Dissecting respiratory disease heterogeneity through the genetics of diffusing capacity

Genome-wide association studies (GWASs) have allowed the robust and replicable identification of novel genomic regions associated with respiratory diseases. For instance, in chronic obstructive pulmonary disease (COPD), nearly all of the described genetic risk regions were not previously known to play a role in COPD pathogenesis [1–9]. However, where sample size is critical to discovery of new genetic risk regions in GWASs, large GWASs of lung function (and lung function extremes) in the general population have made great strides in describing the genetic risk regions contributing to the observed population variability in spirometry, and in risk of COPD [10–16]. Interestingly, the genetic risk regions contributing to the observed population variability of forced expiratory volume in 1 s (FEV₁) and FEV₁ to forced vital capacity (FVC) ratio can be aggregated into genetic risk scores which are predictive of COPD [14, 16, 17].

Despite the wealth of information GWASs of COPD and lung function have provided toward understanding the genetic architecture of COPD, these studies have only marginally addressed the observed clinical and physiological heterogeneity of COPD. Machine learning techniques are one approach to discover enrichment of association of known genetic risk loci with a subgroup of COPD patients with shared clinical features [18]. The differential strength of association of COPD risk variants with COPD-related phenotypes, such as quantitative imaging features, can also be used to explore the genetic basis for COPD heterogeneity [9, 19]. Aside from these techniques, studying the genetics of sub-phenotypes and key physiological features of respiratory diseases may allow us to more directly evaluate the biological processes contributing to the clinical and physiological heterogeneity of these diseases, and in particular, COPD. GWASs of quantitative computed tomography emphysema [20, 21], despite smaller sample sizes and reduced statistical power, have demonstrated that using a more biologically specific phenotype allows discovery of variants not previously discovered in COPD GWASs, including the genome-wide significant identification of variants correlated with the first-described COPD genetic susceptibility variant, the Z allele in SERPINA1. Genetic studies of COPD-related clinical and physiologic features including emphysema pattern [22], emphysema distribution [23], resting oxygenation [24], pulmonary artery enlargement [25], chronic bronchitis [26] and the asthma/COPD overlap syndrome [27], have highlighted a genetic basis for the observed clinical heterogeneity of COPD and invite ongoing work into further describing the genetic susceptibility to specific features of COPD and other respiratory diseases.

In this issue of the European Respiratory Journal, Terzikhan et al. [28] present a heritability analysis and GWAS of the diffusing capacity of the lung measured by carbon monoxide uptake (D_LCO) as well as the transfer coefficient of the lung for carbon monoxide (D_LCO/V_A or K_CO). D_LCO and D_LCO/V_A are physiological traits impaired in the advanced stages of a number of complex respiratory diseases including COPD, interstitial lung disease and pulmonary hypertension. The heritability (the proportion of phenotypic variance explained by genetics) of D_LCO was previously described in adult twins from the Swedish Twin Registry [29]; however, this is the first study examining the genome-wide genetic variant associations with either D_LCO or D_LCO/V_A. In a meta-analysis of 8372 individuals from two European-ancestry studies, the Framingham Heart Study and the Rotterdam Study, the authors identified a genome-wide association of an uncommon variant (rs17280293, gnomAD European MAF 2.8%) in the ADGRG6 (formerly known as GPR126) locus with D_LCO/V_A.

The association of rs17280293 with D_LCO/V_A is given additional credence since variants close to ADGRG6 have previously been associated with FEV₁/FVC ratio [10, 12–14], COPD [9] and various anthropomorphic measurements in large GWASs. Notably, recent lung function GWASs have identified two independent loci at ADGRG6 associated with FEV₁/FVC ratio [13, 14], where the lead variant at the second ADGRG6 locus is in linkage disequilibrium (R²=0.93 in Europeans) with rs17280293, which is associated with D_LCO/V_A in this journal. Terzikhan et al. [28] performed fine mapping and reported a 0.72 posterior probability for rs17280293 to explain the observed D_LCO/V_A association. Further, the authors report that rs17280293 is a missense variant predicted to be deleterious to ADGRG6 protein structure in two different functional impact prediction tools. In a separate lung tissue dataset, the authors reported significantly lower levels of ADGRG6 in lung tissue in association with both COPD case status and lower D_LCO/V_A. These data suggest rs17280293 is the causal variant and ADGRG6 is the causal gene for the D_LCO/V_A association at this locus. ADGRG6 is plausible effector gene, with a role in angiogenesis. Despite this compelling evidence, a more definitive link between rs17280293 and expression or function of ADGRG6 will need to be established.

The ADGRG6 locus is complex. In addition to being previously described in association with lung function and COPD, variants at this locus are also associated with height, and previous studies have described two independent associations at this locus. Terzikhan et al. [28] assessed for possible confounding of the ADGRG6 locus D_LCO/V_A association by adjusting for FEV₁/FVC ratio and showed unchanged effect of rs17280293 on D_LCO/V_A. The association was further shown to be robust to adjustment for height as well as quantitative emphysema in a subset of individuals. These data suggest that the ADGRG6 locus identified by the authors is independently associated with D_LCO/V_A level when considering FEV₁/FVC ratio, height, or amount of emphysema. However, D_LCO is a complex phenotype, affected not just by emphysema but by pulmonary vasculature and haemoglobin; D_LCO/V_A is further influenced by body size but does not scale monotonically with lung volume [30, 31]. The sensitivity analyses performed by the authors are subject to power considerations and further work (for example, using larger sample sizes with more detailed phenotyping and mediation analysis) will need to be done to disentangle relationship of traits at the ADGRG6 locus marked by rs17280293 to determine if traits are truly independent in their association to the ADGRG6 locus or if the associations are explained by un-modelled mediation of the traits by each other.

The work by Terzikhan et al. [28] in this issue of the European Respiratory Journal illustrates how the genetic study of more discrete physiologic traits, such as D_LCO and D_LCO/V_A, impaired at the advanced stages of complex respiratory disease, can contribute to our understanding of the genetic basis for the clinical heterogeneity of respiratory diseases. Additionally, this study demonstrates the importance of sample size for genetic discovery and replication, where only a single loci genome-wide significant association with D_LCO/V_A (and not D_LCO) was described and has yet to be replicated. Despite careful consideration of the explanatory model for the relationship of rs17280293 to D_LCO/V_A levels, this work highlights the difficult task left by genetic discovery, where even genetic associations with high statistical confidence mark the beginning of our journey to biologically cohesive functional descriptions of genetic loci, let alone translation to clinical relevance. In summary, Terzikhan et al. [28] have contributed valuable information to our understanding of complex respiratory diseases and we look forward to future larger investigation of the genetic associations and functional follow-up of D_LCO and D_LCO/V_A.

• Received August 1, 2018.
• Accepted August 15, 2018.

• Copyright ©ERS 2018