PRS-ing Forward to Identify Genetic Risk in Idiopathic Pulmonary Fibrosis

Abstract

Over the last 2 decades, it has become clear that genetic risk is an important determinant for development of idiopathic pulmonary fibrosis (IPF); however, the genetic architecture underlying IPF is complex and remains incompletely understood. Although a variety of common and rare genetic variants have been associated with IPF, an SNP in the promoter of the of the gene encoding MUC5B (Mucin 5B) is the strongest disease risk factor identified to date (1, 2). Across multiple studies, the odds ratio for IPF associated with carrying the T (minor) allele is ∼4.5 (3), making it one of the most impactful disease-associated common variants in humans. In addition to the MUC5B region, 22 other IPF-associated common variant loci were identified in a recent meta-analysis of genome-wide association studies (GWAS) (4). Given the progress in identifying individual genetic variants associated with IPF, a significant question in the field has become whether a predictor of combined genetic risk could be developed that would improve on information obtained by testing for the presence of the MUC5B risk allele. The polygenic risk score (PRS), which was developed in 2007, is increasingly used as a method for analyzing large-scale GWAS data in a variety of diseases (5), including lung diseases such as chronic obstructive pulmonary disease (6) and asthma (7). This approach can combine a large number of common variants, including those below the genome-wide significance threshold for phenotype association, to identify disease risk in individuals. A new study by Moll and colleagues in this issue of the Journal (pp. 791–801) is the first comprehensive investigation of this method in IPF (8). The investigators analyzed data from 14,650 study participants, including 1,970 individuals with IPF and 1,068 individuals with interstitial lung abnormalities (ILAs), from a variety of cohorts to develop PRSs for IPF with or without inclusion of the MUC5B risk allele. Although the PRS calculated with inclusion of MUC5B performed somewhat better than the PRS in the absence of data from the MUC5B region, the top quintile in the PRS (excluding MUC5B) was associated with an odds ratio of ∼7 compared with the lowest quintile, thereby indicating that common variants outside the MUC5B region substantially contribute to risk for IPF. In addition, receiver operating curves for IPF prediction showed that the highest area under the curve was achieved by clinical modeling with a combination of the PRS (excluding MUC5B) and genotype information regarding the MUC5B risk allele. Furthermore, using linkage disequilibrium score regression, the authors estimated observed-scale heritability in IPF at ∼28%, in the range of prior estimates of the impact of genetic predisposition on the development of IPF (1). In addition to IPF, the investigators applied the PRS methodology to ILAs, which are abnormal interstitial changes affecting >5% of lung parenchyma on computed tomography scan (9). ILAs are detectable in approximately 7% of individuals >50 years of age and can in some instances precede development of clinical IPF by several years (10). In ILA studies, the PRS model was associated with presence (odds ratio, 1.25) and progression (odds ratio, 1.16) of ILAs; however, this association was only observed in subjects with European ancestry. Together, the findings in this study reinforce the importance of the MUC5B risk allele in IPF and strongly support the idea that a wide variety of common variants throughout the genome influence disease risk. Although testing for the MUC5B risk allele is the most efficacious single measurement for risk assessment, prediction is modestly improved by adding the PRS generated from the rest of the genome. Therefore, combining the PRS with testing for the MUC5B risk allele and clinical information could be useful in future studies for identification of individuals at the highest risk for development of IPF. An important question is whether genetic risk assessment tools can be used to stratify risk before the onset of disease, particularly to determine which ILAs are most likely to progress to clinical disease, because ILAs are much more frequent than IPF. Although the presence of the MUC5B risk allele has been associated with ILAs (11), the results of this study are somewhat disappointing in this regard, because the association of the PRS with the presence and progression of ILAs was modest at best. To date, PRS methodology has not had a substantial impact on clinical practice, and pitfalls of PRSs in predicting risk of age-related traits have been documented (12). Other issues related to maximizing the power and utility of PRS include limitations in the size of GWAS datasets available for relatively rare diseases, difficulties in extrapolation to ethnicities underrepresented in GWAS datasets, and underlying assumptions of lack of specific gene-by-environment interactions (13, 14), all of which may have implications for application in IPF. Also, PRS methodology does not account for rare genetic variants, which can have a major impact on disease risk in the ∼8.5% of individuals with sporadic IPF who harbor loss-of-function rare variants in telomere maintenance genes (15). Despite these issues, this study represents substantial progress in genetic risk assessment in IPF. Further refinement of PRS calculations is likely in the future as larger and more ethnically diverse genetic data in subjects with IPF become available.