Buffy coats were obtained from healthy blood donors, as anonymously provided by the Comunidad de Madrid blood Bank. Ethical approvals for all blood sources and processes used in this study were approved by the Centro de Investigaciones Biológicas Ethics Committee. All experiments were carried out in accordance with the approved guidelines and regulations. Human PBMCs were isolated from buffy coats over a Lymphoprep™ gradient (#1114545, Axis-Shield PoC AS) according to standard procedures. Monocytes were purified from PBMCs by magnetic cell sorting using human CD14 microbeads (#130-050-201, Miltenyi Biotech). Monocytes (95% CD14⁺ cells) were cultured at 0.5 × 10⁶ cells/ml for 7 days in RPMI 1640 (#21875-034, Gibco) supplemented with 10% inactivated fetal calf serum (FCS) (#S1810-500, Biowest) (complete medium), at 37°C in a humidified atmosphere with 5% CO₂, and containing 1,000 U/ml human GM-CSF (#11343125, Immunotools GmbH) or 10 ng/ml human M-CSF (#11343115, Immunotools GmbH), to generate GM-MØ or M-MØ, respectively. Cytokines were added every 2 days. Blocking anti-activin A Ab (100 ng/ml) (#MAB3381, clone 69403, R&D Systems) or the inhibitors of ALK4, ALK5, and ALK7, SB431542 (10 µM) (#S4317, Sigma-Aldrich) or A-83 (1 µM) (#2039, Tocris) were added every 24 h. Finally, polarized macrophages were treated with ultrapure Escherichia coli 0111:B4 strain LPS (10 ng/ml) (#tlrl-3pelps, Invivogen) for 14–16 h. Exposure to recombinant human activin A (25 ng/ml) (#120-14P, Preprotech) was done for 24 h (monocytes and THP-1 cells) or 48 h (M-MØ). The acute monocytic leukemia cell line THP-1, obtained from ATCC^® (#TIB-202™), was cultured in complete medium at 37°C in a humidified atmosphere with 5% CO₂. A-MØ were obtained from patients undergoing bronchoalveolar lavage (BAL) following the Fundación Jiménez Díaz Medical Ethics committee procedures and after written informed consent from all subjects, in accordance with the Declaration of Helsinki. BAL procedure was performed with a flexible bronchoscope with a total volume of 200 ml of sterile isotonic saline solution at 37°C. BAL fluid fractions were maintained at 4°C and cellular debris removed using a 40 µm cell strainer (45). BAL cells were washed with PBS, centrifuged and resuspended in complete medium containing 100 U/mL penicillin and 100 µg/mL streptomycin (#15140-122, Gibco), 50 µg/ml gentamicin (#G1397, Sigma-Aldrich), and 2.5 µg/ml amphotericin B (#A2942, Sigma-Aldrich). The cells were seeded at 6–8 × 10⁵ cells per well in 12-well plates for 1 h and washed extensively to remove non-adherent cells. Finally, 2 ml of complete medium with antibiotics was added to each well and the adherent cells incubated for 16–18 h before transfection. More than 95% of adherent BAL cells were identified as macrophages according to morphology and phenotypic analysis.

All experiments on mice were conducted according to the Spanish and European regulations on care and protection of laboratory animals and were approved by the Centro de Investigaciones Biológicas animal facility and the Consejo Superior de Investigaciones Científicas Ethics Committee. Bone marrow-derived GM-MØ or M-MØ were obtained by flushing the femurs of 6–10-week-old C57BL/6 mice (provided by the Animal facility at the Centro de Investigaciones Biológicas), and culturing cells during 7 days in DMEM (#41966-029, Gibco) supplemented with 10% FCS and 50 mM 2-ME, containing either murine GM-CSF (1,000 U/ml) (#315-03, PreProtech) or human M-CSF (25 ng/ml) (#11343115, Immunotools GmbH), respectively (46, 47). Cytokines were added every 2 days.

Mouse monoclonal antibodies specific for human CD14 (Alexa Fluor-647-labeled antihuman CD14, #301818, clone M5E2, Biolegend) and human CD163 (PerCP-labeled antihuman CD163, #333625, clone GHI/61, Biolegend) were used. Isotype-matched PerCP-labeled Mouse IgG1 (κ Isotype Ctrl Antibody, #400147, clone MOPC-21, Biolegend) and Alexa Fluor-647 Mouse IgG2a (κ Isotype Ctrl Antibody, #400234, clone MOPC-173, Biolegend) were included as negative controls.

Total RNA was extracted using the total RNA and protein isolation kit (Macherey-Nagel). RNA samples were retrotranscribed with the High-Capacity cDNA Reverse Transcription kit (AB), and individually amplified cDNA was quantified using the Universal Human Probe Roche library (Roche Diagnostics). Oligonucleotides for selected genes were designed according to the Roche software for quantitative real-time PCR (qRT-PCR), and their sequence is indicated in Table S1 in Supplementary Material. qRT-PCR was performed on a LightCycler^® 480 (Roche Diagnostics). Assays were made in triplicates, and results were normalized according to the expression levels of TBP mRNA or/and GAPDH mRNA (for qRT-PCR) or to the mean of the expression level of endogenous reference genes HPRT1, TBP and RPLP0 (for microfluidic gene cards). Results were expressed using the ΔΔCT (cycle threshold) method for quantification.

Macrophage supernatants were tested for the presence of cytokines using commercially available ELISA sets for human TNFα (BD OptEIA Human TNF ELISA set, #555212, BD Biosciences), CCL2 (BD OptEIA Human MCP-1 ELISA set, #555179, BD Biosciences), IL-10 (ELISA MAX Standard set, #430601, BioLegend), IL-6 (ELISA MAX Standard set, #430501, BioLegend), and activin A (DuoSet, #DY338, R&D Systems), following the protocols supplied by the manufacturers.

HEK293-T cells, provided by the Cell culture facility at the Centro de Investigaciones Biológicas, were transfected with an expression vector for PPARγ2 (pBABE-PPARγ2, Addgene) or an empty vector using Superfect transfection reagent (#301305, Qiagen). Human GM-MØ or M-MØ (1 × 10⁶ cells) were transfected using the Human Macrophage Nucleofector^® Kit (#VPA-1008, Lonza) with 1 µg of PPAR reporter DNA mixture (#CCS-3026L, Cignal PPAR Reporter assay kit, Qiagen). This mixture contains a PPAR-responsive firefly luciferase construct and a constitutively expressing Renilla luciferase (40:1) The PPAR-dependent construct encodes the firefly luciferase gene under the control of a minimal CMV promoter and tandem repeats of the PPAR responsive element (PPRE). Firefly and Renilla luciferase activities were determined by using the Dual-Luciferase^® Reporter Assay System (#E1910, Promega).

Cell lysates (40 µg) and nuclear extracts (30 µg) were subjected to SDS-PAGE and transferred onto an Immobilon polyvinylidene difluoride membrane (Millipore, Bedford, MA, USA). After blocking the unoccupied sites with 5% nonfat dry milk, protein detection was carried out with a goat polyclonal against PPARγ2 (G-18, #sc-22020, Santa Cruz Biotechnology), a goat affinity purified polyclonal antibody against Sp1 (PEP2, #sc-59-G, Santa Cruz Biotechnology), or a monoclonal antibody against GAPDH (6C5, #sc-32233, Santa Cruz Biotechnology), and using the SuperSignal West Pico Chemiluminescent system (#34081, Thermo Fisher Scientific).

To silence PPARG gene expression, human GM-MØ, M-MØ (1 × 10⁶ cells) or A-MØ (6–8 × 10⁵ cells) were transfected with a PPARG-specific siRNA (siPPARG) (50 nM) (#s10888, Thermo Fisher Scientific), using HiPerFect transfection reagent (#301705, Qiagen). A negative control siRNA from the same company was used as a transfection control (siControl) (#4390843, Thermo Fisher Scientific). After 6 h of transfection, cells were allowed to recover from transfection in RPMI 1640 medium with 10% FCS and the cells were treated with GW7845 (1 µM) (kindly provided by Jon Collins, Glaxo SmithKline, USA) or DMSO for 18–24 h before assessing for PPARγ markers.

Global gene expression analysis was performed on RNA obtained from three independent samples of GM-MØ that had been transfected with siPPARG or siControl for 48 h, and using a whole human genome microarray from Agilent Technologies (Palo Alto, CA, USA). Only probes with signal values >60% quantile in at least one condition were considered for the differential expression and statistical analysis. Statistical analysis for differential gene expression was carried out using empirical Bayes moderated t test implemented in the limma package¹ and using paired t-test. All the above procedures were coded in R.² Microarray data were deposited in the Gene Expression Omnibus³ under accession no. GSE88768. The differentially expressed genes were analyzed for annotated gene sets enrichment using the online tool ENRICHR⁴ (48, 49). Enrichment terms were considered significant when they had a Benjamini-Hochberg-adjusted p value < 0.05. For gene set enrichment analysis (GSEA) (50), the previously defined “Proinflammatory gene set” and “Anti-inflammatory gene set” (12), which contain the top and bottom 150 probes from the GM-MØ versus M-MØ limma analysis of the microarray data in GSE68061 (ranked on the basis of the value of the t statistic), were used.

Statistical analysis was performed using paired Student’s t-test, and p < 0.05 was considered significant (*p < 0.05, **p < 0.01, and ***p < 0.001).

To initially assess the PPARγ activation-dependent transcriptional profile of GM-MØ and M-MØ, both human macrophage subtypes were exposed for 24 h to the PPARγ agonist GW7845 and the expression of the GM-MØ-specific “Proinflammatory gene set” and M-MØ-specific “Anti-inflammatory gene set” (derived from the data contained in the Gene Expression Omnibus GSE68061) (11, 12) was determined. PPARγ activation upregulated the paradigmatic PPARγ target genes CD36 and FABP4, and downregulated FLT1 and CSF1 expression, in both macrophage subtypes (Figure 1A). However, GW7845 downregulated IL6, IL10, CCL2, HAMP, and CCR2 and enhanced THBS1, exclusively in GM-MØ (Figure 1A). These GW7845-triggered gene expression changes were dependent on PPARγ activation as they were significantly impaired upon siRNA-mediated knockdown of PPARG mRNA (Figures 1B,C). Specifically, PPARG mRNA knockdown inhibited the GW7845-mediated modulation of CD36 and CSF1 expression in M-MØ (Figure 1B) and significantly impaired the GW7845-mediated modulation of CD36, CSF1, FLT1, CCL2, CCR2, IL10, and HAMP in GM-MØ (Figure 1C). Analogous findings were observed in murine bone marrow-derived macrophages, where Pparγ activation modified the expression of a common set of genes in both macrophage subtypes but significantly diminished the expression of Csf1 and Ccr2 only in GM-MØ (Figure 2). Therefore, although PPARγ activation alters the expression of known PPARγ targets in both GM-MØ and M-MØ, it also promotes human macrophage subtype-dependent transcriptional changes because the expression of CCR2, IL10, CCL2, and HAMP is downregulated by GW7845 only in proinflammatory GM-MØ.

To determine whether the distinct transcriptional effects of PPARγ activation in GM-MØ and M-MØ had a functional correlate, the LPS-induced cytokine-producing ability of both macrophage subtypes was evaluated in the presence of GW7845. As expected, LPS stimulation of GM-MØ caused the preferential production of the proinflammatory cytokines TNFα and IL-6, whereas LPS-stimulated M-MØ primarily released IL-10 (9–11) (Figure 1D). In line with the transcriptional results, GW7845 significantly reduced the LPS-induced production of TNFα and IL6 from GM-MØ, but had no effect on the LPS-induced cytokine release from M-MØ (Figure 1D). Importantly, the inhibitory effect of GW7845 on the LPS-induced TNFα production of GM-MØ was PPARγ-dependent, as it was reduced upon PPARγ knockdown (Figure 1E). Therefore, agonist-mediated activation of PPARγ exclusively modulates the LPS-induced cytokine production from proinflammatory human monocyte-derived GM-MØ, further arguing for a polarization-dependent effect of PPARγ in human macrophages.

Given the different effect of PPARγ on GM-MØ and M-MØ, we next determined PPARγ expression and function in both human macrophage subtypes. Transfection of a PPRE reporter construct in both macrophage subtypes revealed that global PPAR-dependent transcriptional activity is higher in GM-MØ than in M-MØ (Figure 3A), thus suggesting that GM-MØ are endowed with a stronger PPARγ-dependent transcriptional activity. Regarding expression, GM-MØ contained higher levels of PPARG1/3 (encoding the ubiquitous PPARγ1 isoform) and PPARG2 (coding for the PPARγ2 isoform) mRNAs than M-MØ (Figure 3B). In fact, the adipocyte-restricted PPARG2 mRNA (40) was barely detectable in M-MØ (Figure 3B). The preferential expression of the PPARγ2-encoding mRNA was also observed in murine bone marrow-derived GM-MØ, whereas mouse M-MØ exhibited significantly higher Pparg1 expression than mouse GM-MØ (Figure 3C), in agreement with a previous report (51) and in line with the distinct gene profiles of monocyte-derived human M-MØ and bone marrow-derived mouse M-MØ (13, 52). Kinetic analysis revealed that PPARG2 mRNA is upregulated in human monocytes exposed to GM-CSF for 3, 5, and 7 days (Figure 3D). Although PPARG1 is expressed at higher levels than PPARG2 mRNA (25-fold approx.), PPARγ2 protein could be detected in whole cell and nuclear extracts from GM-MØ (Figures 3E,F). Therefore, GM-CSF-conditioned proinflammatory human macrophages exhibit a higher expression of PPARγ (PPARγ1 and PPARγ2) than M-CSF-conditioned anti-inflammatory human macrophages.

The GM-CSF-dependent expression of PPARγ is essential for the differentiation of A-MØ (24). Since proinflammatory human GM-MØ polarization is dependent on the autocrine/paracrine action of activin A (11), we next questioned whether activin A contributes to the preferential expression of PPARG mRNAs in GM-MØ. Activin A significantly elevated PPARG1/3 and PPARG2 mRNA levels in M-MØ, monocytes and THP-1 myeloid cells (Figure 4A). Moreover, inhibition of activin A-initiated Smad signaling by either SB431542 (Figure 4B) or A-83 (Figure 4C), or blockade of activin A with an anti-activin neutralizing antibody (Figure 4D), significantly reduced PPARG1/3 and PPARG2 mRNA levels in GM-CSF-dependent proinflammatory GM-MØ. In line with these results, generation of GM-MØ in the presence of A-83 resulted in significantly reduced expression of the PPARγ target gene ABCA1, a gene whose expression is responsive to PPARγ-LXR activation in human macrophages (Figure 4E). Further, analysis of ex vivo isolated human A-MØ revealed the constitutive expression of activin A (Figure 4F), and that the expression of PPARG1/3 and PPARG2 mRNA, as well as the expression of the PPARγ target ABCA1 mRNA, were significantly reduced in the presence of the A-83 Smad signaling inhibitor (Figure 4G). Altogether, these results indicate that activin A is a positive regulator of PPARγ expression and activity in GM-CSF-conditioned macrophages both in vitro and in vivo.

Given the transcriptional effects of PPARγ knockdown (Figure 1C), and to more thoroughly address the role of PPARγ in in vitro generated GM-CSF-conditioned macrophages, we determined the PPARγ-dependent transcriptional profile of GM-MØ in the absence of exogenous agonists. siRNA-mediated PPARγ knockdown significantly modified the transcriptome of GM-MØ, altering the expression of 314 probes (283 annotated genes) (p < 0.003, Table S2 in Supplementary Material). Specifically, PPARγ knockdown led to downregulation of 139 genes and upregulation of 144 genes in GM-MØ (Figure 5A). Twenty-five percent of the genes downregulated by siPPARG (36 out of 139) had been previously predicted as PPAR targets (53), including 20 genes upregulated by long-term rosiglitazone treatment of human monocyte-derived dendritic cells (54) and two genes whose expression is also diminished in mouse Pparγ^−/− macrophages (CD36 and GPD1) (24) (Figure 5B). Similarly, the set of genes upregulated upon PPARγ knockdown contained 19 genes predicted as PPAR targets (53) (Figure 5B), including 5 genes upregulated by rosiglitazone in human dendritic cells (54) and CCL2 and CCL7, whose orthologous genes are overexpressed in murine Pparγ^−/− macrophages (24). Conversely, siPPARG downregulated the expression of MSR1, whose mouse ortholog is overexpressed in Pparγ^−/− macrophages (24). The PPARγ-regulated gene set also included genes whose expression distinguishes A-MØ from other tissue-resident mouse macrophages (KRT79, BCAR3, MAFF, WWTR1) (55) or have been defined as human A-MØ-enriched genes (EDN1, CXCL1, TNFAIP6, IL7R) (56) (Table S2 in Supplementary Material). Therefore, PPARγ knockdown in GM-CSF-conditioned human macrophages allowed the identification of a large set of genes (Table S2 in Supplementary Material) whose expression is specifically modulated by PPARγ in the absence of an exogenous agonist. Besides, and in agreement with the divergent transcriptional profiles of functionally similar human and mouse macrophages (57), the human macrophage PPARγ-dependent gene set in human macrophages only partially overlaps with the list PPARγ-regulated genes previously identified in mouse macrophages.

To gain formation on the biological processes significantly affected after PPARγ knockdown in human macrophages, functional enrichment analysis was performed using GSEA (50). Confirming the validity of the results, PPARγ knockdown led to a very significant reduction in the expression of genes associated with the terms “KEGG_PPAR_Signaling_Pathway” and “KEGG_Peroxisome” (Figure 5C). In line with its known anti-inflammatory function (32, 58), reduction of PPARγ expression in GM-MØ caused a significant increase in the expression of genes within the “Hallmark_Inflammatory Response,” “Hallmark_TNFA signaling via NFKB,” and “GO_Cellular Response to IL1” gene sets (Figure 5D). Also in agreement with the anti-inflammatory activity of PPARγ (32, 58), PPARγ knockdown promoted a significant global upregulation of the GM-MØ-specific “Proinflammatory gene set” (12) (Figure 5D), and specially of two GM-MØ-specific genes like ECSCR and HSD11B1 (11) (Table S2 in Supplementary Material). Unexpectedly, PPARγ knockdown resulted in a very significant increase in the expression of genes within gene sets directly involved cell cycle and proliferation like “KEGG Cell Cycle,” “Reactome Mitotic M G1 Phases,” “Reactome G1 S Transition,” “Hallmark G2M Checkpoint,” “GO Cell Division,” and others (all with FDR q = 0.000) (Figure 5E and not shown). Besides, reduction of PPARγ caused a very significant increase in the expression of genes involved in chemotaxis (Figure 5D), suggesting that PPARγ has a negative regulatory effect on leukocyte mobility. All these results were further supported after ENRICHR analysis (see text footnote 4) (48, 49), which also revealed that PPARγ knockdown specifically impairs the expression of genes whose expression is controlled by a known PPARγ ligand (rosiglitazone, adjusted p = 1.4 × 10⁻¹⁰) and a transcription factor (FOXM1, adjusted p = 8.47 × 10⁻¹⁸) that collaborates with PPARγ in pulmonary inflammation (59, 60) (Table S2 in Supplementary Material and data not shown).

To further validate the correlations found with GSEA and ENRICHR, a representative number of transcriptional changes were assessed after PPARγ knockdown on independent GM-MØ preparations (Figure 5F). PPARγ knockdown significantly reduced the expression of genes encoding PAMP/DAMP receptors (TLR4, CD36), NFκB activators (TNFRSF21, CARD16), the scavenger receptors SR-A1 (MSR1), known PPARγ targets like HSD11B1, and the GM-MØ-specific gene ECSCR (11) (Figure 5F). Conversely, reduction of PPARγ levels increased the expression of genes encoding cytokines (CCL2, CCL7, CCL8, CXCL5), again pointing toward a negative regulatory effect of PPARγ on the expression of genes associated with cell chemotaxis. PPARγ knockdown also enhanced expression of genes encoding various growth-promoting factors (CSF1, TNFSF15, OSM, LIF) (Figure 5F). Moreover, downregulating PPARγ expression led to a significant reduction in the expression of genes that contribute to the significant enrichment signal (“leading-edge”) found after GSEA on gene sets related to cell cycle and proliferation (namely MKI67, BUB1, HMMR, E2F7, and CDKN2C) (Figure 5F, gray-filled bars) and that code for proteins involved in cell cycle regulation like Ki67, E2F7, and CDKN2C. Therefore, and in agreement with the GSEA correlations (Figure 5E), expression of PPARγ in GM-CSF-conditioned macrophages has a positive impact on the expression of genes that directly regulate and mark cell proliferation.

To assess the physiological relevance of the above findings, we next analyzed the functional and transcriptional consequences of siRNA-mediated PPARγ knockdown in human A-MØ isolated from BAL fluids. In agreement with previous reports (61, 62), isolated A-MØ expressed CD163 and exhibited very low levels of cell surface CD14 (Figure 6A). Regarding PPARγ expression, A-MØ exhibited a level of PPARG1/3 and PPARG2 mRNA expression similar to that found in GM-MØ and considerably higher than the expression seen in M-MØ (Figure 6B), and responded to the presence of the PPARγ agonists GW7845 by enhancing the expression of CD36 (Figure 6C). At the functional level, PPARγ knockdown in A-MØ did not significantly modify the LPS-induced production of TNFα or IL-6 (Figure 6D), thus indicating that PPARγ does not regulate the LPS-induced production of proinflammatory cytokines by A-MØ in the absence of an exogenous agonist, a result also seen in monocyte-derived GM-MØ (Figure 1G). By contrast, PPARγ knockdown significantly enhanced CCL2 production by human A-MØ both under basal and LPS-stimulated conditions (Figure 6D). Regarding the transcriptional role of PPARγ in ex vivo isolated human A-MØ, PPARγ silencing in A-MØ yielded similar effects to those previously observed in GM-MØ (Figure 6E). Specifically, PPARγ knockdown caused a significant upregulation of CCL2, CCL8, CSF1, TNFSF15, OSM, and LIF, and a significant downregulation of TLR4, ECSCR, CD36, MSR1, TNFRSF21, and CARD16 (Figure 6E). However, and in contrast with its effects on GM-MØ, downregulation of PPARγ in human A-MØ had no effect on the expression of genes encoding proteins involved in cell cycle regulation (Figure 6E). Altogether, these results indicate that PPARγ significantly contributes to the transcriptional signature of GM-CSF-conditioned human macrophages (either proinflammatory monocyte-derived GM-MØ or ex vivo isolated A-MØ) in the absence of exogenous agonists.