HIF2A as a prognostic and clinical therapeutic target in ovarian clear cell carcinoma

Mengxin Jiang: Conceptualization; methodology; data curation; formal analysis; investigation; project administration; visualization; writing – original draft. Ken Yamaguchi: Conceptualization; methodology; data curation; formal analysis; investigation; funding acquisition; writing – original draft; project administration; resources; software. Kohei Hamada: Investigation; writing – review and editing; methodology; data curation; software; formal analysis. Yuko Hosoe: Methodology; writing – review and editing. Yuka Mise: Conceptualization; writing – review and editing. Sachiko Kitamura: Conceptualization; writing – review and editing. Zhihong Liu: Conceptualization; writing – review and editing. Hairong Wei: Conceptualization; writing – review and editing. Zhenhai Gao: Conceptualization; writing – review and editing. Mana Taki: Conceptualization; writing – review and editing. Koji Yamanoi: Conceptualization; writing – review and editing. Ryusuke Murakami: Formal analysis; software; validation; writing – review and editing. Rin Mizuno: Conceptualization; writing – review and editing. Taito Miyamoto: Conceptualization; supervision; validation; writing – review and editing. Junzo Hamanishi: Conceptualization; funding acquisition; project administration; resources; writing – review and editing. Masaki Mandai: Supervision; writing – review and editing.

This study was approved by the Ethics Committee of the Graduate School and Faculty of Medicine, Kyoto University (reference number G1322), and was conducted in accordance with the principles of the Declaration of Helsinki. Informed consent was obtained from all participants through either an opt‐in approach, where participants signed a printed informed consent document or an opt‐out approach (where participants were informed about the study through the website). All animal studies were conducted in compliance with the Kyoto University Animal Research Committee guidelines.

To evaluate the prognostic potential of HIF2A expression in ovarian CCC, we first evaluated HIF2A mRNA levels using our microarray dataset ( GSE29450 ). Compared with normal ovarian tissue, HIF2A mRNA expression was significantly enhanced in ovarian CCC ( p  = .036, Figure  1A ). Compared with other histologies of ovarian cancer, HIF2A expression was consistently upregulated in CCC and was confirmed by several datasets from tissue samples at our institution and public data, including clinical patient data ( GSE39204 , p <.001, Figure  1B and GSE65986 , p <.001, Figure  1C , respectively) and cell lines ( GSE29175 , p  = .011, Figure  1D ). Nuclear expression of HIF2A is a prognostic factor in ovarian clear cell carcinoma. (A) HIF2A expression between clear cell carcinoma clinical samples ( n  = 10) and normal counterparts ( n  = 10) from the GSE29450 datasets ( n  = 20). (B) HIF2A expression between clear cell carcinoma clinical samples ( n  = 17) and nonclear cell carcinoma clinical samples ( n  = 47) from our datasets ( GSE39204 ). (C) HIF2A expression between clear cell carcinoma clinical samples ( n  = 25) and nonclear cell carcinoma clinical samples ( n  = 30) from GSE65986 . (D) HIF2A expression between clear cell carcinoma cell lines ( n  = 14) and nonclear cell carcinoma cell lines ( n  = 24) from our datasets ( GSE29175 ). (E) Representative images of immunostaining for HIF2A nuclear expression. (F) Kaplan–Meier survival curves comparing progression‐free survival between patients with ovarian clear cell carcinoma with high ( n  = 42) and low ( n  = 22) HIF2A expression treated in Kyoto University Hospital. Log‐rank test indicates a significantly worse prognosis in the high‐expression group ( p  = .003). (G) Kaplan–Meier survival curves comparing overall survival between patients with ovarian clear cell carcinoma with high ( n  = 42) and low ( n  = 22) HIF2A expression treated in Kyoto University Hospital. Log‐rank test indicates a significantly worse prognosis in the high‐expression group ( p  = .005). (H) Kaplan–Meier survival curves showing poorer prognosis with high ssGSEA scores of the PID HIF2A PATHWAY gene set of MSigDB in the Washington University dataset (Kelly L Bolton et al. Clin Cancer Res. 2022) with overall survival times. The log‐rank test indicated a significantly worse prognosis in the high‐expression group ( p  = 0.024). Data are presented as mean ± standard error of the mean. As the functional HIF2A protein is located in the nucleus under hypoxic conditions, 24 we performed IHC on 64 ovarian CCC tumor samples from our institution to further evaluate the association between HIF2A nuclear expression and prognostic outcomes. Representative IHC images showing different intensities of HIF2A nuclear staining are presented in Figure  1E . Kaplan–Meier analyses revealed that high HIF2A nuclear expression was significantly related to decreased PFS and OS ( p  = .003, Figure  1F and p  = .005, Figure  1G , respectively). Using the PID HIF2A PATHWAY gene set of MSigDB based on the Washington University dataset, Kaplan–Meier analysis also revealed that the HIF2A pathway was also highly related to poorer OS in ovarian CCC ( p  = .024, Figure  1H ). Although the FIGO stage showed the most robust association with patient outcomes ( p <.001 for both PFS and OS), HIF2A expression also exhibited an independent prognostic value, with borderline significance for OS ( p  = 0.058 for OS). Given its consistent trend across models, HIF2A is a potentially meaningful biomarker in determining ovarian CCC prognosis (Tables  1A and 1B ). Univariate analysis indicated that high HIF2A nuclear expression and advanced FIGO stages were significantly associated with both reduced PFS (hazard ratio [HR] = 2.774 [95% confidence interval {CI}] 1.217–6.516; HR = 11.960 [95% CI 4.972–30.760], respectively; Table  1A ) and reduced OS (HR = 4.106 [95% CI 1.457–13.200] and HR = 11.490 [95% CI 3.956–37.910], respectively; Table  1B ). Multivariate Cox regression analysis identified advanced FIGO stages as an independent adverse prognostic factor for both PFS and OS (HR = 11.040 [95% CI 4.420–29.420] and HR = 9.852 [95% CI 3.262–33.800], respectively; Tables  1A and 1B ). In the multivariate Cox regression analysis, high HIF2A nuclear expression tended to be an unfavorable factor for PFS and OS (HR = 1.951 [95% CI 0.827–4.754] and HR = 2.904 [95% CI 1.004–9.542], respectively; Tables  1A and 1B ). Validation analysis using the PID HIF2A PATHWAY corroborated these findings in both univariate and multivariate Cox regression analysis, revealing that the high score group of PID HIF2A PATHWAY and advanced FIGO stages were associated with poorer OS (HR = 1.665 [95% CI 1.063–2.613], HR = 6.664 [95% CI 4.199–10.680], respectively; Table  1C ), and both parameters could be independent negative prognostic factors for OS (HR = 1.590 [95% CI 1.003–2.528], and HR = 6.565 [95% CI 4.134–10.520], respectively; Table  1C ). These results indicate that HIF2A signaling possibly contributes to the unfavorable prognosis of ovarian CCC. Univariate and multivariate Cox regression analysis of progression‐free survival (PFS, months) of ovarian clear cell carcinoma based on immunohistochemistry samples. Univariate and multivariate Cox regression analysis of overall survival (OS months) of ovarian clear cell carcinoma based on immunohistochemistry samples. Univariate and multivariate Cox regression with overall survival times (OS months) of ovarian clear cell carcinoma with PID HIF2A pathway gene set from MSigDB in Washington University dataset (Kelly L Bolton et al. Clin Cancer Res. 2022). In parental RMG1 cells, HIF2A protein was almost completely degraded under normoxic conditions but was strongly induced under hypoxia (Supplementary Figure  2A ). Therefore, knockdown of HIF2A expression using a lentivirus in the RMG1 cell line was confirmed by qRT‐PCR and Western blotting under hypoxia condition (Supplementary Figures  2B,C , respectively). Following knockdown of HIF2A, qRT‐PCR analysis revealed that the HIF2A downstream genes ( VEGFA , NDRG1 , GLUiT1 , and IGFBP3 ) tended to be downregulated (Figure  2A–D , respectively). Under 1.5% O 2 for 24 h, RNA sequencing analysis revealed that the sh‐HIF2A cell line is significantly more related to the Gene Ontology (GO) term related to oxidative stress, including “Response To Reactive Oxygen Species” ( p  = .003, P adjusted = 0.106, odds ratio = 25.432) (Table  2 ). The WST‐8 assay using shHIF2A cell lines and shcontrol indicated that the downregulation of HIF2A did not affect cell proliferation in vitro (Supplementary Figure  2D ). HIF2A knockdown enhances ROS production and mitochondrial function and suppresses tumor growth. (A)–(D) mRNA expression levels of HIF2A downstream genes were significantly downregulated in RMG1 cells following HIF2A knockdown under hypoxic conditions (1.5% O 2 ) for 24 h as determined by RT‐PCR. VEGFA (A), NDRG1 (B), GLUT1 (C), and IGFBP3 (D) expressions were normalized by ACTB. (E) Oxygen consumption rate (OCR) measurements in RMG1 cells (control, shHIF2A‐1, and shHIF2A‐2) under hypoxic conditions (1.5% O₂) for 24 h. Downward arrows indicate the injection points of oligomycin (1 μmol/L), FCCP (2 μmol/L), and rotenone/antimycin A (1 μmol/L each), from left to right. (F) Quantification of spare respiration in shcontrol and shHIF2A cell lines under 1.5% O 2 for 24 h. (G) Quantification of ATP production in shcontrol and shHIF2A cell lines under 1.5% O 2 for 24 h. (H) Quantification of maximal respiration in shcontrol and shHIF2A cell lines under 1.5% O 2 for 24 h. Quantification of mitochondrial ROS levels as mean fluorescence intensity (MFI) normalized to Hoechst staining in RMG1 cells under hypoxic conditions (1.5% O 2 ) for 24 h. (J) Representative fluorescence microscopy images of MitoSOX staining in shcontrol and shHIF2A cells, indicating the mitochondrial ROS levels. (K)–(N) HIF2A knockdown significantly inhibited tumor growth in RMG1 mouse xenograft models. Tumor growth curves comparing the shcontrol and shHIF2A groups (shHIF2A‐RMG1‐1 and shHIF2A‐RMG1‐2) are presented. Asterisks above the data points denote significance for shHIF2A‐1 and those below denote significance for shHIF2A‐2 (K). Tumor images from the shcontrol group (L). Tumor images from the shHIF2A‐1 group (M). Tumor images from the shHIF2A‐2 group (N). RMG1 cells transfected with either shcontrol or shHIF2A constructs were intradermally inoculated into mice ( n  = 12 per group). Data are presented as mean ± standard error of the mean. RT‐qPCR data are presented as mean ± standard error of the mean from three independent experiments. * p <.05, ** p <.01, ** p <.001. Enriched terms from the GO_Biological_Process_2025 gene set library, derived from the DEGs upregulated in the shHIF2A group. Note : DEGs were identified based on RNA‐seq data aligned to the human reference genome GRCh38. The overlap between these comparisons was considered as the final set of DEGs. In the mitochondrial functional assays, both shHIF2A‐1 and shHIF2A‐2 showed an obviously higher OCR under 1.5% O 2 for 24 h (Figure  2E ). The spare respiration was significantly upregulated in both the shHIF2A‐1 and shHIF2A‐2 groups ( p <.001 for shHIF2A‐1 and p <.001 for shHIF2A‐2, respectively, Figure  2F ). Similarly, ATP production was significantly upregulated in both shHIF2A cell lines ( p <.001 for shHIF2A‐1 and p <.001 for shHIF2A‐2, respectively, Figure  2G ). Maximal respiration was also significantly enhanced in the shHIF2A‐1 and shHIF2A‐2 cell lines ( p <.001 for shHIF2A‐1 and p <.001 for shHIF2A‐2, respectively, Figure  2H ). These results suggest an improvement in mitochondrial function after HIF2A knockdown. In the mitochondrial ROS detection analysis, the shHIF2A‐1 and shHIF2A‐2 cells showed increased production of mitochondrial ROS compared with the shcontrol cells ( p  = .095 for shHIF2A‐1 and p  = .016 for shHIF2A‐2, respectively, Figure  2I,J ). To establish flank tumor xenograft models, the RMG1 shRNA stable cell lines were implanted in mice. The growth curves demonstrated a significant reduction in the shHIF2A tumors compared with the shcontrol group ( p  = .006 for shHIF2A‐1 and p  = .039 for shHIF2A‐2, Figure  2K ; Figure  2L–N for shcontrol, shHIF2A‐1, and shHIF2A‐2, respectively). Additionally, the tumor weights in the shHIF2A group tended to be reduced compared with those in the control group ( p  = .311 for shHIF2A‐1 and p  = .182 for shHIF2A‐2, Supplementary Figure  2E ). We then investigated the effect of HIF1B , which is also required for HIF transcriptional activity, in parental RMG1 cells under hypoxia conditions. Knockdown of HIF1B resulted in decreased expression of HIF2A downstream genes ( VEGFA , NDRG1 , GLUT1 , and IGFBP3 ) (Supplementary Figure  2F–J ). These results suggest that the suppression of these hypoxia‐responsive genes is a direct consequence of impaired HIF2A‐HIF1B signaling. After treatment with 1 μM NKT2152 under 1.5% O 2 for 24 h in vitro, all four HIF2A downstream genes ( VEGFA , NDRG1 , GLUT1 , and IGFBP3 ) were downregulated in the NKT2152 group compared with DMSO in the RMG1 cell lines (Figure  3A–D , respectively). In the KOC7C cell line, expression of the four HIF2A downstream genes also tended to be downregulated (Figure  3E–H , respectively), suggesting that NKT2152 suppresses HIF2A targeting genes in vitro. NKT2152 enhances ROS production and mitochondrial function and induces an antitumor effect in CDX mouse models. (A)–(D) mRNA expression of HIF2A downstream genes was downregulated after treatment with 1 μM NKT2152 under 1.5% O 2 for 24 h in the RMG1 cell line as determined by RT‐PCR. VEGFA (A), NDRG1 (B), GLUT1 (C), and IGFBP3 (D) expressions were normalized by ACTB . (E)–(H) mRNA expression of HIF2A downstream genes was downregulated after treatment with 1 μM NKT2152 under 1.5% O 2 for 24 h in the KOC7C cell line as determined by RT‐PCR. VEGFA (E), NDRG1 (F), GLUT1 (G), and IGFBP3 (H) expressions were normalized by ACTB . (I) Measurement of OCR (oxygen consumption rate) in the RMG1 cell line treated with different concentrations of NKT2152 under 1.5% O 2 for 24 h (control, 1 μM, and 5 μM), as normalized by the cell protein concentration. Downward arrows show the injection points for oligomycin (1 μmol/L), FCCP (2 μmol/L), and rotenone/actinomycin (both 1 μmol/L) from left to right. (J) Quantification of spare respiration in the RMG1 cell line treated with different concentrations of NKT2152 under 1.5% O 2 for 24 h. (K) Quantification of ATP production in the RMG1 cell line treated with different concentrations of NKT2152 under 1.5% O 2 for 24 h. (L) Quantification of maximal respiration in the RMG1 cell line treated with different concentrations of NKT2152 under 1.5% O 2 for 24 h. (M) Measurement of OCR (oxygen consumption rate) in the KOC7C cell line treated with different concentrations of NKT2152 under 1.5% O 2 for 24 h (control, 1 μM, and 5 μM), as normalized by the cell protein concentration. Downward arrows show the injection points for oligomycin (1 μmol/L), FCCP (2 μmol/L), and rotenone/actinomycin (both 1 μmol/L) from left to right. (N) Quantification of spare respiration in the KOC7C cell line treated with different concentrations of NKT2152 under 1.5% O 2 for 24 h. (O) Quantification of ATP production in the KOC7C cell line treated with different concentrations of NKT2152 under 1.5% O 2 for 24 h. (P) Quantification of maximal respiration in the KOC7C cell line treated with different concentrations of NKT2152 under 1.5% O 2 for 24 h. (Q) Mean fluorescence intensity (MFI) of mitochondrial ROS, normalized to the MFI of Hoechst in the RMG1 cell line treated with different concentrations of NKT2152 under 1.5% O 2 for 24 h (control, 1 μM, and 5 μM). (R) Representative fluorescence images of MitoSOX in the RMG1 cell line treated with different concentrations of NKT2152 under 1.5% O 2 for 24 h (control, 1 μM, and 5 μM), demonstrating mitochondrial ROS levels. (S) Mean fluorescence intensity (MFI) of mitochondrial ROS, normalized to the MFI of Hoechst in the KOC7C cell line treated with different concentrations of NKT2152 under 1.5% O 2 for 24 h (control, 1 μM, and 5 μM). Representative fluorescence images of MitoSOX in the KOC7C cell line treated with different concentrations of NKT2152 under 1.5% O 2 for 24 h (control, 1 μM, and 5 μM), demonstrating mitochondrial ROS levels. (U) Tumor growth curves of xenograft mouse models inoculated with RMG1 cells showing significant antitumor effects of NKT2152 ( n  = 15 in control group and n  = 13 in treatment group). (V) Tumor growth curves of xenograft mouse models inoculated with KOC7C cells showing significant antitumor effects of NKT2152 ( n  = 13, each group). Data are presented as mean ± standard error of the mean. RT‐qPCR data are presented as mean ± standard error of the mean from three independent experiments. * p <.05, ** p <.01, ** p <.001. Compared with the control group, OCR assessment showed a higher production in the RMG1 cell line treated with 1 or 5 μM of NKT2152 under 1.5% O 2 for 24 h (Figure  3I ). Spare respiration, ATP production, and maximal respiration were significantly upregulated in both 1 μM and 5 μM NKT2152‐treated groups ( p <.001 for 1 μM and p  = .001 for 5 μM, Figure  3J ; p  = .010 for 1 μM and p  = .008 for 5 μM, Figure  3K ; p  = .001 for 1 μM and p <.001 for 5 μM, Figure  3L , respectively). Similar results were observed in the KOC7C cell line treated with 1 μM or 5 μM of NKT2152 under 1.5% O 2 for 24 h. In both 1 and 5 μM NKT2152‐treated groups, the OCR level was enhanced (Figure  3M ), and spare respiration, ATP production, and maximal respiration were significantly upregulated ( p  = .022 for 1 μM and p  = .010 for 5 μM, Figure  3N ; p <.001 for 1 μM and p <.001 for 5 μM, Figure  3O ; p  = .001 for 1 μM and p <.001 for 5 μM, Figure  3P , respectively). These findings suggest enhancement of mitochondrial function induced by NKT2152. Mitochondrial ROS production level was also detected under both RMG1 and KOC7C cell lines treated with 1 or 5 μM of NKT2152 under 1.5% O 2 for 24 h. In the RMG1 cell line, both 1 and 5 μM of NKT2152 significantly enhanced mitochondrial ROS production ( p  = .009 for 1 μM and p  = .019 for 5 μM, respectively) (Figure  3Q ). The representative images of each group are presented in Figure  3R . Although the analysis was not significant, mitochondrial ROS production tended to be enhanced in KOC7C cell lines treated with NKT2152 ( p  = .105 for 1 μM and p  = .081 for 5 μM, respectively) (Figure  3S ). The representative images of each group are presented in Figure  3T . To further analyze the clinical therapeutic potential of inhibiting HIF2A in ovarian CCC, we utilized NKT2152 in ovarian CCC CDX models. The tumor growth curves demonstrated a significant anti‐tumor effect in the NKT2152 treatment group compared with the control in both the RMG1 and KOC7C xenograft mouse models ( p  = .015 for RMG1 and p  = .010 for KOC7C; Figure  3U,V , respectively). Tumor weights and sizes were also suppressed in the NKT2152 treatment group, although the difference was not statistically significant ( p  = .105 for RMG1 and p  = .216 for KOC7C, Supplementary Figure  3A , B , respectively). Photos of the tumors are presented in Supplementary Figure  3C , D for RMG1, and Supplementary Figure  3E , F for KOC7C. We generated three PDX models, each with distinct clinical backgrounds, from patients with ovarian CCC treated at our institution and designated them as PDX5, PDX6, and PDX40. PDX5 was obtained from a primary site, and PDX6 was the omental metastasis of PDX5 (PDX5 for Figure  4B–D and PDX6 for Figure  4F–H ) in a 44‐year‐old patient with stage FIGO IVB CCC (Figure  4A–H ). The patient underwent total abdominal hysterectomy (TAH), bilateral salpingo‐oophorectomy (BSO), partial omentectomy (pOM), and appendectomy. The tumors were debulked suboptimally. The tumor size was <10 cm, and ascites cytology was positive. Microscopic findings revealed that the primary tumor was composed of malignant epithelial hobnail cells with a papillary pattern (Figure  4D ). Lymphovascular space invasion (LVSI) was also present. The metastatic sites had malignant epithelial tumors composed of eosinophilic cells with a solid growth pattern (Figure  4H ). Following surgery, recurrence was noted after 2 months. The serum CA125 level at diagnosis was 252.4 U/mL, which increased to 374.6 U/mL upon recurrence. She received several chemotherapy regimens but ultimately died of the disease. PDX40 was a 42‐year‐old female patient diagnosed with early‐stage FIGO IA CCC (Figure  4I–L ). She underwent TAH, BSO, pelvic and para‐aortic lymphadenectomy, and pOM, achieving complete surgical debulking. The microscopic findings revealed that the malignant epithelial tumor was composed of eosinophilic cells with tubulocystic and papillary growth patterns (Figure  4L ). The lymph node and LVSI status were not specified, and no recurrence was reported. NKT2152 treatment resulted in a significant antitumor effect in patient‐derived xenograft (PDX) mouse models. (A)–(H) Representative CT, MRI, and PET‐CT images, along with pathological findings, from PDX patients ID5 and ID6. ID5: Primary site; ID6: Metastatic site. (A) Representative PET‐CT image of PDX patient ID5 (ID6). (B) Representative MRI image of PDX patient ID5. (C) Representative pathological image of PDX patient ID5. (D) Representative H&E staining of tumor tissues from PDX patient ID5. (E) and (F) Representative CT images of PDX patient ID6. (G) Representative pathological image of PDX patient ID6. (H) Representative H&E staining of tumor tissues from PDX patient ID6. (I)–(L) Representative MRI and PET‐CT images, along with pathological findings, from PDX patients 40. (I) Representative PET‐CT image of PDX patient ID40. (J) Representative MRI image of PDX patient ID40. (K) Representative pathological image of PDX patient ID40. (L) Representative H&E staining of tumor tissues from PDX patient ID40. (M) Tumor growth curves of subcutaneous xenograft models established from PDX6 tumors demonstrating significant tumor suppression by NKT2152 ( n  = 7 in each group). (N) Tumor growth curves of subcutaneous xenograft models established from PDX40 tumors demonstrating significant tumor suppression by NKT2152 ( n  = 7 in control group and n  = 6 in NKT2152 treatment group). (O) Tumor figure of the control group of the mouse xenograft models inoculated subcutaneously with PDX6 tumors. (P) Tumor figure of the NKT2152 treatment group of the mouse xenograft models inoculated subcutaneously with PDX6 tumors. (Q) Tumor figure of the control group of the mouse xenograft models inoculated subcutaneously with PDX40 tumors. (R) Tumor figure of the NKT2152 treatment group of the mouse xenograft models inoculated subcutaneously with PDX40 tumors. (S) Tumor weight of the mouse xenograft models inoculated subcutaneously with PDX6 tumors. (T) Tumor weight of the mouse xenograft models inoculated subcutaneously with PDX40 tumors. Data are presented as mean ± standard error of the mean. * p <.05, ** p <.01, ** p <.001. In PDX6 and PDX40, tumor growth in the NKT2152 treatment group was significantly suppressed compared with that in the control group ( p  = .026 for PDX6 and p  = .022 for PDX40; Figure  4M,N , respectively). Photos of the tumors are presented in Figure  4O,P for the control and treatment groups in PDX6, and Figure  4Q,R for the control and treatment groups in PDX40, respectively. The tumor weight in the NKT2152 treatment group was also significantly reduced in PDX6 ( p  = .026, Figure  4S ). There was a reduction in tumor weight in the NKT2152 treatment group in PDX40, but it was not statistically significant (Figure  4T , p  = 0.051). Meanwhile, PDX5 did not suppress tumor growth following NKT2152 administration compared with that in the control group (Supplementary Figure  4A–C ). The tumor weight in the NKT2152 treatment group tended to be lower than that in the control group, but the difference was not significant ( p  = .589, Supplementary Figure  4D ). Interestingly, RNA sequence analysis showed that categories related to inflammation and TNFa signaling were commonly enriched in tumors with stronger responses to HIF2A inhibition under MSigDB Hallmark 2020 pathway analysis ( p  < .001 for “Inflammatory Response,” and p  = .002 for “TNF‐alpha Signaling via NF‐kB” in the DEGs upregulated in PDX40 compared with PDX5, Supplementary Table  10 ; and p  = .001 for “Inflammatory Response,” and p  = .020 for “TNF‐alpha Signaling via NF‐kB” in the DEGs upregulated in PDX6 compared with PDX5, Supplementary Table  11 ). In protein level, IHC of the non‐NKT2152 treated PDX model revealed that the model responsive to NKT2152 treatment showed markedly higher HIF2A expression compared with the non‐responsive model. And among the NKT2152 effective model, PDX6 showed a higher HIF2A expression (PDX6 IRS = 162, PDX40 IRS = 158 and PDX5 IRS = 40). Representative IHC images showing different intensities of HIF2A nuclear staining are presented in Supplementary Figure  4E . These findings indicate the significant anti‐tumor effect of NKT2152 and highlight its potential in being a potential therapeutic target in ovarian CCC.

We systematically demonstrated that the nuclear expression of HIF2A may be a clinically significant prognostic factor in ovarian CCC, providing new insights into its potential as a therapeutic target. Uehara et al. 16 reported that higher HIF2A expression was related to poor outcomes, including PFS, in ovarian CCC, which is aligned with our results. Our study contributed to the current literature by analyzing a larger patient cohort, and it is the first to demonstrate the prognostic value of HIF2A for OS. Aside from ovarian CCC, HIF2A is also linked to a poor prognosis or progression in various cancers, including renal cell carcinoma, 25 neuroblastoma, 26 and HER2‐positive breast cancer. 27 Additionally, high HIF2A expression is significantly associated with poorer outcomes, including OS and PFS, in pan‐cancer. 28 However, the nuclear expression of HIF2A is undetected, indicating the activation of HIF2A signaling. 29 Our findings confirm the results of these studies in that high HIF2A nuclear expression is significantly related to poorer PFS and OS. In ccRCC, HIF2A mRNA levels are significantly upregulated compared to adjacent non‐malignant tissues. 30 Consistent with our data, Michael et al. 9 previously reported higher HIF2A mRNA expression in ovarian CCC compared to high‐grade serous carcinoma. However, their analysis was limited to a single histology subtype. In contrast, our study included additional subtypes such as endometrioid and mucinous carcinoma, providing a more comprehensive histological context and a clearer HIF2A expression profile across ovarian cancer subtypes. Our study not only confirmed the prognostic significance of HIF2A in ovarian CCC but also clarified its functional role. Unlike previous results that focused on angiogenesis and proliferation, ours highlighted the role of HIF2A in promoting ovarian CCC progression through the regulation of mitochondrial metabolism and ROS production. In both in vitro and in vivo experiments, HIF2A knockdown resulted in enhanced mitochondrial function, including increased maximal respiration, spare respiratory capacity, and ATP production, accompanied by increased ROS levels and suppressed tumor growth. These findings provide a novel insight into the oncogenic role of HIF2A in ovarian CCC. At the molecular level, HIF2A knockdown significantly downregulated the expression of several target genes, including VEGFA , GLUT1 , NDRG1 , and IGFBP3 , consistent with previous findings. 10 , 11 , 12 , 31 HIF2A promotes the progression of several cancers, including ccRCC, 12 neuroblastoma, 10 colon cancer, 32 and cervical cancer. 33 HIF2A also contributes to lineage transitions in cancer‐associated fibroblasts during lung cancer metastasis to the brain, 34 and its mutations are associated with larger tumor sizes in pheochromocytoma and paraganglioma. 35 While these studies established HIF2A as a key factor in several malignancies, its role in ovarian CCC remains unexplored despite the fact that ovarian CCC is highly associated with hypoxia. 3 Unlike ccRCC, wherein HIF2A activation is closely associated with VHL mutations, ovarian CCC has a distinct tumor microenvironment characterized by chronic oxidative stress and metabolic adaptations. 3 Our study provides the first systematic investigation of the role of HIF2A in ovarian CCC and highlights its potential as a therapeutic target. By demonstrating that HIF2A is actively involved in ovarian CCC, future studies focused on its specific contributions to ovarian CCC tumor biology and its potential as a therapeutic target should be conducted. RNA sequencing revealed that HIF2A knockdown was associated with ROS under hypoxic conditions, consistent with the findings of Bertoutet al 36 Mitochondria are the primary source of ROS, with mitochondrial complex III playing a key role in ROS generation. 37 In several cancer types, including breast cancer and OCCC, enhanced mitochondrial activity induces ROS production. 17 , 38 Therefore, we hypothesized that HIF2A depletion results in increased mitochondrial respiration and ROS accumulation and contributes to the inhibition of tumor growth. Although HIF2A regulates mitochondrial function in ccRCC, 39 few studies have evaluated the relationship of HIF2A to both mitochondrial activity and ROS. As mitochondria are a major source of ROS and changes in mitochondrial function are associated with altered mitochondrial ROS production in various cancers, including breast cancer and ovarian CCC, 17 , 38 our study is the first to establish a direct connection between HIF2A suppression and mitochondrial activity as well as mitochondrial ROS in ovarian CCC. However, our findings are controversial due to the complicated relationship between mitochondria, ROS, and hypoxia. Moon et al. and Rouault‐Pierre et al. reported that HIF2A knockdown increased ROS levels but damaged mitochondrial function, 40 , 41 which is in contrast with our findings. A possible consideration is the different types of cells involved. Moon et al. 40 studied pancreatic beta cells, and Rouault‐Pierre et al. 41 focused on myeloid leukemia cells, whereas we focused on the solid tumor type. Nevertheless, our study provides novel insights into the role of HIF2A in regulating mitochondrial function in ovarian CCC. Our findings not only clarify the biological role of HIF2A in ovarian CCC but also lay a theoretical foundation for the development of HIF2A‐targeted therapeutics. Our study is the first to evaluate the efficacy of a novel HIF2A inhibitor, NKT2152, in preclinical ovarian CCC models, including both CDX and PDX models, revealing its significant anti‐tumor effects and clinical therapeutic potential. Previous studies have reported on the anti‐tumor effect of other HIF2A inhibitors in ccRCC, Pacak–Zhuang syndrome, pheochromocytoma, paraganglioma, and breast cancer, 38 , 42 , 43 , 44 , 45 , 46 with one being approved for use by the US Food and Drug Administration. 42 Although the clinical data have not been published, the initiation of a phase II trial of belzutifan, a HIF2A inhibitor, for ovarian CCC underscores the translational value of our study and provides further justification for exploring HIF2A inhibition ( NCT06677190 ). NKT2152 has significant anti‐tumor effects in ccRCC and hepatocellular cancer and has begun clinical trials, including combination therapies ( NCT05935748 , NCT05119335 , and NCT04524871 ). These pre‐clinical studies may possibly encourage the conduction of clinical trials regarding NKT2152 use in ovarian CCC. In addition, our findings suggest that baseline HIF2A expression levels may influence the therapeutic response to HIF2A inhibition. Tumors with higher HIF2A expression tended to show better responses to NKT2152, implying that HIF2A expression may serve as a potential biomarker for patient selection in future clinical applications. However, further studies are warranted to explore the broader clinical and biological implications of NKT2152. The observed accumulation of mitochondrial ROS following HIF2A suppression indicates a potential metabolic vulnerability in ovarian CCC that may be therapeutically exploited. However, as our evaluation of NKT2152 was limited to three PDX models, larger studies incorporating a more diverse set of patient‐derived tumors are needed to account for interpatient heterogeneity. Moreover, identifying predictive biomarkers for patient selection and optimizing treatment conditions will be important to maximize the clinical utility of HIF2A‐targeted therapies in ovarian CCC. In conclusion, our findings underscore the clinical and functional significance of HIF2A in ovarian CCC, support its potential as a therapeutic target, and provide a robust theoretical foundation for future clinical trials.

Ovarian clear cell carcinoma (CCC) is the main subtype of epithelial ovarian cancer that has an aggressive phenotype and poor prognosis. 1 The incidence of ovarian CCC in Japan has been increasing and is currently at 25%. 2 Ovarian CCC is associated with endometriosis. 3 Endometriotic cysts contain blood that expose cells to hypoxia, iron abundance, and oxidative stress. 3 , 4 , 5 The stressful carcinogenic environment results in resistance to conventional chemotherapy and ferroptosis and contributes to its aggressive nature. 5 , 6 , 7 , 8 The behavior of ovarian CCC is influenced by several mechanisms, including the IL6‐STAT3‐HIF pathway and hypoxic cell growth. 6 , 9 Few molecular targeted therapies have been approved for ovarian CCC; therefore, novel therapeutic strategies are important for improving its prognosis. Hypoxia is a hallmark feature of several cancers. During hypoxic conditions, hypoxia‐inducible factors (HIFs), mostly containing HIF1A and HIF2A, may be active and form a complex with HIF1B, which translocates to the nucleus and activates downstream genes, including vascular endothelial growth factor A ( VEGFA ), N‐myc downstream regulated 1 ( NDRG1 ), glucose transporter 1 ( GLUT1 , also named as solute carrier family 2 member 1), and insulin like growth factor binding protein 3 ( IGFBP3 ), 10 , 11 , 12 that regulate various processes and promote cancer development. Additionally, HIFs may also regulate mitochondrial function. 13 HIF‐induced mitochondrial alterations, such as reduced oxidative capacity, disrupted biosynthesis, apoptosis, fission, and autophagy, result in mitochondrial dysfunction, further supporting tumor cell adaptation and promoting malignancy. 14 , 15 Among the HIF family, HIF2A is highly expressed in ovarian CCC, contributing to the poor prognosis 9 , 16 and suggesting that it may be a therapeutic target. A HIF2A inhibitor (Belzutifan) is currently being evaluated in a phase II clinical trial for recurrent or persistent ovarian CCC ( NCT06677190 ). Mitochondria play an important role in the behavior of ovarian CCC. 7 Pyruvate dehydrogenase kinase isoform 2 suppresses mitochondrial function and reduces mitochondrial reactive oxygen species (mitochondrial ROS) levels, inducing resistance to conventional chemotherapy in ovarian CCC. 17 Meanwhile, inhibiting hypoxic signaling can restore mitochondrial function; for example, HIF2A inhibition using α ‐ketoglutarate in retinal pigment epithelial cells restored mitochondrial activity and increased mitochondrial ROS production. 18 These findings indicate that HIFs may promote cancer aggressiveness by regulating mitochondrial function and ROS production. Thus, future studies on how HIFs react with mitochondrial function and mitochondrial ROS production are essential in ovarian CCC. NKT2152 is a compound that inhibits HIF2A by blocking its interaction with HIF1B, thereby suppressing HIF2A target genes. In clear cell renal cell carcinoma (ccRCC) and hepatocellular cancer, NKT2152 has demonstrated significant antitumor effects and is currently being evaluated in clinical trials [ NCT05935748 , NCT05119335 , and NCT04524871 ]. As there are similarities between ovarian CCC and ccRCC, 19 it is reasonable to evaluate the potential antitumor effects of NKT2152 in ovarian CCC. This study aimed to evaluate the functional roles that HIF2A plays in ovarian CCC, assess its clinical implications, and explore the therapeutic potential of NKT2152 for ovarian CCC. We also aimed to provide a theoretical foundation to support future research on targeted therapies in ovarian CCC.

Hairong Wei, Zhihong Liu, and Zhenhai Gao are employees and shareholders of NiKang Therapeutics, from which they receive salary. NiKang Therapeutics provided the HIF2A inhibitor (NKT2152) for this research. Ken Yamaguchi and Masaki Mandai reported receiving personal fees from Dumsco Inc. outside the submitted work. Junzo Hamanishi reported receiving Lecture fee from MSD, Eizai, and Astrazeneca; Research fund from Kinopharma, Sumitomo Pharma, ONO, Chugai outside the submitted work. Other authors have no declaration to disclose.

The details are described in the Supplementary Materials and Methods. Surgical specimens were obtained from patients with ovarian CCC who underwent surgery at Kyoto University Hospital between 2001 and 2017. Clinical data, including age, FIGO stage, progression‐free survival (PFS), and overall survival (OS), were retrospectively collected from medical records. For additional validation, PDX tumor samples without treatment of NKT2152 were also subjected to IHC to evaluate HIF2A protein expression. IHC staining was performed on formalin‐fixed paraffin‐embedded tissues using an anti‐HIF2A antibody (NB100‐132, JaICA; 1:300, RRID: AB_10000898), with the antigen retrieval conditions presented in Supplementary Table  1 . The immunoreactive score (IRS) was calculated as the intensity (0–3) × percentage (0%–100%), and groups were defined using a cutoff of 180 (Supplementary Figure  1A ). Two human ovarian CCC cell lines, RMG1 (RRID: CVCL_1662; JCRB1072) and KOC7C (RRID: CVCL_5307; Tottori University), were used in both in vivo and in vitro experiments. All cell lines were authenticated using short tandem repeat profiling within the last 3 years. All experiments were performed with mycoplasma‐free cells. The culture conditions of the cell lines are presented in the Supplementary Materials and Methods. A stable RMG1 knockdown cell line was created via lentiviral transfection using distinct lentiviral shRNAs targeting HIF2A and a nonsilencing control. Shcontrol (VB010000‐0009mxc), shHIF2A‐1 (VB900041‐1610kkn), and shHIF2A‐2 (VB900041‐1616spy) were purchased from Vector Builder. The detailed transfection method was reported previously. 8 The cells transfected by shHIF2As and shRNA targeting a nonsilencing control are represented as shHIF2A‐1, shHIF2A‐2, and shcontrol, respectively. The shRNA sequences are presented in Supplementary Table  2 . Small interfering RNA (siRNA) targeting HIF1B (siHIF1B) was purchased from Invitrogen (Cat. No. 10620318 and Cat. No. 10620319). A non‐targeting control siRNA pool (siGENOME Non‐Targeting siRNA Control Pool #1, Cat. No. D‐001206‐13‐05; Dharmacon) was used as a negative control. The RNA sequences are presented in Supplementary Table  3 . SiRNA was transfected with RMG1 cell line using Lipofectamine 3000(L3000‐015, Invitrogen, Thermofisher scientific). Cell proliferation was evaluated using the WST‐8 assay (Nacalai Tesque) under normoxic conditions. For qRT‐PCR and RNA sequencing, RMG1 and KOC7C cells treated with 1 μM or 5 μM of NKT2152 and shRNA cells (shHIF2A‐1, shHIF2A‐2, and shcontrol cells) under 1.5% O 2 for 24 h were used for RNA extraction. Total RNA was extracted from the original ID5, ID6, and ID40 patient‐derived xenograft (PDX) mouse tumor less than passage 4 without NKT2152 treatment and submitted for RNA sequence. The detailed total RNA extraction method was similar to that in a previous study. 20 The mRNA expression of HIF2A , IGFBP3 , VEGFA , NDRG1 , GLUT1 , and ACTB was evaluated by qRT‐PCR. The primer sequences used in these studies are presented in Supplementary Table  4 . RNA‐seq and differential gene expression analyses were performed according to our previous publication. 21 The processed reads were aligned to the human reference genome GRCh38 or the mouse reference genome GRCm39 using STAR (version 2.7.10a) and quantified with RSEM (version 1.3.1). For in vitro analysis, differentially expressed genes (DEGs) were identified between shcontrol ( n  = 3) and shHIF2A‐1 ( n  = 3) or shHIF2A‐2 ( n  = 3). Genes consistently upregulated in both comparisons were defined as overlapping DEGs (Supplementary Table  5 ), with full DEG lists presented in Supplementary Tables  6 and 7 . A baseMean cutoff was not employed due to the limited DEG yield. For PDX samples, DEGs were identified between the HIF2A inhibitor response group (PDX6 and PDX40, n  = 3 for each) and the nonresponse group (PDX5, n  = 3). A baseMean cutoff of >50 was employed to reduce background noise and improve downstream robustness. DEG results for PDX6 versus PDX5 and PDX40 versus PDX5 were analyzed independently, with results listed in Supplementary Tables  8 and 9 . RNA‐seq data are available in the Gene Expression Omnibus ( GSE297968 for shHIF2A cell line and GSE298088 for PDX tumor), and quality metrics are presented in Supplementary Table  10 . Genes with |log 2 FoldChange|≤ 2 or Padj ≥0.01 were excluded. To ensure interpretability, genes lacking valid HUGO symbols were also removed, with negligible effect on the final DEG results. All DEG sets underwent enrichment analysis using Enrichr. 22 For the shHIF2A dataset, enriched pathways were analyzed based on overlapping DEGs between shHIF2A‐1 and shHIF2A‐2 (Table  2 ). For the PDX dataset, only the commonly upregulated pathways identified in both PDX6 versus PDX5 and PDX40 versus PDX5 were considered (Supplementary Tables  11 and 12 ). Gene expression data were obtained from publicly available microarray datasets, including GSE29450 , GSE39204 , GSE29175 , and GSE65986 . All datasets were analyzed using the Affymetrix Human Genome U133 Plus 2.0 Array or the Affymetrix Human Genome U133A 2.0 Array, and we selected 200878_at as the probe representing HIF2A mRNA expression for subsequent analysis. The dataset from Washington University was analyzed using Kaplan–Meier and Cox regression analyses. 23 To identify the cutoff value, the 192 clinical samples were divided into 13 groups based on their ssGSEA. Based on the peak value, the cutoff was set at 2 (Supplementary Figure  1B ), and the samples were divided into the low HIF2A (ssGSEA ≤2, n  = 111) and high HIF2A (ssGSEA >2, n  = 91) pathway related groups. RMG1 cell line incubated under normoxic or 1.5% O 2 hypoxia condition for 24 h was collected in the 2% SDS (sodium dodecyl sulfate) lysis buffer. Shcontrol and shHIF2A cells treated with 5 μM MG132 (Cat# 1211877‐36‐9, Selleck, Japan) under 1.5% O 2 for 24 h were also lysed in 2% SDS lysis buffer. Lysates were heated at 100°C for 20 min, and protein concentrations were assessed using the bicinchoninic acid assay. Equal amounts of protein were resolved on 4%–12% Bis‐Tris gels and transferred to PVDF membranes (#1620177, BIO‐RAD). After blocking, the membranes were incubated overnight at 4°C with primary antibodies against HIF2A (D9E3, #7096, 1:1000 dilution, Cell Signaling Technology [CST], RRID: AB_10898028) and ACTB (#4967, CST, RRID: AB_330288 1:1000 dilution, Cell Signaling), followed by HRP‐conjugated secondary antibody (#7074, CST, RRID: AB_2099233,1:3000 dilution). Signals were visualized using a chemiluminescent substrate (#34580, Thermo Fisher) and imaged using the ChemiDoc XRS+ system (Bio‐Rad). For cell line‐derived xenografts (CDX), 5–6‐week‐old female athymic BALB/C nude mice (CLEA Japan, Tokyo) were subcutaneously injected with 1 × 10 6 shcontrol or shHIF2A cells using RMG1, 5 × 10 6 KOC7C cells, or 1 × 10 6 RMG1 cells in the abdominal region. For PDX, tumor fragments from patients with ovarian CCC (PDX5, PDX6, and PDX40) were implanted subcutaneously into 6–7‐week‐old female NOG mice (In Vivo Science Inc., Japan). When tumor volumes reached approximately 100 mm 3 (CDX) or diameters reached approximately 5 mm (PDX), mice were randomized into control and treatment groups. The treatment group received 20 mg/kg of the HIF2A inhibitor (NKT2152 [Nikang Therapeutics, USA]) via oral gavage twice daily; controls received vehicle (0.5% methyl cellulose; Wako, Japan). Tumor sizes were measured twice weekly and calculated as (length × width 2 )/2. Treatment continued until a statistically significant difference in tumor volume was observed between the groups or until the tumors reached 2 cm in diameter. The final group sizes varied due to tumor growth and randomization as presented in the figure legends. Mitochondrial stress tests were performed using the Agilent Seahorse XF96 Extracellular Flux Analyzer (Agilent Technologies, CA, USA) following the manufacturer's instructions. The parameters were calculated using Wave software (Agilent Technologies). Mitochondrial ROS levels and membrane potential were assessed using the MitoSOX Red (Thermo Fisher Scientific K.K., Tokyo, Japan) following standard protocols. All statistical analyses were performed using GraphPad Prism (RRID: SCR_002798). Survival differences were assessed using the Kaplan–Meier method with the log‐rank test. Cox regression was used for the univariate and multivariate analyses. Grouped data are presented as mean ± standard error of the mean. The Mann–Whitney U test, one‐way analysis of variance with Dunnett's post hoc test, or Kruskal–Wallis test with Dunn's correction was performed based on data distribution and sample size. A p ‐value <.05 was considered statistically significant.

Data S1: Supporting Information