Molecular profiling reveals novel therapeutic targets and clonal evolution in ovarian clear cell carcinoma

The Institutional Review Board of the Chang Gung Memorial Hospital (Taiwan) approved this study (IRB reference number: 202000143B0), which included a cohort of patients diagnosed with OCCC between 2016 and 2021. Tumor stages were determined using the Federation Internationale de Gynecolgie et d’Obstetrique (FIGO) guidelines [ 21 ]. Formalin-fixed, paraffin-embedded (FFPE) specimens were collected from the Linkou and Kaohsiung branches of the Chang Gung Memorial Hospital. Details of patient enrollment are illustrated in Fig.  1 . OCCC is characterized by a combination of tubulocystic, papillary, and solid pattern with clear, eosinophilic, and hobnail cells [ 22 ]. In cases where peripheral blood samples were unavailable, normal samples were obtained from non-tumoral tissues, such as lymph nodes, fallopian tubes, or the uterus. Clinical and histopathological data, including age, stage, pathological features, and survival outcomes, were retrospectively collected from the patients’ medical records. Fig. 1 Study flowchart. Patients diagnosed with ovarian clear cell carcinoma were retrospectively enrolled based on the registry data. The Systemized Nomenclature of Medicine (SNOMED) codes 87,000-A-M83103, 87,000-B-M83103, and 87,000-C-M83103 were used to identify the relevant cases. FFPE, formalin-fixed paraffin-embedded; WES, whole-exome sequencing Study flowchart. Patients diagnosed with ovarian clear cell carcinoma were retrospectively enrolled based on the registry data. The Systemized Nomenclature of Medicine (SNOMED) codes 87,000-A-M83103, 87,000-B-M83103, and 87,000-C-M83103 were used to identify the relevant cases. FFPE, formalin-fixed paraffin-embedded; WES, whole-exome sequencing Genomic DNA was extracted from 10-μm FFPE sections of tumor and normal tissues using the QIAamp DNA FFPE Tissue Kit (Qiagen, Hilden, Germany), and from peripheral blood using the QIAamp DNA Blood Mini Kit (Qiagen). DNA concentration and integrity were assessed with the Quanti-iT dsDNA HS assay (Invitrogen, Carlsbad, CA, USA) and a fragment analyzer (Advanced Analytical Technologies, Ankeny, IA, USA), respectively. We used the Twist Human Core Exome EF Multiplex Complete Kit Enrichment (Twist Bioscience, South San Francisco, CA, USA) for library preparation. In brief, 50 ng of genomic DNA per sample was subjected to enzymatic fragmentation, and the resulting DNA fragments were subsequently used for library construction. Sequencing was performed on a Novaseq 6000 high-throughput sequencing platform (Illumina, San Diego, CA, USA). Exome reads were trimmed with trimmomatic (version 0.39) to remove adaptors and poor-quality sequences [ 23 ]. Subsequently, they were mapped to the human reference sequence GRCh38.p7 using BWA-mem (version 0.7.15) with default parameters [ 24 ]. Global mapping quality was evaluated using qualimap 2 (version 2.2.1) [ 25 ]. The median read depth for the tumor and normal tissue on the capturing target were 160 × (range: 95 − 255 ×) and 136 × (range: 74 − 452 ×), respectively (Supplementary Table S1). Somatic single-base substitutions (SBSs) and small insertion and deletions (indels) were called using three different tools: MuTect2 (version 4.1.6.0), Strelka2 (version 2.9.2), and Varscan (version 2.3.9). Default parameters were used in all instances [ 26 – 28 ]. Somatic variants were selected based on the following criteria: (i) identification by at least two variant callers, (ii) a minimum read depth of 10 in normal samples and 20 in tumor samples, (iii) a mutated allele fraction of at least 0.1 for C > T and G > A variants (or 0.05 for others), and (iv) at least 7 supporting reads for C > T and G > A variants in the tumor sample (or 5 for others). SBSs and indels were annotated using wANNOVAR [ 29 ]. Gene driver status was determined using the COSMIC database [ 30 ]. Molecular targets were classified following the joint consensus of the Association for Molecular Pathology, American College of Medical Genetics and Genomics, American Society of Clinical Oncology, and College of American Pathologists [ 31 ]. The clinical actionability of the identified genetic alterations was categorized into two tiers. Tier 1 encompassed alterations with direct clinical implications and FDA-approved therapies specifically tailored for the given tumor type. Tier 2 included genetic alterations with FDA-approved treatments for different tumor types, as well as those linked to investigational therapies, and findings derived from meta-analyses, preclinical studies, and case reports. We used freebayes (version 1.3.0) for calling germline variants and wANNOVAR for their annotation in normal samples [ 32 ]. Our analysis focused on 20 cancer-predisposing genes linked to an increased risk of ovarian cancer [ 33 , 34 ], including BRCA1, BRCA2, NTHL1, BRIP1, RAD51C, RAD51D, PALB2, ATM, MLH1, MLH3, MSH2, MSH3, MSH6, PMS1, PMS2, EPCAM, STK11, TP53, and CHEK2 . Germline variants were validated if they had an allele fraction over 0.2, read depth over 20, and were truncating mutations or known pathogenic variants listed in ClinVar [ 35 ]. Mutational signature assignment was conducted using the mSigAct R package (version 2.2) and the COSMIC database (version 3.2) [ 36 , 37 ]. The mSigAct::SparseAssignActivity function was employed to quantify the contribution of SBS signatures in ovarian cancer mutational spectra [ 37 ]. Notably, SBS3, associated with homologous recombination deficiency, was only considered when more than five deletions with microhomology were present, whereas SBS6, SBS26, and SBS44, linked to DNA mismatch repair deficiency, were only applied when over 20 insertions in thymine homopolymer sequences were detected. Analysis of SCNAs was performed using the Sequenza software, version 3.0.0 [ 38 ], with default settings except for the parameter '–het 0.4' and a 200 kbp resolution. The EstimateClonality R package was employed to infer the genome doubling (GD) status [ 39 ]. First, the average copy number (CN) at the chromosomal arm level was calculated from SCNA data. Subsequently, the EstimateClonality::GD.function was used to calculate a p value from 10,000 simulations, based on SCNA probabilities in each tumor, following previously established p value thresholds [ 39 ]. Tumors with p values below these cutoffs were classified as having undergone GD. The genomic instability index (GII) was defined as the fraction of the genome altered by somatic CN gains or losses, with a threshold of ≥ 1 compared to the background ploidy. The genome was partitioned into 5 MB bins by chromosome, and consensus clustering was performed using the ConsensusClusterPlus R package [ 40 ]. The clustering was based on the Euclidean distance with 10,000 iterations. The clonal status of somatic mutations was assessed using a previously published method to determine the timing of their occurrence [ 41 ]. The analysis involved determining the fraction of cancer cells, estimating SCNAs, and evaluating the GD status. Specifically, the MutationTimeR package in R was applied, conducting 1,000 simulations to ascertain the timing of clonal events [ 41 ]. Detailed R scripts and the timing data for all somatic mutations are available at the GitHub repository ( https://github.com/CYHuang-Lab/CGMH-OCCC-WES-project/ ). The TERT gene promoter and 5’ UTR were amplified using the KAPA HiFi HotStart PCR Kit (Roche, Basel, Switzerland) with two primer sets: promoter: 5’-GTCCTGCCCCTTCACCTT-3’ (forward) and 5’-CAGCGCTGCCTGAAACTC-3’ (reverse); 5’ UTR: 5’-AGCCCCTCCCCTTCCTTT-3’ (forward) and 5’-AGCACCTCGCGGTAGTGG-3’ (reverse). The PCR cycling conditions included an initial denaturation at 95 °C for 3 min, followed by 35 cycles at 98 °C for 20 s, 60 °C for 15 s, and 72 °C for 15 s, with a final extension at 72 °C for 1 min. The PCR products were subsequently purified and subjected to Sanger sequencing. Immunohistochemistry (IHC) was performed on a BOND-MAX automated stainer (Leica Biosystems, Nußloch, Germany) using the following primary antibodies and dilutions: WT1 (clone 6F-H2, Cell Marque, Rocklin, CA, USA; 1:100 dilution), napsin A (clone IP64, Leica Biosystems; 1:200 dilution), and HNF1B (catalog number 12533–1-AP, Proteintech, Rosemont, IL, USA; 1:100 dilution). Heat-induced epitope retrieval was carried out at 100 °C using a citrate-based pH 6.0 buffer (BOND Epitope Retrieval Solution 1, Leica Biosystems) for HNF1B and an EDTA-based pH 9.0 buffer (BOND Epitope Retrieval Solution 2, Leica Biosystems) for the remaining antibodies. All analyses were performed in R (version 4.1.0) with two-sided tests, unless specified otherwise. Categorical variables were compared with the Fisher’s exact test, whereas the Wilcoxon rank-sum test was used to analyze continuous variables. Overall survival was measured from the date of pathological diagnosis of OCCC to the date of the last follow-up or death. The survdiff and coxph functions from the R survival package were employed. The p values were adjusted for multiple testing using the Benjamini–Hochberg method, which controls the false discovery rate. In this analysis, p values < 0.05 and q values < 0.1 were considered statistically significant.

Supplementary Table S2 summarizes the clinical outcomes and genomic characteristics of the 102 patients included in the study. The median age of the participants was 52 years (range: 25 − 79 years). Most patients ( n  = 72, 70.6%) presented with stage I disease at diagnosis. The median follow-up duration was 28.9 months (range: 0.9–69.9 months). Figure  2 and Supplementary Fig. S1 depict the genomic profile of OCCC samples, revealing 5370 somatic SBSs and 580 indels impacting 4264 genes and the splicing junctions of 121 genes. The median non-silent mutations count was 46 (range: 3 − 521). Consistent with previous reports [ 16 – 20 ], ARID1A (66.7%) and PIK3CA (49%) were the most commonly mutated driver genes. Promoter or 5’ UTR mutations of the TERT gene were identified in 26 (25.5%) tumors. Mutations were also frequently observed in KRAS (16.7%), PPP2R1A (15.7%), and TP53 (6.9%). Enrichment analysis revealed that mutations in ARID1A , PIK3CA , KRAS , PPP2R1A , and PIK3R1 were more prevalent in OCCC (Supplementary Fig. S2), whereas TP53 has been reported to be frequently mutated in HGSC [ 8 ]. The TP53 -mutated cases in our cohort were corroborated by immunohistochemical negativity for WT1, napsin A, and HNF1B. One patient (tumor OCCC-067) harbored a pathogenic germline mutation in RAD51C without a concurrent somatic TP53 mutation (Supplementary Table S3). Additionally, a single tumor (OCCC-112) exhibited somatic mutations in both MSH2 and MSH6 , suggesting the rare occurrence of DNA mismatch repair deficiency in OCCC. Fig. 2 Genomic landscape of 102 ovarian clear cell carcinoma samples. The rows represent the following data: counts of non-silent mutations, genome instability index, tumor stage, genome doubling (GD) status, mutational signature (MS) group, membership in somatic copy number alteration (SCNA) cluster, and non-silent somatic mutations in established cancer-driver genes (mutated in five or more tumors), as well as mutations in the TERT upstream region. Supplementary Fig. 1 illustrates mutations in known cancer-driver genes at lower frequencies Genomic landscape of 102 ovarian clear cell carcinoma samples. The rows represent the following data: counts of non-silent mutations, genome instability index, tumor stage, genome doubling (GD) status, mutational signature (MS) group, membership in somatic copy number alteration (SCNA) cluster, and non-silent somatic mutations in established cancer-driver genes (mutated in five or more tumors), as well as mutations in the TERT upstream region. Supplementary Fig. 1 illustrates mutations in known cancer-driver genes at lower frequencies The mutational patterns of ARID1A, ATM, and TP53 , which were typical of tumor suppressor genes, were predominantly characterized by truncating mutations and an absence of hotspots (Fig.  3 A, 3E, and 3F). Of the 91 non-silent ARID1A mutations, 91.2% (83 mutations) were truncating, 6.6% (6 mutations) were missense, and 2.2% (2 mutations) occurred at splice sites. In contrast, PIK3CA , KRAS , and PPP2R1A mutations were characterized by the presence of hotspots and an absence of truncating mutations (Fig.  3 B, C, and D). PIK3CA exhibited hotspot mutations primarily at the H1047 ( n  = 17), E542 ( n  = 9), and E545 ( n  = 8) positions (Fig.  3 B), whereas the most common KRAS mutations were G12D ( n  = 7) and G12V ( n  = 6) (Fig.  3 C). The prevalent PPP2R1A mutation identified in our cohort was R183W (11 out of 16) (Fig.  3 D), a finding in line with previous research [ 42 ]. According to the current classification of clinical actionability in oncology [ 31 ], 41 out of 102 OCCC samples (40.2%) exhibited clinically actionable molecular targets. This encompassed tier 1 actionability in two specific tumors (2%): one exhibiting DNA mismatch repair deficiency (dMMR) and another with a somatic BRCA2 mutation. Additionally, tier 2 actionability was identified in 39 tumors (37.5%), including 33 with PIK3CA mutations, three with PIK3CA and ATM mutations, one with a KRAS G12C mutation, one with a RAD54L mutation, and one with an ATM mutation. Fig. 3 Types and frequencies of mutations in six driver genes: ( A ) ARID1A , ( B ) PIK3CA , ( C ) KRAS , ( D ) PPP2R1A , ( E ) ATM , and ( F ) TP53 gene. Selected mutation types were annotated accordingly. Mutations that fall within the tier 1 or 2 clinical actionability are emphasized using asterisk (*) markings. G Loss of tumor suppressor genes in the OCCC tumors. Biallelic loss is defined as having either of the following conditions: 1) double truncating mutations with or without loss-of-heterozygosity (LOH), 2) single truncating mutation with LOH, or 3) complete copy number loss (copy number of zero). Uniallelic loss is defined as having either of the following conditions: 1) single truncating mutation without LOH, 2) Copy number loss without truncating mutation Types and frequencies of mutations in six driver genes: ( A ) ARID1A , ( B ) PIK3CA , ( C ) KRAS , ( D ) PPP2R1A , ( E ) ATM , and ( F ) TP53 gene. Selected mutation types were annotated accordingly. Mutations that fall within the tier 1 or 2 clinical actionability are emphasized using asterisk (*) markings. G Loss of tumor suppressor genes in the OCCC tumors. Biallelic loss is defined as having either of the following conditions: 1) double truncating mutations with or without loss-of-heterozygosity (LOH), 2) single truncating mutation with LOH, or 3) complete copy number loss (copy number of zero). Uniallelic loss is defined as having either of the following conditions: 1) single truncating mutation without LOH, 2) Copy number loss without truncating mutation A total of 91 non-silent mutations were identified in the ARID1A gene, with 83 being truncating mutations, potentially leading to a loss of function. Subsequently, we investigated the frequency of ARID1A deficiency due to biallelic loss using a previously described assessment method [ 43 ]. Biallelic loss was defined as the presence of either two truncating mutations, a single truncating mutation with loss of heterozygosity (LOH), or total copy number loss. The results revealed that 45.1% (46/102) of tumors exhibited biallelic loss of ARID1A (Fig.  3 G), with 43.5% having two truncating mutations and 56.5% a single truncating mutation with LOH. Figure  4 summarizes the clinical actionability and the corresponding potential therapies for the study cohort. Fig. 4 Clinical actionability of mutations identified in ovarian clear cell carcinoma specimens. Tier 1 alterations are those with direct clinical implications for a specific tumor type, including therapies approved by the FDA. Tier 2 includes alterations for which FDA-approved therapies exist for other tumor types, as well as investigational therapies, consensus findings from meta-analyses, preclinical studies, and case reports Clinical actionability of mutations identified in ovarian clear cell carcinoma specimens. Tier 1 alterations are those with direct clinical implications for a specific tumor type, including therapies approved by the FDA. Tier 2 includes alterations for which FDA-approved therapies exist for other tumor types, as well as investigational therapies, consensus findings from meta-analyses, preclinical studies, and case reports TERT mutations were identified in 25.5% (26/102) of OCCC samples (Fig.  5 and Supplementary Table S4), predominantly located in regulatory regions and notably creating novel ETS binding motifs. The majority of TERT mutations (18 of 26, 69%) were identified in the promoter region [ 44 ], whereas a significant proportion (8 of 26, 31%) was found in the less frequently studied 5’ UTR [ 45 ]. These mutations were collectively referred to as TERT upstream mutations. To ensure unbiased detection of TERT upstream mutations, we examined the sequencing coverage of the TERT upstream region (chr5:1,294,990–1,295,146) across our cohort. The median read depth was 99 (range: 26 − 235) for tumors and 60 (range: 21 − 299) for normal samples, with no significant difference between TERT -mutated and not mutated tumors (median: 99 versus 100, p  = 0.867). Furthermore, Sanger sequencing validated 88.5% (23 out of 26) of these mutations, confirming the reliability of our findings (Supplementary Table S4). In our study, the most frequent TERT mutation was c.-124C > T (C228T), detected 15 times, followed by c.-57A > C (A161C) six times, c.-124C > A (C228A) twice, c.-54C > A (C158A) once, and c.-146C > T (C250T) once (Fig.  5 A). We also discovered a novel complex indel (c.-29_-45 GTCCTGCTGCGCACGTG > A) in the 5’ UTR in one tumor. TERT promoter and 5’ UTR mutations were mutually exclusive. Consistent with a previous report [ 44 ], TERT mutations were more common in ARID1A -wild-type tumors (20 of 34, 58.8%) than in ARID1A -mutated tumors (6 of 68, 8.8%; two-sided Fisher’s exact test, p  = 1.3 × 10 –7 ). Tumors with TERT 5’ UTR mutations were associated with shorter overall survival (median, 26.8 months) compared to those without these mutations (median survival not reached, univariate hazard ratio [HR] = 4.49, 95% confidence interval [CI] = 1.4 to 14.34, p  = 0.011, and multivariate HR = 3.86, 95% CI = 1.19 to 12.55, p  = 0.025; Fig.  5 B, Supplementary Tables S5 and 6). These findings suggest that TERT upstream mutations may serve as a prognostic marker in OCCC, warranting further validation in larger cohorts. Fig. 5 TERT upstream mutations identified in ovarian clear cell carcinoma samples. A Classification and prevalence of TERT upstream mutations. B The presence 5’ UTR mutations was associated with significantly worse survival than lack of an upstream mutation or a mutation in the promoter region, as shown in both a univariate (HR = 4.49 versus wild type, p  = 0.011) and a multivariable Cox proportional hazards analysis (HR = 3.86, versus wild type, p  = 0.025, Supplementary Table S5). UTR, untranslated region; CDS, coding sequence TERT upstream mutations identified in ovarian clear cell carcinoma samples. A Classification and prevalence of TERT upstream mutations. B The presence 5’ UTR mutations was associated with significantly worse survival than lack of an upstream mutation or a mutation in the promoter region, as shown in both a univariate (HR = 4.49 versus wild type, p  = 0.011) and a multivariable Cox proportional hazards analysis (HR = 3.86, versus wild type, p  = 0.025, Supplementary Table S5). UTR, untranslated region; CDS, coding sequence We identified a median of 143 SBSs (range: 21 to 1146) per tumor. Figure  6 A and Supplementary Fig. S3 illustrate the distribution of SBS signatures across the examined OCCC samples. The clock-like signatures SBS1 and SBS5 were present in 94 (92.2%) and 86 (84.3%) of tumors, respectively, with a median of 25 (range: 0 to 308) and 89 (range: 0 to 291) mutations attributed to each. APOBEC-related signatures (SBS2 and SBS13) were identified in 42 (41.2%) of tumors. Notably, four malignancies (OCCC-083, 337, 346, and 128) exhibited high APOBEC mutational activity (median: 319 mutations, range: 238 to 844), without correlation to patient age, stage, driver mutations, or overall survival. Fig. 6 A Single-base substitution (SBS) mutation signature activities across all 102 OCCCs. Mutation counts and the proportions of signatures contributing to the mutational spectrum of each tumor were shown in the top two panels. The bottom panel indicates tumors classified by the following “mutational signature groups”: (1) DNA mismatch repair deficiency (OCCC-112), (2) APOBEC-dominant (OCCC-83, 128, 337, and 346). B Genome-wide somatic copy number alteration (SCNA) patterns observed in the OCCC cohort A Single-base substitution (SBS) mutation signature activities across all 102 OCCCs. Mutation counts and the proportions of signatures contributing to the mutational spectrum of each tumor were shown in the top two panels. The bottom panel indicates tumors classified by the following “mutational signature groups”: (1) DNA mismatch repair deficiency (OCCC-112), (2) APOBEC-dominant (OCCC-83, 128, 337, and 346). B Genome-wide somatic copy number alteration (SCNA) patterns observed in the OCCC cohort SCNAs are crucial for cancer progression and treatment response [ 46 ]. In our study, the median GII was 0.24, ranging from 0.01 to 0.89. A significant majority of OCCC samples (94.1%) showed copy number changes exceeding 1 in at least 5% of their genome. GD, a prevalent ploidy abnormality in cancer, was observed in 36.3% of samples. Although TP53 mutations are linked to GD in other cancers [ 39 ], our findings in OCCC revealed no correlation between GD and driver mutations, including TP53 variants (Supplementary Table S7). The most common SCNA in OCCC was 8q amplification, identified in over 60% of the tumors. Additional common amplifications were observed in 3q, 5p, and 17q, while copy-number losses were identified in 1p, 4q, 5q, 6q, 13q, 15q, and 17p (Fig.  6 B). Approximately 20% of tumors exhibited copy-neutral LOH at 1p, affecting the ARID1A gene. SCNAs profiling divided OCCC tumors into three clusters (Supplementary Fig. S4): Cluster 1 (65.7%) with low-to-moderate instability and a low GD rate (13%); Cluster 2 (22.5%) with a high GD rate (95.7%) and extensive copy number loss; and Cluster 3 (11.8%) with a high GD rate, high genomic instability, and LOH. We observed a non-significant trend indicating potentially better survival outcomes for patients with tumors classified in SCNA Cluster 3 (Supplementary Fig. S5). To investigate the timing of driver mutations in OCCC, we employed a previously described method [ 41 , 47 ] that distinguishes mutations as clonal or subclonal based on variant allele fractions and SCNAs status (Fig.  7 A). This approach further classifies clonal mutations as “early clonal”, “late clonal”, or “untimed clonal” depending on their copy number post-amplification (Fig.  7 A). There were 3,098 early clonal mutations, 1,822 late clonal mutations, 8,120 untimed clonal mutations, and 6,628 subclonal mutations. Genes with a high frequency of early clonal and untimed clonal mutations – such as ARID1A , PIK3CA , TERT , and KRAS – are implicated in tumor initiation (Fig.  7 B). Notably, ARID1A , TERT , and PIK3CA mutations were predominantly clonal compared to non-driver genes (Supplementary Table S8, two-sided Fisher’s exact test with a Benjamini–Hochberg correction, q values of 1.044 × 10 –6 , 0.0015, and 0.0746, respectively). Fig. 7 Driver mutation timing estimates in ovarian clear cell carcinoma specimens. A Based on the timing of somatic mutations relative to somatic copy number change at the same locus, we categorized clonal mutation events as “early clonal” (occurring before the copy-number gain), “untimed clonal” (inability to determine the timing relative to the somatic copy-number gain), and “late clonal” (occurring after the somatic-copy-number gain). B Driver genes were grouped according to the timing and clonality of the identified somatic mutations. Color bars to the right of each gene symbol indicate the proportions of early clonal (dark blue), untimed clonal (light blue), late clonal (orange), and subclonal (dark orange) mutations for each gene. The values displayed to the right of the bar graph indicate the number of clonal mutations and the total mutation count within each specified gene. Genes categorized as “early clonal” are identified as potential key factors in the initiation of OCCC Driver mutation timing estimates in ovarian clear cell carcinoma specimens. A Based on the timing of somatic mutations relative to somatic copy number change at the same locus, we categorized clonal mutation events as “early clonal” (occurring before the copy-number gain), “untimed clonal” (inability to determine the timing relative to the somatic copy-number gain), and “late clonal” (occurring after the somatic-copy-number gain). B Driver genes were grouped according to the timing and clonality of the identified somatic mutations. Color bars to the right of each gene symbol indicate the proportions of early clonal (dark blue), untimed clonal (light blue), late clonal (orange), and subclonal (dark orange) mutations for each gene. The values displayed to the right of the bar graph indicate the number of clonal mutations and the total mutation count within each specified gene. Genes categorized as “early clonal” are identified as potential key factors in the initiation of OCCC

Ovarian cancer, primarily consisting of high-grade serous carcinoma (HGSC) and ovarian clear cell carcinoma (OCCC), is a major cause of cancer-related deaths in women, with an annual global death toll exceeding 200,000 [ 1 , 2 ]. OCCC, which is notably associated with endometriosis, comprises 10 − 25% of ovarian cancers in East Asia compared to 5% in Western countries [ 3 – 5 ]. While OCCC is predominantly diagnosed at early stages (specifically I–II), HGSC is typically identified at advanced stages (namely III–IV). Accordingly, approximately two-thirds of OCCC cases are detected at an early stage, in contrast to only about one-fifth of HGSC cases [ 4 ]. Typically, OCCC has a more favorable prognosis than HGSC in the early stages [ 3 – 5 ]. However, as the disease progresses to advanced stages, the clinical outcomes for OCCC become less favorable compared to those of HGSC [ 6 ]. This is primarily attributed to the scarcity of targeted therapies and frequent chemoresistance [ 7 ]. Extensive research has characterized the somatic mutation landscape in HGSC [ 8 , 9 ], demonstrating its prognostic value [ 10 , 11 ]. Notably, mutations in genes such as BRCA1, BRCA2 , PALB2 , and RAD51C have been specifically associated with HGSC [ 12 – 15 ]. The molecular basis of OCCC, however, is less comprehensively understood. In contrast to HGSC, which is characterized by a high prevalence of TP53 mutations, OCCC is distinguished by frequent mutations in the ARID1A, PIK3CA , KRAS , and PPP2R1A genes [ 16 – 20 ]. Nevertheless, most prior studies in this area have been limited by small sample sizes, leading to an incomplete understanding of the comprehensive genomic landscape of OCCC. Within a precision medicine framework, this study aims to address this knowledge gap by investigating the mutation patterns in driver genes, somatic copy number alterations (SCNAs), mutational signatures, and clonal evolution of OCCC. Our research presents findings from a large cohort of 102 OCCC samples, for which whole-exome sequencing (WES) data were available. The current results have the potential to identify novel molecular targets that could inform the development of innovative treatment strategies.

OCCC is notably resistant to conventional chemotherapy, particularly regimens based on platinum and taxanes, which are standard first-line treatments for ovarian cancer [ 3 – 5 ]. This chemoresistance is attributed to several key factors, including the inherently low proliferation rate of OCCC, which limits the efficacy of chemotherapeutic agents. Moreover, OCCC cells frequently display reduced drug accumulation and enhanced drug detoxification mechanisms, which further compromise the clinical efficacy of chemotherapy. Notwithstanding extensive research endeavors, a tailored chemotherapy regimen specifically designed for OCCC remains elusive. Consequently, elucidating the molecular underpinnings of this malignancy is crucial for identifying novel therapeutic targets and developing more effective treatment strategies. Our comprehensive study of OCCC genomic profile revealed that only 2% (2 out of 102) of samples had tier 1 clinically actionable targets, which are suitable for routine clinical use. These included one case with dMMR and another with a BRCA2 mutation. Notably, BRCA2 mutations can respond favorably to poly ADP-ribose polymerase inhibitors, which have been approved for advanced ovarian cancer [ 48 , 49 ]. Similarly, an immune checkpoint inhibitor targeting the PD-1 pathway is available for advanced dMMR malignancies [ 50 ]. Tier 2 clinical actionability – which refers to investigational molecular targets – was more common, being identified in 38.2% (39/102) of the analyzed samples. This category included mutations in the PIK3CA , KRAS , RAD54L , and ATM genes. PIK3CA mutations are of particular interest due to their potential responsiveness to alpelisib, a selective inhibitor of the PI3Kα protein. This drug is currently approved for PIK3CA -mutated breast cancer [ 51 ], and its efficacy in treating ovarian cancer is currently under investigation [ 52 ]. ERBB2 amplification has also emerged as a promising therapeutic target in OCCC [ 53 ]. To expand the feasibility and applicability of targeted therapy clinical trials for OCCC, it is recommended to include ERBB2 amplification as a criterion for patient enrollment in basket trials designed to target this specific genetic alteration. Similarly, covalent inhibitors like sotorasib and adagrasib have shown promise in treating non-small-cell lung cancer harboring the KRAS G12C mutation [ 54 , 55 ]. Moreover, inhibitors targeting the KRAS G12D mutation have demonstrated encouraging clinical responses [ 56 ]. In addition to tier 1 and 2 actionable mutations, our analysis revealed that biallelic loss of ARID1A [ 57 ] was present in 45% of the examined samples. This observation implies that the preclinical investigation of EZH2 inhibitors, which have shown promise in suppressing the growth of ARID1A -mutated cancers [ 58 ], may be a viable avenue for further exploration. Additionally, mutations in PPP2R1A were detected in 16% of the samples, potentially affecting cell growth through alterations in the Aα subunit of serine/threonine phosphatase 2A (PP2A) [ 42 ]. These mutations may be amenable to targeting with ribonucleotide reductase inhibitors [ 59 ]. Our analysis also confirmed the involvement of the TERT gene in the pathogenesis of OCCC [ 44 ]. with mutations in the TERT promoter and 5’ UTR being identified in 17.6% and 7.8% of samples, respectively. This is, to our knowledge, the first study to report the occurrence of TERT 5’ UTR mutations in OCCC. These mutations could lead to increased TERT expression and telomerase activation. Mutations in the TERT promoter create a de novo binding motif for transcription factors, whereas mutations in the 5’ UTR disrupt the connection between the repressor complex MAX/Mad1 and the E-box sequence (CACGTG) [ 60 ]. Notably, OCCC patients with TERT 5’ UTR mutations exhibited poorer survival outcomes compared to those with TERT promoter mutations or no mutations, suggesting the clinical relevance of this molecular pathway. While directly targeting TERT or its promoter mutations is challenging due to the complex role of telomerase in cancer [ 45 ], these variants may influence responses to certain therapies, such as BRAF and MEK inhibitors [ 60 ]. This highlights the potential of TERT mutations in guiding the development of targeted treatments for OCCC. Our study not only identified molecular changes that could serve as a foundation for preclinical exploration of novel therapeutic strategies, but also provided evidence supporting the hypothesis that OCCC originates from endometriotic cells, in line with the “precursor escape” model [ 61 ]. This framework proposes that endometriotic cells gradually amass genetic alterations, eventually giving rise to OCCC. Our analysis of OCCC samples also revealed early clonal mutations in the KRAS and PIK3CA genes. Similar variants have been previously reported in endometriotic epithelium [ 62 ], suggesting these genetic alterations may be involved in OCCC initiation. The molecular context of the 5’ UTR TERT mutations identified in our study can be inferred from the findings of Huang et al. [ 45 ]. They reported that the c.-57A > C (A161C) mutation generates a “(T/A)TCC” sequence, which is a de novo putative ETS-transcription factor binding motif. Similarly, the c.-54C > A (C158A) mutation creates a CCGGAA/T motif, which is a potential binding site for ETS transcription factors. These mutations were found to correlate with upregulated TERT mRNA expression and increased telomerase activity in adult gliomas. Furthermore, we identified a novel complex indel (c.-29_-45 GTCCTGCTGCGCACGTG > A) in the 5’ UTR of TERT , which contains canonical E-box (CACGTG) elements. Previous research has shown that this E-box element mediates repression of TERT transcription in a renal cell carcinoma cell line derivative (RCC23 + 3) with a transferred copy of normal human chromosome 3 [ 63 ]. Taken together, these findings suggest that the 5’ UTR TERT mutations may enhance the transcriptional activity of the TERT core promoter. Regarding frequently altered genes and their prognostic implications in OCCC, our study identified an association between 5’ UTR  TERT  mutations and reduced overall survival. While mutations upstream of the  TERT  promoter are recognized as significant prognostic markers in OCCC [ 44 ], it is important to note that the median follow-up period for our cohort was relatively short (2.5 years) and the occurrence of these genetic alterations was limited. Notably, other genetic mutations, such as those in  ARID1A  and  PIK3CA , have been associated with unfavorable prognosis in a smaller series of 55 patients [ 64 ]. In a separate study, the loss of ARID1A expression was identified as a negative prognostic factor in 9 out of 60 patients with OCCC who received platinum-based chemotherapy [ 65 ]. However, these findings are preliminary and should be interpreted with caution. To derive more reliable conclusions, further validation is essential, particularly in larger cohorts with extended follow-up periods. Moreover, comprehensive molecular investigations are necessary to fully elucidate the role of frequently altered genes in OCCC. In the current study, we found that SCNA was prevalent and heterogeneous in OCCC. In particular, we identified a group of OCCCs with high degree of genomic instability and better survival outcomes in this SCNA cluster 3 group. Previously, TP53 mutation and homologous recombination deficiency may contribute to the genomic instability and genome doubling [ 39 , 66 ]. However, the exact molecular mechanisms underlies the genomic instability is not fully characterized. Our study demonstrated that extent of SCNA in OCCC may represent a potential prognostic marker, necessitating validation in a larger cohort. In conclusion, the current study offers a thorough analysis of the genomic alterations and clonal progression in OCCC, uncovering numerous mutations that could be targeted by novel pharmacological treatments. The study also supports the hypothesis that OCCC originates from endometriotic epithelium. These insights have the potential to form the foundation for advancements in precision oncology, facilitating the development of more effective therapeutic strategies for patients with OCCC.

Supplementary Material 1. Supplementary Material 1.