Homoharringtonine may reduce obesity and insulin resistance linked to diet and aging and help mice live longer

To further verify the senolytic or senomorphic effects of the 110 compounds, we conducted secondary screening by testing these compounds in multiple cell types, including replicatively senescent (RS) HDFs, RS human umbilical vein endothelial cells (RS HUVECs), and PS human retinal pigment epithelial cells (PS hRPEs) (Figs. S1C-H). Through the secondary screening of the 110 compounds, we selected 15 compounds that showed senolytic or senomorphic (hereafter senotherapeutic) activities across all three cell types (Fig. 1A, S1I, and Supplementary Table 1). ABT263 and rapamycin were used as senolytic²⁴ and senomorphic²⁵ positive controls, respectively (Fig. S1I).

Adipose tissue senescence plays a causal role in the development of obesity- and age-associated insulin resistance and metabolic dysfunction^6,19. To validate the effects of the candidate compounds on obesity and associated metabolic dysfunction, we performed pilot in vivo experiments in high-fat diet (HF)-induced obese mice. Among the 15 candidates, two were initially excluded due to known adverse effects. We excluded #1806 (trametinib) due to gastrointestinal symptoms and reduced food intake^26–28, and #231 (everolimus) due to hyperglycemia^29,30. We tested the remaining 13 compounds in the HF-induced obese mice (Fig. 1B and Supplementary Table 2). We found that the #34 compound (HHT) mitigated weight gain and improved glucose regulation compared to PBS control in the absence of food intake differences (Fig. 1C–G and S2A–E). We then performed in vitro analyses to investigate cellular mechanisms underlying this effect in media containing 10% FBS. We found that HHT induced cytotoxicity in RS human visceral preadipocytes (HPAs) and RS HDFs in a dose-dependent manner, and elevated LDH levels in the media of these senescent cells, while reducing senescence-related phenotypes (Figs. S3A–C and S3E–G). This was accompanied by the elevation of cleaved caspase-3 (Casp-3) protein, indicating the induction of apoptotic cell death by HHT treatment in RS HPAs (Fig. S3D) and RS HDFs (Fig. S3H). Interestingly, HHT appeared to reduce cell numbers at 50 nM and 100 nM concentrations in non-senescent (NS) cells (Figs. S3A, S3B, S3E and S3F). Importantly, this did not induce LDH release (Figs. S3C and S3G), suggesting growth arrest rather than cell death, which aligns with the known mechanism of HHT in cancer cells and normal cells^31,32. This finding was further supported by flow cytometry, which demonstrated increased G0/G1 arrest in HHT-treated NS cells (Fig. S3I). Given that HHT induces cell cycle arrest, it was necessary to clarify whether the observed decrease in NS cell numbers was due to actual cytotoxicity or simply growth inhibition. To address this, we repeated the experiments under 0% FBS conditions, where overall cell proliferation is minimized. Under these conditions, HHT induced significant cytotoxicity in senescent HPAs and HDFs while having no effect on NS cell numbers (Fig. 1H–K and S3J–M). These effects of HHT are specific to senescent cells and comparable to well-known senolytics, dasatinib+quercetin (D + Q) and ABT263 (Figs. S3N, S4, and Supplementary Table 3). Additionally, HHT treatment reduced the expression of some SASP proteins in all tested cell types, including HPAs (Fig. S5).

These results suggest that HHT mitigates glucose intolerance in mice, and exhibits senotherapeutic activity across multiple human cell types, including preadipocytes.

Senescent cells accumulate in obese WAT and stimulate the secretion of inflammatory cytokines, immune modulators, growth factors, and proteases (i.e., SASP), leading to adipose tissue dysfunction and metabolic disturbance³⁴. Notably, we found that HHT treatment reduced crown-like structures (asterisks in Fig. 3E, F), suggesting the anti-inflammatory effects of HHT in WAT. HF induces hypertrophic enlargement of adipocytes, characterized by an increase in large-sized adipocytes and a decrease in small-to-medium-sized adipocytes. Treatment with HHT partially mitigated the hypertrophic expansion of the adipocytes, indicating improved WAT function (Fig. 3G). In addition, the levels of proinflammatory cytokines (IL-1β, MCP1) and their regulator (AP1)^35,36 were increased by HF, while HHT treatment suppressed this proinflammatory response (Fig. 3H). At the protein level, key components of the SASP involved in adipose tissue inflammation and fibrosis which were significantly increased by HF (TGF-β1, MMP2, and PAI-1) were reduced by HHT treatment (Fig. 3I). Protein levels of HSL and ATGL, lipolytic enzymes involved in triglyceride (TG) hydrolysis and insulin sensitivity which are known to be reduced in aged subjects³⁷, were also decreased by HF but normalized by HHT treatment (Fig. 3J). These results suggest that HHT attenuates HF-induced WAT inflammation and metabolic dysfunction.

To test whether HHT impacts thermogenesis, we examined the energy expenditure of the HHT-treated mice (Figs. S7A–C). We found that HHT treatment marginally (p = 0.056) increases the energy expenditure (Fig. S7C) and reduces the size of lipid droplets in brown adipose tissue (BAT) with significantly increased uncoupling protein (UCP1) expression in BAT (Figs. S7D and S7E). These data suggest that a moderate increase in energy expenditure by activation of BAT partially contributes to metabolic improvement by HHT treatment.

Taken together, these results suggest that HHT reduces adipose tissue inflammation and improves the thermogenic and lipolytic functions of adipose tissue, likely through the elimination of senescent cells in HF-fed obese mice.

Notably, the proportion of adipocytes (Cluster 5) increased in the HF-HHT-treated WAT compared with the HF-PBS-treated WAT (Fig. 4A, B). This result, along with reduced adipocyte sizes (Fig. 3E, G), suggests increased hyperplastic adipogenesis (i.e., adipocyte cell number increase), which is closely associated with improved WAT metabolic function and enhanced glucose homeostasis. To gain insights into the HHT-mediated molecular changes in adipocytes, we performed DGE analysis between the HF-HHT- and HF-PBS-treated adipocytes and found significantly altered genes presented in the volcano plot (Fig. 4C middle panel). Subsequent gene ontology (GO) enrichment analysis of DGEs in the HHT-treated adipocytes exhibited enriched signatures associated with insulin sensitivity (e.g., Adipor2), lipid and carbohydrate catabolic process, and fat cell differentiation (Fig. 4C, right panel; HHT > PBS plot). In contrast, the PBS-treated adipocytes showed signatures of cell migration and tissue organization (Fig. 4C, left panel; PBS > HHT plot). We further characterized the functional aspect of mature adipocytes (Cluster 5) and found a positive correlation of HHT-treated adipocytes with healthy adipokine profiles, such as Adipoq (Adiponectin), Cfd (Adipsin), and Vegfa (VEGF-A) (Fig. 4D)^42–45. Next, to directly test whether HHT treatment reduces adipose tissue senescence, we analyzed the SenMayo gene set, a panel of 125 genes that are associated with cellular senescence⁴⁶. We examined the SenMayo signature scores in various cell types and observed that the score was significantly lower in adipocytes and myeloid/erythroid cells, while it was increased in macrophages with higher M1-like and lower M2-like signature (Fig. 4E, S8C and S8D). This finding is unexpected and somewhat inconsistent with our histopathological data, which showed a reduced crown-like structure (CLS) following HHT treatment (Fig. 3E). Despite the seemingly contradictory transcriptomic signatures, our immunofluorescent staining with the M1-like macrophage marker iNOS further confirmed that CLS with iNOS-positive pro-inflammatory macrophages was significantly reduced in HHT-treated WAT (Fig. S8E). These findings indicate that HHT attenuates adipocyte senescence and may exert differential effects across various cell types. Furthermore, HHT-treated adipocytes exhibited significantly lower γ-H2AX immunofluorescent signal, indicating lower levels of DNA damage (Fig. 4F). Taken together, these data illustrate that HHT suppresses adipocyte senescence and improves the metabolic function of adipocytes.

Finally, to test whether the improved whole-body glucose handling observed with HHT treatment is a result of improved metabolic function of the adipose tissue, we performed hyperinsulinemic-euglycemic clamp studies. HHT administration restored the whole-body glucose turnover that was decreased by HF (Fig. 5F). We also found increased glucose uptake in the HHT-treated WAT compared with the PBS-treated WAT (Fig. 5G), indicating enhanced insulin sensitivity in WAT following HHT treatment. Additionally, we examined liver tissues and found significantly reduced steatosis in the HHT-treated liver with decreased TG accumulation and alanine aminotransferase (ALT) enzyme (Fig. 5H–J). These findings indicate that HHT-mediated enhancement of APC plasticity, along with reduced adipocyte senescence, promotes healthier adipose tissue remodeling and metabolic homeostasis in HF-fed obese mice.

We showed the senotherapeutic effect of HHT in human preadipocytes, suggesting its potential impact on human WAT. Thus, we further tested the effect of HHT on cellular senescence ex vivo in human WAT. As expected, HHT treatment decreased SAβG staining in human WAT in a dose-dependent manner (Fig. 6M). Consistent with this, HHT decreased the SAβG-positive cells in the tissue sections of the human WAT samples (Fig. S10C) and the levels of p53 and p21 proteins (Fig. 6N) in a dose-dependent manner. These findings indicate that HHT has senotherapeutic activity in human adipose tissue.

To directly assess the physical interaction between HHT and HSPA5, we conducted a Surface Plasmon Resonance (SPR) assay, which allows for the measurement of HHT’s affinity for HSPA5 protein⁵². The SPR assay revealed a significant molecular affinity between HHT and HSPA5 (Fig. 7K). To further investigate the molecular binding between HHT and HSPA5, we employed a binding model prediction. The initial binding model was determined using rigid docking, followed by Molecular Mechanics energies combined with the Generalized Born and Surface Area (MM-GBSA) calculation, a widely used method for estimating the free energy of the binding of small molecules to protein⁵³. HSPA5 is known to play a pivotal role in cell survival, and suppression of HSPA5 by inhibitors that bind to the ATP binding site of HSPA5 can lead to cell death⁵⁴. Therefore, we postulated that HHT may inhibit the function of HSPA5 through binding to the ATP binding site. The predicted binding model of HSPA5 and HHT shows key interactions including four hydrogen bonds (H-bonds) and one pi-cation interaction with HSPA5 residues (Fig. 7L and S11I). These results suggest that HHT can bind to HSPA5 and may act as an inhibitor of HSPA5. To test this hypothesis, we measured the ATPase activity of HSPA5 in the presence of HHT or HA15, a potent and specific inhibitor of HSPA5⁵⁵. We found that the inhibitory activity of HHT was greater than HA15, demonstrating that HHT acts as an HSPA5 inhibitor by reducing its ATPase activity (Fig. 7M). Our data collectively suggest that HHT physically interacts with HSPA5 to inhibit its ATPase activity and induces cytotoxicity in HSPA5 high-expressing senescent cells, subsequently leaving HSPA5 low-expressing cells.

In metabolic tissues, HSPA5 protein levels were significantly increased in both adipose tissue and skeletal muscle from HF-induced obese mice compared with chow-fed controls, whereas its expression was not significantly changed in the liver (Fig. 7N and S11J). The increase in HSPA5 expression was more prominent in adipose tissue than in skeletal muscle, suggesting that adipose tissue is a primary metabolic target organ responsive to HHT treatment. Indeed, HHT treatment in HF-fed obese or aged mice resulted in reduced levels of HSPA5 protein in adipose tissues due to the elimination of HSPA5 high-expressing senescent cells (Fig. 7N–P), providing mechanistic insight into the senescent cell targeted activity of HHT. In addition to its action through HSPA5, HHT is also known to inhibit the translation of ribosomal protein⁵⁶, which may be involved in mediating the senotherapeutic effect of HHT. Thus, to test this possibility, we examined the expression of ribosomal protein, RPS14, which is known to induce cell cycle arrest and senescence⁵⁷. No significant alteration in the level of RPS14 protein was observed in either non-senescent or senescent cells treated with HHT (Fig. S12A). In addition, we tested whether other translation inhibitors, such as cycloheximide (CHX), anisomycin (Aniso), and puromycin (Puro), influence the senescence of RS HPAs. We observed that none of the three translation inhibitors altered cell numbers, expression of aging biomarkers, and SAβG staining in RS HPAs (Figs. S12B–D). These findings indicate that the senotherapeutic activity of HHT is not associated with its translation-inhibitory function in HPAs. Taken together, these data suggest that HSPA5 may be the target molecule mediating the senotherapeutic activity of HHT, independent of HHT’s ribosomal protein inhibition activity.

Mice often develop lymphoma, which is among the leading causes of death in aged mice^60,61. As HHT has been used for the treatment of hematological oncology, the extended lifespan may primarily be due to the anti-cancer effects of HHT rather than its senotherapeutic effects. However, we did not find any differences in blood smear (Fig. S13B), CBC profiles (Supplementary Table 5), the numbers of Peyer’s patches (Fig. S13C), and the weight of inguinal lymph nodes (Fig. S13D) among Zmpste24^+/+, Zmpste24^+/-, and Zmpste24^-/- mice treated with HHT or vehicle (PBS) control. Additionally, no significant histological alterations were observed in the Peyer’s patches, inguinal lymph nodes, or the spleen (Fig. S13E). These results suggest that HHT mitigates aging phenotypes and extends lifespan in the absence of tissue toxicity in progeroid mice, independent of lymphocyte suppression.

As obesity-induced metabolic impairment expedites the aging process and exacerbates age-related diseases^65,66, we employed an additional aging model by a combination of HF and aging. Two-month-old mice were fed a HF diet for 11 months, and HHT was administered by once-weekly injections for 5 months while continuing the HF regimen (Fig. 9F). We found that HHT significantly extended the lifespan of mice, increasing the median lifespan from 509 to 551 days, whereas the maximum lifespan could not be determined because the mice were sacrificed at 558 days for tissue collection (Fig. 9G). In addition, HHT induced weight loss, mainly attributable to fat rather than lean mass reduction (Fig. 9H, I). To investigate the senotherapeutic effects of HHT on various cellular populations in WAT, we performed mass cytometry (CyTOF) analysis^39,67 (Fig. S15A). Intriguingly, despite no significant differences in the number of APCs in perigonadal WAT (pWAT), we found reduced expression of CD142 and ICAM-1, markers for committed APCs, in HHT-treated WAT, indicating that HHT-treated APCs maintain characteristics of stemness^39,67 (Fig. 9J, K). However, we did not find any differences in the adipose tissue immune cell populations, except for the reduction in M1-like macrophages (Fig. S15B), which is consistent with our data shown in Figs. 3E and S8E. Moreover, we did not find any differences in the histopathology of spleen tissue, blood smear, or the number of splenic lymphocytes between PBS- and HHT-treated mice, suggesting that the extended lifespan is not associated with the lymphoma suppression activity of HHT, and that HHT treatment does not cause toxicity to hematopoietic cells, liver, small intestine, or kidney (Figs. S15C–G). We found moderately reduced proteinuria in HHT-treated mice without signs of renal toxicity (Figs. S15H–M). Collectively, these results indicate that HHT delays aging and prolongs lifespan in physiological aging mice.

It is important to acknowledge the limitations of our study. HHT-induced cytotoxicity was not observed in NS cells, particularly under serum-free (0% FBS) conditions. This resistance may be attributed to serum starvation-induced quiescence, which could enable cells to evade HHT’s cytotoxic effects. Additionally, autophagy activation under serum-depleted conditions may have contributed to drug resistance, as autophagy functions as a well-established cytoprotective mechanism against various cellular stresses, including chemotherapeutic agents. We observed the most pronounced in vivo effects of HHT in adipose tissue. A plausible explanation is that HSPA5 is highly expressed in adipose tissue, particularly in visceral adipose tissue, and its expression increases with aging, obesity, and diabetes in both humans and mice⁹⁰. These findings support the notion that adipose tissue may serve as a major target organ of HHT. Another possibility is that HHT preferentially accumulates in adipose tissue compared to other organs. However, tissue distribution and pharmacological action of HHT in mice are currently unavailable. Therefore, future pharmacokinetic and pharmacodynamic studies are warranted to determine the tissue distribution of HHT and to clarify its primary target sites. Since HHT was administered systemically, it may have exerted beneficial effects on other metabolic organs/tissues, such as the liver and skeletal muscle, which could contribute to the observed weight loss and metabolic improvement. Further investigations are needed to elucidate the tissue-specific effects of HHT. Although our preclinical data and human fat explant studies demonstrate strong senotherapeutic effects of HHT, additional translational experiments are required to support future clinical trials. It would be important to examine the effects of HHT in female mice, as our current study focused on male mice. Lastly, further research is warranted to address the potential therapeutic effects of HHT in various age-related conditions.

Animal studies and human adipose tissue research were conducted in compliance with all relevant ethical regulations. Animal experiments were performed according to protocols approved by the Institutional Animal Care and Use Committee (IACUC) of Yeungnam University College of Medicine (YUMC-AEC2019-031, YUMC-AEC2020-025, and YUMC-AEC2022-039) and by the Animal Care Committee of The Centre for Phenogenomics (TCP, AUP #27-0271H). Research involving human adipose tissues was approved by the Institutional Review Board (IRB) of Yeungnam University Medical Center (YUMC 2021-03-052).

Clinically applied compounds (2150 compounds) under phase 1 to 3 clinical trials and FDA-approved were kindly provided by Korea Chemical Bank (Daejeon, Republic of Korea) (https://chembank.org/↗). Cell Counting Kit-8 (CCK-8), and lactate dehydrogenase (LDH) assay kit were obtained from Dojindo Molecular Technologies (Rockville, MD, USA), 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal) from Amresco LLC (Solon, Ohio, USA), doxorubicin from Ildong Pharmaceutical Co. Ltd. (Seoul, Republic of Korea), and HHT from Tocris (Bristol, UK). Dexamethasone (DEX), 3-isobutyl-1-methylxanthine (IBMX), and insulin were obtained from Sigma-Aldrich (St. Louis, MO, USA). Primary antibodies used in this study for Western blotting, immunostaining, and mass cytometry are listed in Supplementary Table 7 and 8. Sequences for primers used in RT-qPCR are listed in Supplementary Table 9. siRNAs were purchased from Life Technologies Corp. (Carlsbad, CA, USA), and detailed information is provided in Supplementary Table 10. HPAs (human visceral preadipocytes, #PT-5005) and HUVECs (human umbilical vein endothelial cells, #C2517A) were purchased from Lonza (Basel, Switzerland). HDFs (human dermal fibroblasts, #PCS-201-012), hRPEs (human retinal pigment epithelial cells, #CRL-2302), and 3T3-L1 mouse preadipocytes (#CL-173) were purchased from American Type Culture Collection (Manassas, VA, USA).

HDFs were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM; Welgene Inc., Gyeongsan, Republic of Korea) supplemented with 10% fetal bovine serum (FBS; Gibco, Carlsbad, CA, USA), while HPAs were maintained in a preadipocyte basal medium with appropriate supplements and growth factors (Lonza, Basel, Switzerland) containing 10% FBS. HUVECs were grown in endothelial cell basal medium-2 enriched with various growth factors, and hRPEs were cultured in DMEM:F-12 medium with 10% FBS. Additionally, 3T3-L1 cells were maintained in DMEM with 10% calf serum (Gibco).

To induce premature senescence (PS), HDFs and hRPEs were treated with 0.5 μM doxorubicin in serum-free media for 4 h⁹⁹. Then, cells were cultured with DMEM (10% FBS) and DMEM/F-12 (10% FBS) for 4 days and induction of cellular senescence was confirmed by measuring the upregulation of p16 and SASP levels as well as an increase in SAβG staining. The percentages of SAβG staining in PS HDFs and PS hRPE were more than 70% and 60%, respectively.

Replicatively senescent (RS) model was established following the protocol outlined in previous studies^99,100. Briefly, HDFs (2 × 10⁵), HUVECs (2 × 10⁵), and HPAs (2 × 10⁵) were cultured in a 100 mm dish and sub-cultured when cells reached 80-90% confluency. The number of population doublings (PDs) was calculated from the geometric equation: PD = log₂F/log₂I, where F is the cell number at the end of incubation time, and I is the cell number at the beginning of the incubation time. We also measured the population doubling time (PDT), which was calculated using the following equation: PDT = (Tf-Ti)/log₂(F/I), where Tf is the final incubation time, Ti is the initial incubation time, F is a final population number, and I is an initial population number. For all experiments, HDFs were used after passage 5 (PD < 20, PDT < 2 days) or passage 23 (PD > 60, PDT > 14 days). HUVECs were used in passage 5 (PD < 20, PDT < 2 days) or passage 23 (PD > 60, PDT > 14 days). HPAs were used in passage 3 (PD < 20, PDT < 2 days) or passage 13 (PD > 60, PDT > 14 days). These are referred to as “non-senescent (NS)” and “senescent” cells, respectively. Cellular senescence in RS HDFs, RS HPAs, and RS HUVECs was confirmed by measuring the upregulation of p53, p16, and SASP levels and an increase in SAβG staining. The percentages of SAβG staining in RS HDFs and RS HUVECs were more than 80% and RS HPAs were more than 60%.

Cell number was assessed using the CCK-8 according to the manufacturer’s instructions and by manual cell counting under light microscopy.

LDH activity was measured in media using an LDH assay kit according to the manufacturer’s instructions.

SAβG activity was measured with minor modifications¹⁰⁰. Cells were washed 2 times with phosphate-buffered saline (PBS) and fixed with 3.7% (v/v) formaldehyde in PBS for 5 min. The human adipose tissues and the epididymal WAT of mice were fixed in 4% paraformaldehyde (PFA) for 30 min and washed twice with PBS. Cells and tissues were incubated in a SAβG staining solution (1 mg/mL 5-bromo-4-chloro-3-indolyl-β-D-galactoside, 40 mM citric acid-sodium phosphate (pH 6.0), 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 150 mM NaCl, and 2 mM MgCl₂) for 18 h at 37 °C. Tissue samples were embedded with an O.C.T compound and cryosectioned to 10 μm at -30 °C. SAβG positive cells stained with blue color were counted and analyzed.

To identify senotherapeutic candidates from a library of 2150 FDA-approved drugs, primary and secondary screenings were conducted. In the primary screening, PS HDFs were seeded at 1 × 10³ cells per well in a 96-well plate and cultured overnight. The cells were then treated with each drug at a concentration of 100 nM and incubated for 4 days at 37 °C in a CO₂ incubator. Cell viability was assessed using the CCK-8 assay, while cellular senescence was evaluated by SAβG staining. Based on these results, 110 compounds were selected for further analysis. In the secondary screening, RS HDFs, RS HUVECs, and PS hRPEs were seeded at 1×10³ cells per well in a 96-well plate, cultured overnight, and treated with each drug at 100 nM for four days. Cell viability was assessed using the CCK-8 assay, and cellular senescence was determined by SAβG staining. To evaluate potential toxicity, non-senescent cells were also treated with 100 nM of each compound for four days, followed by a viability assessment using the CCK-8 assay. ABT263 and rapamycin were used as positive controls for senolytic and senomorphic activity, respectively.

Cells were harvested and fixed in 70% cold ethanol at 4  °C for 1 h. Cells were washed twice with PBS and resuspended in a propidium iodide (50 µg/mL, Sigma-Aldrich, P4170) solution. RNase A was added to a final concentration of 0.5 µg/mL, and the samples were incubated at 4  °C for 4 h. Flow cytometric analysis was performed using a CytoFLEX Flow Cytometer (Beckman Coulter Life Sciences, Brea, CA, USA).

3T3-L1 preadipocytes were cultured in DMEM with 10% calf serum until two days post confluence (Day 0). To induce differentiation, the cells were treated with MDI differentiation medium, consisting of DMEM supplemented with 10% FBS, 0.5 mM 3-isobutyl-1-methylxanthine, 1 μM dexamethasone, and 0.5 μg/mL insulin. On Day 2, the medium was replaced with a maturation medium composed of DMEM containing 10% FBS and 0.5 μg/mL insulin. The cells were maintained in this medium for an additional 6 days, with medium changes every 2 days. On Day 8, adipocyte differentiation and cellular senescence were assessed using Oil Red O staining, SAβG staining, and Western blot analysis.

Cells were fixed with 4% paraformaldehyde for 30 min, rinsed 2 times with PBS, and incubated with an ORO working solution for 1 h. The images were photographed under a microscope, and ORO staining density in cells was quantified using the ImageJ software (National Institutes of Health, Bethesda, MD, USA).

Senescent HPAs were seeded in 60 mm dishes and incubated overnight. siRNAs (20 nM) were transfected into cells using RNAiMax 2000 transfection reagent (Invitrogen, Gaithersburg, MD, USA). Protein knockdowns were confirmed by Western blot analysis. For overexpression study, HPAs were transfected with a pCMV-HSPA5-His plasmid purchased from SinoBiological (Seoul, Republic of Korea) using Lipofectamine 2000 reagent (Invitrogen) and the expression levels of HSPA5 protein were confirmed by Western blotting.

C57BL/6 N mice were purchased from Koatech (Pyungtaek, Republic of Korea), and p16-Luc mice were kindly donated by Eiji Hara (Osaka University, Japan). Zmpste24^-/- mice on a C57BL/6 J background were kindly gifted by Dr. Zhongjun Zhou (University of Hong Kong, China). To examine the effect of senotherapeutic candidates and HHT on adipose tissue senescence and insulin resistance in obesity, 8-week-old male C57BL/6 N mice and p16-Luc were randomly divided into chow diet (Envigo, Indianapolis, IN, USA) or HF groups (Research diets, New Brunswick, NJ, USA) for 6 weeks. HF contains 60.3% of the total calories as fat, 21.3% as carbohydrates, and 18.4% as protein. Then, mice in an HF group were divided into two groups, PBS- or senotherapeutic candidate-treated group, and placed on the same HF for further 6-8 weeks. Senotherapeutic candidates including HHT were dissolved in DMSO and diluted in PBS. Senotherapeutic candidates were injected intraperitoneally three times a week at the indicated dosages (Supplementary Table 2). Mice were given ad libitum access to water and food. To examine the effect of HHT on adipose tissue browning in obese mice, HF-fed mice were exposed to 4 °C for 6 h following PBS and HHT administration for 8 weeks. The HF and drinking water were provided during cold exposure.

To examine the effects of HHT on adipose tissue senescence and insulin resistance in aged mice, 18-month-old male C57BL/6 N mice (DBL, Eumseong, Republic of Korea) were randomly divided into PBS- or HHT-treated groups and were administered intraperitoneally with PBS or HHT three times a week for 14 weeks. They were provided with a chow diet.

To measure the lifespan of aged mice, C57BL/6 J (16 months old) male mice were intraperitoneally injected once a week with HHT for 12 months or PBS (n = 8 in each group). For the lifespan of Zmpste24^-/- mice, Zmpste24^-/- mice (5 weeks old) were administered intraperitoneally once a week for 14 weeks with HHT or PBS (n = 7 for PBS and n = 9 for HHT in male, and n = 4 for PBS and n = 8 for HHT in female mice). For the lifespan of HF-fed aged mice, C57BL/6 J (2 months old) male mice were fed an HF for 11 months, and HHT was administered once a week intraperitoneally for 5 months while continuing the HF regimen.

Mice were housed in a controlled environment with a 12 h light:12 h dark cycle. Food intake and body weight were measured manually once a week during the experimental periods. Mice were anesthetized with 2,2,2-tribromoethanol and tert-amyl alcohol (375 mg/kg), and blood was collected from the orbital plexus. The plasma and tissues were stored at −80 °C until analysis. The in vivo experiments were performed in strict accordance with guidelines and protocols approved by the Institutional Animal Care and Use Committee of Yeungnam University College of Medicine (permit number: YUMC-AEC2019-031, YUMC-AEC2020-025, and YUMC-AEC2022-039) and by the Animal Care Committee of The Centre for Phenogenomics (TCP, AUP #27-0271H).

This study has been approved by the IRB at the Yeungnam University Medical Center (YUMC 2021-03-052). All patients declared their informed consent in written form. Human subcutaneous adipose tissues were obtained from female patients who underwent mammoplasty after mastectomy due to breast cancer. Detailed information was provided in Supplementary Table 11. Adipose tissue samples were cut into 5 × 5 × 5 mm pieces, placed in 48 well plates, and cultivated in DMEM containing 10% FBS and antibiotics at 37 °C in a 5% CO₂ humidified incubator. Human subcutaneous adipose tissue samples were treated with the indicated concentrations of HHT twice every 3 days for 7 days.

Airway responsiveness was determined through whole-body plethysmography (Buxco Research System, Wilmington, NC, USA). Briefly, mice were placed in a whole-body plethysmography chamber. After a few minutes for stabilization, enhanced pause (Penh, an indication of airway resistance) was determined.

The liver was homogenized in a 20-fold solution of chloroform: methanol (2:1) containing 0.01% butylated hydroxyl toluene, and vortexed for about 30 sec. After incubation overnight at 4  °C, the homogenized sample was diluted twice with distilled water and centrifuged at 800 x g for 30 min. The chloroform layer was separated, and the same amount of chloroform containing 1% triton X100 was added. After the evaporation of chloroform, the remaining part was dissolved with distilled water. Triglyceride Reagent and Free Glycerol Reagent (Sigma-Aldrich) were used to measure the level of TG. The concentration was calculated with the standard solution.

An intraperitoneal glucose tolerance test (IPGTT) was conducted after an overnight fasting for 16 h. Before glucose administration, blood was collected from the tail vein to measure basal blood glucose and plasma insulin levels. Animals were injected intraperitoneally with glucose (1 g/kg body weight). Blood glucose levels were measured at 0, 15, 30, 60, 90, and 120 min after injection with blood from the tail vessel, using an ACCU-CHEK Instant glucometer (Roche, Basel, Switzerland). Plasma at 0, 15, 30, and 60 min was stored at -80 °C for insulin measurement. For the insulin tolerance test (ITT), mice were fasted for 6 h and injected with insulin intraperitoneally at 0.75 IU/kg body weight. Blood glucose levels were measured at 0, 15, 30, 60, 90, and 120 min after injection using the same glucometer.

The hyperinsulinemic-euglycemic clamp study was conducted in conscious mice 4 days after vein cannulation. The vein cannulation surgery was performed under isoflurane inhalation anesthesia, during which an indwelling catheter was inserted into the left internal jugular vein. The catheter was externalized through a small incision in the skin at the back of the neck. Following the surgery, the mice were individually housed and carefully monitored until they regained their preoperative body weight, indicating full recovery.

After an overnight fasting, hyperinsulinemic-euglycemic clamp was performed with a priming dose (900 pmol/kg body weight) of human insulin (Lilly, IN, Indianapolis, USA), and constant infusion of insulin at a rate of 24 pmol/kg/min, 20% glucose was infused for maintaining glucose at constant concentrations of 5-6 mM through a Y connector for 2 h using a microdialysis pump (CMA Microdialysis, Kista, Sweden). Blood was withdrawn from the tail vessel, and the plasma glucose levels were measured with a glucose analyzer (Analox, Middlesbrough, UK). The insulin-stimulated whole-body glucose uptake rate was estimated with a continuous injection of [3-³H] glucose (PerkinElmer Life and Analytical Sciences, Shelton, CT, USA) throughout the clamps (0.1 μCi/min). The insulin-stimulated glucose uptake rate in epididymal fat was measured by 2-deoxy-D-[1-¹⁴C] glucose (2-[¹⁴C] DG) administration as a bolus (10 μCi) after starting the clamp at 75 min. After the clamp was over, mice were anesthetized with avertin, and collected plasma and tissue samples were stored at -80 °C for further analysis. To determine tissue 2-[¹⁴C] DG 6-phosphate content, tissue samples were homogenized, and 2-DG-6-phosphate was separated from 2-DG in supernatants using an ion-exchange column (Bio-rad, Hercules, CA, USA). Whole body glucose turnover rate was calculated as the ratio of the glucose infusion rate to the specific activity of plasma glucose. The glucose uptake rate in each tissue was calculated from the plasma 2-[¹⁴C] DG profile using MLAB (Civilized Software, Silver Spring, MD, USA) and the tissue 2-[¹⁴C] DG-6-phosphate content.

Plasma insulin levels were measured with an enzyme-linked immunosorbent assay kit (Merck, Darmstadt, Germany). The liver enzymes, ALT and AST were measured in plasma using the commercial kit (Asan pharm, Seoul, Republic of Korea).

To collect urine, mice were held over the cling wrap and lightly stroked the belly of the animal. Urine was collected into the tube and urine protein assay was performed using a strip test (Medi-Test Combi 10 L, Allentown, PA, USA), and BCA assays (Thermo, Meridian, Rockford, IL, USA).

Urea nitrogen and creatinine in urine and serum were analyzed by using Urea Nitrogen Colorimetric Detection Kit (Arbor Assays, Ann Arbor, Michigan, USA) and QuantiChrom™ Creatinine Assay Kit (BioAssay Systems, Hayward, CA, USA), respectively according to the manufacturer’s instruction. Urine total protein, albumin, creatine, potassium, and sodium were analyzed by the Beckman Coulter AU480 Clinical Chemistry Analyzer (Brea, CA, USA).

Complete blood count analysis was performed using an automated hematology analyzer (Sysmex XN-1000 hematology analyzer, Sysmex, Kobe, Japan) in whole blood samples. Blood samples (200 µL) were collected from the orbital venous plexus of mice under isoflurane anesthesia and placed into heparinized sample tubes, and a complete blood count analysis was performed. For blood smear, following the collection of blood from mice in EDTA-containing tubes, blood smear samples were prepared on glass slides and stained using the Wright-Giemsa method¹⁰¹.

Body composition was assessed in Zmpste24^-/- mice using a DEXA scanner (iNSiGHT VET DXA, OsteoSys Co., Ltd., Seoul, Republic of Korea). Each mouse was anesthetized with an intraperitoneal injection of 240 mg/kg avertin and placed on the scanner bed in the prone position, with the limbs and tail stretched away from the body. Quantitative data within the region of interest were obtained. Body composition of HF-fed aged mice was analyzed using EchoMRI-100 body composition analyzer (Echo Medical Systems, Texas, USA), which quantifies fat and lean mass in live, non-anesthetized mice.

The forelimb grip strength of mice was measured using a BIO-GS3 grip-strength meter (BIOSEB, Pinellas Park, FL, USA). Mice were placed horizontally onto the grid (dimensions 100 × 80 mm, angled 20 °), allowed to grip the grid with their forelimbs only, and then gently pulled away backward until they released their grip. Results were averaged over 5 trials.

Tissues were immediately fixed in 10% neutral formalin. The fixed tissues were divided into an appropriate size to make a paraffin block. Blocks were cut into 5 μm using Microtome (Leica, Wetzlar, Germany). After attaching 2-3 pieces of cut sections to the glass slide, they were dried and sequentially stained with hematoxylin and eosin step. The sections attached to glass slides were first immersed in xylene for 2 min, followed by sequential rehydration in 100% ethanol for 1 minute, 95% ethanol for 2 min, and 90% ethanol for 2 min. After three brief washes in distilled water, the sections were stained with hematoxylin for 7 min, followed by three additional washes. The slides were dipped in acid-ethanol once and immediately washed. Next, the sections were incubated in sodium sulfate for 5 min and rinsed in running water. Eosin staining was performed by immersing the sections in eosin solution for 3 min, followed by washing for 5 min under running water. The sections were then dehydrated through a graded ethanol series (70% to 100%) with three dips at each step. Finally, the sections were dipped in xylene three times and then mounted with a coverslip. The stained samples were observed under a microscope (Olympus, Tokyo, Japan) at 40× and 100× magnification. Pictures were derived using the DIXI eXcope image S/W program (DIXI Science, Daejeon, Republic of Korea), and the size of adipocytes was estimated by measuring the area of all cells in one field at 100 × magnification using the ImageJ program. Crown-like structure (CLS) percentage was analyzed by counting the number of surrounded cells by macrophages at 2 fields and then averaging them in the same samples. Paraffin-embedded tissue sections were stained with Sirius Red (Abcam, ab246832) for Picrosirius Red staining or with a Trichrome Staining Kit (Roche, 06521908001) for Masson’s trichrome staining, followed by light microscopy observation. Blue collagen-positive areas in Masson’s trichome staining were analyzed using an ImageJ software.

For γH2AX, F4/80, and p53 staining, paraffin-embedded tissue sections were deparaffinized and subjected to heat-induced epitope retrieval in citrate buffer (pH 6.0) using a pressure cooker. Sections were then incubated with Protein Block (Agilent, X090930-2) or 5% BSA (MP Biomedicals, 160069) in PBS to reduce non-specific binding, followed by overnight incubation at 4 °C with primary antibodies diluted in DAKO Antibody Diluent (Agilent, S302283-2). Rabbit anti-γH2AX (Fortis Life Science, IHC-00059; 1:1500), rabbit anti-iNOS (Abcam, #ab3523; 1:100), and rabbit anti-p53 (Abbkine, ABP0110; 1:200) were used. Corresponding secondary antibodies - donkey anti-rabbit IgG Alexa Fluor 647 (Invitrogen, A32795), goat anti-rat IgG Alexa Fluor 647 (Invitrogen, A48265), and goat anti-rabbit IgG Alexa Fluor 546 (Invitrogen, A11035) - were applied at room temperature. Nuclear counterstaining was performed with 4’,6-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich, D9542). To reduce tissue autofluorescence, slides were incubated in Sudan Black B (Sigma-Aldrich, 199664) saturated in 70% ethanol for 25 min at room temperature, then mounted with Vectashield Vibrance anti-fade medium (Vector, H-1700). Whole-slide images were acquired at 20× magnification using an Olympus VS120 scanner equipped with a Hamamatsu ORCA-R2 C10600↗ digital camera (Evident, Japan).

Paraffin-embedded tissue sections were deparaffinized, treated with freshly prepared 0.45% H₂O₂ in methanol for 15 min at room temperature in the dark, and subjected to heat-induced epitope retrieval in 10 mM citrate buffer (pH 6.0) at 95–98 °C for 15 min. Cell samples were fixed with 4% paraformaldehyde for 15 min and treated with 3% H₂O₂ for 10 min. Both tissue sections and cell samples were permeabilized with 0.2% Triton X-100 in TBS (tris-buffered saline) for 10 min. Tissue sections were blocked with 5% normal donkey serum containing 1% bovine serum albumin (BSA), and cell samples were blocked with 5% BSA (MP Biomedicals, 160069) in PBS for 1 h. Tissue sections were then incubated with a mouse anti-HSPA5 antibody (Santa Cruz Biotechnology, sc-166490; 1:100) at 4  °C overnight. Cell samples were incubated with a mixture of a mouse anti-HSPA5 antibody (Santa Cruz Biotechnology, sc-166490; 1:100) and a rabbit anti–cleaved caspase-3 antibody (Cell Signaling Technology, #9664S; 1:100) at 4  °C overnight. After three washes with TBS-T (TBS containing 0.05% Tween-20), secondary antibodies were applied for 1 h at room temperature. For tissue sections, a horse radish peroxidase (HRP)-conjugated goat anti-mouse secondary antibody (Cell Signaling Technology, #7076) was used. For cell samples, a mixture of HRP-conjugated goat anti-mouse antibody (Invitrogen, #31430; 1:50) and alkaline phosphatase (AP)-conjugated goat anti-rabbit antibody (Sigma-Aldrich, A3687; 1:50) was applied. Following final TBS-T washes, color development was performed using enzyme-specific substrate systems: HRP activity was visualized with 3,3′-diaminobenzidine (DAB) substrate (Cell Signaling Technology, #7076, or Dako, K5007) to yield brown staining, and AP activity was visualized with Fast Red TR/Naphthol AS-MX (Sigma-Aldrich, F4648) to yield red staining. Finally, tissue sections and cell samples were counterstained with hematoxylin and mounted. Images were acquired using a KF-PRO-005 slide scanner (KFBIO Technology for Health, Yuyao, China), and quantification of stained areas or positive cells was performed using ImageJ software (National Institutes of Health).

Metabolic assessments of each mouse were conducted for a continuous 48 h period using individual closed metabolic chambers (Oxylet; Panlab, Cornellà, Spain). During the measurement period, all mice were provided ad libitum access to food and water to maintain normal physiological conditions. Oxygen consumption (VO₂, mL/min) and carbon dioxide production (VCO₂, mL/min) were sequentially measured for 3 min per mouse, rotating through all animals housed in the system. Room air was analyzed for 3 min after completing one cycle of all mice to establish baseline gas composition. All collected data, including VO₂, VCO₂, and energy expenditure, were analyzed using the manufacturer-provided software (Metabolism; Panlab). To account for inter-individual differences, all values were normalized to body weight.

The genomic DNA of adipose tissue was prepared using a NucleoSpin DNA Lipid Tissue Kit (Machery-Nagel, Duren, Germany), and the identification of differentially methylated CpG (DMCs) using reduced representation bisulfite sequencing (RRBS) data with 500 ng of genomic DNA.

Briefly, the genomic DNA was digested with MspI and ApeKI and then purified using the Qiagen MiniElute PCR Purification Kit. The DNA fragments were directly subject to end-blunting followed by adding a single adenine nucleotide (dA-tailing) to prepare the DNA for adapter ligation. Methylated adapters were ligated to the prepared DNA, and the adapter-ligated fragments, specifically those within the size range of 160 to 420 bp, were isolated and used for bisulfite conversion utilizing the ZYMO EZ DNA Methylation-Gold Kit™, adhering to the manufacturer’s guidelines. The conversion process alters unmethylated cytosines to uracils, enabling subsequent methylation analysis. Following bisulfite treatment, the libraries were amplified via PCR using PfuTurbo Cx Hotstart DNA polymerase (Agilent Technologies, Santa Clara, CA, USA), a step that enriches the library for subsequent sequencing. The final RRBS libraries were evaluated for quality and size distribution using an Agilent 2100 Bioanalyzer (Agilent Technologies). The RRBS libraries were sequenced on an Illumina NextSeq 500 (NextSeq 500, Illumina, San Diego, CA, USA).

Sequencing adapters and low-quality bases present in the raw reads were trimmed using the Skewer tool¹⁰². Following the quality control and adapter trimming, the resulting high-quality reads were aligned to a reference genome that had been converted to a 3-letter code (to represent bisulfite-treated DNA) using the bs_seeker2-align module of BSseeker2¹⁰³. After alignment, the bs_seeker2-call_methylation module of BSseeker2 was employed to determine the methylation status at CpG sites across the genome based on the mapped reads with a minimum coverage of 5 to obtain a detailed profile of methylation patterns, enabling the identification of methylation levels at individual CpG sites across the samples in control and treatment groups.

To estimate epigenetic age (DNAm age), we employed a multi-tissue epigenetic age clock model that utilizes ridge regression model for parameter estimation, leveraging DNA methylation data across various tissues. The parameters for this model, including beta coefficients (weights) and intercepts for methylation CpG sites, were sourced from supplemental information reported in the manuscript¹⁰⁴. These parameters are well-validated and have demonstrated robustness in predicting chronological age from methylation data across different tissues, including adipose tissue. For each sample in our dataset, epigenetic age was calculated by summing the products of methylation levels at specific CpG sites and their corresponding beta coefficients. Recognizing the challenges associated with the variable number of clock sites available in RRBS, and to enhance the accuracy and consistency of our age estimates, we imputed missing methylation data by filling zero coverage sites with the mean methylation level observed for those sites within the control or treatment groups¹⁰⁵. This method ensures a more robust estimation of DNAm age by maintaining a comprehensive dataset, thereby reducing the bias from missing data. The resulting calculation provides a weighted sum of methylation levels, representing an estimation of the epigenetic age in months.

Cells were washed with ice-cold PBS and harvested by scraping in 50 μl of ice-cold radioimmunoprecipitation assay (RIPA) buffer (25 mM Tris-HCl, pH 7.4, 150 mM KCl, 5 mM NaF, and 1 mM phenylmethylsulfonyl fluoride). Cells were ruptured by vortexing two times for 30 sec at 30 min intervals on ice and centrifuged at 16,000 x g for 15 min at 4  °C. To measure SASP in culture media, the culture media were filtered with an Amicon® Ultra-4 Centrifugal Filter, 50 kDa MWCO (Merck KGaA, Darmstadt, Germany) and concentrated using an Amicon® Ultra-4 Centrifugal Filter, 3 kDa MWCO (Merck KGaA). Human adipose tissues were homogenized in 5 volumes of ice-cold RIPA buffer and centrifuged at 16,000 x g for 15 min at 4  °C. Protein concentrations in the supernatants were quantified by the bicinchoninic acid method (Pierce Biotechnology Inc., Rockford, IL, USA). Mouse adipose and lung tissue samples were homogenized in 10 volumes of ice-cold lysis buffer containing 50 mM HEPES, 150 mM NaCl, 50 mM NaF, 1 mM phenylmethylsulfonyl fluoride (PMSF), 1 mM benzamide, 1 mM EDTA, 1 mM EGTA, 1 mM sodium ortho-vanadate, 0.22% β-glycerophosphate, 1% NP40, 10% glycerol, and protease inhibitor cocktail (Santa Cruz Biotechnology., Santa Cruz, CA, USA). The tissue homogenate was centrifuged at 16,000 x g for 10 min at 4 °C, and the supernatant containing protein was collected. Protein concentrations were determined using the Bradford (Bio-rad) assay. An equal amount of protein samples was separated on 10% sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis at 70 V, and then resolved proteins were transferred to a membrane. Membranes were blocked in 1 × TTBS (10 mM Tris-HCl pH 7.5, 150 mM NaCl, and 0.05% Tween-20) containing 5% skim milk for 30 min at room temperature. Primary antibodies were applied overnight at 4 °C, and then HRP-conjugated secondary antibodies were applied for 90 min at room temperature. After rinsing the membranes 3 times with 1 × TTBS for 30 min, antigen-antibody complex was detected using Luminol Reagent (Santa Cruz Biotech Inc). Proteins were visualized with the Fujifilm LAS-3000 image system (Stamford, CT, USA). Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as a control for protein loading. The relative intensities of protein bands, as compared with that of the respective GAPDH signal, were determined by using the Multi Gauge software, version 3.0 (Fujifilm Corp., Stamford, CT, USA), and normalized to 1.0 for control cells by averaging three separate experiments. All phosphorylated forms were normalized to their respective protein forms.

Total RNA was extracted from 80 mg tissue samples using RNeasy Lipid Tissue Mini Kit (QIAGen, Hilden, Germany) according to the manufacturer’s instructions. cDNA was synthesized from 1 µg of RNA using a reverse transcription kit (Applied Biosystems, Foster City, CA, USA). RT-qPCR was performed using the Real-time PCR 7500 System and Power SYBR Green PCR Master Mix (Applied Biosystems). The expression levels of GAPDH were used for sample normalization. Each reaction mixture was incubated at 95 °C for 10 min followed by 45 cycles of 95 °C for 15 sec, 55 °C for 20 sec, and 72 °C for 35 sec.

For adipose tissue single nuclei sequencing (snRNA-seq), frozen tissues were chopped into smaller pieces on dry ice. Chopped tissues were immediately incubated in lysis buffer before homogenization in glass douncers. Isolated nuclei were then washed twice with cold wash and resuspension buffer. Nuclei suspension was filtered through 40 µm filter before proceeding to the next step. Nuclei were stained with DAPI and flow-sorted. After sorting, nuclei were washed and resuspended for counting. Nuclei were stained with SYBR Green II and counted under the microscope. Sorted nuclei were used as input into the 10x Genomics single-cell 3’ v3.1 assay (Pleasanton, CA, USA) and processed as described by the protocol provided by the 10x Genomics. For sequencing, the molarity of each library was calculated based on library size as measured by bioanalyzer (Agilent Technologies) and qPCR amplification data (Roche). Samples were pooled and normalized to 1.5 nM. The library pool was denatured using 0.2 N NaOH for 8 min at room temperature and neutralized with 400 mM Tris-HCl. Library pool at a final concentration of 300 pM was loaded to sequence on Novaseq 6000 (Illumina). Samples were sequenced with the following run parameters: Read 1–28 cycles, Read 2–90 cycles, index 1–10 cycles, index 2–10 cycles. Sequencing target reads for each library were 40,000–50,000 reads/nuclei.

The 10x Genomics’ CellRanger v5.0.0 software was used for sequence alignment and barcode processing with the mouse reference genome (mm10-2020-A). The filtered gene-count matrices from CellRanger were used for subsequent preprocessing. Doublets were identified and removed through the automated ‘scDblFinder’ toolkit using default parameters¹⁰⁶. Low-quality nuclei were removed from the dataset by retaining only nuclei with greater than 400 unique features (genes), less than the 95^th percentile of total UMI counts & features, and less than 0.5 percent of mitochondrial transcripts detected. The percentage of measured gene expression in each cluster attributed to mitochondrial gene (percent. MT) and the number of unique genes found for each nucleus (nFeature_RNA) were within expected values for each cell¹⁰⁷.

The filtered data from HHT treated and control sample was analyzed using the standard integration protocol as implemented in Seurat v4.0.1^108,109. This involved log normalization, variable feature selection and selecting/finding integration features/anchors. Cells were clustered using unsupervised Louvain clustering using the integrated nearest neighbors graph as input. The clustering resolution was determined empirically, as defined by the resolution with the greatest cluster stability (sc3_stability) from the clustree R package^110,111. In the final dataset, there were 5342 and 6392 cells in the HHT and control conditions, with 14 total clusters, each present in both conditions.

Cluster-specific markers were identified using the Wilcoxon Rank Sum test implemented in the FindAllMarkers function in Seurat. Lastly, differential gene expression between HHT and control was modeled using a hurdle MAST model¹¹². We considered genes with an absolute fold change greater than 0.10 and Bonferroni-adjusted p value < 0.05 to be differentially expressed between conditions. These differentially expressed genes were then used as input for Gene Ontology (GO) enrichment analyses. Specifically, GO analysis was conducted for biological processes ontology with a minimum gene set size of 10 and maximum size of 500 using the clusterProfiler package¹¹³. For age-related analysis on snRNA-seq data, we performed additional analysis using SenMayo signature. SenMayo scores were calculated using Seurat’s ‘AddModuleScore’ function¹¹⁴, which computes the average expression levels of each gene set— in this case, the SenMayo set⁴⁶—at the single-cell level. These scores are then normalized by subtracting an aggregated expression of control feature sets. To account for differences in total expression, all analyzed genes were first binned based on their average expression, and control features were then randomly selected from each bin. The SenMayo signature scores were subsequently compared between PBS- and HHT- treated groups using Wilcoxon’s rank sum test, with p values adjusted for multiple comparisons using FDR correction.

The DARTS procedure was performed as previously described⁵¹. Briefly, cell lysates of RS HDFs were incubated with 50 μM HHT for 2 h at room temperature with rotation, and then 2 μl of the indicated protease solution was added to 20 μL samples of lysate. After incubation for 30 min at room temperature, digestion was stopped by adding a protease inhibitor cocktail. Proteins were separated by SDS-PAGE and observed with silver staining. Gel bands showing increased stability were cut out and prepared for mass spectrometry analysis to characterize proteins (Genomine Biotechnology, Pohang, Republic of Korea).

ATPase activity was measured by an enzyme-coupled spectrophotometric assay using ATPase Activity Assay Kit (Sigma-Aldrich, #MAK113) following the manufacturer’s instructions. Briefly, we established a standard curve using phosphate standards (#MAK113B). The reaction solution (40 mM Tris, 80 mM NaCl, 8 mM MgAc₂, 1 mM EDTA (pH 7.5), and 4 mM ATP) was incubated with HSPA5 protein (5 μg) alone or HSPA5 protein and 10 μM inhibitor (HA15) at room temperature for 60 min. Each well was added with 200 µl of reagent (#MAK113A) and then incubated for 30 min at room temperature to terminate the enzyme reaction. The optical density of colorimetric product was measured using a microplate reader (TECAN, Männedorf, Switzerland) at 620 nm. The obtained absorbance was substituted into the standard curve to calculate the ATPase activity.

Measurements of the apparent dissociation constants (KD) between HSPA5 and HHT were carried out using a Biacore T200 biosensor (GE Healthcare Bio-Sciences, Pittsburgh, PA, USA). The recombinant human HSPA5 was bound covalently to the sensor chip CM5 (carboxylated dextran matrix) using an amine-coupling method. The human HSPA5 (20 μg/mL) in 10 mM sodium acetate pH 4.0 was coupled via injection, followed by the injection of 1 M ethanolamine to deactivate residual amines. For kinetic measurements at 20 °C, HHT with concentrations ranging from 1.5625 to 100 μM was prepared by dilution in HBS-P buffer (10 mM of HEPES (pH 7.4), 150 mM NaCl and 0.05% v/v surfactant P20 and 2% DMSO). Based on the results, the association rate constants (ka), dissociation rate constants (kd), and the equilibrium dissociation (binding) constants (KD; kd/ka) were determined.

Possible binding model between HSPA5 and HHT was predicted by docking studies based on the X-ray structures of HSPA5 and its known inhibitor VER155008. The binding site of HHT is presumed to be like VER155008 and the related PDB structure (6CZ1.pdb) was referenced for docking study. The predicted binding model was verified considering the interaction between HSPA5 and VER155008. The initial binding model was determined by rigid docking study and MM-GBSA was performed for the final binding model prediction. MAESTRO (Schrödinger LLC, NY, USA) was used all calculational processes using Glide and Prime modules on Linux environment with default parameters¹¹⁵.

For CyTOF, stromal vascular fraction (SVF) and splenic immune cells were isolated from WAT and spleen, respectively. Visceral WAT and spleen are dissected, and finely minced. Adipose tissue was digested with Adipose tissue dissociation kit (Miltenyi Biotec, Gaithersburg, MD, USA) using gentle MACSTM Octo Dissociator with Heater (Miltenyi Biotec, Gaithersburg, MD, USA). Splenic cells were washed with PBS, neutralized with DMEM/F12 media supplemented with 10% FBS, and filtered using a 100 μm filter. WAT tissues were further digested with Type II collagenase (Worthington; #LS004176↗) in a 37 °C incubator for 30-40 min. Digested tissue was neutralized with DMEM/F12 media supplemented with 10% FBS and filtered using a 100 μm filter. The flow-through was centrifuged for 5 min at 400 x g. After removing the floating adipocyte fraction and supernatant, red blood cell lysis buffer (Sigma-Aldrich, #R7757) was added and incubated at room temperature for 2-3 min. The red blood cell lysis buffer is neutralized with DMEM/F12 media (supplemented with 10% FBS) and centrifuged for 5 min at 400 x g. The dissociated cells from adipose tissue or spleen were washed in a cell staining medium (CSM) and each sample was barcoded before staining according to the protocol described in the Cell-ID 20-Plex Pd Barcoding kit (Fluidigm, San Francisco, CA). Barcoded samples were pooled into 1 sample and stained with metal-tagged antibodies (Supplementary Table 8) as previously described⁶⁷. Fixation and permeabilization were conducted according to the manufacturer’s instructions (Foxp3/Transcription Factor Staining Buffer Set, eBioscience). Finally, samples were washed and suspended in Maxpar Cell Acquisition Solution containing EQ normalization beads (Fluidigm, #201237) and data was collected by Helios instrument. FCS files were normalized using EQ beads by the Helios software. All data processing and analysis was done with Cytobank Enterprise (Beckman Coulter).

All statistical analyses were assessed by using Prism version 8 (Graph Pad Software Inc; La Jolla, CA, USA). The results are presented as the means ± standard error. In in vitro experiments, we conducted at least three independent experiments. Differences among the three or more groups were analyzed via one-way or two-way analysis of variance (ANOVA) followed by Tukey HSD test. Differences between two groups were analyzed by two-tailed Student’s t test. A p-value below 0.05 was considered statistically significant.

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