https://www.sciencedirect.com/science/article/pii/S1568163726002059
A sort of movement in the right direction. As people know I think ageing and development is a program primarily driven by damage to mtDNA which then varies the levels of nuclear acetylation and that drives the phenotype. Hence the programmed ageing vs accumulated damage is a false dichotomy. This paper comes close to that argument as well.
chatGPT(5.6paid):
Overall assessment
This is a thoughtful but highly speculative review proposing a middle ground between “aging is accumulated damage” and “aging is programmed.” Wolfgang Wagner’s central idea is that organisms may not possess a dedicated death program; instead, genetic and epigenetic networks regulate the rate at which stochastic damage and epigenetic drift accumulate.
The strongest part of the paper is this reframing of a program as a regulated rate of deterioration. The weakest part is the further suggestion that aging itself may have been selected because generational turnover increases recombination and species-level adaptability. That evolutionary argument is plausible only under restricted population structures and is not demonstrated by the biological examples presented.
1. Summary of the argument
The traditional divide
The paper begins with two broad classes of aging theory:
- Error-based theories: aging results from accumulated mutations, protein damage, telomere attrition, epigenetic drift and other defects.
- Programmatic theories: aging is partly driven by coordinated, development-related processes.
Classical evolutionary theories—mutation accumulation, antagonistic pleiotropy and disposable soma—explain aging without requiring natural selection to favour aging itself.
Wagner argues that this dichotomy is too rigid.
A broader definition of “program”
He distinguishes a dedicated aging program from a higher-order regulatory architecture.
The proposed model is approximately:
[
\text{regulatory networks}
\longrightarrow
\text{rate of damage production, repair and tolerance}
\longrightarrow
\text{species-specific aging trajectory}
]
Thus, stochastic damage remains the proximate mechanism, but the rate at which it accumulates is biologically regulated and evolutionarily adjustable.
The analogy is not a program specifying each age-related change. It is closer to a program setting the system’s maintenance intensity, error rate and tolerance thresholds.
Why aging might be adaptive
The paper revisits the old Weismannian idea that removal of older individuals may benefit populations. Wagner’s particular argument is that aging:
- removes older generations from resource-limited environments;
- makes space for new individuals;
- increases generational turnover;
- increases the frequency of sexual reproduction and recombination;
- may therefore accelerate adaptation to changing conditions.
Spatially structured populations could, in principle, allow groups with different aging strategies to compete. An aging population might then adapt more quickly than a non-aging population.
Comparative biological examples
Wagner cites several observations as broadly compatible with this argument:
- Lifespan correlates with reproductive schedules and age at sexual maturity.
- Naked mole-rat queens and social-insect queens have exceptional longevity.
- Hydra reproduce predominantly through asexual budding and exhibit negligible senescence under some conditions.
- Germline continuity and cellular reprogramming show that deterioration is not chemically inevitable.
- Species have evolved very different levels of DNA repair and somatic maintenance.
The proposed interpretation is that lifespan is adjusted to ecological and reproductive context rather than determined simply by an unavoidable universal rate of molecular damage.
Epigenetic clocks
The second major strand concerns epigenetic aging.
The paper notes that:
- age-associated DNA methylation changes are reproducible;
- many affected sites are enriched in developmental and Polycomb/PRC2-regulated genes;
- epigenetic age is reset during reprogramming to pluripotency;
- epigenetic clocks tick at different rates in species with different lifespans;
- similar age-related changes occur in histone marks and chromatin accessibility.
These observations make aging look superficially developmental or programmatic.
But Wagner also accepts that methylation aging has a large stochastic component. Imperfect maintenance of methylation states can move CpGs toward site-specific equilibrium values, producing predictable clocks without any clock program.
A possible epigenetic network
The most mechanistically specific section discusses experiments from Wagner’s group in which targeted methylation editing at individual age-associated CpGs produced methylation changes at other loci.
These “bystander” changes:
- were reportedly not explained by direct editing off-targets;
- were enriched at other age-associated CpGs;
- included sites that normally gain and sites that normally lose methylation with age.
Wagner suggests that age-associated CpGs may therefore form a genome-wide regulatory network rather than behaving as completely independent stochastic sites.
Final synthesis
The paper concludes that aging may consist of stochastic errors accumulating at a regulated, species-specific rate. This rate could be set by networks controlling repair, maintenance, metabolism and epigenetic stability, and shaped by ecological selection.
In short:
Aging need not be either an exact genetic program or completely unregulated molecular chaos.
2. What is novel?
This is a conceptual review rather than a report of new experimental work. Its novelty therefore lies mostly in synthesis and terminology.
The “programmed aging rate” formulation
The paper’s clearest contribution is to relocate the programmatic component from the individual lesions to the rate of lesion accumulation.
Under this model:
- mutations need not occur at predetermined positions;
- methylation drift need not follow an exact script;
- individual outcomes can remain stochastic;
- nevertheless, repair capacity, chromatin maintenance and damage tolerance can set a reproducible aging pace.
This formulation reconciles reliable population-level aging trajectories with noisy molecular events.
Integrating stochastic epigenetic clocks with regulation
The paper does not treat the stochasticity of methylation clocks as evidence against regulation. Instead, it argues that biological networks might regulate the statistical properties of drift—the rate, direction, equilibrium and covariance of methylation changes.
That is a more sophisticated claim than saying an epigenetic clock is either a programmed timer or a passive damage meter.
Linking aging to the evolutionary benefit of recombination
The particular group-selection argument emphasizes generational replacement as a means of increasing recombination opportunities. This is a more specific hypothesis than the generic claim that old individuals consume resources.
It generates, at least in principle, testable predictions concerning:
- environmental variability;
- population spatial structure;
- dispersal;
- resource limitation;
- sexual versus asexual reproduction;
- generation time;
- the strength of competition between generations.
Genome-wide epigenetic network proposal
The suggestion that perturbing one age-associated CpG elicits coordinated changes at distant age-associated CpGs is potentially important. If replicated and mechanistically explained, it could shift the interpretation of clocks from collections of correlated markers toward readouts of an interacting chromatin system.
But the review presently gives this idea more weight than the available evidence can support.
3. Strengths
It rejects an unhelpful binary
The paper is correct that “program” and “damage” are not mutually exclusive. DNA repair, proteostasis, mitophagy, antioxidant defences and immune clearance are regulated. Consequently, any aging process driven by failures in these systems must have both stochastic and regulated components.
It distinguishes predictability from determinism
A process can be statistically predictable without each molecular event being predetermined. Epigenetic clocks are an excellent example: accurate age prediction does not prove that their CpGs constitute a causal timer.
It is appropriately cautious about clocks
The author repeatedly acknowledges that:
- epigenetic clocks may be biomarkers rather than causes;
- biological age cannot be represented by one metric;
- partial reprogramming remains controversial;
- the function of most age-associated methylation changes is unknown.
That caution prevents the paper from making the common error of equating clock reversal with organismal rejuvenation.
It emphasizes evolvability of maintenance
The observation that species differ enormously in repair and maintenance capacity is important. Aging rate cannot be explained simply as a fixed physicochemical rate of entropy production. Evolution plainly can modify how organisms resist and tolerate damage.
4. Critique
A. The central evolutionary case is much weaker than the regulatory case
The paper successfully establishes that aging rate is regulated and evolvable. It does not thereby establish that aging was selected for.
There is an important distinction:
[
\text{selection for a particular maintenance allocation}
\neq
\text{selection for aging as an adaptive outcome}
]
Classical disposable-soma and antagonistic-pleiotropy theories already predict species-specific maintenance and aging rates. If high somatic maintenance has costs, selection can favour lower maintenance after the ages at which it improves inclusive fitness. No benefit from death or generational replacement is required.
The proposed synthesis therefore incorporates a strong point already contained within non-programmatic evolutionary theory, then moves from it to a much less well-supported adaptive-aging conclusion.
B. The recombination argument neglects individual-level invasion
Suppose an aging allele benefits the population by freeing resources for future generations. A non-aging or slower-aging mutant in that same population would ordinarily gain an immediate individual advantage:
- it survives longer;
- may reproduce for longer;
- retains resources;
- transmits more copies of its allele.
For adaptive aging to evolve, the group-level benefit must exceed this within-group advantage. That normally requires strong assortment, restricted migration, kin structure, group extinction/recolonisation or another mechanism generating substantial between-group selection.
The paper acknowledges spatial structure but does not quantify whether realistic structures are sufficient. Its island schematic illustrates the desired outcome but does not solve the invasion problem.
C. More turnover does not automatically mean more useful recombination
The argument implicitly equates shorter lifespan with more reproductive cycles. But population turnover depends on:
- age-specific fecundity;
- juvenile survival;
- density dependence;
- generation overlap;
- reproductive suppression;
- parental investment;
- carrying capacity;
- migration.
Killing older individuals may not increase the number of successfully reproducing offspring. Older individuals can also contribute parental care, knowledge, defence and kin support. In humans and some social species, post-reproductive survival may increase the reproductive success of descendants.
Nor is more recombination universally beneficial. Recombination can break up favourable allele combinations, and the advantage depends on linkage, epistasis and environmental change. The paper discusses benefits of recombination but gives insufficient attention to these costs and conditions.
D. Several comparative examples are suggestive but non-diagnostic
Naked mole-rats and eusocial queens
Long-lived queens are compatible with selection maintaining organisms that have high reproductive value. But this supports standard age-specific selection theory at least as readily as adaptive aging.
The longevity of workers also cannot be dismissed merely because they do not reproduce. Workers contribute substantially to inclusive fitness, so their survival can be strongly selected through kin selection.
Hydra
The association between asexual budding and negligible senescence does not demonstrate that absence of recombination removes the benefit of aging. Hydra also possess continuous stem-cell renewal, unusual body architecture, low differentiation complexity and high regenerative capacity. These are major mechanistic confounders.
Moreover, environmental and strain-dependent senescence occurs in some Hydra contexts, so “biological immortality” requires qualification.
Oocyte atresia
The paper asks why reproductive windows are limited if declining selection begins only after reproduction. But reproductive senescence need not be an aging program. Oocyte production, quality control, maternal risk, developmental constraints and early-life optimisation can generate menopause or follicular depletion without selection for organismal aging.
Your earlier point about the mtDNA germline bottleneck is relevant here: extensive follicular atresia could provide mitochondrial and cellular quality selection even though individual follicle loss is regulated. A regulated loss mechanism does not establish a higher-level program for aging.
E. Germline continuity and iPSC rejuvenation are overinterpreted
The germline is not literally free of damage. It is maintained through:
- selection among cells and organisms;
- DNA repair;
- mitochondrial bottlenecks;
- elimination of defective oocytes or embryos;
- turnover between generations.
Continuity of the lineage does not mean that any individual germ cell remains indefinitely undamaged.
Likewise, iPSC reprogramming demonstrates that many epigenetic and cellular aging features are reversible, but it also:
- radically changes cell identity;
- selects successfully reprogrammable cells;
- does not reliably eliminate DNA mutations;
- can introduce genomic abnormalities;
- does not prove that an intact organism can be reset safely.
Reversibility establishes that some aging features are state-dependent; it does not establish that those features were programmed to cause aging.
F. The epigenetic evidence does not yet distinguish program from constrained drift
Enrichment of age-associated methylation at PRC2 targets is interesting, but several non-programmatic explanations remain possible:
- bivalent and Polycomb-regulated regions may be unusually susceptible to maintenance errors;
- chromatin state may constrain which CpGs can drift;
- cell-composition changes may contribute;
- clonal selection may produce reproducible population patterns;
- developmental genes may be epigenetically poised and therefore sensitive to perturbation.
A stochastic process operating on a structured chromatin landscape will produce non-random, reproducible patterns. Consequently:
[
\text{structured age-related change}
\not\Rightarrow
\text{developmental aging program}
]
This is the most important inferential limitation of the epigenetic argument.
G. “Bystander” methylation changes do not yet establish a coordinating network
Changes at distant CpGs after targeted editing could reflect:
- a shared transcription-factor response;
- stress caused by the editing system;
- altered proliferation or cell-state composition;
- chromatin remodelling secondary to local gene-expression changes;
- selection of particular edited clones;
- measurement and regression artefacts.
Calling the result an interconnected epigenetic network is a reasonable hypothesis, but “network” needs mechanistic content: directionality, mediators, reproducibility across cell types and perturbations, time-resolved propagation, and rescue experiments.
A particularly informative test would be whether perturbing several different CpGs moves other loci in a consistent, predictable direction and whether blocking a proposed intermediary abolishes that propagation.
H. The definition of “program” risks becoming too broad
If any genetically regulated rate of imperfect maintenance is called quasi-programmatic, then almost every biological phenotype becomes programmed. This may reconcile competing camps semantically rather than mechanistically.
A useful programmatic theory needs predictions that differ from disposable-soma or antagonistic-pleiotropy models. For example:
- Does the organism actively accelerate damage after a defined life stage?
- Are there coordinated control nodes whose inhibition slows several independent damage processes without merely reallocating resources?
- Does aging increase population fitness under specified ecological conditions?
- Can an aging-promoting allele resist invasion by a longer-lived mutant?
Without such discriminating predictions, “programmed rate of error accumulation” may simply rename evolved maintenance capacity.
I. Histone acetylation and metabolic control are underdeveloped
Given the paper’s emphasis on epigenetic regulation, it concentrates overwhelmingly on DNA methylation. It briefly cites histone and chromatin-accessibility changes but does not integrate:
- acetyl-CoA availability;
- ATP-dependent chromatin remodelling;
- NAD-dependent deacetylation;
- histone acetylation;
- mitochondrial metabolism;
- transcriptional elongation and splicing;
- cell-type-specific metabolic state.
This omission matters because methylation changes might be downstream records of broader metabolic and chromatin-state transitions rather than the principal regulatory clock.
In the context of your citrate → ACLY → acetyl-CoA hypothesis, an alternative model would be:
[
\text{mitochondrial/metabolic deterioration}
\rightarrow
\text{altered nuclear acetyl-CoA and chromatin accessibility}
\rightarrow
\text{transcription/splicing changes}
\rightarrow
\text{secondary methylation remodelling}
]
The paper does not compare this direction of causality with its proposed methylation-network interpretation.
J. The figures are conceptual illustrations, not evidence
Figure 1 shows mechanisms capable of mitigating each hallmark. But the existence of repair or clearance machinery does not mean complete prevention is feasible or selectively advantageous. These systems consume energy, can create cancer risks and may interfere with development or immunity.
Figure 2 illustrates the group-selection scenario but omits competing slow-aging mutants within the aging population.
Figure 3 is a useful summary of the hypothesis, although its arrows do not specify causal mechanisms or measurable variables.
K. Potential conflict and perspective bias
The author reports equity and board membership in Cygenia GmbH and a patent on an epigenetic aging signature. This does not invalidate the arguments, but it is relevant because the review gives particular prominence to DNA-methylation clocks and the author’s own epigenetic-editing work.
5. What the paper establishes—and what it does not
| Claim | Evidential status |
|---|---|
| Aging contains stochastic molecular changes | Strongly supported |
| Organisms regulate damage production, repair and tolerance | Strongly supported |
| Evolution adjusts maintenance and aging rates between species | Strongly supported |
| Epigenetic clocks contain substantial stochastic components | Strongly supported |
| Age-associated methylation is structured by genomic context | Strongly supported |
| Methylation clocks are causal drivers of aging | Not established |
| Age-associated CpGs form a functional genome-wide control network | Interesting preliminary hypothesis |
| Aging was directly selected to increase generational turnover | Speculative |
| Aging improves adaptability by increasing recombination | Conditional theoretical possibility, not demonstrated |
| Reprogramming proves organismal aging is programmed and reversible | Not established |
Bottom line
The paper’s valuable insight is:
Aging can be stochastic in its molecular details while regulated in its overall rate.
That is a useful framework and probably correct. However, it does not require adaptive or species-selected aging. Standard evolutionary theories already allow natural selection to regulate investment in repair, maintenance and reproduction.
The paper therefore provides a convincing reconciliation of regulated maintenance with stochastic deterioration, but only a speculative argument that aging itself exists to accelerate generational turnover. Its epigenetic-network hypothesis is potentially important, yet currently correlational and insufficient to establish either a causal aging clock or a developmental aging program.