Why Do We Age?

A Greenland shark cruising the black water of the Arctic today may have been born before the invention of the telescope. Its flesh is toxic, its metabolism glacial, and by the best radiocarbon estimates from eye-lens proteins, some individuals are pushing four hundred years old. Meanwhile, in a lab dish a few feet from where you might be reading this, a fibroblast cell divides for the fortieth, fiftieth time and simply stops — hunched, swollen, biochemically shouting at its neighbors, refusing to die but refusing to work. Somewhere between the shark and the cell sits every one of us, aging at a rate that feels as inevitable as gravity.

And yet it isn't inevitable, not in the way gravity is. A lobster's mortality risk barely rises with age; a naked mole rat can live thirty years — ten times longer than a similarly sized mouse — while showing almost no increase in death rate for decades. Some jellyfish can reverse their own developmental clock and revert to a juvenile polyp stage, essentially becoming young again. Aging, it turns out, is not a universal law stamped into the physics of matter. It is a biological trait, and biological traits have causes, mechanisms, and — this is the electrifying part — potential off-switches.

The strange thing is that after more than a century of gerontology, and after one of the most productive research explosions in biology's recent history, we still cannot say with confidence why organisms age at all. We know a great deal about how — the molecular wreckage, the exhausted stem cells, the tangled proteins. But whether that wreckage is the cause of aging or merely its wake, whether aging is something evolution built on purpose or something evolution simply failed to prevent, and whether the whole process can in principle be reversed rather than merely slowed — these remain open, contested, and thrilling questions.

svg[Three lenses on the same decline: program, damage, and trade-off, converging on the same measured mortality curve]

If the settled science were the whole picture, the three-camp disagreement would have been resolved decades ago. It hasn't been, because a stack of stubborn anomalies keeps cutting against every clean version of the story.

The telomere problem. Telomere shortening is real, measurable, and mechanistically well understood — but it cannot be the master cause of aging. Mice have telomeres far longer than humans and constitutively active telomerase in most tissues, yet they age faster than we do, not slower. Naked mole rats also have long telomeres and high telomerase activity but live remarkably long lives with minimal aging-related mortality increase, breaking any simple telomere-length-predicts-lifespan rule across species. Within humans, telomere length correlates only loosely and inconsistently with actual lifespan or health outcomes once other factors are controlled for. Telomere biology matters — for cancer suppression, for certain tissue-specific failures like in telomere-related pulmonary fibrosis — but as a universal aging clock, it fails the cross-species test badly.

The damage-accumulation problem. If aging were simply the sum of stochastic molecular damage, we would expect species with similar metabolic rates, similar oxidative burdens, and similar cell-division counts to age at similar rates. They don't. Naked mole rats have metabolic rates comparable to other rodents their size but live roughly ten times longer than a mouse, with a mortality rate that stays nearly flat well into old age rather than rising the way the Gompertz mortality law (the empirical exponential rise in death rate with age that Benjamin Gompertz described in 1825, and which fits humans and most mammals disturbingly well) predicts. Bowhead whales live over two hundred years with body masses and metabolic demands vastly exceeding a mouse's, and yet accumulate cancer and degenerative disease at far lower rates than expected — a puzzle sometimes called Peto's paradox, since larger animals with more cells and more cell divisions should statistically get more cancer, not less. Something is actively suppressing damage or its consequences in these animals beyond what raw thermodynamics would predict, and we do not fully understand what.

The programmed-aging problem. If aging were a designed program, we'd expect evolution to be able to switch it off easily when circumstances favor long life, and indeed something like this does seem to happen — queen bees and worker bees are genetically identical yet differ in lifespan by an order of magnitude, purely through epigenetic and nutritional programming during development (royal jelly feeding activates different gene expression). This looks tantalizingly like evidence for a controllable "aging program." But programmed theories struggle with a harder logical objection: classical evolutionary theory says that group-level benefits (like population control through limited lifespan) are usually unstable against individual-level selection, because a mutant that lives and reproduces longer than its die-on-schedule neighbors should outcompete them, unless some very specific conditions (like strong kin selection or spatial structure) hold. Programmed aging as an adaptation for the species' benefit remains a minority position among evolutionary biologists for exactly this reason, even though certain semelparous organisms — Pacific salmon, some marsupial mice, many annual plants — show a dramatic, hormonally triggered, unmistakably programmed rapid post-reproductive death, driven by a glucocorticoid stress cascade after spawning. So programmed death clearly can evolve, at least in specific reproductive contexts. Whether ordinary human and mammalian aging is the same kind of phenomenon, just slower and less obviously triggered, is exactly what's unresolved.

The reversibility problem. Perhaps the deepest crack in the whole edifice came from a technique nobody expected to bear on aging at all. In 2006, Shinya Yamanaka showed that adult, fully differentiated cells could be reprogrammed into embryonic-like induced pluripotent stem cells (iPSCs) using just four transcription factors — Oct4, Sox2, Klf4, and c-Myc. Strikingly, this reprogramming resets the epigenetic age of the cell back to near zero, as measured by molecular "epigenetic clocks" (patterns of DNA methylation, first systematized by Steve Horvath in 2013, that track chronological age with remarkable precision across tissues). If aging were purely irreversible entropy — molecular rust that can only accumulate — this should be impossible without literally replacing the damaged molecules one by one. Instead, cellular age, at least as captured by these epigenetic markers, appears to be at least partially an information state that can be reset, not merely a pile of physical wreckage that must be hauled away. This single observation is arguably the most theory-shaking fact in the entire field, and it does not sit comfortably in any of the three classical camps without modification.

Out of this tangle, several serious research programs have emerged, each trying to make the reversibility clue and the cross-species anomalies fit a coherent mechanistic story.

The epigenetic information-theoretic view. David Sinclair and collaborators have proposed that aging is fundamentally a loss of epigenetic information — not damage to the DNA sequence itself, but a gradual scrambling of the chemical marks (DNA methylation patterns, chromatin structure) that tell each cell which genes to express and which to silence, originally established during development. Their "information theory of aging" draws an explicit analogy to a compact disc that gets scratched: the underlying data (genome sequence) is intact, but the read head loses track of where it is. In a striking 2023 study, Sinclair's lab showed that partial reprogramming using only three of Yamanaka's four factors (omitting the oncogenic c-Myc) could restore vision in old mice with damaged optic nerves, and reverse several measures of cellular age in multiple tissues, without erasing the cell's identity entirely. The appeal is obvious: it offers a concrete, testable mechanism and an actual reversal, not just slowing. The trouble is that "loss of information" is a slippery, almost thermodynamic metaphor that needs far more rigorous operational definition — what exactly is the information, how is it measured independent of the epigenetic clocks used to detect its loss (a worry about circularity that critics have raised sharply), and why induced deliberate DNA breaks during development (part of Sinclair's proposed mechanism, where repair machinery gets "distracted" from maintaining epigenetic marks) should be the dominant driver across all tissues and species remains unproven.

The damage-repair engineering view. Aubrey de Grey's SENS (Strategies for Engineered Negligible Senescence) framework takes the opposite philosophical stance: aging is fundamentally an accumulation of several distinct types of cellular and molecular damage — including senescent cell accumulation, extracellular protein cross-linking, mitochondrial mutations, and intracellular/extracellular junk aggregates — that in principle could each be repaired or removed with the right biotechnology, much as a vintage car can be kept running indefinitely through part replacement, without needing to understand every detail of why it decays. The appeal is that it sidesteps deep theoretical disputes and focuses on engineering: you don't need to fully understand aging's ultimate cause to build a senolytic drug that clears senescent cells, and indeed such drugs (dasatinib plus quercetin, fisetin, and others) are already in early human trials, following directly from the mouse senescent-cell-clearance results. The trouble is that this is a strategy, not strictly a theory — it doesn't predict why damage of these particular kinds accumulates at the particular rate it does in humans versus bowhead whales, and critics within mainstream biogerontology, including Hayflick himself before his death in 2024, argued that SENS underestimates how deeply interconnected and systemic the failure modes of aging really are, such that fixing each piece separately might not add up to fixing the whole.

The evolutionary optimization / hyperfunction view. Mikhail Blagosklonny has argued for what he calls the "hyperfunction theory": aging is not damage accumulation at all in the primary sense, but a continuation of developmental growth programs (particularly signaling through the mTOR/nutrient-sensing pathway) past the point where they are useful, becoming excessive and pathological — cells don't shrink and fail so much as overgrow and malfunction, as with age-related hypertrophy, arterial stiffening, or hyperplasia. This is a sophisticated modern descendant of antagonistic pleiotropy, made mechanistically concrete: the same growth signals essential for a young organism's development become "hyperfunctional" once development is complete, since selection never had reason to switch them off cleanly. It gains real support from the fact that inhibiting mTOR signaling with the drug rapamycin robustly extends lifespan in every model organism tested so far, from yeast to mice, and is now in human trials for aging-related outcomes. Its weakness is that hyperfunction and damage accumulation are not so easily separated in practice — excess growth signaling plausibly causes damage (through resource depletion, oxidative stress, and metabolic strain) rather than being an alternative to it, so critics see this less as a rival to the damage view than as a refinement of which damage matters most.

The systemic / network view. A newer and less crystallized position, gaining ground especially from single-cell and multi-omic studies of aging tissues, holds that no single hallmark or mechanism is "the" cause, and that aging is better understood as a systems-level breakdown in the regulatory networks — hormonal, immune, neural, microbial — that maintain homeostasis across the whole organism, with different mechanisms dominating in different tissues and species. Studies of "young blood" parabiosis experiments — where surgically joining the circulation of an old and young mouse partially rejuvenates the old animal's stem cell function and even cognition — suggest that some aging phenotypes are driven by circulating systemic factors, not purely by cell-intrinsic damage. This view is empirically well grounded and humbling, but it is, frankly, more a description of the problem's complexity than a solution to it; it resists the kind of crisp mechanistic story that would let us say confidently what to intervene on.

Unlike some frontier questions in physics, where a single crucial experiment can in principle adjudicate between theories, aging research faces a structural obstacle: the gold-standard outcome measure — does a human live substantially longer and healthier — takes decades to observe directly in the very organism we care most about. Almost everything we know about intervention comes from short-lived model organisms, and the translation gap from worm or mouse to human has repeatedly humbled overconfident claims.

That said, several concrete developments would move the needle hard, and some are already underway.

The single most information-dense test would be large, long-duration human trials of geroprotective drugs with mortality or multi-morbidity endpoints, rather than surrogate biomarkers. The TAME trial (Targeting Aging with Metformin), championed by Nir Barzilai, is designed to test exactly this — whether a cheap, already-approved diabetes drug can delay the onset of multiple age-related diseases simultaneously in non-diabetic older adults, treating aging itself, rather than any single disease, as the target. Its results, still pending full completion, would be a landmark regardless of outcome: a positive result would be the first direct clinical evidence that a specific pharmacological intervention slows human aging as such; a null result would sharpen skepticism about metabolic-pathway theories translating cleanly from mice to people.

A second decisive line would be validated, mechanistically grounded biomarkers of biological age that are shown, prospectively and across independent cohorts, to predict future mortality and function better than chronological age — and, crucially, to move in response to interventions in a way that then predicts real health outcomes, closing the current gap where epigenetic clocks correlate with age but their causal role remains unproven. Efforts like the second- and third-generation epigenetic clocks (GrimAge, DunedinPACE) are attempts at exactly this, incorporating physiological and mortality data rather than just chronological correlation, but the field still lacks full consensus that these clocks measure a causal driver of aging rather than a downstream correlate.

A third would be partial reprogramming in primates and eventually humans, done safely — without triggering the cancer risk that comes from the oncogenic factors in the original Yamanaka cocktail, and without erasing cell identity. Several biotech companies (Altos Labs, Retro Biosciences, and academic groups following Sinclair's and Juan Carlos Izpisúa Belmonte's work) are actively pursuing this, with in vivo partial reprogramming already shown to extend lifespan and reverse some aging markers in progeria and naturally aged mice as of the early 2020s. If partial reprogramming can be shown safely to reverse multiple independent hallmarks of aging simultaneously in a large mammal, and if that reversal demonstrably improves function and extends healthy lifespan without elevating cancer incidence, that would be close to the strongest possible evidence for the "aging as reversible information loss" camp — a genuine paradigm-shifting result, likely achievable, if it is achievable at all, sometime in the 2030s given current trajectories, though this is an optimistic estimate and could easily slip by a decade or more given how often biology has humbled such timelines before.

Finally, a deeper theoretical resolution would come from comparative genomics across extreme-longevity species — bowhead whales, Greenland sharks, naked mole rats, some bristlecone pines and clonal organisms — identifying which specific genetic and molecular adaptations suppress the hallmarks of aging in each, and whether these adaptations converge on shared pathways or are wildly different lineage-specific solutions. If very different species turn out to solve long life through the same handful of mechanisms (enhanced DNA repair, superior proteostasis, tumor-suppressor duplication as documented in elephants' extra copies of the p53 gene), that convergence would be strong evidence for a small number of tractable causal levers. If instead every long-lived species has evolved its own idiosyncratic fix, that would support the systemic-network view that aging has no single universal mechanism to target — a genuinely deflationary but important answer.

It's tempting to frame the aging question purely as a biomedical problem — can we cure the diseases of old age — but that undersells what's really at stake. Nearly two-thirds of all human deaths worldwide are attributable to age-related conditions: cardiovascular disease, cancer, neurodegeneration, and their tangled interactions. Every major chronic disease of late life shares aging itself as its single largest risk factor, far outweighing any individual disease-specific risk factor identified so far. This has led a growing faction within biomedicine, including the U.S. FDA's unprecedented 2016 acceptance of a trial (TAME) designed to treat "aging" as a legitimate drug target rather than any single disease, to argue that treating the aging process directly, rather than picking off diseases one at a time in a lifelong game of whack-a-mole, could be the more efficient path to reducing suffering in later life.

But the stakes reach further, into how we understand life itself. If aging is fundamentally reversible information loss, as the epigenetic camp argues, that would mean biological age is not a fixed, one-way arrow but a state that cells can, at least locally, be moved back along — a discovery with implications far beyond medicine, touching on how we think about identity, continuity, and what a "self" persisting through time actually consists of at the molecular level. It would also intersect with the can-we-build-synthetic-life frontier: if we can precisely specify and reset the epigenetic state that constitutes youthful cellular function, we edge closer to being able to design cellular states from scratch rather than merely nudging existing ones, blurring the line between repair and engineering. And it connects tightly to the genotype-to-phenotype-mapping problem, since understanding why a given genome produces a fast-aging mouse or a nearly non-aging naked mole rat is, at bottom, a question about how genotype unfolds into the dynamic, time-extended phenotype of a whole life course, not just a static body plan.

There is also a sobering ethical and social dimension that the science cannot answer by itself. Extending healthy lifespan at scale would reshape almost every institution built around a life course of roughly seventy to ninety years — pensions, career structures, generational turnover in politics and ideas, population and resource dynamics, the very texture of family life across generations. Whether that prospect is a triumph or a danger depends enormously on how — and for whom — such technology would actually be deployed, a question biology alone cannot settle.

The field is unusually rich in vivid, sometimes clashing personalities, which is itself revealing: this is not a question with a quiet, settled consensus, but a live argument being fought in real time between serious scientists who disagree about fundamentals.

physicist[Elizabeth Blackburn|1948–|Australian-American molecular biologist who, with Carol Greider and Jack Szostak, discovered telomerase, revealing the enzymatic mechanism behind chromosome-end maintenance and earning the 2009 Nobel Prize in Physiology or Medicine]

history 1961 — Hayflick and Moorhead demonstrate the finite replicative limit of normal human cells. 1990 — Telomerase activity linked directly to human cellular senescence and immortalization in cancer cells. 2006 — Yamanaka reprograms adult mouse cells into induced pluripotent stem cells. 2011 — Baker and van Deursen show that clearing senescent cells extends healthy lifespan in mice. 2013 — Horvath publishes the first pan-tissue epigenetic clock for human biological age. 2023 — Partial reprogramming restores vision and reduces several aging markers in aged mice.

Stand back far enough and the picture that emerges is not one of ignorance so much as of a field standing at an unusually rich threshold: we have more real, causally validated levers on lifespan and cellular age than at any point in history, and yet we still lack the unifying theory that would tell us which lever is the lever, the one whose pull would matter most, in a human body, over a human lifetime. That is a genuinely strange position for a mature science to be in — awash in mechanism, short on synthesis.

Perhaps that synthesis will never arrive as a single clean theory, and the honest answer will turn out to be plural: aging is simultaneously damage accumulation, and an evolutionary trade-off that permits and even encourages certain damage-promoting genes, and, in some specific and limited sense, information loss that can be locally reversed without needing to reverse the underlying evolutionary logic that shaped it in the first place. Biology rarely honors the sharp either-or categories that theorists reach for first. But even a plural, hedged answer would represent enormous progress over where the field stood even twenty years ago, when senescence was still widely viewed as a passive endpoint rather than an active, druggable process, and when the very idea of reversing cellular age would have sounded like science fiction rather than a replicated laboratory result.

What remains genuinely, thrillingly open:

• Is there a single rate-limiting variable that determines aging speed across species, or does every long-lived lineage solve the problem in its own idiosyncratic way?
• Can partial cellular reprogramming be made safe and precise enough to reverse aging in whole mammals without unleashing cancer, and does that reversal actually extend healthy lifespan rather than just resetting biomarkers?
• Is human aging closer to the "programmed" death of a spawning salmon than we like to admit, or fundamentally a different, unprogrammed kind of decline?
• Will slowing or partially reversing aging in humans reveal new failure modes we've never had the lifespan to encounter — new diseases of a longer life we cannot yet see coming?
• And if aging turns out to be substantially reversible, what does that do to our sense of a life as a story with a natural arc, rather than a state that can, in principle, be edited and rewound?

Ask the tutor: If cellular age can be partly reset by reprogramming a cell's epigenetic marks, does that mean "aging" and "development" are, at bottom, the same kind of biological process running in opposite directions — and if so, what would it take to run the whole process in reverse, safely, in a living human being?