The Effects of Common Genetic Variants in Oncogenes on Ovarian Cancer Survival

Purpose: The 5-year survival rate for invasive epithelial ovarian cancer is <35%. It has been suggested that common, germline genetic variation may influence survival after cancer diagnoses, which might enable the prediction of response to treatment and survival in the clinical setting. The aim of this study was to evaluate associations between common germline genetic variants in the oncogenes BRAF, ERBB2, KRAS, NMI, and PIK3CA, and survival after a diagnosis of epithelial ovarian cancer.

Experimental Design: We evaluated the association between 34 tagging single nucleotide polymorphisms and survival in 1,480 cases of invasive epithelial ovarian cancer cases from three different studies. Cox regression analysis, stratified by study, was used to estimate per rare allele hazard ratios (HR).

Results: The minor allele rs6944385 in BRAF was significantly associated with poor survival [HR, 1.19; 95% confidence intervals (95% CI), 1.02-1.39; P = 0.024]. The association remained after adjusting for prognostic factors (adjusted HR, 1.20; 95 CI, 1.03-1.40; P = 0.018). A haplotype of BRAF was also associated with poor survival (HR, 1.24; 95% CI, 1.02-1.51; P = 0.029) and was more significant after adjustment (HR, 1.44; 95% CI, 1.15-1.81; P = 0.001). We also found evidence of an association between a KRAS haplotype and poor survival in serous subtype (HR, 1.69; 95% CI, 1.21-2.38; P = 0.002), but this was no longer significant after adjustment. Finally, when analyses were restricted to the serous histologic subtype, the rare allele rs10842513 in KRAS, was associated with poor survival (HR, 1.40; 95% CI, 1.10-1.78; P = 0.007).

Conclusion: Common genetic variants in the BRAF and KRAS oncogenes may be important in the prediction of survival in patients with invasive epithelial ovarian cancer.

Ovarian cancer is the most lethal gynecological malignancy; in 2002, there were 204,500 new diagnoses and ∼125,000 deaths from the disease worldwide (1). The 5-year survival rate for stage III/IV disease is 15% to 25% (2). Overall, 5-year survival diagnosed at any stage is ∼35% and has improved little over the last 30 years (3, 4). This poor survival rate is probably due to the late presentation and diagnosis of the disease; ∼70% of cases are diagnosed at stage III/IV (3).

It is well-documented that patients can show different responses to the same treatment, and there is evidence suggesting that germline genetic variation may play a role in chemotherapy resistance (5). Thus, it is plausible that genetic polymorphisms can influence survival in ovarian cancer patients by affecting how well the patient responds to treatment. The identification of genetic polymorphisms that affect survival may aid in the prediction of clinical outcome and response to treatment. Previous studies have shown that genetic polymorphisms may be used to predict clinical outcome; but sample sizes for many of these studies are small (<300 cases), increasing the likelihood that these findings represent false positive reporting.

Activating mutations or overexpression of proto-oncogenes are characteristic features of neoplastic transformation (6, 7). Proto-oncogenes are normally involved in the regulation of cellular processes such as cell division, cell proliferation, survival, motility, and apoptosis. Genetic alterations in these genes have been reported as initiating and early-stage events in tumor development, as well as late-stage events associated with metastatic progression. Variation in the expression of a variety of oncogenes in tumors has also been shown to correlate with clinical features of disease, including survival after diagnosis.

Several oncogenes have been implicated in ovarian cancer development including KRAS, BRAF, PIK3CA, ERBB2 (HER-2/Neu), and NMI. BRAF and KRAS are two of the most commonly mutated and overexpressed genes in ovarian tumors (8) and are also commonly mutated in other cancer types. KRAS acts upstream of BRAF in the mitogen-activated protein kinase (MAPK) pathway. Somatic mutations in KRAS have been associated with a worse survival in patients with colorectal, lung, and pancreatic cancers (9–11). Mutations in BRAF have been associated with poor survival in patients diagnosed with papillary thyroid cancer and colon cancer (12, 13).

PIK3CA, a catalytic subunit of the lipid kinase phosphatidylinositol 3-kinase (PIK3), regulates cell proliferation, adhesion, apoptosis, survival, and motility (14–16). PIK3CA has also been shown to be mutated in several cancer types, including ovary, brain, lung, stomach, colon, and prostate (17). In addition, the gene may be associated with clinical outcome in patients with ovarian, breast, thyroid, lung, and colon cancer (12).

The human epidermal growth factor receptor-2 gene ERBB2 is involved in cell proliferation and cell differentiation (18). ERBB2 is amplified, overexpressed, or mutated in a number of cancers, including ovarian, breast, stomach, liver, bladder, lung, and prostate (18, 19). In a recent study, ERBB2 overexpression was found in 39% of ovarian tumors, and variations in tumor expression were associated with clinical outcome (20). Associations between ERBB2 expression in tumors and survival have also been reported for breast and colon cancer (21).

The NMYC and STAT interactor (NMI) gene is a transcription factor, which interacts with NMYC, CMYC, MAX, FOS, and other transcription factors and also with the breast/ovarian cancer susceptibility gene BRCA1 (22, 23). The gene is also mutated in human leukemias and other cancer types (24).

In a previous study, we evaluated the association between 34 tagging single nuncleotide polymorphisms (tSNP) in the BRAF, ERBB2, KRAS, NMI, and PIK3CA oncogenes and the risks of ovarian cancer in ovarian cancer case-control series from Denmark, United Kingdom, and the United States (Quaye et al.).¹⁰

We found one tSNP (rs11683487) in NMI that may be associated with a reduced risk of ovarian cancer (P_dominant = 0.0039) and associations between ovarian cancer risks and haplotypes in the BRAF, ERBB2, and NMI genes. The aim of the current study was to evaluate whether germline genetic variation in the BRAF, ERBB2, KRAS, NMI, and PIK3CA oncogenes are associated with survival after diagnosis of invasive ovarian cancer, in patients from these same populations.

Study subjects. The individuals in this study were all confirmed cases of invasive epithelial ovarian cancer from three different case-control populations from Denmark, United Kingdom, and the United States. Details of these populations have been described previously and are summarized in Table 1 (25). These studies were as follows: (a) The Danish MALOVA study (445 cases); (b) The UK SEARCH study (728 cases); (c) The Genetic Epidemiology of Ovarian Cancer Study (GEOCS; previously FROC) from Stanford, CA (327 cases). Local ethics committee approval was given for the collections and genotyping in all individuals.

Follow-up data. The MALOVA study comprises cases from regions in and around Copenhagen. The collection took place from December 1994 to May 1999. Follow-up was until 2003. Individuals living in Denmark have a unique personal identification number, which was used to identify patients who were alive, as well as those who had died or emigrated. The cause of death of cases who died during follow-up was determined by matching medical records with a Danish Hospital Reference System. At the time of analysis, there had been 300 deaths (67% of cases).

The GEOCS study comprises cases from the National Cancer Institute's Surveillance Epidemiology and End Results Program, USA. Diagnoses were between 1997 and 2002. Vital status information was obtained from the Greater Bay Cancer Registry, San Francisco, twice during the study. The most recent follow-up occurred in 2004. Computerized hospital tumor registry data or medical records were used for updated vital status by cancer registry staff. The death index of the state was also used to follow the vital status of patients. There was a lag time of ∼18 mo with the death index of the state. At the time of analysis, there had been 147 deaths (45% of cases).

The SEARCH (UK) study comprises cases from East Anglia, West Midlands, and Trent regions of England. Cases were diagnosed from 1991. Active follow-up was conducted 3 and 5 y after diagnosis, and then at 5-y intervals by the Eastern and West Midlands cancer registries. The most current update was on 30th September 2006. Follow-up involved searching hospital information systems for recent visits and contacting general practitioners for the patient's vital status if a recent visit had not occurred. At the time of analysis, there had been 194 deaths (27% of cases).

Tag SNP selection. tSNPs for the genes BRAF, ERBB2, KRAS and PIK3CA, and NMI were selected using HapMap SNP genotype data build 34 from Haploview version 3.32 (26) and Tagger (27). tSNPs were selected with an r² of ≥0.8, a Hardy-Weinberg equilibrium of ≥0.01, and with a minor allele frequency of ≥0.05. For all tSNPs, we selected SNPs with ≥80% genotype call rates for a panel of CEPH HapMap subjects (28).

Genotyping methods. A combination of iPLEX (Sequenom, Inc.) and TaqMan ABI 7900HT Sequence Detection System (Applied Biosystems) were used to genotype samples (28). Samples from MALOVA and SEARCH studies were genotyped with both TaqMan and iPLEX; samples from the GEOCS study were genotyped only by TaqMan. If we were unable to genotype a tSNP, a different tSNP in LD with an r² of >0.8 was genotyped instead. Assays were repeated when the call rates were below 90%. For quality control purposes, duplicate samples and no template controls were included in genotyping studies.

We identified 40 tSNPs in the 5 genes. Of these, six tSNPs (BRAF rs11771946; KRAS rs11047912 and rs17388893; PIK3CA rs1607237, rs6443624, and rs3729692) could not be genotyped or efficiently tagged by any other SNP. Therefore, 34 tSNPs were genotyped in 1,480 ovarian cancer cases. We were unable to genotype three tSNPs (BRAF rs1733832, NMI rs11551174, and PIK3CA rs2865084) in the GEOCS study by TaqMan due to failed assay manufacture. Therefore, there is some variation in the number of tSNPs that could be genotyped in each study (28).

Statistical methods. There was a variable time between diagnosis and patient recruitment, therefore subjects were only considered to be at risk from the date of recruitment (blood draw). This provided an unbiased estimate of the relative hazard, provided that the proportional hazard assumption is not violated. Hence, the survival period was defined as starting at the date of blood draw. All-cause mortality was the only end-point of interest; censoring was at the date a participant was last known to be alive or at 10 y after diagnosis if the participant was still alive. Log-log survival curves were used to ensure that the assumptions of proportional hazards were met. The primary tests were likelihood ratio test for trend (one degree of freedom), based on the number of rare alleles carried. We used Cox regression analysis, stratified by study to estimate the hazard ratio (HR) per rare allele carried. The HRs were adjusted for the prognostic factors age at diagnosis, tumor stage, tumor grade, and histologic subtype. STATA version 8.2 was used for all statistical analyses and to produce the Kaplan-Meier plot of survival estimates. Kaplan-Meier curves were used to estimate survival probabilities for individuals within a particular group, over a period of time.

The TagSNPs program was used to estimate haplotype dosages in the survival analysis. TagSNP implements an expectation-substitution approach to account for the uncertainty caused by the unphased genotype data (29). Haplotypes were defined using the confidence interval option of Haploview (26) with minor adjustments to include adjacent SNPs to increase the cumulative frequency of the common haplotypes (>2% frequency) to >90%. KRAS was the only gene with two haplotype blocks, the other genes had only one haplotype block. Haplotypes with a frequency of >2% in at least one study were considered “common”; “rare” haplotypes were pooled. Cox regression analysis, stratified by study, was used to assess the effect of each haplotype dose on survival.

Thirty-four tSNPs in BRAF, ERBB2, KRAS, NMI, and PIK3CA were genotyped in a combined series of 1,480 confirmed cases of invasive ovarian cancer from three different populations. There were 632 deaths in 5,914 person-years at risk in this series at the time of analysis. Univariate Cox regression analysis showed that there were no associations between genotype frequency and survival for tSNPs in the KRAS, ERBB2, NMI and PIK3CA, oncogenes (P ≥ 0.05; Table 2).

We found an association with ovarian cancer survival for a single tSNP (rs6944385) in the BRAF gene (Table 2; Fig. 1). The hazard associated with the rare allele of rs6944385 increased by ∼20%. The per rare allele HR for rs6944385 was 1.19 [95% confidence interval (95% CI), 1.02-1.39; P_trend = 0.024].

We also evaluated the association between survival and common haplotypes for each gene (Table 3). We found no association for haplotypes in the NMI, ERBB2, and PIK3CA genes. In BRAF, the h32342344 haplotype was associated with a significant increase in hazard (HR, 1.24; 95% CI, 1.02-1.51; P = 0.029); and in KRAS, there was borderline evidence that haplotype h142222 in block 2 was also associated with an increase in hazard (HR, 1.28; 95% CI, 1.00-1.64; P_trend = 0.051).

Age at diagnosis, histologic subtype, tumor grade, and stage were all significantly associated with survival (P < 0.05), when modeled independently. A multivariate model including these variables showed that survival was significantly associated with histologic subtype, age of >50 years, tumor grades 2 and 3, and advanced stage (where the primary tumor had metastasized to the lymph nodes or a distant location). After adjusting for these prognostic factors, we found that the rare allele rs6944385 in BRAF remained significantly associated with survival (HR, 1.25; 95% CI, 1.01-1.50; P = 0.013). The association with reduced survival and the BRAF haplotype h32342344 was stronger when adjusted for these prognostic factors (HR, 1.44; 95% CI, 1.15-1.81; P = 0.001). The other associations were no longer significant (P > 0.05).

When we restricted the analyses to the serous histologic subtype, we found that the tSNP rs10842513 in KRAS was associated with increased hazard (per-rare allele HR, 1.40; 95% CI, 1.10-1.78; P = 0.007). However, this result was no longer significant after adjusting for the prognostic factors (HR, 1.29; 95% CI, 0.98-1.69; P = 0.074). We also found an increased hazard for the h142222 haplotype of KRAS in serous ovarian cancer cancers (HR, 1.69; 95% CI, 1.21-2.38; P = 0.002); but once again, this was no longer significant after adjusting for prognostic factors (HR, 1.31; 95% CI, 0.88-1.96; P = 0.185).

Oncogenes play an important role in the development of cancers. BRAF, ERBB2, KRAS, and PIK3CA are oncogenes that are involved in the development of ovarian cancer, as is the NMI gene through its interaction with the oncogenes NMYC, CMYC, MAX, and FOS. In this study, we investigated the association between survival after diagnosis and germline genetic variation for 34 tSNPs from these 5 genes, genotyped in 1,480 non-Hispanic White patients from Denmark, United Kingdom, and the United States. To our knowledge, this is the first report of epithelial ovarian cancer survival and common tagging polymorphisms in the BRAF, KRAS, and NMI; and the study is substantially larger and better powered than other studies examining the association between survival and variants in the ERBB2 and PIK3CA genes.

No association was found between survival and tSNPs in ERRB2, NMI, KRAS, and PIK3CA with the nominal model. This is in contrast to previous reports suggesting there may be associations with survival for SNPs in ERBB2 and PIK3CA. However, these studies were done in sample sizes of <300 cases, which would have limited power to detect significant associations (5, 20, 30).

The rare allele of the tSNP rs6944385 from BRAF was associated with poor survival. There were 28 rare homozygotes for this tSNP, which represents ∼2% of the total study population. This tSNP is in intron 1 of BRAF and there is no evidence supporting this as a functional target. rs6944385 also tags a nonsynonymous coding SNP in BRAF, rs9648716, with an r² of 1. We used the program Pupasuite (31, 32) to assess the known and predicted functions of rs9648716 but found nothing to provide a functional reason why this SNP might influence survival. It is important to note that this is a prediction, and there remains a possibility that the SNP is functional. It is also possible that the significant association we see with rs6944385 is due to some other, as yet unidentified SNP linked to rs6944385. We also found associations with clinical outcome for haplotypes in BRAF and KRAS. The association with the haplotype in BRAF was still significant after adjusting for prognostic factors.

It is possible that these are chance findings, and so they should be treated with caution, particularly as they have not been adjusted for multiple testing, and the P values are greater than that desired for case-control genetic association studies (P < 10⁻⁴; ref. 33). However, it is intriguing that the positive associations we find are for the BRAF and KRAS oncogenes. Both genes are members of the MAPK pathway, and alterations of these genes are among the earliest somatic genetic changes identified in epithelial ovarian cancers. Alteration of either of these genes seems to influence the clinical features of ovarian tumors, in particular, the histologic subtype of tumors. For example, activating KRAS mutations are particularly associated with ovarian tumors of the mucinous subtype (34).

The MAPK pathway transmits signals for processes such as cell proliferation and cell survival from the cytoplasm to the nucleus (35). The pathway is activated by growth stimulating signals (36), and mutations in BRAF or KRAS lead to the continuous activation of the MAPK pathway. The activation of the MAPK pathway activates downstream cellular targets, including both cellular and nuclear proteins (36). It has been shown that the inhibition of a downstream effector of the MAPK pathway in cell lines with BRAF or KRAS mutations, results in the inhibition of cell growth and promotion of apoptosis (35). It is conceivable then that a functional germline variant in either of these genes could influence a multitude of downstream targets that may affect the biological and clinical characteristics of ovarian cancers.

The issue of combining incidence and prevalence case samples for these studies is also an important one, which could introduce bias into the analyses. In the current study, we combined cases from three different populations: the MALOVA study consists of incident cases only, the GEOCS consists of prevalent cases, and the SEARCH study contains both prevalent and incident cases. It is likely that prevalent cases had treatment before being recruited into the studies; but we accounted for this in the analyses by stratifying the studies, thus reducing the possibility of bias. There is also a chance of ascertainment and survival bias in prevalent case collections, whereby subjects died before blood could be taken for the study, and were therefore excluded. This may have had a strong effect on the fraction of deaths observed in each study. However, the inclusion of prevalent cases does not result in a bias of the HR estimates, as the analysis allows for left-truncation of the data. It has been shown that a test of association using prevalent cohorts are valid (regardless of whether or not left-truncation has been allowed for; refs. 37, 38). Ideally, the incidence and prevalent cases should be analyzed separately, but there were insufficient numbers of cases to make this worthwhile (i.e., without detrimentally limiting the power of the study). Another factor that could influence the results of this study is missing clinical data. This was an issue for the SEARCH study, and reduced the overall size of the study, but was adjusted for in the analyses.

In conclusion, we have evaluated survival after a diagnosis of invasive epithelial ovarian cancer cases and putative associations for 34 common tSNPs in the oncogenes BRAF, ERBB2, KRAS, NMI, and PIK3CA. We found evidence of association with survival in patients with ovarian cancer for a single tSNP and a haplotype in the BRAF gene. These associations were still significant after adjustments for age at diagnosis, tumor stage, grade, and histologic subtype. If validated by other similar studies, these findings could be important for predicting clinical outcome of invasive epithelial ovarian cancer cases, and may suggest biological mechanisms and/or pathways for targeting chemotherapeutic treatments for ovarian cancers.

No potential conflicts of interest were disclosed.

We thank Joan MacIntosh, Hannah Munday, Barbara Perkins, Clare Jordan, Kristy Driver, Mitul Shah, the local general practices and nurses and the East Anglian Cancer Registry for recruitment of the SEARCH cases; the EPIC-Norfolk investigators for recruitment of the Search controls; Craig Luccarini and Don Conroy for expert technical assistance; and to all the study participants who contributed to this research.