Genomic Landscape of Endometrial, Ovarian, and Cervical Cancers in Japan from the Database in the Center for Cancer Genomics and Advanced Therapeutics

Simple Summary This study comprehensively investigated the genomic landscape of >3000 gynecological malignancies (endometrial, cervical, and ovarian cancers) in Japan. The Center for Cancer Genomics and Advanced Therapeutics database used in this study is a useful tool containing real-world data of patients with poor prognoses, as comprehensive genomic profiling tests are limited to patients with cancer who have completed standardized treatments in Japan. Genomic profiling based on histological subtypes, tumor mutational burden, and microsatellite instability highlights actionable mutations for future drug development for each gynecological cancer. Abstract This study aimed to comprehensively clarify the genomic landscape and its association with tumor mutational burden-high (TMB-H, ≥10 mut/Mb) and microsatellite instability-high (MSI-H) in endometrial, cervical, and ovarian cancers. We obtained genomic datasets of a comprehensive genomic profiling test, FoundationOne® CDx, with clinical information using the “Center for Cancer Genomics and Advanced Therapeutics” (C-CAT) database in Japan. Patients can undergo the tests only after standardized treatments under universal health insurance coverage. Endometrial cancers were characterized by a high frequency of TMB-H and MSI-H, especially in endometrioid carcinomas. The lower ratio of POLE exonuclease mutations and the higher ratio of TP53 mutations compared to previous reports suggested the prognostic effects of the molecular subtypes. Among the 839 cervical cancer samples, frequent mutations of KRAS, TP53, PIK3CA, STK11, CDKN2A, and ERBB2 were observed in adenocarcinomas, whereas the ratio of TMB-H was significantly higher in squamous cell carcinomas. Among the 1606 ovarian cancer samples, genomic profiling of serous, clear cell, endometrioid, and mucinous carcinomas was characterized. Pathogenic mutations in the POLE exonuclease domain were associated with high TMB, and the mutation ratio was low in both cervical and ovarian cancers. The C-CAT database is useful for determining the mutational landscape of each cancer type and histological subtype. As the dataset is exclusively collected from patients after the standardized treatments, the information on “druggable” alterations highlights the unmet needs for drug development in major gynecological cancers.


Introduction
Comprehensive genomic profiling (CGP) tests broadly explore treatments based on individual genomic information [1].Until June 2023, three CGP tests have been clinically applicable in Japan, including a tumor-only panel, the FoundationOne ® CDx (F1CDx) assay; a liquid biopsy panel, the FoundationOne Liquid ® CDx assay; and a tumor/normal paired panel, the OncoGuide TM NCC Oncopanel System [2,3].All genomic profiling data and clinical information are transferred to the Center for Cancer Genomics and Advanced Therapeutics (C-CAT) with written informed consent (agreement ratio, 99.7%), and the data are available for research use [3].As the CGP tests under the universal health insurance system in Japan are only applicable to patients who have (already or almost) finished standardized treatments, the dataset is composed of patients with a poor prognosis for all cancer types.Liquid biopsy is limited to patients whose tissue specimens are not available or not suitable for CGP, and to date, F1CDx has been broadly tested (>75%) in Japan.The C-CAT database enables us to understand the mutational landscape, tumor mutational burden (TMB), and microsatellite instability (MSI) status of any type of advanced solid tumor [3].
Both MSI-high and TMB-high (TMB-H, ≥10 mutations/megabase [mut/Mb]) are used as companion diagnostics for an immune checkpoint inhibitor (ICI), pembrolizumab, in solid tumors [13,14].In addition to these tumor-agnostic indications, since December 2022, cemiplimab monotherapy (anti-programmed cell death 1 antibody) has been approved in recurrent cervical cancer as a second-line or later treatment in Japan, regardless of PD-L1 status [15].Since December 2021, lenvatinib (a multi-tyrosine kinase inhibitor) plus pembrolizumab has been approved in Japan for the treatment of advanced/recurrent endometrial cancer, regardless of MSI status [16].Recently, ICI plus platinum-based chemotherapy has shown significantly better overall survival and/or progression-free survival in both endometrial and cervical cancers (either primary advanced or recurrent) [17][18][19].However, the prognostic benefits of ICI-containing regimens are significantly greater in the presence of MSI-H and/or deficient MMR (dMMR) in endometrial cancer and PD-L1 markers in cervical cancer [17][18][19].In ovarian cancer, TMB-H or MSI-H remains the only indication for pembrolizumab, although several ongoing clinical trials include ICIs [20].
In the present study, we aimed to focus on the mutational landscape, TMB, and MSI status of endometrial, cervical, and ovarian cancers in Japanese patients using the C-CAT database of F1CDx (registered from June 2019 to May 2022; https://www.ncc.go.jp/jp/c_ cat/use/index.html,(accessed on 1 June 2022)).

Patient Samples of FoundationOne ® CDx (F1CDx) from the Center for Cancer Genomics and Advanced Therapeutics Database
This Japanese cohort study included 561 endometrial, 839 cervical, and 1606 ovarian cancers that were analyzed using F1CDx under health insurance coverage.The data were obtained from the C-CAT database organized by the National Cancer Center of Japan, which stores the CGP data tests [3].The CGP tests in Japan are limited to patients with solid cancers who have finished (or are expected to finish) standard treatments for advanced unresectable diseases.Therefore, the patients enrolled generally had poor prognoses and were resistant to platinum-based chemotherapies for all three gynecological cancers.We logged into the C-CAT system to collect 3006 of 25,504 patients' F1CDx data for the three gynecological cancers (between June 2019 and May 2022).We accessed the database on 1 June 2022.The workflow of this study is shown in Figure 1.The histological subtypes of each cancer are summarized in Supplementary Table S1.In this study, pure sarcomas were not included in endometrial cancer, whereas 2 sarcomas and 63 non-epithelial tumors were included in cervical and ovarian cancers, respectively.This study was approved by our institutional ethics committee (#2021341G) and the Information Utilization Review Board of C-CAT (#CDU2022-026N).

Patient Samples of FoundationOne ® CDx (F1CDx) from the Center for Cancer Genomics and Advanced Therapeutics Database
This Japanese cohort study included 561 endometrial, 839 cervical, and 1606 ovarian cancers that were analyzed using F1CDx under health insurance coverage.The data were obtained from the C-CAT database organized by the National Cancer Center of Japan, which stores the CGP data tests [3].The CGP tests in Japan are limited to patients with solid cancers who have finished (or are expected to finish) standard treatments for advanced unresectable diseases.Therefore, the patients enrolled generally had poor prognoses and were resistant to platinum-based chemotherapies for all three gynecological cancers.We logged into the C-CAT system to collect 3006 of 25,504 patients' F1CDx data for the three gynecological cancers (between June 2019 and May 2022).The workflow of this study is shown in Figure 1.The histological subtypes of each cancer are summarized in Supplementary Table S1.In this study, pure sarcomas were not included in endometrial cancer, whereas 2 sarcomas and 63 non-epithelial tumors were included in cervical and ovarian cancers, respectively.This study was approved by our institutional ethics committee (#2021341G) and the Information Utilization Review Board of C-CAT (#CDU2022-026N).

F1CDx Testing
F1CDx is a tumor-only panel using DNA isolated from formalin-fixed, paraffin-embedded tumor tissue specimens, which can detect substitutions, insertions, and deletions (indels); copy number alterations (CNAs) in 324 genes; gene rearrangements in 36 genes; and genomic signatures, including MSI and TMB [21].MSI status is reported as "cannot be determined" when the quality is insufficient.TMB by F1CDx is determined by counting all synonymous and non-synonymous variants, except for hotspot genomic alterations, and is considered TMB-H when reported as ≥10 mut/Mb.In our study, all genetic variants,

F1CDx Testing
F1CDx is a tumor-only panel using DNA isolated from formalin-fixed, paraffinembedded tumor tissue specimens, which can detect substitutions, insertions, and deletions (indels); copy number alterations (CNAs) in 324 genes; gene rearrangements in 36 genes; and genomic signatures, including MSI and TMB [21].MSI status is reported as "cannot be determined" when the quality is insufficient.TMB by F1CDx is determined by counting all synonymous and non-synonymous variants, except for hotspot genomic alterations, and is considered TMB-H when reported as ≥10 mut/Mb.In our study, all genetic variants, in-cluding single nucleotide variants, CNAs, and gene fusions, were annotated as pathogenic or likely pathogenic based on CIViC, BRCAExchange, ClinVar, and COSMIC [3].MSI-H and TMB-H are tumor-agnostically approved as CDx for pembrolizumab in solid cancers in Japan.In this study, cases with "cannot be determined" for either TMB or MSI were excluded from the analysis (31 endometrial, 70 cervical, and 80 ovarian cancers).

Statistical Analyses and Graphical Representations
Quantitative variables were analyzed using one-way analysis of variance (ANOVA) (when normality was assumed) and the Kruskal-Wallis H test (when normality could not be assumed) for comparisons among the three groups.Pearson's correlation test was used for correlation analysis between the two groups.All reported p values were two-tailed, and p < 0.05 was considered significant unless otherwise specified.All the graphs, calculations, and statistical analyses were performed using GraphPad Prism software 9.3.0 and R 4.2.0 software.The collation and visual analysis of alteration data were implemented using the "ComplexHeatmap" package in R.

Genomic Alteration Profiles across Cancer Types
We analyzed the genomic alterations (pathogenic or likely pathogenic) in F1CDx from the C-CAT database in 561 endometrial, 839 cervical, and 1606 ovarian cancer samples.The mutational landscape of frequently mutated (pathogenic or likely pathogenic) genes (top 30) in each cancer type and histological subtype is summarized in Supplementary Figure S1, and Figure 2, respectively (A: endometrial, B: cervical, and C: ovarian cancers).
Cancers 2024, 16, x FOR PEER REVIEW 4 of 18 including single nucleotide variants, CNAs, and gene fusions, were annotated as pathogenic or likely pathogenic based on CIViC, BRCAExchange, ClinVar, and COSMIC [3].MSI-H and TMB-H are tumor-agnostically approved as CDx for pembrolizumab in solid cancers in Japan.In this study, cases with "cannot be determined" for either TMB or MSI were excluded from the analysis (31 endometrial, 70 cervical, and 80 ovarian cancers).

Statistical Analyses and Graphical Representations
Quantitative variables were analyzed using one-way analysis of variance (ANOVA) (when normality was assumed) and the Kruskal-Wallis H test (when normality could not be assumed) for comparisons among the three groups.Pearson's correlation test was used for correlation analysis between the two groups.All reported p values were two-tailed, and p < 0.05 was considered significant unless otherwise specified.All the graphs, calculations, and statistical analyses were performed using GraphPad Prism software 9.3.0 and R 4.2.0 software.The collation and visual analysis of alteration data were implemented using the "ComplexHeatmap" package in R.

Genomic Alteration Profiles across Cancer Types
We analyzed the genomic alterations (pathogenic or likely pathogenic) in F1CDx from the C-CAT database in 561 endometrial, 839 cervical, and 1606 ovarian cancer samples.The mutational landscape of frequently mutated (pathogenic or likely pathogenic) genes (top 30) in each cancer type and histological subtype is summarized in Supplementary Figure S1, and Figure 2, respectively (A: endometrial, B: cervical, and C: ovarian cancers).
The TMB value in endometrial cancer was significantly higher than that in cervical cancer (p < 0.001 by one-way ANOVA with the Kruskal-Wallis test) and ovarian cancer (p < 0.001) (Figure 4D).The median TMB values in MSI-H tumors were 21.4 mut/Mb in endometrial, 23.0 mut/Mb in cervical, and 40.4 mut/Mb in ovarian cancers (Figure 4E), with a strong correlation between MSI and TMB in these three cancer types (p < 0.001) (Figure 4E).
The TMB and MSI statuses were distinct among the histological subtypes of each cancer (Supplementary Table S2).
The TMB value in endometrial cancer was significantly higher than that in cervical cancer (p < 0.001 by one-way ANOVA with the Kruskal-Wallis test) and ovarian cancer (p < 0.001) (Figure 4D).The median TMB values in MSI-H tumors were 21.4 mut/Mb in endometrial, 23.0 mut/Mb in cervical, and 40.4 mut/Mb in ovarian cancers (Figure 4E), with a strong correlation between MSI and TMB in these three cancer types (p < 0.001) (Figure 4E).
The TMB and MSI statuses were distinct among the histological subtypes of each cancer (Supplementary Table S2).

Distribution of POLE Genomic Alterations in the Exonuclease Domain among TMB-H and Microsatellite Stable Subsets
All POLE variants (including variants of unknown significance [VUS]) are listed in Table 2.
The ultramutated genotype (TMB > 100 mut/Mb) was identified in eight tumors (five endometrial and three ovarian cancers).In endometrial cancer, all eight (1.4%)POLE exonuclease-mutated tumors were TMB-H (median TMB, 90.78 mut/Mb), of which only one was MSI-H (Table 2).Three MSI-H and TMB-H tumors showed VUS of POLE outside the exonuclease domain, which should be categorized as MSI-H, not as a POLE subgroup (Table 2).Pathogenic/likely pathogenic variants in the POLE exonuclease domain were detected in one case (0.12%) of cervical cancer and three cases (0.19%) of ovarian cancer.None of the POLE variants outside the exonuclease domain were annotated as pathogenic or likely pathogenic (Table 2).
In "TMB-H with MSS" tumors, the ratio of genomic alterations in PIK3CA was the most or the second highest, which was 61.1% in endometrial, 51.4% in cervical, and 31.0% in ovarian cancers (Supplementary Figure S3).Genomic alterations of TP53 were most common in the TMB-H with MSS group in endometrial (61.1%) and ovarian (82.8%) cancers, whereas the rate was 12.0% in cervical cancer (usually human papillomavirus [HPV], which relates to the impairment of TP53 by the ubiquitin-proteasome pathway).The ratio of genomic alterations in CDKN2A and CDKN2B was also high in endometrial and ovarian cancers (Supplementary Figure S3).

Discussion
In this study, we analyzed 3006 endometrial, cervical, and ovarian cancers using a tumor-only panel, F1CDx.The Japanese CGP test dataset is unique in terms of eligible patients and insurance coverage.All the patients have finished or are expected to finish the standardized treatments and take the CGP tests under universal health insurance coverage [3,31].Thus, any poor prognosis in Japanese patients with cancer may allow them to undergo CGP tests.Furthermore, a sufficient number of tumor specimens are usually available through surgery and/or biopsy.Therefore, the C-CAT database is suitable for analyzing the genomic profiles of patients with gynecological cancer with a poor prognosis.
In endometrial cancer, a comparison with the TCGA database highlighted the high incidence of genomic alterations of TP53 (54.4%) and the low incidence of genomic alterations of POLE (1.4%) in this database.This discrepancy supports the significance of the molecular classification of "Proactive Molecular Risk Classifier for Endometrial Cancer" in endometrial cancer by POLE, dMMR, and TP53 [32].Drug development is highly warranted in genomic alterations of the PI3K (PTEN and PIK3CA), RAS (KRAS), and wnt/β-catenin (CTNNB1) pathways in endometrioid carcinomas and TP53, ERBB2, and PIK3CA in non-endometrioid carcinomas (Table 1).A WEE1 inhibitor, adavosertib, showed an objective response rate of 29.4% in recurrent uterine serous carcinomas (usually TP53 mutated), and an international phase IIb study is ongoing [33,34].Further development of precision medicine in endometrial cancer is warranted.
In cervical cancer, the C-CAT dataset was helpful for elucidating the genomic profiling of adenocarcinomas, as the ratio of non-squamous cell carcinomas was significantly lower in the TCGA dataset (19.1%) than in the C-CAT database (53.6%) [8].Key molecular targets, especially in adenocarcinomas, include KRAS, ERBB2, and ARID1A.According to the recently published 5th edition of the World Health Organization classification, cervical cancer is classified as HPV-associated and HPV-independent for each histological type [35].As both the TP53 and RB pathways are impaired by HPV-E6 and HPV-E7 oncoproteins, respectively, genomic alterations of TP53, RB, and CDKN2A/2B are informative for speculating HPV-independent cervical cancers, especially in gastric-type mucinous adenocarcinomas [36,37].
One limitation of the C-CAT database is that data on low-grade serous ovarian carcinomas are mixed with those on high-grade serous carcinomas.Genomic alterations of TP53 in 90% of serous carcinomas suggest that these tumors represent high-grade serous carcinomas.The RAS-MAPK signaling pathway (genomic alterations of NF1 at 16% and KRAS at 12% with mutual exclusivity), the PI3K-mTOR pathway (PIK3CA at 12% and TSC2 at 8%), and certain receptor tyrosine kinases (ROS1 at 9% and ERBB2 at 8%) might be candidates for targeted therapy in serous carcinomas.The pathogenicity of each alteration, especially in BRCA1 and BRCA2, should be carefully addressed [9,30].Drug development targeting ARID1A and PIK3CA in clear-cell ovarian carcinomas is also warranted.Currently, a p110alpha selective inhibitor, CYH33, is under phase 2 clinical trials (NCT05043922, jRCT2031210216), which recruits patients with clear cell ovarian carcinoma with hotspot mutations in PIK3CA (Table 1) [29].Targeting the RAS-MAPK pathway should be key in mucinous carcinomas.
In agreement with previous findings, MSI-H in this study was the main causative genomic finding for TMB-H induction in endometrial cancer, whereas it shared only 10% and 24% of TMB-H in cervical and ovarian cancers, respectively [39,40].A low TMB-H ratio (5.0%) in ovarian cancer may be associated with limited sensitivity to ICIs [41].A comparison between "TMB-H with MSS" and "MSI-H" in each cancer type is informative to elucidate real "driver" alterations.In endometrial and ovarian cancers, the frequency of genomic alterations in TP53 and CDKN2A/2B was significantly higher in the group of "TMB-H with MSS".These findings suggest that TMB-H should be subclassified according to the MSI status.Although pembrolizumab has been approved in any solid cancers with either TMB-H and MSI-H, combination therapies with immune checkpoint inhibitors may be developed separately according to the status of TMB and MSI.
This study has some limitations.First, CGP tests in Japan are reimbursed only for patients who have (almost) finished standardized treatments, suggesting that patients with rapid progression may miss the opportunity to undergo CGP tests.In addition, this study lacks data from patients without medical insurance due to the universal health insurance system in Japan.Second, the response to genome-matched therapies was not analyzed in this study because of the low accessibility of the recommended drugs.Third, the C-CAT database was deposited at designated hospitals located in Japan.Therefore, most of the patients were Japanese.

Conclusions
This study uniquely illustrates the genomic landscape of three major gynecological cancers in the Japanese cohort.It highlights the necessity of future drug developments in each cancer type and each histological subtype.ERBB2, PIK3CA, ARID1A, and KRAS would be key molecular targets in gynecological cancers.Furthermore, the prevalence and correlation between TMB and MSI may influence future immunotherapy, including combination therapies.These insights reinforce the necessity of molecular classification in understanding tumor biology and developing personalized therapies, underlining the potential of genomic profiling in precision oncology.

Supplementary Materials:
The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/cancers16010136/s1, Figure S1.Genomic landscape of three gynecological cancers.Recurrently mutated genes are listed with the status of TMB and MSI and with information about types of alterations and histological subtypes in (A) endometrial, (B) cervical, and (C) ovarian cancers.The upper plot represents the TMB scores by F1CDx.Waterfall plot of genetic alteration profiles in endometrial (A), cervical (B), and ovarian cancer (C). Figure S2.Frequency of genomic alterations in the mismatch repair (MMR) genes according to the MSI and TMB stauts in each cancer type.(A) Frequency of MMR alterations according to the MSI status, (B) Frequency of MMR alterations according to the TMB status.Comparisons between the groups were performed by Fisher's exact test (* p < 0.05; ** p < 0.01; *** p < 0.001).MMR-pv, pathogenic variants in MMR genes, MMR-wt, no pathogenic variants (wild-type) in MMR genes.Figure S3.Genomic landscape according to the TMB and MSI status in (A) endometrial, (B) cervical, and (C) ovarian cancer.Each cancer was categorized as TMB-H with MSS, MSI-H (regardless of TMB status), and TMB-L with MSS.Table S1.Distribution of histological subtypes in each cancer.Table S2.Frequency of MSI-H and TMB-H according to the histological subtypes in each cancer.Table S3.Genomic alterations of MMR genes in each cancer with MSI-H.
Funding: This study was funded by a Grant-in-Aid for Scientific Research (B) (grant number: 21H03074 to K.O.).

Institutional Review Board Statement:
This study was approved by our institutional ethics committee (#2021341G) and the Information Utilization Review Board of C-CAT (#CDU2022-026N).
Informed Consent Statement: Written informed consent has been obtained from all the patients in each hospital to deposit the data to the C-CAT database for research use.

Figure 1 .
Figure 1.A workflow of the study to analyze the C-CAT database.Clinical information from each hospital and genomic data from certified laboratories (i.e., Foundation Medicine, Inc. for F1CDx) are collected and sent to the C-CAT data center.C-CAT reports with annotations of each alteration and therapeutic options are returned to each hospital.The C-CAT datasets with clinical information can be used for research purposes with the permission of institutional ethics committees and the Information Utilization Review Board of C-CAT.

Figure 1 .
Figure 1.A workflow of the study to analyze the C-CAT database.Clinical information from each hospital and genomic data from certified laboratories (i.e., Foundation Medicine, Inc. for F1CDx) are collected and sent to the C-CAT data center.C-CAT reports with annotations of each alteration and therapeutic options are returned to each hospital.The C-CAT datasets with clinical information can be used for research purposes with the permission of institutional ethics committees and the Information Utilization Review Board of C-CAT.

Figure 4 .
Figure 4. Venn diagrams of tumor mutational burden (TMB)-high and microsatellite instability-high in (A) endometrial, (B) cervical, and (C) ovarian cancers.(D) Box plots of TMB levels and (E) scatter dot plots of TMB distribution according to the MSI status in each cancer.** p < 0.01, *** p < 0.001 using one-way analysis of variance with the Kruskal-Wallis test.

Figure 4 .
Figure 4. Venn diagrams of tumor mutational burden (TMB)-high and microsatellite instability-high in (A) endometrial, (B) cervical, and (C) ovarian cancers.(D) Box plots of TMB levels and (E) scatter dot plots of TMB distribution according to the MSI status in each cancer.** p < 0.01, *** p < 0.001 using one-way analysis of variance with the Kruskal-Wallis test.

Figure 5 .
Figure 5. Frequency of (A) microsatellite instability-high and (B) tumor mutational burden-high according to the major histological subtypes in endometrial, ovarian, and cervical cancers.

Figure 5 .
Figure 5. Frequency of (A) microsatellite instability-high and (B) tumor mutational burden-high according to the major histological subtypes in endometrial, ovarian, and cervical cancers.

Table 1 .
Genomic Alterations and Potential Targeted Therapies in Gynecological Cancers.

Table 2 .
List of POLE variants with their pathogenicity, MSI, and TMB status in each cancer.