CORE Metadata, citation and similar papers at core.ac.uk

Provided by Elsevier - Publisher Connector

Cell, Vol. 84, 497–500, February 23, 1996, Copyright 1996 by Cell Press Replicative Senescence: Minireview An Old Lives’ Tale?

Judith Campisi underlying cause of organismic aging. This idea is much Life Sciences Division more controversial and less intuitively obvious. How did Department of Cancer Biology it come about, and what is the evidence for it? Berkeley National Laboratory University of California Berkeley, California 94720 All Men Are Mortal (but What of Their Cells?) The observation that cell division is inherently limited Normal animal cells, with few exceptions, do not divide contradicted an earlier belief, championed by Carrel, indefinitely. This property, termed the finite replicative that vertebrate cells can proliferate indefinitely once re- life span of cells, leads to an eventual arrest of cell moved from the organism. Carrel’s belief stemmed in division by a process termed cellular or replicative se- part from his apparent (but since irreproducible) ability nescence. continually to subculture chick cells. This study and Although predicted and observed earlier, replicative others (see Hayflick, 1965) spawned the idea that cells senescence was first formally described over 30 years may be intrinsically “immortal.” Implicit in some of these ago when Hayflick and his colleagues reported that hu- studies was the idea that understanding cell immortality man fibroblasts gradually and inevitably lost their ability might uncover the basis for organismic mortality (Carrel, to proliferate upon continual subculture (Hayflick, 1965). 1912). With this idea as a backdrop, then, Hayflick as- Since then, many cell types from many animal species serted that “normal human diploid cell strains in vitro have been shown to have a finite replicative life span are in fact ‘mortal’” and suggested that replicative life (see Stanulis-Praeger, 1987). Most of these studies have span in culture reflects processes that occur during used cells in culture. However, a limited number of in the chronological life span, or aging, of the organism vivo experiments, as well as the evidence discussed (Hayflick, 1965). here and elsewhere (Stanulis-Praeger, 1987; Campisi et In retrospect, there was little basis for equating cell al., 1996), strongly suggest that is replicative life span with organismic life span. Certainly, not an artifact of culture. replicatively immortal cells can die just as readily as replicatively mortal cells. Moreover, mammals are repli- catively immortal (through the germline), even though Is Replicative Senescence Physiologically individuals age and die. Why, then, was the parallel Important? between the replicative life span of cells and organismic In higher organisms, particularly mammals, two views life span, or aging, accepted? In fact, it was not. There suggest that replicative senescence may have important were 30 years ago, and remain today, many skeptics. physiologic consequences. Nonetheless, Hayflick and other biologists, perhaps in- One view holds that cellular senescence is a tumor tuitively, pursued the idea that cellular senescence and suppressive mechanism. There is substantial molecular, aging are related. As a result, the idea has garnered cellular, and in vivo evidence to support this idea (Cam- increasing experimental support and increasing interest pisi et al., 1996). This evidence will not be reviewed here, among biologists. but is briefly summarized. First, senescence prevents cells from acquiring the multiple mutations that are needed for malignant transformation. Indeed, many, if Do All Normal Cells Senesce? not most, malignant tumors contain cells that have an When examined carefully, cells that can divide—with a extended or indefinite division potential. Thus, tumori- few notable exceptions—undergo replicative senes- genesis selects for cells that can wholly or partly bypass cence. However, whether cell senescence is a limited senescence. Second, certain oncogenes—both cellular or universal phenomenon has not been adequately ex- and viral—act at least in part by extending replicative plored and, for some cells, may not be easy to de- life span. Thus, oncogenic mutations and the strategies termine. of oncogenic viruses may and do entail mechanisms to Cells from some species—for example, many ro- overcome senescence. Third, among the genes needed dents—spontaneously escape senescence (immortal- to establish and maintain senescence are the and ize) at a measurable frequency (1 in 104–106 cells). Be- retinoblastoma susceptibility genes. These are well-rec- cause immortal cells rapidly overgrow senescing ognized tumor suppressors that, together, are the most cultures, it can be difficult to assess the replicative life commonly lost functions in human cancer. A related span of some cells in mass culture. Nonetheless, several idea suggests that tumor suppression is the adaptive cell types from a wide (but hardly exhaustive) variety of value of senescence, but senescence may have evolved species have been shown to have a finite replicative life to fine-tune tissue modeling during development (Mar- span (Rohme, 1981; see Stanulis-Praeger, 1987). Repli- tin, 1993). Of course, there is also intuitive and teleologic cative senescence is especially stringent in human cells, appeal to the idea that a growth-limiting process may which almost never spontaneously immortalize (McCor- suppress tumorigenesis. mick and Maher, 1988). A second view regarding the physiologic significance Even some single-celled organisms, such as Sacchar- of cell senescence holds that it reflects processes that omyces cerevisiae, clearly senesce when individual cells occur during organismic aging and may constitute an are monitored (Jazwinski, 1993). This suggests that a Cell 498

finite replicative life span may be a very primitive pheno- independent studies show a significant inverse relation- type. There is, however, a major difference in this regard ship between donor age and replicative capacity of cul- between higher eukaryotes and S. cerevisiae. In higher tured cells (Martin et al., 1970; see Stanulis-Praeger, organisms, daughter cells inherit the replicative “age” 1987; Cristofalo and Pignolo, 1993; Campisi et al., 1996). of their mother, minus one division. Thus, mammalian For example, human fetal fibroblasts typically senesce cultures accumulate senescent cells more or less expo- after 60–80 PDs, fibroblasts from young to middle-aged nentially, until they contain only senescent cells. By con- adults may do so after 20–40 PDs, and cells from old trast, S. cerevisiae daughters do not strictly inherit their adults may senesce after 10–20 PDs. mother’s replicative age, and cultures appear immortal. These studies suggest that cells in renewable tissues S. cerevisiae daughter cells are easily identified and may progressively exhaust their replicative life span in separated from mothers, but this is not true for many vivo during aging. If true, tissues and individuals may other cells. We do not know whether immortal cultures vary considerably in the rate at which replicative poten- from other species also contain cells that senesce but tial declines because cell turn over varies widely among do not transmit their replicative age to daughters. tissues and is very likely influenced by disturbances Only two, perhaps three, higher eukaryotic cell types such as infection, inflammation, or injury. may have an unlimited division potential. Certainly the Species Life Span germline is capable of continuous replication (although Interspecies comparisons suggest that cell replicative mature sperm and ova are not). In addition, as noted life span and organismic life span are genetically related. above, tumor cells are often immortal. Finally, some Although limited in scope, these studies show that cells stem cells may be immortal (for example, inner cell mass from short-lived species tend to senesce after fewer cells, spermatogonia, hematopoietic stem cells), but this PDs than cells from long-lived species (Rohme, 1981; has yet to be critically demonstrated. see Stanulis-Praeger, 1987; Campisi et al., 1996). For example, mouse fibroblasts typically senesce after 10–15 PDs, whereas Galapagos tortoise fibroblasts pro- What Is Replicative Senescence? liferate for >100 PDs. Replicative senescence entails an irreversible arrest of These studies suggest that the replicative life span of cell proliferation and altered cell function. It is controlled cells and chronological life span of organisms may be by multiple dominant-acting genes and depends on the controlled by overlapping or interacting genes. number of cell divisions, not time. It also depends on Premature Aging Syndromes the cell type and on the species and age of the donor A genetic link between aging and replicative life span (see Stanulis-Praeger, 1987; Cristofalo and Pignolo, is also supported by studies of hereditary premature 1993; Campisi et al., 1996). aging syndromes in man (Martin et al., 1970; see Sta- Cells acquire three characteristics upon senescence. nulis-Praeger, 1987). The best studied of these is the First, they stably arrest growth with a G1 DNA content, Werner syndrome (WS), caused by a recessive mutation irreversibly losing the ability to enter Sphase in response on chromosome 8p11-p12 (Oshima et al., 1994). WS to physiologic mitogens. The cells remain metabolically patients are fairly asymptomatic early in life. Thereafter, active, and many genes remain mitogen inducible, but they prematurely develop many, but not all, phenotypic there are changes in a few key growth regulators. These correlates of age. These include hair and dermal thin- include the repression of three positive-acting transcrip- ning, atherosclerosis, osteoporosis, and cancer. WS is tional regulators and overexpression of a cyclin-depen- a segmental progeroid syndrome because only some dent protein kinase inhibitor (see Campisi et al., 1996). aging phenotypes are premature. WS cells senesce well Second, senescent cells acquire altered functions. In ahead of age-matched controls. Senescence may also fact, by the criteria of a stable irreversible growth arrest be segmental in WS cells. Senescent WS fibroblasts are and change in function, senescent cells resemble termi- similar to normal counterparts in many ways. However, nally differentiated cells (Goldstein, 1990; Campisi et c-fos repression, a hallmark of normal senescence, does al., 1996). Third, senescent cells acquire resistance to not occur (Oshima et al., 1995). Thus, a single locus apoptotic (programmed) cell death (Wang, 1995) and mutation accelerates age-related processes in vivo and thus are quite stable. The perception that cellular senes- cellular senescence in culture. The WS gene is not yet cence leads to cell death is incorrect. It may stem from cloned but, once identified, may yield extraordinary in- the early finding that senescent cells can be lost when sights into cell senescence, aging, and many age- subcultured (Hayflick, 1965; cf. Linskens et al., 1995) or related diseases. from confusion over the distinction between replicative Physiologic and Molecular Correlates mortality and cell viability. Cell senescence and organismic aging also share physi- ologic and molecular features. For example, senescent cells and aged tissues are more sensitive to a variety The Aging Connection of stresses. At a molecular level, the stress inducibility What is the evidence that cells senesce in vivo, and how of heat shock protein 70 is markedly attenuated in se- might this cause or contribute to organismic aging? nescent human fibroblasts (Choi et al., 1990) and several Donor Age tissues from aged rodents (Heydari et al., 1993; Fawcett Cells cultured from old donors tend to senesce after et al., 1994). In both cases, the attenuation is due to fewer population doublings (PDs) than cells from young reduced binding of a heat shock transcription factor. donors. There is considerable scatter in the data, partic- Similar culture/in vivo parallels exist for the regulation ularly in humans, which may in part be due to genetic of c-fos, certain protease inhibitors, and collagenase or life history differences (or both). Nonetheless, several (see Cristofalo and Pignolo, 1993; Campisi et al., 1996). Minireview 499

Another feature shared by senescence in culture and may have a substantial impact in aged tissue. For exam- aging in vivo is shortening. Telomere shorten- ple (see Campisi et al., 1996), senescent human skin ing is perhaps the best candidate for a cell division fibroblasts overexpress collagenase and underexpress “counting” mechanism in normal somatic cells of higher collagenase inhibitors. This maywell cause or contribute organisms (see Harley and Villeponteau, 1995; Wright to the collagen breakdown and thin dermis typical of and Shay, 1995). In such cells, telomerase is absent or aged skin. Senescent fibroblasts also underexpress in- present at low levels, and telomere length shortens with terleukin-6 and overexpress interleukin-1, cytokines PDs in culture and organismic age. In human fibroblasts, with pleiotropic inflammatory and immune effects. Thus, for example, mean telomere length decreases about 50 senescence-linked changes in differentiation could, at bp per doubling in culture and 15 bp per year of donor least in principle, have rather profound and far-ranging age (Allsopp et al., 1992). By contrast, telomerase is consequences for tissue function. Moreover, relatively relatively abundant in replicating germ cells, tumor cells, few senescent cells would be needed for some of these and lower eukaryotes, where telomere length is stable. effects. Finally, a neutral ␤-galactosidase activity is expressed Finally, resistance to apoptotic death may explain why by several types of human cells upon senescence in senescent cells are not cleared and thus accumulate culture and increases with age in human skin cells (Dimri with age. Indeed, it was recently proposed that caloric et al., 1995). Because this activity is not associated with restriction, which delays many age-related changes, quiescence, terminal differentiation, or immortality, it may do so by increasing the incidence of has provided in situ evidence that senescent cells do (Warner et al., 1995). in fact accumulate in aged tissue in vivo. Thus, age-related decrements in tissue function may, Postmitotic Cells at least in part, derive from an accumulation of senes- Organismic aging entails changes in both proliferative cent cells—which cannot proliferate, which resist apop- and postmitotic (for example, fat, nerve, and muscle) totic death, and which have an altered phenotype. cells. Age-related changes in postmitotic cells cannot, Evolutionary theories of aging have suggested that of course, be due to replicative senescence. Thus, re- traits selected to optimize health during the period of newable and postmitotic tissues may age by different reproductive fitness can have unselected deleterious (albeit interacting) mechanisms. If so, model organisms effects later in life. Replicative senescence may have such as Drosophila melanogaster or Caenorhabditis been selected, at least in mammals, to help ensure the elegans, in which adults are composed entirely of post- relative freedom from cancer that characterizes early mitotic cells, may provide insights into the aging of post- adulthood. However, this trait may be deleterious late in mitotic tissue, but not the aging of renewable tissues. life because dysfunctional senescent cells accumulate. Genes that extend the life span of the postmitotic organ- Moreover, because cancer incidence rises with age, it ism C. elegans have been identified and are discussed also appears that cellular senescence is an imperfect in by Kenyon (1996 [this issue of Cell]). We do not yet tumor suppressive mechanism that fails increasingly know whether these genes also extend the replicative with age. In fact, it would seem that late in life replicative life span in C. elegans embryonic cells, nor whether they senescence wreaks havoc—whether it succeeds or or their homologs alter the rate of aging in mammals. fails. In summary, there is substantial evidence, mostly but not wholly correlative, that senescent cells exist and Selected Reading accumulate with age in vivo. It has not yet been demon- strated that senescent cells cause or contribute to Allsopp, R.C., Vaziri, H., Patterson, C., Goldstein, S., Younglai, E.V., aging. However, the molecular biology of senescence Futcher, A.B., Greider, C.W., and Harley, C.B. (1992). Proc. Natl. Acad. Sci. USA 89, 10114–10118. is just developing, as is the prospect of detecting and controlling senescent cells in vivo. This said, how might Campisi, J., Dimri, G., and Hara, E. (1996). In Handbook of the Biol- ogy of Aging, Fourth Edition, E. Schneider and J. Rowe, eds. (New senescence cells contribute to aging in higher or- York: Academic Press), pp. 121–149. ganisms? Carrel, A. (1912). J. Exp. Med. 15, 516–528. Choi, H.S., Lin, Z., Li, B.S., and Liu, A.Y. (1990). J. Biol. Chem. 265, The Senescent Phenotype—A Triple-Edged 18005–18011. Sword? Cristofalo, V.J., and Pignolo, R.J. (1993). Physiol. Rev. 73, 617–638. One might imagine that cellular senescence could inhibit Dimri, G.P., Lee, X., Basile, G., Acosta, M., Scott, G., Roskelley, C., tissue repair. However, substantial numbers of prolifera- Medrano, E., Linskens, M., Rubelj, I., Pereira-Smith, O., Peacocke, M., and Campisi, J. (1995). Proc. Natl. Acad. Sci. USA 92, 9363–9367. tive cells can often be recovered from very old tissues, and wounds do heal even in very old mammals (although Fawcett, T.W., Sylvester, S.L., Sarge, K.D., Morimoto, R.I., and Hol- brook, N.J. (1994). J. Biol. Chem. 269, 32272–32278. sometimes more slowly). Of course, senescent cells may comprise only a small fraction of aged tissue. Indeed, in Goldstein, S. (1990). Science 249, 1129–1133. situ staining for the senescence-linked ␤-galactosidase Harley, C.B., and Villeponteau, B. (1995). Curr. Opin. Genet. Dev. 5, suggests that senescent cells are sparsely distributed 249–255. in aged skin (Dimri et al., 1995). Thus, the growth arrest Hayflick, L. (1965). Exp. Cell Res. 37, 614–636. of senescent cells may not be a major problem in aged Heydari, A.R., Wu, B., Takahashi, R., Strong, R., and Richardson, A. tissues, although it could certainly compromise the or- (1993). Mol. Cell. Biol. 13, 2909–2918. ganism by delaying repair. Jazwinski, S.M. (1993). Genetica 91, 35–51. By contrast, the altered function of senescent cells Kenyon, C. (1996). Cell 84, this issue. Cell 500

Linskens, M.H., Harley, C.B., West, M.D., Campisi, J., and Hayflick, L. (1995). Science 267, 17. Martin, G.M. (1993). J. Gerontol. 48, 171–172. Martin, G.M., Sprague, C.A., and Epstein, C.A. (1970). Lab. Invest. 23, 86–92. McCormick, J.J., and Maher, V.M. (1988). Mutat. Res. 199, 273–291. Oshima, J., Yu, C.E., Boehnke, M., Weber, J.L., Edelhoff, S., Wagner, M.J., Wells, D.E., Wood, S., Disteche, C.M., Martin, G.M., et al. (1994). Genomics 23, 100–113. Oshima, J., Campisi, J., Tannock, T.C., and Martin, G.M. (1995). J. Cell. Physiol. 162, 277–283. Rohme, D. (1981). Proc. Natl. Acad. Sci. USA 78, 5009–5013. Stanulis-Praeger, B. (1987). Mech. Aging Dev. 38, 1–48. Wang, E. (1995). Cancer Res. 55, 2284–2292. Warner, H.R., Fernandes, G., and Wang, E. (1995). J. Gerontol. 50A, B107–B109. Wright, W.E., and Shay, J.W. (1995). Trends Cell Biol. 5, 293–297.