Longevity Weekly Review 2026-09-09

Week In Review

This was the week biological-age clocks stopped being a side measurement and became the result itself. Three separate reports — from a pharmaceutical company, an academic sleep lab, and a small-cap gene therapy firm — used clock readouts as the headline endpoint rather than as supporting color. Insilico Medicine reported that its AI-designed fibrosis drug rentosertib moved six proteomic aging clocks in a completed phase 2a trial, in Integration of proteomic aging clocks in a phase 2a clinical trial supports simultaneous geroprotective assessment. A UCLA team reported that cognitive behavioral therapy for insomnia slowed the DunedinPACE pace-of-aging measure in a randomized trial, in Cognitive behavioural therapy for insomnia and epigenetic ageing. And Genflow Biosciences announced that its SIRT6 gene therapy trial in aged beagles met its primary endpoint — an endpoint defined as reduction in biological age on a methylation clock. None of these is proof that the underlying aging process was altered. All three are evidence that the field has converged on a common currency for reporting short-term results, which is the practical precondition for trials that do not take thirty years.

The week supplied its own corrective. A Science study following 335 women with repeated blood transcriptomics and metabolomics over as much as eight years found that individual molecular trajectories frequently run opposite to the population-average direction. More than 5,000 genes and 181 metabolites shifted over a median of six years, but the shape of that shift was personal, modulated by genetics, circadian and seasonal rhythms, cell-type composition, and environmental exposures. A clock trained on population averages tells you where a cohort is drifting. It does not necessarily tell you where any one person is going. Read against the three clock-endpoint results above, this is the most important methodological caution the field received this week.

Underneath the measurement story sat a quieter mechanistic theme: mitochondria as garbage, and who takes it out. In Science Immunology, researchers showed that the number of macrophages resident in a tissue scales with that tissue’s mitochondrial activity — heart, skeletal muscle, and brown fat maintain large cleanup crews precisely because they burn through mitochondria fastest, and blocking the cleanup degrades each organ’s function within its own idiom. In Science Advances, a Salk team traced how a bout of mitochondrial superoxide stress confined to embryonic development leaves a durable epigenetic mark that protects the adult heart, with citrate as the messenger. Waste handling and stress signaling turn out to be two faces of the same organelle.

The remaining items point at levers with real human relevance. Two are endocrine: late-life semaglutide extended lifespan in aged female mice in Nature, and disrupting growth hormone signaling at 12 months of age — well past development — extended lifespan in both sexes in Aging Cell. Both undercut the assumption that geroprotective interventions must start early. One is epidemiological: twenty years of SWAN data suggest menopausal hormone therapy begun during the transition is associated with lower cardiovascular risk, a finding that continues the long rehabilitation of a therapy the field over-corrected against. And one relocates a neurodegenerative process entirely: WashU researchers found that the T cells that damage neurons in tauopathy receive their instructions in lymph nodes in the neck, not in the brain — which puts the intervention point outside the blood-brain barrier.

Items

Semaglutide Started Late in Life Extends Lifespan in Female Mice

The GLP-1 receptor agonists have spent three years accumulating evidence that they delay individual age-related diseases. This week’s paper in Nature asks the harder question: do they slow physiological aging itself? Researchers gave semaglutide to 20-month-old female C57BL/6 mice — animals already well into old age, roughly analogous to a human in their sixties — for three months, then continued treatment in a survival cohort.

The treated animals showed improved physiological function across multiple domains, attenuation of several recognized hallmarks of aging, and modulation of nutrient-sensing pathways and conserved genetic regulators of aging. Continued treatment extended lifespan. The Lifespan Research Institute’s summary of the work put the median lifespan increase at more than 12 percent in females.

The comparison the authors drew is the interesting part. Caloric restriction is the reference standard for lifespan extension in rodents, and semaglutide was benchmarked against it directly. The drug matched caloric restriction’s benefits broadly and, in cognition and glucose control, appeared to exceed the caloric-restriction baseline. That is not a claim that a weekly injection is better than eating less; it is a claim that the two act through overlapping nutrient-sensing machinery and that the pharmacological route reaches some endpoints the dietary route does not.

Two caveats travel with the result. The study was conducted in female mice, and sex-specific responses to geroprotective interventions have been a persistent feature of this literature — the growth hormone work covered below found effects in both sexes, but many interventions do not. And a three-month intervention starting at 20 months is a narrow window; whether earlier or intermittent dosing would do better is unknown.

What makes this consequential is the installed base. GLP-1 agonists are already taken by a large fraction of the population for reasons unrelated to aging. If the mouse result reflects something real about mammalian physiology, a natural experiment is already running at scale, and the epidemiology will arrive faster than any purpose-built trial could deliver it.

Source: Nature

Eight Years of Blood Chemistry Show Aging Does Not Follow One Script

Most of what the field knows about how molecules change with age comes from cross-sectional studies: sample many people of different ages once, and infer a trajectory from the differences between them. This Science paper did the harder thing. It profiled whole-blood gene expression and 1,197 serum metabolites at three or more time points, over as much as eight years, in 335 women — following individuals rather than reconstructing an average from strangers.

Over a median of six years, more than 5,000 genes and 181 metabolites changed measurably, many of them associated with cardiometabolic and neurodegenerative disease. That much confirms the population picture. The departure is what happened at the individual level: participants displayed distinct longitudinal trajectories that were sometimes opposite in direction to the population-level trend. A molecule rising with age across the cohort could be falling steadily within a given person.

The authors traced this heterogeneity to identifiable sources rather than leaving it as noise. Cell type composition, host genetics, circadian and seasonal rhythms, environmental exposures including PFAS, and cross-omic connectivity between the transcriptome and metabolome all shaped individual paths. Environmental exposures in particular left what the paper describes as detectable, time-sensitive molecular signatures — meaning the timing of an exposure, not just its presence, is written into the data.

One consistent population-level finding did emerge: a complex remodeling of the immune system, with adaptive immune function declining while innate immune activity was sustained. This matches the direction of travel in the immunosenescence literature and gives the study a stable anchor against which the individual variation can be read.

The implication for the rest of this week’s items is direct. If aging biomarkers are going to serve as trial endpoints, the relevant question is not only whether a clock moves in a treated group, but whether it moves in a given participant relative to their own prior trajectory. The authors frame this as a precision medicine problem, and they are right, but it is equally a clinical trial design problem.

Source: Science

An AI-Designed Drug Moves Six Proteomic Aging Clocks in a Completed Trial

Rentosertib is an experimental treatment for idiopathic pulmonary fibrosis. Its provenance is unusual: Insilico Medicine used its target-discovery platform to nominate TNIK as a target implicated in both fibrosis and aging biology, scored that target against the hallmarks of aging, and then used its generative chemistry system to design the molecule. Both ends of the pipeline — the target and the compound — came out of the AI stack rather than from repurposing an existing drug.

This week’s Nature Biotechnology paper reports a retrospective analysis of the completed phase 2a trial, applying six distinct proteomic aging clocks to patient samples. The clocks indicated a reduction in biological age in treated patients. First author and Insilico CEO Alex Zhavoronkov presented the results at a Nature conference on AI in healthcare at Sorbonne University on September 8.

The framing matters more than the effect size. Nearly every geroprotection trial to date has tested a repurposed generic — rapamycin, metformin, dasatinib — because those compounds are available, cheap, and have known safety profiles. That convenience comes at a cost: none of them was designed for the aging target. Rentosertib is the first clinical candidate to reach this stage where both the target and the molecule were selected with aging biology explicitly in the scoring function.

The obvious limitation is that this was not a geroprotection trial. It was a fibrosis trial, and the aging-clock analysis is secondary and retrospective. IPF is itself an age-related disease with a strong senescence component, so improvement in lung pathology could plausibly move proteomic clocks without the drug doing anything to systemic aging. Distinguishing those two readings requires a trial designed for the purpose.

Set alongside the CBT-I and Genflow results below, the more interesting fact may be that three unrelated groups this week reported clock movement as their primary claim. The field now has a shared reporting convention. Whether that convention tracks anything durable is the open question — and the FDA has not qualified any epigenetic clock as a surrogate endpoint.

Source: Nature Biotechnology

Treating Insomnia Slowed the Pace of Biological Aging in a Randomized Trial

Sleep quality is one of the few aging-relevant variables that is both robustly associated with mortality and genuinely modifiable without a prescription. This UCLA Health study, published in The Lancet Healthy Longevity, tested whether fixing the sleep changes the aging measure.

The design was a randomized controlled trial in adults aged 60 and older with insomnia, assigning roughly a hundred participants to either cognitive behavioral therapy for insomnia or a sleep education control. Blood was drawn at baseline and again at two-year follow-up, and three epigenetic clocks were computed: DunedinPACE, GrimAge, and PC-PhenoAge.

Two findings emerged. Participants receiving CBT-I were nearly three times as likely to achieve insomnia remission as those receiving sleep education — a large effect, consistent with the established efficacy of CBT-I as a treatment. And the CBT-I group showed a significantly slower pace of aging on DunedinPACE, while the sleep education group’s pace of aging increased over the same interval.

DunedinPACE is worth distinguishing from the other clocks here. It was trained not to estimate a person’s age but to estimate the rate at which their physiology is deteriorating — a speedometer rather than an odometer. That design makes it the clock most likely to respond to an intervention on a two-year timescale, and it is also the clock family that the recent literature has identified as most responsive to interventions generally.

The behavioral framing is what gives this result reach. Whatever one concludes about clock validity, CBT-I is an existing, widely deliverable, non-pharmacological therapy with a favorable risk profile. If a course of structured behavioral treatment for a common complaint registers on the same instrument that drug developers are using to report geroprotection, the comparison between the two becomes possible in a way it was not before.

Source: The Lancet Healthy Longevity

A Gene Therapy Trial in Aged Dogs Meets a Biological-Age Endpoint

Genflow Biosciences announced on September 8 that its SLAB trial — Sarcopenia and Longevity in Aged Beagles — met its primary endpoint. The company’s therapy delivers a variant of SIRT6 found in human centenarians, on the hypothesis that the longevity-associated allele confers a functional advantage that can be transferred.

The trial enrolled 24 beagles over ten years of age, randomized and blinded across four arms: two naked-DNA dose levels, a single-dose AAV8 vector group, and a saline control. The primary endpoint was reduction in biological age as measured by the GRIM methylation clock, and the company reports that treated animals showed a clear improvement relative to controls on that measure, with additional improvement across multiple functional and observational secondary measures. Detailed results are to be presented at the Animal Longevity Summit in Toronto on October 1–2.

Companion dogs occupy a genuinely useful position in aging research. They are outbred, share human households and therefore human environmental exposures, develop many of the same age-related pathologies, and live long enough for chronic disease to emerge but short enough for a lifespan study to conclude within a funding cycle. A 24-animal trial is small, and a top-line announcement ahead of a conference presentation is not a peer-reviewed result — but the trial structure, with dose levels, two delivery modalities, and a blinded saline control, is a real trial structure rather than an observational report.

The endpoint choice is the notable thing, and it is the third instance this week. A company set out to demonstrate rejuvenation and defined success in advance as movement on a methylation clock, in a species where a hard survival endpoint would have been achievable within a few more years. That is a bet that regulators and investors will accept the clock as a proxy. Whether the bet pays off depends on evidence the field does not yet have.

For readers tracking the gene therapy angle specifically: the comparison between naked DNA and AAV8 delivery in the same trial is unusual and potentially more informative than the aging readout, since delivery has been the persistent bottleneck for systemic gene therapy in large animals.

Source: BioSpace

Shutting Down Growth Hormone Signaling in Middle Age Still Extends Lifespan

Growth hormone receptor knockout mice are among the longest-lived laboratory mammals known, and the GH/IGF-1 axis is one of the most consistently validated longevity pathways across species. The persistent objection has been developmental: those animals lack GH signaling from conception, are dwarfed, and differ from normal mice in ways that make the lifespan comparison hard to attribute cleanly to aging biology. The relevant question for human translation is whether the pathway can be interrupted in an adult who has already finished growing.

This Aging Cell paper answers it. Using a tamoxifen-inducible system, researchers ablated the growth hormone receptor at 12 months of age — mouse middle age, well past skeletal maturity. Lifespan was significantly extended in both sexes, and without the major effects on body size that confound the constitutive knockouts.

The intervention produced the expected endocrine signature of GH resistance: reduced circulating IGF-1 with elevated GH. Beyond survival, male animals showed improved insulin sensitivity and protection against age-related deterioration in both neuromuscular performance and bone microarchitecture — the second of which is worth flagging, since loss of GH signaling is often assumed to cost bone density.

Single-nucleus RNA sequencing of liver tissue revealed a sexually dimorphic transcriptional response. Both sexes showed a reduction in hepatic B cells, but male hepatocytes shifted toward a feminized gene expression pattern. Since female mice outlive males in most strains and the liver is a central node in GH signaling, a treatment that partially feminizes the male hepatic transcriptome is a mechanistically suggestive result rather than an incidental one.

The practical significance is that this is the second result this week — semaglutide is the other — showing that a metabolic intervention started in an already-aged or middle-aged animal still buys time. The window for geroprotection appears to be wider than the developmental-programming view of the GH literature had implied.

Source: Aging Cell

Tissues Size Their Cleanup Crews to Match How Much Mitochondrial Waste They Make

Macrophages are usually introduced as immune cells — pathogen eaters, inflammation coordinators. This Science Immunology study from groups at UC San Francisco, Yale, Universitat Pompeu Fabra, and Spain’s national cardiovascular research center describes a different job entirely, one closer to municipal sanitation.

The researchers tracked mitochondria in heart muscle, skeletal muscle, and brown fat — the three most energetically demanding tissues in a mouse — and found that spent mitochondria consistently ended up inside macrophages. More striking, the size of the resident macrophage population in each tissue scaled with that tissue’s mitochondrial activity. Tissues that burn through more mitochondria maintain proportionally larger disposal crews.

The mechanism connecting supply to demand runs through fibroblasts, the structural cells usually treated as scaffolding. Fibroblast numbers increased in parallel with tissue mitochondrial activity, and the fibroblasts produced a signal that drove macrophage proliferation. That gives a tissue a way to match its cleanup capacity to its waste output without central coordination — a local feedback loop rather than a systemic set point.

When the team blocked macrophage clearance, each tissue failed in its own characteristic way: the heart’s pumping capacity fell, skeletal muscle weakened, and brown fat lost its ability to generate heat. The failures are organ-specific but the cause is shared, which is exactly the signature of a general maintenance process.

For aging research this reframes a familiar observation. Accumulation of damaged mitochondria is one of the canonical hallmarks of aging, usually attributed to declining mitophagy inside the cell. This work says a substantial part of the disposal happens outside the cell, handled by a separate cell type whose numbers are set by a fibroblast signal — which means age-related mitochondrial accumulation could reflect a breakdown in any of three places, and the macrophage arm is the one most obviously druggable.

Source: Science Immunology

A Brief Mitochondrial Stress Before Birth Protects the Heart for Life

Mitohormesis — the idea that a mild mitochondrial insult provokes an adaptive response leaving the organism more resilient than before — has been one of the more attractive and less mechanistically specified concepts in aging biology. It explains part of why exercise and caloric restriction work, but the chain from transient stress to durable benefit has been mostly hand-waved.

This Salk Institute study in Science Advances fills in the chain. The researchers blocked the mitochondrial antioxidant system, allowing superoxide to accumulate. The superoxide inhibited a key enzyme in the energy-production pathway, which disrupted downstream reactions and caused citrate to accumulate. Citrate escaping the mitochondrion is the messenger: it drives long-term epigenetic changes that establish the adaptive state.

The timing experiment is what makes this striking. The team induced mitochondrial reactive oxygen species only during mouse embryonic development, then let the animals grow up untreated. Those mice were protected in adulthood against doxorubicin cardiotoxicity — a well-characterized, clinically important form of heart damage caused by a widely used chemotherapy drug. A stress confined to embryogenesis produced a cardioprotective phenotype that persisted into adult life.

Citrate is a plausible messenger for exactly this kind of effect. Once in the cytosol it feeds acetyl-CoA production, and acetyl-CoA is the substrate for histone acetylation — a direct route from metabolic state to chromatin state. This is a metabolite carrying information from the mitochondrion to the genome, and the durability of the outcome suggests the message is written rather than merely transmitted.

The authors propose mitohormetic signaling as a therapeutic target for age-related tissue pathology. The obvious question their timing design raises is whether the same citrate-to-chromatin route can be engaged in an adult, or whether the developmental window is doing essential work. Nothing here settles that, but the pathway is now specific enough to test.

Source: Science Advances

The Immune Attack on Tau-Damaged Neurons Is Organized Outside the Brain

Tauopathies — Alzheimer’s disease among them — involve immune cells attacking neurons, and CD8+ T cells accumulating in affected brain tissue have been observed for some time. The assumption has been that whatever activates those T cells happens inside the brain, which is a discouraging premise, because it puts any intervention behind the blood-brain barrier.

Washington University researchers report in Nature Neuroscience that the premise is wrong. In a mouse model of tauopathy, conventional type 1 dendritic cells cross-presented brain-derived antigens in the deep cervical lymph nodes — in the neck, outside the central nervous system. Those dendritic cells primed CD8+ T cells, which then migrated into the brain and promoted neurodegeneration.

The proposed sequence is that tau-driven damage causes brain cells to release material that drains into the cervical lymph nodes, where dendritic cells pick it up and instruct T cells to treat it as a target. The instruction step, in other words, happens at a peripheral site that ordinary drugs can reach.

The intervention experiments support the causal reading. Disrupting the dendritic cells or specifically blocking their cross-presentation activity reduced CD8+ T cell infiltration into the brain and protected against neuronal loss. That is a substantial effect obtained by acting entirely outside the CNS.

The human evidence is partial and the authors are appropriately careful about it: samples from human primary tauopathies show increased brain CD8+ T cells, consistent with the mouse finding, but dendritic cell involvement in humans remains unconfirmed. Even so, this relocates a therapeutic target from one of the hardest compartments in medicine to one of the more accessible ones, for a class of diseases where delivery has been a defining obstacle.

Source: WashU Medicine

Twenty Years of SWAN Data Support Hormone Therapy Started During the Menopause Transition

The Women’s Health Initiative results published in 2002 caused a collapse in menopausal hormone therapy use and shaped clinical practice for two decades. The subsequent reanalysis has been steady and unglamorous: the WHI enrolled women who were on average well past menopause, and timing relative to menopause onset turns out to matter a great deal. What has been missing is evidence about the specific group most likely to seek treatment — women who are still in the transition and are experiencing hot flashes and night sweats.

This JAMA Internal Medicine study, led by researchers at Virginia Commonwealth University and the University of Pittsburgh, addresses that group directly. It draws on the Study of Women’s Health Across the Nation, using clinical data collected between 1997 and 2017 from more than 2,700 women who reported vasomotor symptoms and had no prior cardiovascular events — twenty years of follow-up on a cohort assembled for exactly this purpose.

Hormone therapy initiated during peri- or early postmenopause was associated with a 22 percent reduction in cardiovascular disease events relative to non-users. Among women who began therapy within ten years of menopause onset, the estimated reduction was 27 percent. The protective association was strongest among Black women and among those who started within that ten-year window.

The authors and accompanying commentary are explicit that this is observational: women who take hormone therapy differ from women who do not in ways that statistical adjustment cannot fully resolve, and healthy-user bias is a live concern in exactly this literature. The paper is positioned as laying groundwork for future trials rather than as settling the question.

It nonetheless lands in a changed regulatory environment — the FDA removed boxed warnings from hormone therapy products earlier in 2026. For a field that spends most of its effort on interventions a decade from the clinic, an item about a widely available therapy, a large sex-specific disease burden, and a treatment window that opens and closes on a known schedule is a reminder that some of the available healthspan gains are logistical rather than scientific.

Source: JAMA Internal Medicine