Background
High-grade serous ovarian cancer (HGSC) is the most common ovarian cancer histologic subtype with the majority of patients diagnosed at a late stage [
1]. The standard treatment for HGSC involves both surgical cytoreduction and systemic paclitaxel and platinum chemotherapy with increasing use of targeted treatments of bevacizumab and/or poly ADP-ribose polymerase inhibitors (PARP-i) [
2‐
4]. Despite aggressive frontline therapy, many patients with ovarian cancer eventually relapse, resulting in only 20–30% survival after five years [
5]. These dismal statistics underscore the need for earlier detection and new treatments following diagnosis.
HGSC originates in the fallopian tube [
6,
7] and is characterized by early acquisition of
TP53 variants [
8,
9]. Other characteristics are extensive somatic copy number alterations (SCNA) [
9,
10] with few somatic point variants [
11] and inactivation of DNA homologous recombination repair (HRR) primarily through germline and somatic variants of
BRCA1 and
BRCA2 [
9]. Although less frequent, variants in additional Fanconi anemia pathway genes that play key roles in HRR (e.g.,
BRIP1,
PALB2, RAD51C/D), and
CHEK2, a DNA damage response gene, have been associated with HGSC [
12‐
15]. Amplification of regions containing the oncogenes
CCNE1, MECOM, and
MYC and deletions of regions containing the tumor suppressor genes
RB1,
NF1, and
PTEN are commonly observed [
9,
16,
17]. An integrated analysis combining variant data, copy number changes, and changes in gene expression determined that the main pathways altered in HGSC are the HRR, RB1, PI3K/AKT/mTOR, and NOTCH signaling pathways [
9].
In this study, we identified and characterized germline and somatic genetic alterations in HGSC and their associations with relapse-free and overall survival. From paired germline and tumor DNA samples from 71 HGSC patients treated at City of Hope (COH), we conducted targeted sequencing of 557 genes in pathways involved in response to DNA damage, DNA repair, cell-cycle regulation, programmed cell death, MAPK, and PI3K/AKT/mTOR signaling including known drivers in HGSC tumorigenesis [
18]. In addition, we used the OncoScan assay to determine genome-wide SCNAs and loss of heterozygosity (LOH) in 61 primary ovarian tumors.
Discussion
All germline and somatic genetic variants identified in the 557 sequenced genes are provided in Supplemental Table
10. Consistent with previous studies of HGSC, we observed germline LOF variants in high- and moderate-penetrance HRR pathway genes [
9,
12,
15,
19,
20]. In addition to the germline LOF variants in the HRR genes,
BRCA1,
BRCA2,
MRE11A, and
PALB2, LOF somatic variants in
BRCA1 and
BRCA2 were identified including homozygous deletions in
BRCA1 and
BRCA2. From the combined sequencing and SCNA analysis, we found that approximately one third of the HGSC tumors harbored LOF variants in HRR genes and treatment with PARP inhibitors would be warranted [
3].
We also observed LOF variants in Fanconi anemia pathway genes nominally associated with HGSC [
9,
19]. These variants included germline LOF variants in
FAN1,
FANCD2,
FANCI, and
FANCL. Platinum-based therapies cause interstrand cross links (ICLs) and the Fanconi anemia pathway is critical for resolving ICLs for DNA repair through the HRR pathway [
21]. Therefore, variants in Fanconi anemia genes may affect the efficacy of treatment and resultant survival. Interestingly, we observed a somatic nonsense variant in
BLM/RECQL3 which codes for a RecQ family DNA helicase, whose deficiency results in Bloom’s syndrome which is characterized by a predisposition to early development of multiple forms of cancer [
22].
BLM has been shown to play a critical role in mitosis and multiple steps of HRR-dependent DNA repair [
23]. While several germline LOF variants in
BLM have been observed in HGSC tumors [
19], only one somatic LOF variant was listed in the COSMIC (Catalogue of Somatic Mutations in Cancer) database [
24].
Additional DNA repair pathway variants included germline LOF variants in double-strand break repair (
EME2,
POLQ,
RECQL5, and
SPO11), nucleotide excision repair (
ERCC2 and
POLK), base excision repair (
NEIL3 and
POLG), mismatch repair (
EXO1), and post-replication repair (
RAD18) pathways. Although uncommon, all these genes have documented germline protein truncating variants from other HGSC studies [
19,
25,
26]. Of note, the identical germline splice site variant in
EXO1 and the frameshift deletions in
ERCC2,
POLK, and
POLQ were reported in previous HGSC studies. We also observed somatic LOF variants in other DNA repair pathway genes with one each in
PMS1 (mismatch repair)
, POLK (nucleotide excision repair),
POLL (base excision repair), and
RFC4 (mismatch repair). A protein-truncating variant in
RFC4 was listed in COSMIC HGSC tumors whereas only missense variants were listed in
PMS1,
POLL, and
POLK [
24].
Alteration of the PI3K/AKT/mTOR pathway is a common occurrence in HGSC [
9] and our results support this finding. We observed germline LOF variants in PI3K/AKT/mTOR pathway genes
FKBP11,
FKBP7, and
IFNA5; in previous HGCS genomic studies, germline LOF variants were reported in
IFNA5 and
FKBP7, but not
FKBP11 [
19,
25]. Somatic variants in PI3K/AKT/mTOR pathways genes have been reported in HGSCs [
9,
19,
27] and we identified LOF variants in
CREBBP and
PIK3R1. Our OncoScan analysis identified three somatic homozygous deletions in
PTEN and two in
CREBBP. In addition, we observed one somatic missense variant in
PIK3CA, one in
CREBBP, and one in
PTEN, classified as pathogenic, reinforcing the importance of alteration in the PI3K/AKT/mTOR pathway in HGSC.
We observed variants in several genes involved in cell cycle regulation. As expected, we found a high frequency of somatic
TP53 variants [
8,
9]
. We also observed a germline LOF variant and a potentially pathogenic germline missense variant in the known ovarian cancer gene
CHEK2 [
15,
19]. Our OncoScan analysis identified three somatic homozygous deletions in
RB1 consistent with a previous report that it was a recurrently mutated gene in HGSC [
9]
.
We observed somatic homozygous deletions in the MAPK pathway genes
MAP2K4 and
NF1. The MAPK signaling pathway regulates many cellular processes, including gene expression, cell cycle, cell survival, cell death, and cell movement [
28].
MAP2K4 is listed in COSMIC as potentially having a role in both tumor suppression and as an oncogene [
24]. Supporting its role as a tumor suppressor, homozygous focal deletions in
MAP2K4 have been identified in ovarian and breast cancers and it is suggested that loss of
MAP2K4 could alter function of the JNK (c-Jun N-terminal kinase) pathway [
29,
30]. In addition, somatic copy number loss of
NF1 is common in HGSC [
31,
32]. The NF1 protein has been found to play a role in regulating several intracellular processes and importantly is essential for reducing oncogenic Ras activity, supporting the hypothesis that it acts as a tumor suppressor [
33].
For patients where we had multiple tissues from the primary debulking or tissue from multiple surgeries, in the secondary tumor sites, the somatic point mutations were maintained and there were few newly acquired somatic point mutations suggesting that tumor evolution was not through acquisition of somatic mutations in genes in the pathways we studied. This is consistent with prior reports that gene amplifications rather than somatic point mutations are observed in disease progression [
34,
35]. We were not able to perform OncoScan analysis for multiple tissues from the same patient.
Along with frequent
TP53 variants, a hallmark of HGSC is profound genomic instability [
9]. We identified several amplified regions containing driver oncogenes previously observed in HGSC including
CCNE1,
MECOM,
MYC, and
MYCL1 and deletion regions containing tumor suppressor genes
RB1 and
PTEN [
9,
16,
17,
34,
36‐
38]. GISTIC identified several regions that, although CNAs, are unlikely to be somatic as the genomic regions they encompass include germline copy number variation (Supplemental Table
5). We only analyzed tumor tissue and so could not distinguish germline from somatic CNAs. These regions include the amplifications regions 14q32.33, 2p11.2, and 2p11.2 and the deletion regions 8p11.22, 22q11.23, and 4q34.3. Regardless of CNA source, they may affect outcome similar to LOF variants regardless of whether germline or somatic.
Since cancer-driver SCNAs tend to be shorter in length and higher in amplitude than passenger SCNAs [
39], we sought to identify genes within high-amplitude SCNAs that were associated with relapse-free and overall survival. We determined that amplifications in
NOTCH3,
ZNF536, and
PIK3R2 were significantly (
P < 0.05) associated with shorter relapse-free survival in both our data and TCGA data after adjusting for copy number mutation rate, age at diagnosis, and tumor stage. Debulking status (optimal versus suboptimal) and intra-peritoneal chemotherapy status (yes versus no) were not available in TCGA data. When we included these two variables in our data, the association remained significant for
NOTCH3 (
P = 0.033) and
ZNF536 (
P = 0.012), but not for
PIK3R2 (
P = 0.127). Our finding of the association of
NOTCH3 high copy number gain and tumor recurrence fits with many lines of evidence supporting its role as a potent oncogene [
9]. The amplification of the region encompassing
NOTCH3 (19p13.12) and upregulation of
NOTCH3 expression has been detected in a high percentage of HGSCs [
40]. The upregulation of the NOTCH3 pathway is associated with tumor progression, drug resistance, and HGSC tumor recurrence [
9,
41,
42] suggesting that NOTCH3 inhibitors could be an effective treatment in order to increase sensitivity to platinum-based therapies. There also is growing evidence that
PIK3R2, encoding the p85β regulatory subunit of PI3K, is an oncogene. PIK3R2 induces oncogenic signaling in HGSC [
43] and high expression is correlated with a significant reduction in overall survival in ovarian cancer patients [
44].
The significance of the high copy gain in
ZNF536 is more difficult to explain as it has not been associated with cancer. However, it may be due to its location near (within 500 kbp) of
CCNE1 at 19q12. The most significantly amplified region from our GISTIC analysis was 19q12. Focal amplification of
CCNE1 is a hallmark of HGSC and is associated with chemoresistance [
9,
16,
17,
32]. Interestingly, our Firth’s Cox regression analysis of high-amplitude genes significantly associated with relapse-free survival showed that
ZNF536 was highly significant (
p = 0.014, Table
2) while
CCNE1 was not significant (
p = 0.22, Table S
6). High-level amplification was more frequent in
CCNE1 (11 of 61 patients) than in
ZNF536 (6 of 61 patients), indicating that
CCNE1 amplification may be an earlier amplification event more related to occurrence of HGSC, rather than outcome after therapy.
Knijnenburg and colleagues calculated homologous recombination deficiency (HRD) scores for 33 cancer types from TCGA and found that HGSC had the highest HRD score and that a higher score within HGSCs was associated with better survival [
45] likely due to better response to platinum-based chemotherapies. Although we could not calculate HRD scores, one third of our HGSC patients carried germline or somatic LOF variants in core HRR genes. Compared to patients not carrying these LOF variants, patients with these variants showed significantly better relapse-free survival (
p < 0.05; Supplemental Fig.
2). In both our tumor samples and TCGA tumor samples, LOF variants in HRR genes were associated with increased incidence of high-amplitude SCNAs at multiple chromosome regions (Supplemental Table
6). This observation suggests that LOF variants in HRR genes caused HRD resulting in greater genomic instability as manifest by increased SCNAs.
HGSC tumors are known for the ability to acquire resistance to the killing effects of various chemotherapeutic agents such as platinum and PARP inhibitors. Platinum resistance can develop from multiple mechanisms including reduced intracellular drug accumulation, intracellular inactivation of the agent, increased DNA repair, or impaired apoptotic signaling pathways [
46]. The near ubiquitous somatic mutation of
TP53 is likely the key mechanism in which HGSC tumors evade the triggering of apoptosis.
A strength of this study is the single institution experience with detailed treatment and long-term follow-up data on recurrence and survival. Other strengths are the combined data on germline and tumor DNA sequencing, the OncoScan assay for copy number and large rearrangements, and the availability of multiple tissue sites and/or recurrence debulking tissue for a subset of patients. The primary weakness of the study is the small sample size and that not all samples had the SCNA analysis. Because of limited power, we only ran analyses of LOF variants in BRCA1, BRCA2, and PALB2 with relapse-free and overall survival, as well as genes in amplified regions from tumor analysis. A second weakness is that because we used a targeted gene approach, we did not have the data to generate an HRD score. Lastly, the impact of the genomic alterations on the overall prognosis (relapse-free survival and overall survival) did not account for the use of bevacizumab or PARP inhibitors.