Decreased Glucose Metabolism and Declined Chaperones Are Unique Features Required for the Survival of Senescent Fibroblasts and Pyruvate Dehydrogenase Is a Potent Senolytic Target
Mingzhu Zhang, Ziqi Hu, Shengwen Piao, Yingrui Song, Ying Jia, Jiaxing Liu, Ning Zhao, An Liu, Songbin Fu, Wenjing Sun, Hui Xu, Yu Yang, Steven P. Gygi, Chunshui Zhou
TL;DR
This study identifies unique metabolic and chaperone features in senescent fibroblasts and shows that targeting pyruvate dehydrogenase can selectively kill these cells.
Contribution
The study reveals pyruvate dehydrogenase as a novel and potent target for selectively eliminating senescent cells.
Findings
Decreased glucose metabolism and reduced ATP and alpha-KG production are key features of senescent fibroblasts.
Inhibiting pyruvate dehydrogenase or Hsp90 selectively kills senescent fibroblasts.
Combining TCA cycle inhibition with Hsp90 inhibition enhances the elimination of senescent cells and improves physical function in aged mice.
Abstract
Cellular senescence contributes to aging and age‐related diseases. Deep identifications of the senescence‐specific cellular features are crucial to the better understanding of the survival and maintenance of senescence and the development of novel senolytics against senescent cells. By a global proteomic profiling of senescent human BJ fibroblasts induced by ionizing radiation, 178 cellular proteins with at least 4‐fold or greater changes in abundance were identified, representing the cellular landscape of the senescent fibroblasts. Functional enrichments and biological experiments demonstrated that the decreased glucose metabolism, reduced ATP and alpha‐KG production, and declined chaperones are the most striking features associated with senescent fibroblasts. Moreover, these proteomic features are closely correlated with their transcription alterations confirmed by RT‐PCR.…
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 1
FIGURE 2
FIGURE 3
FIGURE 4
FIGURE 5
FIGURE 6
FIGURE 7
FIGURE 8- —National Natural Science Foundation of China10.13039/501100001809
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 · Genetics, Aging, and Longevity in Model Organisms · DNA Repair Mechanisms
Introduction
1
Cellular senescence with permanent cell cycle arrest and active metabolism is characterized by enlarged and flattened morphologies, SA‐β‐gal positive staining and senescence‐associated secretory phenotype (SASP) such as the secretion of a large number of cytokines, chemokines, growth factors, various proteases and protease inhibitors, etc. Cellular senescence can be induced by telomere shortening (Blasco 2007; Campisi 2013; López‐Otín et al. 2023), as well as a variety of stresses including DNA damage. DNA damage‐induced senescence has more important implications in clinical practice, particularly, the cancer treatment. Conventional radiotherapy and chemotherapy can cause some tumor cells and their surrounding stromal cells at senescence, which is referred to as therapy‐induced senescence (TIS) (Campisi and d'Adda di Fagagna 2007; te Poele et al. 2002). In a short term, TIS may serve as an intrinsic tumor suppression mechanism; however, in the long run, TIS may be the causal source for tumor relapse and therapy resistance (Campisi 2005; Demaria et al. 2017; Milanovic et al. 2018; Schosserer et al. 2017), thus, selectively killing these TIS cells may benefit cancer treatment.
As cellular senescence contributes to aging and aging related pathologies including cancer, the elimination of senescent cells might delay the aging process and the onset of aging related diseases (Baker et al. 2011; Childs et al. 2016, 2021; Yang et al. 2024). As such, senotherapeutics including senolytics (drugs selectively killing senescent cells) and senomorphics (drugs modulating SASP effects) have been developed vigorously (Chaib et al. 2022; Di Micco et al. 2021; Kang 2019). A prominent example is the combination treatment with dasatinib and quercetin, which was proven to kill senescent cells selectively, reduce the secretion of pro‐inflammatory cytokines, alleviate physiological dysfunction, and increase the survival of elderly mice by targeting the survival signaling pathways (Xu et al. 2018). However, several drawbacks may limit their clinical applications. First of all, the selectivity of dasatinib and quercetin is relatively poor because their targets tyrosine kinases and the PI3K‐AKT pathway are also essential for normal cell survival; long‐term treatment may cause serious side effects such as pulmonary edema (Campisi et al. 2019). Second, the levels of their molecular targets vary a lot in different cells, tissues and various senescence triggers (Hernandez‐Segura et al. 2017), such target expression variations may influence the efficacy of senolytics. Third, the dynamic compositions and diverse impacts of SASP also hamper the development of senomorphics (Coppé et al. 2008). Therefore, the deeper identification of the senescent cellular features and better understanding of senescence associated biological pathways are crucial for the development of novel senolytics.
Despite the extensively explored secretion dynamics and transcription alterations of senescent cells (Basisty et al. 2020; Ma et al. 2023; Olinger et al. 2025), the unique cellular characteristics of senescent cells remain largely unclear due to the high heterogeneity of cell types, tissue specificity, and the different nature of senescence‐inducing stimuli (Marmisolle et al. 2025; Tao et al. 2024). In order to elucidate the whole cellular features of senescent cells induced by DNA damage, we employed quantitative proteomics to profile senescent human skin BJ fibroblasts induced by ionizing irradiation. Our profiling has revealed novel biomarkers such as pregnancy zone protein (PZP), alpha‐glucosidase (GAA), as well as several key cellular features including declined glucose metabolism, reduced ATP production, and decreased chaperone levels which are critical for the survival and maintenance of normal senescent cells. In addition, we also demonstrated that co‐inhibition of TCA cycle and Hsp90 activity may provide a better strategy for selective elimination of the senescent cells, particularly, the therapy‐induced senescent cancer cells.
Results
2
The Proteomic Landscape of IR‐Induced Senescent Fibroblasts
2.1
To identify proteomic features of DNA damage induced cellular senescence, we performed a global quantitative profiling of IR‐induced senescent human skin fibroblast BJ cells, a well‐established cellular model for cellular senescence. BJ cells were cultured in SILAC medium, the heavy labeled cells were irradiated with 10 Gy of gamma rays (Cobalt 60 source). After 30 days of culture, the senescence state of the IR treated cells was confirmed by flattened, enlarged and nondividing morphology and 100% positive staining of beta‐galactosidase, as well as the elevated cyclin dependent kinase inhibitor p21 (Figure S1). The senescent cells were harvested and lysed, equal numbers of growing cells (light labeled) were also lysed as control. Both lysates were mixed at a 1:1 ratio and subjected to tryptic peptide preparation and LC–MS/MS analysis (Figure 1A) (Hu et al. 2023). Upon stringent filtering of the resulting quantitative dataset, a total of 178 cellular proteins with at least a 4‐fold of change in abundance in the IR induced senescent BJ cells were identified (for more detailed information see Table S1). Among these candidate proteins, 31 out of 178 proteins are increased in abundance by 4‐fold or greater (Figure 1B), including some well‐known senescent markers such as GLB1 (beta‐galactosidase) (Dimri et al. 1995), SERPINB2/PAI‐2 (serine proteinase inhibitor B2) (Hsieh et al. 2017), supporting the validity of our profiling approach. Interestingly, the abundance of GAA, the alpha‐glucosidase for degradation of glycogen into glucose, is also significantly increased at senescence. Similarly to GLB1, GAA also localizes to lysosomes, hinting GAA might be another robust biomarker for senescence detection. 147 out of 178 proteins are decreased by 4‐fold or greater in the IR induced senescent BJ cells, including some known senescent markers such as LMNA (Lamin A/C) (Ragnauth et al. 2010) and LMNB1 (Lamin B1) (Freund et al. 2012). Moreover, we failed to detect the presence of some prominent senescence biomarkers such as p21, p16, IL6, and IL8, indicating the abundance of those factors might be too low or the stoichiometry of the candidate protein is unbalanced to be detected by our approach in senescent fibroblasts. Therefore, the 178 proteins identified from our profiling should best represent the proteomic landscape of DNA damage induced senescent fibroblasts.
The proteomic features of IR‐induced senescent fibroblasts. (A) Workflow of quantitative proteomic profiling of the senescent BJ fibroblasts induced by ionizing radiation. BJ cells were cultured in SILAC medium for metabolic labeling. Heavy isotope‐labeled cells were exposed to 10 Gy IR and recovered for 30 days prior to the preparation of cell lysates and tryptic peptides. (B) The volcano plot illustrates the 178 candidate proteins based on their abundance changes ranked by the binary logarithm values of fold change and their statistical significance (−log10 p value). The top six of the most up‐regulated or the most down‐regulated proteins are displayed in green or blue. The three senolytic targets HSP90AA1, GLS1, and PDHB revealed and validated in our study are highlighted in red. (C) The biological processes involved in secretion, chaperones, glucose metabolism, fatty acid synthesis, cell cycle control, DNA repair as well as apoptosis are enriched in our data set by Gene ontology (GO) analysis. Each displayed category consists of at least two candidate proteins. (D–K) Network analysis by String and GeneMANIA. The identified interacting modules include cell cycle control (D), TCP‐1 ring complex (TRiC) of chaperones (E), non‐homologous end joining (NHEJ) DNA repair machinery (F), apoptosis regulation (G), protease network (H), protease inhibitor network (I), growth factor network (J), and cytokine network (K). Each functional module is comprised of at least two candidate proteins. Blue color filled circles represent proteins identified in our screen, light color filled circles stand for other proteins known to interact with our candidates.
Most strikingly, based on Gene Ontology (GO) biological process analysis, a large proportion of identified proteins (37 out of the 178, 21%) were still assigned into the SASP category (Figure 1C and Table S1), the largest and extensively explored hallmark of senescent cells, despite the fact that our proteomic profiling was performed by using the whole lysates of senescent cells, not the conditioned media of culture. There were two elegant SASP profilings performed previously; one was conducted by using quantitative secretory proteomics (Basisty et al. 2020) and the other one was done by antibody array covering 120 selected SASP components (Coppé et al. 2008). We compared our results with the two published SASP profiles and found only a few SASP components such as IGFBP5, IGFBP7, LaminB1, SerpinH1, members of the Hsp70 family, and Hsp90 family that were shared among these studies. By contrast, a lot of SASP components including GDF15, STC1, IL6, IL8, and MMPs identified in the two published profiles are missing from our dataset. In addition, some SASP components such as IL14, IL25 and A2M only appeared in our profiling. The shared and distinct composition of SASP is strongly in support of the fact that the SASP is highly complex, heterogeneous, and dynamic, which is largely determined by the nature of senescence inducers (e.g., IR vs. Ras activation) and the type of senescent cells (e.g., fibroblasts vs. epithelial cells). Worth pointing out, our SASP candidates were with a 4‐fold or greater of change in abundance, which is much higher than the cutting‐off (a 1.5‐fold change in abundance) used in Basisty's report. That difference in data filtering may further enhance the discrepancies in the composition of SASP.
Besides the SASP category, we also discovered some other prominent categories, including ER‐Golgi related (22 out of 178, 12%), chaperones (22 out of 178, 12%), mitochondrial metabolism (14 out of 178, 8%), glycolysis (9 out of 178, 5%) and DNA repair (5 out of 178, 3%) (Figure 1C). These categories were significantly enriched in our dataset based on the p values determined by Fisher's exact test for each candidate protein (see Table S1). GeneMANIA networking combined with String interaction analysis revealed that most of the identified proteins can be clustered into some functional modules with several known interacting partners. These functional modules include cell cycle regulation (Figure 1D), TRiC chaperone complex (Figure 1E), non‐homologous end joining (NHEJ) repair machinery (Figure 1F) and apoptosis regulators (Figure 1G). Not surprisingly, a significant proportion of interacting modules fall into canonical SASP, such as proteinases, proteinase inhibitors, growth factors, and cytokines and chemokines (Figures 1H–K). Taken together, these enriched biological process categories and functional interacting modules represent the cellular signature of senescent fibroblasts induced by DNA damage.
Diverse Protein Responses to Senescence and Their Distinct Correlation Patterns With Transcription Alterations
2.2
Regarding the changes in protein abundance, we found proteins belonging to the same family or involved in the same biological process usually exhibit a similar pattern of change in abundance at senescence. For example, we observed the increased levels of enzymes involved in glucoside degradation including GLB1 and GAA (Figure 2A). We also found the decreased levels of non‐homologous end joining (NHEJ) repair members ERCC5 and PRKDC (Figure 2B). On the contrary, upon close examination of some candidates supposed to be the components of SASP, a distinct response pattern was revealed despite that they belong to the same protein family. For instance, the cellular abundance of IGFBP5 (insulin‐like growth factor‐binding protein 5) is up‐regulated by more than 4‐fold, whereas the cellular abundance of IGFBP7 (insulin‐like growth factor‐binding protein 7) is down‐regulated by more than 4‐fold in the senescent BJ cells (Figure 2C). Another case is the serine proteinase inhibitor family, whereas SERPINB2 and SERPINB6 both are accumulated in senescent BJ cells by more than 4‐fold; SERPINH1, pan‐proteinase inhibitors A2M and PZP (pregnancy zone protein) are all decreased by more than 4‐fold (Figure 2D). It is not clear why such distinct response patterns occurred in the same protein family. These observations may indicate distinct functions exist in some protein families.
Protein abundance changes in response to senescence and the correlation with their transcriptional alterations. (A–D) shown the protein response patterns based on protein families. (A) The abundance of both GLB1 and GAA involving in glycoside degradation is increased at senescence. (B) The abundance of NHEJ repair components XRCC5 and PRKDC is decreased at senescence. (C) shown the IGF regulation family with increased IGFBP5 and decreased IGFBP7. (D) depicted the proteinase inhibitor family with increased SERPINB2, SERPINB6, and decreased A2M, PZP, and SERPINH1. (E–G) Three correlation patterns were identified between protein changes in abundance and its transcription alterations. The up‐up positive correlation is represented by the increased transcription of the SERPINB2 gene with the increased level of SERPINB2 protein (E). The down‐down positive correlation is represented by the decreased transcription of the HSP90AA1 gene with the decreased level of Hsp90 protein (F). The reverse correlation is depicted by the increased transcription of PZP gene and the decreased protein level of PZP (G).
Transcription reprogramming is a well‐known alteration upon senescence. Next, we examined the correlation between the changes of protein abundance and their transcription alterations in senescent cells. Two distinct patterns including the positive correlation and the negative correlation were revealed. For example, the up‐regulated transcription of the SerpinB2 gene with concomitant increased SerpinB2 protein was reported in senescent cells previously (Figure 2E) (Hsieh et al. 2017; Zhu et al. 2015). We also observed the down‐regulated HSP90AA1 gene transcription followed by the decreased cellular Hsp90 protein in the senescent BJ fibroblasts (Figure 2F). We designated this kind of response relationship as positive correlations comprising the up‐up and down‐down two subtypes (Figures 2E,F). Furthermore, the transcription and cellular protein correlation of pregnancy zone protein (PZP) was found to be on the other way around (Figure 2G). Our previous study discovered the cellular level of PZP was decreased significantly whereas its transcription was increased at senescence (Hu et al. 2023). This relationship was considered as a reverse correlation. Given the fact that the PZP is an extracellular enzyme, the protein secretory process should account largely for the reverse correlation occurred to PZP. Similarly to PZP, the IGFBP7 also exhibited the reverse correlation at senescence previously (Wajapeyee et al. 2008).
Significantly Down‐Regulated Enzymes in Glycolysis, TCA Cycle, Glutaminolysis, and Fatty Acid Synthesis at Senescence
2.3
Besides the large number of SASP components identified from our profiling, we found glycolytic enzymes including PFKP, ALDOA, TPI1, PKM1/2, and LDHA/B are down‐regulated by more than 4‐fold (Figures 3A,D and Table S1). Most strikingly, both PFKP and PKM1/2 are rate‐limiting enzymes for glycolysis. Western blotting using antibodies against PFKP, PKM2, and ALDOA confirmed their protein levels were decreased in BJ and IMR‐90 senescent cells triggered by IR (Figures 3E,F). In addition, the protein abundance of PDHB, the TCA cycle rate‐limiting enzyme pyruvate dehydrogenase E1 enzyme subunit beta (PDHB), and GLS1, the glutaminase converting glutamine into α‐ketoglutarate during glutaminolysis, also declined in senescent fibroblasts (Figures 3B,D,I). Western blotting confirmed the protein levels of pyruvate dehydrogenase E1 enzyme catalytic subunit alpha (PDHA) and GLS1 dropped in senescent fibroblasts (Figures 3E,F). Besides the enzymes in glycolysis, TCA cycle, and glutaminolysis, enzymes responsible for fatty acid synthesis are also decreased in senescent BJ fibroblasts, evidenced by the more than 4‐fold decrease of ATP citrate lyase (ACLY) and acyl‐CoA thioesterase 7 (ACOT7) (Figures 3C,D). ACLY is essential for the initial step of conversion of glucose into lipid acid, and ACOT7 is crucial for the synthesis of long chain fatty acid. In agreement with our findings, a previous study reported the transcription of TCA cycle genes and fatty acid synthesis genes is reduced in the livers of old humans and aged mice (Zhang et al. 2024).
Enzymes involved in glucose metabolism, glutaminolysis, and fatty acid synthesis are decreased in senescent fibroblasts. (A) The glycolysis pathway and its connections to pentose phosphate pathway (PPP), TCA cycle and glutaminolysis, as well as fatty acid synthesis. Enzymes identified in our screen are depicted in blue ovals. ALDOA: fructose‐bisphosphate aldolase A; G6PD: glucose‐6‐phosphate dehydrogenase; LDHA/B: lactate dehydrogenase A/B; PFKP: 6‐phosphofructokinase platelet type; PKM2: pyruvate kinase 2; TPI1: triosephosphate isomerase isoform 1. (B) Tricarboxylic acid cycle (TCAC) and glutaminolysis in mitochondria. GLS1, glutaminase 1; PDHB, pyruvate dehydrogenase E1 enzyme subunit B. (C) The fatty acid synthesis pathway in cytoplasm and its connection to TCA cycle. Mitochondrial citrates are transported to cytoplasm for fatty acid synthesis. ACOT7, acyl‐CoA thioesterase 7; ACYL, ATP‐citrate lyase. (D) Quantitative summary of the abundance changes of the identified metabolic enzymes in senescent BJ fibroblasts. The abundance change of each enzyme was calculated as the binary logarithm value of the peptide signal in senescent cells versus the peptide signal in control cells. Some peptides identified by more than twice were labeled with error bars. (E, F) The decreased levels of glycolytic enzymes including PFKP, ALDOA, PKM2, TCA cycle rate‐limiting enzyme PDHA and glutaminolysis enzyme GLS1 were confirmed by Western blotting in both senescent BJ and IMR‐90 cells. (G, H) RT‐PCR examinations also confirmed the transcriptional levels of these genes were also decreased. The total protein or RNA was extracted from senescent BJ or IMR‐90 fibroblasts 15 days post 10 Gy irradiation. The products of RT‐PCR were semi‐quantified on agarose gels. β‐Actin was as internal control. (I) The quantitative proteomic data of PDHB and GLS1. GLS1, mitochondrial glutaminase kidney isoform; H, heavy isotope labeling; L, light isotope labeling; ND, not detectable; PDHB, pyruvate dehydrogenase E1 beta subunit; PPM, mass accuracy of the measured peptide; SNR, signal to noise ratio; VQ ConfScore, the quantification quality of each identified peptide pair; XCorr, correlation of the similarity between the measured and the calculated peptides.
To reveal the transcriptional status of the genes encoding the above enzymes, semiquantitative RT‐PCR was performed to examine the transcription of PFKP, PKM2, PDHA, and GLS1 genes at senescence (see Table S2 for primers used in RT‐PCR). Our results clearly demonstrated the transcription levels of these genes were also decreased in both human BJ and IMR‐90 senescent fibroblasts compared to their proliferating counterparts (Figures 3G,H). Taken together, our findings suggest that declined glycolysis, decreased TCA cycle, and low glutaminolysis are the most striking metabolic features of DNA damage induced normal senescent cells; transcription reprogramming might be an important mechanism underlying these metabolic changes upon senescence.
Low Glucose Consumption, Reduced ATP and Alpha‐KG Production and Decreased Cytosolic NAD+/NADH Ratio in Senescent Fibroblasts
2.4
Low levels of enzymes in glycolysis and TCA cycle may hint low consumption of glucose in senescent fibroblasts. Indeed, by measuring the glucose concentration in culture media of senescent fibroblasts, the glucose consumption rate is much lower in the senescent BJ, WI‐38, and IMR‐90 cells compared to their proliferating counterparts (p < 0.001 by Student's t‐test) (Figure 4A). Through glycolysis, TCA cycle and the electron transport chain (ETC), the metabolism of glucose eventually ends up with the production of ATP, the essential energy source for all cellular processes. Next, we measured the ATP levels in senescent BJ, WI‐38, and IMR‐90 cells triggered by IR. We observed the ATP levels decreased by more than 30% in these senescent cells (p < 0.05 by Student's t‐test, Figure 4B). Moreover, the decreased levels of energy metabolism regulator DLD and mitochondrial electron transfer flavoprotein subunit beta ETFB identified in our profiling (for details see Table S1) further support our finding of low ATP production in the senescent fibroblasts. As TCA cycle also produces a lot of metabolic intermediates, which are critical for the biosynthesis of amino acids and nucleotides, etc., the level of alpha‐ketoglutarate (alpha‐KG), one of the prominent TCA intermediates, was measured. The alpha‐KG levels in senescent BJ, WI‐38, and IMR‐90 cells dropped by approximately 60% compared to their proliferating counterparts (Figure 4C). Since the cytosolic NAD+/NADH ratios correlate with the oxygen consumption and the activities of glycolysis and TCA cycle, the cytosolic NAD+/NADH ratios in senescent BJ, WI‐38 and IMR‐90 cells triggered by IR were also measured. We found the NAD+/NADH ratios decreased significantly compared to their proliferating counterparts (Figure 4D). In proliferating cells, the cytosolic NAD+/NADH ratio is relatively stable. Why are the cytosolic NAD+/NADH ratios decreased in senescent cells? A close examination of our proteomic data may shed some light onto this question. Upon senescence, the consumption of NAD+ is increased. For example, we found that aldehyde dehydrogenase 1 family member A1 (ALDH1A1) is one of the most highly up‐regulated enzymes in senescent BJ fibroblasts. Cytosolic ALDH1A1 catalyzes the oxidation of aldehydes such as acetaldehydes and retinaldehydes. As a result of such oxidative reactions, a large amount of cytosolic NAD+ is converted into NADH (Bonnay et al. 2020). On the other hand, previous studies found the de novo synthesis of NAD+ from tryptophan and nicotinamide salvage also decreases with aging (Camacho‐Pereira et al. 2016; Covarrubias et al. 2021; Soma and Lalam 2022). Additionally, we found low levels of LDHA/B from our data indicate less conversion of pyruvate into lactate and thereby less production of NAD+. Collectively, we conclude the increased consumption and decreased production of NAD+ lead to the dropped cytosolic NAD+/NADH ratio in senescent fibroblasts (Figure 4D). Most strikingly, exposure of PDH inhibitor CPI‐613 and GLS1 inhibitor BPTES to the three senescent fibroblast cell lines could lead to a further decrease of the cytosolic NAD+/NADH ratios (Figure 4E), hinting suppression of TCA cycle enhanced the energy shortage and oxidative stress in senescent cells. Furthermore, the levels of alpha‐KG and ATP in the CPI‐613 plus BPTES treated senescent cells were also reduced by 40%–60% compared to their untreated controls (Figures 4F,G), confirming that the lack of intermediate metabolites and ATP might be a primary cause to trigger cell death in senescent cells. Taken together, the low levels of enzymes involving in glycolysis, TCA cycle and ETC revealed in our proteomic profiling may largely account for the low glucose consumption, the decreased ATP and alpha‐KG production and the declined NAD+/NADH ratio in the senescent fibroblasts induced by IR.
*Glucose consumption, ATP and alpha‐KG production, and NAD+/NADH ratio are reduced in senescent fibroblasts, inhibiting TCA cycle and glutaminolysis lead to the selective killing of senescent fibroblasts. (A) The glucose consumption rate is much lower in senescent fibroblasts compared to their proliferating counterparts. The media were collected from 1 × 106 cells cultured for 48 h, glucose concentration was measured by the enzyme mediated colorimetric glucose assay kit and normalized by the fresh medium and cell numbers at 48 h of culture. The consumption rate difference between growing and senescent cells was compared by Student's t‐test, *** depicts p < 0.001. (B) ATP production was reduced in senescent BJ, WI‐38, and IMR‐90 cells compared to their proliferating counterparts. ATP concentration was measured by ATP luminometric assays, presented as means ± SD from three independent experiments. *p < 0.05 by Student's t‐test. (C) Alpha‐KG level was reduced in senescent BJ, WI‐38, and IMR‐90 cells compared to their proliferating counterparts. Alpha‐KG was measured by α‐Ketoglutarate Assay Kit, presented as means ± SD from three independent experiments. ***p < 0.001 by Student's t‐test. (D) The cytosolic NAD+/NADH ratios in senescent BJ, WI‐38, and IMR‐90 cells are decreased compared to their proliferating counterparts. *p < 0.05 or **p < 0.01 by Student's t‐test. (E) CPI‐613+BPTES combination treatment enhanced the decrease of cytosolic NAD+/NADH ratios in senescent BJ, WI‐38, and IMR‐90 cells compared to their untreated counterparts. The measurement was done at 48 h post treatment. *p < 0.05 by Student's t‐test. (F) CPI‐613+BPTES combination treatment enhanced the decrease of alpha‐KG levels in senescent BJ, WI‐38, and IMR‐90 cells compared to their untreated counterparts. *p < 0.05 or **p < 0.01 by Student's t‐test. The measurement was performed at 48 h after treatment. (G) CPI‐613+BPTES combination treatment enhanced the decrease of ATP production in senescent BJ, WI‐38, and IMR‐90 cells compared to their untreated counterparts. *p < 0.05 or **p < 0.01 by Student's t‐test. The measurement was performed at 48 h after treatment. (H) Inhibition of glycolysis by 2‐DG treatment led to decreased cell survival in both proliferating and senescent BJ cells in a dose dependent manner. Cell viability at 72 h of treatment was determined by the CCK8 assay and normalized by the untreated cells. (I) The dose–response curves of CPI‐613 on proliferating and senescent BJ cells. *p < 0.05 by one‐way ANOVA. (J) The dose–response curves of BPTES on proliferating and senescent BJ cells. (K, L) Co‐inhibition of the PDH complex and glutaminolysis by CPI‐613+BPTES enhanced the selective killing of senescent BJ cells. *p < 0.01 at 72 h of treatment tested by Student's t‐test. (M) The morphological changes of growing and senescent BJ cells at the indicated time of treatment under the light microscopy. Senescent BJ cells before CPI‐613+ BPTES treatment were stained with SA‐β‐gal. Cells were imaged at magnification 200×.
Low Levels of Glycolysis, TCA Cycle, and Glutaminolysis Are Essential for the Survival and Maintenance of Senescent Fibroblasts
2.5
Next, we would like to test whether the IR‐induced senescent BJ cells are susceptible to energy depletion treatment, and whether inhibition of glycolysis, TCA cycle, and glutaminolysis could lead to selective killing of senescent cells. Senescent BJ cells triggered by IR were exposed to 2‐deoxy‐glucose (2‐DG, inhibitor of hexokinase 2 in glycolysis) and CPI‐613 (inhibitor of pyruvate dehydrogenase and alpha‐KG dehydrogenase in TCA cycle), respectively. While 2‐DG conferred high cytotoxicity towards the senescent BJ, it failed to show selectivity towards the proliferating BJ cells (Figure 4H). In contrast, a low dose of CPI‐613 alone exhibited higher selectivity towards the senescent BJ cells compared to the proliferating BJ cells (p < 0.05, Figure 4I). Given the fact that glutaminolysis is a major source to replenish TCA cycle intermediate alpha‐KG and blocking glutaminolysis by the GLS1 inhibitor BPTES gives rise to cell death in senescent BJ cells (Figure 4J), we tested the effect of CPI‐613 in combination with the BPTES on senescent BJ cells. Indeed, this combination treatment led to enhanced cytotoxicity towards the senescent BJ cells and much less cytotoxicity to the proliferating BJ cells (Figures 4K–M). Collectively, our data has clearly demonstrated that the low activities of glycolysis, TCA cycle, and glutaminolysis existing in senescent fibroblasts are essential for the survival and maintenance of senescent fibroblasts.
Declined Chaperones Are a Unique Feature of Senescent Fibroblasts and Required for Senescence Survival
2.6
A large number of chaperones with a more than 4‐fold decrease in abundance in senescent BJ cells have also been recovered from our proteomic profiling. These ATP‐dependent chaperones include the chaperonin containing TCP‐1 (TRiC) family (TCP1, CCT2, CCT4, CCT5, CCT7, and CCT8) (Figure 5A), the Hsp70 family (HSPA4, HSPA5, HSPA8, and HSPA9) (Figure 5B), the Hsp90 family (HSP90AA1, HSP90AA2, HSP90AB1, HSP90B1, and Hsp90 associated co‐chaperone CDC37) (Figure 5C), as well as many other chaperone members (Figure 5D). Similarly to the enzymes involved in glucose metabolism, the protein abundance and transcriptional levels of TCP1, HSPA8 and HSPA90AA1 were confirmed to be down‐regulated in senescent BJ and IMR‐90 cells by Western blotting and RT‐PCR (Figures 5E,F). Since HSP90 inhibitor 17‐AAG has been identified as a new class of senolytic agents in a previous study using senescent Ercc1−/− murine embryonic fibroblasts (Fuhrmann‐Stroissnigg et al. 2017), the effects of HSP90 inhibitor 17‐AAG on IR‐induced senescent IMR‐90 and BJ cells were examined. As shown in Figures 5G–I, the decreased survival rate was observed in 17‐AAG treated senescent IMR‐90 and BJ cells in a dose and time dependent manner (p < 0.01 or 0.001 at 80 nM or 120 nM, 72 h treatment), whereas no obvious cytotoxicity of 17‐AAG towards proliferating IMR‐90 and BJ cells was observed. These results confirmed HSP90 inhibitor 17‐AAG can selectively kill senescent cells and implicated low level of chaperones is required for the survival of senescent cells, similarly to the down‐regulated TCA cycle and glutaminolysis at senescence.
*The abundance of chaperones and their transcription are decreased in senescent fibroblasts. Inhibiting Hsp90 by 17‐AAG caused the selective killing of senescent fibroblasts. (A–D) The summarization of protein changes in abundance of TRiC chaperone, Hsp70 family, Hsp90 family, and other chaperones in senescent BJ cells identified by our proteomic profiling. (E) Western blotting confirmed the protein levels of TCP1, Hsp70, and Hsp90 were decreased in senescent BJ and IMR‐90 cells. β‐Actin served as internal control. (F) RT‐PCR confirmed the transcription levels of TCP1, Hsp70, and Hsp90 genes were also decreased in senescent BJ and IMR‐90 cells. The RT‐PCR product of β‐Actin was as internal control. (G) The morphological changes of growing and senescent IMR‐90 cells at the indicated time of 17‐AAG treatment under the light microscopy. The senescent IMR‐90 cells without 17‐AAG treatment were stained with SA‐β‐gal. Cells were imaged at magnification 200×. (H) Inhibiting Hsp90 by 17‐AAG led to the selective killing of senescent IMR‐90 cells in a dose dependent manner. **p < 0.01 by one‐way ANOVA. (I) The dose–response curves of 17‐AAG on proliferating and senescent BJ cells. **p < 0.001 by one‐way ANOVA.
Co‐Targeting TCA Cycle and Chaperones Leads to Enhanced Senolytic Effects on Senescent Fibroblasts and the Senescent Tumor Cells
2.7
The combination therapy often increases drug efficacy, reduces drug resistance and side effects due to the lower dosage in use. Next, we wondered whether these senescent features identified from our profiling could be exploited for selective senolysis. Towards that end, we would like to test the combination of Hsp90 inhibitor with PDH and GLS1 inhibitors simultaneously in senescent BJ cells triggered by IR. Based on the dose–response results of each compound (Figures 4I,J and 5I), the half of each compound's IC_50_ dose in senescent BJ cells was used for the combination treatments, because the half IC_50_ dose not only exerts obvious killing effects on senescent cells, but also has minimal or tolerable cytotoxicity towards the proliferating cells. The half IC_50_ dose of CPI‐613 used alone or in combination was exposed to senescent BJ cells for 72 h and cell viability measurements were performed. While the cell survival rates in the treatment of CPI‐613 alone (50 μM) were 76.38% ± 12.94% (Figure 4I) and 57.07% ± 11.34% in the treatment of CPI‐613 (50 μM) + BPTES (10 μM) (Figure 4L), it dropped profoundly to 42.81% ± 2.77% in the combination treatment of 17‐AAG (60 nM) + CPI‐613 (50 μM) + BPTES (10 μM) (Figures 6A,B) at 72 h after treatment, indicating co‐targeting Hsp90, PDH, and GLS1 could lead to a higher selectivity towards senescent cells. Therapy‐induced senescent (TIS) tumor cells is a common phenomenon during chemotherapy or radiotherapy for cancer treatment; TIS tumor cells may cause drug resistance and even tumor relapses in cancer patients. Thus, we also test whether co‐administration of HSP90, PDH, and GLS1 inhibitors could lead to a selective elimination of TIS tumor cells. TIS lung adenocarcinoma A549 cells and TIS cervical carcinoma HeLa cells were established by doxorubicin induction and confirmed by beta‐Gal staining as well as the elevated p21 protein levels (Figure S1). Both proliferating cancer cells and the TIS cancer cells were exposed to 17‐AAG (60 nM) + CPI‐613 (50 μM) + BPTES (10 μM) for 72 h. The cell survival rates were 26.50% ± 8.59% in TIS A549 and 22.30% ± 9.85% in TIS HeLa, with little cytotoxicity to the growing A549 and HeLa cells (Figures 6C–F). Most intriguingly, the killing efficacy in these TIS tumor cells was even higher than that in the senescent BJ fibroblasts (Figures 6B,D,F,G). Thus, co‐inhibiting Hsp90 and TCA cycle with 17‐AAG + CPI‐613 + BPTES combination (Figures 6B,D,F) leads to the enhanced selective killing of the senescent fibroblasts, particularly, the therapy‐induced senescent cancer cells, hinting that pyruvate dehydrogenase could serve as a novel therapeutic target for the development of senolytics (Figure 6H).
*Inhibiting PDH, GLS1 and Hsp90 by the combination of CPI‐613+BPTES+17‐AAG gave rise to enhanced senolysis on senescent fibroblasts as well as the therapy‐induced senescent tumor cells. (A, B) The effects of CPI‐613+BPTES+17‐AAG combination treatment on proliferating (A) and senescent (B) BJ cells. For the dose of each compound in use, see the results 2.7 section for more details. **p < 0.01 by Student's t‐test. (C, D) The effects of CPI‐613+BPTES+17‐AAG combination treatment on proliferating and Dox‐induced senescent lung adenocarcinoma A549 cells. ***p < 0.001 by Student's t‐test. (E, F) The effects of CPI‐613+BPTES+17‐AAG combination treatment on proliferating and Dox‐induced senescent cervical carcinoma HeLa cells. **p < 0.001 by Student's t‐test. (G) The morphological changes of senescent BJ induced by IR, senescent A549 and HeLa cells induced by Dox at the indicated time of CPI‐613+BPTES+17‐AAG treatment under the light microscopy. The senescent cells without treatment were stained with SA‐β‐gal. Cells were imaged at magnification 200×. (H) The schematic summarization of our findings. The activities of TCA cycle and chaperones are reduced in DNA damage induced senescent cells. Co‐inhibiting Hsp90 and TCA cycle with 17‐AAG+CPI‐613+BPTES combination leads to enhanced selective elimination of senescent cells, hinting TCA cycle and glutaminolysis are novel and potent targets for senolysis.
The Metabolism in Therapy‐Induced Senescent Tumor Cells Is Much More Active Than That in the Senescent Normal Cells
2.8
Why does the 17‐AAG + CPI‐613 + BPTES combination treatment lead to higher efficacy in TIS tumor cells than that in senescent fibroblasts? To answer that question, we examined the levels of the three senolytic targeting proteins Hsp90, PDHA, and GLS1 in TIS A549, TIS HeLa, and TIS U2OS cells induced by Dox treatment. In contrast to that significantly decreased levels of Hsp90, PDHA, and GLS1 in senescent BJ and IMR‐90 fibroblasts (Figures 7A,B and Figure S2), the levels of Hsp90, PDHA, and GLS1 in TIS tumor cells remained unchanged or even elevated at senescence compared to their growing counterparts (Figures 7C,D and Figure S2). Moreover, their transcriptional alterations in TIS cells were nearly the same as their protein abundance (Figure S3 and for the original Western blots and RT‐PCR images see Figures S4 and S5). The distinct levels of drug targets may explain in part why our combination treatment is more effective in TIS tumor cells compared to senescent fibroblasts. A previous study reporting cyclophosphamide (CTX) induced TIS lymphomas exhibited elevated oxygen consumption rate as well as higher ATP production (Dörr et al. 2013) strongly supports our finding that the metabolism in therapy‐induced senescent tumor cells is much more active than that in senescent normal cells.
*Distinct proteomic and transcriptional signatures of metabolic enzymes and chaperones between senescent fibroblasts and the therapy‐induced senescent tumor cells. (A) The abundance of glycolysis‐related enzymes PFKP, ALDOA, PKM2, TCA cycle‐related PDHA, and glutaminolysis‐related GLS1 was decreased significantly in senescent BJ and IMR‐90 cells compared to their proliferating counterparts. (B) The protein levels of chaperones TCP1, Hsp70, and Hsp90 were decreased significantly in senescent BJ and IMR‐90 cells. (C) The abundance of those glycolysis‐related enzymes remained unchanged or even elevated in Dox‐induced senescent A549, HeLa, and U2OS tumor cells compared to their proliferating counterparts. (D) The abundance of these chaperone proteins in Dox‐induced senescent A549, HeLa, and U2OS tumor cells remained nearly unchanged. The relative abundance of each protein was quantified by signal density scanning on Western blots and normalized to the signal of β‐Actin or β‐tubulin. *p < 0.05, *p < 0.01 tested by Student's t‐test.
Co‐Inhibition of TCA Cycle and Hsp90 Reduced Senescent Cells and Alleviated the Physical Dysfunctions in Aged Mice
2.9
Persistent senescence contributes to aging, and the elimination of senescent cells has been shown to delay the aging process (Baker et al. 2016, 2011). Even a small number of senescent cells demonstrated by a previous study were sufficient to induce physical dysfunctions in young mice, and clearance of such senescent cells by the senolytic cocktail dasatinib plus quercetin led to the alleviation of such physical dysfunctions (Xu et al. 2018). Thus, we would like to examine the effect of CPI‐613+BPTES+17‐AAG on the physical dysfunctions in aged mice. D‐galactose induced aged mice were treated with the CPI‐613+BPTES+17‐AAG combination or vehicle intermittently for 1 month (Figure 8A, for the dose of each compound in use, see the methods section for more details). After the combination treatment, the dull and shaggy coat hair and wrinkle formation were ameliorated in the treated mice compared to the vehicle controls (Figure 8B). Then, we checked the expression of p21, a key senescence marker in liver, kidney, and lung tissues from the aged mice, and found the p21 positive rate was much lower in the treated mice compared to the vehicle control (Figures 8C,D), indicating the selective elimination of senescent cells by the CPI‐613+BPTES+17‐AAG combination treatment. In addition, the size and weight of spleens dissected from the two groups were compared (Figures 8E–G). Next, we also looked at the effects of the CPI‐613+BPTES+17‐AAG combination treatment on the physical dysfunctions in aged mice. While there is no significant difference in body weight and food intake between the two groups (p > 0.05 by two‐sided Welch's t‐test, Figures 8H,I), the grip strength (Figure 8J), staying time on the rotarod (Figure 8K), running distance on the treadmill (Figure 8L), and treadmill endurance (Figure 8M) were significantly increased in aged mice treated with CPI‐613 + BPTES+ 17‐AAG (p < 0.05 or 0.01 by two‐sided Welch's t‐test). Our observations are well consistent with the effects of the D+Q cocktail on aged mice reported previously (Xu et al. 2018).
CPI‐613+BPTES+17‐AAG combination treatment reduced the p21 positive senescent cells and alleviated the physical dysfunctions in aged mice. (A) The experimental design for generating aged mice induced by D‐galactose via intraperitoneal injection and the combination treatment in the aged mice. For the dose of each compound in use, see the methods section for more details. (B) Representative images of aging mice on day 5 after the last combination treatment, the dull and shaggy coat hair and wrinkle formation observed in the aged mice were ameliorated in the treated mice. (C, D) Representative images of p21 stained liver, kidney, and lung tissues from aged mice with or without combination treatment and the quantification of p21 positive cells in liver, kidney, and lung tissues. The red arrows indicate the p21 positive cells, the scale bar stands for 30 μm. Data were presented as means ± SD, p value was calculated by two‐tailed Student's t‐test. (E–G) Spleens dissected from 7 mice with or without combination treatment were displayed. The surface area and weight of spleens from each mouse group were measured and plotted in F‐G. The surface area of each spleen was calculated using the following formula: S = 5 × (0.524 × L × W × T)2/3. (H–M) Physical function measurements in aged mice with or without treatment with CPI‐613+BPTES+17‐AAG combination. Changes in body weight (H), food intake (I), grip strength (J), time on the rotarod (K), running distance on the treadmill (L), and treadmill endurance (M) were plotted. Results are shown as box‐and‐whisker plots with the median shown as a line in the middle, whiskers indicate the smallest and largest values. n = 7, 4 female and 3 male. g, gram; KJ, kilojoule; m, meter; N, Newton; s, second. p values were calculated by the two‐sided Welch's t‐test and displayed on each plot.
For clinical applications, in vivo safety and long‐term toxicity of potential therapeutics are two major concerns. Thus, the toxicity of our combination treatment in the treated mice was examined. We found there is no observable difference in the tissue structures of major organs including liver, kidney, and lung (Figure 8C), and no significant difference in the size and weight of spleens (Figures 8E–G), or in the body weight and food intake between the treated and untreated mouse groups (p > 0.05 by two‐sided Welch's t‐test, Figures 8H,I), indicating the combination treatment at the doses we used had minimal or tolerable toxicity to the treated mice.
Taken together, we demonstrated CPI‐613+BPTES+17‐AAG cocktail indeed can alleviate the aging phenotypes by reducing the number of senescent cells in the aged mice, and targeting TCA cycle could be a novel therapeutic approach for the development of novel senolytics against aging.
Discussion
3
Down‐Regulation of Glucose Metabolism Is a Unique Feature of Senescent Normal Cells
3.1
Fibroblasts are one of the most common cell types in connective tissues and are widely distributed in tissues such as skin, bones, muscles, internal organs, and blood vessels. They are critical in maintaining structural integrity, promoting wound healing, and regulating immune responses. Senescent fibroblasts not only play an important role in the aging of tissues and organs, but also are a well‐established cellular senescence model for aging research (Campisi and Robert 2014; Dimri et al. 1995; Kang et al. 2015; Krtolica et al. 2001). Revealing the unique proteomic features of senescent fibroblasts would be helpful for understanding the survival and maintenance mechanisms of senescence in normal cells. In the current study, we employed a quantitative proteomic approach to investigate the cellular landscape of senescent BJ fibroblasts induced by IR. In contrast to the general understanding that senescent cells are much more metabolically active, our profiling clearly demonstrated that the metabolic levels in the IR‐induced senescent BJ fibroblasts are much lower compared to their proliferating counterparts. Given the facts that many cellular processes such as cell cycle progression, DNA replication, and chromatin remodeling are stalled at senescence, there is no need for senescent cells to maintain high levels of glucose metabolism. Therefore, the glucose consumption and ATP production are also decreased in senescent fibroblasts. Most intriguingly, we found the metabolic signature in normal senescent cells is quite different from the therapy‐induced senescent tumor cells (Figures 3 and 7). Why do such distinct metabolic phenotypes exist between normal senescent fibroblasts and senescent cancer cells? First, the abundance of enzymes involved in glucose metabolism remains either unchanged or even elevated in doxorubicin‐induced senescent lung carcinoma A549, cervical carcinoma HeLa and osteosarcoma U2OS cells, indicating high levels of glucose metabolism exist in the therapy‐induced senescent cancer cells. Second, the senescent cancer cells rather than the senescent normal cells may rely more on mitochondrial respiration to sustain ATP production (Wolf 2014). Third, the energy compensation mechanism such as the hyper‐activation of the energy sensor AMPK and transcription factor ATF4 are more active in TIS senescent cancer cells rather than in the normal senescent cells (Shi et al. 2019). Lastly, the elevated glucose consumption and ATP production were observed in the senescent cancer cells induced by cyclophosphamide (Dörr et al. 2013).
Minimal Activities of TCA Cycle and Chaperones Are Required for Senescent Survival and Maintenance
3.2
Though glucose metabolism and ATP production in senescent normal cells are reduced significantly, the low activities of TCA cycle and glutaminolysis are required for the survival of normal senescent cells, as evidenced by inhibition of PDH or GLS1 leading to the selective killing of senescent fibroblasts. The mechanism underlying the selective killing of senescent cells by inhibiting TCA cycle and glutaminolysis is not quite clear. The known role of TCA cycle is to supply intermediate metabolites for cell growth and to produce NADH and FADH for ETC mediated production of ATP. In addition, glutaminolysis is a major source for replenishing TCA cycle intermediate alpha‐KG. Therefore, inhibiting PDH by CPI‐613 in combination with GLS1 inhibitor BPTES will likely block the production of intermediate metabolites of TCA cycle and cause the depletion of ATP production. As such, the shortage of intermediate metabolites and ATP depletion should account for the selective killing of senescent cells. In fact, GLS1 was demonstrated to be essential for the survival of human senescent cells, inhibiting GLS1 by BPTES has been shown to ameliorate age‐associated organ dysfunctions in aged mice (Johmura et al. 2021). More interestingly, both CPI‐613 and BPTES have been tested for cancer treatment. CPI‐613 was proven to be effective for relapsed/refractory AML and metastatic pancreatic adenocarcinoma (Feng et al. 2024; Zachar et al. 2011). BPTES treatment improved the therapeutic outcome in pancreatic ductal adenocarcinoma by removing gemcitabine‐induced senescent cancer cells (Oyama et al. 2024). It is worth pointing out that the doses of CPI‐613 and BPTES used in our senolytic tests on senescent cells are much lower than the established cancer treatment doses (Feng et al. 2024; Oyama et al. 2024).
Besides the large number of enzymes involving in glucose metabolism, we also identified many chaperones with down‐regulated protein levels in senescent fibroblasts, consistent with the previous findings that high levels of chaperones prevent animals from aging. For example, the transgenic worms and flies over‐expressing mitochondrial Hsp22 are long‐lived (Morrow et al. 2004; Walker and Lithgow 2003). Mice deficient in CHIP, a co‐chaperone of the heat‐shock family exhibit accelerated aging phenotypes, whereas long‐lived mouse strains show a remarkable up‐regulation of some heat‐shock proteins (Min et al. 2008; Swindell et al. 2009). The activation of heat‐shock response transcription factor HSF‐1 increases the longevity and thermo‐tolerance in nematodes (Westerheide et al. 2009). On the other hand, the low level of chaperones may lead to the misfolding and aggregation of proteins in age‐related neurodegenerative diseases such as Alzheimer's disease and Parkinson's disease (Campisi and Robert 2014). A couple of mechanisms may account for the low level of chaperones in IR‐induced senescent BJ cell. First, the decreased transcription of chaperone genes leads to less protein translation in normal senescent cells. Second, senescent cells tend to produce more misfolding proteins and cause unfolded protein response (UPR). The UPR, in turn, may suppress the global protein synthesis including chaperones (Hetz et al. 2020). How the Hsp90 inhibitor 17‐AAG affects the survival of senescent cells is unclear. Many client proteins of Hsp90 including EGFR, BRAF, HER2, AKT, CDK4/6, and mutant TP53 are essential for cell survival and resistance to apoptosis, as such, binding of 17‐AAG to Hsp90 may lead to Hsp90 inactivation and destabilization of its client proteins (Newman et al. 2012), which should be accountable for the selective killing of senescent cells by 17‐AAG.
Pyruvate Dehydrogenase May Be a Potent Target for the Development of Novel Senolytics
3.3
Senolytics by targeting and killing senescent cells have shown promise in extending lifespans and ameliorating age‐related diseases (Baker et al. 2011; Childs et al. 2016, 2021). However, the low selectivity and poor efficacy are the major hurdles for clinical applications of the current senolyitcs (Chaib et al. 2022; Di Micco et al. 2021). Thus, identifying more unique senescence features would help the development of novel senolytics with higher selectivity and efficacy. Accumulating evidence suggests TCA cycle plays an important role in promoting cell growth and proliferation. Alpha‐KG, one of the intermediate metabolites of TCA cycle, was shown to increase lifespan and health span in C. elegans and Drosophila through inhibiting mTOR, modulating DNA demethylation and reducing ROS production (Bayliak et al. 2022; Chin et al. 2014). Thus, alpha‐KG could be a potential target for therapeutics against senescence and aging. Low levels of PDH (the initial enzyme to produce alpha‐KG) and GLS1 (to supplement alpha‐KG through glutaminolysis) identified in senescent fibroblasts are well consistent with the pro‐survival activity of alpha‐KG. Despite both PDH and GLS1 being reduced significantly at senescence, they are required for the survival of senescent cells, as senescent cells were more sensitive to CPI‐613 or BPTES treatment, and co‐inhibiting TCA cycle and Hsp90 led to more profound selective killing of senescent cells.
In addition, the CPI‐613 compound does not directly target PDHA, the catalytic subunit of PDH. The senolytic activity of CPI‐613 observed by us most likely originates from the dual targeting of CPI‐613 on PDH and alpha‐KGDH. The dual targeting by CPI‐613 may downplay the PDH uniqueness as a potent senolytic target. Therefore, it is necessary to identify PDHA‐specific inhibitors (which are not available commercially so far) to distinguish the effects of PDHA‐specific inactivation from the additive effects exerted by the dual targeting of CPI‐613 in future investigations. Nevertheless, given the facts that PDH locates upstream of the TCA cycle and is central to supply acetyl‐CoA to the TCA cycle, the inhibition of PDH rather than alpha‐KGDH by CPI‐613 likely plays a dictating role in the selective killing of senescent cells, strengthening the role of PDH as a novel potent senolytic target.
Aging mouse models are widely used for aging research including the development of senolytics, as murine genetics and metabolic processes are similar to those of humans. The natural aging mouse model, ionizing radiation‐induced aging mouse model and D‐galactose induced aging mouse model are most used in aging research. Whereas the natural aging takes a long time to develop and exhibits a lot of individual variations, and ionizing radiation is dangerous and difficult to obtain, we turned to the D‐galactose induced aging mouse model for testing our senolytic targets. Excessive D‐galactose in the mouse body will be oxidized into superoxide anions and oxygen free radicals. As such, the resulting oxidative stress will induce mouse aging. Despite the high survival rate and better experimental reproducibility of the D‐galactose induced aging mouse model, there are still many differences in immunity, behavior and other aspects compared to the natural aging model. Thus, whether the D‐galactose induced aging mouse model can truly reflect the physiological and biochemical changes of aging remains an open question. In this regard, the primate aging model (e.g., cynomolgus monkeys) may be a better choice for senolytic testing in the future investigation (Yang et al. 2024).
Conclusion
4
In summary, by a quantitative proteomic profiling of the cellular landscape of senescent fibroblasts, we discovered that decreased glucose metabolism, reduced production of ATP and alpha‐KG, and declined chaperones are unique cellular features of senescent normal cells including fibroblasts. These cellular features are essential for the survival and maintenance of senescence in normal cells. Co‐inhibition of the TCA cycle and chaperones led to the enhanced selective killing of senescent normal cells as well as the therapy‐induced senescent cancer cells, and alleviated physical dysfunctions in aged mice. Thus, our profiling has identified key cellular features required for the survival and maintenance of senescence in normal cells and demonstrated that pyruvate dehydrogenase is a novel and potent target for the intervention of cellular senescence, shedding new light on the prevention and treatment of age‐related pathologies.
Methods
5
Cell Culture and DNA Damage Treatment
5.1
Human fibroblasts BJ (RRID: CVCL_3653; ATCC Cat#CRL‐2522), IMR‐90 (RRID: CVCL_0347; ATCC Cat#CCL‐186), WI‐38 (RRID: CVCL_0579; ATCC Cat#CCL‐75) and human cancer cell lines HeLa (RRID: CVCL_0030; ATCC Cat#CRM‐CCL‐2), A549 (RRID: CVCL_0023; ATCC Cat#CRM‐CCL‐185), and U2OS (RRID: CVCL_0042; ATCC Cat#HTB‐96) were purchased from American Type Culture Collection (ATCC) and cultured in Dulbecco's modified Eagle's medium (DMEM, Gibco) supplemented with 10% fetal bovine serum (FBS, Invitrogen) and 1% penicillin/streptomycin at 37°C under 5% CO_2_. Mycoplasma contamination was routinely tested. Only mycoplasma‐free cells were used in the experiments. For BJ fibroblasts stable isotope labeling by amino acids in cell culture (SILAC), cells were cultured in the DMEM medium supplemented with either regular lysine and arginine (light labeling) or isotope labeled^13^C_6_‐^15^N_2_‐L‐lysine and^13^C6‐^15^N_4_‐L‐arginine (heavy labeling). For proteomic profiling, heavy labeled BJ cells were irradiated with 10 Gy IR followed by 30 days of recovery culture. For validation experiments, BJ, WI‐38, and IMR‐90 fibroblasts were irradiated with 10 Gy IR followed by at least 15 days of recovery culture. To generate TIS tumor cells, A549, HeLa, and U2OS cells were treated with doxorubicin (MedChemExpress) at 0.75 μM, 0.2 μM, and 0.15 μM for 24 h respectively, and recovered for 24 h. This process was repeated three times for A549 cells. The status of senescence was confirmed by SA‐β‐gal staining and the elevated expression of p21 protein.
Antibodies and Chemical Regents
5.2
Antibodies used in the study: anti‐p21 (Cell Signaling Technology, 2947S) for Western blotting and anti‐p21 (Solarbio, K114558P) for immunohistochemistry; anti‐PFKP (Abways, 13389‐1‐AP), anti‐ALDOA (Abways, CY7206), anti‐PKM2 (Immunoway, YM1322); anti‐PDHA (Santa Cruz Biotech, SC‐377092); anti‐GLS1 (CUSABIO, E1218A), anti‐HSP90 (Proteintech, 13171‐1‐AP), anti‐HSP70 (Proteintech, 10995‐1‐AP), anti‐TCP1 (Abways, CY9414), anti‐β‐Actin (Proteintech, 66009‐1‐Ig), and anti‐β‐Tubulin (Proteintech, 10094‐1‐AP). Doxorubicin, 2‐deoxy‐glucose (2‐DG), CPI‐613, bis‐2‐(5‐phenylacetamido‐1,3,4‐thiadiazol‐2‐yl)ethyl sulfide (BPTES), and 17‐AAG were purchased from MedChemExpress.
SA‐β‐Gal Staining
5.3
The induced senescent cells were fixed in 4% paraformaldehyde for 15 min at room temperature. After that, cells were washed with PBS and incubated in 1 mL of freshly prepared 5‐bromo‐4‐chloro‐3‐indolyl‐b‐D‐galactopyranoside (X‐Gal) staining solution (Beyotime) at 37°C for 8–12 h under darkness. Cells were then washed twice with PBS and imaged under light microscopy.
Quantitative Proteomic Profiling
5.4
Quantitative proteomic profiling was performed as described previously (Hu et al. 2023). Heavy labeled BJ cells at 80% confluence were subjected to 10 Gy IR (γ‐rays), followed by 30 days of recovery culture. Light and heavy labeled lysates were prepared, mixed at a 1:1 ratio, and digested with trypsin (Promega) at 37°C overnight. After purification and fractionation, peptide samples were applied to LC‐MS/MS on an LTQ‐Orbitrap Velos mass spectrometer (Thermo Fisher Scientific). Raw files were converted to mzXML for processing. MS/MS spectra were searched via the SEQUEST algorithm against the human UniProt database (version 2014). Search parameters are allowed for dynamic modifications of oxidation on Met or deamidation on Asn or Gln and up to two missed cleavages and maximum modifications ⩽ 3 per peptide. Matches were filtered with common parameters (mass tolerance ⩽ 0.5 Da, XCorr ⩾ 2.0) to a peptide FDR rate less than 1%. For peptide quantification analysis, column normalization was performed. The abundance of each protein was scaled to a percent of the total, so that the summed S/N for that protein across all channels equaled 100, and the relative abundance was obtained.
Semiquantitative RT‐PCR
5.5
The growing and senescent cell lysates were collected for RNA extraction using TRIzol (Invitrogen) and cDNAs were synthesized by using a Superscript III reverse transcriptase kit (Invitrogen). The DNA concentration was determined by A260 reading (NanoDrop 2000). The transcripts of the examined genes were amplified using gene specific primers (for primer details see Table S2). PCR products were separated on agarose gels. Products of β‐Actin served as internal control.
Cytotoxicity and Cell Viability Assay
5.6
For measuring cell survival rate, the growing or senescent cells were seeded in 96‐well plates and cultured overnight. On the second day, cells were treated with desired drugs at various doses for the desired time. The survival cells were detected by Cell Counting Kit‐8 (Meilun Bio) according to the manufacturer's protocol and the plates were read at 450 nm. Data from 3 independent experiments were presented as means ± SD.
Glucose Consumption, ATP and Alpha‐KG Concentration, and NAD+/NADH Ratio
5.7
Glucose concentration was measured by the enzyme mediated colorimetric glucose assay kit (NJJCBio). Fresh media were added into 1 × 10^6^ cells, after 48 h of culture, media were collected and measured according to the manufacturer's protocol. Glucose consumption rate was normalized by the fresh medium and the cell numbers at 48 h of culture. The desired cells were subjected to desired drug treatment for 48 h and lysed with assay buffer provided in the kits. The concentration of ATP, alpha‐KG, and the ratio of cytosolic NAD+/NADH were determined by using an ATP luminometric assay kit (Beyotime), Amplex Red α‐Ketoglutarate Assay Kit (Beyotime), and NAD+/NADH detection kit (Abbkine Scientific), respectively. Results were analyzed according to the manufacturer's protocol.
Aging Mouse Model
5.8
All animal experiments were performed according to protocols approved by the Committee on the Ethics of Animal Experiments of Harbin Medical University.
C57BL/6 mice were purchased from Beijing Vital River Laboratory Animal Technology Co. Ltd. (Beijing, China). For the induced aging mice, 8‐week‐old mice were administered D‐galactose (500 mg/kg) intraperitoneal injections every day for 8 weeks. CPI‐613 (10 mg/kg), BPTES (7.5 mg/kg), and 17‐AAG (0.5 mg/kg) were dissolved in PBS with 10% DMSO, PEG400, and 5% Tween‐80 and administered via i.p. injection every 5 days for 1 month.
Physical Functions of Mice
5.9
All measurements were performed at least 5 days after the treatment. For the treadmill endurance test, mice were placed onto a motorized treadmill at an incline of 5° (Noldus Information Technology) with a starting speed of 5 m/min for 2 min and an increased speed of 7 m/min for 2 min, then 9 m/min for 1 min. On the test day, mice ran on the treadmill at an initial speed of 5 m/min for 2 min, and then the speed was increased by 2 m/min every 2 min until the mice were exhausted. Running distance was recorded and total work (kJ) was calculated as: kJ = mass (kg) × g (9.8 m/s^2^) × distance (m) × sin (5°).
For the rotarod test, the maximal staying time was measured using an accelerating rotarod system (Noldus Information Technology). Mice were trained on the rotarod for 3 days at speeds of 4, 6, and 8 rpm. On the test day, mice were placed onto the rotarod with rotating speed at 4 rpm. for 2 min. Then, the speed gradually accelerated from 4 to 40 rpm. over a 5‐min period. The maximum staying time was recorded when the mouse dropped off the rotarod. Results were averaged from 3 trials.
For the forelimb grip strength test, a grip strength meter (Noldus Information Technology) was used. The tested mouse was gripped by the tail and allowed it to hold onto the grid with its two front paws. The mouse was pulled away horizontally and the force generated by the mouse was measured. Tests were carried out six times for each mouse, and the maximum read of force was recorded.
Immunohistochemistry (IHC)
5.10
IHC was carried out on paraffin‐embedded sections of liver, kidney, and lung tissues of aged mice induced by D‐galactose with or without senolytic treatment. The expression levels of endogenous p21 were detected using an antibody against mouse p21. Five tissue sections/slides from each mouse were analyzed. Each slide was blinded read under light microscopy by two pathologists. Ten randomly picked view fields from each slide were assessed. The proportion of positively stained cells was calculated and normalized to the total cells in each view field.
Bioinformatics Analysis
5.11
All common bioinformatics analyses were performed using R software including data normalization and plotting (https://www.R‐project.org). UniProt was used as a general source of annotations; GO enrichment was performed using the GOstats package (Falcon and Gentleman 2007). Network graphics were generated by GeneMANIA and STRING analysis.
Statistical Analysis
5.12
All statistical analyses were described in the figure legends. All data were quantified from at least 3 independent results and presented as means ± standard deviation (SD). Student's test or two‐sided Welch's t‐test or one‐way ANOVA test was employed to calculate p value between different testing groups. Statistical significance is depicted by asterisks corresponding to *p < 0.05, **p < 0.01, or ***p < 0.001. GraphPad Prism 7 was employed for statistical analysis.
Author Contributions
M.Z., Z.H., S.P., and Y.S. performed the experiments. Y.J. and J.L. helped the mouse physical function. N.Z., A.L., H.X., and Y.Y. helped bioinformatic analysis. S.F. and W.S. provided the facilities. S.P.G. provided proteomic analysis. C.Z. designed and supervised the study. M.Z., Z.H., S.P., and C.Z. wrote and revised the manuscript. M.Z., Z.H., and S.P. contributed equally to this work.
Funding
This work was supported by National Natural Science Foundation of China (82172622).
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Appendix S1: acel70434‐sup‐0001‐AppendixS1.pdf.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Baker, D. J. , B. G. Childs , M. Durik , et al. 2016. “Naturally Occurring p 16(Ink 4a)‐Positive Cells Shorten Healthy Lifespan.” Nature 530, no. 7589: 184–189. 10.1038/nature 16932.26840489 PMC 4845101 · doi ↗ · pubmed ↗
- 2Baker, D. J. , T. Wijshake , T. Tchkonia , et al. 2011. “Clearance of p 16Ink 4a‐Positive Senescent Cells Delays Ageing‐Associated Disorders.” Nature 479, no. 7372: 232–236. 10.1038/nature 10600.22048312 PMC 3468323 · doi ↗ · pubmed ↗
- 3Basisty, N. , A. Kale , O. H. Jeon , et al. 2020. “A Proteomic Atlas of Senescence‐Associated Secretomes for Aging Biomarker Development.” P Lo S Biology 18, no. 1: e 3000599. 10.1371/journal.pbio.3000599.31945054 PMC 6964821 · doi ↗ · pubmed ↗
- 4Bayliak, M. M. , D. V. Gospodaryov , and V. I. Lushchak . 2022. “Mimicking Caloric Restriction for Anti‐Aging Effects: The Pro‐Oxidant Role of Alpha‐Ketoglutarate.” Current Opinion in Toxicology 30: 100339.
- 5Blasco, M. A. 2007. “Telomere Length, Stem Cells and Aging.” Nature Chemical Biology 3, no. 10: 640–649. 10.1038/nchembio.2007.38.17876321 · doi ↗ · pubmed ↗
- 6Bonnay, F. , A. Veloso , V. Steinmann , et al. 2020. “Oxidative Metabolism Drives Immortalization of Neural Stem Cells During Tumorigenesis.” Cell 182, no. 6: 1490–1507. 10.1016/j.cell.2020.07.039.32916131 · doi ↗ · pubmed ↗
- 7Camacho‐Pereira, J. , M. G. Tarragó , C. C. S. Chini , et al. 2016. “CD 38 Dictates Age‐Related NAD Decline and Mitochondrial Dysfunction Through an SIRT 3‐Dependent Mechanism.” Cell Metabolism 23, no. 6: 1127–1139. 10.1016/j.cmet.2016.05.006.27304511 PMC 4911708 · doi ↗ · pubmed ↗
- 8Campisi, J. 2005. “Senescent Cells, Tumor Suppression, and Organismal Aging: Good Citizens, Bad Neighbors.” Cell 120, no. 4: 513–522. 10.1016/j.cell.2005.02.003.15734683 · doi ↗ · pubmed ↗