Longevity Weekly Review 2026-09-02
Week In Review
This was a week about industrializing the search for interventions — and about being honest regarding what the search has and has not found. The most consequential paper was an infrastructure paper: A cross-species drug-discovery platform to accelerate the identification of lifespan-extending interventions describes a single pipeline that runs survival assays in yeast, nematodes, flies, killifish, and mice, screening hundreds of compounds and asking which effects survive the jump between phyla. Geroscience has never lacked candidate molecules; it has lacked a way to triage them before someone commits a decade to a mouse study. Read alongside Aging vs. biological age: Conceptual considerations for age reversal claims, which argues that the field’s central claim — that biological age can be reversed — is often stated in language too loose to be tested, the pair sketches a discipline trying to tighten both its filters and its vocabulary at once.
Two papers this week made the same uncomfortable point from opposite directions: an intervention that helps one tissue may harm another, and a drug class can be popular without engaging the mechanism it is meant to fix. In NAD+ precursor treatment prevents cardiomyopathy but disrupts erythroid maturation in mitochondrial progeria, nicotinamide riboside improved cardiac function in prematurely aging mice while worsening their anemia — the same molecule, opposite verdicts, depending on whether the tissue divides. And The force-producing fraction: muscle quality as a therapeutic target in sarcopenia argues that decades of drugs designed to make aging muscle bigger have failed because size is not what age takes away; nerve supply, mitochondrial capacity, regenerative reserve, and fat infiltration are, and almost no clinical programme has been built around them.
A third thread reframes cellular senescence as something transmitted rather than merely accumulated. Adipose-cartilage communication via EV-Mito-mtDNA signaling promotes osteoarthritis progression traces mitochondrial DNA travelling from fat cells to cartilage inside extracellular vesicles, where it trips an innate immune alarm; Targeting pro-senescent PTGS2+ macrophages alleviates chondrocyte senescence and osteoarthritis progression finds a macrophage population in aged joints that pushes neighbouring cartilage cells into senescence via a secreted protein. Both point at the same conclusion from different cell types: in an aging joint, some cells are being told to grow old. Core activation program and selective regional responsiveness of microglia during aging and parabiosis supplies the brain’s version — and the caveat that the young-blood signal reaches some brain regions far better than others. Meanwhile, An in vivo resource of age-regulated C. elegans intestinal secretory-pathway proteins catalogues what an aging gut secretes into the space between cells, and finds a conserved enzyme there that extends lifespan when the worm makes more of it.
The week’s human evidence was quieter and more practical. Associations between diet quality, epigenetic aging and epigenome in two population-based cohorts reports that the many competing definitions of a “healthy diet” barely agree on who is eating well — yet they converge on the same epigenetic pathways, which is a more encouraging result than it first appears. And Whole-brain cholinergic modulation following computerized cognitive training in healthy older adults used PET imaging inside a randomized trial to show that ten weeks of speed-based brain training left a measurable neurochemical trace. Neither is a rejuvenation therapy. Both are the kind of result the field will need if the interventions coming out of platforms like the cross-species screen are ever to be evaluated in people rather than argued about.
Items
One Screening Pipeline, Five Species, Four Hundred Compounds
Aging research has a triage problem. Thousands of compounds extend lifespan in something — a yeast culture, a worm, a fly — and almost none of those results predict what will happen in a mouse, let alone a person. The usual reason is not fraud or sloppiness but incompatibility: each model organism has its own assay conventions, its own confounders, and its own definition of death, so results cannot be lined up against each other.
This paper describes an attempt to fix that by building one integrated, high-throughput platform spanning five organisms at once: budding yeast, C. elegans, Drosophila, killifish, and mice. The authors report evaluating more than 400 compounds across thousands of conditions, using automated imaging, miniaturized assays, and deep-learning-based death detection to make survival scoring consistent and cheap enough to run at that scale.
Each tier required its own engineering. In yeast, a miniaturized chronological-lifespan assay read out by flow cytometry made screening scalable — and, usefully, exposed assay confounders that had been distorting smaller studies. In worms and flies, compact imaging rigs paired with YOLO-based object detection allowed animals to be scored without transferring them between plates, which both removes a major source of handling stress and lets the system capture effects that depend on compound, diet, and sex. For killifish — a short-lived vertebrate increasingly used as a bridge between invertebrates and mammals — the team developed an in-house drug-pellet formulation so that large cohorts could receive standardized oral doses. In mice, longitudinal lifespan studies were combined with home-cage activity monitoring to assess late-life interventions, capturing function rather than survival alone.
The scientific payoff is cross-species integration: compounds that work in only one lineage can be separated from those that converge on conserved longevity pathways. The authors report identifying multi-species geroprotectors that hit both known and previously unrecognized conserved mechanisms. That is the point of the exercise — not any single hit, but a framework that ranks candidates by how well their effects travel, which is exactly the property that determines whether an intervention has any chance of translating.
Source: Cell Reports
What Would It Actually Take to Prove Age Reversal?
Few phrases do more work in longevity science, and in longevity marketing, than “biological age reversal.” A test says your biological age is 47 when your birthday says 54; an intervention drops that number by three years; a headline follows. This Cell Metabolism piece steps back to ask a question that is asked less often than it should be: what would the claim have to mean for it to be either true or false?
The authors’ central move is to separate two things routinely conflated. Aging is a process — the accumulation of damage and dysfunction over time. Biological age is an estimate, produced by a model trained on some set of measurements. Lowering an estimate is not the same as reversing a process, and the two can come apart. A clock trained largely on inflammation-linked markers will move when inflammation drops, whether or not anything durable about the organism’s trajectory has changed.
From there the paper works through the burden of proof. If a treatment is claimed to reverse aging rather than merely improve health, what evidence discharges that claim? Restoration of a youthful measurement is weak evidence on its own, because many measurements are reversible in ways that aging is not. Stronger evidence would show the change persists, generalizes across tissues and readouts, and translates into altered risk going forward — which is a considerably harder experiment than reporting a clock delta.
The authors are also explicit that this is partly a problem of language. Terms like rejuvenation, reversal, and slowing are used interchangeably in the literature despite describing different underlying claims with different evidentiary requirements, and that imprecision leaks from papers into press releases into products. It is a foundational rather than an experimental contribution, but it arrives at a useful moment: as platforms make it cheaper to generate candidate interventions, the bottleneck shifts to knowing what a positive result would look like.
Source: Cell Metabolism
Brain Training Leaves a Chemical Trace, Measured by PET
Computerized cognitive training occupies an awkward position in aging research: widely marketed, moderately supported by behavioural trials, and persistently vague about mechanism. If ten weeks of speed-based exercises help, what in the brain has changed? This study, nested in the INHANCE randomized trial, went looking for the answer with a specific molecular imaging target.
The cholinergic system — the acetylcholine-releasing circuitry central to attention, learning, and neuroplasticity — degenerates progressively with age, and is the same system targeted by the cholinesterase inhibitors used in dementia. The researchers used PET imaging with [18F]FEOBV, a tracer that binds the vesicular acetylcholine transporter and therefore indexes the density of presynaptic cholinergic terminals. That gives a physical readout of cholinergic integrity rather than an inference from behaviour.
Ninety-two cognitively intact adults aged 65 and older, mean age 71.9, were randomized in a double-blind, active-controlled design to 35 hours of training over 10 weeks. The intervention arm used speed-based exercises (BrainHQ’s Double Decision and Freeze Frame); the control arm used non-speeded games — Klondike solitaire and a tile-breaking puzzle — matched for engagement and screen time, which is a considerably more rigorous comparison than a no-treatment waitlist. Participants underwent neuropsychological testing and FEOBV PET before and after.
Voxel-wise analysis found that training significantly increased cholinergic binding, with the effect concentrated in the anterior cingulate, medial prefrontal cortex, insula, and orbitofrontal cortex — a set of regions associated with attentional control and salience processing rather than, say, primary sensory areas. That anatomical specificity matters: it is what one would predict if speed-of-processing training were engaging attentional circuitry, and it is harder to explain as a generic effect of showing up.
The result does not establish that cognitive training prevents dementia, and it should not be read that way. What it does is convert a behavioural intervention into a measurable neurobiological one, which is a prerequisite for optimizing dose, duration, and target population instead of guessing.
Source: GeroScience
Young Blood Reaches Some Brain Regions Better Than Others
Parabiosis experiments — surgically joining the circulations of a young and an old mouse — remain among the most striking demonstrations in aging biology, and among the least understood. Old animals exposed to young blood show improvements across multiple tissues, including the brain. The open question has been which cells respond, and where.
This study focused on microglia, the brain’s resident immune cells, which are major contributors to neurodegenerative disease progression and prime suspects in age-related cognitive decline. Using single-cell analysis, the authors first did something the field needed: they identified and benchmarked a set of reproducible microglial states and defined a core set of genes that drive microglial activation in the mouse brain. Microglial nomenclature has proliferated faster than agreement, and a benchmarked core program gives other groups a common reference.
They then asked how those states change with age and with parabiosis-mediated exposure to young and old blood, across four brain regions: cerebellum, cortex, hippocampus, and striatum. The regional comparison is the study’s key design choice, and it paid off. Microglial composition differed by region at baseline, age-related changes differed by region, and — most notably — responsiveness to the parabiosis intervention differed sharply.
The cerebellum consistently emerged as the most responsive region. The striatum showed minimal responsiveness to parabiosis. That is an important asymmetry for anyone hoping to translate systemic rejuvenation signals into therapy: a circulating factor that reaches the bloodstream does not thereby reach every brain region’s immune cells with equal effect, and a treatment evaluated only in hippocampus may look better or worse than it would elsewhere.
The practical implication the authors draw is that microglia-targeted strategies for brain aging need to account for regional vulnerability rather than treating the brain as a single compartment. It also suggests a sharper experimental question than “does young blood help the brain”: which regions are gated, and what does the gating?
Source: Cell Reports
NAD+ Boosting Helped the Heart and Hurt the Blood
NAD+ decline is one of the most commercially successful ideas in aging biology. The coenzyme is central to energy metabolism, its levels fall with age and in degenerative disease, and precursors such as nicotinamide riboside are sold widely as supplements. What has been thin is long-term, tissue-resolved data on what chronic NAD+ boosting actually does across an organism.
This study supplied some, using “mutator” mice — animals carrying a proofreading-deficient mitochondrial DNA polymerase that accumulate mitochondrial mutations and display a premature-aging phenotype. The team gave them nicotinamide riboside and then examined multiple tissues rather than the one they hoped would improve. The results diverged strikingly, and the axis of divergence is instructive: it tracks whether the tissue is proliferative.
In postmitotic cardiac tissue, the treatment worked as advocates would predict. NR enhanced contractility, reduced stress-response markers, and normalized the metabolic profile — a real rescue of the cardiomyopathy that afflicts these animals.
In proliferative bone marrow, the same treatment caused reductive stress: accumulation of NADH and NADPH, along with altered amino acid, nucleotide, and folate levels, and impaired heme biosynthesis. Downstream, erythrocyte maturation defects were aggravated and the animals’ anemia worsened. The mechanism is coherent — flooding a rapidly dividing, biosynthetically demanding tissue with reducing equivalents disturbs the redox balance those pathways depend on — but it is the opposite of the intended effect.
The authors’ conclusion is a general one that extends well past this mouse model: the systemic consequences of NAD+ boosting must be evaluated beyond the primary affected tissue, and the field likely needs tissue-specific metabolic interventions rather than whole-body precursor loading. For a class of compounds already in wide human use, “helps the heart, worsens the anemia” is exactly the kind of finding that deserves to travel.
Source: Cell Reports
An Enzyme Secreted by the Gut That Extends Lifespan
Most aging research looks inside cells. But cells also talk, and the proteins they secrete into the extracellular space coordinate homeostasis and carry signals to distant tissues. That compartment has been comparatively neglected in aging biology, largely because secreted proteins are hard to attribute to a source tissue once they are loose in the body.
This study addressed that with proximity labeling followed by quantitative proteomics — tagging proteins as they transit the secretory pathway of a specific tissue, the C. elegans intestine, so that the resulting catalogue reflects what that tissue exports rather than what is merely present nearby. The authors systematically characterized proteins along the intestinal secretory pathway, identified which of them are modulated with age, and then validated the secretion of those candidates in living animals rather than relying on the labeling alone.
Out of that resource came a specific finding. One age-modulated secreted protein, ACP7, is well conserved in humans. Overexpressing it extended the worms’ lifespan — and, critically, did so in a secretion-dependent manner. That control matters: it shows the longevity effect requires the protein to get out of the cell, ruling out an intracellular explanation and establishing that this is genuinely a signal rather than a housekeeping change.
The authors further show that ACP7 acts as a secreted phosphatase, working on targets in the extracellular space. Extracellular phosphatases are a relatively underexplored regulatory layer, and finding one that sets lifespan opens a mechanism the field has not had much purchase on. Additional proteins along the secretory pathway were also found to regulate lifespan, suggesting ACP7 is one entry in a larger set.
Worm lifespan results famously fail to generalize. But the combination here — a systematic, tissue-specific resource, a human-conserved hit, and a mechanism that depends on secretion — makes this a more testable lead than most, and the resource itself will outlast the particular finding.
Source: Cell Reports
Fat Cells Send Mitochondrial DNA to Cartilage, and Cartilage Reads It as an Alarm
Obesity is a major risk factor for osteoarthritis, and the standard explanation is mechanical: more weight, more load, faster cartilage wear. That story has always been incomplete, since obesity also raises the risk of osteoarthritis in non-weight-bearing joints. This study proposes a biochemical route from fat tissue to joint degeneration, and it runs through one of the most active pathways in current aging research.
The authors identify adipocyte-derived extracellular vesicles enriched in mitochondrial components — EV-Mito — that carry mitochondrial DNA and deliver it to chondrocytes, the cells that maintain cartilage. Once inside, that mtDNA is detected by the cGAS-STING pathway, the cytosolic DNA-sensing system that evolved to spot viral and bacterial genomes. The chondrocyte, in effect, mistakes a neighbouring tissue’s mitochondrial DNA for an infection, and responds accordingly: inflammatory signaling, metabolic dysfunction, senescence, and cartilage degeneration, demonstrated in both murine and human models.
Blocking the pathway blocked the damage. Genetic and pharmacological inhibition of cGAS-STING attenuated cartilage damage — and, notably, also reduced pain behaviours and gait abnormalities, which is the outcome that matters to a patient rather than a histologist.
The human data reinforce the model in a specific way. In population-level analysis, body fat percentage rather than BMI was more closely associated with osteoarthritis risk and severity — precisely what one would expect if adipose tissue itself were the signal source rather than body mass acting through load. And in an exercise cohort, high-intensity interval training selectively reduced adiposity and was associated with lower mtDNA abundance in synovial extracellular-vesicle fractions enriched for adipose markers, reduced synovial cGAMP (the second messenger cGAS produces), and superior improvement in symptoms.
That last chain is the study’s most interesting move: it connects an intervention people can actually do to a molecular readout in the joint fluid, and finds the readout moving in the predicted direction. It also converges with a growing body of work implicating misplaced mitochondrial DNA and cGAS-STING signaling across multiple aging tissues.
Source: Science Advances
A Macrophage Population That Pushes Cartilage Cells Into Senescence
If the previous study showed senescence arriving in the joint from distant fat, this one shows a source much closer to hand. Macrophages are known to contribute to osteoarthritis, but “macrophages are involved” is not a therapeutic target. The question is which macrophage state does what, and by what signal.
Starting from single-cell RNA sequencing of naturally aged mouse joints — aged rather than surgically injured, so the population reflects ordinary aging — the authors identified a macrophage subset with high expression of PTGS2 (the gene encoding COX-2) that was associated with joint senescence. In a surgically induced osteoarthritis model, lineage tracing showed these cells increase markedly. Selectively depleting them, using a diphtheria-toxin-receptor system that kills only the targeted population, attenuated cartilage senescence and slowed disease progression. That is the causal test, and it passed.
The signal turned out to be THBS1, a secreted protein released by these macrophages, which drives chondrocyte senescence through pericellular interactions involving SDC4 and, at least partly, TGF-β-dependent signaling. Deleting Thbs1 specifically in macrophages reduced chondrocyte senescence markers and attenuated osteoarthritis-like degeneration — confirming that the macrophages act on cartilage through this particular messenger rather than through general inflammation.
The mechanistic detail is where the paper becomes unusually interesting. Elevated glycolytic lactate — a hallmark of inflammatory macrophage metabolism — drives lactylation of THBS1 at lysine 262. That modification reduces NEDD4-dependent ubiquitination and therefore stabilizes the THBS1 protein. In other words, the macrophage’s metabolic state directly controls how much of the pro-senescence signal survives to act, linking immunometabolism to senescence propagation through a specific chemical modification.
Practically, the authors show that intra-articular targeting of PTGS2-high macrophages attenuates chondrocyte senescence and osteoarthritis progression. A locally delivered treatment aimed at a defined cell state, rather than a systemic senolytic, is an appealing profile for a disease confined to specific joints — and it sidesteps the dosing and off-target concerns that have dogged whole-body senescent-cell clearance.
Source: Cell Reports
Ten Definitions of a Healthy Diet, One Epigenetic Destination
Nutritional epidemiology has produced a proliferation of diet quality scores — Mediterranean, DASH, various healthy eating indices — each operationalizing “eating well” differently. Whether these are minor variations on a shared idea or genuinely different prescriptions has been surprisingly hard to answer, and it matters for anyone trying to give advice.
Using the Rhineland Study, a large population-based cohort, with validation in EPIC-Potsdam, the authors examined ten diet quality scores at once. Their first finding is deflating: there was minimal overlap among the participants classified in the top quartile of adherence across different scores. The same person’s diet can be excellent by one respected metric and unremarkable by another. Much apparent disagreement in the nutrition literature may reflect this — studies using different scores are not measuring quite the same thing.
The second finding is more encouraging. Adherence to a healthy dietary pattern was associated with reduced epigenetic aging, though the magnitude of that association differed across scores. Diet quality, however defined, tracks with methylation-based markers of biological aging.
The third finding is the one that reconciles the first two. Different dietary patterns were associated with distinct DNA methylation profiles — the epigenome does distinguish between them — but those distinct profiles largely converged onto the same biological pathways. Different roads, same destination.
The authors read this as evidence that general adherence to a healthy dietary pattern promotes health through broadly similar epigenetic mechanisms despite variation in dietary composition. This is observational work, and the usual confounding caveats apply: people who score well on any diet index differ from those who do not in many ways beyond food. But as a structural result about the field’s own instruments, it is valuable — and the convergence finding is a reasonable argument against treating any one dietary framework as uniquely correct.
Source: Nature Communications
Why Drugs That Grow Muscle Have Not Fixed Sarcopenia
Sarcopenia — the age-related loss of muscle strength and function — now has consensus diagnostic criteria on three continents and carries an all-cause mortality hazard ratio near 1.6. It is, by any reasonable standard, a major driver of lost healthspan. And there is no approved drug for it. This review argues that the failure is not a pipeline problem but a biology problem: the field has been targeting the wrong variable.
The pattern it identifies is stark. Agents that reliably enlarge muscle have not improved the strength and gait-speed endpoints that define the condition. Muscle got bigger; patients did not get better. The authors’ response is to decompose muscle quality — the force produced per unit of muscle — into four properties that jointly determine force production and are each druggable in principle: innervation, degraded by motor-unit loss and failed compensatory reinnervation; metabolic competence, degraded by declining mitochondrial content and respiratory capacity; regenerative capacity, degraded as satellite cells drift into senescence within an inflamed niche; and intramuscular lipid, accumulating via fibro-adipogenic progenitor-derived adipogenesis.
Mapping every therapeutic class onto those four axes produces the review’s central indictment. Anabolic agents engage none of them. Classes that do engage a quality axis engage exactly one. And no sarcopenia programme has been designed around innervation or intramuscular lipid at all — two of the four, entirely unaddressed.
Two experiments are offered to make the innervation gap concrete rather than theoretical. Soluble activin receptor IIB, a well-known muscle-enlarging intervention, failed to protect denervated muscle. And follistatin overexpression, added to voluntary exercise, rescued neither motor-unit loss nor neuromuscular junction transmission beyond what exercise achieved alone. Growing the muscle does not maintain the nerve that drives it.
The constructive half of the paper is a design framework for trials that could actually test muscle quality, which requires measuring the axes rather than inferring them from mass. It is a review rather than a new result, but it is the sort of review that redirects a field — and the argument generalizes: mistaking an easily measured proxy for the thing that failed is a recurring hazard across geroscience.
Source: Ageing Research Reviews