A DNA Damage Response Gene Panel for Different Histologic Types of Epithelial Ovarian Carcinomas and Their Outcomes.

The study protocol was approved by the National Taiwan University Hospital Research Ethics Committee (201509042RINA, approved on 24 November 2015 and 201608025RINA, approved on 07 October 2016). Informed consent from all participants was obtained and the methods were performed in accordance with the guidelines and regulations. From December 2015 to October 2018, 172 women diagnosed with epithelial ovarian cancer who had received debulking surgery and adjuvant chemotherapy were enrolled. The cancerous tissue specimens collected during debulking surgery were immediately frozen in liquid nitrogen and stored at −70 °C. A portion of the tissue specimens were sent for pathological examinations to confirm the diagnosis and ensure tumorous tissue sufficient for the following experiments. Clinical data were obtained from medical records, including age, cancer stage, the findings during debulking surgery, treatment course and recurrence. Optimal debulking surgery was defined as a maximal residual tumor size <1 cm following surgery. The tumor grade based on International Union Against Cancer criteria, and cancer stage was based on International Federation of Gynecology and Obstetrics (FIGO) criteria [ 21 ]. All patients received platinum-based adjuvant chemotherapy and regular follow-ups after primary treatments. Recurrence was defined as abnormal results from imaging studies (including computerized tomography or magnetic resonance imaging), elevated CA-125 (more than twice the upper normal limit) for two consecutive tests in 2-week intervals, or a biopsy-proven disease. Progression-free survival (PFS) was defined as the time from the date of primary treatment completion to the date of confirmed recurrence, disease progression or last follow-up. Overall survival (OS) was defined as the period from surgery to the date of death related to EOC or the date of last follow-up. We selected 60 genes involved in DNA damage response (DDR) for the gene panel ( Table 1 ), including genes of homologous recombination (HR), nonhomologous DNA end joining (NHEJ), base excision repair (BER), mismatch repair (MMR), nucleotide excision repair (NER), translesion synthesis (TLS) and cell cycle regulation (CCR) [ 16 , 17 ]. Genomic DNA was isolated using a QIAGEN Genomic DNA extraction kit according to the manufacturer’s instructions (Qiagen Inc., Valencia, CA, US). The purity and concentration of the genomic DNA were checked by agarose gel electrophoresis and the OD 260/280 ratio. The genomic DNA was fragmented with Covaris fragmentation protocol (Covaris, Inc., Woburn, MA, US). The size of the fragmented genomic DNA was checked by Agilent Bioanalyzer 2100 (Agilent Technologies, Inc., Santa Clara, CA, US) and NanoDrop spectrophotometer (Thermo Fisher Scientific, Inc., Wilmington, DE, US). The target gene library was generated with NimblGen capture kits (Roche NimblGen, Inc. Hacienda Dr Pleasanton, CA, US). The samples were sequenced by Illumina MiSeq with paired-end reads of 300 nucleotides. The analysis algorithm was conducted according to our previous protocol [ 22 ]. Briefly, the raw sequencing data were aligned with the reference human genome (Feb. 2009, GRCh37/hg19) with Burrows–Wheeler Aligner software (version 0.5.9) [ 23 ]. SAM tools (version 0.1.18) was used for data conversion, sorting, and indexing [ 24 ]. For single nucleotide polymorphisms (SNPs) and small insertion/deletions (indels), Genome Analysis Toolkit (GATK; version 2.7) was used for variant calling with Base/indel-calibrator and HaplotypeCaller. Pindel or Breakdancer software were used for structural variants larger than 100 bp which cannot be identified by GATK, such as large deletions, insertions and duplications [ 25 ]. After variant calling, ANNOVAR was used for annotation of the genetic variants [ 26 , 27 ]. The dbSNP, Exome sequencing Project 6500 (ESP6500) and the 1000 Genomes variant dataset were used to filter common variants of sequencing results. The sequence variants were classified according to the IARC variant classification [ 28 ]. The pathogenic mutations were defined as large-scale deletion, frame-shift mutation, nonsense mutation, genetic variants associated with uncorrected splicing and mutations affecting protein function demonstrated by functional analyses. The pathogenic and likely pathogenic mutations were used as deleterious mutations in our study. An allele frequency greater than 0.01 in the general population in the 1000 Genomes variant dataset or ESP6500 database were considered benign or likely benign genetic variants. Silent and intronic variants that did not affect splicing were also considered benign or likely benign. Other variants, mainly missense mutations without known functional data, were considered as variants of uncertain significance (VUSs). To reduce their number, bioinformatics analyses, including PolyPhen2 and SIFT, were used to evaluate potential pathogenicity [ 29 , 30 , 31 ]. The VUSs were suspected of being deleterious mutations if they met two criteria: (1) a population frequency of less than 0.01 in the 1000 Genomes and ESP6500 databases and (2) a bioinformatics analysis result with a SIFT score less than 0.05 and a polyphen2 score greater than 0.95. All statistical analyses were performed using the Statistical Package for Social Sciences software package (IBM SPSS Statistics for Windows, Version 22.0. IBM Corp. Armonk, NY, US) and R (version 3.1.2, The R Foundation for Statistical Computing, Institute for Statistics and Mathematics, Wirtschaftsuniversität Wien, Welthandelsplatz Vienna, Austria). One-way ANOVA was used to compare continuous variables and a chi-squared test was used for categorical variables. Survival curves were generated using the Kaplan–Meier method, and differences were calculated using the log-rank test. A multivariate Cox’s regression model was used to evaluate the prognostic factors for progression-free survival (PFS) and overall survival (OS). Statistical significance was set as a p value of less than 0.05.

Epithelial ovarian carcinoma (EOC) is a major cause of death in women worldwide, and patients are usually diagnosed at an advanced stage with a 5-year survival of less than 50% [ 1 , 2 , 3 , 4 ]. Clinical prognostic factors include cancer stage, histological subtypes, tumor grade, residual tumor size after debulking surgery and response to chemotherapy. Despite an initial good response to primary treatments of debulking surgery and adjuvant platinum-based chemotherapy, the majority of patients experience a cancer relapse that is resistant to salvage treatments and eventually die of the disease [ 4 , 5 ]. Precision medicine is the current direction for cancer management depending on the specific genetic or molecular features of cancer. There are several subtypes of EOC—high-grade serous, clear cell, endometrioid, mucinous and low-grade serous—that could be viewed as distinct diseases for their differences in clinical course and pathological features [ 6 , 7 ]. To date, the most promising target therapies for EOC are anti-angiogenesis agents and poly ADP-ribose polymerase inhibitors (PARPi). Bevacizumab in combination with chemotherapy has demonstrated improved progression-free survival, and an overall survival benefit in high-risk patients [ 8 , 9 , 10 ]. Maintenance therapy with PARPi has revised the management of EOC in newly diagnosed and recurrent diseases. The identification of BRCA mutations or homologous recombination deficiency (HRD) status is critical for selecting potential patients, but both positive and negative patients as defined by current HRD assays benefited from PARPi [ 11 , 12 , 13 , 14 , 15 ]. DNA damage response (DDR) is important for maintaining a cell’s genomic integrity, and the DDR pathway is composed of various molecules that detect DNA damage, activate cell-cycle checkpoints, trigger apoptosis, and coordinate DNA repair [ 16 , 17 , 18 ]. Several exogenous or endogenous sources (e.g., oxidative damage, radiation, ultraviolet light, cytotoxic materials, replication errors) may result in DNA damage that may eventually lead to genomic instability and cell death [ 19 ]. DDR consists of several pathways, including base excision (BER), mismatch (MMR) and nucleotide excision repair (NER); translesion synthesis (TLS) for single-strand break repair; homologous recombination (HR) and nonhomologous DNA end joining (NHEJ) for double-strand break repair; and cell cycle regulation (CCR) (27, 28). Homologous recombination is an error-proof repair pathway to restore the original sequence at the double-strand DNA break. BRCA 1/2 genes participating in HR and maintaining PARPi therapy for BRCA -mutated EOC is a good example of synthetic lethality [ 20 ]. Several other DDR genes have been identified as potential targets for novel cancer therapy under clinical investigation [ 16 , 17 ]. Understanding the complex DDR pathways is helpful for exploring the feasibility of novel DDR inhibitors in clinical practice. In the study, we investigated the somatic mutations of DDR genes in 172 EOC patients using a targeted DDR gene panel using a next-generation sequencing method. The correlation of the somatic DDR gene mutations, clinical parameters and outcomes was analysed.

There were 172 EOC patients enrolled: 69 serous, 39 endometrioid and 64 clear cell carcinomas ( Table 2 ). There were 68 high-grade serous carcinomas (type II tumor) and 104 type I tumors. The median age was 52, and the median pre-treatment CA125 value was 400 U/mL; 59.9% were diagnosed at an advanced cancer stage, and 65.1% had undergone optimal debulking surgery; 59.3% had disease recurrence, and 44.2% died of EOC. All patients received adjuvant platinum and paclitaxel chemotherapy. As shown in Table 3 , 114 deleterious somatic mutations were identified from 26 genes of our 60-gene DDR panel in 78 EOC patients: 27 nonsense mutations in 23 patients, 28 frameshift mutations in 20, 28 missense mutations in 26 patients and 31 mutations involving uncorrected splicing in 29 patients. There were single-gene mutations in 57 patients, and multiple-gene mutations in 21: 2 mutations in 14 patients, 3 mutations in 2, 4 mutations in 3, 5 mutations in 1 and 6 mutations in 1 patient ( Figure 1 ). We also identified 109 missense mutations classified as variants of uncertain significance (VUSs) with the potential of being deleterious mutations after searching the database ( http://www.ncbi.nlm.nih.gov/snp , accessed on 28 September 2021) and bioinformatic analyses ( Table S1 and Figure S1 ). The pattern of prevalent mutated DDR genes was different among the histological subtypes ( Figure 1 ). The proportion of wild type DDR genes was 54.7% in all EOC patients; 30.4% in serous carcinoma, 66.7% in endometrioid carcinoma and 73.4% in clear cell carcinoma. The top three prevalent mutated DDR genes were TP53 (27.9%), MUTYH (6.4%) and BRCA2 (5.8%) for all patients. Serous carcinoma— TP53 (56.5%), BRCA2 (5.8%) and RAD51C (5.8%); endometrioid carcinoma— TP53 (15.4%), ATM (12.8%) and MSH2 (7.7%); clear cell carcinoma— MUTYH (9.4%), TP53 (4.7%), BRCA2 (3.1%) and ERCC8 (3.1%). The top three prevalent mutated subgroups of DDR genes were CCR (30.8%), HR (10.5%) and BER (7.0%) for all patients. Serous carcinoma—CCR (58.05%), HR (15.9%) and BER (5.8%); endometrioid carcinoma—CCR (23.1%), MMR (15.4%) and HR (7.7%); clear cell carcinoma—BER (9.4%), CCR (6.3%) and HR (6.3%). For detailed information, please refer to Table S2 and Figure S2 . We evaluated the correlations between the mutation of DDR genes, the clinicopathologic parameters and outcome of the EOC patients. As shown in Table 4 , type II tumors had a higher percentage of HR gene mutations than type I tumors (16.18 vs. 6.73%, p = 0.048, chi-squared test). Endometrioid carcinoma (15.38%) had a higher percentage of MMR mutations than those of serous carcinoma (2.90%) and clear cell carcinoma (4.69%) ( p = 0.03, chi-squared test). Low-grade tumors had a higher percentage of MMR mutations compared with high-grade tumors (17.24 vs. 4.20%, p = 0.009, chi-squared test). Type II tumors had a higher percentage of DSBR mutations than type I tumors (17.65 vs. 6.73%, p = 0.026, chi-squared test). Serous carcinoma (57.97%) had a higher percentage of CCR mutations than those of endometrioid carcinoma (23.08%) and clear cell carcinoma (6.25%) ( p < 0.001, chi-squared test). Type II tumors had higher percentage of CCR mutations than those of type I tumors (58.82 vs. 12.50%, p < 0.001, chi-squared test). The advanced-stage patients had a higher percentage of CCR mutations than the early-stage patients (42.72 vs. 13.04%, p < 0.001, chi-squared test). The recurrent patients had a higher percentage of CCR mutations than those without recurrence (39.22% vs. 18.57%, p = 0.004, chi-squared test). Patients who died of EOC had higher percentages of CCR mutations than living patients (40.79 vs. 22.92%, p = 0.012, chi-squared test). Serous carcinoma (69.57%) had higher percentage of DDR mutations than those of endometrioid carcinoma (33.33%) and clear cell carcinoma (26.56%) ( p < 0.001, chi-squared test). Type II tumors had a higher percentage of DDR mutations than type I tumors (70.59 vs. 28.85%, p < 0.001, chi-squared test). The advanced stage patients had higher percentage of DDR mutations than the early-stage patients (57.28 vs. 27.54%, p < 0.001, chi-squared test). Recurring patients had a higher percentage of DDR mutations than those without recurrence (53.92 vs. 32.86%, p = 0.006, chi-squared test). Patients who died of EOC had a higher percentage of DDR mutations than living patients (59.21 vs. 34.38%, p = 0.001, chi-squared test). EOC patients without DDR gene mutation had longer progression-free survival (PFS) ( p = 0.0072, log-rank test, Figure 2 A) and overall survival (OS) ( p = 0.022, log-rank test, Figure 2 B) than those with 1 DDR or ≥2 DDR mutations. In serous carcinoma, patients with or without DDR mutations had similar PFS ( p = 0.56, log-rank test, Figure 2 C). Patients with ≥2 DDR mutations had a trend of better OS than those with 1 mutation or none, but it was not statistically significant ( p = 0.47, log-rank test, Figure 2 D). In endometrioid carcinoma, patients with ≥2 DDR gene mutations had shorter PFS ( p = 0.0035, log-rank test, Figure 2 E) and OS ( p = 0.015, log-rank test, Figure 2 F) than those with 1 mutation or none. In clear cell carcinoma, patients with ≥2 DDR gene mutations had significantly shorter PFS ( p = 0.0056, log-rank test, Figure 2 G) and OS ( p = 0.0046, log-rank test, Figure 2 H) than those with 1 DDR mutation or none. Tumor recurrence with CCR gene mutation (HR: 1.68 (1.12–2.50), p = 0.011), 1 DDR gene mutation (HR: 1.71 (1.12–2.60), p = 0.013), endometrioid carcinoma (HR: 0.17 (0.08–0.37), p < 0.001), type II tumor (HR: 2.69 (1.81–4.00), p < 0.001), advanced-stage carcinoma (HR: 5.29 (3.16–8.85), p < 0.001), high-grade tumor (HR: 5.57 (2.26–13.70), p < 0.001) and optimal debulking surgery (HR: 0.28 (0.18–0.41), p < 0.001) were significant in the univariate Cox regression model ( Table 5 ). Advanced-stage carcinoma (HR: 3.08 (1.63–5.80), p = 0.001) and optimal debulking surgery (HR: 0.51 (0.32–0.80), p = 0.004) were important prognostic factors in the multivariate analysis. Cancer-related death with TLS gene mutation (HR: 33.76 (3.95–289.00), p = 0.001), 1 DDR gene mutation (HR: 1.96 (1.20–3.20), p = 0.007), endometrioid carcinoma (HR: 0.12 (0.04–0.38), p < 0.001), type II tumor (HR: 1.88 (1.19–2.96), p = 0.007), advanced-stage carcinoma (HR: 6.84 (3.28–14.25), p < 0.001), high-grade tumor (HR: 17.97 (2.50–129.29), p = 0.004) and optimal debulking surgery (HR: 0.26 (0.16–0.41), p < 0.001) were significant in the univariate Cox regression model. Type II tumor (HR: 0.35 (0.20–0.60), p < 0.001), TLS gene mutation (HR: 9.57 (1.08–84.83), p = 0.042), advanced-stage carcinoma (HR: 4.82 (2.09–11.09), p < 0.001) and optimal debulking surgery (HR: 0.38 (0.22–0.64), p < 0.001) were important prognostic factors in the multivariate analysis.

Our study showed that nearly half of the epithelial ovarian cancer (EOC) patients had DNA damage response (DDR) gene mutations with varied proportions of histological subtypes. Two-thirds of serous adenocarcinoma patients, one-third of endometrioid adenocarcinoma patients and one-fourth of clear cell carcinoma patients had DDR gene mutations. Our DDR gene panel consisted of the genes involved in single-strand break repair, double-strand break repair and cell cycle regulation, including the genes recommended by National Comprehensive Cancer Network (NCCN) guidelines as cost-effective tools for assessing the lifetime risk of EOC, such as ATM, BRCA1/2, BRIP1, MLH1, MSH2, MSH6, PALB2, RAD51C and RAD51D [ 32 ]. The major components of DDR gene mutations were CCR in serous, CCR and SSBR in endometrioid and SSBR in clear cell carcinomas; CCR and DSBR in type II tumors (high-grade serous carcinoma in the cohort); and SSBR in type I tumors. A multiple DDR gene panel increased the detection rate of somatic mutation of genes involved in DNA damage repair pathway in comparison with a BRCA test alone. The percentage of BRCA 1/2 somatic mutation in serous carcinoma was 7.2, which was compatible with the 6–7% in previous studies [ 33 , 34 , 35 , 36 , 37 ]. The non- BRCA HR somatic mutation of our study was more than 10% in serous and endometrioid carcinomas, and the MMR somatic mutation was around 15% in endometrioid carcinomas, which was compatible with the previous study [ 38 ]. Our study showed that ovarian clear cell carcinoma patients with DDR gene mutations had an unfavorable survival prognosis. Those who had somatic DDR mutations were significantly associated with advanced-stage carcinomas, tumor recurrence and tumor-related death. The trend was different in histological subtypes as serous carcinomas or type II tumors with DDR mutation showed a better survival trend. Non-serous or type I EOC patients with DDR mutations had a poor prognosis, especially in clear cell carcinoma. Ovarian clear cell carcinoma is an aggressive drug-resistant subtype of EOC in association with endometriosis and glycogen accumulation. It accounts for about 5–13% of all EOCs in Western populations, but up to 20–25% in East Asia, including Taiwan [ 2 , 3 ]. Previous studies showed that the somatic mutations of ovarian clear cell carcinoma (mainly in ARID1A , PIK3CA , KRAS and PPP2R1A ) might be related to chromatin remodeling, cell proliferation, cell cycle checkpointing and cytoskeletal organization [ 39 , 40 , 41 , 42 , 43 , 44 , 45 , 46 , 47 , 48 , 49 ]. However, the frequent mutations of ARID1A, PIK3CA, PPP2R1A or TP53 in ovarian clear cell carcinoma did not correlate well with the prognosis [ 45 ]. Other infrequent gene mutations of clear cell carcinoma included ARID1B, ARID3A, CREBBP, CSMD3, CTNNB1, LPHN3, LRP1B, MAGEE1, MLH1, MLL3, MUC4, PIK3R1, PTEN and TP53 [ 41 , 43 , 46 , 48 , 49 ]. DDR gene mutations in ovarian clear cell carcinoma was unclear in the literature, and our finding of an unfavorable prognosis in clear cell carcinoma patients with DDR gene mutations could provide useful information. Our DDR gene panel could provide a scientific rationale for patient selection in future clinical trials that target DNA damage repair response pathways, especially in clear cell carcinoma. BRCA gene tests or companion HRD assays are currently suggested for PARPi, but there are unmet problems that need to be resolved [ 11 , 12 , 13 , 14 , 15 , 20 ]. The most important one is that the HRD assays cannot consistently identify patients who do not benefit from PARPi therapy. The consensus for the cut-off value was indeterminate because the thresholds of HRD assays were developed from retrospective exploratory analyses [ 11 , 50 , 51 ]. Generally, advance-stage, high-grade serous carcinoma patients with tumor BRCA (t BRCA ) mutations, including germline (g BRCA ) or somatic (s BRCA ), derived the greatest benefit from PARPi maintenance therapy [ 11 , 12 , 13 , 14 , 15 ]. Approximately 11–18% of patients had a g BRCA mutation, and another 6–7% patients had an s BRCA mutation with a negative g BRCA test [ 33 , 34 , 35 , 36 , 37 ]. However, about 5% of g BRCA mutated patients tested negative for t BRCA [ 52 , 53 , 54 ]. The non- BRCA HR gene mutations were usually pooled together to interpret the association with clinical outcomes in previous studies because of their relatively low prevalence [ 35 , 55 , 56 , 57 ]. Twenty-one platinum-sensitive recurrent patients with non- BRCA somatic mutations ( BRIP1 , CDK12 , RAD54L and RAD51B ) derived benefit from olaparib in study 19 [ 58 ]. In ARIEL2, there were 20 patients with non- BRCA HR gene mutations ( ATM , BRIP1 , CHEK2 , FANCA , FANCI , FANCM , NBN , RAD51B , RAD51C and RAD54L ), but the sensitivity in discriminating a rucaparib response was only 11% [ 59 ]. However, BRCA wild type EOC patients still benefitted from PARPi, which indicated that a BRCA test by itself was inadequate for selecting EOC patients for PARPi [ 13 , 14 , 15 ]. It needs to be determined which individual or panel of non- BRCA HR genes could be used to predict a PARPi response, especially in non-serous EOC patients. There were limitations to our study. First, germline gene mutations were not investigated. These not only inform the patients but also identify family members of the possible risk of malignancy [ 52 , 53 , 54 ]. The NCCN suggested germline gene tests of ATM, BRCA1/2, BRIP1, MLH1, MSH2, MSH6, PALB2, RAD51C, RAD51D and STK11 to assess the lifetime risk of EOC [ 32 ], but how many genes should be included in the panel is inconclusive. Second, the numerous variants of uncertain significance (VUSs) identified by multiple gene panels would cause controversy in risk assessment and management [ 60 , 61 , 62 ]. The biological functions and clinical impacts of most individual mutations in the genomic loci have not been well characterized, especially for VUSs [ 63 ]. Even in the well-studied BRCA gene, there is a difference among laboratories in the VUS reporting rate (3–50%), detection protocols and management strategies [ 64 ]. Further sharing and integration of gene sequencing data in an open database might decrease VUSs. Third, the cohort sample was not large enough; only the trends of clinical prognosis that correlated with each DDR pathway were found. Further large-scale investigations are needed.

Our study found that nearly half of the EOC patients had DDR gene mutations of varying proportions in the histological subtypes. Patients with somatic DDR mutations were significantly associated with advanced-stage carcinoma, tumor recurrence and tumor-related death. Type I EOC patients with DDR mutations had an unfavorable prognosis, especially for clear cell carcinoma. A broad multiple-gene DDR panel would provide not only comprehensive information of gene mutations but also a rationale for a future study of a novel therapy target for DNA damage response pathways.