Harvard-Mayo senolytic trial: clearing zombie cells is safe in humans — and the joint data is just as interesting as the brain data.

What senescent cells are and why they matter

Every cell in your body can reach a state called cellular senescence — a permanent arrest of the cell cycle where the cell stops dividing but refuses to die. Senescent cells are sometimes called "zombie cells" for this reason: they are not alive in the functional sense, but they are not dead either.

Under normal circumstances in young tissue, senescent cells serve a purpose — they signal nearby immune cells to clear them out (a process called immune surveillance), and they contribute to wound healing and tumor suppression. The problem begins when this clearance mechanism fails, which happens progressively with age. Senescent cells accumulate. And accumulated senescent cells do not sit quietly.

They secrete a toxic cocktail of pro-inflammatory cytokines, chemokines, proteases, and growth factors collectively called the Senescence-Associated Secretory Phenotype (SASP). The SASP induces senescence in neighboring healthy cells (spreading the dysfunction), recruits chronic low-grade inflammation, degrades extracellular matrix (which drives joint and tissue degradation), and — at sufficient density — can suppress immune surveillance of actual cancer cells.

The two primary biomarkers used to identify senescent cells in tissue and blood are:

• p16INK4a: A cyclin-dependent kinase inhibitor that enforces cell cycle arrest. Its expression is low in young tissue and rises substantially with age. T-cell expression of p16INK4a variants is now used as a clinical biomarker in senolytic trials.
• SA-β-gal (senescence-associated beta-galactosidase): An enzyme that accumulates in senescent cells and can be detected histochemically. It is the standard tissue stain for senescent cell burden in biopsy material.

The SASP is what makes senescent cells harmful. The SASP components — TNF-α (tumor necrosis factor alpha), IL-1α, IL-6, MMP-9, MMP-12 (matrix metalloproteinases) — are the same molecules that drive osteoarthritis joint degradation, neuroinflammation, vascular dysfunction, and fibrosis across organs. Clearing senescent cells removes the SASP source — not just the symptoms downstream of it.

Senescent cells don't die — they sit in tissue and secrete a chronic inflammatory signal. Clearing them removes the source of SASP. That's different from managing inflammation downstream with anti-inflammatory drugs.

The STAMINA trial — what it actually showed

The STAMINA trial (NCT05422885) was a 12-week, single-arm, open-label pilot study run from 2022 to 2024 in the greater Boston area, led by Lewis A. Lipsitz, MD, at Hebrew SeniorLife and Harvard Medical School, with James Kirkland, MD, PhD, of Mayo Clinic — one of the pioneering figures in the senolytic field — as a co-investigator. Cedars-Sinai Medical Center was also involved [1, 2].

Twelve participants completed the protocol. They were 65 years or older (mean age: 77 ± 8 years, 58% female), had slow gait speed below 1 meter per second — a validated marker of physical vulnerability — and had mild cognitive impairment defined by standardized assessment. These were not healthy young volunteers. They were older adults at real risk of functional decline.

The treatment protocol: 100 mg dasatinib (orally) plus 1,250 mg quercetin (orally) for two consecutive days, every two weeks, for six cycles — twelve weeks total. This intermittent dosing schedule is intentional: senolytics do not need to be present continuously because senescent cells take time to re-accumulate after clearance.

The primary findings:

• Safety: One serious adverse event was recorded — an accidental injury, unrelated to the intervention. Six adverse events (out of 81 total recorded over the trial period) were rated as possibly or probably related to treatment, primarily mild reductions in white blood cell count. No serious treatment-related adverse events. Compliance was 99% for dasatinib and 100% for quercetin.
• Cognition: Overall MoCA (Montreal Cognitive Assessment) score showed a mean increase of 1.0 point — non-significant across the full group. In the subgroup with lower baseline MoCA scores (18–25 at enrollment, n=8), the mean increase was 2.0 points, reaching statistical significance (p=0.04).
• The key biomarker finding: Reduction in TNF-α showed a significant inverse correlation with MoCA improvement (r = −0.65, p = 0.02). Participants who cleared more TNF-α showed more cognitive improvement. This is a mechanistic signal — it suggests the cognitive benefit, where it occurred, was mediated by inflammatory reduction consistent with SASP clearance.

The authors were careful to note the absence of a control group and the need for larger trials before conclusions about efficacy can be drawn. This is honest. What the trial establishes is feasibility and safety — and the TNF-α correlation provides a mechanistic hypothesis worth testing in a powered RCT.

How dasatinib + quercetin works

Senescent cells survive despite their damaged state because they upregulate anti-apoptotic proteins — molecular survival mechanisms that prevent their own programmed death. This is called the Senescent Cell Anti-Apoptotic Pathway (SCAP). Senolytics work by transiently disabling SCAP, allowing senescent cells to die while leaving non-senescent cells largely unaffected [3].

Dasatinib is a tyrosine kinase inhibitor originally approved for chronic myeloid leukemia. At the concentrations used in senolytic protocols — lower than oncology dosing — it inhibits the BCR-ABL and Src family kinase pathways that senescent cells rely on for survival. Quercetin, a plant flavonoid, inhibits the PI3K/AKT pathway and Bcl-2/Bcl-xL survival proteins in senescent cells. Together, they hit different nodes of the SCAP network — which may explain why the combination is more effective than either agent alone.

The 2019 Mayo Clinic pilot in diabetic kidney disease patients established the first direct human evidence that D+Q reduces tissue senescent cell burden: within eleven days of treatment, adipose tissue p16INK4a and p21CIP1 expression declined, SA-β-gal activity fell, and circulating SASP factors (IL-1α, IL-6, MMP-9, MMP-12) decreased measurably [4]. The STAMINA trial extends this into an older population with cognitive vulnerability.

It is worth noting that D+Q does not eliminate all senescent cells. A 2025 preprint study found that 30–70% of human senescent preadipocytes in tissue samples were resistant to dasatinib or quercetin. Senolytic resistance is an active research area — "senosensitizers" that prime resistant cells for drug-induced clearance are under investigation [5]. This matters for framing expectations: even with D+Q, a meaningful percentage of the senescent burden may remain.

The joint angle: D+Q and osteoarthritis cartilage

Cartilage degeneration in osteoarthritis (OA) is, in part, a senescent cell problem. Senescent chondrocytes — the cells responsible for maintaining cartilage matrix — accumulate in OA joints and secrete a SASP that includes IL-6, CXCL1, MMP-13, and other catabolic enzymes that degrade the extracellular matrix faster than healthy cells can rebuild it.

A 2025 study published in Aging Cell by Maurer and colleagues tested D+Q on human articular chondrocytes isolated from osteoarthritic cartilage tissue (OARSI grade 1–2) [6]. The results were specific and mechanistically interesting:

• The combination selectively eliminated senescent cells from both cartilage explants and isolated human chondrocytes.
• Post-clearance, the surviving chondrocytes showed significantly increased expression of COL2A1 (type II collagen), ACAN (aggrecan), and SOX9 — the primary markers of a functional, matrix-producing chondrocyte phenotype. These genes are suppressed in OA chondrocytes.
• SASP factors IL-6 and CXCL1 were significantly reduced. Pro-anabolic growth factors — IGF-1, FGF18, and TGFβ2 — increased after clearance.

The translation here is direct: clearing senescent chondrocytes from human OA cartilage tissue allowed the surviving healthy chondrocytes to return to a matrix-building phenotype. The SASP was reduced. The tissue environment shifted from catabolic to anabolic. This happened in human tissue, not a mouse model.

This is the joint angle that matters for the recovery-pain context. Non-surgical chronic joint pain driven by OA is, in part, a SASP-mediated process. Senolytics — if they can be delivered effectively to joint tissue — represent a mechanism-level intervention, not a symptomatic one. The challenge is that oral D+Q at current doses may not achieve joint-tissue concentrations sufficient to produce the effects seen in ex-vivo cartilage studies.

Unity Biotechnology's intra-articular senolytic program (UBX0101) attempted this delivery route in a Phase 2 trial for knee OA and missed its primary endpoint [7]. The company's interpretation was that the underlying biology remains compelling but the drug-target pair and delivery format needed refinement. They subsequently pivoted to ophthalmology. This is not evidence that senolytics don't work in joints — it's evidence that UBX0101 at that dose and schedule, injected intra-articularly in that patient population, did not produce the expected effect.

Fisetin — the natural senolytic

Fisetin is a polyphenol found in strawberries, apples, and persimmons that has demonstrated senolytic activity in preclinical research. It induces apoptosis in senescent cells through Bcl-2/Bcl-xL inhibition and PI3K/AKT pathway suppression — overlapping with the quercetin mechanism but with some distinct binding affinities.

The preclinical data is genuinely interesting. A 2025 study published in Aging Cell (Murray et al.) found that intermittent fisetin supplementation in aging mice improved physical function and reduced skeletal muscle senescent cell burden to a degree comparable with genetic senescent cell clearance and synthetic navitoclax (ABT-263) treatment — a high bar for comparison [8].

In humans, a 2024 Mayo Clinic study in long-COVID patients (n=536, with 44 receiving fisetin) found that 64% of fisetin-treated patients reported significant alleviation in symptoms including fatigue, muscle pain, and orthostatic hypotension. This is observational data within a broader trial and should not be read as a controlled efficacy signal — but it's the most substantial human fisetin dataset published.

Fisetin has real limitations as a senolytic: oral bioavailability is low and variable (typically 5–7%), and achieving tissue concentrations sufficient for senolytic action may require formulations not currently available commercially. The supplements sold as "fisetin" vary enormously in bioavailability. The therapeutic doses tested in animal studies (20–100 mg/kg in mice) do not translate directly to human milligram dosing without pharmacokinetic modeling.

The article on fisetin's clinical trial evidence covers the PK and dosing data in more detail.

What the data does not show

The STAMINA trial has one primary limitation: n=12 with no control group. It cannot distinguish treatment effect from regression to the mean, seasonal variation in cognitive performance, or other confounders. The authors are transparent about this. The trial's purpose was to establish feasibility and generate mechanistic hypotheses — both of which it accomplished.

Additional limitations of the current senolytic evidence base:

• No large RCTs with cognitive or joint outcomes as primary endpoints. STAMINA, the diabetic kidney disease pilot, and the IPF pilot are all small, single-arm, or placebo-controlled only for secondary outcomes. A Phase 3 program with conventional efficacy standards does not yet exist for D+Q in aging-related indications.
• Dasatinib requires a prescription and carries real risks at oncology doses. The longevity doses used in trials (100 mg intermittently vs. 100–140 mg daily in CML) are lower, but dasatinib is not a benign drug. Side effects at higher doses include pleural effusion, QT prolongation, and myelosuppression. Self-dosing based on internet research is not appropriate.
• The senolytic pipeline has produced setbacks. UBX0101 Phase 2 failed. resTORbio's TORC1 immunosenescence program failed Phase 3. These are not proof that senolytics don't work — they are proof that moving from mechanism to clinical outcome is harder than the animal data suggests.
• FDA approval is not imminent. The FDA has not approved any drug specifically for anti-aging senolytic indications. The current regulatory pathway for aging-related indications remains unsettled. Rubedo Life Sciences received IND clearance for a topical senolytic for psoriasis in 2025 — this is the leading edge of FDA action on the class, and it is in dermatology, not systemic aging.

Framework: where senolytics sit in a longevity stack

The human evidence for senolytics is preliminary and promising in the right order: safety is confirmed, mechanistic biomarker changes are confirmed, and the correlation between biomarker changes and functional outcomes (cognition, pain) is emerging. What does not yet exist is a powered efficacy trial with a hard endpoint. For context on a complementary pathway — mitochondrial function rather than senescent cell clearance — the MOTS-c Phase 2a data represents the parallel track of the same longevity-mechanism conversation.