Rationale

Blood eosinophil count in chronic obstructive pulmonary disease (COPD) is a biomarker that predicts increased exacerbations and response to inhaled corticosteroids. However, the molecular pathways altered in the lung tissue of eosinophilic COPD have not been examined. It is also unclear whether there are increased eosinophils in the lung compartments, as studies report conflicting results. We hypothesized that gene expression of lung tissues from eosinophilic COPD would be enriched for eosinophils and eosinophil related pathways and analyzed RNA-sequencing of lung tissue samples from eosinophilic and non-eosinophilic COPD defined by blood eosinophil counts.

Methods

Lung tissue RNA-seq and clinical data for 383 study participants with COPD were obtained from the Lung Tissue Research Consortium (LTRC) study after excluding samples from individuals with systemic corticosteroid use or diagnosis of pulmonary fibrosis. COPD was defined by postbronchodilator ratio of forced expiratory volume in 1 second (FEV1) to forced vital capacity (FVC) <0.7 with FEV1 <80% predicted. Eosinophilic COPD was defined based on a blood eosinophil count of 300 cells/L or greater within one year of lung tissue collection. Differential gene expression analysis was performed with the limma voom R package with adjustment for age, sex, self-reported race, current smoking status, pack-years of smoking, inhaled corticosteroid use and library batch. Relative eosinophil abundance was estimated using Bisque deconvolution and gene set variation analysis (GSVA).

Results

We analyzed 322 non-eosinophilic and 61 eosinophilic COPD lung tissue samples. Individuals with eosinophilic COPD had higher BMI and lower percentage of neutrophils in peripheral blood. The most common pathologic diagnosis was centrilobular emphysema for both groups. There were 2 differentially expressed genes associated with eosinophilic COPD, CRELD2 and UQCC2 (false discovery rate (FDR) <0.05). Pathway analysis of 54 genes with FDR <0.15 using MSigDB was showed enrichment in the glycolysis pathway. There was no association between deconvolution (Bisque or GSVA) estimates of lung tissue eosinophil abundance and measured blood eosinophil count (beta 0.56, p = 0.4 and beta 0.21, p = 0.09). In a stratified analysis lung and blood eosinophils were correlated in non-eosinophilic COPD (beta 0.36, p = 0.04) but not in eosinophilic COPD (beta 0.05, p = 0.8), suggesting differential correlation.

Conclusion

We identified lung gene expression changes associated with eosinophilic COPD. The lack of relationship between lung and blood eosinophils may be due to the different distribution of peribronchial inflammatory eosinophils and parenchymal regulatory eosinophils, which will require testing conducted with matched histology samples.