Senotoxins target senescence via lipid binding specificity, ion imbalance and lipidome remodeling

Abstract

Senescence is a driver of aging and a barrier to tumor progression, but its persistent accumulation drives inflammation and relapse. Thus, the success of chemotherapy could be jeopardized when senescence emerges in the tumor microenvironment. Here we identified the senolytic properties of a pore-forming toxin, sticholysin I (StnI). StnI and our engineered improved form, StnIG, selectively hampers viability of chemotherapy-induced senescent cancer cells, as well as senescent primary cells. We show that its selectivity is mediated by specific binding and lipid ratios associated with senescence, including compromised membrane bilayer asymmetry. Mechanistically, StnIG triggers sodium and calcium influx and an enduring potassium efflux in senescent cells. Calcium triggers the opening of calcium-activated potassium channels, leading to cell death by apoptosis and pyroptosis. Finally we show that StnIG synergizes with senescence-inducing chemotherapy to drive remission of solid tumors in mice. Our findings define StnI and StnIG as senotoxins with translational potential for cancer therapy.

Moral-Sanz, Fernández-Carrasco and colleagues identify senolytic properties of sea anemone-derived pore-forming toxins, with selectivity mediated by senescence-associated lipid profiles. An optimized senotoxin improves the efficacy of chemotherapy in mouse models.

Main

Cellular senescence is a phenomenon characterized by growth arrest, with multifaceted implications ranging from embryonic development and tissue repair to aging and disease¹. Senescent cells accumulate in tissues over time and contribute to chronic inflammation through the secretion of pro-inflammatory factors, a phenomenon termed the senescence-associated secretory phenotype (SASP)². In cancer, although senescence initially acts as a barrier to tumor progression, it can later facilitate tumor growth and metastasis through the promotion of inflammation and tissue remodeling³. Remarkably, chemotherapy acts as a double-edged sword, creating a milieu of senescent cells that contribute to treatment-related adverse effects and potentially support tumor recurrence⁴. Consequently, new antitumoral therapies integrating chemotherapy with the intermittent use of drugs directed against senescent cells, termed senolytics, should reduce tumor resistance and relapse¹. Recently, various classes of senolytics have been identified with diverse targets, including the anti-apoptotic BCL2 family (navitoclax)⁵, the Na⁺/K⁺ ATPase (cardiac glycosides)⁶ and p53 (nutlin-3a)⁷. Other compounds like quercetin and fisetin, alone or in combination with dasatinib, entered clinical trials for senescence-associated diseases but have multiple and still poorly defined targets^8,9. In addition, these compounds tread a fine line, characterized by their narrow therapeutic range with off-target toxicity^10,11, low potency¹² or limited bioavailability¹³. Therefore, improved senotherapeutic approaches and compounds are needed to target senescent cells effectively and improve the management of associated diseases. Several approaches have been taken to address the limitations of these known senolytics, including prodrug modification and use of proteolysis targeting chimera technology to reduce on-target toxicity while improving senolytic activity^14,15.

Venoms and toxins, often associated with predation and defense, have emerged as a valuable resource for the identification of novel therapeutic agents¹⁶. Toxins have evolved to target specific physiological pathways that control tissue homeostasis and essential cellular functions, yielding exceptional specificity and potency¹⁷. For example, certain venoms contain peptides that hinder growth and dissemination of tumor cells^18–20. In addition, toxins serve as crucial tools for identifying fundamental biological processes to drive advancements in cellular physiology and molecular pharmacology.

Sticholysins are members of a group of pore-forming toxins (StnI, StnII and StnIII) originating from the venom of the Caribbean sea anemone Stichodactyla helianthus^21,22. Sticholysins exhibit a robust and potent antitumoral effect tested in vitro in several cellular models^23,24. StnI is a 176-amino-acid-long protein known for its ability to preferentially form pores in membranes that contain a specific balance of sphingomyelin (SM), phosphatidylcholine (PC) and cholesterol^25,26.

Altered lipid arrangements within cell membranes are reported in various human diseases, including cancer²⁷. Lipid composition shapes membrane properties and influences proteins that regulate ion permeability²⁸ and mitochondrial function²⁹. For example, aging and senescence alter lipid metabolism through distinct mechanisms that affect cholesterol, sphingolipids and glycerophospholipids^30–32. Cholesterol esters and SM in human sclera increases with aging due to lipid accumulation and metabolic shifts³⁰. Oncogene-induced senescence, such as in ERBB2-driven models, modifies mitochondrial glycerophospholipids, affecting membrane fluidity³¹. Senescent fibroblasts accumulate ceramides due to elevated sphingomyelinase activity, which reinforces the senescence phenotype³². In addition, lipid droplet composition in senescent cells modulates membrane structures, energy reserves and signaling³³. Overall, lipid alterations contribute to cellular dysfunction and present potential targets for senotherapies.

A key feature of healthy cells is the asymmetric distribution of phospholipids between membrane leaflets, with phosphatidylserine (PS) and phosphatidylethanolamine (PE) primarily restricted to the inner leaflet³⁴. This asymmetry becomes compromised during cellular senescence, leading to the exposure of these lipids on the cell surface^35,36. Given that lipid rearrangements are hallmarks of senescent cell membranes, exploring lipid-protein interactions is essential to understanding their functional consequences.

Actinoporins like StnI exhibit a remarkable ability to form pores in membranes through specific lipid recognition and binding. The crystal structure of a related actinoporin, fragaceatoxin C (FraC), demonstrated that lipids are not merely binding targets but integral structural components of the assembled pore complex³⁷. However, the dynamic interplay between protein residues and specific lipid species, particularly in the context of senescent versus healthy cell membranes, remains poorly understood.

In this article, we investigated the impact of SntI on fundamental cellular processes, elucidated its molecular interactions with lipids, assessed its selectivity for senescent cancer cells and evaluated the in vivo efficacy of an enhanced generated variant against cancer and chemotherapy-induced senescence. Furthermore, we provided proof of concept for a class of natural-derived compounds to target senescent cells, which could potentially set a distinct generation of senolytics. In this context, we introduce the concept of senotoxins.

Results

The sticholysin StnI exhibits senolytic properties in vitro

To investigate the potential senolytic activity of StnI, we chemically induced cellular senescence by inhibition of cyclin-dependent kinases 4 and 6 (CDK4 and CDK6) with palbociclib³⁸. The senescent phenotype included diminished p53 phosphorylation, heightened p21^Cip1 expression (Supplementary Fig. 1) and a notable senescence-associated β-galactosidase (SA-β-gal) activity with an enlarged and flattened morphology (Fig. 1a,d–g). Treatment with StnI (0.1 nM to 3 μM) led to a concentration-dependent reduction in viability of human melanoma SKMel-147 cells (Fig. 1b). We also identified a therapeutic window delimited by different potencies in proliferating (IC₅₀ ≈ 240 nM) and palbociclib-treated SKMel-147 cells (IC₅₀ ≈ 30 nM). This was translated into a senolytic index of 8 for StnI. Similar observations were made when senescence was induced with doxorubicin (Supplementary Fig. 2a). Additionally, we reported a time dependency in the senolysis, as indicated by longer exposures to low concentrations of StnI (10–30 nM; Supplementary Fig. 2b,c). The senolytic efficacy of StnI was comparable to navitoclax (Supplementary Fig. 2d), but with 13 times greater potency (Fig. 1c). The senolytic activity of StnI remained consistently potent across various senescent models induced by chemotherapy (palbociclib, doxorubicin and bleomycin), including human hepatoma (Huh7), lung adenocarcinoma (A549) and neuroblastoma (SH-SY5Y) (Fig. 1d–g). Interestingly, treatment with StnI did not change the senescent phenotype, as evidenced by the expression of p16^Ink4a, p21^Cip1 and p53 phosphorylation (Supplementary Fig. 1).

Fig. 1. The viability of StnI in senescent and proliferating cancer cells.

a, Representative images (left) and quantification (right, n = 4) for SA β-gal staining in SKMel-147 cells under control conditions or treated with palbociclib (5 µM, 7 days). Statistical significance was assessed by an unpaired two-tailed t-test. b, Concentration-dependent response curve for the effects of a 48 h treatment with StnI in the viability of proliferating SKMel-147 (black, control, n = 9) and SKMel-147 cells treated with palbociclib (5 µM, 7 days, n = 9). c, Comparison of StnI (red, n = 9) and navitoclax (purple, n = 4) by viability assays after 48 h of treatment as indicated. d–g, Representative images for the SA β-gal staining (left) and the concentration-dependent response curve (right) for StnI in the viability of proliferating (black) and senescent (red) cells. SKMel-147 treated with doxorubicin (30 nM, 7 days, n = 9) (d), Huh7 cells treated with palbociclib (10 µM, 7 days, n = 6) (e), A549 cells treated with bleomycin (20 µM, 5 days, n = 4) (f) and SH-SY5Y cells treated with palbociclib (10 µM, 7 days, n = 3) (g). Data are shown as mean ± standard error of the mean (s.e.m.). Statistical significance was assessed by one-way analysis of variance (ANOVA) with Šídák’s correction. Image insets in d–g indicate percentage of β-gal-stained cells. n refers to the number of biological replicates. For each biological replicate, three technical replicates were conducted.

Source data

Molecular modeling and improvement of StnI

To determine whether a structural region was responsible for the senolytic properties of StnI, we examined specific domains within the candidate protein (Supplementary Fig. 3). These domains include an amphipathic α-helix at the N terminus (aa 1–29), β-sheet rich (aa 58–101), an aromatic cluster (112–116), α-helix (aa 128–137) and an RGD sequence with potential integrin-binding (aa 142–144). All truncated forms of these chemically synthesized StnI domains failed to induce senolysis and displayed minimal cytotoxicity in either proliferating or senescent cells (Supplementary Fig. 3). This finding suggested that the full-length sequence and/or its conformation are required for the activity of StnI. Supporting this, senolysis was abolished upon protein denaturation (95°C for 45 min; Fig. 2f).

Fig. 2. Concentration-dependent response curve showing cell viability after 48 h.

a–d, StnI wild-type (n = 6) (a), StnII (n = 5) (b), StnI E2AD9A (n = 5) (c) and StnIG (n = 7) (d) in proliferating SKMel-147 cells (black, control) and SKMel-147 treated with palbociclib (5 µM, 7 days). e, Chart illustrates the senolytic index calculated as a ratio of the IC₅₀ in proliferative and senescent cells. f, Dot plot indicates viability of senescent SKMel-147 (palbociclib 5 µM, 7 days) after 48 h with the indicated treatments (n = 3). g–i, Concentration-dependent response curve on the viability of StnIG after 48 h in proliferating (black, control) and SKMel-147 + 30 nM doxorubicin (n = 3) (g), A549 + 20 µM bleomycin (n = 6) (h) and SKMel-147 + 5 µM bleomycin (n = 3) (i). StnI and StnIG denaturation was achieved by incubation at 90°C for 45 min. Data are shown as mean ± s.e.m. Statistical significance was assessed by one-way ANOVA with Šídák’s correction. n refers to the number of biological replicates. For each biological replicate, three technical replicates were conducted. Vh, vehicle.

Source data

We then compared the activity of StnI with a closely related variant (StnII; Supplementary Fig. 4a)²². Despite sharing 93% sequence homology, StnII exhibited 47 times lower cytotoxicity in senescent cells compared to StnI (IC₅₀ ≈ 1.4 µM vs 30 nM) with similar IC₅₀ values in control (≈1.3 µM) and palbociclib-treated cells (≈1.4 µM), yielding a selectivity index (SI) of 0.92 (Fig. 2b,e,f). The N-terminal end of StnI includes a series of acidic residues that contribute to a negative electric charge (E2, D9 and E16), whereas the net change is neutral for StnII (A1, A8 and Q15). Interestingly, substitution of two of the three negatively charged residues at the N terminus (StnI E2AD9A) reduced the cytotoxicity in both senescent (IC₅₀ ≈ 100 nM) and proliferating SKMel-147 cells (IC₅₀ ≈ 400 nM) (Fig. 2c,e,f and Supplementary Fig. 2c). However, this N-terminal modified form maintained certain degree of senolysis (SI = 4; Fig. 2e), with an amino-end hydropathy similar to StnII and the exact same C terminus as StnI (Supplementary Fig. 4). This finding suggests that the properties of the carboxyl end might contribute to the selective cytotoxicity of StnI. We then modified StnI by altering its C terminus polarity and enhancing its purification. This involved introducing a methionine at the amino-end, providing a signal sequence removed upon secretion in recombinant synthesis³⁹ and a 6xHis at the C terminus (StnIG). StnIG showcased a higher C-terminal polarity than StnI and StnII (Supplementary Fig. 4) but had a lower cytotoxicity in proliferating cells (IC₅₀ ≈ 1 µM) while maintaining the senolytic selectivity (IC₅₀ ≈ 30 nM; Fig. 2d,g–i). The senolytic profiling of StnIG was also validated by crystal violet staining (Supplementary Fig. 5). Overall, these modifications improved the senolytic index of StnIG (SI = 31.25) as compared to its native form (SI = 8) or navitoclax (SI = 22.4) (Fig. 2e). Importantly, the impact of a 24-h incubation with StnIG proved irreversible and persisted in the senescent group, whereas proliferating cells exhibited a gradual recovery (Supplementary Fig. 6).

Genome-wide gene analysis reveals that StnI alters cellular metabolism, ion homeostasis and biological pathways

Bulk RNA sequencing of palbociclib-induced senescent cells treated with StnI (100 nM) at 3, 6 and 12 h revealed primarily disruptions in expression of genes important in cell metabolism. The results showed that at 6 and 12 h of StnI treatment consistently deregulated major metabolic pathways (pentose phosphate, oxidative phosphorylation, Krebs cycle, metabolism of lipids, amino acids and nucleotides). The data also showcased downregulation of key signaling pathways in cancer (for example, TGF-β, JAK STAT, MAPK, NOTCH, PPAR and p53) (Extended Data Fig. 1). Furthermore, StnI initiated early deregulation of several key genes in sphingolipid biosynthetic and metabolic pathways. Interestingly, we also observed signs of ionic imbalance (Extended Data Fig. 2a) and deregulation of genes related to ion transport and homeostasis. This included 41 genes linked to membrane potential, affecting chloride, calcium-gated, and voltage-gated potassium channels, as well as 40 genes associated with calcium homeostasis and signaling, such as store-operated calcium modulators and calcium uniporters (Extended Data Fig. 2b–d). Altogether, these findings suggest StnI may induce metabolic and ionic imbalance in senescent cells.

Extended Data Fig. 1. StnIG treatment dynamically modulates key signaling pathways in senescent SKMEL-147 cells.

Gene-set enrichment analysis (GSVA) indicates the most important deregulated signaling pathways in senescent SKMel-147 cells (palbociclib 5 µM, 7 days) under control conditions (left) or after 3-6-12 h treatment with StnI 100 nM. Four biological replicates were done per treatment.

Extended Data Fig. 2. StnIG induces transcriptional changes in membrane transport and calcium signaling pathways in senescent SKMEL-147 cells.

(a) Representative images of proliferating (up) or senescent (down) SKMel-147 treated with vehicle or 100 nM StnIG for 1 h. (b) Enriched terms across the 193 differentially expressed genes involved in membrane transport of molecules and ions. The vertical axis represents the GO category, and the horizontal axis represents the –Log10(P-value) of the substantial GO terms as determined with metascape.org. Bar chart indicates the relative fold change for genes involved (c) ion transport and (d) calcium homeostasis signaling following a 12 h treatment with 100 nM StnI in senescent SKMel-147.

StnI and StnIG induced a long-lasting impairment of mitochondrial respiration

To investigate mitochondrial function, we assessed metabolic fluxes using Seahorse technology and mitochondrial membrane potential by flow cytometry with rhodamine 123 (Rhod123) and TMRM dyes. Acute StnI or StnIG treatment caused a transient increase in cellular respiration within the first 15 min, followed by a steady decline (StnI: −33.5% ± 7%, n = 4; StnIG: −31.1% ± 2%, n = 6), an effect not seen with denatured StnIG (−9.1% ± 1%, n = 6) (Fig. 3a,b). In proliferating cells, StnIG did not reduce basal respiration (Fig. 3b), equating to half the decrease observed in senescent cells (proliferating + StnIG: −15.1% ± 2% versus senescent + StnIG: −31.1% ± 2%; Fig. 3b). Senescent cells exhibited persistent impaired respiration up to 24 h, characterized by increased proton leak, reduced ATP-coupling efficiency, lower maximal respiration, and decreased spare respiratory capacity (Fig. 3c,d). The impaired respiration correlated with mitochondrial membrane depolarization (Fig. 3e,f) and increased apoptosis (Supplementary Fig. 7a). The detection of Annexin V-positive cells indicates the externalization of PS lipids on the outer leaflet of the plasma membrane in senescent SKMel-147 cells, a hallmark of apoptosis initiation. Senescent cells treated with 100 nM StnIG for 1 h showed higher depolarization (Rhod123, low ΔΨ population: control 11.7% ± 2%, n = 9 versus StnIG 21.2% ± 3%, n = 3: P = 0.0043), an effect not seen with PBS, 100 nM StnII or denatured StnIG.

Fig. 3. Metabolic assessment of senotoxins in SKMel-147 cells.

a, Representative traces in senescent SKMel-147 cells for normalized oxygen consumption rates (OCR) to show real-time measurements. Arrows signal addition of compounds to interrogate the respiratory parameters summarized in bar charts. b–d, Plot indicates the registered acute change in OCR normalized to baseline (b), and bars show mitochondrial parameters normalized to basal mitochondrial respiration, including (i) ATP-coupling efficiency, (ii) proton leak, (iii) maximal mitochondrial respiration and (iv) spare mitochondrial respiration calculated for palbociclib-induced senescent cells (c) and proliferating cells following the indicated treatments (d). Vehicle (n = 6), 100 nM StnII (n = 3), 100 nM denatured StnIG (n = 6), 100 nM StnIG (2 h, n = 6; 24 h, n = 5) and 100 nM StnI (n = 4). e,f, Mitochondrial membrane potential determined by flow cytometry with (i) Rhodamine123 and (ii) TMRM dyes for palbociclib-induced senescent cells (e) and proliferating cells (f) following a 24 h incubation with 100 nM StnI, 100 nM StnIG, 100 nM StnII, 100 nM denatured StnIG or 2 h with 50 μM carbonyl cyanide m-chlorophenyl hydrazone (CCCP). Data are shown as mean ± s.e.m. Statistical significance was assessed by one-way ANOVA test with Dunnett’s correction in b–d or by two-way ANOVA test with Dunnett’s correction in e and f. n refers to the number of biological replicates. For each biological replicate, three technical replicates were conducted.

Source data

StnIG promoted a sustained potassium efflux and plasma membrane hyperpolarization in senescent SKMel-147 cells

Senescent cells exhibit marked plasma membrane deopolarization^6,40, consistent with altered intracellular electrolyte composition^40,41 and a potential impaired permeability to K⁺ across the plasma membrane. We performed patch-clamp analysis using 4-aminopyridine (4AP) and tetraethylammonium (TEA) to block voltage-gated (K_v) and calcium-activated (KCa) potassium channels, respectively. Both groups presented slow activating, non-inactivating macroscopic currents (Supplementary Fig. 8a). Senescent cells demonstrated a three- to fivefold elevation in capacitance, indicating larger size (Supplementary Fig. 8b). Given the difference in cell sizes, we compared ion channel contributions both as percentage of basal current and current density (pA/pF). Senescent SKMel-147 cells exhibited more depolarized membrane potential and reduced current density above −40 mV (Supplementary Fig. 8c,d). Pharmacological characterization of the currents highlighted that SKMel-147 cells were more sensitive to TEA than 4-AP (Supplementary Fig. 8e,f), aligning with previous studies^42,43. TEA-sensitive current density was significantly reduced in senescent cells, whereas 4AP sensitivity remained unchanged (Supplementary Fig. 8f). The relative contribution of TEA- and 4AP-sensitive channels was constant across conditions (Supplementary Fig. 8g), indicating KCa channel predominance regardless of senescence.

Senescence induction changed cell size but not potassium channel composition. mRNA analysis showed upregulation of the big-conductance KCa channel (BKCa, KCNMA1) and its auxiliary subunit (KCNMB4) (Supplementary Fig. 8h), the latter being deregulated after senotoxin treatment (Extended Data Fig. 2c). Of relevance, KCNMB4 downregulates BKCa channel activity⁴⁴ and modulates its calcium sensitivity^45,46.

Given the pore-forming ability of sticholysins, we examined the effect of StnIG in cells. Previous reports in artificial membranes showed that sticholysins create pores highly selective for monovalent cations⁴⁷ and dependent upon lipid composition²⁶. However, cellular membranes may behave differently. Acute StnIG treatment in senescent cells led to a major outward current with a time-dependent rectification (Fig. 4a). Under physiological K⁺ gradients (5 mM extracellular, 135 mM intracellular), StnIG shifted the reversal potential leftward (−35 ± 3 mV control versus −64 ± 5 mV StnIG; P = 0.0137; n = 5). Under symmetrical K⁺ conditions, reversal potential approached 0 mV, supporting enhanced K⁺ permeability. Notably, 10 nM StnIG caused membrane hyperpolarization and outward K⁺ currents in both proliferating and senescent cells, with stronger effects in the latter (Fig. 4b,c), showing a distinct current density at the peak effect (40 ± 7 pA/pF vs senescent, 77 ± 5 pA/pF; P < 0.001; n = 4–5).

Fig. 4. Electrophysiological examination of SKMel-147 cells.

a, Representative time course (left) for the voltage-clamped currents in palbociclib-induced senescent SKMel-147 cells measured at the steady state of the voltage step to −100 mV (white) and +60 mV (black) and the associated voltage-clamp protocol and records (right). Numbers indicate matching traces and time points in the time course as follows: (1) before the addition of 10 nM StnIG, (2) at the maximal effects under asymmetrical K⁺, (3) at the maximal effects under symmetrical K⁺, at the maximal effects under asymmetrical K⁺ and (4) 8 min after recovery. b, Change in membrane potential following acute application of 10 nM StnIG in proliferating (black, n = 4) and palbociclib-induced senescent SKMel-147 cells (red, n = 5). Prol, proliferating; Sen, senescent. c–e, Time courses (left) and calculated area under the curve (AUC) for the change in (c) the outward K⁺ current density (n = 5) (c), intracellular content of Na⁺ (n = 3) (d) and Ca²⁺ (n = 3) (e) in proliferating (black) and senescent (red) SKMel-147 cells. **P = 0.0040. Results are expressed as mean ± s.e.m. Statistical significance was assessed by two-sided unpaired Student’s t test. n refers to the number of biological replicates.

Source data

StnIG also increased Na⁺ and Ca²⁺ levels similarly across cell types (Fig. 4d,e), with extracellular Ca²⁺ as the main source (Supplementary Fig. 9). These observations are consistent with the cation selectivity for StnI pores⁴⁷ and suggest preferential K⁺ efflux in senescent cells, possibly due to altered lipid composition^30–32 and membrane dynamics²⁶. Prolonged StnIG exposure led to a sustained hyperpolarization, heightened current density and decreased cell capacitance in senescent cells (Fig. 5a–c). The enduring K⁺ current was sensitive to TEA (Fig. 5d–g). These alterations were specifically observed in StnIG-treated senescent cells (Supplementary Fig. 10) and were consistent with enhanced KCa activity and a sustained loss of K⁺ loss linked to cell-volume decrease⁴⁸.

Fig. 5. StnIG alters electrophysiological properties and viability in senescent SKMel-147 cells.

a–c, Resting membrane potential (a), cell capacitance (b) and current–voltage relationship (c) for palbociclib-induced senescent cells treated for 6–20 h with 100 nM StnIG (brown; n = 10, 12, and 10, respectively in a–c) or parallel senescent controls (red; n = 19, 14 and 13, respectively in a–c). d–g, Representative voltage-clamp protocol and current traces under control conditions and following the sequential application of 100 nM apamin, 1 µM tram-34 and 100 nM charybdotoxin (ChTx) in senescent control (d) and StnIG-treated (e) cells. Dot plots represent the current sensitive to 10 mM TEA (n = 8 for senescent control and senescent + StnIG), 100 nM apamin (n = 5 for senescent control and senescent + StnIG), 1 µM tram-34 (n = 4 for senescent control and n = 5 in senescent + StnIG) and 100 nM charybdotoxin (ChTx, n = 5 for senescent control and senescent + StnIG) in both groups expressed as change in current density (f) and percentage of basal (I₀) current analyzed at +60 mV (g). I_blocker stands for the current measured in the presence of specific blockers, in this case of TEA, apamin, Tram-34 or ChTx. h, Free calcium content analyzed by fluo4 AM staining and flow cytometry as the median fluorescence index (MFI) for the FITC channel normalized by cell size (FITC-A/FSC-A, n = 5 for senescent control and n = 4 in senescent + StnIG). i, mRNA expression of BKCa and KCNMB4 in senescent cells treated with vehicle or StnIG 100 nM (n = 4), expressed relative to control senescent cells. j,k, Viability at 48 h of treatment with StnIG in senescent cells supplemented with 50 mM KCl (n = 5) (j) or treated with 1 mM EGTA (n = 10) or 4 mM CaCl₂ (n = 5) (k). Results are expressed as mean ± s.e.m. Statistical significance was assessed by two-sided unpaired Student’s t test in a, b and f–j, two-way ANOVA test with Dunnet’s correction for multiple comparisons in k and two-sided unpaired Student’s t test with Holm-Šídák correction for multiple comparisons in c. Adjusted P values in c for senescent cells versus senescent + StnIG are as follows: −30 mV, *P = 0.0117; −20 mV, **P = 0.0017; −10 mV, ***P = 0.0004; 0 mV, ***P = 0.0001; +10 mV, **P = 0.0021; +20 mV, **P = 0.0043; +30 mV, **P = 0.0043; +40 mV, **P = 0.0042; +50 mV, *P = 0.0125 and +60 mV, **P = 0.0051. n refers to the number of biological replicates.

Source data

To assess KCa channel contributions, we used selective blockers and found that prolonged StnIG treatment increased IKCa (intermediate; tram-34) and BKCa (big; charybdotoxin), but not SKCa (small; apamin) currents (Fig. 5f,g). StnIG also prolonged intracellular Ca²⁺ elevation and downregulated the BKCa regulatory subunit KCNMB4 (Fig. 5h,i), supporting increased KCa activity and plasma membrane hyperpolarization in senescent cells. Importantly, the senolytic effect of StnIG was partially blunted by EGTA (1 mM, Fig. 5k) and high extracellular K⁺ (Fig. 5j).

StnIG triggers calcium-induced apoptosis and pyroptosis by a sustained BKCa activation

Intracellular calcium levels remained elevated in senescent only cells following StnIG treatment (Fig. 5h and Supplementary Fig. 10g), aligning with sustained BKCa activation and potassium loss. Indeed, potassium is a key intracellular osmolyte, which regulates cell volume, apoptosis^48–50 and caspase activation, including pyroptosis^50–52. The cytotoxic effect of StnIG was potentiated by the BKCa activator BMS-191011 (Fig. 6a and Supplementary Fig. 10i). To dissect the mechanism, we used a BKCa blocker (charybdotoxin) and inhibitors of caspases (z-VAD FMK), pyroptosis (VX-765) and necroptosis (necrostatin-1). StnIG-induced senolysis was independent of necroptosis (100 nM StnIG: 64.2 ± 5.9 versus StnIG + necrostatin: 55.2 ± 6.9, P = 0.059) but sensitive to caspase-1/4 inhibition, calcium chelation (EGTA, BAPTA) and BKCa blockade (charybdotoxin) (Fig. 6b–e). Flow cytometry confirmed that StnIG induces early apoptosis independently of BKCa activation (StnIG 15.5% ± 1.0% versus StnIG + charybdotoxin 12.7% ± 1.8%, P = 0.095). Meanwhile, calcium influx and BKCa activity facilitated cell-volume decrease and late apoptosis in senescent cells, as confirmed with CaCl₂ and charybdotoxin treatments (Fig. 6f,g).

Fig. 6. Role of calcium-activated potassium channels in StnIG-induced apoptosis in senescent SKMel-147 cells.

a–d, Viability of senescent cells analyzed with crystal violet for the effects of StnIG under the control condition (black) and following preincubation with the BKCa activator (30 µM BMS-191011, n = 3) (a), the calcium-activated potassium channel blocker (100 nM charybdotoxin, ChTx, n = 4) (b), in the presence of calcium chelators (1 µM BAPTA, n = 4 or 1 mM EGTA, n = 4) (c), a pan-caspase inhibitor (40 µM z-VAD FMK, n = 4) or a caspase-1/4 inhibitor (10 µM VX-765, n = 4) (d). e, Representative images for crystal violet staining of palbociclib-induced senescent SKMel-147 cells preincubated with 1 mM EGTA or 100 nM charybdotoxin (ChTx) and treated with vehicle or 100 nM StnI1G for 24 h, obtained from experiments in b and c. f,g, Dot plot indicates flow cytometry values for the percentage population of living cells, early apoptosis, late apoptosis and necrosis in palbociclib-induced senescent cells following 24 h of the indicated treatments (f) and the observed changes in cell size expressed as a percentage of the mean FSC-A under control conditions (g) (n = 3 in f and g). Results are expressed as mean ± s.e.m. Statistical significance was assessed by two-way ANOVA test with Dunnet’s correction for multiple comparisons in a–d; or one-way ANOVA test with Dunnet’s correction for multiple comparisons in f and g. n refers to the number of biological replicates. For each biological replicate, three technical replicates were conducted.

Source data

The mediation of lipids on the cytotoxicity of StnIG

Specific lipid species are known to influence senescence and SASP expression³³, but the functional role of lipidome remodeling remains unclear. We compared proliferating and senescent SKMel-147 cells and revealed a distinct lipid profile (Supplementary Fig. 11a), with changes across glycerophospholipids, sphingolipids, storage lipids and lysolipids (Extended Data Fig. 3a–d and Supplementary Fig. 11b). Senescent cells exhibited decreased PS, SM and mitochondrial cardiolipin (CL) but elevated PC, PE ether (PE O-), cholesterol esters (CE) and triacylglycerol (TAG). Interestingly, four out of eight lysolipids were increased during senescence, including LPC (lyso-phosphatidylcholine), LPE-O (lyso-phosphatidylethanolamine-ether), LPI (lyso-phosphatidylinositol) and LPS (lyso-phosphatidylserine) (Extended Data Fig. 3a–d and Supplementary Fig. 11b). These findings suggest a possible role for lyso-phospholipids in SASP-related inflammation⁵³ as well as in membrane and endomembrane reorganization accompanied by enhanced lipid storage during senescence.

Extended Data Fig. 3. Plots indicate lipid content standarized to total lipids (%mol) in SKEML-147 cells.

(a) Glyceraophospholipids, (b) sphingolipids, (c) storage lipids, and (d) lysolipids in proliferating (black, n = 4) and in palbociclib-induced senescent (red, 5 µM, 7 days, n = 4) cells. Concentration-dependent response at 48 h of treatment with StnIG in the viability of (i) proliferating or (ii) senescent cells under control conditions (n = 8) and supplemented with (e) 100 µM palmitoyl-2-oleoylphosphatidylcholine (PC, n = 5), (f) 100 µM sphingomyelin (SM, n = 3), or 100 mU/mL sphyngomyelinase (SMAse, n = 3) and (g) lipoprotein-depleted serum (LPDS, n = 4). Abbreviations: CE, Cholesterol esters; Cer, Ceramide; CL, Cardiolipin; DAG, Diacylglycerol; HexCer, Hexosylceramide; PS, Phosphatidylserine; PG, Phosphatidylglycerol; PI, Phosphatidylinositol; SM, Sphingomyelin; TAG Triacylglycerol; LPA, lyso-Phosphatidate; LPC O-, lyso-Phosphatidylcholine (-ether); LPE O-, lyso-Phosphatidylethanolamine (-ether); LPI, lyso-Phosphatidylinositol; LPS, lyso-Phosphatidylserine; PA, Phosphatidate; PC O-, Phosphatidylcholine (-ether); PE O-, Phosphatidylethanolamine (-ether). Relative abundances were standardized to the total lipid amount (mol%) per sample. Data are shown as mean ± SEM. Statistical significance was assessed by one-way ANOVA test with Šídák’s correction. N refers to the number of biological replicates.

Source data

We next examined PC, SM, and cholesterol in StnIG-induced cytotoxicity. Senescent cells contained higher PC and lower SM (Extended Data Fig. 3a,b and Supplementary Fig. 11b), consistent with increased sphingomyelinase activity³². PC supplementation (100 µM POPC, 1-palmitoyl-2-oleoylphosphatidylcholine) sensitized proliferating cells to StnIG (Extended Data Fig. 3e), mimicking senescent cell responses. SM degradation by exogenous SMAse also potentiated StnIG cytotoxicity (Extended Data Fig. 3f), supplementing SM (100 µM) mitigated StnIG effects in senescent cells but increased toxicity in proliferating cells (Extended Data Fig. 3f). In contrast, lipid depletion using lipoprotein-depleted serum (LPDS) amplified StnIG effects in proliferating cells and reduced them in senescent cells (Extended Data Fig. 3g).

Molecular basis of StnIG preference for senescent cells showcases a selective recognition of exposed PE/PS lipids

To understand the selectivity of StnIG for senescent cells, we modeled its structure based on the high-resolution crystal structure of FraC (PDB: 4TSY), a related and homologous actinoporin (61% sequence identity and 80% similarity). The modeled structure exhibits two alpha-helical segments: the membrane-penetrating helix (α1, Fig. 7a), which spans the lipid bilayer and forms part of the transmembrane pore, and the membrane-facing helix (α2), which lies parallel to the outer leaflet of the bilayer and interacts with lipid headgroups to stabilize the attachment of the protein to the membrane^37,54,55.

Fig. 7. Lipid-specific interactions and membrane dynamics of the StnIG pore complex.

a, Right: equilibrated octameric pore system showing transmembrane α1 and membrane-facing α2 helices (red), β-sandwich core (purple) and membrane phosphate headgroups (orange spheres). Left: α2 helix (orange) with key lipid-interacting residues. Model based on FraC crystal structure (PDB 4TSY). b, Distribution of lipid–protein contact distances between α2 helix residues (Gln130, Tyr133, Glu134, Tyr137 and Tyr138) and phospholipid headgroups. Violin plots show probability density with embedded box plots (median: center line; interquartile range: box; whiskers: 1.5× IQR). Data aggregated over n = 5 independent replicates per system (500 ns each), representing 610,278 contacts defined as ≤4.0 Å between protein and lipid headgroup heavy atoms. c, Glu134–lipid contact stability. Box plots (center line: median; box: 25th–75th percentiles; whiskers: 1.5× IQR) with overlaid individual data points show contact durations (ns, log scale) combining n = 5 replicates per system. Sample sizes: 13,485 episodes (PE 5,838; PS 1,830; PC 5,384; SM 394; PI 39). Two-sided Mann–Whitney U tests: PE versus PC (U = 20,395,678, P = 8.07 × 10⁻¹⁷⁰, ****); PS vs PC (U = 5,750,612, P = 1.40 × 10⁻²⁸, ****). d, Hydrogen bonding between Glu134 and a POPE lipid. Top: structural snapshot showing a hydrogen bond (dashed line) between Glu134 and the PE amine group. Bottom: time series of the donor–acceptor distance showing stable hydrogen bonding interaction (~3 Å) over 150 ns. e, Residue-specific lipid preferences in senescent membranes. Donut charts show lipid contact distribution. Glu134 and Gln130 prefer PE (84.0%, 63.3%). f, Glu134–lipid contact duration distributions. Bar plots show percentage binned by duration from n = 5 replicates per system (asymmetric 1,123 episodes; proliferating 3,610; senescent 3,311). Two-sided chi-square tests (df = 5): asymmetric versus senescent (χ² = 133.13, P = 1.55 × 10⁻²⁶); asymmetric versus proliferating (χ² = 112.10, P = 4.42 ×10⁻²²); senescent versus proliferating (χ² = 10.35, P = 0.197, NS). Two-sided Fisher’s exact test for >10 ns contacts: senescent versus asymmetric (P < 0.0001). g, Phosphate mobility in upper leaflet. Top: spatial RMSF distribution from replicate 1 (last 10 ns). Bottom: data shown as mean ± s.d. fraction across bins across n = 5 replicates with individual replicate data points overlaid. Two-sided Kolmogorov–Smirnov tests: asymmetric versus proliferating (P = 0.044); asymmetric versus senescent (P = 0.018); proliferating versus senescent (P = 0.45). Two-sided chi-square tests (df = 4): asymmetric versus proliferating (P = 0.043); asymmetric versus senescent (P = 0.018); proliferating versus senescent (P = 0.71).

We first embedded the StnIG model in a symmetrical lipid bilayer, mimicking the composition of the senescent membrane lipidomics analysis (Extended Data Fig. 3 and Supplementary Table 1), immersed in 0.15 M KCl solution for 500 ns. The α2 helix emerged as critical for the initial recognition and stable attachment of the toxin to the membrane surface. This helix contains five highly conserved residues among actinoporins, which are positioned to interact directly with membrane lipids: Gln130, Tyr133, Glu134, Tyr137 and Tyr138 (Fig. 7a). Detailed analysis of their interactions with membrane lipids revealed that only Glu134 and Gln130 formed consistent short-range contacts (<3 Å) with PE and PS lipids. Glu134 contributed nearly 30% of all α2 lipid contacts (Fig. 7b), and formed significantly closer contacts with PE/PS lipids compared to PC lipids (median distances 2.81 Å vs 3.82 Å respectively, Mann-Whitney U test, P < 0.001, Cohen’s d = −2.15). Inspection of the trajectories revealed that these short-range contacts were hydrogen bonds formed between the primary amine and serine hydroxyl groups of PE/PS lipids as well as the carboxylate and amide groups of Glu134 and Gln130, respectively (Fig. 7d). These hydrogen bonds were remarkably stable, with significantly longer durations compared to occurring interactions with other lipid types (Fig. 7c,d). This specificity was particularly evident in the contact distribution analysis, where PE lipids constituted 27.8% of the membrane composition but accounted for 84% of Glu134’s contacts and 63.3% of Gln130’s contacts (Fig. 7e), a highly significant overrepresentation (χ² = 271,032, P < 0.001, Cramer’s V = 1.25). In contrast, the three tyrosine residues (133, 137 and 138) showed a more generalized interaction pattern that closely mirrored the overall membrane composition.

The selective interaction of Glu134 and Gln130 with PE/PS lipids stemmed from their ability to form hydrogen bonds with these lipids’ smaller headgroups, whereas the bulky choline group of PC lipids may sterically hinder such close interactions. This observation is particularly relevant, as PE and PS lipids are typically sequestered to the inner leaflet of healthy membranes by lipid flippases³⁴ but become exposed on the outer leaflet during cellular senescence due to disrupted phospholipid asymmetry^35,36.

To explore whether this could explain the selectivity of StnIG to target senescent cells, we constructed two additional membrane models. The first was a symmetric proliferating membrane incorporating the altered lipid ratios observed in the lipidomic analysis (Supplementary Fig. 11) The second was an asymmetric proliferating system where PE and PS lipids in the outer leaflet were replaced with PC lipids to better represent the maintained lipid asymmetry of healthy cells.

Comparative analysis of these three systems revealed distinct patterns of protein-lipid interactions. Although long-lived contacts (>10 ns) between Glu134 and PE/PS lipids were frequently observed in the senescent membrane system, these interactions were significantly reduced in the symmetric proliferating system and nearly absent in the asymmetric system (Fig. 7f). Analysis of phosphate group mobility further highlighted that the asymmetric system had notably higher lipid mobility in the outer leaflet compared to both senescent and proliferating systems (Fig. 7g). These findings suggest that the exposure of PE/PS lipids in senescent membranes may stabilize StnIG, contributing to its enhanced activity against senescent cells.

In vitro reduction of SASP by StnI

To assess the effects of StnI on the SASP, we performed cytokine arrays to identify the main related components modulated in senescent SKMel-147 cells. To eliminate potential confounding effects from cell death, SASP detection was conducted specifically in viable cells post-treatment or control at 24 h. Palbociclib-induced senescence increased several interleukins, chemokines, colony and growth factors (Extended Data Fig. 4a,b) in accordance with the literature⁵⁶. Subsequently, we assessed the potential regulatory impact of StnI on SASP. Low concentrations of StnI (10-30 nM, 96 h) were sufficient to diminish cytokines secretion (IL-6, IL-1β and CCLX10), or fully revert levels comparable to those observed in proliferating cells (IL-1α, CCL2, CCL5, CCL11, MIP-1a, CXCL12 and G-SCF).

Extended Data Fig. 4. SASP assessment in SKMEL-147 cells.

(a) Images of arrays for the SASP secreted by proliferating (control) cells as compared to senescent cells (control, palbociclib 5 µM, 7 days), and senescent cells treated with 10 or 30 nM StnI for 4 days. (b) Heatmap representing raw pixel intensity detected in the array shown in a.

Source data

The senolytic and senomorphic effect of StnIG in human fibroblasts

The senolytic effect of StnIG was evaluated in a range of non-transformed cell lines. We observed no senolytic effect in chemotherapy-induced senescent neonatal foreskin fibroblasts or dermal fibroblasts (Supplementary Fig. 12a,b). However, a clear senolytic effect was observed in replicative senescent fibroblasts (Supplementary Fig. 12c), highlighting a promising potential for StnIG in extending healthspan. To assess the senotherapeutic potential of StnIG, we performed a high-content screening based on the phenotypic SA-β-gal activity in senescent cells. Human IMR90 lung fibroblasts served as the senescent cell model, whereas non-senescent cells were used as controls. StnIG was evaluated in senescent IMR90 cells over a 60-h treatment period using the fluorogenic dye C₁₂FDG to stain for SA-β-gal activity⁵⁷. Senescence levels were quantified by calculating the percentage of C12FDG-positive cells in each sample (Extended Data Fig. 5).

Extended Data Fig. 5. The senomorphic effect of StnIG in IMR90 cells.

(a) Representative images illustrating the senotherapeutic effects of StnIG 11.1 nM and 100 nM on senescent human IMR90 fibroblasts (20 μM etoposide for 24 h) and quantitative analysis (n = 3). Images were acquired using the Cytation 1 system at 4X magnification. Blue fluorescence corresponds to Hoechst 33324-stained nuclei and green fluorescence highlights C12FDG-stained SA-β-gal-positive senescent cells. Data are presented as percentages of relative C12FDG+ cells in total senescent cells (blue) and C12FDG positive cells (geen). Scale bar shown in white (control panel) at 1000 μm. Results are expressed as mean ± SEM. Statistical significance was assessed by two-sided unpaired Student’s t test. (b) Dose-response evaluation of the effects of a 60 h treatment with StnIG in the viability of proliferating IMR90 cells (black, control) and senescent IMR90 cells treated with etoposide (red, 20 μM etoposide 24 hours). Data are shown as mean ± SEM. Statistical significance was assessed by one-way ANOVA test (n = 3). N refers to the number of biological replicates.

Source data

To differentiate between senolytic and senomorphic effects, we determined the effective concentration (EC₅₀) of StnIG for both total and C₁₂FDG-positive IMR90 cells. Compounds that reduced both the total cell population and C12FDG-positive cells were classified as senolytics, whereas those that selectively reduced C12FDG-positive cells without affecting total cell numbers were categorized as senomorphics. StnIG demonstrated a potent senomorphic effect, significantly reducing SA-β-gal activity associated with the senescent phenotype without inducing cell death, with an EC₅₀ value of 10.13 nM. It also exhibited a strong senolytic effect at an EC₅₀ value of 134.6 nM, effectively decreasing the senescent cell population through cell death (Extended Data Fig. 5). These results highlight the distinct senotherapeutic activities of StnIG in IMR90 cells and a promising application to explore beyond cancer research as well as what drives senolysis in non-transformed cell lines.

Validation of StnIG efficacy in zebrafish and mice xenograft models with simultaneous CDK4/6 inhibition

To further validate the senolytic activity of StnIG in vivo, we first examined its efficacy in the AVATAR xenograft tumor zebrafish model with senescent SKMel-147 cells. We demonstrated that 10 nM StnIG showed no toxicity and significantly decreased the accumulation of senescent SKMel-147 cells at 6 days post-injection into the Cuvier circulatory duct of zebrafish embryos. Its efficacy was comparable to 1 µM navitoclax (Fig. 8a,b).

Fig. 8. In vivo validation of StnIG.

a, Representative images of zebrafish embryos’ tails xenografted with palbociclib-induced senescent SKMel-147 cells treated with vehicle or 10 nM StnIG at 6 days post-injection. Images show Dil fluorescence of accumulated senescent SKMel-147 (left) and the bright field merged image. b, Violin plot indicates values for senescent SKMel-147 accumulation normalized as the fold of the mean control in embryos treated with vehicle control (n = 30), 10 nM StnIG (n = 20) or 1 µM navitoclax (n = 36) for 6 days. c, Tumor progression in a xenograft melanoma mouse model of senescence induced by palbociclib (100 mg kg⁻¹, daily, oral gavage) evaluated as the fold change relative to the tumor volume observed before initiation of treatments (day 0) with palbociclib + navitoclax (50 mg kg⁻¹, daily, oral gavage, 8 doses, n = 4), palbociclib + StnIG (2 mg kg⁻¹, every 48 h, i.p., 4 doses, n = 4). Xenograft tumors were measured every day using a digital caliper and applying the algorithm: volume = a × b × b × 0.5, where a is the length and b is the measured breadth of the tumor lump (skin included). Tumor progression was estimated by calculating the fold change in size versus the initial size before the initiation of the treatment (day 0). d, Quantitative PCR analysis of gene expression in tumor samples obtained from groups shown in c. The mRNA levels of p21, IL-6, IL1-α, TNF-α, TGF-β, PAI1, BCL2 and CASP1 genes were measured using qPCR and normalized to ACTB/GAPDH as reference genes. The analyzed groups were vehicle treated (dark green), StnIG-treated (green), senescent (brown) and senescent + StnIG-treated (yellow) mouse groups, respectively. The fold-change expression is presented relative to the vehicle control group and senescent vehicle control group (n = 4). e, Tumor progression in a xenograft liver cancer mouse model of senescence induced by palbociclib as described above. Tumor progression was estimated as tumor percentage volume. Data are means ± s.d. Statistical significance was assessed by Kruskal–Wallis test with Dunn’s correction for multiple comparisons in b, one-way ANOVA in d or two-way ANOVA test with Šídák’s correction for multiple comparisons in c and e, versus vehicle (black) or versus palbociclib (blue).

Source data

In addition, we conducted two distinct studies using xenograft tumor models of SKMel-147 melanoma cells in BALB/c FoxN1 female mice and Huh-7 hepatocarcinoma cells in BALB/c CAnN.Cg-Foxn1^nu/Crl female mice. First, we induced senescence via palbociclib administration in the two mouse groups. Mice were randomly divided in six groups of four (SKMel-147) and six (Huh-7) each, respectively: vehicle, palbociclib (100 mg kg⁻¹, oral gavage, daily), StnIG (2 mg kg⁻¹, i.p., every 48 h), navitoclax (50 mg kg⁻¹, daily, oral gavage), palbociclib + StnIG and palbociclib + navitoclax.

Palbociclib, navitoclax, and StnIG alone demonstrated the expected reduction in tumor growth (Fig. 8c,e). Most importantly, StnIG strongly synergized with the CDK4/6 inhibitor palbociclib for remission of solid tumors in mice, and this effect was comparable to the combined treatment of palbociclib + navitoclax (Fig. 8c,e). We also demonstrated that SA-β-gal staining was reduced in tumors from mice treated with StnIG (Supplementary Fig. 13), further supporting its senolytic effect in vivo.

We measured tumor mRNA levels of key SASP markers (p21, IL-6, IL-1α, TGF-ß, TNF-α and PAI1) along with BCL2 and CASP1 (Fig. 8d). IL-6, BCL2 and CASP1 were significantly reduced in StnIG-treated tumors compared to controls (Fig. 8d).

Importantly, the gross histopathological examination did not reveal any obvious signs of toxicity. However, we noticed a tolerable reduction in body weight of StnIG-treated mice that was ~5% or ~10% as compared to the navitoclax or vehicle group, respectively. To ensure safety regarding the maximum tolerated dose (2 mg kg⁻¹), we conducted detailed histological analyses (H&E staining, Histopathology Report in Supplementary Information) on tissues including brain, heart, lung, kidney and liver. Biochemical assessments of kidney and liver markers, along with serum biochemical analyses, confirmed that StnIG did not induce toxicity compared to control conditions (Supplementary Table 2). Notably, StnIG treatment elevated blood potassium levels exclusively in the senescent-tumor group (palbociclib treated), suggesting that tumor cell death led to a substantial release of intracellular potassium into the bloodstream.

In contrast, mice treated with navitoclax in the xenograft hepatocellular carcinoma model developed bruising and hemorrhage in the tumor area between days 0 and 5, leading to inaccurate measurements. Consequently, these mice and the corresponding time-point measurements were excluded from the graph (Fig. 8e). This observation underscores the issue of navitoclax-induced thrombocytopenia, which limits its clinical application¹⁵.

Discussion

We demonstrated the promising role of the actinoporin StnI and its optimized analogue StnIG as complementary senotherapies in cancer. Their therapeutic potential was supported across diverse cellular cancer models (melanoma, neuroblastoma, adenocarcinoma and hepatoma), various chemotherapy-induced senescence models (palbociclib, doxorubicin, and bleomycin) and in in vivo studies. StnIG selectively targeted senescent cells, leveraging its lipid-binding specificity to facilitate sustained potassium efflux and cell-volume decrease, while importantly attenuating SASP. The proposed mechanism of action suggests a distinct class of senolytics that target ion homeostasis and lipidome, which we introduce as ‘senotoxins’. The senotoxin StnIG demonstrated a robust synergy with the CDK4/6 inhibitor palbociclib, leading to remission of solid tumors in both mice and zebrafish xenograft cancer models. We confirmed the safety of the toxin through a detailed histopathology report and biochemical blood analysis. An observed increase in potassium levels in StnIG-treated mice may be linked to tumor lysis syndrome⁵⁸. Indeed, the toxin induces a complex senolytic effect involving BKCa activation, pyroptosis and apoptosis-driven cell death. Thus, we suggest that potassium elevation could serve as an indirect but measurable biomarker of treatment efficacy, with a sharp rise post-therapy indicating robust tumor cell lysis. This presents a potential noninvasive marker for real-time monitoring of therapeutic success. Of note, our study was limited by the use of immunosuppressed female mice, which prevented analysis of immune system interactions and potential sex-specific effects, as well as the absence of pharmacokinetic and pharmacodistribution studies.

Our results corroborate the anti-tumoral activity of sticholysins^23,24 and offer insights, highlighting that senotoxins preferential targeting senescent cells. To unravel whether a specific structural region was responsible for the senolytic properties of StnI, we created a panel of analogues based on specific regions and amino acid differences between StnI and StnII. However, we did not identify any active region based on the studied structural regions and sequences. Interestingly, reducing N-terminal polarity while keeping an intact C-terminal (StnI E2AD9A) produced a toxin with lower cytotoxicity, which maintained a certain degree of selectivity toward senescent cells. In addition, the senolytic index of StnIG was enhanced by the polar poly-histidine sequence introduced at the C terminus, whereas protein denaturation impaired the senolytic capability of StnI and StnIG. Thus, senolysis is influenced by C-terminal polarity and the overall toxin structure, rather than specific regional sequences.

Our findings indicate that chemotherapy-induced senescence through palbociclib involves a cellular lipid remodeling, with notable alterations in PC, SM and cholesterol. Although our study did not examine the specific lipid distribution at the subcellular level, we demonstrated that differences in the PC:SM balance influenced the cytotoxicity of StnIG. Senolysis by StnIG was diminished by SM supplementation or lipid restriction in senescent cells. In contrast, PC supplementation in proliferating cells enhanced the cytotoxicity of StnIG. This finding is of importance, as sticholysins recognize the phosphocholine hydrophilic head group present in PC and SM^22,26, whereas the pore-forming ability is facilitated by cholesterol and membrane fluidity^25,26.

Our molecular dynamics (MD) simulations provide atomic-level insight into how these lipid alterations may contribute to the selective targeting of senescent cells by StnIG. The membrane-facing α2 helix, particularly through residues Glu134 and Gln130, forms specific and stable hydrogen bonds with PE and PS headgroups. Although these lipids are typically sequestered to the inner leaflet of healthy cell membrane³⁴, their exposure on the outer leaflet during senescence^35,36 provides additional stabilizing interactions for StnIG binding. Similar lipid externalization patterns have been observed in cancer cells, where PS-targeting molecules can achieve selective binding through recognition of these exposed anionic phospholipids⁵⁹.

The asymmetric distribution of these interactions, with long-lived contacts observed predominantly in senescent membrane systems, suggests a molecular mechanism for senotoxin selectivity. This structural basis for selective recognition represents a paradigm for senolytic drug development, where compromised membrane asymmetry could be exploited as a targeting strategy. Furthermore, the identification of specific residue-lipid interactions that drive selectivity could guide the engineering of next-generation senotoxins with enhanced specificity for senescent cells. This study provides a detailed electrophysiological characterization of senescent cells. Both proliferating and senescent cells exhibit predominantly KCa channels, which respond to minor fluctuations in intracellular calcium⁶⁰, and their activation has been associated with antitumoral properties⁶¹. However, their density is diminished in senescence, implying that during the transition to senescence, cells increase in size without corresponding changes in channel content, leading to an overall reduction in KCa density. This notion aligns with the concept of low potassium permeability being required for cell cycle arrest⁶² or serving as an additional mechanism of apoptosis resistance in senescence⁴⁸. Our results demonstrated that StnIG effectively forms pores that allow Na⁺ and Ca²⁺ influx, with a more pronounced hyperpolarizing efflux of K⁺ in senescent cells. Over time, these effects are mitigated or corrected within the context of proliferating cells. In senescent cells, however, we observed accumulation of calcium, an enhanced KCa channel activity and a maintained hyperpolarizing efflux of potassium. It is thus conceivable that StnIG triggers an ion imbalance in senescent cells and promotes other yet more complex effects that hamper the cell adaptability⁶³. Altogether, we confirmed that StnIG-induced senolysis disrupts ionic balance, leading to sustained calcium influx, which activates BKCa channels and drives cell death via apoptosis and pyroptosis. These findings highlight the central role of calcium influx in StnIG-induced cell death specifically in senescent cells. This aligns with our observation that BKCa activation is sustained following StnIG treatment exclusively in senescent cells. Interestingly, previous studies have shown that potassium depletion mediates cell death⁴⁹ and caspase activation, including caspases involved in pyroptosis^51,52. In line with these studies, we showed that the pore-forming StnIG initiates early apoptosis independently of BKCa activation. However, sustained calcium influx and BKCa activation facilitated cell shrinkage and transition to late apoptosis in senescent cells.

Independently of that, the alterations induced by StnIG in senescent cells seem persistent and irreparable following acute application. On the contrary, sticholysin StnI and StnIG exerted milder and transient changes in rapidly proliferating cells in vitro, but sufficient to reduce tumor progression in vivo through repeated administrations. However, it is feasible that factors such as cellular plasticity, ion homeostasis and membrane regenerative capacity might counteract cytotoxicity and confer partial tumor resistance in the absence of senescence.

Our study provides the proof of concept for the potential of StnIG as a therapeutic intervention in cancer, demonstrating improved efficacy in combination with chemotherapy that induces tumor senescence. We also confirmed the senomorphic potential of StnIG in human fibroblasts, raising further questions about its role in aging and age-related conditions. However, the mechanisms underlying the senolytic effects of senotoxins on non-transformed and replicative senescent cells remain unclear. Future studies should also investigate whether the efficacy of StnIG extends to other senescence-associated diseases, such as lung fibrosis⁶⁴, kidney disease⁶⁵ or even improve symptoms in progeroid syndromes⁶⁶.

Methods

The research presented complies with all relevant ethical regulations as approved by the IMDEA Nutrition Research Committee.

Reagents and media

Media for cell culture was purchased from Thermo Fisher Scientific. Drugs were purchase from MedChemTronica and Sigma-Aldrich. Antibodies were obtained from Cell Signaling Technology, including anti-Phospho p53 (#9284), anti-p21 (#2947) and anti-p16 (#92803). Anti- β-actin was obtained from Proteintech. Secondary antibodies were purchased from Sigma-Aldrich (anti-rabbit HRP #A0545) and LI-COR (IRDye 680RD and IRDye 800CW). Lipids were obtained from Avanti Polar Lipids. Hoechst 33342 was purchased from Thermo Fisher Scientific (H1399). C12FDG was purchased from Setareh Biotech (7188). Formaldehyde 32% was purchased from Electron Microscopy Sciences (15714).

Synthesis of StnI and analogues

StnI (UniProtKB P81662), StnII (UniProtKB P07845) and StnI E2AD9A were produced as described previously⁶⁷. Later modifications of StnI included inclusion of methionine at the N-terminal end and six residues of histidine at the C terminus. This new version is referred as StnIG throughout the paper. StnIG was recombinantly produce with pET30a-StnIG (cloned in Ndel-HindIII) and purified in Ni²⁺–NTA columns by GenScript. Endotoxin removal was applied to archived maximum values between 0.1 and 1 Eu mg⁻¹. Purity was tested by SDS-PAGE under reducing condition. The product was stored at −80 °C in 50 mM Tris-HCl, 150 mM NaCl and 10% glycerol, pH 8.0. All other truncated forms were chemically synthesized by GenScript.

Cell culture

All human cancer cell lines and the non-transformed fibroblast lines were maintained in a humidified incubator at 37°C and 5% CO₂ and used between passages 4 and 15. All lines were frequently checked and ensured to be mycoplasma-free. Cells were split twice a week and before archiving an 90% confluency and maintained in DMEM media supplemented with 10% fetal bovine serum, 1% penicillin-streptomycin and 1% non-essential amino acids. IMR90 were cultured in EMEM medium with 10% FBS and pen/strep antibiotics. SKMel-147, SH-SY5Y, Huh7, A549, neonatal foreskin fibroblasts and replicative senescent fibroblasts were donated from collaborators or obtained from Translational Venomics stock. Human dermal fibroblasts and IMR90 lung fibroblasts were purchased directly from the American Type Culture Collection. Cell lines authenticated by Short Tandem Repeat profiling and results are given in Supplementary Table 3.

MTT cell viability assay

Cells were plated in 96-well plates at a density of 3 × 10³ to 8 × 10³ cells per well (proliferative) or 8 × 10³ to 10 × 10³ cells per well (senescent). An MTT assay (Sigma-Aldrich) was used to measure cell viability after 48 to 96 h of treatment^18,20. Absorbance was measured at 570 nm using a microplate reader (Victor Nivo, Perkin Elmer). The effects of vehicle or treatments were expressed as the percentage change of non-treated cells.

Crystal violet cell viability assay

Cells were plated in 96-well plates at a density of 3 × 10³ to 8 × 10³ cells per well (proliferative) or 8 × 10³ to 10 × 10³ cells per well (senescent), allowing at least 16 h for attachment before treatment. After 48 h treatments, cells were washed with PBS and fixed with methanol; 0.5% crystal violet was used for 40 min to stain cells. Cells were then washed in water, and the dye was dissolved in 10% acetic acid. Colorimetric intensity was measured 570 nm excitation using a microplate reader (Victor Nivo, Perkin Elmer). The effects of vehicle or treatments were expressed as percentage change of non-treated cells.

Apoptosis assay

The annexin V-FITC/PI apoptosis detection kit (BD Biosciences) was used to detect different stages of apoptosis. Annexin V has a strong affinity for phosphatidyl serine, which is externalized in the membranes of apoptotic cells^18,20. Mean fluorescence intensity from 10,000 cells was acquired by FACSCelesta flow cytometer (BD Biosciences) using the FITC and PE bandpass filter for annexin V and PI, respectively.

Mitochondrial membrane potential

Mitochondrial membrane potential was determined using rhodamine 123 (Thermo Fisher Scientific) and TMRM (Thermo Fisher Scientific)^18,20. Mean fluorescence intensity from 10,000 cells was acquired with a FACScelesta flow cytometer (BD Biosciences) using a FITC or PE bandpass filters for detection of rhodamine123 or TMRM, respectively.

Bioenergetics

Cellular bioenergetic measurements were performed using the Seahorse XFe96 Analyzer and XFe96 culture microplates (Agilent Technologies) to investigate OCR in proliferating or senescent SKMel-147 cells. Measurements were performed between 2 and 24 h incubation with 100 nM StnI, StnIG or analogue (StnII) in 25 × 10³ proliferating or senescent SKMel-147 cells per well in quadruplicates. OCR was measured in Seahorse XF base medium containing 10 mM glucose, 1 mM sodium pyruvate and 2 mM L-glutamine (pH 7.4) and ECAR in Seahorse XF base medium containing 0.5 mM sodium pyruvate and 1 mM L-glutamine (pH 7.4). Data were expressed as percentage of the basal respiration per well.

Electrophysiological studies

Membrane currents were recorded with an Axopatch 200B and a Digidata 1322 A (Axon Instruments) using the whole-cell configuration of the patch-clamp technique. Cells were superfused at 1 ml min⁻¹ with bath solution (mM: NaCl 130, KCl 5, CaCl₂ 1.8, MgCl₂ 1.2, glucose 10, and HEPES 10 (pH 7.3)) and a pipette (internal) solution (mM: KCl 135, MgCl₂ 1.2, Na₂ATP 5, HEPES 10, and EGTA 0.1 (7.2 pH)). Cells were clamped at −60 mV and potassium currents were assessed by voltage ramps (−100 to +60 mV), single voltage steps (−100 mV and +60 mV) and by acquisition of full I–V relationships for steady state activation (4 s steps from −80 to +60 mV in 10 mV increments). Membrane potential was recorded under the current-clamp mode. Signals were sampled at 10 kHz and low-pass filtered at 2 kHz. Voltage-clamp acquisition and analysis protocols were controlled by Clampex 10.0 software (Molecular Devices). Off-line analysis was performed using Clampfit 10.0 (Molecular Devices). Data were expressed as current density (pA/pF) or I/Izero, where Izero is the current magnitude recorded at the onset of a given experimental intervention. All experiments were performed at room temperature (22–24°C).

Cytokine arrays

Proliferating or senescent (palbociclib 5 µM, 7 days) SKMel-147 cells were plated in a 6-well plate at a confluency of 50 × 10³ cells per well (proliferative) or 300 × 10³ cells per well (senescent). Cells were then treated in triplicates with vehicle (0.1% PBS) or StnI 10 to 30 nM for 4 days. After 4 days of treatment, viable cells were selected using the trypan blue exclusion method and plated at a confluency of 500 × 10³ cells per well (duplicate/group, 6-well plate) in culture media for 24 h. Cytokine levels were analyzed using the Proteome Profiler Human Cytokine Array (R&D Systems, #ARY005B), following the manufacturer’s instructions. Pixel density was determined using the Image Studio Lite Software 5.2 (LI-COR Biosciences).

Senotherapeutic screening

Senescence was assessed by measuring SA-β-gal activity using the C12FDG staining assay⁶⁸. Human IMR90 lung fibroblasts were cultured at 20% O₂ to induce senescence and then plated at a density of 3,000 cells per well in black-walled, clear-bottom 96-well plates, allowing at least 16 h for attachment before treatment. After drug treatment, the cells were incubated at 20% O₂ for 60 h. The medium was then removed, and the cells were treated with 100 nM Bafilomycin A1 in culture medium for 60 min to induce lysosomal alkalinization. This was followed by a 2-h incubation with 20 μM C12FDG (Setareh Biotech) and counterstaining with 2 μg ml⁻¹ Hoechst 33342 (Thermo Fisher Scientific) for 15 min. The cells were then washed with PBS, fixed with 2% paraformaldehyde for 15 min and imaged in six fields per well using the Cytation 1 high-content fluorescent imaging and analysis system (BioTek).

SA-β-gal staining

Approximately 50 to 80 x 10³cells were seeded in coverslips (22 × 22 mm) and left overnight in complete media. The following day, media was removed and cells were washed, fixed and stained overnight as per Cell Signaling SA-β-gal staining Kit instructions (#9860, Cell Signaling Technology). Tumor samples were washed and fixed in a solution of 2% formaldehyde and 0.2% glutaraldehyde in PBS for 1 h on a shaker, washed three times with PBS containing MgCl₂ and subsequently stained an incubator without CO₂ overnight in a 1x X-Gal solution (5 mM K₃Fe(CN)₆, 5 mM K₄Fe (CN)₆·3H₂O, and 20 mg ml⁻¹ X-Gal in PBS). Samples were processed in the Spanish National Cancer Research Center histopathology unit.

Western blots

Cells were lysed in cold RIPA buffer with protease (Merck) and phosphatase (Roche Diagnostics) inhibitors, and lysates stored at −20 °C. Protein concentrations were measured using the Pierce BCA Protein assay kit (Thermo Fisher Scientific). Samples (15 µg protein) were run on sodium dodecyl sulfate–7%–10% polyacrylamide gel under reducing conditions and transferred onto polyvinylidene difluoride membranes (GE Healthcare). Blots were blocked with 5% BSA (Merck) in TBS + 0.1% Tween-20 buffer (1 h, room temperature). They were then incubated with primary antibodies, including β-actin, and then detected used fluorescent anti-rabbit/-mouse antibodies (Supplementary Table 4). Quantification was performed with Odyssey FC and Image Studio Lite Software (LI-COR Biosciences), normalized to β-actin (42 kDa), and expressed as percent relative to control mean.

RNA sequencing

Senescent SKMel-147 cells (570 × 10³ per well) were treated with StnI 100 nM for 0, 3,6 and 12 h. RNA was then extracted using the RNeasy kit (QIAGEN). Integrity was assessed by 2% agarose gels, whereas purity and concentration were measured using a NanoDrop 2000 (Thermo Fisher Scientific). Samples were sequenced by Biomarker Technologies (BMKGENE). For Library preparation, 1 μg RNA per sample was processed using Hieff NGS Ultima Dual-mode mRNA Illumina Kit (Yeasen Biotechnology). mRNA was isolated with poly-T oligo beads, followed by cDNA synthesis, end repair, adaptor ligation (NEBNext), and purification (AMPure XP, Beckman Coulter). USER Enzyme (3 μl, NEB) was used with size-selected, adaptor-ligated cDNA (37°C for 15 min followed by 5 min at 95°C). PCR was then performed with Phusion High-Fidelity DNA polymerase, Universal PCR primers and Index Primer. Products were purified (AMPure XP system), libraries were quality checked on Agilent Bioanalyzer 2100 system, and sequenced on an Illumina NovaSeq platform (150 bp paired-end). Differentially expressed genes (P < 0.05 and FRD < 0.05) were explored using Metascape⁶⁹.

QPCR

Total RNA was extracted from cell and tissue samples using TRIzol reagent (Invitrogen, #15596018). Reverse transcription of up to 1 μg of total RNA was performed using the iScript cDNA Synthesis Kit (Bio-Rad, #1725038). qRT-PCR was performed using GoTaq PCR Master Mix (Promega, #A6002) and specific primers. The reaction was performed in a QuantStudio 6 Flex thermocycler (Applied Biosystems, RRID: SCR_020239). All reactions were performed in triplicates. ACTB and GAPDH were used as endogenous normalization controls, and all data were analyzed through the 2^−ΔΔCT method. The designed primers used, and their sequences are given in Supplementary Table 5.

Lipidomics lipid extraction MS data acquisition and analysis

Lipid extraction for mass spectrometry lipidomics

Lipid extraction was performed on proliferative and senescent SKMel-147 cells (Palbociclib 5 µM, 7 days), treated with 100 nM of StnIG or vehicle for 12 h (4 × 75 cm³ flasks, 1.8 million cells each). Lipid analysis was conducted by Lipotype⁷⁰ using a chloroform/methanol procedure⁷¹. Samples were spiked with internal standards, including CL 14:0/14:0/14:0/14:0 (CL), ceramide 18:1;2/17:0 (Cer), diacylglycerol 17:0/17:0 (DAG), hexosylceramide 18:1;2/12:0 (HexCer), lyso-phosphatidate 17:0 (LPA), lyso-phosphatidylcholine 12:0 (LPC), lyso-phosphatidylethanolamine 17:1 (LPE), lyso-phosphatidylglycerol 17:1 (LPG), lyso-phosphatidylinositol 17:1 (LPI), lyso-phosphatidylserine 17:1 (LPS), phosphatidate 17:0/17:0 (PA), PC 17:0/17:0, PE 17:0/17:0, phosphatidylglycerol 17:0/17:0 (PG), phosphatidylinositol 16:0/16:0 (PI), phosphatidylserine 17:0/17:0 (PS), cholesterol ester 16:0 D7 (CE), SM 18:1;2/12:0;0, triacylglycerol 17:0/17:0/17:0 (TAG). Extracts were dried in a speed vacuum concentrator and resuspended in 7.5 mM ammonium formiate in chloroform/methanol/propanol (1:2:4; V:V:V). Liquid handling was automated using the Hamilton Robotics STARlet platform with Anti Droplet Control.

Samples were analyzed by direct infusion on a QExactive mass spectrometer (Thermo Fisher Scientific) with a TriVersa NanoMate ion source (Advion Biosciences), in both positive and negative modes (MS: R_m/z=200 = 280,000; MS/MS: R_m/z=200 = 17,500). MS/MS was triggered by an inclusion list scanned in 1 Da increments. Ion types monitored included ammonium adducts (CE, DAG, TAG), formiate adducts (LPC, LPC O-, PC, PC O-, Cer, HexCer, SM), and deprotonated anions (CL, LPS, PA, PE, PE O-, PG, PI, PS, LPA, LPE, LPE O-, LPG, LPI). Data were processed using in-house software based on LipidXplorer⁷² and with normalization via a custom data management system. Only lipids with signal-to-noise >5, and 5-fold higher blank intensity were included in downstream analysis.

MD simulations

A homology model of the StnIG octamer was constructed using MODELER 10.5 (ref. ⁷³), based on the FraC crystal structure (PDB: 4TSY, 61% sequence identity, 80% similarity). Model quality was validated with MolProbity⁷⁴, yielding a score of 1.31 (98th percentile), clashscore of 0 and 93.64% residues in favored Ramachandran regions. Membrane embedding and solvation of the model were performed via CHARMM-GUI bilayer builder^75,76, generating systems of approximately 140 Å with lipid compositions detailed in Supplementary Table 3. The system was immersed in 150 mM KCl with 22.5 Å water thickness and neutralized with extra K⁺ ions. CHARMM36m parameters⁷⁷ and TIP3P model were used. Simulations were run in GROMACS 2022 on GPU accelerators⁷⁸ under NPT conditions at 303.15 K using Nose-Hoover (τ = 1.0 ps) and Parrinello-Rahman (τ = 5.0 ps) coupled with semi-isotropic pressure scaling. A 2 fs time step was employed with LINCS constraints on hydrogen atoms. Non-bonded interactions used a Verlet cutoff scheme, van der Waals switching (1.0–1.2 nm) and PME for long-range electrostatics. Each system was simulated for 500 ns in five replicates with different random seeds, following energy minimization and gradual restraint release during equilibration.

Analysis protocols for MD simulations

Trajectory snapshots were saved every 100 ps during production simulations. Analysis was performed using custom Python scripts with MDAnalysis⁷⁹. Protein-lipid contacts were defined by a 4 Å cutoff between protein atoms and lipid headgroups. Contact durations were calculated by over continuous periods, allowing brief interruptions (≤10 frames) to account for thermal fluctuations. Root mean square fluctuations (RMSF) of phosphate groups were computed over 10 ns windows using trajectory-aligned frames. For residue-specific analyses, contacts were grouped by lipid type and normalized by lipid abundance. Statistical comparisons included Mann-Whitney U tests (contact durations) and chi-square tests (contact distributions), with significance set at P < 0.05. RMSF values for phosphate mobility were binned (0–1 Å, 1–2 Å, 2–3 Å, 3–4 Å, >4 Å), and distributions were compared using Kolmogorov-Smirnov and chi-square tests.

In vivo studies

Xenograft model

Mouse experiments were performed at the Institute for Research in Biomedicine, Barcelona, Spain (Approval Number 23-033-PIL) and the National Institute on Aging (Approval Number ASP 277-TGB-2025). Eight-week-old female BALB/c FoxN1 nude mice were sourced from Envigo Spain (melanoma xenografts) or Charles River (hepatocellular xenografts), housed in a pathogen-free facilities with controlled temperature and 12 h light/dark cycles. Mice were randomly assigned to treatment groups and housed in cages (Institute for Research in Biomedicine: four per cage; National Institute on Aging: single-housed) with bedding, food, and water ad libitum. Randomization of animals in intervention studies was performed in Excel using the rand function. Daily monitoring was performed by trained staff and weekly by institutional veterinarians. A pilot study determined the maximum tolerant dose of StnIG via intraperitoneal (i.p.) injection, starting at 40 µg kg⁻¹ and doubling every other day; 2 mg kg⁻¹ was determined as safe. For xenografts, mice were injected subcutaneously with 10⁶ SKMel-147 or 5 × 10⁶ Huh-7 cells in both flanks. When the tumors reached a size of approximately 75 to 100 mm³, mice were randomly divided into six groups (n = 4/group): vehicle, palbociclib, navitoclax, StnIG, palbociclib + StnIG and palbociclib + navitoclax (senolytic control). Palbociclib (100 mg kg⁻¹) was administered by oral gavage daily for 10 days, starting 2 days before other treatments. StnIG (2 mg kg⁻¹) or vehicle (PBS) was given i.p. every 2 days for 8 days. Navitoclax (50 mg kg⁻¹) was administered daily by oral gavage for 8 days. Tumor volumes were measured daily with a digital calliper²⁰. Mice were sacrificed when tumors reached approximately 1000 mm³ by CO₂. However, in some cases, tumors rapidly exceeded this threshold overnight, and mice were euthanized immediately upon detection. Tumors were harvested for histology and molecular analysis. Organs (kidneys, liver, lung, heart and brain) and blood were collected, fixed in 4% PFA or stored at −80°C.

Zebrafish care and handling

Wild-type adult Danio rerio were crossed to obtain embryos. Fish were maintained at one fish per liter in aquaria with a 14:10 h light/dark cycle and water temperature of 28.5°C, following established protocols^18,20,80. All procedures complied with the Animal Care and Use Committee of the University of Santiago de Compostela and Spanish regulations (Directive 2012-63-DaUE). At endpoint, embryos were euthanized by tricaine overdose.

Xenograft assays and image analysis

StnIG toxicity was performed in duplicates and after 6 days of treatment. Mortality of embryos was monitored every 24 h to determine the no observed adverse effect level (NOAEL) of StnIG. Zebrafish embryos were handled as established^18,20. Senescent (palbociclib 5 µM, 7 days) or proliferating SKMel-147 cells, ranging between 100 and 200 were injected into the circulation of fish (Duct of Cuvier) using a microinjector (IM-31 Electric Microinjector, Narishige). At day 1 after injection, embryos were randomly divided into groups for vehicle, 1 µM navitoclax, and 10 nM StnIG. Imaging of injected embryos was performed using a fluorescence stereomicroscope (AZ-100, Nikon) at 6 days after injection to measure the spreading and accumulation of the administered cells in the caudal hematopoietic tissue. Image analysis were performed with Quantifish software 2.1.20 (ref. ⁸¹) and integrated density was calculated for comparisons. A ratio of cells equal to 1 translated to maintenance, a ratio above 1 signified accumulation and a ratio below 1 represented clearance.

FLIPR assays

Cells were seeded at 10,000 per well in black-walled 384-well plates and cultured overnight at 37 °C/5% CO₂. Calcium4 dye (Molecular Devices) was loaded in Ca²⁺/Mg²⁺-free Hanks Balanced Salt solution (Invitrogen), supplemented with either 1.8 mM Ca²⁺ or BAPTA (100 µM) for 30 min at 37 °C/5% CO₂. StnIG (100–0.13 nM) was added using the FLIPR^Penta following 10 reads (1 Hz; excitation 470–495 nm, emission 515–575 nm), and then 300 reads during and following compound addition. A second addition of Ca²⁺ (0.5 mM), and 300 further reads assessed Ca²⁺ influx in nominal Ca²⁺-free conditions. For Na⁺ and membrane potential responses, cells were loaded with 10 µM NaTRIUM Green-2 AM (Abcam) in physiological salt solution (in mM: NaCl 140, glucose 11.5, KCl 5.9, MgCl₂ 1.4, NaH₂PO₄ 1.2, NaHCO3 5, CaCl₂ 1.8, HEPES 10)⁸². Raw fluorescence values were converted to Response over baseline (ΔF) using Screenworks 5.1.1.86 and plotted using GraphPad Prism 10.1.0.

Determination of basal content of free calcium

Intracellular free calcium was measured using Fluo-4AM (HY-101896, MedChemExpress). Cells were stained with 1 µM Fluo-4AM and 0.02% poloxamer 407 (HY-D1005, MedChemExpress) in serum free-DMEM media for 30 min at 37°C. After washing, cells were resuspended in bath solution (as described in electrophysiology methods) and incubated for 30 min in the dark. Fluorescence intensity from 20,000 cells was acquired with a FACScelesta flow cytometer (BD Biosciences) using a FITC bandpass filter and FITC-A/FSC-A ratios.

Statistics and reproducibility

Group size refers to the number of independent biological replicates (n; separate experiments) used for statistical analysis. Within each experiment or group, technical replicates were performed and are reported accordingly. The effects of StnIG were unpredictable in vitro or in vivo, hence, a sample size estimation was not performed. Experiments with mice adhered to the 3 R principles (replace, reduce and refine). Randomization of animals in interventions studies was performed in Excel using the rand function. In vitro experiments were not randomized, and investigators were not blinded to data allocation and outcome assessment. Blinding was applied exclusively during histopathological analyses to minimize bias. Samples were numerically coded before staining and evaluation, ensuring both the personnel at Spanish National Cancer Research Center and the pathologist were blinded to sample identify and experimental groups. Data were excluded only if statistically identified as outliers by GraphPad Prism 10.1.0. Hydrophobicity scales (Zimmerman and Kyte, Doolittle) were analyzed in Expasy (https://web.expasy.org/protscale/). Statistical analysis was conducted for group sizes ≥n = 3, with data from at least three independent experiments performed in triplicate (technical replicates) and expressed as mean ± standard deviation or standard error of the mean. Student’s t-test was used for two group comparisons or one-way ANOVA with Sidak post hoc test for multiple comparisons. Significance was considered at *P < 0.05, **P < 0.01 and ***P < 0.001. The exact P values are given when >0.001. In multigroup studies with parametric variables, post hoc tests were conducted only when the ANOVA F-statistic (or equivalent) reached statistical significance and variance across groups was homogeneous. All sample populations were first assessed for normality. When populations within groups were not normally distributed, a non-parametric Mann-Whitney test was used for single comparisons or Kruskal-Wallis test for multiple comparisons. IC₅₀ values were calculated using a non-linear regression (log(inhibitor) versus response, least squares fit). All analysis were performed using GraphPad Prism 10 (GraphPad Software).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization was performed by M.P.I. and J.M.-S. The methodology was developed and the analysis conducted by J.M.-S., I.F.-C., V.R., A.G., I.K., P.C.-S., E.R.-d.-T., O.E., L.S., R.M.-H., J.L.L.-A., G.C., A.M.-d.-P., I.V., A.C., F.P.-V., J.D.-P., E.R., M.A.F.-R., M.S. and M.P.I. Data interpretation was carried out by J.M.-S. and M.P.I. Statistics were performed by J.M.-S., R.M.-H. and M.P.I. Resources were provided by L.S., A.M.-d.-P., I.V., A.G., J.D.-P., E.R., P.D.R., R.d.C., M.S. and M.P.I. The paper was written by J.M.-S. and M.P.I and reviewed and edited by J.M.-S., I.F-C., V.R., P.C.-S., E.R.-d.-T., L.S., G.C., A.M.-d.-P., I.V., A.C., F.P.-V., J.D.-P., E.R., P.D.R., R.d.C., N.L.P., M.A.F.-R., M.S. and M.P.I. All authors have read and agreed to the published version of the paper.

Peer review

Peer review information

Nature Aging thanks the anonymous reviewers for their contribution to the peer review of this work.

Data availability

All data supporting the findings of the study are provided in this study as source data, publicly available or can be obtained from the corresponding authors upon reasonable request. The GEO code for RNA-seq deposition is GSE290509, and MD simulations are deposited in Zenodo (10.5281/zenodo.17373116)⁸³. Source data are provided with this paper.

Competing interests

M.P.I., J.M.-S., M.S. V.R., I.F.-C. and M.A.F.-R. are named inventors in a filed patent related to senolytic therapies (WO 2025/016917 A1). M.S. is a shareholder of Senolytic Therapeutics, Inc., Life Biosciences, Inc., Rejuveron Senescence Therapeutics, AG, and Altos Labs, Inc. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the paper. The other authors declare no competing interests.

Extended data

is available for this paper at 10.1038/s43587-025-01030-w.

Supplementary information

The online version contains supplementary material available at 10.1038/s43587-025-01030-w.