To investigate the impact of GCs on the macrophage response to LPS, primary mouse bone marrow-derived macrophages (BMDMs) were stimulated with LPS in the absence or presence of 100 nM Dex for four hours. This is similar to the mean serum concentration in healthy volunteers after 7.5 mg oral Dex, and can be considered a moderate dose (25). Steady state mRNA abundance was then assessed by microarray. The data were first filtered for robust gene induction by LPS (greater than five-fold increase of expression, p < 0.05), leading to the identification of 599 LPS-induced protein-coding transcripts. The responses of these transcripts to the addition of Dex were then assessed. As illustrated by volcano plot ( Figure 1A ), the effect of Dex was strongly dominated by inhibition, 169 transcripts being down-regulated and only 13 up-regulated (fold change > 2 and p < 0.05). Several LPS-induced genes were suppressed more than 80% by 100 nM Dex. Cooperative regulation of the Lcn2 (lipocalin 2) gene by LPS and Dex was previously reported (28) and confirmed here ( Figure 1A ).

Gene Ontology (GO) analysis of GC down-regulated genes revealed highly significant enrichment of terms related to inflammation, cytokine production and responses to biotic stimuli ( Figure 1B ; Supplemental Table 1 ). Enrichment of such terms was predictable and uninformative, since the original set of 599 transcripts was selected on the basis of strong induction by LPS. We were interested by the gene sets “Response to virus” (GO:0009615; set 9 in Figure 1B ), “Cellular response to interferon beta” (GO:0035458; set 12 in Figure 1B ), and “Response to protozoan” (GO:0001562; set 14 in Figure 1B ), which displayed 10.4-fold, 31.8-fold and 39.3-fold enrichment, respectively. Enrichment of these terms suggested that Dex inhibited expression of genes involved in antimicrobial responses (which are commonly regulated by IFNβ). However, there was no significant enrichment of GO terms specifically related to responses to intracellular bacteria.

Closer inspection of the microarray data revealed many interferon-stimulated genes (ISGs) amongst the LPS-induced genes that were significantly down-regulated by Dex. To explore this in more detail we made use of two published data sets related to interferon-mediated gene regulation in primary murine macrophages (generated in a similar manner). First, a systems-based analysis of the macrophage transcriptome identified a large number of genes whose delayed induction by lipidA (a highly TLR4-specific LPS moiety) was dependent on autocrine signaling via the interferon receptor IFNAR1 (29). The top 70 IFNAR1-dependent genes from that study were used as a gene set for GSEA (gene set enrichment analysis) ( Figure 1C ). This confirmed extremely strong enrichment of IFNAR1-dependent genes amongst those negatively regulated by Dex in our own data set, with an enrichment score of 0.866 and p value 0.0. From a study of macrophage regulatory networks (26) we then identified genes robustly induced by treatment of BMDMs with IFNβ alone (fold change > 100 and p < 0.05), and used this list of 93 genes in GSEA ( Figure 1D ). Once again there was very strong enrichment of IFNβ-inducible genes amongst those suppressed by Dex in our microarray set, with enrichment score 0.800 and p = 0.0. Removal of genes common to the two lists reduced the enrichment score only marginally (ES = 0.775) and did not change the nominal p value of zero.

We generated a list of 96 genes that were robustly induced by LPS in our own data set and interferon-regulated according to one or both of the above-mentioned data sets (26, 29). Mean responses of these genes to LPS and/or Dex were then plotted ( Figure 1E ). The great majority were expressed in a delayed fashion, with little or no increase of expression 1 h after addition of LPS ( Figure 1E ). At the 4 h time point, expression of 86 out of 92 genes was significantly inhibited by Dex. Importantly, the inhibitory effect of Dex was not quite universal. The IFNβ-responsive gene Acod1 (aconitate decarboxylase 1), previously known as Irg1 (immune-responsive gene 1), mediates production of the anti-inflammatory metabolite itaconic acid (27, 30). The Acod1 transcript was induced more than 1000-fold by LPS, and this response was spared from inhibition by Dex. Two IFNβ-responsive genes were significantly cooperatively regulated by LPS and Dex (the outliers AW112010 and Slc6a12 in Figure 1E : p = 1.73x10^-4 and p = 1.00 x 10^-7, respectively).

These data suggested that LPS treatment of BMDMs induced an antimicrobial state, likely involving autocrine signaling by IFNβ, and that this process was opposed by Dex. The nature of the antimicrobial state was difficult to define because of lack of knowledge of factors that restrict some pathogens, and because many antimicrobial factors are in fact not specific for pathogen class. We therefore focused on antiviral mediators, because many of these have been systematically identified and annotated on the basis of high throughput functional screens [for example (31)]. First we conducted a systematic screen of our microarray data, identifying genes that were induced by LPS (log₂ fold change > 1, p < 0.05); expressed in a delayed manner (at least two-fold higher at 4 h than at 1 h); inhibited by Dex (p < 0.05); labelled as type I interferon-inducible in mouse according to the Interferome database (32); and had plausible roles in the regulation of viral life cycles (at least two supporting publications found in Pubmed using the systematic gene name and “antiviral” as search terms). This generated a list of 107 genes ( Supplemental Table 2 ), which is likely to be under-inclusive because of the use of a single, relatively early time point for analysis of gene expression; shortcomings of the text-mining approach; or shortcomings in functional annotation of genes. For example, the poorly annotated gene AA467197, mouse ortholog of the human C15orf48 gene, has only recently been identified as an interferon regulated gene and a putative antiviral factor (33, 34). It is also important to note that viruses may co-opt cellular antiviral machinery to gain a competitive advantage, therefore not all of the genes in the final list are unequivocally “antiviral” under all circumstances. However, many of these are known to contribute to cell-intrinsic defense against viral pathogens (23, 31). These genes belong to several classes including: GTPases or GTP-binding proteins, which contribute to antimicrobial defense via a number of mechanisms (23, 35) (Gbp2, Gbp4, Gbp5, Gbp9, Gnb4, Iigp1, Irgm2, Mx1, Mx2, Tgtp1, Tgtp2); PRRs involved in the recognition of intracellular pathogens and the initiation of antiviral responses (Aim2, Ddx58 [RIG-I], Dhx58, Eif2ak2 [PKR], Ifi203, Ifi204, Ifih1 [MNDA-5], Mndal, Nlrc5, Pyhin); several members of the poly-ADP ribosyltransferase (PARP) and Schlafen (SLFN) families that contribute to antiviral defenses (36, 37); and diverse restriction factors that act at distinct points of infectious cycles to prevent entry, replication or exit of viruses (Bst2 [tetherin], Ch25h, Cmpk2, Herc6, Ifi35, Ifi44, Ifi47, Ifit1, Ifit2, Ifit3, Isg15, Isg20, Oasl1, Parp14, Rsad2 [viperin], Trim21, Usp18) (23, 24). Patterns of expression of the top 50 most Dex sensitive genes are illustrated in heatmap form in Figure 2A , whilst the full list is included in Supplemental Table 2 .

An unbiased proteomic screen was then used to assess protein levels in BMDMs treated with vehicle, LPS or LPS + Dex for 24 hours. Depth of proteome coverage was good, with more than 6,000 proteins detected, and normalized protein concentrations were similar across conditions, indicating an absence of systemic bias (not shown). We identified proteins whose abundance was increased by LPS (> 1.5-fold increase, Padj < 0.05). The lower threshold for fold change was chosen because LPS-induced changes of abundance were generally lower at the protein than at the mRNA level. This generated a list of 410 LPS-regulated proteins, whose responses to Dex were then examined. According to volcano plot, the effect of Dex was again dominated by suppression ( Figure 2B ). 125 LPS-induced proteins were significantly negatively regulated by Dex, and 96 of these were identified as type I interferon-regulated genes in the Interferome database (32). Amongst these Dex sensitive targets, gene ontology analysis revealed very strong enrichment of terms related to type I interferons (response to interferon-beta, GO:0035456, fold enrichment 41.2, p = 3.2 x 10^-19; cellular response to interferon-beta, GO:0035458, fold enrichment 40.9, p = 9.3 x 10^-17). Gene set enrichment analysis also revealed exceptional enrichment of IFNAR-dependent genes that were negatively regulated by Dex at the protein level ( Figure 2C ). Where data were available in the proteomic data set, we examined the expression of individual protein products of the top 50 Dex-sensitive genes ( Figure 2D ). With a few exceptions, patterns of expression were similar at mRNA and protein levels, with significant increase in response to LPS and significant decrease in response to addition of Dex. The extent of agreement is striking given that mRNAs were sampled at 4 and proteins at 24 hours after stimulation. The selection of the 24 h time point for assessment of protein expression may have led us to miss or underestimate effects of both LPS and Dex in this experiment. Nevertheless, the emerging picture is one of strong and sustained GC-mediated suppression of antiviral programs. Note that some mediators of antiviral responses (e.g., STING1 and MAVS) were constitutively expressed at both mRNA and protein levels and neither up-regulated by LPS nor down-regulated by Dex (data not shown).

Several transcripts encoding secreted or cell-surface immunomodulatory factors were also expressed in delayed fashion, negatively regulated by Dex, and identified as ISGs according to the Interferome database (32) ( Figure 3A ). Il12b, Il6 and Il27 mRNAs were highlighted in Figure 1A . Levels of the corresponding proteins were measured in supernatants of BMDMs treated under the same conditions ( Figure 3B ). All of these cytokines displayed no increase of expression 1 h after an LPS stimulus, strong increase at 4h, and strong inhibition by Dex. Some of the immunomodulatory proteins were measurable in the unbiased proteomic screen, up-regulated by LPS and suppressed by Dex ( Figure 3C ). Here we also confirmed the strong induction of LCN2 protein by the combination of LPS and Dex ( Figure 3C ).

ISGF3 plays a central role in the regulation of ISGs (18). Ligand-activated GR was reported to inhibit expression of ISGs by competing with ISGF3 for an essential transcriptional cofactor (38). To investigate the site of action of Dex we therefore used a mouse macrophage cell line stably transfected with an ISGF3-dependent reporter construct. This construct was strongly activated by either LPS or recombinant IFNβ, and in each case the response was ablated by the broad-spectrum JAK inhibitor Ruxolitinib, consistent with dependence on the phosphorylation of STAT1 and STAT2 ( Figure 4A ). LPS-induced activation of the reporter was dose-dependently inhibited by Dex, whereas the IFNβ-induced response was insensitive to even the highest dose of Dex used (1 μM), suggesting that Dex does not directly interfere with ISGF3 function.

IFNβ caused a rapid increase in the phosphorylation of STAT1, which was evident by 30 minutes and insensitive to Dex ( Figure 4B ). In contrast, LPS caused delayed STAT1 phosphorylation, which was not evident until 2 h, and was impaired by Dex, particularly at later time points. The delayed activation of STAT1 in response to LPS was dependent on IFNAR1-mediated signaling, and could be prevented by either an IFNAR1-neutralising antibody or treatment of cells with Ruxolitinib ( Figure 4C ). These observations suggest that Dex acts upstream of IFNβ biosynthesis to disrupt the IFNβ-mediated autocrine loop that controls macrophage functions. In fact, the Ifnb1 gene appeared in several of the GO terms discussed above ( Supplemental Table 1 ), and its expression was strongly inhibited by Dex in the microarray experiment (see below). Quantitative PCR confirmed rapid, strong and sustained induction of the Ifnb1 gene following LPS treatment of BMDMs ( Figure 4D ). Throughout a 24 h time course, this response to LPS was inhibited at least 70% by 100 nM Dex.

It was previously reported that pre-treatment of the myeloid cell line U937 with Dex impaired the LPS-induced phosphorylation and activation of TBK1 (39). We therefore considered the hypothesis that inhibition of TBK1 function explains the inhibitory effect of Dex on IFNβ-mediated feedback in primary macrophages. However, simultaneous treatment of primary BMDMs with LPS and Dex caused no impairment of early LPS-induced TBK1 phosphorylation under conditions that resulted in strong inhibition of both Ifnb1 and ISGs ( Figure 4E ). Dex weakly but consistently reduced TBK1 phosphorylation at the four-hour time point. However, this late effect was unlikely to contribute to changes in expression of Ifnb1, which were strongly declining by four hours. Dex also had no effect on the phosphorylation of IRF3, which mediates the activation of Ifnb1 transcription downstream of TBK1 ( Figure 4F ).

Glucocorticoids selectively inhibit the expression of certain pro-inflammatory genes by increasing and prolonging the expression of DUSP1 and inhibiting MAPK p38 (7–10). DUSP1 is a negative regulator of Ifnb1 gene expression (40). Both serum IFNβ and hepatic IFNβ-dependent genes (including Ifit2, Rsad2, Iigp1 and many more of the ISGs discussed here) were over-expressed in Dusp1-/- mice following infection with E. coli (41). These observations suggested the hypothesis that Dex impairs expression of IFNβ and IFNβ-dependent genes in LPS-activated BMDMs via the induction of DUSP1. This hypothesis was tested using macrophages generated from Dusp1-/- mice. IFNβ biosynthesis was elevated in Dusp1-/- BMDMs as previously reported (40), but similarly inhibited by Dex in Dusp1+/+ and Dusp1-/- BMDMs ( Figure 5A : note different scales of y axes). A previous publication reported that Dex did not impair LPS-induced IFNβ expression in BMDMs (42). The reasons for this discrepancy are not clear, but may include different sources and doses of LPS or different duration of Dex treatment.

We then systematically examined the impact of DUSP1 depletion on the Dex sensitivity of IFNβ-regulated antiviral genes ( Figure 5B , left) and immunomodulatory genes ( Figure 5B , right). For the gene sets discussed above, Dex sensitivity was calculated as average ratio of expression under conditions (LPS + Dex/LPS). In plots of Dex sensitivity in Dusp1+/+ (x axis) and Dusp1-/- (y axis) backgrounds the dotted diagonal line represents the null hypothesis that disruption of the Dusp1 gene has no impact on responsiveness to Dex. For both gene sets, sensitivity to Dex was closely correlated in Dusp1+/+ and Dusp1-/- BMDMs, supporting the null hypothesis and suggesting a minimal impact of DUSP1 depletion. However, there were several outliers whose response to Dex was altered by Dusp1 gene disruption. Some of these are highlighted in Figure 5B , and primary data from the microarray experiment are illustrated in Figure 5C . In the cases of Irf7, Bst2, Sp100 and Cxcl3, significant inhibitory effects of Dex were completely lost in the absence of DUSP1. We conclude that overall DUSP1 plays a minor and highly gene-specific role in the regulation of IFNβ signaling by Dex. Figure 5C also illustrates the behavior of the Ifnb1 gene itself in the microarray experiment, confirming elevated expression in Dusp1-/- BMDMs and strong suppression by Dex in BMDMs of both genotypes.

Collectively, these observations led to a clear and testable hypothesis: that Dex impairs the responses of macrophages to LPS by inhibiting the expression of IFNβ and disrupting the IFNβ-mediated autocrine feedback loop. We predicted that secondary genes, which are induced by LPS in an IFNβ-dependent manner, would be insensitive to Dex if activated directly by IFNβ, by-passing the site of GC action. To test this hypothesis, we first generated a panel of test genes comprising three intracellular antiviral effectors (Ifit2, Rsad2 and Iigp1) and three secreted cytokines or chemokines (Cxcl9, Il6 and Il27). IFNAR1 dependence of all of these genes was confirmed by reference to a published data-set (29); by our own experiments using Ifnar1-/- BMDMs (data not shown); and by use of an IFNAR1-blocking antibody ( Figure 6A ). We then tested the responses of all six genes to challenge with LPS or IFNβ in the absence or presence of Dex ( Figure 6B ). All three of the antimicrobial effector genes responded as predicted by the hypothesis: their expression in response to LPS was significantly impaired by Dex, whereas their expression in response to IFNβ was insensitive to Dex. Surprisingly, all three of the cytokine/chemokine genes were suppressed by Dex whether induced by LPS or IFNβ. Dex may therefore act both upstream and downstream of IFNβ to impair LPS responses of BMDMs. The IFNβ-induced expression of Ifit2, Rsad2, Iigp1, Il6 and Cxcl9 was very strongly dependent on ISGF3 (at least 100-fold lower expression in IFNβ-stimulated Irf9-/- BMDMs) (43). Ifit2, Rsad2 and Iigp1 were insensitive to Dex, whilst Il6 and Cxcl9 were suppressed by Dex in IFNβ-stimulated BMDMs. This confirms the conclusion from the reporter gene assay ( Figure 4A ); that Dex does not regulate IFNβ signaling in BMDMs via broad suppression of ISGF3 function.

To investigate the same phenomena in human cells we generated monocyte-derived macrophages from peripheral blood of healthy donors and stimulated them with LPS in the absence or presence of 100 nM Dex. Surprisingly there was greater than ten-fold donor-to-donor variation in quantity of IFNβ secreted by LPS-activated MDMs (note logarithmic axis in Figure 7A ). However, IFNβ release was consistently inhibited by 100 nM Dex (mean inhibition ± SEM 52% ± 6%, p = 0.001). Basal expression of some ISGs varied between donors by several hundred-fold (not shown), and magnitudes of response to both LPS and IFNβ were also highly variable ( Figure 7B ). This created a practical problem in generating adequate statistical power to investigate effects of Dex, power calculation suggesting that 100 or more individual donors would be needed for some ISGs. Nevertheless, a preliminary analysis using seven independent isolates of MDMs confirmed that Dex can inhibit expression of the ISGs RSAD2 and CXCL9 whether evoked by either LPS or IFNβ. In human alveolar macrophages the LPS-induced expression of IFNB1 and RSAD2 genes was consistently inhibited by Dex ( Figure 7C ); however, IFNβ protein was below the limit of detection in supernatants of these cells. Scarcity of samples, low yields of mRNA and lack of statistical power prevented us from carrying out an extensive survey of ISG expression. Yet these preliminary data suggest that Dex may also impair type I interferon signaling in alveolar macrophages.

All mice were maintained at the Biomedical Services Unit of the University of Birmingham. Animal care and experimental procedures were performed according to Home Office guidelines and approved by the University of Birmingham Local Ethical Review Committee. Bone marrow was isolated from the femurs of humanely culled 6–12-week-old wild type C57BL/6J mice, and BMDMs (bone marrow derived macrophages) generated essentially as described (80) by differentiation in vitro in RPMI 1640 medium with L-glutamine (Gibco Thermo Fisher Scientific 21875034) supplemented with 10% heat-inactivated fetal bovine serum (Sigma F0392) and 50ng/ml recombinant M-CSF (Peprotech 300-25) for 7 days. BMDMs were plated at a density of 1 x 10⁶/ml in the appropriate cell culture plate at least 1 d prior to stimulation. This method generated primary mouse macrophages with purity approaching 99% as measured by flow cytometry using antibodies against F4/80, CD11b, CD14 and CD64 ( Figure S1 ).

The Dusp1-/- strain was a generous gift from Bristol-Myers Squibb, and was back crossed to C57BL/6 for ten generations prior to experiments described here.

Human monocytes from healthy blood donors were isolated from leukapheresis blood cones supplied by the National Blood and Transplant Service (ethical approval ERN_16-0191). Monocytes were enriched by negative selection using StemCell RosetteSep Human Monocyte Enrichment Cocktail (StemCell 15068; 75μl per ml of blood) and Ficoll-Paque (VWR 17144003). Cells were differentiated for 7 days in RPMI 1640 medium with L-glutamine (Gibco Thermo Fisher Scientific 21875034) supplemented with 5% heat-inactivated fetal bovine serum (Biosera FB-1001) and 50ng/ml recombinant M-CSF (Peprotech 300-25). Differentiated macrophages were plated at a density of 1 x 10⁶/ml in the appropriate cell culture plate at least 1 d prior to stimulation. In our hands this method generates primary human macrophages with purity >98% as measured by flow cytometry using antibodies against CD14, CD64 and CD206 ( Figure S1 ).

Ethical approval was obtained to recruit adult patients scheduled for surgery to remove lung tissue as part of their clinical treatment plan (predominantly lobectomy for lung cancer) at the Thoracic Surgery Unit at the Queen Elizabeth Hospital Birmingham (REC 17/WM/0272). Lung tissue samples distant from any tumor, without macroscopically evident pathology, and surplus to requirement for histopathology, were washed through with 500 ml of sterile 0.9% saline (Baxter, UK) using a 14-gauge needle (Vasofix ^®). The washed through lavage fluid was collected and centrifuged at 4°C and 560g for 10 minutes. Cell pellets were resuspended in 10 ml of RPMI1640 containing 10% FBS, and overlaid onto 10 ml of Lymphoprep™ (StemCell) prior to centrifugation at 800g for 30 min at 4°C. The interphase layer containing mononuclear cells was aspirated into a sterile 50 ml tube containing PBS with 10% FBS, which was then centrifuged at 300g for 10 minutes at 4°C. AMs were resuspended in RPMI1640 containing 10% FBS, 100U/mL penicillin, 100µg/mL streptomycin and 2mM L-glutamine (Sigma-Aldrich), and plated. This method generates primary human alveolar macrophages with purity >98% as measured by a validated combination of cytospin and flow cytometry using antibodies against CD68 (81).

Stimulations were carried out in appropriate culture plates using the following reagents and concentrations unless otherwise stated: LPS (Enzo ALX-581-010-L002; 10ng/ml); recombinant human IFNβ (Peprotech 300-02BC; 10ng/ml); recombinant mouse IFNβ (BioLegend 581306; 10ng/ml); Ruxolitinib (Selleck S1378; 1μM); Dexamethasone (Sigma D8893; 100nM). Where stated, cells were incubated for 2 hours with an IFNAR1 blocking monoclonal antibody (Fisher Scientific MAR1-5A3; 10 μg/ml), or mouse IgG1 kappa isotype control, (Fisher Scientific P3.6.2.8.1; 10 μg/ml) prior to stimulations.

Upon harvesting of in vitro-differentiated cells for counting and seeding for experiments (mouse BMDMs day 7; human MDMs day 6), a fraction of the cell suspension was used for purity assessment by flow cytometry. Cells were stained with eFluor780 fixable viability dye (eBioscience 65-0865, 1/1000 in PBS) for 20min on ice to allow gating for live cells. Fc receptor blocking was performed for 5min prior to antibody staining using Mouse BD Fc Block (BD 553142) or Human BD Fc Block (BD 564220). Antibody staining was performed for 30min on ice in FACS buffer (PBS+2% FCS+1mM EDTA) using the antibodies and dilutions listed below. Cells were washed and resuspended in FACS buffer and run on a BD LSRFortessa X-20. Analysis was performed using FlowJo v10.

Mouse BMDM staining: PE rat anti-mouse F4/80 Antibody (Biolegend 123110, 1/50); PE rat IgG2a κ Isotype Control Antibody (Biolegend 400507, 1/50); FITC rat anti-mouse/human CD11b Antibody (Biolegend 101205, 1/100); FITC rat IgG2b Isotype Control Antibody (eBioscience 11-4031-82, 1/100); PerCP-eFluor710 rat anti-mouse CD14 Antibody (eBioscience 46-0141-80, 1/40); PerCP-eFluor710 rat IgG2a κ Isotype Control Antibody (eBioscience 46-4321-82, 1/40); APC rat anti-mouse CD64 Antibody (Biolegend 139305, 1/20); APC rat IgG1 κ Isotype Control Antibody (eBioscience 17-4301-81, 1/20).

Human MDM staining: PE mouse anti-human CD14 Antibody (Biolegend 367104, 1/100); PE mouse IgG1 κ Isotype Control Antibody (Biolegend 400113, 1/20); PerCP-eFluor710 mouse anti-human CD64 Antibody (eBioscience 46-0649-41, 1/20); PerCP-eFluor710 mouse IgG1 κ Isotype Control Antibody (eBioscience 46-4714-80, 1/20); APC mouse anti-human CD206 Antibody (Biolegend 321110, 1/20); APC mouse IgG1 κ Isotype Control Antibody (Biolegend 400119, 1/20).

The RAW-Blue ISG macrophage cell line stably transfected with an ISGF3-dependent reporter construct was cultured and maintained according to the manufacturer’s instructions (RAW-Blue ISG cells, Invivogen raw-isg). ISGF3 reporter activity was determined from cell culture supernatants using QUANTI-Blue detection reagents according to the manufacturer’s instructions (Invivogen).

RNA was isolated from mouse BMDMs and human monocyte derived macrophages using Norgen Total RNA Purification Plus kit (Geneflow P4-0016) according to manufacturer’s instructions. cDNA was synthesized using the iScript cDNA Synthesis Kit (Biorad 1708891). mRNA was detected by RT-qPCR using SYBR TB Green Premix Ex Taq (Takara RR820W) and primers supplied by Eurofins Genomics or Sigma Aldrich. UBC (human) or B2M (mouse) were used to normalize mRNA measurements via 2^-ΔΔCt method. Primers are listed in Table 1 .

Microarray analysis was performed using SurePrint G3 Mouse GE 8x60K slides (Agilent) and Partek Genomics Suite version 6.6, build 6.13.0315 (Partek) as previously described (82). For generation of volcano plots, transcripts were first filtered for significant upregulation in response to LPS (> 5-fold increase, adjusted p-value < 0.05) and weakly expressed transcripts were removed by application of an arbitrary filter of 200 RMA. Plots (log₂ fold difference of expression vs -log₁₀ ANOVA p-value) were constructed using Prism (GraphPad Software), with subset cut-offs at p-value < 0.05 and fold difference of expression > 2. Microarray data discussed in this paper were deposited at Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) with the accession number GSE68449.

Gene Ontology analysis was performed using the Panther v16 online toolset (83). Redundant GO terms were filtered using REVIGO (84). Gene Set Enrichment Analysis (GSEA) was performed using GSEA_4.1.0 (85).

Conditioned medium samples from cultured macrophages were subjected to custom multiplex Luminex assay (Bio-Rad Custom Bio-Plex Assays), and Invitrogen ProcartaPlex Mouse or Human IFNβ simplexes, (EXP01A-26044-901 or EXP01A-12088-901, respectively), according to manufacturer’s instructions.

For whole cell lysates, cells were harvested in RIPA buffer and samples passed through a Qiashredder column to disrupt genomic DNA (Qiagen 79656). Protein was quantified by Pierce BCA Assay (Thermo Fisher Scientific 23225). Laemmli or XT sample buffer was added and samples heated to 95°C for 5min. Western blotting was performed using Criterion TGX protein gels (BioRad) and Tris-Glycine SDS buffer (Geneflow B9-0032) or 4–12% Criterion™ XT Bis-Tris Protein Gels (BioRad) and XT MES running buffer (BioRad 1610789). Protein was transferred to BioRad Trans-Blot PVDF membranes (BioRad 1704157) using BioRad Trans-Blot Turbo transfer system. Blots were imaged using Clarity Western ECL Substrate (BioRad 1705061) and BioRad ChemiDoc MP Imaging System. Densitometry for Western blot quantification was performed using ImageJ. Antibodies used in western blotting are listed in Table 2 .

Following derivation and stimulation of mouse BMDMs (as described above) cells were washed with PBS and lysed in proteomic lysis buffer (5% SDS, 10mM TCEP, 50mM TEAB). Unbiased proteome analysis was carried out by data-independent acquisition (DIA) mass spectrometry proteomics utilizing S-Trap on-column digestion and purification, following the methods detailed by Baker et al. (86). Protein copy number was determined using Perseus software (86, 87) with normalization for histone protein content. Data were analyzed as described in the above methods for microarray analysis and bioinformatics.

GraphPad Prism software (Version 6) was used for statistical analysis. Mann Whitney U test was used for comparison of two groups. For analysis of multiple groups, ANOVA was used with Bonferroni correction for multiple comparisons. The following marks are used throughout: *,p<0.05; **,p<0.01;***,p<0.005; n.s., not statistically significant. N numbers specified in figure legends indicate biological replicates. In human alveolar macrophages, where expression of ISGs was highly variable between donors, ratio paired t test was used to test for consistent effect of Dex.