Genome-Wide Association Study Identifies Novel Pharmacogenomic Loci For Therapeutic Response to Montelukast in Asthma

Background Genome-wide association study (GWAS) is a powerful tool to identify novel pharmacogenetic single nucleotide polymorphisms (SNPs). Leukotriene receptor antagonists (LTRAs) are a major class of asthma medications, and genetic factors contribute to variable responses to these drugs. We used GWAS to identify novel SNPs associated with the response to the LTRA, montelukast, in asthmatics. Methods Using genome-wide genotype and phenotypic data available from American Lung Association - Asthma Clinical Research Center (ALA-ACRC) cohorts, we evaluated 8-week change in FEV1 related to montelukast administration in a discovery population of 133 asthmatics. The top 200 SNPs from the discovery GWAS were then tested in 184 additional samples from two independent cohorts. Results Twenty-eight SNP associations from the discovery GWAS were replicated. Of these, rs6475448 achieved genome-wide significance (combined P = 1.97 x 10-09), and subjects from all four studies who were homozygous for rs6475448 showed increased ΔFEV1 from baseline in response to montelukast. Conclusions Through GWAS, we identified a novel pharmacogenomic locus related to improved montelukast response in asthmatics.

We hypothesized that we could identify novel loci associated with montelukast response using a GWAS approach. We first tested our hypothesis in a discovery GWAS using genotype and phenotype data from two montelukast treatment arms of the Leukotriene Modifier or Corticosteroid or Corticosteroid-Salmeterol (LOCCS) trial [29] and Effectiveness of Low Dose Theophylline as Add On Therapy for the Treatment of Asthma (LODO) trial [1]. We then tested our top SNP associations for replication in two independent cohorts taking montelukast from the Childhood Asthma Research and Education (CARE) Network trials, the Characterizing the Response to a LT Receptor Antagonist and Inhaled Corticosteroid (CLIC) trial [30] and the Pediatric Asthma Controller Trial (PACT) [31].

Clinical Cohorts and Phenotyping
The discovery cohort included two asthmatic clinical trials with treatment arms evaluating montelukast response, the American Lung Association Asthma Clinical Research Center (ALA-ACRC)supported trials, the Leukotriene Modifier Or Corticosteroid or Corticosteroid-Salmeterol Trial Medicine (CGM) Global Alliance and funding from the BioBank Japan project (supported by the Ministry of Education, Culture, Sports, Sciences and Technology of the Japanese government). Funding for LOCCS and LODO is provided from American Lung Association (ALA) and the ALA's Asthma Clinical Research Centers investigators and research teams for use of LOCCS and LODO data, with additional funding from HL071394 and HL074755 from the NHLBI, and Nemours Children's' Clinic. GlaxoSmithKline supported the conduct of the LOCCS Trial by an unrestricted grant to the ALA. The CARE study is supported by funding through National Heart Lung and Blood Institute (NHLBI), as part of the single nucleotide polymorphism health association resource (SHARe) Asthma Resource Project (SHARP); genotyping of SHARP was supported by the NHLBI (U.S. Federal Government contract number N02-HL-6-4278). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: GlaxoSmithKline supported the conduct of the LOCCS Trial by providing funds to the American Lung Association. The authors received no relevant funding or financial support from GSK. There are no patents, products in development or marketed products related to this manuscript to declare. This does not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials.
(LOCCS) and Effectiveness of Low Dose Theophylline as Add On Therapy for the Treatment of Asthma (LODO) [1,29]. While the LOCCS and LODO clinical trials each analyzed over 400 subjects, for this study, we evaluated a sub-population consisting only of the montelukast treatment arms from these studies that consisted of 133 individuals. For replication, publicly archived, genome-wide SNP data and clinical phenotype information from patients taking montelukast as part of the Childhood Asthma Research and Education (CARE) Network-Characterizing the Response to a LT Receptor Antagonist and an Inhaled Corticosteroid and Pediatric Asthma Controller Trial (CLIC and PACT) (30, 31) (total sample size = 184), were used (dbGaP Study Accession: phs000166.v2.p1 (http://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id = phs000166.v2.p1)). The data evaluated in this study were obtained from four previously published clinical trials (clinicaltrials.gov identifiers: NCT00156819 (LOCCS); NCT00046644 (LODO); NCT00272506 (PACT); NCT00000622 (CLIC)) [1,[29][30][31]. Study participants for these trials provided written informed consent, and this consent procedure was approved by the institutional ethics committee/IRB. The Brigham and Women's Hospital Institutional Review Board approved this study. For all cohorts, subjects were consented for genetic studies and their data was de-identified. Table 1 provides a summary of the populations evaluated in this analysis.
For all populations, the primary outcome phenotype was defined as a change in FEV 1 following 8 weeks of treatment while on montelukast, minus FEV 1 at baseline (ΔFEV 1 ), adjusted for age, gender, and race.

Genotyping and Quality Control (QC)
Genome-wide genotyping of the LOCCS and LODO trials was conducted using the Illumina HumanHap550 chip (San Diego, CA). For CLIC and PACT, genotyping was performed as described (30,31), using the Genomewide Affymetrix SNP 6.0 Array (Santa Clara, CA). The software PLINK v.1.07 [32] was used for QC of genotype data. SNPs with a study-wise missing data proportion above 0.05 were removed from the analysis. SNPs failing to meet Hardy-Weinberg equilibrium (HWE) (P < 0.0001), in addition to SNPs with a minor allele frequency (MAF) < 5% and more than 10% missing genotypes, were also dropped from the analysis. A total of 532,264 SNPs with acceptable quality were genotyped and analyzed in the discovery GWAS for both LOCCS and LODO, and 591,268 SNPs were genotyped and analyzed in both CLIC and PACT.

Statistical Analysis
For the GWAS, an additive genetic association model was evaluated, adjusting for baseline FEV 1 , age, race (self-reported ancestry) and gender as covariates, using PLINK. Due to small sample sizes, both white and non-white subjects were included. However, the genomic inflation factor values for the subset of montelukast treated patients in these populations was 1, indicating that minimal population stratification was present despite population racial heterogeneity. Due to differences in genotyping platforms used, our analysis focused on the SNPs that were genotyped in all four populations. For replication, the one-sided association P values from 261,076 SNPs that had the same direction of effect in the LOCCS and LODO discovery cohorts were combined, and the top 200 SNPs (as ranked by combined P values) were then carried forward for replication in CLIC and PACT. The one-sided P values of the SNPs that had the same direction of effect (β) in LOCCS-LODO and at least one replication cohort, and that also met nominal significance (P < 0.05) [33,34] in at least one replication cohort, were combined using a weighted Z-test [35] in 'R version 3.0.2' (http://www.r-project.org). SNPs with combined P values below the multiple test correction threshold (P = 0.00025) were considered to be replicated. The threshold for genome-wide significance for associated SNPs was determined using the Bonferroni correction (P = 9.40 x 10 −08 ). SNP P values below 10 −05 were considered suggestive of genome-wide significance.

Results
The discovery GWAS was conducted in LOCCS and LODO asthmatic cohorts to evaluate the association of patient genotype with 8-week ΔFEV 1 following treatment with montelukast (133 patients). Plotted results of the discovery GWAS are shown in Fig 1. Non-white subjects were included, and after adjusting for age, race and gender as covariates, plots of the genomic-control adjusted P values demonstrated no evidence of population stratification. In LOCCS, none of the SNPs exceeded the threshold for genome-wide significance (P = 9.40 x 10 −08 ); however, 25 SNPs approached genome-wide significance (P<10 −05 ), of which the top-ranked SNP (rs12659144) achieved a P value of 2.2 x 10 −06 , although it did not also replicate in LODO. In LODO, one SNP, rs2247977, achieved genome-wide significance (P = 4.95 x 10 −08 ), although it did not also replicate in LOCCS. For replication of the discovery SNP associations, the P values of the SNPs with the same direction of effect in LOCCS and LODO were combined, and the top-ranked 200 SNPs from LOCCS-LODO were carried forward for evaluation in CLIC and PACT (S1 Table). Four SNPs, s6475448, rs7794356, rs953977 and rs1364805, survived correction for multiple testing (combined P < 0.00025) ( Table 2). Three of these SNPs, rs6475448, rs7794356, and rs953977, also approached or achieved genome-wide significance ( Table 2).

Discussion
Leukotriene modifier drugs represent a major treatment modality for asthma patients, and the ability of physicians to determine which patients are likely to benefit from these drugs would greatly enhance therapeutic outcomes for asthmatics. We undertook a genome-wide interrogation of 532,264 SNPs to evaluate association of genotype with 8-week ΔFEV 1 following treatment with montelukast in four asthma clinical trials (LOCCS, LODO, CLIC and PACT). We identified four SNPs that replicated in LOCCS-LODO, CLIC and PACT, of which one variant, rs6475448, achieved genome-wide significance (combined P = 1.97 x 10 −09 ) ( Table 2). rs6475448 Definition of abbreviations: "SNP" = single nucleotide polymorphism; "Chr." = chromosome ; "Chr. Location." = chromosomal position of listed SNP; "β" = effect size estimates (ΔFEV 1 , (mL)) for the minor allele.
*  was a novel locus associated with an improvement in response to montelukast in four independent asthmatic populations. rs6475448 is present within MLLT3, which is proposed to regulate cell fates for megakaryocytes and early erythroid cells in humans [36]. Functional and molecular studies have also shown that MLLT3 acts as a positive regulator of erythroid and megakaryocyte differentiation [36]. Red blood cell precursors including megakaryocytes and erythroid cells are capable of transforming arachidonate and LTA 4 to bioactive eicosanoids [37,38]. Megakaryocytes give rise to platelets, which are also activated in asthmatics and contribute to leukotriene production during inflammation [39]. In our study, rs6475448 was associated with a genotype-dependent improved response to montelukast in LOCCS-LODO, CLIC and PACT (Table 2 and Fig 2). While the SNP was intronic, and thus MLLT3 expression was unlikely to be affected, using the web server SCAN [40], we found that this SNP is also an expression quantitative trait locus (cis-eQTL) for SHROOM3, a gene that encodes a cytoskeleton protein responsible for cellular shape during morphogenesis [41], and can affect this gene's expression in the HapMap Improvement in lung function related to montelukast treatment, by rs6475448 genotype. The least-squares (LS) means (adjusted for study, race and gender) and 95% confidence intervals for ΔFEV 1 related to montelukast treatment were generated using R (http://cran.r-project.org/web/packages/lsmeans/ lsmeans.pdf), and plotted for each study (panels), by rs6475448 genotypes: homozygous reference ("GG": lymphoblastoid cell lines (LCLs). Therefore, rs6475448, and its eQTL, SHROOM3, may potentially represent novel candidate loci for asthma, and/or treatment response to leukotriene modifiers.
Our study has several limitations. First, as is common to many pharmacogenomic GWAS, our sample size is modest; however, our sample size is comparable to recently published GWAS of symptomatic response to corticosteroids in asthma [33][34]. In addition, we were able to replicate four SNPs in multiple independent, montelukast-treated populations, providing supportive evidence of true positive associations. Furthermore, because the cohorts evaluated in this study included non-white subjects, racial heterogeneity may also represent a major limitation of the study; however, we accounted for this by including race, age and gender as covariates in our GWAS models, and saw no evidence of population stratification based on genomic inflation factor and Q-Q plot behavior. A third limitation is that the genotyping platforms used to generate the genome-wide genotype data differed among the four cohorts. To overcome this, we focused our analysis on the SNPs in common between platforms. A fourth limitation is that the ages of our replication and discovery populations differed; while LOCCS and LODO evaluated adults, a pediatric population comprised CLIC and PACT montelukast cohorts. While this supports the generalizability of the reported associations, one reason for failure to replicate additional loci may lie in the innate differences in response between children and adults. For instance, we recently described a pharmacogenetic locus for corticosteroid response [33] that was replicably associated in children, but not in adults. A fifth limitation is that, while we were able to identify a novel montelukast treatment-related gene through GWAS, we did not also find SNPs in reported candidate genes for montelukast response (e.g. CYSLTR1) from among the replicated SNP data, which could reflect a limitation of the sensitivity of GWAS, in addition to differences in genotyping platforms used in this study. Finally, additional mechanistic and functional studies will be necessary in order to discern the potential role of MLLT3 in montelukast response.

Conclusions
Through a GWAS of differential montelukast response in four asthmatic cohorts, we have identified a genome-wide significant SNP, rs6475448, which is present within MLLT3. This SNP may represent a novel mechanism for differential responses to leukotriene modifying agents in asthma.
Supporting Information S1