STK15 Polymorphisms and Association with Risk of Invasive Ovarian Cancer

STK15 is a putative oncogene that codes for a centrosome-associated, serine/threonine kinase, the normal function of which is to ensure accurate segregation of chromosomes during mitosis. Amplification of STK15 has been reported in ovarian tumors, suggesting a role in ovarian cancer pathology. STK15 is polymorphic with two single nucleotide substitutions (449t/a and 527g/a) in evolutionarily conserved regions causing amino acid changes (F31I and V57I). Two other nucleotide substitutions (287c/g and 1891g/c) of unknown significance are in 5′ and 3′ untranslated regions (UTR), respectively. To learn more about the involvement of STK15 in ovarian cancer, we genotyped and haplotyped these polymorphisms in three population-based ovarian cancer case-control studies from the United Kingdom, United States, and Denmark with 1,821 combined cases and 2,467 combined controls and calculated risks for developing ovarian cancer. Genotypes of individual polymorphisms in control groups of the United Kingdom, United States, and Denmark conformed to Hardy-Weinberg equilibrium. In combined cases and combined controls, rare allele frequencies were 0.23 and 0.21 for I31, 0.16 and 0.17 for I57, 0.08 and 0.07 for 5′ UTR g, and 0.25 and 0.24 for 3′ UTR c, respectively. Using FF common homozygotes of F31I as comparator, there was increased ovarian cancer risk to FI heterozygotes (odds ratio, 1.18; 95% confidence interval, 1.01-1.36), II homozygotes (odds ratio, 1.25; 95% confidence interval, 0.89-1.75), and I31 allele carriers (odds ratio, 1.17; 95% confidence interval, 1.02-1.35) in the combined group data. For either V57I, 5′ UTR C/G, or 3′ UTR G/C, all genotypic ovarian cancer risks were essentially in unity relative to their respective common homozygotes, VV, cc, or gg. Haplotype analysis of combined group data revealed seven haplotypes with frequencies between 0.02 and 0.5, with c-F-V-g the most common. None of the haplotype-specific risks significantly differed from unity relative to c-F-V-g. These results suggest a model of dominant inheritance of ovarian cancer risk by the I31 allele of F31I and that the I31 allele may be a common ovarian cancer susceptibility allele of low penetrance.

STK15 is a human serine/threonine kinase also known as Aik, ARK1, STK-6, BTAK, aurora 2, aurora A, or HsAIRK1 (1-7). STK15 is located on chromosome 20q13 (Locus Link ID 6790)⁶

and encodes a 46-kDa protein localized to centrosomes (1-4). STK15 RNA, protein, and kinase activity fluctuate with the cell cycle, peaking at G₂-M (2, 4, 5). The protein structure consists of a 274–amino acid COOH-terminal kinase catalytic domain and a 129–amino acid NH₂-terminal domain proposed to mediate interaction with other proteins that facilitate translocation of STK15 to the nucleus during mitosis (7). Overexpression of STK15 in nonmalignant cultured cells leads to centrosome amplification, chromosome instability, transformation, and the ability to induce tumors (4, 5). Amplification and/or overexpression of STK15 has been reported in many types of human cancers and cancer cell lines including those of the breast, colon, prostate, endometrium, and ovary (3-5, 8-13). Overexpression of STK15 mRNA significantly correlated with chromosome instability in breast cancer (10). The mouse homologue of STK15 was mapped as a skin tumor susceptibility gene, with the susceptibility allele overexpressed in normal cells and preferentially amplified in tumor cells (14). These data suggest that a normal role of STK15 is to ensure accurate segregation of chromosomes and that enhanced expression (i.e., gain of function) of STK15 is oncogenic.

Increased copy number of chromosome 20q, the region bearing STK15, has been widely reported in sporadic and hereditary ovarian tumors by comparative genome hybridization (15-24). Moreover, amplification of the putative oncogene, STK15, has been reported in sporadic and hereditary ovarian tumors as well as in ovarian cancer cell lines (9, 12). These data suggest a role for STK15 in the pathogenesis of ovarian cancer.

Two STK15 polymorphisms, F31I (dBSNP reference rs2273535) and V57I (dBSNP reference rs1047972), are in evolutionarily conserved regions of the NH₂-terminal domain (4). F31I and V57I are caused by single nucleotide substitutions, 449t/a and 527g/a, respectively (National Center for Biotechnology Information reference sequence NM003600). The functional significance of V57I is unknown. However, the I31 allele of F31I was preferentially amplified and associated with aneuploidy in human colon tumors (14). The same allele also was associated with faster growth of cultured cells and experimental tumors relative to the F31 allele. Thus, the I31 allele may be more oncogenic than the F31 allele and involved in human tumor susceptibility. To learn more about the involvement of STK15 in the etiology of ovarian cancer, we investigated the F31I and V57I polymorphisms for associations with risk of invasive ovarian cancer in three population-based ovarian cancer case-control studies from the United Kingdom, United States, and Denmark. We also studied two additional polymorphisms: a c/g change at nucleotide 287 in the 5′ untranslated region (UTR; dBSNP reference rs732417) and a g/c change at nucleotide 1,891 in the 3′ UTR (dBSNP reference rs8173). The functional significance of either UTR single nucleotide polymorphism is unknown, but either may have a regulatory function. Furthermore, by genotyping multiple single nucleotide polymorphisms across the gene, we were able to estimate haplotype frequencies in cases and controls and to examine associations of haplotypes with risk.

The SEARCH ovarian cancer study is an ongoing, population-based ovarian cancer case-control study covering the regions served by the East Anglia and West Midlands cancer registries in the United Kingdom. Eligible women are those diagnosed since 1991 with invasive epithelial ovarian cancer under age 70 years. Controls have been randomly selected from the EPIC-Norfolk component of the European Prospective Investigation of Cancer, a prospective study of diet and cancer being carried out in the same population from which the cases have been drawn. The EPIC-Norfolk cohort comprises 25,000 individuals resident in Norfolk (East Anglia), ages 45 to 74 years at first interview in 1993. Blood for DNA extraction was collected during the second health check in 1998 to 2000. The ethnic background of cases and controls is similar, with >98% being White Europeans. Participants are asked to provide written consent, to complete an epidemiologic questionnaire, and to provide a 20 mL whole blood sample. The study was approved by the Anglia and Oxford Multicenter Research Ethics Committee and the Norwich Local Research Ethics Committee. Currently, of 1,785 eligible patients, 1,019 women have agreed to take part. The first 864 cases (with 864 controls) were available for this analysis. DNA was extracted from blood samples by Whatman International Ltd. (Ely, United Kingdom).

The MALOVA study is a population-based, Danish case-control study of ovarian cancer. Eligible cases were women ages 35 to 79 years diagnosed with an ovarian tumor from December 1994 to May 1999. The study included 18 different hospitals from the municipalities of Copenhagen and Frederiksberg as well as the counties of Copenhagen, Frederiksborg, Roskilde, Vestsjælland, Storstrøm, Funen, southern Jutland, and northern Jutland. In total, 692 women with invasive epithelial ovarian cancer, 225 with ovarian borderline tumors, and 1,088 with benign ovarian tumors were enrolled. Controls were drawn from the general female population within the study area (ages 35-79 years) selected at random using the computerized Central Population Register. In total, 2,116 controls were recruited into the study. The study was approved by the local scientific ethical committee. DNA was extracted from blood using the QIAamp DNA Blood Mini Kit (Qiagen, Valencia, CA). Samples from 443 cases and 1,130 controls were available for this study.

The U.S. subjects were ascertained from the Family Registry for Ovarian Cancer of six counties in northern California. Included were ovarian cancer cases (n = 323) from Caucasians (ages 20-64 years) diagnosed between March 1, 1997 and July 31, 2001 and age/racially matched healthy female controls (n = 427). Buccal rinses were obtained from 83 cancer cases and 55 controls and bloods from all other subjects. DNA was purified from peripheral blood leukocytes using the Puregene Kit (Gentra Systems, Minneapolis, MN). DNA was isolated from exfoliated cells in buccal mouthwash rinses as described previously (25).

Genotyping was carried out using Taqman according to manufacturer's instructions. Each assay was carried out using 10 ng DNA in a 5 μL reaction with primers at 900 nmol/L and probes at 200 nmol/L concentrations. Primer and probe sequences and assay conditions used for each polymorphism analyzed are shown in Table 1. All reactions were carried out using 384-well arrays with 12 duplicate samples in each plate for quality control. There were no discordant genotypes in duplicates. Genotypes were called using the Allelic Discrimination sequence detection software (Applied Biosystems, Warrington, United Kingdom). DNA samples that did not give a clear genotype result at the first attempt were not repeated because this is a high-throughput process. Hence, there are variations in the number of samples successfully genotyped for each polymorphism.

Genotyping of DNA from U.S. subjects for F31I and V57I was independently corroborated by pyrosequencing. The pyrosequencing protocol is available by request from the corresponding author.

In the controls, deviations of the genotype frequencies from those expected under Hardy-Weinberg equilibrium were assessed by χ² tests (1 df). Genotype frequencies in cases and controls were compared for each study separately using χ² tests (2 df). Data were pooled and genotype frequencies were compared in cases and controls using unconditional logistic regression with terms for genotype and study and an appropriate likelihood ratio test. Genotype-specific risks with the common homozygote as the baseline comparator were estimated as odds ratios (OR) by unconditional logistic regression. We also tested for rare allele dominance using χ² tests (1 df) for each study separately and unconditional logistic regression for the pooled data. Haplotype frequencies were estimated using the hapipf command implemented in STATA (StataCorp, College Station, TX). This function calculates allele/haplotype frequencies using log-linear modeling embedded within an expectation maximization algorithm (26). The expectation maximization algorithm handles the phase uncertainty and the log-linear modeling allows testing for disease association. STATA also allows haplotype frequencies to be compared between cases and controls while allowing for differences in haplotype frequencies by study using a likelihood ratio test.

Genotype frequencies of all four polymorphisms in each control group (United Kingdom, United States, and Denmark) were consistent with Hardy-Weinberg equilibrium (Table 2). There were no significant differences in the genotype frequencies for F31I between cases and controls either in the individual studies or when the data were pooled (Table 2). Genotypic-specific risks of F31I for the individual studies did not differ significantly from unity, except for rare allele (I31) carriers in the U.S. group (Table 3). There was no evidence of heterogeneity in risk between studies. When the data were pooled, the risk to FI heterozygotes was significantly increased [OR, 1.18; 95% confidence interval (95% CI), 1.01-1.36]. The risk to II homozygotes was also increased but not significantly (OR, 1.16; 95% CI, 0.84-1.60), suggesting a dominant mode of action. Under the dominant model, there was a significant association of I31 with ovarian cancer after pooling the data (P = 0.03) with an increased risk to I31 carriers (OR, 1.17; 95% CI, 1.02-1.35). There were no significant differences in genotype frequencies of either V57I, 5′ UTR c/g, or 3′ UTR g/c between cases and controls (Table 2) and all the genotypic-specific risks were close to unity (Table 3). There was no association of age with genotype in controls (F31I, P = 0.27; V57I, P = 0.94; 5′ UTR, P = 0.94; 3′ UTR, P = 0.68). Moreover, to confirm that the association of genotype with risk was not confounded by age, all genotype risks were unchanged after adjustment for age (data not shown).

Haplotype analysis revealed four haplotypes with a frequency above 0.05 with another three haplotypes with frequencies between 0.02 and 0.05 (Table 4). There were no significant differences in haplotype frequencies in cases and controls (likelihood ratio test P = 0.34) and none of the haplotype-specific risks was significantly different from unity (Table 4).

The known highly penetrant, rare cancer predisposition alleles, such as germ line mutations in the BRCA1 or BRCA2 tumor suppressor genes, are estimated to account for <10% of all ovarian cancer cases and <30% of the excess familial risk of ovarian cancer (27). It is likely that the unexplained heritable component of ovarian cancer susceptibility is due to multiple weakly penetrant alleles.

We have found some evidence that the I31 allele of STK15 is associated with a modest risk of invasive ovarian cancer acting in a dominant manner. However, this result needs to be interpreted with some caution. Firstly, the overall comparison of genotype frequencies in cases and controls was not significant, even at the 5% level. Furthermore, the test for a dominant effect was only marginally significant (P = 0.03 in combined data), so the possibility of a type I statistical error must be considered. Some authors have suggested that stringent criteria should be applied to statistical tests for genetic association (e.g., P < 0.0001) because of the large number of candidate polymorphisms across the human genome. However, 9,000 cases and 9,000 controls would be needed to detect a dominant allele with risk of 1.17 with 80% power at this level of significance. Hidden population stratification is an alternative explanation for a spurious association. This occurs when allele frequencies differ between population subgroups and cases and controls are drawn differentially from those subgroups. However, it seems unlikely that population stratification is relevant in this investigation because the cases and controls in the three studies reported here were drawn from the same ethnic groups. Furthermore, the existence of significant population stratification that has resulted in a false genetic association has never been empirically shown (28).

On the other hand, STK15 is a good candidate for ovarian cancer susceptibility. Moreover, of the four polymorphisms tested (F31I, V57I, 5′ UTR c/g, and 3′ UTR g/c), F31I was the only one with a significant result and is the one most likely to alter protein function. F31I is a structurally nonconservative, aromatic/aliphatic amino acid substitution and occurs in an evolutionarily conserved region of the NH₂-terminal domain of STK15 (4). The NH₂-terminal domain has been proposed to function in translocation of STK15 from cytoplasm to centrosome during mitosis (7). Recently, it was reported that the I31 isoform of STK15 bound less efficiently than the F31 isoform to UBE2N, an ubiquitin-conjugating enzyme that colocalizes with STK15 at centrosomes during mitosis (14). Moreover, the F31 isoform and UBE2N colocalized to centrosomes more efficiently than the I31 isoform and UBE2N. Conceivably, reduced efficiency in subcellular localization of the I31 isoform of STK15 during mitosis may affect normal chromosome segregation, resulting in genomic changes in daughter cells that lead to cell transformation and tumor formation. V57I also occurs in an evolutionarily conserved region of the NH₂-terminal domain of STK15. The effect of this polymorphism on function has not been investigated. However, V57I is a structurally conservative, aliphatic/aliphatic amino acid change and would not be predicted to have a major effect on protein function. Furthermore, we found no evidence from the haplotype analysis that a specific haplotype tagged by a combination of these four single nucleotide polymorphisms was associated with risk and therefore might be acting as a marker for another important single nucleotide polymorphism in linkage disequilibrium.

Clinical genetic testing for cancer is currently limited to genes with the rare, highly penetrant, cancer predisposition alleles such as BRCA1 and BRCA2. Testing for alleles such as F31I is not indicated because the increase in disease risk (if confirmed) is low and the predictive value of the genotype is poor. However, if enough weakly penetrant alleles are eventually identified to account for substantial increased risk, then genotyping a panel of such loci may be useful for risk determination (29). Individuals with a panel genotype profile of increased risk would be candidates for cancer prevention measures or for frequent surveillance for early detection of disease. Current prevention measures for ovarian cancer include periodic oral contraceptive use prior to menopause and prophylactic oophorectomy after child-bearing years (30–33). Present early detection methods for ovarian cancer such as testing for CA125 antigen in blood and ultrasound lack sensitivity and specificity for routine use in the general population but may be effectively employed in women with increased disease risk (31).

We thank David S. Chervinsky, Roger L. Eddy, Hai Shen Chen, and John Mikula for expert technical assistance; Joan MacIntosh, Hannah Munday, Barbara Perkins, Clare Jordan, Kristy Driver, and the East Anglian Cancer Registry for recruitment of the United Kingdom cases; the EPIC-Norfolk Investigators for recruitment of the United Kingdom controls; and all the study participants who contributed to this research.