PPARγ Deficiency Suppresses the Release of IL-1β and α in Macrophages via a Type 1 Interferon Dependent Mechanism

Abstract

Obesity and diabetes modulate macrophage activation often leading to prolonged inflammation and dysfunctional tissue repair. Increasing evidence suggests that the NLRP3 inflammasome plays an important role in obesity-associated inflammation. We have previously shown that activation of the lipotoxic inflammasome by excess fatty acids in macrophages occurs via a lysosomal dependent pathway. However, the mechanisms that link cellular lipid metabolism to altered inflammation remain poorly understood. PPARγ is a nuclear receptor transcription factor expressed by macrophages that is known to alter lipid handling, mitochondrial function, and inflammatory cytokine expression. To undercover novels links between metabolic signaling and lipotoxic inflammasome activation we investigated mouse primary macrophages deficient in PPARγ. Contrary to our expectation, PPARγ knock out (KO) macrophages released significantly less IL-1β and IL-1α in response to lipotoxic stimulation. The suppression occurred at the transcriptional level and was apparent for multiple activators of the NLRP3 inflammasome. RNA sequencing revealed upregulation of IFNβ in activated PPARγ KO macrophages and this was confirmed at the protein level. A blocking antibody against the type 1 interferon receptor restored the release of IL-1β to wild type levels in PPARγ KO cells confirming the mechanistic link between these events. Conversely, PPARγ activation with rosiglitazone selectively suppressed IFNβ expression in activated macrophages. Loss of PPARγ also resulted in diminished expression of genes involved in sterol biosynthesis, a pathway known to influence interferon production. Together these findings demonstrate a crosstalk pathway that influences the interplay between metabolism and inflammation in macrophages.

Introduction

Macrophage dysfunction is a hallmark of diabetes and contributes to several diabetes complications including impaired wound healing, post-myocardial infarction heart failure, atherosclerosis, and non-alcoholic steatohepatitis (1–4). Diabetes is associated with elevated levels of free fatty acids (FFAs) and triglycerides (TAG) in circulation and tissues (5–7). The presence of excess lipids in the nutrient microenvironment can modulate macrophage responses to inflammatory stimuli leading to dysfunction. As an example, we have previously shown that LPS activation of macrophages in the presence of the saturated FAs, such as palmitate, leads to lysosome damage and cell death (8–10). The elaboration of pro-inflammatory cytokines, such as IL-1β, is also increased in this context. In response to excess FAs, lysosome damage facilitates activation of the NLRP3 inflammasome leading to the secretion of IL-1β (11). The combination of premature macrophage cell death and augmented pro-inflammatory cytokine release likely plays a role in the abnormal inflammatory and reparative responses observed in patients with diabetes.

IL-1 medicated inflammation contributes to pathogenesis of metabolic and cardiovascular disease. Prior studies have shown a strong association between IL-1 levels and the risk of diabetes and atherosclerosis in both pre-clinical models and humans (12–15). In fact, the CANTOS trial recently demonstrated that a neutralizing antibody against IL-1β reduces cardiovascular complications in high risk patients with coronary artery disease (16). In light of these observations, the inflammasome has emerged as an important target for therapy in cardiometabolic disease. The NLRP3 inflammasome forms as a cytosolic complex that consists of NLRP3, ASC, and caspase 1(17). There are 2 signals that are necessary for full activation. Signal 1 is typically generated via a toll-like receptor (TLR) ligation, leading to the activation of NF-κB and the transcriptional upregulation of pro-IL-1 β and NLRP3. Signal 2 then stimulates assembly of the inflammasome complex allowing for caspase 1-mediated cleavage of pro-IL-1β. Once cleaved, IL-1 β is biologically active and secreted from the cells. Understanding the mechanisms that regulate signal 1 and signal 2 could facilitate the discovery of new approaches to dampen the activity of this potent inflammatory complex.

It has been increasingly recognized that cellular metabolism is an important regulator of macrophage inflammatory and reparative function. PPARγ is transcription factor that functions as a master regulator of macrophage lipid metabolism and is upregulated in macrophages at sites of inflammation, including atherosclerotic plaques (18–20). In addition to its metabolic effects, PPARγ activation also suppresses macrophage cytokine production (21–23). The anti-inflammatory action of PPARγ on pro-inflammatory cytokines is thought to occur via sumylation of the transcriptional repressor NCOR and/or via the “anti-inflammatory” nature of oxidative mitochondrial metabolism (21, 24). In models of obesity and atherosclerosis, loss of PPARγ from macrophages results in increased insulin resistance and vascular disease, respectively (21, 25–27). These findings support the conclusion that PPARγ in macrophages drives a program that is protective against cardiometabolic disease. Despite these pre-clinical observations, the use of PPARγ agonists in humans with diabetes has been controversial due to unexpected cardiovascular complications (28–30).

Based on our prior studies and the literature, we initially hypothesized that PPARγ would be a potent suppressor of lipotoxic inflammasome activation through effects on both signal 1 and signal 2. We also anticipated that dissecting the mechanism(s) of this response could lead to new insights regarding the interplay between inflammatory signals and lipid metabolism in macrophages. To test this hypothesis, we investigated inflammasome activation in macrophages deficient in PPARγ. Contrary to our expectations, loss of PPARγ from macrophages lead to a significant decrease in IL-1β and IL-1α release in response to SFAs and other NLRP3 activators in vitro. This effect occurred at the level of IL-1 mRNA regulation which resulted in a decrease in pro-IL-1β and pro-IL-1α protein production. RNA sequencing revealed upregulation of type 1 IFN genes in activated PPARγKO macrophages and treatment with PPARγ agonists repressed the expression of these genes. The suppression of IL-1 expression in PPARγKO macrophages was reversed when type 1 interferon signaling was disrupted. Interestingly, PPARγKO macrophages also displayed a defect in the sterol biosynthesis pathway, which has been linked to enhanced IFNβ production via Stimulator of Interferon Genes (STING). Together these data support a model in which the loss of PPARγ dampens de novo sterol biosynthesis and augments IFNβ production, which in turn suppresses the transcription of IL-1α and IL-1β.

Experimental Section

Reagents

L-NIL was from Enzo life sciences (Farmingdale, NY, USA). T0070907 was from TOCRIS (Minneapolis, MN, USA). Rosiglitazone, actinomycin D, α-tubulin antibody, α-actin antibody, and ATP were from Sigma Chemical (St. Louis, MO, USA). IL-1β, IL-1α, NLRP3, STAT1 and phospho-STAT1(#14994) antibodies were from Cell Signaling (Danvers, MA, USA). IFNβ and the IFNβ ELISA were from PBL (Piscataway, NJ, USA). The IL-10 receptor blocking antibody and PE conjugated α-IFNAR antibody were from Biolegend (San Diego, CA, USA). IFNAR blocking antibody (MAR1–5AE) and control IgG were from Leinco Technologies (St. Louis, MO, USA). The α-TRIF antibody, IL-10, and Duoset ELISA kits (IL-1β, IL-1α, TNFα) were from R&D Systems (Minneapolis, MN, USA). Ultrapure E. coli LPS, cGAMP, poly I:C (PIC) and silica were from Invivogen (San Diego, CA, USA). Thioglycollate was from Difco-BD (Franklin Lakes, NJ, USA). Fatty acids were from Nu-Chek Prep (Waterville, MN, USA). Ultrapure-bovine serum albumin (BSA) was from Lampire (Ottsville, PA, USA) and was tested for TLR ligand contamination prior to use by treating primary macrophages and assaying for TNFα release.

Cell Culture

Peritoneal macrophages were isolated from C57BL/6, or the indicated knockout mice 4 days after intraperitoneal injection of 1 ml, 3.85% thioglycollate and plated at a density of 1×10⁶ cells/ml in DMEM containing 10% inactivated fetal serum (IFS), 50 U/ml penicillin G sodium, and 50U/ml streptomycin sulfate (pen-strep). Stimulations were performed on the day after harvest. For flow cytometry experiments, peritoneal cells were cultured on low adherence plates (Greiner Bio-One) to facilitate cell harvest. Cells were removed from low adherence plates by washing with PBS followed by 10 minutes with Cell Stripper (Gibco) and then 10 minutes with EDTA/trypsin (Sigma). Growth medium was supplemented with palmitate or stearate complexed to BSA at a 2:1 molar ratio as described previously and BSA-supplemented media was used as control (31). For cell stimulations, PBS or LPS (100 ng/ml) were added to media containing BSA or BSA-free fatty acid complexes. For triggering the NLRP3 inflammasome by non-lipid activators pMACs were treated with LPS 100 ng/ml for 16h after which they were incubated with ATP (4mM) for 30 min, silica (150 μg/ml) for 6h, or alum (150 μg/ml) for 6h.

Mice

Wild type (WT) C57BL/6 mice were bred in our mouse facility; PPARγflox x LysM-Cre and LysM-Cre controls were from Gwen Randolph (Washington University) and bred in our facility; IFNR1flox x LysM-Cre were from Mike Diamond (Washington University). STING knockin (KI)/golden ticket mice were from Jonathon Miner. All lines were in the C57BL/6 background. Mice were maintained in a pathogen free facility on a standard chow diet ad libitum (6% fat). All animal experiments were conducted in strict accordance with NIH guidelines for humane treatment of animals and were reviewed by the Animal Studies Committee of Washington University School of Medicine.

RNA isolation, Quantitative RT-PCR, and RNA sequencing

Total cellular RNA was isolated using Qiagen RNeasy columns and reverse transcribed using a high capacity cDNA reverse transcription kit (Applied Biosystems). Real-time qRT-PCR was performed using SYBR green reagent (Applied Biosystems) on an ABI 7500 fast thermocycler. Relative gene expression was determined using the delta-delta CT method normalized to 36B4 expression. Mouse primer sequences were as follows (all 5’−3’): 36B4 (forward- ATC CCT GAC GCA CCG CCG TGA, reverse-TGC ATC TGC TTG GAG CCC ACG TT); TNFα (forward-CAT CTT CTC AAA ATT CGA GTG ACA A, reverse-TGG GAG TAG ACA CAA GGT ACA ACC C); IL-1β (forward- AAG GAG AAC CAA GCA ACG ACA AAA, reverse-TGG GGA ACT CTG CAG ACT CAA ACT); IL-1α (forward- TGA GTT TTG GTG TTT CTG GC, reverse- TCG GGA GGA GAC GAC TCT AA); NLRP3 (forward- AAA ATG CCT TGG GAG ACT CA, reverse-AAG TAA GGC CGG AAT TCA CC); CXCL10 (forward-ATC ATC CCT GCG AGC CTA TCC TG, reverse- CGG ATT CAG ACA TCT CTG CTC ATC) IFNβ (forward – GAC GGA GAA GAT GCA GAA GAG TT, reverse – AGT TCA TCC AGG AGA CGT ACA AC);iNOS (forward- ACA TCG ACC CGT CCA CAG TAT, reverse- CAG AGG GGT AGG CTT GTC TC); MX1 (forward- CCA GGT CCT GCT CCA CAC, reverse- TCT GAG GAG AGC CAG ACG AT); ISG15 (forward- AGC GGA ACA AGT CAC GAA GAC, reverse- TGG GGC TTT AGG CCA TAC TC); ACSL1 (forward – ACC ATC AGT GGT ACC CGC TA, reverse – CGC TCA CCA CCT TCT GGT AT); IL-10 (TGG CCT TGT AGA CAC CTT GG, reverse – AGC TGA AGA CCC TCA GGA TG); MVD (forward – ATG GCC TCA GAA AAG CCT CAG, reverse – TGG TCG TTT TTA GCT GGT CCT); PMVK (forward- AAA ATC CGG GAA GGA CTT CGT, reverse- AGA GCA CAG ATG TTA CCT CCA); MVK (forward- GGT GTG GTC GGA ACT TCC C, reverse – CCT TGA GCG GGT TGG AGA C); FPDS (forward- GGA GGT CCT AGA GTA CAA TGC C, reverse- AAG CCT GGA GCA GTT CTA CAC).

The total cellular RNA was also prepared for RNA sequencing with the Clonetech SMARTer kit according to manufacturer’s protocols, ligated with adapters and unique molecular indexes for each sample for every read, and then sequenced on one single-end 50bp lane on an Illumina HiSeq 3000. RNA-seq reads demultiplexed with Illumina’s bcl2fastq2 and were then aligned to the Mus musculus Ensembl release 76 top-level assembly with STAR version 2.0.4b (32). Gene counts were derived from the number of uniquely aligned unambiguous reads by Subread:featureCount version 1.4.5 (33). Sequencing performance was assessed for total number of aligned reads, total number of uniquely aligned reads, genes detected and ribosomal fraction, known junction saturation, and read distribution over known gene models with RSeQC version 2.3 (34).

All gene counts were then imported into the R/Bioconductor package EdgeR and TMM normalization size factors were calculated to adjust for differences in library size across all samples (35, 36). Ribosomal features as well as any feature not expressed in at least two samples above one count-per-million were excluded from further analysis and TMM size factors were recalculated to create effective TMM size factors. The effective TMM size factors and the matrix of counts were then imported into the R/Bioconductor package Limma and weighted likelihoods based on the observed mean-variance relationship of every gene and sample were then calculated for all samples with the voomWithQualityWeights function (37, 38). Performance of the samples was assessed with a Pearson correlation matrix (Supplemental Fig.1A) and multi-dimensional scaling plots (Supplemental Fig.1B). Generalized linear models were then created to test for differentially expressed genes. The results were then filtered for FDR adjusted p-values less than or equal to 0.05.

The biological interpretation of the large set of features found in the Limma results were then elucidated for global transcriptomic changes in known Gene Ontology (GO) and KEGG terms with the R/Bioconductor packages GAGE and Pathview (39, 40). Briefly, GAGE measures for perturbations in GO or KEGG terms based on changes in observed log 2 fold-changes for the genes within that term versus the background log 2 fold-changes observed across features not contained in the respective term as reported by Limma. For GO terms with statistical significance of p-value <= 0.05, heatmaps were automatically generated for each respective term to show how genes co-vary or co-express across the term in relation to a given biological process or molecular function. In the case of KEGG curated signaling and metabolism pathways, Pathview was used to generate annotated pathway maps of any perturbed pathway with an unadjusted statistical significance of p-value <= 0.05. The RNA sequencing data was deposited into the GEO database with the accession # GSE117115, https://www.ncbi.nlm.nih.gov/geo/.

In vivo peritoneal cytokine assessment

WT or mPPARγKO mice were treated with thioglycollate as described above. On day 4, the mice were injected with 10 μg of LPS in a 200 μl volume. After 16h, the mice were sacrificed and the peritoneal cavity was flushed with 5 ml of regular DMEM+serum. The concentration of IL-1β, IL-1α, and TNFα in the peritoneal fluid was determined via ELISA. At the time, a subset of mice was harvested just prior to LPS injection to evaluate cell recruitment to the peritoneum. Both total cell number and macrophage composition by F480 and CD11b expression were similar between the two genotypes.

Western blotting

Total cellular protein was isolated by lysing cells in 150 mM NaCl, 10 mM Tris (pH 8), triton X-100 1% and 1X Protease Complete and phosphatase inhibitors (Thermo-Fisher Scientific). Subsequently, 25 μg of protein from each sample was separated on a TGX gradient gel (4–20%; Biorad) and transferred to a nitrocellulose membrane. For blots of pro-IL-1β, pro-IL-1α, phospho- and total STAT1 transfer was for 1 hr on ice.

ELISA for IL-1β, IL-1α, TNFα, and IFNβ

Supernatants were harvested from macrophage cultures after the indicated stimulations. IL-1β, IL-1α, and TNFα were quantified using a DuoSet ELISA kit (R&D systems) according to the manufacturer’s instructions. IFNβ was quantified using a ELISA kit from (PBL).

IFNR1 flow cytometry

Peritoneal macrophages were removed from low adherence plate and 1 × 10⁶ cells were pelleted in FACS buffer (PBS, 1%BSA) and incubated with Fc block × 5 min on ice followed by incubation with IFNR1-PE (1:200) for 30 minutes on ice in the dark. Samples were analyzed on a FACS caliber flow cytometer (BD).

Metabolism Assays

Cells were plated into 96 well Seahorse plates at density of 75,000 cells/well and stimulated as indicated in the text. After stimulation, the cells were washed and placed in XF media (non-buffered RPMI 1640 containing, 25mM glucose,2mM L-glutamine and 1mM sodium pyruvate) with 10% FCS. Oxygen consumption rates (OCR) and extracellular acidification rates (ECAR) were measured under basal conditions and following the addition the following drugs: 1.5 μM flurorcarbonyl cynade phenylhydrazon (FCCP), and 100 nM rotenone + 1 μM antimycin A (all Sigma). Measurements were taken using a 96 well Extracellular Flux Analyzer (Seahorse Bioscience; North Bellerica, MA, USA)).

Intracellular succinate, lactate, and pyruvate quantification

Two million pMACs/well were plated in 6 well plates. After stimulation, the cells were washed 3 times with PBS and then were snap frozen and scraped in liquid nitrogen. Intracellular metabolites were quantified by LC-MS/MS at Sanford Burnham Metabolomics Core, Medical Discovery Institute (Lake Nona, Fl).

Cholesterol and oxysterol quantification

Macrophages (5×10⁵ cells) were grown in 24 well tissue culture plates. After stimulation with BSA-PBS or palm-LPS for 8h the cells were washed 3 times with PBS and lysates were prepared by scraping cells in 100 μl of ice cold PBS followed by homogenization by 10 passes through a 26 g syringe. The resulting homogenate (50 μL) was mixed 1:1 with methanol for protein precipitation. Oxysterols and free cholesterol listed above were extracted from 50 μL of each macrophage sample with 200 μL of methanol, containing 2ng of deuterated 24-HC-d₇, 25-HC-d₆, 27-HC-d₇, and 2μg of cholesterl-d₇ as internal standards. All oxysterols and cholesterol as well as their deuterated standards were derivatized with N, N-dimethylglycine (DMG) to increase the MS sensitivity.

The sample analysis was performed with a Shimadzu 20AD HPLC system coupled to a tandem mass spectrometer (API-6500⁺Qtrap: Applied Biosystems) operated in MRM mode. The positive ion ESI mode was used for detection of these analytes. The DMG derivatized samples were injected in duplicate for data averaging. Data processing was conducted with Analyst 1.6.3 (Applied Biosystems).

Statistics

Statistical analysis was performed using GraphPad Prism software. All results are expressed as means +/− SE. Groups were compared by paired Student’s t-test or 2-way ANOVA as appropriate. A value of p ≤ 0.05 was considered significant.

Results

PPARγ Deficient Macrophages are defective in IL-1 production

We have previously shown that NLRP3 inflammasome activation and IL-1β release is enhanced in lipid-loaded macrophages activated with the TLR4 agonist LPS (10, 11). In macrophages, PPARγ is robustly expressed and is known to be an important regulator of metabolism and inflammatory function. Therefore, we used PPARγ deficient macrophages as a tool to explore the links between lipid metabolism and the lipotoxic inflammasome. For these experiments PPARγ floxed mice were crossed to LysM-Cre animals to generate myeloid specific PPARγ knockout (mPPARγKO)(18). As expected, these macrophages had reduced expression of PPARγ gene targets (Supplemental Fig. 2)

To simulate the biology of macrophages encountering an inflammatory stimulus in a high lipid environment, we incubated macrophages with the saturated fatty acids (SFAs) palmitate or stearate in combination with the TLR4 ligand LPS. Using this system, mPPARγKO cells released significantly less IL-1β compared to wild type (WT) cells. IL-1α is commonly co-released in response to NLRP3 activators and its release was also decreased in the KO cells. In contrast, TNFα release was not diminished in these macrophages, indicating this is not a global inflammatory defect or a derangement of TLR4 signaling (Fig. 1A-C). To determine if the effects of PPARγ deficiency were specific to lipids, we also assessed non-lipid activators of the NLRP3 inflammasome including ATP, silica, and aluminum. Surprisingly, mPPARγKO macrophages also released less IL-1β and IL-1α in response to these activators, and again TNFα release was not diminished (Fig. 1D-F). To further confirm that PPARγ was required for maximal IL-1 release we treated WT macrophages with the PPARγ antagonist T0070907 for 24h and then similar stimulations were performed. As shown in Figure 1, this compound phenocopied the genetic loss of PPARγ.

Figure 1. PPARγ loss of function is associated with impaired IL-1 cytokine release in primary macrophages.

(A-C) Peritoneal macrophages (pMACs) isolated from WT (open bars) or mPPARγKO mice (filled bars) were treated with control (BSA-LPS), palm (250 μM)-LPS (100 ng), or stearate (150 μM)-LPS for 20h and the release of IL-1β (A), (B) IL-1α, and (C) TNFα was determined by ELISA. (D-F) pMACs isolated from WT (open bars) or mPPARγKO mice (filled bars) were treated with LPS and ATP, alum or silica as described in the methods and IL-1β (D), (E) IL-1α, and (F) TNFα was determined by ELISA. WT pMACs were treated with veh (open bars) or T0070907 (T007; gray bars) for 24h after which they received the indicated stimuli. The release of IL-1β (G), (H) IL-1α, and (I) TNFα was determined by ELISA. Bar graphs report the mean ± standard error (SE) for a minimum of 3 experiments, each performed in triplicate. *, p<0.05 for WT vs. mPPARγKO or veh vs. T007

LPS induced expression of pro-IL-1β and pro-IL-1α is reduced in PPARγ deficient macrophages

The fact that mPPARγKO cells produced less IL-1β and IL-1α in response to a variety of NLRP3 inflammasome activators argued that PPARγ deficiency was impacting a common part of the pathway, such as production of the precursor cytokines and/or NLRP3 itself (signal 1). To investigate the impact of PPARγ on signal 1 in primary macrophages we activated pMACs from both genotypes and assessed expression of pro-IL-1β and pro-IL-1α protein. As seen in figure 2A, the level of both IL-1 family members was reduced at the protein level in mPPARγKO macrophages. Consistent with a transcriptional mechanism, IL-1β and IL-1α mRNA levels were also decreased in macrophages lacking PPARγ, while the level of NLRP3 mRNA was unaffected (Fig. 2B). Kinetic analysis revealed a similar reduction in IL-1β mRNA abundance at all timepoints after palm-LPS treatment. In comparison, TNFα mRNA levels were similar between the genotypes over the same timeframe (Fig. 2C, D). To assess whether this occurred via changes in mRNA stability, WT and mPPARγKO cells were stimulated with LPS for 4 hours to induce IL-1β transcription and then actinomycin D was added to suppression ongoing transcription. The decay of IL-1β transcript was similar between WT and KO cells (Fig. 2E). Thus, macrophages lacking PPARγ produce less IL-1β and IL-1α but have preserved induction of other classical NF-κB regulated cytokines including TNFα and NLRP3.

Figure 2. PPARγ deficiency leads to decreased levels of IL-1 mRNA and protein levels.

(A) WT or mPPARγKO (KO) macrophages were stimulated with BSA-PBS (vehicle) or palm-LPS for 16h and the protein level of pro-IL-1β and pro-IL-1α was assessed by western blotting. Tubulin (tub) is shown as a loading control. (B) pMACs isolated from WT (open bars) or mPPARγKO mice (filled bars) were treated with vehicle or palm-LPS for 8h and mRNA expression of IL-1β, IL-1α, and NLRP3 was assessed by qRT-PCR. (C, D) Kinetic assessment of IL-1β (C) and TNFα (D) mRNA levels following palm-LPS stimulation in WT and mPPARγKO cells. (E) WT or mPPARγ KO cells were treated with palm-LPS for 4h after actinomycin D was added to the culture. The levels of IL-1β mRNA relative to baseline was performed by qRT-PCR. Each genotype was compared to its own baseline which was arbitrarily defined as 1. Bar graphs report the mean ± standard error (SE) for a minimum of 3 experiments, each performed in triplicate. *, p<0.05 for WT vs. mPPARγKO.

LPS-induced IL-1 release in vivo is reduced in myeloid-deficient PPARγ mice

To determine whether a similar phenotype could be observed in PPARγ deficient macrophages in vivo, we elicited macrophages to the peritoneum using thioglycollate and at day 4, when ~85% of the cells are mature macrophages, we injected 10 μg of LPS IP. After 16h, the peritoneum was lavaged with media and the concentrations of IL-β, IL-1α, and TNFα were determined. Consistent with our ex vivo data, lavage fluid from mPPARγKO mice had reduced concentrations of IL-1β and IL-1α compared to WT mice, whereas TNFα levels were similar (Fig. 3A-C). The number of inflammatory cells in the peritoneum was not different between the two genotypes (Fig. 3D).

Figure 3. mPPARγKO macrophages produce less IL-1 cytokines in vivo.

(A-C) WT or mPPARγKO mice were injected with thioglycollate to induce macrophage recruitment to the peritoneal cavity. At day 4, when ~90% of the cells in the peritoneum are macrophages, the mice were given an intraperitoneal injection of LPS (10 μg/200 μl) and 16h later IL-1β (A), IL-1α (B), and TNFα (C) levels in the peritoneal fluid was quantified by ELISA. (D) Day 4 peritoneal cell count for WT vs. mPPARγKO mice following thioglycollate injection. Bar graphs report the mean ± standard error (SE) and the individual dots each represent one mouse. The p values for the comparison of WT vs. mPPARγKO mice are shown.

Reduced IL-1 release in PPARγ KO macrophages occurs independent of alterations in succinate, nitric oxide production, and IL-10

IL-1β and IL-1α signal via the same receptor and have similar biologic effects. As both family members are relevant to cardiometabolic disease and host defense we sought to understand the mechanism linking PPARγ to IL-1 suppression. Recently it has become clear that cellular metabolism can modulate inflammatory cytokine production in macrophages (41, 42). One relevant example is the small molecule succinate, which is an intermediate of the Krebs cycle. Succinate is known to accumulate in LPS activated macrophages and it has been linked to increased IL-1β transcription via HIF-1α, although the mechanism is not completely understood (43). Succinate levels increase after LPS in part due to the inhibition of succinate dehydrogenase by itaconate, a product of the enzyme Irg1, which is induced by TLR4 signaling (Supplemental Fig. 3A) (44). However, Irg1 mRNA expression was similar between the genotypes after activation and the intracellular levels of succinate were slightly increased, rather than decreased, in mPPARγ KO cells (Supplemental Fig. 3B, C). Inhibition of glycolytic flux has also been associated with reduced IL-1β production (45). However, assessment of lactate:pyruvate ratio by metabolomics or direct measurements extracellular acidification rate (ECAR), a surrogate of anaerobic glycolysis, via Seahorse metabolic flux analyzer measurements revealed only a slight reduction of aerobic glycolysis in activated mPPARγKO cells (Supplemental Fig. 3 D,E). Together these data argue against a critical role for these described metabolic pathways in phenotype of PPARγ KO macrophages.

PPARγ is known to negatively regulate expression of inducible nitric oxide synthase (iNOS) in response to LPS in part through a sumylation-dependent mechanism (24). The production of nitric oxide (NO) has been shown to suppress inflammasome activation, leading us to investigate this pathway (46). In agreement with published studies, we observed enhanced expression of iNOS transcript in mPPARγKO macrophages compared to WT cells after palm-LPS stimulation (Fig. 4A). NO production was also slightly increased in PPARγ KO cells compared to WT in response palm-LPS (Fig. 4B). To determine the biologic impact of the NO in this system, a selective inhibitor of iNOS, L-NIL, was added to the cells at the start of the experiment. As a control, we demonstrated that L-NIL completely blocked NO production induced by 500U of IFNβ in combination with palm-LPS (Fig. 4B). However, L-NIL only slightly impacted IL-1β release in WT cells and did not restore IL-1 release to WT levels in PPARγKO cells (Fig. 4C).

Figure 4. The suppression of IL-1 release from PPARγ KO macrophages is independent of nitric oxide and IL-10.

(A) iNOS mRNA levels in WT vs. PPARγKO pMACs 8h after the indicated stimulation. (B) WT or mPPARγKO pMACS were treated with vehicle or palm-LPS and nitrite levels were quantified via colorimetric assay. At the same time, WT macrophages were treated with palm-LPS together with IFNβ (500U) in the presence or absence of the iNOS inhibitor L-NIL (10 μM). (C) pMACs from WT or KO mice were treated with palm-LPS in the presence of vehicle or L-NIL and IL-1β release was quantified by ELISA. (D) WT macrophages were treated with palm-LPS for 20h in the presence of vehicle (open bars) or recombinant IL-10 (20 ng/ml; gray bars) ± α-IL-10 receptor blocking antibody (IL-10Rab) and IL-1β release was measured by ELISA. (E) pMACs from WT or mPPARγKO mice were treated with palm-LPS for 20h in the presence of control antibody or IL-10Rab and IL-1β release was quantified by ELISA. Bar graphs report the mean ± standard error (SE) for a minimum of 3 experiments, each performed in triplicate. *, p<0.05 for WT vs. mPPARγKO, #, p<0.05 for veh vs. L-NIL or IL-10Rab. ND, not detected

The anti-inflammatory cytokine IL-10 has also been linked to the suppression of IL-1 production and we noted that IL-10 mRNA levels were increased slightly in the mPPARγKO cells (47). In our system, the addition of recombinant IL-10 was able to decrease IL-1β release by ~ 50% in activated WT cells and this affect was prevented by an IL-10 receptor blocking antibody (Fig. 4D). In contrast, inhibition of IL-10 signaling in mPPARγKO cells did not significantly alter IL-1β release (Fig. 4E). Taken together this data suggests that the inflammatory phenotype of PPARγ KO macrophages occurs independently of changes in NO or IL-10.

Production of type 1 IFN is enhanced in activated PPARγ deficient macrophages

PPARγ is a transcription factor that can both activate or repress gene expression based on its binding partners. Prior data has established that PPARγ does not bind to the promoter of IL-1β or directly influence its transcription in macrophages (48). Therefore, to understand how PPARγ deficiency might alter IL-1β mRNA we performed RNA sequencing on WT and PPARγ KO cells at baseline or following activation for 8h. We compared differences in gene expression between WT and PPARγKO macrophages stimulated with palm-LPS using a cut off of a 2.0-fold change in expression. GO pathway analysis of the top twenty upregulated pathways in activated PPARγKO compared to WT macrophages revealed an enrichment of pathways involved in anti-viral defense and IFNβ (Fig. 5A). Consistent with this, several pathways related to the release of and response to IFNβ were enhanced in PPARγ deficient cells (Fig. 5B). As illustrated in figure 5C, heatmap analysis of gene expression in the response to IFNβ module revealed a modest, but consistent increase in the expression of IFN target genes, including IFNβ itself, in PPARγKO macrophages. These results suggested that loss of PPARγ may engage the type 1 interferon pathway in macrophages.

Figure 5. RNA sequencing reveals type 1 interferon gene signature in activated PPARγKO macrophages.

(A, B) WT or PPARγKO pMACs were treated with BSA-PBS or palm-LPS for 8h after which RNA was isolated and RNA sequencing was performed. (A) Group summary for pathways related to IFNβ in PPARg vs. WT pMACS treated with LPS. (B) Heatmap expression profile of genes from the cellular response to IFNβ GO biologic processes module. WT and PPARγKO cells are shown under basal conditions (BSA-PBS) and after activation (palm-LPS). The color map for fold change is shown.

To validate these findings, we performed qPCR for several type 1 interferon gene targets. In line with the RNA sequencing data, mRNA levels of MX1, CXCL10, ISG15, and CCL2 were increased in PPARγ KO macrophages after palm-LPS activation. In contrast, acyl-CoA synthetase 1 (ACSL1) expression, which is induced by LPS via a TRIF-dependent, IFN-independent mechanism, was similar between the genotypes (Fig. 6A) (49). Type 1 interferon receptor (IFNAR1) signaling leads to phosphorylation of the transcription factor STAT1, which drives IFN-dependent gene expression. STAT1 phosphorylation was also augmented in mPPARγKO macrophages compared to WT cells (Fig. 6B). The expression of IFNAR was similar between WT and mPPARγKO cells (Fig. 6C, D). However, mRNA levels and secretion of IFNβ were increased in the KO macrophages (Fig. 6E, F). The expression of TRIF was similar between the genotypes (Fig. 6G). Thus, PPARγKO macrophages produce increased IFNβ in response to activation, which results in augmentation activation of the downstream signaling pathways.

Figure 6. Augmented expression and release of IFNβ in PPARγ deficient macrophages in response to palm-LPS.

(A) pMACs isolated from WT (open bars) or mPPARγKO mice (filled bars) were treated with vehicle or palm-LPS for 8h and mRNA expression of the indicated targets was determined by qRT-PCR. (B) Macrophages were treated with vehicle or palm-LPS for 4h or 8h after which cell lysates were isolated and phopho(P)-STAT1 (Y701) was assessed by western blotting. Total (tot) STAT1 and tubulin are shown as controls. (C, D) pMACs from WT or PPARγ KO mice were stained with an antibody the type 1 interferon receptor (IFNAR) and surface expression was assessed by flow cytometry. A representative histogram (C) and grouped MFI data (D) are shown. (E) mRNA levels of IFNβ were quantified in WT and mPPARγKO cells at 1h after palm-LPS treatment via qRT-PCR. (F) IFNβ release from WT and KO pMACs 6h after palm-LPS stimulation. (G) Protein was isolated from WT or PPARγKO macrophages and TRIF protein expression was assessed by western blotting. Bar graphs report the mean ± standard error (SE) for a minimum of 3 experiments, each performed in triplicate. *, p<0.05 for WT vs. mPPARγKO; ns, non-significant

Enhanced type 1 IFN signaling accounts for the suppression of IL-1 release in PPARγ deficient macrophages

Type 1 interferon has been reported to inhibit or activate the NLRP3 inflammasome depending on the context (50). To determine whether increased IFNβ was responsible for the IL-1 phenotype in mPPARγKO macrophages we utilized pharmacologic and genetic approaches. In WT cells, inhibition of type 1 IFN signaling with a IFNAR1 blocking antibody lead to a modest increase in IL-1β release after palm-LPS stimulation. In contrast, IFNAR1 blockade completely reversed the phenotype of PPARγKO cells (Fig. 7A). Consistent with these results, the suppression of IL-1β release by the PPARγ antagonist was blunted in IFNRKO macrophages (Fig. 7B). Western blot and mRNA analysis confirmed a similar phenotype at the level of pro-IL-1β expression (Fig. 7C, D). To evaluate whether PPARγKO macrophages may also have enhanced sensitivity to type 1 IFN, we stimulated WT and KO cells with recombinant IFNβ and assessed STAT1-phosphylation and CXCL10 induction. There were no differences between WT and PPARγ deficient macrophages with either of these assays, further arguing that increased IFNβ release, not enhanced IFNR1 signaling accounts for the observed phenotype (Fig. 7E, F). Thus, increased IFNβ release accounts for the diminished IL-1 production in PPARγKO cells.

Figure 7. Suppression of IL-1 expression occurs via a type 1 interferon dependent mechanism.

(A) WT or PPARγKO pMACs were treated with palm-LPS for 20h in the presence of an IFNAR blocking antibody or control Ig and IL-1β release was quantified by ELISA. (B) WT or IFNAR KO pMACs were treated with veh or the PPARγ antagonist T0070907 24h prior to stimulation with palm-LPS and IL-1β release was quantified by ELISA. (C, D) WT or PPARγKO pMACs were treated with palm-LPS for 8h (mRNA) or 16h (protein) in the presence of an IFNAR blocking antibody or control Ig and IL-1β expression was assessed via qRT-PCR (C) and western blotting (D). (E) WT or KO pMACs were stimulated with palm-LPS or IFNβ (500U) for 4h and P-STAT was assessed by western blotting. (F) pMACs from WT or mPPARγKO mice were treated with IFNβ and mRNA expression of the IFN gene target CXCL10 was assessed via qRT-PCR. Bar graphs report the mean ± standard error (SE) for a minimum of 3 experiments, each performed in triplicate. *, p<0.05 for WT vs. mPPARγKO or; ns, non-significant

Low doses of IFNβ mimic the biologic effects of PPARγ loss of function on IL-1 release

Although IFNβ release was increased in mPPARγKO cells, the absolute amount of type 1 interferon produced was modest. To investigate the impact of low dose IFNβ on the expression of IL-1 we treated WT macrophages with IFNβ in conjunction with palm-LPS and monitored cytokine production. Doses of IFNβ as low as 25U significantly suppressed IL-1β release, while having no effect on TNFα (Fig. 8A, B). Similar to what was observed in PPARγ KO cells, both mRNA and protein levels of pro-IL-1β were reduced with IFNβ treatment (Fig. 8C, D). Moreover, the suppression by IFNβ was prevented with a blocking antibody to IFNAR (Fig. 8E). Conversely, pMACs from mice lacking IFNAR had elevated levels of IL-1β mRNA and protein in response to palm-LPS stimulation (Fig. 8F, G). In sum, this data demonstrates that low concentrations of type 1 interferon can potently suppress IL-1 cytokine production in a manner that recapitulates the findings obtained with PPARγ deficient macrophages.

Figure 8. Low dose IFNβ phenocopies PPARγ loss-of-function in primary macrophages.

(A, B) WT pMACs were treated with palm-LPS and increasing concentrations of IFNβ after which IL-1β (A) and TNFα (B) release was quantified by ELISA. (C) Macrophages were treated with vehicle or palm-LPS for 16h in the presence of IFNβ (50U) and pro-IL-1β and NLRP3 protein levels were assessed by western blotting. Tubulin is shown as a loading control. (D) pMACS were treated as indicated and gene expression of IL-1β and CXCL10 was assessed 8h after stimulation via qRT-PCR. (E) pMACS were stimulated with palm-LPS ± IFNβ in the presence of IFNAR blocking ab or control Ig and IL-1β release was determined by ELISA. (F, G) WT (open bar) or IFNAR KO (gray bar) macrophages were stimulated with palm-LPS and IL-1β mRNA (F) or protein (G) levels were assessed at 8h and 16h, respectively. Bar graphs report the mean ± standard error (SE) for a minimum of 3 experiments, each performed in triplicate. *, p<0.05 for veh vs. IFNβ or WT vs. IFNAR KO; ns, non-significant

PPARγ agonists suppress IFNβ expression

PPARγ can be activated by agonists such as rosiglitazone and pioglitazone. To determine if increasing PPARγ activity could suppress IFNβ production we investigated the impact of rosiglitazone on the expression of IFNβ mRNA and type 1 IFN gene targets. Primary macrophages were incubated with 1 μM rosiglitazone for 16h prior to the addition of veh or palm-LPS. As shown previously, PPARγKO cells had elevated levels of IFNβ mRNA early after stimulation with palm-LPS. Rosiglitazone suppressed IFNβ by ~50% in WT cells, whereas it had no effect on the PPARγKO macrophages (Fig. 9A). A similar pattern was observed for other IFN-regulated genes including iNOS, CXCL10, and MX1 (Fig. 9B-D). In contrast, TNFα expression was not affected by either rosiglitazone or loss of PPARγ. Interestingly, IL-1β mRNA levels were substantially decreased in PPARγKO macrophages, but rosiglitazone did not increase or decrease expression of this cytokine (Fig. 9E, F). When macrophages were treated with a higher concentration of rosiglitazone (10 μM) the expression of multiple IFN gene targets in both WT and PPARγKO macrophages was reduced, demonstrating that at higher doses PPARγ independent effects occur (data not shown)(51).

Figure 9. PPARγ activation suppresses IFNβ.

(A-F) WT or PPARγ KO macrophages were treated with BSA-PBS/veh (open bars), palm-LPS/veh (filled bars), palm-LPS/rosiglitazone 1 μM (gray bars) for 16h followed by stimulation with palm-LPS for 8h in continued presence of the PPARγ agonist. mRNA was isolated from the macrophages and the expression of IFNβ (A) and several of its gene targets (B-D) was assessed by qRT-PCR. In addition, mRNA expression of the pro-inflammatory cytokines IL-1β (E) and TNFα (F) was also determined. (G-H) WT or KO pMACs were treated with rosiglitazone or vehicle as described above followed by palm-LPS for 16h. The release of IFNβ (G) and IL-1β (H) was quantified by ELISA. Bar graphs report the mean ± standard error (SE) for a minimum of 3 experiments, each performed in triplicate. *, p<0.05 for veh vs. rosiglitazone; ns, non-significant

Sterol synthesis genes are suppressed in PPARγ KO macrophages

In an effort to understand the mechanism by which PPARγ influences IFNβ production we returned to our RNA sequencing analysis. Of particular interest, PPARγKO cells had a marked decrease in the expression of several genes involved in the sterol biosynthesis pathway, which has been linked to enhanced IFN production (Fig. 10A, B) (52, 53). To confirm these findings, we performed qRT-PCR to assess transcript levels for several enzymes involved in sterol biosynthesis and found that multiple genes in this pathway were significantly downregulated in PPARγKO macrophages (Fig. 10C). Moreover, rosiglitazone increased the expression of these genes in WT macrophages, whereas it failed to do so in KO cells (Fig. 10D).

Figure 10. Expression of genes involved in sterol biosynthesis are suppressed in PPARγ deficient cells.

(A) Heatmap display of gene expression from sterol biosynthesis pathway obtained from RNA sequencing data macrophages treated with BSA-PBS or palm-LPS for 8h. The brackets indicate genes with particularly low expression in activated PPARγ KO cells (blue). (B) schematic of the metabolic pathway and enzymes (red text) involved in sterol biosynthesis. (C) mRNA expression of the indicated sterol biosynthesis genes was determined via qRT-PCR in WT and PPARγKO cells treated with control or palm-LPS for 8h. (D) pMACs from WT or PPARγ deficient macrophages were treated with 1 μM rosiglitazone (rosi) or vehicle (veh) for 16h after which they were treated with palm-LPS and the expression of sterol biosynthesis genes was assessed by qRT-PCR. (E) pMACs from WT or PPARγ deficient macrophages were treated with BSA-PBS or palm-LPS for 8h after which cells were lysed and free cholesterol, 25-hydroxycholesterol (HC), and 27-HC were measured by GC-MS/MS. Data is presented as normalized ratio an internal standard. Bar graphs report the mean ± standard error (SE) for a minimum of 3 experiments, each performed in triplicate. *, p<0.05 for WT vs. mPPARγKO; #, p<0.05 for veh vs. rosi; ns, non-significant

Based on these observations, we quantified the level of free cholesterol and the enzymatically derived oxysterols 25-hydroxcholesterol (25-HC) and 27-hydroxycholesterol at baseline and with activation (Fig. 10E). 25-HC levels increased in macrophages of both genotypes, consistent with the known induction of 25-hydroxycholesterolase expression by TLR4. Free cholesterol and 27-HC showed a similar pattern in which levels were slightly higher at baseline in macrophages from PPARγKO mice, but dropped more with activation. Consistent with this, the addition of cholesterol did not impact IL-1 release from WT or PPARγKO cells, indicating that cholesterol per se is unlikely to be driver of IFNβ production.

The disruption of sterol biosynthesis in macrophages has been reported to augment IFNβ production through a mechanism that involves enhanced activation of the innate immune receptor STING (53). In contrast to TLR4, STING resides in the ER and activates IFN production independent of the adaptor TRIF, but still requires the downstream transcription factor IRF3 (54). As shown in Figure 11, PPARγ cells treated with the STING activator cGAMP also expressed higher levels of IFN target genes. Interestingly, the augmented response to STING activation became less apparent at increasing doses of cGAMP (Fig. 11D, E). Thus, PPARγ deficiency modulates the sensitivity of macrophages to STING activation without changing the maximal activation response. To confirm that the cGAMP activation was dependent on STING, we treated macrophages with a knockin (KI) of an inactive variant of STING with high dose cGAMP and demonstrated no induction of IFN target genes (Fig. 11F). To investigate whether all IRF3 activation stimuli produce a similar phenotype in PPARγKO macrophages, we also transfected macrophages with poly I:C (PIC) to activate the mitochondrial antiviral signaling protein (MAVS). In contrast to STING activation, PPARγ KO cells transfected with PIC did not have augmented IFN gene expression. In fact, PPARγKO cells expressed lower levels of IFN gene targets relative to WT macrophages (Fig. 11G-I). As MAVS and STING induce IFNβ through same signaling pathway this data suggests that differences in IRF3 activation are unlikely to mediate the PPARγKO phenotype.

Figure 11. PPARγ KO macrophages have enhanced sensitivity to STING ligands.

(A-C) WT or PPARγKO pMACs were stimulated with the STING activator cGAMP (0.5 μg/ml) or palm-LPS and mRNA was harvested 8h later. The expression of type 1 interferon gene targets CXCL10 (A), MX1 (B), and iNOS (C) was assessed via qRT-PCR. (D, E) pMACs from WT or PPARγ deficient macrophages were stimulated with the indicated concentrations of cGAMP and the expression of CXCL10 (D) and iNOS (E) was assessed by qRT-PCR. (F) WT or STING knock-in mice (KI) were treated with PBS or 10 μg/ml of cGAMP and the mRNA expression of CXCL10 was determined at 8h. (G-I) WT or PPARγKO pMACs were treated with cGAMP (0.5 μg/ml) or transfected with PIC for 8h and the expression of the indicated IFN gene targets was assessed by qRT-PCR. Bar graphs report the mean ± standard error (SE) for a minimum of 3 experiments, each performed in triplicate. *, p<0.05 for WT vs. mPPARγKO or STING KI; ns, non-significant

Discussion

In this study, we unravel an unexpected connection between PPARγ and IL-1 expression that is linked by alterations in type 1 interferon signaling. We initially demonstrated that PPARγ KO macrophages produce less IL-1β and IL-1α in response to NLRP3 activators in vitro and in vivo, while having a preserved TNFα response. The suppression of IL-1 cytokines occurred at the transcriptional level and resulted in lower expression of the pre-curser cytokine proteins. RNA sequencing uncovered a type 1 interferon signature in PPARγKO macrophages that was driven by enhanced production of IFNβ. Using gain and loss-of-function approaches we established that IFNβ is required for the suppression of IL-1 expression in PPARγKO cells. We also demonstrate that activation of PPARγ with rosiglitazone diminishes IFNβ mRNA levels. Finally, we provide evidence that this phenotype may be driven by altered sterol biosynthesis and enhanced activity of the innate immune receptor STING.

The NLRP3 inflammasome is a tightly regulated inflammatory complex that requires two distinct signals for activation. Importantly, NLRP3 and IL-1β have been implicated in the pathogenesis of several human diseases including atherosclerosis, diabetes, non-alcoholic steatohepatitis, Alzheimer’s disease, gout, and heart failure (55–59). Relevant to diabetes and vascular disease we have previously shown that excess fatty acids can cause lysosome damage in macrophages triggering NLRP3 assembly and enhanced IL-1β release (11). In this study, we also demonstrate that IL-1α release is increased under these same conditions. This observation is consistent with previous data that the majority of NLRP3 activators also trigger the release of IL-1α (60). As an approach to investigate how lipid metabolism might influence inflammasome activation we assessed IL-1 cytokine release in macrophages lacking the lipid metabolic regulator PPARγ. Contrary to the notion that PPARγ has general anti-inflammatory activity, we discovered that loss of this transcription factor resulted in diminished IL-1β and IL-1α release while have no effect on TNFα secretion. Loss of PPARγ was associated with reduced levels of IL-1β and IL-1α protein and mRNA indicating that priming was attenuated in PPARγ KO macrophages. Moreover, decreased IL-1 cytokine release occurred in response to non-lipid NLRP3 activators as well, demonstrating that this phenotype is broader than just lipotoxicity.

PPARγ is a transcription factor with both activating and repressing capabilities. Prior studies of PPARγ-mediated repression of inflammatory cytokines have shown that this transcription factor does not directly affect IL-1β transcription, suggesting an indirect link between these events (48). Based on prior studies demonstrating that glycolytic flux and succinate accumulation can modulate IL-1β transcription, we initially hypothesized that alterations in cellular metabolism and/or intracellular metabolites might explain the relationship between PPARγ and IL-1 (43). However, these metabolic pathways were not significantly altered in PPARγ deficient macrophages. Instead, using an RNA sequencing approach, we uncovered a signature of enhanced IFNβ signaling in PPARγ KO cells. Using gain and loss of function approaches we demonstrated that the enhanced IFNβ response was responsible for the suppression of IL-1 family members in PPARγ deficient primary macrophages. Our findings of enhanced expression of IFNβ target genes in mPPARγKO macrophages are in line with previous observations made by Welch et al where they used a distinct inducible Cre system to delete PPARγ (51). Although not commented upon in their manuscript, the expression of IL-1β mRNA via northern blot was decreased in LPS treated PPARγ deficient cells. It is unclear how these findings relate to a subsequent study suggesting that IL-1β release is increased in PPARγ KO macrophages (61). However, as that the link between PPARγ and IL-1 occurs via IFN, this interplay could be context dependent. Using the PPARγ agonist rosiglitazone we found that the expression of IFNβ mRNA and its downstream targets was suppressed by PPARγ activation. In contrast, TNFα mRNA expression was not affected. This data is consistent with another study, where the PPARγ agonist troglitazone was shown to suppress IFNβ transcription (62). In sum, our results add to a growing body of literature demonstrating an antagonist relationship between PPARγ expression and interferon mediated inflammatory responses. Also of relevance, PPARγ has been implicated in macrophage alternative activation with IL-4, which is thought to be protective in response to obesity related inflammation (63, 64). Understanding the links between PPARγ, type 1 IFN, and IL-1 has the potential to shed significant insight into macrophage phenotypes and effector responses in vivo.

To gain insight into the mechanism linking PPARγ loss-of function to enhanced release of IFNβ we returned to our RNA sequencing analysis. PPARγKO cells had more robust downregulation of genes involved in sterol biosynthesis at baseline and with activation compared to WT macrophages. Moreover, PPARγ activation augmented expression of these genes. These findings are of interest as recent data has correlated disruption of flux through the sterol biosynthesis pathway with enhanced IFNβ release from macrophages (53). However, direct quantification of free cholesterol levels in PPARγKO macrophages revealed slightly increased levels in the KO cells. This observation may represent defective cholesterol efflux in PPARγ deficient cells (27). Upon activation, KO cells had a more dramatic decrease in free cholesterol compared to WT cells suggesting that sterol pathways are differentially effected; however, the addition of cholesterol to WT or PPARγKO macrophages did not change IFNβ or IL-1 production. Although this data argues against a direct role for cholesterol itself, flux through the de novo sterol biosynthesis pathway also generates other lipid and metabolic intermediates. One such intermediate is the oxysterol 25-HC, which is known to increase with TLR4 activation due to the induction of 25-hydroxycholesterolase. Prior studies have shown that 25-HC can suppress IL-1β by decreasing SREBP activation (52). However, we found that 25-HC levels increased to a similar extent in macrophages from both genotypes in response to palm-LPS. These observations raise the possibility that PPARγ may influence other lipid intermediates/metabolites or directly suppress IRF3.

To gain insight into the role of IRF3 in the IFN phenotype, we compared IFNβ responses in WT and PPARγKO macrophages stimulated with ligands for the intracellular pathogen receptors STING or MAVS. STING resides in the ER whereas MAVS is located in the mitochondrial outer membrane. Both receptors activate IRF3 via a shared signaling pathway; however, STING is also known to regulated by changes in membrane lipid composition. Since IFNβ release was augmented in response to STING ligands, but not MAVS ligands a direct role for PPARγ on IRF3 is less likely. Together this data along with our gene expression results supports a model whereby the IFNβ response is linked to alternations in cellular lipids/sterols.

Regulation of the NLRP3 inflammasome by type 1 interferon is complex with both activating and inhibitory properties described depending on the context. One mechanism by which type 1 IFN can suppress IL-1 production in macrophages is by inducing the expression of IL-10, a potent anti-inflammatory cytokine (65). However, we demonstrated that inhibition of IL-10 signaling did not restore IL-1 release in PPARγ KO cells. Moreover, we determined that the stability of IL-1β mRNA was not affected in PPARγ KO cells suggesting that transcription of this cytokine was directly suppressed by interferon signaling. Although the mechanism(s) underlying IFNβ -mediated suppression of IL-1β and IL-1α in PPARγKO macrophages is not clear, this area remains a focus of ongoing research.

The observation that PPARγ KO macrophages can overproduce type 1 IFN may also be relevant to other metabolic and autoimmune phenotypes described in mPPARγKO mice. In the setting of high fat diet, mPPARγKO mice have worsened insulin resistance and increased atherosclerosis, both of which are associated with chronic inflammation (21, 25–27). Moreover, these mice are also predisposed to autoimmune diseases (66, 67). Our findings raise the intriguing possibility that an enhanced type 1 interferon response may contribute to some of these observations. The fact that type 1 IFN can promote autoimmunity and atherosclerosis is line with this hypothesis (68–70). Future research into this possibility could be of interest as the PPARγ agonists rosiglitazone and pioglitazone are FDA approved and thus could potentially be used to suppress IFNβ production in patients with autoimmune disease (71–73).

Herein we demonstrate that PPARγ deficient macrophages have a defect in IL-1β and IL-1α release in response to treatment with NLRP3 activators. Unexpectedly, we determined that this response was driven by an overproduction of IFNβ in PPARγKO cells. Our findings represent another important example of the complex interplay between metabolic and inflammatory pathways in macrophages that could have relevance to obesity-associated diseases and autoimmunity.