Identification of a novel HIF-1α-αMβ2 Integrin-NETosis axis in fibrotic interstitial lung disease

Neutrophilic inflammation correlates with mortality in fibrotic interstitial lung disease (ILD) however, the underlying mechanisms remain unclear. We aimed to determine whether aberrant neutrophil activation is a feature of ILD and the relative role of hypoxia. We used lung biopsies and bronchoalveolar lavage (BAL) from ILD patients to investigate the extent of hypoxia and neutrophil activation in lungs of patients with ILD. We complemented these findings with ex vivo functional studies of neutrophils from healthy volunteers to determine the effects of hypoxia. We demonstrate for the first time HIF-1α staining in neutrophils and endothelial cells in ILD lung biopsies. Hypoxia enhanced both spontaneous and phorbol 12-myristate 13-acetate (PMA)-induced neutrophil extracellular trap (NET) release (NETosis), neutrophil adhesion, and trans-endothelial migration. Hypoxia also increased neutrophil expression of the αM and αX integrin subunits. Interestingly, NETosis was induced by αMβ2 integrin activation and prevented by cation chelation. Finally, NETs were demonstrated in the BAL from ILD patients, and quantification showed significantly increased cell-free DNA content and MPO-citrullinated histone H3 complexes in ILD patients compared to non-ILD controls. Our work indicates that HIF-1α upregulation may augment neutrophil recruitment and activation within the lung interstitium through activation of β2 integrins. Our results identify a novel HIF-1α-Integrin-NETosis axis for future exploration in therapeutic approaches to fibrotic ILD.


Introduction
The interstitial lung diseases (ILD) are a group of diffuse parenchymal lung disorders that can result in pulmonary fibrosis (PF) 1 . Despite recent advances in diagnostics and therapeutics, ILD is still associated with substantial morbidity and mortality 2 . Neutrophil activation may play a role in ILD and in particular in the most severe fibrotic form, idiopathic PF (IPF). The pathogenesis of IPF is unknown but is thought to involve an aberrant response to repetitive epithelial injury, with associations to genes and proteins linked to epithelial function, integrity and repair. Progressive epithelial damage, and aberrant wound repair leads to extensive scar formation and correlates, clinically, with worsening hypoxia. Increasing desaturation during exercise 3 or sleep 4 is a significant predictor of mortality. Further evidence from animal models suggests that hypoxia may actually contribute to a vicious cycle of disease progression 5 . This has led to the view that hypoxia itself may contribute to worsening of PF but how this occurs is not known.
Hypoxia, a state in which oxygen supply is inadequate for tissue demands, modulates gene expression via transcriptions factors called hypoxia inducible factors (HIF). There are 3 members of the HIF family, HIF-1α, HIF-2α and HIF-3α, which bind conserved DNA sequences or Hypoxia Response Elements (HRE). Although it seems plausible that the IPF lung is hypoxic much of the evidence is indirect. Levels of lactic acid, a metabolite generated in response to hypoxia, are high in IPF lung tissue supporting the concept of a hypoxic microenvironment 6 and HIF-1a and 2a have been shown, ex vivo, to be expressed in lung biopsies from patients with IPF, in some but not all reports 7,8 . Additional genomic studies in IPF patients show up-regulation of hypoxia-related gene signatures, including TGF-β 9 , the key fibrotic cytokine in PF, and of the HIF-1α pathways 8,10 .
The contribution of neutrophils to ILD has also been relatively less studied compared to other inflammatory and fibrotic diseases. Early studies began to explore the potential role of neutrophils in IPF [11][12][13][14] , however research focus has since shifted to other cell types. The number of neutrophils in the bronchoalveolar lavage (BAL) has been shown to predict both disease severity in IPF 15 and the development of PF in patients with hypersensitivity pneumonitis 16 . In addition, neutrophil extracellular traps (NETs) have been shown to indirectly drive PF by stimulation of collagen production from fibroblasts in vitro 17 , and NET release (NETosis) has been associated with PF in older mice in vivo 18 with loss of peptidyl arginine deiminase (PAD)-4, a key neutrophil enzyme for NETosis, being protective 19 . Neutrophils are also associated with disease severity in acute lung injury and acute respiratory distress syndrome (ARDS) 20,21 however, their precise contribution remains uncertain 22 . Neutrophil depletion can ameliorate disease features in mouse models of ARDS 23 and a reduction in neutrophil infiltration 24 , or knock-down of neutrophil elastase (NE) attenuates fibrosis in bleomycin-induced mouse models of PF 25 . Taken together, these studies implicate a contributory role of neutrophils to fibrotic ILD.
Neutrophil survival is a tightly regulated process. Prolonged survival can delay resolution of inflammation and can cause damage to surrounding cells and tissues; however, if apoptosis is premature, antimicrobial function can be compromised 26 . Hypoxia drives neutrophil survival via HIF-1α-dependent NF-κB activation 27 . In addition, HIF-2α has also been shown to be important in regulating neutrophil function 28 . Few reports address the effects of hypoxia upon NETosis.
Inhibition of HIF-1α can reduce NET release 29 , whilst pharmacological stabilisation of HIF-1α increases phagocyte bactericidal activity 30 , implicating a role for down-stream targets of HIF-1α in leukocyte function.
Given the importance of hypoxia and HIF signalling in neutrophil function and the emerging role of neutrophils as key drivers of pulmonary disease, we sought evidence for hypoxia and NETosis in the lungs of patients with ILD and the functional effects of low oxygen levels upon ex vivo neutrophil function and activation.

Patient demographics
BAL was obtained from 11 ILD patients and 7 non-ILD controls undergoing diagnostic bronchoscopies. Demographics, clinical history and treatments at the time of sample collection are listed in Table 1. Within the ILD cohort: 4 (36%) had IPF, 3 (27%) had nonspecific interstitial pneumonia, 3 (27%) had chronic hypersensitivity pneumonitis (HP) and 1 (10%) had unclassifiable ILD. Our non-ILD control group underwent diagnostic bronchoscopy due to: 5 (71%) investigation of haemoptysis, 1 (14.5%) right middle lobe collapse and 1 (14.5%) previous tracheal schwannoma patients undergoing yearly bronchial surveillance. Only the ILD group had lung function tests, as part of standard patient care. None of the patients recruited were taking anti-fibrotic drugs at the time of bronchoscopy. The differential cell population obtained from BAL from patients are listed in supplementary table 1.

Neutrophils and endothelial cells stain positive for HIF-1α in the ILD lung
Given reports of localised hypoxia in pulmonary disease 31

Hypoxic exposure does not affect hydrogen peroxide generation but promotes NETosis
Pharmacological HIF-1α stabilisation has been reported to enhance bacterial killing and NETosis 29,30 , however, these studies were performed using atmospheric oxygen levels. We therefore assessed for any alteration in function, described below, of healthy neutrophils under normoxia (21% oxygen) and hypoxia (1% oxygen). First, we verified hypoxia by examining neutrophil cell lysates for the presence of HIF-1α and HIF-2α at various time points. We

Neutrophil adhesion and trans-endothelial migration are enhanced under hypoxia
Having found an effect on NETosis, we next examined integrin activation and neutrophil adhesion, which are also implicated in NET release 34,35 . We measured neutrophil adhesion to primary human endothelial cells in the absence or presence of PMA (a general integrin activator) or lipopolysaccharide (LPS) (to mimic infectious stimuli), stimuli that activate NETosis via distinct pathways 36 . Hypoxia increased both unstimulated (23.6±4.0% vs 2.7±1.6%, p<0.05) and LPS-stimulated (35.7±4.8% vs 11.3±1.4%, p<0.05) adhesion to resting endothelium, whilst PMA-stimulated adhesion, which was already high, was unaffected ( fig. 3A-C). We then looked at adhesion to endothelium pretreated with TNF-α, to mimic an inflammatory event. Whilst unstimulated neutrophil adhesion to TNF-α activated endothelial cells was not altered by

NETosis is induced by α M b 2 integrin activation
Given reports of reduced NETosis following integrin inhibition and our data showing increased α M b 2 and, to a lesser extent, α X b 2 integrin expression, we tested whether integrin-

BAL from patients with ILD have more NETs than control BAL
Having initially noted evidence of cell-specific hypoxia in ILD lung biopsies and demonstrating that hypoxia enhances neutrophil adhesion, transmigration and NETosis, we hypothesized that aberrant neutrophil activation may also be a feature of the ILD lung. To test this hypothesis, we generated slides with BAL and stained with DAPI to identify cellular DNA.
Interestingly, whilst control BAL neutrophils displayed punctate DAPI staining ( fig. 6A), we observed evidence of NETosis in ILD-BAL neutrophils ( fig. 6B, C). We then obtained BAL from 11 ILD patients (ILD-BAL) and 7 non-ILD controls (control BAL) and quantified levels of cell-free DNA. We found ILD-BAL had 5.5-fold greater cell-free DNA content compared to control BAL (p<0.01) ( fig. 6D). Cell-free DNA content positively correlated with neutrophil counts (% of total cells) isolated from ILD-BAL (p=0.0075) ( fig. 6E), but not in control BAL ( fig. 6F). To verify that these were NETs, we also examined BAL for the presence of MPO-citrullinated histone H3 complexes. Similar to total cell-free DNA, we observed significantly greater values in ILD-BAL, indicating greater levels of NETosis ( fig. 6G).
The precise mechanism neutrophils contribute to ILD pathogenesis, however, is less characterised. Early work from the 1980s began to explore whether neutrophils might contribute to IPF pathology [11][12][13][14] , however this avenue of research lost momentum. Since then, reports have associated increased neutrophil migration and activation with severe pulmonary disease both in animal models 23,53 and man 16,20,21 . We report, for the first time, that neutrophils and endothelial cells in ILD lung biopsies display HIF-1α expression and provide evidence for NETosis in the ILD lung. Given the profound effects hypoxia exerts upon neutrophil survival and function 27,28 , these findings led us to investigate whether hypoxia affects neutrophil extravasation and activation, thus contributing to ILD pathology.
Integrins are adhesive molecules that enable leukocytes to interact with their external environment. Similar to a previous report 54 , we found increased β 2 integrin expression in neutrophils under hypoxia, but specifically found increased α M and, to a lesser extent, α X integrin subunit expression. Interestingly, the α M β 2 and α X β 2 integrins also function as complement receptors, which may be relevant to ILD pathology given that increased levels of complement C3a and C5a and roles for their receptors have been reported in IPF 55,56 . Moreover, studies using the bleomycin-induced mouse model of IPF highlight roles for both C3 and C5 in pulmonary fibrosis 57,58 . Upregulation of β 1 integrins has been described under hypoxia 59 , however, there are no reports assessing expression in neutrophils. A lack of effect may be explained by the relatively low β 1 integrin expression in human neutrophils. Taken together, the evidence indicates that neutrophils predominately engage via β 2 integrins, a mechanism which is enhanced under hypoxia.
Early studies demonstrated that hypoxia enhances neutrophil adhesion to endothelial cells 60 , epithelial cells 61 , and trans-epithelial migration 62 . In support of these findings, our results show altered function of healthy neutrophils with increased neutrophil adhesion and transendothelial migration under hypoxia. In addition, we report that hypoxia enhances NETosis.
Given that the gold standard markers or methods of detection for NETosis have not been We observed tissue-specific HIF-1α expression in ILD lung tissue, mainly restricted to pulmonary endothelial cells and neutrophils, with only minimal upregulation in areas of epithelium and fibrosis, which may hold pathological relevance. Previously, markers of hypoxia have been variably reported in the epithelium of patients with IPF. Several authors have found HIF-2α and CA-IX within the IPF fibrotic reticulum and HIF-1α in the overlying epithelium with IHC 7 , (albeit sometimes in a single patient 8 ). HIF-1α is more readily found in the mouse bleomycin model of PF raising the question of differences between the two species and insults 63 .
Whilst epithelium-specific HIF-1α deletion has no effect upon radiation-induced enteritis, mice with endothelium-specific HIF-1α deletions present with reduced intestinal damage 64 . HIF-1α is known to contribute to the pathology of pulmonary hypertension 65,66 , with some work specifically interrogating endothelial HIF signalling 67 . Neutrophilic inflammation has also been associated with pulmonary hypertension 68 , and believed to drive angiogenesis via NETosis 69 .
Therefore, the pulmonary pathology in ILD patients may in part be attributed to endothelial and neutrophil HIF-1α expression enhancing neutrophil recruitment and NETosis within the lung. In this paradigm, enhanced NETosis would initiate angiogenic signals and drive lung pathology. Interestingly, the model of neo-angiogenesis underlying ILD pathology has attracted interest, and the powerful angiogenic inhibitor, nintedanib, shown to have therapeutic benefits in a range of fibrotic ILDs 70 .
Our findings of hypoxia driving NETosis compliments the increasing evidence that NETs may play a role in many chronic lung diseases 71 , including ILD by stimulation of fibroblasts 17 . In PF, we propose that elevated NETosis may cause epithelial cell damage, dysfunction and death, drive innate and adaptive immune cells activation, and promote a pro-fibrotic environment that ultimately facilitates the progression of pulmonary fibrosis.
In conclusion, we report that the ILD lung contains molecular features of hypoxia, mainly localized to neutrophil and endothelial cells, which may contribute to disease pathology.
Hypoxia enhanced neutrophil β 2 integrin expression, which translated to augmented adhesion and migration across endothelial cells, and NETosis. Our findings are further supported by ex vivo demonstration of NETs within the human fibrotic lung. Taken together, our work begins to elucidate a potential role of hypoxia in driving neutrophil recruitment and activation within the airspace to promote a pro-fibrotic environment. These findings offer a rationale for future translational medicine exploration of a novel HIF-1α-Integrin-NETosis axis as a potential therapeutic target in fibrotic ILD.

Bronchoalveolar lavage
Fibre-optic bronchoscopy with BAL was performed in line with the American Thoracic Society guidelines 72 . BAL fluid was frozen for later NET analysis. None of the patients undergoing bronchoscopy had any infections at the time of procedure.

Immunohistochemistry (IHC)
Lung biopsy specimens were collected as part of routine clinical care. Ethical approval was given by the UK National Research Ethics Committee (13/LO/0900). IHC was performed using the automated Bond-Max system (Leica Biosystems Ltd., Newcastle) with 4µm FFPE sections.

Neutrophil isolation
Neutrophils were isolated as previously described 73 . In brief, neutrophils were isolated by Percoll density centrifugation from sodium citrate anticoagulated blood obtained by informed consent from healthy volunteers. Neutrophils were diluted to 2x10 6 neutrophils/ml in phenol-free RPMI (Thermo Scientific, UK) supplemented with 10% FBS (Thermo Scientific, UK) and 2mM Lgluatamine (Lonza, UK).
Hydrogen peroxide generation H 2 O 2 generation was measured as previously described 73 . Briefly, neutrophils were cultured under normoxia or hypoxia for 1 hour before addition of HRP (Sigma, UK) and Amplex® UltraRed (Invitrogen, UK). H 2 O 2 generation in response to PMA was recorded using a FLUOstar Omega microplate reader (BMG Labetech, Germany) and rates (expressed in nM/sec) determined using Omega Mars Analysis software (BMG Labtech, Germany).

NETosis quantification
NETs was quantified using the Quanti-iT™ PicoGreen® dsDNA kit (Invitrogen, UK). Supernatants were also tested using a capture ELISA. Streptavidin-coated plates (Fisher

NET immunofluorescence
NETs were stained for immunofluorescence microscopy as described 73 . In brief, 5x10 5 neutrophils were added to coverslips, stimulated, and then fixed with 4% PFA. Coverslips were blocked and sequentially incubated with an anti-histone H3 antibody (Abcam, UK) and Alexa Fluor® 488-conjugated goat anti-rabbit IgG secondary antibody (Life Technologies, UK).
Coverslips were washed, mounted, and sealed using with ProLong™ Gold antifade mountant with DAPI (Invitrogen, UK). Slides were visualized using a Zeiss Axio Imager.A1 inverted fluorescence microscope (Zeiss, Germany) and images analysed using Image J.

BAL confocal immunofluorescence
BAL fluid was filtered using a 40µm cell sieve. BAL cells were pelleted, counted and 1x10 5 viable cells were used to produce cytospin slides (Thermo Shandon Cytospin 3, Thermo Scientific). Cytospin slides were fixed in 4% PFA, washed, and blocked overnight in blocking solution (10% goat serum/1% BSA/2mM EDTA/HBSS/0.1% Tween-2). Slides were then washed and incubated with DAPI (Sigma, UK) diluted in blocking buffer for 1 hour. Stained slides were then washed, mounted, sealed and visualised using an Olympus inverted fluorescence confocal microscope and analysed using Image J.

Neutrophil adhesion
HUVEC were cultured in 96-well black tissue culture plates (Thermo Scientific, UK). 24 hours prior to experimentation, HUVEC were subjected to normoxia or hypoxia in the absence or presence of 10ng/ml TNF-α. Neutrophil adhesion in response to 20nM PMA or 100ng/ml LPS were measured as previously described 73 . Briefly, neutrophils were cultured under normoxia or hypoxia for 1 hour, then labelled with 2',7'-bis-(2-carboxyethyl)-5-(and-6)carboxyfluoresceinacetoxymethyl ester (Life Technologies, UK). Neutrophils were then added to wells under normoxia or hypoxia. Fluorescence was measured using a Tecan GENios Spectra FLUOR plate reader (Tecan UK Ltd., UK). Adhesion was calculated by comparing the fluorescence of washed wells to initial fluorescence.

Neutrophil trans-endothelial migration
Trans-endothelial migration assays were performed as previously described 73 . In brief, HUVEC were grown on transwell inserts (Millipore, UK). 24 hours prior to experimentation, HUVEC were cultured under normoxia or hypoxia in the absence or presence of 10ng/ml TNF-α.
Neutrophils were cultured under normoxia or hypoxia for 1 hour and then labelled with CellTracker (Invitrogen, UK). 1x10 6 neutrophils were added to the upper chamber of transwells and allowed to migrate in the absence or presence of 150ng/ml IL-8 in the lower chamber for 90 minutes. Percent transmigration was calculated by comparing the number of cells in the lower chamber and the number of neutrophils added to the upper chamber.

Western blotting
Cell lysates (10µg protein) were resolved by electrophoresis and transferred to a polyvinylidene fluoride membrane (GE Healthcare, UK). Membranes were blocked for 1 hour in 5% skimmed milk/TBS/0.1% Tween-20 and incubated with primary antibodies (1:1000 dilution) overnight at 4 o C. Membranes were then washed, incubated with HRP-conjugated secondary antibodies, and visualised using the Luminata Western HRP substrate system (Millipore, Ireland).

Statistical analysis
Data were evaluated using GraphPad Prism. Data were tested for normality using a Kolmogorov-Smirnov test. In experimental data sets only comparing two groups, a Mann-Whitney test was performed or a Wilcoxon matched pairs test. In data sets with two variables, data were assessed by two-way ANOVA with a Dunnet's multiple comparison test. Correlations were determined by two-tailed Pearson correlation coefficients. A p value below 0.05 was considered significant.