Involvement of the epidermal growth factor receptor in IL‐13–mediated corticosteroid‐resistant airway inflammation

Abstract Background Effective treatment for severe asthma is a significant unmet need. While eosinophilic inflammation caused by type 2 cytokines is responsive to corticosteroid and biologic therapies, many severe asthmatics exhibit corticosteroid‐unresponsive mixed granulocytic inflammation. Objective Here, we tested the hypothesis that the pro‐allergic cytokine, IL‐13, can drive both corticosteroid‐sensitive and corticosteroid‐resistant responses. Results By integration of in vivo and in vitro models of IL‐13–driven inflammation, we identify a role for the epidermal growth factor receptor (EGFR/ERBB1) as a mediator of corticosteroid‐unresponsive inflammation and bronchial hyperresponsiveness driven by IL‐13. Topological data analysis using human epithelial transcriptomic data from the U‐BIOPRED cohort identified severe asthma groups with features consistent with the presence of IL‐13 and EGFR/ERBB activation, with involvement of distinct EGFR ligands. Our data suggest that IL–13 may play a dual role in severe asthma: on the one hand driving pathologic corticosteroid‐refractory mixed granulocytic inflammation, but on the other hand underpinning beneficial epithelial repair responses, which may confound responses in clinical trials. Conclusion and Clinical Relevance Detailed dissection of those molecular pathways that are downstream of IL‐13 and utilize the ERBB receptor and ligand family to drive corticosteroid‐refractory inflammation should enhance the development of new treatments that target this sub‐phenotype(s) of severe asthma, where there is an unmet need.


| INTRODUC TI ON
Asthma is a heterogeneous disease characterized by a diverse profile of symptoms, severity and responses to medications. In mild-to-moderate asthma, treatment with inhaled corticosteroids can significantly reduce inflammation and control symptoms. However, in patients with more severe disease, symptoms can persist despite receiving standard of care treatment, including antibody-based biologics. 1 This subgroup has been defined as "severe refractory" asthma. 2,3 Up to 10% of the asthmatic population are classed as severe; they have higher rates of asthma exacerbations, increased morbidity and account for a disproportionate use of healthcare resources, accounting for more than 60% of the economic burden. 4 Much work has been done to advance the clinical understanding of severe refractory asthma, but there still remains a clear unmet clinical need. 4 The recent focus on disease heterogeneity has raised the concept that asthma consists of multiple phenotypes with distinct underlying mechanisms (endotypes). [4][5][6][7] Initially, the focus was on identification of subgroups based on clinical presentation including exacerbations, persistent symptoms and reduced lung function.
However, with the advent of 'omic technologies (transcriptomics, proteomics, metabolomics, lipidomics) and the recruitment of large patient cohorts, [8][9][10][11] molecular and biological processes can now be evaluated with the aim of identifying distinct endotypes.
A key phenotypic characteristic that is used to define asthma subgroups is the presence or absence of biomarkers of type 2 inflammation. 4 Typical Type-2 biomarkers include bronchial epithelial gene expression of POSTN (Periostin), CLCA1 (Chloride Channel Accessory 1) and SERPINB2 (Serpin Family B Member 2), 12 blood and sputum eosinophils, 13 fractional exhaled nitric oxide (Feno), 14 serum periostin, 12,15 serum IgE 13 and type-2 gene expression (IL4, IL5, IL13) in sputum cells. 16 Elevated levels of these biomarkers are frequently associated with the presence of atopy and bronchial hyperresponsiveness (BHR) 17 and are considered to be sensitive to inhibition by corticosteroids and monoclonal antibodies against IgE, IL-5 or IL-5 receptor. 1,12,17 In contrast, patients who have "non-Type-2" (also known as "Type-2-low") severe asthma are frequently characterized by airway neutrophilia, a Type-17 immune signature involving genes such as CXCL1, CXCL2 and CSF3 18 and disease that is not effectively treated by inhaled corticosteroids or biologics. 6 While this broad classification of corticosteroid responsiveness in relation to Type-2 inflammation is a reasonable generalization, studies with dupilumab, an antibody to IL-4Rα, suggest it is an over-simplification. Thus, treatment of uncontrolled persistent asthmatic patients with dupilumab as an add-on therapy increased lung function and reduced severe exacerbations irrespective of baseline eosinophil count. 19 Subsequent studies confirmed that dupilumab treatment resulted in a reduction in glucocorticoid use and severe exacerbations. 20,21 As these beneficial effects of dupilumab, which inhibits both IL-4 and IL-13 signalling, 22 were seen in patients who were receiving medium-to-high-dose inhaled corticosteroids plus a long-acting β2 agonist, this suggests that these asthma patients have persistent Type-2 responses despite inhaled corticosteroid therapy.
Consistent with this, it has been reported that airway Type-2 inflammation is not suppressed adequately in approximately half of asthmatic patients treated with inhaled corticosteroids 23 and that there are many severe asthma patients with mixed eosinophilic and neutrophilic inflammation. 24 Based on these clinical observations, the aim of this work was to model allergic airway inflammation using a transgenic IL-13 mouse 25 and to test the hypothesis that a subset of IL-13-induced responses are corticosteroid-unresponsive and contribute to ongoing airway symptoms. We show that IL-13-induced airway inflammation involves a corticosteroid-refractory component characterized by

| Mouse tracheal epithelial cell expansion
Tracheal epithelial cells were isolated and expanded as previously described. 27 Isolated cells were grown on Transwells ® (Corning) in submerged culture to allow formation of tight junctions, determined by trans-epithelial electrical resistance (TER). Cells were cultured until the TER was >1000 Ω.cm 2 . IL-13 was induced by addition of 2 µmol/L DOX (Sigma Aldrich) to the culture medium. When required, AG1478 (Sigma Aldrich) and/or Dexamethasone (Hameln) were added to the culture medium, each at a concentration of 10 µmol/L. Apical and basolateral secretions were collected at 48 and 72 hour for KC and CXCL2 protein quantification and cells removed from the transwells in TRIzol and pooled in duplicates for RNA extraction.

| Assessment of lung function
Mice were anaesthetized with 100 μL of anaesthetic containing a 4:1:1 mixture of ketamine, acepromazine and xylazine by intraperitoneal injection. A FlexiVent system (SCIREQ) was used to assess lung function in the form of airway resistance (R) after aerosolized methacholine challenge to provide a measure of BHR, according to the manufacturer's instructions. Airway resistance was measured by forced oscillation technique, with increasing values indicating bronchoconstriction of the lungs. BHR measurements were obtained from individual animals using increasing stepwise concentrations of 0, 2.5, 5 and 10 mg/mL methacholine (Sigma Aldrich).

| Inflammatory cell counts
Bronchoalveolar lavage (BALF) samples were collected by washing the lungs three times with 800 μL sterile PBS. The total volume of the combined fluids was measured and then centrifuged at 300 g for 5 minutes. The BALF supernatants were frozen for cytokine analysis. Red blood cells were lysed from the cell pellets, which were subsequently resuspended in 300 μL PBS. Cells were counted, and 100 000 cells were loaded into a cytospin funnel and centrifuged at 300 g for 5 minutes on to a glass slide. Slides were air-dried, and the cells were stained using a Diff-Quick stain (Siemens). The different inflammatory cell types were counted to a total of 300 cells and expressed as the differential cell count in cells/mL of BALF.

| Immunohistochemistry
The lungs were inflation-fixed at constant pressure with 10% neutral buffered formalin and embedded in paraffin wax. 5μm sections were stained using haematoxylin and eosin (H&E).

| Statistical analyses
Normal distribution of the numeric data was evaluated, and appropriate parametric or non-parametric statistical tests applied.
Parametric data are plotted as mean with one standard deviation (SD) while non-parametric data are shown as boxes representing the 25  was used. For comparison of two or more groups with two independent variables, a two-way ANOVA with Tukey's multiple comparison test was used. * = P < .05, ** = P < .01, *** = P < .001.

| The U-BIOPRED cohort
U-BIOPRED is a multi-centre study that enrolled 311 severe, nonsmoker asthmatics, 88 mild/moderate non-smoking asthmatics and 101 healthy controls. 8 Among these participants, 61 severe asthmatics, 36 mild/moderate asthmatics and 44 healthy volunteers underwent fiberoptic bronchoscopy and epithelial brushing. Epithelial brushings were processed into RNAlater for subsequent analysis on Affymetrix U133 Plus 2.0 microarrays, and only those passing stringent quality control were analysed. 10 All participants provided written informed consent to participate in the study which was approved by national ethics committees. The transcriptomic data are stored as GSE76226. 29 Using this dataset, epithelial expression of EGFR and EGFR ligands was compared between groups using SPSS v24. Paired t tests were applied to the log2-transformed transcriptomic data, while the clinical data were analysed by Kruskal-Wallis test and Mann-Whitney U or Student t tests depending on the type of data distribution; P < .05 was considered significant. Topological data analysis (TDA) was then applied to the epithelial transcriptomic data as described previously, 30,31 with some modification, using the Ayasdi Core software (Ayasdi, Menlo Park, CA) with a norm correlation metric and the neighbourhood lens (resolution, 20 bins; gain, 4.00). Clinical and pathobiological data were then overlaid as metadata onto the generated TDA network to look for associations.

| IL-13 induces BHR and a mixed inflammatory phenotype
IL-13 was significantly elevated ( Figure S1A, B) when Ccsp/Il13 mice were fed DOX to induce transgene expression. As previously reported, 25 when challenged with methacholine, these IL-13-expressing mice showed a significant increase in BHR (P < .002) ( Figure 1A
Furthermore, BALF CXCL1 levels were correlated with the amount of IL-13 in the BALF (r 2 = 0.74, P = .001) ( Figure S3C) consistent with a role for IL-13 in driving airway neutrophilia via CXCL1. CXCL2 protein was also significantly elevated in both the BALF and lung lysate ( Figure S4A,B).

| Corticosteroid treatment does not influence neutrophilic inflammation and BHR
To determine how airway responses are modulated by corticosteroid treatment, IL-13 was induced in Ccsp/Il13 mice for 7 days and dexamethasone (Dex) was given via intraperitoneal injection for the duration of IL-13 induction (7 days) or for the final 5 or 3 days of induction. The control group received DOX to induce IL-13 and were sham-treated with saline for 7 days. Dex significantly reduced eosinophil numbers after 7 days of treatment; furthermore, even shorter treatments given during the final 3 or 5 days of IL-13 induction significantly suppressed eosinophil numbers ( Figure 3A). In contrast, neither infiltration of neutrophils ( Figure 3B) nor BHR ( Figure 3C) were affected by presence or duration of Dex treatment. Although long-term treatment with corticosteroids can have metabolic effects, 32 the dose and duration used had no effect on body weight ( Figure S5). In addition, Dex treatment did not cause any inflammation in the airways of similarly treated sTg mice, nor did it affect body weight.

| The EGFR is a mediator of IL-13-induced corticosteroid-insensitive responses in vitro and in vivo
We have previously reported that EGFR activation drives corticosteroid-insensitive release of IL-8 from human bronchial epithelial cells. 33,34 As IL-13 has been shown to drive epithelial proliferation via an EGFR/TGF-α autocrine loop, 35 we explored whether EGFR activation could contribute to the responses observed in the IL-13 mice.

| Transcriptomic analysis of U-BIOPRED cohort
To explore the relevance of our findings in human asthma, epithelial transcriptomic data from healthy (n = 44), mild-to-moderate asthma (n = 36) and severe asthma (n = 61) were clustered by TDA which provides a general framework to analyse high dimensional data in a manner that is insensitive to the particular metric chosen and provides dimensionality reduction and robustness to noise. 31,38 It has the advantage over standard clustering methodologies in that it provides geometric representations of multi-dimensional data and it is often possible to find subgroups in data sets that traditional methodologies fail to find. The clinical, pathobiological data and log2transformed expression levels of ERBB receptors and EGFR ligands were then applied as metadata ( Figure 7A,B). This revealed a cluster of severe asthmatics with neutrophilia and varying numbers of eosinophils ( Figure 7A). While EGFR expression was down-regulated in this cluster, ERBB3 and several EGFR ligands, including heparinbinding EGF (HB-EGF), epiregulin (EREG) and EGF, were up-regulated ( Figure 7B). This severe asthma cluster was associated with increased expression of CSF3, CXCL2 and CXCL8, genes usually associated with an "IL-17 signature", 39 although it was evident that expression of these genes was heterogeneous across the cluster. Furthermore, within this severe asthma cluster it was possible to identify two subclusters: one was characterized by highest IL13 (Figure 7C), ERBB3 and HB-EGF expression; the other sub-cluster was characterized by or "IL-17 high" phenotypes but not in those subjects with an "IL-13/ IL-17 low" phenotype. In contrast, in the latter group, ERBB4 was the only ERBB family member to show significant modulation (Table 1).
For the EGFR ligands, HBEGF was significantly increased in the "IL-

| D ISCUSS I ON
Severe corticosteroid-refractory asthma is a significant unmet medical. The disease is usually classified based on inflammatory cell profiles and related pathways, giving rise to the dichotomous F I G U R E 2 Expression of IL-13 induces expression of both Type-2 and proneutrophilic mediated markers. Relative mRNA expression in whole-lung lobe lysates from littermate controls (white bars) vs Ccsp/Il13 mice (grey bars) after induction of IL-13 for 7 d for A, Ccl11, B, Postn, C, Serpinb2, E, Cxcl1/KC, F, Cxcl2 and G, Csf3. ELISAs for D, CCL11 and H, CXCL1 protein levels in BALF. For mRNA expression, n = 22 for controls and n = 23 for Ccsp/Il13 mice and n = 7 and n = 10, respectively, for protein, from three independent experiments. Non-parametric data are shown as box plots with medians and 25th to 75th percentiles, and whiskers representing minimum and maximum values with all data points shown; parametric data are shown as mean + SD. Statistical analyses were performed using Mann-Whitney test or Student's t test. ***P < .001 definitions of "Type-2" and "non-Type 2" asthma with, by inference, distinct underlying mechanisms. In this study, we have shown that the lungs of transgenic mice expressing the classical Type 2 pro-allergic mediator, IL-13, exhibited mixed eosinophilic and neutrophilic inflammation and increased expression of both Type-2 and non-allergic, "Type-17" markers, even though Il-17 was not elevated in the lungs of the mice. We also found that the characteristic Type-2 biomarkers can be significantly suppressed (but not ablated) by corticosteroid treatment whereas BHR, neutrophilia and pro-neutrophilic biomarkers were corticosteroid-refractory. Through in vitro mechanistic studies, we demonstrated that these corticosteroid-refractory IL-13-induced pro-neutrophilic responses were sensitive to inhibition of the EGFR and that in vivo inhibition of EGFR signalling in the IL-13 mouse model suppressed pro-neutrophilic cytokine expression and reduced airway neutrophilia and BHR. To relate these findings to human asthma, transcriptomic analysis of epithelial brushings from human volunteers identified a cluster of severe asthmatics displaying eosinophilia and neutrophilia with increased expression several EGFR ligands and ERBB3 whose protein product is known to form heterodimers and signal with EGFR. 40 Importantly, epithelial expression of IL13 was found within a sub-cluster of these asthmatic subjects, suggesting that our transgenic mouse model, which expresses IL-13 in the airway epithelium, mirrors a sub-phenotype of the human disease. Together, our data suggest that epithelial expression of the pro-allergic cytokine, IL-13, can drive corticosteroid-resistant asthma which is mediated by epithelial EGFR/ERBB signalling driving a gene signature and phenotypic responses that are more classically associated with IL-17 and Th17 immune responses.
Furthermore, our data suggest that distinct "Type 17" sub-phenotypes of severe asthma may arise through involvement of different ERBB receptor and ligand combinations.
A range of pathways have been implicated in the pathogenesis of corticosteroid-refractory asthma, including increased activity of kinases, which phosphorylate the glucocorticoid receptor (GR) and prevent its nuclear translocation, and oxidative stress which inhibits the activity of histone deacetylase 2 (HDAC2) which mediates the actions of corticosteroids on pro-inflammatory cytokine expression. 41,42 Other studies have suggested that the neutrophil-high severe asthma phenotype which is poorly responsive to high-dose corticosteroids 43 might be a consequence of the treatment itself, since corticosteroids promote neutrophil survival. 44,45 Other studies have associated the neutrophilic phenotype with bacterial colonization or infection 46 and activation of IL-17-producing helper T cells. 47 However, in a randomized, double-blind, placebo-controlled study of brodalumab, a human anti-IL-17 receptor monoclonal antibody, in moderate-to-severe asthma there was no significant therapeutic effect. 48 Thus, the mechanisms of corticosteroid-refractory disease remain poorly understood, and as a result, no specific treatment is available for this difficult-to-treat group of patients. 42 Previous studies have shown that IL-13 can promote airway neutrophilia. For example, direct instillation of IL-13 into the tracheae of rats results in neutrophilia driven by chemokines including IL-8 49 while transgenic mice expressing IL-13 have been shown to exhibit chronic inflammation involving both eosinophils and neutrophils, as F I G U R E 3 Dexamethasone reduces eosinophil count but does not affect neutrophils or suppress BHR. Differential inflammatory cell counts for A, Eo, eosinophils and B, Neu, neutrophils in BALF after a time-course Dex treatment (grey bars) vs sham treatment (white bars) in Ccsp/Il13 mice. C, Airway resistance in response to increasing concentrations of methacholine (0-10 mg/mL). Non-parametric data are shown as box plots with medians and 25th to 75th percentiles, and whiskers representing minimum and maximum values with all data points shown; parametric data are shown as mean + SD. N = 5 per group from two independent experiments. Statistical analysis was performed using one-way ANOVA or Kruskal-Wallis test with Dunn's test for correction for multiple comparisons. **P < .01, ***P < .001 well as lung remodelling. 25 We have previously shown that EGFR protein expression is increased in the bronchial epithelium of asthmatic patients according to disease severity 54 ; furthermore, EGFR expression correlates with epithelial IL-8 levels and the extent of neutrophilic inflammation in severe asthma. 33,34 Although involvement of the EGFR in inflammatory responses has been demonstrated previously using murine models of ovalbumin 55 or house dust mite-induced allergic inflammation, 56,57 both of these models are corticosteroid sensitive and so cannot address those corticosteroid-refractory responses that are important in severe asthma. Using the IL-13 transgenic mouse model, we identified a subset of IL-13-driven F I G U R E 4 The Type 2 lung inflammation signature is reduced by dexamethasone, but not the proneutrophilic responses. Relative mRNA expression in whole-lung lobe lysates from Ccsp/Il13 mice after induction of IL-13 for 7 days and dexamethasone treatment for all 7, the final 5 or 3 days (grey bars) (n = 5) vs vehicle-treated Ccsp/ Il13 mice (white bars) (n = 5). A, Ccl11, B, Postn, C, Serpinb2, E, Cxcl1/KC, F, Cxcl2 and G, Csf3. ELISAs for D, CCL11 and H, CXCL1 protein levels in BALF. Parametric data are expressed as mean + SD and non-parametric data as box plots showing medians and 25th to 75th percentiles, and whiskers representing minimum and maximum values with all data points shown. Data are from two independent experiments. Statistical analysis was performed using one-way ANOVA or Kruskal-Wallis test with Dunn's test with correction for multiple comparisons. *P < .05, **P < .01 corticosteroid-refractory pro-neutrophilic responses, and by culturing murine epithelial cell cultures in vitro, we identified that IL-13 can induce similar pro-neutrophilic cytokine responses that are refractory to corticosteroids, yet they can be suppressed by EGFR inhibition. We then confirmed these findings in the IL-13 transgenic mouse model by showing that EGFR inhibition with AG1478 in vivo can prevent IL-13-induced corticosteroid-refractory pro-neutrophilic inflammatory responses, airway neutrophilia and BHR. One key difference between the corticosteroid-sensitive models of allergic airways inflammation and the transgenic mouse model is that Il13 is expressed within the bronchial epithelium of the transgenic mouse, rather than in immune cells, as in the allergic models. As discussed below, analysis of the U-BIOPRED cohort also identified IL13 expression in the bronchial epithelium of a subgroup of severe asthmatic subjects, suggesting that the cellular provenance of IL-13 may be an important determinant of disease activity.
The relevance of our findings in human asthma was explored using bronchial epithelial cell transcriptomic data from the U-BIOPRED cohort. TDA analysis revealed that EGFR and ERBB2 were significantly down-regulated while ERBB3 and a number of EGFR ligands, as well as IL13, were up-regulated in severe asthmatics with mixed eosinophilic and neutrophilic inflammation, and this was accompanied by up-regulation of CSF3, CXCL2 and CXCL8. 39 Although the low level of EGFR in this severe asthma cluster appears paradoxical, studies in cancer cells have shown that regulation of EGFR mRNA and protein expression is complex and can occur at multiple transcriptional and post-transcriptional levels in a cell type-specific fashion. 58,59 In our in vitro studies, we showed that IL-13-dependent EGFR activation suppressed Egfr mRNA expression suggesting feedback regulation of EGFR expression upon its activation in bronchial epithelial cells.
As EGFR protein levels are increased in severe asthma, 54 it would be of interest to study EGFR protein expression in the U-BIOPRED cohort and to determine whether post-translational mechanisms are also important for EGFR regulation.
A key finding from our analysis of the U-BIOPRED data was that IL13 mRNA expression was present in bronchial epithelial cells within a sub-cluster of the severe asthma cluster where genes usually associated with an "IL-17" gene signature 39 were also up-regulated. IL-13 expression is normally associated with immune cells such as Type 2 innate lymphoid cells (ILC2s), T helper 2 (Th2) cells mast cells and basophils, with ILC2s and Th2 cells being considered major sources of this cytokine, 60 especially in Type 2 asthma. 61 However, induction of IL-13 mRNA and protein has also been observed in wounded bronchial epithelial cells in vitro, 62 and in our own unpublished work, we also have observed a significant increase in IL-13 protein release from primary bronchial epithelial cells in response to challenge with double-stranded RNA (data not shown). In the published work, release of IL-13 following wounding was shown to enhance epithelial repair via HB-EGF. 62 Thus, it is significant that IL13 expression in the U-BIOPRED severe asthma cluster closely mirrored epithelial HBEGF expression suggesting a wound healing response in this subgroup of patients involving an IL-13/HB-EGF/EGFR axis. Of note, it has been shown that IL-13 receptor α2 (IL-13Rα2) can stimulate epithelial cell HB-EGF production via TMEM219 and that TMEM 219 also contributes to optimal binding of IL-13 to IL-13Rα2. 63 Unlike the Type II IL-4 receptor complex which is a target Dupilumab 22 and binds both IL-4 and IL-13, IL-13Rα2 is a high-affinity receptor for IL-13, but not IL-4. 64 Together, these observations may help to explain the so-called "IL-13 paradox" 65 that while IL-13 is involved in almost all aspects of asthma pathobiology, clinical trials using antibodies F I G U R E 5 IL-13 induces expression of Cxcl1/CXCL1 that can be blocked by the EGFR inhibitor AG1478 in mTEC cultures. A, Relative Cxcl1 mRNA expression and B, CXCL1 protein expression in Ccsp/Il13 mTECS treated, where indicated, with DOX (to induce IL-13), AG1478 or Dex for 72 h. C, D, Relative Cxcl1/CXCL1 mRNA and protein expression, respectively, in control mTECs treated, where indicated, with DOX, EGF, AG1478 or Dex. Data are expressed as mean + SD. Experiments were performed in duplicate and are from three independent experiments. Statistical analysis was performed using one-way ANOVA with Dunn's multiple comparison. *P < .05, **P < .01, ***P < .001 In addition to the IL13 sub-cluster, we also identified a second sub-cluster within the neutrophilic and eosinophilic severe asthma cluster which was characterized by high expression of AREG, EGF and EREG. Given that previous studies have identified an "IL-17 high" severe asthma phenotype which exhibits expression of genes that are reported as altered in psoriasis lesions, 18 it is significant that transgenic expression of AREG in murine skin causes a psoriasis phenotype characterized by marked epidermal hyperplasia, accompanied by neutrophilia and significantly increased CD4+ T cell infiltration. 68,69 Both AREG and HB-EGF interact with heparan sulphate proteoglycans and with members of the tetraspanin family of membrane-associated proteins which can regulate their distribution, bio-availability and action on target cells and can also serve as cell surface co-receptors, facilitating ligand-receptor interactions. 70 Thus, differential regulation of these two EGFR ligands offers the potential to fine-tune their interaction with key target cells to drive distinctive EGFR/ ERBB mediated responses that, in turn, may give rise to differing sub-phenotypes of severe asthma. It is also noteworthy that the "IL-13 low/IL-17 low" severe asthmatic group (Table 1) was the only group to show a significant up-regulation of ERBB4, whose F I G U R E 6 The EGFR inhibitor AG1478 blocks the pro-neutrophilic responses caused by IL-13 expression in vivo. Ccsp/ Il13 mice were fed DOX for 7 d to induce IL-13 and treated with dexamethasone (light grey bars), AG1478 (dark grey bars), dexamethasone + AG1478 (hatched bars) or vehicle control (white bars). A, Airway resistance in response to increasing concentrations of methacholine (0-10 mg/ mL). B and D, Differential inflammatory cell counts in BALF for neutrophils and eosinophils, respectively. C and E, CXCL1 or CCL11 protein expression, respectively, measured by ELISA in whole-lung lobe lysates; data are shown as mean + SD. N = 6 per group from two independent experiments. Statistical analysis was performed using one-way ANOVA. *P < .05, **P < .01 function appears critically involved in reactions that affect cell fate. 71 Its role within this group of severe asthmatics merits further investigation.
Based on the findings from the U-BIOPRED cohort, further analysis of the IL-13 expressing mouse model revealed that corticosteroid treatment caused significant modulation of the EGFR ligand family including Hbegf and Areg, as well as Erbb3. Unlike other family members, ERBB3 is considered "kinase dead" and requires heterodimerization with another family member such as EGFR or ERBB2 for phosphorylation; otherwise, it is refractory to ligand-induced activation. 72 Phosphorylated ERBB3 has multiple binding sites for phosphatidyl inositol-3 kinase (PI3K) 73 whose pathway is central to the development of BHR and inflammation. 74 Furthermore, activation of PI3Kδ has been implicated in corticosteroid resistance as it causes Akt activation and inactivation of HDAC2, one of the mechanisms associated with corticosteroid resistance. 41 In support of this suggestion, EGFR signalling is associated with increased PI3Kδ/Akt activation in ovalbumin-induced F I G U R E 7 Topological data analysis of gene expression obtained from bronchial brushings from the U-BIOPRED cohort. A TDA network was constructed using gene expression data obtained from bronchial brushings from non-smoking severe asthmatics (n = 61), mild-to-moderate asthmatics (n = 36) and healthy controls (n = 44). As indicated, metadata were then applied for (A) asthma severity, sputum neutrophils or eosinophil counts; (B) EGFR, ERBB2, ERBB3, EGFR ligands or IL13; and (C) CSF3, CXCL2 or CXCL8. Nodes are coloured by intensity from blue (low) to red (high). Arrows point to the regions of interest referred to in the Results sections airways inflammation. 55 Our novel finding that ERBB3 is up-regulated in severe asthma requires further investigation as it suggests that EGFR/ ERBB3 heterodimers may contribute to corticosteroid-refractory asthma. As EGFR and corticosteroids also have beneficial effects on epithelial repair and barrier function, 54 discriminating between pro-inflammatory and pro-repair functions for these pathways, and the role(s) of individual ligands and ERBB receptor heterodimers, should help to tailor more effective therapies for severe asthma. 45 Such studies would offer the potential of exploiting the array of small molecule drugs and antibodies that have been developed for cancer therapy. 71,75 In summary, our study suggests that the prototypic Type-2 mediator, IL-13, can give rise to both eosinophilic Type-2 and neutrophilic "Type-17" stereotypic responses, the latter being corticosteroid insensitive and mediated by epithelial EGFR signalling rather than classical immunological Th17 signalling. Based on our TA B L E 1 Comparison of ERBB receptor and ligand expression in "IL-13 high" and "IL-17 high" asthma clusters relative to health in the U-BIOPRED cohort n "IL-13 low/IL-17 low" "IL-17 high" "IL-13 high" Note: Numbers are log2 fold change relative to expression in healthy participants; significant changes are highlighted in bold. The definition of "IL-13 high" and "IL-17 high" was based on that used in. 18

F I G U R E 8
Erbb receptor and ligand mRNA expression in the IL-13-expressing mouse model. Relative mRNA expression in whole-lung lobe lysates from Ccsp/Il13 mice after induction of IL-13 for 7 d (white bars) (n = 23) vs concurrent dexamethasone (grey bars) treatment for 7 d (n = 6) for A, Egfr, B, Erbb2, C, ErbB3 D, Egf, E, Hbegf and F, Areg. Parametric data are expressed as mean + SD and non-parametric data as box plots showing medians and 25th to 75th percentiles, and whiskers representing minimum and maximum values with all data points shown. Data are from three independent experiments. Statistical analyses were performed using Student's t test or Mann-Whitney test. *P<0.05, ***P < .001 current findings, the IL-13 transgenic mouse model should enable further understanding and dissection of the molecular pathways involved in corticosteroid-refractory pathways, especially those involving EGFR/ERBB signalling in promotion of mixed granulocytic or neutrophilic inflammation and distinguishing them from beneficial pro-repair pathways within the epithelium. This should enhance the development of new treatments that target this sub-phenotype(s) of severe asthma.

ACK N OWLED G EM ENTS
We would like to thank Professor Jeffrey A. Whitsett and Cincinnati Children's Hospital Medical Center for the transgenic IL-13 mice. We want to thank the animal facility at the University of Southampton for husbandry of the mice. We also thank Professor Colin D.
Bingle at the University of Sheffield for teaching JFCK how to ex-

CO N FLI C T O F I NTE R E S T
The authors declare the following conflicts of interest: DED and RD report personal fees from Synairgen, which is outside the submitted work. RD also reports receiving fees for lectures at symposia organized by Novartis, AstraZeneca and TEVA, consultation for TEVA and Novartis as member of advisory boards, and participation in a scientific discussion about asthma organized by GlaxoSmithKline. All are outside the submitted work. All other authors have nothing to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available either in the GEO repository (GSE76226) or can be obtained from the corresponding author upon reasonable request.