https://onlinelibrary.wiley.com/doi/epdf/10.1002/brv.70109
chatGPT:
Here’s a structured read of “Why we age” (Ringel, Biological Reviews 2025).
1. Short summary
Ringel argues that all ultimate (evolutionary) theories of ageing fall into three buckets:
- Mechanistic theories – rate-of-living, oxidative damage, hyperfunction, etc., which (as he defines them) do not explain why selection allows ageing.
- Weakening force of selection theories – chiefly Medawar’s mutation accumulation (plus “old age is outside the training set” style ideas).
- Optimisation theories – especially disposable soma, where ageing is an adapted outcome of optimal allocation of limited resources across survival, reproduction, timing, and offspring quality.
He then uses a set of “stress tests” to ask which category fits empirical data:
- Caloric restriction
- Slow-ageing and non-ageing species (turtles/bivalves vs bats/birds, hydra, planaria, Turritopsis)
- Mortality-rate plateaus
- Enormous lifespan of eusocial queens vs workers
- High malleability of lifespan within and between species
His conclusion:
- Mechanistic and weakening-selection theories are each falsified by multiple lines of evidence.
- Optimisation theories are the only class that can accommodate all of these phenomena, often also predicting subtle “nuances” (e.g., why guppies can evolve longer lifespan under higher predation; why non-soma organisms may not age).
A key prediction he emphasises:
If lifespan has been optimised, then mutations that extend lifespan should generally reduce fitness in natural conditions.
He argues that “longevity for free” in C. elegans (daf-2, etc.) is largely an artefact of ignoring post-reproductive parental investment (e.g., intestinal autophagy → yolk → offspring food) and that competition experiments under more natural conditions already show strong fitness costs for daf-2 mutants.
Finally, he suggests that optimisation theory implies:
- Lifespan should be relatively easy to manipulate (lots of plasticity).
- The geroscience hypothesis (modulating ageing modulates multi-morbidity) is likely correct.
- Interventions that tap existing evolved switches (e.g., nutrient sensing, partial reprogramming) should be especially powerful.
2. What the paper actually does, a bit more systematically
2.1 Fitness framing and “how vs why”
- He decomposes fitness into four components:
(1) survival to reproduction,
(2) number of descendants (including inclusive fitness),
(3) generation time,
(4) “quality” of descendants (their own future fitness, shaped by parental investment). - Ageing is defined as age-related decline in any component of this equation.
- He stresses that the standard “survival vs fecundity” trade-off is too crude; timing and quality matter, and can dominate.
2.2 Three categories of theories
-
Mechanistic
- Rate-of-living, oxidative damage, hyperfunction: they posit constraints or programs but either ignore selection or treat ageing as something selection “cannot fix”.
-
Weakening force of selection
- Medawar’s mutation accumulation; Levine’s “old age is out of distribution for evolution”. Selection becomes so weak at late ages that deleterious late-acting variants drift in.
-
Optimisation / disposable soma
- Ageing arises because, under ecological constraints, an optimal allocation of limited energy to maintenance vs reproduction vs timing vs quality will usually underspend on repair, so damage accumulates.
Antagonistic pleiotropy is treated as a mechanism that can sit inside any of these three categories, depending on why the late-life harm isn’t removed (mechanistic linkage, weak selection, or optimal trade-off).
2.3 Empirical “scorecard”
He walks through five phenomena (see Fig. 3) and asks what each class predicts vs what we see.
-
Caloric restriction (CR):
- Mechanistic: CR should slow metabolism/damage → longer life. But CR often doesn’t reduce metabolic rate per mass and may increase it; exercise would then be predicted to shorten life, which is wrong.
- Weakening-selection: CR has no obvious link to the force of selection; also mutation-accumulation logic predicts lifespan shouldn’t be extendable by single levers (too many mutations), which is falsified by CR, single-gene mutants, drugs, and ~10× lifespan extension in some worms.
- Optimisation: famine → shift resources from reproduction to somatic maintenance to “wait out” bad conditions; predicts exactly that mild starvation and many stresses extend lifespan, and that CR effects should be stronger in short-lived species. Matches data and the involvement of nutrient-sensing / diapause genes (daf-2, age-1, mTORC1).
-
Slow-ageing & non-ageing organisms:
- Oxidative-damage style mechanistic theories struggle with fast but long-lived species (bats, birds).
- Weakening-selection: low extrinsic mortality → slower ageing works reasonably, but fails when increased extrinsic mortality produces longer lifespan (predation-intensified guppy lines).
- Optimisation: can handle rockfish with 11 vs >200 year lifespans, turtles vs hummingbirds, non-ageing/immortal-like hydra, planaria, Turritopsis by noting that organisms without a clear soma or that can revert to a germ-like state can avoid ageing altogether; ageing is expected only when survival and reproduction can be decoupled for optimisation.
-
Mortality-rate plateaus:
- Data: many species show late-life plateaus at ~20% hazard per unit time, not an exponential blow-up to 100%.
- Weakening-selection: Hamilton’s formalisation predicts ever-increasing mortality → 100% at sufficient ages; plateaus at low levels are a direct contradiction. Proposed fixes (heterogeneity, modified models) require implausible assumptions or introduce new mismatches.
- Optimisation: Abrams & Ludwig show disposable-soma maths allows plateaus or even declines, but rules out unbounded increase. So a diversity of late-life shapes is expected.
-
Eusocial queens:
- Mechanistic theories predict high function/high metabolism → faster ageing; yet eusocial queens combine huge reproductive output with very long lifespans, often >10× workers with same genome.
- Evolutionary/optimisation view: queens are highly protected from extrinsic mortality; optimising fitness predicts heavy investment in their maintenance. Nuances match: monogynous queens live longer than polygynous; independently founding queens live longer than dependently founding queens – precisely where theory predicts greater investment in queen survival is favoured.
-
Lifespan malleability:
- Between species (rockfish), between castes (queens vs workers), and within species via environment/genetics/drugs (CR, single genes, interventions).
- Weakening-selection theories explicitly claim lifespan should not be very manipulable.
- Optimisation: lifespan is an evolvable, regulated trait; organisms should have levers for plasticity → aligns with the observed breadth of lifespan changes.
2.4 The “key prediction”: fitness cost of longevity
-
In optimisation theories, wild populations sit near a local fitness maximum; any mutation that moves you off that optimum – including lifespan-extending alleles – should usually reduce long-run fitness in natural conditions.
-
This is opposite to:
- Mutation accumulation (longevity mutations would be beneficial but unavailable due to weak selection).
- “Ageing = dysfunction” mechanistic narratives (which imply eliminating ageing would improve fitness).
He then tackles the apparent counter-example: daf-2 worms with extended lifespan but unchanged fecundity and timing when you start RNAi post-reproduction (“longevity for free”).
To argue that this doesn’t really refute optimisation theory, he:
-
Uses Pacific salmon: semelparity and death after spawning make sense when carcasses feed offspring in nutrient-poor rivers; in nutrient-rich environments (Atlantic salmon, estuarine Pacific salmon) you get iteroparity. So “death” itself can be part of offspring quality investment.
-
Applies the same logic to C. elegans:
- Post-reproductive worms undergo intestinal autophagy to make yolk that fills the body; larvae feed on this. This is essentially semelparous parental sacrifice akin to salmon.
- Experiments claiming “no trade-off” ignore this contribution to the quality term in the fitness equation.
He notes we don’t yet have clean multi-generation, natural-condition measurements of long-run growth rates for longevity mutants, but mixed-culture competition experiments already show daf-2 mutants have ~35% of wild-type fitness even under lab conditions, and fare worse in more natural (“scarce food”) environments.
3. Novelty / what’s distinctive
Most of the building blocks (disposable soma, critiques of oxidative damage, Hamiltonian forces of selection, CR theory, guppy experiments, eusocial queens, non-ageing cnidarians) are well-known individually. What feels new or at least distinctive here is:
-
A very clean 3-way classification of ageing theories
- He forces every ultimate theory into one of: mechanistic / weakening-selection / optimisation, and insists they are mutually exclusive and exhaustive at the logical level.
- He explicitly “disassembles” antagonistic pleiotropy and makes you specify which category it really collapses into. That’s tidy and pedagogically useful.
-
Using a targeted empirical “stress test” panel
- The paper isn’t a general narrative review but a structured falsification exercise using five specific phenomena (CR, unusual species, mortality plateaus, eusocial queens, lifespan malleability) presented side-by-side in Fig. 3 as a scorecard.
- The emphasis that mutation-accumulation style theories are formally falsified by mortality plateaus at low hazard is stated more bluntly than in many reviews.
-
Strong, global claim that only optimisation theories survive
- He goes beyond “optimisation fits well” to “it’s the only category consistent with all observations; others effectively require us to believe selection mysteriously doesn’t operate on ageing despite acting on everything else”.
-
Re-framing the “longevity for free” debate with salmon + worm semelparity
- The parallel between semelparous Pacific salmon and late-life C. elegans intestinal autophagy, both as offspring-feeding parental sacrifice, and linking that directly to the “quality” term in the fitness equation, is a neat integrative move.
- He uses this to argue that existing claims of “no trade-off” simply omitted a major fitness component, and that the limited competition experiments we do have actually support the optimisation prediction.
-
Explicit translational / geroscience implications drawn from evolutionary logic
- Lifespan is predicted to be especially manipulable because it’s a regulated trait with pre-existing plasticity.
- Interventions that mimic historically recurrent stresses (famine, thermal stress, hypoxia, infection) are predicted to tap evolved plasticity levers, whereas “out-of-distribution” interventions (e.g., germline ablation) are inherently harder to predict.
- He explicitly links this to the geroscience hypothesis and to partial reprogramming as a way of harnessing germline-like rejuvenation machinery.
So: conceptually this is more a synthesis and reframing than a brand-new model, but it’s unusually sharp in its classification, falsification framing, and in the insistence that ageing is an optimisation outcome, not a failure of selection.
4. Critique
4.1 Conceptual / theoretical issues
-
“Mechanistic” vs “evolutionary” is a bit of a straw man.
- The way mechanistic theories are defined here (as effectively non-evolutionary) doesn’t quite match how many modern proponents see them. Hyperfunction theory and much work on damage accumulation are explicitly embedded in life-history and optimisation frameworks; they’re mechanisms within an evolutionary theory, not alternatives to it. Lumping them as “mechanistic” and then criticising them for not addressing selection risks attacking a simplified version of the field.
-
Mutation-accumulation models are presented in a rather rigid form.
- He treats Hamilton/Medawar-style models as predicting an inexorable exponential mortality increase to 100% at some extreme age, then uses plateaus to “falsify” the whole class. But there is significant theoretical work extending mutation-accumulation and mixed models (MA + antagonistic pleiotropy, heterogeneous frailty, state-dependent selection) that can produce plateaus or deceleration without the extreme parameter pathologies he emphasises. He mentions some of these but dismisses them quickly.
-
“Only optimisation theories fit all data” is stronger than the evidence really supports.
- Most of his empirical arguments show that pure mutation-accumulation or pure naive mechanistic theories don’t work in isolation. But in practice, many evolutionary gerontologists treat ageing as a mix: early-life optimisation + late-life weakening of selection + specific mechanistic constraints.
- The paper doesn’t really engage with hybrid models; it mostly compares extreme “pure” versions and then crowns optimisation the winner.
-
Fitness measure is idealised and not actually computed.
- The 4-component fitness decomposition is conceptually useful, but the paper never actually calculates long-run growth rates under alternative life-history schedules; it relies instead on qualitative argument. In places (e.g., daf-2 fitness) the step from “there’s a likely cost to offspring quality” to “therefore total fitness is lower” is plausible but not shown quantitatively.
4.2 Empirical / data-interpretation issues
-
Reliance on a relatively small empirical base for the “fitness cost of longevity” claim.
- The central prediction (longevity mutations reduce fitness) is supported mainly by mixed-culture daf-2 worm experiments in particular lab setups. That’s important, but very narrow. Other systems (e.g., long-lived Drosophila lines, yeast) could offer different patterns, and the paper doesn’t systematically review them.
- He explicitly notes that fully natural conditions and many-generation experiments haven’t been done; yet the conclusion is stated strongly.
-
CR and lifespan malleability discussion is quite intervention-heavy but somewhat species-light.
- The paper leans heavily on short-lived model organisms, with only brief acknowledgement that CR in long-lived primates shows mixed or modest benefits. That’s consistent with his optimisation argument (CR levers are stronger in short-lived species), but a more balanced treatment would grapple more with the human/higher-mammal data.
-
Mortality plateaus as a “fundamental falsification” might be overstated.
- Late-life hazard estimation is notoriously sensitive to cohort heterogeneity, censoring, and pooling; not all species show clear plateaus, and the level (~20%) and universality he cites are still debated.
- A more cautious statement would be that classic Hamilton-Medawar models in their simplest form are inconsistent with several well-documented mortality patterns, rather than that weakening-selection logic per se is disproven.
-
Non-ageing species are still contentious.
- Hydras, planaria, and Turritopsis are often used as “non-ageing/immortal” examples, but the actual data on age-specific mortality and replication in natural settings is thin and messy. The paper leans on them as strong cases without really detailing the experimental limitations.
4.3 Scope and omissions
-
Limited engagement with alternative evolutionary frameworks.
-
There is little discussion of:
- Reliability theory (ageing as failure of redundant systems),
- State-dependent life-history theory,
- Multi-level selection (e.g., colony-level selection in eusocials),
- Explicit demographic genetics models that mix optimisation and mutation.
-
These could either strengthen his optimisation framing (by showing they’re compatible) or complicate the neat 3-way partition.
-
-
Mechanistic detail is deliberately thin.
- That’s partly by design (the paper is about “why”, not “how”), but it also means some claims—like “ageing is quasi-regulated via sub-maximal repair”—stay at a conceptual level. For readers looking for tight links to specific hallmarks or pathway data, the paper is suggestive rather than demonstrative.
4.4 Translational / geroscience angle and conflicts of interest
- The paper ends on a distinctly optimistic translational note: lifespan is “relatively easy to manipulate,” geroscience is likely correct, and tapping into built-in plasticity mechanisms is the right path.
- That is plausible, but wording like “should be relatively easy” is stronger than many would be comfortable with given the translational failures we’ve already seen (e.g., antioxidants, mixed CR data, mTOR inhibitors side-effect profiles).
- The author transparently discloses that he is employed by Life Biosciences and sits on several longevity-related boards, which reasonably raises the bar for sceptical readers when it comes to claims about how straightforward lifespan modulation will be in humans.
5. Bottom line
As a synthesis, it’s:
- Conceptually clean and well-argued on the ultimate side.
- A strong, pro-optimisation, pro-disposable-soma manifesto.
- A useful corrective against both naive mechanistic “ageing is just damage” narratives and simplistic “selection just stops caring” stories.
But:
- It downplays hybrid theories and nuanced mutation-accumulation work,
- Overstates the degree to which alternative classes are “falsified,”
- And leans a bit harder than the data warrants on the ease and generality of lifespan extension and on the universality of fitness costs for longevity mutations.
If you’d like, next step could be to map this against, say, Omholt & Kirkwood 2021, or recent hyperfunction papers, and see exactly where their predictions genuinely diverge.