Comprehensive analysis of serum tumor markers and BRCA1/2 germline mutations in Chinese ovarian cancer patients

Abstract Background The serum tumor markers has been widely used in ovarian cancer diagnosis. BRCA1/2 germline mutations are the most common predisposing factors for ovarian cancer development. This study aimed to comprehensively investigate serum tumor markers and BRCA1/2 germline mutations and analyze their associations with ovarian cancer. Methods Levels of 11 serum tumor markers were examined in ovarian cancer patients and controls with benign gynecologic diseases. By integrating multiplex PCR and next‐generation sequencing technologies, BRCA1/2 germline mutations were analyzed and confirmed by Sanger sequencing. The discriminative models with serum tumor markers and BRCA1/2 mutation status were constructed for ovarian cancer detection and patient stratification. Results Among 11 markers, six of them were significantly elevated and only beta‐human chorionic gonadotropin (β‐HCG) was significantly reduced in ovarian cancer patients. A total of 54 (23.3%) ovarian cancer patients were found to harbor BRCA1/2 deleterious mutations, and BRCA1/2 mutations were significantly associated with Hereditary Breast and Ovarian Cancer‐related tumors and family history of cancer. Carbohydrate antigen 125 showed a good performance in ovarian cancer detection as a single marker (AUC = 0.799), while a panel of eight markers showed a good performance in BRCA1 mutation detection with an AUC value of 0.974. In addition, a panel of five serum tumor markers combined with BRCA1/2 mutation status showed a good performance in lymph node metastasis prediction (AUC = 0.843). Conclusions We found the association between BRCA1/2 germline mutation status and serum tumor marker levels, and identified discriminative models that combined serum tumor markers with BRCA1/2 mutation status for ovarian cancer detection and patient stratification.


| INTRODUCTION
Ovarian cancer is the third most commonly diagnosed cancer and the first leading cause of the gynecologic malignancies related deaths, with 238,700 new cases and 151,900 deaths worldwide in 2012 (Lu & Chen, 2014;Torre et al., 2015). The incidence of ovarian cancer in China has increased annually during the past 10 years, with up to 52,100 new cases and 22,500 deaths in 2015 (Chen et al., 2016). Ovarian cancer is a heterogeneous disease that can originate from surface epithelial cells, specialized stromal cells or germ cells, which can be subdivided into several histological subtypes including serous, mucinous, endometrioid, and clear cell tumors (Weiderpass & Tyczynski, 2015). In addition, ovarian cancer is one of the most lethal gynecologic malignancies to diagnose and difficult to detect at early stage (Buzolin et al., 2017).
Serum tumor markers play a crucial role in ovarian cancer diagnosis. Carbohydrate antigen 125 (CA125), a transmembrane glycoprotein produced by coelomic epithelium, is the first recommended serum marker for early detection and monitoring relapse of ovarian cancer (Fan et al., 2016). The expression of CA125 can be used to evaluate the treatment of ovarian cancer, which has been considered to be the most helpful clinical serological marker of ovarian cancer, especially in serous carcinoma (Esselen et al., 2016). Human epididymis protein 4 (HE4) is a protein that consists of a single peptide and two whey acidic protein domains containing a "four-disulfide core" with eight cysteine residues (Clauss, Lilja, & Lundwall, 2002). CA125 and HE4 have been approved by the FDA for monitoring treatment and detecting of ovarian cancer (Montagnana et al., 2011). Besides CA125 and HE4, many other serum markers have been studied in ovarian cancer. For example, carbohydrate antigen 15-3 (CA15-3) was over-expressed in a wide variety of ovarian cancer (Williams et al., 2014) and elevated in approximately 70% of ovarian cancer patients, predominantly in those with advanced disease (Jeschke et al., 2012). The level of serum β-HCG was associated with ovarian cancer, with high levels in advanced FIGO stages (III and IV), regardless of histological type of tumor (Djurdjevic, Maksimovic, Pantelic, Golubovic, & Curcic, 2011). These markers in ovarian cancer detection have been individually investigated in many laboratories and clinical tests. However, the sensitivity and specificity of these markers still need further evaluation in Chinese patients, and it is meaningful to find tumor marker panels with high sensitivity and specificity in ovarian cancer diagnosis.
BRCA1 and BRCA2 are two high-susceptibility genes for familial ovarian cancer (Miki et al., 1994;Wooster et al., 1995). Individuals who inherit BRCA1/2 germline mutations showed a high lifetime risk and early onset of ovarian cancer (Choi et al., 2015). Genetic testing for BRCA1/2 mutations has been proved to be a key step in the risk assessment, prognosis, treatment, and prevention of ovarian cancer (Pan & Xie, 2017). Recently, screening BRCA1/2 mutations has been applied as a companion diagnostic test guiding clinical medication for ovarian cancer patients (Daly et al., 2017). Poly ADP-ribose polymerase (PARP) inhibitors are novel targeted drugs, which have recently been approved to treat advanced ovarian cancer patients carrying BRCA1/2 mutations (Shi et al., 2017). Ovarian cancer patients with BRCA1/2 mutations will most likely benefit from PARP inhibitors compared to non-BRCA mutated patients (Dizon, 2017). BRCA1/2 mutation detection could become a routine clinical practice for evaluation of women with ovarian cancer for personalized medicine (Spriggs & Longo, 2016).
To investigate serum tumor markers and BRCA1/2 germline mutations in Chinese ovarian cancer patients, we examined the concentrations of serum tumor markers in 232 ovarian cancer patients and 219 controls with benign gynecologic diseases and screened the entire coding exons and exon-intron boundaries of BRCA1/2 in ovarian cancer patients. The discriminative models with serum tumor markers and BRCA1/2 mutation status were also established for ovarian cancer detection and patient stratification. to 2017. The 219 female controls were diagnosed with benign gynecologic diseases including uterine fibroids, ovarian cyst, ovarian benign mass, ovary teratoma, endometriosis. The clinicopathological characteristics of all ovarian cancer patients were summarized in Table 1, and that of benign controls were summarized Table S2. All ovarian cancer patients were diagnosed pathologically by experienced gynecologic pathologists.

| Sequencing experiments and mutation analysis
Genomic DNA was extracted from each blood sample using TIANamp Genomic DNA Kit (TianGen Biotech, Beijing, China) and quantified using Nanodrop 2000 (Thermo Fisher Scientific, Wilmington, DE). Target enrichment and library preparation were performed by PCR using Human BRCA1/2 Sequencing Panel Kit (Sansure Biotech, Hunan, China) according to the manufacture's instruction, which can amplify the entire coding exons and exon-intron boundaries of BRCA1/2 genes simultaneously. The library PCR products were purified with AMPure XP system (Beckman Coulter, Brea, CA), and quantified using Qubit ® dsDNA HS Assay Kit (Life Technologies, Foster City, CA) then pooled into one complete library with equimolar ratio. The prepared libraries were sequenced on MiSeq system (Illumina, San Diego, CA) using MiSeq Reagent Kit v2 (500 cycles). Analysis of sequencing data, mutation annotation, and mutation confirmation by Sanger sequencing were performed using our previously described methods (Li et al., 2017). The sequence numbering was based on transcript and protein sequence of BRCA1 (NM_007294.3 and NP_009225.1) and BRCA2 (NM_000059.3 and NP_000050.2), respectively.

| Statistical analysis
Continuous data were summarized using mean and standard deviation and the difference was determined by the t test. The χ 2 test was used to compare categorical variables between groups across clinicopathological characteristics except age at diagnosis. Alternatively, Fisher's exact test was used when χ 2 test was violated. The Wilcoxon rank sum test was used to analyze the difference of age at diagnosis and levels of serum tumor markers between BRCA1/2 mutation carriers and non-BRCA mutation carriers. The obtained p values were considered statistically significant if the p value is <0.05. The false discovery rate procedure was used to adjust p values for multiple testing (Holm, 1979). Logistic regression analysis was used to construct discriminative models with serum tumor markers and BRCA1/2 mutation status for ovarian cancer detection and patient stratification. The predictive performance of logistic regression models was evaluated by sensitivity, specificity, accuracy, and the area under the ROC curve (AUC) measurements. The leave-one-out cross-validation prediction error was also estimated as a performance measurement for these models. All of the computations were performed using the R software (version 3.3.3, http://www. cran.r-project.org).

| Levels of serum tumor markers in ovarian cancer and the associations with clinicopathological characteristics
Among the 11 serum tumor markers, AFP, CA125, CA19-9, CA242, CEA, Ferritin, NSE and HE4 showed elevated levels and CA15-3, HGH, and β-HCG showed reduced levels in ovarian cancer patients ( Table 2). The levels of CA125, CA19-9, CEA, Ferritin, NSE, and HE4 were significantly elevated (p = 2.82E-24, 0.021, 0.029, 6.98E-17, 1.92E-06, and 1.63E-07, respectively) and the level ofβ-HCG was significantly reduced (p = 1.33E-07) in ovarian cancer patients. The serum levels of CA125, Ferritin, β-HCG, and NSE were displayed in Figure 1a. The mean level of CA125 was elevated more than fivefold in ovarian cancer patients when comparing to that in controls (296.14 U/mL vs. 55.74 U/mL). In contrast, the mean level of β-HCG was dramatically reduced in ovarian cancer patients when comparing to that in controls (0.9 ng/ml vs. 26.69 ng/ml). In addition, the mean level of HE4 was also elevated more than sixfold in ovarian cancer patients (271.29 ng/ml vs. 40.33 ng/ml), although it was measured only in 61 ovarian cancer patients and 60 controls.
In addition, CA19-9 and Ferritin was significantly higher in elder and younger patients. CA125, CA15-3, and Ferritin level was significantly associated with Histological subtype. CA125, Ferritin, NSE, and β-HCG level was significantly associated with FIGO stage. In high grade tumor the CA125, CA15-3, and Ferritin were significantly high expressed. Moreover, an association between Ferritin level and Menopausal age was found in ovarian cancer patients.

BRCA1/2 in ovarian cancer patients
Among the 232 ovarian cancer patients, a total of 43 deleterious germline mutation loci were identified in 54 ovarian cancer patients (Table 3). The entire frequency of BRCA1/2 deleterious germline mutations in this study was 23.3% (54/232). Among these 43 mutations, there were 30 frameshift mutations (six insertions, 21 deletions, three deletion-and-insertions), six stop-gain mutations, six splicing mutations, and one missense mutation. Except for the six splicing mutations, all of the BRCA1/2 mutations were illustrated on the protein structures in Figure 2. All deleterious mutations were confirmed by Sanger sequencing ( Figure S1).

| Associations of BRCA1/2 mutations with clinicopathological characteristics and serum tumor markers
The associations of BRCA1/2 deleterious germline mutation status with clinicopathological characteristics of the 232 patients were summarized in Table 1. Regarding to histological subtype of ovarian cancer patients, there was statistically significant difference between BRCA1/2 mutation carriers and non-BRCA mutation carriers in histological subtype (p = 0.011). In addition, the FIGO stage of BRCA1 mutation carriers and BRCA1/2 mutation carriers were significantly different from non-BRCA mutation carriers (p = 0.019 and 0.007 respectively). Notably, BRCA1 mutation carriers and BRCA1/2 mutation carriers were significantly more likely to be positive in lymph node metastasis, when compared with non-BRCA mutation carriers (p = 0.025 and 0.024, respectively). We also investigated whether deleterious germline mutations were associated with Hereditary Breast and Ovarian Cancer (HBOC) related tumor (Table 1). In this study, 38 patients showed family history of cancer and 11 patients had HBOC-related tumors. Overall, 9 in 11 patients with HBOC-related tumor had BRCA1/2 deleterious mutations. Compared to non-BRCA mutation carriers, BRCA1, BRCA2, and BRCA1/2 mutations carriers all showed significant association with HBOC-related tumors (p < 0.001, p = 0.004, and p < 0.001, respectively). BRCA1 mutation carriers and BRCA1/2 mutation carriers also showed significant association with family history of cancer, when compared with non-BRCA mutation carriers (p = 0.049 and 0.014, respectively). The associations between BRCA1/2 deleterious germline mutation status and serum levels of tumor markers were summarized in Table S1. The levels of CA125 in BRCA1 and BRCA1/2 mutation carriers were significantly higher than that in non-BRCA mutation carriers (p = 0.010 and 0.040, respectively). BRCA1 mutation carriers had significantly higher levels of CA15-3, β-HCG (p = 0.035 and 0.027 respectively), and significantly lower level of HGH (p = 0.027) than non-BRCA mutation carriers. No significant difference was found in other serum tumor markers.

| DISCUSSION
Serum tumor markers and BRCA1/2 germline mutations are crucial factors in cancer diagnosis, treatment, and prognosis. In this study, we comprehensively investigated multiple serum tumor markers and BRCA1/2 germline mutations in a Chinese cohort of ovarian cancer patients. The controls we used in this study were from women who visited hospital with benign gynecologic diseases, which might represent real clinic situation better than using healthy woman as control.
We found that several serum tumor markers were associated with BRCA1/2 mutation status. BRCA1/2 mutation status could improve serum tumor markers performance in ovarian cancer discriminative models. BRCA1/2 mutation status provides important information for the identification of patients that are most likely to benefit from treatment with PARP inhibitors and guides treatment decisions. In this study, the overall frequency of BRCA1/2 deleterious germline mutation was 23.3% (16.38% in BRCA1 and 6.90% in BRCA2) in patients from Hunan province in mid-south China. This frequency was close to two multicenter studies from eastern and north China ovarian cancer patients by NGS (Shi et al., 2017;Wu et al., 2017) 16.7% (13.1% in BRCA1 and 3.9% in BRCA2) and 28.4% (20.8% in BRCA1 and 7.6% in BRCA2), respectively. HBOC is syndrome primarily associated with mutations in BRCA1 or BRCA2 genes. Same as previous study, the BRCA1/2 germline mutations were significantly associated with HBOC-related tumor or family history (Bolton et al., 2012;Maistro et al., 2016;Shi et al., 2017). These ovarian cancer patients carrying BRCA1/2 mutations could benefit from PARP inhibitors for targeted treatment (Shi et al., 2017). Serum tumor markers play a crucial role in ovarian cancer diagnosis, prognosis prediction, and monitoring disease progression (Richards et al., 2015). CA125 as a single marker showed good performance in ovarian cancer detection (AUC = 0.799), but with low sensitivity in ovarian cancer detection model (Yang, Lu, & Bast, 2017). To overcome the limitation of single marker in ovarian cancer detection, combined detection of multi-tumor markers has been suggested (Bian et al., 2014). In this study, we found a panel of eight serum tumor markers showing improved performance in ovarian cancer detection (AUC = 0.873) than single CA125 marker.
HE4 was another serum marker that has been approved by FDA. Overexpression of HE4 plays direct biological role in promoting ovarian cancer cell proliferation. In our ovarian cancer detection model, the prediction performance for ovarian cancer was improved with AUC value up to 0.916 when HE4 was included. In addition, the panels including HE4 could predict BRCA1/2 mutation status, especially BRCA1 mutation status with an AUC value as 0.974. A recent study reported that ovarian patients with BRCA1 gene mutations have relatively low serum HE4 levels (Chudecka-Glaz, Cymbaluk-Ploska, Strojna, & Menkiszak, 2017). We also observed lower HE4 level in BRCA1 mutation carriers than that in noncarriers, although the difference is not significant. The mechanism underneath this correlation is worth further investigation in the future. A limitation must be mentioned that HE4 detection was not included in the Protein Chip we used here, and the detection of HE4 was not applied in our hospital until 2017. HE4 level was only detected in 121 samples in our cohort and these results still need future validation in large cohorts.
It is very interesting that by adding lymph node metastasis status as a parameter in patient stratification, the prediction of BRCA1/2 mutation status (AUC = 0.917) was much improved than only using serum markers (AUC = 0.881), and BRCA1/2 mutation status together with five serum markers could predict lymph node metastasis status (AUC = 0.843) better than serum markers alone (AUC = 0.735). It has been reported that ovarian cancer patients with BRCA1/2 mutations had significantly more bulky lymph nodes than patients with wild type BRCA1/2 (Petrillo et al., 2017). Our results also indicated that BRCA1/2 mutation carriers had more lymph node metastasis than nonmutation carriers, especially in BRCA1 mutation carriers. Further investigation of the relationship between lymph node metastasis and BRCA1/2 mutation status in ovarian cancer might provide more information for patient stratification.
In accordance with a previous study , we observed that the elevated serum level of CA125 was also associated with BRCA1/2 mutation. The serum level of HGH was significantly lower in BRCA1 mutation carriers than nonmutation carriers in our study. This may be related to proliferation mechanism that HGH levels in serum do not facilitate tumor cells proliferation (Santovena, Farina, Llabres, Zhu, & Dannies, 2010). Elevated serum level of CA15-3, β-HCG was significantly higher in BRCA1 mutation carriers than nonmutation carriers. Although the difference is not significant after adjusted, it still could be observed that there were differences in levels of CA15-3, β-HCG, HGH between BRCA1/2 mutation carriers, and non-BRCA mutation carriers. Both genetic variations and serum tumor markers are crucial features of cancer. Combined genetics and serum tumor marker variation detection would improve cancer diagnosis and treatment in clinical. A recent research provided inspiring result by combining 12 tumor markers and 16 genes mutation analysis for cancer early detection including ovarian cancer (Cohen et al., 2018). The crucial tumor markers and mutations, such as AFP, CA125， CEA, CA199, HE4 and TP53 (OMIM:191170), PIK3CA (OMIM:171834), EGFR (OMIM:131550), BRAF (OMIM:164757) mutations were included in their cancer SEEK panel. The combined analysis of tumor markers and genetic variations would be a promising method for ovarian cancer detection and provide guideline for treatment. The panels we reported here could be alternative biomarkers for patient stratification using targeted therapy in ovarian cancer such as PARP inhibitor.
In conclusion, we analyzed serum tumor markers and the prevalence of BRCA1/2 germline mutations in Chinese ovarian cancer patients. The identified panels of serum tumor markers showed a good performance in ovarian cancer prediction. In addition, the serum tumor markers combining BRCA1/2 mutation status could also predict metastasis status. These findings provided important information for ovarian cancer prediction and patient stratification, which eventually would benefit the diagnosis and treatment for ovarian cancer patients.

ETHICAL STANDARDS
We declare that the experiments performed in this study comply with the current laws of the People's Republic of China.