Comparative analysis of senolytic drugs reveals mitochondrial determinants of efficacy and resistance
Masahiro Wakita, Koyu Ito, Kaho Fujii, Dai Sakamoto, Takumi Mikawa, Sho Sugawara, Xiangyu Zhou, Jeong Hoon Park, Hideka Miyagawa, Daisuke Motooka, Emi Ogasawara, Naotada Ishihara, Akiko Takahashi, Hiroshi Kondoh, Eiji Hara
TL;DR
This study compares 21 drugs that target aging cells and finds that mitochondrial health affects how well these drugs work, suggesting ways to improve their effectiveness.
Contribution
The study identifies mitochondrial integrity as a key factor in resistance to senolytic drugs and proposes metabolic interventions to enhance their efficacy.
Findings
ABT263 and ARV825 were the most effective senolytic drugs across multiple cell models.
Senolytic resistance is driven by mitochondrial integrity maintained via V-ATPase activity.
Metabolic interventions like ketogenic diet or SGLT2 inhibition improved senolytic efficacy in mice.
Abstract
Cellular senescence contributes to aging and disease, and senolytic drugs that selectively eliminate senescent cells hold therapeutic promise. Although over 20 candidates have been reported, their relative efficacies remain unclear. Here we systematically compared 21 senolytic agents using a senolytic specificity index, identifying the Bcl-2 inhibitor ABT263 and the BET inhibitor ARV825 as most effective senolytics across fibroblast and epithelial senescence models. However, even upon extended treatment with these most potent senolytics, a proportion of senescent cells remained viable. We found that senolytic resistance was driven by maintenance of mitochondrial integrity through V-ATPase-mediated clearance of damaged mitochondria. Imposing mitochondrial stress via metabolic workload enhanced the senolytic efficacies of ABT263 and ARV825 in vitro, and in mouse models, ketogenic diet…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 10
Figure 11
Figure 12
Figure 13
Figure 14
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9- —https://doi.org/10.13039/100009619Japan Agency for Medical Research and Development (AMED)
- —https://doi.org/10.13039/501100002241MEXT | Japan Science and Technology Agency (JST)
- —https://doi.org/10.13039/501100001691MEXT | Japan Society for the Promotion of Science (JSPS)
- —https://doi.org/10.13039/501100001700Ministry of Education, Culture, Sports, Science and Technology (MEXT)
- —https://doi.org/10.13039/501100004398Mitsubishi Foundation
- —The University of Osaka OU Master Plan Implementation Project
- —https://doi.org/10.13039/501100008886Princess Takamatsu Cancer Research Fund
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsTelomeres, Telomerase, and Senescence · Protein Degradation and Inhibitors · RNA regulation and disease
Main
Cellular senescence is a stable form of cell-cycle arrest triggered by a variety of potentially oncogenic stresses, such as telomere erosion, oxidative stress, radiation or oncogene activation^1,2^. While this response prevents the expansion of cells at risk of malignant transformation and thereby acts as a tumor-suppressive mechanism, senescent cells also have a darker side^3^. They secrete pro-inflammatory factors collectively known as the senescence-associated secretory phenotype (SASP)^4–6^, which can promote tissue dysfunction and disease depending on the biological context^7–9^. Accordingly, the targeted elimination of senescent cells by ‘senolytic’ drugs has emerged as a promising therapeutic strategy^10,11^. Over the past decade, more than 20 candidate senolytic drugs have been reported, spanning diverse mechanisms of action^12^. However, despite this growing list, there has been no systematic head-to-head comparison of their efficacies and specificities^10–12^. As a result, it remains unclear which agents most effectively eliminate senescent cells while sparing non-senescent counterparts^12^. Furthermore, even the most potent senolytic drugs fail to eliminate a subset of resistant cells, but the mechanisms underlying this resistance remain poorly understood^10–12^. In this study, we systematically compared 21 senolytic agents using a quantitative senolytic specificity index (SSI), identified ABT263 (ref. ^13^) and ARV825 (ref. ^14^) as the most effective senolytics across fibroblast and epithelial models, and uncovered mitochondrial quality control as a key determinant of resistance to senolysis.
Results
Systematic comparison of reported senolytic drugs
We conducted a comparative analysis of the activities and specificities of 21 reported senolytic drugs (ABT199, ABT263, ABT737, ARV825, BPTES, CB839, dasatinib and quercetin (D + Q), digoxin, fisetin, gingerenone A, IMP1088, JQ1, nintedanib, OTX015, PCC1, PCLX-001, P5091, RG7112, R406, 17-DMAG and 25-hydroxycholesterol)^13–25^ in targeting and eliminating senescent cells (Fig. 1a). To this end, we utilized IMR-90, a representative line of normal human diploid fibroblasts (HDFs) that has historically served as a prototypical model in cellular senescence research^26^, and the epithelial cell line RPE-1, widely used for studying epithelial cell senescence^27^, which were both induced into senescence (replicative and doxorubicin (DXR) induced, respectively; Fig. 1a). The senescent state was confirmed by several markers^28^, including increases in p16^INK4a^ and p21^WAF1/CIP1^, reduction in lamin B1 and loss of EdU incorporation, in both replicative and DXR-induced models (Supplementary Fig. 1). Senescent and non-senescent cells were treated with varying concentrations of these compounds, including the optimal doses reported previously^13–25^. To quantify specificity, we introduced an SSI, defined as SSI = (percentage senescent cells eliminated on day 3 relative to day 0)/(percentage reduction in control cell number on day 2 relative to untreated control cells).Fig. 1. Comparative analysis of senolytic drugs revealed ABT263 and ARV825 as the most potent compounds.a, Outline of the comparative analysis of senolytic drug. Early-passage HDFs (IMR-90 cells) and RPE-1 epithelial cells were rendered senescent by serial passaging (Rep-Sen cells) or treatment with 150 ng ml^−1^ DXR for 10 days (DXR-Sen cells), respectively. Senescent and non-senescent (control) cells were treated with 21 compounds at multiple concentrations for 3 days. Relative cell numbers were determined throughout the experiments. If the reduction of control cells on day 2 was <1%, the denominator was set to 1. Created with BioRender.com. b,c, For each compound, the SSI value at the concentration yielding the highest index is shown. (See Extended Data Figs. 1 and 2 for actual data). d–f, Early-passage pre-senescent normal HDFs (IMR-90 (d), TIG-3 (e), and TIG-1 (f)) were rendered senescent by serial passaging (Rep-Sen cells; e) or by treatment with 250 ng ml^−1^ DXR for 10 days (DXR-Sen cells; d,e), or by treatment with 100 µM etoposide for 2 days followed by a 10-day recovery period (ETO-Sen cells; f). Control and senescent cells were treated with the indicated compounds for 3 days (d–f). Relative cell numbers were determined throughout the experiments. g, Early-passage (control) TIG-3 cells were rendered senescent by treatment with 250 ng ml^−1^ DXR for 10 days (DXR-Sen cells) or by serial passaging (Rep-Sen cells). These senescent and control cells were then incubated with ABT263 (0.5 µM) or ARV825 (25 nM) for 7 days. Relative cell numbers were determined by daily cell counts. Data are presented as the mean ± s.d. (d–g). Representative results from three independent experiments are shown. DMSO, dimethylsulfoxide.Source data
Because control cells typically reach confluence by day 3, their cell number could not be reliably assessed at this time point; therefore, we used their cell counts on day 2. A higher SSI indicates greater specificity toward senescent cells, reflecting efficient elimination relative to adverse effects on controls. Using this metric, we confirmed that most compounds displayed senolytic activity, albeit to varying degrees (Fig. 1b,c and Extended Data Figs. 1 and 2). However, at concentrations that did not affect the proliferation and survival of non-senescent (control) cells in both IMR-90 and RPE-1 cells, only the Bcl-2 family inhibitors ABT263 (ref. ^13^) and ABT737 (ref. ^19^) and the BET inhibitor ARV825 (ref. ^14^) demonstrated high efficacy (Fig. 1b,c and Extended Data Figs. 1 and 2). Moreover, although the results may vary depending on cell type and senescence-inducing method, ABT263 and ARV825 consistently achieved the highest SSI values in both IMR-90 and RPE-1 cells (Fig. 1b,c and Extended Data Figs. 1 and 2). Similar results were obtained with senescence induced by other methods, as well as in other fibroblast lines (Fig. 1d–f). Notably, using control cells treated with the concentration of each compound yielding the maximal SSI, Annexin V staining, which detects phosphatidylserine exposure (a hallmark of early apoptosis), revealed apoptotic responses in a subset of cells for most compounds, except ABT199 (ref. ^19^), ABT263 (ref. ^13^), ABT737 (ref. ^19^), ARV825 (ref. ^14^), digoxin^17,18^ and OTX015 (ref. ^14^), particularly in IMR-90 cells (Extended Data Fig. 3). These results further support ABT263 and ARV825 as the most promising senolytic candidates in the models examined here, although senolytic sensitivity can vary depending on cell type and biological contexts. Indeed, preadipocytes have been reported to be relatively resistant to BCL-XL inhibition, including that by ABT263 (ref. ^29^). Importantly, even with these potent agents, complete elimination of senescent cells was not achievable, as approximately 20% to 30% of the senescent cells survived even at prolonged time points (Fig. 1g).
V-ATPase-dependent mitochondrial quality control promotes resistance to ABT263 and ARV825
To investigate why a subset of senescent cells remained resistant, we next examined the characteristics of the surviving population. RNA-sequencing (RNA-seq) analysis revealed 51 upregulated genes and 71 downregulated genes in common, in senescent fibroblasts that survived ABT263 or ARV825 treatment (Extended Data Fig. 4a). Differentially expressed genes (DEGs) and Gene Ontology (GO) analyses showed that many upregulated genes encoded SASP factors (Extended Data Fig. 4b). This was confirmed by quantitative PCR with reverse transcription (RT–qPCR) and immunofluorescence analysis (Extended Data Fig. 4c,d). The persistence of senescent cells with heightened SASP expression may diminish therapeutic benefit, underscoring the need to clarify resistance mechanisms and develop strategies to overcome them. Reanalysis of gene expression suggests increased oxidoreductase activity, including elevated expression of AKR1C1–AKR1C3, which are involved in detoxification of reactive carbonyl species and reduction of oxidative stress (Extended Data Fig. 4b,e). Although some of these genes might simply become upregulated by the reduction of BRD4 (ref. ^30^) in ARV825-treated cells, this expression pattern is suggestive of preserved mitochondrial membrane potential and reduced reactive oxygen species (ROS) accumulation (Extended Data Fig. 4b,e). Notably, ABT263, a BH3 mimetic, inhibits anti-apoptotic Bcl-2 family proteins, thereby enabling activation of BAX/BAK and subsequent mitochondrial outer membrane permeabilization and cytochrome c release^13^. We therefore hypothesized that ABT263-resistant cells maintain mitochondrial function. Indeed, MT-1 staining, an indicator of mitochondrial membrane potential^31^, showed that most senescent cells displayed reduced mitochondrial membrane potential 2–3 days after ABT263 treatment, but many of the resistant cells that persisted after 7 days retained strong MT-1 signals (Extended Data Fig. 5a). Conversely, ROS levels, which typically increase with mitochondrial dysfunction and contribute to cellular damage^32^, rose on days 2–3 but declined by day 7 in surviving cells (Extended Data Fig. 5b). Furthermore, treatment with the antioxidant N-acetyl-L-cysteine reduced ABT263-induced senolysis (Extended Data Fig. 5c), supporting a role for preserved mitochondrial function in the survival of ABT263-resistant senescent cells.
Because MT-1 could not be combined with apoptosis markers such as Annexin V for double staining, we searched for an alternative indicator of mitochondrial function. To this end, we performed single-cell RNA sequencing (scRNA-seq) to identify candidate genes for this purpose. The scRNA-seq analysis revealed that a series of genes encoding subunits of the V-ATPase, which is required for the lysosomal clearance of damaged mitochondria^33^, such as ATP6V0E1, were among the top commonly expressed genes in both ABT263-resistant and ARV825-resistant senescent cells (Fig. 2a–c). Notably, the ATP6V0E1 knockdown significantly reduced MT-1 signals and cell survival in ABT263-treated senescent cells (Extended Data Fig. 6a–c), whereas treatment with EN6, a V-ATPase activator, produced opposite effects (Extended Data Fig. 6d,e), suggesting an association between ATP6V0E1 expression and mitochondrial membrane potential. We then performed double staining with MT-1 and ATP6V0E1 and found that senescent cells displayed substantial heterogeneity in both signals (Fig. 2d). After ABT263 treatment, cells with very low MT-1 fluorescence (≤15 fluorescence units) increased between days 1 and 5 but decreased by day 7. Up to this time, the surviving cells had shown a remarkable increase in the correlation coefficient between MT-1 signal intensity and ATP6V0E1 expression (Fig. 2d). Furthermore, double staining for ATP6V0E1 and Annexin V demonstrated that cells with very low ATP6V0E1 expression (≤5 fluorescence units) became Annexin V-positive from days 1 to 3 after ABT263 treatment but were no longer detectable on day 7 (Fig. 2e). It is also important to note that the depletion of mitochondria^34,35^ significantly attenuated the senolytic activity of ABT263, indicating that the accumulation of damaged mitochondria is a critical determinant of senolysis in this experimental setting (Extended Data Fig. 7). Importantly, comparable results were also obtained with ARV825 (Fig. 2 and Extended Data Figs. 5–7). Taken together, these results suggest that senescent cells are heterogeneous with respect to V-ATPase expression. Cells with low V-ATPase expression may fail to clear damaged mitochondria upon treatment with ABT263 or ARV825, leading to ROS accumulation and cell death, whereas cells with high V-ATPase expression are likely to remove damaged mitochondria more efficiently, preventing ROS accumulation and thereby surviving (Fig. 2 and Extended Data Figs. 5–7).Fig. 2. Heterogeneity of senescent cells underlies resistance to ABT263 and ARV825.a, Uniform manifold approximation and projection (UMAP) embedding of scRNA-seq profiles from early-passage pre-senescent (control) TIG-3 cells and DXR-induced senescent (DXR-Sen) TIG-3 cells treated with or without ABT263 or ARV825 for 7 days. b, Top ten enriched biological processes from Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of upregulated DEGs in group A. KEGG pathway enrichment analysis was performed using a one-sided Fisher’s exact test, with P values adjusted for multiple testing using the Benjamini–Hochberg false discovery rate (FDR) method; pathways with FDR < 0.05 were considered significant. c, Genes belonging to the top-ranked KEGG pathway ‘OXPHOS’. Differential gene expression analysis was performed using a two-sided MAST hurdle model implemented in Seurat. P values were adjusted for multiple testing using the Benjamini–Hochberg FDR method. d,e, DXR-Sen TIG-3 cells were incubated with ABT263 (0.5 µM) or ARV825 (25 nM) for the indicated durations and subjected to MT-1 staining together with ATP6V0E1 RNA in situ hybridization (d), or Annexin V staining together with ATP6V0E1 RNA in situ hybridization (e). Nuclei were counterstained with DAPI. Representative images at each time point are shown; enlarged views of the regions outlined with yellow dashed lines are shown in the lower-left corners. Scale bars, 10 μm. Signal intensities of individual cells are plotted. Areas with both MT-1 and ATP6V0E1 fluorescence signals ≤ 15 (d), or with Annexin V fluorescence signals ≥5 and ATP6V0E1 signals ≤ 5 (e), are outlined with red dashed lines, and the percentage of cells within each area is indicated in red. The r values at the top of each graph indicate the correlation coefficient between the x-axis and y-axis variables. At least 50 cells were scored per group. Representative data from three independent experiments are shown. AFUs, arbitrary fluorescence units.Source data
Glycolysis-to-OXPHOS shift raises mitochondrial load, boosting senolysis by ABT263 and ARV825
Because BRD4, a major target of ARV825, has been implicated in regulating mitochondrial gene expression^36^, we examined this possibility but did not observe a notable effect in senescent HDFs (Supplementary Fig. 2). Instead, ARV825 is known to downregulate XRCC4 (ref. ^14^), which functions with DNA ligase IV in nonhomologous end joining^14,37^, via the reduction in BRD4 (Extended Data Fig. 8a–c). Interestingly, XRCC4 has also been reported to localize to mitochondria, where it associates with DNA ligase III^38^, which shares structural similarity with DNA ligase IV. In line with this, we found that XRCC4 binds to DNA ligase III and colocalizes with mitochondria in senescent cells (Extended Data Fig. 8d,e). Consistent with this, both XRCC4 and DNA ligase III knockdowns closely phenocopied the effects of ARV825 and ABT263 treatments (Extended Data Fig. 8f–k). Collectively, these findings indicate that ARV825 promotes senescent cell death, at least in part, by compromising mitochondrial function, suggesting that mitochondrial quality control is likely a key determinant of resistance to ABT263-induced and ARV825-induced senolysis. To further confirm this idea, we applied mitochondrial stress by short interfering RNA (siRNA)-mediated depletion of DNA polymerase subunit gamma (PolG), the only polymerase responsible for mitochondrial DNA (mtDNA) replication^39,40^, in DXR-induced senescent cells. Notably, PolG depletion alone caused a slight but statistically significant decrease in mitochondrial membrane potential (Extended Data Fig. 9a–d). However, combining PolG depletion with ABT263 or ARV825 further reduced mitochondrial membrane potential and increased ROS levels (Extended Data Fig. 9a–e), resulting in a marked increase in senescent cell death (Extended Data Fig. 9f). Importantly, these effects were not observed in non-senescent control cells (Extended Data Fig. 9). Together, these results suggest that imposing mitochondrial stress aggravates the decline in mitochondrial function induced by ABT263 and ARV825, thereby enhancing senolysis.
We therefore sought a more physiological means to impose mitochondrial stress and focused on the metabolic shift from glycolysis to oxidative phosphorylation (OXPHOS) induced by a low-carbohydrate diet^41,42^. To mimic this effect in cultured cells, we initially used low-glucose medium. However, this approach also reduced the expansion of actively proliferating cells^43^, raising concerns about nonspecific effects. Accordingly, we tested an alternative strategy: inhibition of GLUT1 signaling. BAY876 (ref. ^44^), a highly specific GLUT1 inhibitor, suppresses glycolysis and thereby forces a compensatory shift toward mitochondrial dependence, resulting in increased TCA cycle activity, elevated oxygen consumption rate (OCR) and increased reliance on OXPHOS. Treatment of DXR-induced senescent TIG-3 fibroblasts with BAY876 in normal medium induced this metabolic shift from glycolysis to OXPHOS (Fig. 3a). Moreover, a significant decrease in mitochondrial membrane potential and an increase in ROS levels were observed when ABT263 or ARV825 was combined with BAY876 treatment in senescent cells, but not in control cells (Fig. 3b,c). Furthermore, the efficiency of senescent cell death induced by ABT263 or ARV825 was substantially enhanced by cotreatment with BAY876 (Fig. 3d). Similar results were also observed when cellular senescence was induced in different cell types (Supplementary Fig. 3). These results, in conjunction with the observation that glycolysis is accelerated in senescent cells and that redox homeostasis is maintained^45^, lead us to propose that a metabolic shift from glycolysis to OXPHOS increases mitochondrial workload and stress, thereby substantially enhancing the senolytic effects of ABT263 and ARV825.Fig. 3. Metabolic shift from glycolysis to OXPHOS by BAY876 enhances the senolytic activity of ABT263 and ARV825.a, Early-passage pre-senescent (control) or DXR-Sen TIG-3 cells were incubated with or without 7 µM BAY876 for 3 days and then subjected to Seahorse analysis to measure OCR following sequential injection of oligomycin (O, 1.5 µM), FCCP (F, 1 µM) and rotenone plus antimycin A (R/A, 0.5 µM). Basal respiration, maximal respiration and ATP production rates are shown to the right. b–d, Control TIG-3 cells and DXR-Sen TIG-3 cells were treated with the indicated compounds and subjected to MT-1 staining to assess mitochondrial membrane potential (b), CellROX staining to measure intracellular ROS levels (c) or cell number quantification (d). MT-1 and CellROX analyses were performed 2 days after treatment; 58 cells (b), 60 cells in control cells and 51 cells in DXR-Sen cells (c) were scored per group. Data are presented as the mean ± s.d. (a*–*d). Statistical significance was determined by two-sided Student’s t-test (a), one-way analysis of variance (ANOVA) followed by Sidak’s test (b and c) or two-sided Welch’s t-test (d). Scale bars, 10 µm (b and c). Although data shown are from technical replicates (a–d), experiments were independently repeated at least once to confirm reproducibility.Source data
Carbohydrate restriction enhances the in vivo efficacy of mitochondria-targeting senolytic therapies
To determine whether metabolic imposition of mitochondrial stress enhances the senolytic efficacy of ABT263 and ARV825 in vivo, we next tested this concept in mouse models using male mice. Senescent cells are known to promote cancer growth and metastasis through secretion of SASP factors, depending on the biological context^3^. In particular, B16 melanoma cells transplanted via the tail vein reportedly accumulate preferentially in the lungs of older mice, where senescent cells are more abundant, than in younger mice^46^. Therefore, we investigated whether administering a low-carbohydrate ketogenic diet^47,48^ together with ABT263 or ARV825 treatment in aged mice would further reduce the number of senescent cells in the lungs and decrease the accumulation of B16 melanoma cells. To exclude the possibility that the ketogenic diet and/or senolytic drugs directly affect the B16 melanoma cells, both interventions were discontinued 3 weeks before the transplantation of B16 melanoma cells (Fig. 4a,b). In aged mice fed a normal diet, the numbers of p16^INK4a^-positive and p21^WAF1/CIP1^-positive but Ki67-negative cells decreased in both the ABT263-treated and ARV825-treated lungs (Fig. 4c and Extended Data Fig. 10a,b). Moreover, there was a trend toward reduced accumulation of B16 cells in the lungs compared with vehicle-treated mice (Fig. 4d), and these effects were more pronounced in mice fed a ketogenic diet (Fig. 4c,d and Extended Data Fig. 10a,b). Notably, the expression of Cxcl2 and Cxcl12, genes encoding SASP factors known to promote B16 cell migration^49,50^, was substantially reduced in mice receiving both a ketogenic diet and senolytic drug treatment (Fig. 4e). Although a recent report showed that ketogenic diets can induce cellular senescence in several mouse tissues^51^, we did not observe such an effect at least in the lungs of aged mice. This suggests that the impact of a ketogenic diet on cellular senescence may be age dependent. Taken together, these findings indicate that combining a ketogenic diet with mitochondria-targeting senolytic drugs such as ABT263 or ARV825 can effectively reduce the burden of senescent cells, at least in the lungs of aged mice, thereby attenuating cancer cell accumulation.Fig. 4. Metabolic intervention enhances the inhibitory effects of senolytic drugs on tumor growth and metastasis in mice.a, Experimental timeline (n = 9–10 per group). Red arrows indicate administration of ABT263, ARV825 or vehicle for 5 consecutive days each week for 2 weeks. Black arrows denote daily feeding with a normal diet (ND) or ketogenic diet (KD); blue arrows indicate daily feeding with ND only. b, Blood glucose concentrations measured at week 94. c–e, Mice were euthanized at week 100, and tumor-free areas of the lungs were dissected and subjected to the following analyses: c, Expression levels of p16^INK4a^ and p21^WAF1/CIP1^ by RT–qPCR. d, Representative macroscopic lung images and quantification of metastatic nodules. e, Expression levels of Cxcl2 and Cxcl12 by RT–qPCR. f, Experimental timeline. On day −7, nude mice received subcutaneous transplantation of HCT116 cells. The blue arrow indicates daily treatment with DXR for 7 consecutive days. Red arrows denote administration of ABT263, ARV825 or vehicle for 5 consecutive days each week for 2 weeks. The black arrow indicates continuous access to dapagliflozin (SGLT2 inhibitor) or vehicle. g, Blood glucose concentrations on day 18. h, Tumor volumes measured every other day with calipers in biologically independent animals: control + vehicle (n = 9), control + ABT263 (n = 9), control + ARV825 (n = 9), SGLT2i + vehicle (n = 8), SGLT2i + ABT263 (n = 9) and SGLT2i + ARV825 (n = 8). i,j, Mice were euthanized, and serial sections of xenograft tumors were analyzed by immunofluorescence for senescence markers (p21^WAF1/CIP1^-positive or 53BP1-positive but Ki67-negative cells). Nuclei were counterstained with DAPI. Enlarged views of regions outlined with yellow dashed lines are shown in the lower-left corners; histograms to the right show the percentages of positive cells, quantified from four fields per mouse across five mice per group. Although data shown are from technical replicates (b–e and g–j), experiments were independently repeated at least once to confirm reproducibility. Data are presented as the mean ± s.d. (b–e and g–j). Statistical significance was assessed by a two-sided Student’s or Welch’s t-test (b–e and g–j). Scale bars, 10 mm (d) and 10 µm (i and j).Source data
To further explore the enhanced effects of combining a ketogenic diet with ABT263 or ARV825 in an animal model of cancer therapy, we used a therapy-induced senescence model^28^, as surviving cancer cells after chemotherapy can drive recurrence and progression. Our previous report showed that the concurrent administration of DXR, an anticancer drug known to induce cellular senescence, with ARV825 modestly improved the antitumor efficacy of DXR in a xenograft model using HCT116 colorectal cancer cells^14^. Moreover, at the concentrations used in our senolysis assays, neither ARV825 nor ABT263 alone inhibited the proliferation of HCT116 cells. A reduction in cell numbers was observed only when cells had been pretreated with DXR, consistent with senolysis-dependent cell death (Supplementary Fig. 4). We therefore investigated whether a ketogenic diet could further augment the efficacy of DXR combined with ARV825 or ABT263 in the same xenograft mouse model. However, when nude mice bearing HCT116 xenografts were treated with DXR while fed a ketogenic diet, many died. This outcome may reflect the known risk of severe side effects, such as ketoacidosis, in individuals on an extremely low-carbohydrate diet, especially under stress or in type 2 diabetes^52^. To overcome this limitation, we tested the sodium–glucose cotransporter 2 (SGLT2) inhibitor, which reduces blood glucose by blocking renal reabsorption, as an alternative means to lower glucose availability^53^. This approach decreased blood glucose levels without causing lethality in the xenograft model (Fig. 4f,g). Notably, the administration of DXR followed by a combined treatment with an SGLT2 inhibitor and ARV825 or ABT263 resulted in significantly stronger inhibition of tumor growth compared to individual treatments (Fig. 4h). In parallel, tumors exhibited a significant reduction in p21^WAF1/CIP1^-positive and 53BP1-positive but Ki67-negative cells, markers of senescence, together with increased ROS and apoptosis markers (Fig. 4i,j and Extended Data Fig. 10c,d). Collectively, these results suggest that ARV825 and ABT263 may exhibit further enhanced senolytic activity and improved suppression of tumor growth and metastasis when combined with a carbohydrate signaling blockade (see the model in Supplementary Fig. 5). Nevertheless, we cannot fully exclude the possibility that the observed antitumor effects may partially involve senolysis-independent actions of ABT263 and ARV825.
Discussion
Our study systematically compared reported senolytic drugs by the SSI and identified ABT263 and ARV825 as the most effective agents in our models (Fig. 1b,c and Extended Data Figs. 1 and 2). Both targeted mitochondrial function, yet a subset of senescent cells survived by maintaining mitochondrial integrity, at least in part, through V-ATPase-mediated clearance of damaged mitochondria (Fig. 2 and Extended Data Fig. 6). These findings indicate that senescent cell heterogeneity^54–56^, increasingly recognized in the field, is a key determinant of senolysis resistance. Thus, the mechanisms underlying senolytic efficacy and resistance may vary depending on cell type and context. Nevertheless, we demonstrate that metabolic interventions already in clinical use, such as ketogenic diets^47,48^ or SGLT2 inhibitors^53^, can increase mitochondrial workload and thereby induce mitochondrial stress, enhancing the senolytic efficacy of ABT263 and ARV825 in vivo (Fig. 4 and Extended Data Fig. 10). While our findings provide proof of principle that senolytic efficacy can be modulated in vivo, the effects of metabolic modulation are likely context dependent^51,52^, with potential consequences for SASP^3,10–12^, immune surveillance of senescent cells^57^ and tissue homeostasis^9,58^. Approaches using antibodies^59–62^ or chimeric antigen receptor T cells^63^ also underscore the potential of senolysis, although compound-based strategies may be advantageous for reversible modulation. Notably, a recent study by Fielder et al. reported that mitochondrial stress can potentiate the senolytic activity of ABT263 (ref. ^64^). In that report, mitochondrial uncoupling or galactose culture increased mitochondrial workload and sensitized senescent cells to ABT263 (ref. ^64^). This is consistent with our conclusion that mitochondrial quality control is a central determinant of senolytic resistance to mitochondria-targeting senolytic drugs, and our findings extend this concept by suggesting the potential for translational dietary and pharmacological interventions that are already clinically available. One limitation of this study is that all in vivo experiments were performed using male mice; therefore, the generalizability of our findings to female mice remains to be determined. Collectively, these observations suggest possible avenues for developing rational combination senotherapies.
Methods
Ethical approval
All animal experiments were approved by the Animal Research Committee of the Research Institute for Microbial Diseases, The University of Osaka.
Chemicals
The following chemicals were used: ABT263 (S1001, Selleck Chemicals), ABT737 (S1002, Selleck Chemicals), ABT199 (16233, Cayman Chemical), ARV825 (HY-16954, MedChemExpress), JQ1 (HY-13030, MedChemExpress), OTX015 (S7360, Selleck Chemicals), 17-DMAG (S1142, Selleck Chemicals), BPTES (19284, Cayman Chemical), CB839 (S7655, Selleck Chemicals), P5091 (S7132, Selleck Chemicals), RG7112 (S7030, Selleck Chemicals), IMP1088 (HY-112258, MedChemExpress), PCLX-001 (E1217, Selleck Chemicals), nintedanib (S1010, Selleck Chemicals), R406 (S2194, Selleck Chemicals), 25-hydroxycholesterol (HY-113134, MedChemExpress), digoxin (S4290, Selleck Chemicals), procyanidin C1 (HY-N2342, MedChemExpress), gingerenone A (HY-120912, MedChemExpress), fisetin (S2298, Selleck Chemicals), dasatinib (S1021, Selleck Chemicals), quercetin (10005169, Cayman Chemical), DXR (046-21523, FUJIFILM Wako Chemicals), etoposide (341205, Sigma-Aldrich), DAPI (340-07971, Dojindo), Hoechst 33342 (H342, Dojindo), N-acetyl-L-cysteine (013-05133, FUJIFILM Wako Chemicals), BAY876 (S8452, Selleck Chemicals), dapagliflozin (S1548, Selleck Chemicals) and EN6 (S6650, Selleck Chemicals).
Cell culture
IMR-90, TIG-1 and TIG-3 human fibroblasts, as well as B16BL6 mouse melanoma cells, were obtained from public bioresource banks. RPE-1 and HCT116 cells were obtained from Lonza and the American Type Culture Collection, respectively. IMR-90, TIG-3, RPE-1, HCT116 and B16BL6 cells were maintained in Dulbecco’s Modified Eagle’s medium supplemented with 10% fetal bovine serum and 100 U ml^−1^ penicillin–streptomycin. TIG-1 cells were cultured in Modified Eagle’s medium supplemented with 10% fetal bovine serum and 100 U ml^−1^ penicillin–streptomycin. All cell lines were routinely tested and confirmed to be negative for mycoplasma contamination. None of the cell lines used in this study are listed as misidentified or cross-contaminated by the International Cell Line Authentication Committee. Cell line authentication was not performed by the authors but was guaranteed by the supplier. For in vitro studies, cells were seeded and randomly allocated to experimental groups.
Mice
C57BL/6J and nude (nu/nu) mice were obtained from a commercial vendor (CLEA Japan). Mice were housed under specific pathogen-free conditions at 23 °C ± 2 °C and 55% ± 15% humidity on a 12-h light–dark cycle, with ad libitum access to either a normal diet (CE-2, CLEA Japan; 12 kcal% fat, 29 kcal% protein, 59 kcal% carbohydrates) or a ketogenic diet (D10070801, Research Diets; 90 kcal% fat, 10 kcal% protein, 0 kcal% carbohydrates). Experiments were performed in parallel under identical conditions. The maximal permitted tumor burden approved by the institutional Animal Research Committee was a tumor volume of ≤1,500 mm³ or a maximum diameter of ≤15 mm. Tumor size was monitored regularly, and animals were euthanized at the end of the experiments. The maximal permitted tumor size/burden was not exceeded in any experiment.
Lung metastasis assay
Before tumor inoculation, 87-week-old male C57BL/6J mice (n = 9 to 10 per group) were fed a normal diet or a ketogenic diet for 5 weeks. Mice then received oral administration of ABT263 (100 mg per kg body weight) or vehicle, dissolved in a mixture of 10% ethanol, 30% polyethylene glycol 400 and 60% Phosal 50 PG^65^, or intraperitoneal administration of ARV825 (5 mg per kg body weight) or vehicle, dissolved in 10% DMSO, 40% PEG300, 5% Tween-80 and 45% saline^14^, for 5 consecutive days per week over 2 weeks. After this treatment, ketogenic diet-fed mice were switched to a normal diet for 3 weeks. Subsequently, mice were injected with 1 × 10^5^ B16BL6 cells into the tail vein. Three weeks later, lungs were collected, surface metastatic nodules were counted, and the tissues were subjected to RT–qPCR and immunofluorescence analyses.
Xenograft tumor assay
HCT116 cells (5 × 10^5^) were mixed at a 1:1 (vol/vol) ratio with Matrigel (BD Biosciences) and injected subcutaneously into 6–8-week-old male nude mice (n = 8 to 9 per group). After intraperitoneal administration of DXR (1 mg per kg body weight) for 7 consecutive days, mice received dapagliflozin (SGLT2 inhibitor; 0.0125 mg ml^−1^ in drinking water, approximate daily dose of 2.5 mg per kg body weight) together with ABT263 (100 mg per kg body weight)^65^, ARV825 (5 mg per kg body weight)^14^ or vehicle, for 5 consecutive days per week over 2 weeks. Tumor size was measured using a Vernier caliper, and tumor volume (mm^3^) was calculated as (length × width^2^)/2. Tumor tissues were subjected to immunofluorescence analyses.
Blood glucose measurement
Blood glucose levels were measured using a Freestyle Glucose Monitoring System (71386-80, 80224-75; Abbott Laboratories).
Induction of senescence in vitro
For DXR-induced senescence, IMR-90, TIG-3, RPE-1 and HCT116 cells were incubated with DXR at 250 ng per ml, 250 ng per ml, 100 ng per ml and 200 ng per ml, respectively, for 10 days. For etoposide-induced senescence, TIG-1 cells were incubated with etoposide at 100 µM for 2 days. For replicative senescence, IMR-90 and TIG-3 cells were serially passaged until they exceeded 73 population doublings.
Cell proliferation and survival assay
Cells were seeded in 12-well plates, and cell numbers were counted daily at identical grid positions using gridded culture dishes. Relative cell numbers were calculated by normalizing to the initial count on day 0, which was set as 100%.
RNAi
RNA interference (RNAi) was performed by transfecting previously validated siRNA oligonucleotides using Lipofectamine RNAiMAX (13778150; Thermo Fisher Scientific) according to the manufacturer’s instructions. The siRNAs used were as follows: LIG3 (1: Dharmacon, ON-TARGETplus siRNA, ID: L-009227-00-0005; 2: Thermo Fisher Scientific, Silencer Select Pre-Designed siRNA, ID: s8177); POLG (1: Dharmacon, ON-TARGETplus siRNA, ID: L-012649-00-0005; 2: Thermo Fisher Scientific, Silencer Select Pre-Designed siRNA, ID: s10789); XRCC4 (1: Dharmacon, ON-TARGETplus siRNA, ID: L-004494-00-0005; 2: Thermo Fisher Scientific, Silencer Select Pre-Designed siRNA, ID: s14951); ATP6V0E1 (Dharmacon, ON-TARGETplus siRNA, ID: L-011559-01-0005); BRD4 (Dharmacon, siGENOME siRNA, ID: M-004937-02-0005); and control (Dharmacon, ON-TARGETplus Non-targeting siRNA, ID: D-001810-0X).
Western blotting
Cell pellets were lysed in RIPA buffer containing protease inhibitor cocktail (25955-11; Nacalai Tesque). Protein concentrations were determined using a Protein Assay kit (740967.250; Takara Bio). Samples were denatured in Laemmli sample buffer for 5 min at 95 °C, separated by SDS–PAGE, and transferred onto polyvinylidene difluoride membranes (EMD Millipore). After blocking with 5% milk, membranes were incubated with primary antibodies: β-actin (1:2,000 dilution; A5316, Sigma-Aldrich), lamin B1 (1:1,000 dilution; ab16048, Abcam), p16 (1:1,000 dilution; sc-56330, Santa Cruz), p21 (1:1,000 dilution; 2947, Cell Signaling Technology), XRCC4 (1:1,000 dilution; sc-271087, Santa Cruz), LIG3 (1:1,000 dilution; sc-135883, Santa Cruz), POLG (1:1,000 dilution; ab128899, Abcam), ATP6V0E1 (1:1,000 dilution; PA5-1114887, Thermo Fisher Scientific) and BRD4 (1:1,000 dilution; 13440, Cell Signaling Technology). Membranes were then incubated with horseradish peroxidase-conjugated secondary antibodies (1:2,000 dilution; 7074 and 7076, Cell Signaling Technology) and visualized using Amersham ECL Prime/Select (GE Healthcare), followed by detection with ImageQuant 800 (Cytiva).
Immunoprecipitation
Immunoprecipitation was performed using Dynabeads Protein G (10004D, Thermo Fisher Scientific). Protein lysates were pre-cleared by incubation with Dynabeads Protein G at 4 °C for 30 min in immunoprecipitation buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, and protease inhibitor cocktail). The pre-cleared lysates were then incubated overnight at 4 °C with 3 µg of antibody against XRCC4 (sc-271087, Santa Cruz Biotechnology), LIG3 (sc-135883, Santa Cruz Biotechnology) or control IgG (5415, Cell Signaling Technology). Immune complexes were captured with 30 µl Dynabeads Protein G at room temperature for 2 h, washed and analyzed by immunoblotting as described above.
Immunohistochemistry
Dissected tissues were fixed in Bouin’s solution for 2 h (for p16 staining)^66^ or in 4% paraformaldehyde (PFA) overnight at 4 °C (for all other staining), then embedded in paraffin and sectioned at 5 µm. Sections were deparaffinized and subjected to heat-mediated antigen retrieval in citrate buffer (pH 6.0). After permeabilization with 0.1% Triton X-100, sections were blocked with BLOXALL (SP-6000, Vector) and 2.5% normal horse serum, followed by incubation with primary antibodies.
Primary antibodies used were p16 (1:1000; ab211542, Abcam) for Bouin’s-fixed mouse samples^66^, and p21 (1:1,000 dilution, 2947, Cell Signaling; 1:1,000 dilution, ab188224, Abcam), Ki67 (1:1,000 dilution; 14-5698-82, Thermo Fisher Scientific) and 53BP1 (1:1,000 dilution; NB100-304, Novus Biologicals) for 4% PFA-fixed samples. Sections were incubated with secondary antibodies (ImmPRESS Polymer Anti-Rabbit IgG, Vector, MP-7401) for 30 min. Signal amplification was performed with the TSA TMR System (AKOYA Biosciences, NEL702001KT) following the manufacturer’s instructions. Sections were then incubated with secondary antibodies (Alexa Fluor 488, 1:2,000 dilution; A-21208, Thermo Fisher Scientific) and DAPI for 1 h, and mounted with Fluoromount-G (SouthernBiotech, 0100-01). Fluorescence images were acquired with an All-in-One Fluorescence Microscope (BZ-710; Keyence), and the number or area of positive cells was quantified automatically using ImageJ (version 1.54) software.
TUNEL analysis
TUNEL staining was performed on 4% PFA-fixed mouse tissues using the One-step TUNEL In Situ Apoptosis Kit (Red, Elab Fluor 555, E-CK-A325; Elabscience) according to the manufacturer’s instructions. Nuclei were counterstained with DAPI (4′,6-diamidino-2-phenylindole). Fluorescence images were acquired with an All-in-One Fluorescence Microscope (BZ-710; Keyence), and the number or area of TUNEL-positive cells was quantified automatically using ImageJ (version 1.54).
Dihydroethidium analysis
Frozen tumors embedded in Tissue-Tek optimal cutting temperature compound (4583, Sakura Finetek) were sectioned at 10 µm using a cryostat at −20 °C. To assess tissue ROS levels, sections were incubated with 5 µM dihydroethidium (HY-D0079, MedChemExpress) for 30 min at 37 °C. After washing with PBS, nuclei were counterstained with DAPI. Fluorescence images were acquired with an All-in-One Fluorescence Microscope (BZ-710, Keyence), and the area or number of dihydroethidium-positive cells was quantified automatically using ImageJ (version 1.54) software.
Cytochemistry
For fluorescence probe staining, cultured cells were incubated with MT-1 (1:1,000 dilution; MT13, Dojindo), 2.5 µM CellROX Deep Red (C10422, Thermo Fisher Scientific) or MitoTracker Red CMXRos (1:1,000 dilution; M7512, Thermo Fisher Scientific) for 30 min and then washed with PBS. After staining with MT-1, cells were fixed with 4% PFA. The MT-1 signal was stable after fixation. Fixed cells were either imaged directly using a TRITC filter or subjected to in situ hybridization for ATP6V0E1 mRNA based on the hybridization chain reaction (HCR) method. Imaging was performed using identical exposure times and gain settings across all experiments. Fluorescence intensity was measured with ImageJ (version 1.54) after background subtraction. For immunofluorescence staining, cultured cells were fixed with 4% PFA, permeabilized with 0.1% Triton X-100 for 10 min and blocked with 5% horse serum before incubation with primary antibodies (XRCC4, 1:200 dilution, sc-271087, Santa Cruz; CXCL2, 1:50 dilution, AF-452-NA, R&D Systems; GROα/β/γ, 1:50 dilution, sc-365870, Santa Cruz; CXCL5, 1:50 dilution, AF254, R&D Systems; CCL20, 1:50 dilution, AF360, R&D Systems; MMP3, 1:50 dilution, 66338-1-lg, Proteintech) overnight at 4 °C. The secondary antibody conjugated with Alexa Fluor 488 or Alexa Fluor 555 was applied for 1 h at room temperature. Nuclei were counterstained with DAPI. Fluorescence images were acquired using an All-in-One Fluorescence Microscope (BZ-710; Keyence), and the area or number of positive cells was quantified automatically using ImageJ (version 1.54).
EdU incorporation assay
EdU incorporation was assessed using the Alexa 555 Click-It EdU kit (C10338, Thermo Fisher Scientific) with modifications to the manufacturer’s protocol. Cells were incubated with 50 µM EdU for 7 days before fixation. After fixation, two sequential 1-h click reactions were performed with freshly prepared solutions. Coverslips were washed three times with 3% BSA in PBS, and nuclei were counterstained with Hoechst 33342. Fluorescence images were acquired with an All-in-One Fluorescence Microscope (BZ-710, Keyence), and the number or area of positive cells was quantified automatically using ImageJ (version 1.54).
Annexin V immunofluorescence staining
Cultured cells were washed with Annexin-binding buffer (10 mM HEPES, 140 mM NaCl, 2.5 mM CaCl_2_, pH 7.4), then incubated with Annexin V Conjugate (A13201, Thermo Fisher Scientific) for 15 min. After incubation, the cells were washed with Annexin-binding buffer and then fixed. Fixed cells were either imaged directly or subjected to in situ hybridization for ATPV0E1 mRNA based on the HCR method. Nuclei were counterstained with DAPI as described above. Fluorescence images were acquired with an All-in-One Fluorescence Microscope (BZ-710; Keyence), and the number or area of positive cells was quantified automatically using ImageJ (version 1.54).
HCR-based in situ hybridization
Cultured cells were fixed overnight at 4 °C in 4% PFA. After fixation, the cells were treated with methanol for 10 min at room temperature. Hybridization and amplification were performed using the ISHpalette Short Hairpin Amplifier A161-Cyan5 (IPL-B-A161, Nepa Gene) according to the manufacturer’s instructions. Nuclei were counterstained with DAPI as described above. Fluorescence images were acquired with an All-in-One Fluorescence Microscope (BZ-710, Keyence), and the number or area of positive cells was quantified automatically using ImageJ (version 1.54).
Oligonucleotide probe sequences used for detecting ATP6V0E1 mRNA are shown in Supplementary Table 1.
Real-time qPCR analysis
Total RNA was extracted using TRIzol (15596018, Thermo Fisher Scientific) or the RNeasy Mini Kit (74106, Qiagen) according to the manufacturer’s instructions. cDNA was synthesized with a commercially available RT Reagent Kit with gDNA Eraser (RR047A; vendor blinded for review). Real-time qPCR was performed on a StepOnePlus PCR System (Applied Biosystems) using TB Green Premix Ex Taq II (RR820A; vendor blinded for review). Expression levels of target genes were normalized to β-actin using the standard ΔΔCt method and expressed relative to the average value of the control cells, which was set to 1. β-actin was used as the internal reference gene following standard qPCR practice. The PCR primer sequences used are shown in Supplementary Table 2.
mtDNA copy number analysis
DNA was extracted from TIG-3 cells using the DNeasy Blood & Tissue Kit (69506, Qiagen). qPCR was performed on a StepOnePlus PCR System (Applied Biosystems) with TB Green Premix Ex Taq II (RR820A, Takara Bio). Standard curves were generated using 0.5 ng, 1 ng, 2 ng, 4 ng and 8 ng of TIG-3 cell DNA for mtDNA amplification, and 1 ng, 2 ng, 4 ng, 8 ng and 16 ng for nuclear DNA amplification. The primer sequences were as follows:
• mtDNA, forward 5′-CTTCTGGCCACAGCACTTAAC-3′ and reverse
5′-GCTGGTGTTAGGGTTCTTTGTTTT-3′;
•β2M (nuclear DNA), forward 5′-GCTGGGTAGCTCTAAACAATGTATTCA-3′ and
reverse 5′-CCATGTACTAACAAATGTCTAAAATGGT-3′.
RNA-seq analysis
RNA isolation was performed using TRIzol. Library preparation was performed using a TruSeq stranded mRNA Library Prep kit (Illumina) according to the manufacturer’s instructions. Sequencing was performed on an Illumina NovaSeq 6000 sequencer (Illumina) in 101-base paired-end mode. Sequenced reads were mapped to the human reference genome sequences (hg19) using TopHat version 2.2.1. The number of fragments per kilobase of exon per million mapped fragments was calculated using Cufflinks version 2.2.1. The DEG and GO analysis was performed using the iDEP.96 webtool.
scRNA-seq analysis
scRNA-seq was performed using the Chromium Fixed RNA Kit, Human Transcriptome (10x Genomics). Approximately 10⁶ fixed TIG-3 cells were hybridized with whole-transcriptome probe pairs for 18 h at 42 °C. Following post-hybridization washes, the cells were encapsulated into gel beads-in-emulsion (GEMs) using the Chromium X system. Within each GEM, ligated probe pairs were barcoded with 10x GEM barcodes and unique molecular identifiers, followed by heat denaturation and recovery. Barcoded products were pre-amplified and purified using SPRIselect beads. Final libraries were generated by sample index PCR with the Dual Index Kit TS Set A and purified by size selection. Library quality and size distribution were assessed using an Agilent Bioanalyzer, and concentrations were determined with the KAPA Library Quantification Kit (Roche). Libraries were sequenced on an Illumina NovaSeq X Plus platform with 151-base pair paired-end reads. Demultiplexing, alignment and gene expression quantification were performed using Cell Ranger v9.0.1 (10x Genomics), aligned to the human reference genome (GRCh38-2024-A). Filtered feature–barcode matrices were generated for downstream transcriptomic analyses.
OCR analysis
OCR was measured using a Seahorse XFe24 extracellular flux analyzer (Agilent Technologies). Cells (0.5–1 × 10⁵ per well) were seeded in XFe24 cell culture microplates (Agilent Technologies) and incubated with or without 7 µM BAY876 for 3 days. OCR was measured at 37 °C in Seahorse assay medium (5 mM HEPES, 10 mM glucose, and 10 mM pyruvate, pH 7.4). Each measurement cycle consisted of 3 min mixing, 30 s waiting and 3 min measurement. The first three cycles were used to assess basal respiration. Subsequently, 1.5 μM oligomycin, 1 μM FCCP, 0.5 μM rotenone and 0.5 μM antimycin A were sequentially injected, and OCR was recorded in IMR-90, TIG-3 and RPE-1 cells. After measurements, cells were stained with Hoechst 33342, and OCR values were normalized to viable cell numbers.
Statistics and reproducibility
All data are presented as the mean ± s.d. Statistical analyses were performed using GraphPad Prism software (version 10). Intergroup comparisons were performed using a two-tailed Student’s or Welch’s t-test, Kruskal–Wallis test or one-way ANOVA followed by Tukey’s or Sidak’s post hoc test, depending on the experimental design. Statistical details are provided in the figure legends. The criterion for statistical significance was set at P < 0.05. All experiments, except those shown in Fig. 2a–c and Extended Data Fig. 4a,b, were repeated at least two independent times with similar results. Data distribution was assumed to be normal, but this was not formally tested. Individual data points are shown in the figures. Data collection and analysis were not performed blind to the conditions of the experiments. Mice were randomly assigned to experimental groups before treatment using simple randomization, with group allocation stratified by genotype and/or treatment. No statistical methods were used to predetermine sample sizes for the lung metastasis and xenograft tumor assays, but the sample sizes used are similar to those reported in previous publications (refs. ^14^,^46^, respectively). No animals or data points were excluded from the analysis, except for aged mice bearing spontaneous cancers, which were excluded before analysis.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Supplementary information
Supplementary InformationSupplementary Figs. 1–5. Reporting Summary Supplementary Table 1. Probes used for HCR-based in situ hybridization. Supplementary Table 2. Primers used for RT–qPCR analysis. Supplementary DataStatistical source data of supplementary figures.
Source data
Source Data Fig. 1. Statistical source data. Source Data Fig. 2. Statistical source data. Source Data Fig. 3. Statistical source data. Source Data Fig. 4. Statistical source data. Source Data Extended Data Fig. 1. Statistical source data. Source Data Extended Data Fig. 2. Statistical source data. Source Data Extended Data Fig. 3. Statistical source data. Source Data Extended Data Fig. 4. Statistical source data. Source Data Extended Data Fig. 5. Statistical source data. Source Data Extended Data Fig. 6. Statistical source data. Source Data Extended Data Fig. 6. Unprocessed western blot. Source Data Extended Data Fig. 7. Statistical source data. Source Data Extended Data Fig. 7. Unprocessed western blot. Source Data Extended Data Fig. 8. Statistical source data. Source Data Extended Data Fig. 8. Unprocessed western blot. Source Data Extended Data Fig. 9. Statistical source data. Source Data Extended Data Fig. 9. Unprocessed western blot. Source Data Extended Data Fig. 10. Statistical source data.