Senolytic‐Resistant Senescent Cells Have a Distinct SASP Profile and Functional Impact: The Path to Developing Senosensitizers

Preadipocytes (also known as adipose‐derived stromal cells or adipose mesenchymal "stem" cells [MSCs]) were isolated from abdominal subcutaneous fat biopsies obtained from subjects donating kidneys for transplantation aged 38.4 +/− 11.4 years (mean +/− SD), median 37 (25–68), with a body mass index 28.3 +/− 3.9 (mean +/− SD), median 27.2 (23–40.46) or undergoing bariatric surgery aged 39.25 +/− 8.5 years (mean +/− SD), median 40.5 (28–48) with a body mass index 39.7 +/− 5.03 (mean +/− SD), median 37.29 (36.6–47.1), both males and females. We primarily used preadipocytes obtained from kidney transplant patients. Preadipocytes from both kidney transplant and bariatric surgery patients were used in Figures 1, 4, and 5. All subjects gave informed consent. The protocol was approved by the Mayo Clinic Institutional Review Board for Human Research. Using methods we previously reported (Tripathi, Nchioua, et al. 2021): (1) human preadipocytes were isolated and cultured; (2) all human preadipocyte lots that were not induced to become senescent retained proliferative capacity over multiple passages with low markers of senescence, and were therefore considered to be non‐senescent/control cells; and (3) human preadipocyte senescence was induced by 20 Gy X‐irradiation (X‐Rad320, PRECISION) and experiments were performed 30 days after radiation. At this time point, more than 70% of cells were positive for SA‐β‐gal staining, exhibited minimal proliferation, and had markedly elevated expression of SASP factors (Tripathi, Nchioua, et al. 2021). To minimize inter‐donor variability, we employed a paired experimental design in Figures 1B,D, 4E, and 5B, Figure S4C,D. Preadipocytes from each donor were divided into three groups: one served as a non‐irradiated control (non‐senescent) and the others were subjected to irradiation to induce senescence followed by vehicle or senolytic treatment. In HUVECs purchased from Lonza (CC‐2519, San Diego, California), senescence was induced by 10 Gy X‐irradiation and experiments were conducted 10–14 days after irradiation. The MDA‐MB‐231 human breast cancer cell line was obtained from ATCC. Cells were treated with 1 μ/mL Cisplatin for 24 h and maintained in culture for 1 week. The cells were then treated with Dasatinib (200 nM) or vehicle for 24 h followed 24 h later by a TLR3 agonist, polyinosine‐polycytidylic acid (10 μg/mL), or vehicle daily for 3 days. After treatment with the TLR3 agonist or vehicle, cells were again exposed to Dasatinib or vehicle for 24 h.

Dasatinib (SML2589) and Quercetin (Q4951) were purchased from Sigma (Sigma, St. Louis, MO, USA). The JAK inhibitor Ruxolitinib (tlrl‐rux) was purchased from InvivoGen (Sigma, St. Louis, MO, USA). Lipopolysaccharide (LPS; tlrl‐3pelps) was purchased from InvivoGen (San Diego, CA, USA).

For most studies, cells were washed with PBS and RNA was isolated using Trizol and chloroform (Xu et al. 2015). Concentration and purity of samples were assayed using a Nanodrop spectrophotometer. Each cDNA sample was generated by reverse transcription using 1–2000 ng RNA following the manufacturer's recommended protocol (High‐capacity cDNA Reverse Transcription Kit; Cat #4368813, Thermo Fisher Scientific, Waltham, MA, USA). A reverse transcription program was used (10 min at 25°C, 120 min at 37°C, 5 min at 85°C, held at 4°C). TATA‐box binding protein (TBP) served as a control for gene expression analyses. A list of primers is provided in Table 2. For RNA sequencing, library preparation was performed using a TruSeq Stranded mRNA kit prior to sequencing on the Illumina NovaSeq 6000 platform.

Raw reads were trimmed with fastp (v0.20.1) (Bolger et al. 2014). Trimmed reads were then mapped to the reference genome GRCh38 with STAR (v2.7.9a) (Kim et al. 2015) and gene expression is estimated by generating a count matrix with STAR‐counts. The raw gene counts were adjusted for the sample pairing covariate using ComBat‐Seq within the R package, sva v3.50.0 (Zhang et al. 2020). Differential expression analysis was performed using the R package, edgeR v4.0.16 (Robinson et al. 2009). Raw gene counts were first pre‐filtered to keep only those genes with greater than one average count per million in either comparison group. The raw p values were adjusted using the Benjamini–Hochberg procedure (Benjamini and Hochberg 1995) to control the false discovery rate, and those genes with an adjusted p value ≤ 0.05 were considered to be significant. Differentially expressed genes were used for heatmap visualization and hierarchical clustering. Functional enrichment analyses were performed using clusterProfiler v4.14.0 (Wu et al. 2021).

CM were prepared by exposing cells to RMPI 1640 containing 1 mM sodium pyruvate, 2 mM glutamine, MEM (minimum essential medium) vitamins, MEM nonessential amino acids (all Gibco), and Streptomycin (Life Technologies, Carlsbad, CA, USA) (Xu et al. 2015). CM were collected 24–48 h after the cells had been exposed to the media.

Cellular apoptosis was assessed using the In Situ Cell Death Detection Kit (Roche, 11684795910) according to the manufacturer's instructions. Samples were counterstained with DAPI (Life Technologies). Apoptotic and nonapoptotic cells were examined by fluorescence microscopy (Nikon ECLIPSE Ti, NIS Elements AR 5,20,02).

Cell survival was assayed by crystal violet. Briefly, cells were washed with PBS twice and fixed with 4% PFA in PBS for 15 min and then washed again with PBS after incubation. Cells were then treated with 0.5% crystal violet in 25% methanol. After incubation for 10 min, crystal violet was removed and plates were washed with water until washings were clear of visible dye. Plates were dried and crystal violet was dissolved in 100% methanol. Absorbance at 570 nM was quantified.

SAβgal activity was assayed as previously described (Baker et al. 2011). In brief, primary preadipocytes were washed with PBS and then fixed for 5 min in PBS containing 2% (vol/vol) formaldehyde (Sigma‐Aldrich, St. Louis, USA) and 0.25% glutaraldehyde (Sigma‐Aldrich, St. Louis, USA). Following fixation, cells or tissues were washed with PBS before being incubated in SAβgal activity solution (pH 6.0) at 37°C for 16–18 h. The enzymatic reaction was stopped by washing cells or tissues 3–5X with ice‐cold PBS. Bright field microscopy (Nikon Eclipse Ti) was used for imaging.

Total DNA was extracted from media using the QIAmp‐Blood & Tissue kit (Qiagen, Germantown, MD, USA). mtDNA copy number was measured by quantitative rtPCR using primers for mitochondrially encoded NADH dehydrogenase subunit 2 (MT‐ND2) and cytochrome c oxidase subunit III (MT‐COX3) of the electron transport chain.

Cells were fixed with 4% (vol/vol) paraformaldehyde for 10 min and permeabilized using PBS containing 0.25% Triton X‐100 for 10 min. After being washed with PBS, cells were blocked with PBS and Tween‐20 (PBST) containing 1% BSA for 30 min. Cells were incubated with primary antibodies in 1% BSA in PBST (blocking buffer) overnight at 4°C, washed with PBS and then incubated with secondary fluorescent antibodies (Life Technologies, Carlsbad, CA, USA) in the blocking buffer for 1 h in the dark. DAPI (Life Technologies, Carlsbad, CA, USA) was used to stain nuclei for cell counting. Fluorescence microscopy (Nikon Eclipse Ti) was used for imaging. For EdU incorporation assays, cells were incubated in 10 μM EdU for 24 h in growth medium. All subsequent EdU detection steps were carried out using Click‐iT EdU Alexa Fluor Imaging Kits (C10337, Thermo Fisher Scientific, Waltham, MA USA) according to the manufacturer's instructions. The following antibodies were used: p16 (1:200 dilution, Cat# 705‐7493, Cintech Histology); p21 (1:100 dilution, Cat# 109199, Abcam, Cambridge, MA); and Phospho‐Histone H2A.X (Ser139) (20E3) (Rabbit; 1:200 dilution, Cat #9718, Cell Signaling, Danvers, MA).

Two‐month‐old SCID‐beige mice (Charles River Laboratories, Wilmington, MA, USA) were used for the SC transplantation studies. Mice were maintained in a pathogen‐free facility at 23°C–24°C under a 12‐h light, 12 h dark regimen with free access to water and a chow diet (standard mouse diet with 20% protein, 5% fat [13.2% fat by calories], and 6% fiber; Lab Diet 5053, St. Louis, MO). For the obese mouse studies, animals were maintained on a 60% (by calories) fat diet (D12492, irradiated; Research Diets, New Brunswick, NJ). All mouse studies were approved by the Mayo Clinic Institutional Animal Care and Use Committee.

Human preadipocyte senescence was induced by 20 Gy x‐irradiation. SCs or control nonirradiated cells were collected by trypsinization. Cell pellets were washed with PBS once and resuspended in PBS for transplantation. Recipient SCID‐beige mice were anesthetized using isoflurane and cells were injected intraperitoneally in 150–200 μL PBS through a 22 G needle (Xu et al. 2018).

Forelimb grip strength was determined in the transplanted mice using a grip strength meter (Columbus Instruments, Columbus, Ohio) (Xu et al. 2018). Results were averaged from 3 to 5 trials. For the wire hanging test (Xu et al. 2018), mice were placed on a 2 mm thick metal wire, which was placed 35 cm above a padded surface. Mice were allowed to grab the wire only with their forelimbs. Hanging time was normalized to body weight as hanging duration in seconds × body weight (grams). Results were averaged from 3 trials for each mouse. In this experiment, mice were housed in cages. This may introduce clustering effects due to the dependency of the mice in the same group. The authors acknowledge the need to properly account for clustering due to the risk of underestimating standard errors and inflating the Type I error rates. However, we do not expect a significant impact of the clustering effects in this study because animals from all three groups (treated with non‐senescent cells, senescent cells, and senolytic‐resistant senescent cells) were together in the same cages. Thus, any within‐cluster effects on the variability of the outcome would impact all mice rather than a specific mouse group. Additionally, findings rely more on treatment group responses, not individual mouse responses.

CM were filtered and cytokine and chemokine protein levels in CM were assayed using Luminex xMAP technology. Multiplexing analysis was performed using a Luminex 100 system (Luminex, Austin, TX, USA) by Eve Technologies Corp. (Calgary, Alberta, Canada). Data are represented as pg/ml for each SASP factor as a function of cellular density.

An antibody panel was designed based on surface markers, transcription factors, and cytokines. Each antibody was tagged with a rare metal isotope (Table 1) and function verified by mass cytometry according to the manufacturer's manual (Multi Metal labeling kits; Fluidigm South San Francisco, CA, USA). A CyTOF‐2 mass cytometer (Fluidigm, South San Francisco, CA, USA) was used for data acquisition. Acquired data were normalized based on normalization beads (Ce140, Eu151, Eu153, Ho165, and Lu175 as in (Palmer et al. 2019)). Preadipocytes were isolated from mouse epididymal white adipose tissue. Collected cells were incubated with metal‐conjugated antibodies for testing intracellular proteins including transcription factors and cytokines. Fixation and permeabilization were conducted according to the manufacturer's instructions (Foxp3/Transcription Factor Staining Buffer Set, eBioscience, San Diego, CA, USA). CyTOF data were analyzed by Cytobank (Santa Clara, CA, USA). Cell populations were clustered and visualized using viSNE, an implementation of the t‐distributed stochastic neighbor embedding (t‐SNE) algorithm based on the Barnes‐Hut approximation, available within the Cytobank platform (Van Der Maaten 2014). Total preadipocytes were considered to be CD45⁻, CD31⁻, and Sca1⁺, macrophages F4/80⁺ and CD11b⁺, endothelial cells CD31⁺ and CD146⁺, and T lymphocytes CD4⁺ or CD8⁺ as in (Palmer et al. 2019).

GraphPad Prism version 10.0.3 was used for statistical analyses. Results are presented as mean ± SD or SEM as stated in the figure legends. p < 0.05 was considered to be statistically significant and p < 0.05, p < 0.01, p < 0.005, and p < 0.001 are indicated as *, **, ***, and ****, respectively. Statistical analyses were performed using two‐tailed tests. Statistical tests include Student's t‐tests or two‐tailed Mann–Whitney tests comparing two groups, one‐way ANOVA and post hoc multiple comparison test with Tukey's Honestly Significant Difference (HSD) comparing more than two groups of pairs as detailed in the figure legends. Due to the limited sample size and the dependence of the methods used on the distributional assumptions of normality, a sensitivity analysis was then conducted using nonparametric tests based on taking the ranks of the variables then analyzing the ranks. Statistical significance of the results was diminished upon nonparametric testing as opposed to using standard t‐tests, supporting the desirability of conducting future confirmatory studies with larger sample sizes.

The JAK/STAT inhibitor (JAKi), Ruxolitinib, attenuates the SASP of senescent human preadipocytes (Xu et al. 2015). Reducing the SASP by Ruxolitinib or si‐RNA mediated knockdown of JAK1 attenuated the ability of Dasatinib to clear senescent human preadipocytes (Figure 1A,B, Figure S1A,B). Pathogen‐associated molecular pattern (PAMP) factors, such as lipopolysaccharide (LPS), which is in the outer membrane of gram‐negative bacteria, have been shown to exacerbate pro‐inflammatory SASP factor expression (Figure S1C) (Camell Christina et al. 2021). Here, we found that LPS pre‐treatment of SCs enhanced their killing by senolytics (Figure 1C,D). Neither JAK inhibitor nor LPS changed the senescent markers (Figure S1D,E). Hence, interference with SASP factor expression related to the JAK/STAT pathway decreases the impact of senolytic treatment on human preadipocytes, while "senosensitizing" microenvironmental factors such as those linked to infections increase susceptibility of senolytic‐resistant SCs to senolytic treatment (Tripathi, Nchioua, et al. 2021; Camell Christina et al. 2021).

Senolytics were developed based on the observation that those SCs that are tissue‐damaging employ antiapoptotic, pro‐survival pathways (SCAPs) to protect themselves from their own proapoptotic SASP (Zhu et al. 2015). Different senolytics discovered using this hypothesis‐driven approach target 30% to 70% of senescent cells in vitro (Zhu et al. 2015, 2017). Here we tested if it is indeed the senescent cells with high pro‐inflammatory/proapoptotic SASP factor expression that are selectively vulnerable to senolytics in vivo. Fourteen‐ to 16‐month‐old male mice that had been on a high‐fat diet (see Methods) for 8 months were administered Dasatinib (5 mg/kg) plus Quercetin (50 mg/kg) or vehicle (60% Phosal, 10% ethanol, and 30% PEG‐400) for five consecutive days by oral gavage. Mice were euthanized 3 days after the final dose. Organs were harvested and single‐cell suspensions from epididymal fat tissue were analyzed by CyTOF (Figure 2A). Changes in cellular composition caused by senolytics are shown in Figure 2B. Senolytics decreased preadipocyte and immune cell abundance (black arrows). Markers of senescence, including p16^Ink4a, p21^Cip1, CENP‐B, and γ‐H2AX, were detected in CD34⁺ cells (preadipocytes) and F4/80⁺ cells (macrophages; Figure 2C). Among preadipocytes, cluster c, a pro‐inflammatory/proapoptotic subset of senescent cells, was decreased by senolytics (Figure 2D,E). With the sample size used in our study, senolytic treatment may have had a slight but yet nonsignificant effect on macrophages, including those that express senescence markers and pro‐inflammatory factors (Figures S2A,B and S3A). Natural killer (NK) cells were reduced (Figure S3A), but these NK cells did not have detectable expression of the DNA damage marker, γ‐H2AX (Figure 2C). Senolytic treatment did not lead to statistically significant changes in endothelial cell abundance, which may relate to their lower expression of senescence markers (Figure 2D). As we reported previously in cultured cells and tissue explants (Xu et al. 2018; Zhu et al. 2015), these findings suggest senolytics primarily target those SCs with a pro‐inflammatory/apoptotic SASP in vivo and that the presence of pro‐inflammatory SCs correlates with immune cell infiltration; however, this needs further investigation.

To test whether those cells that are resistant to senolytics are indeed senescent, several established markers of senescence were analyzed. We previously showed that the efficacy of different senolytics depends on the cell type of origin (Zhu et al. 2015). For example, Dasatinib targets senescent preadipocytes, while Quercetin targets senescent HUVECs (Zhu et al. 2015). Senescent preadipocytes and HUVECs were treated with Dasatinib or Quercetin, respectively, for 24 h. Next, the remaining cells were washed with PBS 4X and subsequently incubated for 3 days. The resulting senolytic‐resistant SCs were proliferatively arrested as confirmed by BrdU labeling (Figure 3A). Senolytic‐resistant (only cells remaining after exposure to senolytics) were compared to total (senolytic‐resistant plus senolytic‐sensitive) SC populations because it is not yet feasible a priori to isolate only senolytic‐sensitive SCs. In both the total and senolytic‐resistant SC populations, we did not detect statistically significant differences in the markers associated with cell cycle inhibition (p16^INK4a and p21^CIP1) and the DNA damage marker, γ‐H2AX (Figure 3B–D and Figure S4A–C). Like the total SC population, senolytic‐resistant SCs were SAβgal+ (Figure 3E). Furthermore, senescence markers in Quercetin‐treated senescent endothelial cells (HUVECs) were similar to the total senescent HUVEC population (Figure S4D–F). Hence, consistent with our previous reports (Xu et al. 2018; Zhu et al. 2015), we propose, at least in vitro, that SC populations comprise two subsets, the one being susceptible to senolytics and the other resistant.

To test further if senolytic treatment indeed targets pro‐inflammatory senescent subtypes, bulk RNA‐sequencing was conducted of preadipocytes cultured with Dasatinib (senolytic‐resistant SCs) or vehicle (total SCs) from six healthy human kidney transplant donors after inducing senescence by X‐radiation. Clustering of the differential gene expression profiles between senolytic‐resistant versus total SCs is shown in the heat map and volcano plots of key genes (Figure 4A,B). A detailed list of all differentially expressed genes is provided in Table S1. Potential differences in biological function between the senolytic‐resistant versus total human preadipocyte SC populations were ascertained by Gene Ontology over‐representation analysis of the differentially expressed genes (Figure 4C,D, Table S2). Senolytic‐resistant SCs exhibited enrichment of biological processes associated with growth and repair and tissue development compared to the total SC population (Figure 4D). Repressed functions in senolytic‐resistant SCs included activation and migration of immune cells, as further suggested by gene expression and protein analyses (Figure 4C,E, Figure S5), including genes encoding chemokines/cytokines such as CXCL1, CXCL5, and CXCL8. Furthermore, senolytic‐resistant SCs had higher glycoprotein non‐melanoma‐B (GPNMB) expression, a recently identified senescence marker (Saade et al. 2021; Suda et al. 2022, 2021), compared to the total SC population (Figure 4E). GPNMB can impact innate and adaptive immune function, decrease inflammation, promote tumor growth, and shield some cancers from immune/inflammatory defenses (Maric et al. 2013; Lazaratos et al. 2022).

Analogously to preadipocytes, senescent HUVECs resistant to Quercetin had differences in gene expression compared to the total HUVEC SC population, with a less inflamed gene expression pattern in the senolytic‐resistant HUVECs than the total senescent HUVEC population (Figure S6A). With the sample size used in our studies, unlike SCs, non‐senescent control human preadipocyte or HUVEC cultures exposed to senolytics did not have statistically significant reductions in their already low pro‐inflammatory SASP factor expression compared to cells exposed to vehicle (Figure S6B,C). Although some SASP factors were upregulated in non‐senescent cells treated with senolytics, the levels of these pro‐inflammatory factors remained lower than SCs. This observation suggests that while senolytics clear SCs and reduce inflammation over the long term, they can also elicit stress adaptation responses by short‐term, transient activation of pro‐inflammatory signals in non‐senescent cells. These findings support the possibility that senolytics primarily target the SC subtype with an inflammatory, apoptosis‐inducing SASP profile, consistent with the hypothesis‐driven approach that we used to discover the first senolytics (Zhu et al. 2015).

Senescent preadipocytes can spread an inflammatory state to non‐senescent cells and activate immune cells, in part through SC mtDNA production and release (Xu et al. 2015; Iske et al. 2020). This induction of inflammation by SC mtDNA was recently confirmed (Zhu et al. 2017; Victorelli et al. 2023). Given the differences in SASP profiles between senolytic‐resistant and the total SC populations, whether the SC subtypes induce inflammation to the same extent in non‐senescent cells was tested (Figure 5A). Non‐senescent preadipocytes from healthy donors had lower expression of inflammatory factors when exposed to CM from senolytic‐resistant cells compared to those cultured with CM prepared from the total SC population (Figure 5B). Furthermore, cell‐free mtDNA content in CM, marked by expression of NADH dehydrogenase subunit 2 (MT‐ND2) and cytochrome c oxidase subunit III (MT‐COX3) of the electron transport chain (known to be proapoptotic (Iske et al. 2020; Victorelli et al. 2023)) tended to be higher in cultures of the total SC population than in cultures of the senolytic‐resistant SC population (Figure 5C), unlike cell‐free nuclear DNA (Figure S7). Hence, senolytic‐resistant SCs appear to provoke less inflammation than the total population of SCs (senolytic‐resistant plus senolytic‐sensitive SCs).

Transplanting small numbers of SCs into mice is sufficient to induce features of frailty, suggesting a role of SCs in causing physical dysfunction (Xu et al. 2018). To test whether the SC subtypes differ in their impact on physical function in vivo, 1 × 10⁶ senolytic‐resistant or total population SCs were transplanted intraperitoneally into 2‐month‐old male young SCID‐Beige mice (Figure 6A). After 1 month, physical function variables were tested. In previously healthy mice that were transplanted with total SC population cells, reductions in grip strength (Figure 6B) and hanging endurance (Figure 6C) were more evident than in mice transplanted with senolytic‐resistant SCs, indicating differences between the impacts of the SC subtypes on function.

Lastly, we explored the potential application of senosensitizers in the context of cancer. Triple‐negative breast cancer (TNBC) remains one of the most challenging malignancies to treat due to its poor prognosis and resistance to conventional therapies. It has been found that chemotherapy can induce a senescent phenotype in cancer cells and that senolytics can eliminate some of these therapy‐induced senescent (TIS) cells (Pacifico et al. 2024). Based on these points and our findings above about senolytic‐sensitive versus‐resistant SCs, we hypothesized senosensitizers could enhance the effectiveness of senolytics in removing TIS. We previously found TLR3 agonists (which activate the PAMP pathway also activated by coronaviruses) upregulate pro‐inflammatory/proapoptotic SASP factors in senescent cells (Tripathi, Nchioua, et al. 2021). Since LPS is not likely to become a feasible senosensitizing intervention in humans, we used a TLR3 agonist as a senosensitizer. As a preliminary test of our hypothesis, we treated human TNBC MDA‐MB‐231 cells with Cisplatin, followed by Dasatinib, then the TLR3 agonist polyinosinic‐polycytidylic as a senosensitizer, and finally Dasatinib again (Figure S9). This "1:2:3:4" stepwise regimen (chemotherapy → senolytic → senosensitizer → senolytic) resulted in a ~90% further reduction in TNBC cell survival than cells treated with chemotherapy alone (Figure S9).