Transcriptomic Signature of Asthma–Chronic Obstructive Pulmonary Disease Overlap in Whole Blood

To the Editor:

There is a subset of patients who present clinically with features of both asthma and chronic obstructive pulmonary disease (COPD), known as asthma–COPD overlap (ACO). Because asthma and COPD may be composed of multiple endotypes (1), characterization of ACO using molecular phenotypes may result in a more specific categorization. Because bronchoscopy is not part of the routine management of COPD, it will be important to identify more easily accessible biomarkers. Therefore, we quantified whole-blood gene expression by RNA sequencing and measured total IgE levels in ACO to identify a transcriptomic profile characteristic of the phenotype and to refine our understanding of molecular mechanisms underlying the disease.

The Genetic Epidemiology of COPD (COPDGene) study is a multicenter observational study including 10,192 smokers with and without COPD (2). Complete blood counts and blood for RNA sequencing were collected at the second visit, ∼5 years after initial recruitment. COPD was defined by a postbronchodilator ratio of forced expiratory volume in 1 second (FEV₁) to forced vital capacity < 0.7 with FEV₁ < 80% predicted (Global Initiative for Chronic Obstructive Lung Disease stages 2–4). ACO was defined as COPD with self-report of physician-diagnosed asthma before the age of 40, as previously described (3, 4).

Gene expression was measured in 1,199 current and 1,456 former smokers including 793 individuals with COPD, 120 subjects with ACO, 79 subjects with asthma, and 962 control subjects with normal spirometry (see Table E1 in the data supplement). There was no significant difference in cell differential counts in ACO versus COPD; however, the percentage of neutrophils was elevated in subjects with ACO compared with control subjects (61.4% vs. 57.2%; P = 1.95 × 10⁻⁴), whereas the percentage of lymphocytes was decreased (26.5% vs. 31.4%; P = 8.00 × 10⁻⁷). IgE levels quantified in 691 subjects using previously described methods (5) showed elevated IgE levels in subjects with ACO compared with subjects with COPD (log₁₀ IgE, 1.9 vs. 1.6; P = 0.02) and with control subjects (1.9 vs. 1.6; P = 0.01) (see Figure E1 in data supplement).

Differential gene expression analysis was performed using the voom (6)/limma (7) R package (R Foundation for Statistical Computing) including library preparation batch, race, age, sex, current smoking, and pack-years of smoking history as covariates. There were 3,687 genes with differential expression (false discovery rate < 10%) in ACO compared with control subjects, demonstrating a large amount of transcriptional dysregulation in these subjects. Similarly, 3,320 genes were differentially expressed (false discovery rate < 10%) in subjects with COPD compared with control subjects. Although there were 1,994 genes that were altered in both subjects with ACO and subjects with COPD compared with smoking control subjects, 1,692 genes were specific to ACO (Table 1 and Figure E2A in the data supplement). In sensitivity analyses, we found that fold changes and test statistics were highly correlated in analyses performed with and without the inclusion of cell proportions (r = 0.84 and 0.82, respectively) or inhaled corticosteroid usage (r = 0.98 and 0.98, respectively) as covariates, indicating that the differential expression is not driven by differences in cell counts or medication use. The most significantly enriched pathway in the genes dysregulated in ACO but not COPD was the taste transduction pathway, with nine genes being differentially expressed; these genes had similar but slightly smaller fold changes in the comparison of ACO with COPD (Tables E2–E3). The T2R genes are expressed in multiple airway cell types, including airway smooth-muscle cells, airway epithelial cells, and immune cells (8), functioning in relaxation of airway smooth muscle (8). Therefore, it has been proposed that T2R genes may be ideal targets to improve bronchodilation and modulate immune dysregulation and remodeling changes (8).

Table 1.

Results of Gene Expression Analysis in COPDGene Whole Blood

Analysis	Comparison	Number of Differentially Expressed Genes	Top Differentially Expressed Genes	Top Pathways
Identification of genes dysregulated in ACO	Unique to ACO (differentially expressed in subjects with ACO compared with smoking control subjects but not differentially expressed in subjects with COPD compared with smoking control subjects)	1,692	EPHX2	Taste transduction
			EMR1	Hematopoietic cell lineage
			LOC105377623	Huntington disease
			ANTXRLP1	Oxidative phosphorylation
			MGST1	Insulin signaling pathway
	Unique to COPD (differentially expressed in subjects with COPD compared with smoking control subjects but not differentially expressed in subjects with ACO compared with smoking control subjects)	1,326	ADARB1	B-cell receptor signaling pathway
			HLF	RNA transport
			PANK2	Cytokine–cytokine receptor interaction
			FBXW4	NOD-like receptor signaling pathway
			DGCR8	Selenocompound metabolism
Genes associated with IgE levels	—	61	ADORA3	Hematopoietic cell lineage
			IL1RL1
			DACH1
			P2RY2
			RAB44

Definition of abbreviations: ACO = asthma–COPD overlap; ADORA3 = adenosine receptor A3; COPD = chronic obstructive pulmonary disease; COPDGene = Genetic Epidemiology of COPD; IL1RL1 = IL-1 receptor–like 1.

IgE levels were quantified in 691 subjects with gene expression, including 76 subjects with ACO and 215 control subjects with COPD. We identified 61 genes associated with IgE levels (Tables E1 and E4). The “Hematopoietic cell lineage” Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway was significantly enriched in these genes (Bonferroni-adjusted P = 5.21 × 10⁻⁵). The top differentially expressed gene in this analysis was ADORA3 (adenosine receptor A3), which is known to play a role in smooth-muscle contraction and neutrophil degranulation (9) and is a member of the sphingolipid signaling pathway (KEGG pathway ko04071). Another top gene was IL1RL1 (IL-1 receptor–like 1), which has been reproducibly associated with asthma susceptibility (10), and its expression levels have been associated with blood eosinophils and markers of T-helper cell type 2 (Th2) inflammation (11, 12).

A previous study demonstrated that airway epithelial expression of a set of 100 genes representative of Th2 inflammation in asthma was associated with increased disease severity and “asthma-like” features in COPD but was not associated with clinical history of asthma (13). We investigated whether this signature could identify a subset of subjects with COPD who had asthma-related features in whole blood by calculating a blood Th2 score using Gene Set Variation Analysis (14). Of the 100 genes, 53 had detectable expression levels (>1 count per million in 50% of subjects) (Table E5). We found that Th2 scores were elevated in subjects with ACO compared with both subjects with COPD and smoking control subjects (P = 0.017 and 3.98 × 10⁻⁵, respectively) (Figure 1A). The Th2 score was also associated with a lower FEV₁% predicted (β = −13.0; P = 1.35 × 10⁻⁵), an increased eosinophil percentage (P = 6.18 × 10⁻²³), and exacerbation frequency (P = 0.0044) (Figures 1B–1D) but was not associated with IgE levels. Hierarchical clustering of gene expression values of the 53 Th2-signature genes expressed in whole blood revealed five distinct clusters (Figure 1E). We found that the three largest clusters were associated with changes in immune-cell proportions, lung function, and exacerbation frequency (Table E6). This suggests that there is a subset of patients with COPD with a Th2 immune response and asthma-like phenotype that could not be identified through asthma history alone, supporting the argument that the use of molecular phenotypes could help identify specific COPD subtypes that potentially differ in clinical characteristics and therapeutic responses.

Figure 1.

T-helper cell type 2 (Th2) blood gene expression scores in the COPDGene (Genetic Epidemiology of COPD) study. (A) Th2 score according to disease group. P values were calculated using ANOVA and the Tukey honest significant differences test. (B) Association of Th2 scores with FEV₁% predicted. (C) Association of Th2 score with eosinophil percentage. (D) Exacerbation frequency is associated with a higher Th2 score. (E) Heatmap of expression in 1,954 subjects of 53 Th2-score genes with detectable expression levels in whole blood. Columns are ordered by hierarchical clustering. Genes are in rows, and samples are in columns. Red indicates high relative expression; blue indicates low relative expression. Circles represent subjects and red lines represent linear regression lines. Box-and-whisker plots show minimum values, the first data quartile, the median, the third data quartile, and the maximum value. ACO = asthma–COPD overlap; COPD = chronic obstructive pulmonary disease; FEV₁ = forced expiratory volume in 1 second.

Potential weaknesses of this study include the characterization of gene expression in whole blood instead of lung tissue, although the changes that we have detected are likely reflecting transcriptional changes in circulating immune cells, which are relevant to both asthma and COPD; the determination of ACO based on self-report of doctor diagnosis of asthma, which could be subject to recall bias or diagnostic misclassification, although our previous use of this diagnosis has identified both clinical and genetic associations (3, 4); and the lack of replication of differentially expressed genes and pathways in an independent data set. Furthermore, we did not study the effect of long-acting bronchodilator use on gene expression.

In conclusion, we identified genes that are altered in ACO, including genes involved in the taste 2 receptor pathway, as potential avenues for further investigation. Furthermore, we found a subset of patients with COPD with an elevated blood Th2-response signature associated with asthma-like phenotypes and reduced lung function.