Proteome-wide Mendelian randomization analyses to identify potential therapeutic targets for lung function and respiratory disease

RATIONALE Impaired lung function predicts mortality and is a diagnostic criterion for chronic obstructive pulmonary disease (COPD). Proteins are often the target of pharmacological interventions, therefore identifying causal links between proteins and lung function could inform understanding of COPD pathophysiology and suggest therapeutic targets. We aim to infer the potential impact of circulating protein levels on lung function, using strictly defined cis protein quantitative trait loci (cis-pQTLs) as genetic instrumental variables for Mendelian randomisation (MR).

METHODS We applied two-sample MR by integrating protein GWAS data (2,923 proteins, 48,195 UK Biobank European participants) with lung function GWAS data (four lung function traits, 149,166 European participants from 36 non-UK Biobank cohorts). We selected strictly defined cis-pQTLs, within 100 kilobase pairs of a transcription start site and strongly associated (P≤5×10⁻⁹) with protein levels, and applied single-cis-MR analysis (Wald ratio method). Sensitivity analyses included colocalization analysis (to distinguish causal effects from genomic confounding by linkage disequilibrium), and bidirectional MR to explore possible reverse causation. Replication analysis was conducted where possible. We used the Drug-Gene Interaction Database and phenome-wide association studies (PheWAS) to inform biological and clinical interpretation of identified proteins.

RESULTS We curated 1,841 proteins with a suitable cis-pQTL instrument, and evaluated evidence for causal effects of these proteins on four lung function traits. The single-cis MR analysis implicated 16 proteins for lung function at a Bonferroni-corrected threshold (Wald ratio estimator P<1.71×10-5), with evidence from colocalization. Of these, 10 proteins have been previously implicated either by lung function GWAS, or from other MR analyses with colocalization. Surfactant protein D (SFTPD) has been highlighted in previous respiratory MR analyses and variants in SFTPD have been previously reported to be associated with emphysema; our PheWAS suggested that this variant has a relatively specific effect on lung function as it was associated with no non-respiratory traits at a FDR<1%. In contrast to previous expression QTL evidence, our study suggested that ITGAV inhibition could reduce FEV1/FVC; we note that reduced lung function was also seen in a recent trial of an ITGAV inhibitor (NCT01371305). Our MR analysis implicated six proteins not implicated by previous lung function GWAS or MR (DTD1, PILRA, PTPRK, TDPRK, GRHPR, NUDT5).

CONCLUSIONS Our protein-based approach identified proteins that may be causally related for lung function variability. We highlight known protein drug targets, and identify several new proteins which are potentially therapeutic targets but warrant further follow up for potential utility and safety.