This article is a US Government work and, as such, is in the public domain of the United States of America.
Cervical cancer (CxCa) is caused by persistent human papillomavirus (HPV) infection; genetic predisposition is also suspected to play a role. Our study is a targeted candidate gene follow-up based on: (i) strong clinical evidence demonstrating that mutations in the TMC6 and TMC8 (EVER1 and EVER2) genes associate with the HPV-associated disease epidermodysplasia verruciformis (EV) and (ii) recent epidemiological data suggesting a genetic susceptibility conferred by polymorphisms in such genes for skin and CxCa. Clarifying the association of the TMC6/8 genes with risk of CxCa will help in understanding why some HPV-infected women develop persistent infection, cervical lesions and eventually cancer while others do not. Twenty-two single nucleotide polymorphisms (SNPs) harboring the TMC6/8 genes were genotyped in 2,989 cases with cervical intraepithelial neoplasia grade III or invasive CxCa and 2,281 controls from the Swedish population. Association was evaluated in logistic regression models. Two SNPs displayed association with cervical disease: rs2290907 [odds ratio (OR)GGvsAA = 0.6, 95% confidence interval (95% CI): 0.3–0.9, p = 0.02)] and rs16970849 (ORAGvsGG = 0.8, 95% CI: 0.66–0.98, p = 0.03). The present data support the involvement of the TMC6/8 region in CxCa susceptibility but further analyses are needed to replicate our findings, fully characterize the region and understand the function of the genetic variants involved.
Cervical cancer (CxCa) is the second most common cancer in the world among women and caused by persistent infection by oncogenic mucosatropic human papillomaviruses (HPV)1 of the alpha genus. Although HPV infection is a necessary cause of CxCa, it is not sufficient and the majority of infected women never develop cancer. It has been shown that genetic susceptibility factors are also important,2 and the association of genes in immunological pathways with susceptibility to CxCa has been demonstrated in several studies.3–6
Mutations in the TMC6 and TMC8 genes, also known as the EVER1 and EVER2 genes, are implicated in the development of the rare autosomal recessive disease epidermodysplasia verruciformis (EV).7–9 EV is characterized by extreme susceptibility to infection by cutaneous HPV of the beta genus.10 A polymorphism in the TMC8 gene has also been associated with increased risk of squamous cell carcinoma of the skin and the presence of HPV antibodies in control subjects.11 The TMC6/8 genes may be involved in controlling HPV gene expression and replication in epidermal keratinocytes or they may directly affect innate and adaptive immune responses which control the clearance of HPV-infected keratinocytes in EV cases.12
The link between the TMC6/8 genes and vulnerability to HPV makes these genes interesting candidate genes in CxCa. The genes are located adjacently on chromosome 17q25. A recent report by Wang et al.13 indicated an association between genetic variation in the TMC6/8 region and persistent HPV infection as well as risk of CxCa. The findings warrant replication in larger studies and in different populations. There is also a need for further investigations of genetic variation in the TMC6/8 genes to understand the possible mechanisms of susceptibility to persistent HPV infection and development of CxCa.
Our study aims to evaluate the association of single nucleotide polymorphisms (SNPs) in the genomic region harboring the TMC6 and TMC8 genes with susceptibility to CxCa. A previous observation13 indicated that the TMC6/8 genes may influence risk of CxCa, and our study provides comprehensive data on the association of polymorphisms in the TMC6/8 region with CxCa in the Swedish population.
Material and Methods
Our study comprises two Swedish cohorts, here denoted as cohorts 1 and 2, and the total sample set consisted of 2,989 cases and 2,281 controls. Of the cases, 2,767 (93%) were diagnosed with cervical intraepithelial neoplasia grade III (CINIII) and 222 (7%) with invasive cervical carcinoma (ICC).
Cohort 1 was based on a nested study within the Northern Sweden Health and Disease Study Cohort and has been previously described.6 It is derived from the Swedish province of Västerbotten, which comprises the Monitoring of Trends and Determinants in Cardiovascular Disease (MONICA) Project and the Västerbotten Intervention Program. The MONICA Project contains material from population-based screenings for cardiovascular disease that was carried out in 1986, 1990, 1994, 1999 and 2004. There were approximately 14,000 sampling occasions for approximately 9,000 persons, 50% of whom are included in the Västerbotten Intervention Program. Established in 1985, the Västerbotten Intervention Program is a long-term health promotion project that by 2004 included approximately 74,000 persons, of whom approximately 70,000 had donated blood. The samples are collected annually from all residents aged 40, 50 and 60 years in Västerbotten County and are stored as plasma, erythrocytes and buffy coats at −80°C at the university hospital of the University of Umeå, Umeå, Sweden. CxCa cases were traced through the Swedish Cancer Registry. For the current investigation, 1,002 cases (867 CINIII and 135 ICC) and 1,734 controls were included. All controls were selected from female subjects alive and free of CxCa at the time of diagnosis of the case in each respective subcohort, and matching the index case on age at diagnosis (±5 years). Mean ages of the cases and controls were 48.0 years (range, 20.1–69.7) and 49.6 years (range, 26.3–70.0), respectively. Genomic DNA was extracted from buffy coats using standard procedures, and the study was approved by the Institutional Review Board of the Umeå University.
Cohort 2 has also been described previously14 and included 1,987 cases from families with at least two affected women (1,900 CINIII, 87 ICC) from all of Sweden. Our study includes 942 individuals with one first degree relative (mother, sister or daughter) and 39 individuals with two first degree relatives also participating in the study. The remaining 1,006 individuals were the sole participants of their family. The families were originally identified by cross-linking the Swedish Cancer Registry and the National Family Registry. The registry identified 5,400 individuals from families with at least two affected. A total of 4,145 individuals were invited, and blood was available from 2,135 (51%).2, 14 All cases were from the Swedish population. To avoid any potential overlapping of individuals between cohorts, all participants in cohort 2 coming from the region of Västerbotten (Umeå pathology department) were excluded. Cohort 2 also included 547 unrelated controls (265 male and 282 female). These were blood donors at Uppsala University Hospital (n = 257) and healthy nonobese adolescents from a Swedish obesity study (n = 289).15 The controls were of Swedish descent (both parents of Swedish origin). DNA was extracted from peripheral blood using standard methods. The study was approved by the regional ethical review board in Uppsala, and all cases provided written informed consent.
The study was designed to capture variations across the 58 kb genomic region on chr. 17q25 harboring the TMC6 and TMC8 genes (17:73600594–73659033). The region was defined as 20 kb upstream from TMC6 gene and 8.4 kb downstream of the TMC8 gene. Tag SNPs were selected using data from the HapMap CEU (Utah residents with Northern and Western European ancestry from the Centre d'Etude du Polymorphisme Humain [CEPH] collection) in The International HapMap Project (Rel24/phase II Nov08, on The National Center for Biotechnology Information (NCBI) B36 assembly) and the Tagger algorithm in Haploview.16, 17 Aiming for 100% coverage of the region resulted in 25 selected SNPs, which tag a total of 68 SNPs (Supporting Information Fig. 1). A multiplex genotyping assay was developed according to previously described methods.6 Briefly, the assay was based on newly designed multiplex polymerase chain reaction (PCR) and hybridization using allele-specific probes coupled to fluorescently labelled polystyrene beads in a Luminex system. Genotypes were assigned by plotting the ratio of both hybridization signals (expressed as median fluorescence intensity values) of the allele-specific probe pairs for each tested sample using the software Spotfire 2.1 (TIBCO Software, CA). The accuracy of the assay was validated using a HapMap reference panel of 91 samples. In total, 22 SNPs were successfully included in the assay (Table 1) resulting in 88.5% coverage of the genetic variation in the target region. To ensure robustness of genotyping, 192 samples were typed twice. Additional quality control samples such as negative PCR controls (no template) and CEPH reference samples were randomly distributed among the study samples. Genotyping of samples from both cohorts was performed simultaneously in one single laboratory.
Table 1. SNP genotyped in our study
Basic statistical analysis was performed using SAS 9.1.3 (SAS Institute, Cary, NC) and the analysis tool PLINK (http://pngu.mgh.harvard.edu/purcell/plink/).18 Initial quality control included calculating the genotyping success rate for the SNPs and individuals and assessing the Hardy–Weinberg equilibrium in controls with p = 0.05 as threshold. Testing for heterogeneity among the two cohorts was performed using the Breslow–Day test, as described previously.19
Risk estimates were computed as genotype odds ratios (ORs) with 95% confidence intervals (95% CIs) in unconditional logistic regression models. To account for differences between cohorts, multivariate analysis was performed including cohort as a categorical covariate in the model. In addition, analysis was restricted to cases diagnosed with ICC (n = 222) versus female controls. Haplotype blocks were identified as per the definition of Gabriel et al.,20 using Haploview software.17 Then haplotype analyses were performed in SAS/GENETICS using the PROC HAPLOTYPE procedure, where all haplotype blocks were tested adjusting by cohort. This procedure uses the expectation maximization algorithm to predict the maximum likelihood estimates of the haplotype frequencies assuming Hardy–Weinberg equilibrium.
The statistical power to find genetic susceptibility factors was increased by inclusion of cases from a family cohort, as these samples are more likely to share such factors than randomly selected sporadic cases. However, the use of related subjects may lead to inflated associations simply because of family structure. The software tool PedGenie provided in Genie 2.7.221, 22 can be used to accommodate a study design which combines data from related and unrelated individuals. PedGenie calculates empirical p values and CIs corrected for family structure based on N simulations. The input to PedGenie is genotype data and pedigree information describing the family structure. Unrelated individuals are entered as being the only member of a family. PedGenie calculates allele frequencies based on genotype data and assigns genotypes to all founders accordingly. For each simulation, a genotype distribution is simulated based on founder genotypes and then combined with the true phenotype to calculate the simulated statistic. Summarizing all simulated statistics that are larger than the observed statistic and dividing by N gives p values and CIs corrected for family structure. In our study, PedGenie was used to investigate the influence of family structure in a separate analysis of the Uppsala cohort. Empirical 95% CI for genotype-based OR was calculated by 10,000 gene-drop simulations in PedGenie and compared to the 95% CI from standard unadjusted analysis. The empirical and the unadjusted p values were comparable; therefore, no correction for relatedness was implemented in the pooled analysis presented in the following sections (data not shown).
The aim of our study was to analyze a candidate region previously implicated in CxCa susceptibility;13 therefore, no correction for multiple testing was applied, and p < 0.05 was considered statistically significant.
The robustness of genotyping was evaluated by repeated typing of 192 samples yielding >99% consistent genotypes. Analysis of the genotyping success rate resulted in removal of 124 (2.4%) samples due to <90% call rate. In the remaining samples, all SNPs were successfully genotyped, and the total genotyping rate was 99.5%. None of the SNPs displayed deviations from Hardy–Weinberg equilibrium in controls of the combined and the separated cohorts.
Large p values for the Breslow–Day test (between 0.07 and 0.97, data not shown) indicates no significant cohort differences for all the tested SNPs, except for the SNP rs3813026 (p =0.01). In addition, to avoid possible heterogeneity among the two Swedish cohorts, CxCa risk estimates for all markers were calculated by logistic regression modeling with adjustment for cohort (Table 2). Two markers remained significantly associated with CxCa after controlling for cohort differences: rs2290907 (ORGGvsAA = 0.6, 95% CI: 0.3–0.9, p = 0.02) and rs16970849 (ORAGvsGG = 0.8, 95% CI: 0.66–0.98, p = 0.03).
Table 2. Polymorphisms associated with CxCa (CINIII + ICC) in combined and adjusted analysis
Figure 1 shows the location of all genotyped SNPs on chromosome 17. The two associated SNPs are located in the intronic regions of the TNRC6C (SNP rs2290907) and TMC8 genes (SNP rs16970849).
When the analysis was restricted to cases diagnosed with ICC (n = 222), the ORs for the two SNPs associated with overall disease were similar, although statistical significance was not reached due to lack of power: rs2290907 (ORGGvsAA = 0.2, 95% CI: 0.03–1.5, p = 0.1) and rs16970849 (ORAGvsGG = 0.8, 95% CI: 0.5–1.3, p = 0.3). Risk estimates for CXCa and ICC did not differ when analysis was done excluding males from the controls group of cohort 2 (data not shown). Haplotype analysis did not show any additional association (data not shown).
The aim of our study was to evaluate genetic variation in the genomic region harboring the TMC6 and TMC8 genes with susceptibility to CxCa. Our approach was a targeted candidate gene follow-up based on: (i) strong clinical evidence that demonstrates a genetic predisposition to EV acquisition in patients carrying mutations in these genes and (ii) epidemiological data coming from skin and CxCa studies suggesting the involvement of polymorphisms in the mentioned genes. Clarifying the potential association of the TMC6/8 genes with risk of CxCa will help in understanding why some HPV-infected women develop persistent infection, cervical lesions and eventually cancer, while others do not.
Some evidence for an association between TMC6/8 genes and CxCa was observed in the Swedish population. The TMC6/8 region on chromosome 17 displayed limited linkage disequilibrium indicated by the fact that 25 SNPs were needed to completely capture the variation provided by 68 SNPs. When results were adjusted by cohort, two markers located in the TNRC6C (rs2290907) and TMC8 (rs16970849) genes remained associated with disease. The effects of the associated SNPs were modest but consistent in analyses restricted to invasive cancers and controls.
The present work is the largest study to date to investigate the role of EVER polymorphisms in CxCa susceptibility. It comprises a total sample size of 2,989 cases and 2,281 controls from a single population. We estimate to have covered around 92% of the genomic variation in the TMC6/TMC8 region. An additional benefit is the inclusion of cases from a family study enriched for genetic susceptibility factors. The overall genotyping performance was good for all tested markers; call rates were >99.7% for the two CxCa-associated SNPs in our study.
The present investigation was instigated by the report of Wang et al.12 who studied common genetic variants in 305 candidate genes and reported an association in the TMC6/8 region. In total, they investigated 26 SNPs in this region in 416 cases and 425 controls from a population-based cohort in Guanacaste, Costa Rica. They found that EVER region has a statistically significant association with progression to CINIII/cancer, and inside the region one particular SNPs, the rs9893818 displayed the strongest association (ORACvs.CC = 2.76 95% CI: 1.59–4.77, ptrend = 0.0003). This SNP was not included in our study as it is monomorphic in HapMap-CEU population. The SNP rs9893818 is located downstream the TMC8 gene and is approximately 8 kb from our associated SNP rs16970849, which is located in the TMC8 gene. Examination of this region in HapMap-CEU and The Yoruba in Ibadan, Nigeria (YRI) populations revealed no marker in high or even moderate linkage equilibrium (LD) with rs9893818. Furthermore, we have observed among the CEU and YRI populations the formation of five haplotypes blocks along the TMC6/8 region, which are overlapped among both populations. Interestingly, the marker associated in the Costa Rican population lie outside of these extended block haplotypes (Supporting Information Fig. 1). Overall, 13 SNPs from the previous study were typed in our study. As expected, the allele frequencies for these SNPs vary between the studies. The populations are ethnically distinct, and only four variants displayed less than 5% allele frequency difference among controls.
HPV16 as well as other high-risk HPV are recognized as causal factors for the development of CxCa. Recent studies suggest a direct link between TMC gene products and the HPV E5 protein. The TMC6 and TMC8 proteins form a complex with the Zn2+ transporter 1 (ZnT-1) in the ER membrane. HPV16 E5 binds to the TMC/ZnT complex and prevents complex-mediated inhibition of the Zn-sensoring transcription factor MTF-1 and c-jun activity.23 Following a model proposed by Lazarczyk et al.,12 the TMC proteins constitute a cellular barrier for HPV replication. The interaction of the E5 protein of mucosal alpha-papillomaviruses like HPV16 with the TMC/ZnT complex in keratinocytes would result in increased intracellular free Zn2+ levels and increased activity of transcription factors such as AP-1 which would then lead to increased viral replication. Genetic polymorphisms in the TMC6/8 genes that result in amino acid changes might affect TMC protein function, or changes in regulatory regions might affect expression levels, RNA stability or splice variants. Either change could affect HPV replication and thereby the viral load.
In a recent population-based case–control study of head-and-neck cancer in the greater Boston-MA area, homozygotes for the T allele of TMC8 SNP rs7208422, encoding an amino acid change from Asn to Ile at position 306, was found to be associated with seropositivity to HPV 16L1 (ORTT: 0.49, 95% CI 0.27–0.86) and also to be protective against head-and-neck cancer (ORAT: 0.70, 95% CI 0.52–0.93; ORTT: 0.62, 95% CI 0.44–0.86; Nelson et al., unpublished data). In contrast with this report, but in agreement with Wang et al., we did not find any statistically significant association of this SNP with CxCa after adjusting by cohort. The discordant findings may be explained by differences on the rs7208422 allele frequencies among different studies and ethnic groups. If the effects of this polymorphism vary with HPV type, then analysis stratified by HPV type might reveal associations with CxCa susceptibility. However, one of the limitations of our study is the lack of HPV data among cases to allow for analysis stratified by HPV type or phylogenetic groups. Although HPV information exists for about 50% of the cases, the number of cases positive for HPV16 (the most common HPV type detected in CxCa) was not sufficient to allow for a proper evaluation of effects by HPV status of cases.
The EVER SNPs found associated with CxCa in our study and their tagged variants all are located in intronic regions or 3′-untranslated sequences and their putative functional consequences are currently unknown. The variant rs1048591 located in the 3′-UTR of TNRC6C and the variant rs16970849 located in an intron of TMC8 tag several other SNPs (Supporting Information Fig. 1) located also in intronic regions of the TMC genes. Although, there is evidence that TNRC6C gene plays a role in RNA-mediated gene silencing by micro-RNAs (miRNAs),24–26 there is no observation for a possible participation of this gene in cancer development.
Several steps have been taken to account for differences among the two pooled Swedish cohorts in our study. (i) We performed all analysis by using a Monte Carlo significance testing (metagenetic) which did not show any differences on the risk estimates for all associated and nonassociated SNPs. (ii) We tested for heterogeneity among the cohorts by using the Breslow–Day test. This analysis did not result in any significant heterogeneity for the included SNPs, except for the SNP rs3813026 (p =0.01), which showed no association with disease. (iii) We calculated all risk estimates adjusting for Cohort differences. Nonetheless, we acknowledge the need for caution in interpreting our findings, and the need for additional replication efforts before the association between EVER genes and CxCa is considered established.
In conclusion, our observations support the involvement of the TMC6/8 region in susceptibility to CxCa. Caution in the statistical analysis for genetic studies is needed even when studying very homogeneous populations such as ours. More work such as deep sequencing of the region is required to establish the full spectrum of variation and identify the functional variants. Future studies would benefit from including HPV type in the analysis. It would also be of interest to investigate the contribution of associated variants in other HPV-related cancers and to consider the association of TMC6/8 variation with HPV antibody responses. The latter would be relevant not only for clarifying the observed associations of TMC6/8 genes with HPV seropositivity against alpha and beta-types but also for understanding the mechanisms of these genes in HPV infection.
The authors acknowledge the valuable work of Ms. Åsa Agren from the Department of Public Health and Clinical Medicine/Nutritional Research, University of Umeå, Umeå, Sweden, in keeping track of the samples and data handling. They also thank Prof. Matti Lehtinen of Tampere University, Tampere, Finland, for his advice during the application for Umeå samples. This work was supported by the Swedish Cancer Foundation to U.G.