cGAS breaks the geroscience script — and senescence's gene lists don't agree with each other
cGAS has spent the past two months looking like a clean geroprotective target. This week it stopped being one: Gorbunova's group reports that cGAS-knockout mice age faster — shorter median lifespan, more frailty, LINE1 derepression and a smoothed H3K9me3 landscape — through a nuclear, non-catalytic chromatin role that an active-site inhibitor would not spare. Three days later, in the same journal, oocyte-specific cGAS deletion and STING inhibition rescue ovarian aging, which puts the field's defensible position at 'compartment-dependent,' not 'inhibit it.' Elsewhere: an FDA-approved antiretroviral (maraviroc) emerges as a myelopoiesis-normalizing geroprotective candidate in aged mice, Nature adds a mitochondrial metabolic checkpoint (SLC25A1) to SASP control, and a preprint unifying the field's four standard senescence gene sets finds that exactly two genes — IL6 and JUN — appear in all four.
Absolute reduction in 4-year dementia incidence with recombinant zoster vaccine (18.8% vs 24.6%; ~1 case avoided per 17 vaccinated) — observational, n=509,926
Where the human evidence is
Count by statusEvidence-stage map
Evidence stage × investigator promise (1–10)cGAS–STING: the same target, opposite signs
cGAS-knockout mice age prematurely: shortened median lifespan, LINE1 derepression and a smoothed H3K9me3 landscape
Martinez, Morandini and colleagues characterized aging in cGAS-knockout mice and found the opposite of the field's working assumption: cGAS KO animals show a shortened median lifespan, increased frailty and induced inflammation across multiple organs. Mechanistically, losing cGAS derepresses LINE1 retrotransposons — reduced DNA methylation at LINE1 elements, elevated cytoplasmic LINE1 cDNA — and smooths the H3K9me3 heterochromatin landscape with increased chromatin accessibility; the phenotype is recapitulated by cGAS knockdown in vitro. Critically, the authors attribute this to a nuclear heterochromatin-maintenance function that is independent of cGAS's cytoplasmic DNA-sensing role and independent of its catalytic activity, meaning a catalytic-site inhibitor would not spare it. Read alongside the naked mole-rat cGAS work covered on July 19 and the reduced-function cGAS variant in long-lived families covered on July 26, this is a direct constraint on systemic cGAS inhibition as a geroprotective strategy — and an argument that STING-level or tissue-restricted targeting is the safer design.
Read the paperOocyte mitochondrial DNA leakage activates cGAS–STING and propagates cGAMP through gap junctions to drive ovarian aging
Lei, Zhu and Yang show that aging oocytes accumulate cytoplasmic mitochondrial DNA through increased mtDNA leakage, activating cGAS to produce cGAMP and trigger STING signalling — and that oocyte-derived cGAMP passes through CX37 gap junctions into surrounding granulosa cells, propagating sterile inflammation across the follicle. Oocyte-specific Tfam knockout recapitulated mtDNA leakage, STING activation in both compartments, inflammation and accelerated ovarian dysfunction, with Opa1 knockdown and Pink1 deletion as complementary mitochondrial-stress models. Oocyte-specific Cgas deletion in Tfam mutants, or pharmacological STING inhibition with H-151, ameliorated ovarian dysfunction. The instructive contrast with the Gorbunova paper is compartmental: cGAS–STING signalling looks pathogenic in the aging follicle and protective at the level of the whole animal, which argues that tissue-restricted delivery — not systemic inhibition — is the translatable form of this target.
Read the paperSenescence: a metabolic checkpoint, and a definitional problem
Mitochondrial acetyl-CoA supply is a required checkpoint for SASP transcription; SLC25A1 inhibition dampens inflammation and improves healthspan in aged mice
Building on the mtDNA- and mtRNA-leakage work that established innate immune activation as the SASP trigger, this paper identifies a second, metabolic layer of control. In senescent cells the mitochondrial pyruvate–citrate–acetyl-CoA axis is upregulated, and the resulting acetyl-CoA supply is required for histone acetylation at SASP loci and for robust SASP transcription: mtDNA-driven signalling activates the inflammatory transcription factors, but acetyl-CoA availability determines whether the programme is actually executed. Raising acetyl-CoA promotes SASP expression; inhibiting SLC25A1, the mitochondrial citrate exporter, reduces histone acetylation and chromatin accessibility at SASP loci, dampens inflammation and improves healthspan in aged mice. This is a senomorphic mechanism with a defined druggable node that does not require killing the senescent cell — but SLC25A1 is broadly expressed and citrate export is not senescence-specific, so therapeutic-window selectivity is the unanswered question.
Read the paperUnifying the field's four standard senescence gene sets: only 173 of 1,250 genes are corroborated by two or more resources, and just two (IL6, JUN) by all four
This preprint maps CellAge, GenAge, the SenMayo signature and the Reactome cellular senescence pathway onto a single canonical identifier (Ensembl gene ID), collapsing 1,460 summed source entries into 1,250 unique genes and removing 210 redundancies, then layers on cross-species conservation, tissue and cell-type expression, and high-confidence protein–protein interactions (reaching 95.8%, 97.9% and 93.0% of genes respectively; 89.4% across all three). The headline number is the concordance: only 173 genes are supported by two or more of the four resources, and exactly two — IL6 and JUN — are supported by all four. The work is presented as a resource rather than a critique, and it is unreviewed, so treat the annotation layers as provisional. But the concordance figure is the finding that matters, and it lands one week after the SA-β-gal antibody investigation: when a paper, a supplement label or a clinic report cites 'senescence gene signature' enrichment, which signature was used is now a material variable, not a formality.
Read the paperHuman evidence: what actually moves in people
Recombinant zoster vaccine associated with 24% lower 4-year dementia incidence in 509,926 skilled-nursing-facility residents (target trial emulation)
Hayes and colleagues at Brown applied target trial emulation to Medicare claims and electronic health records from 509,926 adults aged 66+ admitted to more than 5,500 skilled nursing facilities between 2017 and 2022, all vaccine-eligible and free of prior dementia. Over four years, dementia was diagnosed in 18.8% of those receiving at least one Shingrix dose versus 24.6% of the unvaccinated — a 5.8-point absolute difference, roughly one case avoided per 17 vaccinated. Three things temper it: only 8,843 of 509,926 participants were vaccinated, so healthy-vaccinee confounding remains a live concern despite adjustment (vaccinated patients were younger and healthier, and the authors report that adjustment did not fully explain the association); the design is observational and cannot establish causation; and the authors disclose GlaxoSmithKline funding, with GSK reporting no control over design, analysis or the decision to publish. It is directionally consistent with the Welsh natural-experiment data on the older live vaccine, and it is the first such analysis restricted to the recombinant vaccine in a frail institutionalized population.
Read the paperSex-stratified clinical aging clocks from 172 measures in >100,000 people show divergent midlife trajectories that reconverge in later life
Li, Gao and Zhang profiled 172 clinical measures across more than 100,000 participants aged 18–98 at three centers of the Multicentric Chinese Aging Study, then built sex-specific clinical aging clocks. Women and men follow measurably divergent trajectories through midlife that converge again in later life. Phenome-wide analysis nominated accumulating metabolic factors — LDL, triglycerides, glucose, uric acid — and tumor markers including carcinoembryonic antigen and HE4 as age-accumulating features; the authors then showed these factors induce senescence-related phenotypes in human endothelial cells, and used a high-fat-diet mouse model with dietary reversal to argue the metabolic burden is modifiable. The practical value is that this is a clock built from measures already on a standard panel rather than from methylation or proteomics. Two caveats before porting it: the human design is cross-sectional, so trajectories are inferred rather than observed, and the cohort is entirely Chinese, so reference ranges should not be transferred to other populations without recalibration.
Read the paperSleep duration and 23 biological aging clocks: a U-shape with the lowest age gaps at 6.4–7.8 hours (with a provenance note)
Wen and colleagues related self-reported sleep duration to 23 biological aging clocks derived from in vivo imaging, plasma proteomics and metabolomics in UK Biobank participants aged 37–84. A U-shaped relationship emerged across nine brain and body systems and all three omics layers, with the lowest sample-specific biological age gaps between 6.4 and 7.8 hours, varying modestly by organ and sex. Short (<6 h) and long (>8 h) sleep were both associated with increased systemic disease risk and all-cause mortality; for late-life depression the pathways differed, with aging clocks partly mediating the long-sleep association while short sleep showed a more direct link. Mendelian randomization did not provide strong evidence that disease causally affects sleep but could not fully exclude reverse causality, and sleep duration is self-reported throughout. Provenance note, because it matters for how this is being cited: this paper appeared in Nature in May 2026. The wave of coverage circulating in late August is a press re-release, not a new result.
Read the paperInterventions in the pipeline: a strain, a metabolite, and a graft
Bifidobacterium pseudocatenulatum and its metabolite 5-aminovaleric acid betaine counteract inflammaging and extend healthspan in aged mice
Lu, Ping and Zhang mapped enterotype-specific gut microbial remodeling across aging in multiple Chinese cohorts, built a microbiome-based aging clock (MicroAge), and identified Bifidobacterium pseudocatenulatum as a candidate geroprotective species consistently depleted with age in both sexes. In naturally aged mice, oral B. pseudocatenulatum monotherapy restored intestinal homeostasis, mitigated multi-organ inflammaging, improved cognitive–motor performance and extended healthspan. The mechanistic handle is 5-aminovaleric acid betaine (5-AVAB), a strain-derived metabolite whose levels decline physiologically in aging humans and whose supplementation partially reproduced the systemic benefits on its own. This is a credible strain-plus-postbiotic hypothesis, not a validated human intervention: the human data are associational, the interventional data are murine, the endpoint is healthspan rather than lifespan, and no commercial B. pseudocatenulatum or 5-AVAB product currently has human efficacy evidence. Strain identity matters — generic 'Bifidobacterium' supplements are not this.
Read the paperSelf-contracting subcutaneous myografts deliver systemic exercise-like benefits — and double as a therapeutic protein depot — in aged and obese mice
Liu, Yao and Shyh-Chang transplanted differentiated autologous myocytes into the subcutaneous space, producing mature, vascularized grafts that contract continuously without volitional input. The myografts improved whole-body muscle mass and function, metabolic regulation and regenerative outcomes in both aging and obese mouse models, framed explicitly as an option where exercise is contraindicated, inaccessible or insufficient. The less obvious result is that the grafts served as a stable depot for virally transduced therapeutic proteins including parathyroid hormone and growth hormone, counteracting bone and muscle loss with no side effects observed. Distance to clinic is substantial: autologous cell manufacture, a surgical implant, unresolved graft durability and tumorigenicity questions, and — for the gene-therapy depot version — a continuous growth-hormone exposure that would need a very careful risk case in an older population. Notable as a concept, not as a near-term option.
Read the paper