A Review Of Why And How We Age: A Defense Of Multifactorial Aging

By Charles Bernard Olson


Mechanisms of Ageing and Development, 41 (1987) 1-28
Copyright (c) 1987 Elsevier Scientific Publishers Ireland Ltd.
(Received January 16th, 1987; Revision received June 12th, 1987)
Key words: Senescence; Longevity; Evolution; Multifactorial; Pleiotropy; Repair

 Send me email      Return to Home

SUMMARY

Part 1:  Longevity is optimized such that reproduction is maximized. Williams (Evolution, 11 (1957) 398-411) proposed pleiotropic genes with beneficial effects during youth and harmful effects at older ages. Because of environmental death (e.g. predation, disease, accidents), even a small increase in younger reproduction could outweigh a large harmful effect at older ages. Guthrie (Perspect. Biol. Med., 12 (1969) 313-324) and Kirkwood (Nature, 270 (1977) 301-304; New Sci., 81 (1979) 1040-1042; Physiological Ecology: An Evolutionary Approach, Blackwell, Oxford, 1981, pp. 165-189; Hum. Genet., 60 (1982) 101-121; Proc. R. Soc. Lond., B205 (1979) 531-546; Handbook of the Biology of Aging, 2nd Edn., Von Nostrand Reinhold New York, 1985, pp. 27-44) proposed that additional longevity requires a further investment of resources when young, thereby reducing the resources available for reproduction when young. The gene(s) controlling this partitioning of resources between younger and older reproduction are a good example of Williams's pleiotropic genes. Population biology provides a great deal of evidence of a tradeoff between younger and older reproduction. A "marginal longevity theorem" is proposed which states that for a population in equilibrium with its environment a marginal change in any gene affecting longevity should cause equal and opposite marginal changes in younger and older expected reproduction. Senescence is not irrelevant in the wild; rather, the amount of senescence in the wild results from a balance between its marginal costs to older reproduction and its associated marginal benefits in younger reproduction.

Part 2:  The wide variety of damage prevention processes in the body are subject to the problem of diminishing returns. Consequently, a broad spectrum of damage occurs in the body, varying in frequency, harmfulness, and ease of repair. The types of damage which are prevented or repaired tend to be more frequent, harmful, and easily prevented or repaired. In contrast, aging damage (which accumulates) consists of a large number of different types of damage which (when considered separately) are relatively infrequent, less harmful, and/or more difficult to repair. Only when these types of damage accumulate to become very numerous do they (when considered collectively) become significant. Since the selective advantages associated with senescence apply to all parts of the body, primary aging damage occurs in all tissues, cells, and subcellular organelles. The distribution of metabolic resources among the various damage repair and prevention processes is optimized. The phenotype of aging reflects not only the accumulation of aging damage but also regulatory and/or programmed changes (such as menopause) which mitigate the impact of senescence, optimizing the organism's performance given the accumulating damage. The inverse correlation between longevity and specific metabolic rate suggests that a significant portion of aging damage results from metabolic processes.

PART 1: WHY WE AGE

Aging is one of the great paradoxes of nature. As the distinguished evolutionary biologist George Williams [1] has pointed out, "It is indeed remarkable that after a seemingly miraculous feat of morphogenesis a complex metazoan should be unable to perform the much simpler task of merely maintaining what is already formed". (p. 398). All characteristics of our marvelously complex human bodies have been selected for by the process of evolution -- what could be the advantage of the gradual deterioration which we call senescence?

Resolving the paradox depends upon a clear understanding of the evolutionary process -- natural selection, or "survival of the fittest". We think of fitness as health, strength, long life, etc., but fitness in an evolutionary sense has a distinctly different meaning: it is the ability to successfully reproduce [9, pp. 158-161]. Reproductive ability in the natural environment is what is maximized by evolution. All other characteristics, including health, strength and longevity, are optimized such that reproductive ability is maximized.

An increase in health or strength or longevity (i.e. maximum potential lifespan) will be advantageous solely to the extent that it results (on average) in an increase in the reproduction of the organism. If an increase in any of these traits decreases total reproduction, then it is not an advantage.

An important point regarding longevity and reproduction concerns the effect of environmental death [10,11]. Due to the probability of death resulting from starvation, disease, accidents, or predation, the amount of reproduction anticipated for a particular genotype must be discounted: as increasing ages are considered, the probability that an individual organism will survive to reproduce at that age decreases.

But even if the amount of expected reproduction is discounted for the effect of environmental death, wouldn't an increase in longevity still necessarily increase reproduction? After all, an organism that can live longer can reproduce longer. However this reasoning entails an implicit and mistaken assumption: that increasing longevity (which presumably increases reproduction when older) will have no effect upon reproduction when younger.

Williams [1] proposed that aging may result from pleiotropic genes (genes which have more than one effect) which have beneficial effects during youth and detrimental effects at older ages (see also Medawar [10, p. 38; 11, pp. 64-651 and Bidder [12]). This idea of pleiotropic genes having different effects on different ages is a powerful concept and, as we will see below, one which seems capable of explaining why senescence evolved. However, it is also very general and it gives no clue to what a typical pleiotropic gene might deal with or how it might operate. In his paper, Williams gave only one hypothetical example: a mutation which has "a favorable effect on the calcification of bone in the developmental period but which expresses itself in a subsequent somatic environment in the calcification of the connective tissue of arteries" [1, p. 402].

A far more general and convincing example of a pleiotropic gene is seen in the works of Guthrie [2] and Kirkwood [3-8]. They proposed that additional longevity requires an investment of resources when young. Consequently an organism that ages will have more resources available for reproduction when young than an organism that ages less rapidly or not at all. (See also Weismann* [13, p. 141; See Kirkwood and Cremer [6] for an excellent review of August Weismann's contributions to gerontology.], Lack [14, p. 307], Williams [9, pp. 158-192;15] and Calow [16, pp. 453,4551.)

Guthrie [2] gives an example which illustrates the basic idea. He postulates two genotypes which result in the same weight, litter size, developmental time, etc., but with the following important difference: the A genotype reproduces once, and then dies; the B genotype is able to survive to a second reproduction before it dies. "Since B must prepare for continued existence past the first reproduction, it necessarily requires more energy than A in its pre-reproductive development" (p. 315). Guthrie points out that the higher energy requirements of the longerlived B could result in increased pre-reproductive mortality through increased susceptibility to starvation or greater exposure to predation (while searching for the extra food). Also, "A could be expected in time of stress to raise more of her first litter to reproductive maturity than B, even though they both have the same number of offspring per litter, because B phenotypes will have had to channel some of their energy customarily spent on reproduction into preparations for post-reproductive survival and another reproductive period" (pp. 315-316). For an organism to evolve increased longevity without requiring an increase in the "energy" needed, it would have to "reallocate reproductive energy, which would either postpone maturity, decrease litter size, increase offspring mortality, increase the time between litters, or some combination of these" (p. 316).

Similarly, Kirkwood points out that a "disposable soma" (i.e. a body that ages) may be "cheaper" to produce than an eternally youthful one. "If it costs more to build a soma that potentially lasts forever than one that senesces at an age when few, if any, individuals can expect to remain alive, the latter strategy must be preferable.... The fundamental answer to the perennial question of why we grow old may simply be that to do otherwise, to stay forever young, would require us to invest resources in somatic maintenance, that, from our genes' selfish point of view, are better spent on reproduction" [4, p. 1042].

Thus these authors propose that there is a direct tradeoff between younger and older reproduction because an increase in longevity requires an increased investment of resources when young. Clearly any gene or genes which control this partitioning of resources between reproduction at different ages would be an excellent example of what William's pleiotropy theory proposed.

Experimental and theoretical support for this idea of a tradeoff between younger and older reproduction comes from the field of population biology. Cole [17] introduced the idea that characteristics such as age of maturation, longevity, broods per season, and size of brood, evolve together as an ensemble of "lifehistory tactics". Increased survivorship in the wild (which should correlate with increased older reproduction) has been found to correlate with slower growth, later maturation, and a lower level of reproductive effort (the proportion of resources devoted to current reproduction), all of which should decrease reproduction when younger [18-25; 26, pp. 114,136; 27, pp. 164-166]. For theoretical analyses of life-history tactics, see Refs. 9,15,17,28,29; for a review, see Stearns [30). Organisms can be classified as either r-selected (early maturity, rapid achievement of peak reproduction, high reproductive effort, short lifespan) or K-selected (late maturity, low rise to peak reproduction, low reproductive effort, long lifespan) (for a review, see Pianka [31]).

The idea of a tradeoff between younger and older reproduction can be expressed in a graphical form. Figure 1 shows the "reproductive velocities" (reproduction* per unit time) for two genotypes over the course of the lifespan. [Reproduction does not correspond exactly to the number of healthy fertile offspring. For example, a bird which raises four offspring is not necessarily more successful than one which raises three: due to a higher level of parental care and feeding, the three could have greater total reproductive success than their four somewhat less vigorous competitors. Reproduction corresponds to the extent of the organism's contribution to the genetic pool in the long-term future.]

2K: Reproductive velocity vs Age, in the absence of environmental death

Fig. 1. Reproductive velocities in the absence of environmental death. Curves S' and L' indicate the average or expected reproduction per unit time for two different genotypes (i.e. potential organisms characterized by their genetic make-up). The probability of environmental death has not been taken into account, and therefore the curves represent the average reproductive velocities at various ages assuming survival to the ages in question. See Fig. 2 for the curves with environmental death taken into account. Figure 5 shows both sets of curves on the same graph. Note that the shorter-lived genotype (curve S') has an earlier age of maturation and a higher peak reproductive velocity than the longer-lived genotype (curve L). Organisms which die at different ages (e.g. X, Y and Z) will be subject to different selective pressures. For example, organisms which die at age X will experience selection for a shorter lifespan. This is because those with the S genotype will have achieved greater reproduction than those with the L genotype (i.e. the area under curve S' up to age X is greater than the area under curve L' up to age X). Conversely, organisms which die at age Z will experience selection for a longer lifespan. There is a "break-even age" (age 1) at which the two genotypes will expect equal reproduction. See the text for a discussion of the balance of the opposing selective forces acting on organisms dying before versus after the break-even age, and the effect of changes to the survival curve upon that balance.

This is not a comparison of two specific organisms (whose lives would be affected greatly by chance), but rather of two potential organisms (characterized by their genetic make-up) averaged over all of the possible hazards and opportunities of their environment. The figure does not however incorporate the probability of environmental death. Rather, it indicates the average reproductive performance at various ages assuming survival to the ages in question.

The curves in Fig. 1 illustrate the concept that longevity has a cost in terms of younger reproduction. The two genotypes differ in their age of maturation, their peak reproductive velocity, and in their longevity. The longer-lived genotype (curve LI must divert resources away from reproduction while younger, thereby delaying maturation and/or lowering its expected rate of reproduction.

A comparison of the total expected reproduction of the two genotypes in a natural environment requires that environmental death be taken into account, as shown in Fig. 2. The curves in Fig. 2 correspond to the curves in Fig. I multiplied by their respective survival curves (Fig. 5 shows both sets of curves in a single graph).


2K: Reproductive velocity vs Age, with environmental death

Fig. 2. Reproductive velocities with environmental death taken into account. Curves S and L indicate the expected reproductive velocities of the two genotypes (shorter-lived and longer-lived) of Fig. 1, incorporating the discounting effect of environmental death. For example, if shorter-lived organisms have a 25% chance of surviving to a particular age, then their individual reproductive velocity at that age would be on average four times of that indicated on the graph (i.e. the corresponding curve in Fig. I would be four times higher at that point). Thus, the curves in Fig. 2 (S and L) correspond to the curves of Fig. I (S' and L) multiplied by their respective genotypes' survival curves (i.e. their cumulative probability of death in their natural environment). Figure 5 shows the four curves on the same graph. Total expected reproduction for each genotype corresponds to the area beneath the respective curves. Therefore a comparison of the sizes of the striped and dotted areas will reveal which genotype has the greater reproduction. For the purposes of the discussion, it is assumed that the two areas are equal, and therefore that the two genotypes are equally advantageous (at the given level of environmental death).


The total area beneath each curve in Fig. 2 corresponds to the total expected reproduction for the respective genotype. Therefore a comparison of the striped and the dotted areas will show which of the two genotypes will have a greater total expected reproduction. The genotype with the greater reproduction will increase its representation in the gene pool relative to the other genotype. [This assumes that the population is not changing in size, i.e. "r" (the malthusian parameter) equals zero. If r is not zero, then reproduction at different ages must be weighted accordingly. For example, in a growing population, early reproduction is worth relatively more than later reproduction due to "compounding" of reproduction. The assumption that r = 0 merely simplifies the argument, and does not change the outcome. Indeed, Hamilton [32, pp. 23-27] has shown that even in the absence of death some degree of senescence would be advantageous, i.e. mutations which sacrificed older reproduction in favor of younger reproduction would eventually achieve numerical predominance in such a population.]

The rate of environmental death, which is reflected in the "steepness" of the survival curve, has a dramatic effect upon the relative reproductive success of genotypes with differing longevities. Figure 3 shows two survival curves: curve A with 50% survival per year, and curve B with 25% survival per year. An exponential curve reflects a constant probability of death [10,11], and closely approximates adult mortality, at least until relatively advanced ages [33], for a number of species in the wild [18] (although there are some apparent exceptions [34], however see also Ref. 18 p. 173). Note that a change in the "steepness" of the survival curve does not affect different ages equally, but rather affects older ages to a greater extent. For example, I year olds are half as numerous in curve B (relative to curve A), while 3 year olds are one eight as numerous.

Let us assume that the striped and dotted areas in Fig. 2 are equal in size. This would mean that the two genotypes (shorter- and longer-lived) would be equally advantageous. If the survival curve were to become steeper, both early and late reproduction (the striped and dotted areas) would decrease, but late reproduction (the dotted area) would decrease by a larger (compounded) factor. Consequently, the shorter-lived genotype would be selected for. Conversely, a shallower survival curve would increase both early and late reproduction, but late reproduction would increase to a greater extent, thereby favoring the longer-lived genotype. Thus, changes in the environmental death rate should produce changes in the relative advantages of genotypes of differing longevity, with harsher conditions resulting in shorter-lived organisms and less severe conditions leading to longer-lived organisms.

2K: Survival curves of differing steepness

Fig. 3. Survival curves of differing steepness. Curve A reflects 50% survival per year; curve B reflects 25% survival per year. At age 1, there are twice as many "A" survivors as "B" survivors. At age 2, the ratio has increased to 4. At increasing ages, the ratio increases further. Thus changes in the steepness of an exponential survival curve affect older survival (and reproduction) disproportionately. Note that the area beneath an exponentially declining curve is finite. For example, the area under curve B from age 2 to infinity is only 1/15 of the obviously finite area from age 0 to 2. This finiteness is relevant: it means that the infinitely long potential reproductive period of an eternally young organism may nonetheless result in a small amount of expected reproduction once environmental death is taken into account (see curve E in Fig. 5). The above survival curves do not take into account a higher level of infant and juvenile mortality.


There is a second way of viewing the balance of selective pressures upon longevity. Note that individual organisms which die accidental deaths at different ages will be subject to different selective pressures on longevity. For example, among individuals which die at age X (see Fig. 1), those with the shorter-lived genotype will have reproduced more (the area under curve S' up to age X is greater than the area under curve L' up to age X). In contrast, among those individuals which die at age Z, the longer-lived have reproduced more (compare the areas under curves S' and L' up to age Z). There is some "break-even age" (age Y) at which the two genotypes will expect equal reproduction. Thus, individuals which die before age Y will experience selection for a shorter lifespan, while those dying after age Y will experience selection for greater longevity. If, as was suggested regarding Fig. 2, the two genotypes have equal expected reproduction, then the selective pressures described above must be in balance.

A shift in the environmental death rate will change the proportions of individuals dying before and after the break-even age, and thereby will result in selective pressure to change longevity. For example, increased environmental death would increase the numbers dying before age X (increasing selective pressure for a shorter lifespan) and would decrease the numbers dying after age Z (decreasing selective pressure for a longer lifespan). As a shorter lifespan evolved, a shift in the break-even age to younger ages would reduce and eventually eliminate the imbalance between the opposing selective pressures, thereby resulting in a new optimal longevity.

When a population is in equilibrium with its environment (with respect to longevity), then the average longevity of the population should be at an optimal value. Figure 4 illustrates this concept. Significant decreases in longevity should decrease total reproduction due to an underinvestment in longevity. (Note that this is "investment" in a broader sense than merely the allocation of energy or nutrients; it also includes any sacrifice in performance through which longevity may be increased. See the Discussion.) A shorter-lived organism would be gaining an amount of reproduction when younger, but it would be sacrificing a larger amount of reproduction when older (due to increased senescence). Conversely, significant increases in longevity should also decrease total net reproduction due to an overinvestment of resources in longevity -- a longer-lived organism would be sacrificing an amount of reproduction when young which is larger than that gained by the chance of reproduction when older. Due to environmental death, the proportion of organisms which survive to reproduce at the greater ages would be too small to compensate for the loss of reproduction by the far greater proportion of organisms which are alive at younger ages. Thus the aging rate of a population is optimized with respect to its environment. [It is interesting to think of the physiological reaction to stress as a temporary increase in the aging rate in response to a more hazardous environment; i.e. an organism's aging rate may be adjustable so that it can be more closely optimized with respect to changing environmental conditions.]

2K: Reproduction vs Longevity

Fig. 4. Longevity is optimized such that reproduction is maximized. Lower than optimal investment in longevity results in lower reproduction because the loss of reproduction when older (due to senescence) is greater than the increase in reproduction when younger (achieved with the additional resources). Conversely, higher than optimal investment in longevity lowers reproduction because the loss of reproduction when younger (due to the diversion of resources for greater longevity) is greater than the corresponding increase in reproduction when older. At the optimum, marginal changes in longevity do not change total reproduction (see tangent). MR = maximum reproduction; OL = optimal longevity.


Marginal changes in longevity at the optimum should result in no change in total net reproduction (see the tangent in Fig. 4). This means that any such marginal shift in longevity must involve equal and opposite changes in younger and older reproduction. This "marginal longevity theorem" may be stated as follows:

When a population* is in equilibrium with its environment, the following condition should hold: Any marginal change in any heritable trait which increases expected reproduction at older ages will cause an equal and opposite change in expected reproduction at younger ages.

*Since the individual organisms within a population will show variation in their many traits, the marginal longevity theorem necessarily deals with the averages.

It is easy to see why this must be the case. If the changes in older and younger reproduction were not equal and opposite, then there would be a resulting change in total reproduction. Consequently a marginal change in the heritable trait in the appropriate direction (such that total reproduction increases) would be advantageous and would be selected for. This contradicts the assumption that the organism is in equilibrium with its environment.

As described above, the relationship between younger and older reproduction is sensitive to changes in the environmental death rate. The marginal longevity theorem assumes a particular environmental death rate. If the survival curve were to change, then the equality between marginal changes in younger and older reproduction would no longer hold (as described above), and a new equilibrium (and a new optimal lifespan) would have to be reached.

A common misconception is that senescence is irrelevant in the wild (e.g. Comfort [26, pp.44-48]). This idea is supported by the apparent absence of senile organisms in natural environments. However, as Williams point out, it is a mistake to equate senescence with senility -- "No one would consider a man in his thirties senile, yet, according to athletic records and life tables, senescence is rampant during this decade" [1, p. 399]. Indeed some aspects of senescence begin before puberty, e.g. fatty streaks in the aorta [35,36].

Rather than being irrelevant, the amount of senescence in the wild is of central importance in determining the optimal longevity of a species. Consider what a total absence of senescence in the wild would imply in the light of the tradeoff between longevity and youthful reproduction. All organisms would be dying totally senescence-free, and therefore each organism would have sacrificed some reproduction when younger for the sake of a useless longevity. Figure 5 shows graphically the negative impact of "eternal youth" upon expected reproduction.

Genetic changes which increase younger reproduction at the expense of eternal youth (i.e. resulting in senescence) would be selected for. Senescence would increase not until it begins to have an impact in the wild (decreasing late reproduction), but rather until the marginal cost of an increase in senescence (i.e. a decrease in older expected reproduction) equals the corresponding marginal benefit (i.e. an increase in younger expected reproduction). This is in fact a restatement of the marginal longevity theorem described above.

Thus, the amount of senescence in the wild results from a precise balance between its marginal costs to expected reproduction when older and its associated marginal benefits to expected reproduction when younger.

3K: Reproductive velocity (dR/dt) vs Age: The maladaptiveness of eternal youth

Fig. 5. The maladaptiveness of eternal youth. Curves S', L', and E' show the expected reproductive velocities for three genotypes (shorter-lived, longer-lived and eternally young) in the absence of environmental death. Note the tradeoff between longevity and younger reproduction (age of maturation, peak reproductive velocity). Curves S, L and E take the cumulative probability of environmental death into account. The area under curve E (which corresponds to the total expected reproduction of the E genotype) is much smaller than that under curves S or L, despite the infinitely long curve. The area under an exponentially declining curve is finite -- see the legend of Fig. 3. See Medawar [11, pp. 64-65] and Hamilton [32, pp. 23-27] for other arguments for the maladaptiveness of eternal youth.

SUMMARY (PART 1)

Longevity is optimized such that reproduction is maximized. Williams [1] proposed pleiotropic genes with beneficial effects during youth and harmful effects at older ages. Because of environmental death (e.g. predation, disease, accidents), even a small increase in younger reproduction could outweigh a large harmful effect at older ages. Guthrie [2] and Kirkwood [3-8] proposed that additional longevity requires a further investment of resources when young, thereby reducing the resources available for reproduction when young. The gene(s) controlling this partitioning of resources between younger and older reproduction are a good example of Williams's pleiotropic genes. Population biology provides a great deal of evidence of a tradeoff between younger and older reproduction. A "marginal longevity theorem" is proposed which states that for a population in equilibrium with its environment a marginal change in any gene affecting longevity should cause equal and opposite marginal changes in younger and older expected reproduction. Senescence is not irrelevant in the wild; rather, the amount of senescence in the wild results from a balance between its marginal costs to older reproduction and its associated marginal benefits in younger reproduction.

PART 2: HOW WE AGE: A DEFENSE OF MULTIFACTORIAL AGING

Aging appears to be an accumulation of damage. For example, at the cellular level, DNA accumulates strand breaks and a variety of other damage [37, p. 148; 38], proteins accumulate post-translational modifications [39-41], mitochondria lose DNA [42,43], and their inner membrane accumulates peroxidized lipids [44, p. 108]. At the organismal level, damage accumulates in each of the many bodily tissues [45].

Aging damage accumulates despite the human body's great variety of processes for repairing or preventing damage, such as:

Attempts to understand aging are complicated by the question of whether any particular instance of aging damage is caused by prior aging damage; in other words, which damage is primary and which is secondary. Primary aging damage occurs even when all of the damage prevention and repair processes in the body are fully functional. For example, humans age between 20 and 30 years of age despite the proper functioning of the body during that time. (If this were not the case, then the 30-year-old would be equivalent to the 20-year-old and would not subsequently age.)

Secondary aging damage is damage which occurs as a result of previous aging damage (which may be primary and/or secondary). An accumulation of primary aging damage would be expected to cause an increasing amount of secondary aging damage, and the accumulation of total damage would accelerate resulting inevitably in death. Despite the importance of secondary aging damage, this paper will not deal with it further, but will instead focus upon the more fundamental issue of primary aging damage -- what it consists of and where and why it occurs.

A major thesis of this paper is that the reduction of damage accumulation in the body is subject to diminishing returns. That is, to borrow the terminology of economics, each additional unit of metabolic currency (e.g. energy, space, time and nutrients, usually in the form of enzymes) invested in the reduction of damage accumulation yields a progressively smaller return. Each additional enzyme will reduce damage accumulation by a smaller amount.

For example, within cells, a certain amount of hydrogen peroxide is generated per unit time as a side product of essential metabolic reactions [53]. As a consequence of a particular concentration of catalase (the enzyme which breaks down hydrogen peroxide) in the cell, there is a resulting steady-state concentration of hydrogen peroxide. Increasing the concentration of catalase will further reduce the steady-state concentration; however, due to the fact that each additional enzyme is competing with a larger number of others, the reduction will become progressively smaller. This problem of diminishing returns applies to all types of damage prevention of this sort (i.e. the reduction of the steady-state concentration of any harmful substance produced in the cell).

Diminishing returns applies as well to damage repair. A steady-state concentration of damage will exist, determined by the rate of damage production and the level of repair processes. Additional repair enzymes will lower that concentration by diminishing amounts.

Although types of damage which are repaired do not directly contribute to aging (i.e. they do not accumulate), they can contribute to aging indirectly: many if not most types of damage can cause further damage. For example, DNA loses bases spontaneously due to hydrolysis [46, pp. 18-20]. DNA repair enzymes rapidly recognize and repair the resulting apurinic sites, leading to a low steadystate concentration. A second instance of damage on the opposite DNA strand (which by itself would be repairable) adjacent to or near an existing apurinic site could result in a mutation. Thus, although the apurinic sites are constantly repaired and do not accumulate, the mutations which they help create do accumulate.

Furthermore, repair processes can indirectly favor damage production. For example, excision repair of damaged DNA exposes single-stranded DNA, which has a far greater susceptibility to base hydrolysis and cytosine deamination than double-stranded DNA [60, pp. 147,174]. This exacerbates the problem of diminishing returns.

There are many sources of damage in our bodies. Some harmful substances, such as free radicals [61] and hydrogen peroxide [53], are generated by our metabolism. Some essential nutrients such as oxygen and glucose have significant harmful effects. For example, glucose has a aldehyde group which reacts with primary amine groups (e.g. on proteins or single-stranded DNA [62]) throughout the body [63). This complex reaction, called nonenzymatic browning or the Mallard reaction, occurs at a high rate in diabetics and at a lower rate in normal humans [64]. Even water causes damage through hydrolysis. Other sources of damage include amino acid racemization [41], radiation [65], and errors in macromolecular biosynthesis [66]. As one considers less and less frequent damage, the list of damage sources becomes progressively longer. For example, normal constituents of the cell (e.g. amino acids), which are relatively inert over short periods (hours, days), are reactive when viewed over longer periods of time (months, years).

The above list of sources of damage gives a misleading picture of the complexity of damage in the body -- it makes matters appear far too simple. Each source of damage listed above does not generate only one type of damage; rather the precise type of damage created depends upon not only the damage source but also the molecules which are being damaged. For example, hydrolysis can cleave membrane phospholipids, split bases from DNA and deaminate cytosine and adenosine [46, pp. 9-20], and break proteins into fragments and deaminate asparagine and glutamine [67]. From the perspective of the problem of how to prevent or detect and repair the damage, it would be misleading to call these events a single "type" of damage. This is true even of different hydrolytic events in proteins, for their consequences will depend upon not only upon the protein cleaved but also upon the particular fragments created. Some fragments might be easily recognized by selective protein degradation systems within the cell, while others (due to their particular amino acid sequence and conformation) might resemble normal cellular constituents sufficiently well to avoid degradation for long periods. Other classes of damage, such as free radical damage, in which a large number of highly reactive molecules can react with any molecule in the body in a variety of ways, can comprise a practically limitless number of types of damage.

In short, damage occurring in the body is phenomenally complex. It varies widely in degree of harmfulness, frequency of occurrence, and the ease with which it can be prevented or repaired. For example, two instances of aging damage which differ enormously in harmfulness are (1) a carcinogenic mutation, and (2) an alteration of a cytosolic enzyme which inactivates it but in such a way that the enzyme will not be broken down (i.e. resulting in a permanent piece of intracellular "garbage"). Frequency varies from the most common reactions downward without lower limit. For example, a particular type of damage might occur on average less than once per cell over an organism's lifetime, and thus only a fraction of cells would suffer this type of damage. Ease of prevention or repair varies, for example, according to the number of additional genes which must be maintained and the number of enzymes which must be produced. (The provision of enzymes for damage prevention or repair is not the only means of reducing damage production in the body. To be more general, "ease of prevention or repair" should include any degree of performance which can be sacrificed to bring about lower damage production. Note that if some aspect of damage production can be lowered without sacrificing performance (which includes providing additional enzymes) then such a modification would be immediately selected for -- such changes are presumably already incorporated in our designs.)

Thus there is an enormous contrast between the simplicity of the task of preventing damage from a particular source (e.g. the inactivation of a damageproducing agent) and the complexity of the task of repairing the damage which occurs despite the existence of damage prevention processes. Clearly it would be preferable to prevent all of the damage before it occurs. But the only way to keep molecules from reacting is to keep them apart. The phenomenal degree of compartmentalization within the body (e.g. cellular, mitochondrial, and nuclear membranes, the blood/brain barrier) reflects this strategy, but it of course has limitations: barriers are not perfect, and they have an associated cost in terms of space and metabolic resources. Thus, because this and other damage prevention strategies are subject to diminishing returns (as discussed above), the rate of damage production can be lowered but not eliminated. Consequently, the cell (and the body) is faced with the problem of repairing a broad spectrum of types of damage, which vary enormously in relative harmfulness, frequency, and ease of repair.

Certainly, in the process of evolving mechanisms to deal with this damage, one would expect the types of damage which are more frequent, harmful, and easily repaired to be dealt with first. For those types of damage, the benefits of repair will be large relative to the costs of repair. As one considers less and less frequent, harmful, and/or easily repaired types of damage, the potential benefits of additional repair processes decline in relation to the costs. Thus diminishing returns also applies to the evolution of additional repair processes to deal with increasingly insignificant, infrequent, and/or difficult-to-repair types of damage.

Turnover (periodic complete replacement) is a unique repair process because it deals with all types of damage, harmful and trivial, frequent and infrequent, easy and difficult to repair. Turnover is therefore an important exception to the problem of diminishing returns*, and, if an organism were constructed such that turnover occurred throughout its body, then it could be able to avoid the problem of senescence. There are apparent examples of an absence of senescence in both the plant and animal kingdoms. Some plants propagate vegetatively (i.e. not using sexual reproduction) indefinitely [70, pp. 10-11]. Sonneborn [71] experimented with a species of flatworm which periodically divides into anterior and posterior segments. He showed that lineages of posterior segments apparently do not senesce whereas lineages of the anterior segments do -- perhaps due to structures in the head, such as the central nervous system, which do not turn over.

*Since stem cells are only a small proportion of the cells of the tissue, higher expenditures on repair will be a proportionally smaller cost to the organism. Also, the metabolism of stem cells could be simplified and kept at a low level to reduce damage production. For example, mouse spermatogonia have "empty" mitochondria [68, p. 253] which suggests that they are not functioning, thereby avoiding the metabolic production of the highly reactive oxygen metabolites. Furthermore, selection could act upon the stem cells to keep them free of damage, perhaps in conjunction with a mechanism for asymmetrical distribution of damage between daughter cells [69, p. 382].

Unfortunately for humans, natural selection has resulted in a design for our bodies which includes many parts in which complete turnover does not occur, e.g. the central nervous system, kidneys, lungs, cardiac muscle, skeletal muscle, the lens of the eye, collagen, elastin, basement membranes. Consequently, due to the problems of diminishing returns discussed above, damage accumulation can only be reduced but not eliminated, resulting in a compromise where more frequent, harmful, and easily repaired damage is repaired or lowered in frequency but the less frequent, less harmful, and less easily repaired damage accumulates (i.e. senescence).

The tradeoff described in Part I between younger and older reproduction applies not only to the organism as a whole but also to each organ and tissue within it. If enhancement of the performance of the organ when young (and hence enhancement of the performance and reproduction of the young organism) can be obtained at the expense of a gradual accumulation of damage, which lowers performance (and hence reproduction) of the organism when old, then such (primary) senescence will be selected for up to the optimum described by the marginal longevity theorem (see Part 1), i.e. the point where a marginal increase in the rate of senescence produces a decrease in older reproduction which is equal to the corresponding increase in younger reproduction.

The possible savings (in reproduction when young) associated with senescence apply to any of the organs or tissues in the body, including tissues with organizations compatible with complete turnover. Since the organization of tissues with stem cells (mitotic tissues) allows relatively inexpensive reduction of senescent damage accumulation, they would be expected to evolve in the direction of a lower rate of senescence to a considerable extent, and hence mitotic tissues should show relatively less senescence than non-mitotic ones; however, given a finite lifespan due to the non-mitotic tissues in the body, the reduction of senescence in the mitotic tissues would be expected to stop short of complete elimination as the benefits of reducing senescence become negligible.

The above argument also applies at the subcellular level. The possible benefits of senescence (in terms of younger performance) apply to all systems and components within the cell, and consequently each should suffer primary aging damage. Just as it would not make sense for natural selection to eliminate all damage except that resulting from a single organ system, it would not make sense to eliminate cellular senescence except that resulting from a single subcellular organelle or component.

Likewise, it would not make sense for natural selection to eliminate all damage other than that due to one particular molecular reaction. Instead, the degree to which damage resulting from a particular organ system, subcellular organelle, or molecular process contributes to aging will depend upon the "economics" of its production, prevention and repair. The problem of diminishing returns ensures that the optimum achieved by natural selection will involve a finite rate of damage accumulation in each of the many parts of the body and the cell as a result of each of the many sources of damage.

Thus it is suggested that primary aging damage occurs independently in each tissue, cell and subcellular organelle. However, aging itself does not occur independently in each part of the body due to the occurrence of secondary aging damage. For example, immune system dysfunction can cause damage throughout the body via auto-immunity, DNA damage can cause damage throughout the cell via absent or altered gene products, etc.

One would expect the distribution of metabolic resources (e.g. energy, space, time and nutrients, usually in the form of enzymes) between the different repair or prevention processes to be optimized such that total expected reproduction is maximized. In other words, any marginal shift of resources between different damage repair or prevention processes should result in no net change in total reproduction. In view of the marginal longevity theorem described in Part 1, this makes sense: a marginal shift of resources between different repair and/or prevention processes consists of a marginal increase in expenditure for one and a marginal decrease for the other. According to the marginal longevity theorem, each of these marginal changes (considered separately) result in equal and opposite changes in younger and older reproduction. Consequently, their sum cannot change total reproduction. If this were not true, then a simple shift of resources from one repair process to another could increase total reproduction presumably such a shift would already be selected for.

To a considerable extent, the progression of senescence in young adults is hidden from view by the enormous surplus capacity of the mechanisms within our bodies. No change in the organism's performance is seen as a result of a small reduction in that surplus capacity except under the most strenuous of conditions (e.g. athletic contests). As damage accumulates, this surplus capacity can be brought into operation. For example, individual mitochondria may function more actively to compensate for a reduction in their numbers [72]. A similar effect has been seen among synapses [73], where a reduction in their numbers is partially compensated for by an increase in their individual size.

Furthermore, it would be naive to expect that the phenotype of aging results solely from an accumulation of damage. One would expect the cell and the body to monitor levels of damage and damage-producing agents [58, p. 328; See Cutler [74, pp. 100-108; 75, pp. 413-423] and Sohal et al. [76] for evidence of compensative regulation of tissue antioxidant levels], and to respond in ways which mitigate the impact of the accumulating damage, optimizing the performance of the cell and the organism given the problems resulting from senescence. Thus it is plausible that some of the physiological changes which take place during aging are not direct consequences of aging damage, but rather are regulatory and/or programmed changes.

The most dramatic example of this is menopause, which is thought to be a programmed cessation of ovulation which allows the older woman to focus her efforts on helping her existing children and grandchildren rather than undergo additional risky pregnancies [1, pp. 407-408]. Further potential examples of programmed changes include other hormonal changes [45, pp. 289-408], shifts in body composition such as the increased proportion of fat with age [45, pp. 417-421], changes in the regional distribution of fat in the body, changes in sleep [45, pp. 33-34], and a decline in basal metabolic rate with age [77] (at least part of this decline is due to the increased proportion of fat [78], and it is questionable whether any decline in the metabolic rate of individual tissues occurs at all [45, pp. 411-439]).

For example, consider the changes which occur during human aging in the body's proportion of fat and its regional distribution. While it is difficult to attribute these changes to an accumulation of damage, it is easier to see them as programmed changes which make the body better suited to an increasingly sedentary lifestyle.

Some cases of cellular and protein "aging" may in fact be examples of "programmed obsolescence" [79-80] whose purpose is to mark the cell or protein for selective degradation. A "senescent cell antigen" appears on the surface of senescent and damaged red blood cells which when recognized by an antibody in serum initiates their removal from the circulation by macrophages [81]. Similarly, deamination of asparagine residues may be a "programmed" first step in the degradation of aging triosephosphate isomerase [82]. A need for more rapid turnover in the presence of a higher specific metabolic rate may explain the inverse correlation observed between specific metabolic rate and the thermodynamic stability of myoglobin [83].

Likewise, the cessation of proliferation in in vitro aging is apparently not due to the inability of the cells to divide. Rather, it appears to be caused by a regulatory protein which blocks DNA synthesis [84], perhaps as a regulatory response to accumulating damage. The continued viability of in vitro aged fibroblasts is shown by their ability to survive for long periods (e.g. 2 years) in the post-mitotic state [85, p. 90; 86, pp. 706-707] and the fact that viral yields of post-mitotic fibroblasts are not reduced in quantity or quality [87-89]. Similarly, Daniel et al. [90] have shown a reinitiation of growth in senescent serially transplanted mouse mammary gland epithelium in response to cholera toxin, which elevates intracellular cyclic-AMP levels. Note that an organism would be expected to benefit if cells with more than a certain level of damage refrain from dividing: not only would the resulting population of cells be less damaged, but also the probability of carcinogenesis would be reduced

One additional fact about mammals and their longevity merits discussion: the well known inverse correlation between maximum lifespan and specific metabolic rate [91,92]. Additional evidence includes experiments in which maximum lifespan has been extended by calorie-restriction [93,94], hypothyroidism induced by neonatal thyroxine injection [95], and reduction of ambient temperature in poikilotherms [96]. However, the extension in lifespan is not always in proportion to the reduction in caloric consumption [97]. And Sohal [96] has criticized the assumption that ambient temperature changes cause parallel changes in metabolic rate.

In a review of the rate of living theory, Sohal [96] redefined the "rate of living" theory as follows: "the rate of aging is directly related to the rate of unrepaired molecular damage by the by-products of oxygen metabolism, and it is inversely related to the efficiency of antioxidant and reparative mechanisms" (p. 41). However, as discussed above, there is no reason to believe that the byproducts of oxygen metabolism are the sole cause of primary aging damage in the body. First of all, metabolism comprises far more than just oxygen metabolism, e.g. glucose metabolism, purine and pyrimidine metabolism, amino acid metabolism, fatty acid metabolism, etc. Each of these different areas of metabolism generate a variety of reactive intermediates which are potential sources of damage. Here are a few examples of aldehydes which are intermediates in some major metabolic pathways: D-glyceraldehyde-3-phosphate in glycolysis and the pentose phosphate pathway, D-erythrose-4-phosphate and D-ribose-5-phosphate in the pentose phosphate pathway, 5'-phosphoribosyl-N-formylglycineamide in purine biosynthesis, etc. [98]. Secondly, not all damage sources are directly linked to metabolism (e.g. radiation, racemization, glucosemediated crosslinking [Glucose-mediated crosslinking would depend upon the concentration of glucose, which does not necessarily correlate with its rate of consumption.], hydrolysis). Nonetheless, the correlation does suggest that a significant proportion of aging damage is linked to metabolism (in the broader sense).

SUMMARY (PART 2)

The wide variety of damage prevention processes in the body are subject to the problem of diminishing returns. Consequently, a broad spectrum of damage occurs in the body, varying in frequency, harmfulness and ease of repair. The types of damage which are prevented or repaired tend to be more frequent, harmful, and easily prevented or repaired. In contrast, aging damage (which accumulates) consists of a large number of different types of damage which (when considered separately) are relatively infrequent, less harmful, and/or more difficult to repair. Only when these types of damage accumulate to become very numerous do they (when considered collectively) become significant. Since the selective advantages associated with senescence apply to all parts of the body, primary aging damage occurs in all tissues, cells and subcellular organelles. The distribution of metabolic resources among the various damage repair and prevention processes is optimized. The phenotype of aging reflects not only the accumulation of aging damage but also regulatory and/or programmed changes (such as menopause) which mitigate the impact of senescence, optimizing the organism's performance given the accumulating damage. The inverse correlation between longevity and specific metabolic rate suggests that a significant portion of aging damage results from metabolic processes.

DISCUSSION

Williams [1] proposed that there is a tradeoff between younger and older reproduction, and hence younger and older vigor. Evidence from population biology supports this hypothesis, showing a correlation between increased survivorship (hence increased older reproduction) and slower growth, later maturation and decreased reproductive effort (hence decreased younger reproduction) [18-25; 26, pp. 114,136; 27, pp. 164-166]. This evidence generally involves comparisons between different species and/or differing environments; however, Snell and King [20] compared individual rotifers under laboratory conditions. Data comparing reproductive output under standardized conditions throughout lifespan for different individuals of the same species may already exist for some species of birds and mammals (e.g. production of eggs or milk by laying hens or dairy cows throughout life).

The most important feature of William's pleiotropy theory is that it shows how senescence (due to its pleiotropic linkage to youthful vigor) can be a selective advantage for an individual organism. In contrast, the popular view that aging is a positive adaptation which removes older individuals from a population and ensures a more rapid turnover in generations [13, pp. 22-24; 99, pp. 590-601] depends upon group selection (i.e. senescence would be advantageous for the population or species but not for the individual). For an excellent critique of group selection in general, see Lack [14, pp. 299-312]. Also, Williams's explanation of aging does not maintain that senescence occurs so late in life that it is not subject to significant selective pressure (Medawar [10,11]); see Part I for a discussion of the relevance of senescence in the wild.

Guthrie [2] and Kirkwood [3-8] have proposed that additional longevity requires increased investment of resources in somatic maintenance when young, thereby reducing resources available for reproduction when young. The gene(s) controlling the partitioning of resources between somatic maintenance and reproduction when young provide an excellent example of the pleiotropy which Williams [1] proposed, with opposite effects upon reproduction at younger and older ages.

Guthrie [2] suggested that the advantage of senescence was a savings of "energy". Kirkwood (e.g. [3,7]) initially suggested that the energy savings allowed by reduced longevity were due to a reduced accuracy of macromolecular biosynthesis (e.g. protein synthesis) which could result in an "error catastrophe" [63,97]. He has subsequently broadened the scope of the energy savings to encompass all types of somatic maintenance and repair [5,8].

The benefits of reduced longevity could however include any type of enhanced performance, of which a lower energy requirement is only one. Other possibilities are lower requirements for all other types of essential nutrients (e.g. vitamins, essential amino acids, essential fatty acids) and increased performance in the whole range of physiological functions of the organism (e.g. wound healing, blood clotting, cold or heat tolerance, maximal running speed, endurance). For example, imagine a modification to the cardiovascular system which increases some aspect of its performance (e.g. clotting speed) at the expense of a slow gradual accumulation of vascular damage such as atherosclerosis. In principle, any aspect of youthful performance could trade off against longevity. However, the energy savings from reduced somatic maintenance could plausibly account for most of the benefits associated with accelerated senescence.

The above explanation of aging (called the "disposable soma theory" by Kirkwood) predicts a positive correlation between species longevity and levels of damage prevention and repair processes (or any other physiological modifications which are found to reduce damage accumulation). Such a correlation has been found for ultraviolet-induced DNA excision repair [101-103] and for the relative lack of metabolism of polycyclic hydrocarbons [104,105]. The intracellular concentration of superoxide dismutase, which catalyzes the removal of the superoxide radical, has been found to correlate better with "lifespan energy potential" (the specific metabolic rate times the maximum lifespan, which equals the maximum amount of energy consumed during the lifespan per gram body weight) than with maximum lifespan [106]. This makes sense because the generation of the superoxide radical is presumably more dependent upon the rate of metabolism than upon the passage of time. Cutler [50] has documented a number of antioxidants in plasma (urate, carotenoids, alpha-tocopherol, and ceruloplasmin) which also correlate with lifespan energy potential.

Previous authors have argued for the multifactorial nature of aging, e.g. Williams [1], Maynard Smith [107], Weiss [108], and Wallace [109]. Williams [1, p. 406] argued that "Senescence should always be a generalized deterioration, and never due largely to changes in a single system.... [I]f the adverse genic effects appeared earlier in one system than any other, they would be removed by selection from that system more readily than from any other. In other words, natural selection will always be in greatest opposition to the decline of the most senescence-prone system". A generalized deterioration of all organs and systems is in fact seen in mammalian aging [45].

Cutler [110] has made an evolutionary argument against multifactorial aging:

Maximum lifespan potential was found to have increased approximately 2-fold over the past 3 million years, reaching a maximum rate of increase of 14 years per 100000 years about 100000 years ago. It is estimated that about 0.6% of the total functional genes have received substitutions leading to one or more adaptive amino acid changes during this 100000-year time-period. This suggests that aging is not the result of an expression of a large number of independently acting processes (p. 4664).

A major problem with this argument is the assumption that the evolutionary changes have resulted predominantly from base substitutions, ignoring the possible role of other types of genomic changes, e.g. insertions and deletions of DNA. [An appealing concept is mutation of gene regulation via DNA length changes, i.e. insertions or deletions in a non-coding segment of DNA (within a gene or gene complex; possibly an intron, but not necessarily) whose length affects the functioning (e.g. timing and/or degree of expression) of that gene Or gene complex. Like a rheostat, it allows fine adjustment over a wide range; for example, a 1000 base sequence can be increased or decreased in 0.1% increments. Furthermore, up to half of all small length change mutations in the "rheostat segment" would be expected to be favorable, in contrast to the very low proportion of adaptive mutations in a coding sequence. Evidence in support of this speculation is the rapid evolutionary change sometimes seen in intron length, where mutations which change intron length can be twice as frequent as base substitutions [111, 112].] Interestingly, while recent primate evolution displays a relatively low rate of neutral base substitution [113], it is also characterized by a large number of insertion and/or deletion events [114].

An important point, which Cutler [50, p. 17] states explicitly, is the distinction between the complexity of the aging process and the complexity of changing the rate of the aging process. Even assuming that aging is highly complex and multifactorial, it may nonetheless be possible to extend longevity in a simple fashion, e.g. by changing the metabolic rate and/or the developmental rate.

As Maynard Smith [107, p. 5] has pointed out, one implication of multifactorial aging is that "in most arguments about the nature of 'the aging process' both sides will be partly right". Thus, theories which localize the causes of aging to one source of damage within the body can be seen as each representing part of the truth. The wealth of existing theories explaining how aging damage can be generated itself gives testimony to the multifactorial nature of aging. Such theories exist for:


The idea that primary aging damage consists of a large number of different types of damage which tend to be relatively infrequent and harmless (when considered separately) is implicit in the very concept of multifactorial aging. A similar idea, though limited to somatic mutations, was proposed by Fulder [120], who described senescence as an "insidious accumulation of mildly deleterious events" (p. 110). However, as discussed above, there is no reason to believe that primary aging damage is restricted to DNA.

How might additional evidence for multifactorial aging be found? New sources of damage within the body may be discovered. Furthermore, analysis of broad categories of damage (e.g. "free radical damage") may show that they consist of many different types of damage which pose different repair and prevention problems. To study primary aging damage (as opposed to a mixture of primary and secondary aging damage), researchers will need to study damage which occurs before "middle age", i.e. damage which has accumulated despite the proper functioning of the body's damage repair and prevention mechanisms.

An aspect of aging which deserves far more research attention in the future is microheterogeneity. Most research methodology enforces a strong bias in favor of homogeneity [133]. For example, after homogenizing a tissue, one can only deal with averages -- it destroys any information concerning differences between the various cell types within a tissue and between individual cells of the same type. Yet due to the stochastic nature of the occurrence of molecular damage, each individual cell should experience somewhat different damage. Wilson et al. [134] used electron microscope cytochemistry to show differences among rat liver hepatocytes as a result of senescence (see also Wilson [135]). Microheterogeneity is also evident in in vitro aging [136,137].

In conclusion, aging consists of wear and tear which accumulates not because the damage is unrepairable but rather because the cost of its repair (largely due to its great heterogeneity) is greater than the corresponding benefit.

REFERENCES

Thanks for visiting

    Send me email      Return to Home