Association of HLA-DRB1, interleukin-6 and cyclin D1 polymorphisms with cervical cancer in the Swedish population—A candidate gene approach

Authors


Abstract

High-risk human papillomavirus (hrHPV) infection is the major risk factor for cervical cancer (CxCa). The role of genetic susceptibility in the disease has been suggested, but the existing data lack consistency. We conducted a nested case-control study on 973 CxCa cases and 1,763 matched controls, from two Swedish population-based cohorts to examine the association of common genetic variants with CxCa risk. Human leukocyte antigen (HLA) alleles and 24 other polymorphisms in 14 genes were selected on the basis of reported association or mechanistic plausibility with an HPV infection or cervical cancer development. Genotyping was conducted using multiplex PCR and Luminex technology. A significant association of CxCa with various polymorphisms was observed: rs1800797 in the IL-6 gene (odds ratio [OR] = 0.88, 95% confidence intervals [CI]: 0.79–0.99); rs1041981 in the LTA gene (OR = 0.87, 95% CI: 0.78–0.98), and rs9344 in the CCND1 gene (OR = 1.14, 95% CI: 1.02–1.27), for those individuals carrying the rare allele. Additionally, the alleles 0401 and 1501 of the HLA class II DRB1 locus were associated with an increased risk (OR = 1.23, 95% CI: 1.04–1.45 and OR = 1.29, 95% CI: 1.11–1.50, respectively), and allele 1301 was associated with decreased risk (OR = 0.59, 95% CI: 0.47–0.73). The effects of CCND1 and the HLA*DRB1 alleles were independent of the effect of smoking. We did not find any association of risk with polymorphisms in genes related to the innate immune system. In conclusion, our study provides evidence for genetic susceptibility to CxCa due to variations in genes involved in the immune system and in cell cycle. © 2009 UICC

Cervical cancer (CxCa) is the second most common cancer in women worldwide. It accounts for 250,000 deaths per year and at least 80% of them occur in developing countries.1, 2 Abundant experimental and epidemiological evidence indicates that infection with high-risk human papillomavirus (hrHPV) is the major risk factor of CxCa.3 hrHPV infection is one of the most common sexually transmitted diseases and most women in the world are infected at least once with hrHPV.4 Most infections are, however, cleared by the immune system, and only a minority persist and develop to CxCa,5, 6 suggesting a complex host–virus relationship.

Genetic factors contribute to the susceptibility of acquiring cervical intraepithelial neoplasia and invasive CxCa. This is indicated by familial aggregation for CxCa7, 8 and heritability estimates that are comparable with ovarian and breast cancers.8 Although the heritability estimates differ between studies and familial aggregation is partially due to environmental factors shared between relatives, it is suggested that around 60–70% of the familial risk is attributable to a heritable component.9

Several studies, mainly based on a candidate gene approach, have investigated genetic markers in various biological pathways of cervical carcinogenesis (immune response, cell cycle and metabolic processes). Currently, more than 74 genes have been reported to be associated with CxCa (http://www.hugenavigator.net). Significant associations with HLA alleles DRB1*1501, DRB1*13, DRQB1*0602 have been verified in meta-analyses.10, 11 In addition, immunohistochemical studies have demonstrated downregulation of HLA molecules in neoplastic cervical cells.12 Other immune-related genes may, however, also contribute to the genetic susceptibility to cervical cancer. These include interleukin-6 (IL-6),13 interleukin-4 receptor (IL4-R),14 interleukin-10 (IL-10),15 cytotoxic T-lymphocyte antigen 4 (CTLA-4),16FAS (CD95 or APO-1),17 lymphotoxin-alpha (LTA),18 tumor necrosis factor-alpha (TNF-alpha).19

Exploiting the high-throughput, low-density array Luminex platform to study a large cohort of individuals representing homogeneous and well-established population-based biobanks from Sweden, we aimed at a comprehensive investigation of genetic markers located in the immune-related genes. In addition, cell cycle genes (TP-53, cyclin-dependent kinase inhibitor 1A (P-21 or CDKN1A) and cyclin D1 (CCND1)) were studied to improve our knowledge on the genetic susceptibility to CxCa, and to reassess previous contradictory or nonconfirmed findings.

Material and methods

Study population

We conducted a case-control study nested within The Northern Sweden Health and Disease Study Cohort through 2 subcohorts: the Västerbotten Intervention Program (VIP) and the Mammography Screening in Vasterbotten (http://www.p3gobservatory.org/catalogue.htm?studyId=566). The VIP is a long-term project intended for health promotion of the population of Västerbotten. Since 1985 all individuals 40, 50 and 60 years of age in the population of the county are invited for screening and the cohort included by April 2007, 99,000 sampling occasions from 77,000 unique individuals. Since 1995 the VIP material is supplemented with blood samples and questionnaires from the mammary screening program (29,000 unique individuals and 54,500 sampling occasions). From 1997, repeated screening was started within the mammary screening program with sampling every second year in the age group 50–69 years within the county. Therefore, both cohorts belong to the same ethnically homogenous population.20, 21 The study population consisted of 2,736 participants coming from the 2 population-based cohorts. A total of 973 cases defined as women with invasive cervical carcinoma (196) or cervical intraepithelial neoplasia grade 3 (777) as the primary cancer diagnosis among the participants of the 2 cohorts were traced through automatic record linkages with the Swedish Cancer Registry. Additionally, 1,763 controls selected from the same cohort as the corresponding case were also included in the study. Cases were matched to the controls on age at diagnosis (±5 years). At the time of the cases diagnosis, the controls had to be alive and without any previous cancer diagnosis. 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 in the biobank laboratory using standardized protocols. The DNA samples of cases and controls were randomly divided on 96-well plates and the laboratory personnel were blinded to the case-control status throughout the study. The study was approved by the Institutional Review Board of the Umeå University.

Genotyping of markers

Single nucleotide polymorphisms (SNP) included in this study (Table I) had an estimated minor allele frequency of ≥4% in Caucasians (The National Center for Biotechnology Information) and they were selected from reported candidates assumed to be mechanistically associated with an HPV infection or cervical cancer development. SNP typing was done by a newly designed multiplex PCR and subsequent hybridization. A total of 10 ng of human genome DNA were used per PCR reaction using the multiplex PCR kit (Quiagen, Hilden, Germany). As one of the primers in a primer pair was 5′-biotinylated, amplimers with a size of 100 to 200 bp could be detected by subsequent hybridization to probes coupled to fluorescently labelled polystyrene beads (Luminex Corp., Austin, TX) with conditions described recently.22, 23 For each allele of the selected SNP, a specific probe was designed by locating the polymorphism in the middle of an 17- to 18-nt-long oligonucleotide. After hybridization, the complexes were labeled by Strep-PE conjugate, washed and analyzed in the Luminex 100 analyzer. The Luminex 100 analyzer contains 2 lasers to identify the bead set by the internal bead color and to quantify the reporter fluorescence on the bead surface. Results are expressed as the median fluorescence intensity (MFI) of at least 100 beads analyzed per set and per reaction. The assay was calibrated with a reference panel of 90 DNA samples from the European population cohort of the HAPMAP Project,24 which have been previously genotyped (data not shown). Genotypes were assigned by plotting the ratio of both MFI values of the specific allele probes for each tested sample using the software Spotfire 2.1 (TIBCO Software, CA). Rates of genotyping completion ranged between 93.7 and 100%. For quality control, 10% of randomly selected samples containing both cases and controls were reanalyzed by the matrix-assisted laser desorption/ionization time-of-flight (Maldi-TOF) mass spectrometry and the iPLEX Gold Assay (Sequenom, Hamburg, Germany). In this validation assay, less than 1.2% discrepancies were observed.

Table I. Polymorphisms Evaluated in this Study
GeneNucleotide allele major/minorrs numberVariation1Minor allele frequency in controls
  • Rare alleles, HLA-DRB1 alleles 0102, 0103, 0305, 0402, 0403, 0405, 0407, 0408, 0410, 0802, 0803, 0810, 0901, 1001, 1101, 1102, 1103, 1104, 1201, 1303, 1401, 1402, 1502 and 1601.

  • 1

    Amino acid exchange, position in the promoter or intron.

Immune response:
 CTLA-4T>Crs5742909Promoter0.07
 CTLA-4A>Grs231775T17A0.49
 FAST>Crs1800682Promoter0.44
 FASG>Ars2234767Promoter0.10
 IL-10A>Grs1800896Promoter0.46
 IL-10C>Ars1800872Promoter0.26
 IL4-RA>Grs1805010V7510.47
 IL4-RT>Crs1805015S503P0.24
 IL4-RA>Grs1801275Q576R0.29
 IL-6G>Trs1554606intron 30.50
 IL-6G>Ars1800797Promoter0.46
 LTAC>Ars1041981T60N0.44
 TMC-8A>Trs7208422N306I0.47
Cell cycle:
 CCND1G>Ars9344P241P0.48
 P-21C>Ars1801270R31S0.04
 TP-53G>Crs1042522R72P0.30
Innate response:
 TIRAPC>Trs8177374S180L0.17
 TLR-3T>Ars5743305Promoter0.33
 TLR-3C>Trs3775291L412F0.34
 TLR-8A>Grs3788935Promoter0.24
 TLR-8A>Grs3761623Promoter0.46
 TLR-9T>Crs5743836Promoter0.14
 TLR-9G>Ars352139Intron 10.41
 TLR-9A>Grs352140P545P0.42
HLA-DRB1 alleles:
 DRB1*0101   0.08
 DRB1*0301   0.10
 DRB1*0401   0.13
 DRB1*0404   0.08
 DRB1*0701   0.05
 DRB1*0801   0.05
 DRB1*1301   0.11
 DRB1*1302   0.05
 DRB1*1501   0.17
 DRB1*rare alleles   ∼ 0.11

Genotyping of HLA DRB1 family

Medium resolution genotyping of the HLA-DRB1 alleles was done by the Finnish Red Cross Blood Service using the LABType kit and the Luminex technology (One Lambda, Los Angeles, CA), according to the manufacturer's instructions except that the volume of the PCR and the hybridization reactions were reduced by 50%. The alleles were analyzed with the HLA Visual software (One Lambda, Los Angeles, CA).

Statistical analysis

The observed genotype frequencies in the controls were tested for Hardy-Weinberg equilibrium. The difference between the observed and the expected frequencies was tested for significance using the χ2 test. The association of cervical cancer with the genetic markers was analyzed by conditional logistic regression and odds ratios (OR) with the 95% confidence intervals (CI) were estimated with and without adjusting for smoking status. We calculated a global p value (Global p) on the basis of the 3 or 2-level categorical genotype or allele variable (0, 1, 2 or 0, 1, respectively), and a p trend value on the basis of the 3-level ordinal genotype variable (0, 1, 2) in the logistic regression model. Analyses that account for the multiple comparisons were not done because of the selection criteria of the markers based on their reported association with HPV infection or CxCa development. For all genetic analyses, considering a p < 0.05 as statistically significant, our study had 85% power to detect an OR of 1.35 for a polymorphism with a risk genotype frequency higher than 10% (PS software for power and sample size calculation, http://biostat.mc.vanderbilt.edu/twiki/bin/view/Main/PowerSampleSize). Additional, stratified analyses based on disease severity were done to test the association of invasive cervical carcinoma (ICC) or cervical intraepithelial neoplasia grade 3 (CIN3) and the genetic markers. To limit the number of comparisons in the analysis, only markers that were significantly associated with CxCa in the whole population were included.

The selection of the genetic model (genotype, dominant, recessive and additive model) for the possible gene–gene or gene–environmental interactions was done by the negative log likelihood and statistical hypothesis testing. Additionally, for the gene–gene interactions between the SNPs and HLA-associated markers and the gene-smoking interaction, different levels of risk to cervical cancer were defined for each of the variables. Further stratified analyses were done by smoking and CxCa-associated HLA markers. All analyses were performed by using SAS (version 9.1; SAS Institute, NC). Haplotypes for the associated SNPs were constructed using the SAS/Genetics Software (SAS Institute 2002), which estimates haplotypes by using an expectation-maximization (EM) algorithm.

Results

The distribution of the genotypes for all but one polymorphism was in agreement with the Hardy-Weinberg equilibrium (HWE). The SNP rs1801270 was found to deviate slightly from HWE among controls (p = 0.04), therefore it was excluded from the analysis. The SNP description and the minor allele frequencies for each marker are shown in Table I. Three of the 24 SNPs investigated were found significantly associated with cervical cancer (CIN3+ICC) development (Table II). The SNP rs9344 (CCND1 gene) was significantly associated with an increased risk for CxCa, and the other 2 SNPs, rs1800797 (IL-6 gene) and rs1041981 (LTA gene), were associated with a decreased risk. Forthe HLA-DR locus, 2 alleles, DRB1-0401 and DRB1-1501, were significantly associated with an increased risk of CxCa (Table II). On the contrary, the DRB1-1301 allele was found to be a protective factor. Risk estimates for CxCa, genotype and allele frequencies of the other markers not associated with the disease are presented in the Supporting Information Table I.

Table II. Risk Estimates (Odds Ratio, or and 95% Confidence Intervals, 95% CI) for Cervical Cancer According to Genetic Markers Identified as Risk Factors
Risk factorModelControlsCasesOR95% CI2Global pp trend3
n1%n1%
  • 1

    Numbers for individual genotypes may not add up to the total because of missing data.

  • 2

    95% lower and upper confidence intervals.

  • 3

    p trend on basis of 3-level ordinal variable(0,1,2) of the genotypes in the model.

  • ng, frequency for this genotype equal cero; Hom, homozygous for that allele; Het, hetrozygous for that allele; –Other, a genotype formed by other alleles.

LTA-rs1041983 Genotypes       0.620.02
 AA33719.715416.20.750.600.95  
 CA81347.645647.90.920.771.10  
 CC55732.634135.91.00  
 Allele         
 A1,48743.676440.20.870.780.980.02 
 C1,92756.41,13859.81.00  
IL-6-rs18007973 Genotypes       0.10.03
 AA37021.618018.90.770.610.98  
 GA84149.047349.60.900.741.08  
 GG50429.430131.61.00  
 Allele         
 A1,58146.183343.70.880.790.990.03 
 G1,84953.91,07556.31.00  
CCND1-rs93443 Genotypes       0.080.03
 AA41124.026027.31.291.031.61  
 GA83748.946348.61.090.891.33  
 GG46527.122924.11.00  
 Allele         
 A1,6590.488930.491.141.021.270.03 
 G1,7670.529210.511.00  
HLA-DRB1*04013 Genotypes       0.040.01
 Hom-0401211.3192.11.640.863.12  
 Het-040139524.125627.81.231.021.49  
 Other1,22674.764670.11.00  
 Allele         
 Carrier4370.132940.161.231.041.450.01 
 Non-carrier2,8470.871,5480.841.00  
HLA-DRB1*13013 Genotypes       <.0001<.0001
 Hom-1301251.570.80.430.181.00  
 Het-130130218.410711.60.600.480.76  
 Other1,31580.180787.61.00  
 Allele         
 Carrier3520.111210.070.590.470.73<.0001 
 Non-Carrier2,9320.891,7210.931.00  
HLA-DRB1*15013 Genotypes       0.0020.0008
 Hom-1501472.9333.61.490.942.35  
 Het-150147128.731834.51.351.131.62  
 Other1,12468.557061.91.00  
 Allele         
 Carrier5650.173840.211.291.111.500.0008 
 Non-carrier2,7190.831,4580.791.00  

Smoking was strongly associated with CxCa. The risk for CxCa in individuals reporting to be former smokers or current smokers compared to non (never) smokers was about 2-fold (OR = 1.8 and 95% CI: 1.35–2.41; OR = 3.01 and 95% CI: 2.38–3.80), respectively (Table III). To stratify the genetic marker analysis by smoking status, 40.6% of cases and 38.8% of controls had to be excluded because of missing smoking history. The risk classification for smoking status was a baseline risk (never smokers), a medium risk (former smokers) and a high risk (current smokers). The point estimates found for the CCND1 polymorphisms and the 2 HLA alleles remained statistically significant and were even stronger after adjusting by smoking. However, for the IL-6 and LTA markers the significant association vanished. On the contrary, the SNP rs1801275 (IL4-R gene) emerged as a factor protective against CxCa (OR = 0.83 and 95% CI: 0.69–0.99) in an additive model (data not shown).

Table III. Adjusted Risk Estimates (Odds Ratio, or and 95% Confidence Intervals, 95% CI) for Cervical Cancer According to Genetic Markers Identified as Risk Factors
Risk factorsModelControlsCasesOR95% Cl2Global pp trend3
n1%n1%
  • 1

    Numbers for individual genotypes may not add up to the total because of missing data.

  • 2

    95% lower and upper confidence intervals.

  • 3

    p trend on the basis of 3-level ordinal variable (0,1,2) of the genotypes in the modeling, frequency for this genotype equal cero; Hom, Homozygous for that allele; Het, heterozygous for that allele; –Other, a genotype formed by other alleles.

Smoking3 Categories       <.0001<.0001
 Never64561.422038.81.00  
 Former16916.110418.31.801.352.41  
 Current23722.524342.93.012.383.80  
LTA-rs10419813 Genotypes       0.990.9
 AA19718.99116.10.980.701.37  
 CA50548.328550.40.990.771.28  
 CC34332.819033.61.00  
 Allele       0.9 
 A89943.046741.30.990.841.16  
 C1,19157.066558.71.00  
IL-6-rs18007973 Genotypes       0.20.08
 AA2152.511420.10.750.541.05  
 GA51749.227849.00.830.641.09  
 GG31830.317530.91.00  
 Allele       0.09 
 A94745.150644.60.870.741.02  
 G1,15354.962855.41.00  
CCND1-rs93443 Genotypes       0.010.007
 AA22921.816028.31.561.132.15  
 GA52450.027047.71.070.811.42  
 GG29628.213624.01.00  
 Allele       0.007 
 A98246.859052.11.251.061.47  
 G1,11653.254247.91.00  
HLA-DRB1*04013 Genotypes       0.080.05
 Hom-0401121.2122.22.531.006.41  
 Het-040125325.016229.51.200.921.56  
 Other74873.837568.31.00  
 Allele         
 Carrier27713.718616.91.240.991.560.06 
 Noncarrier1,74986.391283.11.00  
HLA-DRB1*13013 Genotypes       0.0050.001
 Hom-1301151.530.50.190.040.90  
 Het-130118618.45910.70.650.460.92  
 Other81280.248788.71.00  
 Allele         
 Carrier21610.7655.90.570.410.790.0006 
 Noncarrier1,81089.31,03394.11.00  
HLA-DRB1*15013 Genotypes       0.030.02
 Hom-1501272.7162.91.190.582.43  
 Het-150127727.318734.11.411.091.82  
 Other70970.034663.01.00  
 Allele         
 Carrier33116.321919.91.281.031.580.02 
 Noncarrier1,69583.787980.11.00  

The risk classification defined for an HLA genotype was as follows: −2, −1, 0 (baseline risk), +1, +2 and +3 for the combinations of the associated alleles assuming an additive effect for the DRB1 alleles 0401 and 1301 and a dominant effect for the allele 1501 (Table IV). The best fitting model using the predefined HLA risk scores −2 (1301/1301), −1 (1301/allele not associated), 0 (2 alleles not associated or 1301/0401), 1 (0401/allele not associated or 1501/1301), 2 (1501/1501) and 3 (1501/0401) was used. It gave a statistically significantly increased point estimate for the HLA-DRB1 genotype associated risk, when it was evaluated on a continuous scale (Table IV). An increase of 23% (OR = 1.23, 95% CI: 1.14–1.32) in the CxCa risk estimate was observed when the HLA genotype changed from the protective alleles to the susceptible ones (Table IV). A similar risk estimate (OR = 1.25, 95% CI: 1.13–1.40) was observed when the analyses were done adjusting by smoking (data not shown).

Table IV. Risk Estimates (Odds Ratio, or and 95% Confidence Intervals, 95% CI) of Cervical Cancer According to Predefined Risk Levels given for a Specific HLA-DRB1 Genotype
ModelAICDFGlobal pOR95% Cl1
  • 1

    95% confidence interval;

  • 2

    HLA scores are as follows: −2 (1301/1301), −1 (1301/allele not associated), 0 (two alleles not associated or 1301/0401); 1 (0401/allele not associated or 1501/1301), 2 (1501/1501) and 3 (1501/0401). –AIC, Aikaiks's information criterion; DF, degrees of freedom.

Six cateories21749.955<.0001   
 3 vs. 0   1.280.891.86
 2 vs. 0   1.481.191.84
 1 vs. 0   1.401.121.75
 0   1.00
 −1 vs. 0   0.610.430.85
 −2 vs. 0   0.540.231.28
Six categories(continuous)1751.881<.00011.231.141.32
Three categories1745.932<.0001   
 High risk vs. baseline   1.391.171.66
 Baseline   1.00
 Low risk vs. baseline   0.580.420.79

Analyses for gene–gene interactions were done generating risk classification for each of the CxCa-associated genetic factors. The risk classification for HLA-DRB1 was included in the models as a continuous variable. The risk classification defined for individuals having a specific genotype of CCND1, IL-6, IL4-R and LTA genes, included as an additive effect in the model was: a baseline risk (absence of the risk allele), a medium risk (heterozygosity for the risk allele) and a high risk (homozygosity for the risk allele). The LTA gene showed interactions with CCND1 and HLA-DRB1 but they were of borderline statistical significance (p value = 0.045 and p value = 0.103, respectively). All gene-smoking interactions were not significant (data not shown). Additionally, analyses stratifying by smoking and CxCa-associated HLA alleles did not show any differences between the stratified groups (data not shown). Frequencies for each of the haplotypes of IL-6 and IL4-R genes are shown in Table V. No association with the disease was found for IL-6 or IL4-R haplotypes.

Table V. Risk Estimates (Odds Ratio, or and 95% Confidence Intervals, 95% CI) of Cervical Cancer According to IL-6 and IL-4R Haplotypes
GeneHaplotypes1TotalCasesControlsOR95% Cl3
n2%n2%n2%
  • 1

    Haplotypes formed from SNPs rs1805010, rs1805015 and rs1801275 in the IL-4R gene and the SNPs rs1554606 and rs1800797 in the IL-6 gene.

  • 2

    Numbers for individual genotypes may not add up to the total because of missing data.

  • 3

    95% lower and upper confidence intervals.

IL4-RATA1,61530.457230.21,04330.61
 ATG911.7291.5621.80.850.541.34
 ACA40.130.210.035.470.5752.70
 ACG78614.828114.850514.81.020.851.21
 GTA2,17641.079642.01,38040.51.050.921.20
 GTG1853.5603.21253.70.880.631.21
 GCG4498.51558.22948.60.960.771.20
IL-6GG2,70250.798852.11,71450.01
 GA40.130.210.035.200.5450.09
 TG2224.2874.61353.91.120.841.48
 TA2,40045.082043.21,58046.10.910.811.02

The 6 polymorphisms associated with CxCa development in the whole population were analyzed further for associations with CIN3 and ICC, independently (Supporting Information Tables 2 and 3). Four polymorphisms remained significantly associated with CIN3: the SNP rs9344 in the CCND1 gene (OR = 1.37, 95% CI: 1.07–1.76 and global p = 0.02, for the AA homozygous variant, and OR = 1.17, 95% CI: 1.04–1.33 and global P = 0.01, for the allele variant), the SNP rs1041981 in the LTA gene (OR = 0.72, 95% CI: 0.55–0.93 and global p = 0.04, for the AA homozygous variant, and OR = 0.85, 95% CI: 0.74–0.96 and global p = 0.01, for the allele variant), the polymorphism HLA-DRB1*1301 (OR = 0.55, 95% CI: 0.42–0.72 and global p < 0.001, for heterozygous genotype), and the polymorphism HLA-DRB1*1501 (OR = 1.42, 95% CI: 1.16–1.73 and global p = 0.002, for heterozygous genotype). For the ICC-restricted analysis, the OR showed the same tendencies, but statistical power was lower due to only 196 cases. Only the polymorphism rs9344 in the CCND1 gene remained significantly associated with the disease (OR = 3.01, 95% CI: 1.39–6.50 and global p = 0.01, for the AA homozygous variant, and OR = 1.72, 95% CI: 1.19–2.50 and global p = 0.004, for the allele variant).

Discussion

Our study was based on a large, homogeneous population and it aimed at identifying the possible role of immune response- and cell cycle-related candidate genes in the development of CxCa. The multiplex approach addressed genetic markers previously suggested to be associated with disease susceptibility in Caucasians.10, 11, 13, 14, 25 Three of 24 polymorphisms (in the IL-6, LTA and CCND1 genes) were significantly associated with CxCa. Additionally, 3 alleles of the HLA class II DRB1 locus had a significant association with CxCa. The genetic effects of CCND1 and the HLA-DRB1 alleles in the susceptibility to CxCa were independent of the effect of smoking.

IL-6 is a multifunctional cytokine that can regulate immune and inflammatory responses. We found a protective effect for the IL-6 gene against CxCa in those individuals carrying the A allele of the rs1800797 polymorphism. The association of this marker with the disease seems to be independent of the presence of other CxCa-associated genes, HLA-DRB1 alleles and smoking status. Other studies to investigate this variant of the IL-6 gene in CxCa have not been conducted. A previous study in Brazilian women has reported an association between −174G>C (rs1800795), another polymorphism in the IL-6 gene and cervical neoplasia.13 We did not include the latter variant in our SNP set due to limitations in the primer sets for the multiplex PCR. These 2 alleles seem to contribute in an opposite fashion to the susceptibility of developing CxCa.

The role of variations in the IL-6 gene in susceptibility for thedevelopment of CxCa can be explained by, first, CxCa cancerbeing a process where the HPV infection may favor the development of local chronic inflammation with a more evident inflammatory response in severe lesions than in normal cervix.26 Second, a promoting effect in tumor cell growth by autocrine and/or paracrine processes has been demonstrated for IL-6 in cervical cancer.27 Third, variations in the IL-6 gene, such as the rs1800795 polymorphism, have been found to be associated with the plasma levels of the protein.28 We propose that the protective association observed might be an effect in the level, location or timing of the gene expression, which may decrease the local levels of IL-6, and thereby limit its participation in a proinflammatory reaction.

The LTA rs1041981 (exon-3) variant, that results in the amino-acid substitution Thr60Asn, also showed a protective effect in the present study. This association supports the finding of a previous Japanese case-control study, where the polymorphism Thr60Asn decreased the risk of cervical squamous cell carcinoma.18 A second polymorphism (rs909253 in the intron 1 of the LTA gene) has been shown to be in complete linkage disequilibrium with the Thr60Asn polymorphism and to confer a similar effect for CxCa and endometrial cancer.18 There is evidence that both wild-type (60Thr) and variant (60Asn) LTA stimulates mRNA expression of vascular cell-adhesion molecule 1 (VCAM1) in ischemic stress.29 A similar effect has been observed for the mRNA expression of E-selectin in human artery smooth-muscle cells.30 If this is the case for CxCa, the enhancement of the vascular adhesion molecules presented in high-grade squamous intraepithelial lesion (SIL) but not in low-grade SIL12 may be in part explained by the presence of the studied polymorphism. Such polymorphisms may change the ability of LTA to recruit immunocompetent cells and to stimulate the antineoplastic immune response. LTA gene is located within the large haplotype block of HLA-DQB1 and we can not exclude the possibility that the observed effect of the Thr60Asn polymorphism in CxCa in our Swedish cohort can be explained by linkage disequilibrium with other polymorphisms in the haplotype block. However, according to our gene–gene interaction analysis, it seems to be independent of the presence of the susceptibility alleles of HLA-DRB1.

We found significant associations for 3 HLA-DRB1 alleles that have been extensively reported in other studies, supporting the hypothesis that genetic diversity in the HLA molecules affects the risk of CxCa. The alleles DRB1*1501 and DRB1*0401 increased the risk of CxCa and the allele DRB1*1301 decreased it. A previous association of the HLA-DRB1*1501 allele with an increased susceptibility to CxCa in the Swedish population has been demonstrated.31–33 Additionally, in Scottish,34 Hispanic,35 Brazilian36 and other non-European Caucasian populations,37 similar findings have been described. It is important to highlight that the effect of the allele 1501 is mainly observed in the haplotype DRB1*1501-DQB1*0602,31, 32 and in some particular studies the association seems to be HPV-dependent.34 However, there are discrepant results with this allele (or haplotype), for populations such as Asians38, 39 or US Caucasians,11, 40 where no significant associations were found or a negative association was reported.41 The protective effect of the DRB1*1301 allele found in this study supports many other findings where DRB*1301 allele alone16, 31, 32, 35, 40 or in the presence of the allele DQB1*0603,40 that is in strong linkage disequilibrium, has shown a significant association with CxCa. Again, some contradictory findings exist for these alleles, and also lack of association has been observed.34, 38–41 Finally, in contrast to a few previous investigations,36, 38 but in agreement with others,34, 38, 40 we observed a positive association between the presence of DRB1*0401 and CxCa.

Overall, our study indicates an association of the HLA-DRB1 alleles 0401, 1501 and 1301 with CxCa in the Swedish population independently of other reported CxCa susceptibility genes analyzed and independently of smoking. However, we cannot exclude the contribution of other HLA or non-HLA genes such as TNF-alpha,19TAP242, 43 or MICA43 that may be in linkage disequilibrium with the genetic markers tested. It has also been demonstrated that at least 36 HLA allele combinations are associated with a susceptibility to CxCa in the US Caucasian population.40 This suggests that co-occurrence of multiple HLA risk alleles gives more informative explanation for susceptibility to CxCa, than individual alleles. Furthermore, the association between HLA polymorphisms and risk of CxCa has been suggested to be influenced by HPV type and variation of HPV types and type-variants between populations.44, 45

In our attempt to look for possible gene–gene interactions, a risk classification for the carriers of CxCa-associated HLA genotypes was done. An open question is determining whether the effect of 1 susceptible allele (e.g., DRB1*0401) is dominant over the other protective allele (e.g., DRB1*1301). We proposed the first approach to answer this question, when we produced the risk categories assuming as a baseline risk (category 0), the genotype formed by both susceptible and protective alleles. However, we cannot exclude more complex genetic models inside the DRB1 genotypes. Additionally, biological mechanisms of efficiency in the antigen presentation of a specific HLA allele and heterozygosity advantage in HPV infection and cervical cancer, needs further investigations.46

We could not demonstrate a genetic susceptibility by the innate immune system in cervical cancer. To the best of our knowledge, this is the first study testing an association of CxCa with polymorphism in toll-like receptors (TLR), therefore, additional investigations with different approaches are needed.

In addition to the immune-related genes, 1 polymorphism (rs9344 in the CCND1 gene) of 3 selected candidate genes, which participate in cell cycle was positively associated with the risk of CxCa. Those individuals who carry the allele A had a higher risk for the disease than those with the G allele, a result in contrast to a previous Portuguese study where the G allele gave a positive association with the disease.25 Absence of association between the rs9344 polymorphism and CxCa has been reported in a Korean population.47 The large number of subjects tested in our study may have given a higher power to detect such associations in the Swedish study contrary to the results found in the Korean study where only 222 cases and 314 controls were tested.47 This polymorphism has been associated with an increase in the alternative splicing of the RNA.48 In addition, an excessive cyclin D1 expression and/or activity in human cancers can deregulate the mitotic cell cycle and promote oncogenic transformation.49, 50 Cyclin D1 is essential for HPV-induced cell transformation.51 The results in the present study support a genetic susceptibility in a stage of CxCa development that is not related with immune response against the HPV infection or its persistence. Therefore, the observed susceptibility may contribute as a cofactor of HPV in the disruption of the cell cycle checkpoints and initiation of carcinogenesis. The observed effect of CCND1 variant in CxCa seems to be independent of the smoking history. The absence of a significant interaction with smoking, and additional stratified analyses, are in line of smoking as an independent cofactor to the development of CxCa.52, 53 This differs from previous findings in lung cancer, where the rs9344 polymorphism54 and increased expression of the protein55 were linked to a heavy smoking history.

The validated Luminex platform allowed rapid testing of 24 markers in 2,700 samples with a minimum of DNA material. About 30% of the possible markers that were to be included in the system initially, however, were eliminated because of restrictions in the design and performance of the multiplex PCR. This technology has been previously described and validated in several studies that have used the platform for a high-throughput multiplex genotyping.19, 56 Strengths of our study included also the population-based study design,21 with a correct disease classification by the use of the Swedish Cancer Registry, that has a completeness of nearly 100%.57 And a large set of samples that allowed statistical power to distinguish an association between the included markers and CxCa, and a smoking history that allowed testing for gene–environmental interactions.

Some limitations of the present study are as follows: First, the possible confounding by HPV type in the tumor. Unfortunately, for most of the study cases tumor tissues are not available within reasonable effort. However, a small sample of the cases (133 cases) have been analyzed for HPV DNA, in which less than 10% of cases had other high-risk HPV types than 16 or 18 (HPV31 or 33).58 On the other hand, this kind of assessment should go further to the level of HPV variants. However, such sequencing data is not available from sizeable cohorts of women with and without CxCa. Second, although the study is based on a population of around 106,000 individuals, only 973 women developed CxCa and the power is still small for such further stratifications. When the genetic markers were evaluated in the CIN3 and ICC independent analyses, most of the associations observed in the overall analysis for CxCa are attributed to the CIN3 cases. Third, the modest magnitude of the genetic effects observed would be lost in multiple comparisons correction. However, keeping in view the biological relevance of the SNPs included in the study, true nature of modest effects observed cannot be excluded.

In summary, results of our Swedish population-based study provide evidence on immunogenetic susceptibility to CxCa but also a possible predisposition to the “inside disease” processes such as the cell cycle. Our findings as well as the previous reports are still not conclusive about how this genetic susceptibility is mechanistically linked to HPV infection and its progression to CxCa. As our study is one of the largest genetic susceptibility studies on CxCa published so far, it may contribute for further studies and meta-analyses.59 Harnessing of biobank cohorts in combination with high-throughput genotyping methodology to perform strongly powered studies toward these aims is warranted.

Acknowledgements

The authors thank Dr. Tim Waterboer from the Division of Genome Modifications and Carcinogenesis, DKFZ, Heidelberg, Germany, for his advice in the luminex method and data management. They thank Dr. Heljiä-Marja Surcel from the National Public Health Institute, Oulu, Finland, for her assistance in the HLA decisions. They also recognize and thank 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 thank Dr. Alexandra Nieters from the Division of Cancer Epidemiology, for reference reagents and Dr. Alex Weber from the Junior Research Group Toll-like receptors and cancer, for discussions on TLR polymorphisms, DKFZ, Heidelberg, Germany.

Ancillary