As key cells of the immune system, macrophages coordinate the activation and regulation of the immune response. Macrophages present a complex phenotype that can vary from homeostatic, proinflammatory, and profibrotic to anti-inflammatory phenotypes. The factors that drive the differentiation from monocyte to macrophage largely define the resultant phenotype, as has been shown by the differences found in M-CSF– and GM-CSF–derived macrophages. We explored alternative inflammatory mediators that could be used for in vitro differentiation of human monocytes into macrophages. IFN-γ is a potent inflammatory mediator produced by lymphocytes in disease and infections. We used IFN-γ to differentiate human monocytes into macrophages and characterized the cells at a functional and proteomic level. IFN-γ alone was sufficient to generate macrophages (IFN-γ Mϕ) that were phagocytic and responsive to polarization. We demonstrate that IFN-γ Mϕ are potent activators of T lymphocytes that produce IL-17 and IFN-γ. We identified potential markers (GBP-1, IP-10, IL-12p70, and IL-23) of IFN-γ Mϕ and demonstrate that these markers are enriched in the skin of patients with inflamed psoriasis. Collectively, we show that IFN-γ can drive human monocyte to macrophage differentiation, leading to bona fide macrophages with inflammatory characteristics.
Macrophages are unique cells of the immune system involved in host defense, homeostasis, and tissue repair (1). Macrophages can be categorized based on ontogeny into two groups, monocyte-derived macrophages or tissue-resident macrophages, which originate from cell progenitors in the yolk sac or fetal liver monocytes (2). The proportion of macrophages from these origins varies depending on the tissue and inflammatory status (2). Core macrophage functions include the phagocytosis of pathogens and dead cells and activation of other immune cells (3). Macrophages are key in the pathogenesis of various diseases in which they can either resolve or worsen disease outcomes (1). These cells contribute to excessive inflammation in various immune-mediated diseases and in chronic inflammatory diseases like rheumatoid arthritis (RA), atherosclerosis, inflammatory bowel disease, nonalcoholic steatohepatitis, and psoriasis (4–8). Psoriasis is characterized by lesional areas in the skin where there is a thickening of the epidermis, enhanced keratinocyte proliferation, leukocyte infiltration, and inflammation (9). Macrophages acquire a proinflammatory phenotype and produce cytokines such as IL-23, TNF, and IL-1β. The IL-23/IL-17 axis is overactive in psoriasis, and biologics targeting these cytokines have proven to be efficacious in severe disease. IL-23, exclusively derived from myeloid cells, activates T cells to produce cytokines such as IL-17 and IFN-γ, which further polarize macrophages, leading to a continuous inflammatory state (10–12).
Macrophages, as shown by transcriptomic studies, can adopt a spectrum of activation states depending on the environmental stimuli (13–15). The concept of macrophage heterogeneity is important not only for activation of macrophages already within the inflamed tissue, but it also impacts the differentiation of infiltrating monocytes. The general consensus is that monocyte to macrophage differentiation is driven by either M-CSF or GM-CSF (16), and these macrophages present distinct phenotypes, morphology, and function (14, 17). GM-CSF–derived macrophages are considered to be more proinflammatory compared with M-CSF–derived macrophages (18–21), and this is in line with the detrimental role of GM-CSF and macrophages across multiple inflammatory diseases (22).
Within the tissue, monocytes and macrophages will also encounter other inflammatory mediators in addition to M-CSF and/or GM-CSF. One important cytokine released in numerous diseases is IFN-γ, which belongs to the type II IFN protein family and is often used to polarize macrophages either alone or in combination with LPS (23). IFN-γ is produced by activated lymphocytes, namely CD4 (Th1) cells, NK cells, and CD8 cytotoxic T cells (24). Elevated circulating levels of IFN-γ have been found in patients with psoriasis, sarcoidosis, lupus nephritis, and juvenile idiopathic arthritis, among other diseases (25–30).
Thus far, most human macrophage models use in vitro–differentiated macrophages due to the difficulties associated with isolating human macrophages ex vivo. Although differentiation protocols are usually restricted to M-CSF or GM-CSF with only limited use of alternative differentiation factors, other factors can yield functional macrophages, too (31). Considering the importance of IFN-γ as a potent regulator of human disease, we hypothesized that IFN-γ would drive monocyte differentiation into proinflammatory macrophages. In the current study, we find that IFN-γ is sufficient to drive monocyte to macrophage differentiation even in the absence of M-CSF. IFN-γ–differentiated macrophages have a stronger proinflammatory phenotype compared with either M-CSF– or GM-CSF–derived macrophages. We show that the principal proteins and axis upregulated in IFN-γ–derived macrophages are also found in patients with psoriasis and that these macrophages are distinct from M-CSF/GM-CSF macrophages in the proteomic and gene expression of certain chromatin modifiers. We conclude that the use of IFN-γ as a differentiation factor is a useful model to generate human macrophages that resemble a heightened inflammatory disease phenotype, such as that seen in psoriasis.
Materials and Methods
In order to culture macrophages, monocytes were isolated from blood of healthy donors. The human biological samples were sourced ethically, and their research use was in accordance with the terms of the informed consent under an Institutional Review Board/Ethics Committee–approved protocol. Blood from healthy volunteers was collected into tubes containing sodium heparin anticoagulant. Blood was diluted 1:1 with PBS and layered on top of 15 ml Lymphoprep (GE Healthcare). Tubes were spun down at 1500 rpm for 20 min without break and acceleration at room temperature (RT). PBMCs were collected from the ring fraction. Monocytes were isolated from PBMCs using CD14-positive magnetic beads following the manufacturer’s instructions (Miltenyi Biotec). Cells were cultured in RPMI 1640 (without glutamine and HEPES) (Life Technologies) with 5% FCS, 2 mM l-glutamine, penicillin (100 U/ml), and streptomycin (100 mg/ml) (all from Life Technologies).
For the first experiment (Fig. 1A), monocytes were cultured for 5 d at 37°C and 5% CO2 with M-CSF (50 ng/ml) (R&D Systems), GM-CSF (5 ng/ml) (R&D Systems), LPS (100 ng/ml) (Escherichia coli 0111:B4; Sigma-Aldrich), IFN-γ (31 ng/ml) (R&D Systems), TNF (50 ng/ml) (R&D Systems), or IL-17 (25 ng/ml) (R&D Systems) alone or in combination with M-CSF. For the second experiment (Fig. 1B, 1C, 1E), monocytes were cultured with 0, 0.1, 0.5, 1, 5, 10, 50, or 100 ng/ml IFN-γ alone or in combination with 50 ng/ml M-CSF. For all other experiments, monocytes were cultured with 100 ng/ml M-CSF, 5 ng/ml GM-CSF, or 50 ng/ml IFN-γ. For the stimulations, cells were treated for 24 h with LPS (100 ng/ml), IFN-γ (50 ng/ml) (alone or in combination with LPS), and IL-4 (10 ng/ml) (R&D Systems). For every experiment, the culture conditions were tightly controlled (using flat-bottomed plates, with the same plating density [1 × 106 cells/ml], cell culture medium, and donor-matched monocytes).
A total of 10 ml human blood from patients (n = 10 patients and n = 7 healthy volunteers) with either moderate or severe psoriasis was collected into serum gel vacutainers. Blood was supplied by the Clinical Unit Cambridge for GlaxoSmithKline-funded research. All donors provided written informed consent for the collection and use of their samples in accordance with the protocol of the local Institutional Review Board (Cambridge Research Ethics Committee). Tubes were centrifuged at 2500 rpm for 5 min at RT. The plasma was harvested from each tube and stored in cryopreservation tubes for cytokine measurement.
Monocytes were differentiated into M-CSF–, GM-CSF–, and IFN-γ–derived macrophages on slide chambers (Nunc Lab-Tek II Chamber slide system) at 4 × 105 cells/chamber. After 5 d of differentiation, cells were fixed for 15 min in 100 µl 4% formaldehyde (Pierce; Thermo Fisher Scientific). Afterwards, cells were washed three times with PBS and blocked for 30 min with 200 µl blocking buffer (1% BSA and 0.1% Tween in PBS). After blocking, cells were incubated with a CD68 Ab (Abcam) (1:100 in blocking buffer) overnight at 4°C. The next day, cells were washed three times with PBS and incubated for 1 to 2 h in the dark with a secondary Ab (Goat Anti-Rabbit IgH H&L, Alexa Fluor 488; 1:1000 in blocking buffer; Abcam). The Ab was washed, and ProLong Gold antifade mount with DAPI (Thermo Fisher Scientific) was added. Microscopic pictures were taken from three donors (DFC450; Leica) using FITC and Cy5 filters for the CD68 and DAPI with a ×10 original magnification with the LAS software version 22.214.171.124. For the analysis, the cells expressing CD68+/DAPI+ were counted as positive cells, and the percentage was then calculated from the total number of cells (CD68+/DAPI+ and CD68−/DAPI+).
After the different stimulations, the supernatant was collected to measure cytokine secretion. To measure IL-1β, IL-12p70, IL-6, and IL-10, a Human Proinflammatory 7-Plex Tissue Culture Kit (Meso Scale Discovery) was used following the manufacturer’s instructions. M-CSF was measured using an MSD kit (Meso Scale Discovery) following the manufacturer’s instructions with 1:1 diluted sample. A Meso Scale Discovery kit for TNF-α was used following the manufacturer’s instructions, diluting the samples 1:10. This was also done for IFN-γ response protein 10 (IP-10) and IL-17 measurements using the V-PLEX Human IP-10 Kit (Meso Scale Discovery) and the MSD 5 PLEX (N751B; Meso Scale Discovery) following the manufacturer’s instructions. For IP-10 and IL-17 measurement, samples were undiluted. This kit was also used to measure IFN-γ concentrations in the coculture experiments, using undiluted samples. CCL-18 was detected by ELISA (R&D Systems) following the manufacturer’s instructions, and M-CSF samples were diluted 1:10. MCP-1 (Meso Scale Discovery) and CCL-1 (Meso Scale Discovery) kits were also used following the manufacturer’s instructions. For CCL-1, samples were diluted 1:1 and undiluted for MCP-1.
For IL-23, an MSD assay was developed in-house using an MSD Standard Bind assay plate (Meso Scale Discovery). All of the Abs used in the IL-23 MSD assay belong to the IL-23 Human ELISA kit (Mabtech). Briefly, the plate was coated with 25 µl/well IL-23 capture Ab at 1 µg/ml in PBS overnight at 4°C. The next day, the plate was washed three times with washing buffer (0.05% Tween 20 in PBS) and incubated with blocking buffer (3% BSA in PBS) for 1 h at RT on a shaker (700 rpm). After the incubation, the plate was washed, and 25 µl diluted sample (1:10 in culture media) or 25 µl each of the concentrations of the eight-point standard curve (10,000–0 pg/ml at 1:4 dilution between each point) for IL-23 was added to the plate and incubated for 2 h at RT on a plate shaker at 700 rpm. The plate was washed, and 25 µl secondary Ab (1 µg/ml in blocking buffer) was added to the plate for 1.5 h. The plate was washed and 25 µl SULFO-TAG detection reagent (0.1 μg/ml in blocking buffer; Meso Scale Discovery) was added to the plate and incubated for 1 h at RT on a plate shaker at 700 rpm. Finally, the plate was washed, and 150 µl 2× Read buffer (Meso Scale Discovery) was added. The plate was read by the MSD SECTOR Imager 6000.
Viability was determined by measuring total ATP content in monocyte/macrophage cultures via the CellTiter-Glo kit and following the manufacturer’s instructions. Data are presented as mean and SEM of total luminescence units per condition.
M-CSF–, GM-CSF–, or IFN-γ–derived macrophages were cultured in 24-well plates at a concentration of 1 × 106 cells/well. Macrophages were detached using a Cell Dissociation buffer (Sigma-Aldrich) followed by two washes in PBS. Cells were stained with a Live/Dead stain (1:1000; BD Biosciences) for 15 min at RT in dark. Cells were washed twice with FACS buffer (420201; BioLegend) and incubated with an Fc receptor blocking agent (Human TruStain FcX, 422302; BioLegend) for 10 min at RT, prior to incubation with the Abs. CD14 (V500, 562693; BD Biosciences), CD16 (BV421, 302038; BioLegend), CD80 (PE, 305208; BioLegend), CD86 (FITC, 374204; BioLegend), CD64 (PerCP Cy5.5, 305024; BioLegend), and CD11b (APC, 301350; BioLegend) Abs were all diluted 1:100 in FACS buffer and incubated for 30 min at RT in the dark. Flow cytometric analysis was performed on a BD FACSCanto II flow cytometer. Data was analyzed using the FlowJo software v10. Macrophages were gated by removing doublets and afterward selecting viable cells. Median fluorescence intensity (MFI) was quantified using fluorescence minus ones as control.
T cell cocultures
The cocultured CD4+ T cells and macrophages were isolated from the same donor. Monocyte isolation and macrophage culture were performed as described above. T cells were isolated from the CD14-negative flow-through using the CD4+ isolation kit (130-096-533; Miltenyi Biotec) following the manufacturer’s instructions. T cells were kept in culture for 6 d in RPMI 1640 (without HEPES and l-glutamine) (Life Technologies) with 10% FCS, 2 mM l-glutamine, penicillin (100 U/ml), and streptomycin (100 mg/ml), supplemented with 2 ng/ml IL-7 (207-IL/CF; R&D Systems). Macrophages were cultured for 5 d as described above. Once the cells had differentiated, they were stimulated for 24 h with LPS (100 ng/ml) in fresh media. After 24 h, the supernatant was collected (100 µl) and replaced by new media. Supernatant was transferred to an empty 96-well plate. T cells were centrifuged and resuspended at 1 × 107 cell/ml in media without IL-7. Cells were added to the wells of the plates containing either the macrophages with fresh media (1:1 ratio) or the supernatant from the stimulation. In another plate, the supernatant was not removed following the 24-h LPS stimulation, and T cells (1:1 ratio) were added to the macrophages together with the supernatant. After 3 d of coculture, supernatant was collected and cytokines were measured.
Monocytes (1 × 105 cells/well) were differentiated into the different subtypes of macrophages for 5 d in black clear-bottom 96-well tissue culture–treated plates (3603; Costar). To measure phagocytosis, pHrodo Green E. coli Bioparticles (P35366; Thermo Fisher Scientific) were resuspended in a concentration of 2 mg/ml in 0.9% saline solution. Cells were incubated at 37°C, 5% CO2, and every 30 min, E. coli bioparticles were added at 200 µg/ml every time point. After 3 h, cells were fixed with unbuffered saline solution with 2% glutaraldehyde (G7651; Sigma-Aldrich) and 1% Parafix (PRC/R/38; Pioneer Research Chemicals) for 20 min at RT. After washing, cells were permeabilized with 0.1% Triton X-100 (T8787; Sigma-Aldrich) in PBS for 30 min. The buffer was removed, and CellMask (C37608; Molecular Probes) was added at 1 µg/ml with Hoechst at 2 µg/ml for 30 min. Cells were washed, and images were taken with an IN Cell 2200 at a ×10 original magnification. Images were analyzed using Columbus software (version 2.8.0), identifying individual cells and measuring the MFI per cell.
Gene expression panel
RNA was isolated from the different subtypes of macrophages using RNeasy Mini Kit (74104; Qiagen) following the manufacturer’s instructions. cDNA was generated using SuperScript III First-Strand Synthesis SuperMix for qRT-PCR (11752-250; Thermo Fisher Scientific) according to the manufacturer’s instructions. The 84-gene panel RT2 Profiler PCR Array Human Epigenetic Chromatin Modification Enzymes (PAHS-085ZE; Qiagen) was run according to the manufacturer’s instructions. For the analysis, the Software Array Studio (version 10.1) was used. Threshold cycle (Ct) values were converted to abundance and transformed to Log2. Values were normalized to the housekeeping genes (ACTB, B2M, GAPDH, HPRT1, and RPLP0), and principal component analysis (PCA) was performed. The threshold for statistical significance of genes in the comparison, a threshold of Log2 fold change (FC) > 0.58 < −0.58 and raw p value <0.05, was set. ΔΔCt to M-CSF values were calculated, and for statistical analysis of these data, one-way ANOVA with Dunnett correction was conducted.
Macrophages (1 × 106 cells/well) were lysed using MQ water in combination with NuPage LDS nonreducing sample buffer (1×) (ThermoFisher NP0007) and 1× NuPage sample reducing agent (ThermoFisher NP0009). A total of 5 µl of the ladder SeeBlue Plus2 Standard (LC5925; Invitrogen) and 12 µl samples/well in a NuPAGE 4–12% Bis-Tris Mini gel were used. The gel was run in MOPS buffer (Thermo Fisher Scientific) at 100–120 V for 2 to 3 h. After that, proteins were transferred to a nitrocellulose membrane using the iBlot Gel Transfer system following the manufacturer’s instructions. Membranes containing protein were cut above the housekeeping protein (Histone H3) and blocked with 3% milk in PBS (blocking buffer). The different parts were incubated overnight at 4°C with the appropriate Ab, guanylate-binding protein 1 (GBP-1) (ab131255; Abcam) diluted 1:2,000 or H3 (ab1791; Abcam) 1:20,000 diluted in blocking buffer. Postincubation, membranes were washed (PBS and 0.05% Tween 20) and then incubated with a secondary Ab, Alexa Fluor 680 donkey anti-rabbit IgG (H+L) (A10043; Invitrogen) diluted at 1:8000 in wash buffer for 1 h in the dark. Membranes were washed and visualized with the Odyssey Infrared Imaging system.
Cell surface proteomics
Surface protein enrichment followed by quantitative mass spectrometry was performed as previously described (32). In brief, glycosylations of cell surface proteins on live cells were oxidized using sodium metaperiodate (1 mM, 10 min) and biotinylated using Alkoxyamine-PEG4-Biotin (Thermo Fisher Scientific) via an oxime ligation reaction (1 mM, 10 min). Cells were washed, harvested, and lysed in 50 mM Tris 4% SDS. Biotinylated proteins were enriched using neutravidin beads (Thermo Fisher Scientific), on-bead digested with trypsin, labeled with tandem mass tags (TMT10; Thermo Fisher Scientific), and subsequently pooled. Samples were derived from four different donors. In each tandem mass tag experiment, all samples from one donor (monocytes, M-CSF, M-CSF plus LPS, M-CSF plus IFN-γ, GM-CSF, GM-CSF plus LPS, IFN-γ, and IFN-γ plus LPS) were combined. Samples were fractionated into five fractions using stage-tip–based strong cation exchange fractionation (33) and measured on a Q Exactive mass spectrometer using 120-min gradients. Raw files were processed using an in-house pipeline (34), and spectra were searched against the IPI database using Mascot 2.5 (Matrix Science, Boston, MA). Protein abundance was calculated relative to the M-CSF channel. Proteins used for further analysis were filtered for localization to the plasma membrane (manually curated and extended Swissprot annotation) and, for at least three quantified spectra, matched as well as two unique peptides to ensure high-quality identification and quantification. Differential protein abundance was tested individually for each contrast by a two-sample two-sided t test with multiple hypothesis testing correction using the Benjamini-Hochberg procedure. Proteins fulfilling the following criteria were regarded as significantly affected: (I) p value fulfills the criteria:
where logFC is the Log2 ratio of the pairwise comparison, “s” the SD of the dataset defined as 3 × median SD of protein abundance estimations for all proteins between replicates, and “P” the minimal accepted log10-transformed p value for a protein set to p = 0.01, (II) logFC ≥ s, and (III) adjusted p value <0.1.
Patient dataset analysis
Datasets were obtained from the National Center for Biotechnology Information Gene Expression Omnibus (35): 39 independent human transcriptome-wide datasets for 10 skin diseases were reviewed and selected. All selected datasets are derived from human skin biopsy samples from published clinical trials and included comparisons of diseased lesional to nonlesional and/or normal skin biopsies. Where comparisons are conducted for patients on treatment, the longitudinal time points are included, and the data are provided as Log2 FC relative to baseline (prior to treatment initiation). Studies were curated to ensure technical quality and comparability across studies, and diseases and probes were remapped to allow cross-platform comparisons.
Transcriptomic studies used in the meta-analysis were: psoriasis: GSE2737, GSE6710, GSE13355, GSE14905, GSE26866, GSE30999, GSE34248, GSE41662, GSE41663, GSE50790, GSE51440, GSE52471, GSE53431, and GSE54456; acne: GSE6475 and GSE53795; rosacea: GSE65914; alopecia areata: GSE45512 and GSE58573; atopic dermatitis: GSE5667, GSE16161, GSE27887, GSE32924, GSE36842, GSE58558, and GSE59294; vitiligo: GSE53146 and GSE65127; lichen planus: GSE38616 and GSE52130; actinic keratosis: GSE2503 and GSE32628; dermatomyositis: GSE1551, GSE5370, GSE11971, and GSE46239; lupus: GSE52471; hidradenitis suppurativa: GSE72702; anti–IL-17 in psoriasis: GSE31652 and GSE53552; anti–IL-23 in psoriasis: GSE51440; and anti-TNF in psoriasis: GSE57376, GSE11903, and GSE41663.
The meta-analysis method used in this study has been described previously in Qu et al. (36). To reanalyze and integrate the available clinical transcriptomics datasets, we obtained the raw mRNA expression data from the Gene Expression Omnibus. We then preprocessed each dataset by applying the robust multiarray average algorithm in combination with a method that accounts for evolving transcript definitions by remapping microarray probes to the most current gene annotations (37), as described previously (38). Differentially expressed gene changes (i.e., lesion versus normal, lesion versus nonlesion, or time posttreatment versus baseline) were determined by linear model fit accounting for paired study designs where applicable. Additionally, variance estimates were derived by applying an empirical Bayes methodology (39). We then integrated the resulting individual processed datasets by matching the Entrez gene identifiers that correspond to the respective probe set identifiers.
Data were analyzed using GraphPad Prism version 5.0 (GraphPad Software, La Jolla, CA). Data represent the mean ± SEM. Statistical differences were analyzed using a two-way ANOVA with Bonferroni post hoc test analysis for grouped analysis and one-way ANOVA with Dunnett post hoc test or t test for column analysis. The p values <0.05 were considered statistically significant: *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.
IFN-γ alone is sufficient to induce monocyte differentiation into macrophages
To determine whether factors other than M-CSF or GM-CSF are able to generate genuine and responsive macrophages, we cultured human monocytes for 5 d in the presence of different inflammatory mediators (LPS, IFN-γ, TNF, or IL-17) alone or in combination with M-CSF (Fig. 1A). After 5 d, the resultant cells were stimulated for 24 h with LPS to measure cytokine production. We observed that M-CSF– and GM-CSF–differentiated macrophages exhibited a strong response to LPS with production of classical inflammatory cytokines (Fig. 1A). As expected, monocytes left in culture without any growth factor or stimulation did not survive and acted as an internal control for other conditions that were insufficient for monocyte survival, such as IL-17. In the absence of M-CSF, IFN-γ–differentiated macrophages exhibited a strong response to LPS by producing the classical proinflammatory cytokines, TNF, IL-1β, IL-6, and IL-12p70. Similar effects were observed in macrophages that were differentiated with IFN-γ in the presence of M-CSF. Macrophages generated in the presence of LPS alone responded to a subsequent LPS stimulation with high IL-6 and MCP-1 but not TNF, which is in line with a tolerized phenotype (40). Macrophages generated in the presence of TNF exhibited a weak LPS response, with IL-12p70, MCP-1, and TNF being detected above the monocyte-only culture. Based on these results and the importance of IFN-γ in inflammatory diseases (41), we decided to focus on macrophages generated in the presence of IFN-γ alone. First, we wanted to identify the optimal concentration of IFN-γ needed for differentiation of monocytes into macrophages. We cultured monocytes for 5 d with a range of IFN-γ concentrations (0.1–100 ng/ml) in the presence or absence of suboptimal M-CSF concentrations. In the absence of M-CSF, phase bright cells were only observed when ≥1 ng/ml IFN-γ was used for differentiation, and these macrophages exhibited a clustered and round morphology (Fig. 1B). Although M-CSF–differentiated macrophages typically display an elongated morphology, the presence of 1 ng/ml IFN-γ hindered this morphology, indicating dominance of the IFN-γ signaling pathway. Monocytes differentiated with 50–100 ng/ml IFN-γ appeared morphologically similar and were found to be distinct from monocytes and other monocyte-derived in vitro–differentiated macrophages (Supplemental Fig. 1A). We measured total cellular ATP content as an orthogonal marker of cell survival and found that ATP levels increased with increasing IFN-γ concentrations, reaching a plateau at 50–100 ng/ml IFN-γ (Fig. 1C). Monocytes differentiated with 50–100 ng/ml IFN-γ had similar total ATP levels to those generated in the presence of M-CSF plus IFN-γ or M-CSF alone. Monocytes differentiated with IFN-γ alone did not rely on endogenous production of M-CSF as a survival factor (Supplemental Fig. 1B). Based on these data, we selected 50 ng/ml IFN-γ as the optimal concentration to generate these macrophages, hereafter termed IFN-γ Mϕ. IP-10 is a well-known IFN response cytokine (42). We wanted to test whether the response to IFN-γ was robust and irrespective of the presence of M-CSF over the 5-d differentiation period and therefore decided to measure IP-10 in the supernatants of these cells. We found enhanced IP-10 levels with increasing IFN-γ concentrations irrespective of the presence of M-CSF (Fig. 1D). We further tested the selectivity and rapidity of IP-10 production and assessed production throughout the differentiation period and across the three macrophage subtypes. We found that IP-10 was specifically expressed by IFN-γ Mϕ and produced as early as 6 h into differentiation, and this further increased over time (Fig. 1E). Next, we wanted to confirm that these cells expressed CD68, a prototypical macrophage marker (43). We compared CD68 immunostaining across monocytes, IFN-γ Mϕ, and the well-characterized M-CSF (M-CSF Mϕ) and GM-CSF (GM-CSF Mϕ) macrophages and found no difference across the three macrophage subtypes (Fig. 1F).
These data demonstrate that IFN-γ alone is sufficient to induce human monocyte differentiation into macrophages as shown by CD68 staining. These IFN-γ Mϕ are viable, morphologically distinct, and LPS responsive.
IFN-γ Mϕ exhibit a hyperinflammatory phenotype while retaining phagocytic capacity
As a core macrophage function, we compared the phagocytic capacity of the three Mϕ subtypes using pHrodo Green E. coli beads. Surprisingly, we observed that IFN-γ Mϕ had the same phagocytic function as M-CSF Mϕ, and there was no difference in the rate or extent of phagocytosis across the three Mϕ subtypes (Fig. 2A).
To better understand and compare the functional capabilities of IFN-γ Mϕ, we assessed cytokine production following stimulation with LPS, LPS plus IFN-γ, or IL-4. We found that IFN-γ Mϕ produced significantly more IL-12p70 and IL-23 after stimulation (LPS alone of LPS plus IFN-γ), whereas production of TNF and IL-1β was not statistically different across the three macrophage subtypes (Fig. 2B). M-CSF Mϕ produced IL-12p70 after stimulation with LPS plus IFN-γ, although at much lower levels than GM-CSF and IFN-γ Mϕ. It is interesting to note that M-CSF Mϕ stimulated with LPS plus IFN-γ do not produce the same cytokine response as IFN-γ Mϕ plus LPS, demonstrating that the differentiation conditions have produced an inherently differential baseline. In line with this greater proinflammatory response, stimulated IFN-γ Mϕ did not produce IL-10. IL-4 stimulation of these macrophages evoked a strong CCL-18 response from M-CSF Mϕ, whereas this was significantly reduced in GM-CSF Mϕ and absent in IFN-γ Mϕ, providing evidence that exposure to IFN-γ during differentiation has altered the resulting macrophages to be less amenable to resolution signals such as IL-4.
Next, we wanted to study any differences in cell surface marker expression evident at baseline or following stimulation with LPS (for details, see Supplemental Fig. 1C, 1D). At baseline, IFN-γ Mϕ expressed more CD64 in comparison with M-CSF and GM-CSF Mϕ ((Fig. 2C). CD14 and CD16 are known to be highly expressed in M-CSF Mϕ (44); however, both IFN-γ and GM-CSF Mϕ exhibit significantly reduced expression in comparison. CD11b, the macrophage adhesion marker (45), although expressed across all three Mϕ subtypes, was considerably higher in GM-CSF Mϕ. Under inflammatory conditions, both IFN-γ and GM-CSF Mϕ induced expression of the costimulatory molecules CD80 and CD86. These data indicate that both IFN-γ and GM-CSF Mϕ exhibit a large overlap in cell surface markers at baseline and in response to LPS compare with M-CSF Mϕ, with a key feature being induction of costimulatory molecules CD80 and CD86.
We wanted to follow up on the functional consequences of increased CD80 and CD86 expression and hypothesized that IFN-γ and GM-CSF Mϕ may have enhanced T cell activation capacity relative to M-CSF Mϕ. To assess this, we cocultured macrophages with autologous T cells. We found that naive macrophages cocultured with T cells for 3 d did not induce T cell activation as assessed by IFN-γ and IL-17 production (Fig. 2D). LPS stimulation of GM-CSF and IFN-γ Mϕ, but not M-CSF Mϕ, led to T cell activation with a strong Th1 response, as characterized by IFN-γ production. To determine whether T cell activation was driven via macrophage-derived secreted factors, we transferred supernatants from activated macrophages to T cells and found suboptimal IFN-γ production. We then assessed whether cell–cell interactions alone were sufficient for T cell activation and used LPS-activated Mϕ in fresh cell culture medium and cocultured with T cells to find only partial T cell activation as measured by IFN-γ production.
We additionally assessed whether other T cell subsets were induced through measuring supernatants for IL-13 and IL-17. Although IL-13 was not detectable (data not shown), a strong IL-17 response was induced via IFN-γ Mϕ (Fig. 2D). IL-17 was induced via IFN-γ Mϕ–secreted factors together with direct cell–cell interactions and was not evident in M-CSF and GM-CSF Mϕ.
In summary, these data show that IFN-γ Mϕ are phagocytic and have a similar activation profile to GM-CSF Mϕ, but exhibit a hyperinflammatory phenotype that can induce autologous T cell activation with both Th1 and Th17 responses.
GBP-1 is a potential marker for IFN-γ Mϕ
We next set out to identify a selective marker for IFN-γ Mϕ using cell surface proteomics analysis. Selective labeling of plasma membrane proteins via an oxime ligation utilizing the N-linked glycosylations terminating in sialic acids is a powerful method to increase the sensitivity for detection for typically low abundant cell surface proteins (32). Human primary monocytes derived from four donors were differentiated into macrophages by M-CSF, GM-CSF, or IFN-γ and polarized by either LPS or IFN-γ. Cell surface proteins were labeled on live cells, and enriched glycosylated proteins were analyzed by quantitative mass spectrometry. On average, we identified and quantified 1200 plasma membrane annotated proteins per individual donor (Supplemental Table I).
We compared each of the three unstimulated Mϕ subsets to each other and found that the greatest difference was between M-CSF and IFN-γ Mϕ (Fig. 3A). There were only 6 significantly upregulated proteins but 66 downregulated proteins in IFN-γ Mϕ relative to M-CSF Mϕ. This trend was also visible in the comparison between IFN-γ and GM-CSF Mϕ, in which only 5 (GBP-1, LCK, RDX, STX11, and ITGAL) of the total 25 differential proteins were upregulated in IFN-γ Mϕ. Proteins upregulated in IFN-γ Mϕ compared with M-CSF Mϕ included GBP1, IFIT5, STX11, G6PD, HSP90AA1, and SARS. We displayed these data in a heat map and included an additional condition of M-CSF Mϕ stimulated with IFN-γ (Fig. 3B). The heat map shows differences in the cell surface proteome relative to M-CSF Mϕ and clearly demonstrates the overall decrease in several classes of proteins in IFN-γ Mϕ. It is important to highlight that the clustering of IFN-γ–activated M-CSF Mϕ does not exhibit the same pattern as an IFN-γ Mϕ, meaning that the differences come from the differentiation process rather than the activation. A list of the results of the proteomic study has been included (Supplemental Table I).
Next, we wanted to identify proteins that were selectively upregulated in IFN-γ Mϕ so that they can be used as markers for these macrophages. We compared the FC expression of the protein dataset normalized to M-CSF Mϕ across the different conditions. We found one protein that was selectively upregulated in IFN-γ Mϕ: IFN-induced GBP-1. We compared the expression of this potential marker across the three macrophage subtypes under stimulation conditions to determine whether it was sensitive to activation state. We found that GBP-1 was highly expressed under basal conditions in IFN-γ Mϕ, but it was also induced by activation with LPS or IFN-γ in both M-CSF and GM-CSF Mϕ (Fig. 3C). To independently validate the proteomics data for GBP-1, we generated additional biological replicates for all conditions and used Western blotting to confirm high GBP-1 expression in IFN-γ Mϕ under basal conditions (Fig. 3D). In these samples, we similarly observed that activation of M-CSF or GM-CSF Mϕ with LPS or IFN-γ also led to similar levels of GBP-1 as in IFN-γ Mϕ. It is worth noting in this study that the proteomics experiment specifically probed cell surface presence, whereas the Western blot quantified total protein abundance in the cell. Nevertheless, these data suggest that GBP-1 alone cannot be used as a reliable marker for IFN-γ Mϕ but rather a marker of proinflammatory macrophages.
IFN-γ Mϕ as a relevant model for psoriasis
We wanted to test the translational relevance of the in vitro IFN-γ Mϕ model and used GBP-1 expression as a starting point. We conducted a meta-analysis for GBP-1 in a transcriptomics database of inflammatory skin diseases from published clinical studies using whole-skin punch biopsies (as described in the Materials and Methods section). We found that GBP-1 was consistently and significantly upregulated in lesional skin from patients with psoriasis (Supplemental Fig. 2A). We wanted to explore this further using the functional characteristics identified earlier (high IL-12p70, IL-23, and IP-10 leading to upregulation of T cell production of IFN-γ and IL-17) and determined expression in skin of a patient with psoriasis. We further queried the psoriasis dataset and found that IFN-γ Mϕ markers were significantly enriched in lesional areas of skin of a patient with psoriasis (Fig. 4A). Although expression of GBP-1, IP-10, and IL-23 was significantly higher, it was reassuring to see that other macrophage markers, CD68 and CD14, were largely unchanged, indicating that enhanced expression of IL-23 was not due to increased numbers of macrophages in lesional skin. We tested whether the secreted mediators IP-10, IFN-γ, and IL-17 were also elevated in serum from patients with psoriasis. IP-10 levels were significantly higher in patient serum versus age-matched control subjects (Fig. 4B, left), although highly variable. IL-17 (Fig. 4B, right) and IFN-γ (data not shown) levels in patient serum were more variable and did not reach statistical significance, which may reflect differences in systemic inflammatory mediators versus local production, as the RNA levels in inflamed lesions were more robust. Using the markers identified earlier, we tested whether treatment reversed this signature in patient skin biopsies. Strikingly, GBP-1 was consistently and significantly downregulated in patients treated with biologics targeting IL-17A, IL-23, or TNF (Fig. 4C). A similar effect was seen with IL23A and IP10 but not other macrophage markers like CD14, CD68, and CD16b, suggesting that macrophage populations may not be reducing but the hyperinflammatory phenotype was being resolved. Additionally, IFNG and IL17A were also consistently suppressed following treatment, indicating the overall axis that we have shown for IFN-γ Mϕ is responsive to known efficacious treatments in patients with psoriasis. Although GBP-1 and IP-10 are not exclusively produced by IFN-γ Mϕ, these data indicate that the IFN-γ–IL-23–IL-17 axis observed in this in vitro model could be considered as a relevant model when investigating diseases such as psoriasis, in which this axis is prominent. Furthermore, GBP-1 expression may provide a valuable marker of hyperinflammatory macrophages in disease.
Differential gene expression of chromatin modifiers
IRF5 is a transcription factor that has been linked to driving proinflammatory macrophages (21, 46). We wanted to test whether IRF5 expression was elevated in our model of proinflammatory IFN-γ Mϕ. We checked basal expression levels after differentiation in the three macrophage subtypes and also after activation. Already under basal conditions, the levels were significantly higher in IFN-γ Mϕ compared with M-CSF or GM-CSF Mϕ (Supplemental Fig. 2B). In line with their proinflammatory phenotype, we found that IFN-γ Mϕ expressed higher levels of IRF5, and this was not further increased upon stimulation with LPS and/or IFN-γ. However, IRF5 was induced in M-CSF and GM-CSF Mϕ following stimulation with IFN-γ alone or in combination with LPS to levels similar to those found in unstimulated IFN-γ Mϕ.
The chromatin landscape plays a critical role in macrophage activation and phenotype (47–49). Considering this, we wanted to understand whether chromatin modifiers were differentially expressed across the three macrophage subtypes and could account for the strong phenotypic differences. We analyzed a standard RT2 Profiler PCR 84-gene array panel of chromatin modifiers in M-CSF, GM-CSF, and IFN-γ Mϕ. When analyzing the PCA, the data showed that the samples clustered depending on the differentiation factor used and that M-CSF and GM-CSF Mϕ were more similar than IFN-γ Mϕ (Fig. 5A). We generated volcano plots that showed significantly upregulated and downregulated genes (threshold of Log2 FC ± 0.58 and p value of 0.05) in the different comparisons (Fig. 5B). Gene lists for statistically up- and downregulated genes are shown in Supplemental Table II. From the gene lists obtained, Venn diagrams were generated for the genes in the comparison of IFN-γ Mϕ to either M-CSF or GM-CSF Mϕ. From this analysis, we observed that six genes were commonly upregulated in IFN-γ Mϕ versus both M-CSF and GM-CSF Mϕ (CIITA, RPS6KA5, SETDB2, NCOA3, KDMC4, and SETDB1). Moreover, six genes were downregulated in comparison with both M-CSF and GM-CSF Mϕ (DNMT3A, WHSC1, AURKA, AURKB, HDAC9, and ESCO2). Interestingly, no genes were specifically upregulated in IFN-γ Mϕ compared with M-CSF, whereas 21 were specifically upregulated when compared with GM-CSF Mϕ. In the downregulated genes, seven were selectively suppressed when compared with M-CSF Mϕ, and only one gene was specifically downregulated in IFN-γ versus GM-CSF Mϕ (NEK6). We checked the ΔΔCt to M-CSF Mϕ values of GM-CSF and IFN-γ Mϕ for specific epigenetic genes that have been widely linked to the control of inflammation, such as histone deacetylases (50). Of the histone deacetylase members investigated, only HDAC5 was upregulated in IFN-γ Mϕ. CIITA and SETBD2 were upregulated in IFN-γ Mϕ, whereas KAT2A, an acetyltransferase, and ESCO2, AURKB, and HDAC9 were downregulated in IFN-γ Mϕ (Fig. 5D).
All of these data thus demonstrate that specific epigenetic modifiers are differentially expressed in IFN-γ Mϕ and may be relevant as future targets for inflammatory disease.
The tissue microenvironment that monocytes encounter in disease situations is complex and contains a wide range of inflammatory mediators. Therefore, if we want to generate translatable in vitro models that represent this complexity, the use of a wider range of these mediators during the differentiation process is necessary. Knowing the importance of IFN-γ in (auto-) immune diseases and its higher levels in the tissue/serum in those conditions, we wanted to understand how IFN-γ would impact monocyte differentiation and characterize the resultant macrophages. We identified an optimal concentration of IFN-γ required for differentiation into viable macrophages and demonstrated that IFN-γ signaling was dominant over M-CSF signaling. IP-10 promotes the migration of T cells to inflammatory sites (42) and was rapidly induced during monocyte differentiation induced by IFN-γ but not M-CSF or GM-CSF. We further demonstrated that IFN-γ Mϕ were not only viable and exhibited macrophage characteristics through CD68 expression (51), but also exhibited phagocytic functions to the same extent as M-CSF Mϕ.
Further functional characterization demonstrated that in comparison with M-CSF and GM-CSF Mϕ, IFN-γ Mϕ had a stronger proinflammatory phenotype together with decreased anti-inflammatory/alternative response. IFN-γ Mϕ also expressed higher levels of proinflammatory surface markers CD86, CD80, and CD64 compared with M-CSF Mϕ. Finally, as regulators of the immune system, these macrophages have a higher capacity to activate Th1/Th17 responses in autologous T cells than M-CSF or GM-CSF Mϕ. This may be driven through higher expression of IL-12/IL-23 and CD80/CD86, which are important for T cell activation (52–54).
In the course of this study, we found that IFN-γ Mϕ exhibited a more proinflammatory phenotype compared with M-CSF Mϕ activated with IFN-γ (either alone or in combination with LPS). This suggests that exposure to IFN-γ during monocyte differentiation imprints a stronger proinflammatory phenotype than with activation alone.
To further characterize these cells, we performed a proteomics study and found that the IFN-γ Mϕ are significantly different to both M-CSF and GM-CSF Mϕ. One of the aims of this experiment was to find a protein that could be used as a marker to identify similar macrophages in samples from patients with immune disease. In the proteomics data, we identified the plasma membrane–associated protein, IFN-induced GBP-1, that was specifically upregulated under basal conditions in IFN-γ Mϕ, although also present following activation of the other macrophage subtypes. GBP-1 has been associated with inflammatory conditions, in which its expression is induced in epithelial cells by proinflammatory mediators, such as IL-1β, TNF, and IFN-γ (55), innate responses in defense to pathogens (56), and also elevated in patients with chronic inflammatory diseases like RA, systemic lupus erythematosus, and systemic sclerosis (57). In cancer, it has been demonstrated to have immunosuppressor activities in colorectal cancer (58, 59). The conflicting role and broader expression pattern of GBP-1 presents caveats for its use as a single specific identification marker of proinflammatory/IFN-γ Mϕ in inflammatory disease.
Taking into account the difficulties to define a single specific marker for IFN-γ Mϕ, we identified a panel of phenotypic characteristics that were indicative of a strong IFN-γ–driven effect on monocytes/macrophages, used a signature of enhanced GBP-1, IP-10, and IL-23/IL-12, and suppressed CD14 and CD16 levels. We also included increased IFN-γ and IL-17 levels in this signature given the well-established link between IL-12/IL-23 activation of T cells to produce IFN-γ and IL-17 (54, 60, 61). Searching databases for patients with these signatures, we found that inflamed skin from patients with psoriasis exhibited this pattern (high IP-10, IL-23, and IL-17 and low CD14), whereas there was no change in CD68, suggesting that macrophage numbers were not driving this increase but rather a proinflammatory macrophage population. Some of these markers (GBP1, IP10, IL23A, IFN-G, and IL17A) were also consistently downregulated in patients treated with biologics targeting IL-17A, IL-23, or TNF. Elevated, although variable, IP-10 levels in serum from patients with psoriasis also corroborated this finding, although it seems that elevated IP-10 expression is more robust locally than systemically. Our results suggest that IFN-γ Mϕ could be used as a more relevant in vitro macrophage model for the study of inflammatory diseases like psoriasis.
We also investigated potential drivers for this hyperinflammatory phenotype and focused on chromatin modifiers. Despite PCA plots showing that the differentiation factor was the key driver for differences in gene expression, there were no genes specifically upregulated in IFN-γ versus M-CSF Mϕ. However, 21 epigenetic enzymes were differential between IFN-γ versus GM-CSF Mϕ, which included the lysine demethylase 6B (KDM6B). This gene has previously reported to be upregulated after IFN-γ stimulation (62) and in proinflammatory conditions (e.g., LPS), and its depletion shows abrogation of proinflammatory cytokine production (63, 64). Genes upregulated in IFN-γ Mϕ relative to both M-CSF and GM-CSF Mϕ included CIITA [MHC CIITA known to be induced by IFN-γ (65, 66)] and RPS6KA5 [known to be induced in activated macrophages (67)], and a single nucleotide polymorphism in this gene is associated with superior efficacy of anti-TNF treatment in patients with RA (68, 69). HDAC5 and SETDB2 were specifically upregulated in IFN-γ Mϕ. HDAC5 has been reported to regulate macrophage activation and TNF production via the NF-κB pathway (70–72). SETDB2 has been reported as an IFN-induced gene and expressed in inflammatory macrophages (73). Unsurprisingly, IRF5 expression was upregulated in IFN-γ Mϕ, and this transcription factor has been strongly linked to a proinflammatory phenotype in macrophages and reported to control expression of IL-12B and IL-23A among others (21, 46). IRF5 has also been linked to inflammatory diseases (for example, in patients with juvenile idiopathic arthritis) and contributes to the pathogenesis of macrophage activation syndrome (74–76).
Genes downregulated in IFN-γ Mϕ included ESCO2 [associated with Roberts syndrome (77)], AURKB [upregulated in psoriasis (78, 79) and in RA patients (80, 81)], HDACs [HDAC9, which is linked to inflammatory macrophages in atherosclerosis (82, 83), HDAC7, and HDAC11, which is reported to block IL-10 expression in APCs (84)], and KAT2A [known to be suppressed by IFN-γ (62)].
The development of new in vitro models that better recapitulate the human in vivo environment is important not only to develop new treatments, but also to study disease pathogenesis. In this study, we show that using IFN-γ we can generate proinflammatory macrophages that have clear characteristics seen in human inflammatory disease. Thus, this may represent an important model for studying disease. The use of alternate differentiation factors based on their relevance to disease offers a potential translational advantage and could unravel novel features in these in vitro models, leading to an improved understanding of diseases.
We thank Andrea Wolf for laboratory support and Manuela Kloes-Hudak, Tatjana Ruedi, and Kerstin Kammerer for expert mass spectrometry assistance.
This work was supported by the Fondation Leducq (Transatlantic Network Grant CVD-16 to M.P.J.d.W.), the Netherlands Heart Foundation (CVON 2011/B019, CVON 2017-20, and 2019B016 to M.P.J.d.W. and 2020T029 to A.E.N.), Amsterdam Cardiovascular Sciences (to J.V.d.B., A.E.N., and M.P.J.d.W.), the Netherlands Heart Foundation and Spark-Holding BV (2015B002 to M.P.J.d.W.), and the European Union’s Horizon 2020 Research and Innovation Program under Grant ITN-2014-EID-641665 (ITN-grant EPIMAC to M.P.J.d.W.).
The online version of this article contains supplemental material.
D.C.A., M.K., S.B., H.C.E., C.A., J.F., C.S., I.R., R.K.P., and P.K.M. are employees and/or shareholders at GlaxoSmithKline. The other authors have no financial conflicts of interest.