Exome sequencing of extreme phenotypes identifies DCTN4 as a modifier of chronic Pseudomonas aeruginosa infection in cystic fibrosis

Abstract

Exome sequencing has become a powerful and effective strategy for the discovery of genes underlying Mendelian disorders. However, use of exome sequencing to identify variants associated with complex traits has been more challenging, partly because the sample sizes needed for adequate power may be very large. One strategy to increase efficiency is to sequence individuals who are at both ends of a phenotype distribution (those with extreme phenotypes). Because the frequency of alleles that contribute to the trait are enriched in one or both phenotype extremes, a modest sample size can potentially be used to identify novel candidate genes and/or alleles. As part of the National Heart, Lung, and Blood Institute (NHLBI) Exome Sequencing Project (ESP), we used an extreme phenotype study design to discover that variants in DCTN4, encoding a dynactin protein, are associated with time to first P. aeruginosa airway infection, chronic P. aeruginosa infection and mucoid P. aeruginosa in individuals with cystic fibrosis.

Fig. 1

A) QQ-plot of p-values for a rare variant test of association (Online Methods) between each gene and extreme P. aeruginosa infection phenotypes. The most significant association (p=2.2×10⁻⁶⁾ ) is with DCTN4 with a collapsed score based on two variants (rs11954652 and rs35772018). B) Kaplan-Meier curves comparing age-at-onset of chronic P. aeruginosa infection by presence of DCTN4 variants among children in quintiles 2 and 3 of enrollment age among those reaching the endpoint (enrollment ages 1.6 to 6.7). Because of the need for analysis stratified on enrollment age, it is not possible to create a representative time-to-event curve with all individuals at once. This curve showing the middle quintiles is representative of the effect size over all strata combined: the HR for this subgroup is 2.3 (95% CI=[1.3, 4.5]), similar to the estimate of 1.9 (p=0.004) over the entire analysis set. C) Kaplan-Meier curves comparing age-at-onset of chronic P. aeruginosa infection by presence of DCTN4 variants among children who were not selected for a negative P. aeruginosa history in the EPIC validation sample. Blue line: children without DCTN4 variants (n=246); red dotted line: children with DCTN4 variants rs11954652 and/or rs35772018 (n=34). HR=2.7 [1.4, 5.3] with p=0.004 (Online Methods). Comparison with 1B shows a larger baseline hazard for these children (both curves more steep than in 1B illustrates the need for stratification on enrollment age when implementing the Cox model. D) Kaplan-Meier curves comparing age-at-onset of chronic P. aeruginosa infection among all enrollment strata by DCTN4 variant group; blue line: no DCTN4 variants (n=565); red dotted line: rs11954652 heterozygotes (n=78); green line rs11954652 homozygotes and rs35772018 heterozygotes combined (n= 22). Individuals in the latter group have higher risk than those in either of the other two groups (HR = 3.3, p=0.002 compared to baseline). Differences between groups appear somewhat compressed relative to the Cox model hazard ratio estimates because all enrollment strata are shown together in this plot: there are too few individuals in the third group to visualize differences within strata, but it is notable that a strong difference can be seen even without stratification.