Human Peripheral Blood Mononuclear Cells (PBMCs) were isolated from buffy coats from anonymous healthy donors (provided by the Centro de Transfusiones de la Comunidad de Madrid) over a Lymphoprep (Nycomed Pharma) gradient according to standard procedures. Monocytes were purified from PBMC by magnetic cell sorting using anti-CD14 microbeads (Miltenyi Biotec). Monocytes (>95% CD14⁺ cells) were cultured at 0.5 x 10⁶ cells/ml in Roswell Park Memorial Institute (RPMI 1640, Gibco) medium supplemented with 10% fetal bovine serum (FBS, Biowest) for 7 days in the presence of 1000 U/ml GM-CSF or 10 ng/ml M-CSF (ImmunoTools) to generate GM-CSF-polarized macrophages (GM-MØ) or M-CSF-polarized macrophages (M-MØ), respectively (46). Cytokines were added every two days and cells were maintained at 37°C in a humidified atmosphere with 5% CO₂ and 21% O₂. Where indicated, macrophages were treated at different time points with one dose of LXR agonist GW3965 (31) (1 μM, Tocris), LXR inverse agonist GSK2033 (32) (1 μM, Tocris) or both, using dimethyl sulfoxide (DMSO) as vehicle. In the dual condition, the inverse agonist was added 1-hour prior to agonist treatment. In some experiments, cells were exposed to the GSK3β inhibitor CHIR99021 (2 μM) or DMSO as control. When indicated, Ascitic Fluid from cancer patients (Tumor-derived Ascitic Fluid, TAF) was added to monocytes (0.5:1 in culture medium), and cultures were maintained for 72 h. Ascitic fluids from four metastatic tumors (ovary cancer with peritoneal metastasis, renal carcinoma, and two from gastric carcinoma patients) were kindly provided by Dr. M^a Isabel Palomero (Oncology Department, Hospital General Universitario Gregorio Marañón) after patients had provided informed consent (Approval was obtained from the ethics committee of Hospital General Universitario Gregorio Marañón, and the procedures used in this study adhere to the tenets of the Declaration of Helsinki). Samples were centrifuged (4000g, 15 min) to remove cells and particulate material, sterile-filtered, aliquoted, and stored at -80°C until use. The patients provided informed consent and the Hospital General Universitario Gregorio Marañón ethics committee approved the study. For macrophage activation, cells were treated with 10 ng/ml E. coli 055:B5 lipopolysaccharide (Ultrapure LPS, Sigma-Aldrich). Human cytokine production was measured in macrophage supernatants using commercial ELISA [(TNF-α (BD Biosciences), IL-10, Activin A, CCL19, IL1β, IFNβ (R&D Systems)] according to the procedures supplied by the manufacturers.

Total RNA was extracted using the total RNA and protein isolation kit (Macherey-Nagel). RNA samples were reverse-transcribed with High-Capacity cDNA Reverse Transcription reagents kit (Applied Biosystems) according to the manufacturer’s protocol. Real-time quantitative PCR was performed with LightCycler^® 480 Probes Master (Roche Life Sciences) and Taqman probes on a standard plate in a Light Cycler^® 480 instrument (Roche Diagnostics). Gene-specific oligonucleotides ( Supplementary Table 1 ) were designed using the Universal ProbeLibrary software (Roche Life Sciences). Results were normalized to the expression level of the endogenous references genes TBP, HPRT1 or GAPDH and quantified using the ΔΔCT (cycle threshold) method.

M-MØ and GM-MØ cell lysates were subjected to SDS-PAGE (30-50 μg unless indicated otherwise) and transferred onto an Immobilon-P polyvinylidene difluoride membrane (PVDF; Millipore). After blocking the unoccupied sites with 5% non-fat milk diluted in Tris-Buffered Saline plus Tween 20 (TBS-T), protein detection was carried out with antibodies against LXRα (PPZ0412; Biotechne), LXRβ (PPK8917; Biotechne), MAFB (HPA005653, Sigma Aldrich) or c-MAF (sc-7866; Santa Cruz Biotechnology). Protein loading was normalized using an antibody against GAPDH (sc-32233; Santa Cruz Biotechnology) or vinculin (V9131; Sigma-Aldrich). Quimioluminiscence was detected in a Chemidoc Imaging system (BioRad) using SuperSignal™ West Femto (ThermoFisher Scientific).

RNA was isolated from M-MØ generated from monocytes exposed to a single dose of DMSO, GW3965, GSK2033 or both at the beginning of the 7-day differentiation process. Alternatively, RNA was isolated from CD14⁺ monocytes treated with DMSO (vehicle) or 1 μM GW3965 for 1 hour, and then cultured for 3 days in RPMI 1640 with 10% FBS supplemented with 50% Tumor-derived Ascitic Fluid (TAF). Sequencing was done on a BGISEQ-500 platform (https://www.bgitechsolutions.com). RNAseq data were deposited in the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) under accession GSE156783 and GSE181313. On average, 88.04 M reads per sample were generated and clean reads were mapped to the reference (UCSC Genome assembly hg38) using Bowtie2 (average mapping ratio to reference genome, 91.82%) (47). Gene expression levels were calculated by using the RSEM software package (48), and differential gene expression was assessed by using the R-package DESeq2 algorithms using the parameters Fold Change>2 and adjusted p value <0.05. Plots generated with the ggplot2 package, and heatmaps and clustering were done using the Genesis software (http://genome.tugraz.at/genesisclient/) (49). Differentially expressed genes were analyzed for annotated gene sets enrichment using ENRICHR (http://amp.pharm.mssm.edu/Enrichr/) (50, 51), and enrichment terms considered significant with a Benjamini-Hochberg-adjusted p value <0.05. For gene set enrichment analysis (GSEA) (http://software.broadinstitute.org/gsea/index.jsp) (52), gene sets available at the website, as well as gene sets generated from publicly available transcriptional studies (https://www.ncbi.nlm.nih.gov/gds), were used.

For comparison of means, and unless otherwise indicated, statistical significance of the generated data was evaluated using the paired Student t test in GraphPad Prism 8. In all cases, p<0.05 was considered as statistically significant.

We have previously defined gene sets whose expression not only marks human monocyte-derived GM-MØ (“Pro-inflammatory gene set”) and M-MØ (“Anti-inflammatory gene set”) (13, 20) (GSE68061) ( Figure 1A ), but discriminates the gene profiles of pro-inflammatory and immunosuppressive macrophages in vivo (24). In fact, the transcriptome of M-MØ shows a very strong enrichment of genes preferentially expressed by large tumor-associated macrophages (large TAM) ( Figure 1B ), whose presence associates with a lower disease-free survival rate in colorectal liver metastasis (42), while the transcriptome of GM-MØ resembles the specific gene profile of “small TAM” (42) and rheumatoid arthritis synovial fluid macrophages (RASF-MØ) (53) ( Figure 1B ). Remarkably, the LXR pathway has been reported to be highly upregulated in large TAM from colorectal liver metastasis (42) but also in pro-inflammatory RASF-MØ (44), thus raising the question of the role of LXR on human macrophage polarization. To clarify this apparent discrepancy, we sought to determine the role of LXR in the generation of monocyte derived macrophages under the influence of either GM-CSF (pro-inflammatory GM-MØ) or M-CSF (anti-inflammatory M-MØ). Initial experiments on a large number of independent samples revealed a higher expression of LXRα (encoded by NR1H3) in GM-MØ, and a higher level of NR1H2-encoded LXRβ in M-MØ ( Figure 1C ). Besides, RNA-seq (GSE188278) ( Figure 1D ), and the expression of desmosterol-upregulated genes in GM-MØ or M-MØ (54) ( Figure 1E ), showed that the expression of LXR targets is strongly regulated during monocyte differentiation into either GM-MØ or M-MØ ( Figure 1D ), with significantly higher expression of ABCG1, LPL and APOE in GM-MØ, and greater expression of ABCA1 and ARL4C in anti-inflammatory M-MØ ( Figures 1D, E ).

To directly address the involvement of LXR dependency in the acquisition of the M-MØ transcriptome, monocytes were exposed to a single dose of either the LXR agonist GW3965, the LXR inverse agonist GSK2033 or both, at the beginning of the differentiation process with M-CSF ( Figure 2A ). RNAseq on the resulting macrophages (GW-M-MØ, GSK-M-MØ and GW/GSK-M-MØ) showed that GW3965, and to a lower extent GSK2033, notably altered the acquisition of the transcriptome of M-MØ ( Figure 2B ). GSEA revealed that the most unambiguous effect was observed in GW-M-MØ, whose transcriptome indicated a strong under-representation of the M-MØ-specific “Anti-inflammatory gene set” and a very positive enrichment of the “Pro-inflammatory gene set” ( Figure 2C ). Moreover, comparison of the differentially expressed genes between GW-M-MØ and CNT M-MØ identified a significant number of genes of the “Anti-inflammatory gene set” (with reduced expression in GW-M-MØ, clusters 3 and 4) and the “Pro-inflammatory gene set” (with enhanced expression in GW-M-MØ) ( Figures 2D, E ). Indeed, analysis of a validation set of samples confirmed that differentiation in the presence of GW3965 impairs the expression of genes associated to the anti-inflammatory activities of M-MØ (IGF1, FOLR2, CD163L1, CCL2) and augments the expression of paradigmatic genes of the “Pro-inflammatory gene set” (INHBA, PPARG1) (13, 20, 55, 56) ( Figure 2F ). Conversely, the opposite effects were seen in the case of GSK-M-MØ gene profile ( Figures 2C–F ). Since expression of LXR target genes (e.g., ABCA1) (54) confirmed the strong efficacy of GW3965 and GSK2033 in our system ( Figures 2D, F ), these results indicate that LXR activation in monocytes impairs the acquisition of the anti-inflammatory transcriptional profile during the generation of monocyte-derived macrophages in response to M-CSF, and shifts the M-CSF-dependent differentiation of M-MØ towards the pro-inflammatory side. Interestingly, gene ontology analysis (Enrichr) of the 313 genes downregulated in GW-M-MØ showed a strong enrichment in MAF- and MAFB-dependent genes, as well as a highly significant enrichment of the genes that define the profile of “large TAM” from colorectal liver metastasis (42) (GSE131353, Figure 2G ), thus lending further relevance to the monocyte-conditioning ability of LXR modulators.

The functional significance of the transcriptional changes observed in GW-M-MØ and GSK-M-MØ was next weighed by comparing their respective cytokine profile in resting conditions and after activation with LPS. Compared to CNT M-MØ, GW-M-MØ produced lower levels of the anti-inflammatory cytokine IL-10 and higher levels of the immuno-stimulatory CCL19 chemokine ( Figure 3A ). Besides, and compared to GW-M-MØ, IL-1β production was significantly lower in GSK-M-MØ and GW/GSK-M-MØ ( Figure 3A ). This pro-inflammatory trend was confirmed after LPS stimulation, as GW-M-MØ secreted higher levels of TNF and IL-1β than CNT M-MØ ( Figure 3B ). Moreover, although LPS-treated GW-M-MØ also secreted higher levels of IL-10, the TNF/IL-10 ratio was considerably higher in GW-M-MØ than in CNT M-MØ ( Figure 3C ). Thus, since the pro-inflammatory effects of GW3965 treatment were impaired or abolished by GSK2033 (GSK-M-MØ and GW/GSK-M-MØ) ( Figures 3A–C ), these results confirm that modulation of LXR activity impacts the inflammatory activity of M-MØ, and that LXR activation prompts monocytes towards the generation of monocyte-derived macrophages with a higher pro-inflammatory transcriptional and cytokine profile.

Since LXR synthetic ligands cause a chronic and long-lasting modulation of LXR activity, we next assessed their effects at different time points in the monocyte-to-M-MØ differentiation process. To that end, LXR modulators (GW3965, GSK2033, or both) were added at day 0, day 2 or day 5 along M-MØ differentiation ( Figure 4A ), and the expression of the M-MØ-specific “Anti-inflammatory gene set” was determined. As shown in Figure 4B , the highest inhibitory effect of GW3965 on the expression of the “Anti-inflammatory gene set” (e.g., IGF1, FOLR2, CD163L1, CCL2) was found at the beginning of the differentiation process, and was not significant when added at later time points. By contrast, although the inverse agonist GSK2033 had minor effects, it completely prevented the effect of the agonist at all assayed time points ( Figure 4B , GW/GSK-M-MØ). These results imply that the involvement of LXR on the M-CSF-dependent acquisition of the transcriptional profile of M-MØ decreases along M-MØ differentiation, being maximal at the start of the monocyte-to-M-MØ differentiation process.

As we have previously identified the genes whose expression is modulated along monocyte-to- M-MØ differentiation (GSE188278) ( Figure 4C ), we next checked whether modulation of LXR activity also affected their expression. As shown in Figure 4D , 15% of the genes upregulated in the transcriptome of GW-M-MØ (61 out of 402) were specifically downregulated along monocyte-to- M-MØ differentiation. Conversely, 25% of the genes downregulated in the GW-M-MØ transcriptome (77 out of 313) corresponded to genes specifically upregulated along monocyte-to- M-MØ differentiation ( Figure 4D ). In fact, and in agreement with the analysis of M-MØ-specific genes ( Figure 2 ), a very significant percentage of the differentiation-associated genes with reduced expression in GW-M-MØ were MAFB/MAF-dependent, IL-10-regulated, and preferentially found in “large TAM” from colorectal liver metastasis (42) ( Figure 4D ), again emphasizing that LXR activation results in impaired expression of genes directly linked to the MAF/MAFB-dependent generation of anti-inflammatory M-MØ.

MAFB and MAF are master regulators for the differentiation of anti-inflammatory M-MØ (46, 57, 58), whose generation is prevented by activin A (13, 20). Gene ontology analysis using Enrichr ( Figure 2G ) or GSEA ( Figure 5A ) indicate an under-representation of MAF- and MAFB-dependent genes in the transcriptome of GW-M-MØ. Compared to CNT M-MØ, protein analysis revealed that GW-M-MØ has significantly reduced levels of MAF ( Figure 5B ) and MAFB ( Figure 5C ), although the latter did not reach statistical significance. Moreover, GSK-M-MØ expressed higher levels of both MAFB and MAF than CNT M-MØ ( Figures 5B, C ). Altogether, these results agree with the predictions of gene ontology analysis and demonstrate that the pro-inflammatory outcome of LXR activation (GW-M-MØ) correlates with reduced expression of the factors that shape the transcriptional and functional profile of M-MØ, whereas LXR inhibition (GSK-M-MØ) results in enhanced expression of both MAFB and MAF. Furthermore, GW-M-MØ were also found to produce slightly higher levels of activin A ( Figure 5D ), a factor that impairs the differentiation of M-MØ, further supporting that the modulation of LXR activation in monocytes ends up altering the expression of factors that shape the transcriptional and functional profile of M-MØ.

To find out whether MAF/MAFB mediate the macrophage polarizing action of LXR, the expression of both factors was enhanced by inhibiting GSK3β, which controls the stability of large MAF transcription factors through phosphorylation of their transcriptional activation domains (59). Thus, differentiating M-MØ were treated with the GSK3β inhibitor CHIR-99021 before exposure to the LXR agonist GW3965 ( Figure 5E ). As expected, CHIR-99021-treated M-MØ exhibited an enhanced expression of MAFB after 48 hours ( Figure 5F ). More importantly, the GW3965-induced reduction in the expression of M-MØ-specific genes (FOLR2, CD163, IL10, IGF1, LGMN) was lower in CHIR-99021-treated M-MØ, an effect that was statistically significant in the case of IL10 ( Figure 5G ). Therefore, the pro-inflammatory influence of the LXR agonist can be impaired by hindering GSK3β activity, suggesting that MAF and MAFB mediate the inhibitory effect of LXR on the expression of genes of the M-MØ-specific “Anti-inflammatory gene set”.

LXR target gene expression has been previously found to be the most enriched pathway in large TAM from colorectal liver metastasis (42), whose gene profile greatly resembles M-MØ ( Figure 1B ). Thus, to find out the potential relevance of modulating LXR activity under pathological settings, we evaluated whether altering LXR activity was also capable of modifying the macrophage-polarizing ability of tumor-derived ascitic fluids (TAF) of distinct origin. To that end, monocyte-derived macrophages were differentiated in the presence of TAF, and either with or without GW3695 ( Figure 6A ). Comparison of the resulting macrophages (TAF-MØ, GW-TAF-MØ) revealed that LXR activation greatly modifies the transcriptome of macrophages generated under the influence of tumor-derived ascitic fluids, as the expression of almost 1000 genes was significantly altered ( Figure 6B ). As expected, GW-TAF-MØ significantly over-expressed genes regulated by LXR and SREBP ( Figures 6C and 7A ) as well as of genes upregulated in GW-M-MØ and downregulated in GSK-M-MØ ( Figures 6C and 7B ). More importantly, GSEA revealed that the gene profile of GW-TAF-MØ shows a very strong under-representation of M-MØ-specific genes ( Figure 6D ) as well as of MAFB-dependent genes ( Figure 6E ), a result further supported by gene ontology analysis using Enrichr ( Figure 6F ). Indeed, and as shown in Figure 6G , the expression of representative MAF, MAFB and MAFB-dependent genes was reduced by GW3965 in all GW-TAF-MØ samples, where the expression of GW3965-upregulated genes was higher. Finally, analysis of the transcriptome of GW-TAF-MØ evidenced a very significant downregulation of the genes that define the gene profile of “large TAM” (adjp, 7.54e-27), and a positive enrichment of the genes that characterize “small TAM” (9.13e-14) from colorectal liver metastasis (42) ( Figures 6F and 7B, C ). These results confirm that LXR activation exerts a similar effect on macrophages generated in the presence of either M-CSF or tumor-derived ascitic fluids, and demonstrate that LXR activation opposes the polarizing action of pathological tumor-derived fluids by impairing the acquisition of the genes that characterize anti-inflammatory (M-CSF-dependent) macrophages and enhancing the expression of genes that define pro-inflammatory macrophages.