COPD patients with chronic bronchitis and higher sputum eosinophil counts show increased type‐2 and PDE4 gene expression in sputum

Chronic obstructive pulmonary disease (COPD) patients with higher eosinophil counts are associated with increased clinical response to phosphodiesterase‐4‐inhibitors (PDE4i). However, the underlying inflammatory mechanisms associated with this increased response is not yet elucidated. This post hoc analysis focused on sputum gene expression in patients with chronic bronchitis who underwent 32‐day treatment with two doses of the inhaled PDE4i CHF6001 (tanimilast) or placebo on top of triple therapy. Biological characterization and treatment effects were assessed between patients with different sputum eosinophil levels (eosinophilhigh ≥ 3%; eosinophillow < 3%) at baseline (primary samples) or at the end of the treatment of the placebo arm (validation samples). Forty‐one genes were differentially expressed in primary samples (p‐adjusted for false discovery rate < 0.05); all up‐regulated in eosinophilhigh patients and functionally enriched for type‐2 and PDE4 inflammatory processes. Eleven out of nineteen genes having immune system biological processes annotations including IL5RA, ALOX15, IL1RL1, CLC, GATA1 and PDE4D were replicated using validation samples. The expression of a number of these inflammatory mediators was reduced by tanimilast treatment, with greater effects observed in eosinophilhigh patients. These findings suggest that type‐2 and PDE4 overexpression in COPD patients with higher sputum eosinophil counts contribute to the differential clinical response to PDE4i observed in previous clinical trials.


| INTRODUC TI ON
Chronic obstructive pulmonary disease (COPD) patients with higher sputum or blood eosinophil counts have a greater response to inhaled corticosteroids (ICS), 1-3 with blood eosinophil counts being able to predict the ICS effect on exacerbation prevention. Furthermore, post hoc analyses of randomized controlled trials (RCTs) involving COPD patients with chronic bronchitis who received the PDE4 inhibitor roflumilast or placebo in addition to maintenance ICS and long-acting bronchodilators also demonstrated an association between higher blood eosinophil counts and greater effects of roflumilast on exacerbation prevention. 4 The mechanisms responsible for these differential drug effects are unknown but may relate to an increased presence of type-2 (T2) inflammation in COPD patients  6 In preclinical studies and RCTs, tanimilast showed a potent topical anti-inflammatory effect which was devoid of class-related systemic adverse effects. [5][6][7][8][9][10] In recent post hoc analyses, it was shown that tanimilast significantly reduced the exacerbation rate in the subgroup of COPD patients with chronic bronchitis and eosinophil count ≥ 150cells/µl after 24 weeks of treatment. 10 In a biomarker RCT conducted in COPD patients with chronic bronchitis receiving triple therapy (ICS/ long-acting β 2 agonist therapy (LABA) / long-acting muscarinic antagonist (LAMA)), tanimilast showed clear anti-inflammatory effects by modulating a range of airway biomarkers and inflammation pathways after 32 days of treatment. 8,11 Moreover, the ability of tanimilast to reduce sputum eosinophil counts was increased in patients with higher eosinophils levels in sputum (≥3%). 12 These data are compatible with previous RCT results with roflumilast showing inhibition of sputum eosinophil counts accompanied by a reduction in bronchial mucosal eosinophil numbers. 13 We have performed a post hoc analysis using samples from the tanimilast biomarker RCT. 8 Samples obtained before randomization (primary samples) and at the end of the treatment of the placebo arm (validation samples) were used to stratify patients according to sputum eosinophil counts. A threshold of 3% was used to define a subset of patients associated with the phenotype of eosinophilic COPD, that is 'eosinophil high ' (≥3%) versus 'eosinophil low ' (<3%). [14][15][16] Differential whole-genome gene expression analysis was carried out in whole blood and sputum cells by microarray with the aim to characterize the underlying biology and to evaluate the effect of the treatment on the two groups of patients.

| Study objective and design
This post hoc analysis was conducted on samples from patients being treated with triple therapy who were randomized to one of the 32-day treatment periods (800 or 1600 μg twice daily (BID); total daily doses of 1600 or 3200 μg or placebo) in a crossover study (ClinicalTrials.gov: NCT03004417), results of which have been previously reported. 8 Samples obtained before receiving the investigational drug (at screening visit; baseline samples for primary analysis) and at the end of the placebo treatment (latest available collection on Day 20, 26 or 32; placebo samples for validation analysis) were used to stratify patients into two subgroups using a sputum eosinophil threshold of 3%, that is 'eosinophil high ' (≥3%) and 'eosinophil low ' (<3%). Microarray differential whole-genome gene expression analysis between subgroups and the effect of the treatment on the identified significant genes both in the primary and validation analyses was carried out in whole blood and/or sputum cells.
Methods for the sputum and blood sample collection, and processing for the ribonucleic acid [RNA] assessments, extraction and amplification, sample profiling, microarray data quality control and pre-processing have been previously reported. 11 The raw data analysed in this publication have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE133513 (https://www.ncbi.nlm.nih.gov/geo/query/ acc.cgi?acc=GSE13 3513).

| Patients
Patients were male or female, ≥40 years of age, current or exsmokers with a smoking history ≥ 10 pack-years, a diagnosis of COPD, post-bronchodilator FEV 1 ≥ 30% and < 70% predicted, ratio of FEV 1 to forced vital capacity (FVC) <0.70, COPD Assessment Test (CAT) score ≥ 10, a history of chronic bronchitis (defined as chronic cough and sputum production for more than three months per year for at least two consecutive years), and treated with inhaled triple ICS/LABA/LAMA therapy for at least two months prior to enrolment. All patients provided written informed consent prior to any study-related procedure. The key exclusion criteria were a moderate or severe COPD exacerbation within six weeks prior to entry or between screening and randomization, and the use of PDE4i within two months prior to entry. 8

| Processing and data analysis
Microarray preparation and data processing were described in detail in a previous manuscript. 11 Briefly, samples were pre-processed using Robust Multichip Algorithm (RMA). Probe-level intensity measurements (CEL files) were background corrected, normalized and summarized as expression measurements using RMA. Pre-processed data were filtered to remove control transcripts and any uninformative transcripts (lowly expressed, invariant probe sets). Differential expression analysis between subgroups was performed in R version 4.0 (R Core Team, Vienna, Austria, 2020). The eBayes algorithm of the Linear Models for Microarray Data (LIMMA) Bioconductor package v3.44.1, 17 with Benjamini-Hochberg multiple testing correction, 18 was used to identify significant probe sets from filtered microarray data. All probe sets with P-value adjusted for false discovery rate (pFDR)<0.05 were considered to produce lists of significant differentially expressed genes (DEGs) for each analysis. When multiple probe sets were associated with the same gene, the probe set with the lowest pFDR value was considered in the DEG list.
The treatment response for the DEGs identified in the differential expression analyses between subgroups was evaluated as change

| Hierarchical clustering and functional enrichment analysis
Hierarchical heatmap clustering was performed in R version 4.0.
Differentially expressed probe set lists were annotated with gene identifiers using the latest annotation provided by Affymetrix for the Plus 2.0 array. To further understand the underlying biological significance of DEGs, g:Profiler 19 was used for functional enrichment analysis to produce KEGG, Reactome and Wiki pathways as well as gene ontology (GO) functional annotation (biological process (BP), cellular component (CC) and molecular function (MF)). DEG lists were used as input. Only DEG lists comprising more than 30 unique genes were used as input for the analysis. Entities were ranked according to a statistically derived enrichment score and were adjusted for Benjamini-Hochberg multiple testing (pFDR < 0.05).

| Molecular interaction network analysis
The genes of interest were input into the PSICQUIC (Proteomics Standard Initiative proposed the Proteomics Standard Initiative Common QUery InterfaCe) web service to explore the molecular interaction based on Reactome, 20 and IMEx databases. 21 Network diagrams were drawn with Cytoscape version 3.8.0. genes, 22 were included in the analyses.
In blood, only 2 probe sets were significantly (pFDR < 0.05) up-regulated in the eosinophil high group, translating for peripheral myelin protein-22 (PMP22) and phospholipase A acyltransferase-5 (PLAAT5) ( Figure 1A). In contrast, in sputum cells, there were 61 probe sets corresponding to 41 DEGs significantly differentially expressed (pFDR < 0.05) between the two groups ( Figure 1B) with one gene in common with blood (PLAAT5). All probe sets were statistically significantly up-regulated in the eosinophil high compared to eosinophil low population with fold change > |1.3| and pFDR < 0.05 (Table 2). Hierarchical heatmap clustering of the 61 significant probe sets highlighted one up-regulated cluster enriched for eosinophil high patients and one down-regulated cluster enriched for eosinophil low patients ( Figure 1C). Functional analysis of the DEGs resulted in an enrichment (pFDR < 0.05) of 104 GO biological processes, 1 GO molecular function, 1 GO cellular component, 1 reactome pathway and 3 KEGG pathways. The most associated common terms were immune system processes, cytokine signalling, interleukin-5 (IL5) production and cellular membrane components ( Figure 2; Table S1).
Overall, functional enrichment analysis and directionality of differential expression showed up-regulation of inflammatory genes in sputum cells of eosinophil high patients.  Table 2; Table S1). Notably these inflammatory genes were all up-regulated in the eosinophil high group and were associated with T2 inflammation or PDE4 pathways ( Figure 3). Furthermore, by exploring molecular network associations, major connections were found, with IL5RA, PDE4D, CCL26, S1PR1, NTRK1 and YES1 playing a central role involving the majority of interactions. Notably, PDE4 was found to be connected to the T2 inflammatory network via IL5RA, and S1PR1 with the cyclic AMPdependent protein kinase A (PRKACB) and adenylate cyclase isoforms (ADCY) acting as bridging molecules in the network ( Figure 3).
To explore if smoking status and gender could explain the biological differences observed between eosinophil high and eosinophil low patients, differential gene expression analysis was conducted between ex-and current smokers (43% and 57%, respectively) and male and female patients (71% and 29%, respectively). This analysis highlighted major biological differences in sputum caused by active smoking; 750 probe sets, corresponding to 497 DEGs, were significantly differentially expressed (pFDR < 0.05) in sputum, while no probe sets were significant in blood ( Figure S1; Table S2). Only two genes (both up-regulated in ex-smokers) were in common with DEGs between eosinophilhigh and eosinophil low patients; IL1RL1 and the family member 1, transcriptional corepressor (TLE). Functional analysis of DEGs in sputum resulted in an enrichment (pFDR < 0.05) of 318 GO biological processes, 16 GO molecular functions, 25 GO cellular components, 2 KEGG and 12 Wiki pathways (Table S3). The most common terms associated with the top entities were response to chemical stimulus and cytokine activity.

TA B L E 1 (Continued)
Differential gene expression analysis between male and females in comparison to eosinophilic status did not show any overlapping significant probe set ( Figure S2; Table S4; Table S5, Table S6).

| Validation analysis; Placebo samples
Stratification using the available placebo sputum samples resulted in thirteen out of fifty patients (26%) being eosinophil high ; 77% of whom maintained levels ≥ 3% from baseline.
Key results observed in the primary analysis using baseline samples were reproduced in this validation analysis using placebo samples. Specifically, 77 probe sets, corresponding to 48 DEGs, were significantly differentially expressed (pFDR < 0.05) between eosinophil high and eosinophil low patients in sputum cells (Table 2; Figure S3).
Hierarchical heatmap clustering of the 77 significant probe sets highlighted one up-regulated cluster enriched for eosinophil high patients and one down-regulated cluster enriched for eosinophil low patients ( Figure S4). Seventy-six out of 77 probe sets were statistically significantly up-regulated in the eosinophil high compared to eosinophil low population, all with fold change > |1.3| and pFDR < 0.05.
Functional analysis of DEGs resulted in an enrichment (pFDR < 0.05) of GO biological processes and GO molecular functions associated with immune system processes, cytokine and interleukin-5 (IL5) signalling (Table S7).
Considering the DEGs identified in the primary and validation analyses (41 and 48 DEGs, respectively, corresponding to 65 unique genes; Table 2), there was a strong correlation between the corresponding fold change values from the two data sets (Pearson-r = 0.9, P < .0001; Figure 4; Table S8). In particular, three key T2 inflammatory DEGs identified in the validation analysis, namely cysteinyl leukotriene receptor 2 (CYSLTR2), prostaglandin D2 receptor 2 (PTGDR2) and CCAAT-enhancer-binding protein epsilon (CEBPE) when compared to the primary analysis showed fold changes of

| Tanimilast effect on gene expression in eosinophil high/low patients
Analysis in sputum of the pre-to post-dose gene expression fold change for tanimilast versus placebo treatments on the DEGs identified in the primary or in the validation analysis (Table 2)

| D ISCUSS I ON
The analysis presented here focused on sputum gene expression in Differentially expressed genes (DEGs) in common between the primary and validation analyses.

TA B L E 2 (Continued)
consistently and significantly reduced in the eosinophil high population and the overall population by both tanimilast doses but with a more pronounced effect in the eosinophilic group. It has previously been reported that PDE4 inhibitors have a greater effect in COPD patients with higher eosinophil counts 4,10 and that PDE4 inhibition can reduce airways eosinophil numbers. 12,13 The results presented here demonstrate biological effects of PDE4 inhibition on inflammation processes associated with increased eosinophil counts in COPD patients, supporting these previous clinical observations.
These primary and validation results demonstrate up-regulation of a specific T2-and PDE4-related fingerprint that is associated with the phenotype of eosinophilic COPD.
Network analysis showed a set of genes (IL5RA, PDE4D, S1PR1, NTRK1 and YES1) playing central roles as interacting molecules within a network. Notable T2 cytokines and chemokines within this network were IL5RA which plays a key role in eosinophil differentiation, recruitment, activation and survival, 29 IL4 which is secreted by T2 cells and is involved in the accumulation of eosinophils at sites of inflammation, B cell differentiation and T2 cytokine production, 29 and CCL26 which acts as a ligand for C-C motif chemokine receptor 3 (CCR3) which is expressed predominantly on eosinophils and mediates the chemotactic response to several chemokines. 29 Furthermore, CCR3 receptors are strongly up-regulated by the sphingolipid inflammatory mediator S1PR1 which is critically involved in eosinophils activation and recruitment. 30 Other DEGs were also functionally linked to T2 inflammation, 31-37 including ALOX15 and the tyrosine kinase receptor NTRK1 which play important roles in immune responses including the cellular response to interleukin-13. 38,39 IL1RL1 is the receptor for interleukin-33 (IL-33) which acts as a selective chemoattractant of Th2 cells, 29 and elicits IL5-dependent eosinophilia. 40 The signal transducer CD24 is an adhesion antigen expressed at the surface of eosinophils, B lymphocytes, T cells, dendritic cells and neutrophils. CD24 was recently shown to bind a variety of danger-associated molecular patterns, such as high-mobility group box protein-1 (HMGB1), members of the heat-shock-protein (HSP) family and nucleolins. 41 The transcription factor GATA1 is critically involved in T2 cell maturation, activation and granulopoiesis. 42 Finally, CLC is a lysophospholipase expressed in eosinophils and basophils whose activity mediates extracellular cytotoxicity and inflammation. 43 The expression of CLC and ALOX15 that we show here to be consistently associated with eosinophil counts and that is reduced by the effect of the PDE4 inhibitor tanimilast in sputum cells was also previously shown to be associated with T2-high gene expression signature in bronchial epithelial brush. 27,28 In the validation analysis, we identified other two key receptors which are known pharmacological target of T2 inflammatory conditions, namely cysteinyl leukotriene 2 and prostaglandin D2. 44,45 Notably, assessment of the fold change values between the primary and validation analyses for the identified DEGs highlighted a strong correlation between the corresponding values from the two data sets.
We also identified a significant up-regulation of the gene coding for the enzyme PDE4D in the eosinophilic population. This isoform, which is strongly inhibited by both roflumilast and tanimilast, 46,47 was shown to reduce the expression of adhesion molecules, airways F I G U R E 2 Top 20 gene ontology (GO) biological processes (black), GO molecular functions (red), GO cellular components (green), KEGG pathways (orange) and Reactome pathways (blue) identified by functional enrichment analysis of the significant (pFDR < 0.05) differentially expressed genes (DEGs) in sputum cells between eosinophil high and eosinophil low patients (primary analysis; baseline samples) reactivity and enhance muco-ciliary clearance. 48 The novel association found in the present analysis between PDE4D up-regulation and eosinophilic inflammation was also supported by the network analysis of the inflammatory genes showing interactions between PDE4D and the T2 network via IL5RA and S1PR1 with the cyclic AMP-dependent protein kinase A (PRKACB) and adenylate cyclase isoforms (ADCY) acting as bridging molecules. Differently from sputum, the differential expression in blood did result in only two genes significantly up-regulated in the eosinophil high group of patients.
Notably one of these genes, PLAAT5, which is involved in phospholipase A1/2 and acyltransferase activities was also significantly up-regulated in sputum.
Our findings indicate that eosinophil high COPD patients display a specific profile of T2-related airway inflammation despite the concomitant use of ICS. In the control groups of two recent meta-analyses of roflumilast and mepolizumab, in which COPD patients were using maintenance ICS and bronchodilators, there was an increase in exacerbations at higher blood eosinophil counts. 4,49 Overall, these findings suggest ongoing T2-and PDE4-related inflammation associated with eosinophil numbers in COPD patients treated with ICS that may be targeted with PDE4i.
In our study population, eosinophil high COPD patients were characterized by a higher proportion of males and ex-smokers.
These findings are in line with some other studies which showed a higher prevalence of males in COPD patients with higher eosinophil counts, 50-53 and a role of active smoking in decreasing the number of eosinophils in the airways with a possible impact on local T2 inflammation. 54 Notably, we show that only one inflammatory gene, up-regulated in ex-smokers (IL1RL1), overlapped with the differentially expressed genes between eosinophil high and eosinophil low patients. This indicates that biological differences between patients F I G U R E 3 Inflammatory network molecular interaction analysis of the significant (pFDR < 0.05) differentially expressed genes (DEGs) associated with immune system GO biological processes (primary analysis; baseline samples; yellow nodes). Each node represents a single protein-coding gene locus, edges represent the inferred type of association between nodes; light-blue nodes, genes biologically associated with the differentially expressed genes (yellow nodes). Edges: a blue line, a direct interaction; a red line, phosphorylation reaction; a black line, biological association, which indicates interaction between molecules that may participate in formation of one physical complex, describing a set of molecules that are co-purified in a single pull-down or coimmunoprecipitation; a green line, biological physical association, indicating an interaction between molecules within the same physical complex, suggesting that the molecules are in close proximity but not necessarily in direct contact with each other. FC: fold change F I G U R E 4 Correlation between fold change values of the primary and validation analyses for the significant (pFDR < 0.05) differentially expressed genes (DEGs) identified in the two analyses, with regression line and 95% confidence intervals

*** **
A threshold of 3% in sputum is widely used to identify a phenotype characterized by eosinophilic inflammation and comprises approximately 20-40% of the whole COPD population. [14][15][16] This proportion is in line with our observations that patients with sputum eosinophils ≥ 3% accounted for approximately one third of the whole study population. We recently showed, in this cohort of patients, that blood eosinophils predict sputum eosinophilia with an accuracy of approximately 80%. 12 This association between blood and sputum eosinophils was also previously observed in other studies, 14,15 indicating that blood eosinophil counts can be a good surrogate marker of eosinophilic inflammation in sputum, which we show here is also associated with greater T2 inflammation.
We acknowledge that this analysis has some limitations. The limited sample size, in particular for the group of patients with higher sputum eosinophil counts, might have restricted the pool of genes significantly differential expressed preventing a complete biological differentiation. In addition, the use of complementary techniques (eg protein assessments) can add value to validate further gene expression findings.
In conclusion, recent studies showed that the effect of oral and inhaled PDE4 inhibitors on exacerbations in COPD patients with chronic bronchitis appears to be greater at higher blood eosinophil counts. 4,10 Furthermore, we recently showed that tanimilast significantly reduced sputum eosinophil numbers in the eosinophilhigh group. 12 These previous results coupled with our current data strongly suggest that differential responses to PDE4 inhibition may relate to an increased presence of features of T2 inflammation and PDE4-related pathways in the eosinophilic COPD phenotype.

ACK N OWLED G EM ENTS
The authors would like to thank Brendan Colgan (Celerion, UK) and all the investigators and patients at the investigative sites for their support of this study. Dave Singh is supported by the National

Institute for Health Research (NIHR) Manchester Biomedical
Research Centre (BRC).

DATA AVA I L A B I L I T Y S TAT E M E N T
Chiesi commits to sharing with qualified scientific and medical re-