RESEARCH ARTICLE Open AccessLimited overlap in significant hits betweengenome-wide association studies on twoairflow obstruction definitions in the samepopulationDiana A. van der Plaat1,2, Judith M. Vonk1,2, Lies Lahousse3,4, Kim de Jong1,2, Alen Faiz5,2, Ivana Nedeljkovic3,Najaf Amin3, Cleo C. van Diemen6, Guy G. Brusselle3,4,7, Yohan Bossé8, Corry-Anke Brandsma5,2, Ke Hao9,Peter D. Paré10, Cornelia M. van Duijn3, Dirkje S. Postma11,2 and H. Marike Boezen1,2*AbstractBackground: Airflow obstruction is a hallmark of chronic obstructive pulmonary disease (COPD), and is defined aseither the ratio between forced expiratory volume in one second and forced vital capacity (FEV1/FVC) < 70% or< lower limit of normal (LLN). This study aimed to assess the overlap between genome-wide association studies(GWAS) on airflow obstruction using these two definitions in the same population stratified by smoking.Methods: GWASes were performed in the LifeLines Cohort Study for both airflow obstruction definitions innever-smokers (NS = 5071) and ever-smokers (ES = 4855). The FEV1/FVC < 70% models were adjusted for sex, age,and height; FEV1/FVC < LLN models were not adjusted. Ever-smokers models were additionally adjusted forpack-years and current-smoking. The overlap in significantly associated SNPs between the two definitions andnever/ever-smokers was assessed using several p-value thresholds. To quantify the agreement, the Pearsoncorrelation coefficient was calculated between the p-values and ORs. Replication was performed in theVlagtwedde-Vlaardingen study (NS = 432, ES = 823). The overlapping SNPs with p < 10− 4 were validated in theVlagtwedde-Vlaardingen and Rotterdam Study cohorts (NS = 1966, ES = 3134) and analysed for expressionquantitative trait loci (eQTL) in lung tissue (n = 1087).Results: In the LifeLines cohort, 96% and 93% of the never- and ever-smokers were classified concordantly basedon the two definitions. 26 and 29% of the investigated SNPs were overlapping at p < 0.05 in never- and ever-smokers, respectively. At p < 10− 4 the overlap was 4% and 6% respectively, which could be change findings asshown by simulation studies. The effect estimates of the SNPs of the two definitions correlated strongly, but thep-values showed more variation and correlated only moderately. Similar observations were made in theVlagtwedde-Vlaardingen study. Two overlapping SNPs in never-smokers (NFYC and FABP7) had the same directionof effect in the validation cohorts and the NFYC SNP was an eQTL for NFYC-AS1. NFYC is a transcription factor thatbinds to several known COPD genes, and FABP7 may be involved in abnormal pulmonary development.(Continued on next page)* Correspondence: h.m.boezen@umcg.nl1Department of Epidemiology, University of Groningen, University MedicalCenter Groningen, Hanzeplein 1, 9700, RB, Groningen, The Netherlands2Groningen Research Institute for Asthma and COPD (GRIAC), University ofGroningen, University Medical Center Groningen, Groningen, TheNetherlandsFull list of author information is available at the end of the article© The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Plaat et al. BMC Pulmonary Medicine           (2019) 19:58 https://doi.org/10.1186/s12890-019-0811-0(Continued from previous page)Conclusions: The definition of airflow obstruction and the population under study may be important determinantsof which SNPs are associated with airflow obstruction. The genes FABP7 and NFYC(-AS1) could play a role in airflowobstruction in never-smokers specifically.Keywords: Genome-wide association study, Genetics, Airflow obstruction, COPDBackgroundChronic obstructive pulmonary disease (COPD) is amajor cause of morbidity and mortality in the world andencompasses emphysema, chronic bronchitis, and smallairways disease [1, 2]. The diagnosis of COPD is largelybased on the presence of airflow obstruction, measuredby the spirometric assessment (post-bronchodilator) ofthe ratio between forced expiratory volume in one sec-ond and forced vital capacity (FEV1/FVC). The Globalinitiative for chronic Obstructive Lung Disease (GOLD)recommends to use a fixed cut-off for defining airflowobstruction, namely an FEV1/FVC ratio below 70% [3],whereas the American Thoracic Society/EuropeanRespiratory Society (ATS/ERS) guidelines recommend todefine airflow obstruction as FEV1/FVC below the lowerlimit of normal (LLN) [4]. The LLN is a reference valuebased on sex, age, height and ethnicity and is calculated asthe lower fifth percentile of a healthy reference population[5]. There is a considerable controversy about which def-inition should be used in research and clinical practice,since both may lead to misclassifications [5–8]. This hasimportant implications, since misclassifications may leadto inappropriate medication and therapies [9, 10].It is generally accepted that both genetic susceptibilityand environmental factors contribute to airflow obstruc-tion. Genetic variants associated with airflow obstructionhave been identified by several genome-wide associationstudies (GWAS), but different definitions of airflowobstruction and populations were used. [11–16] As anillustration, the case-control study including onlysmokers with > 2.5 pack-years by Pillai et al. used thefixed ratio (FEV1/FVC < 70%) to define airflow obstruc-tion, while the population based study including bothever- and never-smokers by Wilk et al. used the lowerlimit of normal (LLN) [15, 16]. Only few regions wereidentified in both studies, namely the CHRNA5/3 andHHIP regions. We therefore aimed to assess the geneticoverlap between the two definitions of airflow obstruc-tion in the same individuals. We stratified by smokingstatus to assess the overlap between the two airflowobstruction definitions in never- and ever-smokers sep-arately. We used the Lifelines Cohort Study as discoverysample and the Vlagtwedde-Vlaardingen study to repli-cate our observations. In addition, genetic loci associatedwith both airflow obstruction definitions could indicaterobust genetic associations with airflow obstruction,which could potentially be novel loci. We therefore, as asecondary aim, validated the top overlapping single-nu-cleotide polymorphisms (SNPs) between the two airflowobstruction discovery analyses in an independent SNPvalidation sample and assessed if they were acting as ex-pression quantitative trait loci (eQTLs) in a lung tissuesample.Materials and methodsStudy populationsTo study the overlap between the two airflow obstruc-tion definitions, all subjects with available genotypic datawere included from the Dutch LifeLines Cohort Study(discovery sample) and the Vlagtwedde-Vlaardingenstudy (replication sample) [17–19]. In addition, subjectsfrom the Vlagtwedde-Vlaardingen study and the threeindependent cohorts of the Rotterdam Study (RS I toIII) were selected to validate the top overlappingSNPs from LifeLines (SNP validation sample), therebyincreasing the SNP validation sample size [20]. Allsubjects provided written informed consent and thestudies were approved by local medical ethics com-mittees. Smoking status was based on self-reportedsmoking history and pack-years smoked. In thestratified analyses never-smokers having smoked 0pack-years and ever-smokers having smoked > 5pack-years were included, thereby excluding subjectswith > 0 and ≤ 5 pack-years. Subjects were defined ashaving airflow obstruction based on having apre-bronchodilator FEV1/FVC ratio (%) < 70% or <LLN (based on Global Lung Initiative 2012(GLI-2012)) [21]. All subjects completed pulmonaryfunction testing according to ATS or ERS criteria [22].Additional details are provided in Additional file 1.GenotypingThe IlluminaCytoSNP-12 arrays were used to genotypeblood samples in LifeLines and the Vlagtwedde-Vlaar-dingen study. SNPs with a genotype call-rate ≥ 95%,minor allele frequency ≥ 1% and Hardy-Weinbergp-value ≥10− 4 were included. Non-Caucasian samplesand first-degree relatives were excluded based onself-reporting, outlier (Identity By State) and principalcomponent analysis. After quality control, 227,981 geno-typed SNPs were included in the discovery analyses(LifeLines) and 242,926 genotyped SNPs were includedPlaat et al. BMC Pulmonary Medicine           (2019) 19:58 Page 2 of 12in the replication analyses (Vlagtwedde-Vlaardingen).Only genotyped SNPs were included in the analyses toprevent introducing bias, since it is known that imput-ation can reduce the effect size estimation, especially ifhealthy controls are used as reference [23]. Bloodsamples in the Rotterdam study were genotyped withthe 610 K and 660 K Illumina arrays and similar QCcriteria as in the other cohorts were applied.Statistical analysisFour separate GWASes were performed assessing thegenetic associations between the two definitions ofairflow obstruction, stratified by smoking status, for boththe LifeLines (discovery) and Vlagtwedde-Vlaardingen(replication) studies. Logistic regression (additive geneticmodel) was performed using PLINK (v1.07) [24]. The“FEV1/FVC < 70%” model was adjusted for sex, age andheight. The “FEV1/FVC < LLN” model was not adjustedfor these variables, since they are included in the LLNcalculation. In ever-smokers, the models were addition-ally adjusted for pack-years and current-smoking. Weused different p-value thresholds to assess the numberof overlapping SNPs between the two definitions. Inaddition, to quantify the agreement of the resultsbetween the two definitions and between never-andever-smokers, we calculated the Pearson correlationcoefficient between the p-values and between the ORs.Power simulationsOr study has a relative small sample size (n = 5070) andtherefore relative low power. We assessed the effect oflow power on the overlap between the two definitions byincreasing our never-smoking discovery sample (Life-Lines) 2 (n = 10,140) and 4 (n = 20,280) times. Inaddition, to assess if our results were spurious, we usedour never-smoking discovery sample and randomly allo-cated 10 times the airflow obstruction cases but keepingthe same distribution as in our original dataset (FEV1/FVC < 70%: n = 548, FEV1/FVC < LLN: n = 401, overlap-ping cases: n = 371 (64%)). For both simulation studies,we repeated the GWAS analyses on both airflowobstruction definitions in the created datasets and com-pared the number of overlapping SNPs.Validation of overlapping SNPsOnly the top overlapping SNPs between the two airflowobstruction definitions in the discovery sample(LifeLines) were evaluated in the SNP validation sample,the Vlagtwedde-Vlaardingen study and RS I to III. A fix-ed-effects meta-analysis of the effect estimates weightedby the inverse of the standard errors from all four valid-ation cohorts was performed using METAL (v2011) [25].We considered replication if the meta-analysis p-valuewas below the Bonferroni corrected p-value defined as0.05/number of overlapping SNPs and, in addition, hadthe same direction of effect in all cohorts. Inaddition, SNP*ever-smoking interactions were esti-mated and we assessed if the overlapping SNPs wereassociated with gene expression levels in lung tissuewithin a 4 Mb window around the SNP (2 Mb on ei-ther side of the SNP), using data from the lung eQTLconsortium [26]. In total, 1087 subjects were includedin the linear regression model, adjusted for diseasestatus, age, sex, smoking, and cohort specific principalcomponents. SNPs with a p-value below theBonferroni corrected threshold (p = 0.05/number ofprobesets) were considered significant eQTLs. SeeAdditional file 1 and GEO accession numbersGSE23546 and GPL10379 for additional information.ResultsPopulation characteristicsThe LifeLines cohort (discovery sample) included5070 never-smokers and 4855 ever-smokers withcomplete data on all covariates (see Table 1). Of thenever-smokers in LifeLines, 96% had a concordantairflow obstruction classification for the two defini-tions: 89% did not have airflow obstruction and 7%did have airflow obstruction. The remaining 4% had adiscordant classification (see Additional file 1: Table S1A).Figure 1a shows that of all never-smoking subjects withairflow obstruction based on at least one airflow ob-struction definition (n = 578), 36% had a discordantairflow obstruction classification. Of the ever-smokers,93% was classified concordantly: 77% did not haveairflow obstruction and 17% did have airflow obstruc-tion. The remaining 7% had a discordant classification(see Additional file 1: Table S1B). Of all ever-smokingsubjects with airflow obstruction based on at leastone definition (n = 1138), 30% had a discordant air-flow obstruction classification (see Fig. 1a). Subjectswith an FEV1/FVC < 70% and > LLN were aged be-tween 41 and 85, and subjects with an FEV1/FVC >70% and < LLN were aged between 22 and 43. Theseand other characteristics of the airflow obstructiongroups separately for never- and ever-smokers inLifeLines are shown in Additional file 1: Table S2.The Vlagtwedde-Vlaardingen study (replication sample)included 432 never-smokers and 823 ever-smokers (seeTable 1). Of the Vlagtwedde-Vlaardingen study, 94% and90% of the never- and ever-smokers were classifiedconcordantly based on the two definitions (see Additionalfile 1: Table S1 C-D).The SNP validation sample used for the SNP valid-ation meta-analysis included 1966 never-smokers and3134 ever-smokers from the Vlagtwedde-Vlaardingenstudy and RS I to III (see Table 1).Plaat et al. BMC Pulmonary Medicine           (2019) 19:58 Page 3 of 12GWAS resultsThere was minimal population stratification in all ana-lyses of LifeLines, indicated by the genomic inflation fac-tor lambda (λ: 1.0002–1.0217, see Additional file 1:Figure S1). The results based on a p < 10− 4 of all fouranalyses in LifeLines are given in See (Additional file 1:Tables S3-S6), including the Manhattan plots (seeAdditional file 1: Figures S2 and S3). For comparison,the effect estimates of both airflow obstruction definitionare given in these tables. Summary statistics (p values,betas, and standard errors for all SNPs that were tested) ofthe GWAS result of both the Lifelines Cohort Study andthe Vlagtwedde-Vlaardingen study are provided inAdditional file 2.Overlap between the resultsWe used several p-value thresholds to assess the overlapbetween the GWAS results of both airflow obstructiondefinitions separately in never- and ever-smokers ofLifeLines (see Table 2). A threshold of 0.05 resulted inthe observation that 26% and 29% of the SNPs wereoverlapping between the two airflow obstruction defini-tions in never- and ever-smokers, respectively. Threepercent of the SNPs were overlapping between never-and ever-smokers for both definitions. A smaller p-valuethreshold resulted in a lower percentage of overlap e.g. athreshold of p < 10− 4 resulted in 4% and 6% overlappingSNPs between the two airflow obstruction definitions innever- and ever-smokers, respectively (see Fig. 1b), andzero overlap between never- and ever-smokers using thesame definition of airflow obstruction. Similar observa-tions were made in the replication sample theVlagtwedde-Vlaardingen study (see Table 2), since atp < 0.05 the overlap between the definitions was 24%and 25% in never- and ever-smokers, respectively, and2% of the SNPs were overlapping between never- andever-smokers for both definitions.The correlations between the SNP-specific p-valuesand ORs from the two airflow obstruction definitionswere 0.48 (p-value) and 0.78 (OR) in never-smokers,and 0.51 (p-value) and 0.81 (OR) in ever-smokers (seeFig. 2). Between never- and ever-smokers the correla-tions of the SNP-specific p-values were 0.0008 forFEV1/FVC < 70% and 0.002 for FEV1/FVC < LLN, andfor the OR the correlation was − 0.02 for bothdefinitions. Similar observations were made in thereplication sample, the Vlagtwedde-Vlaardingen study.The correlations between the two definitions were0.45 (p-value) and 0.76 (OR) in never-smokers, and0.41 (p-value) and 0.74 (OR) in ever-smokers.Between never- and ever-smokers the correlationswere − 0.001 (p-value) and 0.015 (OR) for FEV1/FVC< 70% and − 0.003 (p-value) and 0.004 (OR) for FEV1/FVC < LLN.Table 1 Characteristics of never- and ever-smokers included in the current studyNever-smokers Ever-smokersLifeLines Vla-Vlab RS I RS II RS III LifeLines Vla-Vlab RS I RS II RS IIIN with no missing data 5070 432 408 379 747 4855 823 640 583 1088Males, N (%) 1942 (38) 103 (23) 81 (20) 124 (33) 308 (41) 2312 (48) 590 (72) 375 (59) 343 (59) 531 (49)Age (yrs), median(min-max)46 (18–89) 54 (36–79) 78 (72–94) 71 (65–98) 62 (51–93) 49 (22–85) 53 (35–79) 79 (72–95) 71 (65–93) 62 (52–93)Height (cm), mean (SD) 174 (9) 166 (9) 163 (9) 166 (9) 171 (9) 175 (9) 173 (8) 168 (9) 171 (9) 172 (9)Current-smokers, N (%) – – – – – 2171 (45) 478 (58) 99 (16) 108 (19) 298 (27)Pack-years (yrs), mean (SD) – – – – – 17 (11) 27 (21) 29 (21) 29 (21) 27 (20)Pulmonary function, mean (SD)FEV1%predicted (%),mean (SD)a98 (13) 91 (13) 101 (18) 102 (16) 101 (16) 94 (14) 85 (15) 98 (23) 96 (21) 99 (19)FEV1/FVC (%),mean (SD)b78 (7) 76 (7) 77 (7) 79 (6) 78 (6) 75 (8) 72 (10) 74 (9) 74 (9) 75 (8)FEV1/FVC <70%, N (%) 548 (11) 77 (18) 51 (13) 27 (7) 75 (10) 1107 (23) 287 (35) 163 (26) 151 (26) 235 (22)FEV1/FVC < LLN, N (%) 401 (8) 49 (11) 9 (2) 9 (2) 25 (3) 833 (17) 212 (26) 63 (10) 62 (11) 122 (11)Both FEV1/FVC <70%and < LLN, N (%)371 (7) 49 (11) 9 (2) 9 (2) 25 (3) 802 (17) 207 (25) 63 (10) 62 (11) 122 (11)Moderate/severe COPD,N (%)c106 (2) 26 (6) 21 (5) 9 (2) 16 (2) 310 (6) 145 (18) 90 (14) 90 (15) 104 (10)Discovery sample = LifeLines cohort study, replication samples = Vlagtwedde-Vlaardingen (Vla-Vla) study, and SNP validation sample = Vla-Vla and RS I to IIIFEV1 forced expiratory volume in one second, FVC forced vital capacityaFEV1%predicted is based on the reference equation by GLI-2012 [21]bFEV1%IVC (inspiratory vital capacity) for the Vlagtwedde-Vlaardingen study (Vla-Vla)cCOPD GOLD stage 2 and up (FEV1/FVC <70% and FEV1%p < 80% based on pre-bronchodilator measurements)Plaat et al. BMC Pulmonary Medicine           (2019) 19:58 Page 4 of 12Power simulationsWe found that the percentage of overlap increased whenwe expanded our never-smoking identification sample 2and 4 times (see Additional file 1: Table S7). The overlapbetween SNPs with p < 10− 4 was 3.6% in the original data-set, 16.5% in the 2x dataset and 26.6% in the 4x dataset. Inaddition, when we randomly allocated cases 10 times inour identification sample, we found the percentage ofoverlap between the two definitions at p < 10− 4 variedbetween 0 to 16%, compared to 4% in the original dataset(see Additional file 1: Table S8).Validation of overlapping SNPsIn never-smokers of LifeLines, two SNPs were overlap-ping between the FEV1/FVC < 70% and < LLNdefinitions at a threshold of p < 10− 4 (see Table 3). Thefirst SNP (rs7519348) is located in an intron of the genenuclear transcription factor Y subunit C (NFYC), andthe second SNP (rs6913003) is located in an intron of fattyacid binding protein 7 (FABP7, see Additional file 1:Figure S4 and S5 for LocusZoom plots). The minor allelesof both SNPs were associated with a higher risk of airflowobstruction and had comparable odds ratios in bothanalyses. The SNP in NFYC (rs7519348) was significantlyassociated with FEV1/FVC < LLN in the SNP validationmeta-analysis (p = 0.034), but did not pass the multipletesting correction (0.05/2 = 0.025), and was not signifi-cantly associated with FEV1/FVC < 70% in the SNPvalidation meta-analysis (p = 0.07). The SNP in FABP7(rs6913003) was not significantly associated with FEV1/FVC < 70% or < LLN in the SNP validation meta-analyses(p = 0.08 in both), although the direction of effect was thesame in all independent cohorts. Both SNPs did not reachgenome-wide significance according to the Bonferroni-corrected threshold (p < 2.19 × 10− 7) in the discovery ana-lysis (LifeLines) or meta-analysis of both the discovery andSNP validation samples (see Table 3 and Additional file 1:Table S9). Yet, the odds ratios were comparablebetween all analyses (see Additional file 1: Table S10and Figure S6). These two overlapping SNPs were notassociated with airflow obstruction in ever-smokersand these associations were significantly differentFig. 1 Venn diagrams showing the overlap between the two definitions of airflow obstruction for the number of subjects classified as havingairflow obstruction (a) and the number of identified SNPs with p < 10− 4 (b) in LifeLines (discovery sample)Plaat et al. BMC Pulmonary Medicine           (2019) 19:58 Page 5 of 12between ever- and never-smokers as shown in theinteraction analysis (see Additional file 1: Table S11).In ever-smokers of LifeLines, three SNPs were overlap-ping between the two analyses in at p < 10− 4 (see Table 3).The first SNP (rs13118083) is annotated to hedgehoginteracting protein (HHIP, 342 kb away), but is locatedwithin the long non-coding RNA LOC105377462 accord-ing to the SNP database by NCBI (https://www.ncbi.nlm.-nih.gov/SNP/). The second SNP (rs7074210) is locatedapproximately 62 kb from ST8 Alpha-N-Acetyl-Neurami-nide Alpha-2,8-Sialyltransferase 6 (ST8SIA6), and the lastSNP (rs4930390) is annotated to Chromosome 11 OpenReading Frame 80 (C11orf80). The minor alleles of thefirst 2 SNPs were associated with a higher risk of airflowobstruction and the minor allele of rs4930390 with a lowerrisk. The effect was significantly different between never-and ever-smokers for SNP rs4930390 according to bothdefinitions and for rs7074210 in the FEV1/FVC < 70%analyses (see Additional file 1: Table S11). The three SNPswere not replicated in the SNP validation sample (seeTable 3 and Additional file 1: Tables S9-S10).Gene expression in lung tissueThe minor allele (G) of rs7519348 (overlapping SNP innever-smokers) was associated with higher gene expres-sion of NFYC Antisense RNA 1 (NFYC-AS1) in lungtissue (Fig. 3). Summary statistics of the eQTL analysisfor all overlapping SNPs at p < 10− 4 are provided inAdditional file 2.DiscussionWe investigated the genetic overlap between GWASesusing two airflow obstruction definitions in the samepopulation (FEV1/FVC < 70 or < LLN). We expected areasonable overlap in associated SNPs between the twodefinitions, since 96% of the never-smokers and 93% ofthe ever-smokers were classified the same way in thediscovery sample LifeLines. Surprisingly, only a very smallproportion (4% and 6%) of SNPs was overlapping at p <10− 4 (see Fig. 1). Even with different significance thresh-olds the overlap was limited (26% and 29% at p < 0.05) (seeTable 2). The same observation was made in the replica-tion sample, the Vlagtwedde-Vlaardingen study. In this co-hort, 94% and 90% of the never- and ever-smokers,respectively, were classified concordantly, but at p < 0.05only 24% or 25% of the SNPs were overlapping. Inaddition, the effect estimates for the two airflow obstruc-tion definitions correlated strongly in both cohorts but thep-values showed more variation and correlated only mod-erately resulting in different top-hits depending on the ob-struction definition (see Fig. 2). Thus, the chosen strategyand definition of airflow obstruction had a substantial in-fluence on the GWAS results. This implies that in adiscovery-replication design with a predetermined selec-tion p-value, different genetic variants would befollowed-up depending on the definition used. In addition,there was no correlation between the p-values nor betweenthe ORs of never- and ever-smokers in both cohorts. Noneof the selected SNPs overlapped between never- andTable 2 Table showing the number of SNPs with a p-value below the mentioned threshold for both FEV1/FVC < 70% and < LLNanalysis and the overlapThreshold Never-smokers Ever-smokers (> 5 py) Overlap never- and ever-smokers70% LLN Overlap 70% LLN Overlap 70% LLNDiscovery analysis (LifeLines)< 0.05 11,377 11,475 4755 (26%) 12,114 12,366 5445 (29%) 625 (2.7%) 616 (2.7%)<0.01 2232 2197 673 (18%) 2522 2493 824 (20%) 23 (0.5%) 21 (0.4%)< 10−3 222 233 54 (13%) 246 266 76 (17%) 1 (0.2%) 0< 10− 4 31 27 2 (4%) 21 29 3 (6%) 0 0< 10−5 4 4 0 3 2 1 (25%) 0 0< 10− 6 1 0 0 0 0 0 0 0<Bonferroni* 0 0 0 0 0 0 0 0Replication analysis (Vlagtwedde-Vlaardingen study)<0.05 10,702 10,295 4026 (24%) 12,592 12,571 4976 (25%) 487 (2.1%) 488 (2.2%)<0.01 1857 1807 544 (17%) 2567 2609 754 (17%) 17 (0.4%) 12 (0.3%)< 10−3 164 174 38 (13%) 223 262 41 (9%) 0 0< 10−4 18 19 6 (19%) 12 21 0 0 0< 10−5 7 3 2 (25%) 0 0 0 0 0< 10−6 1 0 0 0 0 0 0 0<Bonferroni* 0 0 0 0 0 0 0 0*P < 2.19 × 10−7Plaat et al. BMC Pulmonary Medicine           (2019) 19:58 Page 6 of 12ever-smokers at p < 10− 4, and at p < 0.05 the overlap wasonly 3% in LifeLines (discovery sample) and 2% inVlagtwedde-Vlaardingen (replication sample, see Table 2).The current study therefore also highlights the importanceof stratifying the analysis according to smoking status.The difference between results from the two definitionsmight be explained by the fact that obstructive airwaydiseases are heterogeneous diseases with multiple pheno-types, symptoms and comorbidities. It might thus bebeneficial for future GWA studies to focus more onspecific COPD subtypes rather than on a broad definitionof airflow obstruction or COPD that can be causedby multiple underlying physiologic and genetic mech-anisms. In previous GWA studies, in mainly smokers,on classical COPD phenotypes like emphysema andchronic bronchitis, the well-known general COPDgenes (HHIP, CHRNA and FAM13A) were consistentlyidentified [27–32]. Perhaps, to identify specific geneticpathways underlying specific COPD phenotypes weshould not study the classical COPD phenotypes, butrather clinical COPD subtypes based on symptoms,comorbidities or pathology.Fig. 2 Pearson correlation between the p-values (a/c) or OR (B/D) of FEV1/FVC < 70% and < LLN analyses separately for never- (a/b) andever-smokers (c/d) in LifeLines (discovery sample)Plaat et al. BMC Pulmonary Medicine           (2019) 19:58 Page 7 of 12The CHRNA5/3 and HHIP regions were overlappingbetween six previous GWA studies on airflow obstruc-tion, using different airflow obstruction definitions andpopulations [11–16]. In the current study, two of theidentified SNPs in ever-smokers were located in theCHRNA5 and HHIP regions as well, pointing towards arobust genetic association of these regions with airflowobstruction and COPD (see Additional file 1: Table S6).Likewise, most of previously identified regions associatedwith airflow obstruction or COPD were nominal signifi-cant (p < 0.05) in the current study (see Additional file 1:Table S12). Out of the 22 loci identified by the study ofHobbs et al, SNPs in 18 loci were associated with at leastone of the airflow definitions at a nominal significance(10 SNPs in never-smokers and 12 SNPs in eversmokers) [14]. In never-smokers, 6 of the 10 SNPs weresignificantly associated with both definitions and inever-smokers 7 of the 12 SNPs were significantly associ-ated with both definitions. Some SNPs were significantin both never- and ever-smokers (e.g. HHIP, PID1 andTHSD4), while others were either only significant innever-smokers (e.g. FAM13A, DSP and RIN3) or inever-smokers (e.g. CHRNA5, TET2 and ADGRG6). Inaddition, many of the loci previously associated withlung function outcomes (FEV1, FVC, and FEV1/FVC)were also nominal significant (p < 0.05) in the currentstudy (see Additional file 1: Table S13). Specifically, ofthe loci reported by Wain et al., 23 out of 28 loci forFEV1, 10 out of 17 loci for FVC and 38 out of 51 loci forFEV1/FVC were associated with at least one of theairflow definitions at a nominal significance [33]. Lastly,we also checked if the top overlapping SNPs were asso-ciated with lung function outcomes in our previousGWA studies on FEV1, FEV1/FVC and FEF25–75 [34, 35].A SNP annotated to HHIP was associated with FEV1/Table 3 Results of the overlapping SNPs identified in both genome-wide association studies on FEV1/FVC < 70% and FEV1/FVC <LLN in never- and ever-smokersDiscovery analysis SNP validation meta-analysis Direction ofeffect in theindependentcohortsaSNP Chr A1 MAF Gene Test OR SE P OR SE P I2Never-smokers (n = 5070) (n = 1966)rs7519348 1 A 33% NFYC (intronic) <70% 1.36 0.07 4.92 × 10−6 1.21 0.11 0.07 0.0 ++++ 0<LLN 1.37 0.07 2.27 × 10−5 1.40 0.16 0.03 0.0 +++++rs6913003 6 T 4% FABP7 (intronic) < 70% 1.90 0.13 8.99 × 10−7 1.48 0.22 0.08 0.0 +++++<LLN 1.83 0.14 1.83 × 10−5 1.72 0.31 0.08 0.0 +++++Ever-smokers (n = 4855) (n = 3134)rs13118083 4 A 45% HHIP (342 kb 5′) <70% 1.23 0.05 4.36 × 10−5 0.97 0.07 0.60 17.0 + 0 − − +<LLN 1.26 0.05 2.27 × 10−5 1.05 0.08 0.57 46.7 ++ 0 − +rs7074210 10 G 18% ST8SIA6 (62 kb 5′) <70% 1.35 0.06 3.08 × 10−6 1.05 0.08 0.57 64.7 + − +++<LLN 1.33 0.07 3.23 × 10−5 0.99 0.10 0.95 64.5 + − + − +rs4930390 11 G 24% C11orf80 (intronic) <70% 0.76 0.06 9.40 × 10−6 1.02 0.07 0.74 0.0 − + 0 – 0<LLN 0.73 0.07 6.68 × 10−6 1.01 0.09 0.91 10.1 −++ − −SNPs were selected based on having a p-value < 10−4 in both the discovery analyses on the fixed ratio of 70% and LLN. The logistic regression model of FEV1/FVC< 70% was adjusted for sex, age and height, the LLN model was not adjusted. Ever-smoking models were additionally adjusted for pack-years and current-smoking. Discovery sample = LifeLines cohort study, and SNP validation sample = Vlagtwedde-Vlaardingen and RS I to III. A1 =minor allele (effect allele),MAF =minor allele frequency, OR = Odds Ratio, SE = standard error and P = p-value, I2 = heterogeneity measurea Order: LifeLines, Vlagtwedde-Vlaardingen, and Rotterdam Study I to III. + represents an OR > 1, − represents an OR < 1, and 0 represents is an OR between 0.95and 1.05 (no effect)Fig. 3 Results of eQTL analysis in lung tissue for rs7519348, anoverlapping SNP in never-smokers. The unadjusted mean log2microarray intensity and 95% CI are plotted, obtained from ameta-analysis of three cohorts included in the lung eQTL datasetPlaat et al. BMC Pulmonary Medicine           (2019) 19:58 Page 8 of 12FVC and FEF25–75 in both never- and ever-smokers(results were replicated) and the CHRNA5/3 region wasonly associated with FEV1/FVC in ever-smokers. TheNFYC and FABP7 regions were associated with FEV1/FVC (p = 4.40 × 10− 4 and p = 1.87 × 10− 4) innever-smokers, and the FABP7 SNP was also associatedwith FEF25–75 levels (p = 0.026). Interestingly, the NFYCregion was also overlapping between the current studyand the study by Pillai et al. We identified multiple SNPsannotated to NFYC, whereas Pillai et al. identified a SNP(rs3767943) in the gene KCNQ4, which is located on theright side (3′) of NFYC [15]. The NFYC region mighttherefore be an interesting region to further study theunderlying mechanisms of its association with airflowobstruction.A SNP in the intron of NFYC and a SNP in FABP7were the two overlapping SNPs between the airflow ob-struction definitions at p < 10− 4 in never-smokers andshowed the same direction of effect in the five independ-ent cohorts. The minor allele of the SNP in NFYC(rs7519348) was associated with a higher risk of airflowobstruction. This gene is a highly conserved transcrip-tion factor that is predicted by GeneGlobe to bindpromoter regions of 218 genes (see Additional file 1:Table S14) including genes previously associated withlung related outcomes, like ADORA2B, AKAP9, CD163,ELMOD2, HLA-DPB1, ITPR2, KLF10 and SERPINA6[27, 36–42]. In more detail, HLA-DPB1 is a knownCOPD gene related to disease severity, SERPINA6 wasassociated with emphysema, a deletion in ADORA2Bwas shown to be associated with a decrease in lungfibrosis and pulmonary hypertension, and ELMOD2 is acandidate gene for familial idiopathic pulmonary fibrosis[27, 36, 39, 40]. The identified SNP was not associatedwith expression levels of NFYC in lung tissue, but wasan eQTL for a probeset annotated to NFYC-AS1. Thefunction of this specific antisense-RNA, which are gen-erally thought to have a regulatory role, is still unknown.The minor allele of the SNP in FABP7 (rs6913003)was also associated with a higher risk of airflow ob-struction in never-smokers. This SNP was not associ-ated with the expression of FABP7 or other genes inlung tissue. FABP7 is an intracellular lipid-bindingprotein, involved in long-chain fatty acids transportand cell proliferation [43]. It may be involved in ab-normal pulmonary development, since lower expres-sion of FABP7 was found in patients with congenitalcystic adenomatoid malformation [44]. In addition,higher expression of FABP7 was seen in clear cellrenal cell carcinoma and the authors suggested thatthe gene activates the ERK and STAT3 signallingpathways [45]. STAT3 was implicated to play a role inpulmonary inflammation and thus FABP7 mightindirectly be involved in airflow obstruction [46].We were aware of the risk for spurious findings due tothe low power of our study and thus we validated ourtop overlapping SNPs in 4 independent validationcohorts. We furthermore investigated the effect of lowpower on the overlap between the two definitions by in-creasing our dataset 2 and 4 times. We found that thepercentage of overlap increases when the sample size in-creases, but still the number of SNPs that do not overlapremains high, i.e. 73.4% when the sample size increased4-fold. So even when the study power is greatly in-creased, different SNPs will be found depending on theairflow obstruction definition tested. We also performeda simulation study by 10 times randomly allocatingairflow obstruction cases and based on this simulation,we have to conclude that the differences and overlap wefound could be chance findings, but that is why we vali-dated the overlapping SNPs in 4 independent validationcohorts.We only assessed a modest number of SNPs (n =227,981 SNPs) compared to previous large GWAS studies(n > 1 million SNPs), since we only included genotypedSNPs to prevent any bias by imputation. The disadvantageof this approach is that we may have a lower genomiccoverage. Another limitation of the current study is theuse of pre-bronchodilator measurements to define airflowobstruction, which preferably should be based onpost-bronchodilator measurements. Especially subjectswith asthma could be misclassified as having airflow ob-struction, but the results of the overlapping SNPs did notchange in a sensitivity analysis excluding asthmatics oradjusting for asthma (see Additional file 1: Table S15).Moreover, only a low number of never-smoking subjectshad an FEV1/FVC < LLN in the three Rotterdam Study co-horts, but nevertheless results were replicated in thesenever-smokers. Finally, the “FEV1/FVC < 70%” model wasadjusted for sex, age and height, but the “FEV1/FVC <LLN” model was not adjusted for these variables, sincethey are included in the LLN calculation. If we do howeveradjust the “FEV1/FVC < LLN” model for these variables,the results do not change. The top SNPs are the same andthe correlation between p-values for the LLN models ad-justed and not adjusted is 0.98. In addition, the reportedcorrelation in never-smokers between the two definitionswas 0.48 for p-values and 0.78 for OR. If we use the LLNadjusted model the correlation is 0.48 and 0.79, respect-ively. This confirms that we used appropriate models toassess the genetic overlap between the two airflowdefinitions.ConclusionsThe definition of airflow obstruction and the populationunder study may be important determinants of whichSNPs are associated with airflow obstruction, and thuson which variants are selected for replication. It isPlaat et al. BMC Pulmonary Medicine           (2019) 19:58 Page 9 of 12therefore important to use the same definition of airflowobstruction in future studies, especially in consortia. Inaddition, future studies should focus more on specificCOPD subtypes and subgroups (e.g. based on smokingstatus), since there was no overlap in results betweennever- and ever-smokers, pointing towards possibledifferent underlying mechanisms. Finally, our resultssuggest that the genes FABP7 and NFYC(-AS1) couldplay a role in the pathogenesis of airflow obstruction innever-smokers.Additional filesAdditional file 1: Supplementary methods, tables and figures.(DOCX 1790 kb)Additional file 2: GWAS summary statistics. (XLSX 139809 kb)AbbreviationsCOPD: Chronic Obstructive Pulmonary Disease; eQTL: Expression quantitativetrait locus; FABP7: Fatty acid binding protein 7; FEV1/FVC: Ratio of forcedexpiratory volume in one second to forced vital capacity; GWAS: Genome-wide association study; MAF: Minor allele frequency; NFYC: Nucleartranscription factor Y subunit C; SNP: Single-nucleotide polymorphismAcknowledgmentsWe thank Rob Bieringa, Joost Keers, René Oostergo, Rosalie Visser, and JanSchouten for their work related to data collection and validation in theLifeLines cohort study and the Vlagtwedde-Vlaardingen study. The authorswould like to thank the staff at the Respiratory Health Network Tissue Bankof the FRQS for their valuable assistance with the lung eQTL dataset at LavalUniversity. We are grateful to the study participants and all staff involved inthe LifeLines cohort study, Vlagtwedde-Vlaardingen Study, Rotterdam Studyand lung eQTL database. We would like to thank Anis Abuseiris, Karol Es-trada, Dr. Tobias A. Knoch, and Rob de Graaf as well as their institutions Bio-physical Genomics, Erasmus MC Rotterdam, The Netherlands, and especiallythe national German MediGRID and Services@MediGRID part of the GermanD-Grid, both funded by the German Bundesministerium fuer Forschung undTechnology under grants #01 AK 803 A-H and # 01 IG 07015 G for access totheir grid resources.FundingThis study is sponsored by grant number 4.1.13.007 of Lung Foundation(Longfonds), the Netherlands. DAvdP, KdJ and NA are supported by grantnumber 4.1.13.007 of Longfonds. LL is a Postdoctoral Fellow of the ResearchFoundation - Flanders (FWO). The LifeLines Biobank initiative has been madepossible by funds from FES (Fonds Economische Structuurversterking), SNN(Samenwerkingsverband Noord Nederland) and REP (Ruimtelijke EconomischProgramma). The Vlagtwedde-Vlaardingen cohort study was supported bythe Ministry of Health and Environmental Hygiene of the Netherlands andthe Netherlands Asthma Fund (grant 187) and the genetic data of the cohortwere funded by the Netherlands Asthma Fund (grant no. 3.2.02.51), theStichting Astma Bestrijding, BBMRI-NL (Complementiation project) and theEuropean Respiratory Society COPD research award 2011 to H.M. Boezen.The Rotterdam Study was supported by the Erasmus MC and ErasmusUniversity Rotterdam; the Netherlands Organisation for Scientific Research(NWO); the Netherlands Organisation for Health Research and Development(ZonMw); the Research Institute for Diseases in the Elderly (RIDE); theNetherlands Genomics Initiative (NGI); the Ministry of Education, Culture andScience; the Ministry of Health, Welfare and Sports; the European Commission(DG XII); and the Municipality of Rotterdam. Genotyping and gene expressionfor the lung eQTL study was funded by Merck & Co. Inc. The lung eQTL studyat Laval University was supported by the Fondation de l’Institut universitaire decardiologie et de pneumologie de Québec, the Respiratory Health Network ofthe FRQS, the Canadian Institutes of Health Research (MOP - 123369). Y.B. holdsa Canada Research Chair in Genomics of Heart and Lung Diseases. The sponsorsof this study played no role in the design of the study, data collection, analysis,interpretation or in the writing and submission of the manuscript.Availability of data and materialsSummary statistics of all analyses are available in Additional file 2. The eQTLlung tissue datasets analysed during the current study are available at theGene Expression Omnibus (GEO) repository, GEO accession numbersGSE23546 and GPL10379. LifeLines data is available (at costs) to all scientists.Scientists can apply for access to Lifelines data and samples by submitting aresearch proposal to the LifeLines biobank (www.lifelines.net).Authors’ contributionsDAvdP participated in the study design, analysis and interpretation of thedata, and drafting of the manuscript, tables and figures. HMB, DSP, CCvD andCMVD obtained funding. KdJ, HMB, DSP, JMV, CCvD, CMVD, IN and NAdetermined the study design, participated in the analysis and interpretationof data, and critically supervised writing of the manuscript. LL and GGBparticipated in collecting and analysing the Rotterdam Study data. AF, YB,CAB, DSP, KH, and PDP participated in setting up the lung tissue databaseand analyses. All authors approved the final version of the manuscript.Ethics approval and consent to participateWritten informed consent was provided by all subjects of all includedcohorts. The Lifelines Cohort Study and the Vlagtwedde-Vlaardingen studywere approved by the Medical Ethics Committee of the University MedicalCenter Groningen, Groningen, the Netherlands. The Rotterdam Study wasapproved by the medical ethics committee of Erasmus University. The eQTLlung tissue database was approved by the ethics committees of the Institutuniversitaire de cardiologie et de pneumologie de Québec (Laval) and theUBC-Providence Health Care Research Institute Ethics Board (UBC). The studyprotocol was consistent with the Research Code of the University MedicalCenter Groningen and Dutch national ethical and professional guidelines.Consent for publicationNot applicable.Competing interestsLL reports personal fees from Boehringer Ingelheim GmbH, non-financialsupport from Novartis, grants from AstraZeneca, grants and non-financialsupport from European Respiratory Society and Belgian Respiratory Society,outside the submitted work. The University of Groningen has receivedmoney for Professor Postma (DSP) regarding a grant for research from AstraZeneca, Chiesi, Genentec, GSK and Roche. Fees for consultancies were givento the University of Groningen by Astra Zeneca, Boehringer Ingelheim, Chiesi,GSK, Takeda and TEVA. All other authors declare no competing interests.Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.Author details1Department of Epidemiology, University of Groningen, University MedicalCenter Groningen, Hanzeplein 1, 9700, RB, Groningen, The Netherlands.2Groningen Research Institute for Asthma and COPD (GRIAC), University ofGroningen, University Medical Center Groningen, Groningen, TheNetherlands. 3Department of Epidemiology, Erasmus Medical Center,Rotterdam, The Netherlands. 4Department of Respiratory Medicine, GhentUniversity Hospital, Ghent, Belgium. 5Department of Pathology and MedicalBiology, University of Groningen, University Medical Center Groningen,Groningen, The Netherlands. 6Department of Genetics, University ofGroningen, University Medical Center Groningen, Groningen, TheNetherlands. 7Department of Respiratory Medicine, Erasmus Medical Center,Rotterdam, The Netherlands. 8Department of Molecular Medicine, Institutuniversitaire de cardiologie et de pneumologie de Québec, Laval University,Québec, Canada. 9Merck Research Laboratories, Boston, MA, USA.10Department of Medicine, Center for Heart Lung Innovation and Institutefor Heart and Lung Health, University of British Columbia, St. Paul’s Hospital,Vancouver, Canada. 11Department of Pulmonary Diseases, University ofGroningen, University Medical Center Groningen, Groningen, TheNetherlands.Plaat et al. BMC Pulmonary Medicine           (2019) 19:58 Page 10 of 12Received: 13 June 2018 Accepted: 11 February 2019References1. World Health Organisation (WHO). The top 10 causes of death, Fact sheet N°310. http://www.who.int/mediacentre/factsheets/fs310/en/ .2. Postma DS, Kerkhof M, Boezen HM, Koppelman GH. Asthma and chronicobstructive pulmonary disease: common genes, common environments?Am J Respir Crit Care Med. 2011;183:1588–94. https://doi.org/10.1164/rccm.201011-1796PP.3. Global initiative for chronic Obstructive Lung Disease (GOLD). GlobalStrategy for the Diagnosis, Management and Prevention of COPD 2017.2017. http://www.goldcopd.org/.4. Pellegrino R, Viegi G, Brusasco V, Crapo RO, Burgos F, Casaburi R, et al.Interpretative strategies for lung function tests. Eur Respir J. 2005;26:948–68.5. Mohamed Hoesein FAA, Zanen P, Lammers J-WJ. Lower limit of normal orFEV1/FVC <0.70 in diagnosing COPD: An evidence-based review. RespirMed. 2011;105:–907, 15. https://doi.org/10.1016/j.rmed.2011.01.008.6. Medbo A, Melbye H. Lung function testing in the elderly--can we still useFEV1/FVC<70% as a criterion of COPD? Respir Med. 2007;101:1097–105.7. Hardie JA, Buist AS, Vollmer WM, Ellingsen I, Bakke PS, Morkve O. Risk ofover-diagnosis of COPD in asymptomatic elderly never-smokers. Eur RespirJ. 2002;20:1117–22.8. Roberts SD, Farber MO, Knox KS, Phillips GS, Bhatt NY, Mastronarde JG, et al.FEV1/FVC ratio of 70% misclassifies patients with obstruction at theextremes of age. Chest. 2006;130:200–6.9. Sorino C, D’Amato M, Steinhilber G, Patella V, Corsico AG. Spirometriccriteria to diagnose airway obstruction in the elderly: fixed ratio vs lowerlimit of normal. Minerva Med. 2014;105(6 Suppl 3):15–21.10. Ramsey SD. Suboptimal medical therapy in COPD: exploring the causes andconsequences. Chest. 2000;117(2 Suppl):33S–7S.11. Cho MH, Boutaoui N, Klanderman BJ, Sylvia JS, Ziniti JP, Hersh CP, et al.Variants in FAM13A are associated with chronic obstructive pulmonarydisease. Nat Genet. 2010;42:200–2. https://doi.org/10.1038/ng.535.12. Cho MH, Castaldi PJ, Wan ES, Siedlinski M, Hersh CP, Demeo DL, et al. Agenome-wide association study of COPD identifies a susceptibility locus onchromosome 19q13. Hum Mol Genet. 2012;21:947–57. https://doi.org/10.1093/hmg/ddr524.13. Cho MH, McDonald ML, Zhou X, Mattheisen M, Castaldi PJ, Hersh CP, et al.Risk loci for chronic obstructive pulmonary disease: a genome-wideassociation study and meta-analysis. LancetRespiratory Med. 2014;2:214–25.https://doi.org/10.1016/S2213-2600(14)70002-5.14. Hobbs BD, de Jong K, Lamontagne M, Bosse Y, Shrine N, Artigas MS, et al.Genetic loci associated with chronic obstructive pulmonary disease overlapwith loci for lung function and pulmonary fibrosis. Nat Genet. 2017;49:426–32.https://doi.org/10.1038/ng.3752.15. Pillai SG, Ge D, Zhu G, Kong X, Shianna KV, Need AC, et al. A genome-wideassociation study in chronic obstructive pulmonary disease (COPD):identification of two major susceptibility loci. PLoS Genet. 2009;5:e1000421.https://doi.org/10.1371/journal.pgen.1000421.16. Wilk JB, Shrine NR, Loehr LR, Zhao JH, Manichaikul A, Lopez LM, et al.Genome-wide association studies identify CHRNA5/3 and HTR4 in thedevelopment of airflow obstruction. Am J Respir Crit Care Med.2012;186:622–32. https://doi.org/10.1164/rccm.201202-0366OC.17. Stolk RP, Rosmalen JG, Postma DS, de Boer RA, Navis G, Slaets JP, et al.Universal risk factors for multifactorial diseases: LifeLines: a three-generationpopulation-based study. Eur J Epidemiol. 2008;23:67–74. https://doi.org/10.1007/s10654-007-9204-4.18. van der Lende R Gezondheidsorganisatie T.N.O. R te G. Epidemiology ofchronic non-specific lung disease (chronic Bronchitis). A critical analysis ofthree field surveys of CNSLD carried out in the Netherlands. Van Gorcum; 1969.19. Gosman MM, Boezen HM, van Diemen CC, Snoeck-Stroband JB, LapperreTS, Hiemstra PS, et al. A disintegrin and metalloprotease 33 and chronicobstructive pulmonary disease pathophysiology. Thorax. 2007;62:242–7.20. Hofman A, Brusselle GG, Darwish Murad S, van Duijn CM, Franco OH,Goedegebure A, et al. The Rotterdam Study: 2016 objectives and design update.Eur J Epidemiol. 2015;30:661–708. https://doi.org/10.1007/s10654-015-0082-x.21. Quanjer PH, Stanojevic S, Cole TJ, Baur X, Hall GL, Culver BH, et al. Multi-ethnic reference values for spirometry for the 3–95-yr age range: the globallung function 2012 equations. Eur Respir J. 2012;40:1324–43. https://doi.org/10.1183/09031936.00080312.22. Miller MR, Hankinson J, Brusasco V, Burgos F, Casaburi R, Coates A, et al.Standardisation of spirometry. Eur Respir J. 2005;26:319–338.23. Khankhanian P, Din L, Caillier SJ, Gourraud PA, Baranzini SE. SNP imputationbias reduces effect size determination. Front Genet. 2015. https://doi.org/10.3389/fgene.2015.00030.24. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, et al.PLINK: a tool set for whole-genome association and population-basedlinkage analyses. Am J Hum Genet. 2007;81:559–75.25. Willer CJ, Li Y, Abecasis GR. METAL: fast and efficient meta-analysis ofgenomewide association scans. Bioinformatics. 2010;26:2190–1. https://doi.org/10.1093/bioinformatics/btq340.26. Hao K, Bosse Y, Nickle DC, Pare PD, Postma DS, Laviolette M, et al. LungeQTLs to help reveal the molecular underpinnings of asthma. PLoS Genet.2012;8:e1003029. https://doi.org/10.1371/journal.pgen.1003029.27. Kong X, Cho MH, Anderson W, Coxson HO, Muller N, Washko G, et al.Genome-wide association study identifies BICD1 as a susceptibility gene foremphysema. Am J Respir Crit Care Med. 2011;183:43–9. https://doi.org/10.1164/rccm.201004-0541OC.28. Manichaikul A, Hoffman EA, Smolonska J, Gao W, Cho MH, Baumhauer H,et al. Genome-wide study of percent emphysema on computed tomography inthe general population. The Multi-Ethnic Study of Atherosclerosis Lung/SNPHealth Association Resource Study. Am J Respir Crit Care Med. 2014;189:408–18.https://doi.org/10.1164/rccm.201306-1061OC.29. Castaldi PJ, Cho MH, San Jose Estepar R, McDonald ML, Laird N, Beaty TH,et al. Genome-wide association identifies regulatory Loci associated withdistinct local histogram emphysema patterns. Am J Respir Crit Care Med.2014;190:399–409. https://doi.org/10.1164/rccm.201403-0569OC.30. Cho MH, Castaldi PJ, Hersh CP, Hobbs BD, Barr RG, Tal-Singer R, et al. AGenome-Wide Association Study of Emphysema and Airway QuantitativeImaging Phenotypes. Am J Respir Crit Care Med. 2015;192:559–69. https://doi.org/10.1164/rccm.201501-0148OC.31. Boueiz A, Lutz SM, Cho MH, Hersh CP, Bowler RP, Washko GR, et al.Genome-Wide Association Study of the Genetic Determinants ofEmphysema Distribution. Am J Respir Crit Care Med. 2017;195:757–71.https://doi.org/10.1164/rccm.201605-0997OC.32. Lee JH, Cho MH, Hersh CP, McDonald ML, Crapo JD, Bakke PS, et al.Genetic susceptibility for chronic bronchitis in chronic obstructivepulmonary disease. Respir Res. 2014;15:112–3. https://doi.org/10.1186/s12931-014-0113-2.33. Wain LV, Shrine N, Artigas MS, Erzurumluoglu AM, Noyvert B, Bossini-CastilloL, et al. Genome-wide association analyses for lung function and chronicobstructive pulmonary disease identify new loci and potential druggabletargets. Nat Genet. 2017;49:416–25. https://doi.org/10.1038/ng.3787.34. van der Plaat DA, de Jong K, Lahousse L, Faiz A, Vonk JM, van Diemen CC,et al. Genome-wide association study on the FEV1/FVC ratio in never-smokers identifies HHIP and FAM13A. J Allergy Clin Immunol.2017;139:533–40. https://doi.org/10.1016/j.jaci.2016.06.062.35. van der Plaat DA, de Jong K, Lahousse L, Faiz A, Vonk JM, van Diemen CC,et al. The well-known gene HHIP and novel gene MECR are implicated insmall airway obstruction. Am J Respir Crit Care Med. 2016;194. https://doi.org/10.1164/rccm.201604-0843LE.36. Karmouty-Quintana H, Philip K, Acero LF, Chen NY, Weng T, Molina JG, et al.Deletion of ADORA2B from myeloid cells dampens lung fibrosis andpulmonary hypertension. FASEB J. 2015;29:50–60. https://doi.org/10.1096/fj.14-260182.37. Oldenburger A, Poppinga WJ, Kos F, de Bruin HG, Rijks WF, Heijink IH, et al.A-kinase anchoring proteins contribute to loss of E-cadherin and bronchialepithelial barrier by cigarette smoke. Am J Physiol Physiol. 2014;306:C585–97.https://doi.org/10.1152/ajpcell.00183.2013.38. Kaku Y, Imaoka H, Morimatsu Y, Komohara Y, Ohnishi K, Oda H, et al.Overexpression of CD163, CD204 and CD206 on alveolar macrophagesin the lungs of patients with severe chronic obstructive pulmonarydisease. PLoS One. 2014;9:e87400. https://doi.org/10.1371/journal.pone.0087400.39. Hodgson U, Pulkkinen V, Dixon M, Peyrard-Janvid M, Rehn M, Lahermo P,et al. ELMOD2 is a candidate gene for familial idiopathic pulmonary fibrosis.Am J Hum Genet. 2006;79:149–54. https://doi.org/10.1086/504639.40. Savarimuthu Francis SM, Larsen JE, Pavey SJ, Duhig EE, Clarke BE, BowmanRV, et al. Genes and gene ontologies common to airflow obstruction andemphysema in the lungs of patients with COPD. PLoS One. 2011;6:e17442.https://doi.org/10.1371/journal.pone.0017442.Plaat et al. BMC Pulmonary Medicine           (2019) 19:58 Page 11 of 1241. Wilker EH, Alexeeff SE, Poon A, Litonjua AA, Sparrow D, Vokonas PS, et al.Candidate genes for respiratory disease associated with markers ofinflammation and endothelial dysfunction in elderly men. Atherosclerosis.2009;206:480–5. https://doi.org/10.1016/j.atherosclerosis.2009.03.004.42. Koczulla AR, Jonigk D, Wolf T, Herr C, Noeske S, Klepetko W, et al. Kruppel-like zinc finger proteins in end-stage COPD lungs with and without severealpha1-antitrypsin deficiency. Orphanet J Rare Dis. 2012;7:29. https://doi.org/10.1186/1750-1172-7-29.43. Haunerland NH, Spener F. Fatty acid-binding proteins--insights from geneticmanipulations. Prog Lipid Res. 2004;43:328–49. https://doi.org/10.1016/j.plipres.2004.05.001.44. Wagner AJ, Stumbaugh A, Tigue Z, Edmondson J, Paquet AC, Farmer DL, etal. Genetic analysis of congenital cystic adenomatoid malformation reveals anovel pulmonary gene: fatty acid binding protein-7 (brain type). Pediatr Res.2008;64:11–6. https://doi.org/10.1203/PDR.0b013e318174eff8.45. Zhou J, Deng Z, Chen Y, Gao Y, Wu D, Zhu G, et al. Overexpression ofFABP7 promotes cell growth and predicts poor prognosis of clear cell renalcell carcinoma. Urol Oncol. 2015;33:113.e9–113.17. https://doi.org/10.1016/j.urolonc.2014.08.001.46. Ruwanpura SM, McLeod L, Miller A, Jones J, Vlahos R, Ramm G, et al.Deregulated Stat3 signaling dissociates pulmonary inflammation fromemphysema in gp130 mutant mice. Am J Physiol Cell Mol Physiol.2012;302:L627–39. https://doi.org/10.1152/ajplung.00285.2011.Plaat et al. BMC Pulmonary Medicine           (2019) 19:58 Page 12 of 12