The study population involved in genotype analysis comprised of 671 unrelated individuals (311 asthmatic children and 360 healthy controls) of the Hungarian (Caucasian) population. Asthma was diagnosed by respiratory medicine specialists according to the recent Global Initiative for Asthma guidelines (http://www.ginasthma.org/). Asthmatic patients were recruited from the Allergic Outpatient Consultation of the Budai and Heim Pál Children's Hospitals, Budapest. Atopy was determined by skin prick tests (Lofarma SpA, Milan, Italy) and specific IgE blood tests (RAST test, Roche kit, Roche Diagnostics, Mannheim, Germany) to environmental aeroallergens and food allergens. Atopy was defined by positive skin prick test to at least 1 allergen (wheal diameter 3 mm greater than saline control) and/or positive total or specific IgE levels. Total and specific serum IgE levels to more than 100 allergens were determined by 3gAllergy™ blood tests in Immulite 2,000 Immunoassay System (Siemens Healthcare Diagnostics; Deerfield, IL, USA). Serum IgE levels were classified as normal or high according to the following age-specific reference ranges (kU/L): 0-1 year, <15; 1-5 years, <60; 5-10 years, <90; and adult, <100.

The control children were randomly selected from the Orthopedic Department in the Budai Children's Hospital or from the Urological Department of Heim Pal Hospital, Budapest. Children in the control group had mild musculoskeletal alterations (like pes planus or scoliosis), phimosis, or other minor urogenital problems, showed no symptoms of asthma and required no medication. Older subjects were also included as controls. They were adult blood donors without asthma, and according to questionnaires they had not experienced asthma symptoms earlier (Table 1). Statistical adjustments and verifications were carried out to account for potential population stratification issues resulting from the slight mean age difference between cases and controls.

Thirty-four adults were enrolled in the study to measure their gene expression levels of the selected genes in induced sputum, but 11 of them were excluded from the analysis because of the quality of the sputum samples (the study population was comprised of 13 asthmatic patients and 10 controls). Their detailed characteristics were previously described by Ungvári et al.12 The asthmatic patients had mild to moderate stable asthma, with neither other lung diseases nor lower respiratory tract infection. Patients were required to have FEV1 of greater than 70% of predicted, baseline methacholine PC20 (the provocative concentration of methacholine causing a 20% fall in FEV1) of less than 16 mg/mL. Healthy volunteers were recruited from the staff and students of the participating Hungarian universities. They gave no history of respiratory diseases, and had a FEV1/FVC>80% and normal methacholine airway responsiveness (PC20>16 mg/mL). The included groups did not differ statistically in age, sex, smoking status, and allergy.

The study was conducted according to the principles expressed in the Declaration of Helsinki and approved by the Ethics Committee of the Hungarian Medical Research Council (ETT TUKEB) (Ethics Committee of the Hungarian Medical Research Council [http://www.ett.hu/tukeb/tukeb.htm]). Written informed consents were obtained from all patients, or the parents or guardians of the minors involved in the study.

The current study is based on our former gene expression microarray analysis of the ovalbumin-induced mouse model of asthma. The former experiment used female 6-8 weeks old, female, pathogen-free, BALB/c mice kept on an ovalbumin-free diet. Groups of mice were sensitized (on days 0 and 14) and challenged with ovalbumin (on days 28, 29, and 30), and 1 group was sensitized and challenged with placebos. On day 31, 24 hours after the third (last) allergen challenge, mice were anesthetized, bronchoalveolar lavage fluid was isolated, and the lungs were removed for further analysis. RNA was isolated from the lung tissues and gene expression patterns were measured with Agilent Whole Mouse Genome Oligo Microarray 4×44 K chips (record number GSE11911 in the GEO database). Microarray data was analyzed with unpaired t-test and one-way analysis of variance (ANOVA). Over thousand transcripts showed > 2.0-fold statistically significant differential expression in the 3 groups relative to the control group. See Tölgyesi et al.10 for further details.

The gene expression results of the mouse model were used to identify genes whose orthologues may be candidates for human asthma or atopy. A gene was selected for further consideration if its expression level in the lung changed significantly during the asthmatic process (corrected P value under 0.02, see the paper of Tölgyesi et al.10). The candidate list was ranked based on the strength of the regulation change, the potential role in asthma, earlier findings in the scientific literature, and their scientific novelty. The final list of the 60 selected genes is presented in Supplementary Material 1.

UCSC Table Browser was used to select SNPs.13 In the final SNP sorting, we preferred the promoter, missense, and UTR region polymorphisms. To reach the best coverage of a studied gene or genomic region, linkage disequilibrium (LD) data was involved in the selection process (International HapMap Project,14 Haploview software15).

Genomic DNA was isolated from whole blood samples using iPrep PureLink gDNA Blood Kit, iPrep Purification Instrument (Invitrogen).

A total of 90 SNPs were genotyped using the Sequenom iPLEX Gold MassARRAY technology at McGill University and Génome Québec Innovation Centre, Montréal, Canada.

Sputum induction was carried out as previously described by Ungvári et al.12,16 RNA was isolated with Qiagen Mini Rneasy Kit (Qiagen, MD USA) and transcribed to cDNA with High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA). Real-time quantitative PCR was performed for the selected genes using the ABI 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). The relative mRNA levels in each sample were normalized to its β-actin content, which proved to be a stable control reference.

Hardy-Weinberg equilibrium (HWE) was tested for the control population using the chi-square goodness-of-fit test (results in Supplementary Material 2) implemented in the online DeFinetti HWE application (Helmholtz Zentrum München, Institut für Humangenetik http://ihg.gsf.de/cgi-bin/hw/hwa1.pl). The significance level was set at 0.01. The same software was used to perform Pearson's chi-squared tests of association to asthma and calculate odds ratios (using single marker allele, recessive and dominant models). Confidence intervals (CI) were calculated at the 95 percent level. For each SNP, missing genotypes were imputed by sampling from the univariate distribution. The Haploview application15 was used to perform single marker and haplotype association tests. To account for multiple hypothesis testing, we performed permutation tests in Haploview with both single marker and haplotype associations with a minimum of 1,000 permutations. The single marker associations and the dominant-recessive models were also verified with IBM SPSS Statistics V20 (Pearson's chi-squared test); to account for multiple hypothesis testing in this case, we applied the more stringent Bonferroni correction.

In all multivariate logistic regression analyses (IBM SPSS Statistics V20), age was included as a covariate to apply sufficient statistical adjustment for the mean age difference between cases and controls (-11 years).

Additional statistical verifications were also carried out (IBM SPSS Statistics V20) to avoid false positive associations resulting from population stratification. We used binary logistic regression models with and without age as a covariate and assessed how the polymorphisms' P values and odds ratios differ. If the P values and odds ratios differed considerably, that would be a clear sign of a confounded relation between asthma and the polymorphism. However, the assessed P values and odds ratios were nearly identical in all of the models; the preventive or risk status of the predictor remained the same (the largest odds ratio differences were in the range of 20%-30%), thus no evidence for a population stratification was found.

For the analysis of the RT PCR measurements, we used the delta-delta-CT algorithm to normalize CT values for housekeeping genes and for the control samples. Samples with RNA Integrity Number of poor quality or extreme CT values were excluded from the study. We used Student's t-test on the normalized fold change values to assess statistical significance.

The basic characteristics of the studied SNPs (containing gene, rs id, functional classification and alleles) are presented in Supplementary Material 1, whereas genotype frequencies in controls and cases, odds ratios, and results for statistical tests of HWE and of population association are available in Supplementary Material 2. The total percentage of missing genotype data was 7.1%. Patients with over 80% missing genotype data were excluded (n=9, ~1%), and polymorphisms with over 20% missing ratio were also excluded (rs2432561, rs3814896, rs1821142, rs17027410). Out of the remaining 86 polymorphisms 2 showed significant deviation from the HWE in the control population, thus they were also excluded (rs11016071, rs2857600).

Between asthma cases and controls, the genotype distributions of 4 SNPs of 2 genes differed significantly in all of the applied statistical tests: SCIN (rs2240572, rs2240571, rs3735222) and PPARGC1B (rs32588). The polymorphisms of 6 other genes (ITLN1, FABP3, MAT1A, OSGIN, LY9, LGMN) showed statistically borderline differences (they were significant with most of the statistical tests but failed multiple testing correction with permutation testing: 0.05<P value <0.5). However, based on its statistical significance in the genotyping study, our results in mice, and its possible role in the asthmatic process, we included one of these genes in our further investigations: ITLN1 (rs4656958). We examined if the P values of the dominant and recessive models show a difference compared to the single marker allele associations tests, but none of the uncorrected P value differences were notable, thus these models were not examined further (Table 2). The single marker results and the dominant-recessive model ones were also verified by IBM SPSS Statistics V20 with Pearson's chi-squared test combined with Bonferroni correction (the corrected significance threshold for 90 polymorphisms was 0,00055). The results are in a complete agreement with our earlier findings, in which the same polymorphisms proved to be significantly associated with asthma.

To see if the polymorphisms have greater combined effects compared to what was seen in single SNP analyses, we tested for their association to asthma at the level of haplotype blocks. This analysis confirmed the present single SNP findings, but the haplotype-level P values did not differ significantly from the corresponding single marker results: the association of the SCIN haplotypes was very significant, and the weak association of ITLN1 was stronger compared to the single marker results, but the latter was still not significant after correction for multiple hypothesis testing (Table 3). We calculated LD coefficients between the different SNPs (Table 4) and found that, in both genes, the significantly associated SNPs were in strong linkage with each other. This suggests that, in each gene, there is only 1 causal variant, which mediates the association of all nearby SNPs to asthma. In PPARGC1B, only 1 polymorphism was studied; therefore, no haplotype was investigated.

We analyzed the multivariate relevance of the polymorphisms investigated with logistic regression using age and gender as covariates. All the 3 previously selected SNPs (SCIN: rs2240572, PPARGC1B: rs32588, ITLN1: rs4656958) remained significantly related to asthma with protective effects. Similar to our previous results, a strong haplotype level association was not observed (1 SNP from each gene got into the model) (Supplementary Material 3).

We also examined the effects of these polymorphisms within each gender group using logistic regression. The effect of SCIN: rs2240572 (P=0.002, OR=0.41) and PPARGC1B: rs32588 (P=0.013, OR=0.48) was highly significant in women, but that of ITLN1: rs4656958 was not significant in this group. In men, ITLN1: rs4656958 (P=0.005, OR=0.59) was highly significant, but not the other 2 polymorphisms. Logistic regression also revealed that ITLN1: rs4656958 had a significant (P=0.007, OR=0.665) protective effect in the severe atopic asthmatic subgroup (Supplementary Material 3).

Taken together, these results indicate association to asthma of 3 genes in our study: SCIN, PPARGC1B, and ITLN1.

Next, we investigated whether gene expression levels in induced sputum samples corroborate the association of SCIN, PPARGC1B, and ITLN1. Indeed, the expression level of all 3 genes was significantly lower in the sputum sample of asthmatics than of controls.

We compared our results to the third measurement (24 hours after the third/last allergen challenge) of the original asthmatic mouse model study.10 The results were somewhat unexpected: the expression of all 3 genes changed in the opposite direction, i.e., the expression levels were significantly higher in the lung of mice with ovalbumin-induced airway inflammation (Figure).

We can apply Fisher's method to combine the P values of the statistically independent studies, such as the genotyping study and the gene expression studies in our case. This method quantitatively combines the strength of the evidence from both experiments in order to rank the results (Table 5). According to these, the SCIN gene had the lowest combined P value, indicating the strongest evidence for true association. The PPARGC1B also had a relatively low combined P value. It must be noted, however, that an insignificant combined P value does not mean the absence of association in this respect.