B95-01 Prioritizing Therapeutic Targets for Progressive Pulmonary Fibrosis: A Causal Mediation Analysis

Abstract Rationale Progressive pulmonary fibrosis (PPF) is a common manifestation of interstitial lung disease (ILD) that leads to declining lung function and death. While protein biomarkers of ILD survival have been identified, it remains unclear which are causally related to PPF and should be prioritized for therapeutic targeting. Methods A secondary analysis of proteomic and phenotypic data from ten prospectively recruited pulmonary fibrosis cohorts across the US, UK and Canada was performed to identify proteins that cause organ fibrosis in laboratory models and associate with transplant-free survival (TFS) in humans through a PPF-mediated causal pathway. After quantile normalization, circulating plasma proteins associated with three-year TFS were identified in a discovery (n = 2011) cohort. Those causally related to organ fibrosis in laboratory models were advanced for causal mediation analysis, which estimated the effect of each protein (exposure) on restricted mean TFS time (outcome) through a PPF-related pathway (mediator). The PPF measure was defined using an ordinal composite score of annualized decline in forced vital capacity and diffusion capacity, with each unit increase in score associated with worse TFS (Figure 1A). Proteins mediated by this PPF measure in the discovery cohort were advanced for testing in an independent validation cohort (n = 1180), with those showing sustained mediation considered candidate therapeutic targets. Results Of 2921 proteins evaluated, 124 were associated with TFS in the discovery cohort, including 51 shown to be causal of organ fibrosis in laboratory models. Of these 51 proteins, PPF mediated the TFS association for 41. When tested in an independent validation cohort, PPF continued to mediate the TFS association for 13 proteins, which were classified as candidate therapeutic targets (Figure1B). Proteins with the strongest mediated effect were amphiregulin (AREG), integrin beta six (ITGB6) and latent transforming growth factor beta binding protein 2 (LTBP2), with higher mediated effects as protein concentration increased (Figure 1C). Confounding sensitivity analysis suggested that results were robust to unmeasured confounding, with E-values 2 for AREG and ITGB6 and 1.5 for six additional proteins. After stratification by ILD subtype, only AREG, LTBP2, matrix metalloproteinase 10 (MMP10) and syndecan-1 (SDC1) were consistently mediated by PPF across common ILD subtypes. Conclusions Here we provide epidemiological evidence for a potentially causal relationship between 13 circulating proteins and PPF, which corroborates laboratory-based findings and supports prioritization of these proteins for therapeutic targeting. This abstract is funded by: NIH / NHLBI