Targeting Senescence with Apigenin Improves Chemotherapeutic Efficacy and Ameliorates Age‐Related Conditions in Mice

In an effort to discover novel natural agents that can effectively regulate senescent cells, we performed comprehensive drug screening with a NMA library comprising 66 natural products (Table S1, Supporting Information), mostly secondary metabolites derived from plants, microbes, and certain animals. To this end, PSC27, a primary normal human prostate stromal cell line, was selected as an experimental cell model for extensive studies. PSC27 develops a canonical SASP upon senescence induced by inherent stress and/or exterior stimuli, including exhaustive replication (replicative senescence, RS), genotoxic chemotherapy, and ionizing radiation (therapy‐induced senescence, TIS).^[11, 26, 27, 28^] We exposed cells to a sub‐lethal dose of bleomycin (BLEO, 50 µg mL⁻¹), with a typical cellular senescence phenotype being observed 8–10 d after treatment, as evidenced by elevated senescence‐associated β‐galactosidase (SA‐β‐Gal) staining positivity, decreased BrdU incorporation and increased DNA damage repair (DDR) foci (Figure S1A–C, Supporting Information). To streamline our approach, we designed a screening procedure to evaluate the effects individual natural agents exerted on the in vitro viability of senescent cells and expression of their SASP (Figure1A).

Induction of apoptosis in senescent cells represents a central feature and the most valuable capacity of senolytics, as exemplified by studies involving senolytic compounds such as ABT‐263 (navitoclax), ABT‐737 and procyanidin C1 (PCC1).^[11, 29, 30^] Alternatively, a major advantage of senomorphics, for example, rutin and resveratrol,^[28, 31^] is the ability to selectively downregulate the SASP expression. To determine the relevant anti‐senescence potentials in these NMA components, we analyzed the effect of these natural agents on senescent PSC27 cells, by first confirming its experimental feasibility as a cell‐based model for geroprotective compound screening purposes. Preliminary results indicated that some agents either killed senescent cells (while preserving their proliferating counterparts), or suppressed the expression of CXCL8, a canonical SASP factor, thus further validating the potential and appropriateness of PSC27 for extended investigations (Figure 1B–E, Figure S1D,E, Supporting Information). Despite a thorough screening of the NMA library, we failed to discover any new senolytic agents, which are supposed to functionally resemble the positive controls ABT‐263 and PCC1 (Figure 1B,C; Figure S1D,E, Supporting Information). The results also implied the rareness of natural senolytic agents and the challenge in expanding the present reservoir of senotherapeutic agents particularly those of senolytic potential. Nevertheless, we noticed that several natural compounds generated a significant senomorphic effect, which warrants in‐depth studies (Figure 1D,E).

Among the medicinal agents, rutin, resveratrol, and apigenin displayed strong senomorphic capacity (Figure 1D,E). Resveratrol is a natural product with SASP‐inhibiting activity and targets PI3K/Akt signaling pathway.^[31^] As another phytochemical molecule recently unraveled with senomorphic efficacy, rutin attenuates the acute stress‐associated phenotype (ASAP) and affects the SASP development afterward, mainly through disrupting the interactions between ATM and HIF‐1α as well as between ATM and TRAF6.^[28^] In this study, we chose to specifically focus on apigenin, a plant‐derived natural flavonoid, which can downregulate the expression of typical SASP factors upon bleomycin‐induced senescence in BJ fibroblasts by inhibiting NF‐κB p65 activity through targeting IRAK1/IκBα signaling and IκBζ expression, with in vivo activity of SASP suppression identified in the kidney of aged rats.^[32^] Apigenin can also restrain the SASP by blockade of IL‐1α signaling through IRAK1, p38MAPK, and NF‐κB, thus alleviating pro‐tumorigenic effects of the SASP in breast tumors.^[25^] However, despite these intracellular activities, which are relatively downstream in the SASP signaling network, a wider influence of apigenin on the transcriptome‐wide expression of senescent cells, especially the more upstream or even direct targets after exposure of these cells to apigenin, remains basically unknown.

Given that apigenin inhibits the expression of CXCL8, a SASP hallmark factor, we further asked whether it downregulates the majority of SASP factors, or even counteracts cellular senescence. Therefore, we performed SA‐β‐Gal staining and BrdU incorporation assays in vitro to determine the anti‐senescence effects of apigenin. The data indicated that SA‐β‐Gal staining profiles and mitotic inactivity remained largely unchanged, in both proliferating cells and their senescent counterparts (Figure2A,B).

A subset of hallmark SASP factors, including IL6, CXCL8, IL1α/1β, MMP1/3, GM‐CSF, TIMP1, and WNT16B displayed a dose‐dependent decline upon exposure of senescent cells to apigenin, with a concentration of 10 µM appearing most effective (Figure 2C). Furthermore, transcriptomic data from RNA‐seq showed that expression of the canonical SASP factors was markedly upregulated upon cellular senescence but significantly downregulated upon apigenin treatment (Figure 2D). Immunoblot assays indicated that apigenin abrogated SASP expression in senescent cells, as evidenced by the decreased expression of IL1α, IL1β, IL6, CXCL8, and SAA2 (Figure 2E). Given the striking effect of apigenin in restraining the SASP expression, we performed GSEA analysis for signatures of SASP and NF‐κB, the latter plays a significant role in activating the ASAP and sustaining the SASP.^[27^] Our results suggest that both molecular signatures were apparently suppressed upon apigenin treatment (Figure 2F,G).

Gene ontology (GO) enrichment analysis of apigenin‐downregulated genes indicated that chemokine activity, estrogen 16‐alpha‐hydroxylase activity, double‐stranded RNA binding and CCR/CXCR chemokine receptor binding activities were among the most significantly suppressed molecular functions (MF) (Figure 2H). In terms of biological processes (BP), type I interferon signaling pathway, defense response to virus, response to viruses, and inflammatory response were most dramatically affected (Figure S2A, Supporting Information). The representative cellular components (CC) most associated with downregulated genes were those residing in extracellular region and space, although some proteins were transported to plasma membrane, platelet dense granule lumen, as well as calcium‐ and calmodulin‐dependent protein kinase complex (Figure S2B, Supporting Information). KEGG pathway analysis suggested that cytokine‐cytokine receptor interaction, viral protein interaction with cytokine receptor, TNF signaling and NF‐κB signaling pathways were most significantly inhibited upon treatment of senescent cells by apigenin (Figure 2I). Altogether, all these data confirm the capacity of apigenin to suppress the expression of pro‐inflammatory cytokines and chemokines, a subset of central SASP factors, starting from the transcriptomic level.

As the vast majority of SASP components are soluble factors secreted into the extracellular space, we collected the conditioned media (CM) from human stromal cells and analyzed them with mass spectrometry (MS). The resulting data suggest that among the SASP factors upregulated in senescent cells, a large number exhibited a reversal change pattern, namely downregulation, upon apigenin treatment (Figure S2C,D, Supporting Information). Some typical SASP components such as IL6, CXCL1/3/8, CSF1, CCL2, MMP1/2/3, IGFBP7, GDF15, HGF, AREG, and EREG, which are frequently reported by literatures to be secreted by human senescent stromal cells,^[33, 34^] appeared in the top list with statistical significance. Therefore, as a senomorphic agent, apigenin can restrain the expression of a large subset of SASP factors, although not the full spectrum SASP, implying the presence of other signal transduction pathway(s) involved in mediating signals that support the SASP development.

To expand, we further assessed the effect of apigenin on senescent cells induced by other means, including inherent stress or environmental stimulation, as illustrated by replicative senescence (RS) and ionizing radiation (RAD), respectively. As expected, apigenin holds the potential to diminish the expression of the majority of canonical SASP factors, regardless of the types of senescence‐inducing modality (Figure, Supporting Information). Additionally, we performed further experiments with WI38 and IMR90, two typical fibroblast lines derived from human embryonic lung tissues. The results largely resembled that of PSC27, a human prostate‐derived stromal line, suggesting that apigenin‐exerted suppression of the SASP is essentially organ type‐independent (Figure, Supporting Information). Together, our studies demonstrated a salient capacity of apigenin in restraining SASP expression, which is neither senescence type‐ nor organ type‐dependent. S2E,F S2G,H

Given the critical role of apigenin in suppressing the SASP, we next queried its functional mechanism in targeting senescent cells. To examine the potential effect of apigenin on DDR signaling, which is usually the driving force of a persistent SASP, we first investigated the extent of molecular pathway perturbation in senescent cells upon exposure to apigenin. Induction of senescence by BLEO caused notable activation of ATM, a central modulator of the DDR signaling cascade, as evidenced by immunoblot assays (Figure3A). As a response to external challenge or endogenous stimuli, proliferating cells first develop an acute stress‐associated phenotype (ASAP), characterized by phosphorylation and nucleus‐to‐cytoplasm translocation of ATM, TRAF6 auto‐ubiquitination and TAK1 activation, a series of prompt intracellular reactions to damage before senescence markers become evident^[27^] (Figure 3A). Upon cell genotoxic treatment, TAK1, a key cytoplasmic kinase that modulates the SASP, was rapidly phosphorylated, a change allowing its functional involvement in the transition of the ASAP toward the SASP via dual feedforward mechanisms.^[27^] We further observed phosphorylation/activation of p38MAPK, a molecule downstream of TAK1, followed by engagement of PI3K/Akt/mTOR, a signaling pathway that supports the formation of a chronic secretory phenotype, the SASP. However, in the presence of apigenin, these sequential molecular events were generally abrogated, although phosphorylation of ATM and TAK1 remained largely unaffected. In accordance with this molecular pattern, we would surmise that the potential target(s) of apigenin may be mechanistically located downstream of ATM and/or TAK1, but upstream of p38MAPK, PI3K/Akt/mTOR and other factors that may be implicated in the transition of the ASAP to the SASP during cellular senescence.

To unravel the mechanism regulating the function of apigenin in targeting senescent cells, we conducted a proteome‐scale profiling of interactors with capacity to interact with both ATM and p38MAPK, two molecules that are known to be involved in the transduction of signals that allow development of the SASP but lack direct correlation in senescent cells. Bioinformatics suggested 198 ATM‐ and 234 p38MAPK‐uniquely interacting molecules in human cells, with 19 such interactors shared by both ATM and p38MAPK (Figure 3B, Figure S3A,B, Supporting Information). Among these interactor candidates, we noticed that HSPA8 and NPM1, intracellular molecules that are involved in cellular stress responses, particularly senescence,^[35, 36^] and hold the potential to mediate the interaction between ATM and p38MAPK, thus deserving further investigation. p‐ATM‐mediated co‐immunoprecipitation (Co‐IP) revealed that each of HSPA8 and NPM1 can directly interact with p‐ATM and p38MAPK. However, the interaction of HSPA8, but not NPM1, with p‐ATM or p38MAPK was remarkably attenuated by apigenin, indicating that apigenin specifically hinders the interaction of HSPA8, rather than other molecules such as NPM1, with p‐ATM and p38MAPK (Figure 3C,D).

HSPA8, an important member of the heat shock protein 70 family, is functionally involved in chaperone‐mediated autophagy, one of the main pathways of the lysosome‐autophagy proteolytic system, whereas deficient protein degradation compromises cellular proteostasis and activates signaling pathways to culminate in induction of cellular senescence, a major feature of aging.^[37^] HSPA8 activates NF‐κB signaling by destabilizing IκBβ protein in the absence of lipopolysaccharide (LPS) or facilitating its nuclear translocation in the presence of LPS, an activity that can be synergized by thioredoxin domain‐containing protein 5 (TXNDC5) to exacerbate the inflammatory phenotype of synovial fibroblasts in rheumatoid arthritis via NF‐κB signaling.^[38^] HSPA8 is also essential for VEGF‐induced Akt phosphorylation, while downregulation of HSPA8 abolishes Akt phosphorylation.^[39^] Therefore, we speculate that HSPA8 plays a key role in orchestrating the functional involvement of a number of factors responsible for maintaining senescence and SASP development. To test this hypothesis, we evaluated expression of a series of SASP factors or senescence‐associated markers. Our study indicated that in the presence of VER155008 (hereafter VER), a selective HSPA8 inhibitor, expression of canonical SASP factors was dampened in a concentration‐dependent manner (Figure 3E). Further data indicated that cellular senescence was accompanied by IκBα degradation in the cytoplasm as well as cytoplasm‐to‐nucleus translocation of NF‐κB subunits (p65, p50), a tendency that was generally reversed upon apigenin or VER treatment, suggesting a mechanism shared by these agents in curtailing activation of the NF‐κB complex (Figure 3F,G). Interestingly, treatment with VER also suppressed the occurrence of a cascade of molecular events including activation of p38MAPK and engagement of PI3K/Akt/mTOR upon genotoxic stress (Figure 3H). We further observed markedly reduced interactions between ATM/p38MAPK and HSPA8 upon inhibition of HSPA8 activity with the small molecule inhibitor VER, as evidenced by data from IP and immunoblot assys (Figure S3C,D, Supporting Information).

Thus, interfering with crosstalk between ATM and its critical targets such as HSPA8 and p38MAPK, exemplified by the natural agent apigenin, is able to attenuate SASP expression while maintaining cellular senescence or growth arrest, a distinctive feature that is basically in line with effects of senomorphics. As the majority of natural flavonoid polyphenols have antioxidant activities, including reducing ROS overproduction and alleviating oxidative stress, we next asked whether apigenin has a similar effect in senescent cells. When exposed to a genotoxic agent (BLEO) for 8–10 d, stromal cells had markedly elevated ROS levels compared with their proliferating counterparts. However, in the presence of apigenin (or VER as an experimental control), production of ROS was significantly inhibited, largely consistent with the oxidative radical‐neutralizing capacity of apigenin, although the ROS level remained largely unaffected in normal cells (Figure 3I,J).

Although our data suggested that apigenin inhibits SASP expression through disrupting the interactions of HSPA8 with other critical factors that mediate SASP signal transduction, resulting in a blocked transition of ASAP to the SASP (Figure 3C–E), the potential direct target(s) of apigenin in senescent cells causing a senomorphic effect remains unclear. Therefore, we applied two affinity‐based approaches, biotin‐tagged apigenin (hereafter Bio‐APIG) and drug affinity responsive target stability (DARTS),^[40, 41^] to reveal the potential direct target(s) of apigenin (Figures4A, S4A–C, Supporting Information). In the probe‐based route, upon incubation with senescent cell lysates, Bio‐APIG bound to proteins before streptavidin‐agarose beads were applied to pull down the Bio‐APIG‐protein complex for assay by MS. We observed 61 potential target proteins, which were identified with distinct enrichment by Bio‐APIG but competitively inhibited by unlabeled apigenin (unique peptides ≥2, LFQ intensity ≥1.5 for enrichment or ≤0.67 for competitive inhibition), also evidenced by protein sliver staining (Figure 4B, Figure S4D, Supporting Information). To further discover bona fide target proteins, we followed another approach, namely DARTS, to substantiate the findings by target identification. The data indicated that 16 proteins were markedly upregulated upon incubation with apigenin, relative to vehicle (unique peptides ≥2, LFQ intensity ≥1.2) (Figure 4C). Among the proteins (PRDX6, SEPT11, and YWHAE) identified by both approaches, we speculated that apigenin may directly bind to PRDX6, a hypothesis that needs to be experimentally proved (Figure 4D).

We next interrogated the biophysical properties underpinning the interaction between PRDX6 and apigenin. As an approach to analyze protein thermal stabilization after ligand binding, cellular thermal shift assay (CETSA) was performed in a temperature range from 30 to 50 °C. We observed an increase of the stability of PRDX6 upon incubation with apigenin compared to the vehicle (Figure 4E,F). We further examined the binding effect by performing CETSA at concentrations ranging from 0 to 400 µM of apigenin either at the fixed temperature of 50 °C, or the ratio of enzyme (pronase) to protein fixed at 1: 300. Unsurprisingly, there appeared a marked stability elevation of PRDX6 as reflected by both approaches in a concentration‐dependent manner (Figure 4G–J). Results from surface plasmon resonance (SPR) assay, a label‐free direct optical biosensor approach, showed a distinct binding of recombinant human (rh) PRDX6 in the presence of apigenin, yielding a predicted dissociation constant (KD) of 0.237 µM (Figure 4K). To identify the residue(s) of PRDX6 responsible for the binding by apigenin, we performed in silico molecular modeling, with the results illustrating that apigenin establishes direct hydrogen bond connections with PRDX6 via the Cys91 and Glu210 of PRDX6 (Figure 4L, Figure S4E, Supporting Information). We then performed microscale thermophoresis (MST) with exogenous PRDX6 protein linked with a GFP tag. Similar to the SPR assay, MST analysis displayed evident interaction between PRDX6 and apigenin with an estimated Kd of 916 nM (Figure 4M,N). Since we speculated Cys91 and Glu210 of PRDX6 may be critical for the binding between target and ligand, we generated constructs that allow to express mutants at these two sites individually or simultaneously. In contrast to wild type PRDX6, mutation of either Cys91 or Glu210 to alanine increased their Kd by approximately tenfold. However, when both sites were mutated, binding between PRDX6 and apigenin was basically abrogated (Figure 4M,N, Figure S4F, Supporting Information). Altogether, our data indicate that the binding of PRDX6 and apigenin may generate a significant effect during cellular senescence.

Previous studies reported that PRDX6 has unique dual‐function enzyme activities within the peroxidase family, with its phospholipase A2 (PLA2) activity causing cells to generate arachidonic acid (AA), a molecule that often induces inflammation.^[42^] As a peroxidase, PRDX6 facilitates the conversion of H₂O₂ into water in the presence of NADPH to minimize oxidative damage. Thus, we engineered a construct to overexpress PRDX6 and assessed the capacity of apigenin to remove H₂O₂, especially when compared with NAC, a selective inhibitor of peroxidase (Figure S4G,H, Supporting Information). In contrast to NAC, which remarkably reduced the peroxidase activity of purging H₂O₂ (IC50 = 15.5 µM), the efficacy of clearing H₂O₂ by apigenin was less notably compromised (IC50 = 49.5 µM) (Figure S4I, Supporting Information). The differential consequences generated by PRDX6 and NAC may be attributable to the multiple biological effects caused by NAC, a competent antioxidant that also elevates intracellular Cys level and allows for sulfane sulfur production, stimulating mitochondrial bioenergetics, actions that are beyond its well‐known function of attenuating oxidative stress and protecting cells against oxidative damage, mainly as a ROS scavenger.^[43^]

However, upon measurement of PLA2 activity, we noticed a distinct reduction, implying that the capacity of apigenin to dampen the expression of pro‐inflammatory factors may be attributed to its capacity to block the PLA2 activity of PRDX6 (Figure 4O). Moreover, we applied MJ33, a selective PRDX6 PLA2 inhibitor, to examine its efficacy to constrain SASP expression, which is correlated with inflammatory activity of senescent cells. We found that MJ33 treatment lowered expression of the canonical SASP factors in a concentration‐dependent manner (Figure 4P). Proteomics profiling showed that, unlike the vehicle, MJ33 treatment significantly inhibited the PI3K/Akt signaling pathway, further substantiating its senomorphic effect (Figure S4J, Supporting Information). We analyzed the protein levels of human PRDX family, which encompasses six homologs (PRDX1‐6), in senescent cells in the absence or presence of apigenin. We observed neither increased nor decreased expression of these homologs (Figure S4K, Supporting Information). To further dissect the mechanism underlying SASP downregulation by apigenin‐induced suppression of the PLA2 activity of PRDX6, we performed PRDX6‐mediated co‐immunoprecipitation (co‐IP) followed by MS analysis. The results suggested that HSPA8 indeed does interact with PRDX6, suggesting functional interference with either of these molecules may explain the senomorphic effect of apigenin we observed in this study (Figure S4L,M, Supporting Information). Altogether, the data support the possibility that apigenin dampens SASP expression through directly binging to PRDX6 and blocking its PLA2 activity.

Given that PRDX6 is a direct target of apigenin, we queried whether PRDX6 plays an important role in mediating the SASP development. To address this, we chose to deplete PRDX6 from human stromal cells with small hairpin RNAs (shRNAs) (Figure, Supporting Information). Upon treatment of cells with BLEO, we observed markedly weakened upregulation of SASP factors (Figure, Supporting Information). Of note, the addition of apigenin to cell culture did not further reduce expression level of the SASP, as evidenced by the case of key SASP factors such as IL6, CXCL8, IL1α, and ILβ, suggesting that apigenin‐caused inhibitory effect of the SASP was mainly through targeting PRDX6. S5A,B S5C

Many soluble SASP factors secreted by senescent stromal cells favor malignant changes of cancer cells, acting as a significant force that fuels tumor progression as demonstrated by previous studies.^[26, 44, 45, 46^] Herein, we examined the capacity of SASP factor‐containing CM in promoting cancer cell proliferation, a basic feature of activated stroma in the tumor microenvironment (TME). The stromal cell line PSC27 was induced to become senescent by BLEO, a chemotherapeutic agent frequently used in cancer treatment, in the presence or absence of apigenin, with CM collected 8–10 d afterward and used to incubate prostate cancer (PCa) cell lines (PC3, DU145, M12, LNCaP) (Figure5A). Not surprisingly, proliferative potential was substantially enhanced in all PCa lines after treatment with CM from senescent stromal cells, a tendency consistent with increased migration and invasion of these cells (Figure 5A–C). To exclude potential effects of proliferation on cell migration, we performed wound healing assays during a shorter period, 18 h after cell scratch. Apigenin exhibited a remarkable ability to inhibit cell migration, even though no significant increase in proliferation was observed within such a timeframe. However, the malignant phenotype was markedly attenuated by apigenin treatment (Figure6A–C).

Our previous studies revealed that genotoxicity‐damaged stromal cells in TME hold remarkable potential to confer resistance on remnant cancer cells, as evidenced by the SASP factors WNT16B, AREG, and EREG.^[26, 45, 47^] Whether or not the ability of the SASP factors to enhance cancer chemoresistance can be reduced by apigenin remains unknown. We found that the viability of cancer cells was markedly elevated upon co‐culture with CM derived from senescent stromal cells, but diminished almost to the basal level as compared to their normal stromal cell counterparts upon treatment with apigenin (Figure 5D). The capacity of apigenin in counteracting chemoresistance was also in line with the distinct shift of cell survival curves, which was based on the percentage of cell viability across MIT concentrations in the range of 0.1–10 µM, a clinically relevant window of circulating concentrations in cancer patients^[48^] (Figure 5E,F). Such decreased cancer cell viability may be explained by the elevated pro‐apoptotic effect of apigenin on cancer cells via its SASP‐dampening effects, as reflected by the increased activity of caspase 3/7 (Figure 5G). We used docetaxel (DTX), another chemotherapeutic agent, to further confirm the capacity of apigenin in preventing chemoresistance. Replacement of MIT with DTX largely recapitulated the effects of MIT by decreasing apoptosis and increasing chemoresistance, while both were considerably suppressed by apigenin (Figure S6B–D, Supporting Information). These results suggest that senescent stromal cell‐derived CM conferred cancer cells with remarkable therapeutic resistance, a tendency that was attenuated upon apigenin treatment of stromal cells.

Metformin, a widely used antidiabetic drug, is a repurposable agent and can exert multiple beneficial effects including reduction of the SASP without causing substantial cytotoxicity to senescent cells,^[49, 50^] thus holding the potential as a senomorphic agent. The results from our metformin‐involving assays generally phenocopied those derived from apigenin assays in lowering cancer cell malignancy by downregulating the activity of stromal cell CM, indicating a common pattern shared by both agents (Figure S6E–H, Supporting Information). Thus, pharmacologically targeting SASP development with a senomorphic agent such as apigenin, can substantially deprive cancer cells of these stromal CM‐conferred gain‐of‐functions, specifically resistance to a chemotherapeutic drug. These findings imply the possibility of developing relevant strategies to improve the efficacy of current anticancer regimens.

Given the efficacy of apigenin in attenuating SASP expression in senescent stromal cells and restraining malignant phenotypes of cancer cells in vitro, it is tempting to ask if this agent holds potential for controling senescence‐related pathologies in vivo. To this end, PSC27 sublines were mixed together with PC3, a typically malignant prostate cancer line, at a pre‐optimized ratio (1:, 4) to generate tissue recombinants prior to subcutaneous inoculation to the hind flank of mice with non‐obese diabetes and severe combined immunodeficiency (NOD/SCID). Tumor sizes were determined 8 weeks later for pathological appraisal. In comparison to tumors made up of PC3 and PSC27^Naive, xenografts comprising PC3 and PSC27^SEN were remarkably enlarged. However, treatment of PSC27^SEN cells with apigenin prior to tissue recombinant construction substantially reduced tumor volumes (P < 0.01). Notably, the efficacy of apigenin was largely reproduced by rapamycin, which exerted a durable in vivo effect even when senescent human cells were treated only once by such a SASP‐inhibiting agent in vitro (Figure S7A, Supporting Information).^[51^]

To simulate clinical settings, we designed a preclinical regimen involving genotoxic agents and/or apigenin (Figure 6A, Figure S7B, Supporting Information). Following a 2‐week period after subcutaneous implantation with generally observable tissue recombinant uptake in host animals, a single dose of placebo or MIT was administered at the beginning of the 3rd, 5th, and 7th week, with apigenin given 7 days after each of these administrations, until the end of a 8‐week regimen (Figure S7B, Supporting Information). Although no significant benefits were observed in the apigenin group, MIT administration caused remarkable tumor shrinkage (57.8% reduction in volume), confirming the effectiveness of MIT as a chemotherapeutic agent. When apigenin was administered after MIT, we noticed an additional reduction in tumor size by 51.1%, corresponding to a total decrease of 74.9% compared with the placebo group. We next queried whether cellular senescence occurred in xenografts. As expected, in the TME of PC3/PSC27 recombinant tumors, stromal cells exhibited a markedly elevated expression of canonical SASP factors including IL‐1α/IL‐6, CXCL8, MMP1/MMP3, ANGPTL4 and AREG, a pattern arising in parallel with the upregulation of typical senescence markers p16^INK4a and p21^CIP1 in the MIT group, indicating senescence induction in vivo and SASP expression upon chemotherapeutic treatment (Figure 6C,D, Figure S7C, Supporting Information). Moreover, histologic staining indicated enhanced SA‐β‐Gal positivity in xenografts of mice exposed to MIT, a feature in sharp contrast to apigenin, which neither promoted nor suppressed cellular senescence (Figure 6E), a tendency in line with our results acquired from in vitro assays (Figure 2A). The alterations were predominantly observed in stromal cells instead of adjacent cancer cells, indicating the potential of residual cancer cell repopulation and resistance acquisition in treatment‐damaged TME. However, upon delivery of apigenin, SASP expression was dampened, as evidenced by data from transcript‐based quantitative assays (Figure S7D, Supporting Information).

We next interrogated how the expression of the SASP induces therapeutic resistance of cancer cells. To explore underlying mechanisms, we dissected tissues from animals 7 days after treatment initiation, a time point just prior to the emergence of resistant colonies. In contrast to the placebo group, MIT treatment per se enhanced DDR signaling (as measured by γH2AX) and caused cell death particularly apoptosis (as measured by caspase 3 cleavage (CCL), an apoptosis marker) in cancer cells, when apigenin was combined with MIT, while the administration of apigenin alone triggered neither alteration, implying limited effectiveness of apigenin when used as a single agent for tumor treatment (Figure 6F–H). Upon co‐administration of apigenin and MIT, intensities of both DNA damage and apoptosis were further elevated, indicating increased cytotoxicity upon combinatorial therapy. Upon ELISA analysis of the circulating SASP factors, AREG, and EREG, we found MIT treatment resulted in elevated circulating levels of these factors, a tendency that was essentially reversed upon apigenin administration (Figure 6I).

As an additional point, it remains undetermined whether PRDX6 is functionally involved in regulation of the SASP in cancer cells per se. To this end, we generated PC3 sublines, wherein PRDX6 was depleted by shRNAs (Figure S7E, Supporting Information). The growth of xenografts comprising only PC3 was measured in a way similar to that of tumor composed of both PSC27 and PC3. However, we noticed that apigenin treatment did not confer significant benefits on tumor regression (Figure S7F, Supporting Information), suggesting stromal cell‐related sensitivity to apigenin cannot be simply reproduced by cancer cells. Although there seems to be a trend for PC3 tumors depleted of PRDX6 to be smaller upon apigenin treatment, the changes were not statistically significant (P > 0.05). Our datasets suggest that in response to MIT‐mediated chemotherapy, cancer cells also become senescent, but they do not express a typical SASP, as the expression levels of a number of key SASP factors such as IL1α, IL6, CXCL8, and MMP1 remained largely unchanged (Figure 6C,D). Moreover, we found that the DNA damage/apoptosis indices of tumors composed entirely of cancer cells, failed to show significant changes upon apigenin treatment (Figure S7G, Supporting Information). Thus, the tumor‐regression effect conferred by apigenin is not detectable in absence of stromal cells, suggesting a critical role of senescent stromal cells in the development of cancer resistance during chemotherapy.

To establish the safety and effectiveness of our therapeutic strategy, a systematic appraisal of the physiology of experimental animals was performed. Of note, gross assessment data indicated that both single and co‐administration were well tolerated, as evidenced by the absence of body weight loss between different groups during the entire regimen (Figure, Supporting Information). Further, no remarkable fluctuation was observed in the serum level of creatinine, urea, or metabolic status including alkaline phosphatase (ALP) and alanine aminotransferase (ALT) activities, which represent a set of biochemical measurements of the function of major organs such as liver and kidney (Figure, Supporting Information). Hence, co‐administration of a senomorphic agent and canonical chemotherapy holds the potential to achieve maximal anticancer effects while minimizing severe systemic toxicities. S8A S8B

In the tissue microenvironment, the SASP is a major executor of the paracrine effects of senescent cells and mediates various local and systemic biological effects, including those underlying pathologies and impacting overall healthspan.^[34^] To determine the efficacy of apigenin in regulating senescence‐related in vivo changes including organismal aging, we chose to employ irradiation‐challenged mice for age‐related studies. Briefly, wild type (WT) mice were exposed to a sublethal dose of whole body irradiation (WBI) to induce substantial senescence within tissues, followed by senotherapeutic intervention with apigenin or vehicle, the latter a therapeutic control (Figure7A). As a result, animals experiencing WBI manifested an aberrant physiological profile, as exemplified by apparently graying fur (Figure 7B). However, this condition was markedly improved upon treatment by apigenin.

Age‐dependent expansion of the pulmonary alveolus, an important tissue structural alteration that contributes to pulmonary dysfunction, was found in WBI‐treated mice, but much attenuated in animals receiving apigenin treatment (Figure 7C). Our data suggest a benefit of apigenin supplementation in preventing the appearance of age‐related pulmonary abnormalities, as changes in the composition of the airways and the alveoli may cause respiratory dysfunction and eventually promote chronic lung disorders during chronological aging.^[52^] Upon histological assessment of spleens of aged mice, we observed atrophy and disarrangement of white pulp, the major immune organ comprising various immune cell subtypes, including DC cells, CD4⁺, CD8⁺ T cells, and CD19⁺ B cells, abnormal changes of which may contribute to age‐associated immune dysfunction (Figure S9A, Supporting Information). However, apigenin exhibited capacity to improve the structural profile of spleen tissues by preventing such abnormalities. Furthermore, the density of spleen‐derived immune cell populations, particularly CD4⁺ T and CD19⁺ B cells, which mainly reside in the periarteriolar lymphocyte sheath (PALS) and the region between PALS and marginal sinuses, respectively,^[53^] markedly decreased in the WBI group, but the change was reversed upon apigenin treatment.

Subsequent analysis indicated the emergence of senescent cells in several major organs, as demonstrated by increased SA‐β‐Gal positivity in kidney and spleen tissues of these animals post WBI‐treatment (Figure 7D, Figure S9B, Supporting Information). As compared with vehicle‐treated mice, the tendency of SA‐β‐Gal staining positivity remained largely unchanged in the apigenin group, suggesting treatment with such a senomorphic agent cannot prevent cellular senescence in these animals. We noted that apigenin failed to downregulate the expression of key senescence markers including p16^INK4a and p21^CIP1 (Figure 7E). The data were further validated by immunofluorescence analysis of p21^CIP1 expression in the tissue microenvironment (Figure S9C, Supporting Information). However, in contrast, a subset of SASP factors including IL6, CXCL8, IL1α, IL1β, CXCL1/3, and MMP1/3/9, which were significantly upregulated after exposure of mice to WBI, exhibited remarkable decline in animals treated with apigenin, as exemplified by the data from analysis of kidney tissues (Figure 7E). This observation was largely consistent with data acquired from mice involved in chemotherapeutic regimens (Figure S7D, Supporting Information). We examined the biospecimens at protein level. Our data indicated that the upregulation of the canonical SASP marker IL6 in tissues such as the kidney after animals underwent WBI treatment, was partially reversed in animals treated by apigenin (Figure S9D,E, Supporting Information). Upon performance of ELISA assays to measure the circulating levels of AREG and EREG, we observed an apparent elevation of circulating concentration of these SASP factors in irradiation‐challenged animals, a tendency that was counteracted upon apigenin administration (Figure S9F,G, Supporting Information).

Next, we assessed whether preclinical interventions altered the physical function in these animals. Unsurprisingly, mice experiencing WBI manifested reduced exercise performance and muscle strength, as gauged by grip strength, hanging endurance, rotard rod duration, and balance cross tests (Figure 7F–I). However, declines in each of these activities were partially but significantly reversed upon treatment of prematurely aged animals with apigenin, as compared with the vehicle group. We questioned whether apigenin could generally improve health conditions including short term memory and anxiety. Notably, we observed a basic recovery of short‐term memory in aged animals that received apigenin treatment, as evidenced by data from Y‐maze tests (Figure 7J). Data from open field assays indicated that apigenin prolonged duration of exploring the central zone of a wide arena, suggesting anxiety of aged mice was alleviated in aged mice compared with the vehicle group (Figure 7K). The data suggest that apigenin administration represents an effective approach to alleviate physical dysfunction and avoid cognitive impairment during premature aging, promoting overall health.

To establish the safety of our interventional strategy in these immunocompetent animals (C57BL/6J), a systemic and routine evaluation of blood cell counts and serum biochemical indices was performed. As a result, we observed a slight decrease in the counts of both white blood cells (WBCs) and red blood cells (RBCs) in WBI‐treated mice (Figure, Supporting Information). Although the difference appeared insignificant, it reflected a partial decline of hematopoietic capacity and immune competency. Upon apigenin administration, these indices remained generally unchanged, validating the safety of apigenin treatment in the hematopoiesis system (Figure, Supporting Information). We found no obvious abnormality in the serum levels of creatinine, urea, and activities of ALP and ALT, liver, and renal biochemical parameters (Figure, Supporting Information). Altogether, our preclinical data support that therapeutic regimens involving a senomorphic agent, such as apigenin, in prematurely aged mice, hold potential to alleviate dysfunction without causing systemic cytotoxicity. S9H–J S9H‐L S9M–P

The primary normal human prostate stromal cell line PSC27 was a gift from Dr. Peter Nelson (Fred Hutchinson Cancer Center) and maintained in stromal complete medium as described previously.^[26^] Human fetal lung stromal lines WI38 and IMR90 were from ATCC and cultured with F‐12K medium supplemented with 10% FBS. Prostate cancer epithelial cell lines PC3, DU145, and LNCaP and the human embryonic kidney line 293T were from ATCC and cultured with RPMI 1640 (10% FBS). The prostate cancer epithelial line M12 was kindly provided by Dr. Stephen Plymate (University of Washington), which was originally derived from the benign line BPH1 but phenotypically neoplastic and metastatic.^[72^] All lines were tested for mycoplasma contamination and authenticated with STR assays.

Stromal cells were grown until 80% confluent (CTRL) and treated with 50 µg/mL bleomycin (BLEO). Cells were rinsed briefly with PBS and maintained for 8–10 days prior to performing examinations. For drug screening of potential senolytics, natural candidates (totally 66 in the NMA library) (Selleckchem L8300‐TCM subset) were tested each at 3 µg mL⁻¹ on survival of 5.0 × 10³ control and senescent cells for 3 d. Subsequent evaluation of effects of these candidate agents (each applied at 1 µg mL⁻¹) was conducted to test the extent of SASP inhibition (assay for senomorphics). For those with the potential to act as effective and safe senomorphics, further assessments were done. The phytochemical agent apigenin was examined in the range of 1–10 µM, with 10 µM determined to be the lowest effective concentration for senomorphics activity. The positive control senolytics ABT‐263 (navitoclax) and PCC1 were applied at 1.25 and 50 µM, respectively. The small molecule inhibitor of p38MAPK, SB203580, and the PRDX6‐PLA2 inhibitor MJ33 were employed at 10 µM, to treat senescent cells in culture before collection for lysis and expression analysis.

Full length human PRDX6 gene sequence was cloned into pHBLV‐U6‐MCS‐CMV‐ZsGreen‐PGK‐Puro (Hanbio Biotechnology). Small hairpin RNAs (shRNAs) targeting PRDX6 (1#, sense strand 5′‐ GATCCGCTCCCAACTTTGAGGCCAATACCATTCAAGAGATGGTATTGGCCTCAAAGTTGGGAGCTTTTTTG‐3′, anti‐sense strand 5′‐ AATTCAAAAAAGCTCCCAACTTTGAGGCCAATACCATCTCTTGAATGGTATTGGCCTCAAAGTTGGGAGCG‐3′, 2#, sense strand 5′‐ GATCCGATTGCCCTTTCAATAGACAGTGTTTTCAAGAGAAACACTGTCTATTGAAAGGGCAATCTTTTTTG‐3′, anti‐sense strand 5′‐ AATTCAAAAAAGATTGCCCTTTCAATAGACAGTGTTTCTCTTGAAAACACTGTCTATTGAAAGGGCAATCG‐3′, scramble, sense strand 5′‐ GATCCGTTCTCCGAACGTGTCACGTAATTCAAGAGATTACGTGACACGTTCGGAGAATTTTTTC‐3′, anti‐sense strand 5′‐ AATTGAAAAAATTCTCCGAACGTGTCACGTAATCTCTTGAATTACGTGACACGTTCGGAGAACG‐3′) were cloned in pLKO.1‐Puro vector (Addgene). Upon production by 293T cells, lentivirual titers were adjusted to infect ≈90% of cells. Stromal cells were infected overnight in the presence of polybrene (8 µg mL⁻¹), allowed to recover for 48 h, and selected for 72 h before being used for further analysis.

Total RNA samples were prepared from senescent PSC27 cells cultured with regular DMEM or DMEM containing apigenin (10 µM) for 3 consecutive days. Sample quality was validated by Bioanalyzer 2100 (Agilent), and RNA was subjected to sequencing by Illumina NovaSeq 6000 with gene expression levels quantified by the software package RSEM (https://deweylab.github.io/RSEM/↗). Briefly, rRNAs in the RNA samples were eliminated using the RiboMinus Eukaryote kit (Qiagen), and strand‐specific RNA‐seq libraries were generated using the TruSeq Stranded Total RNA preparation kit (Illumina, San Diego, CA, USA) according to the manufacturer's instructions before deep sequencing.

Pair‐end transcriptomic reads were mapped to the reference genome (GRCh38.p14) (Genome Reference Consortium Human Build 38, INSDC Assembly GCA_0 00001405.28, 12.2013) (http://asia.ensembl.org/Homo_sapiens/Info/Index↗) (ensembl_105) with reference annotation from GENCODE v42 using the Bowtie tool. Duplicate reads were identified using Picard tools (1.98) script mark duplicates (https://github.com/broadinstitute/picard↗), and only non‐duplicate reads were retained. Reference splice junctions are provided by a reference transcriptome (Ensembl build 73).^[73^] FPKM values were calculated using Cufflinks, with differential gene expression called by the Cuffdiff maximum‐likelihood estimate function.^[74^] Genes of significantly changed expression were defined by a false discovery rate (FDR)‐corrected P value < 0.05. Only Ensembl genes 73 of status “known” and biotype “coding” were used for downstream analysis.

Reads were trimmed using Trim Galore (v0.6.1) (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/↗) and quality assessed using FastQC (v0.10.0) (http://www.bioinformatics.bbsrc.ac.uk/projects/fastqc/↗). Differentially expressed genes were subsequently analyzed for enrichment of biological themes using the DAVID bioinformatics platform (https://david.ncifcrf.gov/↗) and the Ingenuity Pathways Analysis program (http://www.ingenuity.com/index.html↗). Raw data of bulk RNA‐seq were deposited in the NCBI Gene Expression Omnibus (GEO) database under the accession code GSE273159↗.

Venn diagrams and associated empirical P‐values were generated using the USeq (v7.1.2) tool IntersectLists.^[75^] The t‐value used was 22008, as the total number of genes of status “known” and biotype “coding” in Ensembl genes 73. The number of iterations used was 1000.

For each gene, the FPKM value was calculated based on aligned reads, using Cufflinks.^[74^] Z‐scores were generated from FPKMs. Hierarchical clustering was performed using the R package heatmap.2 and the distfun = “pearson” and hclustfun = “average”.

Profiling of ATM‐ and p38MAPK‐interactive molecules was performed with BioGRID (v4.4.237), a biomedical interaction repository with data compiled through comprehensive curation efforts and used as a public database archiving and disseminating genetic and protein interaction data from all major model organisms and humans (thebiogrid.org).^[76, 77^] BioGRID searches 85467 publications for 2800228 protein and genetic interactions, 31144 chemical interactions, and 1128339 post‐translational modifications from major model organism species, with the human selected as the target species throughout this study.

To assess whether apigenin targets PRDX6 protein, CETSA‐immunoblot experiments were carried out. CETSA is used to examine the binding efficiency of drug and protein. Briefly, PSC27 cells in culture were trypsinized and precipitated by centrifugation, lysed on ice with RIPA buffer containing protease inhibitor cocktail, and centrifuged at 14 000 × g at 4 °C for 20 min. Each lysate was then equally aliquoted into several parts in EP tubes, and incubated with apigenin at room temperature for 1 h. Samples were heated for 3 min under different temperatures (30, 35, 40, 45, and 50 °C). The precipitated proteins were separated from the soluble fraction by centrifugation, and boiled at 95 °C for 10 min. Samples were subsequently used for immunoblot analysis.

DARTS assays were performed to measure the interaction stability of the small molecule apigenin with its targeting protein(s), and performed according to a previously reported protocol.^[78^] PSC27 cells were lysed with IP lysate buffer on ice with protease and phosphatase inhibitors, before being centrifuged at 4 °C and quantified by the BCA method. Lysates were allowed to rapidly warm up to room temperature and DMSO or apigenin (100 µM) was added to the lysates for 1 h incubation in a rotator. Samples were digested by pronase (1, 300 (w/w), 10 µg mL⁻¹) for 30 min at room temperature. Digestion was stopped by 0.5 M EDTA (pH 8.0), with samples boiled in loading buffer for SDS‐PAGE and immunoblot evaluation, or directly subjected to MS analysis for proteomic profiling. The resultant proteomic data of DARTS were deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org↗) through the iProX partner repository with a unique dataset identifier PXD055761.

The binding affinity of apigenin to recombinant human PRDX6 (rhPRDX6) was determined at 22 °C with a Biacore T200 instrument equipped with CM5 sensor chips (GE Healthcare).^[79^] Briefly, rhPRDX6 protein (SinoBiological) was diluted in sodium acetate solution (pH 4.5) to a final concentration of 20 µg mL⁻¹. PRDX6 protein was immobilized on a CM5 sensor chip (GE Healthcare) by amine, allowed to couple with the final immobilization densities of 8000 RU. Immobilized PRDX6 protein was then applied to capture apigenin. The running buffer contained PBS‐P (10 mM sodium phosphate, 150 mM NaCl, 0.05% surfactant P20, pH 7.3) and 5% DMSO. The pH was changed from 7.0 to 5.5 to adjust the binding affinity under several pH environments. Experiments were performed at 25 °C, with a series of concentrations (12.5, 6.25, 3.125, 1.56, and 0.78 µM) of apigenin analyzed at a flow rate of 30 µL min⁻¹, with an injection time of 60 s and a dissociation time of 60 s in each binding cycle. A blank immobilization was performed on one channel surface of the chip to adjust the binding response. Association and dissociation constants were acquired using the Biacore T200 Evaluation software (v3.0 GE Healthcare). Data was exported to GraphPad Prism to plot the final curves.

PSC27 cells were lysed in IP lysate buffer on ice with protease and phosphatase inhibitors, then centrifugated at 4 °C for 10 min, the supernatant was collected and subjected to protein quantification by BCA method. Then, the supernatant was divided into 4 equal parts, wherein 2 of them were pretreated with 10 and 20‐fold concentrations of unlabeled apigenin, respectively, at room temperature for 30 min. Subsequently, upon treatment by biotin or biotin‐apigenin (100 µM) with unlabeled apigenin in the IP buffer overnight at 4 °C. The next day, 30 µL of streptavidin‐agarose beads (Sigma–Aldrich) were added to bind for 2 h at 4 °C. After rinsed thrice with IP buffer, proteins in the beads were eluted, separated by SDS‐PAGE and stained with silver or directly subjected to sample separation for MS analysis.

Peptides were resolved in 0.1% formic acid, with the BCA protein quantification kit employed to measure peptide concentration. Samples were separated utilizing an analytic column (75 µm × 15 cm, 3 µm, 100Å) (C‐18, Thermo Scientific) on a nanoflow HPLC Easy‐nLC 1200 system (Thermo Scientific), using a 120 min LC gradient at 300 nL min⁻¹. Buffer A comprises 0.1% (v/v) formic acid in H₂O and Buffer B comprises 0.1% (v/v) formic acid in 80% acetonitrile. The solution gradient was as follows, 1–5% B in 6 min, 5–26% B in 88 min, 26–40% B in 22 min, 40–100% B in 5 min, 90% B in 5 min. MS analyses were conducted on a Q Exactive HF‐X mass spectrometer (Thermo Scientific). MS scan was performed in positive ion mode with spray voltage at 1900 V, with ion transfer tube temperature at 320 °C. Xcalibur software was used to acquire profile spectrum data in a data‐dependent acquisition pattern (DDA). The MS1 full scan was set at a resolution of 60000 @ m/z 200, AGC target 3e6, and maximum IT 50 ms by orbitrap mass analyzer (350‐1800 m/z), followed by ‘top 15′ MS2 scans generated by HCD fragmentation at a resolution of 15000 @ m/z 200, AGC target 1e5 and maximum IT 30 ms. Isolation window was set at 1.6 m/z. The normalized collision energy (NCE) was set at NCE 28%, with the dynamic exclusion time as 45 s. Precursors with charge 1, 7, 8, and > 8 were excluded from MS2 analysis.

All raw mass spectrometric data were analyzed using MaxQuant 2.0.3.0 against the human Swiss‐Prot database comprising 20423 sequences (2023). Carbamidomethyl of cysteine was selected as a fixed modification. Oxidized methionine, protein N‐term acetylation, lysine acetylation, asparagine And glutamine (NQ) deamidation were selected as variable modifications. Enzyme specificity was set as trypsin. The tolerances of first search and main search (Andromeda search engine) for peptides were set at 20 ppm and 6 ppm, respectively. Minimum cutoff for peptide length was set at 7 amino acids, with maximum permissible missed cleavage set at 2. Maximal FDR for peptide spectral match, proteins, and site was set to 0.01. A minimum of 2 sequence‐unique peptides was required for identification. Feature matching between runs was done with a retention time window of 2 min, with the label‐free quantitation (LFQ) function enabled. The MaxQuant peptide and protein quantification results from the “peptides.txt” and“proteinGroups.txt” files were imported into Perseus software (version 1.5.1.6) for further analysis. Statistical significance between the groups was evaluated using Student's t tests (p < 0.05), Proteins were defined as differentially expressed if the ratios were ≥ 1.5 or ≤ 0.67 in the experiment group compared with the control.

The amino acid sequence of PRDX6 (code, P30041↗) was acquired from UniProt. The structures of apigenin were derived from PubChem (https://pubchem.ncbi.nlm.nih.gov/↗) and then were energy‐minimized by Molecular Operating Environment (MOE) (v2019.01.02), using MMFF94 force field with the gradient convergence set to 0.1 kcal mol⁻¹. The X‐ray diffraction microscopy structure of PRDX6 (PDB ID, 5B6M) was obtained from RCSB Protein Data Bank (PDB, https://www.rcsb.org/↗). The water molecules farther than 4.5 Å from the receptor were removed. Subsequently, all protein structures were prepared by the built‐in MOE structure preparation and Protonate3D software tools with the default parameters. The binding site of each protein was generated by the SiteFinder module in MOE. The placement method was set to Triangle Matcher with the London dG scoring function, with the refinement score algorithm set to GBVI/WSA dG. Ultimately, the binding mode with the best docking score was selected for the next analysis. The docking score denotes the affinity between the protein receptor and the docking ligand. The lower docking score suggests better affinity.^[80^]

For confirmation of the binding of apigenin to wild type and mutant PRDX6, PRDX6 constructs with GFP protein or mutant (Cys91 or/and Glu210 to alanine) cloned in pcDNA3.1‐PRDX6‐EGFP‐C2 were emgineered and overexpressed in 293T cells, which were lysed 48 h afterward for MST analysis using Nano Temper Monolith NT Labelfree instrument (Nano Temper Technologies). A series of concentrations of apigenin (200, 100,50, 25, 12.5, 6.25, 3.12, 1.6, 0.8, 0.4 and 0.2 µM) were mixed with PRDX6 protein solution. After incubation at room temperature for 5 min, the mixtures were loaded into MO‐AK002 capillaries. MST assessments were carried out with 20% MST power and 20% LED power with nano‐BLUE pattern. MST datasets were processed using the NT analysis 1.5.41 software (Nano Temper Technologies).

Peroxiredoxin (Prdx) activity was determined by Hydrogen Peroxide Assay Kit (Beyotime). Briefly, PRDX was overexpressed in PSC27 cells, which were then lysed and quantified by BCA. Cell lysates were incubated with a series of concentrations of apigenin at room temperature for 10 min. H₂O₂ was added to a final concentration of 50 µM, with samples incubated for 5 min. H₂O₂ assay agents were added, with samples incubated for 30 min at room temperature, the remaining H₂O₂ was determined by assessment of the absorbance value at 560 nm. N‐acetylcysteine (NAC) was used as a positive control.

Phospholipase A2 activity was determined according to the manufacturer's instructions (EnzChek Phospholipase A2 kit, Invitrogen). Briefly, PSC27 cells overexpressing PRDX6 were harvested, lysed, and quantified by BCA. For standard curve preparation, PLA2 stock solution (500 U/mL) was diluted with 1 × reaction buffer to make a series of concentrations (0–10 units/mL) of PLA2. For sample measurement, equal amounts of protein were diluted with 1 × PLA2 reaction buffer (1, 1, v/v) up to 50 µL volume, 50 µL of the substrate liposome mix were added to each black microplate for incubation with a total volume of 100 µL for 10 min. The fluorescence of standards, controls, and samples was measured at Ex460 nm/Em515 nm.

Whole cell lysates were prepared using RIPA lysis buffer supplemented with protease/phosphatase inhibitor cocktail (Biomake). Nitrocellulose membranes were incubated overnight at 4 °C with primary antibodies, and HRP‐conjugated goat anti‐mouse or ‐rabbit served as secondary antibodies (Vazyme). For immunofluorescence analysis, cells were fixed with 4% formaldehyde and permeabilized before incubation with primary and secondary antibodies, each for 1 h. Upon counterstaining with DAPI (0.5 µg mL⁻¹), samples were examined with an Imager A2. Axio (Zeiss) upright microscope to analyze specific gene expression.

Levels of intracellular ROS were determined using a ROS Assay Kit (Beyotime) which employs dichloro‐dihydro‐fluorescein diacetate (DCFH‐DA) as a probe. Briefly, cells were cultured in 6‐well plates for 24 h at 37 °C and were then washed twice with serum‐free medium. Medium containing 10 µM DCFH‐DA was added. Cells were then incubated for 20 min at 37 °C, with light avoided during incubation. After incubation, the cells were washed thrice with serum‐free medium, then observed and photographed using a fluorescence microscope (Nikon). The fluorescence intensity was measured using ImageJ software (v1.51, NIH).

For proliferation assays of cancer cells, 2 × 10⁴ cells were dispensed into 6 well‐plates and co‐cultured with stromal cell‐derived conditioned medium (CM). Three days later, cells were digested and counted with hemocytometer. For migration assays, cells were added to the top chambers of transwells (8 µm pore), while stromal CM was added to the bottom. Migrating cells in the bottom chambers were stained by crystal violet 12–24 h later, with samples examined with an Observer A1. Axio (Zeiss) inverted microscope. Invasion assays were performed similarly with migration experiments, except that transwell were coated with basement membrane matrix (phenol red free, Corning). Alternatively, cancer cells were subject to wound healing assays performed in 6‐well plates, with healing patterns graphed with a brightfield microscope. For chemoresistance assays, cancer cells were incubated with stromal CM, with the chemotherapeutic agent MIT (mitoxantrone) provided in wells for 3 days at each cell line's IC50, a value experimentally predetermined. Cell viability was assayed by a CCK‐8 kit, with the absorbance at 450 nm measured using a microplate reader.

Animals were maintained in a specific pathogen‐free (SPF) facility, with NOD/SCID (Model Animal Research Center of Nanjing University) mice at an age of approximately 6 weeks (≈20 g body weight) used. Ten mice were incorporated in each group, with xenografts subcutaneously generated at the hind flank under isoflurane inhalation. Stromal cells (PSC27) were mixed with cancer cells (PC3) at a ratio of 1: 4 (i.e., 250 000 stromal cells admixed with 1 000 000 cancer cells to make tissue recombinants before implantation in vivo). Animals were sacrificed at 2–8 weeks after tumor implantation, according to tumor burden or experimental requirements. Tumor growth was monitored weekly, with tumor volume (v) measured and calculated according to the tumor length (l), width (w), and height (h) by the formula, v = (π/6) × ((l+w+h)/3)^[3, 81^] Freshly dissected tumors were either snap‐frozen or fixed to prepare FFPE samples. Resulting sections were used for IHC staining against specific antigens or subject to hematoxylin/eosin staining.

For chemoresistance studies, animals received subcutaneous implantation of tissue recombinants as described above and were given standard laboratory diets for 2 weeks to allow tumor uptake and growth initiation. Starting from the 3^rd week (tumors reaching 4–8 mm in diameter), MIT (0.2 mg kg⁻¹ doses) or vehicle controls were administered through intraperitoneal injection (therapeutic agents via i.p. route), on the 1st day of 3rd, 5th and 7th weeks, respectively. Apigenin was delivered 7 days after each MIT or vehicle administration, at doses of 10 mg kg⁻¹. Upon completion of the 8‐week therapeutic regimen, animals were sacrificed, with tumor volumes recorded and tissues processed for histological evaluation.

At the end of chemotherapy and/or targeting treatment, animals were anesthetized and peripheral blood was gathered via cardiac puncture. Blood was transferred into a 1.5 ml Eppendorf tube and kept on ice for 45 min, followed by centrifugation at 9000 x g for 10 min at 4 °C. Clear supernatants containing serum were collected and transferred into a sterile 1.5 ml Eppendorf tube. All serum markers were measured using dry‐slide technology on IDEXX VetTest 8008 chemistry analyzer (IDEXX). About 50 µl of the serum sample was loaded on the VetTest pipette tip before securely fit on the pipettor and the manufacturer's instructions were followed for further examination.

All animal experiments were performed in compliance with NIH Guide for the Care and Use of Laboratory Animals (National Academies Press, 2011) and the ARRIVE guidelines, and were approved by the Institutional Animal Care and Use Committee (IACUC) of the Shanghai Institute of Nutrition and Health (no. SINH‐2023‐SY‐1/SINH‐2024‐SY‐1), Chinese Academy of Sciences. Every effort was made to minimize the number of animals used and to reduce potential suffering.

For SA‐β‐Gal staining, frozen sections were dried at 37 °C for 20–30 min before being fixed for 15 min at room temperature. The frozen sections were washed thrice with PBS and incubated with SA‐β‐Gal staining reagent (Beyotime) overnight at 37 °C. After completion of SA‐β‐Gal staining, sections were stained with eosin for 1–2 min, rinsed under running water for 1 min, differentiated in 1% acid alcohol for 10–20 s, and washed again under running water for 1 min. Sections were dehydrated in increasing concentrations of alcohol and cleared in xylene. After drying, samples were examined under a bright‐field microscope.

Preclinical specimens from mouse tissues were fixed overnight in 10% neutral‐buffered formalin and processed for paraffin embedding. For histological assessment, standard staining with hematoxylin/eosin was performed on sections of 5–8 µm thickness cut from each specimen block. For immunohistochemistry, tumor specimens were first fixed in 4% paraformaldehyde and embedded in paraffin. For appraisal, tissue sections were cut from FFPE chunks, de‐paraffinized, and incubated in 10 mM sodium citrate buffer (pH 6.5) at 95 °C for 40 min for antigen retrieval. Peroxidase activity was quenched with 3% H₂O₂ and tissues were blocked in 5% bovine serum albumin for 30 min, before being incubated with primary antibodies (e.g., against IL6 or cleaved caspase 3 [CCL3]), for SASP expression appraisal or cell apoptosis evaluation, respectively) overnight at 4 °C. After 3 washes with PBS, tissue sections were incubated with biotinylated secondary antibody (1, 200 dilutions, Vector Laboratories) for 1 h at room temperature then washed thrice, after which streptavidin‐horseradish peroxidase conjugates (Vector Laboratories) were added and the slides incubated for 45 min. DAB solution (Vector Laboratories) was then added and slides were counterstained with hematoxylin. Alternatively, frozen tissue sections were used for immunofluorescence staining to probe target protein expression (e.g., p21 induction) in animals receiving various preclinical treatments, with counterstaining performed with DAPI. Images were then captured with epifluorescence microscope.

For routine blood examination, 100 µL fresh blood was acquired from each animal and mixed with EDTA immediately. The blood samples were analyzed with Celltac Alpha MEK‐6400 series hematology analyzers (Nihon Kohden). For serum biochemical analyses, blood samples were collected and clotted for 2 h at room temperature or overnight at 4 °C. An aliquot of approximately 50 µl serum was subject to analysis for creatinine, urea, alkaline phosphatase (ALP), and alanine transaminase (ALT) by an VetTest 8008 chemistry analyzer (IDEXX) as reported previously.^[81^] For blood cell tests, 50 µl fresh blood was collected from each mouse and mixed with EDTA immediately. Circulating levels of hemoglobin, white blood cells, lymphocytes, and platelets of the blood samples were analyzed by Sysmex XN‐1000 series automated hematology analyzers (Sysmex, XN‐1000).

All animal experiments were conducted in compliance with the NIH Guide for the Care and Use of Laboratory Animals (National Academies Press, 2011) and the ARRIVE guidelines, and were approved by the IACUC of Shanghai Institute of Nutrition and Health, Chinese Academy of Sciences. For each preclinical regimen, animals were monitored for conditions including hypersensitivity (changes in body temperature, altered breathing, and ruffled fur), body weight, mortality, and changes in behavior (i.e., loss of appetite and distress), and were disposed of appropriately according to the individual pathological severity as defined by relevant guidelines.

The open‐field test was employed to assess the locomotor activity and anxiety of experimental animals when placed in an unfamiliar environment.^[82^] The open‐field chambers (l × w × h, 60 cm × 60 cm × 60 cm) were made of Plexiglas with a non‐reflective square base. Total distance in the whole open‐field chamber, distance traveled in the center zone (30 cm × 30 cm, 50% of the total area) and in the edge zone (20% of the total area, along edges) were recorded. Less activity in the center zone indicated anxiety‐like behaviors.

Spontaneous alternation behavior indicates the tendency for mice to alternate their (conventionally) nonreinforced choices on successive opportunities.^[83^] The Y‐maze spontaneous alternation was employed to assess short‐time spatial recognition and working memory in mice by measurement of spontaneous alternations.^[84^] The maze consisted of three arms, with each arm 31 cm long, 5 cm wide, and 10 cm high. Each arm had markers of different colors as distinct visual cues. After being placed individually at the center of the apparatus, mice were allowed to explore freely through the maze during an 8‐min session. Alternation was defined as successive entries into all three arms on overlapping triplet sets. The number of arm entries and alternations were recorded visually to calculate the percentage of the alternation behavior with the following formula, % Alternation = [Number of Alternations/ (Total number of arm entries ‐2)] x 100. Spontaneous alternation (%), defined as successive entries into the three arms on overlapping triplet sets, is associated with spatial short‐term memory.

All in vitro experiments were performed in triplicate, while animal studies were conducted with at least eight mice per group except for assays involving PC3‐only xenografts, wherein three mice were used per group. Data are presented as mean ± SD except where otherwise indicated. Sample size (n) for each statistical analysis was indicated wherever possible. Statistical methods used to assess significant differences with sufficient details were provided. GraphPad Prism (9.5.1) was used to collect and analyze data, with statistical significance determined according to individual settings. Cox proportional hazards regression model and multivariate Cox proportional hazards model analyses were performed with SPSS statistical software. Statistical significance was determined by unpaired two‐tailed Student's t‐tests, one‐ or two‐way ANOVA, Pearson's correlation coefficients tests, Kruskal‐Wallis, log‐rank tests, Wilcoxon–Mann–Whitney tests, or Fisher's exact tests. When ANOVA indicated significant differences, post hoc Tukey's test was applied to compare mean values. For all statistical tests, a P value < 0.05 was considered significant. The sample size for each experiment was not predetermined by statistical method, and no data were excluded from analyses.