A PDO-based epigenetic inhibitor screen identified EZH2 as a key regulator of PARPi resistance in BRCA1-deficient EOC
Epigenetic modification is reported to be necessary for chemotherapy resistance in solid tumors, and reversing epigenetic reprogramming holds considerable promise as a strategy to overcome therapeutic resistance. Therefore, we aimed to identify potential epigenetic regulators associated with PARPi resistance. A systematic screen of inhibitors of epigenetic modifications (Supplementary Fig. S1a) was conducted in PARPi-resistant PDOs (Fig. 1a and Supplementary Fig. S1b), and only tazemetostat, a specific EZH2 enzyme activity inhibitor22, effectively attenuated drug resistance in EOC, as evidenced by a significant reduction in the survival rate of PARPi-resistant PDOs. However, an inhibitor of the formation of the PRC2 complex (astemizole, which blocks EZH2-EED protein‒protein interactions)23 and inhibitors targeting other epigenetic regulators had significantly weaker inhibitory effects (Fig. 1b, c). We established two PARPi-resistant BRCA1-deficient cell lines, OVCAR8-OlaR and UWB1.289-OlaR, to characterize the mechanism of EZH2 in PARPi resistance (Fig. 1d). The half-maximal inhibitory concentration (IC50) for olaparib in OlaR cells was determined to be more than five times higher than that in the original parental cells (Fig. 1e). Western blot results confirmed that EZH2 expression was substantially increased in drug-resistant cells compared with that in parental cells or normal ovarian epithelial cells (Fig. 1f). Furthermore, an analysis of the KM-Plotter database revealed that patients with high levels of EZH2 expression had significantly worse clinical outcomes (HR = 1.26; P = 0.0025; Fig. 1g). OVCAR8 and UWB1.289 cell lines exhibit extremely poor tumorigenicity; we tested subcutaneous, orthotopic, and peritoneal dissemination models, but none of these models resulted in tumor formation. Therefore, we generated EZH2-knockdown HEYBRCA1-KO cells, established a xenograft model in nude mice and monitored tumor growth over a 4-week period to validate the function of EZH2 in vivo (Fig. 1h and Supplementary Fig. S1c). The mice were sacrificed 26 days after tumor cell transplantation, and we found that EZH2 knockdown significantly increased the sensitivity of the tumours to PARPis (Fig. 1i, j). Together, these results indicate that EZH2 is a novel driver gene for PARPi resistance in BRCA1-deficient EOC cells.
Fig. 1: A PDO-based epigenetic inhibitor screen identified EZH2 as a key regulator of PARPi resistance in BRCA1-deficient EOC.
a Schematic illustration showing the experiment used to assess the sensitivity of PDOs to drugs regulating epigenetics. b Heatmap of the drug screening results (n = 3 independent experiments). Statistical significance was determined using one-way ANOVA. Tazemetostat, a specific EZH2 enzyme activity inhibitor, significantly reduced the survival rate of PARPi-resistant PDOs. c Schematic diagram of the PDO drug screening results. d Schematic illustration showing the establishment of olaparib-resistant BRCA1-deficient cell lines (OVCAR8-OlaR and UWB1.289-OlaR). e IC50 values for PARPi-resistant and parental BRCA1-deficient EOC cells according to the results of CCK-8 assays. f Expression of the BRCA1 and EZH2 proteins in the indicated cells, as determined by Western blot. g Kaplan‒Meier analysis of the effect of EZH2 on progression-free survival according to the KM-Plotter database. Patients with high levels of EZH2 expression had significantly worse clinical outcomes (HR = 1.26; P = 0.0025). h Schematic diagram of tumor xenograft experiments using EZH2 knockdown BRCA1-deficient EOC cells. i Tumor volumes measured on the indicated days are shown (n = 6 mice per group). Statistical significance was determined using two-way ANOVA. j Image of subcutaneous transplanted tumors (n = 6 mice per group). k Tumor weights were measured and are presented as the means ± S.D. (n = 6 mice per group). Statistical significance was determined using one-way ANOVA.
EZH2 drives PARPi resistance in BRCA1-deficient EOC cells independent of the PRC2 complex
Using a truncated EZH2 mutant (lacking the methyltransferase domain), we further confirmed that EZH2 primarily mediates PARPi resistance through its enzymatic activity (Supplementary Fig. S1d–f). Additionally, we determined the expression levels of EZH2 and H3K27me3 in tumor tissues from sensitive and resistant patients by performing immunohistochemistry (IHC). Analyses of the clinical cohort confirmed that EZH2 expression was significantly increased in PARPi-resistant EOC compared with that in sensitive PARPi-sensitive EOC (Fig. 2a, b); however, the total H3K27me3 level was comparable between these two groups (Fig. 2c), and the levels of EZH2 and H3K27me3 were not significantly correlated (Fig. 2d). Moreover, western blot results confirmed that the H3K27me3 level was almost identical between resistant cells and parental cells (Fig. 2e). The above results indicate that EZH2 drives PARPi resistance in BRCA1-deficient EOC through a mechanism that does not rely on histone modification. We further confirmed the role of EZH2 in promoting PARPi resistance by systematically evaluating multiple EZH2 inhibitors with distinct mechanisms of action in BRCA1-deficient resistant cell lines (Fig. 2f–h) and PDOs (Fig. 2i, j). Although targeted degradation24 (MS1943), the inhibition of catalytic activity (tazemetostat), and inhibition of PRC2 complex disruption (astemizole) significantly reduced H3K27me3 levels, astemizole did not produce a therapeutic benefit. These findings further indicate that EZH2 promotes PARPi resistance in a PRC2 complex-independent manner. Notably, knocking down EED to disrupt the assembly of the PRC2 complex also failed to abolish the PARPi resistance induced by EZH2 overexpression (Fig. 2k–m). Collectively, these findings indicate that EZH2 promotes PARPi resistance in BRCA1-deficient EOC through a noncanonical mechanism involving the methylation of a nonhistone substrate.
Fig. 2: EZH2 drives PARPi resistance in BRCA1-deficient EOC through a mechanism that does not depend on its histone methyltransferase activity.
a Schematic representation of EZH2 and H3K27me3 IHC staining in clinical PARPi-sensitive and -resistant BRCA1-deficient EOC samples. b Statistical graph of the histochemical scores for EZH2 staining. Statistical significance was determined using Student’s t-test. c Statistical graph of the histochemical scores for H3K27me3 staining. Statistical significance was determined using Student’s t test. d Analysis of the correlation between EZH2 and H3K27me3 levels in the FUSCC cohort (n = 30). Correlations between EZH2 and H3K27me3 levels were assessed using the nonparametric Spearman rank correlation analysis. e Western blot analysis of EZH2 and H3K27me3 levels in parental and PARPi-resistant cells. f Western blot analysis showing the effects of different EZH2 targeting methods on EZH2 protein levels. The cells were treated with tazemetostat, astemizole, or MS1943 (each at 5 μM) for 48 h. g, h Results of the colony formation assay using cells treated with multiple EZH2 inhibitors (n = 3 independent experiments). Statistical significance was determined using two-way ANOVA. i, j Drug sensitivity in PDOs treated with multiple EZH2 inhibitors (n = 3 independent experiments). Statistical significance was determined using two-way ANOVA. k Western blot analysis showing the effects of EED silencing on the PARPi response. l, m Results of the colony formation assay showing the effects of silencing EED on the PARPi response (n = 3 independent experiments). Statistical significance was determined using two-way ANOVA.
YES1-induced tyrosine phosphorylation dictates the function of EZH2 in BRCA1-deficient EOC
We next aimed to determine why EZH2 functions in a PRC2-independent manner. Previous studies have shown that posttranslational modifications (PTMs) serve as central regulatory mechanisms governing the function and mode of EZH225,26,27. Notably, our GSEA of the OVCAR8 and OVCAR8-OlaR transcriptome sequencing data revealed that protein phosphorylation-related genes were significantly enriched in drug-resistant cells (Supplementary Fig. S2a), leading us to hypothesize that phosphorylation may serve as a critical determinant of the functional mode of EZH2 in BRCA1-deficient EOC. We validated this hypothesis by performing immunoprecipitation (IP) of EZH2, followed by western blot with antibodies against phosphorylation and other PTMs for screening. The results revealed that tyrosine phosphorylation was the most significantly altered PTM of EZH2 in drug-resistant cells (Fig. 3a and Supplementary Fig. S2b). Similarly, elevated EZH2 protein expression and tyrosine phosphorylation levels were consistently detected in BRCA1-deficient, drug-resistant clinical EOC samples (Fig. 3b, c and Supplementary Fig. S2c, d). Further experiments revealed that the level of the tyrosine-phosphorylated EZH2 protein in the PRC2 complex (IP-EED) did not differ significantly between parental and resistant cells (Supplementary Fig. S2e). These findings indicated that EZH2 in the PRC2 complex lacks tyrosine phosphorylation, which suggested that tyrosine phosphorylation might enable EZH2 to exert its drug resistance-promoting effect independent of histone modification.
Fig. 3: YES1-induced tyrosine phosphorylation dictates the function of EZH2 in BRCA1-deficient EOC.
a Antibodies against different PTMs were used to identify the types of PTMs (posttranslational modifications) that occur in EZH2. b EZH2 protein levels in primary (n = 5) and recurrent tumor (n = 6) tissues from patients with BRCA1-deficient EOC. Statistical significance was determined using two-way ANOVA. c Phosphorylated EZH2 protein levels in primary (n = 5) and recurrent tumor (n = 6) tissues from patients with BRCA1-deficient EOC. Statistical significance was determined using two-way ANOVA. d EZH2 antibodies were used to enrich for EZH2 proteins and EZH2-interacting proteins in parental and PARPi-resistant cells, and a mass spectrometry analysis was subsequently performed. e IP‒MS results showing that YES1 is a potential kinase regulating EZH2 phosphorylation. A total of 4 YES1-specific peptide segments were detected in resistant cells. f Co-IP results showing that YES1 interacts with EZH2 in resistant cells. The YES1 inhibitor CH6953755 (5 μM) can inhibit the binding of YES1 to EZH2. g Western blot analysis showed that YES1 can regulate the phosphorylation of EZH2. Treatment with the YES1i CH6953755 (5 μM) reduced the level of tyrosine-phosphorylated EZH2. h, i Drug sensitivity in PDOs treated with the YES1i CH6953755 (5 μM, n = 3 independent experiments). Statistical significance was determined using two-way ANOVA. j Schematic diagram of the tumor xenograft experiments. k Tumor volumes were measured on the indicated days (n = 7 mice per group). Statistical significance was determined using two-way ANOVA. l Tumor weights were measured and are presented as the means ± S.D. (n = 7 mice per group). Statistical significance was determined using two-way ANOVA. m Body weight was measured on the indicated days (n = 7 mice per group). n, o Results of the colony formation assay for the functional validation of the EZH2-YES1 regulatory axis. Treatment with the EZH2i MS1943 (5 μM) blocked YES1-induced PARPi resistance (n = 3 independent experiments). Statistical significance was determined using two-way ANOVA.
We conducted co-Immunoprecipitation (co-IP) followed by mass spectrometry and a proteomic analysis to explore the key kinase responsible for EZH2 phosphorylation. YES proto-oncogene 1 (YES1) was the only tyrosine kinase that bound to EZH2 in PARPi-resistant cells (Fig. 3d, e). YES1 is a member of SRCs of non-receptor tyrosine kinases that is involved in various physiological and pathological processes. Phosphorylation of SRCs at Tyr416 (corresponding to Tyr426 of YES1) in the activation loop of the kinase domain upregulates enzyme activity28. Co-IP confirmed the increased interaction between YES1 and EZH2 in PARPi-resistant cells (Fig. 3f). In clinical samples, the level of the active form but not the total YES1 level was increased in PARPi-resistant samples (Supplementary Fig. S2f, g), and both tyrosine-phosphorylated EZH2 (p-Tyr EZH2) and total EZH2 levels were positively correlated with the level of active YES1 (pY416 SFK) (Supplementary Fig. S2h, i), indicating a significant effect of YES1 on EZH2 phosphorylation. A specific YES1 inhibitor, CH6953755, was introduced to verify this result, and the YES1i substantially reduced both total and phosphorylated EZH2 levels (Fig. 3f). YES1 overexpression significantly increased the level of EZH2 phosphorylated on tyrosines, whereas the YES1i treatment markedly attenuated this effect (Fig. 3g and Supplementary Fig. S2j). Moreover, YES1i treatment increased PARPi sensitivity in both resistant PDOs (Fig. 3h, i) and cell lines (Supplementary Fig. S2k, l). We further validated the effect of phosphorylated EZH2 on the sensitivity of BRCA1-deficient EOC cells to PARPis in vivo by examining the growth of tumors transplanted into nude mice after treatment with EZH2i MS1943, and the YES1i CH6953755 (Fig. 3j). We found that combination therapy with EZH2i or YES1i and olaparib synergistically inhibited tumor growth (Fig. 3k, l) but did not significantly affect the body weight of the mice (Fig. 3m). Notably, YES1 overexpression increased PARPi resistance, whereas the inhibition of EZH2 eliminated these effects (Fig. 3n, o). These data support the hypothesis that YES1-induced p-Tyr EZH2 plays a vital role in regulating the PARPi response in BRCA1-deficient EOC.
The translocation of phosphorylated EZH2 to mitochondria promotes mitochondrial fusion and inhibits mitochondrial apoptosis
We subsequently investigated the potential mechanism by which p-Tyr EZH2 promotes PARPi resistance. Currently, researchers believe that PARPis exert their pharmacological effects by inducing apoptosis29. Flow cytometry results showed that combined treatment with the EZH2i or YES1i significantly increased PARPi-induced apoptosis in BRCA1-deficient resistant EOC cells (Fig. 4a, b). GSEA also revealed that the negative regulation of intrinsic apoptosis pathways was significantly enriched in drug-resistant cells (Supplementary Fig. S3a). Additionally, we examined extrinsic (cleaved Caspase-8) and intrinsic (cleaved Caspase-9) markers of apoptosis by performing Western blot analyses, and the results showed that pharmacological inhibition of either YES1 or EZH2 induced Caspase-3 and Caspase-9 activation, while Caspase-8 cleavage remained unaltered (Supplementary Fig. S3b, c). IF staining also revealed that pharmacological inhibition of either YES1 or EZH2 promoted the mitochondrial translocation of BAX, a vital process for triggering intrinsic apoptosis (Fig. 4c, d). Our findings suggest that targeting phosphorylated EZH2 in resistant cells might increase PARPi sensitivity by inducing the mitochondrial apoptosis pathway.
Fig. 4: YES1 phosphorylates EZH2 and induces its translocation to the mitochondria to promote mitochondrial fusion and inhibit the mitochondrial apoptosis pathway.
a, b Apoptosis detection using flow cytometry. Targeting phosphorylated EZH2 significantly increased PARPi-induced apoptosis (n = 3 independent experiments). Statistical significance was determined using two-way ANOVA. c, d Colocalization of BAX and mitochondria after treatment with the EZH2i MS1943 (5 μM) or YES1i CH6953755 (5 μM). Targeting phosphorylated EZH2 could increase the translocation of BAX to the mitochondria. Statistical significance was determined using two-way ANOVA. Cells in at least 20 fields of view were counted. e Representative images of sensitive and resistant cells obtained using a fluorescence confocal microscope. f Statistical analysis of mitochondria of different lengths in sensitive and resistant cells. Statistical significance was determined using two-way ANOVA. Cells in at least 20 fields of view were counted. g Representative transmission electron microscopy images of sensitive and resistant cells. h Statistical analysis of mitochondria of different lengths in sensitive and resistant cells. Statistical significance was determined using two-way ANOVA. Cells in at least 20 fields of view were counted. i Effects of the PARPi olaparib, the EZH2i MS1943 (5 μM) and the YES1i CH6953755 (5 μM) on mitochondrial morphology. Pharmacological inhibition of EZH2 or YES1 reversed PARPi-induced mitochondrial fusion, inducing mitochondrial fission. Statistical significance was determined using two-way ANOVA. Cells in at least 20 fields of view were counted. j Apoptosis detection using flow cytometry. The EZH2i MS1943 (5 μM) blocked YES1-induced apoptosis resistance in cells subjected to PARPi treatment (n = 3 independent experiments). Statistical significance was determined using two-way ANOVA. k Colocalization of BAX and mitochondria. The EZH2i MS1943 (5 μM) blocked YES1-induced BAX translocation to mitochondria in cells treated with the PARPi. Statistical significance was determined using two-way ANOVA. Cells in at least 20 fields of view were counted. l Statistical analysis of mitochondria of different lengths within cells. The EZH2i MS1943 (5 μM) blocked YES1-induced mitochondrial fusion in PARPi-treated cells. Statistical significance was determined using two-way ANOVA. Cells in at least 20 fields of view were counted. m Differences in the localization of EZH2 in mitochondria between parental and resistant cells. n Statistical analysis of EZH2 colocalization with mitochondria in parental and resistant cells. Statistical significance was determined using Student’s t-test. Cells in at least 20 fields of view were counted. o Subcellular fractionation experiment. Resistant cells exhibited increased extranuclear EZH2 localization. p YES1 overexpression significantly increased the mitochondrial localization of EZH2, and treatment with the YES1i CH6953755 (5 μM) markedly reduced this translocation. q GSEA results showed that genes enriched in the terms “mitochondrion organization” and “protein localization to mitochondrion” were significantly differentially expressed between parental and resistant cells. r Active YES1 in the cytoplasm might interact with EZH2, where it phosphorylates EZH2, and this process prevents EZH2 nuclear translocation. Excess EZH2 subsequently undergoes mitochondrial translocation, triggering mitochondrial fusion in resistant cells.
We next aimed to elucidate how p-Tyr EZH2 regulates mitochondrial apoptosis. Since mitochondrial fusion and fission regulate the mitochondrial apoptosis pathway30,31,32, we evaluated the dynamic changes in mitochondria using Mito Tracker staining. We observed increased mitochondrial fusion in PARPi-resistant EOC cells using confocal scanning microscopy (Fig. 4e, f), which was further validated by transmission electron microscopy imaging (Fig. 4g, h). Western blot analyses also revealed reduced expression of mitochondrial fission markers in resistant cells (Supplementary Fig. S3d, e). Notably, we observed increased integrity of mitochondrial branching and network structures upon olarparib treatment in resistant cells, whereas pharmacological inhibition of EZH2 or YES1 reversed this phenotype, inducing mitochondrial fission (Fig. 4i and Supplementary Fig. S3f, g). Accordingly, YES1 overexpression reduced mitochondrial apoptosis and mitochondrial BAX localization but induced mitochondrial fusion, whereas EZH2 inhibition abrogated these changes (Fig. 4j–l). These results suggested that mitochondrial fusion (Supplementary Fig. S3h) is the “core” biological event contributing to p-Tyr EZH2-induced apoptosis resistance in drug-resistant cells.
Intriguingly, immunofluorescence staining and Western blot analyses revealed that resistant cells exhibited increased extranuclear EZH2 localization (Fig. 4m–o). Subcellular fractionation experiments demonstrated that activated YES1 accumulated in the cytoplasm of PARPi-resistant cells, and a portion of EZH2 was also detectable in the cytoplasm of resistant cells. Thus, we inferred that active YES1 might interact with EZH2 in the cytoplasm, where it phosphorylates EZH2, and this process prevents the nuclear translocation of EZH2. The excess EZH2 subsequently undergoes mitochondrial translocation, triggering mitochondrial fusion. The overexpression of YES1 significantly increased the mitochondrial localization of EZH2, whereas treatment with YES1i markedly reduced this translocation (Fig. 4p and Supplementary Fig. S3i) and increased PARPi-induced apoptosis (Supplementary Fig. S3j). GSEA also revealed that the terms “mitochondrion organization” and “protein localization to mitochondrion” were significantly enriched in drug-resistant cells (Fig. 4q).
Previous studies have indicated that translocases of the outer membrane (TOMs) serve as the major entry channels into mitochondria. IP‒MS revealed that phosphorylated EZH2 may interact with TOM20 and TOM5 in drug-resistant cells (Supplementary Fig. S3k). Subsequent experiments demonstrated that knocking down TOM20 expression significantly inhibited the translocation of cytoplasmic EZH2 to mitochondria (Supplementary Fig. S3l). Co-IP experiments further confirmed the increased interaction between TOM20 and EZH2 in resistant cells (Supplementary Fig. S3m). Moreover, inhibiting EZH2 translocation to the mitochondria by knocking down TOM20 effectively demonstrated the efficacy of PARPi therapy (Supplementary Fig. S3n, o). These data suggest that TOM20 is involved in the translocation of phosphorylated EZH2 to the mitochondria (Supplementary Fig. S3p).
Taken together, the above findings indicated that activated YES1 catalyzes EZH2 phosphorylation to maintain EZH2 in the cytoplasm, which facilitates its mitochondrial translocation mediated by TOM20, where it potentially drives PARPi resistance by promoting mitochondrial fusion and suppressing mitochondrial apoptosis (Fig. 4r).
YES1 drives drug resistance by inducing the phosphorylation of EZH2 at tyrosine 728
We employed NetPhos 3.1 (https://services.healthtech.dtu.dk/services/NetPhos-3.1/) to identify the specific tyrosines of EZH2 that are phosphorylated by YES1 and identified 7 potential candidate sites (Fig. 5a). Therefore, we generated and expressed dephospho-mimetic mutations targeting these potential residues in EZH2-knockdown PARPi-resistant EOC cells. Immunofluorescence staining revealed that only the dephospho-mimetic mutation at Y728 (Y728F) abolished EZH2 translocation to mitochondria (Fig. 5b and Supplementary Fig. S4a), and this residue was evolutionarily conserved across species (Supplementary Fig. S4b). Furthermore, EZH2-Y728F was the only mutation that failed to restore mitochondrial fission caused by EZH2 knockdown (Supplementary Fig. S4c). We validated the function of p-Y728 EZH2 by generating a phospho-mimetic EZH2-Y728D construct. Immunofluorescence staining showed that compared with EZH2-Y728F, EZH2-Y728D resulted in increased mitochondrial localization and pro-fusion function (Fig. 5c–e). Nuclear-mitochondrion separation experiments in 293 T cells also revealed that EZH2-Y728D can be specifically localized to mitochondria (Supplementary Fig. S4d). However, in EZH2-knockdown PARPi-resistant cells, both the EZH2-WT and EZH2-Y728D proteins translocated to the mitochondria, possibly because YES1 was activated in resistant cells (Supplementary Fig. S4e). Both flow cytometry and IF assays showed that, unlike EZH2-Y728F, EZH2-Y728D could block the mitochondrial translocation of BAX and abrogate the apoptosis induced by EZH2 knockdown during PARPi treatment (Fig. 5f, g). A colony formation assay also confirmed that EZH2-Y728D could completely abolish the sensitization caused by EZH2 knockdown (Fig. 5h, i). We further generated EZH2-knockdown HEYBRCA1-KO cells that were cotransfected with 3 different plasmids (EZH2-WT, EZH2-Y728D, and EZH2-Y728F), established a xenograft model in nude mice, and monitored tumor growth over a 4-week period to validate the results in vivo. We found that EZH2-Y728D exerted similar effects on promoting PARPi resistance to EZH2-WT. However, EZH2-Y728F resulted in drastic tumor shrinkage after PARPi therapy (Fig. 5j, k).
Fig. 5: YES1 drives drug resistance by inducing the phosphorylation of EZH2 at tyrosine 728.
a Potential tyrosine phosphorylation sites of EZH2 predicted by NetPhos 3.1. Cut-off value = 0.5. b Colocalization analysis of EZH2-His containing mutations at different sites with mitochondria. Statistical significance was determined using two-way ANOVA. Cells in at least 20 fields of view were counted. c, d Localization of EZH2 phosphorylated and dephosphorylated at the Y728 site to mitochondria. Statistical significance was determined using two-way ANOVA. Cells in at least 20 fields of view were counted. e Effect of EZH2pY728 on mitochondrial morphology. Statistical significance was determined using two-way ANOVA. Cells in at least 20 fields of view were counted. f IF staining showing that EZH2pY728 blocked the mitochondrial translocation of BAX. Statistical significance was determined using two-way ANOVA. Cells in at least 20 fields of view were counted. g Flow cytometry results showing that EZH2pY728 inhibited PARPi-induced apoptosis (n = 3 independent experiments). Statistical significance was determined using two-way ANOVA. h, i Colony formation assays revealed that EZH2pY728 reversed PARPi resistance. Statistical significance was determined using two-way ANOVA. j Tumor volumes measured on the indicated days (EZH2-WT, n = 6; EZH2-Y728D, n = 7; EZH2-Y728F, n = 8). Statistical significance was determined using two-way ANOVA. k Tumor weights were measured and are presented as the means ± S.D. (EZH2-WT, n = 6; EZH2-Y728D, n = 7; EZH2-Y728F, n = 8). Statistical significance was determined using two-way ANOVA. l Molecular docking results from AlphaFold3 illustrating the interface between YES1Y426D and EZH2. m, n Schematic illustration and results of the in vitro phosphorylation assay. o Typical images of IHC staining for pY728 EZH2 and active YES1 (pY416 SFK). p Correlations between the levels of pY728 EZH2 and active YES1 were assessed using the nonparametric Spearman rank correlation analysis. q, r Comparison of the expression patterns of pY728 EZH2 and active YES1 between PARPi-sensitive and PARPi-resistant EOC tissues. Statistical significance was determined using Student’s t-test.
We further performed molecular docking with AlphaFold3 and found that the C3 site of YES1Y426D (YES1 activating mutant) can form a covalent hydrogen bond with the Y728 site of EZH2 (Fig. 5l). In vitro phosphorylation assays demonstrated that YES1-Y426D specifically catalyzed the phosphorylation of the EZH2 Y728 site. However, when the Y728 site was mutated, activated YES1 failed to catalyze EZH2-Y728 phosphorylation (Fig. 5m, n). Next, we generated a specific antibody targeting pY728 EZH2 and detected its levels in clinical samples. The results indicated a positive correlation between YES1pY426 and EZH2pY728 levels (Fig. 5o, p), and both YES1pY426 and EZH2pY728 were highly expressed in tissues from patients with PARPi-resistant BRCA1-deficient EOC (Fig. 5q, r). In summary, the results of this study indicate that Y728 is the key residue involved in YES1-induced EZH2 phosphorylation, which mainly promotes the mitochondrial translocation of EZH2 and subsequent mitochondrial fusion.
Mitochondrion-localized EZH2 methylates MYO19 to enhance mitochondrial fusion
The regulatory mechanism underlying EZH2-induced mitochondrial membrane fusion in resistant EOC remains incompletely understood. We addressed this issue by performing a comparative mass spectrometry analysis of EZH2 interactomes in parent and PARPi-resistant cells, which revealed increased binding between EZH2 and mitochondrial fission-related proteins in the resistant population according to the results of the Gene Ontology (GO) enrichment analysis (Fig. 6a). Similarly, a subsequent screen identified myosin XIX (MYO19), a mitochondria-anchored actin-based motor protein that regulates organelle dynamics and oxidative phosphorylation (OXPHOS) metabolism33, as a binding partner of EZH2 in resistant cells (Fig. 6b and Supplementary Fig. S5a). Functional validation revealed that MYO19 depletion could abrogate EZH2i-induced mitochondrial fragmentation and apoptosis (Fig. 6c–e). These results indicated that MYO19 mainly inhibited mitochondrial fusion in EOC cells and was negatively regulated by EZH2. Hence, we further explored how EZH2 regulates the ability of MYO19 to reprogram the dynamics of mitochondrial fusion. First, Western blot revealed that EZH2 did not regulate MYO19 protein expression and vice versa (Supplementary Fig. S5b). IF staining showed that PARPi treatment promoted mitochondrial fusion and reduced the mitochondrial localization of MYO19, whereas EZH2 knockdown restored the positioning of MYO19 in mitochondria and increased mitochondrial fission events (Supplementary Fig. S5c, d). Moreover, EZH2i increases the mitochondrial localization of MYO19 in resistant cells (Fig. 6f). Collectively, we speculated that EZH2 mainly prevents the mitochondrial localization of MYO19 to promote mitochondrial fusion.
Fig. 6: Mitochondrion-localized EZH2 regulates mitochondrial morphology by methylating the K928 site of MYO19.
a IP‒MS and GO enrichment analysis results showing that EZH2 binds to multiple mitochondrial fission-related proteins in drug-resistant cells. b MYO19 was verified as a substrate for phosphorylated EZH2. c Effect of interventions targeting MYO19 on mitochondrial morphology. Interference with MYO19 blocked the effects of the EZH2i MS1943 (5 μM). Statistical significance was determined using two-way ANOVA. Cells at least 20 fields of view were counted. d MYO19 depletion abrogated the EZH2 inhibitor-induced mitochondrial translocation of BAX, as determined by IF staining. Statistical significance was determined using two-way ANOVA. Cells at least 20 fields of view were counted. e Apoptosis detection using flow cytometry (n = 3 independent experiments). Interference with MYO19 blocked the effects of the EZH2i MS1943 (5 μM). Statistical significance was determined using two-way ANOVA. f Treatment with the EZH2i MS1943 (5 μM) increased the localization of MYO19 to mitochondria. Statistical significance was determined using Student’s t test. Cells in at least 20 fields of view were counted. g, h EZH2pY728 reduced the mitochondrial localization of MYO19. Statistical significance was determined using two-way ANOVA. Cells in at least 20 fields of view were counted. i MYO19 contains an evolutionarily conserved histone H3K27-like EZH2 recognition motif (R-K-S) centred at lysine 928 (K928). j Conservation of MYO19 K928 across species. k Electrostatic potential map of the EZH2-MYO19 interaction interface. l, m Verification of methylation at the MYO19 K928 site in drug-resistant cells with endogenous MYO19 knockdown. n The effect of point mutations at the K928 site on the mitochondrial localization of MYO19 was assessed using IF staining. Compared with that of WT MYO19, the localization of the K928 demethylated mutant to mitochondria was increased. Statistical significance was determined using two-way ANOVA. Cells in at least 20 fields of view were counted. o Effect of point mutations at the MYO19 K928 site on mitochondrial morphology. Statistical significance was determined using two-way ANOVA. Cells in at least 20 fields of view were counted. p The effect of MYO19 K928 methylation on mitochondrial translocation of BAX was evaluated by IF staining. Statistical significance was determined using two-way ANOVA. Cells in at least 20 fields of view were counted. q The effect of MYO19 K928 methylation on PARPi-induced apoptosis was evaluated using flow cytometry (n = 3 independent experiments). Statistical significance was determined using two-way ANOVA. r, s Results of the in vitro methylation assay. t, u The effect of MYO19 K928 methylation on the PARPi resistance of BRCA1-deficient EOC cells was evaluated by performing a colony formation assay (n = 3 independent experiments). Statistical significance was determined using two-way ANOVA. v Interactions between YES1, EZH2 and MYO19 and their downstream effects.
As we confirmed that EZH2 functions via its enzymatic activity, we hypothesized that MYO19 might be a novel substrate of EZH2 in mitochondria. As expected, phosphorylated EZH2 catalyzed MYO19 trimethylation (Supplementary Fig. S5e). Moreover, transfection with EZH2 phosphorylation and dephosphorylation mimic plasmids (EZH2-Y728D and EZH2-Y728F) also demonstrated that EZH2pY728 was essential for MYO19 trimethylation and reduced its localization to mitochondria (Fig. 6g, h and Supplementary Fig. S5f). Based on these findings, we propose the following mechanism in BRCA1-deficient, PARPi-resistant EOC cells: phosphorylated EZH2 can translocate to mitochondria, where it catalyzes the trimethylation of MYO19, a critical regulator of mitochondrial fission. This posttranslational modification promotes mitochondrial fusion, ultimately promoting cellular survival during PARPi treatment.
Phosphorylated EZH2 promotes mitochondrial fusion by catalysing trimethylation of the K928 residue of MYO19
We further explored the EZH2‒MYO19 interaction pattern. We discovered that MYO19 contains an evolutionarily conserved histone H3K27-like EZH2 recognition motif (R-K-S)34 centred at lysine 928 (K928) (Fig. 6i, j). We further performed molecular docking with AlphaFold3 and screened 13 potential binding sites to validate this site. In the prediction results, excluding the binding site between wild-type (WT) EZH2 and MYO19, 6 specific binding sites were identified between EZH2Y728D and MYO19 (Supplementary Fig. S5g). A further detailed docking model indicated that EZH2Y728D might form a negatively charged catalytically active pocket that binds to the positively charged MYO19-RKS and catalyzes trimethylation at the K928 site (Fig. 6k). Endogenous MYO19 expression was silenced, and the WT-MYO19 or K928R-MYO19 mutant was subsequently re-expressed in resistant cells to verify this docking model. Western blot assays showed that in PARPi-resistant cells, the WT MYO19 protein was methylated, whereas the K928R mutant was not (Fig. 6l, m). Functional assays revealed that the demethylation-mimetic mutant MYO19-K928R localized normally to mitochondria and supported fission (Fig. 6n, o). Consistent with these findings, the K928R mutant promoted the translocation of BAX to mitochondria (Fig. 6p) and enhanced PARPi-induced apoptosis (Fig. 6q). In vitro methylation assays confirmed that EZH2Y728D specifically catalyzed MYO19 methylation at the K928 site (Fig. 6r, s). Colony formation assays indicated that the introduction of MYO19-K928R significantly reduced PARPi resistance (Fig. 6t, u). Taken together, these findings suggest that the PARPi-induced phosphorylation of EZH2 promotes its translocation to mitochondria, where it catalyzes the trimethylation of K928 in MYO19. This posttranslational modification increases mitochondrial fusion, suppresses mitochondrial apoptosis, and ultimately drives PARPi resistance in BRCA1-deficient EOC cells (Fig. 6v).
PARPi activates YES1 to increase EZH2 protein accumulation in PARPi-resistant EOC cells
As described above, EZH2 was upregulated in PARPi-resistant cells and tissues (Fig. 2b, e), which raised the following question: what causes the elevated expression of EZH2? We addressed this question by evaluating the expression pattern of EZH2. The results revealed that the EZH2 protein was highly expressed in both PARPi-resistant cells and tissues, but the results of RNA-Seq and qPCR revealed that the mRNA expression did not differ significantly between sensitive and resistant tissues or cells (Fig. 7a and Supplementary Fig. S6a). In addition, we observed that olaparib treatment induced a time-dependent increase in the EZH2 protein level in PARPi-sensitive cells, whereas this phenomenon was not observed in resistant cells. Moreover, the EZH2 mRNA levels in both types of cells remained unchanged after PARPi treatment (Fig. 7b, c), indicating that the regulation of EZH2 expression occurred at the posttranslational level. Since we found that the EZH2 protein level decreased after treatment with YES1i in the previous experiments (Fig. 3f), we speculated that EZH2 is phosphorylated during PARPi resistance, which affects EZH2 protein accumulation. We first conducted experiments by applying different doses of YES1i for different times to verify this hypothesis, and found that YES1i could reduce EZH2 protein levels but not EZH2 mRNA levels (Fig. 7d, e and Supplementary Fig. S6b). We also applied dasatinib, an approved drug targeting YES1 kinase activity, and the results showed that dasatinib treatment and YES1i treatment reduced EZH2 protein levels to a similar extent in PARPi-resistant cells. More importantly, dasatinib and CH6953755 abrogated the olaparib-induced increase in EZH2 protein levels in parental cells (Supplementary Fig. S6c). Our results confirmed that the increase in EZH2 protein expression was mediated by YES1-induced phosphorylation. We next investigated how YES1 was activated in EOC cells. Using a pY416 SFK antibody28,35, we found that PAPRi treatment activated YES1 and induced EZH2 phosphorylation. Interestingly, YES1 was maintained in an active state in drug-resistant cells even in the absence of a PARPi, which may have been caused by the long-term effects of the PARPi (Fig. 7f). Combined with previous findings concerning the active YES1 states in clinical samples (Supplementary Fig. S2f–i), we inferred that PARPi induction can lead to YES1 activation, which subsequently catalyzes EZH2 tyrosine phosphorylation to increase its protein level in PARPi-resistant cells.
Fig. 7: PARPi can activate YES1, thereby inducing phosphorylation of the EZH2 Y728 site and increasing EZH2 protein stability.
a Relative EZH2 mRNA levels in the BRCA1-deficient EOC tissues shown in Supplementary Fig. S2c, d were determined using qPCR. Statistical significance was determined using two-way ANOVA. b Effect of different exposure times of the PARPi olaparib (20 μM) on EZH2 protein expression. c Effect of the PARPi olaparib (20 μM) on EZH2 mRNA levels, as detected by qPCR. Statistical significance was determined using two-way ANOVA. d Effect of different concentrations of the YES1i CH6953755 on EZH2 protein expression. e The effect of the YES1i CH6953755 (5 μM) on EZH2 mRNA levels was detected using qPCR. Statistical significance was determined using two-way ANOVA. f Effects of the PAPRi olaparib (20 μM) on the phosphorylation of YES1 and its target EZH2. g, h Effects of the PARPi olaparib (20 μM) and the tyrosine kinase inhibitors CH6953755 (5 μM) and dasatinib (1 μM) on EZH2 protein stability. Statistical significance was determined using two-way ANOVA. i Effects of the PARPi olaparib (20 μM) and the tyrosine kinase inhibitors CH6953755 (5 μM) and dasatinib (1 μM) on EZH2 ubiquitination levels. j–l Effect of specific phosphorylation at the Y728 site on EZH2 stability. m Effect of specific phosphorylation at the Y728 site on the ubiquitination of EZH2 and the strength of the interaction between EZH2 and TRIM4. n The strategy used to screen potential E3 ubiquitin ligases. o Verification of the interaction between TRIM4 and EZH2 in parental and resistant cells. p Effects of the PARPi olaparib (20 μM) and YES1i CH6953755 (5 μM) on the strength of the EZH2-TRIM4 interaction. q Phosphorylation of EZH2 at the Y728 site inhibits its binding to TRIM4, reduces its ubiquitination and increases its protein stability.
The phosphorylation of the Y728 residue of EZH2 prevents it from binding to the E3 ligase TRIM4 and increases its protein stability
Since EZH2 expression is regulated at the posttranslational level, we performed cycloheximide (CHX) chase assays to evaluate EZH2 protein stability. The results showed that PARPi treatment prolonged the half-life of the EZH2 protein and that the inhibition of EZH2 tyrosine phosphorylation blocked this effect (Fig. 7g, h). Moreover, a proteasome inhibitor (MG132) reversed the decrease in EZH2 levels caused by the inhibition of its phosphorylation (Supplementary Fig. S6d), suggesting that phosphorylation of EZH2 affects its protein stability via the proteasome pathway. Ubiquitination experiments revealed that the level of EZH2 ubiquitination was reduced after treatment with the PARPi, which could be abolished by specific YES1i or dasatinib treatment (Fig. 7i). We further verified the relationship between EZH2 phosphorylation and protein stability using the Y728D/Y728F mutant plasmid. Western blot results showed that EZH2-Y728D had the longest half-life and lowest ubiquitination level, whereas EZH2-Y728F showed the opposite phenotypes (Fig. 7j–l). Moreover, a proteasome inhibitor reversed the decrease in EZH2 levels caused by the dephospho-mimetic mutation (Supplementary Fig. S6e). The above results indicated that YES1-induced phosphorylation of EZH2 at Y728 diminished its ubiquitination and subsequent proteasomal degradation, which resulted in the accumulation of the EZH2 protein in BRCA1-deficient EOC cells and ultimately led to PARPi resistance.
The next goal was to identify the key E3 ligase responsible for EZH2 ubiquitination. We screened for interacting partners of unphosphorylated EZH2 by performing IP‒MS and subsequently compared the screening results with those of the E3 ubiquitin ligase pool (UbiBrowser 2.0) and found that EZH2 specifically bound to two E3 ligases, TRIM4 and TRIP12, in parental cells (Fig. 7n). Further co-IP experiments confirmed that EZH2 interacted with TRIM4 but not TRIP12 in PARPi-sensitive EOC cells and that this interaction was greatly weakened in PARPi-resistant cells (Fig. 7o). Notably, the binding of EZH2 to TRIM4 was substantially reduced by PARPi treatment; however, YES1i treatment abrogated this interaction (Fig. 7p). Ubiquitination experiments and co-IP experiments further verified that EZH2pY728 specifically inhibited its binding to TRIM4 (Fig. 7m), as evidenced by the successful binding of the EZH2-Y728F mutant to TRIM4, whereas the EZH2-Y728D mutant disrupted this interaction. These results indicate that phosphorylation of the EZH2 Y728 site inhibits its binding to TRIM4, reduces EZH2 ubiquitination and increases its protein stability (Fig. 7q).
