Background
Rheumatoid arthritis (RA) is characterized by synovial inflammation [
1] as well as the production of autoantibodies such as rheumatoid factor and antibodies against self-proteins (anticitrullinated peptide antibodies [ACPA]) that underwent citrullination, a posttranslational modification generated by the peptidylarginine deiminase [
2]. The synovial tissue in inflamed joints undergoes hyperplasia and forms the so-called pannus [
3]. It is well described that activated synovial fibroblasts within the pannus strongly contribute to inflammation and tissue degeneration [
4‐
6]. Furthermore, there is an infiltration of immune cells, such as B cells, T cells, dendritic cells, and macrophages, into the synovium. These activated inflammatory cells produce cytokines, chemokines, matrix metalloproteinases (MMPs), and osteoclast-promoting factors [
7‐
9], resulting in perpetuation of the inflammation, cartilage damage, and bone destruction that are characteristic of RA.
Toll-like receptors (TLRs) have been shown to contribute to the inflammatory response in RA [
10,
11]. A few TLRs (TLR2, TLR3, TLR4) have been found to be upregulated in synovial tissue in RA [
12] but not in osteoarthritis [
13,
14] or in synovial versus peripheral monocytes from patients with RA [
15]. Upregulation of TLR2 and TLR4 has been demonstrated in synovial macrophages from patients with RA but not in monocyte-derived macrophages from healthy donors (HD) [
16]. Synovial macrophages from patients with RA also showed increased tumor necrosis factor (TNF)-α and interleukin (IL)-8 expression, mediated through TLR2 and TLR4, compared with macrophages from patients with other forms of inflammatory arthritis.
Granulocyte-macrophage colony-stimulating factor (GM-CSF) and macrophage colony-stimulating factor (M-CSF) expression are known to be upregulated at sites of inflammation and autoimmunity [
17], and both are increased in synovial fluid (SF) of patients with RA. Specifically, elevated levels of TNF-α and IL-1β in patients with RA can promote the production of GM-CSF as well as M-CSF by synovial fibroblasts and chondrocytes [
17‐
19]. Several studies and reports have demonstrated that GM-CSF-, interferon (IFN)-γ-, lipopolysaccharide (LPS)-, and TNF-α-differentiated monocytes display inflammatory M1 properties, whereas M-CSF, immunoglobulin G (IgG), IL-10, IL-4, and IL-13 lead to anti-inflammatory M2 macrophages [
20,
21]. Furthermore, cluster of differentiation 14 (CD14) and CD163 [
20] as well as gene markers such as heme oxygenase 1 (
HMOX1), folate receptor β (
FOLR2), or solute carrier family 40 member 1 (
SLC40A1) characterize anti-inflammatory M2 macrophages [
22,
23]. Resident macrophages in RA were shown to exhibit a more M1-like proinflammatory activity; however, they also express M2 markers such as CD163 or
HMOX1 [
24,
25]. Hence, it remains unclear whether classical M1 or M2 or an as yet undefined macrophage population predominates numerically and functionally in RA [
19,
26,
27]. The process of M1 and M2 polarization displays a high grade of plasticity [
28], and the phenotype and activation state of polarized macrophages can be altered in a special local microenvironment or can even be reversed under pathophysiological conditions. In our study, we aimed at assessing the functional plasticity of conventional macrophage subsets under inflammatory conditions usually present in RA, such as abundant TLR ligands in synovia as a result of increased tissue damage [
1]. We therefore investigated “naive” monocytes from peripheral blood of healthy individuals or patients with RA and differentiated them into M1-like and M2-like macrophages in vitro by using GM-CSF or M-CSF, respectively. These polarized macrophage populations were then challenged with different TLR ligands (Pam3, LPS) and compared with classical cytokine activation via IFN-γ/LPS. To evaluate the functional and phenotypical reaction of the generated M1 and M2 subsets on TLR stimulation, we assessed cytokine release, expression of characteristic gene markers, and alteration in cell surface markers.
We report that TLR2 engagement impairs the anti-inflammatory activity of M2-like macrophages derived from healthy or RA monocytes without changing the expression profile of the conventional M2 cell surface markers CD14 and CD163, but altering the expression of M2-specific gene markers HMOX1, FOLR2, and SLC40A1 toward an M1-specific profile. Thus, our study implies the emergence of a “chimeric” M2 subset that exerts decreased anti-inflammatory functions and possibly even constitutes a factor that promotes the inflammatory conditions in a disease setting such as RA.
Methods
Isolation, in vitro differentiation, and stimulation of monocytes and monocyte-derived macrophages
Monocytes were isolated from peripheral blood donated from healthy individuals (blood supply center, SRK beider Basel, Basel, Switzerland) or patients with RA (Department of Rheumatology, University Hospital Basel, Basel, Switzerland). RA was determined as defined by the 2010 American College of Rheumatology/European League Against Rheumatism classification criteria. All blood donors gave informed consent to participate in the study. The studies were approved by the regional ethics review board. Monocytes were isolated from peripheral blood mononuclear cells by CD14 microbead separation (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany) and differentiated into M1-like and M2-like macrophages by culturing them in standard medium [RPMI 1640, 10% FCS, 1% glutamine, 1% antibiotics, 1% 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES)] in the presence of 50 ng/ml GM-CSF or M-CSF (PeproTech, Hamburg, Germany), respectively, for 8–10 days. Freshly prepared GM-CSF and M-CSF medium was added every 2–3 days. For M0, CD14+ separated cells were either directly processed for surface marker staining or kept in standard medium for 1–3 days for subsequent TLR stimulation experiments. Stimulation of cells was performed for 24 h with 300 ng/ml Pam3CysSerLys4 (Pam3), 100 ng/ml LPS, or 10 μg/ml polyinosinic-polycytidylic acid [poly(I:C)] (all from InvivoGen, San Diego, CA, USA). We used IFN-γ/LPS (20 ng/ml and 100 ng/ml, respectively; PeproTech) as a macrophage activation control.
Fluorescence-activated cell sorting analysis
After stimulation, cells were washed once with cold filtered PBS/0.5% bovine serum albumin (fluorescence-activated cell sorting [FACS] buffer) and stained with fluorescently labeled antibodies CD14-allophycocyanin-cyanine 7 (APC-Cy7), CD163-fluorescein isothiocyanate (FITC), CD206-BV421, CD86-phycoerythrin (PE), and CD80-FITC (all from BD Biosciences, Allschwil, Switzerland) for 30 minutes on ice in the dark. Cells were then washed three times with FACS buffer and fixed with 1% formaldehyde in FACS buffer. Cells were analyzed by FACS (BD Fortessa; BD Biosciences, San Jose, CA, USA) using FlowJo software (FlowJo, Ashland, OR, USA) for analysis. Results are presented as either the percentage of positive stained cells among the total cell population or as mean fluorescence intensity (MFI), calculated as ΔMFI = MFIspecific surface marker − MFIcorresponding unstained control and normalized to the basal MFI of unstained control cells.
Gene expression
Monocytes (M0) or M1- and M2-differentiated macrophages were cultured in 24- or 48-well plates with 270,000 cells/well or 150,000 cells/well, respectively, and stimulated as described above. RNA was isolated using the miRNeasy Micro kit (Qiagen, Hilden, Germany), and 1 μg total RNA was reverse-transcribed with the RealMasterScript SuperMix Kit (5Prime GmbH, Hilden, Germany). Gene expression of TLR2, TLR3, TLR4, HMOX1, FOLR2, and SLC40A1 was measured by qRT-PCR using TaqMan StepOnePlus (Applied Biosystems/Thermo Fisher Scientific, Foster City, CA, USA). Values were normalized to ubiquitin C (UBC) or TATA-box binding protein (TBP) messenger RNA (mRNA) levels and are presented as 2−ΔCT.
Enzyme-linked immunosorbent assay
Cells were cultured in either 48- or 96-well plates with either 150,000 cells/well or 50,000 cells/well, respectively. After stimulation, supernatants were collected and cytokine, chemokine, and MMP3 release was measured by enzyme-linked immunosorbent assay (ELISA) (IL-6, IL-8, IL-1β, and TNF-α ELISAs, eBioscience, San Diego, CA, USA; MMP3 ELISA, R&D Systems, Minneapolis, MN, USA). Values are presented either as concentration in picograms per milliliter or as a ratio.
Western blot analysis
Monocytes (4 × 106) were cultured in 60 × 15-mm dishes under M1- or M2-like differentiation conditions as described above and stimulated either (1) for 1 h to assess nuclear shuttling of NF-κB and interferon regulatory factors 3 and 7 (IRF3/7) or (2) for 30 minutes for mitogen-activated protein kinase (MAPK) phosphorylation detection. Whole-cell protein extracts were isolated by adding 300 μl of lysis buffer containing 50 mM HEPES (pH 7.5), 450 mM NaCl, 15% glycerol, 2 mM ethylenediaminetetraacetic acid, 1 mM phenylmethylsulfonyl fluoride (PMSF), and freshly added protease/phosphatase inhibitor cocktail (Cell Signaling Technology, Danvers, MA, USA). Protein was harvested by incubation on a roller shaker for 10 minutes and subsequent centrifugation for 30 minutes at 14,000 rpm. For nuclear extraction, cells were washed with cold PBS and lysed with lysis buffer containing 10 mM HEPES (pH 7.9), 1.5 mM MgCl2, 10 mM KCl, 1 mM PMSF, 5 mM dithiothreitol (DTT), and 0.1% IGEPAL (Sigma-Aldrich, Buchs, Switzerland) with freshly added protease/phosphatase inhibitor cocktail for 5 minutes on ice. Nuclear pellets were harvested after centrifugation at 10,000 rpm for 5 minutes at 4 °C and lysed for 20 minutes in buffer containing 20 mM HEPES (pH 7.9), 1.5 mM MgCl2, 420 mM NaCl, 25% glycerol, 1 mM PMSF, and 5 mM DTT with freshly added protease/phosphatase inhibitor cocktail. Protein concentrations were determined using Coomassie Plus Protein Assay reagent (Thermo Fisher Scientific, Rockford, IL, USA). Equal amounts of protein were loaded onto 12% sodium dodecyl sulfate-PAGE gels and transferred onto PVDF membranes (Bio-Rad Laboratories AG, Cressier, Switzerland). Membranes were incubated over night at 4 °C with either rabbit anti-p38, rabbit anti-phospho-p38 (Thr180/Tyr182), mouse anti-extracellular signal-regulated kinase 1/2 (anti-ERK1/2), rabbit anti-phospho-ERK1/2 (Thr202/Tyr204), rabbit anti-stress-activated protein kinase (SAPK)/c-Jun N-terminal kinase (JNK), rabbit anti-phospho-SAPK/JNK (Thr183/Tyr185), mouse anti-NF-κB p65, rabbit anti-IRF3, rabbit anti-IRF7, and mouse anti-histone H3 (all 1:1000; Cell Signaling Technology). After three washes with Tris-buffered saline with Tween 20 (TBS-T) (0.05% Tween; ROTH AG, Arlesheim, Switzerland) for 10 minutes, membranes were incubated for 1 h with secondary goat anti-rabbit IgG IR800 antibody or goat anti-mouse IgG IR700 antibody (all 1:10,000; Azure Biosystems, Dublin, CA, USA) and washed again three times for 10 minutes with TBS-T and additionally for 5 minutes with TBS before analysis. Fluorescence was detected using an Odyssey CLx imaging system (LI-COR Biosciences GmbH, Bad Homburg vor der Höhe, Germany).
Statistical analysis
All statistical analyses were carried out using Prism 7 software (GraphPad Software, La Jolla, CA, USA). Data distribution was first assessed for normality using the Kolmogorov-Smirnov test. Parametric analysis of normally distributed data was performed by ordinary one-way analysis of variance (ANOVA) using Dunnett’s multiple comparisons test. Nonparametric data were analyzed using the Kruskal-Wallis test with Dunn’s multiple comparisons test. Multiple-group analysis was carried out by ordinary two-way ANOVA using the Holm-Sidak multiple comparisons test. A p value < 0.05 was considered statistically significant. All data are presented as mean ± SD.
Discussion
Macrophages play an important role in the pathogenesis of RA. Depending on the local microenvironment, they can be polarized toward either proinflammatory M1 or anti-inflammatory M2 macrophages [
20,
21]. In the present study, we aimed to investigate the phenotypical and functional plasticity of predifferentiated M1 and M2 macrophage subtypes under conditions associated with RA, such as the presence of abundant TLR agonists. By differentiating monocytes from peripheral blood of HD or patients with RA into M1 and M2 macrophages and exposing them to TLR ligands Pam3 and LPS, we anticipated to elucidate the processes that affect infiltrating monocytes in inflamed synovial tissue. Using this experimental design, we demonstrate that M2-polarized macrophages derived from monocytes of HD or patients with RA display an impaired anti-inflammatory activity profile under TLR2 engagement compared with TLR4 stimulation. Thus, following TLR2 stimulation by its ligand Pam3, the M2 population secreted the proinflammatory cytokines IL-6 and IL-8 at levels comparable to Pam3-stimulated M1-polarized macrophages. Despite this shift toward a proinflammatory M1 function, M2 macrophages continued to express the typical M2 cell surface markers CD14 and CD163. However, gene expression of
HMOX1,
FOLR2, and
SLC40A1, three characteristic markers of an anti-inflammatory M2 phenotype, were reduced toward M1 levels, thus correlating with the promoted proinflammatory cytokine profile seen in M2 following TLR2 stimulation. Somehow, unexpectedly, we found that TLR4 stimulation by LPS also led to prominent downregulation of these M2 genetic markers, even though signaling through TLR4 resulted in strong anti-inflammatory activity as measured by ratio of IL-10 to IL-6 and to IL-8 and as expected for M2 macrophages. Thus, in conditions of abundant TLR2 stimulation, a “chimeric” M2 seems to emerge, displaying an M2-like phenotype defined by surface markers while obtaining M1-like functions as defined by genetic markers and cytokine secretion.
As reported in other publications [
39], it is possible that ex vivo monocytes from peripheral blood differ in certain aspects between HD and patients with RA. Indeed, we found that basal expression of the surface marker CD206 differed in freshly isolated monocytes from patients with RA compared with HD. However, we found that the discrete macrophage subsets generated from peripheral monocytes of patients with RA displayed similar if not equal subset-specific phenotypical and functional responses upon TLR2 or TLR4 treatment as compared with HD.
To date, in only a few studies have researchers analyzed macrophage subsets in RA [
24,
25]. Ambarus et al. [
24] compared different surface markers (CD14, CD163, CD68, CD32, CD64, CD200R, CD80) on macrophages in synovial tissue or in monocyte-derived macrophages from RA versus spondyloarthritis. In line with our study, their data indicate that, in an inflammatory environment, there exist macrophages with a mixed M1/M2 phenotype. In a second study, Soler Palacios et al. [
25] undertook phenotypic and transcriptomic characterization of ex vivo isolated CD14
+ RA SF macrophages and compared them with M1 (GM-CSF) and M2 (M-CSF) macrophages generated in vitro. Their presented data showed that RA SF macrophages exhibit a rather mixed phenotype expressing several M1-like proinflammatory markers but also including M2-like markers. Interestingly, several aspects are compatible with our M2-derived “chimeric” macrophages following TLR2 engagement. Thus, they also demonstrated that
FOLR2 and
SLC40A1 gene expression levels in RA SF macrophages were low and corresponded to the generated M1 (GM-CSF) macrophages. Instead, the expression of
HMOX1, a third genetic marker for an anti-inflammatory M2 phenotype, was similar in RA SF macrophages compared with the generated M2 subset. In our M2 generated macrophages, stimulation with LPS resulted in significant downregulation of
HMOX1, whereas Pam3 only mildly reduced
HMOX1 expression. Nevertheless, LPS induced strong anti-inflammatory activity with high IL-10/IL-6 and IL-10/IL-8 ratios. These results suggest that low expression levels of
HMOX1 alone cannot discriminate M1 from M2 macrophages in terms of an anti-inflammatory cytokine secretion profile. In addition, RA SF macrophages in the study of Soler Palacios et al. [
25] also exhibited a tendency to express several markers apparent in an M2 cell type, such as increased CD14 and CD163 levels (as measured by MFI) or superior IL-10 expression compared with their generated M1 cell type. Our data therefore indicate that, in contrast to what has been conventionally proposed, surface markers as well as individual gene expression markers do not correlate with proinflammatory or anti-inflammatory cytokine expression in M1 and M2 macrophages under inflammatory conditions, which presumably consist of combined TLR2 and TLR4 stimuli. In this context, it is noteworthy that, under certain circumstances, M-CSF-generated macrophages (M2-like) can exhibit proinflammatory activity, as demonstrated by stimulation with ACPA [
40] or TLR ligands in combination with IgG [
30]. Likewise, Vogelpoel et al. [
30] reported synergistic upregulation of proinflammatory cytokines in M2 macrophages exposed to IgG and TLR ligands, which did not differ between macrophages derived from HD or patients with RA. Furthermore, in addition to pathogen-associated molecular patterns (PAMPs), also damage-associated molecular patterns (DAMPs) such as extra domain A fibronectin, tenascin-C, serum amyloid A, high-mobility group box 1 protein, and gp96 are potent agonists on TLRs. All of these DAMPs have been found at elevated levels in synovia from patients with RA [
41], and tissue-resident macrophages are potentially exposed to a mixture of PAMP- and DAMP-related ligands. In the context of DAMPs, it is also noteworthy that TLR2, but not TLR4, signaling induces strong MMP3 secretion, which was found to be a critical factor in the progression of cartilage and bone erosion in advanced RA [
37].
In our in vitro study, we could demonstrate that in the presence of abundant TLR2 ligands, M2 macrophages derived from peripheral blood of HD or patients with RA lose their anti-inflammatory activity. IL-10 expression is significantly lower in M2 upon TLR2 engagement than with TLR4 stimulation. IL-10 is a major regulator of immunity to infection [
42]; it inhibits the activity of Th1 cells, natural killer cells, and macrophages and limits the production of proinflammatory cytokines and chemokines. ERK1/2 is part of the signaling cascade that is activated in macrophages and promote the production of IL-10 [
43,
44]. In our study, we observed that both Pam3 and LPS activate ERK1/2, p38, and JNK to a similar extent, despite the differential effect of the applied ligands on pro- and anti-inflammatory cytokine levels. These observations point to a regimen of regulatory steps that govern the inflammatory and anti-inflammatory responses of M1 and M2 macrophages upon TLR stimulation. Thus, TLR stimulation might generate a broad MAPK signaling that then will subsequently be discriminated at different regulatory checkpoints, such as fine-tuning of downstream target gene expression by a specific set of microRNAs. Of note, all investigated MAPKs were activated to a higher degree in M2 than in M1 after stimulation with TLR2 and TLR4 ligands. These observations might have implications for the use of MAPK inhibitors as anti-inflammatory therapy [
45] in RA, because administration of such drugs would reduce the function not only of proinflammatory M1 but also of anti-inflammatory M2 subsets.
TLRs have been shown to be highly expressed in rheumatoid synovial tissue or synovial macrophages from patients with RA [
12‐
16], and the stimulation of these receptors plays a role in the pathogenesis of RA [
10,
46,
47]. The importance of TLR signaling for the pathogenesis of RA has been suggested by studies with murine arthritis models. Abdollahi-Roodsaz et al. [
48] found that development of streptococcal cell wall-induced arthritis in mice was dependent on TLR2 during the acute phase, and this effect shifted to TLR4 dependency during the chronic phase. They also showed that administration of a TLR4 antagonist suppressed clinical and histologic characteristics of arthritis in a mouse model of collagen-induced arthritis [
49]. Pierer et al. [
50] revealed a significantly lower incidence of collagen-induced arthritis in TLR4-deficient mice. In a model of zymosan-induced arthritis, it was demonstrated that TLR2-deficient mice showed a decrease in early and late phases of joint inflammation [
51]. These studies indicate that both receptors play an important role in the development of arthritis.
TLRs are discussed as therapeutic targets for inflammatory diseases but also for cancer [
52,
53]. Several agonists and antagonists are under development and are already in different clinical phases. Therapeutic effects of treatment with anti-TLR2 and anti-TLR4 monoclonal antibodies have been investigated in a mouse study against polymicrobial sepsis [
54]. Interestingly, a single administration of either anti-TLR2 or anti-TLR4 increased the survival rate and decreased peritoneal, serum, and lung TNF-α levels more efficiently than a combinatorial approach. Our results indicate that therapeutic administration of TLR4 antagonists in RA may result in less beneficial treatment outcomes because the anti-inflammatory activity of M2 macrophages might be impaired by the blockage of TLR4. In addition, depletion of TLR4 signaling would possibly render macrophages more prone to activation of TLR2 by binding of DAMPs present in the synovium of patients with RA. By contrast, the use of TLR2 antagonists as potential RA therapeutics might decrease the release of proinflammatory cytokines from M1-like macrophages while increasing the anti-inflammatory properties of M2-like macrophages by allowing TLR4-mediated IL-10 secretion.