The effects of inhaled corticosteroids on healthy airways

Background The effects of inhaled corticosteroids (ICS) on healthy airways are poorly defined. Objectives To delineate the effects of ICS on gene expression in healthy airways, without confounding caused by changes in disease-related genes and disease-related alterations in ICS-responsiveness. Methods Randomised open-label bronchoscopy study of high dose ICS therapy in 30 healthy adult volunteers randomised 2:1 to i) fluticasone propionate 500 mcg bd daily or, ii) no treatment, for 4 weeks. Laboratory staff were blinded to allocation. Biopsies and brushings were analysed by immunohistochemistry, bulk RNA sequencing, DNA methylation array and metagenomics. Results ICS induced small between-group differences in blood and lamina propria eosinophil numbers, but not in other immunopathological features, blood neutrophils, FeNO, FEV1, microbiome or DNA methylation. ICS treatment upregulated 72 genes in brushings and 53 genes in biopsies, and downregulated 82 genes in brushings and 416 genes in biopsies. The most downregulated genes in both tissues were canonical markers of type-2 inflammation (FCER1A, CPA3, IL33, CLEC10A, SERPINB10 and CCR5), T cell-mediated adaptive immunity (TARP, TRBC1, TRBC2, PTPN22, TRAC, CD2, CD8A, HLA-DQB2, CD96, PTPN7), B cell immunity (CD20, immunoglobulin heavy and light chains), and innate immunity, including CD48, Hobit, RANTES, Langerin and GFI1. An IL-17-dependent gene signature was not upregulated by ICS. Conclusions In healthy airways, 4-week ICS exposure reduces gene expression related to both innate and adaptive immunity, and reduces markers of type-2 inflammation. This implies that homeostasis in health involves tonic type-2 signalling in the airway mucosa, which is exquisitely sensitive to ICS. Registered at ClincialTrials.gov: NCT02476825


INTRODUCTION
Inhaled corticosteroids (ICS) are the cornerstone of asthma treatment.They attenuate eosinophilic airway inflammation (1,2), improve lung function, and reduce asthma symptoms, exacerbations, and mortality (3).However, their use is also associated with an increased risk of pneumonia (4).Corticosteroids modulate the expression of many molecular pathways at the level of gene transcription, through direct upregulation of anti-inflammatory molecules and βadrenoceptors (transactivation), and suppression of pro-inflammatory genes, either through direct DNA-binding or via inhibition of pro-inflammatory transcription factor binding (transrepression) (5).In people with severe asthma, ICS are relatively ineffective even at high doses, but the mechanisms behind this corticosteroid insensitivity are poorly understood and likely multi-factorial (6,7).
When considering the underlying molecular pathways driving both severe asthma and relative corticosteroid insensitivity, it is unclear to what extent this is driven by pathways that are not responsive to corticosteroids, as opposed to inhibition of corticosteroid signalling.Gene expression profiling in asthmatic epithelial bronchial brushings and bronchial biopsies has identified several molecular pathways present in subgroups of patients with mild, moderate and severe asthma (8,9).Approximately 80% of people with steroid-naïve "mild" asthma demonstrate evidence of blood or airway eosinophilia, with a concomitant increase in the fraction of exhaled nitric oxide (FeNO) (10,11).This phenotype is characterised by increased expression of an airway gene expression signature driven by IL-4 and IL-13 (8,9,12,13); tissue eosinophilia is also dependent on IL-5 (14).Together these cytokines are described as Th2 or type-2 cytokines (T2).T2 expression and the accompanying eosinophilia are suppressed by ICS in mild asthma (15).In severe asthma, a persistent T2 gene signature is evident in about 25% of patients, suggesting corticosteroid insensitivity (8,9).In addition, in severe asthma, .about 25% of patients have evidence of an IL-17-dependent gene signature, which is seen only in people on ICS (8,9), and mutually exclusive with the T2 signature.It is therefore not clear whether this IL-17 activity represents an independent corticosteroid-insensitive pathway driving severe asthma or a consequence of ICS therapy.Approximately 50% of people with severe asthma have neither a T2-or IL-17-dependent airway gene signature, and the mechanisms driving their persistent disordered airway physiology remains unknown.So, while gene expression profiling provides insight into the abnormalities present in severe asthmatic airways, the multiple effects of ICS on airway gene expression make it difficult to disentangle the changes due to the disease as opposed to the treatment, and whether corticosteroid activity is inhibited or not.
The effects of ICS on healthy airways are poorly defined.We hypothesise that gene expression data in severe asthma will be more interpretable if we can delineate and thus allow for the effects of high dose ICS therapy.We have therefore performed a randomised open label bronchoscopy study of high dose ICS therapy in healthy adult volunteers, with the aim of understanding transcriptional consequences of ICS therapy without the confounding effects due to disease-related processes. .

METHODS
Detailed methods are provided in the online data supplement.This prospective study was approved by the East Midlands-Leicester Central Research Ethics Committee (reference:15/EM/0313) and registered at clinicaltrials.gov(NCT02476825).
Participants gave written informed consent.

Participant population
Healthy volunteers aged 18-65 were eligible, were current non-smokers with <10 pack year smoking history, and had no prior history or clinical evidence of lower respiratory disease with normal spirometry.Participants with a history of rhinitis were required to have a PC20 methacholine >16 mg/ml.

Study design
This was a randomised, open-labelled, bronchoscopy study designed to assess the effects of 4 weeks treatment with fluticasone propionate on airway gene expression and cellularity in healthy adult volunteers.The primary endpoint was the corticosteroid-inducible gene expression pattern in healthy airways.Secondary endpoints included the relative change from baseline in airway cellularity.To ensure there were sufficient data for analysis, if a subject withdrew before completion of the study, a further subject(s) was randomised after the first 30 randomisations until a total of 30 subjects had completed the study.

Bronchoscopy
Subjects underwent bronchoscopy conducted according to British Thoracic Society guidelines (16).Mucosal biopsies and brushes were collected from 2 nd -5 th generation bronchi under direct vision as per study procedure manual.

Tissue processing, immunohistochemistry and assessment of immunopathology
Please see the online supplement.
All pathological data were assessed by an observer blinded to the identity and treatment allocations of the participants.

RNA sequencing
Please see the online supplement.

Bisulphite conversion and DNA methylation arrays, DNA methylation data quality control and normalization, Differential DNA methylation analysis, Expression quantitative trait methylation (eQTM) analysis
Please see the online supplement.

Microbiota Sequence Data Generation, Processing, and Analysis
Please see the online supplement.

Transcriptomic analysis
Sequences in fastq files (in single and pair ends) were aligned using STAR aligner (version 2.7.1a) to the human reference genome GRCh38; R package (17)Rsubread was employed for quantification of reads assigned to genes.
Raw count pre-processing, normalisation and differential gene expression analysis was performed using R packaged DESeq2, edgeR, and limma.Gene pathways enrichment analysis was executed using R packages fgsea, clusterProfile, and ReactomePA.Heatmaps, volcano and MA plots were generated using in-home R code, ggplot2 and plotly packages (17,18).

Statistical analysis
Basic summary statistical analysis was performed using GraphPad Prism version 7.03 (GraphPad Software, San Diego).Parametric and non-parametric data are presented as mean (standard deviation [SD]) and median (interquartile range [IQR]) respectively unless otherwise stated.

Clinical characteristics
We recruited 44 healthy participants.32 proceeded to bronchoscopy but one patient withdrew from the study after the first bronchoscopy.31 completed 2 bronchoscopies, but one was subsequently withdrawn due to both <80% medication adherence and an intercurrent asymptomatic bronchitis evident at the 2 nd bronchoscopy (with bronchial wash samples positive for a non-Covid coronavirus and Staphylococcus aureus).The clinical characteristics of the 30 participants completing the study are shown in table 1.

Effects of ICS on biomarker and physiological measurements
There was a significant increase in blood eosinophil numbers after 4 weeks in the observation group compared to people using ICS, but this was not related to atopic status (Supplementary figure E1A).There were no significant between-group differences for changes in blood neutrophils, FeNO or FEV1 (Supplementary figure E1B-D).

Effects of ICS on airway inflammatory and structural cells
Suitable paired samples for immunohistochemical analysis of the lamina propria were available from 17 participants receiving ICS and 8 undergoing observation (Figure 1 and Supplementary Figure E2).Although there was no significant change in lamina propria eosinophil counts within either group from the 1 st to 2 nd bronchoscopy, there was a significant between-group difference in the changes (p=0.01),due to a non-significant increase in the observation group (Figure 1A).There was no correlation between the change in blood eosinophils versus tissue eosinophils in the observation group (rs= -0.024, p=0.97).Comparing the changes in measurements from the 1 st to 2 nd bronchoscopy for the ICS versus the observation group, there were no significant differences between treatment groups for lamina propria neutrophil, tryptase+ or chymase+ mast cell counts, airway smooth muscle (ASM) and .

13
epithelial area expressed as a percentage of biopsy area, or reticular basement membrane depth (Figure 1B-G)(Supplementary Figure E2).

Effects of ICS on airway gene expression
Suitable paired brush and biopsy samples were available for 15 and 20 participants respectively receiving ICS.We observed significant differential expression of genes amongst participants at week 4 compared with baseline in the ICS-treatment group, with upregulation of 72 genes in brushings and 53 genes in biopsies, and downregulation of 82 genes in brushings and 416 genes in biopsies (Figure 2, Supplementary Tables E1-E4, Supplementary Figures E3, E4).
Amongst participants in the observation-only group there were no significant changes in gene expression observed between baseline and week 4 (Supplementary Figure E5).There was a close correlation between epithelial brush and bronchial biopsy gene expression, with 20 genes common to the top 24 most significantly differentially upregulated genes in both airway compartments (Table 2) and 20 genes common to the top 41 most significantly downregulated genes in both compartments (Table 3).
Consistent with these findings, pathway analysis of bronchial brushings with Reactome showed strong ICS-related downregulation of innate and adaptive pathways, including 'Immunoregulatory interactions between a lymphoid and non-lymphoid cell', five pathways related to TCR signaling, the immunological synapse or co-stimulation, with weaker signals for 'Generation of second messenger molecules', 'PD-1 signaling', and 'Chemokine receptors bind chemokines', showing the potent ability of ICS to suppress local adaptive T cell immunity (Supplementary Figures E6, Supplementary Table E5).Parallel analysis of the bronchial biopsies again showed effects on lymphoidnon-lymphoid interactions and TCR signaling, but also strong suppression of 'Extracellular matrix organization', 'Cell surface interactions at the vascular wall' and 'Integrin cell surface interactions', suggesting the potential for suppression of inflammatory cell recruitment to the mucosa (Supplementary Figure E7, Supplementary Table E6).ICS did not upregulate the IL-17-dependent gene signature identified previously in people with moderate-severe asthma (Supplementary Fig. E8) Next we compared our set of differentially upregulated genes with a set of 26 genes previously reported as induced by 10 weeks of inhaled fluticasone in bronchial brushings from participants with mild asthma (19).Geneset enrichment analysis profiles showed very close agreement between genesets (Supplementary Figure E9), particularly for brushings, which were directly .15 comparable between studies, showing very similar ICS-induced gene induction in health and mild asthma.

Effects of ICS on the airway microbiome
Minimal effects of ICS treatment were observed on the airway microbiome (details in Supplementary Results and Supplementary Figure E10).

Effects of ICS on airway DNA methylation
There were minimal effects of ICS treatment on DNA methylation (details in Supplementary Results and Supplementary Figure E11). .

DISCUSSION
In this open label, randomised study in healthy adult volunteers, we have defined genes that are altered directly by medium-term (4 weeks), twice daily high-dose ICS therapy independent of the confounding effect of asthma pathophysiology.Changes in airway cellularity were minimal between ICS treatment and no treatment, and there were no measurable biologically significant effects on the airway microbiome or DNA methylation.
By contrast, we observed widespread changes in gene expression following ICS use in these healthy individuals, which predominantly comprised downregulated genes.A striking finding was that the most significantly downregulated genes statistically were canonical markers of T2-driven inflammation.Notably the most highly downregulated gene, in both airway brushes, which predominantly sample the airway epithelium, and mucosal biopsies, was FCER1A which encodes the α chain of the high affinity IgE receptor, a key effector component of mast cell activation by allergens and of T2 immunity, which is highly upregulated in asthma (20,21).In addition, there was downregulation of genes encoding mast cell tryptases (TPSB1, TPSAB1) and carboxypeptidase A3 (CPA3), the beta chain of the high affinity IgE receptor (MS4A2), and prostaglandin D2 synthase (PTGDS1).We did not see a reduction in mast cell numbers in the airway mucosa of these healthy individuals, so the predominant effect of ICS appears to be on gene transcription, rather than mast cell survival, but nevertheless suggests that ICS have a potentially important dampening effect on mast cell function.In biopsies, there was also downregulation of the important mast cell chemoattractants CXCL10 and CXCL11 which promote mast cell migration to the ASM in mild steroid-naive asthma through the airway mast cell chemokine receptor CXCR3 (22).In people with severe asthma using high dose ICS +/-oral corticosteroids, mast cells are not increased within the ASM (23,24), suggesting suppression of mast cell chemoattractants from ASM may remain corticosteroid-sensitive in severe disease.This is in contrast to the release of mast cell-derived proteases and autacoids which demonstrate ongoing release in severe disease (23,25).
Other highly downregulated T2-related genes included IL33, an airway epithelial cell alarmin which acts to promote initiation of T2 responses (26), CLEC10A which is a marker for alternatively activated macrophages (27), SERPINB10 which is highly upregulated on airway epithelium by IL-13 and inhibits Th2 cell apoptosis, and CCR5 whose ligands include CCL5 (RANTES), a chemoattractant for eosinophils and mast cells.Moreover PTGS1, which encodes cyclooxygenase-1 was the 7 th most downregulated gene in brushings.This enzyme converts arachidonate to prostaglandins and its dysregulation or inhibition is implicated in salicylate-sensitive asthma and sino-nasal eosinophilic inflammation (28).Similarly in biopsies, periostin (POSTN), an IL-13-induced epithelial gene associated with T2 high asthma (29), was also downregulated.
Taken together, the ability of ICS to downregulate this extensive set of T2-related genes implies that homeostasis in health involves a low level of tonic T2 signalling in the airway mucosa that is very sensitive to ICS.This would be consistent with the observation that IL-4and IL-5-positive cells are present in healthy airways (30), as are mast cells and dendritic cells expressing FcεRIα (31).Constitutive T2 signalling is also evident in primary epithelial cell cultures grown at air-liquid interface, where we previously observed that STAT6-dependent genes including POSTN, CLCA1 and SERPINB2 were repressed by NF-dependent cytokine stimulation or dexamethasone (32).Thus differentiated bronchial epithelial cells have a low level of tonic STAT6 dependent signalling in the absence of exogenous IL-4 or IL-13.However, FeNO was not reduced by ICS in these healthy volunteers, and NOS2, which regulates FeNO production, was not altered by ICS therapy, suggesting tonic T2 signalling is below the threshold required for pathological FeNO generation.
. The role of low-level, tonic, constitutive T2 signalling in the airways is uncertain but likely beneficial for tissue homeostasis.T2 immunity exhibits many host-protective functions, including maintaining metabolic homeostasis, suppressing excessive T1 inflammation, maintenance of barrier defence and regulation of tissue regeneration (33)(34)(35)(36).For example, IL-33 is pleiotropic and can promote type 2 inflammation but in other contexts it can be immunoregulatory (26), and preserves epithelial integrity during influenza infection in a mouse model (37,38).Thus the effects of type 2 immunity may be context specific and affected by the cellular source and concomitant inflammatory milieu, and at steady state may promote airway epithelial barrier function.Inhibition of homeostatic tonic T2 signalling might therefore have deleterious effects which may have an impact clinically, and might explain an enhanced propensity to proteobacterial colonisation or infection, as was observed in the MEX (39) and RASP-UK(40) studies amongst participants with very low T2 biomarkers.
In addition to the inhibition of T2-driven genes, there was marked suppression of molecules involved in both protective innate immunity (e.g.CD48, CEACAM5) and adaptive immunity, with suppression of genes related to dendritic cells (CD207), T cells (e.g.CD2/3/6/8/96, TRBC1/2, TRAC), and B cell function (e,g, CD20/79, most heavy and variable light chains for IgA, IgG and IgM, JCHAIN).This potential impairment of innate, cell-mediated and antibodydependent immunity likely explains the reproducible dose-dependent increased pneumonia risk in people with asthma and COPD who are using ICS (4).A subgroup of people with asthma also become colonised with certain fungi (41), and it is possible that this is also a consequence of the mucosal immunosuppression observed here.
An important aim of this study was to provide information on the activity of ICS in healthy airways, to remove the confounding that might occur due to disease-related changes in gene expression or inherent ICS responsiveness, which should facilitate analyses of gene expression .changes in asthma.Studies of severe asthma to-date (42) have attempted to account for the effects of ICS based on gene expression changes derived from in vitro cell cultures, from an interventional study of gene expression changes in people administered ICS for chronic obstructive pulmonary disease, and from one previous intervention study which used microarrays to assess acute airway transcriptional consequences 6 hours after a single inhaled dose of budesonide (1600 µg) in 12 healthy, steroid-naïve men (43).Our findings are complementary to the previous budesonide study in that some differentially regulated genes are common to our two studies (including upregulation of FKBP5, ZBTB16, PHACTR3, TSC22D3 and downregulation of CD207, FCER1A, IL33), but there is a striking difference in the overall consequences of a single dose versus 4 weeks twice daily ICS use.The predominant effect of a single large acute dose of ICS was upregulation of genes (transactivation), with 68 genes upregulated and only 28 downregulated.Many of the upregulated genes were proinflammatory including growth factors, chemokines, chemokine receptors, cytokines, growth factors, and coagulation factors.By comparison in our chronic high dose exposure study only 53 genes were upregulated in biopsies, whilst 416 genes were downregulated (transrepression).This highlights the critical and often overlooked importance of the complex temporal dynamics of corticosteroid effects.Early during an acute infectious or traumatic challenge a surge in endogenous corticosteroids will tend to enhance protective inflammatory and procoagulant responses, but as time passes these responses will attenuate and slower but potent transrepressive effects will dominate to curtail uncontrolled inflammatory cell recruitment and activation, to prevent bystander tissue damage, and to direct resolution of tissue homeostasis.It is this latter situation which is of greatest clinical relevance to long term asthma management.Of note, comparing our data to the study by Woodruff(12), where people with mild asthma received inhaled placebo or fluticasone propionate (500 µg) twice daily for 8 .weeks, there was a strong correlation between the studies for ICS-inducible genes, suggesting that in mild asthma at least, ICS-dependent transactivation is preserved.
When considering previous studies of gene expression in severe asthma, an IL-17dependent gene signature is expressed in a subset of people, and mutually exclusive with a T2 gene signature (8,9).This IL-17-dependent signature was only seen in people using ICS, but it has been unclear whether this is disease-or treatment-related, and whether it is harmful or potentially protective.We did not see upregulation of this IL-17 signature after 4 weeks of ICS treatment in this study, thus it seems most likely a feature of disease.Similarly, there is upregulation of CEACAM family members in severe asthma, notably the IL-13-dependent gene CEACAM5 (44,45), and CEACAM6 which was increased on both the epithelium and neutrophils in bronchial biopsies (44).Here we found that CEACAM6 expression did not change with ICS treatment, while CEACAM5 expression was reduced (tables E2, E4), in keeping with the inhibition of other T2-related genes.Therefore the upregulation of these CEACAMs in severe asthma appears to be a feature of the disease rather than treatment.In the recent U-BIOPRED bronchoscopy study (24), ICS-inducible genes such as FKBP5 were upregulated in severe asthma, suggesting that their participants were adherent to treatment, and importantly, that ICS transactivation appears to be preserved in severe asthma, but many pathological pathways that should be sensitive to transrepression by ICS such as T2 signalling are not responsive in a subset of patients (8,9,23).
There are some limitations to our work.Firstly the effects of ICS in healthy airways at 4 weeks, while likely representative of the steady state in long term therapy, might not be fully representative of longer term therapy.However it is not reasonable to ask healthy volunteers to take ICS for a year, and adherence would likely wane.Secondly our analyses use bulk sequencing of airway brushes and biopsies, and ICS effects are likely to be highly cell-specific, .so future studies using spatial sequencing, from central and peripheral airways as well as nasal tissues and peripheral blood would be informative.For example, as we did not enumerate all of the cell types implicated in the transcriptomic analyses, we cannot conclude whether the changes in T-and B-lymphocyte related gene expression are due to transcriptional changes within lymphocytes, changes in the relative proportions of lymphocytes in the samples, or a combination of the two.
In summary, we defined genes altered directly by ICS therapy without confounding by disease.We provide evidence that IL-17-dependent signalling in asthma is disease-dependent rather than ICS-dependent, and demonstrate downregulation of canonical markers of T2 inflammation, implying that homeostasis in health involves tonic T2 signalling in the airway mucosa, which is exquisitely sensitive to ICS.There was also broad suppression of innate and adaptive immunity, in keeping with known immunosuppressive effects of corticosteroids. .
30 participants were randomised by a blinded investigator (MR) in a 2:1 ratio to one of two study groups: i) fluticasone propionate 500 mcg b.i.d.via Accuhaler (Diskus) for 4 weeks (n=20), or ii) no treatment (observation) for 4 weeks (n=10).Bronchoscopy was performed at baseline prior to the start of treatment/observation, and at the end of week 4. Genentech and Leicester laboratory support staff were blinded to treatment allocation.. 10

Table 1 -Demographics of the study participants
.

Table 2 -Top upregulated differentially expressed genes
Top 20 upregulated genes common to both epithelial brushes and bronchial biopsies occurring in the top 24 differentially upregulated genes when ranked by FDR p value.Table ranks genes by arithmetic mean of FDR p values across both tissues.Family With Sequence Similarity 107 Member A; role in actin and microtubule organisation.

Table 3 -Top downregulated differentially expressed genes 600 601
Top 20 downregulated genes common to both epithelial brushes and bronchial biopsies occurring in the top 41 biopsy differentially 602 downregulated genes when ranked by FDR p value.Table ranks genes by arithmetic mean of FDR p values across both tissues.