An optimized Protocol for Human M2 Macrophages using M-CSF and IL-4/IL-10/TGF-β Yields a Dominant Immunosuppressive Phenotype

Monocytes are highly abundant circulatory effector cells and play a vital role in driving or resolving inflammatory processes depending on their activation phenotype. We investigated and compared a panel of polarization protocols of blood-derived monocytes to achieve a stable, optimal and effective regimen for in vitro induction of immunosuppressive human macrophages, evaluating their surface receptor expression, cytokine profile, scavenging function and ability to suppress T-cell proliferation. Importantly, we assessed the effect of copolarization or secondary pro-inflammatory stimulation of a primary anti-inflammatory activation phenotype. A combination of IL-4/IL-10/TGF-β yielded a relatively stable and dominant immunosuppressive phenotype characterized by higher IL-10 production and down-regulated TNF-α, IL-6, CD86, CD274 and MHC II expression. Functionally, IL-4/IL-10/TGF-β-stimulated macrophages (M2) had a potent deactivating effect on a subsequent pro-inflammatory LPS/IFNγ-activated macrophage (M1) stimulation and significantly suppressed T-cell proliferation. Monocytes derived from patients with chronic inflammatory diseases could be induced to be anti-inflammatory using this protocol. Pre-differentiation with GM-CSF or M-CSF was further demonstrated to enhance final M1/M2 activation status. Our findings indicate a robust polarization protocol for generation of specific immunosuppressive human monocyte-derived macrophages.


Introduction
Monocytes are highly dynamic and versatile mononuclear phagocytes and are involved in steady-state homeostasis, innate immune surveillance, establishment and resolution in inflammation [1]. When appropriately activated, monocytes conduct context-dependent functions in inflammatory, antimicrobial or antitumour responses [2], the local environment phenotypically polarizing the monocytes into macrophages to enable these specific functions [3]. Distinct murine subsets (Ly6C high /Ly6C low ) have specific roles in these processes [4], the spleen being a major reservoir of undifferentiated Ly6C high monocytes that are readily mobilized during inflammatory processes [5]. In contrast to blood monocytes, major tissue resident macrophage populations are now understood to originate from the yolk sac or foetal liver [6,7], to proliferate locally [8,9] and to transiently relocate early during focal inflammation [10].
Plasticity and diversity are the hallmarks of macrophage lineages. Two major extreme activation phenotypes have been described in both rodents and humans, represented by M1 (classical or pro-inflammatory) and M2 (alternative or anti-inflammatory). It has become increasingly clear that there is an overlapping spectrum of these activation phenotypes [11]. We have previously compared induced macrophage activation phenotypes from autoimmuneresistant and autoimmune-susceptible rodent strains and have determined that autoimmune-susceptible strains have a common pro-inflammatory phenotype that leads to perpetuation of inflammation instead of its resolution [12]. Similar genetically determined aberrant macrophage phenotypes have also been reported for Type 1 diabetic NOD mice [13]. Taken together, this implies that there is a genetic predisposition for autoimmune susceptibility linked to monocyte-macrophage functionality.
The outcome of a focal inflammatory response will be the net result of the relative functions, activation states and numbers of differentially activated infiltrating and resident myeloid cells under the auspices of genetic influences and local environmental cues. In states of chronic inflammation such as autoimmune diseases, there is an apparent imbalance in the pro-inflammatory (pathogenic) and antiinflammatory (resolution) properties of myeloid cells.
Macrophage colony-stimulating factor (M-CSF) and granulocyte-macrophage colony-stimulating factor (GM-CSF) are not only important hematopoietic growth factors but also potent cytokines that are needed for cell survival, proliferation, differentiation and activation [14]. While circulating monocytes are exposed to M-CSF in the blood, exposure to GM-CSF will occur primarily in tissues upon their infiltration into sites of focal inflammation. Functional heterogeneity induced by colony-stimulating factors in monocyte-derived macrophages is well-documented [15], GM-CSF-and M-CSF-derived macrophages being hypothesized to be polarized to M1 and M2 states, respectively [16][17][18].
Gene expression profiling studies have reported important differences between human and mouse monocyte subsets [19], between different human monocyte subsets [20,21], and distinct genetic fingerprints have even been reported for single human monocyte subsets [22]. We previously defined a protocol for induction of potent anti-inflammatory macrophages in mice [23]. The purpose of the current study was thus to refine this protocol to induce potent anti-inflammatory macrophages from human monocytes.
Monocyte isolation and purification. Human monocytes were isolated and purified from peripheral blood mononuclear cells (PBMC) of buffy coats obtained from healthy volunteers (Clinical Immunology and Transfusion Medicine Department of Karolinska Institutet, Sweden) using Ficoll-Hypaque (GE Healthcare, Uppsala, Sweden) according to the manufacturer's instructions. After isolation from the PBMC using a CD14 + selection kit (Miltenyi, Biotech, GmBH, Germany), monocytes were cultured at a concentration of 2 9 10 5 /ml in RPMI 1640 (Gibco, Grand Island, NY, USA), supplemented with 10% heat-inactivated foetal bovine serum (FCS; Sigma), 100 U/ml penicillin, 100 lg/ml streptomycin, 2 mM L-glutamine and 20 lM b-mercaptoethanol (all reagents from Life Technologies, Stockholm, Sweden). A total of 25 independent donors were used with minimal variation observed, and for each analysis, at least four different donors are represented.
Peripheral blood mononuclear cells were also obtained from 12 relapsing-remitting multiple sclerosis patients (seven males and five females with median age 39 years) from the in-house biobank of the Neurology Clinic of Karolinska Hospital, Sweden. Blood samples of nine spondylarthritis (SpA) (six female and three male with median age 43 years) were collected from the Rheumatology Clinic of Karolinska Hospital, Sweden. All patients were receiving medication and gave their informed consent, and the studies were approved by the local ethical committee.
Monocyte polarization, differentiation and activation. Overnight-plated monocytes were polarized for 24 h to M1 macrophages using LPS and IFNc and to M2 macrophages by different combinations of IL-4, IL-10, IL-13 and TGFb, or left untreated for the duration of the culture (M0). The cytokine doses were 20 ng/ml and LPS was used at 50 ng/ml. Monocytes were also pre-differentiated into macrophages by culture for 6 days in RPMI/10% FCS supplemented with 50 ng/ml of either M-CSF or GM-CSF in 6-well culture plates at a cell concentration of 2 9 10 6 /ml.
Flow cytometry. Twenty-four hour-stimulated monocytes were stained with a cocktail of mAbs (PE-CD163, FITC-CD80, PerCP-Cy TM 5.5-CD86, APC-CD206, PE-Cy TM 7-CD274, and Pacific blue-HLA-DR) and dead cell markers (Life Technologies) with appropriate isotype control stainings. Samples were acquired with a Gallios flow cytometer (Beckman Coulter, Brea, CA, USA) and analysed using Kaluza v1.1 software (Beckman Coulter). Expression was quantified using median fluorescence intensity (MFI) of the marker of interest. Isotype controls were applied to control for background signals.
Endocytosis assay. Endocytosis of polarized monocytes was measured by cellular uptake of 10 lg/ml Dextran Alexa Fluor 647 (Life Technologies) for 2 h at 37°C and quantified by flow cytometry. Propodium iodide was used to exclude dead cells. MFI of Alexa-dextran was quantified by flow cytometry.
ELISA cytokine analysis. ELISA kits for detection of secreted TNF-a, IL-6, TGF-b and IL-10 in cell culture supernatants were purchased from R&D Systems and used according to the manufacturer's instructions.
T-cell suppression assay. Optimization of the ratio between macrophages and PBMC determined that a 1:4-1:8 macrophage/ PBMC ratio was effective for T-cell suppression. For the assay, 25 9 10 3 /well macrophages and 10-20 9 10 4 autologous PBMC were incubated for 72 h in a total volume of 200 ll of complete media (RPMI+FCS) with 1 lg/ml anti-CD3 (OKT3). Cells were pulsed with 1 lCi/well [methyl-3 H] thymidine (Amersham, Aylesbury, UK) for 18 h before harvesting and analysis using a liquid b-scintillation counter.
RNA isolation, cDNA synthesis and RT-PCR. RNA was isolated from 1 9 10 6 cells/ml stimulated monocytes and RT-PCR performed as described [24]. Data were analysed using Bio-Rad's CFX Manager 2.0, and several housekeeping genes (i.e. GAPDH, HPRT, b-actin and 18S) were used for normalization. All experiments were related to untreated control. Primer sequences are listed in Supplementary Fig. 2. Heatmaps were generated in R (www. r-project.org/) using donor median values. Rows and columns were subjected to unsupervised clustering using the distance function: 1-cor(t(x)) applying Spearman's correlation. The colour gradients are scaled for each gene.
Statistical analysis. Statistical significance was determined by Mann-Whitney U-test unless stated otherwise. PCR data were analysed using one-way ANOVA with Tukey's multiple comparison correction. A P < 0.05 (*) was considered significant. Statistical analysis was conducted using GraphPad software (San Diego, CA, USA).

Monocyte polarization with IL-4/IL-10/TGF-b induces a unique suppressive M2 macrophage phenotype
We first screened phenotypes of healthy blood donorderived monocytes with a panel of polarization protocols. We used a combination of LPS/IFNc as a potent M1 activation protocol, IL-4/IL-13 as a wound-healing M2 activation protocol and IL-4/IL-10 with or without TGFb as deactivating M2 protocols. We assessed expression of surface receptors including costimulatory receptors (CD80, CD86, CD273 and CD274) together with MHC II and CD206 (mannose receptor). Our results revealed that CD206 was more highly expressed in M2 cells activated by IL-4/IL-10 or IL-4/IL-10/TGF-b (Fig. 1A), while CD274 ( Fig. 1B), CD86 (Fig. 1C) and MHC II (Fig. 1D) were expressed at significantly higher levels in LPS/IFNcstimulated cells than in M2 cells. The polarization protocols thus induce distinguishable macrophage phenotypes from human monocytes. As IL-10 is a potent anti-inflammatory cytokine and plays a vital role in resolving inflammation, we investigated whether any of the M2 induction protocols had the ability to induce IL-10 secretion. The combination of IL-4/IL-10/ TGF-b induced significantly higher IL-10 secretion (Fig. 1E).
Endocytosis is involved in many important cellular functions during inflammation, including clearance of apoptotic cells and cellular debris from the local environment as well as antigen presentation. We therefore investigated how the panel of stimulants affected the endocytic properties of monocytes by measuring uptake of fluorescent dextran. The results clearly demonstrate that the combination of IL-4/IL-10 or IL-4/IL-10/TGF-b induces endocytic activity (Fig. 1F).
Suppression of T-cell proliferation has been reported as a key feature of M2 cells in both tumour and inflammation biology settings. To address the suppressive ability of the M2 monocytes, we cocultured them with anti-CD3activated PBMC. M2 monocytes stimulated with IL-4/IL-10/TGF-b significantly suppressed T-cell proliferation compared to M1 monocytes (LPS/IFNc stimulated) that potentiated proliferation (Fig. 1G).
Taken together, these data indicate that the same cytokine combination (IL-4/IL-10/TGF-b) that was previously determined to be optimal for mouse macrophages could also induce human monocytes to adopt a potent antiinflammatory M2 macrophage phenotype.

Hierarchy of macrophage activation states
We next investigated the hierarchy of M1/M2 activations through firstly copolarizing monocytes with both protocols simultaneously and then using sequential M2/M1 stimulations. We specifically addressed the net effects on proinflammatory M1 pathways in both of these experimental settings.
In the copolarization setting TNF-a ( Fig. 2A), IL-6 ( Fig. 2B) and IL-12 (Fig. 2C) levels were consistently significantly reduced compared to the LPS/IFNc-stimulated M1 control by the synergistic effects of IL-4/IL-10/ TGF-b. To determine the relative stability of the M2 phenotype following a secondary pro-inflammatory stimulation, we next first induced monocytes with M2 protocols for 24 h and then after washing stimulated with LPS/IFNc for an additional 24 h. Following this secondary challenge, IL-4/IL-10/TGF-b-stimulated cells displayed elevated expression of MHC II (Fig. 2D) and CD86 (Fig. 2E), but importantly had low levels of TNF-a production (Fig. 2F) and secreted significantly higher levels of IL-10 (Fig. 2G).
These data thus indicate that IL-4/IL-10/TGF-b treatment counteracts M1 activation during copolarization and induces a robust M2 phenotype in monocytes that does not appreciably switch to an M1 phenotype following secondary pro-inflammatory stimulation.

mRNA expression in M2 monocyte-derived cells
To extend the phenotypic analyses, we screened expression of a selected range of M1 and M2 transcriptional (mRNA) markers representative of both activation states following polarization of monocytes for 24 h (Fig 2H). The established M1-and M2-associated transcripts faithfully clustered away from each other as judged from the dendrogram (Fig 2H, left). The M2-associated markers (represented in the lower half), namely CD206, IL1R2, CIITA, VEGF and HMOX1, were most highly induced following polarization with IL-4/IL-10/TGF-b.
To extend this analysis, primarily M2-polarized cells were secondarily challenged with LPS/IFNc (M2?M1, Fig 2H right-hand panel). A clear response to M1 stimulation was that M1-associated transcripts (upper half) were highly induced. However, in all M2-polarized cells, the levels of M1 markers were significantly lower, the reduction being most evident with IL-4/IL-10/TGF-b polarized cells. Furthermore, M2 markers such as CD206 or IL1R2 were retained highest with IL-4/IL-10/TGF-b polarization. Taken together, these data demonstrate firstly that distinct M2 phenotypes are induced depending on the combination of applied cytokines and secondly that IL-4/ IL-10/TGF-b polarization is relatively resistant to subsequent M1 conversion.

M2 monocytes are capable of suppressing T cells from individuals with autoimmune diseases
In states of chronic inflammation such as autoimmune diseases, the assumption is that there is a local dominance of pro-inflammatory immune activity over anti-inflammatory activity in the target organs of individuals with disease. One therapeutic possibility could thus be to adoptively transfer autologous, pre-activated, anti-inflammatory macrophages to down-regulate the pro-inflammatory pathogenesis locally, an approach we have successfully employed in different experimental models of autoimmune disease [25,26]. A prerequisite for this approach is that it is possible to induce an anti-inflammatory phenotype in the monocytes of individuals with disease. We thus next investigated this possibility in a small cohort of samples from healthy controls or individuals with either multiple sclerosis (MS) or Spondylarthritis (SpA) using M0 (unstimulated), M1 (LPS/IFNc) or M2 (IL-4/IL-10/TGF-b) polarization protocols.
Overall, we observed no consistent defects in the in vitro polarization of patient-derived monocytes in the selected targets assessed by RT-PCR as most followed the same trends as in the healthy controls ( Fig 3A). These data demonstrate that comparable macrophage phenotypes to healthy controls, both M2 anti-inflammatory (upper row) and M1 pro-inflammatory (lower row), can be induced from monocytes from individuals with pro-inflammatory diseases using the employed polarization protocols. We next assessed whether monocytes isolated from individuals with the chronic inflammatory conditions MS or SpA could be induced by the IL-4/IL-10/TGF-b polarization protocol to exert anti-inflammatory effects. Our results demonstrated that IL-4/IL-10/TGF-b-activated M2 cells could significantly suppress proliferation of autologous T cells in samples from both types of disease, respectively (Fig. 3B). Secondly, we observed a marked suppression of pro-inflammatory TNF-a secretion in M2 macrophages following simultaneous M1 activation (Fig. 3C). This demonstrates that in vitro polarized M2 cells from patients with chronic pro-inflammatory diseases are still capable of being induced to exert immunosuppressive activities. At this point, it is not well studied to what degree there is heterogeneity in monocyte/macrophage phenotypes in healthy populations or to what degree current patient treatments or disease states affect their monocyte/macrophage phenotypes. We did, however, observe modestly elevated CD36 levels in M0 macrophages of both patient groups. In light of the restricted sample size, the results should be regarded as a proof-of-principle. Samples for an extended macrophage activation proteomics study accounting for current disease state, treatment and age in much larger patient cohorts are currently being collected.

The effect of GM-CSF and M-CSF pre-differentiation on monocyte-derived macrophage activation phenotypes
The bone-marrow-derived mouse macrophages utilized in our previous study were standardly grown in M-CSF during initial differentiation into macrophages. As human monocytes are reported by some authors to represent M1 cells if differentiated with GM-CSF or M2 cells if differentiated with M-CSF, we systematically addressed the phenotypes following (1) only differentiation with each growth factor or (2) combining growth factor-induced differentiation with subsequent specific M1 or M2 activation. Flow cytometric analyses revealed that pre-differentiation with GM-CSF or M-CSF had different effects on resultant macrophage phenotypes. CD206 expression was elevated following M2 activation, with highest expression on M-CSF pre-differentiated macrophages (Fig. 4A). CD86 was elevated following M1 activation, with highest expression on GM-CSF pre-differentiated macrophages (Fig. 4B). CD274 was elevated following M1 activation, but somewhat surprisingly highest with M-CSF predifferentiation (Fig. 4C). These results indicate that while M1 and M2 activation states can be induced irrespective of pre-differentiation signals, the balance of pre-differentiation growth factors will ultimately define the final level of macrophage activation phenotype. We also screened the mRNA expression of a selected panel of polarized M1 and M2 cytokines, chemokines and surface receptors following initial M-CSF and GM-CSF pre-differentiation and subsequent polarization (Fig. 4D). The left-hand panel of the heatmap depicts primary polarization with IL-4/IL-10/TGF-b, and the right-hand panel the subsequent secondary stimulation with LPS/ IFNc. The genes cluster into distinct groups, M1 markers located at the top and M2 markers at the bottom, respectively.
In conclusion, these data demonstrated that the most robust IL-4/IL-10/TGF-b M2 activation phenotype is induced following initial pre-differentiation with M-CSF.

Discussion
We conducted the current study to address how best to induce in vitro an immunosuppressive human macrophage phenotype. We compared and contrasted selected activation protocols that we have previously tested with mouse cells and concluded that a combination of IL-4/IL-10/ TGFb yielded a phenotype characterized by: (1) expression of M2-associated surface markers and gene transcripts (e.g. CD206), (2) immunosuppressive cytokines (e.g. IL-10), (3) ability to retain the phenotype on secondary pro-inflammatory activation and (4) ability to suppress autoreactive pathogenic T-lymphocyte proliferation. Moreover, our data clearly indicate that there are both similarities and differences in final monocyte activation states depending on whether M-CSF or GM-CSF is used during predifferentiation. Following specific subsequent activation, the most potent M2 phenotype was attained in M-CSF predifferentiated monocytes, while the most potent M1 phenotype was attained in GM-CSF pre-differentiated monocytes, similar to what we have determined with murine bone marrow macrophages. Our conclusion is that we can generate a comparable IL-4/IL-10/TGF-b-polarized immunosuppressive macrophage phenotype with respect to both immunological phenotyping and functional activity in both human and murine systems.
Monocytes represent circulating cells that in steady state are exposed to M-CSF in the blood. Given appropriate chemotactic signals, they will infiltrate into tissues, transforming into pro-inflammatory tissue macrophages under the influence of GM-CSF. The tissue-destructive properties of pro-inflammatory M1 macrophages is evident in many settings of chronic inflammatory autoimmune diseases, yet immunosuppressive M2 properties are equally apparent in settings of tumours [27] and helminth infections [28]. While significant large numbers of M1 cells are typical of foci of tissue destruction in the former setting, more modest numbers are associated with the latter function. In a given pro-inflammatory focus, the relative balance of M1 cells driving disease and the M2 cells attempting to naturally regulate the immune response will determine the final outcome of either chronic inflammation or healing. Manipulation of this M1/M2 stochastic balance through adoptive transfer of in vitro preactivated macrophages thus represents a relatively logical immunomodulatory strategy and a growing body of researchers explore this possibility [29]. How long the transferred cells remain potent in their immunosuppressive functional state requires further investigation. Translation into the clinic has already been attempted, with the first human trials using immunosuppressive macrophages being conducted in a setting of spinal cord injury [30]. While the therapeutic effects in animal models of this pathology have been promising, translation into the human setting has been less efficacious.
An important aspect will be stability of the induced anti-inflammatory phenotype, which must not switch on entering a pro-inflammatory focus. Macrophage activation plasticity has been investigated previously, with a first indication of the capability of macrophages to switch phenotype in vitro being reported [31]. However, another study conversely concluded that the nature of the first of two different consecutively applied stimulants decides the final activation state of the cells [32]. Our M2 induction protocol using a combination of IL-4/IL-10/TGF-b yielded a phenotype that buffered subsequent pro-inflammatory secondary stimulation. Compared to the strong and persistent challenging stimuli we used in vitro, how M2-polarized cells behave in in vivo situations remains to be explored.
In clinical settings of pro-inflammatory diseases such as secondary progressive MS or advanced Type 1 diabetes, myeloid cell therapy using anti-inflammatory macrophages is an attractive concept. As the source of macrophages is from the patient themselves there should be ample possibilities for repeated treatments, and this would compensate for the potential short-term immunosuppressive action the transferred cells might have in vivo. While a heightened pro-inflammatory monocyte activation has typically been reported from individuals with autoimmune diseases [33][34][35][36], our data demonstrate that given an appropriately efficient polarization protocol an effective anti-inflammatory activation state can still be induced in cells recovered from individuals with disease. We consider that the use of stringently optimized polarization protocols to ensure maximal retention of functional immunosuppressive capacity post-transfer is a key consideration for macrophage therapy in pro-inflammatory diseases.