Background
The airways of patients with asthma display a number of characteristic structural changes that are collectively referred to as airway remodeling. These structural changes include an increase in airway smooth muscle mass, goblet cell hyperplasia/metaplasia leading to increased mucus production, epithelial fragility [
1,
2], and increased numbers of fibroblasts and myofibroblasts in the subepithelial region that are associated with increased extracellular matrix protein deposition and thickening of the lamina reticularis [
3,
4].
Although it was long thought that airway remodeling developed only after years of disease expression, bronchial biopsy studies have now established that components of airway remodeling can be demonstrated in early childhood, even prior to the formal diagnosis of asthma being made [
5]. Several studies have demonstrated features of airway remodeling, including thickening of the lamina reticularis, increased smooth muscle mass, increased mucus gland area and angiogenesis in pre-school age children [
6‐
9]. Interestingly, such features were not observed in symptomatic infants with reversible airflow obstruction [
10]. This suggests that airway remodeling occurs in response to some inciting stimulus (or stimuli) in early life, and is not a congenital phenomenon.
Human rhinovirus (HRV) infections in early life are a common trigger for childhood wheezing illnesses [
11,
12], and HRV-induced wheezing illnesses in the first 3 years of life are a major risk factor for subsequent development of asthma [
13]. Longitudinal analysis has shown that pre-school age children have about six HRV infections per year [
14], and serial viral infections can lead to recurrent wheezing episodes [
15]. Given the concurrent time frame of recurrent HRV-induced wheezing episodes and the development of AR, it is possible that HRV infections may be one stimulus that could contribute to the initiation and progression of airway remodeling in asthma.
The human airway epithelial cell is the primary site of HRV infection and previous studies have shown that infected epithelial cells release a number of growth factors and proteins linked to airway remodeling [
16‐
18]. The epithelium has also been suggested as a potential source of the increased numbers of mesenchymal cells in the airways during a number of disease conditions through the process of epithelial to mesenchymal transition (EMT), in which typical epithelial phenotypic proteins and morphology are lost and features of mesenchymal cells are acquired [
19‐
22]. In the current studies, we tested the hypothesis that HRV infection of airway epithelial cells could contribute to airway remodeling by inducing EMT. Moreover, because the cytokine transforming growth factor-β (TGF-β) has been described to be a key mediator of EMT in human epithelial cells [
20,
23], and is present in increased amounts in asthmatic airways [
24,
25], we also examined potential interactions of HRV with TGF-β in inducing EMT.
Methods
Epithelial cell culture
Major group HRV species, which use intercellular adhesion molecule-1 (ICAM-1) as a receptor, do not infect more than 5-10% of primary human bronchial epithelial cells in culture [
26]. Therefore, we chose to perform the current studies using the BEAS-2B human bronchial epithelial cell line, which uniformly expresses ICAM-1 [
27]. This cell line was derived [
28], and kindly provided by Dr. Curtis Harris (NCI, Bethesda, MD). The BEAS-2B cell line has been widely used in the literature to study EMT [
23,
29‐
31], as well as epithelial response to HRV [
18,
32‐
34]. BEAS-2B cells were cultured in Bronchial Epithelial Basal Medium (BEBM) (Lonza, Walkersville, MD) supplemented with bovine pituitary extract, epidermal growth factor, epinephrine, gentamicin and amphotericin B, hydrocortisone, insulin, trans-retinoic acid, transferrin, and triiodothyronine to make Bronchial Epithelial Growth Medium (BEGM) on 6-well culture plates at 37 °C and 5% CO
2.
HRV generation and purification
Two serotypes of HRV were used, both of which are members of the HRV-A group. Stocks of major group HRV-16, or of minor group HRV-1A, which uses members of the low-density lipoprotein receptor (LDLR) family for cell entry, were originally obtained from American Type Culture Collection (Rockville, MD). HRV-16 was propagated in WI-38 fetal lung fibroblasts cells (American Type Culture Collection, Rockville, MD), while minor group HRV-1A was propagated in H1-HeLa cells (American Type Culture Collection, Rockville, MD). Both strains of HRV were purified by sucrose density centrifugation to remove ribosomes and soluble factors as previously described [
35,
36]. Viral titers were determined by using WI-38 or H1-HeLa cells grown in 96-well plates as previously described [
36].
Treatment of epithelial cell culture
When cells achieved 60% confluence, culture medium was removed and replaced with BEGM from which hydrocortisone had been removed (BEGM no HC), for 24 hours prior to stimulation. Fresh BEGM no HC was then added and BEAS-2B cells were infected with 104.5 50% tissue culture-infective dose (TCID50) U/ml of either purified HRV-16 or purified HRV-1A alone, 10 ng/ml recombinant TGF-β1 (R&D Systems, Minneapolis, MN), or the combination of each HRV and TGF-β1, and subsequently incubated at 34 °C in 5% CO2 for specified time points. In 120-hour experiments, the medium and TGF-β1 were replaced at 48 and 96 hours. We demonstrated that treatment with HRV alone, TGF-β1 alone, or the combination, did not reduce cell viability using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) viability assay (data not shown). For studies examining activation of mitogen activated protein kinase (MAPK) signaling pathways, BEAS-2B cells were pretreated for 24 hours in BEGM no HC then placed in BEBM for one hour prior to experiment treatment to reduce basal activation. BEBM was used for treatment medium.
Replication-deficient HRV
To render HRV replication-deficient, stocks of purified HRV-16 and HRV-1A were exposed for 5 min to a Spectroline Model XX-15 F high intensity short wavelength (254 nm) ultra-violet (UV) lamp (Spectronics Corp., Westbury, NY) at a distance of 5 cm. It was confirmed that treated virus was replication-deficient by showing that the ability to cause lysis in appropriate host cells (WI-38 or H1-HeLa) was lost. UV-treated virus was used in experiments at doses identical to those for intact (replication competent) HRV in all experiments.
Pharmacological Inhibitors
Commercially available pharmacological inhibitors, SB203580 and PD95089 from InvivoGen (San Diego, CA), were used to inhibit p38 MAPK and MEK1/MEK2 (ERK 1/2) MAPK pathways, respectively. For these studies, BEAS-2B cells were pretreated with the inhibitor of interest for 1 hour prior to addition of HRV-16, HRV-1A, TGF-β1, and the combination of each HRV serotype with TGF-β1. Cells were harvested for whole-cell lysates at 24 hours.
Whole-cell lysates
Post infection, BEAS-2B cells were lysed in ice-cold lysis buffer (1% Triton X-100 in 1 X MES buffered saline pH 7.4, containing anti-protease tablets, 50 nM sodium orthovanadate, 0.4 M sodium pyrophosphate, and 1 M sodium fluoride). Lysates were analyzed for protein content using a DC Protein Assay (BioRad Laboratories, Mississauga, ON, Canada) as manufacturer recommended.
Western blots
Equivalent amounts of whole cell lysate were separated by 10% SDS-PAGE, and then transferred to a 0.45 μm nitrocellulose membrane. Membranes were blocked with 5% skim milk, and incubated overnight at 4 °C with the following antibodies: E-cadherin (BD Biosciences, San Jose, CA), cytokeratin-18 and cellular fibronectin (both from Sigma-Aldrich, Oakville, ON, Canada), vimentin, phospho-ERK1/2 mitogen activated protein kinase (MAPK), total-ERK1/2 MAPK, phospho-p38 MAPK, total-p38 MAPK, SLUG, SNAIL, or phospho-SMAD 2/3 (all from Cell Signaling, Danvers, MA). Membranes were then stripped and re-probed for the housekeeping protein, β-tubulin (Sigma-Aldrich, Oakville, ON, Canada).
Densitometry analysis
Densitometry analysis was performed using ImageJ software (version 1.41, National Institute of Health, Bethesda, MD). The fold change in expression of the protein of interest was determined by normalizing the raw pixel count of the protein of interest to the raw pixel count of the respective housekeeping protein. This was then compared to the normalized value for medium control. Data is expressed as fold change over medium control unless specified otherwise.
Assessment of morphological changes
To determine the percentage of cells exposed to each treatment that acquired a mesenchymal morphology, we used ImageJ software to measure the length and width of 75 BEAS-2B cells grown under normal medium conditions. The mean and standard deviation of the length to width ratio was then calculated for these 75 cells. We then calculated the mean length to width ratio + 10 standard deviations and established this number as a threshold that must be exceeded before a cell was considered sufficiently long and thin (spindle-like) to be classified as of a mesenchymal morphology. Using a grid reticule, we then used the same software to count a minimum of 50 cells from each treatment condition in each of three separate experiments and calculate the percentage of cells (mean ± SEM) that showed a mesenchymal phenotype in response to each treatment.
Statistical analysis
All statistics were performed using GraphPad Prism 6 (GraphPad Software, CA). Data is presented as mean ± standard error of the mean. Non-Gaussian distributed data was analyzed by Kruskal-Wallis one-way analysis of variance (ANOVA) with Dunn’s post hoc analysis. Synergy between groups was determined by summing the values obtained in response to each individual treatment per experiment, and comparing that to the value obtained in response to the combined treatment using Wilcoxon matched-pairs analysis. For all statistical tests, a two-tailed p value of ˂0.05 was considered significant.
Discussion
Recurrent HRV-induced wheezing illnesses in the first 3 years of life are a major risk factor for subsequent development of asthma [
13]. Given that this time frame also coincides with the development of airway remodeling, we hypothesized that HRV infections may contribute to aspects of airway remodeling by altering epithelial cell biology. We, and others, have previously shown that HRV infection of human airway epithelial cells, both
in vitro and in
vivo, can lead to the induction of a number of growth factors linked to airway remodeling [
16‐
18], as well as of chemotactic agents that can cause migration of mesenchymal cells towards the epithelial layer [
43]. We now provide direct evidence that two different serotypes of HRV are also able to drive epithelial phenotypic and morphological changes that are indicative of EMT, further supporting a potential role of HRV infections in airway remodeling. Our data contrast somewhat with a recent report that HRV-39 could induce EMT-like marker protein changes in regenerating, but not normal epithelial cell cultures [
44].
The pleiotropic cytokine, TGF-β
1 has been reported to be a key initiator of EMT in profibrotic processes occurring in the airways in asthma [
20,
23,
45], and levels of TGF-β
1 are elevated in the airways of both children and adults with asthma [
24,
25]. Given the central role ascribed to TGF-β
1 in triggering EMT, we were surprised to observe that both serotypes of HRV tested were more effective than TGF-β
1 at inducing the loss of the epithelial phenotypic marker proteins, E-cadherin and cytokeratin-18. This data contrasts with prior reports of TGF-β
1 triggering the loss of epithelial markers in both BEAS-2B cells [
23] and primary human airway epithelial cells [
20]. The reasons for this discrepancy are unclear, although differences in culture medium and growth conditions could potentially explain the difference between our data and those of Doerner and colleagues who grew BEAS-2B cells in a keratinocyte culture medium [
23]. It is not feasible that BEAS-2B cells used in our studies do not express receptors for TGF-β
1, given that this cytokine was clearly effective in inducing mesenchymal markers. Our data were consistent with those of Heijink and coworkers who also found that TGF-β
1 had only a modest effect on the loss of epithelial markers [
46]. Nonetheless, to examine this issue further we used activin A, another member of the TGF-β superfamily that we have previously shown to be induced in both BEAS-2B and primary airway epithelial cells in response to HRV infection [
16] as an alternative ligand for the TGF-β receptor. Consistent with the data obtained using TGF-β
1, activin A also caused only a modest loss of E-cadherin but synergized with HRV-16 in inducing loss of this epithelial marker.
It is not clear why exposure to HRV is so effective in downregulating epithelial markers. One possible explanation relates to observations that, in addition to triggering some of the signal pathways studied in the current work, both the major (ICAM-1) and minor (LDL-receptor) HRV receptors are known to be associated with lipid rafts [
47], which, in turn, are known to associated with the actin cytoskeleton [
48]. Indeed, it has been known for some time that ICAM-1 is associated with the actin cytoskeleton [
49]. More recently, it has been shown that binding of HRV to ICAM-1 is known to recruit the cytoskeletal linker protein ezrin to ICAM-1 [
37]. Thus binding of HRV to it receptor (ICAM-1 or LDLR) likely leads to rearrangements in the actin cytoskeleton and/or microtubules, as has been shown for other picornaviruses [
50]. Since the actin cytoskeleton is integral to the retention of an epithelial phenotype, and since cytokeratin and junctional proteins, such as E-cadherin, are closely affiliated with the actin cytoskeleton, this may lead to an increased capacity of HRV to induce loss of these epithelial markers, setting the stage for mesenchymal transition.
In contrast to data for epithelial markers, TGF-β
1 alone was more effective than either serotype of HRV alone in inducing expression of the mesenchymal markers, fibronectin and vimentin. There was a clear interaction between the two stimuli, however, as the combination of HRV and TGF-β
1 clearly induced synergistic changes in both epithelial and mesenchymal markers. The combination of HRV and TGF-β
1 was also most effective in causing morphological changes consistent with EMT. This is the first demonstration of an interaction between HRV and TGF-β
1 in causing EMT-like changes, but our data show some similarities to those of Heijink and co-workers, who also found that TGF-β
1 had only modest effects on reducing expression of epithelial markers, but synergized with house dust mite to induce changes consistent with EMT [
46]. Moreover, TGF-β
1 has also been reported to synergize with nicotine to induce EMT [
51].
EMT is generally considered to be a slow and progressive process, with intermediate stages in which cells have the characteristics of both epithelial and mesenchymal cells [
52]. It was unexpected, therefore, to find that changes associated with EMT, and interactions with TGF-β
1, could already be observed within 24 h of HRV exposure, and to find that viral replication was not required for HRV to exert its effects. This implied a role for early, receptor-dependent signaling, rather than replication dependent signaling through double stranded RNA pattern recognition receptors. Given that both major and minor group HRV serotypes induced phenotypic changes, relevant signals must be sent via either ICAM-1 or LDLR. Because little is known about early signaling through LDLR, we examined activation of the p38 and ERK1/2 MAPK pathways, as these are known to be activated both by TGF-β
1 [
53,
54], and via ICAM-1 signaling [
37,
41,
42], to determine if these may also contribute to signaling via the LDLR. As expected, HRV-16 and TGF-β
1 each induced activation of the p38 MAPK pathway and we also provide the first evidence that HRV-1A also induced p38 activation. The combination of each serotype of HRV with TGF-β
1 led to an approximately additive activation in each case. Despite a prior report of ERK1/2 activation by TGF-β
1 in intestinal epithelial cells [
54], TGF-β
1 alone did not induce appreciable activation of the ERK1/2 pathway in BEAS-2B cells. As expected, HRV-16 induced robust activation of ERK1/2 and this was slightly enhanced in the presence of TGF-β
1. By contrast, HRV-1A was considerably less effective in activation of ERK1/2 and there was no additional response in the presence of TGF-β
1.
To determine the potential role of MAPK pathways in EMT induced by HRV and TGF-β1, selective inhibitors of each pathway were used to examine effects on fibronectin expression. Inhibition of the p38 pathway significantly inhibited fibronectin expression induced by either serotype of HRV alone, and in combination with TGF-β1. Surprisingly, inhibition of this pathway had little effect on responses to TGF-β1 alone. Although TGF-β1 did not induce a marked activation of ERK1/2, inhibition of this pathway significantly reduced fibronectin induction in response to the combination of HRV-16 and TGF-β1 without affecting the modest response to HRV-16 alone. Interestingly, fibronectin induction in response to HRV-1A, either alone or combined with TGF-β1, was enhanced by inhibition of the ERK1/2 pathway. Thus the p38 MAPK pathway appears to play a role in induction of EMT by both major and minor group HRV strains, alone and in conjunction with TGF-β1, but the ERK1/2 pathway exerts different effects depending upon the HRV strain receptor involved.
To further examine the underlying basis of the potential mechanisms by which HRV may interact with TGF-β
1 to induce EMT, we examined pathways implicated in TGF-β
1 signaling. We provide the first evidence that both HRV-16 and HRV-1A are not only able to induce expression of the E-cadherin transcriptional repressor, SLUG, but that the combination of each HRV serotype with TGF-β
1 induces activation of SLUG that is greater than that seen with TGF-β
1 alone. This provides a potential mechanism to support the reduced expression of E-cadherin seen with HRV alone, and the synergistic induction observed with the combination of HRV and TGF-β
1. Interestingly, consistent with an earlier report [
44], while each HRV serotype appeared to modestly induce activation of a second, related transcriptional repressor SNAIL, the combination of HRV and TGF-β
1 did not induce any activation above that observed with TGF-β
1 alone. Finally, we showed that neither HRV-16 nor HRV-1A induced activation of the SMAD2/3 pathway, and the combination of either strain of virus with TGF-β
1 had no effect above that seen with TGF-β
1 alone.
A limitation of this study is the use of the BEAS-2B bronchial epithelial cell line. This was chosen because less than 10% of primary cells can be infected by major group HRV strains [
26], raising concerns that insufficient “signal-to-noise” would be obtained. As noted, the BEAS-2B cell has been commonly used to examine EMT [
23,
29‐
31], and has been shown to reproduce effects observed with HRV in primary cells in a number of studies [
16,
55‐
57]. Nonetheless, it will be important to extend these studies to primary epithelial cells. This should be feasible having established that HRV-1A induced EMT as minor group HRV can infect a greater percentage of primary cells than major group strains [
58]. It will also be important to determine if HRV-induced EMT is enhanced in epithelial cells from asthmatic children or adults.
It should be noted that, while EMT is a well-recognized process in both physiological and pathophysiological conditions, it remains unclear what percentage of fibroblasts and myofibroblasts in the asthmatic airways derive from EMT. Studies using genetic fate tracking of lung epithelial cells in a mouse model of allergic asthma reported that approximately 30% of murine lung fibroblasts were derived from epithelial cells that underwent EMT [
59], but it is far from clear if this is the case in humans.