1. Introduction
Ovarian cancer (OC) is the most severe gynecologic malignancy with stable incidence and mortality. The most frequent OC types (85–95%) are epithelial tumors, which are high-grade (HG) serous in 70% of cases [
1,
2]. Because of the nonspecific symptoms and a lack of presymptomatic screening modalities, most women are diagnosed with an advanced disease, having a dismal 25% 5-year survival rate [
3].
The overall OC lifetime risk oscillates around 2% in the general female population in developed countries. Central and Eastern Europe, including the Czech Republic, represented a region with the highest OC incidence (11.9 ASRW per 100,000 females) and mortality (6.0 ASRW per 100,000 females) worldwide in 2018 (
http://gco.iarc.fr). In the Czech Republic alone, annual OC incidence and mortality in 2018 reached 9.5 and 6.7 ASRW per 100,000 females, respectively.
Genetic predisposition for OC is unusually high and is reported in up to 25% of cases [
4,
5,
6]. The most frequent germline mutations affect the
BRCA1 and
BRCA2 genes, conferring 24% and 8.4% OC lifetime risks, respectively [
7]. The
BRCA1 and
BRCA2 mutation carriers frequently but not exclusively develop HG serous OC [
8]. Carriers of mutations in these major OC predisposition genes have also very high risk of breast cancer (BC) development. A high OC risk has also been associated with germline mutations in
RAD51C, RAD51D, Lynch syndrome genes, and
STK11; a moderate OC risk with
BRIP1 [
9,
10,
11,
12,
13]. Risks associated with germline mutations in genes with anticipated BC and/or OC predisposition (incl.
ATM, BARD1, CDH1, CHEK2, NBN, PALB2,
PTEN, and
TP53) and in other candidate genes remain to be determined [
14,
15,
16,
17]. The identification of presymptomatic women at high risk who can benefit from risk-reducing salpingo-oophorectomy (RRSO) is of critical importance, as demonstrated by the reduced OC mortality in
BRCA1 and
BRCA2 mutation carriers undergoing preventive surgery [
18].
In this report, we aim to establish an association of germline mutations with OC in the Czech patients belonging to the Slavic population that has not been systematically analyzed for OC predisposition. Seven Czech genetic laboratories participated in the analysis of 1333 Czech OC patients by the identical procedure using CZECANCA panel (CZEch CAncer paNel for Clinical Application) targeting 219 genes [
19]. Prevalence of variants in genes affected in OC patients was assessed in 2278 population-matched controls. This analysis enabled us to comprehensively determine mutations frequency and clinicopathological characteristics of OC in carriers of mutations in genes with known OC predisposition but also to analyze contribution of population-specific variants in other candidate genes to OC predisposition.
3. Discussion
The analysis of 1333 Czech OC patients and 2278 population-matched controls provides the most comprehensive view of the genetic architecture of OC predisposition in the Slavic population. From 18 OC/BC predisposition genes listed in current NCCN breast/ovarian familial cancer guidelines, mutations in 10 genes were significantly associated with OC risk in our population being present in 399/1333 (29.9%) OC patients and 31/2278 (1.4%) PMC (
Figure 1). Mutations in eight remaining genes were extremely rare (
CDH1, PTEN, STK11, and
TP53) or absent (
CDKN2A and
NF1) or did not significantly differ in frequency among cases and controls (
ATM, PALB2, and
CHEK2). Mutations in
BRCA1/2, RAD51C/D, and Lynch syndrome genes were associated with a high OC risk, while mutations in
BRIP1 were associated with a moderate OC risk in our study (
Table 1), in concordance with previous reports [
9,
10,
20,
21]. The
BRCA1 and
BRCA2 mutations, present in 84.0% of all mutation carriers, were by far the most frequent alterations found in 17.9% and 7.4% of our patients, respectively. Mutations in other eight genes leaded by
RAD51C/RAD51D/BRIP1 affected additional 5.0% of patients, as shown also by others recently [
5,
6,
22]. Germline mutations in Lynch syndrome genes together associated with high OC risk. Mutations in
MLH1 prevailed similarly as in Lynch syndrome patients diagnosed with colorectal cancer [
23].
In contrast to previous studies, our results suggest increased OC risk in carriers of
NBN and
BARD1 mutations [
12,
24]. We did not find significant increase of OC risk for carriers of mutations in
ATM and
PALB2, as noticed previously [
12,
24,
25]. However, further analyses considering very large population-matched studies or studies considering families of mutation carriers can better disclose moderate risk associations, as shown for
PALB2 mutations recently [
26].
Overrepresentation of mutations in the
CHEK2 gene in OC patients in this study was marginally nonsignificant in contrast to our previous report where we identified moderately increased OC risk for
CHEK2 mutation carriers [
27]. However, last four
CHEK2 coding exons were not targeted in our gene panel omitting possible deleterious
CHEK2 alterations identified in our previous study in which last four coding exons were analyzed separately in both cases and controls. Mutations in
NF1 were absent and were extremely rare in
CDH1 and
PTEN, just like
STK11 mutations found in a patient with nonepithelial OC, a characteristic Peutz–Jeghers syndrome manifestation [
9]. Altogether, the high overall frequency of mutations in OC predisposition genes in our study is in agreement with some previous studies [
4,
5,
6,
28] and may contribute to a high OC incidence in our population.
Multigene testing revealed 13 carriers of multiple pathogenic mutations (1.0% of patients). Similar frequency of individuals with this multilocus inherited neoplasia alleles syndrome (MINAS) [
29] was shown also in previous analyses of OC patients [
30,
31].
We analyzed available phenotype characteristics in 1320 OC patients with one pathogenic mutation at the most in 10 genes associated with OC risk in our study (
Figure 2). While the highest prevalence of
BRCA1/2 mutation carriers was in patients diagnosed with double primary OC and BC, mutations in
RAD51C/RAD51D/BRIP1 prevailed in patients diagnosed with OC only (
Figure 2B); nevertheless, their distribution among histological subtypes was similar to that in
BRCA1/2 mutation carriers (
Figure 2E). In contrast to Castera et al. who found mutations in
RAD51C/RAD51D/BRIP1 dominantly in French OC patients with a positive family OC history [
32], we identified mutations in these genes in 1/116 (0.9%) and 22/587 (3.7%) carriers in HOC patients and in patients with a negative family cancer history, respectively. Further, we have noticed a surprisingly high frequency of OC-predisposing mutations in older patients. Their prevalence in patients ≥ 60 years was 23.6%, whereas Harter et al. found in this age group 18.9% mutation carriers even though frequency of mutation carriers in patients <60 years in both studies was comparable (32.6% and 33.2%, respectively) [
28].
BRCA1 mutations dominated in patients <60 years over
BRCA2 mutations, while in patients ≥ 60 years, their frequencies were comparable. Moreover, we revealed 29
BRCA1/2 mutation carriers (13.9% of patients) in 208 OC patients diagnosed at ≥60 years with no family cancer history, while Morgan and colleagues detected only two (4.3%)
BRCA1/2 mutations in 46 sporadic OC patients ≥ 60 years [
33]. Even in the oldest subgroup of our OC patients diagnosed at ≥70 years, the frequency of
BRCA1/2 mutation carriers exceeded 18%, while in other studies,
BRCA1/2 mutations’ frequency in this age category was below 10% [
34,
35]. This high frequency of
BRCA1/2 mutations in our patients ≥70 years contrasted with a low frequency in women diagnosed at <30 years (18.1% vs. 3.6%;
p = 0.003;
Figure 2A). The difference was even more apparent in “sporadic” OC cases (with no family cancer history), where
BRCA1/2 mutations were found in 6 out of 45 (13.3%) women ≥70 years but in none of 52 cases diagnosed at <30 years. It should be emphasized that although rare histological OC types were more frequent in the subgroup of 52 patients diagnosed with sporadic OC at <30 years, 32 (65.3%) of 49 informative cases developed invasive epithelial OC.
Mutations in OC predisposition genes significantly prevailed in subgroups with high-grade/ nonspecified serous, borderline, and endometrioid tumors over subgroup with low-grade serous, mucinous, clear cell, or other rare histologic types (
Figure 2E). Surprisingly, the overall mutation frequency in patients with borderline tumors was comparable with that of in HG serous OC (32.2% and 36.7%, respectively;
Figure 2E). Thus, we compared mutation frequency in patients with no family cancer history diagnosed with these histological tumor types, and we found that although the mutation frequency in sporadic borderline tumors was half in comparison to sporadic HG serous (
Figure 3), it still largely exceeded 10% in both hereditary and sporadic cases, justifying the genetic testing of borderline tumors. The large proportion of borderline tumors with positive family cancer history in our study suggested that this OC subtypes belong to a possible manifestation of a cancer predisposition. However, our observation needs to be confirmed in other populations as current reports about borderline tumors in
BRCA1/2 mutation carriers are limited.
The multigene panel enabled us to identify other candidate genes associating with increased OC risk. We noticed many rare truncating variants episodically affecting various genes and clustering into
PPM1D,
NAT1, and
SHPRH in OC patients. The
PPM1D gene, coding for WIP1 phosphatase, was the only candidate associated with OC risk following multiple testing correction. Similarly to the previous studies describing its mosaic variants in OC patients [
36,
37,
38], we also found mosaic gain-of-function mutations resulting in increased WIP1 phosphatase activity [
38]. All
PPM1D mutations in our patients were identified in postchemotherapy treatment blood samples suggesting their somatic origin [
39]. Germline mutations in
NAT1 have not been analyzed for OC predisposition so far. However, several polymorphisms in
NAT1 (coding for arylamine N-acetyltransferase 1 engaged in carcinogen metabolism and detoxification) were shown to modify the risk of various cancers [
40,
41]. The
SHPRH gene codes for E3 ubiquitin-protein ligase targeting PCNA upon DNA damage [
42]. Contribution of
SHPRH germline variants to OC risk remains elusive. Overall, low mutation frequencies found in gene candidates in our study precluded its precise OC risk estimations and will require large, multiethnic, case-control studies, segregation analyses in affected families, and functional analyses. Alongside variants clustering to a few candidate genes, we identified rare mutations in a gene family coding for Fanconi anemia (FA) proteins involved in the repair of DNA interstrand crosslinks [
43]. Several FA genes belong to established OC predisposition genes, including
BRCA1 (
FANCS),
BRCA2 (FANCD1),
RAD51C (FANCO),
PALB2 (FANCN), and
BRIP1 (FANCJ). Except these, we found rare mutations in other FA genes (
FANCG,
FANCD2, and
FANCA) in 11 (0.83%) of 1333 OC patients compared to 5 in 2278 PMC (0.2%), with cumulative OR = 3.78 (95% CI 1.21–13.91;
p = 0.02). Interestingly, these rare mutations were detected almost exclusively in patients without mutations in other OC predisposition genes.
The strengths of this study include an identical NGS analysis and bioinformatics pipeline in all patients, a careful curation of clinical data, and an ethnically homogeneous set of patients and controls representing the largest sample set from the region of Central and Eastern Europe. Despite that, the number of individuals still did not allow the precise OC risk calculations in rarely mutated genes. Although all OC cases in the Czech Republic are eligible for genetic testing, OC patients with positive family cancer history and earlier-onset individuals were enriched in our study, especially in a small subgroup enrolled before 2015 (in the Center A only).
Whether the high prevalence of clinically important germline mutations in OC patients justifies population-wide screening is a vivid matter of debate [
44,
45,
46,
47,
48]. We emphasize that we found
BRCA1/2 mutations in 14.5% of OC patients with no family cancer history who would currently not be revealed presymptomatically without population screening. We assume that careful application of germline testing in all OC patients and their relatives would reduce OC burden in our population. Moreover, the mutations in
BRCA1/2 [
49,
50] and other OC predisposition genes [
51,
52] represent valuable predictive biomarkers improving OC chemotherapy.
4. Materials and Methods
Analyzed patients (N = 1333) were enrolled in 2010–2018 and included all OC cases regardless of familial cancer history or OC histology subtypes. As knowledge about germline mutations’ frequency in women diagnosed with BTO is limited, we included these histological subtypes to our study. Clinicopathological data were obtained during genetic counselling or retrieved from the patients’ records. OC patients with a positive cancer family history were stratified into (i) hereditary ovarian cancer (HOC) families with OC and other nonbreast cancer (BC) in the family history; (ii) hereditary breast and/or ovarian cancer (HBOC) families with BC and OC or other cancer in the family history, and (iii) multiple cancer families with non-OC and non-BC in the family history. Index patients were tested in seven centers: (A) First Faculty of Medicine, Charles University, Prague (N = 637); (B) Masaryk Memorial Cancer Institute, Brno (N = 357); (C) Gennet, Prague (N = 273); (D) AGEL Laboratories, Novy Jicin (N = 34); (E) GHC Genetics (N = 12); (F) Pronatal (N = 11), and (G) University Hospital Olomouc (N = 9).
Population-matched controls (PMC;
N = 2278) included 616 noncancer controls collected in centers A (
N = 344), B (
N = 150), and D (
N = 122), and 1662 unselected controls provided by the National Center for Medical Genomics (
http://ncmg.cz). The noncancer controls were volunteers (78 males and 538 females) aged ≥ 60 years without a personal or family cancer history (in first-degree relatives). The unselected controls (1170 males and 492 females; median age 57 years, range 18–88 years) were unrelated individuals analyzed by whole-exome sequencing (WES) for various noncancer conditions.
All patients and controls were Caucasians of a Czech origin. Written informed consent was obtained from all patients and controls. The study was approved by the Ethics Committee of the General University Hospital in Prague; ethics approval number was 92/14. The study was performed in accordance with the Declaration of Helsinki.
4.1. Next-Generation Sequencing
Germline blood-derived DNA was analyzed by the CZECANCA (CZEch CAncer paNel for Clinical Application; custom-made SeqCap EZ choice panel; Roche) panel NGS targeting 219 genes on MiSeq (Illumina), as described in details previously [
19]. Sequencing reads were aligned by Novoalign v2.08.03 to the human reference genome (hg19). Variants were identified using GATK and Pindel, CNVs using CNV score [
19]. The entire diagnostic pipeline was successfully tested using European Molecular Genetics Quality Network schemes (EMQN) and validated as we have described recently [
19].
4.2. Variant Classification
We first analyzed 18 genes considered clinically relevant to the HBOC syndrome (MIM #604370) by NCCN, namely,
ATM, BARD1, BRCA1, BRCA2, BRIP1, CDH1, CHEK2, MLH1, MSH2, MSH6, NBN, PALB2, PTEN, RAD50, RAD51C, RAD51D, STK11, and
TP53. Germline variants (with frequency ≤ 0.01 and ≤0.05 in 1000 Genomes project and noncancer PMC, respectively) were classified into three groups: i) pathogenic/likely pathogenic, ii) variants of unknown significance (VUS), and iii) likely benign/benign, based on recommendations from the ENIGMA consortium (
https://enigmaconsortium.org). All nonsense/frameshift/splicing (± 1–2 bp) mutations/CNVs were considered pathogenic/likely pathogenic unless classified as other in the ClinVar database; whole gene duplications were considered VUS. The other types of mutations were considered pathogenic/likely pathogenic only if classified as such in ClinVar by at least two submitters. TP53 variants were classified using the IARC TP53 database (
http://p53.iarc.fr/),
CHEK2 VUS using a recently published functional assay [
27].
Subsequently, we analyzed variants in another 201 genes targeted by the CZECANCA panel. Nonsense/frameshift/splicing (± 1–2 bp) mutations/CNVs (except whole gene duplications) with frequency ≤0.01 and ≤0.05 in 1000 Genomes project and in noncancer PMC, respectively, were considered pathogenic.
All pathogenic/likely pathogenic mutations in patients and noncancer PMC were confirmed by Sanger sequencing and CNVs by MLPA (if available) or by qPCR (protocol available on request), and they were submitted to ClinVar under the submission ID SUB5822876.
4.3. Statistical Analysis
The odds ratio (OR) for particular gene was calculated using Fisher’s exact test, and p values <0.05 were considered significant. The multiple mutation carriers were excluded from the OR calculations. For the identification of other OC candidate genes, the Bonferroni correction was employed. The associations between mutation status and clinicopathological characteristics were estimated using Fisher’s exact test, and p values <0.05 were considered significant.