RNAseq analysis of bronchial epithelial cells to identify COPD-associated genes and SNPs

Abstract

Background: There is a need for more powerful methods to identify low-effect SNPs that contribute to hereditary COPD pathogenesis. We hypothesized that SNPs contributing to COPD risk through cis-regulatory effects are enriched in genes comprised by bronchial epithelial cell (BEC) expression patterns associated with COPD.

Methods: To test this hypothesis, normal BEC specimens were obtained by bronchoscopy from 60 subjects: 30 subjects with COPD defined by spirometry (FEV1/FVC < 0.7, FEV1% < 80%), and 30 non-COPD controls. Targeted next generation sequencing was used to measure total and allele-specific expression of 35 genes in genome maintenance (GM) genes pathways linked to COPD pathogenesis, including seven TP53 and CEBP transcription factor family members. Shrinkage linear discriminant analysis (SLDA) was used to identify COPD-classification models. COPD GWAS were queried for putative cis-regulatory SNPs in the targeted genes.

Results: On a network basis, TP53 and CEBP transcription factor pathway gene pair network connections, including key DNA repair gene ERCC5, were significantly different in COPD subjects (e.g., Wilcoxon rank sum test for closeness, p-value = 5.0E-11). ERCC5 SNP rs4150275 association with chronic bronchitis was identified in a set of Lung Health Study (LHS) COPD GWAS SNPs restricted to those in putative regulatory regions within the targeted genes, and this association was validated in the COPDgene non-hispanic white (NHW) GWAS. ERCC5 SNP rs4150275 is linked (D' = 1) to ERCC5 SNP rs17655 which displayed differential allelic expression (DAE) in BEC and is an expression quantitative trait locus (eQTL) in lung tissue (p = 3.2E-7). SNPs in linkage (D' = 1) with rs17655 were predicted to alter miRNA binding (rs873601). A classifier model that comprised gene features CAT, CEBPG, GPX1, KEAP1, TP73, and XPA had pooled 10-fold cross-validation receiver operator characteristic area under the curve of 75.4% (95% CI: 66.3%-89.3%). The prevalence of DAE was higher than expected (p = 0.0023) in the classifier genes.

Conclusions: GM genes comprised by COPD-associated BEC expression patterns were enriched for SNPs with cis-regulatory function, including a putative cis-rSNP in ERCC5 that was associated with COPD risk. These findings support additional total and allele-specific expression analysis of gene pathways with high prior likelihood for involvement in COPD pathogenesis.

Keywords: Bronchial epithelial cells; CAT; CEBPG; COPD; ERCC5; GPX1; GWAS; KEAP1; TP73; XPA; cis-regulation; eQTL.

Fig. 1

Schematic description of research design. ¹RNAseq: RNA sequencing by next generation sequencing; ²BEC: bronchial epithelial cell; ³COPD, chronic obstructive pulmonary disease; ⁴GWAS, genome wide association study; ⁵ASE: allele-specific expression; ⁶LHS GWAS: Lung Health Study Genome Wide Association Study; ⁷DAE: differential allelic expression; ⁸COPDgene NHW: COPDgene Non-Hispanic White Cohort

Fig. 2

Network of bivariate correlation among genes (transcript abundance values) for control and COPD cohorts. Each line represents Pearson r-value with p-value < 0.05. Left: Control, Right: COPD. (See Additional file 1: Table S3 for r- and p-value of each gene pair)

Fig. 3

Inter-gene correlation differences in control vs COPD cohorts. a, b TP73–2 vs ERCC5

Fig. 4

Receiver operating characteristic curve (ROC) (a) and summary of performance of classifier (b) in 30 control and 30 COPD subjects

Fig. 5

Inter-individual variation allelic ratio for cDNA compared with gDNA. Each symbol represents results from a single heterozygous individual. a CAT-rs1049982, b CEBPG-rs3745968, c ERCC5-rs17655, d KEAP1-rs1048287, e TP73-rs1801174