The cBioPortal (http://www.cbioportal.org/) (14,15) was used to retrieve public whole exome sequencing data and mRNA sequencing data for ovarian serous carcinoma. Initially, ‘Ovary/Fallopian Tube’ was selected in the ‘Select Studies for Visualization & Analysis’ and the Ovarian Serous Cystadenocarcinoma dataset [The Cancer Genome Atlas (TCGA), PanCancer Atlas; https://www.cancer.gov/tcga] was chosen. This dataset from TCGA contains whole exome sequencing data from 585 cases, and mRNA sequencing data from 300 cases of ovarian serous carcinoma. Genomic alteration, mRNA expression and survival data were analyzed via the cBioPortal website by submitting a query regarding CRFs, including ARID1A, AT-rich interaction domain 1B, actin-like protein 6A (ACTL6A), SMARCA1, SMARCA2, SMARCA4, SMARCA5, SMARC subfamily b member 1 (SMARCB1), SMARC subfamily c member 1 (SMARCC1), SMARCC2, SMARC subfamily d member 1, SMARC subfamily e member 1, helicase-like transcription factor, chromodomain-helicase-DNA-binding protein 1 (CHD1), CHD2, CHD3, CHD4, CHD5, inositol-requiring 80 and bromodomain-containing protein 9. The genetic alterations in the ‘OncoPrint’ module were analyzed and mRNA expression levels in the ‘mRNA’ module of ‘Comparison/Survival’ were displayed.

Moreover, the association of individual CRF genes with prognosis was compared, and patients within the dataset were categorized into two groups: One with CRF gene amplification, and the other without genetic alteration of CRFs. The prognostic value of individual mRNA expression levels of CRF genes in the two groups was compared using median normalized RNA-seq by expectation maximization values (cut off values: ACTL6A, 2,117; SMARCC2, 3,607; CHD4, 6,891). Survival curves were constructed using the Kaplan-Meier method, and the log-rank test was performed using the ‘Survival’ module of ‘Comparison/Survival’ on cBioportal to analyze overall survival (OS).

The present retrospective study adhered to the principles outlined in The Declaration of Helsinki. This study was approved by the Ethics Committee of Kyushu University (Fukuoka, Japan; approval nos. 21120-01, 21037-02 and 23005-00). The case records of the Department of Anatomic Pathology, Kyushu University from 1988-2020 were accessed to identify cases of ovarian serous carcinoma with available clinical data and formalin-fixed and paraffin-embedded (FFPE) blocks of ovarian tissue. This search yielded 318 cases of surgically resected ovarian serous carcinoma. Cases other than primary ovarian cancer (12 cases involving fallopian tube and 9 cases involving peritoneal cancer) were excluded, as were cases where neoadjuvant chemotherapy had been administered (91 cases). All cases were independently reviewed by two pathologists (NM and TI). The typical histological structure of OHGSC was confirmed, including papillary or solid proliferation, severe nuclear atypia and frequent mitotic figures, in accordance with the World Health Organization Classification of Female Genital Tumors (16). Additionally, three cases diagnosed with low-grade serous carcinoma were excluded. Ultimately, this analysis included 203 cases of OHGSC. All tumor samples were FFPE. For both immunohistochemistry (IHC) and copy number assays, one representative FFPE block was used for each case. Clinical data, including age, International Federation of Gynecology and Obstetrics (FIGO) 2014 stage (17), presence of metastases, adjuvant therapy and prognosis, were obtained from medical records. Furthermore, one normal skeletal muscle tissue was collected from the Department of Anatomic Pathology, Kyushu University as the control for copy number assays.

The primary antibodies used for IHC staining are listed in Table SI. Staining conditions were optimized by testing numerous approaches. FFPE tissue was cut into 3 µm sections for further processing. The paraffin-embedded sections were deparaffinized in xylene and rehydrated in ethanol series (99, 90 and 80%). Antigen retrieval was performed by boiling the slides at 98 or 110°C in Target Retrieval Solution (pH 9.0; Dako; Agilent Technologies, Inc) for ARID1A, SMARCA4, SMARCB1, SMARCC2, tri-methylation of lysine 27 of histone H3 (H3K27me3) and p53 staining, while 10 mM sodium citrate (pH 6.0) was used for SMARCA2, CHD4 and ACTL6A. Subsequently, 3% hydrogen peroxide was used for blocking endogenous peroxidase activity at room temperature for 5 min and PBS was used for washing. Sections were then incubated with the primary antibodies at room temperature for 90 min (CHD4, ACTL6A, H3K27me3 and p53) or at 4°C overnight (ARID1A, SMARCA2, SMARCA4, SMARCB1 and SMARCC2). The EnVision-kit (Dako; Agilent Technologies, Inc) was used for ARID1A, SMARCA2, SMARCB1, SMARCC2, H3K27me3, CHD4, ACTL6A and p53 staining, and the EnVision Flex-kit (Dako; Agilent Technologies, Inc) was used for SMARCA4 staining to achieve better specificity, in accordance with the manufacturer's instructions. Sections were dehydrated in an ethanol series (95, 99 and 100%) and cleared in xylene. Nuclear staining of stromal cells and vascular endothelium was used as a positive internal control and stroma was used as a negative internal control to evaluate the staining. In several cases, p53 immunostaining was performed to support histological diagnosis. The evaluation of IHC slides was conducted using a light microscope by two pathologists (NM and TI) who were blinded to the details of the patients.

The expression of ARID1A, SMARCA2, SMARCA4, SMARCB1, SMARCC2 and H3K27me3 was considered ‘lost’ when there was a complete absence of nuclear staining in tumor cells, while the surrounding normal cells exhibited consistently preserved nuclear staining. By contrast, ACTL6A, SMARCC2 and CHD4 expression were evaluated using an H-score, calculated by multiplying the proportion and intensity of tumor cells displaying nuclear staining. The proportion score was determined by assessing the percentage of tumor cells with positive nuclear staining relative to all tumor cells on the slide (0–100%). The intensity score was assessed using the intensity of the nuclear staining, categorized as follows: 0, not stained; 1, weak; 2, moderate; and 3, strong. The resulting H-score ranged from 0–300. The threshold value between the high and low protein expression groups was determined using the median H-score. Notably, stromal cells and vascular endothelium exhibited positivity and served as internal positive controls across all cases.

Tumor DNA from each of the 203 cases was extracted from FFPE blocks using the DNAstorm FFPE Kit (Biotium, Inc.), according to the manufacturer's protocol. However, the copy number assay could not be conducted in some cases due to limited sample volume. Tumor DNA was successfully extracted in 154 cases for ACTL6A, 143 cases for SMARCC2 and 140 cases for CHD4. Normal DNA extracted from a normal skeletal muscle tissue was used as the control. TaqMan™ Copy Number Assay (cat. no. 4400291; Thermo Fisher Scientific Inc.) was used to conduct DNA copy number analysis. A predesigned primer and probe mix specific for ACTL6A (Assay ID. Hs02294862_cn), SMARCC2 (Assay ID. Hs02933711_cn) and CHD4 (Assay ID. Hs02108296_cn) was used according to the manufacturer's instructions (Thermo Fisher Scientific Inc.). The Rnase P gene (cat no. 4316844; Thermo Fisher Scientific Inc.) was used as the endogenous control. Quantitative polymerase chain reaction (qPCR) was performed with THUNDERBIRD™ Probe qPCR Mix (Toyobo Life Science) using the ∆Cq method (18) on a QuantStudio 3 Real Time PCR System (Thermo Fisher Scientific Inc.). The thermocycling conditions were as follows: Hot start at 50°C for 2 min and 95°C for 1 min; followed by 50 cycles of denaturation at 95°C for 15 sec, annealing at 60°C for 60 sec and extension at 68°C for 30 sec. The relative quantities of DNA obtained from the PCR were analyzed using CopyCaller (Thermo Fisher Scientific Inc.) to determine the copy numbers. Cases with negative results for Rnase P were excluded from the analysis. Copy number alterations (CNAs) were categorized as follows: Copy number <1, deep deletion; 1, shallow deletion; 2, diploid; 3–4, gain and >4, amplification.

All statistical analyses were performed using the JMP statistical software version 17 (SAS Institute, Inc.). The data were analyzed using Wilcoxon's rank-sum test. Multiple comparisons of DNA copy number data and immunohistochemical expression data were analyzed using the Wilcoxon rank-sum test with Bonferroni correction. For survival analysis, Kaplan-Meier analysis and the log-rank test were used. P<0.05 was considered to indicate a statistically significant difference.

The gene alterations of CRFs in ovarian serous carcinoma are summarized in Fig. 1A. Among 585 cases, 57% demonstrated CRF gene alterations, primarily in the form of gene amplification, which was observed in 208 (35.6%) cases. The most prevalent genetic alteration among CRFs in ovarian serous carcinoma was ACTL6A amplification (19.5%). However, this genetic alteration did not significantly affect OS (P=0.620; Fig. 1B). Patients with SMARCC2 amplification (3.1%) had a significantly shorter median OS compared with patients with unaltered SMARCC2 (P=0.005; Fig. 1C). Patients with CHD4 amplification (5.7%) had a notably shorter OS compared with patients with unaltered CHD4; however, this association was not statistically significant (P=0.169; Fig. 1D). Furthermore, there was no significant association between ACTL6A, SMARCC2 and CHD4 mRNA expression levels and OS (Fig. S1A-C). However, ACTL6A mRNA expression was significantly higher in patients with ACTL6A amplification compared with in patients without alterations in the ACTL6A gene (P<0.001; Fig. S1D).

Table I presents the clinical characteristics of patients with OHGSC included in the present histological study. The age of the patients ranged from 28–87 years (mean, 58.3 years; median, 58 years). Out of the 203 cases, 148 (72.9%) had available 5-year clinical follow-up data, with a mean follow-up duration of 61.9 months (1–233 months). The majority of patients (78.4%) presented with advanced stage disease (III–IV). In total, there were 124 cases (61.1%) with recurrence, and 81 cases (39.9%) resulted in disease-related death.

Regarding treatment, all cases underwent surgical resection. Additionally, 195 cases (96.1%) received adjuvant chemotherapy, consisting of various regimens, such as paclitaxel and carboplatin, docetaxel and carboplatin, cyclophosphamide, doxorubicin and cisplatin, cyclophosphamide, epirubicin and cisplatin, or paclitaxel and cisplatin. Furthermore, 9 cases received maintenance treatment, which supplemented chemotherapy with bevacizumab treatment.

Figs. 2, S2 and S3 display representative images of IHC staining, while the summarized IHC results are presented in Table II. ARID1A, SMARCA2, SMARCA4, SMARCB1, SMARCC2 and H3K27me3 were predominantly expressed in the nucleus of the tumor cells (Fig. S2). Nuclear staining of CRFs was lost in 8.9% of cases. Specifically, loss of nuclear ARID1A staining occurred in 2.5% of cases, SMARCA2 in 2.5% of cases and SMARCA4 in 3.9% of cases (Table II; Fig. 2A-C). Additionally, 2.5% of cases exhibited loss of H3K27me3 expression (Table II; Fig. 2D). The nuclear staining of these markers in stromal cells and vascular endothelium served as a positive internal control. Loss of ARID1A, SMARCA2 and SMARCA4 protein expression was mutually exclusive, and the loss of H3K27me3 was not related to loss of the protein expression of these CRFs. However, SMARCC2 and SMARCB1 protein expression was retained in all cases (Table II). The intensity of ACTL6A, SMARCC2 and CHD4 protein expression was higher in tumor cells compared with in stromal cells or lymphocytes (Fig. 2E-G). In addition, the nucleus of OHGSC tumor cells was strongly and diffusely positive for p53 (Fig. 2H).

The H-score of ACTL6A, SMARCC2 and CHD4 ranged from 140–220 (mean 169.66), 140–210 (mean 180.39), and 60–210 (mean 168.82), respectively. The median H-scores: 170 for ACTL6A, 180 for SMARCC2 and 170 for CHD4, were used as the cutoff values for distinguishing low and high CRF expression; representative images of high and low ACTL6A, SMARCC2 and CHD4 staining are shown (Fig. S3A-F).

Fig. 3A presents a summary of the DNA copy numbers of ACTL6A, SMARCC2 and CHD4 in OHGSC. Results were obtained for 154 cases for ACTL6A, 143 cases for SMARCC2 and 140 cases for CHD4, out of a total of 203 cases. For comparison, normal skeletal muscle tissue was used as the control and the copy numbers were normalized to Rnase P, which served as the internal control. The copy numbers of ACTL6A, SMARCC2 and CHD4 ranged from 1–8 (mean 3.09), 1–7 (mean 2.27), and 0–5 (mean 1.34), respectively.

Notably, CNAs with increased copy numbers were predominant in ACTL6A and SMARCC2, while CHD4 CNAs primarily exhibited decreased copy numbers. Among the cases assessed, 102 out of 154 (66.2%) demonstrated ACTL6A copy number gain or gene amplification, 48 out of 143 (33.5%) showed SMARCC2 copy number gain or gene amplification, and 99 out of 140 (70.7%) demonstrated CHD4 shallow deletion or deep deletion. The relationship between ACTL6A, SMARCC2 and CHD4 copy numbers and the immunohistochemical expression of these proteins was assessed. However, no statistically significant association was observed between the copy numbers and the protein expression of ACTL6A (Fig. 3B), SMARCC2 (Fig. 3C) and CHD4 (Fig. 3D).

The survival analyses for copy numbers of CRFs are summarized in Fig. S4, while the association of protein expression with OS is presented in Fig. 4.

Regarding the CNAs of ACTL6A, SMARCC2 and CHD4, changes in copy number of these genes demonstrated no statistically significant association with the OS of patients (P=0.434, P=0.629 and P=0.578, respectively; Fig. S4A-C). Similarly, there was no significant association between the copy numbers of these CRFs and the FIGO stage (P=0.506, P=0.862 and P=0.974, respectively; Fig. S4D-F). However, although not significant (P=0.094), in patients with FIGO stage III/IV OHGSC, copy number gain or amplification in either ACTL6A, SMARCC2 or CHD4 demonstrated unfavorable outcome trends compared with patients with diploid ACTL6A, SMARCC2 and CHD4 (Fig. S4G).

Moreover, although not significant (P=0.128), patients with FIGO stage III/IV with shallow or deep deletions in either ACTL6A, SMARCC2 or CHD4 demonstrated unfavorable outcome trends compared with those with ACTL6A, SMARCC2 and CHD4 diploids (Fig. S4H). However, due to the small number of cases, the effect of CRF CNAs on FIGO stage I/II cases could not be analyzed.

In cases with decreased protein levels of ARID1A, SMARCA2 or SMARCA4, similar outcome trends were demonstrated in OS compared with those with retained expression levels, and there was no statistically significant difference (P=0.879; Fig. 4A). Notably, patients with high ACTL6A protein levels demonstrated a statistically longer OS compared with patients with low ACTL6A protein levels (P=0.027; Fig. 4B).

Regarding the association with FIGO stage, higher SMARCC2 protein expression was detected in patients with a higher FIGO stage, but this relationship was not statistically significant (P=0.198; Fig. 4D). Likewise, the difference between ACTL6A and CHD4 protein levels and FIGO stage was not found to be significantly different (P=0.315 and P=0.775, respectively; Fig. 4C and E).