• Open Access

The non-existent aging program: how does it work?


  • Dale E. Bredesen

    Corresponding author
    1. Buck Institute for Age Research, University of California, San Francisco, 8001 Redwood Blvd, Novato, CA 94949, USA
    Search for more papers by this author

Dale E. Bredesen, Buck Institute for Age Research, University of California, San Francisco, 8001 Redwood Blvd, Novato, CA 94949, USA. Tel.: +1 415 209 2000; fax: +1 415 209 2005; e-mail: dbredesen@buckinstitute.org


Aging and lifespan determination have been viewed, in the most well-accepted theories, as nonprogrammatic, and are thought to result from the evolutionary selection for early fitness at the expense of late survival. Here, recent data implicating potentially programmatic aspects of aging and lifespan determination are discussed, and analogies between programmed cell death and programmed organismal death are offered. It is hoped that the recognition of at least the possibility of a programmatic aspect, or aspects, to the determination of longevity and the process of aging will help to optimize our chances to identify appropriate therapeutic targets both for longevity enhancement and disease prevention.

‘No man becomes a fool until he stops asking questions.’

Charles Steinmetz

The interorganismal variation in lifespan extends over at least several orders of magnitude, and may be indefinite, given the rare cases in which lifespan appears not to display a clear limit. Of the various theories for the limitation of lifespan, perhaps the one with the greatest current experimental support is the theory of antagonistic pleiotropy (Rose & Graves, 1989). This theory is based on the Darwinian theory of evolution, and argues that evolution is associated with the collection of mutations that confers fitness early in life but has deleterious effects late in life (e.g. Kirkwood & Rose, 1991).

It has been argued cogently on theoretical grounds that there is unlikely to be a program for aging, i.e. physiological development is unlikely to include a programmed organismal senescent step (Rose & Graves, 1989). Indeed, it is not at all clear how an organism subjected to the evolutionary selection process would evolve a program for its own demise, given that other organisms lacking this program would presumably be at a survival advantage, or, at the least, no disadvantage. On the other hand, recent data from a number of laboratories support the notion that longevity may be regulated by a genetic program or programs, at least in invertebrate model systems such as Caenorhabditis elegans and Drosophila melanogaster (and possibly in the mouse, as well, although the data from experiments in the mouse are not as extensive, nor as clear cut, as those from the invertebrate systems) (Johnson & Wood, 1982; Lin et al., 1998; Guarente & Kenyon, 2000; Holzenberger et al., 2003). One of the striking findings supporting this notion is that the inhibition or inactivation of specific genes enhances longevity rather markedly, implying that the associated gene products normally have the effect of limiting lifespan (Kenyon et al., 1993). Does this imply the existence of a program that limits longevity, i.e. of an ‘aging program’?

The purpose of the current contribution is to address this question, in part to attempt to reconcile the emerging experimental results with the theoretical predictions, and in part to ask whether such a reconciliation is in fact required. As a starting point, an analogy between the altruistic biological program of cell death (PCD) and the altruistic putative program(s) of organismal death (POD) will be considered. Theoretically, there should be little or no similarity: PCD is well described (over 60 000 manuscripts describing and characterizing this phenomenon have been published), and, although altruistic at the cellular level, it is clearly advantageous at the supracellular (in this case, organismal) level; POD, on the other hand, would be predicted to be altruistic at the organismal level but advantageous at the supraorganismal (presumably, group or societal) level. It has been argued that such supraorganismal selection is uncommon or non-existent, especially in higher organisms; therefore, whereas the theoretical basis for PCD is clear, apparently no such theoretical basis for POD exists (at least given known biological conditions). It is of interest, then, that at least superficially there are numerous similarities between these putative programs.

What are the key aspects of programmed cell death?

Some of the major features of PCD are listed in Table 1. In summary, there exists a set of biochemical programs that effect cellular suicide. These programs play roles in development, cell turnover, neoplasia, neurodegeneration, osteoporosis, atherosclerosis, and other processes. It is thus clear that cell death programs play important roles in age-associated diseases, but the role, if any, of PCD in the aging process itself is unclear. Cell death programs are evolutionarily conserved, genetically controlled, mediated by classical signal transduction pathways and induce phenotypes of relative morphological uniformity. The most well defined of these programs is referred to as apoptosis, from the Greek apo (away from) and ptosis (falling). However, other, clearly distinct forms of PCD exist, such as autophagic PCD and paraptotic PCD (Sperandio et al., 2000).

Table 1.  Comparison of some of the features of programmed cell death and (putative) programmed organismal death
 Cell deathOrganismal death
Genetic control+ (Ced-3, IGFR, etc.)+ (IGFR, etc.)
May be triggered by DNA damage, ROS, stress, etc.++
Endogenous setting with moderate range‘Apostat’ (Bax/Bcl-2, etc.)‘Longevistat’ (hormonal setting?)
Evolutionary conservation++
IGFR/Akt/FKH pathway involved++
Relative morphological uniformity of phenotype++
Multiple pathways++

PCD is dependent on signal transduction pathways that include receptors (e.g. Fas), transcription factors (e.g. p53), proteases (e.g. caspase-3), adaptors (e.g. FADD), and other molecules (e.g. Bcl-2 and related family members). Importantly, these PCD pathways may be activated both by endogenous and by exogenous triggers. Reactive oxygen species may be both inducers and mediators of PCD, but do not appear to be required for PCD. In the case of apoptosis, these pathways converge on the activation of a set of cysteine proteases specific for a subset of aspartic acid residues; these proteases are referred to as caspases (cysteine aspartyl-specific proteases) (Salvesen & Dixit, 1997). Other forms of PCD may be caspase independent, but it is not yet clear whether other proteases, such as cathepsins or calpains, are required to drive these alternative pathways.

The propensity for the cell to activate apoptosis – the apostat (Bredesen, 1996a,b; Salvesen & Dixit, 1997) – may be set over a rather broad range. Hormones such as insulin, and trophic factors such as insulin-like growth factor I (IGF-I), alter the apostat, as do many other signalling molecules. In addition, internal signals such as heat shock proteins and other chaperones, antioxidants and forkhead family transcription factors also participate in the setting of the apostat.

What are the analogies between PCD and any putative program for organismal death? More to the point, do any apparent analogies between PCD and POD offer insight into the aging process? Certainly the immediate concern is that any similarities may turn out to be superficial, and therefore may not lead to any substantive insights into the aging process itself. Nonetheless, a consideration of potential similarities is warranted here. The first and most obvious analogy is that PCD occurs not only in multicellular organisms but also in single-celled organisms, and in the latter case PCD and POD are synonymous (Ning et al., 2002; Bayles, 2003; Burhans et al., 2003; Debrabant et al., 2003; Nystrom, 2003; Rice & Bayles, 2003; Zhang et al., 2003). Although PCD in unicellular organisms is not yet as well characterized as in multicellular organisms, it has become clear that it does indeed occur in multiple unicellular organisms – both prokaryotic and eukaryotic – and that it bears a number of morphological and biochemical similarities to PCD in multicellular organisms, such as protease activation and DNA fragmentation.

A second potential relationship between PCD and a putative organismal death program is the apparent overlapping of gene products that affect the two processes: for example, p53, Sir2, Akt, FOXO family transcription factors, proteins of the IGFR signalling pathway, chaperonins and antioxidant proteins, among others, exert effects on longevity and on PCD. In addition to proteins, hormones (e.g. ecdysone) play important roles in both programmed cell death and lifespan determination (Simon et al., 2003). However, most of these hormones and proteins have multiple effects, and some proteins with major effects on PCD have no apparent effect on longevity [e.g. Bcl-2 (although Bcl-2 has been shown to extend the survival of yeast in stationary phase (Longo et al., 1997), and Bcl-2 null mice do undergo early greying (Veis et al., 1993))], so the overlap between proteins with effects on PCD and those with effects on longevity may turn out to be simply a coincidence. However, if so, the coincidence extends to caloric restriction and to chaperone proteins, both of which enhance organismal and cellular longevity (Mattson et al., 2002; Walker & Lithgow, 2003).

A third analogy between PCD and any putative organismal death program is in the setting of the apostat, i.e. the cell's setting of its own propensity to end its life actively. Recent data have made it very clear that an analogous longevistat exists, i.e. a setting of the likelihood of ending life within a given increment of time, with a remarkably broad range of settings (Arantes-Oliveira et al., 2003). Such a longevistat is not predicted by the theory of antagonistic pleiotropy, and neither is it categorically excluded, although the arguments on which the theory is based offer no insight into why such a longevistat would evolve. The theory of antagonistic pleiotropy argues that aging is an ‘accident’, an undirected by-product of selection for early fitness genes by an evolutionary sieve that opens (i.e. fails) progressively in the post-reproductive period. In contrast, the identification of a co-ordinately controlled longevistat suggests that aging is a finely regulated biological process, much like PCD or oxidative phosphorylation. This is not at all to suggest that external processes – such as time and stresses, among others – do not impinge on or even drive the process. Indeed, the apostat sets the propensity for both endogenous and exogenous processes to trigger PCD; and in an analogous fashion, it is conceivable that the longevistat will turn out to set the propensity for both endogenous and exogenous processes to trigger the putative POD.

Thus both PCD and the putative program for organismal death appear to be adjusted by a complex, internal rheostat that integrates signals from a number of different sources. Both are evolutionarily conserved; respond to DNA damage, reactive oxygen species and other insults; are modulated by p53, IGFR, Akt and associated forkhead family transcription factors; result in morphologically relatively uniform phenotypes; and are likely to be mediated by multiple pathways, such that the inactivation of any one does not prevent the death of the cell/organism.

What are the major objections to the notion of an organismal death program?

One of the common explanations offered to reconcile the notion of antagonistic pleiotropy, as the proposed underlying cause of the aging process, with the identification of genes having major effects on longevity, is to argue that ‘that is not what they are there for’, i.e. the genes with a major effect on longevity do not have as a primary function an effect on longevity (Johnson et al., 2000). The implication is that these genes have as their primary functions effects within previously described signal transduction pathways, such as insulin-related signalling, and do not function primarily as part of a biochemical longevistat. However, what these genes are ‘there for’ may be more of a philosophical issue than a scientific one. Perhaps more to the point, if these genes with major effects on longevity were not‘there for’ controlling longevity – for example, if antagonistic pleiotropy were indeed the cause of the aging process, and mutations in these genes were simply incidental causes of reduced fitness – then the associated gene products would not be expected to be organized into complex, hierarchical, tightly controlled biochemical pathways in which most or all of the members control longevity. Instead, what has been observed is typical of biochemical pathways of (presumably) known function. For example, a recent report from Kenyon and colleagues (Hsu et al., 2003) showed that daf-16, a FOXO-family transcription factor with a major effect on longevity, controls a large number of genes – including cellular stress-response, antimicrobial and metabolic genes, and specific ‘life-shortening’ genes – many of which have smaller effects on longevity, but, in the aggregate, exert a major effect. Thus longevity control by daf-16 looks nothing like a random process – whether the otherwise seemingly disparate genes controlled by daf-16 will turn out to have something in common other than longevity control remains to be determined, but the finding that longevity control is a common thread for a surprisingly diverse set of genes within this pathway is a strong argument against randomness, against longevity control as a non-primary effect of this biochemical pathway, and therefore against the notion that longevity control is not what these genes are ‘there for’. If, as suggested by the theory of antagonistic pleiotropy, aging arose from random mutations enhancing early fitness at the expense of late, one would not expect well-coordinated biochemical pathways controlling longevity, but that is what has been observed. One would expect many unrelated genes with modest effects, rather than coordinated, hierarchical pathways directing longevity.

Another argument against any putative program for organismal death is that Darwinian evolution dictates that organisms do not evolve counter-selective traits. A similar argument would render PCD in unicellular organisms equally unlikely. In both cases, however, the assumption is made that we understand what drives the selection. It has been assumed that the selective pressure is unidirectional – toward greater early life fitness – but what if the selective pressure is toward something more complicated, such as flexibility or control? For example, if modulation of the putative longevistat were not programmed, then a changing environment would result in a slow adjustment of the longevistat over generations, rather than the rather rapid response that has been observed following, for example, caloric restriction or hormesis. It appears that, rather than evolve toward shorter or longer lifespan, organisms have evolved toward having it both ways – focusing resources on fecundity when that is advantageous, and on longevity when that is advantageous. Does this contradict the claim that longevity cannot be selected for?

What are the basic aspects of a putative program for organismal death?

Overall, the data currently available support the notion that organismal death, much like cell death, results from an integrational event, contributions to which are derived from three processes: a non-programmatic component driven both by external and by internal insults; a programmatic component that effects a relatively uniform phenotype, has a set point capable of a relatively wide range of settings and utilizes multiple signal transduction pathways; and the uni- or bi-directional interaction of the programmatic and non-programmatic components. The following points form the underpinnings for such a model:

  • 1Based on recent data from multiple organisms, there exists a biochemical program that affects longevity.
  • 2Because the inhibition of some mediators of this program extends lifespan, and the activation of the program can limit lifespan, this program can bring about programmed organismal death (although this program alone may not necessarily be sufficient to bring about POD).
  • 3Based on the current data, it may be argued that this effect is not a ‘program for organismal death’, but rather a ‘program for inhibiting organismal death’, or a ‘program for modulating the response to internal and external stressors that, in combination, lead to the organismal alterations referred to as “aging”’. However, from a practical point of view, there is an important common implication: the program is subject to experimental, and potentially therapeutic, intervention.
  • 4Why such a program exists is a different question than whether it exists. As noted above, one possibility (but by no means the only one) for why such a program may exist is the potential evolutionary advantage of having the ability to adapt to an inconstant environment, by shifting toward high fecundity or toward delayed reproduction, depending on the conditions.
  • 5The programmatic component displays at least two distinguishable subcomponents: one that sets the net endogenous rate of accumulation of insults (including the generation of reactive oxygen and nitrogen species, resultant DNA damage, other macromolecular damage, protein misfolding, etc.), and another that sets the response to such an accumulation (analogous to the setting of the apostat in various cell types by p53 status and other apoptosis regulators). Thus the former may be thought of as programmatic modulation of the aging process, whereas the latter may be considered to be a programmatic response to the aging process (Fig. 1). An important distinction between these two subcomponents is that the former ratchets longevity and tumorigenesis in opposing directions, i.e. a decrease in insult generation and accumulation increases longevity and decreases tumorigenesis, whereas the latter ratchets longevity and tumorigenesis in similar directions.
  • 6As noted above, interaction between the programmatic and non-programmatic components may turn out to be uni- or bi-directional: for example, setting the longevistat toward greater longevity leads to a set of cellular biochemical events that results in a decrease in the rate of accumulation of misfolded proteins and oxidatively damaged DNA (among other effects), arguing that the programmatic aspects may affect the non-programmatic ones. Whether there is also interaction in the opposite direction, i.e. whether an increase in misfolded proteins or damaged DNA, etc., actually contributes to the triggering of a programmed organismal death, as it clearly does for PCD, is less clear. However, the recent results from studies of p53 (e.g. Tyner et al., 2002) suggest that response to damage, as opposed to the simple accumulation of damage, may indeed be important in lifespan determination (just as it is for PCD), and this in turn supports the notion of an interaction between damage accumulation and the triggering of an organismal death program.
Figure 1.

Lifespan determination: sites of potential programmatic aspects. Note that some genes control multiple aspects of the longevity determination set point (‘longevistat’), through regulation of a set of diverse and otherwise apparently unrelated genes. Note also that the evidence for programmatic effects on lifespan modulation (subcomponent 1) is much more clear than the evidence for programmatic effects on the response (subcomponent 2).


As suggested by the Hegelian dialectic, progress in the form of a new synthesis may result from the identification and description of an apparently contradictory thesis and antithesis. The evolutionary theory of aging has been remarkably successful at explaining many of the findings associated with organismal aging. However, as detailed above, some recent findings are not readily explained by this theory, and support the existence of a programmatic component to lifespan determination, mediated by a hierarchical, tightly controlled biochemical pathway or pathways. It is important to point out, however, that the evolutionary theory of aging does not necessarily exclude the possibility of a programmatic component, and that programmatic and non-programmatic components may collaborate to fashion the aging phenotype. It is hoped that the accommodation of these recent findings and implications by the current theories of aging will result in a more accurate model of lifespan determination and the aging process.