Introduction
Asthma is a chronic inflammatory lung disease with common symptoms of wheezing and shortness of breath associated with airway remodeling and hyperresponsiveness (AHR) [
1‐
3]. Current therapies including bronchodilators and corticosteroids can be effective at managing symptoms and newer type 2-targeted biologics, such as duplizumab, may also be effective in subsets of patients with type 2-driven disease [
4]. However, gaps remain in our understanding of what drives key disease features and current therapies may not be effective in all patients.
Transcriptomic studies in asthma have provided valuable information in the whole lung context [
5‐
7], however, deciphering the individual contributions of the airway and parenchyma in disease pathogenesis may expedite the development of novel tissue-specific treatment strategies. A significant challenge in the study of gene expression patterns is deciphering the relationships between a large number of differentially expressed genes and overall functional effects. To address this, a network-based approach is being increasingly adopted to systematically explore the complex and dynamic transcriptional changes underlying diseases such as asthma and chronic obstructive pulmonary disease [
8‐
13]. By utilizing bioinformatics tools such as weighted gene correlation network analysis (WGCNA) [
14], co-expression network analysis can provide a holistic view of the global connectivity structure and functional organization of the gene expression program and identify subtle disease-associated changes in correlation structure that are not detected by conventional differential gene expression analysis [
11,
15]. Furthermore, since disease-associated genes often interact to form disease modules [
8], insights gained from co-expression network analysis can be leveraged to infer upstream module-driving genes that control entire disease pathways and can become promising therapeutic targets [
11,
16].
In this study, we performed airway and parenchyma transcriptomics using a house dust mite (HDM)-induced model of experimental asthma that replicates key features of the human disease [
17,
18]. Interrogation of HDM-induced genes using WGCNA upstream module analysis identified several transcription factors that regulate airway and/or parenchymal gene expression, including transcription factor PU.1 (SPI1). SPI1 inhibition using specific small molecule inhibitors, DB1976 or DB2313, had no protective effects on airway inflammation. However, DB2313 (8 mg/kg) decreased airway mucus secreting cell numbers, and DB2313 (1 mg/kg) and DB1976 (2.5 mg/kg and 1 mg/kg) reduced small airway collagen deposition. Both compounds decreased AHR. This study reveals important differences between the airway and parenchyma transcriptomic profiles in a HDM-induced model of asthma. We also demonstrate that SPI1 is important in HDM-induced experimental asthma pathogenesis and that its pharmacological inhibition reduces HDM-induced collagen deposition and AHR.
Discussion
The individual contributions of the airways and parenchyma in asthma pathogenesis is incompletely understood. Here, we performed transcriptomic analysis of the airways and parenchyma in a HDM-induced model of experimental asthma. Our approach, which was sufficiently powered (n = 10 per group) to assess upstream drivers of disease, highlights key similarities and differences between the airways and parenchyma. Notably, SPI1 was similarly upregulated in both the airways and parenchyma. Using pharmacological inhibitors of SPI1, DB1976 and DB2313, we show that treatment with these compounds has differential effects at different doses on airway-associated mucus secreting cell numbers and small airway collagen deposition, with no protective effects on airway inflammation. Significantly, both compounds at either dose decreased airway hyperresponsiveness, with DB2313 at 8 mg/kg having the strongest effect. Our study demonstrates the importance of assessing airway and parenchyma transcriptomics separately in HDM-induced experimental asthma. We define SPI1 as an important disease-associated transcription factor in both tissues and that its pharmacological inhibition may be a novel treatment strategy for the most intractable features of chronic asthma.
We first showed that the separation of airway and parenchyma data and in response to PBS or HDM on the PCA plot, are likely consequences of differences in the baseline gene expression profiles of the two tissue types, and HDM exposure. This is likely due to the different cellular compositions of these two compartments, including airway smooth muscle cells in the airway compartment, and ATII cells in the parenchyma. We next showed that there are distinctive sets of differentially expressed genes in the airways and parenchyma in HDM-induced experimental asthma. Gene cluster analysis showed that some similar pathways are upregulated in the airways and parenchyma, including leukocyte migration and regulation of cytokine responses, and these are comprised of numerous genes known to be involved in the early phase of allergen-induced airway inflammation [
35]. Interestingly, downregulated pathways, including muscle development and muscle system processes, were found in both the airways and parenchyma. It is well established that airway smooth muscle plays key roles in asthma, including in airway hyperresponsiveness [
36]. However, less is known about the role of vascular smooth muscle, although it has been shown that glucocorticoid treatment in asthma can increase airway vascular smooth muscle contraction and decrease airway blood flow [
37]. Our analysis shows that the downregulation of genes involved airway and vascular smooth muscle development and processes may be common to both processes and dysregulated in asthma. Furthermore, gene cluster changes in blood circulation and circulatory system processes were found to be up- and down-regulated in parenchyma samples and downregulated in airway samples. These data highlight a complex relationship between vascular smooth muscle, blood circulation and circulatory system processes in experimental asthma that warrants further investigation.
We next utilized WGCNA to unveil mechanisms that may not be detected through standard differential gene expression analysis. The three largest median increase in expression with HDM exposure modules in the airways and parenchyma consisted of innate immune, cell cycle and molecule transport/localization responses. We then used ChEA3 [
16] to rank transcription factors based on pooled information across six transcription factor target gene set libraries to identify those responsible for the observed changes in gene expression. Interestingly, the top 10 transcription factors by mean rank were similar in cell cycle modules in both the airways and parenchyma, however, there were some differences in the innate immune and molecule transport/localization responses in the airways and parenchyma.
We next interrogated the role of the transcription factor
Spi1 that is involved in the innate immune system module. Spi1 was upregulated in both the airways and parenchyma, and belongs to the erythroblast transformation specific (ETS) family of transcription factors that mediate a wide range of cellular processes, including growth, differentiation, signaling, and apoptosis [
38].
Spi1 also has known roles as a master regulator of hematopoiesis [
39,
40], and is expressed in the respiratory tract, brain, endocrine tissues, gastrointestinal tract, liver and gallbladder, pancreas, kidney and urinary system, reproductive tract, muscle tissues, skin, bone marrow and lymphoid tissues. DB1976 and DB2313 are small molecule SPI1 inhibitors that are heterocyclic diamidines that bind with high specificity to AT-rich sequences in the DNA minor groove that is selective for the SPI1 protein and thus competitively inhibit SPI1 binding [
41,
42]. In acute myeloid leukemia, inhibition of SPI1 using these inhibitors resulted in the downregulation of SPI1 downstream targets and reduced tumor burden [
42]. Furthermore, another study showed that DB1976 treatment downregulated pro-fibrotic gene expression in fibroblasts and had both preventative as well as therapeutic benefits in a bleomycin-induced skin fibrosis model [
43].
In our study, treatment with DB1976 increased airway eosinophils and decreased small airway collagen deposition but had no effect on mucus secreting cell numbers. In contrast, DB2313 increased neutrophil numbers, and reduced mucus secreting cell numbers and completely inhibited collagen (at 1 mg/kg but not 8 mg/kg). Significantly, treatment with both drugs at either dose reduced airway hyperresponsiveness, with almost complete inhibition with DB2313 at 8 mg/kg. SPI1 is proposed to modulate the expression of over 3,000 genes in hematopoietic cells [
44‐
46], therefore the differences in the effects of DB1976 and DB2313 may be linked to the specific downstream genes that were regulated, such as
IL1RN,
PTAFR,
ITGAX and
CLEC7A. Interestingly, some studies show that
Spi1 can suppress Th2 cytokine expression, and that
Spi1 expression can be upregulated through selective antagonism of microRNA (miR)-126 [
47‐
49]. Likewise, the deletion of miR-155 is associated with the upregulated expression of
Spi1 and reduced allergen-induced type 2 inflammation [
50]. Whether these pathways are affected in HDM-induced experimental asthma warrant further investigation. Importantly, the expression of the majority of genes that can be regulated by
Spi1 can also be controlled by other transcription factors, which complicates measuring genes that are proposed to be
Spi1-dependent as a readout of DB2313 and DB1976 target specificity. It is also important to consider that we administered DB2313 and DB1976
via the intranasal route to direct these compounds into the lungs, but we cannot exclude the possibility that these compounds entered the circulation and exerted effects on
Spi1 at peripheral sites. This is unlikely to affect the molecular specificity of these compounds to their target, but rather highlights that these extrapulmonary effects may promote feedback responses that exert effects on lung inflammation, structure, and function.
Whilst the effects of DB1976 and DB2313 on inflammation in HDM-induced experimental asthma were surprising, our data support the concept that inflammation, remodeling and AHR in allergic asthma can occur and be regulated independently [
30,
51]. It is now well established that there is a disconnect between inflammation and AHR in asthma patients and that they are regulated by different processes. In support of our findings, a study of patients with chronic asthma showed no correlation between the numbers of inflammatory cells present in BALF and the level of AHR [
52]. Our study, as well as those from others, provide strong evidence that airway remodeling can lead to constricted airways, which may drive the variation in AHR in asthma patients [
52‐
54]. Currently, biologics for the treatment of severe asthma primarily focus on targeting inflammation. Our findings suggest that a broader set of drug targets remain to be discovered that can selectively target inflammation, remodeling, or both.
Interestingly, treatment with both DB1976 and DB2313 reduced features of airway remodeling and AHR, suggesting that HDM-induced Spi1 responses could drive these potentially inter-connected processes. Indeed, another study showed that DB1976 treatment of TGFβ-stimulated fibroblasts decreased
COL1A1, α-SMA expression and F-actin [
43], supporting the concept that Spi1 responses can promote remodeling. Taken together, the data show that it is likely that increased Spi1 responses can have context-dependent, dichotomous roles. Of note, our bioinformatics approach utilized a database of published findings to identify SPI1 as a key upstream driver of genes involved in the innate immune module, and our functional studies demonstrate that SPI1 is important in airway remodeling and AHR. Collectively, our data show that our approach has the capacity to identify new pathways and the functional effects of factors that may cross over several modules and can inform further studies that interrogate novel relationships. Future studies that examine the effects of different treatment regimens with DB1976 and DB2313, including treatment after establishment of disease, may provide additional insights into how these compounds affect remodeling and/or AHR.
A limitation of our study is the different high doses of DB1976 (2.5 mg/kg) and DB2313 (8 mg/kg) that were dictated by drug solubility and preclude direct dose-to-dose comparisons between the two highest doses of these drugs. However, at 1 mg/kg, the disparities in the physiological responses to DB1976 and DB2313 may partially be due to differences in their chemical structure and subsequent binding affinity. Indeed, reported IC
50 values are lower for DB2313 compared to DB1976, indicating that DB2313 exhibits higher binding affinity for its target [
42]. Whilst the selectivity of DB1976 and DB2313 for SPI1 has been established [
41,
42], a rigorous examination of the specificity of these compounds against a large range of targets has not been performed. Differences in the specificity of these compounds could explain the differences in the physiological responses that were observed in vivo, and this will be important to examine in future studies. Additionally, DB1976 exhibits increased cellular toxicity in Lin
−Sca1
+c-Kit
+ bone marrow cells from wild type mice compared to DB2313 [
42], and DB2313 has more pronounced effects on reducing cell viability in THP-1 cells at high doses compared to DB1976 [
42]. This suggests that high doses of these compounds have differential effects on cell viability depending on the type of cell that is exposed. However, which specific cell types are affected by intranasal drug administration in our model is unknown and a rigorous investigation of this is beyond the scope of our study but would be interesting to assess in future work by performing systemic, rather than lung local, administration. Furthermore, our study has used female mice, and it is well known that are sex-specific effects to allergen exposure [
55], thus future studies examining the efficacy of DB1976 and DB2313 in male mice are warranted.
Our study demonstrates the advantages and value of using airway and parenchymal transcriptomics to assess and interrogate the roles of gene expression modules and upstream transcription factors at the lung local level. Importantly, we show that targeting one of these factors, Spi1, by direct administration of novel pharmacological agents allows us to delineate whether these modules are drivers or consequences of disease, and our data highlight the potential for targeting increased Spi1 responses in the lungs as a novel therapeutic strategy to reduce remodeling and AHR in asthma patients.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.