A101-19 Longitudinal High-resolution Gene Expression Profiling Identifies a Blood-based Biomarker for COPD Exacerbations

Abstract Introduction Exacerbations in chronic obstructive pulmonary disease (COPD) are common and associated with worsened morbidity, healthcare utilization, and mortality. Despite their relative prevalence and clinical impact, little is known about how these events are initiated or progress over the course of the exacerbation cycle. We hypothesized that a prospective study employing high frequency, longitudinal transcriptomics could identify key gene regulators of exacerbations. Methods Six exacerbation-prone COPD participants underwent comprehensive clinical assessments, spirometry, and blood collection, then self-performed weekly blood collection for 24 weeks coupled with daily EXAcerbations of Chronic pulmonary disease Tool (EXACT) questionnaires. RNA sequencing was performed on all samples, and differential expression was assessed per subject using ImpulseDE, which models all time points as a trajectory. Differentially expressed genes were analyzed with weighted gene correlation network analysis (WGCNA) to identify a network module. We input module genes into adaptive LASSO (adaLASSO) using cross-validation with one subject per fold to develop an exacerbation risk score. Using data from the Genetic Epidemiology of COPD cohort (COPDGene) cohort, we validated this score in GOLD 2-4 participants with 5-year RNA-sequencing and prospective exacerbation data. The selected genes were reweighted and tested in a 50/50 train-test split using logistic regression with zero vs. non-zero exacerbation as the outcome, adjusting for age, sex, smoking status, FEV1, and exacerbation history. As a reference, developed a de novo adaLASSO score in COPDGene in parallel. Results Complete data (110 samples) were available for 4 individuals (64.8±2.2 years, 75% Female; 50% White), FEV1=53±19%, FEV1/FVC=0.5±0.1, 2.3±0.5 exacerbations over the study period. ImpulseDE found 29 genes that overlapped in ≥ 2 subjects; genes rose in the week prior to the clinical onset of exacerbation and sharply decreased after the clinical peak; pathway enrichment identified genes associated with neutrophil degranulation, viral response, and interferon-ɑ/Β signaling. We identified a 28-gene module that tracked closely with exacerbations, of which 6 genes were selected by adaptive LASSO as an exacerbation risk score (Figure 1). In COPDGene (testing set: n = 555), the reweighted 6-gene risk score (RSPH9, RP4-1024N4, TTC21A, LGALS9, PLSCR1, TRIM38) was associated with prospective exacerbations in the test set (β = 0.310, 95% CI: (0.098, 0.526), p = 0.005), while a de novo score was not (p = 0.579). Conclusion We identified unique gene signatures associated with exacerbations representing responses to viruses and neutrophil activation. The network module identified genes then were predictive of exacerbations in an external COPD cohort that could not be identified by traditional risk prediction approaches. This abstract is funded by: UAB Lung Health Center Pilot Grant