From the previously published literature, we searched for ovarian cancer studies that performed germline testing of 171 genes ( Supplementary Table S1 ) at the BGI Shenzhen Clinical Diagnostic Laboratory. All samples were obtained with informed consent, and a total of 3 studies met the criteria (21–23). Mutation data was obtained from the authors based on a reasonable request. And mutation detection methods were described in detail in the Supplementary Material 1 . The mutation results and clinical information from the three studies were pooled, and samples with missing age, family history of cancer, or regional information were excluded. The data of 1171 ovarian cancer samples that can represent ovarian cancer in the Chinese population were retained. The average depth was over 100X and the coverage at 30X exceeded 95% for each sample. The sequencing coverage and quality statistics for each sample were summarized in Supplementary Table S2 .

Single nucleotide variations (SNVs), and insertions and deletions (INDELs) were selected that were localized in all exons and intron-exon boundaries (± 20 bp) of the genes. Mutations with fewer than 5 supporting reads, a mutation frequency less than 20%, or a population frequency greater than 5% in the population polymorphism databases GnomAD, ExAC, and 1k Genomes were excluded (24–26). The final mutation results for each sample were obtained. According to the genetic variation classification standards and guidelines jointly developed by the American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology (AMP) (27, 28), mutations were classified into five categories, pathogenic, likely pathogenic, variant of unknown significance (VUS), likely benign and benign. Patients with likely pathogenic or pathogenic mutations were defined as those carrying deleterious mutations.

All the statistical analyses and plots were performed using R (version 3.6.3). Pearson’s χ2-test and Fisher’s exact test were used to determine the statistical significance of categorical variables. Student’s t test was used to compare continuous variables, such as age at diagnosis, between two groups. All P values reported were two-sided, and a P value of less than 0.05 was considered statistically significant. False discovery rates were calculated using the Benjamini-Hochberg procedure; FDR < 0.05 was used as the threshold for significance after correction for multiple hypothesis testing.

The study included 1171 patients with qualified testing data ( Table 1 ). These samples include all seven regions in China and can represent the characteristics of the Chinese ovarian cancer population. The age range of the enrolled patients was 12~86 years old ( Supplementary Figure S1 ), the median age was 54 years old, and 247 patients had a family history of cancer. According to the statistics of clinical data, there was no significant correlation between family history of cancer and age of cancer onset. BRCA-positive patients were defined as those with likely pathogenic and/or pathogenic mutations in BRCA1/2, and 103 (8.8%) out of 1171 ovarian cancer patients were BRCA-positive.

Among the 1171 patients, a total of 19435 mutations were detected ( Supplementary Table S3 ), with an average of more than 16 mutations per patient. These mutations contained 5657 types, of which 94.38% (5345) of which were SNPs, and the remaining 312 were INDELs. Categorized by mutation effect, the most common types were missense mutations (3105), followed by synonymous mutations (1704), 491 splice-site mutations, 177 frameshift mutations, 72 INDELs within coding frames, and 108 nonsense mutations. In the cohort, there were a total of 48 mutations present in 39 genes, each carried by more than 5% of the patients, and the five highest frequency mutations were in the following order: MC1R, c.359T>C; ERCC5, c.1586G>C; PRSS1, c.72C>T; BARD1, c.1075_1095del; and KIT, c.1638A>G. Meanwhile, we applied fit Chi-square calculation to evaluate these mutations with ovarian cancer risk and found MC1R:c.359T>C, ERCC5:c.1586G>C and PRSS1:c.72C>T that significantly fails to conform Hardy-Weinberg equilibrium (P-value<0.05). Notably, nearly 48% (23/48) of the high-frequency mutations were synonymous. We used the PATHVIEW software package (29) for the 39 genes and found that they were significantly enriched in DNA damage response repair pathways such as fanconi anemia pathway, homologous recombination, mismatch repair, nucleotide excision repair, and base excision repair ( Supplementary Figure S2A ).

Mutations were detected in each of the 171 genes within the assay ( Supplementary Table S4 ). The most common gene, PRKDC, was detected in 612 patients (approximately 52%), and the least common gene, MAX, was detected in only 1 patient. Genes with a high prevalence of mutations in more than 10% of the cohort were selected, and a total of 30 genes were obtained ( Figure 1 ). The top 10 genes with high incidence were PRKDC, BRCA2, BRCA1, FANCI, ERCC5, SLX4, PTCH1, MC1R, BARD1, and RET. The 30 genes were significantly enriched in the fanconi anemia pathway (P value 2.55e-13, FDR 5.15e-11) and mismatch repair (P value 2.40e-10, FDR 2.42e-8) ( Supplementary Figure S2B ). An exploration of the distribution of mutation sites within genes showed single or multiple hotspot mutations in high-incidence mutated genes ( Figure 2 ), such as BRCA1: c.2566T>C (p.Y856H), BRCA2: c.10234A>G (p.I3412V), ERCC5: c.1586G>C (p.C529S), FANCI: c.2011A>G (p.I671V), PRKDC: c.10681T>A (p.L3561M), and BARD1: c.1075_1095del (p.L359_P365del), suggesting that these sites were associated with ovarian cancer occurrence in China.

Compared with other high-frequency mutated genes ( Figure 1 ), more INDEL mutations were detected in BRCA1, BRCA2, PRKDC, BARD1, MSH6, RECQL4, and MSH3 genes, and more splice-site mutations were detected in the BRCA1, MSH6, MUTYH, and TSC2 genes, which tended to cause greater functional changes. BRCA1, BRCA2, and BARD1 are core genes of the homologous recombination repair pathway (30), and MSH6 and MSH3 are core genes of the mismatch repair pathway (31). The National Comprehensive Cancer Network (NCCN) guidelines (32) recommended screening for mutations in RAD51D, EPCAM, and RAD51C, which were detected in 5%, 4%, and 2% of the population, respectively. However, only the high-frequency mutated genes, BRCA1 and BRCA2, were within the scope of NCCN-recommended screening. It implied that the mutated genes of ovarian cancer in the Chinese population differed significantly from those recommended by NCCN for ovarian cancer screening in both European and American populations.

According to the classification of mutation function, a total of 954 (16.9%) were annotated as likely pathogenic or pathogenic mutations in 1041 patients ( Supplementary Table S3 ). The top 10 genes with most deleterious mutations in the Chinese ovarian cancer population were MC1R (12%), MLH1 (9%), PRKDC (8%), KIF1B (8%), FANCM (7%), FANCI (6%), PRSS1 (6%), SDHA (6%), BRCA2 (6%), and CFTR (5%) ( Supplementary Figure S3 ). In addition, deleterious BRCA1 mutations were detected with a frequency of about 3%, and BRCA1 and BRCA2 were detected in 103 patients (8.8%) patients. Pathway enrichment of these high-frequency genes showed that they were significantly enriched in the fanconi anemia pathway (P value 1.03e-06, FDR 2.00e-04). Analysis of the distribution of mutation sites within genes revealed the presence of single or multiple hotspot mutations in high-frequency mutation genes ( Supplementary Figure S4 ), such as MC1R: c.359T>C (p.I120T), MLH1: c.1151T>A (p.V384D), PRKDC: c.10681T>A (p.L3561M), ERCC5: c.640C>T (p.R214C), ERCC5: c.767A>G (p.Q256R), PTCH1: c.2222C>T (p.A741V), and RECQL4: c.212A>G (p.E71G), suggesting that these sites were related to the occurrence and progression of ovarian cancer in China. The mutations observed were primarily missense mutations and were predominantly heterozygous, indicating that the dosage of the mutation had a significant impact on normal functional execution.

Considering only the NCCN-recommended ovarian cancer screening genes, it was noticed that no mutations were detected in the recommended EPCAM and STK11 genes. Compared with the two American cohorts (cohorts A and B) (19, 33), and the four Chinese cohorts (cohorts C-F) (13, 18, 34, 35) ( Supplementary Table S5 ), it was observed that the detection rates of the two cohorts in the United States were 18.10% and 18.81%, respectively. For the four cohorts in the Chinese population, the detection rate of Cohort C was 27.20%, and the other three cohorts were tested for only the BRCA1 and BRCA2 genes, with detection rates of 22.40%, 17.00%, and 28.40%, respectively. The detection rate of this study cohort was 28.35%, which was close to that of previous Chinese cohorts. Further analysis of the deleterious mutations in NCCN-recommended screening genes in foreign and domestic cohorts revealed that the BRCA1 and BRCA2 genes were generally detected at high frequencies in the American and Chinese populations ( Figure 3 ), and the difference was that the MLH1 gene carried the highest frequency of deleterious mutations in this cohort. Meanwhile, it was discovered that all high-frequency mutation genes in the American population were within the screening scope recommended by the NCCN guidelines, while the high-frequency mutation genes in the Chinese cohort of ovarian cancer, such as MC1R, PRKDC, KIF1B, FANCM, FANCI, PRSS1, and SDHA, were not included in the scope of the NCCN guidelines, indicating that the NCCN guidelines based on the European and American populations were not suitable for ovarian cancer genetic screening in the Chinese population.

Among the 130 patients without detecting deleterious mutations, the high-frequency mutation genes were similar to those of all cohort populations, and pathway enrichment analysis showed that these genes were also significantly enriched in the fanconi anemia pathway (P value 1.60e-06, FDR 3.24e-04). Compared with the 1041 patients carrying deleterious mutations, there was no significant difference in age distribution (t-test P value 0.84) or family history of cancer (chi-square test P value 0.70). At the same time, it was identified that the genes with significantly high frequency in the group carrying deleterious mutations were mainly enriched in the homologous recombination (P value 1.02e-05, FDR 2.06e-03) and fanconi anemia pathway (P value 7.12e-05, FDR 7.20e-03).

A total of 190 BRCA1 mutations and 169 BRCA2 mutations were detected in the enrolled ovarian cancer population. Among the high-frequency BRCA1 and BRCA2 mutations, except for BRCA1:c.5470_5477delATTGGGCA (5.85% of all BRCA1 mutations, 20/342), which was a pathogenic mutation, the others were all missense mutations annotated as benign. Further analysis revealed that BRCA1: c.2566T>C (p.Y856H) accounted for 12.87% of all BRCA1 mutations ( Supplementary Figures S5A, B ), and BRCA2: c.8187G>T (p.K2729N) constituted 9.36% of all BRCA2 mutations ( Supplementary Figures S5C, D ), were significantly higher in the East Asian population than in other populations, and were geographically characteristic BRCA1/2 mutation in the Chinese ovarian cancer population. However, BRCA2:c.10234A>G (p. I3412V) comprised 23.15% of all BRCA2 mutations ( Supplementary Figures S5E, F ), with a frequency of 2.40% in East Asian populations and 8.03% in this ovarian cancer cohort. Notably, its prevalence exceeds 10% in normal populations in the Americas and Africa, displaying that it was not a geographically characteristic mutation in the Chinese population.

Co-mutation in BRCA1 and BRCA2 occurred in 99 patients, representing 8.5% (99/1171) of the entire cohort. Analysis of the characteristics of the BRCA1 and BRCA2 co-mutation group and other patients, it was observed that there was no statistical difference in family history distribution (chi-square test, P value 0.68). However, in terms of age ( Supplementary Figure S6 ), the group carrying BRCA1 and BRCA2 co-mutations had a significantly earlier age of cancer onset than the rest patients (mean age, 51 VS 54.5 years, Wilcoxon test P value 7.30e-03). Similar results have been reported in multiple studies (34, 36), implying that BRCA1 and BRCA2 were associated with the early onset of cancer.

Comparative analysis of the difference in ovarian cancer mutation genes between the BRCA1/2 mutation group and the non-mutation group uncovered that FAM175A, EMSY, PTCH1, and HNF1B were significantly highly mutated in the group without BRCA1/2 mutations ( Supplementary Figure S7A ), particularly PTCH1: c.3907C>T (p.R1303C) ( Supplementary Figure S7B ). When considering only the difference in deleterious mutations between the two groups, it was found that FAM175A was equally significant in the group without BRCA1/2 mutations ( Supplementary Figure S7C ).

Ovarian cancer exhibited a notable pattern of family inheritance and aggregation, primarily attributed to the transmission of tumor-associated germline mutations to the offspring along with the reproductive process. The enrollment cohort consisted of 247 samples with a family history of cancer and 924 samples without. By comparing the differences in mutation genes between the two groups ( Figure 4A ), it was observed that the XPA and NF2 genes were significantly more frequently mutated in the group with a family history of cancer (P <0.05), while the ERCC5 and TSC1 genes were more common in the group without a family history of cancer (P <0.05). Analysis of the mutation sites in these genes revealed that the TSC1 gene had a significantly higher frequency of mutation c.250G>A (p.A84T) in the population without a family history ( Figure 4B ).

The analysis of hereditary mutations in each sample revealed a decrease in the total number of mutations as onset age increased within the patient population ( Supplementary Figure S8A ), suggesting that the occurrence of ovarian cancer in the younger age group was mainly related to genetic factors. According to the age distribution of the cohort, the patients in the first quartile of the age range (47 years and younger) were categorized into the younger age group and were compared with the group older than 47 years. Fisher’s exact test was performed on the mutations of each gene in the two age groups ( Supplementary Figure S8B ). The results showed that ERCC4, CHEK2, and PDGFRA were significantly more mutated in the younger group (P value<0.05), while POLH was significantly more common in the older group.

Regional characteristics of cancer-related mutation genes were highly prevalent due to differences in the ancestral genetic backgrounds of various regions. In the cohort, a Fisher’s exact test was conducted on each mutated gene between a single region and other regions ( Figure 5A ). The results showed that MEN1 was significantly more highly mutated in the East China group, MXI1 gene in the Southwest China group, TMEM127 and FH in the South China population, RPA1 in the North China population, and CDH1, CHEK2, FAM175A, and EXT2 in the Northeast population. ERCC4 and HMMR were significantly more common in the non-North China population, while POLH was more common in the non-South China population. The pathway enrichment revealed that high-frequency mutated genes in North China were significantly enriched in the fanconi anemia pathway (p-value 6.20e-07, FDR 1.25e-04), homologous recombination (p-value 7.15e-07, FDR 7.22e-04), nucleotide excision repair (P value 3.49e-05, FDR 2.35e-03), mismatch repair (P value 4.82e-04, FDR 0.02), while in South China, in the fanconi anemia pathway (P value 2.12e-05, FDR 4.28e-03). When only deleterious mutations were considered ( Supplementary Figure S9 ), CDH1 and FAM175A were significantly more frequently mutated in the Northeastern population, while ERCC4 was in non-North China populations, and the PRKDC in non-South China populations. The results indicated that differences in the geographic scope of China also led to differences in cancer susceptibility genes for ovarian cancer.

By analyzing the mutation carrier rates of the differential genes, we found that MEN1: c.209A>C (p.D70A) was highly prevalent in the East China population ( Figure 5B ), while MEN1: c.1523G>A (p.G508D) and c.527G>A (p.R176Q) in the populations of other regions. Furthermore, a significant high-frequency mutation, POLH: c.1433C>T (p.T478M) ( Figure 5C ), was observed in non-South China populations and ERCC4: c.241G>T (p.V81F) in non-North China populations ( Figure 5D ). The pathogenic PRKDC: c.10681T>A (p.L3561M) mutation was significantly more common in the non-South China population ( Figure 5E ). The presence of highly mutated genes in different regions suggested that precise treatment of ovarian cancer in the Chinese population should be based on characteristic mutation data specific to this population.