Two reviews, ten years apart, from the same five-author core in the same journal. The 2013 paper enumerated nine hallmarks. The 2023 follow-up listed twelve. No cell changed in the interval; what changed was the evidence and the bar for admission. That distance between a list and a mechanism is what the hallmarks of ageing mean in practice, and it is the point most popular coverage skips.
This guide is written for readers who have to judge a paper rather than a headline. It sets out what the framework claims, which results have survived independent replication, which remain associations, where synthetic and analytical chemistry actually enter, and what the field still cannot measure. It contains no health advice and describes no human use of any compound.
What the hallmarks of ageing mean, and what they do not claim
A hallmark is a classification device, not a cause. The 2023 review states its admission rules openly: a hallmark should manifest with age, should accelerate ageing when experimentally accentuated, and should offer the opportunity to decelerate, stop or reverse ageing when therapeutically targeted. Those three premises are a filter for candidates. They are not evidence that any single candidate drives the process.
The 2013 paper was explicitly provisional in its own wording. It described nine tentative hallmarks as common denominators of ageing across organisms, with emphasis on mammals, and it named the unsolved problem in its own abstract: dissecting the interconnectedness between the candidate hallmarks and their relative contributions. A decade later that is still the central research question, not a settled matter.
Two consequences follow for anyone reading the literature. A hallmark is not automatically a target, because a process that rises with age may be a repair response rather than the damage itself. And a result in yeast, worms or flies does not transfer to mammals by default; the 2013 review was framed around conserved pathways precisely because conservation has to be demonstrated case by case.
From nine hallmarks in 2013 to twelve in 2023
The 2013 list was: genomic instability, telomere attrition, epigenetic alterations, loss of proteostasis, deregulated nutrient sensing, mitochondrial dysfunction, cellular senescence, stem cell exhaustion, and altered intercellular communication.
The 2023 list keeps all nine and adds three: disabled macroautophagy, chronic inflammation, and dysbiosis. The same review also connects the hallmarks to a parallel set of proposed hallmarks of health, which include spatial compartmentalisation, maintenance of homeostasis, and adequate responses to stress.
| Aspect | Hallmarks of Aging, 2013 | Hallmarks of aging, 2023 |
|---|---|---|
| Number of hallmarks | Nine | Twelve |
| Added categories | None | Disabled macroautophagy, chronic inflammation, dysbiosis |
| Stated admission criteria | Common denominators across organisms, described as tentative | Three explicit premises: age-associated manifestation, acceleration when accentuated, deceleration when targeted |
| Stated open problem | Interconnectedness and relative contributions of the hallmarks | Hallmarks described as interconnected and linked to proposed hallmarks of health |
| Journal and year | Cell, 2013 | Cell, 2023 |
The honest summary of that change is not that three new mechanisms were discovered. It is that three processes accumulated enough experimental support to clear a bar the authors had written down. The framework grew because the evidence grew, which also means it can grow again.
Which ageing results failed to replicate, and why
The most instructive failure in modern ageing biology is the sirtuin story. Overexpression of sirtuins, a family of nicotinamide adenine dinucleotide dependent protein deacetylases, had been reported to increase lifespan in budding yeast, in Caenorhabditis elegans and in Drosophila melanogaster. A 2011 re-examination in Nature found that standardising the genetic background and using appropriate controls abolished the apparent effects in both the worm and the fly.
The detail matters more than the headline. In C. elegans, outcrossing a high-expressing line abrogated the longevity increase without abrogating the overexpression itself; longevity co-segregated instead with a second-site mutation affecting sensory neurons. In Drosophila, a strain with ubiquitous overexpression was long-lived relative to wild-type controls, as previously reported, but not relative to the appropriate transgenic controls. The same study found that dietary restriction extended fly lifespan independently of the sirtuin gene. The authors did not conclude that sirtuins are irrelevant to lifespan; they concluded that the previously reported effects were not robust to genetic background and to the mutagenic effects of transgene insertion.
That is the general lesson for reading this field. When an effect disappears under better controls, the informative question is which control was missing, not whether the molecule is currently fashionable.
Why lifespan studies use genetically heterogeneous mice
The counterexample is rapamycin. In 2009 a multi-site study reported in Nature that rapamycin fed from 600 days of age extended median and maximal lifespan in male and female mice. On the basis of age at ninety per cent mortality, the increase was fourteen per cent for females and nine per cent for males. The effect was seen at three independent test sites, in genetically heterogeneous mice chosen to avoid genotype-specific effects on disease susceptibility.
A second report, published in 2011 by an overlapping group and funded under the same National Institute on Aging awards, gave rapamycin in food from the age of nine months and found median survival extended by an average of ten per cent in males and eighteen per cent in females, with significant increases including maximum life span at each of three test sites. In that study resveratrol and simvastatin had no significant effect on survival in either sex.
Three design features carry that evidence: genetic heterogeneity, three independent laboratories, and survival as the endpoint. The sirtuin episode failed on the first of them, which is why the design, and not the compound, is the part worth copying. None of this is a statement about human ageing, and a mouse survival curve is not a licence to extrapolate.
Where chemistry actually enters the picture
The molecules used as probes in this field have almost nothing structurally in common. Rapamycin, also called sirolimus, is a macrolide of molecular formula C51H79NO13 with a molecular weight near 914, listed in PubChem as CID 5284616 with CAS Registry Number 53123-88-9. Spermidine is a triamine of formula C7H19N3 and molecular weight near 145, PubChem CID 1102. Metformin is a biguanide of formula C4H11N5 and molecular weight near 129, PubChem CID 4091.
A fifty-one-carbon macrocyclic lactone and a seven-carbon polyamine are not the same analytical problem. The first is large, highly substituted and rich in stereocentres. The second is small, strongly basic, poorly retained on conventional reversed-phase columns and effectively invisible to ultraviolet detection without derivatisation. The hallmark framework groups such compounds by the pathway they perturb, never by their chemistry, and that mismatch is a frequent source of confusion in secondary coverage.
The practical consequence is unglamorous and decisive: a biological result is only as trustworthy as the identity and purity of the material that produced it. If a probe compound carries an unassigned impurity, the activity reported may belong to the impurity rather than to the named molecule. Characterisation data, not a catalogue name, is what makes a research chemical usable as evidence, which is also why our terms of supply treat every item as laboratory material and nothing else.
Why epigenetic clocks are correlation, not causation
The multi-tissue epigenetic clock published in 2013 was developed using about eight thousand samples from eighty-two DNA methylation array datasets, encompassing fifty-one healthy tissues and cell types, and it rests on 353 CpG sites. Its reported properties are striking: estimated age close to zero for embryonic and induced pluripotent stem cells, correlation with cell passage number, a highly heritable measure of age acceleration, and applicability to chimpanzee tissues. Across six thousand cancer samples from thirty-two datasets, all twenty cancer types considered showed significant age acceleration, averaging thirty-six years.
Here is the limit that headlines drop. A clock trained to predict chronological age measures the thing it was trained on. Deviation from that prediction, usually called age acceleration, is an association with outcomes unless an experiment demonstrates otherwise. The author of the 2013 clock proposed that it measures the cumulative effect of an epigenetic maintenance system, which is a hypothesis about what the marker reflects, not a demonstration that moving the marker moves the biology. The difference between chronological age and biological age, as used in this literature, is a modelling result rather than a directly measured quantity.
Do longer telomeres protect against disease?
Telomere attrition appears in both hallmark lists, so the intuitive reading is that longer telomeres must be better. A Mendelian randomisation study published in 2017 in JAMA Oncology tested that intuition using germline genetic variants as instrumental variables, with summary data available for 35 cancers and 48 non-neoplastic diseases, corresponding to 420 081 cases and 1 093 105 controls.
Increased telomere length due to germline genetic variation was generally associated with increased risk of site-specific cancers. The strongest odds ratios per standard deviation of genetically increased length were reported for glioma at 5.27, serous low-malignant-potential ovarian cancer at 4.35, lung adenocarcinoma at 3.19 and neuroblastoma at 2.98. For psychiatric, autoimmune, inflammatory and diabetic conditions there was generally little evidence of association, with exceptions running the other way: coronary heart disease at 0.78, abdominal aortic aneurysm at 0.63, coeliac disease at 0.42 and interstitial lung disease at 0.09.
The authors concluded that longer telomeres likely increase risk for several cancers while reducing risk for some non-neoplastic diseases, including cardiovascular disease. A hallmark can therefore point in opposite directions depending on the outcome examined, which is the clearest single argument against reading the hallmark list as a list of quantities to maximise or minimise.
Is cellular senescence a cause or a consequence of ageing?
Senescence is the hallmark with the cleanest causal experiment behind it. A 2011 Nature study used a transgene called INK-ATTAC to eliminate p16Ink4a-positive cells on demand in a BubR1 progeroid mouse background. Life-long removal of those cells delayed the onset of age-related changes in adipose tissue, skeletal muscle and eye, and late-life clearance attenuated the progression of disorders that were already established. The authors read this as evidence that cellular senescence is causally implicated in generating age-related phenotypes in that model.
The human evidence is far thinner. A first-in-human, open-label pilot published in 2019 enrolled fourteen participants with idiopathic pulmonary fibrosis. Its primary endpoints were retention and completion of planned clinical assessments, not efficacy. Physical function measured as six-minute walk distance, four-metre gait speed and chair-stands time improved significantly; pulmonary function, clinical chemistries and reported health were unchanged; one serious adverse event was reported; and effects on circulating senescence-associated secretory phenotype factors were inconclusive. There was no control group, and the authors framed the study as supporting feasibility and warranting larger randomised controlled trials.
A causal result in a progeroid mouse and a feasibility pilot in fourteen patients are not the same kind of evidence. The distance between them is where most of the overstatement in this field lives.
What counts as a validated biomarker of ageing
Nothing yet, in the sense of a consensus-qualified measure. A 2023 framework paper in Cell on biomarkers of ageing states the obstacle plainly: the current lack of standards and consensus on the properties of a reliable ageing biomarker hinders their further development and validation for clinical applications. The paper advances terminology, classification and potential clinical use cases, discusses validation steps, and marks the remaining challenges as areas needing future research.
That is the state of play worth holding onto when reading any longevity claim. A 2019 Nature review describes ageing research as entering a new era with medical, commercial and societal implications, which is a statement about momentum rather than about proven interventions. Momentum and validation are different things, and only one of them belongs in a specification.
Frequently asked questions
How many hallmarks of ageing are there now?
Twelve, according to the 2023 update published in Cell. The 2013 original proposed nine: genomic instability, telomere attrition, epigenetic alterations, loss of proteostasis, deregulated nutrient sensing, mitochondrial dysfunction, cellular senescence, stem cell exhaustion and altered intercellular communication. The 2023 revision retains all nine and adds disabled macroautophagy, chronic inflammation and dysbiosis as separate hallmarks.
What is the difference between the 2013 and 2023 hallmarks of aging?
The 2023 version adds three hallmarks and states explicit admission criteria: a candidate should manifest with age, should accelerate ageing when experimentally accentuated, and should allow ageing to be decelerated, stopped or reversed when therapeutically targeted. The 2013 paper called its nine hallmarks tentative and identified their interconnectedness and relative contributions as the major unresolved challenge.
Did the sirtuin lifespan results replicate?
Not under standardised conditions. A 2011 re-examination published in Nature found that standardising genetic background and using appropriate transgenic controls abolished the reported lifespan extension from sirtuin overexpression in both C. elegans and Drosophila. In the worm, longevity co-segregated with a second-site mutation affecting sensory neurons rather than with the overexpression itself.
Does rapamycin extend lifespan in mice?
Yes, in genetically heterogeneous mice, at three independent test sites, in two separate reports. Fed from 600 days of age, rapamycin raised age at ninety per cent mortality by fourteen per cent in females and nine per cent in males. Given in food from nine months of age, median survival rose by an average of eighteen per cent in females and ten per cent in males.
Are epigenetic clocks a measure of biological age?
They are predictors trained on chronological age, so deviations from prediction are associations until an experiment shows otherwise. The 2013 multi-tissue clock uses 353 CpG sites, was developed from roughly eight thousand samples across fifty-one healthy tissues and cell types, and reported average age acceleration of thirty-six years in cancer tissue. Prediction accuracy is not a demonstration of mechanism.
Do longer telomeres protect against disease?
Not uniformly. A 2017 Mendelian randomisation analysis covering 35 cancers and 48 non-neoplastic diseases found that genetically increased telomere length was generally associated with higher risk of site-specific cancers, including glioma and lung adenocarcinoma, while being associated with lower risk of coronary heart disease, abdominal aortic aneurysm and interstitial lung disease. Direction of effect depends on the outcome.
Is there a validated biomarker of ageing?
No consensus-qualified biomarker exists. A 2023 framework paper in Cell states that the lack of agreed standards and consensus on what makes an ageing biomarker reliable is itself the obstacle to further development and validation for clinical applications. That paper proposes terminology, classification and validation steps rather than endorsing any specific existing measure as validated.
Why do lifespan studies use genetically heterogeneous mice?
To stop a single genotype from dictating the result. Inbred strains carry strain-specific disease susceptibilities that can masquerade as an ageing effect, and transgene insertion sites can do the same, as the sirtuin re-examination showed. Both multi-site rapamycin reports used genetically heterogeneous mice, chosen explicitly to avoid genotype-specific effects on disease susceptibility, at three independent test sites.
References
- Lopez-Otin C, Blasco MA, Partridge L, Serrano M, Kroemer G. The Hallmarks of Aging. Cell, 2013;153
- Lopez-Otin C, Blasco MA, Partridge L, Serrano M, Kroemer G. Hallmarks of aging: An expanding universe. Cell, 2023;186
- Burnett C, et al. Absence of effects of Sir2 overexpression on lifespan in C. elegans and Drosophila. Nature, 2011;477
- Harrison DE, et al. Rapamycin fed late in life extends lifespan in genetically heterogeneous mice. Nature, 2009;460
- Miller RA, et al. Rapamycin, But Not Resveratrol or Simvastatin, Extends Life Span of Genetically Heterogeneous Mice. Journal of Gerontology Series A, 2011;66A
- Horvath S. DNA methylation age of human tissues and cell types. Genome Biology, 2013;14
- Haycock PC, et al. Association Between Telomere Length and Risk of Cancer and Non-Neoplastic Diseases: A Mendelian Randomization Study. JAMA Oncology, 2017;3
- Baker DJ, et al. Clearance of p16Ink4a-positive senescent cells delays ageing-associated disorders. Nature, 2011;479
- Justice JN, et al. Senolytics in idiopathic pulmonary fibrosis: results from a first-in-human, open-label, pilot study. EBioMedicine, 2019;40
- Moqri M, et al. Biomarkers of aging for the identification and evaluation of longevity interventions. Cell, 2023;186
- Campisi J, et al. From discoveries in ageing research to therapeutics for healthy ageing. Nature, 2019;571
- PubChem Compound Summary CID 5284616, sirolimus. National Library of Medicine
Research use only. Nonsensia Lab supplies analytical reference standards for laboratory and research applications. This article is published for scientific and educational purposes. It is not medical advice, it does not describe any use in humans, and nothing in it should be read as a recommendation to administer any substance to a person or animal.
Filed under: Longevity Science
Nonsensia Lab supplies the compounds discussed in this guide as analytical reference standards for laboratory and research use.