Abstract
The effects of phosphodiesterase (PDE) 4 inhibitors on gene expression changes in BEAS-2B human airway epithelial cells are reported and discussed in relation to the mechanism(s) of action of roflumilast in chronic obstructive pulmonary disease (COPD). Microarray-based gene expression profiling failed to identify mRNA transcripts that were differentially regulated by the PDE4 inhibitor 6-[3-(dimethylcarbamoyl)benzenesulphonyl]-4-[(3-methoxyphenyl)amino]-8-methylquinoline-3-carboxamide (GSK 256066) after 1, 2, 6, or 18 hours of exposure. However, real-time polymerase chain reaction analysis revealed that GSK 256066 was a weak stimulus, and the negative microarray results reflected low statistical power due to small sample sizes. Furthermore, GSK 256066, roflumilast, and its biologically active metabolite roflumilast N-oxide generally potentiated gene expression changes produced by the long-acting β2-adrenoceptor agonists (LABAs) salmeterol, indacaterol, and formoterol. Many of these genes encode proteins with antiviral, anti-inflammatory, and antibacterial activities that could contribute to the clinical efficacy of roflumilast in COPD. RNA-sequencing experiments established that the sensitivity of genes to salmeterol varied by ∼7.5-fold. Consequently, the degree to which a PDE4 inhibitor potentiated the effect of a given concentration of LABA was gene-dependent. Operational model fitting of concentration-response curve data from cells subjected to fractional, β2-adrenoceptor inactivation determined that PDE4 inhibition increased the potency and doubled the efficacy of LABAs. Thus, adding roflumilast to standard triple therapy, as COPD guidelines recommend, may have clinical relevance, especially in target tissues where LABAs behave as partial agonists. Collectively, these results suggest that the genomic impact of roflumilast, including its ability to augment LABA-induced gene expression changes, may contribute to its therapeutic activity in COPD.
Introduction
Phosphodiesterase (PDE) 4 inhibitors entered clinical development in the 1980s as potential antidepressant drugs (Zeller et al., 1984) and, since that time, have suffered a high level of attrition due to a low therapeutic ratio and weak efficacy (Giembycz, 2008). Nevertheless, in April 2010, roflumilast became the first, selective, orally active PDE4 inhibitor to be approved for human use, with chronic obstructive pulmonary disease (COPD) being a primary indication (Giembycz and Field, 2010; Gross et al., 2010; Wedzicha et al., 2016). The 2019 Global Initiative for Chronic Obstructive Lung Disease guidelines recommend that roflumilast be used as an add-on therapy in a specific subgroup of patients with COPD. These are categorized as high-risk patients having severe, symptomatic disease in whom exacerbations occur despite regular treatment with a combination of a long-acting β2-adrenoceptor agonist (LABA), a long-acting muscarinic receptor antagonist, and an inhaled corticosteroid (ICS) (http://goldcopd.org). In this COPD phenotype, the therapeutic activity of roflumilast relies on its ability to improve airway caliber. However, PDE4 inhibitors do not promote acute bronchodilatation (Grootendorst et al., 2003), suggesting that the gain in lung function and associated reduction in exacerbation frequency are unrelated to direct airway smooth muscle relaxation. Instead, preclinical studies and trials of roflumilast in human subjects suggest a primary mode of action is to suppress inflammation (Gamble et al., 2003; Grootendorst et al., 2007; Hatzelmann et al., 2010; Moodley et al., 2013; Giembycz and Newton, 2014).
Another selective, orally active PDE4 inhibitor, apremilast, was approved in 2015 for the treatment of plaque psoriasis and psoriatic arthritis (Fala, 2015). Similar to the COPD phenotype for which roflumilast is indicated, these disorders are characterized by a chronic, systemic dysregulation of cytokine generation with attendant inflammation, implying that PDE4 inhibitors may share the same or a similar mechanism of action (Pincelli et al., 2018).
At a molecular level, inhibition of PDE4 increases the intracellular concentration of cAMP in target cells and tissues. Although the downstream signaling pathways that ultimately lead to improved clinical outcomes are ill defined, cAMP is known to modulate gene expression by activating a family of transcription factors of which cAMP response element (CRE)–binding protein and activating transcription factor-1 are prototypical examples (Zhang et al., 2005). Recently, we reported that the LABAs indacaterol and salmeterol promoted significant, and potentially beneficial, gene expression changes in BEAS-2B airway epithelial cells and human primary bronchial epithelia by mechanisms that involve canonical, Gsα/adenylyl cyclase/cAMP-dependent signaling (Yan et al., 2018). Therefore, logic dictates that PDE4 inhibitors may also provide clinical benefit by modulating gene expression (Tannheimer et al., 2012; Moodley et al., 2013; Giembycz and Maurice, 2014; BinMahfouz et al., 2015; Joshi et al., 2017). A genomic, anti-inflammatory mechanism of action also accommodates the likelihood that airway smooth muscle is but one of several tissues that are therapeutic targets of orally active PDE4 inhibitors. In this respect, the airway epithelium, which is considered a major player in COPD pathogenesis (Crystal, 2014), and extrapulmonary tissues including circulating leukocytes, vascular endothelium, and bone marrow are attractive additional candidates. Indeed, the need for systemic exposure may help explain why PDE4 inhibitors developed for inhaled administration have, without exception, failed in clinical trials of COPD.
In this study, we hypothesized that PDE4 inhibitors work, in part, by genomic mechanisms and interact in an additive or synergistic manner with LABAs. To test this idea, the transcriptomic signatures of two highly selective PDE4 inhibitors, roflumilast N-oxide (RNO), the active metabolite of roflumilast (Hatzelmann et al., 2010), and GSK 256066 (6-[3-(dimethylcarbamoyl)benzenesulphonyl]-4-[(3-methoxyphenyl)amino]-8-methylquinoline-3-carboxamide; Tralau-Stewart et al., 2011), were obtained in BEAS-2B human airway epithelial cells treated alone and concurrently with an LABA. In addition, the impact of PDE4 inhibitors on the operational efficacy, magnitude of response, and duration of action of LABA-induced gene expression changes was determined. BEAS-2B cells were selected for this investigation because they display gene expression profiles that mirror, to a large degree, those obtained in human primary bronchial epithelial cells treated with a variety of stimuli, including LABAs (Yan et al., 2018).
Materials and Methods
Drugs and Reagents.
GSK 256066, indacaterol, and β2A (8-hydroxy-5-((R)-1-hydroxy-2-methylaminoethyl)-1H-quinolin-2-one) were provided by Gilead Sciences (Seattle, WA). Salmeterol and formoterol were donated by GlaxoSmithKline (Stevenage, UK) and AstraZeneca (Mölndal, Sweden), respectively. Roflumilast and RNO were from Nycomed (Konstanz, Germany). DCITC (5(2-(((1′-(4′-isothiocyanatephenylamino)thiocarbonyl)amino)-2-methylpropyl)amino-2-hydroxypropoxy)-3,4-dihydrocarbostyril) was a gift from Dr. Stephen Baker (University of Florida). All drugs were dissolved in DMSO and diluted in serum-free medium (SFM). The highest concentration of DMSO used in these experiments (0.2%, v/v) did not affect any output measured.
Generation of a CRE Reporter in BEAS-2B Cells.
Cells were transfected with 8 µg of plasmid DNA (pADneo2-C6-BGL) using Lipofectamine 2000 (Invitrogen, Burlington, ON, Canada) to generate 6×CRE BEAS-2B luciferase reporter cells as described previously (Meja et al., 2004).
Submersion Culture of BEAS-2B Cells.
BEAS-2B cells were cultured under a 5% CO2/air atmosphere at 37°C in 12- or 24-well plastic plates (Corning Life Sciences, Lowell, MA) containing keratinocyte-SFM (Thermo Fisher Scientific, Burlington, ON, Canada) supplemented with epidermal growth factor (5 ng/ml), bovine pituitary extract (50 μg/ml), penicillin (100 mg/ml), and streptomycin (100 IU/ml). When confluent, cells were growth-arrested for 24 hours in keratinocyte-SFM without supplements (Greer et al., 2013) and processed as described later. For RNA-sequencing (RNA-seq) and subsequent validation experiments, BEAS-2B cells were grown in Dulbecco’s modified Eagle’s/Ham’s F12 medium containing 10% fetal bovine serum, 2.5 mM l-glutamine, and 14 mM NaHCO3 (all Invitrogen) until confluent and for a further 24 hours in SFM.
Measurement of Luciferase Activity.
6×CRE BEAS-2B reporter cells were treated with PDE4 inhibitor (GSK 256066, roflumilast, or RNO) or LABA (indacaterol, salmeterol, or formoterol) alone and in the combinations indicated in the text. After 6 hours, cells were lysed in 1× firefly luciferase buffer (Biotium, Hayward, CA), and luciferase activity was measured by luminometry. Data are expressed as fold increase in enzyme activity relative to vehicle-treated samples matched for time.
Western Blot Analyses and ELISAs.
Confluent BEAS-2B cells at 37°C were incubated with RNO (1 μM) and/or salmeterol (100 nM). At 60 minutes, the culture medium was decanted and cells were lysed in HCl (0.1 M). cAMP in the resulting lysates was measured by ELISAs (Enzo Life Sciences, Farmingdale, NY) according to the manufacturer’s instructions. Alternatively, cells were incubated with GSK 256066 and indacaterol alone and in combination at the concentrations indicated. At 6 hours, supernatants were collected and interleukin-6 (IL-6) was measured by ELISA (D6050; R&D Systems, Minneapolis, MN). Cells were lysed in 1× Laemmli buffer supplemented with phosphatase inhibitors (Sigma-Aldrich) and 1× complete protease inhibitor cocktail (Roche, Indianapolis, IN). The cell lysates were size fractionated on 10% acrylamide gels, electrotransferred onto reinforced 0.2-µM nitrocellulose membranes (GE Healthcare, Waukesha, WI), and blocked with 5% milk in Tris-buffered saline containing 1% Tween 20. Subsequently, membranes were probed with antibodies against NR4A2 (PP-N1404-00), NR4A3 (PP-H7833-00; both Perseus Proteomics Inc., Tokyo, Japan), and glyceraldehyde-3-phosphate dehydrogenase (MCA4739; Bio-Rad, Hercules, CA). After washing, membranes were incubated with horseradish peroxidase–conjugated, anti-mouse immunoglobulin (115-035-003; Jackson ImmunoResearch Laboratories Inc., West Grove, PA). Proteins were detected by chemiluminescence using SuperSignal West Pico PLUS chemiluminescent substrate (#34580; Thermo Fisher Scientific), visualized by autoradiography, and expressed relative to glyceraldehyde-3-phosphate dehydrogenase. Preliminary studies verified the identity of NR4A2 and NR4A3 by gene silencing (data not shown).
Measurement of Gene Expression by Real-Time Polymerase Chain Reaction.
BEAS-2B cells were treated with PDE4 inhibitor and/or LABA as described earlier. Total RNA was extracted (RNeasy Mini Kit; Qiagen Inc., Mississauga, ON, Canada) and reverse transcribed using a qscript cDNA synthesis kit according to the manufacturer’s instructions (Quanta Biosciences, Gaithersburg, MD). Real-time polymerase chain reaction (PCR) analysis of cDNA was performed using the primer sequences shown in Supplemental Table 1 as described previously (Joshi et al., 2017; Yan et al., 2018).
Gene Expression Profiling by Microarray.
BEAS-2B cells were cultured for 1, 2, 6, and 18 hours (N = 4 at each time point) with vehicle, GSK 256066 (10 nM), and a concentration of indacaterol (10 nM) that maximally activated 6×CRE reporter cells (Supplemental Fig. 1). Cells were also treated with indacaterol and GSK 256066 in combination (Ind+GSK, both 10 nM) under identical conditions. Total RNA was extracted (vide supra) and processed for gene expression profiling (Yan et al., 2018). The microarray images were scaled and normalized using the probe logarithmic intensity error algorithm in Transcriptome Analysis Console (version 4.0; Affymetrix, Santa Clara, CA) and stored as .chp files. Signals from the four replicates for each probe set were averaged, and the relative expression patterning was implemented in Transcriptome Analysis Console. At each time point, data from all treatments were analyzed concurrently and visualized by generating volcano plots. The P statistic was adjusted using the Benjamini and Hochberg false discovery rate (FDR), with step-up procedure, and significance was set to the <0.1, <0.05, and <0.01 probability levels as indicated.
Gene Expression Profiling by RNA-seq.
BEAS-2B cells were treated for 2 hours with vehicle, RNO (1 µM), and two submaximal concentrations of salmeterol [0.3 nM (Salm0.3) and 0.5 nM (Salm0.5)] alone and in the presence of RNO (1 µM). A maximally effective concentration of salmeterol [100 nM (Salm100)] was examined in parallel to define maximum responses. Total RNA was extracted as described earlier, and a total of 28 samples (N = 4 per treatment group) were submitted to the Centre for Health Genomics and Informatics, University of Calgary, for sequencing.
RNA sequencing libraries were prepared using the NEBNext Ultra II Directional kit (New England Biolabs, Ipswich, MA) with the poly(A) mRNA magnetic isolation module as described by the manufacturer. The libraries were validated by using the D1000 Screen Tape assay on an Agilent 2200 TapeStation system (Agilent Technologies, Santa Clara, CA) and quantified with a Kapa qPCR Library Quantification kit for Illumina (Kapa Biosystems, Boston, MA). The libraries were pooled and sequenced across two consecutive, single-end, 75-cycle sequencing runs on a NextSEq 500 instrument (Illumina, San Diego, CA) according to the manufacturer’s instructions, generating approximately 33 million reads per sample.
Demultiplexing of the sequencing data and read quality of each sample was performed using bcl2fastq conversion software (version 2.18.0.12; Illumina) and FastQC (version 0.10.1, Babraham Bioinformatics, Babraham Institute, Cambridge, UK), respectively. Good-quality reads were mapped to the reference human transcriptome (GRCh37/hg19 version) and quantified using Kallisto (version 0.42.4) (Bray et al., 2016) with 100 bootstraps per sample. Differential expression analysis was performed at the transcript and gene level using the R package Sleuth (version 0.30.0) (Pimentel et al., 2017). Pairwise comparisons were performed between vehicle-treated and salmeterol-treated (100 nM) samples, and differentially expressed genes (DEGs) were identified based on an FDR-corrected P value of ≤0.05. Induced (≥2-fold) and repressed (≤0.5-fold) genes were filtered to remove those expressed at ≤1 transcript per million after Salm100 and vehicle, respectively, before subsequent analyses. Pairwise comparisons of these DEGs were performed between vehicle and all other treatments using the Wald test in Sleuth to estimate fold changes (derived from β values). Data were also expressed as a change in transcript per million as indicated.
Analysis of Gene Expression Profiles.
The microarray and RNA-seq data have been deposited with the National Center for Biotechnology Information’s Gene Expression Omnibus and are freely available using accession codes GSE106710 and GSE126981, respectively. Unless stated otherwise, genes are referred to by the official Human Genome Nomenclature Committee symbols supplied by the National Center for Biotechnology Information (www.ncbi.nlm.nih.gov). Functional annotation clustering of indacaterol- and Ind+GSK-regulated genes including associated enriched gene ontology (GO) terms was performed with the Database for Visualization and Integrated Discovery (DAVID) bioinformatics resources (version 6.8) at medium stringency (Huang et al., 2009). Results are reported using the GO term that describes the biologic process (GOTERM_BP_DIRECT). When this descriptor was absent from a given gene cluster, molecular function (GOTERM_MF_DIRECT), cellular component (GOTERM_CC_DIRECT), UniProt sequence feature, UniProt keyword, InterPro, or Kyoto Encyclopedia of Genes and Genomes pathway was reported. Pseudogenes, hypothetical genes, noncoding RNAs, and uncharacterized sequences lacking annotation were excluded from all analyses.
Curve Fitting.
Monophasic E/[A] curves were fit by least-squares, nonlinear, iterative regression to the following form of the Hill equation (eq. 1; Prism 6; GraphPad Software Inc., San Diego, CA):(1)where E is the pharmacological effect; Emin and Emax are the basal response and maximum response, respectively; p[A] is the negative log molar concentration of the compound of interest; p[A]50 is a location parameter equal to the negative log molar concentration of compound producing (Emax − Emin)/2; and nH is the Hill coefficient of the E/[A] curve at the p[A]50 level.
Applying Fractional Receptor Depletion to Quantify the Impact of PDE4 Inhibition on the Efficacy of β2-Adrenoceptor Agonists.
E/[A] curves were constructed for salmeterol and β2A, a quinolinone-based orthostere (Yoshizaki et al., 1976) present in many β2-adrenoceptor agonists, in the absence and presence of a PDE4 inhibitor. These experiments were performed in cells that had been pretreated (60 minutes) with vehicle or the alkylating agent DCITC (100 nM) and then washed in SFM (Deyrup et al., 1998). Each family of E/[A] curves was fit simultaneously to the operational model of agonism (eq. 2), which describes a theoretical relationship between E and agonist concentration (Black and Leff, 1983). Algebraically:(2)where Em is the theoretical maximum response of the tissue, KA is the agonist equilibrium dissociation constant, [A] is agonist concentration, n is the slope of the relationship between the concentration of agonist-receptor complexes and response, and τ is the operational efficacy of the agonist. In these analyses, only τ was allowed to vary between individual E/[A] curves; for all other parameters (i.e., Em, KA, and n), a common value was assumed (Black and Leff, 1983; Leff et al., 1990).
Determination of Receptor Reserve.
Occupancy-response curves in the absence and presence of a PDE4 inhibitor were constructed using KA values determined by β2-adrenoceptor depletion. At each concentration of agonist, fractional receptor occupancy was determined assuming that the binding of ligand to the β2-adrenoceptor was a noncooperative process (eq. 3), where RA and Rt represent the number of agonist-occupied receptors and total number of receptors, respectively:
(3)Statistics.
Data are presented as the mean ± S.E.M. or as box and whisker plots of N independent determinations. Differences in CRE reporter activity and gene expression changes were evaluated by using Student’s t test or repeated-measures, one-way ANOVA as indicated. When the ANOVA F test P value was <0.05, differences between groups were identified by using Tukey’s multiple comparison test without Greenhouse-Geisser correction (Lew, 2007). The relationship between global changes in gene expression produced by two treatment interventions was determined by Pearson product moment correlation. Least-squares perpendicular major axis (Deming) regression (Cornbleet and Gochman, 1979) was used to verify differences between treatments or treatment methods. In the text, the terms synergy and synergistic refer to a change in gene expression produced by a combination of an LABA and a PDE4 inhibitor that is greater than the sum of their individual effects. The null hypothesis was rejected when P < 0.05.
Results
Effect of GSK 256066 on Global Gene Expression Changes.
The microarray results of GSK 256066–treated BEAS-2B cells are displayed as volcano plots in Supplemental Fig. 2, where induced and repressed genes are represented as red circles (>1.5-fold) and blue circles (<0.67-fold), respectively. Although many gene expression changes were apparent and exceeded P values of 0.05 obtained by ANOVA, no probe set on the array was significantly different from vehicle at a Benjamini and Hochberg FDR of <10%. Previously, we reported that PDE4 inhibitors can potentiate LABA-induced gene expression in BEAS-2B cells, suggesting that a common pool of cAMP may regulate transcription (Moodley et al., 2013; BinMahfouz et al., 2015). Thus, the results reported here may indicate that GSK 256066 is a weak stimulus in these cells, and that the small sample size lacked statistical power. To address this prospect, indacaterol (10 nM)-regulated gene expression changes at 1, 2, and 6 hours were correlated against their counterparts in GSK 256066–treated BEAS-2B cells using FDRs of <10%, <5%, and <1% (Fig. 1A). This approach revealed significant correlations between the two drugs, which strengthened as the stringency of the FDR was increased. Furthermore, slopes of Deming regressions were shallow (0.23–0.36) at all time points, which is consistent with weak transcriptional activity of GSK 256066 relative to indacaterol. Using the same RNA, the expression of 18 genes (labeled A–R) that were significantly (FDR <5%) upregulated by indacaterol at 2 hours (yellow circles in Fig. 1; Supplemental Fig. 2) was tested to determine their sensitivity to GSK 256066 by real-time PCR. These results are shown in Supplemental Fig. 3, where the expression of each gene is plotted over time together with data from the probe set that gave the most robust increase on the microarray. Pearson’s analyses indicated significant correlations between the microarray- and PCR-generated data at 1, 2, and 6 hours (Fig. 1B), although changes in gene expression were modest and did not reach statistical significance (Supplemental Fig. 3). However, repeating this experiment with a larger sample size established that most of these genes were significantly upregulated in BEAS-2B cells exposed to GSK 256066 (10 nM) and roflumilast (1 µM) for 1 and/or 2 hours (Supplemental Fig. 4).
Effect of GSK 256066 on Expression of the Indacaterol-Regulated Transcriptome.
Microarrays were also used to determine the effect of GSK 256066 (10 nM) on the number of genes that were significantly (FDR <10%) induced or repressed at 1, 2, 6, and 18 hours (Fig. 2A) by a concentration of indacaterol (10 nM) that maximally activated 6×CRE reporter cells (Supplemental Fig. 1). Relative to time-matched, vehicle-treated cells, 181 DEGs were either upregulated (135) or downregulated (46) at 1 hour by >1.5-fold and <0.67-fold, respectively. At 2 hours, the number of DEGs had increased to 351 (304 induced and 47 repressed) and had declined to 251 (206 induced and 45 repressed) by 6 hours. At 18 hours, nine genes were upregulated and 1 gene was downregulated by indacaterol.
In the presence of GSK 256066, the number of indacaterol-regulated genes at 1 and 2 hours was similar to the gene count in BEAS-2B cells exposed to indacaterol alone (Fig. 2A). In contrast, at 6 and 18 hours, the number of significant gene expression changes was considerably greater in cells exposed to Ind+GSK than those exposed to indacaterol (Fig. 2A). The Venn diagrams in Fig. 2B show that 41%, 87%, 86%, and 60% of all genes induced or repressed by indacaterol were similarly regulated by Ind+GSK at 1, 2, 6, and 18 hours, respectively. The corresponding values for genes induced or repressed by Ind+GSK that were also regulated by indacaterol were 56%, 86%, 55%, and 30%. Analyses of data at all time points revealed that ≥70% of all genes were regulated by both interventions (Fig. 2B).
Ontological Analysis and Functional Annotation Clustering.
Categorization of indacaterol- and Ind+GSK-regulated transcripts (by probe set) at 1, 2, 6, and 18 hours was performed manually using six generic descriptors: (1) transcriptional regulators (red); (2) transporters, ion channels, and membrane receptors (orange); (3) metabolic proteins (yellow); (4) general signaling molecules, including translational regulators (green); (5) other functions (blue); and (6) not assigned (purple). Annotated probe sets meeting the expression criteria (>1.5-fold, <0.67-fold; FDR < 10%) were assigned Human Genome Nomenclature Committee gene symbols and are listed in Supplemental Tables 2 and 3. The number of induced and repressed genes within each general term expressed as a percentage of the total number of significant gene expression changes is also presented as a pie chart at each time point for each intervention.
To explore how downstream function may change with time, the number of DEGs assigned to each of the six terms described earlier was enumerated at 1, 2, 6, and 18 hours (Supplemental Fig. 5). Apart from repressed genes assigned to the term receptors, ion channels, and transporters, the gene count in each of the other categories mirrored the global effect of GSK 256066 on the indacaterol-regulated transcriptome (Fig. 2A). Breaking down these changes by category showed that indacaterol promoted a transient burst (peaking at 2 hours) in induced and repressed genes that are associated with transcriptional regulation and signaling. At 1, 2, and 18 hours, the gene count was not materially affected by GSK 256066, whereas at 6 hours it was markedly increased (Supplemental Fig. 5). Indacaterol-regulated gene expression changes in the other categories occurred more slowly, with the highest number recorded at 6 hours. Again, GSK 256066 did not affect the number of DEGs at the early or late time points; only those appearing at 6 hours were increased in number (Supplemental Fig. 5).
Functional annotation clustering of genes that were differentially expressed in response to indacaterol and Ind+GSK across all time points determined that the numbers of clusters and enriched GO terms within each cluster were similar (Supplemental Tables 4 and 5). Many of the most highly enriched terms, such as positive regulation of transcription from RNA polymerase II promoter (GO:0045944), relate to gene transcription and contain transcripts that encode sequence-specific transcription factors, coactivators, (co)repressors, and allied regulators of gene expression. Of these, several encode transcriptional repressors, including KLF2, KLF4, KLF15, NR4A1, NR4A2, and NR4A3. Other enriched GO terms, such as transcription, DNA templated (GO: 0006351); extracellular region (GO:0005576); integral component of plasma membrane (GO:0005887); peptidyl-threonine dephosphorylation (GO:0035970); and type I interferon signaling pathway (GO:0060337), contained genes that may attenuate cytokine production (CD200, DUSP1, SOC3), protect against COPD exacerbations (DMBT1, CRISPLD2), and regulate oxidative stress, fibrosis, and mucus secretion (EGR1, TXNIP).
Enriched GO and Kyoto Encyclopedia of Genes and Genomes terms including extracellular space (GO:0005615), cell-cell signaling (GO:0007267), cytokine activity (GO:0005125), positive regulation of tyrosine phosphorylation of stat3 protein (GO:0042531), and tumor necrosis factor signaling pathway (HSA:04688) were populated with a variety of adverse effect (AE) genes, notably AREG, BDNF, CCL2, CCL20, CXCL2, CXCL3, CTGF, EDN1, IL6, IL11, IL15, and IL20.
A comprehensive analysis of indacaterol-regulated transcripts in BEAS-2B cells including functional annotation clustering has been reported previously (Yan et al., 2018), and those findings were largely replicated here in BEAS-2B cells treated with Ind+GSK (Supplemental Tables 2–5).
Effect of GSK 256066 on the Duration of Indacaterol-Induced Gene Expression Changes.
The global effect of PDE4 inhibition on the expression of the indacaterol-regulated transcriptome was further interrogated by comparing the magnitudes of all significant gene expression changes (by probe set) induced by Ind+GSK at 1, 2, 6, and 18 hours (no fold threshold at the FDR indicated) with the corresponding indacaterol data (Fig. 3). This analysis revealed significant correlations between the two treatments at each time point. At 1 and 2 hours, GSK 256066 had little effect on those genes upregulated by indacaterol (slopes of Deming regressions ∼1), whereas at the two later time points, slopes of regressions were shallow in favor of Ind+GSK (Fig. 3, C and D). The overall effect of this interaction is presented in Fig. 3, E–H as areas under the curves of the mean change in gene expression over the total period of exposure (AUC0–18h). For example, taking the 158 probe sets that were significantly upregulated by Ind+GSK at 1 hour; plotting the overall mean induction of the same probe sets at 2, 6, and 18 hours; and comparing these results with the indacaterol counterparts (Fig. 3E) revealed that GSK 256066 produced an overall enhancement of gene induction. Similar data were obtained for those genes significantly upregulated by Ind+GSK at the other time points (Fig. 3, F–H). These outcomes were attributable to the persistence of gene expression changes at 1 and 2 hours, as well as the induction of a greater number of genes at 6 and 18 hours. In each case, the AUC0–18h values for the 1-, 2-, 6-, and 18-hour data sets were 23%, 21%, 20%, and 28% greater, respectively, in cells treated with Ind+GSK compared with indacaterol alone (Fig. 3, E–H). Indacaterol-repressed genes were affected similarly (Supplemental Fig. 6).
The impact of GSK 256066 on the AUC0–18h of the 259 probe sets that were significantly induced (>2-fold; FDR <10%) by indacaterol at any time point was converted to a fold change and presented as a heat map. As shown in Fig. 4, the effects of PDE4 inhibition were gene-dependent and formed a continuum that ranged from 2.5-fold for DNAI1 to 0.21-fold for UGT1A8/9. Using cutoff levels of >1.1-fold and <0.9-fold, GSK 256066 variably increased and decreased, respectively, the AUC0–18h of 157 (61%) and 34 (13%) indacaterol-induced transcripts. The AUC0–18h values of the remaining 68 (26%) transcripts were unaffected (Fig. 4).
The effects of GSK 256066 on 18 indacaterol-induced genes (labeled A to R in Fig. 4) that spanned the AUC0–18h continuum were validated by real-time PCR using the same RNA (Fig. 5). For the majority (13/18) of these genes, GSK 256066 variably enhanced the indacaterol AUC0–18h by maintaining transcript expression at the 6- and 18-hour time points. However, the kinetics of other gene expression changes (i.e., C5AR1, CRISPLD2, DMBT1, SOCS3) were unaffected by GSK 256066 or even abbreviated (e.g., BMP2), which suggests gene-dependent differences in regulation by cAMP.
Effect of PDE4 Inhibition on the Magnitude of Gene Expression Changes Produced by a Submaximal Concentration of an LABA.
The experiments described in the previous section explored the impact of RNO in cells treated with a maximally effective concentration of indacaterol (10 nM; Supplemental Fig. 1). As this may have precluded an assessment of whether these drugs could interact in an additive or synergistic manner, the effects of a lower concentration of indacaterol (1 nM; [A]32) on the expression of nine genes were determined in the absence and presence of GSK 256066 (10 nM) and roflumilast (1 μM). In most cases, at 1 and/or 2 hours, the effect of the drugs in combination was greater than the LABA alone (Fig. 6A). When changes in expression produced by the drugs in combination were plotted against the sum of their individual effects, lines of Deming regressions deviated from unity, raising the possibility that the PDE4 inhibitor and LABA interacted synergistically (Fig. 6B). A similar and more striking interaction occurred when indacaterol was substituted with a higher effective concentration of formoterol (30 pM; [A]45; Fig. 6, A and B). These data are also presented as box and whisker plots to illustrate the variability in response to LABA and PDE4 inhibitor (Supplemental Fig. 4). The ability of GSK 256066 to enhance the effect of indacaterol was reproduced at the protein level using NR4A2, NR4A3, and IL-6 as representative examples (Fig. 6C).
The Salmeterol-Regulated Transcriptome and the Effect of RNO.
RNA-seq was used to establish if PDE4 inhibitors augmented the expression of all LABA-regulated genes or a subpopulation. For these experiments, RNO and salmeterol were substituted for GSK 256066 and indacaterol to provide clinical applicability and to gain further evidence that this interaction represents a class effect of LABAs and PDE4 inhibitors. Initially, the sensitivity of genes that comprise the LABA-regulated transcriptome to agonist was determined. This was achieved by comparing global gene expression changes in BEAS-2B cells treated for 2 hours with two submaximal concentrations of salmeterol (Salm0.3 and Salm0.5), which equated to [A]14 and [A]36 on the 6×CRE reporter, respectively, with a concentration of salmeterol (Salm100) that defined maximal responses (Supplemental Fig. 1).
At an FDR of ≤5%, 180 and 16 genes were significantly induced (≥2-fold) and repressed (≤0.5-fold), respectively, by Salm100. Changes in gene expression produced by Salm0.3 expressed as a percentage of the corresponding Salm100 data formed a continuum that ranged from 11.6% (NPTX1) to 83% (TCF21) for upregulated genes (Fig. 7A) and from 37.6% (C10orf10) to 71.6% (KRTAP2-3/KRTAP2-4) for genes that were repressed (Supplemental Table 7). Assuming that: 1) salmeterol is a full agonist on all DEGs, 2) gene expression is described by symmetrical E/[A] curves with nH = 1.8 (Supplemental Fig. 1), and 3) Salm100 maximally induced or repressed all DEGs, then the sensitivity of genes within the salmeterol-regulated transcriptome varied by ∼7.5-fold (Fig. 7, A and B). In contrast, Salm0.5 was a strong stimulus in BEAS-2B cells and, unlike its modest effect on the 6×CRE reporter (Supplemental Fig. 1), promoted gene expression changes that were ≥58% of their respective maxima (Fig. 7A).
In BEAS-2B cells treated for 2 hours with RNO (1 μM), 16 genes were differentially regulated (≥2-fold, ≤0.5-fold; FDR ≤ 5%), consistent with the superior sensitivity of RNA-seq over microarrays. Analysis of the 196 Salm100-regulated genes (Supplemental Table 7) revealed strong correlations between gene expression changes produced by RNO and all concentrations of salmeterol tested (Supplemental Fig. 7). The additional finding that RNO augmented salmeterol-induced cAMP accumulation (Fig. 7C) implies that both drugs can regulate gene expression by a common mechanism that involves the activation of cAMP-dependent protein kinase (Yan et al., 2018). To explore that possibility, gene expression changes produced by salmeterol and RNO in combination and the sum of their individual effects were analyzed by Deming regression (Fig. 8, A and B). On most genes, the activity of salmeterol was augmented by RNO (1 μM) in a synergistic manner, which was reflected by slopes that were steeper than the line of identity in favor of Salm0.3 + RNO (1.98) and Salm0.5 + RNO (1.24; Fig. 8, A and B). However, the magnitudes of these interactions varied and formed a continuum due to differences in the sensitivity of genes to salmeterol (Supplemental Tables 7 and 8). This is illustrated in Fig. 8, C and D, which shows simulated salmeterol E/[A] curves with nH fixed to a value of 1.8 (vide supra) in the absence and presence of a concentration of RNO that displaced the control curve 3-fold to the left. It can be seen that the degree to which RNO augments a given response depends on where the measurement is made on the salmeterol E/[A] curve and how it is calculated. On NPTX1 and TCF21, which lie toward the extremes of the salmeterol sensitivity spectrum (Fig. 7A), the impact of RNO differed markedly (Fig. 8, E and F). Thus, RNO augmented Salm0.3-induced NPTX1 and TCF21 expression from 11.6% to 51% and from 83% to 97% of their maximum responses, respectively.
Quantifying the Impact of RNO on the Efficacy of Salmeterol.
Pretreatment of 6×CRE BEAS-2B cells with RNO (1 μM; 30 minutes) produced a modest activation of the reporter (<2-fold) and a 4.5-fold sinistral displacement of the mean salmeterol E/[A] curve without affecting the maximum response (Fig. 9A). To mimic a therapeutic target where receptor number is limiting, the effect of RNO was examined in cells subjected to fractional, irreversible β2-adrenoceptor inactivation with DCITC (100 nM; 60 minutes). In these experiments, the upper asymptote of the mean E/[A] curve was significantly depressed (by 49%), and the potency of salmeterol was reduced by a factor of 8.5-fold. In the presence of RNO, the effects of receptor depletion were partially rescued; there was an increase in the potency of salmeterol and in the maximum response attained (Fig. 9A). Analyzing this family of E/[A] curves by operational model fitting determined that RNO had doubled the efficacy of salmeterol in the absence (τ: from 10.5 to 24) and presence (τ′: from 0.9 to 1.7) of DCITC (Table 1).
The operational model also provides a measure of agonist affinity. For salmeterol-induced reporter activation, this was calculated to be 3.7 nM (Table 1). Substituting this value in eq. 3, which is a statement of the law of mass action, provides a description of the relationship between receptor occupancy and response (Kenakin, 2016). For activation of 6×CRE BEAS-2B reporter cells, this relationship deviated significantly from the line of identity (Fig. 9B). Interpolation from the mean occupancy-response curve showed that 4%, 13%, and 25% receptor occupancy was required to generate 20%, 50%, and 80% of the maximal response, respectively, and is consistent with a receptor reserve. In the presence of RNO, the deviation from linearity was more pronounced; the salmeterol KA/[A]50 value was increased from 8 to 32, and the generation of 20%, 50%, and 80% of the maximal response now required only 1%, 4%, and 13% β2-adrenoceptor occupancy, respectively (Fig. 9B).
In cells treated with DCITC, salmeterol-induced reporter activation collapsed to a linear function of receptor occupancy. The KA/[A′]50 value was ∼1, and the receptor reserve present under control conditions was lost (Fig. 9B). In contrast, the sensitivity of receptor-depleted cells treated with RNO to salmeterol was partially restored. The occupancy-response relationship returned to a shallow rectangular hyperbola, where 15%, 43%, and 77% binding generated 20%, 50%, and 80% of the maximal response, respectively, and the KA/[A′]50 value was increased from 0.74 to 1.59 (Fig. 9B; Table 1).
GSK 256066 (10 nM; 30 minutes) had a similar impact on the operational efficacy and receptor reserve of β2A (Fig. 9, C and D; Table 1), the functionality that confers β2-adrenoceptor agonism (Yoshizaki et al., 1976) in indacaterol, carmoterol, and abediterol. Thus, the interaction between salmeterol and RNO shown in Fig. 9A is likely to be generic to LABAs and PDE4 inhibitors.
On bona fide genes (i.e., CRISPLD2, NR4A2, RGS2), RNO produced sinistral displacements of the salmeterol E/[A] curves in the absence and presence of DCITC, consistent with increases in efficacy and receptor reserves (Fig. 9, E–G). However, the quality of the data was not sufficiently robust for quantification by operational model fitting.
Discussion
The results of large-scale, phase III clinical trials indicate that the PDE4 inhibitor roflumilast is beneficial in a subset of individuals with severe, bronchitic COPD in whom frequent exacerbations occur despite ICS/LABA combination therapy, even in the presence of a long-acting muscarinic receptor antagonist. These results are important because they illustrate that the ceiling of benefit, following an additional drug intervention, had not been attained (Martinez et al., 2015, 2018). While the mechanism of action of roflumilast is unclear, its ability to improve lung function and reduce exacerbation frequency (Wedzicha et al., 2016), in the absence of direct bronchodilatation (Grootendorst et al., 2003), implies that suppression of inflammation plays a role (Gamble et al., 2003; Grootendorst et al., 2007). We have reported previously that LABAs promote changes in gene expression in human airway epithelial cells that may contribute to: 1) their clinical efficacy in obstructive lung diseases, especially when combined with an ICS (Giembycz and Newton, 2011, 2015; Rider et al., 2018); and 2) the AEs that are associated with chronic β2-adrenoceptor agonist monotherapy (vide infra; Yan et al., 2018). This study extended those findings by establishing that PDE4 inhibitors also promoted changes in gene expression in BEAS-2B cells and, perhaps more importantly, increased the operational efficacy, enhanced the magnitude of response, and variably altered gene expression kinetics induced by LABAs. Thus, if suppression of airway inflammation underpins the clinical activity of PDE4 inhibitors in COPD, genomic mechanisms may be involved.
The PDE4-Regulated Transcriptome.
Immediate and delayed targets of cAMP signaling could equally contribute to the efficacy of PDE4 inhibitors. Accordingly, DEGs were identified in BEAS-2B cells exposed to GSK 256066 for 1, 2, 6, and 18 hours. GSK 256066 was chosen for this experiment because it is a potent, pan-PDE4 inhibitor with considerable selectivity (>30,000-fold) over all other PDE families and the Cerep panel of receptors (Tralau-Stewart et al., 2011). At a concentration 3800× greater than its KI for the inhibition of human PDE4B1 (Joshi et al., 2017), GSK 256066 did not affect gene expression even at a FDR <10%. While these data suggested that genomic mechanisms play little role in the mechanism of action of PDE4 inhibitors, further analyses ascertained that GSK 256066 was a weak stimulus in BEAS-2B cells, and the small sample size used for the arrays lacked statistical power to detect small changes in gene expression. Indeed, real-time PCR confirmed that many genes induced by indacaterol were indeed upregulated by GSK 256066 and roflumilast.
The weak transcriptional activity of PDE4 inhibitors may reflect low basal adenylyl cyclase activity in BEAS-2B cells and questions the significance of genomic mechanisms in vivo. This is an important consideration given that roflumilast monotherapy was beneficial in clinical trials of COPD (Calverley et al., 2009). However, the modest in vitro effects reported here are likely amplified in vivo because adenylyl cyclase activity in target cells and tissues will be higher. Indeed, many Gs-coupled receptors are under tonic activation by various endogenous ligands, including catecholamines, adenosine, and prostaglandins (Kaur et al., 2008; Wilson et al., 2009; Greer et al., 2013; Moodley et al., 2013; BinMahfouz et al., 2015). Furthermore, any genomic effects of PDE4 inhibitors could be further enhanced by endogenous glucocorticoids given that these drugs can often summate or even synergize at a transcriptional level (Moodley et al., 2013; BinMahfouz et al., 2015). Thus, the clinical efficacy of roflumilast monotherapy may reflect its ability to enhance the activity of various endogenous ligands to produce a more robust gene expression signature than these in vitro data suggest.
Effect of PDE4 Inhibition on the LABA-Regulated Transcriptome.
Consistent with this in vivo prediction, RNO and GSK 256066 enhanced the expression of a panel of formoterol- and indacaterol-induced genes in BEAS-2B cells, and this was reproduced at the protein level using NR4A2, NR4A3, and IL-6 as representative examples. Operational model fitting determined that PDE4 inhibition doubled the efficacy of LABAs and, thereby, increased the β2-adrenoceptor reserve. Clinically, this finding could be described as “LABA sparing” where, in the presence of a PDE4 inhibitor, a lower agonist concentration is required to produce the same degree of gene induction or repression. To establish if this interaction extended to the LABA-regulated transcriptome or just a subpopulation of DEGs, the effects of RNO on global gene expression changes produced by submaximal concentrations of salmeterol were determined by RNA-seq. In these experiments, RNO augmented the expression of the majority of DEGs. These included those with AE and therapeutic potential (vide infra), although the magnitude of effect varied because of the estimated 7.5-fold difference in their sensitivity to salmeterol. This was defined by NPTX1 and TCF21, which lied toward the extremes of the sensitivity spectrum. Thus, Salm0.3 alone and in the presence of RNO equated to [A]11.6 and [A]51 for NPTX1 and [A]83 and [A]97 for TCF21, respectively. Collectively, these results support the idea that Gs-dependent signaling plays a dominant role in regulating β2-adrenoceptor–mediated gene expression. While noncanonical mechanisms cannot be excluded (see Penn et al., 2014), cis-acting CREs for the transcription factor cAMP response element-binding protein are found in the promoter regions of a large number of cAMP-regulated genes, which supports this proposal (Zhang et al., 2005). An additive or synergistic interaction of a PDE4 inhibitor with an LABA may be particularly important in target cells and tissues that express low β2-adrenoceptor numbers or where receptor-effector coupling efficiency is weak (Rabe et al., 1993). Indeed, modeling this scenario by rendering salmeterol and β2A (the functionality that confers agonism in many LABAs) partial agonists with DCITC revealed that PDE4 inhibitors partially rescued the associated loss in operational efficacy, increased the maximum response, and produced sinistral displacements of LABA E/[A] curves as the law of mass action predicts.
GSK 256066 also prolonged the duration of many indacaterol-induced gene expression changes. This was prominent at 6 hours and often persisted to 18 hours when the level of mRNA transcripts induced or repressed by Ind+GSK was still greater than with indacaterol alone. A simple explanation of these data is that by maintaining the cAMP signal, PDE4 inhibitors likewise sustain gene transcription. However, GSK 256066 did not prolong the expression of all LABA-induced genes equally; in fact, for some genes, the AUC0–18h was unaffected or even decreased by GSK 256066. Thus, additional mechanisms that accommodate variable effects of PDE4 inhibition on gene expression kinetics must be entertained. One possibility is that the initial and delayed components of gene induction are regulated by temporally distinct transcriptional events that involve feed forward loops (Mangan and Alon, 2003). This type of regulation has been described for many genes, including DUSP1 (Lu et al., 2008), and can equally explain both prolongation and retardation of gene expression kinetics.
Genes with Therapeutic Potential.
The LABA- and LABA+PDE4 inhibitor–regulated transcriptomes in BEAS-2B cells contained genes that may help reduce exacerbation frequency and the associated inflammation (Perera et al., 2007) that characterizes the COPD phenotype that responds to roflumilast. In particular, several upregulated genes encode proteins that suppress proinflammatory cytokine generation, such as CD200, DUSP1, and SOCS3 (Liu et al., 2007; Snelgrove et al., 2008; Yoshimura et al., 2018), whereas others may directly protect against bacterial and viral exacerbations. For example, the multifunctional, secreted protein DMBT1 binds Gram-positive and Gram-negative bacteria and so could defend against microbial pathogens; it also inhibits the infectivity of human influenza A and immunodeficiency viruses (Madsen et al., 2010). Likewise, CRISPLD2 encodes a secreted lipopolysaccharide-binding protein (Wang et al., 2009) that can neutralize the pathogenicity of Gram-negative bacteria and, thereby, downregulate TLR4-mediated inflammation (Zhang et al., 2016). LABAs and PDE4 inhibitors also induced several genes that may encode negative feedback regulators of inflammation that include KLF2, KLF4, KLF15, and the NR4A family of transcription factors (Rodríguez-Calvo et al., 2017; Sweet et al., 2018). Many of these genes are further upregulated when an LABA and/or a PDE4 inhibitor are combined with a glucocorticoid (Moodley et al., 2013; Rider et al., 2018), which may be relevant to understanding how these drugs work in a clinical setting (Giembycz and Newton, 2014).
Noteworthy DEGs downregulated by indacaterol and Ind+GSK include EGR1. This gene is induced by cigarette smoke and is elevated in the lungs of subjects with severe COPD (Ning et al., 2004) with potential to promote mucus hypersecretion (Wang et al., 2017), inflammation, fibrosis, and remodeling in the airways (Lee et al., 2004; Cho et al., 2006). TXNIP is another cigarette smoke–sensitive gene (Sun et al., 2018) that may contribute to the free radical burden that often prevails in COPD airways (Domej et al., 2014) by encoding an inhibitor of the oxidoreductase, thioredoxin 1 (Nishiyama et al., 1999). Therefore, the ability of PDE4 inhibitors and LABAs to repress EGR1 and TXNIP could help arrest proinflammatory and fibrotic changes in the lungs and reduce oxidative stress, respectively.
Adverse-Effect Genes.
Chronic β2-adrenoceptor agonist monotherapy in asthma is associated with an increased risk of serious AEs (Pearce et al., 1991; Cates et al., 2014). The apparent absence of similar toxicity in subjects with COPD is, therefore, a paradox (Decramer et al., 2013). If genomic mechanisms contribute to these AEs in asthma as previously proposed (Ritchie et al., 2018; Yan et al., 2018), then how can these discrepant responses to treatment be rationalized? This question is pertinent given that the expression of putative AE genes was exaggerated by a PDE4 inhibitor. The profound differences in etiology, pathogenesis, inflammation, and causes of exacerbations between the two diseases may provide an explanation (Decramer et al., 2013) such that these gene expression changes in COPD have less pathologic relevance than in asthma. Alternatively, cells exposed chronically to cigarette smoke in a Th1-like inflammatory environment may respond to an LABA by expressing a less harmful transcriptome.
Conclusions
The results of this study implicate widespread changes in gene expression in the mechanism of action of PDE4 inhibitors in COPD. This effect may be particularly relevant when added on to an LABA or an ICS/LABA combination therapy. Nevertheless, a comparable genomic signature must be confirmed in airway epithelia harvested from individuals with COPD for this assertion to gain traction. Such in vivo investigations are necessary because they will reveal the genomic capacity of these drugs at therapeutically relevant doses on a background of airways and extrapulmonary inflammation that is common in individuals exposed chronically to cigarette smoke. They will also provide valuable information on the extent to which cultured airway epithelial cells predict the genomic behavior of their in vivo counterparts in response to drug interventions under pathologic conditions. Finally, the anti-inflammatory activity of oral apremilast in plaque psoriasis and psoriatic arthritis suggests that systemic exposure could contribute to the mechanism of action of roflumilast and may help explain why inhaled PDE4 inhibitors are inactive in COPD.
Acknowledgments
The authors acknowledge Sylvia Wilson for preparing the RNA samples used for microarray-based gene expression profiling, and Dr. Paul M. K. Gordon, Centre for Health Genomics and Informatics, University of Calgary, for assistance with bioinformatics.
Authorship Contributions
Participated in research design: R. Joshi, Yan, Hamed, Newton, Giembycz.
Conducted experiments: R. Joshi, Yan, Hamed, T. Joshi.
Performed data analysis: R. Joshi, Yan, Hamed, Mostafa, T. Joshi, Giembycz.
Wrote or contributed to the writing of the manuscript: R. Joshi, Yan, Hamed, Mostafa, T. Joshi, Newton, Giembycz.
Footnotes
- Received November 28, 2018.
- Accepted April 25, 2019.
↵1 Current affiliation: Global Development Operations, Novartis Healthcare Pvt. Ltd. Salarpuria-Sattva Knowledge City, Raidurg, Hyderabad, India.
This study was supported by a project grant from the Canadian Institutes for Health Research (PJT 152904), the Lung Association of Alberta & NWT, and an unrestricted research grant from Gilead Sciences Inc., Seattle. O.H. and T.J. are recipients of studentships awarded by the Lung Association of Alberta & NWT. D.Y. was supported by Alberta Innovates. Real-time polymerase chain reaction was facilitated by an equipment and infrastructure grant from the Canadian Fund of Innovation and the Alberta Science and Research Authority.
The authors state no conflict of interest.
↵This article has supplemental material available at molpharm.aspetjournals.org.
Abbreviations
- β2A
- 8-hydroxy-5-((R)-1-hydroxy-2-methylaminoethyl)-1H-quinolin-2-one
- AE
- adverse effect
- AUC0–18h
- area under the curve from 0 to 18 hours
- COPD
- chronic obstructive pulmonary disease
- CRE
- cAMP response element
- DCITC
- 5(2-(((1′-(4′-isothiocyanatephenylamino)thiocarbonyl)amino)-2-methylpropyl)amino-2-hydroxypropoxy)-3,4-dihydrocarbostyril
- DEG
- differentially expressed gene
- FDR
- false discovery rate
- GO
- gene ontology
- GSK 256066
- 6-[3-(dimethylcarbamoyl)benzenesulphonyl]-4-[(3-methoxyphenyl)amino]-8-methylquinoline-3-carboxamide
- ICS
- inhaled corticosteroid
- IL
- interleukin
- Ind+GSK
- indacaterol and GSK 256066 in combination
- LABA
- long-acting β2-adrenoceptor agonist
- PCR
- polymerase chain reaction
- PDE
- phosphodiesterase
- RNA-seq
- RNA-sequencing
- RNO
- roflumilast N-oxide
- Salm0.3
- salmeterol 0.3 nM
- Salm0.5
- salmeterol 0.5 nM
- Salm100
- salmeterol 100 nM
- SFM
- serum-free medium
- Copyright © 2019 by The American Society for Pharmacology and Experimental Therapeutics