IL1RAP expression and the enrichment of IL‐33 activation signatures in severe neutrophilic asthma

Abstract Background Interleukin (IL)‐33 is an upstream regulator of type 2 (T2) eosinophilic inflammation and has been proposed as a key driver of some asthma phenotypes. Objective To derive gene signatures from in vitro studies of IL‐33‐stimulated cells and use these to determine IL‐33‐associated enrichment patterns in asthma. Methods Signatures downstream of IL‐33 stimulation were derived from our in vitro study of human mast cells and from public datasets of in vitro stimulated human basophils, type 2 innate lymphoid cells (ILC2), regulatory T cells (Treg) and endothelial cells. Gene Set Variation Analysis (GSVA) was used to probe U‐BIOPRED and ADEPT sputum transcriptomics to determine enrichment scores (ES) for each signature according to asthma severity, sputum granulocyte status and previously defined molecular phenotypes. Results IL‐33‐activated gene signatures were cell‐specific with little gene overlap. Individual signatures, however, were associated with similar signalling pathways (TNF, NF‐κB, IL‐17 and JAK/STAT signalling) and immune cell differentiation pathways (Th17, Th1 and Th2 differentiation). ES for IL‐33‐activated gene signatures were significantly enriched in asthmatic sputum, particularly in patients with neutrophilic and mixed granulocytic phenotypes. IL‐33 mRNA expression was not elevated in asthma whereas the expression of mRNA for IL1RL1, the IL‐33 receptor, was up‐regulated in the sputum of severe eosinophilic asthma. The mRNA expression for IL1RAP, the IL1RL1 co‐receptor, was greatest in severe neutrophilic and mixed granulocytic asthma. Conclusions IL‐33‐activated gene signatures are elevated in neutrophilic and mixed granulocytic asthma corresponding with IL1RAP co‐receptor expression. This suggests incorporating T2‐low asthma in anti‐IL‐33 trials.


| INTRODUC TI ON
Asthma is a heterogenous disease characterized by chronic airway inflammation affecting over 350 million people globally. Asthma exhibits many clinical phenotypes which remain poorly understood apart from the severe eosinophilic phenotype characterized by recurrent exacerbations and underlined by a type 2 inflammation with the expression of interleukin (IL)-4, IL-13 and IL-5. 1 Phenotypes associated with non-T2 pathways remain unclear although there is evidence for a neutrophilic inflammatory component of asthma that may be driven by Th-17 and Th1 pathways. 2 Current research, particularly in severe asthma (SA), aims to identify disease-driving mechanisms that would constitute a molecular phenotype or endotype of non-T2 pathways and, thereby enable the discovery of targeted therapies. Results: IL-33-activated gene signatures were cell-specific with little gene overlap.
Individual signatures, however, were associated with similar signalling pathways (TNF, NF-κB, IL-17 and JAK/STAT signalling) and immune cell differentiation pathways (Th17, Th1 and Th2 differentiation). ES for IL-33-activated gene signatures were significantly enriched in asthmatic sputum, particularly in patients with neutrophilic and mixed granulocytic phenotypes. IL-33 mRNA expression was not elevated in asthma whereas the expression of mRNA for IL1RL1, the IL-33 receptor, was up-regulated in the sputum of severe eosinophilic asthma. The mRNA expression for IL1RAP, the IL1RL1 co-receptor, was greatest in severe neutrophilic and mixed granulocytic asthma.
Conclusions: IL-33-activated gene signatures are elevated in neutrophilic and mixed granulocytic asthma corresponding with IL1RAP co-receptor expression. This suggests incorporating T2-low asthma in anti-IL-33 trials.

K E Y W O R D S
gene set variation analysis, IL-33, severe asthma

G R A P H I C A L A B S T R A C T
Stimulation of human mast cells, basophils, ILC2 cells and HUVECs with IL-33 generates distinct gene expression patterns but similar pathways are activated. We produced IL-33-activated gene signatures and used GSVA and showed enrichment of these signatures in patients with neutrophilic and mixed granulocytic asthma. These patients had the highest expression of the IL-33 receptor co-receptor IL1RAP. Abbreviations: BAS, basophil; HUVECs, human umbilical vein endothelial cells; IL, interleukin; ILC2, type 2 innate lymphoid cell; IL1RAP, IL-1 receptor accessory protein; JAK, Janus kinase; MC, mast cell; STAT, signal transducer and activator of transcription; Th, T helper; TNF, tumor necrosis factor Interleukin (IL)-33 is an alarmin released from airway epithelial cells in response to cellular injury or stress resulting from infection, allergy or pollutants. 3 Upon release, IL-33 binds to a heteromeric receptor consisting of the IL-1 receptor-like 1, IL1RL1, or ST2 and its co-receptor IL-1 receptor accessory protein (IL1RAP). This results in activation of downstream signalling pathways such as nuclear factor-κB (NF-κB) and mitogen-activated protein kinase (MAPK). 3 IL-33 activation is associated with type 2 (T2) inflammation and the release of the T2 cytokines, IL-5 and IL-13, from T-helper 2 (Th2) and group 2 innate lymphoid cells (ILC2s). IL-33 is also involved in the maturation of Th2 cells and acts as a Th2 cell chemoattractant. 4,5 IL-33 can also activate mast cells, basophils, eosinophils and natural killer cells. 6 However, during viral infection, IL-33 can both drive 3 and dampen 7 type 1 immunity. IL1RL1 is constitutively expressed in Th2 cells and transiently expressed in virus-exposed Th1 cells in vitro and in vivo. 8 Furthermore, free IL-33 binds to CD8 + T cells to potentiate cytotoxic functions in the presence of Th1 inflammation. 9 Associations between IL-33 and IL1RL1 genetic variants have been reported for asthma. 10 An IL-33 loss-of-function mutation protects children from asthma and has been associated with a lower blood eosinophil count. 11 IL-33 mRNA and protein expression is elevated in the airways of SA patients, 12 with a greater release of IL-33 in vivo during rhinovirus-induced asthma exacerbations. 13 IL1RL1 is characteristically expressed on Th2 cells which, in part, distinguishes them from Th1 cells 10 and was one of 20 genes distinguishing eosinophilic from non-eosinophilic asthma in sputum. 14 IL1RL1 is also a feature of patients enriched for the molecular phenotype transcriptomic associated cluster (TAC)1 that is associated with very high sputum eosinophilia in SA. 14 However, drugs directed against the IL-33/IL1RL1 axis have proved less efficacious in SA patients with high blood eosinophil counts than in those with low blood eosinophils. 15,16 In order to elucidate the type of asthma driven by IL-33 and its interaction with its heteromeric receptor, we obtained and derived several cell-specific IL-33-activated gene signatures. We used these to determine their expression levels in the transcriptome of cell compartments in blood and in the airways of asthmatic participants in the Unbiased Biomarkers for the Prediction of Respiratory Disease Outcomes (U-BIOPRED) cohort. Using this approach, we describe the phenotypes driven by IL-33, which consisted of the severe neutrophilic and mixed granulocytic asthma.
Patients were compared by asthma severity (MMA, SAs/ex, SAns), by granulocytic subtype (P = Paucigranulocytic, E = Eosinophilic, N = Neutrophilic and M = Mixed) and by molecular phenotype or TAC where TAC1 is high sputum eosinophilic, TAC2 is neutrophilic associated with inflammasome activation and TAC3 is paucigranulocytic or mixed phenotype linked with metabolic/mitochondrial pathways. 14 The TAC groups were derived from clustering of differentially expressed genes in sputum from asthmatic subjects with eosinophilic versus non-eosinophilic disease. 14 The ADEPT asthma cohort, 18 including severe (SA, n = 51), moderate (ModA, n = 55) and mild (MldA, n = 52) asthma, was used as a validation cohort. Transcriptomic data were available from the NCBI GEO database 19 for ILC2 cells cultured with IL-2/IL-7 in the presence or absence of IL-33 (50 ng/ml for 7 days, GSE72433), IL-33-stimulated basophils (20 ng/ml for 24 h, GSE64639), 22 IL-33 treated human umbilical vein endothelial cells (HUVECs) (50 ng/ml for 4 h, GSE37624) 23 and Treg cells engineered to constitutively express IL1RL1 and stimulated with IL-2 and anti-CD3/CD28 in the presence and absence of 20 ng/ml IL-33 for 16h (GSE117481). 24

| In vitro IL-33 stimulation of mast cells
Human peripheral blood mast cells from 6 donors were isolated as previously described incorporating a CD117 magnetic bead purification step. 25 All patients gave written informed consent, and the protocol was approved by the University of Manchester research ethics committee (UREC 2018-2696-5711). After 8 weeks, mature primary human mast cells were cultured at 0.5-1 × 10 6 in 500 ml supplemented Iscove's Modified Dulbecco's Medium 25 (see Appendix S1) in the presence or absence of IL-33 (50 ng/ml) for 24 h. RNA-seq data were generated using Illumina HiSeq 4000 as previously described. 26

| Development of IL-33 activated gene signatures
STAR, 27 Kallisto 28 and featureCounts 29 were used in in DNAnexus® for processing RNA-seq data generated in-house using IL-33-stimulated mast cells. R 3.5.0 (http://www.r-proje ct.org/) was used to perform statistical analysis. Differential gene expression (DEG) analysis between baseline and treated in vitro cell transcriptomics was performed using Limma for public array datasets and DESeq2 for the mast cell RNA-seq data. 30 Up-and down-regulated DEGs were identified by a log 2 fold change ≥1 or ≤ −1 and a false discovery rate 31 adjusted p value of ≤0.05. Gene signatures were cross-checked against published singlecell data for cell-specific genes for activated endothelium and mast cells in healthy subjects and patients with mild asthma (Table S1A). 32 In addition, we compared human neutrophil signatures from healthy donors (Table S1B) 33 and human eosinophil signatures from both healthy subjects (Shridar eosinophil 2) and patients with asthma (Table S1C). 34 Any overlap was removed. The final list of gene signatures used can be seen in Table S1. Removed genes are indicated in red.

| Identifying a consensus IL-33-activated gene signature
Overlap between signatures was investigated to attempt to establish a consensus IL-33-activated gene signature. A Venn diagram of all gene signatures was generated using public software (http://bioin forma tics.psb.ugent.be/webto ols/Venn/). All differential genes seen in at least n = 2 experiments were pooled together and analysed in enrichment analysis to attempt to identify any common features across cell types following IL-33 stimulation. human gene set database. 36 Gene set variation analysis (GSVA) was used to determine patient-wise enrichment scores (ES) indicating the relative collective expression of genes within the gene signatures for patients relative to the rest of the cohort of patients in a given transcriptomics dataset. 37 The ggpubr package (https://CRAN.Rproje ct.org/packa ge=ggpubr) was used to perform Wilcoxon ranksum tests to establish significance of individual gene differences between groups of subjects and display statistical comparisons on plots with Hommel multiple testing p value adjustment. 38 Plots were generated using ggplot2 (https://ggplo t2.tidyv erse.org).

| Generation of IL-33-activated gene signatures
IL-33-activated gene signatures were generated successfully from mast cell, basophil, ILC2 and HUVECs (Table S2). No differential gene signature was produced from the Tregs analysis. Minimal cellspecific genes from mast cells, activated endothelium and neutrophils were present in the IL-33-activated gene signatures generated (Table S1).
Pathway analysis of each cell-specific IL-33-activated upregulated gene signature (Tables S3-6 Although the cell-derived IL-33-activated gene signatures overlapped with respect to pathway enrichment, the gene lists were highly cell-specific with minimal overlap between them (Table S7) F I G U R E 1 Venn diagram plot of the different cell type generated IL-33 signatures for (A) up-regulated (B) and down-regulated signatures. No consensus overlap is seen; however, there is noted overlap between pairs of signatures and we were unable to identify a consensus IL-33-up-regulated gene signature ( Figure 1A). However, an integrative analysis of the overlapping genes within the signatures identified 81 genes that were present in 2 or more signatures (Table S8 A). Pathway analysis of these 81 genes confirmed enrichment of TNF, NF-κB, IL-17, JAK/ STAT signalling pathways and Th17, Th1 and Th2 differentiation pathways in addition to pathways associated with rheumatoid arthritis and inflammatory bowel disease (Table S9).

| Cell-specific down-regulated IL-33-activated gene signatures
As with the up-regulated signatures, minimal overlap was seen in  (Table S8B). Pathway enrichment analysis of these 14 genes did not reveal any significant pathways (Table S10).
No further analysis of these down-regulated gene signatures was performed.

| Enrichment of IL-33-up-regulated gene signatures in the blood and sputum transcriptome
There was a trend for higher ES in peripheral blood for IL-33-upregulated gene signatures according to asthma severity (Figure 2A-D) that reached significance (p < 0.05) for the ILC2 signature for all asthma cohorts against HC. There was no significant enrichment in blood by either granulocytic or TAC subtype.
The ES for IL-33-up-regulated signatures also showed enrichment with asthma severity in sputum ( Figure 2E-H) with significant enrichment for SAns (p < 0.0001) and MMA (p < 0.01) compared with HC for the IL-33 mast cell signature ( Figure 2E). Significant enrichment was seen in SAns against HC (p < 0.05) for the HUVEC IL-33-up-regulated signature ( Figure 2H). The ES in sputum were highly significantly elevated for all signatures in neutrophilic and mixed asthma subtypes ( Figure 3A-D). The mast cell signature was also significantly enriched (p < 0.01) in eosinophilic asthma compared to HC ( Figure 3A). Furthermore, the TAC1 and TAC2 subtypes, previously derived unbiased clustering of differentially expressed genes between sputum eosinophilic and non-eosinophilic asthmatics, 14 (Table S7).
There was no significant enrichment of any IL-33-activated signatures in bronchial and nasal brushings or bronchial biopsies when analysed according to asthma severity or granulocyte phenotype

| Validation of IL-33-activated gene signatures in the ADEPT cohort
There was no significant difference in ES of cell-specific IL-33 signatures by asthma severity across the ADEPT cohort ( Figure 4A-D).

| IL1RL1 (ST2) gene expression
There was a wide spread of IL1RL1 mRNA expression in blood that was significantly elevated in peripheral blood cells of MMA (p < 0.001) and SAns (p < 0.05) compared with HC ( Figure 5D). There was no difference in IL1RL1 mRNA expression across asthma severity which agreed with the blood eosinophil levels previously reported in these patients. 17 Sputum IL1RL1 mRNA levels were significantly higher in SA compared with HC (p < 0.001) and MMA (p < 0.0001)( Figure 6D). IL1RL1 was significantly elevated in eosinophilic and mixed subtypes (both p < 0.0001) compared with paucigranulocytic and neutrophilic asthma subtypes and against HC subjects ( Figure 6E). IL1RL1 mRNA was also elevated in neutrophilic asthma when compared to HC ( Figure 6E). IL1RL1 mRNA expression was significantly elevated in TAC1 against TAC2 and 3 (p < 0.0001)( Figure 6F).
However, IL1RL1 mRNA levels were significantly lower in SAs/ ex compared with SAns and MMA and HC (all p < 0.05) in bronchial biopsies ( Figure S4D). No significant difference in IL1RL1 mRNA expression was seen by granulocyte ( Figure S4E) or TAC ( Figure S4F) status.

| IL1RAP gene expression
IL1RAP mRNA expression in blood was unaffected by asthma severity, granulocytic grouping or TAC subtype ( Figure 5G-I). In contrast, IL1RAP mRNA expression was significantly increased in sputum in SAns and MMA (both p < 0.05) compared with HC ( Figure 6G). Furthermore, IL1RAP mRNA levels were significantly elevated in neutrophilic and mixed asthma compared to HC (both p < 0.001) ( Figure 6H) and in TAC2 asthmatics compared with TAC1 and TAC3 asthmatics (both p < 0.0001)( Figure 6I).  Figure S5G).
One-way analysis of variance (ANOVA) testing reveals that blood neutrophils, blood lymphocytes and blood monocytes do not alter the distribution of blood IL1RAP gene expression across the granulocytic groups (p = 0.874962). However, when these are included as interaction terms with regard to TAC subtype, it attains significance (p = 0.0227). Adjustment of the potentially confounding effect of these cells would result in a more TAC2 distribution of IL1RAP mRNA, a finding which further supports our investigation's finding of IL1RAP being associated with neutrophilic asthma. The IL-33 receptor is a complex receptor constructed of IL1RL1 and IL1RAP. IL1RL1 and IL1RAP mRNA expression was not differentiated amongst asthma cohorts in blood but IL1RL1 was elevated F I G U R E 5 Boxplots of U-BIOPRED blood asthma gene expression for IL-33, IL1RL1 (ST2) and IL1RAP (co-dimer of the IL-33 receptor) by severity cohort (HC = healthy control; MMA = mild-moderate asthma; SAs/ex = severe asthma smoking/ex-smoking; and SAns = severe asthma non-smoking) (A, D, G), granulocytic subtype (HC = Healthy Control, P = Paucigranulocytic, E = Eosinophilic, N = Neutrophilic and M = Mixed) (B, E, H), and transcriptome associated cluster (C, F, I) which reflects eosinophilic and non-eosinophilic phenotypes of asthma (TAC1 = eosinophilic asthma, TAC2 = neutrophilic asthma and TAC3 = mixed or paucigranulocytic asthma) in the sputum of severe eosinophilic asthma. IL1RAP mRNA expression was highest in severe neutrophilic and mixed granulocytic asthma. Together with the enrichment of the IL-33-stimulated cell signatures, these data provide evidence for IL-33 pathway activation in severe neutrophilic asthmatics in addition to those with eosinophilic asthma. This supports evidence from recent clinical trials where anti-IL-33-targeted therapies worked better in asthmatics with lower blood eosinophil levels. 15,16 Our findings of high enrichment of the IL-33-activated signatures in neutrophilic patients contrasts with the concept IL-33 being a driver of Th2 biology and of blood eosinophils being biomarker of IL-33 pathology. 10 We identified the JAK/STAT pathway as being activated downstream of IL-33 interaction with its heterodimer receptor. The JAK-STAT pathway can be activated in both T2 and non-T2 pathologies 42 which is further supported by our finding of IL1RAP mRNA being elevated in neutrophilic subjects. In a previous study, the expression of IL1RAP mRNA in sputum was significantly associated with sputum neutrophilia and airflow obstruction in asthma. 43 Furthermore, the human protein atlas shows high levels of IL1RAP protein in neutrophils (https://www.prote inatl as.org/ENSG0 00001 96083 -IL1RA P/tissu e/granu locytes).

| DISCUSS ION
IL1RAP mRNA is increased in PBMCs from children with asthma compared with healthy controls, and its expression is enhanced with increasing asthma severity and during exacerbations. 44 Selective blocking of IL1RAP with a monoclonal antibody (mAB-mR3) inhibits signalling not only by IL-33 but also IL-1 and IL-36 pathways. 45 In an ovalbumin-driven murine model of asthma, MAB-mR3 reduced lung neutrophilia as well as suppressing eosinophilia. This is in line with the predominant neutrophilic inflammation induced by IL-33 in the skin and peritoneal cavity and the ability of MAB-mR3 to suppress F I G U R E 6 Boxplots of U-BIOPRED sputum asthma gene expression for IL-33, IL1RL1 (ST2) and IL1RAP (co-dimer of the IL-33 receptor) by severity cohort (HC = healthy control; MMA = mild-moderate asthma; SAs/ex = severe asthma smoking/ex-smoking; SAns = severe asthma non-smoking) (A, D, G), granulocytic subtype (HC = healthy control, P = paucigranulocytic, E = eosinophilic, N = neutrophilic and M = mixed) (B, E, H), and transcriptome associated cluster (C, F, I) which reflects eosinophilic and non-eosinophilic phenotypes of asthma (TAC1 = eosinophilic asthma, TAC2 = neutrophilic asthma and TAC3 = mixed or paucigranulocytic asthma) neutrophils in a monosodium urate (MSU) crystal-induced model of gout. 45 The IL1RAP single nucleotide polymorphism (SNP) rs10513854 is associated with childhood wheeze whilst the rs9290936 SNP is associated with persistent wheeze. 46 These data support a functional role for the raised levels of IL1RAP seen in this study in patients with TAC2 or neutrophilic asthma. Furthermore, it suggests that these patients may respond to anti-IL1RAP-directed therapies. 45 IL1RL1/ST2 mRNA expression being significantly up-regulated in the sputum of eosinophilic asthmatics is consistent with the report of a strong association of IL-33 and its receptor with severe eosinophilic asthma. 10 This may be a reflection of the high expression of IL1RL1/ST2 on eosinophils 10  polymorphisms present. 10 The role of these polymorphisms needs to be further investigated.
We did not see significant enrichment of IL-33-activated signatures in the tissue compartments studied (bronchial and nasal brushings and bronchial biopsies). This may reflect the preferential enrichment of immune cells in sputum compared with biopsy samples. Furthermore, it may be a result of the effect of medication in suppressing tissue inflammation as previously described. 39 It was surprising to see enrichment of an endothelial cell signature in sputum but there is evidence for the presence of endothelial progenitor cells 49 and endothelial cell microparticles 50 in the sputum of COPD patients.
There are some limitations to this study including incomplete matching of tissue and sputum samples within U-BIOPRED with less than 5 TAC2 subjects having had bronchial biopsies or brushings collected, that may account for the lack of enrichment of the IL-33 signature in the airway samples compared to sputum rather than the use of activation signatures per se or the reflection of other cellular/molecular processes. In addition, the lower numbers of samples within the ADEPT sputum samples and the relative paucity of neutrophilic (n = 7) and mixed granulocytic (n = 5) subjects may account for only a partial validation of the U-BIOPRED data. IL-33 activation signatures derived from bulk RNA sequencing of cells may potentially reflect contamination with other cell types which will be resolved using single cell RNA-sequencing-derived signatures. Future studies involving severe asthma molecular phenotyping should strive to obtain samples from different compartments from all subjects to reduce the low overlap between compartments seen with many studies. The data presented here are a step towards the molecular classification of asthma. This type of approach has proved critical for indicating personalized targeted therapies in other chronic inflammatory and immune-based diseases. 51 The low expression of IL-33 mRNA in all sample types, aside from bronchial biopsies and brushings, the lack of differentiation of IL-33 gene expression between patient groups and the lack of corroboration with the enrichment of the IL-33 signatures indicates that IL-33 gene expression alone is insufficient to demonstrate IL-33 activation. Indeed, it is evident that IL-33 is not predominantly regulated at the level of mRNA expression, which is preformed and stored intracellularly within the nucleus where it may have additional roles. 20,21 Without a specific assay of IL-33 protein expression and activity, it is important to make use of IL-33 cell activation signatures to indicate IL-33 presence. This approach could be used in other studies where cellular activation status is more important than cell counts alone and where alarmin detection is also problematic as with thymic stromal lymphopoietin (TSLP). Analysis of TSLP-stimulated cell signatures in asthma should be undertaken in future studies. In addition, RNA-seq data from airway tissues, sputum or BAL cells should be obtained before and after successful, or even unsuccessful, treatment with drugs such as tezepilumab 52,53 to define a responder signature that could be used to help clarify potential responder populations.
In summary, we have shown an enrichment of IL-33-stimulated cell signatures in the sputum of severe neutrophilic asthmatics, extending the current genetic association studies and concepts of IL-33 playing a preferential role in severe eosinophilic asthma. This analysis highlights the need to examine the effect of IL-33-targeted treatments in non-T2 subjects.

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 openly available in Gene Expression Omnibus (GEO) reference numbers GSE76262 and GSE69683.