Prior investigations into the pharmacogenetics of asthma have generally been limited to one or a few SNPs from one gene. We investigated 844 SNPs from 111 candidate genes selected from asthma β-agonist and corticosteroid pathways, and from our prior candidate gene studies, and screened these SNPs for association with BDR using a family-based screening algorithm that allowed us to rank the SNPs based on estimated power for replication. We then genotyped 99 SNPs from 17 genes in a population-based cohort of patients with asthma, who participated in an asthma clinical trial. Finally, we genotyped 83 SNPs in seven genes in two separate cohorts of subjects with asthma. We found SNPs in the ARG1 gene to be associated with BDR in these three asthma populations, after adjusting for multiple comparisons.
has recently been implicated in asthma. Zimmermann and colleagues (29
) reported increased expression of ARG1
in murine lung, and also found increased arginase 1 protein expression from human asthma bronchoalveolar lavage cells. Variants in ARG1
were associated with atopy in a cohort of Mexicans with asthma (30
maps to chromosome 6q23 and encodes one isoform of the enzyme arginase, which metabolizes l
-Arginine homeostasis is involved in the regulation of airway function, because the availability of this amino acid to nitric oxide synthase (NOS) determines the production of the endogenous bronchodilator nitric oxide (NO) (31
). Changes in l
-arginine homeostasis may contribute to many of the features of asthma, such as airway hyperresponsiveness, airway inflammation, and airway remodeling (32
). Intracellular l
-arginine levels are regulated by at least three distinct mechanisms (reviewed by Maarsingh and colleagues ): (1
) cellular uptake by cationic amino acid transporters, (2
) recycling from l
-citrulline, and (3
) metabolism by NOS and arginase. Arginase is postulated to be involved in asthma by depleting stores of l
-arginine, an NOS substrate, which leads to decreased production of NO, a potent bronchial smooth muscle relaxer (33
), and it has been shown to inhibit airway smooth muscle relaxation (35
). Finally, RNA interference of arginase 1 in the lungs resulted in complete loss of airway hyperresponsiveness to methacholine due to IL-13 treatment (37
). This correlated with arginase 1 expression, which suggests that the polymorphisms involved with the current findings in human asthma may cause a loss of expression or function of arginase 1.
We used a gene-based strategy to select SNPs to take forward for replication. In this method, after ranking SNPs from 1 (most power) to 844 (least power), we grouped all SNPs for each gene and calculated the median SNP rank for that gene. Thus, whereas some genes had one or two SNPs that were assigned high ranks, these genes may not be taken forward because the median SNP rank did not meet the predetermined cutoff. We adopted this strategy because we were not sure that LD patterns across the four asthma populations would be similar. It is interesting to note that ADRB2
, a gene that has been widely studied in asthma pharmacogenetics (38
), was not one of the genes that was selected using this strategy, despite including 18 SNPs from this gene in the screening analysis. It is possible that there was insufficient power in the screening stage, because we only analyzed the 209 trios in the placebo group in CAMP. However, it should be noted that a prior analysis using all 400 trios also did not find an association with any of the ADRB2
SNPs and BDR (39
). Furthermore, the phenotype that we investigated is different from that reported in other studies reporting on the pharmacogenetic effect of ADRB2
). We are currently performing additional genotyping and analyses using an SNP-based strategy for replication, rather than the gene-based strategy that we used here, to see if we identify important SNPs in this gene and others for association with BDR.
Our analysis used the phenotype of acute response to a short-acting β2-agonist, albuterol, in part because this was the phenotype that was common to all asthma cohorts. In the screening algorithm, we used the information from repeated measures of BDR among the 209 white children randomized to the placebo group in the CAMP study over the 4 years of the trial. This was done to increase the power for the screening method. In contrast, for the replication cohorts, we only used the information on BDR response on entry into the respective studies, to standardize the phenotype. Thus, our results may not be applicable to patients with asthma who are on regular β2-agonist treatment (either short- or long-acting). We also did not address interactions with any other class of asthma medication, because baseline medication was different for all the populations: BDR was performed in both CAMP and the Adult Trial populations after several weeks of being off all asthma medications; LOCCS subjects were receiving inhaled corticosteroids for 4–6 weeks before BDR testing; and drug regimens for LODO subjects were not changed before entry into the trial.
We used a novel method of screening a large number of SNPs for association analysis (11
). This method has been successfully used to identify disease-susceptibility genes (12
). Because this method has only been developed for family-based studies and not population-based studies, we used the CAMP population for screening the original 844 SNPs. The traditional method would have been to analyze all the SNPs in one population, determine which SNPs were associated with BDR at a predetermined level of significance, then test these SNPs in the replication populations. However, if we had used this usual method for gene finding, we would then have had to adjust our overall results for the 844 SNPs that were originally tested, and it is likely that no finding would have survived this adjustment for multiple testing, even if the association was real. In our method, because we screened on power and not P
value, we only needed to adjust for the 99 tests in the first replication step. Thus, this screening method allows the use of modest-sized populations for gene discovery because it limits the number of tests that are actually being performed.
The population to which we applied our screening algorithm was a cohort of childhood asthmatics, whereas the three asthma replication cohorts were composed predominantly of adults with asthma. As we stated previously, the rationale for this is that the screening method was developed for the setting of family-based studies and not for population-based studies. There is no similar screening method that has yet been developed for population-based studies. Because our replication populations were of small to modest sizes, we applied the screening method as a means of minimizing the number of tests. Although there were only 209 parent–child trios included in the screening analysis, we maximized the power in the screening stage by using the 11 repeated measures of BDR over the 4 years of the trial. In addition, there were differences in the asthma severity and in the magnitude of the BDR between the populations as shown in . Despite these differences, we were able to detect associations between SNPs in ARG1 and BDR in each of the three replication populations. Although the association between these SNPs and BDR in CAMP was not statistically significant, the effect sizes of each SNP were of sufficient magnitude for them to be selected based on power in the screening analysis. We can only surmise at this point that there may be age-related effects associated with the SNPs in this gene. We believe, therefore, that the results of the association between ARG1 polymorphisms and BDR are robust and applicable to both children and adults with asthma in a variety of settings.
The three ARG1
SNPs that were associated with BDR were all in the promoter region of the gene and were in tight LD with each other. Genotyping of all known SNPs in the gene or resequencing of the gene will need to be performed to determine if these three promoter SNPs are in LD with the functional mutation. There is mounting evidence for “cross-talk” between pathways involved with the relaxation and constriction of airway smooth muscle (40
). It is thus not entirely unexpected that a gene involved with airway hyperresponsiveness in mouse studies is associated with a bronchodilator response in human asthma. However, arginase 1 has not been previously identified as one of the proteins involved in such cross-talk. This unexpected finding shows the potential value of whole-genome coverage to study drug response to uncover novel genetic determinants. The screening method that we used for this analysis would be easily applicable to the case of whole-genome association.
In summary, we have identified SNPs in ARG1 as novel BDR determinants. Further studies will need to identify the functional SNP or SNPs in this gene. Other pharmacogenetic studies using long-acting β2-agonists, either alone or in conjunction with corticosteroids, and investigation of other phenotypes (e.g., FEV1, peak flow) are needed to clarify the effects of variants in this gene. Our analysis shows the utility of a family-based algorithm to effectively screen SNPs for replication in other cohorts. This method is easily applicable to the case of whole-genome association.