(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



A unified framework for evolutionary genetic and physiological theories of aging [1]

['Jean-François Lemaître', 'Université Lyon', 'Cnrs', 'Laboratoire De Biométrie Et Biologie Évolutive Umr', 'Villeurbanne', 'Jacob Moorad', 'Institute Of Ecology', 'Evolution', 'School Of Biological Science', 'University Of Edinburgh']

Date: 2024-03

A second physiological theory of aging with an evolutionary basis is the DTA. This theory asserts that aging results from suboptimal regulation of age-specific gene expression in adulthood and thus emerges as a by-product of strongly selected and regulated developmental processes [24–26]. From a broad evolutionary perspective, the DTA views the evolution of aging as the result of weakening selection with age resulting in “physiological processes that are optimized for early-life development, growth, and reproduction and are not sufficiently optimized for late-life function” [26].

Recent verbal models of the DTA invoke direct effects of gene expression on function and fitness and frame the model as distinct from those invoking indirect effects of damage or repair processes [ 45 ]. We argue this is unnecessary from an evolutionary perspective. Aging could foreseeably evolve via DTA through direct, damage-independent effects on fitness (as envisaged in [ 24 , 45 , 46 ]). For example, non-optimized mammalian TOR pathway expression in late life could directly cause cellular dysfunction and disease. However, the DTA does not preclude a role for damage or failure of repair/replacement function in late life. For example, hypofunction could manifest through loss of proteostasis, a key hallmark of aging [ 7 , 8 ]. Genes controlling the production of molecular chaperones and protein degradation (ubiquitin–proteasome system and autophagy) may be optimally expressed in early life for development and fitness but insufficiently expressed to prevent the ultimate accumulation of misfolded proteins later in life. This could result in aging through indirect effects of a gene (or genes) optimized for early-life fitness but insufficiently expressed to prevent functional failure at some point in late life ( Fig 3 ). Tissues or biological structures which have a fixed form or size once development is complete (e.g., tooth size in some mammals, wing or body structure in some insects) but show deterioration via environmental wear and tear may represent similar examples of hypofunction (e.g., tooth wear limiting feeding and wing damage limiting flight [ 50 , 51 ]). The key difference here with the DST lies in the critical fact that resource allocation trade-offs are not involved. A fundamental question for the DTA theory is whether it purely invokes direct effects of age-specific gene expression or whether it can involve indirect effects of damage accumulation as well. We would argue these are important mechanistic distinctions, but that the DTA currently encompasses both and can remain distinct from the DST as long as resource allocation trade-offs are precluded.

The DTA suggests that selection does not sufficiently optimize age-specific biological function because the force of selection on traits declines with increasing age. The DTA covers a broad range of scenarios where the biological function is higher or lower than optimal for a given age. The “selection shadow” reflects late life, when selection is weak or absent under natural conditions. Three potential patterns are illustrated. First, consistent continuous biological function can become deleterious in late ages, but natural selection is too weak (“selection shadow”) to result in the evolution of a modifier gene that will down-regulate (or sufficiently down-regulate) late-life expression of the focal gene (hyperfunction; A ). Second, it could be beneficial to increase biological function with age, but selection in late life is not strong enough to achieve this, and so expression evolves to be below the optimal level in late life (hypofunction; B ). Finally, it could be optimal to maintain biological function at a constant level with age, but weak selection in late life results in misregulation and either over (hyperfunction) or under (hypofunction) expression in old age ( C ). DTA, developmental theory of aging.

Hyperfunction is just one potential way in which the DTA might manifest. Alternatively, the switching-off or suppression of a developmental pathway in late life or the failure to maintain expression at adequate level after development is complete could contribute to aging [ 24 , 26 ]. This mechanism has been termed “hypofunction” [ 26 ]. Putative examples include the rapid suppression of heat shock protein expression in worms during early adulthood that results in a weakened adult stress response [ 26 , 48 ], sarcopenia [ 24 ], or an absence of tooth replacement in late life in mammals such as elephants, where cycles of tooth rotation occur [ 49 ]. Fig 3 illustrates the different ways in which the DTA might manifest, clarifying that hyperfunction and hypofunction are 2 sides of this theoretical coin, depending on whether expression lies above or below the fitness optima in late life.

The most widely discussed process through which the DTA might manifest is the continued expression of cellular pathways which favor growth and development, and thereby early-life fitness, but which become deleterious to fitness in late life and ultimately contribute to increased aging [ 24 ]. This is termed the “hyperfunction theory of aging” [ 46 ]. Blagosklonny [ 46 ] specifically associates the term with a particular pathway—the target of rapamycin (TOR) pathway—suppression of which extends lifespan and delays aging across distantly related laboratory model organisms [ 46 , 47 ]. He argues that TOR expression is essential in early life for development, growth, and anabolic cellular function but continued high levels of expression in later adulthood are associated with late-onset pathologies including osteoporosis and myocardial infarction in humans [ 46 ]. In his paper that introduces the AP model of aging, Williams illustrates with an example of a similar process. He proposes a gene promoting calcification that enhances bone growth in early life but, with continued expression, results in a loss of arterial elasticity and subsequently heart disease in late life [ 19 ].

We can have 2 vital rates to phenotype functions that apply at 2 different ages but have identical shapes. However, because the “selection shadow” that leads to less selection at late ages applies, the fitness to late-age phenotype function will appear flatter, or more compressed along the fitness axis, than the corresponding fitness to early-age phenotype function ( Fig 4 ). One consequence of this is that directional selection (the mean slope experienced by the population) will push the early phenotype more strongly towards the early-age optimum than the corresponding late phenotype, provided that they both are evaluated at the same distance from their optima. Second, diminishing fitness returns implied by the shape of the functions means that there will be early-life phenotypes close to their corresponding optimum that experience the same selection pressures as late-life phenotypes that are further from their optimum. As we argue below, this feature is important in determining aging rates that evolve through an MA process.

These slopes can be interpreted in terms of natural selection by placing these functions into the context offered by the hierarchy presented in Fig 1 . By multiplying the function described above (phenotype to vital rate) by the function that Hamilton used to describe selection for vital rates (fitness to vital rates), we get a new function that directly assesses the sensitivity of fitness to changes in the phenotype ( Fig 4 , outer functions). This is exactly a model of “optimizing selection” (sensu [ 52 ]), in which intermediate phenotypes have the highest fitness and extreme phenotypes (in either direction) have the lowest fitness. Optimizing selection is a description of the global function that relates fitness to the phenotype. The strength and direction of directional selection determines how the force of natural selection contributes to the change in the mean value of the phenotype over one generation. This is equal to the average slope of the global fitness function taken over all members of the population (which is equal to the slope at the mean phenotype when the distribution of individual trait values is Gaussian) [ 53 ]. Directional selection thus describes the slope of the local fitness function.

This figure illustrates how the proposed optimality model that links phenotypes to vital rates, and captures verbal models of the DTA, fits into the hierarchy illustrated in Fig 1 and how it yields age-specific directional selection. The optimality functions (lower 2 plots in the center) are assumed to be identical at both ages. However, because selection emphasizes improvements to vital rates more in early life than at late life (upper 2 plots in the center), the sensitivity of fitness to changes in the early-age phenotypes (left) is greater than the fitness sensitivity to changes in the late-life phenotype (right). These outer functions define age-specific directional selection, which is the slope at the population-mean phenotype under our model assumptions. The blue circles represent these means, and the blue lines are the slopes at those points. Note that the same amounts of directional selection can exist at both ages, provided that the late-age phenotype is further removed from its optimum. DTA, developmental theory of aging.

Although increasingly discussed and tested in the biogerontological literature, the DTA has never been formally framed within the ETA. However, much like the DST, the DTA is has been viewed as an AP mechanism [ 25 , 26 ]. As we show below, however, the DTA can work via both MA and AP. Verbal models of DTA present aging as evolving due to a failure of optimization of key physiological processes by natural selection in late life. We can integrate this mechanism into the ETA by linking age-specific manifestations of these physiological processes (phenotypes) to vital rates at those same ages. The age-specific vital rates are then optimized at some specific phenotypic value; phenotypes that are below or above these values lead to lower vital rates. The phenotypic location of that vital rate optimum can itself be age specific. Let us also imagine that vital rate improvement diminishes as the phenotype approaches the optimum (a diminishing return); this can be visualized as a Gaussian function. The slope of this function at any point along the phenotypic axis quantifies the sensitivity of the vital rate to changes in the phenotype (see the 2 bottom, central functions in Fig 4 ).

Age-specific optimality selection

In the following discussion, we use a model of age-specific optimality selection to illustrate how hyperfunction and hypofunction can act through both MA and AP mechanisms to cause the evolution of aging. Here, we summarize these descriptions using the biological hierarchy introduced in Fig 1. In Fig 5, we present 2 scenarios through which this conceptualization of the DTA might lead to the evolution of aging when genetic effects on early and late life are negatively correlated (AP; Fig 5A and 5B). Interestingly, aging can also evolve under this model when genetic effects are strongly positively correlated and young and old phenotypes sit on opposite sides of their age-specific optima (Fig 5C and 5D). Whether hyperfunction or hypofunction evolves simply depends upon which sides of the peak the late-life phenotype ends up.

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 5. The DTA with age-dependent pleiotropy. Blue circles identify the phenotypic value of the population mean before senescence evolves, the orange stars identify its value afterwards, and “g” refers to a single allele that increases senescence. A and B illustrate cases where that allele has effects that are negatively associated across the age-specific phenotypes and the population mean phenotypes at early and late ages begin at suboptimal values that are on the same side of the optimum. (A) Hypofunction: the population mean phenotypes are less than their optimal values. Natural selection favors the increase of the trait early in life despite its costs to vital rates late in life. (B) Hyperfunction: the optima are below the initial population means. Selection favors a decrease in the early-life phenotype, and this causes an increase in the late phenotype. C and D illustrate cases where that allele has effects that are positively associated across the age-specific phenotypes and the population mean phenotypes at early and late ages begin at suboptimal values that are on opposite sides of the optimum. (C) Hypofunction: the population mean phenotype at early and late life are higher and lower than their respective optima, respectively. Natural selection will act to reduce early-life values, which causes late-life values to become lower and therefore more suboptimal. (D) Hyperfunction: the population mean phenotype at early and late life are lower and higher than their respective optima, respectively. Natural selection will act to increase early-life values, which causes late-life values to become higher and therefore more suboptimal. DTA, developmental theory of aging. https://doi.org/10.1371/journal.pbio.3002513.g005

Aging could also evolve in the DTA framework via MA processes (Fig 6A and 6B), contrary to the widely held view that the DTA is necessarily a specific case of the AP model. To illustrate this, we imagine that de novo mutations arise every generation and these mutations have age-specific effects on phenotypes that tend to move the mean phenotype in either a positive or negative direction. At the same time, directional selection will cause the population mean phenotypes to migrate up the vital rate gradient at both ages. We can expect that if the properties of these new mutations are the same for both ages (e.g., age-specific mutations tend to be equally damaging to vital rates at the different ages), then the populations will reach evolutionary equilibria at points where the early-age phenotype is closer to its vital rate optimum than the late-age phenotype is to its optimum. When the mean effects of new mutations on the phenotype are negative, then this senescence is described by hypofunction (Fig 6A), and when the mean effects are positive, then senescence via hyperfunction evolves (Fig 6B). This conceptualization of the DTA as age-specific optimizing selection is novel, but we feel it captures most verbal models of the theory in the literature adequately. It importantly presents the DTA as a process of weakened optimizing selection in late life, which could evolve through either AP or MA processes.

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 6. The DTA as an MA model. For the MA model, we imagine that there are 2 mutations with strictly age-specific effects (g y and g o ) that recur every generation with some equal and unspecified frequency. The joint effects of mutations and selection will cause the population to equilibrate with age-specific phenotypes that are closer to the early-age optimum than to the late-life optimum. If the effects of the mutations tend to reduce the phenotype, then hypofunction evolves (A). Alternatively, hyperfunction evolves if mutations tend to increase the phenotype (B). DTA, developmental theory of aging; MA, mutation accumulation. https://doi.org/10.1371/journal.pbio.3002513.g006

To illustrate these ideas in more depth, we begin by imagining that at every age, there is a standard Gaussian function that relates individual values of a phenotype to a vital rate manifested at that same age, and this vital rate is maximized at some intermediate phenotypic value (Fig 4). This age-specific optimal set of vital rates may change with age, but here we standardize it by assigning to each optimum a phenotypic value of “0.” Individual aging is indicated when age-specific values are at different distances from their respective means. “Senescence” is defined here to mean an age-related change that is deleterious with respect to vital rates. In this model, phenotypic senescence is found when the late-age phenotypic mean is further from its optimum in any direction than the early-age phenotypic mean is distant from its optimum. Likewise, equidistance between mean age-specific phenotypes and their corresponding optima equates to no aging, and negative senescence for that phenotype exists when the late-age phenotypic mean is closer to its optimum that its early age counterpart. This model suggests 2 forms of senescence associated with the DTA. Hypofunction implies that the late-acting phenotypic mean is further from its optimum because its value is too small, and hyperfunction follows when this mean is further from its optimum because it is too large. Below, we describe how this model causes phenotypes to evolve senescence from an initial non-aging state by the action of natural selection.

Let us assume for simplicity that the phenotype at a given age only affects a vital rate at that same age, and the distribution of individuals’ phenotypes within the population is also Gaussian. The “vital rate gradient” is the slope of this function found at the population mean phenotype. The strength of “directional selection,” or simply β that favors an increased phenotypic value is equal to the product of this vital rate gradient and the gradient that describes the strength of selection for the vital rate [30]. This latter gradient is independent of the phenotype and, as the ETA shows us, it weakens with age [20]. As we have assumed that the shape of the function that associates vital rates to trait values is the same across ages, it follows that the magnitude of β will be less at older age if phenotypes at the 2 ages are equidistant from their respective optima. However, it should also be noted that because the local curvature of the function experienced by the population changes with the mean phenotype, the vital rate gradient will tend towards zero as the population approaches the optimum. To summarize, directional selection for the age-specific phenotype, β, is the product of: directional selection for the vital rate, which must decrease with age, and the vital rate gradient, which must decrease as the population mean approaches its age-specific optimum.

A non-aging population will necessarily experience stronger β for the early phenotype because the vital rate gradient is defined to be identical for both age-specific phenotypes. Assuming equal amounts of heritable variation for the phenotypes at both ages, this difference in β will cause the evolution of incipient senescence because the early-age phenotypes will move more effectively towards its optimum than the late-age phenotypes. The same process will occur in the next generation, but because the early-age mean will begin closer to its optimum than the late-age mean, the vital rate gradient will become stronger for the late-age phenotype. This will reduce the difference between β at each age, and less new senescence will evolve in this generation. We can expect this process to continue to increase evolved senescence over subsequent generations until the population reaches an evolutionary equilibrium dictated by the genetic constraints specific to the 2 age-specific phenotypes. The nature of these genetic constraints differs between classical models of MA and AP, and we apply genetic models of both to the fitness functions implied by our optimization models in what follows.

Age-dependent de novo mutations will arise in every new generation. In keeping with classical population genetic models of MA [54], we assume that mutations have strictly age-specific effects. Every generation, these mutations will change the mean age-specific phenotypes in some direction and magnitude that is constant over time (the mean mutational effect). We assume that this change is the same for both age-specific phenotypic means. We note that the direction of change in the mean phenotype can be positive or negative, and these directions can be shown to lead to hyperfunction and hypofunction, accordingly. To demonstrate, let us first change our perspective of age-specific vital rate to phenotype functions from the univariate view (Figs 4–6) to a bivariate view of fitness given in Fig 7A. This figure describes the function that links fitness to the underlying age-specific phenotypes at early and late ages (the x and y axes, respectively). A bivariate fitness optimum is arbitrarily set at the (0,0) phenotype. When viewed from the perspective, directional selection is proportional to the vector that finds the steepest fitness gradient at the point of the bivariate mean. Over a single generation, the population mean will change in both dimensions according to the effects of natural selection and new mutations. These processes can have effects on the phenotypes that act in concert or in antagonism, but an evolutionary equilibrium is found only when these causes of phenotypic change are equal and opposite.

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 7. Reconciling the DTA and the ETA via optimizing selection. (A) Mutation accumulation. The contour plot illustrates the function that relates fitness (the contours) and early and late phenotype values (the x and y axes, respectively) when corresponding age-specific vital rates are optimized at intermediate phenotype values. This is a bivariate perspective of the univariate fitness functions illustrated in Fig 4. Negative values indicate regions of low fitness and the “+” identifies the phenotype combination that maximizes fitness. The y = x and y = -x axes define phenotypes that combine to describe no aging; the intersection of these axes creates quadrants of parameter space that describe different types of aging. Hyperfunction and hypofunction correspond to the top and bottom quadrants, respectively, and reverse senescence is indicated by phenotypes in the left and right quadrants. The distribution of phenotypes in a population is represented by gray circles, and the mean values are located at their centers. The arrows indicate the evolutionary progression of these populations from initial states defined by no aging to states consistent with senescence. (B) AP. The initial non-aging populations are represented by ellipses with bivariate means located on the y = x axis and indicated by points. The lines along the major axis of each identify the allowable phenotypic combinations that are permitted given the strict genetic constraints placed on the model. The arrows indicate the paths that these populations follow as they evolve towards the phenotype combinations with maximum allowable fitness. The final populations are described by the gray ellipses centered on this point. Senescence can evolve either with negative genetic correlations across phenotypes (the bottom-left and upper-right distributions that correspond to Fig 5A and 5B, respectively) or with positive genetic correlations across phenotypes (the bottom-right and upper-left distributions that correspond to Fig 5C and 5D, respectively). AP, antagonistic pleiotropy; DTA, developmental theory of aging; ETA, evolutionary theory of aging. https://doi.org/10.1371/journal.pbio.3002513.g007

Let us consider first the case when mutations tend to have a negative effect on both phenotypes. We can begin with an example of a non-aging population where both age-specific phenotypes are above the optima. This corresponds to population “1-” in Fig 7A. In this case, directional selection will cause the phenotypes to decrease (as this will increase fitness), and the effect of mutations will move the population in the same general direction. The population will quickly approach the optimum phenotypic values, but recurrent mutations will eventually push the age-specific phenotypes into values that are suboptimal for both ages, at which point selection will now favor phenotype increases. This process is illustrated by the arrow leading from “1-.” An equilibrium will be found at the point where increases caused by selection exactly cancel decreases caused by mutation. Recall that the changes due to mutation must be the same at both ages; this requires that selection must also be equal at equilibrium. This equilibrium point (indicated by the population with the dot at its center; Fig 7A) resides at a point where the ratio of selection gradients for early-age vital rates to late-age vital rates is equal to the ratio of vital rate gradients for late-age phenotypes to early-age phenotypes. This condition is met only when the late-age phenotype is further removed from its optimum than the early-age phenotype. This defines senescence, and because the late-age displacement is negative, this is hypofunction. This corresponds to the univariate perspective on MA and hypofunction illustrated in Fig 7A. Note that the starting point makes no difference to the location of the equilibrium. We had imagined that a non-aging population began with equally above optimal age-specific phenotypes, but we could just as easily imagine them as equally below optimal. In this case (population 2-), selection and mutation act in antagonism from the outset, the evolved change is smaller, but the equilibrium point is the same as before. The process is like that above if we assume instead that mutations tend to increase both age-specific phenotypes. However, regardless of whether our non-aging population begin with below optimal (2+) or above optimal (1+) phenotypes, the evolutionary equilibrium is found at the same phenotypes that describe senescence by hyperfunction (consistent with Fig 7B).

In the previous model (MA), new mutations were assumed to have effects on phenotypes at one age that were independent of effects on the other age. This allowed the age-specific phenotypes to evolve independent of each other. Viewed from the bivariate perspective provided in Fig 7A, natural selection was free to drive the population means directly up the fitness gradient in the direction of the global maximum. However, AP models assume that single alleles have effects on traits at multiple ages, and these effects have opposing effects on the vital rates at early and late ages. The effects on the age-specific phenotypes are contextual because the relationship between changes to phenotypes and, changes to vital rates depend upon where the population phenotypic means are located relative to their optima. For example, an allele that increases fitness and senescence by increasing age-specific phenotypes at both early and old ages can be qualify as an “AP” allele if that increase is advantageous when young and deleterious when old (Fig 7B). The optimality model implies that alleles with correlated effects at the 2 ages can be AP when one of 2 conditions are met: the age-specific phenotype means are on the same side of their optima and, the alleles have opposing effects on the phenotype; or the means are on different sides of their optima and, the alleles have effects on the phenotype that are in the same direction. We will demonstrate that both can lead to the evolution of senescence by either hypofunction or hyperfunction.

We consider first the former scenario. As before, we assume an initially non-aging population. The population mean can either begin equally below optimal for both age-specific phenotypes or equally above optimal (Fig 7B, bottom-left and top-right distributions). The population means are represented by dots, and because the allelic effects are highly positively correlated across the age-specific phenotypes, the joint distribution of genetic values that underly the population are depicted as being highly eccentric. In both cases, natural selection is not free to drive the population directly up the fitness gradients because very little genetic variation exists in that direction (there are no single alleles that can decrease or increase both phenotypes). Instead, the highly correlated allelic effects will constrain evolution to proceed along the axis of these distributions. For the case where both age-specific phenotypes begin above optimal, the population will find a position that maximizes fitness in obtainable parameter space that is near optimal for the early phenotype but above optimal for the late phenotype (hyperfunction). When both age-specific phenotypes are initially below optimal, selection will drive the population towards senescence via hypofunction.

If the population instead begins with one below optimal and one above optimal age-specific phenotype, and the alleles have the same effects on both (Fig 7B, bottom-right and top-left distributions), then the evolution of the phenotypes are constrained along axes that are orthogonal to those in the previous example. As before, the maximum fitness that is obtainable is located at age-specific phenotypes that demonstrate senescence. In this latter scenario, hyperfunction evolves from populations which begin with below-optimal early phenotypes and above-optimal late phenotypes. Hypofunction evolves from populations with the opposite initial configurations.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3002513

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.