Clinicopathological correlation of ARID1A status with HDAC6 and its related factors in ovarian clear cell carcinoma

Ovarian clear cell carcinoma (OCCC) is associated with a frequent loss in ARID1A function. ARID1A reportedly suppresses histone deacetylase (HDAC)6 in OCCC directly. Here, we evaluated the clinical significance of HDAC6 expression and its related factors in terms of ARID1A status. Immunohistochemical expression of HDAC6, hypoxia inducible factors-1α (HIF-1α), programmed death-1 ligand (PD-L1), CD44 (cancer stem cell marker), and ARID1A was analysed for 106 OCCC patients. High nuclear HDAC6 expression correlated with patient death (p = 0.038). In the multivariate analysis of overall survival, surgical status (complete or incomplete resection) (hazard ratio (HR) = 17.5; p = <0.001), HDAC6 nuclear expression (HR = 1.68; p = 0.034), and PD-L1 expression (HR = 1.95; p = 0.022) were the independent prognostic factors. HDAC6 upregulation and ARID1A loss did not necessarily occur simultaneously. High HDAC6 expression was associated with poor prognosis in OCCC with ARID1A loss; this was not observed without ARID1A loss. HDAC6 expression showed a significant positive correlation with HIF-1α, PD-L1, and CD44. In OCCC, HDAC6 involvement in prognosis depended on ARID1A status. HDAC6 also led to immuno- and hypoxia- tolerance and cancer stem cell phenotype. HDAC6 is a promising therapeutic target for OCCC with loss of ARID1A.

Scientific RepoRts | (2019) 9:2397 | https://doi.org/10.1038/s41598-019-38653-0 of programmed death-1 ligand (PD-L1), which leads to cancer immune tolerance 17 . Hypoxia inducible factor-1α (HIF-1α) protein expression, transcriptional activity 18 , and tumour angiogenesis 19 are induced by HDAC6, and the cancer stem cell phenotype is maintained by HDAC6 via CD44 20 . HDAC6-selective inhibitors are currently in clinical trials for multiple myeloma 21,22 . Recently, Bitler et al. 23 identified that ARID1A directly suppresses HDAC6 in OCCC and provided evidence that HDAC6 may be a promising therapeutic target in ARID1A-mutated cancers. However, the significance of the association between HDAC6 and ARID1A has not been well documented in clinical samples. Herein, we investigated the significance of HDAC6 as a therapeutic target in OCCC and the usefulness of ARID1A as its biomarker using clinical samples.

Results
Clinicopathological characteristics and immunohistochemical (IHC) expressions. Table 1 lists the characteristics of the patients. The age of patients ranged from 32 to 80 years, with the average being 55.7 years. All patients included in the study were Japanese. Patients with OCCC were classified after surgery as International Federation of Obstetrics and Gynaecology (FIGO) stage I (n = 71), stage II (n = 16), stage III (n = 17), and stage IV (n = 2). A total of 90 (84.9%) patients underwent complete surgical resection, while 85 (80.2%) patients underwent adjuvant chemotherapy after surgery. Lymphadenectomy was performed in 62 patients (58.5%), 9.7% of which had lymph nodes metastasis. All adjuvant chemotherapies administered to patients contained platinum-agents, 84.7% of which were combined with taxane-agents. The median follow-up duration was 54.2 months for survivors (range, 10-121 months). Recurrence and death were observed in 32 (30.2%) and 23 (21.7%) patients, respectively. A positive expression of ARID1A was observed in 54 patients (50.9%), while a high expression of HDAC6 was observed in 59 (55.7%) and 20 (18.9%) (nucleus and cytoplasm, respectively) patients; high expression was observed in 60 (56.6%), 7 (6.6%), and 24 (22.6%), patients for HIF-1α, PD-L1, and CD44, respectively. The correlations between patient characteristics and IHC expressions are shown in Table 2. No correlation was found between all IHC expressions (HDAC6, HIF-1α, PD-L1, ARID1A, and CD44) and age, FIGO stage, and surgical status. A high expression of CD44 was correlated with recurrence (p = 0.018), while a high expression of HDAC6 (nucleus), HIF-1α, PD-L1, and CD44 correlated with death (p = 0.038, 0.047, 0.038, and 0.036, respectively). There was no significant correlation between ARID1A loss and any of the available clinicopathological parameters.
Correlation with survival and IHC expressions. In the univariate analysis using the Cox proportional hazards model, high expression of PD-L1 and CD44, FIGO stage, and surgical status were found as the prognostic factors for progression free survival (PFS) and overall survival (OS) ( Table 3). In the multivariate analysis, high expression of HIF-1α (hazard ratio (HR) = 1.75; 95% CI, 1.17 to 2.61, p = 0.006), PD-L1 (HR = 2.34; 95% CI, Figure 1. Scheme of HDAC6 functions: When ARID1A loss occurs, suppression of HDAC6 is relieved. In the nucleus, HDAC6 destabilizes p53 by deacetylation, and suppresses apoptosis. As a member of the HDAC family, HDAC6 inactivates chromatin by deacetylation of the core histones. These effects of HDAC6 in the nucleus are responsible for the resistance to platinum agents. In the cytoplasm, HDAC6 leads to cell proliferation, angiogenesis, glucose metabolism, glucose transport, and CSC phenotypes via tubulin, HIF-1α, PD-L1, and CD44. In addition, HDAC6 result in tolerance to taxane agents, cytotoxic T cells, immuno-checkpoints inhibitors, and hypoxic stress.  Correlation among IHC expressions. ARID1A loss also showed a significantly positive correlation with the high expression of PD-L1 (Table 4); however, this was not observed with the high expression of HDAC6 (nucleus, p = 0.431, Fig. 3A; cytoplasm, p = 0.258, Fig. 3B) and HIF-1α (p = 0.510, Fig. 3C). The nuclear high expression of HDAC6 also showed a significantly positive correlation with HIF-1α (p = <0.001, Fig. 3D). The cytoplasmic high expression of HDAC6 showed a significantly positive correlation with PD-L1 (p = 0.010, Fig. 3E) and CD44 (p = 0.043, Fig. 3F).

Discussion
In the present study, OCCC patients with high nuclear expression of HDAC6 had a poor prognosis regardless of FIGO stage and surgical status, the latter of which is a well-known important prognostic factor in EOC. These results suggest that HDAC6 is one of the refractory factors to the standard treatments in OCCC. The standard chemotherapy for EOC is a combination of platinum and taxane agents; however, OCCC patients are resistant to this combination. The deacetylation of alpha-tubulin, induced by HDAC6, decreases the effect of taxane agents as a microtubule-stabilizing agent 24 . When HDAC6 is inhibited, taxane resistance is reversed in EOC cell lines 24,25 . HDAC6 upregulation leads to tumour cisplatin resistance, and depletion of HDAC6 enhances cisplatin-induced DNA damage and apoptosis 26 . HDAC6-selective inhibitors exhibit an anti-tumour effect in breast cancer 27,28 , gastric cancer 19 , multiple myeloma 21,22 , and lymphoma 29 . Therefore, we suggest that HDAC6 is a potentially key therapeutic target for OCCC. Notably, HDAC6-selective inhibitors are well-tolerated and show minimal toxicity in clinical trials 21,22 . HDAC6-selective inhibitors may therefore improve the efficacy and adverse effects such as kidney failure 30 and peripheral neuropathy 31 that often accompany the standard chemotherapy for EOC. The present study also showed the coexistence of an upregulation in HDAC6 and ARID1A loss, leading to a shorter survival for OCCC patients than for patients having either one of the two factors; these activities do not necessarily happen simultaneously. Bitler et al. 23 showed that an HDAC6-selective inhibitor (ACY-1215, ricolinostat) suppressed the proliferation of ARID1A-mutated OCCC cell lines and improved the survival of mice bearing ARID1A-mutated OCCC compared to that of mice bearing ARID1A-wild type OCCC. Fukumoto et al. 32 also reported that the pan-HDAC inhibitor improved the survival of mice bearing ARID1A-mutated OCCC, while Gupta et al. 33 demonstrated that HDAC inhibitors responded to ARID1A-mutated urothelial cancers when compared to that in ARID1A-wild type urothelial cancers, in two clinical trials. These observations therefore indicate that ARID1A status is an important biomarker in the treatment of HDAC. A high PD-L1 expression demonstrated a positive correlation with high HDAC6 cytoplasmic expression and ARID1A loss. When HDAC6 is inhibited, immunotherapy response is enhanced with PD-L1 blockage [34][35][36] . Shen et al. 37 showed that treatment with an anti-PD-L1 antibody reduced tumour burden and prolonged survival of ARID1A-mutated mice; this was not observed in the ARID1A-wild type EOCs. HIF-1α and CD44 also showed a positive correlation with HDAC6. Therefore, HDAC6 may serve as a therapeutic target for OCCC with ARID1A loss associated with PD-L1, HIF-1α, and CD44 expression. Given that a loss in ARID1A is frequent in cancers 9 , the present findings may have implications beyond the established OCCC. Our study had several limitations. The sample size used in this study was small, and the survival analysis was only performed with a few events. However, when considering the low incidence of OCCC, the present study included a relatively large number of patients. Before drawing a conclusion based on the results of this study,   further confirmation is warranted via a multi-ethnic population study and on a larger scale. Secondly, the present study consisted solely of semi-quantitated IHC analysis and lacked both quantitative protein analysis and molecular correlations; the molecular correlations in OCCC between ARID1A and HDAC6 have already been reported 23 . The novelty of the present study is its verification of the findings in clinical samples. However, further studies are required to quantitatively analyse HDAC6 protein and mRNA expression.
In conclusion, the involvement of HDAC6 in OCCC prognosis was demonstrated to depend on the ARID1A status. HDAC6 was observed to function as a promising therapeutic target for OCCC with ARID1A loss, in close association with immuno-modulation, response to hypoxia, and cancer stem cell phenotype. HDAC6-selective inhibitors are expected to have safe and synergistic effects when combined with the current standard chemotherapy for EOC.

Patients and samples. Patient electronic medical charts from the Saitama Medical University International
Medical Centre from 2007 to 2016 were reviewed under the approval of the institutional review board (IRB number, 16-257). All methods were performed in accordance with the 1975 Declaration of Helsinki. All tumour specimens in the pathological analysis were obtained with informed consent (or a formal waiver of consent) with an approval from the Ethics Committee of our hospital. We recruited 106 patients with OCCC, whose tumours were surgically resected and pathologically confirmed. The clinicopathological characteristics of these cases, such as age, recurrence/PFS, death/OS, FIGO stage, surgical status (complete resection or incomplete resection), and treatment methods were reviewed. IHC staining. IHC expression of HDAC6, ARID1A, HIF-1α, PD-L1, and CD44 was analysed using tissue microarray (KIN-2, AZUMAYA, Tokyo, Japan). Tissue microarray was generated from 2 cylindrical cores that were 3.0 mm in diameter in each block; these were punched out of paraffin-embedded tissue blocks corresponding to the representative histological findings, and then inserted into a recipient block. A total of 106 tissue blocks were cut into 4-μm serial sections, and were each run through an automated system by Dako Autostainer Link 48 (Agilent technologies, CA, USA) as per the manufacturer's protocol. The following primary antibodies were used: polyclonal rabbit anti-HDAC6 (dilution, 1:500; ab1440, Abcam, Cambridge, UK), polyclonal rabbit anti-HIF-1α (dilution, 1:100; NB100-479; Novus Biologicals, CO, USA), monoclonal rabbit anti-PD-L1 (dilution, 1:100; 28-8 pharmDx; Dako North America, CA, USA), monoclonal rabbit anti-ARID1A (dilution, 1:1000; ERP13501-73; Abcam, Cambridge, UK), and monoclonal mouse anti-CD44 (dilution, 1:200; 156-3C11; Thermo Fisher Scientific, MA, USA). For all antibodies, the Target Retrieval Solution (pH 9.0, HDAC6 and CD44; pH 6.0, Interpretation of IHC results. IHC evaluation was performed by one pathologist (Masanori Yasuda) and one physician (Mitsutake Yano) with subspecialties in gynaecological oncology; both of them were blinded to the clinicopathological parameters (Fig. 4). The following four-tiered scoring scheme was used: negative (0%), weak (1-50%), moderate (51-80%), and marked (81-100%). To optimize the PFS and OS differences, the raw data were binarised for statistical analysis as follows: in HDAC6, HIF-1α, PD-L1, and CD44. The moderate and marked expression were grouped as high-level, whereas the completely negative and weak expression were grouped as low level. For ARID1A, the completely negative expression was categorised as the loss-group, whereas the weak, moderate, and marked expressions were categorised as positive groups. This categorisation was based on the evidence that the complete absence of ARID1A expression is significantly correlated with its mutation status 38 .
Statistical analysis. IHC expressions and the clinicopathological parameters were assessed using the Pearson chi-square test or the Fisher exact test. Spearman's rank correlation was used to determine whether there was a positive or a negative correlation between the factors. Univariable survival analysis was performed by generating Kaplan-Meier curves, and differences between the groups were assessed using the log rank statistic. Univariable and multivariable survival analyses were performed using the Cox proportional hazards model. All analyses were performed using SPSS v24.0 (SPSS Inc, IL, USA). P values < 0.05 were considered significant.