Functional interactors of three genome-wide association study genes are differentially expressed in severe chronic obstructive pulmonary disease lung tissue

Abstract

In comparison to genome-wide association studies (GWAS), there has been poor replication of gene expression studies in chronic obstructive pulmonary disease (COPD). We performed microarray gene expression profiling on a large sample of resected lung tissues from subjects with severe COPD. Comparing 111 COPD cases and 40 control smokers, 204 genes were differentially expressed; none were at significant GWAS loci. The top differentially expressed gene was HMGB1, which interacts with AGER, a known COPD GWAS gene. Differentially expressed genes showed enrichment for putative interactors of the first three identified COPD GWAS genes IREB2, HHIP, and FAM13A, based on gene sets derived from protein and RNA binding studies, RNA-interference, a murine smoking model, and expression quantitative trait locus analyses. The gene module most highly associated for COPD in Weighted Gene Co-Expression Network Analysis (WGCNA) was enriched for B cell pathways, and shared seventeen genes with a mouse smoking model and twenty genes with previous emphysema studies. As in other common diseases, genes at COPD GWAS loci were not differentially expressed; however, using a combination of network methods, experimental studies and careful phenotype definition, we found differential expression of putative interactors of these genes, and we replicated previous human and mouse microarray results.

Gene expression profile of human lung in a relatively early stage of COPD with emphysema

“…It is not known whether ACER2 with increased expression was associated with apoptosis of the emphysema in the present study. FAM13A, known as a COPD-related gene, was significantly upregulated in COPD with emphysema group ( P -value 0.00096), and this is consistent with many previous genetic studies 9,10…”

Integrated transcriptomic correlation network analysis identifies COPD molecular determinants

“…Then, the number of the 145 GWAS genes included in the DEGs was zscorenormalized and the p-value for the given z statistics was calculated; the p-value of 0.2 indicates that the number of differentially expressed GWAS genes is equal to what expected by chance. This observation is in accordance with the results obtained in [Morrow 2017], where the authors showed that COPD GWAS genes were not differentially expressed in lung tissue samples. The DEG matrix of 2097 rows (DEGs) and 327 columns (219 COPD cases + 108 control samples) was used as input to SWIM in order to build the COPD gene correlation network based on the Pearson correlation coefficient, where a threshold is set for the absolute value of the minimum correlation coefficient necessary to draw an edge (see Materials and Methods section).…”

“…Instead, the gene KCND3 codes for potassium channel subfamily D member 3 and is located in a genome-wide significant COPD GWAS region, although the COPD-associated top SNP is located about 575 Kb from the transcription start site. The expression of PRDX4 is strongly positively correlated in the COPD correlation network with SERPINE2, CD79A, and POUF2AF1, which were previously considered as putative interactors of genes at COPD GWAS loci [Morrow 2017]. The expression of KCND3 is strongly positively correlated with two of them (SERPINE2 and CD79A).…”

“…To further assess the validity of the SWIM analysis in identifying disease genes and modules associated to COPD status, we applied the SWIM software on the GSE76925 dataset (test set), which contains microarray gene expression profiling of lung tissue samples from 111 COPD cases and 40 control smokers with normal lung function [Morrow 2017]. In this case, starting from 22631 genes, we obtained 887 significantly DEGs at 15% FDR, of which 493 (56%) were down-regulated in COPD cases, while the remaining 394 (44%) were up-regulated ( Supplementary Table 1).…”

“…Here, we applied SWIM to lung tissue gene expression data from two well-characterized COPD case-control populations to study the differences between lung samples from normal subjects (represented by smokers with normal spirometry) and COPD cases. We used the dataset with a larger number of lung tissue samples (i.e., GSE47460 with 219 COPD cases and 108 controls) as the "training set" for running SWIM and the dataset of Morrow and colleagues [Morrow 2017] with a smaller number of samples (i.e., GSE76925 with 111 COPD cases and 40 controls) as the "test set" for validating the results.…”