HMGB1 amplifies ILC2-induced type-2 inflammation and airway smooth muscle remodelling

Type-2 immunity elicits tissue repair and homeostasis, however dysregulated type-2 responses cause aberrant tissue remodelling, as observed in asthma. Severe respiratory viral infections in infancy predispose to later asthma, however, the processes that mediate tissue damage-induced type-2 inflammation and the origins of airway remodelling remain ill-defined. Here, using a preclinical mouse model of viral bronchiolitis, we find that increased epithelial and mesenchymal high-mobility group box 1 (HMGB1) expression is associated with increased numbers of IL-13-producing type-2 innate lymphoid cell (ILC2s) and the expansion of the airway smooth muscle (ASM) layer. Anti-HMGB1 ablated lung ILC2 numbers and ASM growth in vivo, and inhibited ILC2-mediated ASM cell proliferation in a co-culture model. Furthermore, we identified that HMGB1/RAGE (receptor for advanced glycation endproducts) signalling mediates an ILC2-intrinsic IL-13 auto-amplification loop. In summary, therapeutic targeting of the HMGB1/RAGE signalling axis may act as a novel asthma preventative by dampening ILC2-mediated type-2 inflammation and associated ASM remodelling.


Introduction
Asthma can start at any time in life, although most often begins in early childhood [1]. Wheezy viral bronchiolitis is a major independent risk factor for subsequent asthma [2], yet key knowledge gaps exist in relation to the pathogenic processes that link the two diseases. As the primary site of viral replication and a rich source of type-2 'instructive' mediators (e.g. IL-33), the airway epithelium is now recognised to be pivotal to disease inception through the local activation of type-2 innate lymphoid cells (ILC2) and mucosal dendritic cells. IL-33 is thought to be the main driver of ILC2 expansion and activation, however a growing body of evidence is beginning to implicate high-mobility group box 1 (HMGB1), another nuclear alarmin whose extracellular expression correlates with asthma severity [3][4][5]. In preclinical models of allergic asthma, anti-HMGB1 or genetic ablation of RAGE (receptor for advanced glycation endproducts), one of the receptors ligated by HMGB1, is protective [6][7][8]. Intriguingly, anti-HMGB1 or RAGE deficiency also diminishes IL-33-induced type-2 inflammation [6,9], suggesting that the HMGB1/RAGE axis contributes to a feed-forward circuit that amplifies type-2 inflammation. However, whether this pathway relates to ILC2 activation remains unknown.
One of the teleological roles of type-2 inflammation is to mediate wound repair [10,11]. For example, ILC2s can promote re-epithelialisation of the airway mucosa following viral infection or helminth infestation [12,13]. However, ILC2s can promote aberrant tissue repair, as observed in pulmonary fibrosis and liver fibrosis [14,15]. Curiously, it is not known whether ILC2s contribute to airway smooth muscle (ASM) remodelling, a cardinal feature of asthma and one of the main contributing factors to airway narrowing, loss of lung function, and airways hyperreactivity. Notably, structural alterations to the airway wall are evident in schoolage children with asthma [16,17], and ASM area is increased in wheezy 'pre-asthmatic' children [18], highlighting the need for early intervention strategies to prevent airway remodelling, which is thought to be less plastic in later life [19].
Susceptibility to severe viral infections in infancy and the later development of asthma may stem from an impaired and/or delayed antiviral immune responses [20,21]. Indeed, a deficient interferon (IFN)-lambda response and greater epithelial sloughing to viral infections in children predicts later asthma persistence [22], thus linking impaired immunity to tissue damage. Downstream of toll-like receptor 7 and other virus-sensing pattern recognition receptors, IFN regulatory factor (IRF)7 regulates the antiviral IFN response to respiratory syncytial virus (RSV) and rhinovirus (RV) infection [23][24][25]. In a recent landmark paper comparing asthma risk in Amish (low risk) and Hutterite (4x higher risk) farm children [26], IRF7 was identified as a differentially expressed hub gene. In another clinical study, unbiased network analysis separated exacerbating children into IRF7 hi and IRF7 lo phenotypes, the latter associating with a deficient IFN response and shorter time to exacerbation recurrence [27]. Collectively, these findings suggest that impaired antiviral immunity predisposes to the inception of later asthma, however the pathogenic cellular and molecular processes that underpin this relationship remain ill-defined.
In this study, to determine whether HMGB1 is a contributing factor we used a preclinical model of neonatal viral bronchiolitis. We found that IRF7 deficiency has a profound effect on the airway epithelium: namely increased viral burden, nuclear-to-cytoplasmic translocation of HMGB1, followed by epithelial sloughing into the lumen. The subsequent release of HMGB1 has a profound effect on the number and activation of lung ILC2s, which we show to promote ASM cell proliferation in an ILC2-ASM co-culture system. Although both ILC2s and ASM cells produce HMGB1, we show that ILC2-derived HMGB1 acts in an autocrine manner to amplify type-2 cytokine production. Together, our findings uncover an important role for HMGB1 in ILC2 activation and implicate ILC2s as important effectors of both type-2 inflammation and ASM thickening in early-life, predisposing to later asthma.

Severe viral bronchiolitis induces epithelial damage and alarmin release
To assess the effect of impaired antiviral immunity on the development of severe viral bronchiolitis, we inoculated WT ( +/+ ) and IRF7-deficient ( -/-) neonatal (7 day old) mice with pneumonia virus of mice (PVM). As PVM is a mouse-specific pathogen (same genus as respiratory syncytial virus; RSV) it replicates efficiently in mice (unlike hRSV) and allows for the temporal study of host-pathogen interactions [28,29]. Compared to WT neonatal mice, viral load was massively elevated and clearance delayed in IRF7 -/neonatal mice ( Fig 1A). This was associated with diminished production of antiviral cytokines, IFN-α ( Fig 1B) and IL-12p40 (S1A Fig). As occurs clinically, the severe bronchiolitis that developed in IRF7 -/neonatal mice was associated with greater neutrophilic inflammation, tissue oedema, and pronounced sloughing of the airway epithelium (Fig 1C-1E). This epithelial damage was associated with elevated IL-33 and HMGB1 levels in the bronchoalveolar lavage fluid (BALF) and lungs of IRF7 -/neonatal mice (Fig 1F and 1G and S1B Fig). In contrast, other known inducers of type-2 inflammation, such as IL-25 and TSLP, were below the limit of detection. Mirroring the pattern of viral load in the epithelium, nuclear to cytoplasmic translocation of HMGB1 peaked at 7 dpi and remained elevated at 10 dpi ( Fig 1H). Although HMGB1 levels in BALF persisted at 14 dpi, epithelial cell cytoplasmic HMGB1 had waned, suggesting that extracellular HMGB1 was derived from sloughed epithelial cells or another cellular source at this time. The functional activity of HMGB1 is affected by its oxidative state; all-thiol HMGB1 acts as a chemokine, whereas disulphide HMGB1 acts as a cytokine through the activation of MD-2/TLR4 [30]. Using Western blot to interrogate the isoforms, we observed that both all-thiol and disulphide HMGB1 were significantly elevated in the lung homogenate of infected IRF7 -/mice compared to their WT counterparts ( Fig 1I). Thus, impaired antiviral immunity predisposed to airway epithelial damage, IL-33 release, and a sustained period of elevated HMGB1 expression.

PLOS PATHOGENS
HMGB1 amplifies airway remodelling strains, ILC2s were largely negative for AmCyan (reporting for IL-4; Fig 2C). Changes to the size of ASM area can occur in the airway wall of wheezy, pre-asthmatic infants [18]. We found an increase in ASM area and proliferation in IRF7 -/compared to WT infected mice ( Fig 2E  and S1F Fig), suggesting that ASM alterations are initiated in response to a severe respiratory infection in early-life. TGF-β is a potent ASM mitogen [32]. We found that active TGF-β levels in BALF were significantly higher in IRF7 -/mice compared to WT mice at both 7 and 10 dpi (Fig 2F). Probing the lung tissue for TGF-β expression by immunohistochemistry revealed the airway epithelium and ASM (clearly evident by morphology) to be a rich source of TGF-β following infection, and that its expression in ASM cells was significantly greater in IRF7 -/mice compared to WT mice ( Fig 2G). Together, these data demonstrate that, in a susceptible host, a respiratory viral infection can lead to the induction of type-2 inflammation and ASM remodelling in early life.

Secondary Pneumovirus infection or cockroach allergen exposure induces an asthma-like phenotype in IRF7-deficient mice
Frequent viral infections are associated with progression to asthma [33], which we have previously modelled in TLR7 -/mice and pDC-depleted mice [28,34]. Here, we identified that reinfection of IRF7 -/mice (at 7 weeks of age) induces an increase in ASM area (Fig 3A), and that this phenotype persists for at least 8 weeks (S2A Fig). Viral challenge also increased collagen deposition in the airway mucosa (S2B Fig) and increased airway hyperreactivity in the IRF7 -/mice ( Fig 3B). Although this response was maximal at 7 days, it remained significantly elevated at 21 days post viral challenge (S2C Fig), presumably as a consequence of changes to the airway wall architecture. These responses were not associated with viral load, since the virus was not detectable by immunohistochemistry, even at 1 dpi. However, as with the primary infection, the induction of airway remodelling coincided with increased type-2 inflammation (IL-13 production, airway eosinophilia, mucus hypersecretion), which waned by 21 dpi (Fig 3C-3E). Again, the type-2 inflammation was preceded by an increase in IL-33 and HMGB1 expression in BALF of the IRF7 -/mice ( Fig 3F). Additionally, we identified that IRF7 -/mice that had experienced severe bronchiolitis in infancy were predisposed to develop experimental allergic asthma in later life following exposure to low-dose cockroach allergen extract (CRE; S2D and S2E Fig). In contrast, PVM/CRE co-exposed WT mice did not develop asthma-like pathologies (S2E Fig). Importantly, in the absence of the viral infection in earlylife, exposure to CRE alone did not induce experimental asthma in IRF7 -/mice.

Neutralisation of HMGB1 or IL-33 diminishes type-2 inflammation and ASM growth in IRF7 -/mice
IL-33 has been implicated in ILC2 proliferation and airway remodelling; however, the contribution of HMGB1 to these processes remains an open question. To investigate this, we treated IRF7 -/mice with anti-HMGB1 (clone 2G7), which neutralises both all-thiol and disulphide HMGB1, or an isotype-matched control (S3A Fig), then killed the mice at 10 dpi when the pathologies associated with asthma onset were most prominent. For comparative purposes, we treated a separate group of mice with anti-IL-33. Anti-HMGB1 decreased HMGB1 but not IL-33 levels in BALF, whereas blockade of IL-33 lowered both HMGB1 and IL-33 ( Fig 4A and  4B), suggesting that IL-33 operates upstream of HMGB1. Anti-HMGB1 and anti-IL-33 treatment of IRF7 -/mice decreased ILC2 numbers in the lung, type-2 cytokine responses in BALF, and airway eosinophilia (Fig 4C and 4D and S3B Fig). Moreover, both treatments arrested ASM growth (Fig 4E), an effect that was associated with lower TGF-β expression in the ASM layer ( Fig 4F). Of note, treatment of IRF7 -/mice with glycyrrhizin, a putative HMGB1 antagonist, reproduced the findings observed with anti-HMGB1 (S3C- S3I Fig). Collectively, these data indicated that in a host with impaired with antiviral immunity (i.e. IRF7 -/mice), IL-33 and HMGB1 contribute to increased ILC2 numbers, type-2 cytokine production and the development of ASM remodelling in response to an early-life viral infection.

Activated ILC2s from IRF7 -/mice induce ASM proliferation in vitro in an HMGB1-dependent manner
ILC2s contribute to tissue remodelling by promoting fibrosis, collagen deposition, and epithelial cell regeneration [12][13][14][15], however, they have yet to be implicated in the expansion of ASM area, a cardinal feature of asthma. To confirm a role for ILC2s, we next developed an in vitro co-culture assay to test whether ILC2s directly induce ASM cell proliferation. Primary tracheal ASM cells were isolated from neonatal WT mice, loaded with a proliferation dye and cultured in an FCS-low (0.25%) media prior to stimulation. Testing the proliferative effects of IL-33 and IL-13 (relative to 10% FCS) revealed that IL-33 is a poor mitogen (S4A Fig), whereas IL-13 is a potent mitogen; with an EC 50 below the level of IL-13 detected in BALF (Fig 2A). To
The requirement for TLR4 inferred a role for disulphide-HMGB1 [30,36]. Consistent with this theory, stimulation of ASM cells with disulphide HMGB1, but not the non-oxidisable mutant all-thiol HMGB1 [30], induced their proliferation ( Fig 5B). As IL-13 but not IL-33 stimulated ASM cells produced HMGB1 (Fig 5C), and because IL-13-induced ASM cell proliferation was attenuated by TLR4 or RAGE antagonism (S4E Fig), this implicated a key role for ASM cell-derived HMGB1. In light of these finding, we re-examined the lung sections probed with anti-HMGB1. Whereas vehicle-inoculated mice did not express HMGB1 in the ASM, its expression increased incrementally over the course of infection in the ASM bundles of IRF7 -/-

PLOS PATHOGENS
HMGB1 amplifies airway remodelling but not WT mice (Fig 5D), similar to the pattern of ASM growth (Fig 2E). In summary, our findings suggest that in a genetically susceptible host (i.e. IRF7 -/mice in this context), ILC2produced IL-13 induces the release of HMGB1, leading to the activation of TLR4 and/or RAGE signalling, which in turn, increases TGF-β production, ASM proliferation, and amplifies the production of IL-13.

ILC2-derived HMGB1 in IRF7 -/mice acts in an autocrine manner to amplify IL-13 production
Demonstrating that ASM cells release HMGB1 did not rule out the possibility that ILC2-derived HMGB1 might contribute to ASM proliferation. Following IL-2/IL-33 stimulation of ILC2s purified from IRF7 -/mice, HMGB1 was elevated in the cell cytoplasm and detected in the supernatant (Fig 6A [left]), and this release was significantly diminished upon IL-13 neutralisation ( Fig 6A [right]). Together with our findings from the ILC2-ASM cell co-culture system (Fig 5B), where anti-HMGB1 ablated IL-13 levels, these data suggested a feedforward loop whereby IL-13-induced HMGB1 amplifies the production of IL-13. Consistent with this theory, IL-2/IL-33-stimulated ILC2s produced IL-13 in a biphasic manner, with the second phase occurring after the peak of HMGB1 release ( Fig 6B). As we had previously shown that RAGE -/mice generate attenuated type-2 inflammatory responses to both viral-or allergen-triggered experimental asthma [6,37], we speculated that HMGB1/RAGE ligation promotes type-2 cytokine production by ILC2s. To determine whether ILC2s express RAGE, we used RAGE -/mice in which the functional Ager gene has been replaced by GFP. Compared to non-stained WT ILC2s (grey), ILC2s derived from RAGE -/mice (black) were GFP bright, indicating that ILC2s express RAGE at steady state (Fig 6C). To assess the role of HMGB1 directly, we stimulated lung ILC2s in culture with disulphide-HMGB1 or non-ox-HMGB1 in the presence of IL-2 for 6 days. Non-Ox-HMGB1 but not disulphide-HMGB1 induced an increase in IL-5 and IL-9 production. IL-13 production was numerically higher but this effect was not significant (Fig 6D). To determine whether RAGE or HMGB1 contribute to IL-33-induced type-2 cytokine expression by ILC2s, we stimulated ILC2s (isolated from the lungs of infected 4C13R IRF7 -/mice) with IL-2/IL-33 for 4 days in the presence of anti-HMGB1, LPS-RS (a TLR4 antagonist) or FPS-ZM1 (a RAGE antagonist) [38]. The increased intensity of the dsRed signal (IL-13 expression) following IL-2/IL-33 stimulation was diminished by anti-HMGB1 or RAGE antagonism, but not TLR4 antagonism (Fig 6E), implicating an ILC2 intrinsic HMGB1/RAGE signalling pathway that amplifies IL-13 production. Alterations in ILC2 cell death or proliferation amongst the treatment groups did not account for this effect (S4F Fig). In addition to attenuating IL-13 expression, anti-HMGB1 and RAGE antagonism ablated the production of IL-5 and IL-9 (Fig 6F) in the culture supernatant, suggesting that HMGB1-mediated activation of RAGE contributes to the regulation of ILC2 type-2 cytokine production.
Finally, to confirm that RAGE contributes to the development of type-2 inflammation in PVM-infected IRF7 -/mice, we administered the RAGE antagonist, FPS-ZM1. Consistent with the lack of effect of RAGE deficiency on ILC2 proliferation (S4F Fig), lung ILC2 numbers were unaffected. However, RAGE antagonism significantly decreased IL-5 levels in BALF, diminished airway eosinophilia, and critically, prevented aberrant ASM growth in infancy (Fig 6G-6J).

Discussion
Severe viral bronchiolitis is an independent risk factor for asthma. To elucidate novel pathogenic processes that link the two diseases, we simulated a clinically relevant gene-environment interaction by infecting IRF7 -/mice with PVM. In the absence of IRF7, viral burden increased PLOS PATHOGENS HMGB1 amplifies airway remodelling

PLOS PATHOGENS
HMGB1 amplifies airway remodelling substantially in the airway epithelium, leading to epithelial sloughing (i.e. cell death), HMGB1 release, expansion of IL-13-producing ILC2s, and ASM growth. Anti-HMGB1 decreased ILC2 numbers and IL-13 production, and attenuated ASM growth. Similarly, anti-HMGB1 attenuated ILC2-induced ASM proliferation in a co-culture model. In addition to the airway epithelium and underlying smooth muscle, we identified that ILC2 also release HMGB1 and that this contributes to an IL-13-induced auto-amplification loop via the activation of RAGE. Thus, our data suggest that by dampening type-2 inflammation and attenuating virus-associated ASM remodelling, biologics or small molecules targeting HMGB1 and/or RAGE will serve as novel preventatives for the treatment of asthma.
Several lines of evidence support the concept that impaired antiviral immunity in early-life is associated with the later development of asthma. To identify novel pathogenic mechanisms that occur consequent to defective antiviral host defence, we inoculated IRF7 -/neonatal mice and followed host-pathogen interactions in a temporal manner. In IRF7 -/mice, the peak of infection (7 dpi) was associated with increased nuclear-to-cytoplasmic HMGB1 translocation in the epithelium, increased IL-33 levels in the airway lumen, and the infiltration of neutrophils, which can process IL-33 to a more biologically active form [39]. This response was followed (10 dpi) by increased epithelial sloughing, elevated HMGB1 in the airway, ILC2 expansion, and growth of the underlying ASM layer. Of the lung ILC2s, approximately half were producing IL-13, presumably as a consequence of the cytokine milieu, i.e. high alarmin (IL-33/HMGB1) concentration in the absence of counter regulatory signals from type I IFNs or IL-12 [34,[40][41][42]. HMGB1 neutralisation attenuated the type-2 inflammatory response and arrested ASM growth, suggesting a role for HMGB1 and ILC2 in the development of ASM remodelling.

PLOS PATHOGENS
HMGB1 amplifies airway remodelling IL-33-stimulated ILC2s (in the absence of ASM cells) produce HMGB1 downstream of IL-13, and that the neutralisation of HMGB1 or blockade of RAGE, but not TLR4, attenuates the production of IL-13 by ILC2 (as assessed by dsRed intensity and IL-13 protein production). Similarly, antagonism of the HMGB1/RAGE axis ablated IL-5 and IL-9 production, thus identifying a novel autocrine role for HMGB1 in the production of type-2 cytokines by ILC2s. We also demonstrated that non-oxidisable HMGB1 induces type-2 cytokine production by ILC2s. A recent paper also implicated a role for RAGE downstream of IL-13 (and IL-4, IL-5), although the nature of the RAGE ligand remained unclear, with HMGB1 not tested exhaustively [45]. Consistent with a role for HMGB1/RAGE in mediating type-2 inflammation, we found that RAGE antagonism dampened type-2 inflammation and ASM remodelling in infected IRF7 -/mice. However, RAGE antagonism was not as effective as anti-HMGB1, suggesting that HMGB1 exerts RAGE-independent pathways. Indeed, it was also noteworthy that TLR4 antagonism diminished IL-13 production in the ILC2-ASM co-cultures, but had no effect on IL-13 production when the ILC2s were cultured alone, suggesting that the activation of TLR4 on ASM cells induces a factor which feeds back to the ILC2s. One possibility is TGFβ, which supports lung ILC2 [46]. If such bidirectional crosstalk between inflammatory cells and structural cells sustains the inflammatory response and underlies disease chronicity, then breaking these circuits offers an opportunity for new and effective therapeutic interventions.
Asthma is a polygenic disorder, underpinned by gene-environment interactions. As individual SNP effects on disease risk are small, we elected to use knockout mice, which we have used previously to identify novel pathogenic mechanisms [34,37,47]. We believe this approach is preferable to using an extremely high inoculum in WT mice that can induce non-physiological responses and increase the potential for false positives. However, by using IRF7 -/mice, we do not presume that our findings are relevant to all cases of viral bronchiolitis in infancy; indeed we agree with view that there are different endotypes of bronchiolitis [48], as is now accepted to be the case for asthma. On the other hand, we do not take the view that our findings are only amenable to individuals with a genetic variant affecting IRF7, not least since exacerbating children can be subdivided into IRF7 hi and IRF7 lo phenotypes based on gene expression [27]. In the in vitro ILC2: ASM cell co-culture model, we used exogenous IL-33 rather than encapsulate the lung cytokine microenvironment, and thus, there is no merit in exploring whether IRF7-sufficient ILC2s would promote ASM proliferation (i.e. the phenotype is IL-13 dependent, so an IL-33-stimulated IRF7-sufficient ILC2s will self-evidently induce ASM proliferation).
Taken together with our previous findings, and reports from other groups, it is increasingly apparent that HMGB1 is an important mediator in the pathogenesis of viral bronchiolitis and asthma [4][5][6]8,9,44]. Intriguingly, PVM infection of RAGE -/mice induces viral bronchiolitis characterised by impaired antiviral immunity and ASM remodelling but not type-2 inflammation [37]. Notably, HMGB1 neutralisation or the backcrossing of RAGE -/mice with TLR4 -/mice protects against the development of ASM remodelling, implicating disulphide-HMGB1/ TLR4 activation in ASM remodelling. Our findings here support a role for the same ligandreceptor pairing in mediating ILC2-induced ASM growth. In contrast, HMGB1/RAGE activation appears more critical to the amplification of type-2 inflammation.
Collectively, our findings provide a mechanistic basis to explain the effectiveness of anti-HMGB1 treatment or RAGE deficiency in acute and chronic models of allergen-and viralinduced experimental asthma. HMGB1/RAGE contributed to ILC2-mediated type-2 cytokine production and type-2 inflammation, and ILC2-produced IL-13 activates a HMGB1/TLR4/ RAGE signalling pathway in ASM cells leading to increased TGF-β expression and ASM remodelling.

Ethics statement
The ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines in the EQUA-TOR (Enhancing the Quality and Transparency of Health Research) Network library were followed for this report. All animal studies were performed in accordance with the Animal Care and Ethics Committees of the University of Queensland Animal Ethics Committee (AEC209/ 13) and QIMR Berghofer Medical Research Institute Animal Ethics Committee (P2319 A1706-609M) and are consistent with guidelines provided by the Australian National Health and Medical Research Council.

Sample extraction and processing
Following euthanasia, BAL was performed and lung lobes excised and processed as follows: the left lung lobe was digested immediately for flow cytometry, the superior right lobe (for histological analysis) was fixed in formalin neutral buffer overnight before storage in 70% ethanol, the post-caval and inferior lobes (for protein analysis by ELISA) were pooled and snap-frozen, and the inferior right lobe (for quantification of mRNA expression) was snap frozen. All snap frozen lungs were stored at -80˚C.

Flow cytometry
Single cell suspensions of lung cells were generated as described [28]. Following Fc block, cells were incubated with various antibodies to allow for immunophenotyping. These

Histology and Immunohistochemistry
Five micron thick lung tissue sections were stained with Periodic acid-Schiff or picrosirius red to enumerate mucous-secreting cells and collagen respectively. Immunohistochemistry for PVM G protein (gift from U. Buchholz, National Institutes of Health, Bethesda, MD), HMGB1 (polyclonal; Abcam Cat#ab18256, RRID:AB_444360), and Ly6G + neutrophils (1A8; BD Biosciences Cat# 551459, RRID:AB_394206) and quantification of immunoreactivity was performed as described previously [6,28,34,51]. ASM was detected using α-smooth muscle actin (1A4; Sigma-Aldrich Cat# A5691, RRID:AB_476746) monoclonal antibody and ASM area around the small airways (defined as a circumference <500 μm for neonates and <800 μm for mice >7 weeks old) was measured using Scanscope XT software and expressed as area per micrometer of basement membrane. AECs sloughing was measured using Scanscope XT software and expressed as percentage of epithelial denudation of airway basement membrane [34]. Oedema was assessed by point counting of fluid-filled airspaces [42]. To detect TGF-β1, tissue sections were pre-treated with 10% normal goat serum then incubated with anti-TGF-β1 (polyclonal; Abcam Cat# ab92486, RRID:AB_10562492) overnight, followed by incubation with anti-rabbit IgG-alkaline phosphatase (Sigma-Aldrich) for 60 min. Immunoreactivity was developed with Fast Red (Sigma-Aldrich), followed by counterstaining with Mayer's hematoxylin. TGF-β1 expression in the ASM is measured by the fraction of the airway perimeter that has an adjacent TGF-b immunoreactive ASM cell. For each mouse, 5 airways were measured, and the mean value used in the graph. To identify proliferating ASM cells, lung sections were sequentially incubated with mouse anti-α-SM actin ((1A4; Sigma-Aldrich Cat# A5691, RRID:AB_476746), goat anti-mouse AF488 (Thermofisher), anti-PCNA-biotin (PC10; Thermo Fisher Scientific Cat# 13-3940, RRID:AB_2533017), and streptavidin-AF647 (Thermofisher), then counterstained with DAPI (Sigma-Aldrich). To enumerate airway eosinophils, a cytospin of BAL cells was generated then incubated with May-Grünwald Giemsa stain. To visualise airway epithelial basement membrane thickening, sections were incubated with picrosirius red, and the stained area quantified using ImageJ (NIH). Mucus was stained with periodic acid-schiff staining and scored as the percentage of mucus-secreting AECs. Researchers who performed histological analysis and scoring were blinded to the sample identity.

Western Blot
Total protein in lung homogenates was determined using the Pierce BCA protein assay kit (Thermo Fisher Scientific). Equivalent amounts of protein were loaded and separated on 4-20% Mini-PROTEAN TGX Precast Protein gels (Bio-Rad Laboratories) followed by transfer on to a PVDF membrane (Bio-Rad Laboratories). After blocking with 5% skim milk in 0.1% Tween 20/TBS, the membranes were probed with anti-HMGB1 (1:500, rabbit polyclonal; Abcam Cat#ab18256, RRID:AB_444360) overnight at 4˚C, washed with TBS-T, then incubated for 1 hr with IRDye 680RD donkey anti-rabbit IgG H+L (1:10,000, Millenium Science/Li-Cor). Semi-quantification of the bands was performed with a LI-COR Odyssey scanner and software (LI-COR Biosciences).