Cystic Fibrosis: Using genetic association to identify modifiers of disease variability in cystic fibrosis

One of the greatest challenges in modern medicine is to understand why patients with the same disease can have different outcomes. Identifying the causes of disease variation can be of considerable prognostic value. Furthermore, factors that modify a disease may be more amenable to therapeutic manipulation than the underlying cause of the disease. Thus, the recent report that severity of life limiting pulmonary disease in cystic fibrosis (CF) patients is associated with variation in the transforming growth factor β1 (TGFβ1) gene has generated considerable attention.¹ This interest is warranted on several grounds. First, genetic variation in TGFβ1 was estimated to account for 15–20% of the variability in lung disease (as measured in their study) suggesting that it could be an important prognostic factor. Second, although the CF transmembrane conductance regulator (CFTR) was identified as the molecular defect in CF 16 years ago, the disease has proven to be challenging to treat at the molecular level. Finally, the TGFβ signaling pathway is a critical determinant of lung homeostasis, and therapies are already available to manipulate this pathway.

The study design used by Drumm et al is similar to case–control association studies that test whether selected genetic variants alter susceptibility to complex traits. Variants in 10 candidate modifier genes previously shown to be associated with CF disease outcome were tested for association with patients grouped according to ‘mild’ or ‘severe’ lung disease. Only variants -509C/T and codon 10 T/C in the TGFβ1 gene showed significant differences in genotype distribution among the two groups of patients. Three features distinguish this study from preceding work; the criteria used to define disease severity, the number of subjects studied, and replication of association.

Classifying patients according to severity of a chronic disease manifestation can be challenging, especially if progression of the illness is nonlinear, as is the case for CF lung disease. To address this issue, Drumm et al used an objective measure of pulmonary mechanics, the volume of air in the first second of a forced exhalation (forced exhalation volume in one second or FEV₁). This cross-sectional measure is predictive of CF survival and is used for monitoring progress and response to treatment of CF patients.^{2, 3} Patients were classified as ‘mild’ or ‘severe’ if their age-adjusted FEV₁ at the time of enrollment fell in the upper 25% or the lower 25%, respectively, of FEV₁ values of patients in the US CF Registry in 1999. To control for allelic effects at the CFTR locus, only FEV₁ values from patients homozygous for the common CFTR mutation ΔF508 in the CF Registry were utilized thereby matching the CFTR genotype of the enrolled patients. As cross-sectional measures of an on-going process may lead to misclassification, the authors also utilized FEV₁ measures over the preceding five years and a regression model⁴ to predict pulmonary function at 20 years of age for each subject. The aforementioned measure was found to be consistent with the initial classification for 97.3% of the enrolled patients. The use of cross-sectional and longitudinal measures to define disease severity is a major strength of the paper and sets a worthy precedent for genetic association studies of chronic disorders.

The other distinguishing features are not unique to this study but are highly desirable in association studies. Adequate power to detect associations between alleles of moderate effect and phenotype require substantial numbers of well-characterized patients. Multicenter collaboration enabled enrollment of over 800 patients for the Drumm study. While the authors did not provide an analysis of the power of their study to detect association of single or multiple modifiers, the number of individuals enrolled far exceeds most candidate gene modifier studies of single gene disorders in humans. Second, validation of a genetic association via replication is an important step to excluding false positive associations.⁵ Constructive replication studies performed by a different set of investigators, using a separate set of patients with different genotyping methods and a different statistical model provide the most objective form of replication.⁶ The replication study by Drumm et al had a number of these attributes. An independent set of patients were recruited with a broad range of CFTR genotypes and without bias regarding disease severity classification. The authors used both cross-sectional and longitudinal measures of FEV₁ but divided patients into ‘severe’ and ‘mild’ groups using a different method than the initial study.

What are the next steps in evaluating genetic modifiers of CF lung disease and in particular, the role of genetic variation in TGF β1? Although the study by Drumm et al suggests that TGFβ1 alleles appear to account for a reasonable fraction of variance in CF lung function, the relative contribution of genetic and non-genetic factors to this trait is unknown. Hence, it is not possible to determine whether TGFβ1 variation accounts for all or part of the genetic contribution to FEV₁ variation. To address this issue, association studies searching for modifiers of CF lung disease should assess the contribution of novel candidates to trait variance relative to TGFβ1 alleles. Second, as noted in the editorial accompanying the Drumm paper,⁷ patients recruited according to lung disease severity also had a significant difference in body mass index (BMI). Thus, TGFβ1 alleles could be associated with FEV₁, BMI or a combination of the two. Intriguingly, analysis of CF twins and CF siblings revealed that genetic factors contributed to a composite measure of FEV₁ and BMI, but not to FEV₁ alone.⁸ Finally, stratification of patient groups still lingers as a possible explanation for association observed by Drumm et al.⁹ Stratification has been observed in Caucasian Americans that can be traced to different European ancestry.¹⁰ In the case of CF, the frequency of the CF alleles differs considerably depending upon the geographic and ethnic origin of European patients. The frequency of TGFβ1 alleles might also be stratified in European Americans. The replication study by Drumm et al demonstrated association between TGFβ1 genotypes and age-adjusted FEV₁ above or below 68% predicted that was independent of CFTR genotype and independent of substantial variation in environment (patients were enrolled at several centers in North America). These findings suggest that adequately powered studies of independent collections of CF patients should detect association between TGFβ1 alleles and CF lung disease severity as defined in the replication study of Drumm et al. Replication of the findings of Drumm et al in other CF patient populations or in family based transmission disequilibrium studies could reduce or eliminate the concern of false positive association due to population stratification. Such studies could also be structured to determine if TGFβ1 variants affect pulmonary status, body mass index or a combination of both. Although the path from association to causation may be tortuous, affirmation of the findings of Drumm, Knowles and the CF Gene Modifier Group would provide a new tool in the fight against CF▪