In the first stage we conducted a GWA analysis using a DNA-pooling approach. Genotyping was performed using the Illumina HumanHap 300 Whole-Genome Genotyping BeadChip. The genome-wide analysis was conducted in three different sets of pooled samples: (1) subjects with atopy and asthma; (2) subjects with atopy and without asthma, and (3) control subjects with neither asthma nor atopy (Table 1). Samples for this analysis were randomly selected from United Kingdom, Spain and Germany cohorts of the European Community Respiratory Health Survey (ECRHS) study 9 . Atopy was defined as sensitization (IgE levels >0.35 kU/L) to specific allergens (D. Pteronyssinus, cat, timothy grass. or C. herbarum). Asthma was defined as the presence of attacks of asthma in the last 12 months or taking currently medication for asthma. All asthmatics subjects had also atopy. For pool construction, DNA samples were diluted and measured twice by double stranded DNA quantification using PicoGreen^® dsDNA reagent Kit (Invitrogen), and then normalized at 50 ng/μl. After visual inspection in 0.7% alkaline agarose gels, selected DNA samples were pooled in three independent pools of 75 individuals each one. We have controlled the quality of the Illumina genotyping process including a sample from HapMap (GM12873) with known genotype for the panel of 300K markers. We verified the concordance of genotypes with the HapMap database (concordance = 99.91%) and inconsistent single nucleotide polymorphisms (SNPs) were excluded from the analysis (n = 296, 0.09%). A pool of HapMap (n = 57) subjects were also performed and genotyped to validate the allele frequency estimation. SNPs with a poor performance in the estimation of allele frequency were excluded from the analysis (n = 3271, 1.03%) using a threshold of > 12% of difference, as shown in previous reports 7 10 , Finally, three replicates of each pool of ECRHS subjects were constructed to account for sampling errors in the analysis.

Allele frequencies for each SNP were estimated from pooled samples correcting by the ratio of intensity of both alleles 11 . Allele frequencies in each pool were compared using the 1 df chi-square T statistic for testing differences between two proportions accounting for experimental and sampling errors 11 . Multiple testing was controlled using a false discovery rate (FDR) at 5% 12 . Allele frequency estimation was validated using a pool of 57 samples from HapMap whose individual genotypes for 300K Illumina panel were previously known. Predicted and real allele frequencies showed a strong correlation (Pearson correlation coefficient, r = 0.99).

In a second stage, replication was performed by individual genotyping of subjects included in the pools (n = 225) and in an additional set of 429 atopic subjects (46% with asthma) and 222 controls subjects randomly selected from the ECRHS study. Only promising SNPs that were still significant after correction by multiple testing were followed up. Target regions were delimited by non significant SNPs (p > 0.05) upstream and downstream from the SNP significant associated after multiple testing corrections in the first stage (see Additional file 1: table S1). Complex regions with genomic structural variants were discarded for this study (chr 1 and chr 5) 13 . For the saturation we used a tag SNP approach forcing the inclusion of the polymorphisms detected in the first stage, or their perfect tags (r² ≥ 0.8). From the HapMap project data set, we utilized genotypes from the public release 21a (phase II; NCBI35, dbSNPb125) corresponding to 90 individuals from the CEPH 30 trios of European descent. The selection of tag SNPs was performed using the pair-tagging strategy implemented in Haploview software (v.4.0) with a MAF ≥ 0.05 and an r² ≥ 0.8 14 . In addition to HapMap SNPs, four other polymorphisms previously described to be associated with atopic dermatitis and situated in one of the targeted regions (COL29A1) were included 15 . The 53 SNPs selected were genotyped in individual samples using SNPlex technology (Applied Biosystems).

Genotyping quality was controlled by including negative controls and internal positive controls consisting in four replicates of two HapMap reference samples. Both, genotype concordances with the HapMap database and among replicates, were verified. Genotype data caring was done using SNP analysis to results software (SNPator) http://www.snpator.com/. Forty-five out of 225 pooled samples and 145 out of 651 replication samples were dropped from the individual analysis due to low volume of DNA available or low concentration (Table 1). Seven out of 53 genotyped SNPs were excluded from the analysis since deviation from Hardy-Weinberg equilibrium (HWE) (in controls p < 0.05) or showed a genotyping rate under 80%. The final replication set included 46 SNPs with an overall genotyping rate of 97% in 686 samples (Table 1). Genotyping assays were performed at the Barcelona Node of the "Centro Nacional de Genotipado" (CeGen) in Spain.

The statistical analysis for data derived from individual genotyping was performed assuming an additive genetic model and using logistic regression using R statistical software package 16 . Models were adjusted for age, sex and smoking status. Country of origin was considered in the regression models in order to discard confounding by population stratification 17 . Linkage disequilibrium (LD) blocks were estimated according to Gabriel et al. 18 method implemented in Haploview.

A human RNA tissue panel (Stratagene) including brain, heart, kidney, liver, lung, placenta, skeletal muscle, spleen, testis, ovary and thymus was used to evaluate the expression of SGK493. One μg of RNA was converted to cDNA using the SuperScript III First-Strand Synthesis System for RT-PCR (Invitrogen). SGK493 and the GAPDH genes were amplified in independent PCRs using the following primers SGK493-F: 5'-AAGTGACGGACCTGGATGAC-3' and SGK493-R: 5'-TGTGAGGCAGGAGGTATGTG-3' and GAPDH-F: 5'-ACCCAGAAGACTGTGGATGG-3' and GAPDH-R: 5'-TGCTGTAGCCAAATTCGTTG-3'. The expected fragment sizes according to hg18 are 189 bp and 415 bp, respectively. The PCR was performed on a final volume of 25 μl using 0.5 U of BioTherm DNA polymerase (GeneCraft), 2 mM of Mg, 50 μM of each dNTP, 0.12 μM of each primer and 1 μl or 0.5 ul of cDNA (SGK493 and GAPDH, respectively). Finally 10 μl of the SGK493 PCR and 5 μl of the GAPDH PCR were resolved on a 2% agarose gel.

The study was conducted in accordance with the Declaration of Helsinki Principles. Ethical approval was obtained for each centre from the appropriate institutional ethics committee and all subjects participating in this study provided informed consent.

A genome wide analysis was performed with three pools of DNA from atopic, atopic asthmatics and control individuals. Population characteristics are shown in Table 1. Significant results after correction for 5% FDR in the GWA of pooled DNA are shown in Table 2 (see Additional file 1: table S1 for the genome-wide surrounding SNPs). Seven SNPs were associated with atopy and one with atopic asthma. From these results, five regions were selected for individual genotyping at a second stage. Three of these regions contained genes putatively related to asthma or atopy: collagen type XXIX alpha 1 (COL29A1, also known as COL6A5, a new member of the collagen family 19 ), mitogen-activated protein kinase kinase kinase 5 (MAP3K5, also known as apoptosis signal-regulating kinase 1 (ASK1)), collagen type XVIII alpha 1 (COL18A1) and solute carrier family 19 member 1 (SLC19A1). The other two regions contained genes for which no obvious functionality could be inferred: Homo sapiens protein kinase-like protein SgK493 (SGK493) and a predicted gene in chromosome 8 (NT_007995.50) which presented two consecutive SNPs associated with atopy in the pooled analysis. Additional SNPs surrounding these six markers were genotyped at an individual level (see Additional file 1: table S2, for the results of the 53 SNPs genotyped at an individual level).

A summary of the positive results on individual samples for atopy is shown in Table 3. SNPs found to be significant for atopy at individual level in both pooling samples and replication set were located in SGK493 and MAP3K5 genes. Variants in COL18A1 and COL29A1 were not consistently replicated in both samples, although a variant in the COL18A1 (at 5' of the SLC19A1 region) was significant in pooling samples (p = 0.005) and border significant in the replication (p = 0.06). For variants listed in Table 3, the combined analysis of samples from pooling and the additional case control set, led to lower p values, indicating that genetic effects are in the same direction in both sets 20 .

The region in chromosome 2 containing the SGK493 gene showed the SNPs with the most significant p values in the association with atopy, and similar results were observed for the subset of atopic individuals with asthma (see Additional file 1: table S2). The variant rs4952590 remain significant even after Bonferroni correction (46 SNPs, p ≤ 0.001) in the both pooling and replication samples. This variant was associated with a reduction in risk of atopy in both samples (odds ratio (OR) = 0.18, 95% confidence interval (CI) 0.07-0.47 for pooling sample and OR = 0.52, 95%CI 0.35-0.79 for replication sample). In addition, variant rs13409978 in the pooling sample and three SNPs (rs11124858, rs13409978, rs4952590) remained significant in the combined analysis after Bonferroni correction. Figure 1 shows the representation of the linkage disequilibrium (LD) pattern for the 6 SNPs of SGK493 genotyped for all individuals of the sample (n = 699). Two blocks of disequilibrium were defined according to Gabriel et al. 18 method implemented in Haploview. Complete LD pattern of the region derived from HapMap data is summarized in the Additional file 1: figure S1. Haplotype analysis using the six SNPs (Table 4) confirms the results obtained in the single marker analysis. The only haplotype containing C allele of rs13409978 and T allele of rs4952590 is the only significantly associated with atopy (p = 0.00005). These SNPs are in high LD (D'= 0.97 and r² = 0.88). The expression pattern of SGK493 presented in Figure 2 shows that SGK493 is ubiquitously expressed in all the human tissues analyzed.

The intronic SNP rs9494554, located within MAP3K5, was associated with atopy in the pooling and replication samples, and remained significant in the combined analysis after Bonferroni correction. A similar association for rs9494554 was observed in the subsample of atopic asthmatic individuals (see Additional file 1: table S2). Other SNP in MAP3K5 (rs9402839) was nominally significant for atopy in pooling sample and border significant in the replication step.