The rationale for the work begins with a paradox. Fisetin, a naturally occurring flavonoid found in strawberries and other plants, has shown promise in preclinical studies, reducing senescence markers, alleviating inflammation, and extending healthspan in aged mice. Those findings have fueled multiple clinical trials testing fisetin in chronic kidney disease, osteoarthritis, skeletal health, COVID-19, and frailty. Yet fisetin is, by the authors’ own measurements, not a strong senolytic. It requires high doses in cell culture to kill senescent cells, and it suffers from poor solubility and absorption, properties that could limit its clinical effectiveness. The team set out to fix these weaknesses without abandoning the flavonoid scaffold entirely.

Because fisetin is promiscuous—activating or suppressing a wide range of pathways involved in antioxidant defense, apoptosis, proliferation, inflammation, and angiogenesis—traditional structure-based drug design, which depends on a predefined protein target, was poorly suited to the problem. Instead, the researchers turned to phenotypic drug discovery, screening compounds for their ability to change a visible cellular phenotype: the activity of senescence-associated beta-galactosidase, a classic marker of senescent cells. Using a high-content fluorescent imaging platform, they screened a focused library of natural flavonoids spanning flavones, flavonols, flavanols, anthocyanidins, and other subclasses in senescent mouse embryonic fibroblasts lacking the DNA repair protein ERCC1, alongside matched non-senescent controls.

The screening results were revealing. Fisetin showed only moderate senolytic activity, with an EC50 of roughly 22.74 micromolar and a low selectivity index of 1.53, meaning it was barely more toxic to senescent cells than to healthy proliferating ones. Luteolin, a flavone found in many vegetables, emerged as the most potent and selective natural flavonoid in the screen, with an EC50 of 14.21 micromolar. From these data, the team extracted preliminary structure–activity relationship rules: planar scaffolds retaining the C2–C3 double bond and C4-oxo group of flavones and flavonols were consistently active, while isoflavones, flavanones, and anthocyanidins were largely inert. A 3′,4′-dihydroxyl pattern on the B ring proved essential, a 5-hydroxyl on the A ring hurt activity, electron-donating methoxy groups helped, and any glycosylation was detrimental.

Armed with these guidelines, the researchers launched an iterative cycle of design, synthesis, and biological testing, ultimately creating 35 novel flavonoid analogs. Most outperformed both fisetin and luteolin. Substituents at the 6-position of the A ring proved more favorable than those at the 7-position, and bulky groups such as tert-butyl and isopropyl markedly enhanced activity. Two leads emerged. SR29384, bearing an isopropyl substituent, was the most selective compound in the entire series. SR31133, carrying a five-membered alicyclic ring, was the most potent, with an EC50 of 0.8 micromolar—a more than 27-fold improvement in senolytic activity over fisetin. Both analogs also showed improved computed drug-like properties, including greater lipophilicity, lower topological polar surface area consistent with better blood–brain barrier permeability, and reduced predicted toxicity and carcinogenicity, though the authors caution that such in silico predictions require empirical longitudinal validation.

The optimized analogs proved to be broad-spectrum senolytics. They reduced the number of senescent cells across a diverse panel of models, including Ercc1-deficient fibroblasts, wild-type fibroblasts stressed by oxidative or genotoxic damage, senescent human IMR90 lung fibroblasts, and replicatively senescent human umbilical vein endothelial cells. Treatment also significantly downregulated key senescence and inflammatory markers, including p16Ink4a, p21Cip1, and interleukin-6, outperforming fisetin. Notably, a DPPH radical scavenging assay showed that all three compounds had comparable antioxidant activity, indicating that the enhanced senolytic power of the new analogs was not simply a matter of better antioxidant chemistry but reflected genuinely altered interactions with cellular targets.

The in vivo results were the most striking. In 28-month-old wild-type C57BL/6J mice—an age at which fisetin would not be expected to be significantly effective—the team administered a deliberately low dose of 20 milligrams per kilogram per day for five consecutive days, a stringent comparison against the 100 milligrams per kilogram used in earlier fisetin studies. Fisetin had only marginal effects on senescence and SASP markers. The two analogs, particularly SR29384, significantly decreased expression of p16Ink4a and p21Cip1 as well as multiple SASP factors, including Il6, Il1β, Mcp1, and Pai1, in tissues including kidney, brain, lung, spleen, fat, and muscle. The systemic reduction of senescence markers in diverse organs, including the brain, served as functional evidence that the compounds possess sufficient oral bioavailability to engage their targets, even though direct pharmacokinetic measurements were not performed.

SR29384 was then tested in Ercc1-deficient progeroid mice, a model of accelerated aging. Treated three times per week for five weeks at 20 milligrams per kilogram by oral gavage, the compound suppressed the composite score of aging symptoms, with particularly notable improvements in dystonia and ataxia, two indicators of aging-related neurodegeneration. Gene expression analysis showed significant reductions or trends toward reduction in senescence and SASP factors across the kidney, liver, lung, and brain. The authors note that the effects were tissue- and marker-dependent, and that SR29384 did not show uniform superiority across all tissues, an honest caveat that underscores the complexity of translating senolytic activity from cell culture to living organisms.

Perhaps the most scientifically valuable contribution of the study is its mechanistic dissection. By integrating bulk RNA sequencing, machine-learning-based pathway prediction, structure-based target prediction, protein–protein interaction network analysis, and molecular docking, the team converged on PARP1 and CDK2 as the key shared targets of both analogs, with BCL-xL implicated as a possible third. Docking simulations suggested that both analogs bind the PARP1 catalytic cavity more strongly than fisetin, which suffered unfavorable donor–donor clashes with key residues, and that they compete with ATP in the CDK2 activation domain. Enzyme assays confirmed these predictions: SR31133 inhibited PARP1 with an IC50 of 2.51 micromolar, outcompeting the standard inhibitor 3-AB, and inhibited CDK2 with an IC50 of 12.73 micromolar. Neither analog directly inhibited BCL-xL at the concentrations tested, suggesting that the observed perturbation of BCL-xL signaling networks occurs indirectly. Supporting the functional relevance of these targets, the PARP1 inhibitor olaparib reduced SASP gene expression and modestly reduced SA-β-gal positivity in senescent IMR90 cells, while the CDK2 inhibitor staurosporine reduced SASP genes and p16Ink4a expression.

The proposed mechanism is a multi-hit assault on the pro-survival machinery of senescent cells. Senescent cells resist apoptosis through upregulated anti-death pathways, and the new analogs appear to exploit this vulnerability by disrupting DNA damage detection through PARP1 inhibition, interfering with cell cycle and stress signaling through CDK2, and potentially undermining BCL-xL-mediated survival through crosstalk among these proteins. The two compounds also showed distinct secondary profiles: SR29384’s exclusive targets were enriched for inflammatory signaling and extracellular matrix remodeling, which may explain its remarkable selectivity, while SR31133’s exclusive targets pointed toward histone deacetylation processes. The authors emphasize that formal pharmacokinetic, tissue-distribution, and target-engagement studies are needed next, and that the compounds remain preclinical. Still, by pairing a phenotypic screening platform with modern computational biology, the study demonstrates how an aging-related natural product with murky mechanisms can be transformed into rationally optimized drug candidates—and brings the field a step closer to senotherapeutics that could one day treat multiple age-related diseases at once.

Subject of Research: Development of novel flavonoid senolytic drugs that clear senescent cells and improve healthspan markers in mouse models of aging

Article Title: Novel Flavonoid Senotherapeutics Identified by Phenotypic Drug Discovery Reduce Senescence and Improve Multiple Markers of Healthspan

Article References: Zhang, L. J., Salekeen, R., Soto‐Palma, C., Wu, Q., He, Y., Elsallabi, O., Bdiri, B., Hughes, B., Nunes, A., Xu, W., Zhang, B., Mohamed, A., McGowan, S. J., Angelini, L., O'Kelly, R., Kamenecka, T. M., Niedernhofer, L. J., & Robbins, P. D. (2026). Novel Flavonoid Senotherapeutics Identified by Phenotypic Drug Discovery Reduce Senescence and Improve Multiple Markers of Healthspan. Aging Cell, 25(10), Article e70736. https://doi.org/10.1111/acel.70736

Image Credits: AI Generated

DOI: 10.1111/acel.70736

Keywords: senolytics, cellular senescence, fisetin, flavonoids, SASP, PARP1, CDK2, phenotypic drug discovery, healthspan, aging, Aging Cell, preclinical

The central conceptual move in the paper is a reframing of what distinguishes a quiescent cell from a senescent one. Rather than treating the two states as fundamentally different biological categories, Yilmaz and Shay describe them as positions along a spectrum of arrested states, defined by two parameters: the stability of the arrest and the magnitude of the perturbation required to reverse it. A deeply stable senescent cell and a shallowly arrested quiescent cell are, in this view, points on the same landscape of cell-cycle exit decisions, separated not by kind but by degree. The reframing does not erase the distinctive biology of senescence. The authors retain the senescence-associated secretory phenotype, known as SASP, along with metabolic reprogramming and chromatin remodeling, as active, context-dependent components of the senescent program. What changes is the interpretation: senescence is no longer a unique or exceptional state, but one outcome among several within a broader decision-making architecture.

To give this continuum quantitative teeth, the authors introduce what they call the Maintenance Threshold Model, or MTM. The model proposes that the stability of any arrested state emerges from a balance between two competing quantities: the cell’s maintenance capacity, meaning its ability to repair and renew its own components, and the accumulated cellular stress it carries. When maintenance capacity exceeds stress, an arrest remains reversible; when stress overwhelms maintenance, the cell tips into stable senescence. Critically, the model does not pool these inputs into a single average. Instead, it proposes that they are gated in series, a design principle with a striking technical implication: proteostatic and organellar maintenance capacity determines whether an arrested cell can complete division independently of whether the Rb-E2F restriction point is traversed. In other words, having an intact cell-cycle brake is not enough; a cell must also possess the protein-quality and organelle-renewal machinery to survive what comes after reactivation.

This serial gating has deep roots in systems biology. Switch-like, bistable behavior in stress-response networks is what allows cells to integrate signals about cellular stress, metabolic capacity, and quality-control machinery into discrete fate outcomes rather than graded drift. The review argues that such switch-like regulatory behavior, not categorical state identity alone, governs whether an arrested cell stays reversible, progresses toward stable senescent arrest, or resumes proliferation. Bistability, in this framework, means that two stable outcomes can coexist for the same set of underlying conditions, with the direction a cell takes determined by threshold crossings rather than by continuous adjustment. That architecture explains why cells exposed to similar levels of damage can end up in radically different fates, and why intervention at the right moment can, in principle, push an arrested cell back toward a reversible state or forward into irreversible arrest.

One of the model’s most consequential claims concerns why senescent cells persist. The authors propose that senescent-cell persistence is set not by one threshold but by two independent ones, both of which become less reliable with age. The first is cell-autonomous: the intrinsic maintenance capacity of the cell itself, its proteostasis, its macroautophagy, its organellar quality control. The second is non-cell-autonomous: immune surveillance, the ability of the surrounding organism to recognize and clear arrested cells. A senescent cell accumulates when its own maintenance systems fail to restore reactivation competence and when the immune system fails to remove it. Aging degrades both thresholds simultaneously, which, in the MTM framework, explains the age-dependent accumulation of senescent cells far more naturally than any single-mechanism account. The model thereby connects intracellular biochemistry and organism-level immunology in a single quantitative picture of tissue decline.

The review’s synthesis spans three traditionally separate research communities: aging biology, stem cell biology, and cancer research. In aging, the model provides a mechanistic rationale for why senescent cells accumulate in old tissues and why clearing them can restore function. In stem cell biology, it clarifies how tissue reserves maintain the option of re-entry into the cycle, a property essential for homeostasis and regeneration. In cancer, it speaks directly to a long-standing clinical puzzle: therapy-induced senescence. Many cancer treatments, from chemotherapy to radiation, work in part by driving tumor cells into a senescence-like arrest. But therapy-induced senescent cells are not always a victory. Some eventually escape arrest and resume proliferation, often with more aggressive behavior, and their SASP can remodel the tumor microenvironment in ways that support disease progression. The MTM offers a principled way to think about when arrested tumor cells will stay locked down and when they will rebound.

Because fate is plastic under this framework, the authors argue it can be deliberately manipulated, and they lay out a three-part therapeutic framework that exploits that plasticity. The first strategy is state-locking: pushing arrested cells firmly past the point of reversibility so that tumor cells driven into arrest by treatment never escape. The second is metabolic inflexibility targeting: senescent cells typically rely on particular metabolic configurations, and attacking those dependencies selectively eliminates them. The third, and perhaps most conceptually adventurous, is program hijacking: co-opting the senescent program itself, including its secretory machinery, for therapeutic benefit rather than simply suppressing it. Each strategy treats the arrest-state continuum as a control surface with tunable dials rather than a fixed landscape, aiming interventions at the thresholds that determine outcomes.

The same logic runs in reverse for aging interventions. If senescent-cell burden is set by two aging-sensitive thresholds, then interventions could target either side: bolstering cell-autonomous maintenance capacity, through approaches that support proteostasis and macroautophagy, or strengthening non-cell-autonomous immune surveillance, so that the aging immune system regains its ability to find and clear arrested cells. The framework also suggests caution for senolytic approaches: rather than viewing all senescent cells as targets for elimination, a threshold-based view encourages asking whether individual arrested cells sit near the reversible end of the continuum, where restoring maintenance capacity might be preferable to killing. Fate plasticity cuts both ways, offering opportunities to restore youthful function as well as risks of reactivating the wrong cells.

The technical vocabulary of the model is worth pausing on, because it encodes the review’s central bet. The Rb-E2F pathway is the canonical molecular restriction point governing G1-to-S transition, long treated as the decisive gate for cell-cycle re-entry. By insisting that proteostatic and organellar maintenance is gated in series with, and independent of, that restriction point, the MTM makes a falsifiable prediction: cells that pass the cell-cycle gate but lack maintenance capacity should fail to complete division or collapse into stress, while cells with robust maintenance should remain reactivation-competent even after prolonged arrest. Experiments that independently titrate CDK2 activity, autophagic flux, and proteostatic load against one another would directly test this claim. Similarly, the two-threshold account of senescent-cell persistence predicts that impairing immune surveillance should raise senescent-cell burden multiplicatively with intrinsic maintenance decline, a prediction testable in aging animal models.

What emerges from the review is a decision-centric framework in which quiescence, senescence, and proliferation are dynamically stabilized states shaped by tunable regulatory thresholds rather than fixed endpoints. Cellular fate, on this account, is a dynamic balance between damage accumulation and the capacity of quality-control systems to preserve reactivation competence, and the balance can be tipped deliberately. For a field that has spent decades cataloging markers of senescence and debating definitions, the shift from taxonomy to dynamics is significant. If the thresholds the model identifies can be measured, modulated, and eventually targeted in patients, the arrested cell may stop being a biological verdict and become, instead, a clinical choice, one that physicians could push toward clearance, commitment, or recovery depending on what the disease and the patient demand.

Subject of Research: Threshold control of cell fate decisions between quiescence and senescence in aging and cancer

Article Title: Threshold control of growth arrest in quiescence and senescence: implications for cell fate plasticity in aging and cancer

Article References: Threshold control of growth arrest in quiescence and senescence: implications for cell fate plasticity in aging and cancer. (n.d.). https://doi.org/10.1186/s12943-026-02782-8

Image Credits: AI Generated

DOI: 10.1186/s12943-026-02782-8

Keywords: cellular senescence, quiescence, cell fate plasticity, Maintenance Threshold Model, SASP, Rb-E2F, CDK2, proteostasis, macroautophagy, immune surveillance, therapy-induced senescence, aging