Human Genetics

, Volume 132, Issue 3, pp 301–312

Polymorphisms of the Interleukin 6 gene contribute to cervical cancer susceptibility in Eastern Chinese women

Authors

  • Ting-Yan Shi
    • Cancer InstituteFudan University Shanghai Cancer Center
    • Department of OncologyShanghai Medical College, Fudan University
  • Mei-Ling Zhu
    • Cancer InstituteFudan University Shanghai Cancer Center
    • Department of OncologyShanghai Medical College, Fudan University
  • Jing He
    • Cancer InstituteFudan University Shanghai Cancer Center
    • Department of OncologyShanghai Medical College, Fudan University
  • Meng-Yun Wang
    • Cancer InstituteFudan University Shanghai Cancer Center
    • Department of OncologyShanghai Medical College, Fudan University
  • Qiao-Xin Li
    • Cancer InstituteFudan University Shanghai Cancer Center
    • Department of OncologyShanghai Medical College, Fudan University
  • Xiao-Yan Zhou
    • Cancer InstituteFudan University Shanghai Cancer Center
    • Department of PathologyFudan University Shanghai Cancer Center
    • Department of OncologyShanghai Medical College, Fudan University
  • Meng-Hong Sun
    • Department of PathologyFudan University Shanghai Cancer Center
    • Department of OncologyShanghai Medical College, Fudan University
  • Zhi-Ming Shao
    • Department of Breast SurgeryFudan University Shanghai Cancer Center
    • Department of OncologyShanghai Medical College, Fudan University
  • Ke-Da Yu
    • Department of Breast SurgeryFudan University Shanghai Cancer Center
    • Department of OncologyShanghai Medical College, Fudan University
  • Xi Cheng
    • Department of Gynecologic OncologyFudan University Shanghai Cancer Center
    • Department of OncologyShanghai Medical College, Fudan University
    • Department of Gynecologic OncologyFudan University Shanghai Cancer Center
    • Department of OncologyShanghai Medical College, Fudan University
    • Cancer InstituteFudan University Shanghai Cancer Center
    • Department of EpidemiologyThe University of Texas M.D. Anderson Cancer Center
Original Investigation

DOI: 10.1007/s00439-012-1245-4

Cite this article as:
Shi, T., Zhu, M., He, J. et al. Hum Genet (2013) 132: 301. doi:10.1007/s00439-012-1245-4

Abstract

Interleukin 6 (IL6) encodes a cytokine protein, which functions in inflammation, maintains immune homeostasis and plays important roles in cervical carcinogenesis. Single nucleotide polymorphisms (SNPs) in IL6 that cause variations in host immune response may contribute to cervical cancer risk. In this two-stage case–control study with a total of 1,584 cervical cancer cases and 1,768 cancer-free female controls, we investigated associations between two IL6 SNPs and cervical cancer risk in Eastern Chinese women. In both Study 1 and Study 2, we found a significant association of the IL6-rs2069837 SNP with an increased risk of cervical cancer as well as in their combined data (OR 1.27 and 1.19, 95 % CI 1.08–1.49 and 1.04–1.36, P = 0.004 and 0.014 for dominant and additive genetic models, respectively). Furthermore, rs2069837 variant AG/GG carriers showed significantly higher levels of IL6 protein than did rs2069837 AA carriers in the target tissues. Using multifactor dimensionality reduction (MDR) and classification and regression tree (CART) analyses, we observed some evidence of interactions of the IL6 rs2069837 SNP with age at primiparity and menopausal status in cervical cancer risk. We concluded that the IL6-rs2069837 SNP may be a marker for susceptibility to cervical cancer in Eastern Chinese women by a possible mechanism of altering the IL6 protein expression. Although lacked information on human papillomavirus (HPV) infection, our study also suggested possible interactions between IL6 genotypes and age at primiparity or menopausal status in cervical carcinogenesis. However, larger, independent studies with detailed HPV infection data are warranted to validate our findings.

Abbreviations

HPV

Human papillomavirus

IL6

Interleukin 6

SNP

Single nucleotide polymorphism

FUSCC

Fudan University Shanghai Cancer Center

TZL

Taizhou longitudinal study

BMI

Body mass index

FIGO

International Federation of Gynecology and Obstetrics

LN

Lymph node

LVSI

Lympho–vascular space invasion

ER

Estrogen receptor

PR

Progesterone receptor

UTR

Untranslated region

MAF

Minor allele frequency

LD

Linkage disequilibrium

TFBS

Transcription factor binding site

IHC

Immunohistochemistry

MDR

Multifactor dimensionality reduction

CART

Classification and regression tree

OR

Odds ratio

CI

Confidence interval

FPRP

False positive report probability

HWE

Hardy–Weinberg equilibrium

CVC

Cross-validation consistency

TN

Terminal node

ChIP

Chromatin immunoprecipitation

Introduction

Cervical cancer is the third most commonly diagnosed cancer and the fourth leading cause of cancer death in women, accounting for 9 % (529,800) of all new cancer cases and 8 % (275,100) of all cancer deaths among women in 2008 in the world (Jemal et al. 2011). More than 85 % of these cases and deaths occur in developing countries, including China (Jemal et al. 2011). A large body of molecular epidemiology research supports the hypothesis that persistent infection with an oncogenic human papillomavirus (HPV) type and chronic inflammation is the primary cause of cervical cancer, deemed as a necessary cause for the disease (Jemal et al. 2011; Walboomers et al. 1999). On the other hand, host immune response plays a role in eliminating viral infection and preventing progression to cervical cancer, in which immune cells are recruited before or after inflammation and tumor progression (Bedoya et al. 2012).

Interleukin 6 (IL6), as an important inflammation-related gene, has been mapped to chromosome 7p21 and consists of five exons spanning ~4.9 kb of the genomic DNA (Fig. 1a). IL6 is one of the interleukin family genes, encoding a phosphorylated and glycosylated cytokine protein with 212 amino acids (Fig. 1b), which is involved in inflammation, maturation of B cells and maintenance of immune homeostasis (Gruol and Nelson 1997). Therefore, IL6 implicates in a wide variety of inflammation-related disease states (Kishimoto et al. 1995).
https://static-content.springer.com/image/art%3A10.1007%2Fs00439-012-1245-4/MediaObjects/439_2012_1245_Fig1_HTML.gif
Fig. 1

Structural characteristics and protein expression levels of IL6. a The IL6 gene consists of five exons spanning ~4.9 kb of genomic DNA, with two IL6 SNPs (rs2069837 and rs2069840) being labeled; b IL6 is a phosphorylated glycoprotein and consists of 212 amino acids, containing an IL6 superfamily domain; c pairwise linkage disequilibrium (LD) among two selected SNPs of IL6, which capture the other nine untyped SNPs in the same gene. The value within each diamond represents the pairwise correlation between SNPs (measured as r2) defined by the upper left and the upper right sides of the diamond. The red-to-white gradient reflects higher to lower LD values, and the diamond without a number corresponds to r2 = 1; d IL6 protein expression levels were detected in the target tissues with immunohistochemistry (IHC). The rs2069837 variant AG/GG genotypes were significantly associated with an increased level of IL6 protein expression, compared with the rs2069837 AA genotype (χ2-test, P = 0.036). The graphs on the right show representative IHC results

Previous data have showed an over-expression of IL6 mRNA and protein in cervical cancer cells (Abdelwahab et al. 2012; Vazquez-Ortiz et al. 2005; Veerapaneni et al. 2009; Wei et al. 2001). Moreover, IL6 levels also increased in cervico-vaginal secretions of patients with cervical cancer, which are associated with the severity of disease (Tjiong et al. 1999). For example, high micro-environmental IL6 levels may promote tumor angiogenesis and cancer development (Wei et al. 2001), which could be genetically determined at an individual level. Previously published studies have suggested that potentially functional single nucleotide polymorphisms (SNPs) in IL6 may modify gene function and lead to susceptibility to human cancers (Kwon et al. 2011; Lim et al. 2011). However, few studies have investigated the roles of functional SNPs in cervical cancer, and the published studies had either a relatively small sample size or no further validation of their findings. Although there was a much higher incidence rate of cervical cancer in China than that in Western populations, few studies of genetic susceptibility have been performed in Chinese populations (Castro et al. 2009; Gangwar et al. 2009; Nogueira de Souza et al. 2006).

Therefore, in the present study, we used an in silico approach to select possibly functional SNPs in the IL6 gene and performed a two-stage case–control study with a relatively large sample size to investigate their associations with cervical cancer risk in Eastern Chinese women.

Materials and methods

Study subjects

The first case–control study (Study 1) consisted of 458 cervical cancer patients between February 2008 and February 2009 in Fudan University Shanghai Cancer Center (FUSCC), and 519 population-based cancer-free female controls from the Taizhou longitudinal study (TZL) conducted at the same time period, as described previously (Wang et al. 2009). The replication study (Study 2) included 1,126 consecutive cervical cancer patients between March 2009 and March 2011 from FUSCC, and an additional 1,249 cancer-free female controls, who were enrolled from women coming to the Outpatient Department of Breast Surgery at FUSCC for breast cancer screening. All subjects were genetically unrelated, who were from Eastern China where they lived. The tumors were histopathologically confirmed as primary cervical cancer independently by two gynecologic pathologists as routine diagnosis at FUSCC. The controls, with the selection criteria including no individual history of cancer, were frequency matched to the cases on age (±5 years) and residential areas.

During an in-person interview, all potential subjects provided information about their demographics and known risk factors with an approximate response rate of 95 % for the cases as well as 90 and 95 % for the controls from TZL and the Outpatient Department of Breast Surgery at FUSCC, respectively. Because most Chinese women are non-smokers and non-drinkers, our study populations were restricted to women who did not smoke cigarettes or drink alcohol. For the cases, clinical and pathological information was further extracted from the patients’ electronic database at FUSCC, including tumor histology (WHO 2010), FIGO stage (International Federation of Gynecology and Obstetrics) (Pecorelli 2009), tumor size (the largest tumor diameter of the primary tumor), pelvic lymph node (LN) metastasis, lympho–vascular space invasion (LVSI), depth of cervical stromal invasion and the expression of estrogen receptor (ER) and progesterone receptor (PR). Each participant provided a one-time 10 ml of venous blood sample (collected by the Tissue Bank of FUSCC after the diagnosis and before the initiation of treatment for the cases), and samples were kept frozen till DNA extraction for genotyping. A written informed consent was obtained from all recruited individuals, and the research was approved by the Institutional Review Board of FUSCC.

Selection of IL6 SNPs and genotyping

The SNPs were selected from the NCBI dbSNP database (http://www.ncbi.nlm.nih.gov/projects/SNP) and the International HapMap Project database (http://hapmap.ncbi.nlm.nih.gov/) based on four criteria: (1) located at the coding sequence, 5′-untranslated region (UTR), 3′-UTR and intron regions of the IL6 gene, (2) minor allele frequency (MAF) of at least 5 % in Chinese populations, (3) with low linkage disequilibrium (LD) using an r2 threshold of <0.8 for each other, and (4) predicted as a potentially functional SNP using the SNP function prediction (FuncPred) software (http://snpinfo.niehs.nih.gov/snpfunc.htm). As a result, two [i.e., IL6-rs2069837 and IL6-rs2069840, predicted to be located in transcription factor binding sites (TFBS)] were selected, which captured other nine untyped SNPs within the same gene (Fig. 1c).

Genomic DNA extraction and genotyping were conducted as described previously (He et al. 2012), with a successful genotyping rate of 99.6 % using the Taqman assay (the probes are listed in Supplementary Table S1). The discrepancy rate in 10 % of duplicated samples was less than 0.1 %, and a few samples were randomly selected to be sequenced and the genotypes were confirmed.

IL6 protein expression levels by IL6 genotypes in the target tissues

The IL6 protein expression level was performed by the immunohistochemistry (IHC) assay on 5-μm-thick tissue sections prepared from formalin-fixed, paraffin-embedded tissue from the constructed tissue microarray block, using the antibody against IL6 [sc-130326, mouse monoclonal antibody, Santa Cruz Biotechnology (Inc., Santa Cruz, CA), 1:50 dilution] and ChemMate™ EnVision™ detection kit (DAKO, Glostrup, Denmark). A known positive sample was included as the assay control. For the negative control, the primary antibody was replaced with non-immune mouse serum. The IHC staining results were independently scored by two researchers (T-Y. S and Q-X. L), who were blinded to patient information, with a scoring system based on both the percentage of positive tumor cells and staining intensity, as described previously (Cheng et al. 2011). The assessment of the protein expression was defined as low (≤3+) and high (>3+ to 6+) levels, and for cores that were uninterpretable because of tissue loss or lack of tumor cells, a score of not applicable (N/A) was given. We performed Pearson’s χ2-test and logistic regression analysis to estimate the correlation between IL6 genotypes and IL6 protein expression levels.

High-order interaction analysis

To explore high-order gene-environment interactions in cervical cancer risk, we performed multifactor dimensionality reduction (MDR) (Ritchie et al. 2001) and classification and regression tree (CART) (Zhang and Bonney 2000) analyses using the MDR V2.0 beta 8.2 program (http://www.multifactordimensionalityreduction.org/) and SAS software (version 9.1; SAS Institute, Cary, NC), respectively. To search for the best n-factor model in the MDR analysis, the n-dimensional space formed by all possible combinations of n factors in a given set is reduced to a single dimension by reclassifying each factor as either high risk or low risk, as described previously (Ritchie et al. 2001). CART creates a decision tree that depicts how well each genotype or environmental factor predicts disease (Zhang and Bonney 2000). A splitting rule based on generalized GINI index (Therneau et al. 1997) was used in the present study to stratify data into subsets of individuals, which are represented in the CART decision tree as nodes.

Statistical analysis

The differences in selected demographic variables between cases and controls were evaluated by the Pearson’s χ2-test. The associations of IL6 genotypes with cervical cancer risk were estimated by computing odds ratios (ORs) and their 95 % confidence intervals (CIs) from both univariate and multivariate logistic regression models. The association was also evaluated in subgroup analyses stratified by demographic and clinic variables. For all significant genetic effects observed in this study, we calculated false positive report probability (FPRP) with prior probabilities of 0.0001, 0.001, 0.01, 0.1 and 0.25 to test false positive associations (Wacholder et al. 2004). A FPRP value <0.2 indicated an association that remained robust for a given prior probability and was considered a noteworthy finding. Statistical power was estimated to detect an OR of 1.50/0.67 (for a risk/protective effect), with an α level equal to the observed P value (Wacholder et al. 2004). All statistical analyses were performed with SAS software, unless stated otherwise. All P values were two-sided with a significance level of P < 0.05.

Results

Population characteristics

Among all the studied subjects, 12 controls failed to be genotyped after repeated assays, likely due to poor quality of DNA. Thus, the final analysis included a total of 1,584 cases and 1,756 controls. There was no significant difference in distributions of age between cases and controls in Study 1, Study 2 and all subjects combined (P = 0.389, 0.408 and 0.208, respectively). However, the cases were more likely to be premenopausal, thinner and younger at primiparity than the controls (Supplementary Table S2). Because the differences in age at primiparity, menopausal status and body mass index (BMI) were significant between cases and controls, these variables were further adjusted for any residual confounding effect in later multivariate logistic regression analyses.

Association of IL6 SNPs with cervical cancer risk

The genotype frequencies of the IL6 rs2069837 and rs2069840 SNPs and their associations with cervical cancer risk in Study 1 and Study 2 are summarized in Table 1. All observed genotype distributions among the 511 and 1,245 controls of Study 1 and Study 2, respectively, agreed with the Hardy–Weinberg equilibrium (HWE, P = 0.935 and 0.239 for rs2069837 and 0.936 and 0.292 for rs2069840, respectively). In Study 1, we found a significant association between the rs2069837 SNP and an increased risk of cervical cancer in the additive genetic model (adjusted OR 1.26, 95 % CI 1.00–1.58, P = 0.046) with adjustment for age, age at primiparity, menopausal status and BMI. In Study 2, this association was validated by a significantly increased OR of 1.27 and 1.18 (P = 0.009 and 0.031) in both dominant and additive genetic models, respectively, which remained unchanged after all subjects were combined (adjusted OR 1.27 and 1.19, 95 % CI 1.08–1.49 and 1.04–1.36, P = 0.004 and 0.014 for both dominant and additive genetic models, respectively) with adjustment for age, age at primiparity, menopausal status and BMI. However, these associations were not observed for the rs2069840 SNP. In stratification analyses, as shown in Table 2 and Supplementary Table S3, by assuming a dominant genetic model, the significantly increased risk associated with rs2069837 AG/GG variant genotypes was more evident in women who were younger (≤46 years), premenopausal, thinner (BMI < 25 kg/m2) or younger at primiparity (≤24 years) (P = 0.004, 0.012, 0.001 and 0.008, respectively), which was also observed for subgroups of squamous cell carcinoma, FIGO stage I, depth of cervical stromal invasion more than 1/2 and negative expression of ER and PR. However, homogeneity tests suggested that there was no difference in risk estimates between subgroups of the strata.
Table 1

Logistic regression analysis of associations between IL6 variant genotypes and cervical cancer risk in Eastern Chinese women

IL6 variant genotypes

Study 1

Study 2

All subjects

Cases/controls

Crude OR (95 % CI)

P

Adjusted OR (95 % CI)a

Pa

Cases/controls

Crude OR (95 % CI)

P

Adjusted OR (95 % CI)a

Pa

Cases/controls

Crude OR (95 % CI)

P

Adjusted OR (95 % CI)a

Pa

Subjects

458/511

    

1,126/1,245

    

1,584/1,756

    

rs2069837

 AA

271/335

1.00

 

1.00

 

706/841

1.00

 

1.00

 

977/1,176

1.00

 

1.00

 

 AG

165/158

1.29 (0.99–1.69)

0.064

1.26 (0.96–1.66)

0.100

378/357

1.26 (1.06–1.50))

0.010

1.30 (1.08–1.56)

0.005

543/515

1.27 (1.10–1.47)

0.002

1.29 (1.09–1.53)

0.003

 GG

22/18

1.51 (0.79–2.87)

0.209

1.57 (0.82–3.03)

0.176

42/47

1.06 (0.69–1.63)

0.775

1.04 (0.67–1.62)

0.869

64/65

1.19 (0.83–1.69)

0.349

1.11 (0.74–1.65)

0.621

 AG/GG

187/176

1.31 (1.01–1.71)

0.041

1.29 (0.99–1.68)

0.059

420/404

1.24 (1.05–1.47)

0.013

1.27 (1.06–1.52)

0.009

607/580

1.26 (1.09–1.45)

0.001

1.27 (1.08–1.49)

0.004

 Additive model

 

1.27 (1.01–1.58)

0.037

1.26 (1.00–1.58)

0.046

 

1.17 (1.01–1.35)

0.037

1.18 (1.02–1.37)

0.031

 

1.20 (1.06–1.35)

0.004

1.19 (1.04–1.36)

0.014

rs2069840

 GG

1/3

1.00

 

1.00

 

3/9

1.00

 

1.00

 

4/12

1.00

 

1.00

 

 CG

52/74

2.10 (0.21–20.76)

0.524

2.13 (0.22–21.09)

0.517

143/162

2.64 (0.70–9.94)

0.151

2.65 (0.69–10.16)

0.156

195/236

2.47 (0.79–7.79)

0.122

2.86 (0.86–9.51)

0.086

 CC

405/434

2.79 (0.29–26.93)

0.374

2.82 (0.29–27.18)

0.371

980/1,074

2.73 (0.74–10.11)

0.132

2.69 (0.71–10.15)

0.144

1,385/1,508

2.75 (0.89–8.54)

0.080

3.02 (0.92–9.89)

0.068

 CG/CC

457/508

2.69 (0.28–25.95)

0.391

2.72 (0.28–26.22)

0.388

1,123/1,236

2.72 (0.74–10.06)

0.134

2.69 (0.71–10.13)

0.145

1,580/1,744

2.71 (0.87–8.42)

0.084

3.00 (0.92–9.82)

0.069

 Additive model

 

1.36 (0.95–1.95)

0.094

1.35 (0.94–1.94)

0.100

 

1.10 (0.88–1.38)

0.410

1.09 (0.86–1.37)

0.487

 

1.17 (0.97–1.41)

0.110

1.13 (0.92–1.40)

0.258

The results were in bold, if P < 0.05

Study 1, the subjects were recruited between 2008 and 2009; Study 2, the subjects were recruited between 2009 and 2011

aAdjusted for age, age at primiparity, menopausal status and BMI in logistic regression models

Table 2

Stratification analysis for associations between IL6 variant genotypes and cervical cancer risk under dominant genetic models in all subjects of Eastern Chinese women

Variables

rs2069837 (cases/controls)

Crude OR (95 % CI)

P

Adjusted ORa (95 % CI)

Pa

Phom

rs2069840 (cases/controls)

Crude OR (95 % CI)

P

Adjusted ORa (95 % CI)

Pa

Phom

AA

AG/GG

GG

CG/CC

Age, years

 ≤46 (mean)

537/609

357/296

1.37 (1.13–1.66)

0.002

1.38 (1.11–1.71)

0.004

0.199

3/6

891/899

1.98 (0.49–7.95)

0.334

2.00 (0.47–8.55)

0.349

0.480

 >46 (mean)

440/567

250/284

1.13 (0.92–1.40)

0.241

1.17 (0.92–1.50)

0.210

 

1/6

689/845

4.89 (0.59–40.70)

0.142

5.27 (0.60–46.15)

0.134

 

Age at primiparity, years

 ≤24 (mean)

590/478

351/220

1.29 (1.05–1.59)

0.015

1.34 (1.08–1.66)

0.008

0.595

4/8

937/690

2.72 (0.82–9.06)

0.104

2.53 (0.74–8.62)

0.137

0.798

 >24 (mean)

338/435

215/233

1.19 (0.94–1.50)

0.149

1.18 (0.92–1.52)

0.182

 

0/2

553/666

0.980

0.980

 

Menopausal status

 Premenopausal

690/556

451/278

1.31 (1.09–1.58)

0.005

1.28 (1.06–1.55)

0.012

0.248

3/5

1,138/829

2.29 (0.55–9.60)

0.258

2.32 (0.55–9.81)

0.254

0.677

 Postmenopausal

281/368

150/182

1.08 (0.83–1.41)

0.573

1.19 (0.89–1.58)

0.240

 

1/5

430/545

3.94 (0.46–33.89)

0.211

5.10 (0.56–46.23)

0.148

 

BMIb, kg/m2

 <25

741/766

477/366

1.35 (1.14–1.60)

0.001

1.40 (1.16–1.69)

0.001

0.090

4/5

1,214/1,127

1.35 (0.36–5.02)

0.659

1.85 (0.49–7.03)

0.365

0.235

 ≥25

220/409

117/214

1.02 (0.77–1.34)

0.909

0.98 (0.72–1.35)

0.919

 

0/7

337/616

0.978

0.980

 

Histologyb

 CINIII

99/1,176

63/580

1.29 (0.93–1.80)

0.131

1.35 (0.94–1.93)

0.105

0.832

0/12

162/1,744

0.985

0.984

0.739

 Squamous

777/1,176

478/580

1.25 (1.07–1.45)

0.004

1.28 (1.08–1.52)

0.005

 

3/12

1,252/1,744

2.87 (0.81–10.20)

0.103

3.24 (0.87–12.04)

0.079

 

 Non–squamous

96/1,176

66/580

1.39 (1.00–1.94)

0.048

1.34 (0.94–1.91)

0.103

 

1/12

161/1,744

1.11 (0.14–8.57)

0.922

0.94 (0.12–7.61)

0.953

 

FIGO stage

 I

448/1,176

294/580

1.33 (1.11–1.59)

0.002

1.32 (1.08–1.61)

0.006

0.290

4/12

738/1,744

1.27 (0.41–3.95)

0.680

1.31 (0.40–4.33)

0.660

0.387

 II

346/1,176

199/580

1.17 (0.95–1.43)

0.133

1.23 (0.99–1.54)

0.066

 

0/12

545/1,744

0.975

0.975

 

 III–IV

34/1,176

14/580

0.84 (0.45–1.57)

0.575

1.13 (0.55–2.31)

0.741

 

0/12

48/1,744

0.987

0.990

 

Tumor size, cm

 <4

595/1,176

358/580

1.22 (1.04–1.44)

0.018

1.23 (1.03–1.48)

0.025

0.829

1/12

952/1,744

6.55 (0.85–50.45)

0.071

6.90 (0.86–55.07)

0.068

0.139

 ≥4

310/1,176

192/580

1.26 (1.02–1.54)

0.030

1.32 (1.05–1.66)

0.016

 

3/12

499/1,744

1.14 (0.32–4.07)

0.835

1.25 (0.34–4.66)

0.737

 

Pelvic LN

 Negative

722/1,176

432/580

1.21 (1.04–1.42)

0.015

1.23 (1.03–1.46)

0.021

0.639

1/12

1,153/1,744

7.93 (1.03–61.09)

0.047

8.21 (1.03–65.26)

0.047

0.047

 Positive

214/1,176

137/580

1.30 (1.03–1.64)

0.031

1.40 (1.08–1.81)

0.011

 

3/12

348/1,744

0.80 (0.22–2.84)

0.728

0.87 (0.23–3.27)

0.839

 

LVSI

 Negative

565/1,176

331/580

1.19 (1.00–1.41)

0.045

1.22 (1.01–1.47)

0.036

0.230

2/12

894/1,744

3.08 (0.69–13.77)

0.142

3.26 (0.70–15.19)

0.132

0.522

 Positive

268/1,176

182/580

1.38 (1.11–1.70)

0.003

1.45 (1.14–1.83)

0.002

 

2/12

448/1,744

1.54 (0.34–6.90)

0.573

1.71 (0.36–8.00)

0.498

 

Depth of cervical stromal invasion

 ≤1/2

440/1,176

251/580

1.16 (0.96–1.39)

0.122

1.20 (0.98–1.47)

0.078

0.325

1/12

690/1,744

4.75 (0.62–36.58)

0.135

4.91 (0.62–39.01)

0.132

0.423

 >1/2

480/1,176

311/580

1.31 (1.10–1.56)

0.002

1.36 (1.12–1.65)

0.002

 

3/12

788/1,744

1.81 (0.51–6.41)

0.361

2.10 (0.56–7.81)

0.269

 

ER expression

 Negative

473/1,176

280/580

1.20 (1.00–1.43)

0.045

1.24 (1.02–1.51)

0.030

0.800

3/12

750/1,744

1.72 (0.48–6.10)

0.403

1.98 (0.53–7.37)

0.306

0.716

 Positive

41/1,176

26/580

1.29 (0.78–2.12)

0.326

1.37 (0.80–2.32)

0.248

 

0/12

67/1,744

0.984

0.989

 

PR expression

 Negative

495/1,176

299/580

1.23 (1.03–1.46)

0.023

1.27 (1.04–1.54)

0.016

0.276

3/12

791/1,744

1.81 (0.51–6.43)

0.358

2.09 (0.56–7.77)

0.271

0.320

 Positive

19/1,176

7/580

0.75 (0.31–1.79)

0.512

0.75 (0.29–1.93)

0.551

 

0/12

26/1,744

0.990

0.990

 

Phom = P value for Homogeneity test

The results were in bold, if P < 0.05

BMI body mass index, FIGO International Federation of Gynecology and Obstetrics, CIN cervical intraepithelial neoplasia, LN lymph node, LVSI lympho–vascular space invasion, ER estrogen receptor, PR progesterone receptor

aObtained in logistic regression models with adjustment for age, age at primiparity, menopausal status and BMI

bAccording to the current WHO recommendations

We then calculated FPRP values for all observed significant associations. When the assumption of prior probability was 0.01, the association with the rs2069837 SNP (AG/GG vs. AA) was still noteworthy in all subjects (FPRP = 0.091) as well as in the subgroups of age ≤46 years, BMI < 25 kg/m2, FIGO stage I and depth of cervical stromal invasion >1/2 (FPRP = 0.188, 0.097, 0.177 and 0.175, respectively; Supplementary Table S4).

Correlations of IL6-rs2069837 genotypes with protein expression levels in the target tissues

Given the observed significant association between rs2069837 and cervical cancer risk, we further performed the IHC staining for IL6 protein expression in the target tissues. Out of the 151 cases with available cervical tissues, 90, 59 and two were high, low expression levels of IL6 and score of N/A, respectively. The genotype distributions of the IL6 genotypes in these cases with IL6 protein expression levels are shown in Table 3. Assuming a dominant genetic model, we found that individuals who carried rs2069837 AG/GG genotypes showed significantly higher levels of IL6 protein expression than those who carried rs2069837 AA genotypes (χ2-test, P = 0.036; adjusted OR 2.59; 95 % CI 1.14–5.86, P = 0.023; Fig. 1d) with adjustment for age, age at primiparity, menopausal status, BMI, tumor histology, FIGO stage, tumor size, pelvic LN, LVSI and depth of cervical stromal invasion.
Table 3

Logistic regression analysis of correlation between IL6 variant genotypes and IL6 protein expression in cervical cancer

Variant genotypes

IL6 protein expression levels

Pa

Crude OR (95 % CI)

P

Adjusted OR (95 % CI)b

Pb

Score (mean ± SD)

High N (%)

Low N (%)

All patients

3.19 ± 1.64

90 (60.4)

59 (39.6)

  

rs2069837

 AA

3.13 ± 1.61

47 (53.4)

41 (46.6)

0.110

1.00

 

1.00

 

 AG

3.24 ± 1.71

35 (70.0)

15 (30.0)

 

2.04 (0.98–4.25)

0.058

2.52 (1.06–5.98)

0.037

 GG

3.55 ± 1.63

8 (72.7)

3 (27.3)

 

2.33 (0.58–9.35)

0.235

3.00 (0.54–16.79)

0.208

 AG/GG

3.30 ± 1.69

43 (70.5)

18 (29.5)

0.036c

2.08 (1.04–4.16)

0.038

2.59 (1.14–5.86)

0.023

 Additive model

    

1.76 (1.00–3.08)

0.049

2.11 (1.07–4.16)

0.031

rs2069840d

 CG

2.82 ± 1.71

12 (54.6)

10 (45.5)

0.543

1.00

 

1.00

 

 CC

3.26 ± 1.62

78 (61.4)

49 (38.6)

 

1.33 (0.53–3.30)

0.544

0.96 (0.35–2.63)

0.942

The results were in bold, if P < 0.05

aχ2-test for genotype distributions between high and low expression levels of IL6 protein

bAdjusted for age, age at primiparity, menopausal status, BMI, tumor histology, FIGO stage, tumor size, pelvic LN, LVSI and depth of cervical stromal invasion in logistic regression models

cFor dominant genetic models

dNone was observed for the rs2069837 GG genotype

Association of high-order interactions with cervical cancer risk

To further explore high-order interactions, we performed MDR and CART analyses by including the two IL6 SNPs (i.e., rs2069837AG/GG and rs2069840 CG/CC) and three risk factors (i.e., age at primiparity, menopausal status and BMI). In the MDR analysis, age at primiparity was the best one-factor model with the highest CVC (100 %) and the lowest prediction error (38.2 %) among all five factors. More interestingly, the four-factor model had a maximum CVC (100 %) and a minimum prediction error (34.4 %), which showed a better prediction than one factor (Table 4 and Supplementary Table S5). Consistent with the MDR result of the best one-factor model, the initial split of the root node by the CART analysis was age at primiparity, indicating that age at primiparity was the strongest risk factor for cervical cancer among the factors considered. Further inspection of the tree structure revealed distinct interaction patterns. Premenopausal women with young age at primiparity and rs2069837 AG/GG genotypes (terminal node 2, TN2) had a 2.07-fold increased risk of cervical cancer (P < 0.001; Fig. 2), compared with the group with the lowest risk (reference group, TN4).
Table 4

MDR analysis for the cervical cancer risk prediction with and without IL6 variant genotypes

Number of risk factors

Best interaction models by MDR analysis

Cross-validation

Average prediction error (%)

P for permutation test

1

Age at primiparity

100/100

38.2

<0.0001

2

Age at primiparity, BMI

100/100

36.3

<0.0001

3

Age at primiparity, menopausal status, BMI

100/100

34.5

<0.0001

4

Age at primiparity, menopausal status, BMI, rs2069837AG/GG

100/100

34.4

<0.0001

5

Age at primiparity, menopausal status, BMI, rs2069837AG/GG, rs2069840CG/CC

100/100

34.4

<0.0001

The multi-locus model with maximum cross-validation consistency (CVC) and minimum prediction error rate is indicated in bold

https://static-content.springer.com/image/art%3A10.1007%2Fs00439-012-1245-4/MediaObjects/439_2012_1245_Fig2_HTML.gif
Fig. 2

Classification and regression tree analysis of age at primiparity, menopausal status and genetic variations in IL6. Terminal nodes are thick bordered. W wild-type genotype, V variant genotype, TN terminal node, #P value < 0.05

Discussion

To the best of our knowledge, this is the first study that has investigated whether the potentially functional IL6 SNPs (i.e., rs2069837 and rs2069840) are associated with cervical cancer risk in Eastern Chinese women. In this two-stage case–control study with a total of 1,584 cervical cancer cases and 1,768 female controls, the rs2069837 variant AG/GG genotypes were found to be associated with a significantly increased risk of cervical cancer with a statistical power of 99.1 %. We further validated the functionality of the rs2069837 variant AG/GG genotypes by detecting its protein expression levels in the target tissues and found that the carriers of these variant genotypes showed significantly higher levels of IL6 protein expression than did rs2069837 AA carriers, which provides additional biological evidence to support the IL6 rs2069837 A>G change at the TFBS as the underlying mechanism of the observed association between the rs2069837 SNP and risk of developing cervical cancer.

It is well-known that host immune response and chronic inflammation play roles in preventing progression to cervical cancer (Bedoya et al. 2012). IL6, an important inflammation-related gene, encodes a phosphorylated glycoprotein that regulates immune, inflammatory response and hematopoiesis (Kishimoto et al. 1995) as well as stimulates NK-cell-mediated killing of tumor cells and acts as an autocrine stimulator of cervical neoplastic cell growth (Iyori et al. 2011). Two well-studied IL6 SNPs [rs1800797 (Castro et al. 2009) and rs1800795 (Gangwar et al. 2009; Nogueira de Souza et al. 2006)] were reported to be associated with cervical cancer risk in Caucasians and Brazilians, but these SNPs were not observed in Chinese populations (MAF = 0) based on the HapMap genotyping database. In the present study, we used an in silico approach to select and genotype two SNPs in IL6 for Chinese populations. The rs2069837 SNP was first reported as a genetic risk factor of Alzheimer’s disease that is accompanied by a chronic inflammatory process (Combarros et al. 2009). Moreover, another association study of 433 cases and 1,375 controls demonstrated that the rs1800796 SNP, in high LD with rs2069837 (Fig. 1c), was significantly associated with lung cancer risk in Singaporean Chinese non-smoking females (Lim et al. 2011). Consistently, our finding provides further support for the effect of the IL6-rs2069837 SNP on cervical cancer risk in Chinese women.

To further explore high-order multiple-factor interactions in associations with cervical cancer risk, we performed MDR and CART analyses (Chen et al. 2007) and found that the rs2069837 variant AG/GG genotypes appeared to modify cervical cancer risk associated with age at primiparity and menopausal status. Menopausal status and age at primiparity were found to be the best one-factor model in Study 1 and Study 2, respectively (Supplementary Table S5). This inconsistency might be due to the use of two different normal control groups, and the combined analysis was necessary, in which possible risk factors or confounders were adjusted in multivariate analyses that minimized the selection bias. For menopausal status, epidemiological studies have shown that estrogen metabolite levels are associated with severity of cervical intra-epithelia lesions and may predict cervical cancer risk (Sepkovic et al. 1995). Young age at primiparity is also found to be one of risk factors for cervical cancer (International Collaboration of Epidemiological Studies of Cervical Cancer 2007), which is likely to be associated with a chronic inflammation as well, in which cytokines, like IL6, may increase the risk of developing cervical cancer (Macintyre et al. 2012). Moreover, one previous study with 1,757 Alzheimer cases and 6,295 controls had showed that IL6 genotype interactions did contribute to the dysregulation of inflammation (Combarros et al. 2009). Therefore, it is biologically plausible that IL6 genetic variations may modify the risk associated with inflammatory stimulation and interact with other risk factors for the development of cervical cancer.

Our further genotype–phenotype correlation analysis showed that rs2069837 variant AG/GG carriers had significantly higher levels of IL6 protein expression in the target tissues than did rs2069837 AA carriers. Accumulated data have showed that IL6 protein played important roles in cervical cancer development (Verhoog et al. 2011), and its over-expression was observed in cervical cancer cells (Abdelwahab et al. 2012; Vazquez-Ortiz et al. 2005; Veerapaneni et al. 2009; Wei et al. 2001). Meanwhile, IL6 levels in cervicovaginal secretions were associated with the severity of cervical cancer (Tjiong et al. 1999). Our findings further suggest that the rs2069837 SNP may increase the risk of cervical cancer by modifying IL6 protein expression. Although located at the intron region, the IL6 rs2069837 SNP was predicted to be at a TFBS of the gene that might participate in gene regulatory networks (MacQuarrie et al. 2011), including inflammation-related pathways. This kind of spacer sequences in intron regions may contain some unidentified functional elements, such as TFBS for known or uncharacterized transcription factors or perhaps other structural features not yet understood (He et al. 2011). However, whether the IL6 rs2069837 SNP is a functional one, or just a tagging one, needs to be determined by further functional studies, such as chromatin immunoprecipitation (ChIP)-ChIP, ChIP-sequencing assays in cell lines or to be tested in animal models (MacQuarrie et al. 2011), which may reveal the exact mechanisms underlying the observed association with the development of cervical cancer. On the other hand, cancer is a complex and multifactorial disease, and any single SNP may not be sufficient for the prediction of the overall risk (Galvan et al. 2010). Future studies should include more genes and more SNPs in genes, especially functional ones, involved in inflammatory pathways or other related genes associated with cervical cancer risk.

Several limitations of our study need to be addressed. First, this hospital-based case–control study may have selection bias and information bias, which may be minimized by frequency-matching cases and controls as well as the adjustment for potential confounding factors in final analyses. Second, due to the retrospective nature of the original study design and the lack of routine HPV screening, we did not have enough information on other risk factors, especially HPV infection, which could be potential confounders. This was because the hospital did not perform HPV and related subtype tests routinely for the diagnosis of all cervical cancer cases, let alone for the controls. A recent meta-analysis has demonstrated that the high-risk subtypes of HPV16, 18 and 45 accounted for a greater or equal proportion of HPV infections in cervical cancer compared with normal cytology, but others, like high-risk HPV33, 51 and 58, were in a reverse direction (Guan et al. 2012). Therefore, the HPV infection status could be a confounder in estimating the risk associated with genetic factors.

In summary, in this two-stage case–control study of 1,584 cases and 1,768 controls, we found the IL6-rs2069837 SNP to be associated with cervical cancer risk in Eastern Chinese women, and this genetic variant led to differential IL6 protein expression. Our findings also suggest some possible interactions between IL6 genetic variations and age at primiparity or menopausal status. However, well-designed large, prospective studies with detailed information about HPV infection are required to validate our findings.

Acknowledgments

This study was supported by the funds from “China’s Thousand Talents Program” Recruitment at Fudan University and by the Shanghai Committee of Science and Technology, China (Grant No. 12DZ2260100). We would like to thank Ya-Jun Yang and Jiu-Cun Wang from Fudan University for the DNA samples of 511 controls originated from the TZL. We also thank Yu-Hu Xin and Hong-Yu Gu from Fudan University Shanghai Cancer Center for the technical and immunohistochemical support, respectively.

Conflict of interest

We declare that we have no conflict of interest.

Supplementary material

439_2012_1245_MOESM1_ESM.doc (342 kb)
Supplementary material 1 (DOC 342 kb)

Copyright information

© Springer-Verlag Berlin Heidelberg 2012