Introduction

Asthma is the most common chronic disorder in children, and exacerbation of asthma is a major cause of childhood morbidity and hospitalization. The prevalence of childhood asthma in Japan is 5.1% among infants, 6.4% among children, and 3.2% among adults, and 1.146 million patients received ongoing medical care for asthma in 1996 (Makino et al. 2005). Asthma is a complex and chronic disease in which allergen-induced inflammatory processes in the airways contribute to the development of symptoms, such as wheezing, cough, dyspnea, and breathlessness. Histopathologic studies of patients with asthma revealed infiltration of the airway wall by inflammatory cells such as T cells, eosinophils, and mast cells, and structural changes, i.e., thickening of all components of the airway wall, in asthmatic airways (Saetta and Turato 2001).

Asthma is believed to be a complex disorder involving genetic and environmental factors, and several asthma susceptibility loci have been identified through genome-wide screening (Daniels et al. 1996; Ober et al. 1998; Wjst et al. 1999; Dizier et al. 2000; Yokouchi et al. 2000; Laitinen et al. 2001). Tumor necrosis factor (TNF) gene is located on chromosome 6p21 within the major histocompatibility complex class III region (Carroll et al. 1987), which has been linked to asthma by several genome-wide linkage studies (Daniels et al. 1996; Wjst et al. 1999; Yokouchi et al. 2000; Nicolae et al. 2005). TNF is a potent pro-inflammatory cytokine and is released during allergic responses by both macrophages and mast cells (Pennica et al. 1984; Thomas et al. 1996; Thomas 2001). TNF blockade inhibited late-phase airway hyperresponsiveness and airway eosinophilia in a mouse model of asthma (Choi et al. 2005), and anti-TNF therapy provided clinical improvement in patients with severe asthma (Howarth et al. 2005). These accumulated data support the idea that TNF plays a major role in the pathogenesis of asthma.

One TNF polymorphism, a G-to-A transition at position −308 (−308G/A, also referred to as TNF-α-308*1 and 308*2), has been investigated extensively for its association with asthma and asthma-associated phenotypes. Some studies have supported an association with asthma (Moffatt and Cookson 1997; Albuquerque et al. 1998; Chagani et al. 1999; Li Kam Wa et al. 1999; Winchester et al. 2000; Witte et al. 2002; Shin et al. 2004; Wang et al. 2004), whereas other studies have not (Louis et al. 2000; Zhu et al. 2000; Buckova et al. 2002). Moffatt and Cookson (1997) first reported that −308G/A was significantly associated with asthma in an Australian population (Moffatt and Cookson 1997). The −308G/A polymorphism was also reported to be associated with bronchial hyperreactivity (Li Kam Wa et al. 1999). Chagani suggested that the −308 G/A polymorphism may be a risk factor for asthma, but that it does not increase the risk of a fatal asthma attack (Chagani et al. 1999). To date, there have been no studies examining the association between the −308G/A polymorphism of TNF and asthma in the Japanese.

The goal of the present study was to investigate whether the −308G/A polymorphism is associated with asthma. We performed association studies in two independent Japanese populations and a meta-analysis of all available studies, and examined a possible association between asthma susceptibility and the −308G/A polymorphism.

Materials and methods

Study populations

Family study

The probands were mite-sensitive asthmatic children who visited the Pediatric Allergy Clinic of the University Hospital of Tsukuba (Japan). A full verbal and written explanation of the study was given to all family members interviewed, and 165 trio families (father, mother, and one affected child, 495 members), including 47 families used for our genome-wide screening (Yokouchi et al. 2000) gave informed consent and participated in the study.

Each family member was questioned regarding allergic symptoms and underwent a physical examination performed by a participating pediatrician. Asthma was diagnosed in participants according to the criteria of the National Institutes of Health, USA (NIH 1997). Patients had to satisfy the two following criteria:

  1. 1.

    Two or more episodes of wheezing and shortness of breath during the previous 12 months

  2. 2.

    Reversibility of the wheezing and dyspnea, either spontaneously or by bronchodilator treatment

Methacholine challenge testing was not done because of the young age of the asthma patients; however, differential diagnosis of asthma in the affected children was made by participating physicians or pediatricians who had treated the children for more than 2 years, and they confirmed each diagnosis of asthma. Because wheezing is often associated with viral respiratory infection in young children (Martinez et al. 1995), we excluded children younger than 3 years of age. The young adult patients included in this study had suffered from chronic asthma since childhood. Total serum IgE levels and IgE levels specific for Dermatophagoides farinae were determined with the Pharmacia CAP System (Uppsala, Sweden). This study was approved by the Ethics Committee of the University of Tsukuba.

Case–control study

All case–control participants were Japanese and gave written informed consent to participate in the study in accord with the rules of the Process Committee of the SNP Research Center, the Institute of Physical and Chemical Research (RIKEN), Japan.

The participants were 461 children with asthma (mean age 10.5 years), and 465 healthy individuals (mean age 44 years) who had neither respiratory symptoms nor a history of asthma-related disease. The children with asthma were patients at Osaka Prefectural Habikino Hospital or National Sagamihara Hospital. The 465 healthy control subjects were recruited through physicians’ interviews about whether they had been diagnosed with asthma and/or atopy. All patients with asthma were diagnosed according to the criteria described above. Clinical characteristics of the patients and control subjects are given in Table 1.

Table 1 The clinical details of the study populations
Full size table

Genotyping

Genomic DNA was extracted from peripheral blood leukocytes. Polymerase chain reaction (PCR) of the −308G/A polymorphism of TNF was carried out with primers 5′-AGGCAATAGGTTTTGAGGGCCAT-3′ and 5′-TCCTCCCTGCTCCGATTCCG-3′. Amplification conditions were 94°C for 10 min followed by 40 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 30 s, and a final extension of 72°C for 7 min. PCR products were digested with NcoI endonuclease at 37°C for 16 h. Digested PCR fragments were subjected to agarose gel electrophoresis and visualized by ethidium bromide staining and ultraviolet transillumination. Expected product sizes for the G allele were 87 and 20 bp, and the expected product size for the A allele was 107 bp.

Statistical methods

Genotype distributions were compared with those expected for samples from populations in the Hardy–Weinberg equilibrium with an exact test (Wigginton et al. 2005). Transmission disequilibrium testing (TDT) was performed with the ASPEX package (http://www.aspex.sourceforge.net/). The significance of differences in the allele and genotype frequencies between the case and control groups was determined by the Chi-squared test. P<0.05 was considered significant. The statistical power of TDT for detecting a polymorphism that directly affected risk was calculated according to the method proposed by Risch and Merikangas (1996). Statistical power for the case–control study was calculated as described previously (Schlesselman 1974).

Meta-analysis

MEDLINE and PubMed database searches were performed to identify studies related to asthma and the −308G/A polymorphism of TNF. For computer-based searches, we used a combination of key words related to asthma and TNF. Keywords used for the database search were as follows: TNF, tumor necrosis factor, polymorphism, variants, and asthma. Articles of any language were considered. To be eligible for inclusion in our study, studies had to meet the following criteria:

  1. 1.

    Design, case–control trial

  2. 2.

    Participants, asthmatic

  3. 3.

    Methods, genotype of −308G/A polymorphisms

  4. 4.

    Results, genotype data

We included our present data in the meta-analysis. To be included in the meta-analysis, asthma could be either physician-diagnosed or self-reported. The following studies were excluded:

  1. 1.

    Studies with incomplete genotype and phenotype data

  2. 2.

    Studies analyzing intermediate phenotypes such as bronchial hyperresponsiveness or wheezing

  3. 3.

    Studies of occupational asthma

  4. 4.

    Family-based association studies

Meta-analysis was performed by the Mantel–Haenszel approach, assuming a fixed effects model, and by the DerSimonian and Laird method, assuming a random effects model, with the Rmeta package (http://www.cran.r-project.org/bin/windows/base). We used allele frequency data from each eligible study to examine the association of asthma with −308G/A. Heterogeneity among studies was tested with Chi-squared statistics obtained by adding the weighted squares of the deviations of each estimate from the pooled estimate. Publication bias was examined by plotting a funnel of the reported effect assessed with the odds ratio against the standard error of the natural logarithm of the odds ratio (P>0.10; Egger et al. 1997). We estimated the odds ratios by comparing asthmatic participants with control subjects in the same study and calculated odds ratios under the prior hypothesis that individuals who had the −308A allele were more susceptible to asthma. To guard against the possibility that any one study contributed disproportionately to the summary odds ratios, we conducted a sensitivity analysis by recalculating the meta-analysis after omitting one study at a time (Azzam and Mathews 2003). The heterogeneity test was performed for each analysis to determine whether there was evidence of heterogeneity.

Results

TDT in asthmatic families

In the present study, we genotyped two independent asthmatic samples, family, and case–control samples. In the family study, frequencies of the −308GG, GA, and AA genotypes in the parents were 0.96, 0.04, and 0 respectively, and the distribution did not deviate from the expected Hardy–Weinberg equilibrium (P=1.0). Preferential transmission of the −308G or −308A allele was not observed by TDT (5 transmitted vs. 7 nontransmitted for −308A, P=0.78).

Case–control study

The results of the case–control study are shown in Table 2. The genotype distribution among control subjects did not deviate from the expected Hardy–Weinberg equilibrium (P=0.07 for case–control and P=0.78 for TDT families). Case and control subjects among families who underwent TDT were identified, and an artificially constructed case population consisting of transmitted parental alleles to the affected child and a control population of nontransmitted parental alleles, as described by Kirov et al. (1999). No significant differences in the genotype and allele distributions were observed between the case and control groups (P>0.05).

Table 2 Genotype distributions and odds ratio of the TNF-308G/A polymorphism in childhood asthma and controls
Full size table

Statistical power of the TDT and of the case–control samples was 0.54 and 0.74 respectively, at the alpha level of 0.05 if the relative risk of asthma in those persons carrying a putative risk allele was 1.8 compared with that in persons without the allele.

Meta-analysis

The meta-analysis included a total of 2,477 asthmatic and 3,217 control participants (Fig. 1; Moffatt and Cookson 1997; Albuquerque et al. 1998; Louis et al. 2000; Winchester et al. 2000; Buckova et al. 2002; Witte et al. 2002; El Bahlawan et al. 2003; Sandford et al. 2004; Shin et al. 2004; Wang et al. 2004; Bilolikar et al. 2005; Gupta et al. 2005). Several studies were not included because allele frequency data were not available (Chagani et al. 1999; Li Kam Wa et al. 1999; Trabetti et al. 1999; Lin et al. 2002; Di Somma et al. 2003; Li et al. 2006) or because of an incomplete phenotype (probable asthma; Zhu et al. 2000). We also excluded the family-based association study (Randolph et al. 2005) and a study of occupational asthma (Beghe et al. 2004) from our meta-analysis. The list of articles retrieved, and characteristics of the studies included are shown in Table 3. The genotype distribution among control subjects in each study did not deviate from the expected Hardy–Weinberg equilibrium (P>0.05). No publication bias was observed by funnel plot of the reported effects of −308G/A (P=0.77). Overall, the combined odds ratio with a fixed effects model for asthma was 1.29 (95% CI, 1.13–1.48, P=0.00012), and that with a random effects model was 1.31 (95% CI, 0.99–1.74, P=0.06). The result of a heterogeneity test was statistically significant (Q=55.94, df=14, P<0.00001) and was considered evidence of heterogeneity. Therefore, sensitivity analysis was performed by repeated calculation of the combined odds ratio estimate after omission of one study at a time.

Fig. 1
figure 1

Odds ratios (with 95% confidence intervals) of asthma with TNF-308G/A polymorphisms

Full size image
Table 3 Characteristics of the studies for the meta-analysis
Full size table

The result of sensitivity analysis is shown in Table 4. The result of the heterogeneity test was no longer significant after omission of the study by Shin et al. (2004), indicating that their study contributed disproportionately to the summary odds ratio, suggesting the existence of heterogeneity. The combined odds ratio with the fixed effects model and with the random effects model after removal of that study was 1.46 (95% CI, 1.27–1.68, P=0.0000001) and 1.46 (95% CI, 1.20–1.77, P=0.00014) respectively.

Table 4 Odds ratios (OR) and sensitivity analysis of meta-analysis for the TNF-308G/A polymorphism and asthma
Full size table

Meta-analysis of six Caucasian populations revealed significant evidence of heterogeneity (Q=15.47, df=5, P=0.0085). After omission of the study by Albuquerque et al. (1998), the homogeneity test was not significant (P=0.36). The combined odds ratio with the fixed effects model and random effects model for asthma in the Caucasian populations was 1.57 (95% CI, 1.27–1.92, P=0.000022), and 1.58 (95% CI, 1.27–1.96, P=0.000033) respectively.

Meta-analysis of seven Asian populations also showed significant evidence of heterogeneity (Q=36.14, df=6, P<0.00001). After omission of the study by Shin et al. (2004), the homogeneity test result was not significant (P=0.60). The combined odds ratio with the fixed effects model and random effects model in the Asian population was 1.71 (95% CI, 1.29–2.25, P=0.0002) and 1.70 (95% CI, 1.29–2.25, P=0.00019) respectively.

Discussion

In the present study, we did not detect a significant association between the −308G/A polymorphism of TNF and childhood asthma in two independent Japanese populations. However, meta-analysis revealed that the polymorphism was significantly associated with asthma, supporting involvement of the −308G/A polymorphism in the development of asthma.

Conflicting results are often obtained in studies of the genetics of complex traits; some studies may show an association between a polymorphism and a disease, whereas others fail to replicate the result. Discrepancies may be due to differences in the populations studied, difficulty in defining the phenotype, or lack of statistical power. Asthma is a complex disorder in which multiple genes and environmental factors are involved, and each genetic factor contributes modestly or weakly to the development of asthma. In the present study, the frequency of the −308A allele in the Asian populations was low, and the power to detect an association was not so high, especially in TDT. Therefore, we performed a meta-analysis to increase the power of the analysis for asthma and the −308G/A polymorphism and found an association between the −308G/A polymorphism and asthma susceptibility. To see the effect of the −308G/A polymorphism in specific ethnicities, we performed separate meta-analysis of Caucasian and Asian populations. The studies by Albuquerque et al. (1998) in Caucasian populations and Shin et al. (2004) in Asian populations contributed to heterogeneity in each ethnic group. After omission of these studies from our meta-analysis, results were consistent between the two ethnic groups, indicating that the −308A allele is associated with an increased risk of asthma.

Heterogeneity is considered when a given collection of studies is more heterogeneous than would be expected on the basis of sampling variance alone. This suggests that each of the studies measures an effect that is different from that in the other studies and that the variance comes from different study designs or populations in addition to the usual sampling variance. Sensitivity analysis revealed significant heterogeneity in the present meta-analysis (P < 0.00001), and the study by Shin et al. (2004) contributed to the heterogeneity (Table 4). After exclusion of that study, the test of heterogeneity for −308A was not significant (P = 0.052), and the summary odds ratio changed from 1.29 to 1.46. There are a number of reasons for the heterogeneity, such as differences in patients and control subjects and ethnicity, or differences in methodology. However, as shown in Table 3, the definition of case and control subjects in the study by Shin et al. (2004) did not differ from that in other studies, and differences in ethnicity or genotyping methods is unlikely to have caused the discrepancy in the present analysis.

It has been reported that the estimate of a genetic effect in the first report to describe an association is upwardly biased (winner’s curse phenomenon; Lohmueller et al. 2003). When the initial study showing an association of TNF with asthma (Moffatt and Cookson 1997) was excluded from our meta-analysis, the combined odds ratio for the −308A allele was reduced from 1.46 to 1.41, but the association with asthma was still statistically significant (P = 0.0000088 for the fixed effects model and P=0.001 for the random effects model). This indicates that the initial positive study did not substantially influence the results of the meta-analysis of the association of asthma with the −308G/A polymorphism.

In our previous study, the −1037C/T (−857C/T) polymorphism of TNF, −1211T/−1043C/−1037C (−1031T/−863C/−857C) haplotype of TNF, and the LTA/TNF haplotypes were associated with asthma (Noguchi et al. 2002; Migita et al. 2005). The present meta-analysis data showed that asthma is associated with the −308G/A polymorphism of TNF, further supporting the importance of the TNF region in the development of asthma. Screening of the −308G/A polymorphism of TNF may be beneficial in identifying individuals susceptible to asthma and for personalized health risk assessments.