Herein, we observed significant associations of common IPO13 polymorphisms with airway responsiveness (the most dynamic treatment response phenotype in the CAMP clinical trial) among children with mild-to-moderate asthma. These associations were observed exclusively among those children who were not randomized to inhaled GCs, and persisted over the ~4.5 years of clinical observation. The genetic effects conferred by these IPO13 variants were clinically significant, with an average 1.5–2.1 fold increase in mean PC20 among carriers of IPO13 variants.
IPO13 has been functionally characterized as a primary regulator of GC-bound GR transport across the nuclear membrane. Inhibition of lung epithelial cell IPO13 production inhibits nuclear translocation of GR from the cytoplasm and subsequent GC-mediated silencing of inflammatory cytokine production[13
], suggesting that the normal anti-inflammatory response induced by GC is dependent on normal IPO13 function. In light of these observations, it is curious that the genetic effects observed in the current study were observed only among subjects who were not taking daily corticosteroids. We propose two possible mechanisms to explain these findings. The first is developmental. IPO13 was first identified in studies of lung development, where IPO13 (initially known as LGL2) was found to be differentially expressed in fetal rat lung cell culture[12
]. It is conceivable that IPO13 variation impacts airway hyperresponsiveness by altering airway anatomy through changes during airway morphogenesis and development. Though plausible, the absence of association of IPO13 variants with baseline lung function in the current study (data not shown) suggests that this is not likely the case. However, given the diverse roles of glucocorticoid during lung morphogenesis (see Introduction) and the considerable impact on lung development by aberrant perinatal glucocorticoid exposure[27
], we are hesitant to completely discount this possibility at this time.
A second possible mechanism of action is that IPO13 variants improve airway hyperresponsiveness by enhancing the local anti-inflammatory effects of circulating, endogenous GCs by facilitating increased GR transport into the nucleus and thus increasing the effective bioavailability of endogenous GC. Supporting this hypothesis is the observation that carriers of IPO13 variants who were not on inhaled steroid exhibited improvements in airway responsiveness over the course of the clinical trial that approached those for subjects who were taking inhaled steroids (see Figure ), suggesting that that IPO13 variation enhances endogenous GC nuclear availability to therapeutic levels. It has been previously demonstrated that nucleocytoplasmic shuttling of IPO13 is developmentally regulated and highly variable in rat lung[12
]. Sequence variation could potentially influence this regulation by increasing nuclear membrane availability through increased IPO13 expression or by altering the kinetic properties of GR transport by influencing GR binding affinities. We note that the paucity of coding variation at the IPO13 identified through our resequencing efforts suggests that the genetic effects observed are likely due to regulatory variation rather than structural changes. Though studies are currently ongoing to explore these possibilities, we note that because none of the airway responsiveness-associated non-coding variants map to highly conserved genetic sequence and are not predicted to harbor transcription factor binding sites, it is unlikely that we have as of yet identified a putative functional variant.
It is noteworthy that the associated haplotype block not only spans the IPO13 locus, but two neighboring genes as well: ATP6V0B and DHP2 (Figure ). Though it is not possible to completely exclude these other genes as functionally responsible for the observed associations, it is unlikely to be the case. The motivation for studying these polymorphisms was the recognition of IPO13 as the primary nuclear transporter for steroid-bound glucocorticoid. Had we identified these variants through a hypothesis-free approach (i.e. a genome-wide study), the pretest probability for each gene in the region would be similar. However, because these SNPs were chosen due to the biologic prior on IPO13, it would be improbable that the true functional effects would be mediated through a neighboring gene. Unlike IPO13, there is little biological evidence to support either ATP6V0B or DHP2 in either the pathogenesis of airways responsiveness or glucocorticoid pharmacogenetics. ATP6V0B is a subunit of the vacuolar-type H(+)-ATPase (V-ATPase) multisubunit enzyme, responsible for organelle acidification. DHP2 encodes a protein involved in diphthamide biosynthesis that confers resistance in yeast to the effects of diphtheria toxin. Though surveys of genomic databases (including UniGene and GEO) suggest that ATP6V06 is ubiquitously expressed and DHP2 is weakly expressed in the lung, there is little reason to suspect either as the responsible locus.
In genetic association studies of complex traits such as airway hyperresponsiveness and pharmacogenetic responses, it is important to consider a variety of methodological and statistical issues that can hamper proper interpretation of observed findings. Two features of the current study warrant particular attention: phenotype misclassification and statistical power. Most observational studies of the genetics of asthma attempt to avoid confounding of lung phenotype measurements by medication use by performing spirometric and airway hyperresponsiveness measurements following a short period (less than 24–48 hours) off anti-asthma controller medications. However, due to safety consideration, long-term avoidance of asthma controller medication is typically not permitted. The randomized, placebo-controlled trial is perhaps the only setting in which such confounding can be eliminated, and is a major strength of the study presented here. It is directly a result of the availability of ~4.5 years of repeated methacholine challenge measurements off anti-inflammatory agents among more than two-thirds of CAMP participants who were randomized to placebo or nedocromil that enabled detection of the observed associations in this study. In light of the differences in genetic effect observed across treatment arms, we stress that future attempts to replicate our findings should use approaches that address this issues, as replication may only be possible when proper adjustment for glucocorticoid use are made.
Subjects randomized to inhaled budesonide represent only 1/3 of the cohort. We recognize that this relatively smaller sample size is inadequately powered to detect strong genetic effects with alleles of modest frequency, and that it is possible that an effect similar to that observed among those not on steroids could potentially have been observed among steroid users had the number of subjects in this latter group approached those of the former. However, we note that sample size impacts only the ability to claim a statistically significant difference in genotype effect but does not in any way influence the absolute effect observed. In this study, the trends of association among those on budesonide were quite dissimilar to those observed among subjects not on budesonide, with all variants actually conferring greater airway responsiveness (though the confidence intervals in this group are quite broad and all span the null). Therefore, while we stress that our conclusions reported herein apply only to those subjects in the non-budesonide arm, it is likely that differential effects of these alleles are present between subjects who were and were not taking steroids.