An International Case-Control Study of Interleukin-4Rα, Interleukin-13, and Cyclooxygenase-2 Polymorphisms and Glioblastoma Risk

Glioblastoma, an aggressive primary malignant brain tumor, is the most common type of adult glioma and has a poor prognosis with only ∼3% of glioblastoma patients surviving 5 years after diagnosis (0.3% for people older than 65 years at diagnosis; ref. 1). Inherited syndromes account for only a small proportion of glioblastoma but familial aggregation of this tumor has been observed (2). The strongest known environmental risk factor for glioma is exposure to therapeutic doses of ionizing radiation (3); however, this risk factor accounts for only a small proportion of cases. In addition, evidence from one cohort and nine case-control studies (e.g., refs. 4-6) is consistent for an inverse association between self-reported asthma and allergic conditions and risk of glioma. Wiemels et al. (7) found that glioma patients have lower serum IgE levels than controls; however, the possibility that immunosuppression by the tumor itself or by its standard treatment with immunosuppressive drugs may lower IgE levels or eliminate allergies cannot be excluded.

IL-4 and IL-13 are genes that share a common IL-4Rα chain on their receptors and code cytokines (immunoregulatory proteins) that share functions. These cytokines play a central role in allergy by stimulating IgE synthesis (8), show strong antitumor activity in mice, and inhibit proliferation of astrocytoma and low-grade glioma in human cell lines (9, 10). The anti-inflammatory properties of IL-4 and IL-13 are probably based on their ability to induce regulatory T-cell expression (11) or, in the brain, to enhance cyclooxygenase (COX)-2 expression, thereby causing cell death of activated microglia (major inflammatory cells of the central nervous system; refs. 12-14). In addition to its potential role in inhibiting brain inflammation, COX-2 expression is associated with glioma angiogenesis (15). Consistent with this evidence, Sivak-Sears et al. (16) found an inverse association between nonsteroidal anti-inflammatory drug use and glioblastoma risk.

Studies of germ line polymorphisms of the IL-4Rα and IL-13 genes provide relatively strong support for a role of these genes in IgE production or allergy. For example, in a study of 4,570 DNA samples from a British cohort, Maier et al. (17) found “robust confirmation” for an association between allelic variation of the IL-13 gene and variation in total serum IgE levels. They further state that the strength and consistency of their findings show that this association is valid and not a false-positive result. These authors also report a statistically significant association (P = 0.04) between an IL-4Rα germ line variant (rs1801275) and IgE levels. In addition, a recent meta-analysis of seven studies of the association between this same IL-4Rα variant and allergic asthma (18) found evidence for an association [OR, 1.54; 95% confidence interval (95% CI), 1.14-2.08]. In the absence of functional studies, it is not possible to identify precisely which single nucleotide polymorphisms (SNP) are important in determining the association between the IL-4Rα and IL-13 genes and IgE or allergy (due in part to the correlation among SNPs on the same gene); however, these two genes seem to play a central role in IgE production and allergic disease.

In a previous analysis of data from a Swedish population-based case-control study, Schwartzbaum et al. (19) used four germ line polymorphisms on the IL-4Rα and IL-13 genes as indicators of allergy risk. The rationale for using SNPs related to allergies rather than self-report of allergy or measurement of IgE after glioma diagnosis is that, unlike these two estimates of the probability that an individual has allergic disease, SNPs provide an indicator that is not subject to differential case-control error. Schwartzbaum et al. found that genetic variants that increase the risk of IgE or allergies also decrease glioblastoma risk. Their finding for the IL-13, rs1800925 SNP was confirmed for glioma by Wiemels et al. (20) in a large case-control study of glioma (456 cases, 541 controls). Furthermore, Wiemels et al. report that the same IL-13 SNP was inversely associated with IgE levels among controls (P = 0.04) as was the IL-13 rs20541 SNP (P < 0.001). However, Wiemels et al. did not observe the relationships between the IL-13 rs20541 SNP and glioma nor did they find the same associations between the IL-4Rα SNPs and glioma previously seen for glioblastoma by Schwartzbaum et al. Differences between the two studies may be attributed to chance or variations in the distribution of SNPs among populations. Therefore, it is important that the association between these allergy-related SNPs and glioma be evaluated in diverse populations.

The purpose of the present study was to determine whether the same associations between IL-4Rα and IL-13 SNPs and glioblastoma that had previously been observed by Schwartzbaum et al. (19) in the relatively small Swedish data set (110 cases, 430 controls), would be found in a data set that included three additional population-based case-control studies from Southeast England, Denmark, and Finland. In addition to examining individual SNPs, as had Schwartzbaum et al., we also considered whether IL-4Rα or IL-13 haplotypes might be related to glioblastoma. A haplotype consists of sets of alleles on the same chromosome that are inherited together and may have more functional importance than individual SNPs. Therefore, evaluating haplotypes rather than SNPs may increase statistical power. Furthermore, because our selection of IL-4Rα and IL-13 SNPs was based on the central role of these genes in IgE-mediated allergies, we estimated associations between their haplotypes and self-reported allergies among general population controls.

The previous Schwartzbaum et al. study also evaluated the −765GC COX-2 polymorphism for evidence of an association with glioblastoma. Although they observed no statistically significant association in this study, the number of cases was relatively small and because there is evidence for a role of the COX-2 gene in brain inflammation and glioma angiogenesis, it seemed worthwhile to repeat the initial analysis in a larger study.

Four population-based case-control studies of primary brain tumors were carried out in Sweden, Denmark, Finland, and Southeast England. These studies were conducted in the Stockholm, Gothenburg, Lund, and Umeå regions of Sweden; throughout Denmark; all regions except Northern Lapland and Åland in Finland; and the Thames regions of Southeast England. All four studies followed the core protocol of the INTERPHONE Study, coordinated by the IARC (21), but with extensions to the study design such as a wider age range of patients, an extended questionnaire and collection of blood samples. Results from the Swedish study were reported previously (19); however, since the previous study was published, the sample of participants whose blood we had collected has changed slightly. One misclassified glioblastoma case was removed from the sample and 14 controls that had been inadvertently excluded were added. Nonetheless, because of the similarity of the published Swedish data set and the Swedish data set that we now analyze, inferences based on results of analyses of both samples are the same.

With the aim of reducing phenotypic heterogeneity, we restricted the present study to glioblastoma patients whose histologic diagnoses were coded 9440, 9441, or 9442 using criteria in the International Classification of Diseases for Oncology, third edition, 2000. These patients were identified through neurosurgery, neuropathology, oncology, and neurology centers. Lists of case patients were also obtained from the appropriate population-based cancer registries to ensure completeness of ascertainment. Eligible case patients were individuals diagnosed with primary glioblastoma between September 1, 2000, to February 29, 2004 (the dates of case ascertainment within this period vary among centers), at ages 20 to 69 years in the Nordic countries and 18 to 59 years in England, and resident in the study region at the time of diagnosis. We restricted the present analyses to cases diagnosed with glioblastoma, the most common type of adult glioma, to reduce possible etiologic heterogeneity among different categories of glioma.

Control participants in the Nordic centers were randomly selected from the population register for each study area and frequency matched to all adult brain tumor cases (including all glioma and meningioma cases) on age, sex, and region. In England, frequency-matched controls were randomly selected from general practitioners' practice lists. Control participants were subject to the same age and residence criteria as case patients and the criterion that they had not been previously diagnosed with a brain tumor.

Eligible cases were approached either by mail, or personally at the clinics with written information about the study, and asked to participate in the study, whereas the controls were first approached by mail. If the subjects contacted by mail did not respond, another letter was sent or the subject was approached by telephone. Before asking for participation, study subjects received both an invitation letter and written information about the study. Informed consent was obtained from all study participants. Each study was approved by local ethics committees and informed consent was obtained from all study participants before the start of interview.

Computer-assisted personal interviews were conducted in participants' homes or other convenient locations (e.g., hospital rooms, offices, etc.) by trained interviewers. Information collected on allergic conditions included questions concerning previous diagnoses of asthma, hayfever, or eczema and whether these conditions were still present at the time of interview. In Sweden, Southeast England, and Denmark, a blood sample was drawn from cases and controls who participated in the interview and agreed to give blood. Finnish investigators did not attempt to collect blood from all interviewed participants; rather, they drew blood from a convenience sample, of predetermined size, from patients in three hospitals or controls living in the area served by these hospitals.

Dynamic allele-specific hybridization (22-26) was conducted to identify the IL-4Rα (rs1805015, rs1801275) and IL-13 (rs20541, rs1800925) genotypes. For this, two PCR primers and one dynamic allele-specific hybridization probe per target SNP were designed by means of custom software (27) supplied by DynaMetrix Ltd. To verify successful PCR amplification, several randomly chosen samples were examined on a 3.0% low-melt agarose gel. Dynamic allele-specific hybridization analysis of the PCR product was then conduced on membrane macroarrays, using the dynamic allele-specific hybridization-2 protocol (25). A random sample of 15% of all DNA samples was reassayed and the genotype assignment was confirmed. An error rate of <1% was observed.

We assessed country-specific odds ratios (OR) using unconditional logistic regression adjusted for the variables sex and age at diagnosis for cases (or for the controls, their age at their interview date with adjustment for the interval between the diagnostic and interview date of the cases and the difference between the mean interview dates of cases and controls). We also examined whether region within each country, on which controls were also matched (except in Denmark), should be included in the country-specific models. To do so, we conducted within-country analyses adjusted for age, sex, and region and found that results were almost identical to within-country analyses adjusted for age and sex only. We therefore did not adjust for region within country of residence as a potentially confounding variable.

To find out whether we could legitimately estimate the combined effects of polymorphisms across countries of residence, we conducted global tests for heterogeneity among countries. We used log-likelihood ratio tests to compare models with glioblastoma as the dependent variable, including main SNP and country effects and SNP-country interaction terms to models with only SNP and country main effects. Results suggested heterogeneity among countries for both IL-4Rα SNPs (P < 0.10) but none for the IL-13 SNPs or the COX-2 SNP. ORs for two IL-4Rα haplotypes and one IL-13 haplotype also showed evidence of heterogeneity among countries. We therefore present country-specific IL-4Rα SNP results as well as total and, for consistency, also show country-specific as well as total findings for the IL-13 SNPs and all seven IL-4Rα and IL-13 haplotypes. When we restricted the analysis to controls and used all allergies combined as the dependent variable, we found no evidence for heterogeneity. We therefore do not present country-specific findings for the haplotype analyses of allergies among glioblastoma general population controls.

To pool the SNP and haplotype data over countries, we used two-stage random effects meta-analysis (28) where country-specific ORs were first adjusted for sex and age and then summarized using random-effects meta-analysis. For our data, ORs using this meta-analytic method were similar to those obtained by excluding the Finnish data (Finnish results for the IL-4Rα SNP seemed to differ from those of the other three countries) and pooling data from the remaining three countries. However, because CIs for the ORs produced by the meta-analytic method were slightly wider than those produced by pooling data from three of four countries, we report only ORs from the meta-analysis. Our rationale for this choice is based on the high prevalence of false-positive genotyping findings in the literature. However, because our meta-analytic method rests on country-specific results, we did not have adequate numbers of observations to evaluate genotype trends.

Haplotype probabilities for the two IL-4Rα (rs1805015, rs1801275) and two IL-13 (rs20541, rs1800925) sequence variants were estimated using Proc Haplotype's (SAS Genetics) implementation of the expectation-maximization algorithm. Each individual in the sample is assigned a probability of having each haplotype or combination of SNPs (e.g., for IL-13 haplotypes, G-C, G-T, A-C, and A-T). The probability assigned by the program is based on individual genotypes together with the prevalence of their SNPs in the population. These probabilities are then used as weights to estimate haplotype frequencies. We excluded the IL-4Rα C-A haplotype from the analysis because it was found in less than 5% of the total population and therefore was more likely to produce an unstable result. We next estimated country-specific sex and age-adjusted ORs for each haplotype using unconditional logistic regression and summarized them using random-effects meta-analysis (28). Each haplotype is included in a separate model with the reference category consisting of individuals who did not have that specific haplotype.

To find out whether the IL-4Rα or IL-13 haplotypes were associated with self-reported allergies among controls in our data set, we divided controls into inclusive categories representing controls reporting any allergy, asthma, eczema, or hayfever and a comparison group representing controls reporting no allergies. The dependent variable for each of four unconditional logistic regression models consisted of one allergic condition category and the control group consisted of people who reported no allergies. Independent variables were three IL-4Rα and four IL-13 haplotypes, country, sex, and age.

In Table 1, the percentage males, median age at diagnosis, and associations between any allergy and glioblastoma are comparable for case and control participants in each country who did and did not have their blood drawn. Age and sex differences among cases and controls reflect our use of participants who were initially matched to the age and sex distribution of all glioma and meningioma case patients (see Materials and Methods).

In Table 2, summary ORs for both IL-4Rα polymorphisms provide no evidence for an association with glioblastoma (rs1805015 OR, 1.08; 95% CI, 0.70-1.66; rs1801275 OR, 1.00; 95% CI, 0.79-1.69) although ORs from the Southeast English data set (rs1805015 OR, 1.36; 95% CI, 0.87-2.12; rs1801275 OR, 1.45; 95% CI, 0.94-2.24) are consistent with previously reported Swedish findings (ref. 19; rs1805015 OR, 1.57; 95% CI, 1.01-2.45; rs1801275 OR, 1.56; 95% CI, 1.02-2.39).

Table 3 shows associations between haplotypes and glioblastoma risk. Only the T-G IL-4Rα haplotype is clearly related to glioblastoma risk (summary OR, 2.26; 95% CI, 1.13-4.52). Swedish, Southeast English, and Danish findings are consistent with this result but the Finnish findings are not, possibly because they are based on smaller numbers of observations.

As can be seen by comparing ORs in Table 4 with those in the last column of Table 3, 26 of 28 allergy-haplotype ORs are on the opposite side of the null from the corresponding glioblastoma-haplotype ORs. The T-G IL-4Rα haplotype ORs in Table 3 that is associated with glioblastoma (OR, 2.26; 95% CI, 1.13-4.52) is also related to a decreased risk of hayfever in Table 4 (OR, 0.38; 95% CI, 0.14-1.06).

We did not confirm associations between two IL-4Rα and two IL-13 polymorphisms and glioblastoma risk that were originally observed in the Swedish data set (19). We did, however, find an increased risk of glioblastoma among people with the T-G IL-4Rα haplotype. Whether this effect is merely a false-positive finding is not known, but it is noteworthy that a similar raised risk was apparent in several, although not all, countries. The inverse OR for hayfever, which would be expected if the T-G IL-4Rα haplotype accounts, in part, for the reduced risk of glioblastoma among people reporting hayfever, provides some additional plausibility for this finding.

In their large case-control study of glioma, Wiemels et al. (20) also found suggestive evidence for an association of borderline statistical significance between an IL-4Rα haplotype and glioma (OR, 1.49; 95% CI, 0.99-2.25). However, their haplotype includes four SNPs in addition to the two that we evaluated. When they restricted their haplotype analysis to the same IL-4Rα SNPs that we examined, they found no evidence for an association with glioma (OR, 1.13; 95% CI, 0.83-1.53). Although they do not report specific results of this analysis for glioblastoma cases (which would have allowed direct comparison with our findings), overall, they state that they observed no difference in outcomes when their analyses were limited to glioblastoma cases.

An additional finding of Wiemels et al. that may support results of the present study is the association that they observed between the IL-13 rs1800925 SNP and glioma (OR, 0.76; 95% CI, 0.57-1.00), which is similar to the result that we observed in the present study between the same SNP and glioblastoma (OR, 0.78; 95% CI, 0.59-1.04). However, they found no evidence for a relation between the two IL4-Rα SNPs that we studied and glioma.

The majority of haplotype-allergy ORs are on the opposite side of the null from the haplotype-glioblastoma ORs. Whether this pattern is attributable to sampling variation is not known; however, if it is not, then it suggests that the IL-4Rα and IL-13 haplotypes play a role in both conditions. Wiemels et al. observed an inverse pattern for both IL-13 SNPs and IgE levels among the controls; however, they did not find associations between the IL-4Rα SNPs and IgE levels. Furthermore, this inverse association observed by Wiemels et al. is consistent both with prior knowledge about the link between the IL-4Rα and IL-13 SNPs and allergies (29, 30) and the reported inverse relation between allergies and glioblastoma (4-6).

Although our sample is one of the largest reported glioblastoma case-control data sets, it is still possible that this sample was not large enough to identify case-control differences in the distribution of IL-4Rα or IL-13 SNPs. The pooled OR for the total sample was close to the null value and the previously reported IL-4Rα SNP Swedish findings (19) are replicated only in the Southeast English data set, which has approximately the same number of cases (n = 107) as the Swedish sample (n = 110). The Danish (n = 66) and Finnish (n = 44) samples are smaller and provide no evidence for these associations. Possibly more important in accounting for our failure to duplicate the IL-4Rα SNP Swedish findings than sample size was the evidence for heterogeneity of the IL4Rα SNP ORs among countries. This heterogeneity may reflect the role of unknown and therefore uncontrolled environmental factors in obscuring the effects of the allergy SNPs. Our failure to replicate our findings may be attributed to the differing prevalence among countries of genotypes on the IL-4/IL-13 pathway with which the IL-4Rα or IL-13 SNPs may interact. It is becoming clear that SNPs or even individual genes cannot be examined for their effect on disease in isolation but rather whole genetic pathways must be simultaneously investigated as well as pathways that perform similar functions.

Of central importance in determining the validity of our findings is the presence of prior knowledge (31). There are several lines of evidence that suggest a role for IL-4 and IL-13 cytokines in glioma etiology. In addition to their central role in allergic conditions, these cytokines resolve brain inflammation by inducing COX-2 and killing microglia. Both IL-4 and IL-13 suppress cell proliferation in the normal astrocytic and low-grade astrocytoma cell lines (9, 10), possibly by blocking angiogenesis (32). Additional evidence for an association between IL-13 and glioblastoma includes the fact that glioblastoma tissue is characterized by the presence of IL-13Rα2 decoy receptors that limit the biological availability of IL-13 (33).

Yet, in spite of a relatively strong biological rationale, we were unable to provide evidence for associations between the IL-4Rα and IL-13 SNPs and glioblastoma originally observed in the Swedish data set. We did, however, find an association between the T-G IL-4Rα haplotype and glioblastoma risk and also a suggestion of an inverse relation between IL-4Rα and IL-13 haplotypes and allergies. Potential reasons for our lack of confirmation of previous results include the absence of a relationship between the cytokines coded by these genes and glioblastoma risk or our failure to simultaneously consider other genes on the IL-4/IL-13 pathway or to identify the specific SNPs on the IL-4Rα and IL-13 genes involved in glioblastoma development. Cousin et al. (34) reported that at least six SNPs are required to characterize the variability of the IL-13 gene, whereas (35) four are needed to represent that of IL-4Rα. However, we evaluated only two IL-4Rα and two IL-13 SNPs in the present article. Although our findings weakly indicate an association between the IL-4Rα gene and glioblastoma risk, the relative strength of the previous biological evidence suggests that further research is needed to evaluate the possible role of the entire IL-4/IL-13 pathway in glioblastoma development.

We thank all patients, controls, interviewers, and team collaborators in the study in all the participating countries, as listed in Schwartzbaum et al. (36), for their invaluable contribution to this work.