Background
Asthma is a chronic inflammatory disease of the airways, characterized by symptoms of breathlessness, wheezing and cough, associated with variable airflow limitation and airway hyperresponsiveness (AHR). Asthma is also characterized by airway remodeling which includes goblet cell metaplasia, epithelial damage, subepithelial fibrosis, basement membrane thickening, and increased airway smooth muscle (ASM) mass [
1]. The pathogenesis of asthma is believed to involve an interaction between the innate and adaptive immune systems [
2], and phenotypic changes within the epithelial-mesenchymal trophic unit [
3]. Genetic and genomic analyses have been used to discover the molecular mechanisms underlying these phenotypic changes. Large genome-wide association studies (GWAS) have reproducibly identified single nucleotide polymorphisms (SNPs) in or near genes predominantly expressed in the airway epithelium and immune cells as susceptibility factors for asthma. The specific genes include gasdermin B (
GSDMB) [
4], interleukin 33 (
IL33), and thymic stromal lymphopoietin (
TSLP) [
5]. Another candidate gene
ADAM33, expressed in ASM, has been shown to be associated with asthma in linkage analysis [
6]. Genetic variants associated with susceptibility for asthma may exert their effect by altering gene expression levels; indeed many of the SNPs associated with asthma and AHR have been shown to be expression quantitative trait loci (eQTL) in lung tissue, epithelial and blood cells, and altered protein expression of some of these genes has been found in cells and tissue from asthmatic individuals [
7].
In this hypothesis driven study we used a reproducible multiplexed technology (Nanostring®) to quantify the expression of 334 genes potentially involved in phenotypic changes in asthmatic airways. This technology is highly sensitive, reproducible, and is suitable for archived tissue specimens as it is insensitive to RNA degradation [
8]. We hypothesized that changes in the expression patterns of genes involved in ASM contraction, the cytoskeleton, epithelial barrier function, innate/adaptive immunity, fibrosis and remodeling, and epigenetics would be present in the airway tissue of asthmatics compared to non-asthmatics. Our results suggest that alterations in the expression of genes involved in cell-cell and cell-matrix interactions may contribute to the pathogenesis of asthma, particularly severe asthma. The identification of altered gene co-expression networks may identify changes in transcriptional regulation that could be pathogenic and missed with commonly used analyses for differential expression.
Discussion
In this study the Nanostring® platform was used to quantify the airway expression of candidate genes hypothesized to be important in the pathophysiology of asthma. mRNA was obtained from the airways of asthmatic (mainly fatal asthmatics) and non-asthmatic donor lungs and measures of airway remodeling were made on the sampled airways. Candidate genes were grouped into the following categories: ASM contraction, the cytoskeleton, epithelial barrier function, innate/adaptive immunity, fibrosis/remodeling, and epigenetics. 51 genes (15%) were nominally differentially expressed (p.unadj <0.05) in asthmatic airway tissue and included many genes important in cell-cell and cell-matrix interactions (COL1A1, COL3A1, ITGB6, LAMC2, RAC1). Of the 51 genes differentially expressed based on a nominal p-value, only 3 were significant following multiple comparison correction (ITGB6, COL1A1, COL3A1).
Cell-cell junctions are altered in the airway epithelium of asthmatics [
29] and this may result in greater permeability of the epithelial layer and ultimately hypersensitivity of the ASM to agonist challenge [
30]. The most significant differentially expressed gene in the data set was ITGB6. ITGB6 rapidly accumulates following injury to the epithelial layer and is considered to be important for normal wound healing [
31]. In the mouse loss of ITGB6 causes an increase in the number of B-cells and T-cells around the airways, an increase in IL-4 production, and airway hyperresponsivenss in naïve mice without allergen challenge [
31]. Additionally, influenza virus has been shown to interact with ITGB6 to cause epithelial cell death and collagen deposition in a TGF-ß dependent manner [
32]. Another gene with lower abundance in asthmatics that did not meet the adjusted
p-value cut off was RAC1. RAC1 has been shown to be important in the formation of tight junctions in an EGFR dependent manner [
33], specifically by regulating tight junction protein 1 (or zona occluden 1) [
34] which was also lower in abundance in the asthmatic samples. Changes in these genes could also affect epithelial-mesenchymal transition (EMT) [
35], cytoskeletal stability, actin filament assembly/disassembly, cell stiffness and/or cell migration [
36]. In addition, a protein-protein interaction network highlighted that down-regulated genes were enriched in pathways for cell-cell communication and integrin cell surface interactions.
The most significantly up-regulated gene in asthmatic samples was
COL1A1 which codes for the alpha chain in type 1 collagen. Collagen 1 is the major type of collagen in basement membrane and a major protein found in remodeled airways. Collagen 1 is important in airway remodeling [
37], in particular thickening of the subepithelial space which is associated with worsening of asthma symptoms [
38]. Collagen type 3 was also significantly elevated in asthmatic subjects and is also significantly elevated in the basement membrane of asthmatic subjects [
39]. Type 1 and 3 collagen have both been associated with worsening lung function in a horse model of asthma [
40]. Beyond the ability of collagens to affect distensibility of the airways, collagen 1 has been shown to stimulate ASM to produce MMP1 [
41] and proliferate, in conjunction with FAK [
42]. Collagen 1 and 3 expression has also been shown to be unaffected by corticosteroid usage in severe asthmatics [
43] and collagen 1 and 3 contribute to the loss of the anti-mitotic effect of corticosteroids [
44]. Enrichment of pathways involved in collagen remodeling was seen in our network analysis and highlights the importance of understanding fibrosis as it relates to fiber production, degradation, and organization and how this impacts normal cell function.
ADAM33 has been described in candidate gene studies to be associated with asthma [
6] and was significantly elevated in asthmatic airways.
ADAM33 has been implicated in smooth muscle development, cell-cell connections, and cell proliferation and differentiation [
45]. Over-expression of
ADAM33 in asthmatic airways has been described [
46] and may be an important determinant of disease progression. There is evidence that ADAM33 can stimulate angiogenesis ex vivo and in vivo and by this mechanisms may contribute to airway remodeling [
47]. Furthermore, ADAM33 family member TACE/ADAM17 can mediate release of TNF-α and fracktalkine (or CX3CL1) from the cell membrane [
48,
49] and ADAM9 may mediate the release of growth factor HB-EGF [
50]. If ADAM33 has a similar capacity for cytokine or growth factor cleavage this could make it a major contributor to airway remodeling in asthma.
Other genes in the 5 most differentially up or down regulated genes that were nominaly significant include: Cyclic ADP Ribose (CD38), Interleukin 13 Receptor Alpha 1 (IL13RA1), Prostaglandin F Receptor (PTGFR), Heat Shock Protein Beta 1 (HSPB1), and Interferon Induced with Helicase C Domain 1 (IFIH1). CD38 is a protein that generates the second messenger cADPR to cause calcium release. Recent work has explored the role of CD38 in asthma and has suggested that increased CD38 expression causes hypercontractility in ASM cells from asthmatics [
51] although in our samples we saw no CD38 staining in the ASM layer (Additional file
1: Fig. S1 and S2). Additionally, CD38 deficient mice have reduced AHR following ovalbumin challenge [
52]. Decreased IL13RA1 expression is surprising in the context of asthma but this could be due to a compensatory response to continued eosinophilia and IL13/IL4 exposure in the lung [
53]. Prostaglandin’s can be both pro and anti inflammatory but there is little research on the role of prostaglandin F in the context of asthma. HSPB1 (or heat shock protein 27) is a chaperone protein that has been implicated in cellular differentiation, apoptosis, and smooth muscle contraction [
54]. Up-regulation of the gene could contribute to ASM hypercontractility in asthma but further work investigating the phosphorylation state and activity of HSP27 in asthma is needed to answer this question. IFIH1, also known as MDA5, is a DEAD box double stranded (ds) RNA helicases that can detect intracellular viral dsRNA and lead to the production of interferons [
55]. MDA5 and TLR3 signaling have been shown to be deficient in bronchial epithelial cells from asthmatic subjects [
56] and this could be responsible for the defective epithelial release of interferon I and III in response to rhino virus infection [
57,
58].
Co-expression of genes does not imply interaction between their proteins but instead may suggest similarities in their regulation by transcription factors or epigenetic mechanisms [
59]. Co-expression analyses can reveal changes in the regulation of gene expression [
60] and have been used to identify epigenetic changes that affect gene co-expression in cancer [
61]. In our study, genes that were differentially co-expressed between asthmatics and non-asthmatics were significantly enriched for pathways involved in virus recognition and regulation of interferon production (Table
4). The genes enriched in these pathways were from cluster 16 and were RIG-I (
DDX58), RIG-1-like receptor 3 (
DHX58), and interferon induced with helicase C domain 1 (
IFIH1). This finding suggests that a central molecular mechanism may regulate diverse antiviral immune molecules in response to viral infections that may trigger asthma exacerbations and/or pathogenesis [
62]. IFIH1, as discussed earlier, was also differentially expressed.
One of the most intriguing results was the identification of a single transcriptional repressor, CTCF, that controls the expression of all but one of the differentially co-expressed pairs of genes. CTCF influences gene expression through chromatin modifications [
63] resulting in insulation of the target regions [
64]. CTCF is an architectural protein that mediates inter- and intra-chromosomal interactions at distant genomic sites, and regulates three-dimensional genome architecture [
63]. There are examples of CTCF silencing one gene while activating another [
63]. Specific to asthma, differential expression at the
ZPBP2/
GSDMB/
ORMDL3 locus was identified resulting from allele-specific chromatin remodeling mediated by CTCF [
65]. A SNP in
ZPBP2 created a CTCF binding site resulting in increased expression of ZPBP2 but diminished expression of
GSDMB and
ORMDL3 [
65]. Additionally, CTCF is highly sensitive to DNA methylation at CTCF binding sites [
63]; changes to the methylome can have direct effects on the regulation of CTCF target genes. CTCF could play a crucial role in controlling the many gene expression changes observed in the airways of asthmatics and is worthy of more intense research that is beyond the scope of this paper.
There are several limitations of this study. Firstly, the use of whole airway RNA rather than RNA from specific cell types precludes us from conclusively identifying the site of gene expression. Secondly, the majority of the asthmatic patients were fatal asthmatics and experienced hypoxia and treatment with steroids during their fatal attack which can affect mRNA expression in tissues taken for research purposes. We addressed this by controlling for steroid use in our analysis of differential gene expression. Additionally, the use of more severe asthmatics may mean that these results are not generalizable to asthmatics as a whole. But considering severe asthmatic populations have the most hospital visits and are most at risk for exacerbations, we believe our results provide significant insight into the genes that are altered in fatal disease. Finally, the relatively small sample size limits our ability to detect differences in gene expression less than ~1.5 fold on average, although this also means that the changes we see are likely to be real. Procurement of donor lungs is time consuming and costly so increasing the number of patients for this study was not feasible, however future studies in asthmatic biopsies or cell culture experiments could confirm these results with the ability for much larger sample size. A limited number of donors with non-fatal asthma (n = 4) means we were unable to test for differences between these two groups of donors (fatal vs. non-fatal).