Oncogenic pathway landscape of ovarian cancer and correlation with clinical prognosis.

Ovarian cancer is the fifth most common cancer in women and the leading cause of death among gynecologic malignancies [ 1 2 ]. With a better understanding of genetic mutations and the mechanisms of disease development, clinical research is currently underway to identify mutations that could serve as potential therapeutic targets in ovarian cancer and to evaluate targeted therapies for these mutations [ 3 ]. Several bioinformatics platforms are available to provide rapid access to the most up-to-date information regarding genomic mutations and the corresponding targeted drugs [ 4 5 ]. Next-generation sequencing (NGS), a method that can reveal genomic abnormalities by simultaneously analyzing multiple genes, is primarily utilized to confirm the genetic profile of cancer patients. However, most papers on NGS published to date only focused on the frequency of detecting genetic mutations in cancer patients [ 6 7 ]. This limitation is believed to be associated with the genetic characteristics of ovarian cancer. Unlike cancers in other sites where driver mutations exist, ovarian cancer is a complex and polygenic disease, which makes it difficult to analyze the relationship between individual genetic alterations and clinical prognosis. Since a mutation in a gene indicates a problem in a related pathway, analyzing oncogenic pathways can effectively determine the clinical significance of mutations in specific genes. Considering the characteristics of ovarian cancer, which involve polygenic mutations, analyzing the oncogenic pathway involved rather than individual genetic alterations allows for more accurate analysis and is also an effective method for selecting the appropriate target agents that specifically target the identified oncogenic pathways. In this study, we aimed to identify the main oncogenic pathway by histological type in patients with ovarian, tubal, and primary peritoneal cancer, based on NGS test results and to determine the correlation with clinical prognosis.

A total of 420 patients were identified who underwent NGS testing after being diagnosed with ovarian, tubal, and primary peritoneal cancer during the study period. The analysis was performed on a total of 378 patients after excluding those who met the exclusion criteria. The baseline characteristics of these patients are summarized in Table 1 . The median age at diagnosis was 55 years, and 304 (80.4%) patients had advanced-stage disease (Stages III/IV). An average of 6.8 rounds of chemotherapy were administered as the first-line treatment, and 295 (78.0%) patients experienced recurrence. The median PFS and OS were 23.7 months (interquartile range [IQR]=13.6–37.8) months and 52.4 months (IQR=32.6–75.6), respectively. NGS testing was performed during the initial treatment in 239 (63.2%) cases and after recurrence in 139 (36.8%) cases. Primary tissue was used as the specimen for NGS testing in 338 (89.4%) cases, while recurrent tissue was used in 40 (10.6%) cases. The median tumor cellularity was 60% (IQR=40%–70%), and the median tumor mutational burden was 12.5 mutation/Mb (IQR=9.4–17.2). Histologic types of study patients are shown in Fig. S1 . Values are presented as median (interquartile range) or number (%). CA-125, cancer antigen 125; CR, complete remission; FIGO, the International Federation of Gynecology and Obstetrics; IDS, interval debulking surgery; ORR, objective response rate; OS, overall survival; PDS, primary debulking surgery; PFS, progression free survival; PR, partial response; PD, progression disease; SD, stable disease. Fig. 1 shows the frequency of oncogenic pathways associated with each histology. TP53 mutation was the main oncogenic pathway in the majority of patients with high-grade serous carcinoma (HGSC) (256/276, 92.8%), carcinosarcoma (7/8, 87.5%), and mucinous carcinoma (6/11, 54.5%). Alteration of the DNA damage response pathway was identified in 39.5% (109/276) of HGSC patients, 25% (3/12) of low-grade serous carcinoma (LGSC) patients, and 24% (6/25) of endometrioid carcinoma patients. The main oncogenic pathway in patients with LGSC was MAP kinase signaling, and alterations in this pathway were identified in 58.3% (7/12) of patients, with 4 patients having KRAS mutation, 3 patients having BRAF mutation, and 1 patient having NRAS mutation. In patients with mucinous carcinoma, alterations in MAP kinase signaling were identified in 54.5% (6/11) of cases, and all 6 patients had a KRAS mutation. The presence of a wider range of oncogenic pathways was observed in patients with endometrioid carcinoma and clear cell carcinoma, and alterations in the PI3K-AKT-mTOR signaling and SWI/SNF family pathways were the most common in both groups. ARID1A mutation was identified in 28% (7/25) of patients with endometrioid carcinoma and 59.4% (19/32) of patients with clear cell carcinoma. In endometrioid carcinoma, alterations in the WNT/β-catenin signaling pathway (WNT1, CTNNB1, APC, FZD1) and other oncogenic pathways were confirmed in 7 patients (CTNNB1 5, APC 2). Additionally, although the number of patients was small, a FOXL2 mutation was confirmed in all three adult granulosa cell tumor (GCT) patients. HGSC, high-grade serous carcinoma; LGSC, low-grade serous carcinoma. Figs. 2 and 3 show the PFS and OS of patients with HGSC according to the alteration of each oncogenic pathway. A Cox proportional hazards model was used, and adjustments were made for age, FIGO stage, initial treatment ( N -acetylcysteine or primary debulking surgery), residual disease after debulking surgery, and first-line chemotherapy regimen. Pathway #4, which had the lowest mutation frequency (8 patients), was not included in the graph, and the analysis results indicated that it was not statistically significant in terms of PFS (hazard ratio [HR]=0.78; p=0.55) and OS (HR=1.28; p=0.63). For pathway #2, the results showed an association with better PFS (HR=0.737; 95% confidence interval [CI]=0.563–0.965; p=0.026), but not with OS (HR=0.738; 95% CI=0.509–1.071; p=0.109). Pathway #6 did not show a significant association with PFS (HR=1.162; 95% CI=0.676–1.998; p=0.587); however, it did show an association with better OS (HR=0.290; 95% CI=0.091–0.916; p=0.035). For all other pathways, no statistical correlation could be found with PFS and OS. HGSC, high-grade serous carcinoma; PFS, progression free survival. HGSC, high-grade serous carcinoma; OS, overall survival. Fig. S2 shows the prognosis of HGSC patients based on CCNE1, PIK3CA, and MYC amplification. In patients with HGSC, amplification of CCNE1, PIK3CA, and MYC was confirmed in 27 (9.8%), 12 (4.3%), and 53 (19.2%) patients, respectively. PIK3CA showed a potential association with worse PFS (HR=1.794; 95% CI=0.976–3.298; p=0.057), but not with OS (HR=1.484; 95% CI=0.694–3.190; p=0.311). For CCNE1 and MYC amplification, no statistical correlation could be found with PFS and OS. Additionally, we assessed whether alterations in a specific oncogenic pathway were associated with the age of onset, origin site of cancer, and tumor grade in patients with HGSC. However, the statistical analysis did not reveal any significant results.

TP53 mutation is the main oncogenic pathway commonly found in HGSC, carcinosarcoma, and mucinous carcinoma. High frequency of TP53 mutation, BRCA1/2 inactivation, and widespread copy number changes are hallmarks of HGSC. TP53 mutation is identified in 94% of high-grade serous ovarian cancers [ 19 ]. The absence of a TP53 mutation potentially suggests the possibility of false negative results due to low tumor cell purity, because almost all high-grade serous ovarian carcinomas have an oncogenic TP53 mutation. Among the 276 HGSC patients included in this study, TP53 mutation was not found in 20 (7.2%) patients, and in 6 of them, no mutations were detected. In LGSC and mucinous carcinoma, the MAP kinase signaling pathway was identified as the main oncogenic pathway in more than 50% of patients. In a recent paper analyzing genomic alterations in the MAPK-ERK pathway in gynecologic malignancies, the highest incidence was observed among patients with mucinous ovarian (71%), low-grade serous ovarian (48%), and endometrioid ovarian (37%) tumors [ 20 ], which is largely consistent with the results of this study. KRAS is a GTPase that functions as an upstream regulator of the MAPK and PI3K pathways, and it is therefore critical for cell proliferation, survival, and differentiation. Oncogenic KRAS mutation is one of the frequent molecular alterations in mucinous ovarian tumors, which can be identified in about two-thirds of the cases [ 21 ]. Laboratory studies and preliminary clinical data suggest that cancers with KRAS mutations may be sensitive to MEK or ERK inhibitors [ 22 ]. Currently, a phase 2 study of VS-6766 (RAF/MEK clamp) alone and in combination with defactinib (FAK inhibitor) in recurrent low-grade serous ovarian cancer (ENGOT-ov60/GOG-3052/RAMP 201) is in progress [ 23 ]. All changes in MAP kinase signaling identified in mucinous carcinoma resulted from KRAS mutation; however, in the case of LGSC patients, three patients with BRAF mutation were identified. As reported in a study by Memorial Sloan Kettering, BRAF mutations are detected less frequently in low-grade serous cancers (5%) than in serous borderline tumors (45%) [ 24 ]. The presence of the BRAF V600E mutation in serous borderline/LGSC is associated with early-stage disease and an improved prognosis (low risk of recurrence). The BRAF V600E mutation is known to be oncogenic. The FDA has granted accelerated approval to the RAF-targeted inhibitor dabrafenib in combination with the MEK1/2-targeted inhibitor trametinib for the treatment of unresectable or metastatic solid tumors with BRAF V600E mutation. Among the study patients, there were two cases of BRAF V600E mutation; however, treatment using targeted agents was not administered. Both endometrioid carcinoma and clear cell carcinoma exhibit alterations in various oncogenic pathways, with changes in PI3K-AKT-mTOR signaling being confirmed in nearly 50% of cases. The PI3K-AKT-mTOR signaling pathway functions in cell growth and proliferation, protein translation and synthesis, and the regulation of apoptosis. In addition, activation of the PI3K signaling pathway is important for cancer development. A number of drugs targeting PI3K signaling are currently being evaluated in clinical trials, and it is becoming increasingly clear that PI3K inhibitors are effective in inhibiting tumor progression [ 25 ]. While alpelisib (PI3Kα-selective inhibitor) in combination with fulvestrant (selective estrogen receptor down-regulator) is FDA-approved for the treatment of patients with hormone receptor-positive, HER2-negative, PIK3CA-mutated advanced or metastatic breast cancer, the clinical utility of this combination in patients with PIK3CA-mutant ovarian cancer types has yet to be defined. Alterations in the SWI/SNF family pathway were most frequently found in clear cell carcinoma and endometrioid carcinoma. ARID1A mutation was identified in 59.4% of clear cell carcinoma patients and 28% of endometrioid carcinoma patients. In a recent paper analyzing the molecular landscape of Müllerian clear cell carcinomas, the most frequently mutated genes were ARID1A (48%), PIK3CA (45%), TP53 (23%), and PTEN (10%), which is almost identical to the results of this study [ 26 ]. ARID1A, a tumor suppressor involved in transcriptional regulation, is frequently inactivated by mutation in endometriosis-associated ovarian cancer. Temsirolimus (small molecule inhibitor of the PI3K/AKT pathway and GSK126 (EZH2 inhibitor) have demonstrated proliferation inhibition in ARID1A-mutated ovarian clear cell adenocarcinoma, and many related clinical studies are currently underway [ 27 ]. In endometrioid carcinoma, alterations in WNT/β-catenin signaling were confirmed in 7 (28%) patients (CTNNB1 5, APC 2). CTNNB1 mutations and abnormal β-catenin expression have a negative prognosis in endometrial endometrioid carcinoma, and recent evidence also indicates that β-catenin plays a prognostic role in ovarian endometrioid carcinoma. In a recent study, CTNNB1 mutations and nuclear β-catenin expression were found to be correlated with earlier FIGO stages and associated with better PFS in patients with endometrioid ovarian cancer [ 28 ]. Several clinical trials related to this signaling pathway are being conducted, but no clear results have been published yet. Although the number of patients was small, the FOXL2 C134W mutation was confirmed in all three adult GCT patients. In a previous study, FOXL2 mutation was found to be present in the majority of primary (89.2%) and metastatic adult GCT (80.0%) cases, and the authors suggested that this mutation could be a valuable tool in diagnosing the disease and identifying metastatic lesions with an unknown primary origin [ 29 ]. For pathway #2 (DNA damage response), the results of the correlation analysis showed an association with better PFS, but not with OS. The results of the SOLO1/GOG 3004 Trial demonstrated that first-line PARP inhibitor (PARPi) maintenance therapy was helpful in improving not only PFS but also OS [ 30 ]. However, no clear results related to OS improvement with the use of PARPi have been reported in platinum-resistant recurrent ovarian cancer, which can be explained by the shared molecular characteristics and clinical predictors of platinum and PARPi sensitivity, which imply an overlap of resistance mechanisms [ 31 ]. Among the HGSC patients included in this study, a total of 115 patients used PARPi, of whom only 28 (24.3%) received PARPi maintenance therapy as a first-line treatment. This could explain the lack of a significant correlation between the alteration of pathway #2 and OS. Based on this, it is thought that PARPi treatment should be performed at first-line rather than after relapse, if possible. For pathway #6 (RTK signaling family), correlation analysis showed an association with improved OS despite no association with PFS. Compared to the pathway #1 mutation group, which is the most representative group in HGSC patients, the pathway #6 mutation group showed a similar pattern in the PFS ( Fig. 2 ), while OS showed some differences ( Fig. 3 ). Recurrence occurred in 14 out of 16 patients with pathway #6 mutation, but only 3 of them died, and 2 of them survived for more than 5 years. Among the 14 patients with recurrence, 8 patients maintained platinum-sensitive status for more than 5 years. Based on this data, it is assumed that platinum sensitivity lasts relatively long in patients with pathway #6 mutations due to their biological characteristics. Based on the findings of this study, it is anticipated that the use of MEK or ERK inhibitors in LGSC and mucinous carcinoma will yield positive outcomes in clinical trials. In patients with endometrioid carcinoma and clear cell carcinoma, clinical prognosis can be improved by implementing targeted treatments such as PI3K inhibitors, MEK or ERK inhibitors, and EZH2 inhibitor-based combinatorial strategies tailored to the patient’s genetic profile through NGS testing. Additional research is needed to determine whether the involvement of the RTK signaling family pathway is indeed associated with better OS and to identify the underlying reasons for this association. This is the first study to identify the landscape of oncogenic pathways in ovarian cancer patients and analyze their association with clinical prognosis. This study had some limitations that should be considered. In cases of ovarian cancer in which polygenic mutations occur, it is often unclear which mutation is the driver mutation. Moreover, because a mutation in one gene can simultaneously affect multiple pathways, there are inherent limitations in clearly distinguishing and analyzing individual pathways. Additionally, since cancer is heterogeneous, even if the region where NGS was conducted is representative, it may not accurately reflect the genetic mutation profile of all other metastatic regions. We attempted to compare PFS and OS according to alterations in each major oncogenic pathway in patients with cell types other than HGSC; however, due to the limited number of patients, only the pathways with a high frequency were analyzed, and no significant results were obtained. Unlike cancers in other sites with clear driver mutations, ovarian cancer exhibits a wide range of cell types, and due to its molecular diversity, ovarian cancer cannot be simply classified using only 2 or 3 biomarkers. So, it is necessary to develop insight by identifying the overall oncogenic pattern and understanding the general characteristics of the tumor. The clinical prognosis may be improved by implementing targeted treatment tailored to the patient’s genetic profile through NGS testing.

We conducted a retrospective review of the NGS test results and medical records of all patients diagnosed with ovarian, fallopian tube, or primary peritoneal cancer who underwent NGS testing at Asan Medical Center (Seoul, Korea) between June 1, 2017 and May 31, 2021. Cases with incomplete medical records, loss of follow-up, and cases in which chemotherapy was not administered were excluded from the analysis. Among clinical research participants, those whose drug treatment was terminated early due to the study being discontinued at the decision of the sponsor were excluded from the corresponding analysis because the efficacy of the drug could not be confirmed. After reviewing the matched hematoxylin and eosin-stained slides from each FFPE tissue section, a pathologist selected two to five 6-μm-thick slices from each specimen for genomic DNA extraction. The number of slices chosen depended on the sample size and tumor cellularity. After treatment with xylene and ethanol for deparaffinization, genomic DNA was isolated using a NEXprep FFPE Tissue kit (NexK-9000; Genes Laboratories, Seongnam, Korea), following the manufacturer’s protocol. Briefly, tissue pellets were completely lysed by incubating them with proteinase K in the lysis buffer overnight at 56°C, followed by an additional incubation for 3 minutes with magnetic beads and solution A at 37°C. After incubating for 5 minutes on a magnetic stand, the supernatants were removed and washed three times with ethanol. After air-drying the beads for 5 minutes, DNA was eluted in 50 μL of nuclease-free water and quantified using a Qubit dsDNA HS Assay kit (Thermo Fisher Scientific, Waltham, MA, USA). Targeted NGS was performed using the NextSeq 550DX platform (Illumina Inc., San Diego, CA, USA) with OncoPanel AMC version 4.5, which was designed by Asan Medical Center through SureDesign (Agilent Technologies, Santa Clara, CA, USA) using the GRCh37 reference version. This panel was approximately 1.2 Mbp with 33,524 probes targeting a total of 343 genes, including entire exons of 244 genes, 110 hot spots, and partial introns for eight genes that are often rearranged in cancer. Entire exons for four MMR genes, MLH1, MSH2, MSH6, and PMS2, were covered by the OncoPanel AMC version 4.5 panel ( Table S1 ). Two hundred nanograms of gDNA were fragmented by sonication (Covaris Inc., Woburn, MA, USA), resulting in a mean size of 250 bp. The fragmented DNA was then subjected to size selection using Agencourt AMPure XP beads (Beckman Coulter, High Wycombe, UK). A DNA library was prepared by sequentially performing end repair, A-tailing, and ligation with a TruSeq adaptor using a SureSelect XT Reagent kit (Agilent Technologies). Each library was labeled with sample-specific barcodes consisting of 6 bp and quantified using Qubit. Twenty-four libraries were pooled for hybrid capture using an Agilent SureSelectXT custom kit (Agilent Technologies). The concentration of the enriched target was measured using quantitative PCR (Kapa Biosystems, Woburn, MA, USA), and the sample was loaded onto the NextSeq 550Dx platform (Illumina Inc.) for paired-end sequencing. Sequenced reads were aligned to the human reference genome (National Center for Biotechnology Information build 37) using BWA (0.5.9) with default options [ 8 ]. Demultiplexing was performed using the MarkDuplicates tool in the Picard package to remove PCR duplicates (Broad Institute, Cambridge, MA, USA; http://broadinstitute.github.io/picard , last accessed February 14, 2018). Deduplicated reads were realigned at known indel positions using the GATK IndelRealigner tool [ 9 ]. Next, base qualities were recalibrated using the GATK BaseRecalibrator tool ( https://software.broadinstitute.org/gatk/download ). Somatic single-nucleotide variants (SNVs) and short indels were detected with an unmatched normal using VarDict and the SomaticIndelocator tool in GATK (Broad Institute) [ 9 10 11 ]. Common and germline variants from somatic variant candidates were filtered out with the common dbSNP build 141 (found in >1% of samples), the Exome Aggregation Consortium release 0.3.1 ( http://exac.broadinstitute.org ), and the Korean Reference Genome database ( http://152.99.75.168/KRGDB ) and an in-house panel of normal variants [ 12 13 ]. Final somatic variants were annotated using Variant Effect Predictor version 79 [ 14 ] and converted to maf file format using vcf2maf (GitHub, https://github.com/mskcc/vcf2maf , last accessed February 14, 2018). False-positive variants were manually curated using Integrative Genomics Viewer [ 15 ]. Genetic testing results were classified according to the American Society of Clinical Oncology guidelines into a four-tier classification system. Tier I included variants with strong clinical significance (evidence level A [FDA-approved therapy included in professional guidelines] or B [well-powered studies with consensus from experts in the field]). Tier II included variants with potential clinical significance (evidence level C [FDA-approved therapies for different tumor types or investigational therapies; multiple small published studies with some consensus] or D [preclinical trials or a few case reports without consensus]). Tier III included variants with unknown clinical significance. Lastly, tier IV included variants that are benign or likely benign [ 16 ]. Identified mutations were categorized into seven pathways or functional groups of genetic alteration most frequently associated with ovarian cancer as follows [ 17 18 ]. Pathway #1: p53; pathway #2: DNA damage response; pathway #3: cell cycle; pathway #4: PI3K-AKT-mTOR signaling; pathway #5: MAP kinase signaling; pathway #6: RTK signaling family; pathway #7: SWI/SNF family ( Table S2 ). Descriptive analysis is used to analyze demographic and clinical information. Overall survival (OS) is defined as the date of death or last follow-up from the date of diagnosis of ovarian, fallopian tube, or primary peritoneal cancer. The survival rate was estimated using the Kaplan-Meier method, and differences in survival were evaluated using a log-rank test. The Cox proportional hazards model was used to identify which covariates were statistically significantly associated with progression free survival (PFS) or OS. The multivariable Cox model was used to adjust for confounding variables, including age, FIGO stage, initial treatment, and postoperative residual. All reported p-values were two-sided, and p-values less than 0.05 were considered significant. All statistical analyses were performed using SAS® version 9.4 (SAS Institute Inc., Cary, NC, USA) software and R 3.6.1 ( https://www.r-project.org/ ). The study protocol was approved by Institutional Review Board of the Asan Medical Center (IRB No. S2023-1065-0001).