CD11b⁺Ly-6C^hi Suppressive Monocytes in Experimental Autoimmune Encephalomyelitis¹

Experimental autoimmune encephalomyelitis (EAE)³ is an autoimmune inflammatory disease of the CNS, induced by immunization with myelin Ags (1). It is an important model for multiple sclerosis, and is also valuable for studying the pathogenesis of autoimmunity in general (2). EAE is initiated by the peripheral activation and proliferation of myelin-specific CD4⁺ T cells, which then migrate to the CNS. The T cells are reactivated in the CNS and produce an array of proinflammatory cytokines and chemokines (3, 4). Microglia and astrocytes are also activated, and a large number of macrophages and neutrophils are recruited from the periphery, resulting in CNS inflammation (5, 6, 7).

Endogenous regulatory mechanisms in EAE include cytokines, regulatory T cells, negative costimulatory pathways, and NO. Although EAE severity is usually correlated with the presence of Th1 cytokines in the lesions, IFN-γ is not necessary for the disease induction (8, 9). In fact, IFN-γ knockout mice develop more severe EAE, suggesting that IFN-γ has an important immune regulatory function (10). The production of Th2 cytokines such as IL-4, IL-10, and IL-13, as well as TGF-β is generally associated with EAE recovery and tolerance induction (11, 12, 13, 14, 15). Exogenous administration of IL-10 can ameliorate disease in some EAE models (16, 17). However, studies investigating how cytokines regulate innate immune cells in EAE are less extensive.

The programmed death-1 (PD-1)/PD-1 ligand pathway is an important negative costimulatory pathway for activated T cells. We have shown that PD-L1 and PD-L2 are important in maintaining EAE resistance and in suppressing the chronic disease, but specific PD-1 ligands play differential roles depending on the genetic background (18). In BALB/c mice, the blockade of PD-L1 but not PD-L2 increases EAE incidence. We found that purified T cells from anti-PD-L2-treated mice have greatly increased spontaneous proliferation, but coculture with splenic CD11b⁺ cells from either anti-PD-L2 or control-treated mice strongly suppressed T cell proliferation, suggesting that myeloid cells can suppress highly activated T cells induced by PD-L2 blockade.

Accumulating evidence indicates that adaptive immunity is initiated, programmed, and constantly regulated by innate immune cells. In recent years, a population of myeloid cells suppressing immune responses against tumor Ags has been described. These cells express CD11b and Gr-1, and have been named myeloid-derived suppressor cells (MDSC) (19, 20). CD11b⁺Gr-1⁺ cells are a heterogeneous population including immature macrophages, granulocytes, dendritic cells, and other myeloid cells (21, 22). MDSCs may suppress anti-CD3/anti-CD28 induced T cell proliferation (23), down-regulate CD3ζ-chain expression (24), inhibit the cytotoxicity of CD8⁺ T cells (25), and induce T cell apoptosis (26). MDSCs express NO synthase 2 (NOS2) and arginase 1, and both NO production and arginine degradation are involved in T cell suppression (27). CD11b⁺Gr-1⁺ suppressive cells have also been identified in parasite infection models and in a trauma model (28, 29, 30, 31). Because both mononuclear cells and granulocytes express CD11b and Gr-1, the specific suppressive population within the CD11b⁺Gr-1⁺ cells is not clear.

In this study, we found that a population of inflammatory monocytes is expanded in the periphery after EAE immunization. Phenotypically, we characterized these cells as CD11b⁺Ly-6C^highLy-6G⁻. They are also F4/80⁺ and CD93⁺, but CD31⁻. These cells are highly suppressive for both activated CD4⁺ and CD8⁺ T cells, and induce T cell apoptosis through the production of NO. In addition, CD11b⁺Ly-6C^high cells are present at a high frequency in the CNS during clinical EAE, suggesting they have important immune regulatory functions.

Female BALB/c and C57BL/6 mice, C.Cg-Tg(DO11.10)10Dlo/J mice, C.129S7(B6)-Ifng^tm1Ts/J mice, B6.129P2-Nos2^tm1Lau/J mice were obtained from The Jackson Laboratory, and housed according to local and National Institutes of Health guidelines. All animals were used at 6–8 wk of age. Nonspecific NOS inhibitor N^G-monomethyl-l-arginine (l-NMMA), arginase 1 inhibitor N^ω-hydroxyl-nor-l-arginine (nor-NOHA), specific NOS2 inhibitor N6-(1-iminoethyl)-l-lysine (l-NIL) were obtained from Calbiochem. Lipopolysaccharides from E. coli 055:B5 (LPS), peptidoglycan from Staphylococcus aureus (PGN), indoleamine 2,3-dioxygenase (IDO) inhibitor 1-methyl tryptophan (1-MT) and caspase inhibitor z-VAD-fmk were obtained from Sigma-Aldrich. CpG oligodeoxynucleotide was from Invivogen. Various cytokines, FACS Abs and anti-IFN-γ neutralizing Ab (clone XMG1.2) were purchased from BD Biosciences and eBioscience. Chicken OVA peptide 323–339 (ISQAVHAAHAEINEAGR) was synthesized by Quality Controlled Biochemicals.

MOG35–55 peptide (MEVGWYRSPFSRVVHLYRNGK) was synthesized by Quality Controlled Biochemicals and was purified to over 99% by HPLC. Mice were immunized subcutaneously in the flank with the emulsion made of 75 μl of Ag peptide (150 μg of MOG35–55) and 75 μl of complete Freund’s adjuvant containing 0.3 mg heat-inactivated Mycobacterium tuberculosis (H37Ra; Difco Laboratories). Each animal also received 200 ng of pertussis toxin (List Biological Laboratories) through i.v. injections on day 0 and 2 postimmunization.

Splenocytes were isolated after RBC lysis. The first method for subset isolation involved staining with anti-CD11b and anti-Gr-1 Abs, and four major populations were purified by cell sorting. In the second isolation method, CD11b⁺ cells were positively selected using CD11b MACS beads (Miltenyi Biotec), and then stained with anti-Ly-6C and anti-Ly-6G Abs. Three myeloid cell populations were isolated by cell sorting. Purified myeloid cells were cultured in X-VIVO serum-free medium or in RPMI 1640 medium containing 10% FBS, with the addition of glutamine, 2-ME, sodium pyruvate, nonessential amino acid, and antibiotics (BioWhittaker).

Flat bottom 96-well plates were coated with anti-CD3/anti-CD28 (both 1 μg/ml) for 3 h at 37°C. Splenic CD4⁺ and CD8⁺ T cells were purified using specific MACS beads (Miltenyi Biotec), and then stimulated with plate-bound anti-CD3/anti-CD28 at 2 × 10⁵ cells/well for 24 h. These activated T cells were either cultured alone or cocultured with various myeloid cell populations, in the presence of plate-bound anti-CD3/anti-CD28 stimulation. After 24 or 48 h, 1 μCi [³H]thymidine was added into each well, and cells were harvested 16 h later. For the proliferation assay of splenocytes from DO11.10 TCR transgenic mice, splenocytes were isolated from nonimmunized animals, and cultured at 2 × 10⁵ cells/well with 0, 1, 5, or 25 μg/ml OVA_323–339 peptide in a round-bottom 96-well plate. Myeloid cells were added at the beginning of culture. After 24 h, 1 μCi [³H]thymidine was added into each well, and cells were harvested 16 h later. To measure the cytokine concentration in culture, supernatants were collected before adding [³H]thymidine, and assays were conducted using Beadlyte mouse multicytokine detection system (Upstate Biotechnology), according to the manufacturer’s protocol.

For cell surface marker staining, isolated cells were blocked with 10 μg/ml Mouse BD Fc Block at 4°C for 5 min, and labeled with various fluochrome-labeled Abs and 7-aminoactinomycin D (7-AAD) including proper isotype controls for 30 min at 4°C. After two steps of washing, cells were analyzed on FACSCaliber (BD Biosciences). Intracellular perforin staining was performed according to eBioscience staining protocol. For CFSE labeling, CD4⁺ cells were incubated in 5 μM CFSE for 5 min at room temperature, followed by four times of wash. Annexin V and 7-AAD staining was conducted following the BD Biosciences protocol.

The cells were collected on the slides using a cytospin machine. After air-drying for 30 min, the slides were stained with a Hema 3 staining kit (Fisher Scientific).

This assay was performed using the nitrate/nitrite colorimetric assay kit from Cayman Chemical Company. After converting nitrate in the culture supernatant to nitrite with nitrate reductase, Griess Reagents were added to convert nitrite into a deep purple compound. The absorbance at 550 nm was examined with a microplate reader (Bio-Rad). The concentration of nitrate/nitrite was determined by comparing with standards.

The animals were sacrificed and perfused with 10 ml of PBS. Spinal cord tissues were dissected out, cut into small pieces, and digested with collagenase IV for 20 min at 37°C. The samples were then passed through 70 μm meshes, resuspended in 30% percoll, and loaded onto 70% percoll. After centrifuge at 1300 × g for 20 min, the CNS inflammatory cells were retrieved from the 30/70% percoll interface, and washed in PBS.

Total RNA was extracted using TRIzol reagent (Invitrogen Life Technologies). Complementary DNA was synthesized using SuperScript III first-strand synthesis system for RT-PCR (Invitrogen Life Technologies). NOS2, Arginase-1, and β-actin mRNA expression were examined using TaqMan gene expression assays. Triplicate samples were examined in each condition. A comparative threshold cycle (C_T) value was normalized for each sample using the formula: ΔC_T = C_{T (gene of interest)}– C_{T (β-actin)}, and the relative expression was then calculated using the formula 2^−ΔCT.

The data in text represent the mean ± SEM, and the error bars in figures also represent SEM. Unpaired two-tailed t tests were used to compare the significance of differences between two groups, and one-way ANOVA followed by Bonferroni tests were used to analyze data with more than two groups.

In normal BALB/c mice, 5.9 ± 0.5% of splenocytes were CD11b⁺, but this frequency was increased to 34.8 ± 3.6% 10 days after MOG_35–55 immunization (p < 0.001, Fig. 1,A). This increase was not specific to MOG_35–55 peptide, because immunization with PBS/CFA emulsion plus pertussis toxin treatment induced a similar increase in spleen CD11b⁺ cell frequency (data not shown). To investigate the function of these cells, we isolated splenic CD11b⁺ cells from immunized BALB/c mice on day 10 postimmunization, and cocultured them with CD4⁺ T cells that had been activated with plate-bound anti-CD3/anti-CD28 for 24 h. Proliferation assays after 24 and 48 h of coculture show that spleen CD11b⁺ cells can suppress CD4⁺ T cell proliferation (Fig. 1 B).

To better define the subpopulation of CD11b⁺ cells with T cell suppressive function, we stained splenocytes from immunized mice with anti-CD11b and Gr-1 Abs, and found four major populations of CD11b⁺ cells (Fig. 2,A). They are as follows: CD11b^intGr-1^−/low (G1), CD11b^highGr-1^int (G2), CD11b^intGr-1^high (G3), and CD11b^highGr-1^high (G4) cells. We sorted each population and tested their ability to suppress pre-activated CD4⁺ T cells as described above. As shown in Fig. 2,B, CD11b^highGr-1^int (G2) cells were strongly suppressive, and CD11b^highGr-1^high (G4) cells were mildly suppressive. Other two populations did not show significant suppressive effects. Thus, CD11b⁺Gr-1⁺ population in EAE model is heterogeneous in function. Because the Gr-1 Ab recognizes both Ly-6C and Ly-6G epitopes (32), CD11b⁺ population was further characterized with anti-Ly-6C and anti-Ly-6G Abs. We found three populations in the CD11b⁺ cell gate, i.e., Ly-6C^lowLy-6G⁻ cells, Ly-6C^intLy-6G⁺ cells, and Ly-6C^highLy-6G⁻ cells (Fig. 3,A). We sorted each population and cocultured the cells with pre-activated CD4⁺ T cells at 1:1 ratio for 24 h. The data show that Ly-6C^highLy-6G⁻ cells completely inhibited T cell proliferation, while the other two populations did not have any suppressive function (Fig. 3,B). CD11b⁺Ly-6C^highLy-6G⁻ cells were highly effective in suppressing CD4⁺ T cells, because a 1:4 coculture could still suppress ∼90% of the proliferation, and 1:8 ratio culture suppressed about half of the T cell proliferation (Fig. 3,C). Similar suppressive effects were observed when CD11b⁺Ly-6C^highLy-6G⁻ cells were cocultured with anti-CD3/anti-CD28-stimulated CD8⁺ T cells (Fig. 3,D). These data suggest that CD11b⁺Ly-6C^highLy-6C⁻ cells from immunized animals are able to efficiently suppress CD4⁺ and CD8⁺ T cell proliferation in an Ag-nonspecific manner. In addition, CD4⁺ T cell production of various cytokines, which include IFN-γ, TNF-α, IL-17, IL-4, and IL-10, was also suppressed by CD11b⁺Ly-6C^highLy-6G⁻ cells (Fig. 3 E). The less profound suppression of cytokines compared with T cell proliferation is likely due to the accumulation of cytokines during the 24 h CD4⁺ T cell pre-activation.

To characterize the morphology of three groups of CD11b⁺ cells, we performed Wright-Giemsa staining (Fig. 4, A–C). CD11b⁺Ly-6C^highLy-6G⁻ cells are exclusively large mononuclear cells (Fig. 4,A). CD11b⁺Ly-6C^intLy-6G^high cells are mostly mature neutrophils, or immature neutrophils with ring-shaped nuclei (Fig. 4,B). CD11b⁺Ly-6C^lowLy-6G⁻ cells include mixed size mononuclear cells as well as eosinophils (Fig. 4 C). We will call the CD11b⁺Ly-6C^highLy-6G⁻ cells “inflammatory monocytes” (IMC) to be consistent with the literature (33, 34).

Morphologically, the suppression of proliferation by CD11b⁺Ly-6C^high cells (IMC) was accompanied by massive T cell death (Fig. 4, D–G). Although control CD4⁺ T cell culture (Fig. 4,D) and coculture with CD11b⁺Ly-6C^low (Fig. 4,E) or CD11b⁺Ly-6C^int cells (Fig. 4,F) show normal T cell blasts, the coculture with IMC shows that the vast majority of T cell blasts became fragmented apoptotic bodies with condensed nuclei (Fig. 4,G). Annexin V and 7-AAD staining shows that 36 h of coculture with IMC significantly increased apoptotic cell death in CD4⁺ T cells (61.2 ± 0.7% in coculture vs 27.4 ± 0.6% in CD4⁺ T cells only, p < 0.0001. Fig. 5, A and D). We believe that the degree of T cell apoptosis induced by IMC is under-represented in Annexin V/7-AAD staining because many dead cells were lost during the staining procedure. When CD4⁺ T cells were further characterized with proliferation-induced CFSE dilution, we found that proliferating CD4⁺ T cells were preferentially eliminated (Fig. 5, B and E). Consistent with this, the remaining live CD4⁺ T cells in the coculture with IMC showed fewer proliferation cycles (Fig. 5, C and F). In addition, Annexin V/7-AAD staining confirmed that CD11b⁺Ly-6C^low cells, which are another mononuclear cell population, did not increase CD4⁺ T cell apoptosis (data not shown).

To investigate the possible role of NOS, arginase 1, or IDO in mediating T cell death, we used their respective inhibitors l-NMMA, nor-NOHA, and 1-MT in the cocultures. l-NMMA treatment fully restored CD4⁺ T cell proliferation, while nor-NOHA and 1-MT did not have any significant effect (Fig. 6,A). Furthermore, z-VAD-fmk, a pan-caspase inhibitor, could also fully restore T cell proliferation (Fig. 6 A). These data suggest that NO produced by IMC induces T cell apoptosis through the activation of caspases.

To examine the specific role of inducible NOS (NOS2) and directly measure NO production, we cocultured activated CD4⁺ T cells with different ratios of IMC from immunized animals, and treated with a specific NOS2 inhibitor N6-(1-iminoethyl)-l-lysine (l-NIL) in 1:2 IMC/CD4 T cell coculture (Fig. 6,B). We found that specific NOS2 inhibition completely reversed suppression. Furthermore, the nitrate and nitrite production in the culture supernatants correlated with the numbers of seeded IMC and the degree of T cell suppression. l-NIL completely blocked NO production from IMC (Fig. 6 C). Similarly, treatment of the coculture with anti-IFN-γ neutralizing Ab also reversed the suppression and blocked NO production from IMC. These data suggest that both IFN-γ production in the coculture and NOS2 activity are required for NO production from IMC and their suppressive function. In contrast, treatment with catalase (a H₂O₂ scavenger), or the blockade of PD-1, PD-L1, PD-L2, CD54, or F4/80 had no effect on IMC-induced T cell suppression (data not shown).

In addition, the suppression by IMC was also effective in an Ag-specific assay in the presence of functional APCs. Thus, IMC abolished the proliferation of DO11.10 TCR transgenic splenocytes stimulated with OVA_323–339 (Fig. 6,D). NOS2 blockade with l-NIL similarly reversed suppression (Fig. 6 E).

To confirm the essential role of NOS2 activity in IMC suppression, IMC were isolated from MOG_35–55-immunized wild type and NOS2^−/− B6 mice, and cocultured with anti-CD3/anti-CD28 pre-activated CD4⁺ T cells from normal B6 mice (Fig. 7,A). The data show while IMC from wild-type B6 mice are functional, IMC from NOS2 knockout mice could not suppress CD4⁺ T cell proliferation. Because IFN-γ neutralization abrogated T cell suppression, we examined whether IFN-γ from IMC or from CD4⁺ T cells was required. We found that IMC from IFN-γ knockout mice (BALB/c background) were fully functional in T cell suppression (Fig. 7,B), but IMC from wild-type BALB/c mice did not suppress IFN-γ knockout CD4⁺ T cells (Fig. 7 C). Collectively, these data suggest that IFN-γ produced by CD4⁺ T cells induces NOS2 activity in IMC, which is necessary for IMC-mediated T cell suppression.

We examined NOS2 and arginase 1 expression in IMC after ex vivo purification from MOG_35–55-immunized BALB/c mice, but neither enzyme could be detected by real time PCR (Fig. 8, A and B). After 24 h of in vitro culture without additional treatment, only NOS2 was weakly induced. When IMC were treated with various cytokines, we found that IFN-γ and GM-CSF strongly up-regulated NOS2 expression, but IL-4 could antagonize IFN-γ-induced NOS2 up-regulation. Treatment with other cytokines including IL-17 or TLR agonists LPS, CpG and peptidoglycan (PGN) did not strongly induce NOS2 expression (Fig. 8,A). We also examined arginase 1 expression after cytokine and TLR agonist treatment, and only IL-4 induced arginase 1 expression. Interestingly, IFN-γ treatment also antagonized arginase 1 up-regulation by IL-4 (Fig. 8 B). These data suggest that spleen IMC do not express significant level of NOS2 or arginase 1 after 10 days of immunization, but they have the plasticity to up-regulate NOS2 or arginase 1 upon encountering different cytokines. IFN-γ and IL-4 can antagonize each other in the induction of NOS2 and arginase 1 expression.

We examined the frequency of CD11b⁺Ly-6C^high IMC in the bone marrow, blood, spleen, lymph nodes, and CNS in naive BALB/c mice and during the course of EAE (Fig. 9). In the bone marrow, blood, and spleen, the frequency of IMC increased 5–10-fold after immunization, and reached a peak around day 10 postimmunization. In contrast, the frequencies of CD11b⁺ cells and IMC were <2% and 0.2% respectively in the lymph nodes, including the inguinal draining lymph nodes. In the CNS, the peak of CD11b⁺Ly-6C^high cell frequency was delayed compared with the periphery. IMC constituted ∼5–10% of CNS inflammatory cells before EAE onset (day 7 and day 10), but they constituted ∼30% of the infiltrating cells at the peak of clinical EAE disease (day 14). The IMC frequency in the CNS gradually decreased after EAE recovery. Therefore, IMC are recruited into the CNS in parallel to the development of CNS inflammation.

Phenotypic characterization shows that CD11b⁺Ly-6C^high cells express F4/80 Ag, but do not express CD3, CD11c, CD19, or DX-5 (Fig. 10 and data not shown), consistent with a macrophage lineage. IMC specifically express CD93 (AA4.1), which is known expressed by immature B cells and other hemopoietic cells in the bone marrow (35, 36). However, IMC do not express CD31, an Ag expressed on some hemopoietic progenitor cells as well as mature leukocytes (37). IMC express some adhesion molecules, such as CD11a, CD11b, CD49d, and low levels of CD62L. They do not express IA/IE, CD80, CD86, PD-1, PD-L1, PD-L2, B7-H3, or B7-H4, suggesting that they are not mature APCs. IMC do not express Fas ligand, CD25, CD122 or intracellular perforin (data not shown), suggesting that these molecules may not participate in T cell suppression. In contrast, many CD11b⁺Ly-6C^lowLy-6G⁻ cells express F4/80, CD11c, IA/IE, CD80, CD86, CD40, or PD-L1, and thus may have Ag-presenting functions. In summary, the phenotype of IMC suggests that they are macrophage lineage cells recently emigrated from the bone marrow, and they are not yet mature APCs.

In this study, we found that CD11b⁺Ly-6C^highLy-6G⁻ cells, which are markedly increased in the spleen and blood after immunization, are highly suppressive for activated CD4⁺ and CD8⁺ T cells. Activated T cells produce IFN-γ, which induces NOS2 expression in IMC. IMC then produce NO and induce T cell apoptosis. These cells are mononuclear in morphology, express higher levels of Ly-6C and CD93 than other CD11b⁺ subsets, but lack the expression of MHC class II or costimulatory molecules. They up-regulate NOS2 or arginase 1 in response to different cytokines.

In BALB/c EAE model, we determined that the suppressive myeloid cells are exclusively CD11b⁺Ly-6C^highLy-6G⁻ cells. Ly-6C can be detected on monocytes/macrophages, endothelial cells, plasma cells, thymocytes, NK cell, and T cell subsets (38), while Ly-6G is expressed on granulocytes in both the bone marrow and periphery (32). The function of these molecules is not clear yet. We found that in the EAE model, CD11b⁺ cells can be divided into Ly-6C^high, Ly-6C^int and Ly-6C^low populations. Although Ly-6C^intLy-6G⁺ cells are mature and immature neutrophils in morphology, Ly-6C^highLy-6G⁻ cells have clear mononuclear morphology. Sunderkotter et al. (34) reported that bone marrow monocytes representing earlier developmental stages express a high level of Ly-6C, while blood monocytes at steady state express lower levels of Ly-6C. Ly-6C^high monocytes were significantly increased in the circulation following blood monocyte depletion and during parasite infection (34). We found that CD11b⁺Ly-6C^high cells express a higher level of CD93 (AA4.1) than the other two subsets of CD11b⁺ cells, which is another immature hemopoietic cell marker (35, 39). Ex vivo purified IMC did not express significant levels of NOS2 or arginase mRNA, but they showed plasticity in upregulating these molecules after IFN-γ or IL-4 treatment. This would suggest that spleen IMC had not differentiated into M1 or M2 like cells in vivo. Based on their morphology, and expression of Ly-6C, CD93, CD11b, and F4/80, but lack of other lineage marker or costimulatory molecule expression, we conclude that the suppressive myeloid cells are immature inflammatory monocytes recently developed from hemopoietic progenitor cells. In addition, we found that EAE-primed C57BL/6 mice developed the same Ly-6C^high suppressive population, which constitute ∼15% in CD11b⁺ splenocytes. As shown in Fig. 7 A, B6 IMC virtually eliminated CD4⁺ T cell proliferation after 24 h coculture. However, we have noticed that B6 IMC were less suppressive than BALB/c IMC when cultured at lower ratios with CD4⁺ T cells (data not shown). We are currently investigating the mechanisms of this difference.

NO has both protective and pathogenic functions in EAE. Blockade of NOS2 activity in the priming phase of active EAE model accelerated disease onset and increased EAE severity (40). Similarly, NOS2 knockout mice developed worse EAE (41, 42), and increased numbers of CD4⁺ and CD8⁺ memory T cells persisted in the late EAE stage (43). In addition, EAE-resistant strains such as in PVG and Brown Norway rats may produce higher amount of NO upon immunization, and this mechanism may directly contribute to disease resistance (44, 45). Consistently, enhanced NO generation by CFA pre-immunization or treatment with a NO donor, SIN-1, significantly suppressed EAE development (46, 47). These data indicate that NO may play a protective role in EAE. However, NOS2 inhibition during adoptive transfer EAE model alleviated the disease, suggesting that NO overproduction in the CNS may have a pathogenic role in causing tissue damage (40, 48). Our data show that the Ly-6C^high subset of CD11b⁺ cells suppresses T cells through the production of NO. Thus, CD11b⁺Ly-6C^high cells may represent the NO-producing regulatory myeloid cells in EAE. Interestingly, ex vivo purified IMC from EAE immunized mice did not express significant level of NOS2 mRNA. This may be due to the rapid generation of IMC in vivo, and the possibility that recently emigrated IMC have not exposed to activated T cells for a sufficient time to up-regulate NOS2. It is also possible that IMC activated by T cell-derived IFN-γ are prone to cell death after producing a large amount of NO. In this regard, we found that >80% of IMC died after overnight culture with CD4⁺ T cells, a rate much higher than GM-CSF treated IMC (data not shown). In contrast, we found that ∼30% of CNS inflammatory cells isolated from EAE animals were CD11b⁺Ly-6C^high cells, suggesting that IMC can efficiently migrate into the inflammatory tissues. It is likely that IMC have both protective and pathogenic functions in vivo: IMC may suppress activated T cells both in the periphery and in the CNS, contributing to the EAE resistance in certain strains and to EAE recovery, but NO over-production in the CNS may also enhance neural tissue damage. The role of IMC in inflammatory CNS may be similar to that of activated microglia (49).

Although our data suggest that both IFN-γ and GM-CSF can strongly up-regulate NOS2 mRNA in IMC, neutralization of IFN-γ or coculture with IFN-γ^−/− CD4 cells completely abrogated IMC suppression, suggesting IFN-γ is an essential physiological factor inducing NO production in IMC. Because IFN-γ knockout mice had exacerbated EAE (10), it is likely that in these mice IMC lose their NO producing capacity and thus their regulatory function.

Myeloid-derived suppressor cells have been described in tumor (19, 20), parasite and fungus infection (29, 50, 51, 52), and trauma models (30). Most of these studies use CD11b⁺Gr-1⁺ as a phenotypic marker. Recently, Huang et al. (53) reported that Gr-1⁺CD115⁺ cells showed stronger suppressive function than CD11b⁺Gr-1⁺ cells, and Gallina et al. (22) described that CD11b⁺IL-4Rα⁺ could be a more specific phenotypic markers for MDSC. We found that CD11b⁺Gr-1⁺ cells were increased to 20–25% in the spleen and blood of immunized animals, but only CD11b⁺Ly-6C^high cells, which constitute 4–5% of splenocytes and blood leukocytes, are strongly suppressive. It is unclear whether suppressive myeloid cells in other disease models have the same CD11b⁺Ly-6C^high phenotype.

There are some differences between MDSC in tumor models and IMC in our EAE model. First, MDSC derived from spleen of tumor-bearing mice are not suppressive for CD4⁺ cells unless they are cultured in vitro for a few day or treated with IL-4 before coculture with T cells (25, 27). IMC from EAE immunized mice are very efficient in T cell suppression once purified. Secondly, MDSC from tumor models express both NOS2 and arginase 1, and only inhibiting both enzymes reverses the suppression (27, 54). In contrast, IMC from EAE model preferentially use NOS2 for T cell suppression. Thirdly, MDSC from tumor model are not dependent on IFN-γ activation because MSDC from IFN-γ receptor knockout mice were still suppressive (55), while IMC from EAE model depend on T cell-derived IFN-γ signaling for suppression. Lastly, MDSC may use other mechanisms such as oxygen radical production for T cell suppression depending on the specific model (25). In contrast, catalase treatment in IMC from EAE model did not reverse suppression (data not shown). In general, these differences may suggest that IMC in EAE model may represent a different developmental stage comparing with MDSC in tumor model, perhaps due to the Th1-biased environment in EAE model.

We propose that targeted manipulation of IMC differentiation and activation in vivo may help to control autoimmune responses. It has been proposed that macrophages may be activated in at least three different ways. Classical activation of macrophages with IFN-γ either alone or with LPS, induces M1 cells. Alternative activation of macrophages with IL-4 and IL-13 induces M2 cells, which are associated with high arginase activity, increased production of IL-1 receptor antagonist and IL-10, but reduced production of proinflammatory cytokines (56, 57, 58). In addition, macrophages can be inactivated by culture with IL-10 and TGF-β, and this is associated with decreased MHC class II expression and increased production of anti-inflammatory cytokines (59, 60, 61). These different activation pathways are likely associated with different transcriptional factors similar to what occurs in different helper T cell lineages. Our data show that NOS2 and arginase 1 expression in IMC can be differentially regulated by IFN-γ and IL-4 respectively. IMC receive IFN-γ signaling in the vicinity of activated T cells, and thus they may function as a Th1-specific negative feedback mechanism to reduce the frequency of encephalitogenic effector/memory T cells. In contrast, it is possible to use IL-4, IL-10, and TGF-β to antagonize the NOS2 induction in CNS when NO over-production is detrimental to CNS homeostasis. Prevention of IMC differentiation into mature APCs in the CNS may also have therapeutic potential.

In summary, our study defines a specific population of innate immune cells that regulates activated T cells in an autoimmune disease model. Further characterization of their generation, migration, differentiation, as well as their in vivo function may have clinical implications for human autoimmune diseases.

We thank Drs. Lester Kobzik (Department of Pathology, Brigham and Women’s Hospital, Boston, MA), and Jeremy Duffield (Department of Medicine, Brigham and Women’s Hospital, Boston, MA) for helpful discussions. We also thank Drs. Li Yang, Yue Wang, and Pia Kivisäkk (Center for Neurologic Diseases, Brigham and Women’s Hospital, Boston, MA) for experimental assistance.

The authors have no financial conflict of interest.