1. Introduction
Aging is, if you look at it honestly, the single largest unaddressed risk factor in medicine — it rarely causes any one disease outright, yet it quietly loads the dice for nearly all of them (Sun, 2023). For much of the twentieth century, clinicians treated cardiovascular disease, neurodegeneration, cancer, and type 2 diabetes as though they told separate biological stories, each with its own textbook chapter and its own specialist (Paez-Ribes et al., 2019; Wyles et al., 2022). That separation is starting to look less like biology and more like convenience. The geroscience hypothesis — still argued over in places, but increasingly hard to dismiss — proposes that these seemingly unrelated conditions share upstream drivers, and that intervening on those shared drivers might blunt several diseases at once rather than chase them one at a time (Chaib et al., 2022; Cossarizza, 2026).
Among the candidate drivers, cellular senescence has probably drawn the most sustained attention, in part because it sits at an unusually convenient intersection: it is mechanistically tractable, and — unlike, say, epigenetic drift — it is a cell state, which means it can in principle be targeted and cleared (Chmielewski, 2026; Fu & Zhou, 2025). The phenomenon itself is not new. Leonard Hayflick and Paul Moorhead described it in 1961, watching human diploid fibroblasts simply stop dividing after a finite number of passages — the so-called Hayflick limit (Hayflick & Moorhead, 1961; Sun, 2023). For decades this was treated as a curiosity of the culture dish. It is now understood rather differently: as a stress-induced, essentially irreversible cell-cycle arrest that unfolds in vivo in response to a wide range of insults — telomere attrition, persistent DNA damage, oncogene activation, mitochondrial dysfunction, reactive oxygen species, and therapy-induced genotoxic stress among them (Chaib et al., 2022; Kusmanto, 2025; Ozdemir et al., 2025). Senescent cells do not die quietly; they resist apoptosis and remain, somewhat perversely, highly metabolically active (Liao et al., 2021).
There is, frustratingly, no single molecular flag that marks a senescent cell wherever it happens to be. Researchers instead lean on a multiparametric signature — elevated cyclin-dependent kinase inhibitors, lysosomal senescence-associated β-galactosidase activity at pH 6.0, cellular hypertrophy, loss of lamin B1 at the nuclear envelope, persistent γ-H2AX foci marking unresolved DNA damage, chromatin remodeling, and intracellular lipofuscin accumulation (Chmielewski, 2026; Cohn et al., 2023; Erusalimsky, 2021; Gasek et al., 2021). None of these, taken alone, is diagnostic. Together, they are workable.
It would be a mistake, though, to think of senescence as simply bad. Transient senescence does real physiological work — it helps pattern the embryo, closes wounds, and, notably, acts as an autonomous brake on tumor formation (Alum et al., 2025; Paez-Ribes et al., 2019). The trouble starts when senescent cells stop clearing on schedule and instead accumulate. This is driven largely by the senescence-associated secretory phenotype, or SASP — a loosely regulated, tissue- and context-dependent cocktail of pro-inflammatory cytokines (IL-6, IL-1β, TNF-α), chemokines such as MCP-1, matrix-degrading metalloproteinases, and extracellular vesicles (Chmielewski, 2026; Cossarizza, 2026; Erusalimsky, 2021). Persistent SASP output produces a low-grade, sterile inflammatory state that has come to be called “inflammaging” — a term that has aged well, so to speak — which erodes tissue architecture, exhausts local stem-cell niches, and pushes neighboring healthy cells into so-called bystander senescence (Cossarizza, 2026; Fu & Zhou, 2025).
This is where the field runs into its central complication, and arguably the reason this review exists at all: senescent cells are not one thing. They differ, sometimes dramatically, by tissue of origin, by the nature of the inciting stressor (replicative exhaustion versus oncogene-induced versus therapy-induced versus metabolic stress), by how long they have been senescent, and even by the sex of the organism (Mansfield et al., 2024; Ozdemir et al., 2025; Sun, 2023; Zhang et al., 2026). A senescent pre-adipocyte and a senescent endothelial cell do not share a transcriptome, a SASP composition, or, crucially, a dependency on the same survival pathway (Sun, 2023; Zhang et al., 2026). This variability sat largely unresolved until single-cell and single-nucleus RNA sequencing, spatial transcriptomics, and machine-learning classifiers such as SenCID, SenePy, and SenMayo made it possible to actually see it — to watch senescent cells move through heterogeneous trajectories rather than collapse into a single uniform cluster (Cohn et al., 2023; Fu & Zhou, 2025; Mansfield et al., 2024). The practical upshot is sobering: because a universal senescent marker does not exist, a clearance strategy that works beautifully against one subpopulation may do nothing — or, worse, cause harm — against another (Cohn et al., 2023; Mansfield et al., 2024).
In response, researchers have built senotherapeutics, broadly split into two families. Senolytics kill senescent cells outright by transiently disabling the senescent-cell anti-apoptotic pathways (SCAPs) — Bcl-2/Bcl-xL, PI3K/Akt, p53/p21, ephrins, HSP90 — that senescent cells rely on to survive their own toxic secretome (Chaib et al., 2022; Gasek et al., 2021). The first generation of these agents — dasatinib plus quercetin (D+Q), fisetin, and the Bcl-2-family inhibitor navitoclax (ABT-263) — supplied the field’s original proof of concept, improving physical function, easing fibrosis, and extending healthspan in animal models, with early human trials in idiopathic pulmonary fibrosis, diabetic kidney disease, and Alzheimer’s disease confirming at least safety and target engagement (Chaib et al., 2022; Fu & Zhou, 2025; Riessland et al., 2024; Zhang et al., 2026). But these agents are blunt instruments: navitoclax causes dose-limiting thrombocytopenia and neutropenia, and cell-type specificity remains incomplete across the class (Fu & Zhou, 2025; Zhang et al., 2026).
Senomorphics take a gentler route — dialing down the inflammatory output of the SASP through mTOR, NF-κB, JAK/STAT, or p38 MAPK inhibition without forcing the cell to die (Ozdemir et al., 2025; Wyles et al., 2022). Repurposed drugs like metformin and rapamycin fall into this camp (Cossarizza, 2026; Paez-Ribes et al., 2019). The trade-off is dosing: because the cells survive, senomorphics need continuous administration, which trades the intermittent “hit-and-run” convenience of senolytics for a longer exposure window and, presumably, more chances for something to go wrong (Cohn et al., 2023; Fu & Zhou, 2025).
These limitations have pushed the field toward precision senotherapy — immune-based senolysis using CAR-T cells against uPAR or senolytic vaccines against GPNMB and CD153; targeted proteolysis via PROTACs that route Bcl-xL or BRD4 to E3 ligases for degradation while sparing platelet-rich tissue; and galactose-caged prodrugs and nanoparticles that stay inert until cleaved by the high SA-β-Gal activity unique to senescent lysosomes (Kusmanto, 2025; Paez-Ribes et al., 2019; Riessland et al., 2024; Zhang et al., 2026).
To connect this mechanistic detail to something clinically useful, this review asks three questions: (1) how does in vivo transcriptomic heterogeneity dictate a senescent cell’s susceptibility to a given SCAP inhibitor, and can tissue-specific “senotypes” predict responsiveness; (2) can precision platforms — uPAR-CAR-T, Bcl-xL PROTACs, SA-β-Gal-responsive prodrugs — genuinely sidestep the toxicities and resistance mechanisms (e.g., Mcl-1 upregulation) that limit first-generation agents; and (3) what are the long-term safety implications of chronic SASP suppression relative to intermittent senolysis, particularly for regeneration, wound healing, and immune surveillance. Guided by these questions, we set out to (a) systematically map layers of senescent heterogeneity using recent multi-omics and spatial data, (b) critically compare first- and next-generation senotherapeutics on efficacy, pharmacokinetics, and safety, and (c) sketch a translational roadmap that confronts the biomarker bottleneck and the design of trials suited to geriatric and oncology populations.
