Eplicated in the two replication sets. eQTLSNPs on chromosome 4q31 are subdivided in two strong LD blocks (Figure S2). The strongest eQTL in Laval dataset, validated in both replication sets, was rs7667092 with BC029578 (Figure 6). The expression levels of the gene increased with the number of T alleles in all cohorts. In the three cohorts, this SNP explained 7.6 to 12.5 of the gene expression variance of BC029578. However, this polymorphism was not in LD with SNPs previously associated with COPD (r2 = 0.016). Two SNPs (rs1828591, rs13118928) previously associated with COPD were found to affect the expression of HHIP. Rs1828591 was the most significant SNP associated with HHIP in the Laval dataset. This eQTL was replicated in UBC, but not in Groningen (Figure 7). The G allele was associated with lower expression of HHIP in the Laval and UBC datasets. The same pattern was observed in the Groningen set, but the association was not significant.Table 2. SNPs associated with COPD in previous GWAS.Locus 4qSNP rs1964516 rsSNP positionStudy89,875,909 Cho et al. 2012. Human Molecular Genetics.11 89,883,979 Cho et al. 2010. Nature Genetics.10 Cho et al. 2012. Human Molecular Genetics.rs1903003 4q31 rs89,886,297 Cho et al. 2010. Nature Genetics.10 145,480,780 Cho et al. 2010. Nature Genetics.10 Pillai et al. 2009. PLoS Genetics.Lung eQTLs in the 19q13 LocusOn 19q13, 739 SNPs and 95 probesets covering 45 different genes were tested. The expression levels of RAB4B, MIA and CYP2A6 were not available in our datasets. 222 eQTLs were detected (Figure 8 and Table S3). 174 SNPs were Epigenetic Reader Domain regulating 11 probesets located on 10 genes (ZNF780A, SERTAD3, NUMBL, EGLN2, CYP2G1P, AXL, B3GNT8, LOC100505495, CEACAM21, CEACAM4). 210 eQTLs were validated in both replication cohorts. SNPs associated with gene expression were distributed across four LD blocks (Figure S3). 26 SNPs were associated with the expression levels of CEACAM21 and LOC100505495, and 3 others SNPs were associated with CEACAM21 and CEACAM4. The eQTLs on 19q13 were mainly located in two discrete foci one distal and one proximal to the COPD susceptibility locus RAB4B/rs145,486,389 Cho et al. 2012. Human Molecular Genetics.11 Pillai et al. 2009. PLoS Genetics.rs13141641 19q13 rs2604894 rs145,506,456 Cho et al. 2012. Human Molecular Genetics.11 41,292,404 Cho et al. 2012. Human Molecular Genetics.11 41,302,706 Cho et al. 2012. Human Molecular Genetics.doi:10.1371/journal.pone.0070220.tRefining COPD Susceptibility Loci with 23727046 Lung eQTLsFigure 1. Lung eQTLs on 4q22 in the Laval dataset. Each dot represents an association test between one SNP and one probeset. Only dots above the red line are significant (p,5.1061026). Significant SNPs were regulating the expression levels of PPM1K in red, GPRIN3 in blue, SNCA in green and MMRN1 in purple. The SNP with the smaller p-value is indicated. SNPs previously associated with COPD are presented at the bottom. doi:10.1371/journal.pone.0070220.gEGLN2/MIA/CYP2A6 (Figure 8). These eQTL-SNPs were not in LD with the COPD SNPs Autophagy rs7937 and rs2604894. The latter twoSNPs were in strong LD (r2 = 0.82) and rs7937 was genotyped in our lung eQTL dataset. Rs7937 was not associated withFigure 2. Boxplots of gene expression levels in the lung for PPM1K according to genotype groups for SNP rs17013978. The left y-axis shows the mRNA expression levels for PPM1K. The x-axis represents the three genotyped groups for SNP rs17013978. The right y-axis shows the proportion of the gene expression.Eplicated in the two replication sets. eQTLSNPs on chromosome 4q31 are subdivided in two strong LD blocks (Figure S2). The strongest eQTL in Laval dataset, validated in both replication sets, was rs7667092 with BC029578 (Figure 6). The expression levels of the gene increased with the number of T alleles in all cohorts. In the three cohorts, this SNP explained 7.6 to 12.5 of the gene expression variance of BC029578. However, this polymorphism was not in LD with SNPs previously associated with COPD (r2 = 0.016). Two SNPs (rs1828591, rs13118928) previously associated with COPD were found to affect the expression of HHIP. Rs1828591 was the most significant SNP associated with HHIP in the Laval dataset. This eQTL was replicated in UBC, but not in Groningen (Figure 7). The G allele was associated with lower expression of HHIP in the Laval and UBC datasets. The same pattern was observed in the Groningen set, but the association was not significant.Table 2. SNPs associated with COPD in previous GWAS.Locus 4qSNP rs1964516 rsSNP positionStudy89,875,909 Cho et al. 2012. Human Molecular Genetics.11 89,883,979 Cho et al. 2010. Nature Genetics.10 Cho et al. 2012. Human Molecular Genetics.rs1903003 4q31 rs89,886,297 Cho et al. 2010. Nature Genetics.10 145,480,780 Cho et al. 2010. Nature Genetics.10 Pillai et al. 2009. PLoS Genetics.Lung eQTLs in the 19q13 LocusOn 19q13, 739 SNPs and 95 probesets covering 45 different genes were tested. The expression levels of RAB4B, MIA and CYP2A6 were not available in our datasets. 222 eQTLs were detected (Figure 8 and Table S3). 174 SNPs were regulating 11 probesets located on 10 genes (ZNF780A, SERTAD3, NUMBL, EGLN2, CYP2G1P, AXL, B3GNT8, LOC100505495, CEACAM21, CEACAM4). 210 eQTLs were validated in both replication cohorts. SNPs associated with gene expression were distributed across four LD blocks (Figure S3). 26 SNPs were associated with the expression levels of CEACAM21 and LOC100505495, and 3 others SNPs were associated with CEACAM21 and CEACAM4. The eQTLs on 19q13 were mainly located in two discrete foci one distal and one proximal to the COPD susceptibility locus RAB4B/rs145,486,389 Cho et al. 2012. Human Molecular Genetics.11 Pillai et al. 2009. PLoS Genetics.rs13141641 19q13 rs2604894 rs145,506,456 Cho et al. 2012. Human Molecular Genetics.11 41,292,404 Cho et al. 2012. Human Molecular Genetics.11 41,302,706 Cho et al. 2012. Human Molecular Genetics.doi:10.1371/journal.pone.0070220.tRefining COPD Susceptibility Loci with 23727046 Lung eQTLsFigure 1. Lung eQTLs on 4q22 in the Laval dataset. Each dot represents an association test between one SNP and one probeset. Only dots above the red line are significant (p,5.1061026). Significant SNPs were regulating the expression levels of PPM1K in red, GPRIN3 in blue, SNCA in green and MMRN1 in purple. The SNP with the smaller p-value is indicated. SNPs previously associated with COPD are presented at the bottom. doi:10.1371/journal.pone.0070220.gEGLN2/MIA/CYP2A6 (Figure 8). These eQTL-SNPs were not in LD with the COPD SNPs rs7937 and rs2604894. The latter twoSNPs were in strong LD (r2 = 0.82) and rs7937 was genotyped in our lung eQTL dataset. Rs7937 was not associated withFigure 2. Boxplots of gene expression levels in the lung for PPM1K according to genotype groups for SNP rs17013978. The left y-axis shows the mRNA expression levels for PPM1K. The x-axis represents the three genotyped groups for SNP rs17013978. The right y-axis shows the proportion of the gene expression.