Role of common genetic variants in ovarian cancer susceptibility and outcome: progress to date from the ovarian cancer association consortium (OCAC)


Simon A. Gayther, Department of Preventive Medicine, Keck School of Medicine, University of Southern California, 1450 Biggy St., Rm 2517G, LG591 MC9601, Los Angeles, CA 90033, USA.
(fax: 323 442 7787; e-mail:


Abstract.  Bolton KL, Ganda C, Berchuck A, Pharaoh PDP, Gayther SA (National Cancer Institute, Rockville, MD, USA; EGA Institute for Women’s Health, University College London, London, UK; Duke University Medical Center, Durham, NC, USA; University of Cambridge, Cambridge, UK; and University of Southern California, Los Angeles, CA, USA). Role of common genetic variants in ovarian cancer susceptibility and outcome: progress to date from the ovarian cancer association consortium (OCAC) (Review). J Intern Med 2012; 271: 366–378.

In this article, we review the current knowledge of the inherited genetics of epithelial ovarian cancer (EOC) susceptibility and clinical outcome. We focus on recent developments in identifying low-penetrance susceptibility genes and the role of the ovarian cancer association consortium (OCAC) in these discoveries. The OCAC was established to facilitate large-scale replication analyses for reported genetic associations for EOC. Since its inception, the OCAC has conducted both candidate gene and genome-wide association studies (GWAS); the latter has identified six established loci for EOC susceptibility, most of which showed stronger association with the serous histological subtype. Future GWAS and sequencing studies are likely to result in the discovery of additional susceptibility loci and may result in established associations with clinical outcome. Additional rare and uncommon ovarian cancer loci will likely be uncovered from high-throughput next-generation sequencing studies. Applying these novel findings to establish improved preventative and clinical intervention strategies will be one of the major challenges of future work.

Epithelial ovarian cancer (EOC) is the sixth most common cause of death because of cancer in women and is the leading cause of death from gynaecological malignancy in the developed world [1]. The high mortality rate associated with EOC is largely attributable to the nonspecific nature of symptoms in early-stage disease and the lack of effective screening tools. As a result, most women are diagnosed with advanced-stage disease when 5-year survival rates are <50%. In contrast, diagnosis of early-stage disease confined to the ovaries is associated with a markedly better 5-year survival rate of about 90%. Thus, the development of effective screening and early detection methods could substantially reduce disease mortality.

Most ovarian cancers are epithelial in origin, referred to as EOC, and will be the focus of this review. There are four major histological subtypes of EOC – serous, mucinous, endometrioid and clear cell. An increasing amount of evidence suggests that ovarian cancer is comprised of multiple disease entities with distinct aetiologies. Histological subtypes vary with respect to clinicopathologic features, biological behaviour, and response to chemotherapy and show distinct underlying molecular changes.

Multiple reproductive, demographic and lifestyle factors are known to influence a woman’s risk of developing EOC, but the strongest known risk factor is a positive family history; having a single affected first-degree relative is associated with a two- to three-fold increased risk of disease [2]. Whilst familial aggregation can be explained by both genetic and shared environmental factors, twin studies have reported an increased disease concordance in monozygotic twins compared with dizygotic twins, suggesting that genetic factors play a more substantial role [3]. From these studies, it has been estimated that ∼22% of the risk of ovarian cancer is attributable to heritable factors. Germline, or inherited, genetics may play a role over the entire course of ovarian cancer, from its inception to the response of patients to chemotherapy (Fig. 1).

Figure 1.

Categories of genes influencing ovarian cancer susceptibility and outcome. Genes related to the initiation of carcinogenesis could impact genomic instability, DNA repair and cellular destruction pathways or could influence the hosts response to environmental and lifestyle risk factors. These processes would continue to be relevant for the maintenance of the tumour. Genes related to invasion could impact the development of local invasion and more distant metastasis. These genes could also influence the host tolerance of the metastasis. Genes influencing response to therapy could impact the extent of tumour toxicity, the bioavailability of the therapy at the cancer site(s) and the type or magnitude of side effects. Genes related to both carcinogenesis and invasion could be related to the risk of invasive ovarian cancer development. Genes from all three cases could impact the likelihood of survival after diagnosis of ovarian cancer.

Efforts to uncover the genetic factors that underlie EOC have identified a few high-penetrance susceptibility genes. Families containing several cases of ovarian cancer, or breast and ovarian cancer occurring together, are most often associated with mutations in the BRCA1 and BRCA2 genes. These genes are the strongest known genetic risk factors for EOC; mutation carriers are estimated to have a cumulative lifetime risk of ovarian cancer of 40–50% for BRCA1 and 20–30% for BRCA2 [4, 5]. Other known high-penetrance genes include the DNA mismatch repair (MMR) genes that predispose to Lynch syndrome, or hereditary nonpolyposis colorectal cancer. Inherited MMR gene mutations are estimated to confer cumulative lifetime risks of ovarian cancer ranging from 6.7% to 12% by age 70 [6–8]. Recently, Meindl et al. [9] described highly penetrant mutations in RAD51C in hereditary breast–ovarian cancer families not accounted for by BRCA1 or BRCA2. However, deleterious mutations in these genes are rare in most populations (e.g. BRCA mutations occur in <0.5% of individuals [10]), and the known high-penetrance susceptibility genes explain <40% of the excess familial risk of ovarian cancer [11]. This suggests that high-penetrance genes are likely to account for a small fraction of all cases and that other EOC susceptibility genes exist. The characteristics of susceptibility alleles that account for the unexplained familial risk are not known. High-penetrance alleles similar to BRCA1 or BRCA2 are unlikely to exist as almost all multi-case, multi-generation ovarian cancer families are explained by these genes. A polygenic disease model, with multiple common genetic variants (with minor allele frequencies of >5%), uncommon variants (1–5% frequency) and rare variants (≤1% frequency), each conferring low-moderate variations in disease risk, may account for the missing heritability for EOC [12].

Germline genetic variation may also influence clinical features of disease development and outcome. Metastasis and its complications are the usual cause of EOC-related deaths. Thus, factors related to tumour aggressiveness, response to therapy and the underlying health state of the patient can be used to predict survival. Factors associated with poor survival in patients with ovarian cancer include advanced-stage, high-grade serous disease, large residual tumour volume after surgery and advanced age at diagnosis [13]. Overall survival (OS) is the traditional end-point for studies of ovarian cancer prognosis. Progression-free survival (PFS) (i.e. the length of time during and after treatment in which a patient’s disease does not worsen) is sometimes preferred because it is less influenced by variation in secondary therapeutic regimes [14]. Progression is commonly defined by a rise in CA125 levels although evidence of tumour growth from imaging studies or physical examination is also used [15]. A tumour’s response to therapy, most often defined by RECIST criteria [16], is commonly used to establish biological activity in phase 1 and 2 trials and is a direct measure of cytotoxicity.

There is evidence that the genetic background of the patient may influence the propensity of a tumour to metastasize, and a patient’s response to treatment. For example, germline genotype influences metastatic potential in mouse models of breast cancer [17, 18]. Patterns of gene expression in human breast carcinomas are also associated with survival [19]. These molecular signatures are present early in tumorgenesis before clinically apparent metastasis has occurred, suggesting that tumour aggressiveness may be determined at the early or initiating stages of tumour development [20]. Several studies have indicated germline genetic variants influence response to therapy [21–29] and the likelihood of adverse drug reactions [30, 31]. For EOC, rare germline mutations in the BRCA1/2 genes appear to improve OS and PFS [32, 33], which may be due to an improved response to platinum-based therapy [34, 35].

Candidate gene association studies for ovarian cancer

The most powerful approach to identify common genetic variants conferring low-moderately elevated risks of complex multifactorial disease traits is the genetic association study. The aim is to search for genetic associations between a disease phenotype and genetic variants, by comparing the allele frequencies of variants between affected (case) and unaffected (control) individuals. The most common type of DNA variant is the single nucleotide polymorphism (SNP), of which there are in excess of 10 million scattered throughout the genome [36]. The development of very high-throughput technologies over the last decade that enable rapid genotyping of several thousands of SNPs in large sample sizes has revolutionized genetic studies that aim to identify novel susceptibility alleles associated with disease. Initially, these studies focused on genetic variants within candidate genes selected on the basis of a known or predicted function in disease development. This included nonsynonymous SNPs in coding regions with suspected function, and noncoding SNPs in coding and noncoding regions that may represent markers in linkage disequilibrium (LD) with the true causal variant.

The past decade has seen a wealth of published literature describing candidate gene studies aimed at identifying common risk variants of modest effect associated with EOC risk. Together, these studies have examined several thousands of SNPs in hundreds of genes. Candidate genes have generally been selected from biological pathways that could plausibly lead to ovarian carcinogenesis. Much of the focus has been on genes belonging to the steroid hormone (e.g. PGR, ESR1, CYP3A4, CYP19A1, SRD5A2), DNA repair (e.g. BRCA1, BRCA2, BRIP1, ERCC1, XRCC2, RAD51, MMR genes) and cell cycle control pathways (e.g. ABL1, CCND1, CDK genes) and known oncogenes (e.g. BRAF, KRAS, NMI) and tumour suppressors (e.g. TP53, RB1). The results from these studies have often been conflicting with very few, if any, robust genetic associations emerging. Because of the low probability that a given SNP is associated with disease, stringent P-value cut-offs for significance are required to reduce the likelihood of false-positive results [37]. There is debate regarding the specific level of significance, but P < 0.0001 has been previously suggested as an appropriate threshold for candidate gene studies [38]. Initial nominally significant associations (P < 0.05) in studies of a few hundred cases and controls have usually failed to replicate in larger studies consisting of thousands of subjects.

One example is the progesterone receptor gene (PGR). Originally, a 306-bp. Alu insertion in intron 7, termed PROGINS, was reported to be associated with an increased risk of EOC [39]. Later, SNPs in exons 4 (V660L) and 5 (H770H) were identified and shown to be perfectly correlated with the Alu insertion. Numerous subsequent studies focused on the PROGINS complex (or a subset of its markers) or an additional putative functional SNP (+331C/T) in the promoter region, but results were inconclusive [40–50]. For example, Pearce et al. [51] genotyped 17 PGR SNPs in 267 cases and 397 controls and found significantly increased risks associated with the PROGINS variant along with a SNP 3′ of PGR.

Similarly inconsistent results have been observed for genes of the DNA repair pathway; although early studies reported significant risk associations for SNPs in some genes (e.g. BRCA2 N372H [52]; XRCC2 R188H [53]), subsequent studies failed to confirm the initial findings [54, 55]. Most of the original studies reporting risk associations were conducted using small sample sizes and had limited power to detect associations at the modest effect sizes expected (i.e. odds ratios (ORs) <1.2). It is therefore likely that many of the ovarian cancer risk associations reported in the literature are chance findings. Further genotyping of putative candidate SNPs in many thousands of additional cases and controls will be required to distinguish the few true associations from the many false-positive associations.

There are many published reports describing the impact of common genetic variation in candidate genes on EOC prognosis. The focus has been primarily on genes involved in DNA repair or drug metabolism and also genes whose somatic alteration is linked to oncogenesis or metastasis. Nominally statistically significant associations (P < 0.05) have been reported between EOC prognosis and SNPs in TP53 [56–62], the GST [63–65] family, LTBP-1L [66], the MRP-1 [67], HER2 [68], MMP-1 [69], CCND2 and CCNE1 [70], XPA and XPG [71], MDR-1 [72], ERCC1 [73], AIB1 [74], BRAF and KRAS [75], VEGF [76–78], TGF-β [66], CYP17 and SRD5A2 [79]. These studies have used a variety of different end-points including OS, PFS and tumour response; considerable variability exists both between and within studies in how the latter two were defined. None of the reported associations have reached stringent levels of statistical significance, and most are likely false positives.

Replicating genetic associations for EOC in large samples size

It is clear from the earlier association studies that large-scale replication analyses are required to clarify the reported genetic associations for EOC. The Ovarian Cancer Association Consortium (OCAC) was established for this purpose. Since its inception in 2005, OCAC has brought together ∼40 ovarian cancer studies worldwide which, when combined, can provide >15 000 cases and >15 000 controls for genotyping studies. This is the largest consortium for genetic association studies of ovarian cancer. The OCAC has conducted several studies aimed at replicating candidate gene associations. This includes several studies that have examined nonsynonymous coding SNPs and functionally relevant noncoding SNPs within or near candidate genes, and studies that have evaluated multiple sets of tagging SNPs (tSNPs) that define candidate gene ‘regions’ in terms of haplotypes of SNPs in LD with each other. Generally, these studies have shown that the vast majority of previously reported genetic associations for EOC are likely to either be false-positive findings, or at best associations conferring very weak effects. To date, only 12 of the 57 SNPs that have been followed up with additional genotyping by the OCAC continued to show association with EOC risk (Table 1 and Table S1), but none have reached the level of significance to be considered definitive.

Table 1. Summary of candidate gene SNPs followed up by the OCAC
GeneSNPInitial positive studyFollow-up study through the OCAC
CasesaControlsHetOR (95% CI)HomOR (95% CI)OR (95% CI)bP-valueReferencesCasesaControlsHetOR (95% CI)HomOR (95% CI)OR (95% CI)bP-trendReferences
  1. BF, Bayes factor; EOC, epithelial ovarian cancer; OCAC, ovarian cancer association consortium.

  2. Fifty-seven SNPs that show some evidence of association with EOC risk have been followed up by genotyping additional cases and controls through the OCAC. Listed are the 12 SNPs that continue to show some evidence of association, after replication, in a combined analysis at the 5% level; some are limited to specific histological subtypes of EOC. A full list of the 57 SNPs can be found in Table S1.

  3. aNumbers are for all histological subtypes of EOC, except where SC, serous; EC, endometrioid; MC, mucinous; CCC, clear cell. bPer-allele OR unless otherwise stated. cDominant OR (95% CI). dPosterior median OR (95% posterior probability intervals). eCombined serous risk estimate for all races; OR = 1.17 (1.08–1.27), P = 7.2 × 10−5.

ABL1rs28551928259401.26 (1.03–1.55)0.02[1]232242691.03 (0.91–1.16)1.59 (1.08–2.34)0.02[1]
CDKN1Brs2066827∼1500∼25000.88 (0.77–1.01)0.68 (0.51–0.90)0.004[2]361857190.98 (0.89–1.07)0.79 (0.65–0.95)0.04[2]
CDKN2A/2Brs3731257∼1500∼25000.90 (0.78–1.03)0.80 (0.59–1.07)0.046[2]360157050.89 (0.81–0.97)0.87 (0.73–1.03)0.008[2]
ESR1rs2295190112818661.24 (1.06–1.44)0.006[3]640793161.11 (1.04–1.19)0.001[3]
NMIrs11683487146425640.80 (0.69–0.93)0.87 (0.71–1.02)0.04[4]256143560.87 (0.8–0.99)c0.03[4]
PGRPROGINS2673961.19 (0.85–1.68)3.23 (1.19–8.75)0.02[5]651 (EC)67941.17 (1.01–1.36)0.04[6]
RB1rs2854344131863960.73 (0.61–0.89)c0.0009[7]381765840.88 (0.79–1.00)0.04[8]
TERTrs7726159675 (SC)11621.22 (1.06–1.40)0.007[9]1771 (SC)40221.14 (1.01–1.25)0.003e[9]
TP53rs2287498193 (SC)441Not given<0.05[10]1859 (SC)59271.30 (1.07–1.58)d165.7 (BF)[10]
TP53rs2287498391 (SC)786Not given<0.05[10]       
TP53rs12951053391 (SC)789Not given<0.05[10]2585 (SC)85081.19 (1.01–1.38)d47.8 (BF)[10]
TYMSrs4951398269411.1 (0.8–1.3)1.4 (1.0–1.8)0.10[11]354 (MC)99621.19 (1.03–1.39)0.02[12]
TYMSrs4951398269411.1 (0.8–1.3)1.4 (1.0–1.8)0.10[11]821 (EC)99620.90 (0.81–0.99)0.04[12]
TYMSrs4951398269411.1 (0.8–1.3)1.4 (1.0–1.8)0.10[11]409 (CCC)99620.86 (0.75–0.88)0.04[12]
VDRrs2228570711442.36 (1.23–4.53)c0.01[13]182034791.09 (1.01–1.19)0.04[14]

In one study, Ramus et al. [80] genotyped seven statistically significant candidate SNPs (P < 0.05) from previous reports in up to 4624 cases and 8113 controls from the OCAC; only one SNP (in RB1) remained marginally significantly associated with disease risk (P = 0.04). Similarly, an association for a SNP in the GALNT1 gene was not validated in a large follow-up study of 13 OCAC studies [81]. For PGR, the three candidate SNPs from previous studies showed no evidence of association with EOC risk in 4788 cases and 7614 controls from the OCAC, but the large sample size allowed stratification by histological subtype, and a marginally significant association was observed for the PROGINS variant with the endometrioid subtype (n = 651) [82]. Previous risk associations in VDR [83] and TYMS [84] also showed weak associations in OCAC follow-up studies.

The most compelling associations for EOC from candidate gene association studies have come from the more empirical tSNP approach. The OCAC has also performed validation of multiple tSNPs for which previous studies had shown evidence of risk association. In one study, two tSNPs in the TP53 gene were associated with an increased risk of serous ovarian cancer in up to 2585 cases and 8508 controls (per-minor allele OR = 1.30, 95% CI 1.07–1.57, Bayes factor [BF] = 165.7 for rs2287498; per-allele minor OR = 1.19, 95% CI 1.01–1.38], BF = 47.8 for rs12951053) [85]; the association of TP53 was also observed in a subsequent study of multiple DNA damage response and repair genes [86]. Another borderline significant association was found for a SNP near ESR1 in 6407 cases and 9316 controls (P = 0.001). The ability of OCAC to perform histological subtype analyses for candidate SNPs also revealed associations with the serous (n = 3545; P = 0.025) and mucinous (n = 447, P = 0.002) subtypes for the ESR1 SNP [87]. Other studies include an analysis of 13 cell cycle control genes in 11 OCAC case–control studies, which reported significantly reduced EOC risks associated with SNPs in CDKN1B (P = 0.04) and CDKN2A/2B region (P = 0.008) [88], but a subsequent study of 39 cell cycle genes failed to confirm these associations. Finally, a recent study of genetic variants in stromal–epithelial cross-talk genes showed that a SNP in the telomerase gene TERT was associated with an increased risk of serous ovarian cancer (per-minor allele OR = 1.17, 95% confidence interval [CI] 1.08–1.27, P = 7.2 × 10−5) [38]. This remains the most highly significant risk association reported from candidate gene studies of ovarian cancer.

Genome-wide association studies for ovarian cancer

Another approach to genetic association studies is the genome-wide association study (GWAS). GWAS use dense arrays of SNPs distributed throughout the genome that capture a large proportion of the common genetic variation (i.e. tSNPs), which are genotyped and tested for associations with the trait of interest. The main advantage of the GWAS is that it does not rely on any functional rationale in selecting genetic variants. There has been an explosion in the published literature over the last 3 years of GWAS successfully identifying novel susceptibility loci for a wide range of human characteristics and pathologies, from breast cancer risk [89] to human height [90]. These studies have provided overwhelming empirical support to the hypothesis that many traits are driven, at least in part, by common genetic variation.

Ovarian cancer susceptibility

There has been one published GWAS (summarized in Fig. 2) aimed at finding common risk alleles for EOC [91–93]. Like many other GWAS, a staged design was used in which hundreds of thousands of SNPs are genotyped in a restricted set of cases and controls in a first stage to identify a top-ranked series of SNPs reaching a statistical threshold that suggests some evidence of disease association. These SNPs are then followed up in one or more replication stages to confirm if any SNPs reach a stringent level of statistical significance to confirm their disease association. The ovarian cancer GWAS comprised roughly 1800 EOC cases and 2300 controls genotyped for half a million SNPs in stage 1, of which approximately 23 000 SNPs were followed up in about 9000 additional subjects (4200 cases and 4800 controls). Additional replication was then performed through the OCAC for the top 35 SNPs in a further 4300 cases and 6000 controls. These analyses identified five novel susceptibility loci for EOC (at 9p22, 8q24, 2q31, 3q25 and 17q21) reaching significance levels of <5 × 10−7. A prognosis GWAS that was performed in parallel identified an additional susceptibility locus, and so in total, six risk loci have now been identified for EOC (Table 2). An important feature of these loci is that they confer only modest effects, with per-allele relative risks ranging from around 0.8 to 1.2.

Figure 2.

Study design for genome-wide association study of ovarian cancer.

Table 2. Common genetic susceptibility loci identified through the OCAC GWAS
LocusSNPSubtypeNo. casesNo. controlsMAFOR (95% CI)aP-value
  1. GWAS, genome-wide association studies; MAF, Minor allele frequency; OCAC, ovarian cancer association consortium.

  2. aPer-minor allele odds ratio (95% confidence interval [CI]).

2q31rs2072590All10 40616 3400.321.16 (1.12–1.21)4.5 × 1014
Serous592516 3401.20 (1.14–1.25)3.8 × 10−14
Mucinous79616 3401.30 (1.17–1.44)7.3 × 10−7
3q25rs2665390All10 40617 3690.081.19 (1.11–1.27)3.2 × 10−7
Serous589617 3691.24 (1.15–1.34)7.1 × 10−8
8q24rs10088218All10 46216 3620.120.84 (0.80–0.89)3.2 × 10−9
Serous591716 3620.76 (0.70–0.81)8.0 × 10−15
9p22rs3814113All876111 8310.280.82 (0.79–0.86)5.1 × 10−19
Serous484711 8310.77 (0.73–0.81)4.1 × 10−21
17q21rs9303542All10 24213 0910.271.11 (1.06–1.16)1.4 × 10−6
Serous581413 0911.14 (1.09–1.20)1.4 × 10−7
19p13rs2363956All10 48013 1760.451.12 (1.07–1.17)1.2 × 10−7
Serous589513 1761.18 (1.12–1.25)3.8 × 10−11

As described previously, there is substantial evidence suggesting that germline genetics contribute to disease heterogeneity for ovarian cancer. This appears to be true for common low penetrance as well as high-penetrance genes. Of the six susceptibility loci identified, five were most strongly associated with the serous subtype disease, which partly reflects the greater frequency of serous over other subtypes in the discovery phase of the GWAS. Some loci may also predispose to multiple subtypes of disease. For example, at the 2q31 locus, common variants were associated with risks of serous disease but also the mucinous subtype. It is also likely that novel genetic variants exist that predispose specifically to other (nonserous) EOC subtypes. Nonserous subtypes are, however, much rarer than serous EOC, and so, the power to detect alleles conferring the modest risks seen so far will be limited. These types of study will only be possible through the continued and expanded collaborations developed by consortia such as the OCAC.

Ovarian cancer survival

Alongside the GWAS to identify risk alleles for EOC, a GWAS to identify germline variants associated with survival after EOC diagnosis was also performed. Whereas a case–control comparison is used to identify risk alleles, a case only analysis is used to identify prognostic associations. Out of the cases in the phase 1 of the GWAS, 1700 had survival time data available. The ∼5000 top-ranked survival SNPs from the phase 1 analysis were genotyped in the phase 2 sample set. The top two loci from the combined phase 1 and 2 analysis (at 13q32 and 19p13) were then genotyped in an additional 4300 cases. Neither locus showed evidence with survival time in this replication set. However, SNPs at the 19p13 did show an association with EOC risk, with the strongest association in cases of the serous subtype (P = 4 × 10−11). This locus is of further interest because simultaneously, and for the same SNPs, it was reported to be associated with the risk of breast cancer in BRCA1 mutation carriers and is also associated with an increased risk of double (ER−, PR−) and triple-negative (ER−, PR−, HER2−) breast cancer [94].

The fact that the GWAS did not identify survival loci reaching genome-wide significance levels could be related to sample size; the power of this study to detect SNPs of moderate effect was limited. Larger EOC case studies with more complete information collected on clinical outcomes may yet identify genetic associations for prognosis in future. However, defining the most suitable analytic approach for these studies is also critical. Evaluating OS is the most accessible end-point to study, and perhaps, the most versatile as a variety of mechanisms are reflected in OS including treatment response and tumour aggressiveness. Disease-specific survival may also be used but is hindered by misclassification in determining cause of death. PFS and tumour response may be more useful for studies of response to therapy, but neither is commonly available for large numbers of cases outside of clinical trials. Varying definitions of progression-free survival may also lead to misclassification and therefore loss of power. The choice of whether to adjust for known prognosticators of ovarian cancer survival must be treated with caution. In terms of power for detection, an advantage will only be gained if (i) the SNP is associated with both the covariate of interest and survival, and (ii) if the SNP and covariate impact survival in the opposite direction. Examples of this do occur (e.g. BRCA1/2), but the cost could be failure to detect variants that influence survival through a mechanism related to the covariate and loss of power because of missing covariate data.

Future perspectives

Further characterization of ovarian cancer susceptibility loci

None of the genetic variants identified as common low-penetrance susceptibility alleles for EOC have any obvious function that suggests a causal role in disease development. In all likelihood, the most significantly associated risk alleles at these loci represent markers in LD with one or more functional variants. GWAS to date have been limited in their ability to identify the true causal variants for disease because only a fraction of the genetic variation throughout the genome has been screened. Therefore, two immediate goals of follow-up studies of susceptibility loci are as follows: (i) detailed fine mapping of susceptibility regions to find the most significant risk-associated alleles; (ii) elucidation of the underlying biology at these susceptibility loci.

If the risk associations are driven by a single causative variant, the strength of the effect size for the causative SNP will be larger than its correlated marker [95]. Therefore, identification of the functional variant would have the potential of improving risk prediction and the power to detect possible interaction between environmental and other genetic factors. Further insight into the biological basis of the association would also be gained through discovery of the causal variant. Because of the prohibitively large numbers of samples that are required to discriminate between highly correlated variants [96], fine mapping studies often need to include multiple ethnic groups showing differing patterns of LD. At its most successful, fine mapping studies are likely to reduce the number of possible causal variants from hundreds to a handful of SNPs. A critical next step involves experimental and bio-informatic approaches to identify SNPs with plausible functional effects.

The following are the two key questions in addressing function: (i) How do risk variants affect gene expression and gene function? and (ii) How do the gene targets contribute to ovarian cancer development? The effect sizes for associated SNPs are small relative to effects seen in monogenic diseases. Thus, the functional effects of SNPs are likely to be subtle, and it is unclear whether current molecular biology methods have enough resolution to differentiate such small effects. Identifying the candidate genes that are associated with disease relies on developing suitable models that accurately recapitulate the disease in vitro and in vivo, and systematic approaches will need to be developed to evaluate multiple candidate target genes within susceptibility regions.

For the six susceptibility loci associated with EOC risk, whilst the SNPs have no obvious functional effects, the regions themselves, or genes within the regions, possibly offer some intriguing insights into the functional role of these loci in EOC development. For example, the 8q24 locus is recognized as a multiple cancer susceptibility locus for many malignancies including breast, prostate and colorectal cancer. The SNP associations for EOC lie ∼700 kb 3′ of the cMYC oncogene, and the suggestion is that SNPs in this region influence distal regulation of this gene. The 19p13 susceptibility region contains the putative candidate gene MERIT40, which has been shown to interact directly with the BRCA1 protein, and TiPARP on 3q25, a poly (ADP-ribose) polymerase (PARP) belonging to a pathway that acts as an alternative DNA repair mechanism and small molecule therapeutics target in BRCA1/2-deficient cells.

Clinical implications and future studies

The greater understanding of the molecular mechanisms of ovarian cancer development provided by the identification of new susceptibility genes and biological pathways could lead to the development of novel, more effective therapies that could be individually tailored on the basis of genotype. The paradigm is the identification of the BRCA1/2 genes; following the functional characterization of these genes, a novel therapy for BRCA1/2-associated cancers has been developed based on inhibition of the PARP DNA repair pathway. Results from a phase 1 clinical trial of an oral PARP inhibitor (olaparib) have been promising, showing significant antitumour activity in ovarian cancers with BRCA1/2 mutations and few of the adverse effects of conventional chemotherapy [97]. There is also some evidence that patients with high-grade serous ovarian cancer also respond favourably to olaparib, perhaps suggesting the development of BRCA or homologous recombination–associated disease through somatic mechanisms (e.g. methylation, somatic mutations; copy number deletions) [98]. The identification of common genetic susceptibility variants could also be used to establish population-based genetic risk prediction strategies. Currently, genetic risk assessment for ovarian cancer can only be performed for individuals with a family history of disease and/or where known or suspected BRCA1/2 mutations segregate within families. In such cases, prophylactic bilateral salpingo-oophorectomy is an intervention approach that almost completely removes disease risk. Could such an intervention be applied on a more population-based level using common low-penetrance alleles? Testing for single low-penetrance SNPs is of little use, but recent studies have described the potential value of using combinations of low-risk alleles for risk stratification in the context of population-based screening programs in breast [99] and colorectal cancers [100]. A similar targeted screening approach based on genetic risk prediction could be applied to ovarian cancer, but effective screening tools for ovarian cancer must first be established. Additionally, many more loci will need to be identified as the handful of confirmed risk alleles identified thus far do not provide sufficient information. We used the method of Park et al. [101] to estimate the number of susceptibility SNPs for serous subtype ovarian cancer and the power of future GWAS studies to detect additional loci. We estimate that there are 71 susceptibility loci within the range of effect sizes seen in our current GWAS and that these loci could explain 15% of the total genetic variance for ovarian cancer. The number of these additional loci that would be expected to be found in future GWAS for serous ovarian cancer of varying sample sizes are shown in Fig. 3.

Figure 3.

Sample size required to detect loci in future genome-wide association studies (GWAS) for serous EOC. Included in this analysis were the six loci for ovarian cancer attaining significance levels of <5 × 10−7. Because some of these loci would not have been detected without a serous-only analysis of the combined phase 1 and 2 datasets, we calculated the power to detect the serous-only effect in the combined 1 and 2 analysis, conditional on the loci attaining the significance cut-off for the un-stratified phase 1 analysis. The sample size is for a one-phase GWAS of an equal number of serous ovarian cancer cases and controls.

Identifying additional EOC susceptibility loci

As the cataloguing of human genetic variation continues, and new platforms start to enable larger and more comprehensive GWAS to be performed, so it will become possible to identify additional common susceptibility variants for EOC. However, it seems unlikely that common variants alone will explain all the remaining heritability of ovarian cancer. Additional heritability may be explained by gene–environment interactions, gene–gene interactions, epigenetics and rare (<1%) or uncommon (1–5%) variants. Few established examples in cancer exist for the former three, and more study will be needed to determine their relative importance. Rare and uncommon variants have already been shown to influence susceptibility to some cancers (e.g. breast, prostate). Association studies of uncommon variants will benefit from recently released genotyping chips that include content from the 1000 Genomes Project pilot data [102] and provide increased low-frequency variant data. To capture rare susceptibility variants with effects too small to be detected by linkage, it will be necessary to sequence the genomes of ovarian cancer cases. High-throughput next-generation sequencing studies, although still prohibitively costly, will start to accelerate the pace with which these are identified for multiple traits including ovarian cancer.


Researchers are slowly unravelling the basis of genetic susceptibility to ovarian cancer. Following the initial work of familial-based studies to identify high-penetrance susceptibility genes, large population-based studies have now uncovered common low-penetrance susceptibility loci. This has only been made possible through the collaborative work of a multi-centre consortium, the OCAC. The continuing work of the OCAC and availability of new genotyping and sequencing technologies will likely result in the discovery of additional common, uncommon and rare susceptibility loci for the disease. This may also lead to the discovery of genetic variants that influence clinical features of EOC including disease heterogeneity and outcome. Finally, two of the major challenges in future will be developing approaches to understand the functional relevance of susceptibility loci that emerge, and applying these findings to establish improved preventative and clinical intervention strategies.

Conflict of interest

No conflict of interest to declare.