (or even effects on aging), exploitation of QTL as clues to physiological mediators of aging and disease is likely to be more productive in the short term.

The use of QTL studies to guide physiological analysis is likely to pay dividends even without, and well before, the positional cloning of the relevant genes. It should be possible, for example, to compare mice with differing alleles at these QTL to see whether they differ in specific enzyme or hormone levels that are thought to play important roles in aging. These results can also be used to guide searches for human genes on syntenic chromosomes that might play equivalent roles in modulating disease risks in aging people. (Chromosomes are syntenic if they contain long stretches of homologous genes, suggesting that they are derived from a common ancestral chromosome. 

6. The genes that really count: genes that slow aging in dogs, humans, porcupines, and naked mole rats. It is possible to construct a rodent that lives 20 to 30 years: Sumatran crested porcupines top out at 27.75 years, and naked mole rats (30-g beasts that are about 25% jaw and teeth) last 20 to 30 years, depending on which anecdotes you believe. The genes that do the heavy lifting--the ones we needed to acquire to outlive baboons, lemurs, and chimps, not to mention our prosimian ancestors--are the ones we need to learn about if we wish to do more than nibble around the edges of the aging/gene problem. Unfortunately, these loci are not going to be identified by mating mice to porcupines or hominids to prosimians.

There are two schools of thought about how many genes need to be modified to make very large species-specific changes in aging rates. Both schools of thought agree that when a species finds itself in a low-risk environment--by learning to fly, by accidental isolation on a predator-free island, by growing sharp quills, or by learning to gossip and throw rocks--circumstances shift so that genomes that slow down life history patterns now find themselves at an advantage. Life history patterns that produce smaller litters over a long period of time now do better than the old-fashioned system of betting the farm on a single, big, early litter. The genes that support this slowed-down maturational pattern get selected and produce the kind of tougher creature that can last years or decades instead of months.

But how many genes need to change to bring about this new state of affairs? One view (not my own) argues that the number might be very large: The transition could require changes that lead to postponement of cancer; others that lead to delays in muscle and bone degeneration; and still others that lead to the slowing of cataract formation, to changes in liver gene expression, or to longer-lasting resistance to infectious agents. The main challenge to this model is that of synchrony: How is it, exactly, that eyes and bones and immune systems and antineoplastic precautions all happen to last about 10 years in dogs and 30 years in chimps but only 2 years in mice? The proffered explanation is that evolution can accomplish a great deal through successive, fine-scale adjustments, producing carefully tailored patterns of cell and gene function that last just long enough to get the reproductive job done, but not so long as to involve an unnecessary investment of the limited resources available.

My own hunch is that the number of genes that control the rate of aging across (mammalian) species is a good deal smaller--perhaps a dozen or fewer--and that these act to time the pace of multiple developmental and senescent processes in much the same way that a rheostat can control multiple independent and interacting electric circuits throughout a building complex. By my interpretation, the evidence from caloric restriction experiments and from studies of single-gene mutations, such as those that produce long-lived dwarf mice, implies that a very wide range of age-related processes can indeed be slowed in a coordinated fashion by flipping levers placed at a few strategic spots. While acknowledging the remarkable powers of evolution working over geologic time, I suspect that aging poses a special case in that the risk of so many potentially lethal problems develops exponentially over the last third of the possible life-span for each species. If an early primate, with its life history focused on reproduction between the ages of 4 to 8 years and whose survival to 12 years is rather unlikely, finds itself in a niche sufficiently plush to support reproduction between the ages of 8 and 16 years (and possible survival to age 24 or so), it seems to me that mutations that merely delay cataracts would have little value, because an individual with the best eyesight in the world cannot survive without bones or muscles or teeth. Nor would mutations that delay the development of all forms of cancer have much effect on reproductive output in the new critical age range (8 to 16 years), because cancer-free individuals with an exponentially increasing risk of immunodeficiency and bone loss and cataracts and cardiovascular compromise would not be in much better shape. Mutations in (hypothetical) genes that regulate the pace of multiple developmental and degenerative processes are, in my view, much more likely to produce coordinated deceleration of age changes than are mutations in many genes, each of whose functions are of more limited scope. The (hypothetical) loci that accomplish this coordination may or may not work through the same biochemical and cellular pathways that are tweaked by caloric restriction, by dauer-control and stress-resistance genes that do the trick for worms, and by the insulin-like growth factor-1 responses implicated in longevity control in mice and dogs. Finding the genes that adjust life history patterns and aging rates during the evolution of slow-aging species will require an unusual degree of interaction among comparative biologists, one-species geneticists, comparative geneticists, and gerontologically oriented cell biologists and physiologists.

So, in summary, I postulate that there are six different kinds of genes that can be or have been called "longevity genes." One of these six classes (the first) represents a red herring--and probably a "straw" herring; another (the sixth), if it exists at all, is where the action is. The situation calls for a kind of Pascal's wager (16): Genes that make major changes in the aging rate, enough to worry life insurance agents, will not be found at all except by researchers who make the assumption that there are such loci and set out to triangulate their position by imaginative use of both interspecies and intraspecies comparisons.

November 28, 2001


References cited:

   1. D. E. Harrison, Potential misinterpretations using models of accelerated aging. J. Gerontol. Biol. Sci. 49, B245-B245 (1994).

   2. M. Kuro-o et al., Mutation of the mouse klotho gene leads to a syndrome resembling ageing. Nature 390, 45-51 (1997).

   3. G. M. Martin, in Genetic Effects on Aging, D. Bergsma, D. E. Harrison, Eds. (Liss, New York, 1978), pp. 5-39.

   4. R. A. Miller, in Handbook of the Biology of Aging, E. J. Masoro, S. N. Austad, Eds. (Academic Press, San Diego, CA, 2001), chap.14.

   5. S. J. Olshansky, B. A. Carnes, A. Desesquelles, Prospects for human longevity. Science 291, 1491-1492 (2001).

   6. S. J. Olshansky, B. A. Carnes, C. Cassel, In search of Methuselah: estimating the upper limits to human longevity. Science 250, 634-640 (1990).

   7. K. Flurkey, J. Papaconstantinou, R. A. Miller, D. E. Harrison, Lifespan extension and delayed immune and collagen aging in mutant mice with defects in growth hormone production. Proc. Natl. Acad. Sci. U.S.A. 98, 6736-6741 (2001).

   8. D. W. Walker, G. McColl, N. L. Jenkins, J. Harris, G. J. Lithgow, Evolution of lifespan in C. elegans. Nature 405, 296-297 (2000).

   9. L. Guarente, C. Kenyon, Genetic pathways that regulate ageing in model organisms. Nature 408, 255-262 (2000).

  10. J. Leips, T. F. Mackay, Quantitative trait loci for life span in Drosophila melanogaster: interactions with genetic background and larval density. Genetics 155, 1773-1788 (2000).

  11. A. U. Jackson, A. T. Galecki, C. Chrisp, D. T. Burke, R. A. Miller, Mouse loci associated with life span exhibit sex-specific and epistatic effects. J. Gerontol. Biol. Sci. 57:B9-B15 (2002).

  12. R. H. Ebert, V. A. Cherkasova, R. A. Dennis, J. H. Wu, S. Ruggles, T. E. Perrin, R. J. Shmookler Reis, Longevity-determining genes in Caenorhabditis elegans: chromosomal mapping of multiple noninteractive loci. Genetics 135, 1003-1010 (1993).

  13. D. R. Shook, T. E. Johnson, Quantitative trait loci affecting survival and fertility-related traits in Caenorhabditis elegans show genotype-environment interactions, pleiotropy and epistasis. Genetics 153, 1233-1243 (1999).

  14. R. A. Miller, S. Austad, D. Burke, C. Chrisp, R. Dysko, A. Galecki, V. Monnier, Exotic mice as models for aging research: polemic and prospectus. Neurobiol. Aging 20, 217-231 (1999).

  15. R. A. Miller, R. Dysko, C. Chrisp, R. Seguin, L. Linsalata, G. Buehner, J. M. Harper, S. Austad, Mouse (Mus musculus) stocks derived from tropical islands: new models for genetic analysis of life history traits. J. Zool. 250, 95-104 (2000).

  16. Pascal's Wager, Stanford Encyclopedia of Philosophy (2001).








Wear and Tear