CSF-1–Dependent Red Pulp Macrophages Regulate CD4 T Cell Responses

Macrophages (MΦs) play a critical role in innate and acquired immunity and can contribute proinflammatory or anti-inflammatory responses (1). Recently, some resident MΦs populations possessing regulatory function were discovered in the airway interstitium and intestinal lamina propria (2, 3), indicating that these suppressive MΦs play important roles in the maintenance of immune homeostasis. However, in the steady state, the role of resident splenic MΦs in T cell immune responses is poorly understood.

Once the circulating monocytes (M_o) enter various tissue compartments, they exhibit a high degree of heterogeneity and acquire specific functions (4, 5). Spleen is known to contain various MΦ subsets, such as white pulp tingible-body MΦs, marginal zone-residing MΦs, and red pulp MΦs (RPMs) (4–6). RPMs remove self-Ags, such as blood-borne particulate matter and aged erythrocytes, from the blood (6). Although T cell responses usually occur in the T cell zone of the white pulp, it is well known that numerous small patches within red pulp parenchyma (making up a total volume comparable to that in the white pulp in humans) contain mainly T cells, B cells, and RPMs (6–8), indicating the possibility of interaction between RPMs and T cells in the red pulp. Therefore, T cell responses occurring in the red pulp should be strictly regulated, because the majority of Ags are derived from self.

CSF-1, also known as M-CSF, is constitutively produced by several types of cells, including fibroblasts, endothelial cells, stromal cells, and MΦs (9, 10). The major function of CSF-1 is to elicit the differentiation and development of various tissue-resident MΦs (9, 10). CSF-1–deficient mice, which are also known as op/op mutant mice, have a residual MΦ population (CSF-1–independent MΦs). Previous studies showed that op/op mice were capable of inducing normal humoral and cellular immunity, but they contained lower levels of TNF-α and G-CSF following immunization with sheep RBCs, suggesting that the CSF-1–independent MΦ population is primarily responsible for the classical APC population, whereas the CSF-1–dependent MΦ population has the potential to regulate immune response (11, 12). Consistent with this notion, a recent report showed that in a model of graft-versus-host disease, mice in which CSF-1–dependent MΦs were depleted using an Ab against CSF-1R had a decreased number of Foxp3⁺ regulatory T cells (Tregs), which resulted in accelerated pathology and exaggerated donor T cell activation (13). Although RPMs in op/op mice were reduced to ∼50% of those in the normal littermates, these mice still had F4/80⁺ RPMs, suggesting that RPMs are composed of CSF-1–dependent and -independent populations (14–16). However, the phenotypical and functional differences between CSF-1–dependent and -independent RPMs remain largely unknown.

Previous studies showed that splenic F4/80^hiMac-1^low MΦs exist in the red pulp of the spleen (3, 5, 17–20). In this study, we first isolated splenic F4/80^hiMac-1^low MΦs and analyzed how T cell responses are regulated by F4/80^hiMac-1^low MΦs in vivo and in vitro. F4/80^hiMac-1^low MΦs showed a different expression of surface molecules compared with splenic F4/80⁺Mac-1^hi M_o/MΦs, which are capable of inducing strong T cell immune responses. Purified F4/80^hiMac-1^low MΦs produced suppressive cytokines, such as TGF-β and IL-10, and suppressed CD4⁺ T cell proliferation in vitro. Moreover, F4/80^hiMac-1^low MΦs induced the differentiation of naive CD4⁺ T cells into Foxp3⁺ Tregs via a TGF-β–dependent mechanism. Additionally, the differentiation of F4/80^hiMac-1^low MΦs was highly regulated by CSF-1, because F4/80^hiMac-1^low MΦs were not present in op/op mice. In vitro-generated bone marrow-derived MΦs induced by CSF-1 (M-MΦs) showed an expression pattern of surface molecules similar to F4/80^hiMac-1^low MΦs. Furthermore, M-MΦs suppressed CD4⁺ T cell proliferation and induced the generation of Foxp3⁺ Tregs in vitro and in vivo. These data suggested that CSF-1–dependent F4/80^hiMac-1^low MΦs are a subpopulation of RPMs and that they regulate T cell immune responses by several distinct mechanisms and maintain peripheral immune homeostasis.

C57BL/6 and BALB/c male mice were purchased from Japan SLC (Shizuoka, Japan). C57BL/6-Tg (Tcra2D2,Tcrb2D2)1Kuch/J mice (transgenic for the TCR Vα3.2 and Vβ11 chains reactive to myelin oligodendrocyte glycoprotein [MOG]_35–55), commonly known as 2D2 mice (21), and B6;C3Fe a/a-Csf1^op/J mice, heterozygous for the osteopetrotic mutation (op/wt) (14), were purchased from Jackson Laboratory. Foxp3-EGFP knockin mice (BALB/c background), produced by a standard method based on a previous report (22), were obtained from Kyoto University. All mice used were 7–9 wk old and were maintained under specific pathogen-free conditions in our animal facility and were studied using a protocol approved by the Hokkaido University Committee for Animal Use and Care.

To obtain F4/80^hiMac-1^low MΦs and F4/80⁺Mac-1^hi M_o/MΦs, splenocytes were prepared by collagenase D treatment (Roche) and incubated with biotinylated anti-mouse α9 integrin mAb for 20 min. After washing, IMag Streptavidin Particles (BD Bioscience) were added to the cell suspension. The tube containing this labeled cell suspension was placed within the magnetic field of a BD IMagnet. The α9 integrin⁺ fraction contained F4/80^hiMac-1^lowα9⁺ MΦs with ∼80% purity. Because the α9 integrin⁻ fraction contained many non-M_o/MΦ cells, Mac-1⁺ cells were enriched using mouse CD11b (Mac-1) MicroBeads (Miltenyi Biotec), according to the manufacturer’s protocol (these cells are Mac-1⁺α9⁻ cells). Next, these two populations were stained with anti-F4/80 and Mac-1 mAbs, and the F4/80^hiMac-1^lowα9⁺ (F4/80^hiMac-1^low MΦs) and F4/80⁺Mac-1^hiα9⁻side scatter (SSC)^low (F4/80⁺Mac-1^hiM_o/MΦs) cells were purified using a FACS Vantage SE (BD Bioscience). F4/80⁺Mac-1^hi cells are composed of SSC^low M_o/MΦs and SSC^hi eosinophils (18). Because the two populations cannot be distinguished only by the expression of F4/80 and Mac-1, we excluded SSC^hi eosinophils from F4/80⁺Mac-1^hi M_o/MΦs using an SSC parameter in all experiments. Naive CD4⁺ T cells were isolated from 2D2 mice with anti-CD4 microbeads (Miltenyi Biotec), according to the manufacturer’s protocol, and then were stained with anti-CD4, CD62L, or CD25 mAbs. The CD4⁺CD62L^hi (2D2-CD4) cells were sorted using a FACSVantage SE. In some experiments, we sorted the CD4⁺CD25⁻ cells from the spleen of 2D2 mice. Purity of all sorted populations was >95%. The splenic APCs that we used were obtained using a CD4-microbeads negative selection system, and the cells were irradiated by x-ray (3000 rad).

Isolated MΦs were adjusted to 1.5 or 2.0 × 10⁶ cells/ml in TIL media (IBL), supplemented with 10% FCS and cultured in 96-well flat-bottom dishes at 37°C at 5% CO₂ for 48 or 72 h. The cultured supernatants were used for ELISA (48 h) and T cell-stimulation experiments (72 h). For TLR stimulation of MΦs, purified F4/80^hiMac-1^low MΦs and F4/80⁺Mac-1^hi M_o/MΦs were adjusted to 1.0 × 10⁶ cells/ml in TIL media, supplemented with 10% FCS and LPS (1 μg/ml; IBL) or CpG-B dinucleotides (ODN1668, 3 μg/ml; system science) and cultured for 48 h. The culture supernatant was harvested and used for ELISA. To generate M-MΦs or bone marrow-derived dendritic cells (GMDCs), bone marrow cells were isolated from the femurs and tibias of mice and cultured in RPMI 1640 media, supplemented with 10% FCS, 100 U/ml penicillin, and 100 μg/ml streptomycin in the presence of 20 ng/ml CSF-1 or GM-CSF (WAKO). After 3 d, the adherent cells were collected and cultured for 2 more days in the presence of fresh CSF-1 or GM-CSF (20 ng/ml) and then the cells were collected by cell scrapers and used in subsequent experiments. For the stimulation of lymphocytes, APCs, such as purified F4/80^hiMac-1^low MΦs, F4/80⁺Mac-1^hi M_o/MΦs, CD4 T cell-deleted splenocytes as positive control, M-MΦs, or GMDCs, were cultured with 2D2-CD4 T cells (1 × 10⁵) and MOG₃₅₋₅₅ peptide (MEVGWYRSPFSRVVHLYRNGK: 20 μg/ml) in 96-well round-bottom plates. For measuring the proliferation, [³H]methyl thymidine was added to cells during the last 18 h of culture. In some studies, anti–TGF-β mAb (1D11 from R&D Systems), isotype control Ab, or human rTGF-β (PeproTech) was included in the cultures at the indicated concentration. In other experiments, splenic M_o/MΦs were cultured with 2D2 naive CD4⁺CD62L^hi T cells and MOG_35–55 peptide in the absence or presence of F4/80^hiMac-1^low MΦs or culture supernatant (100 μl). For suppression assays, sorted CD4⁺EGFP⁺ Tregs were cultured with 5 × 10⁴ freshly isolated CD4⁺CD25⁻ naive T cells, 5 × 10⁴ irradiated APCs, and 0.5 μg/ml anti-CD3 mAb (Biolegend). Cells were cultured for 96 h, and the T cell proliferation was determined as described above.

For in vivo CD4⁺ T cell priming, 5 × 10⁶ MACS-sorted naive 2D2-CD4⁺ T cells were injected i.v. into normal C57BL/6 recipients. In some experiments (Foxp3-expression assay), 2 × 10⁶ FACS-sorted CD4⁺CD25-(2D2) T cells were injected into the C57BL/6 recipients. At 24, 48, and 72 h after T cell transfer, the recipient mice received 0.5–1 × 10⁶ M-MΦs or GMDCs pulsed with MOG_35–55 peptide. These cells were prepared as described above and pulsed with MOG_35–55 peptide (50 μg/ml) for 2.5 h at 37°C before transfer. At 7 d following T cell transfer, the spleens were removed, and Ag-specific CD4⁺ T cells were detected by flow cytometry (FCM) using TCR Vα3.2 and Vβ11 mAbs. In another experiment, 2 × 10⁶ MACS-sorted naive 2D2-CD4⁺ T cells were injected i.v. into normal C57BL/6 recipients. On days 1–3 after T cell transfer, recipient mice received 0.5–1 × 10⁶ M-MΦs pulsed with MOG_35–55 peptide, followed by GMDC injection (1 × 10⁶) on days 7–9. Fourteen days after T cell transfer, the spleens were removed, and the numbers of Ag-specific CD4⁺ T cells were analyzed by FCM.

mAbs to α9 integrin (18R18D) (23) and CCR2 (clone MC21) (24) were used. Anti-mouse F4/80 (Cl:A3-1) and Dectin-2 (D2.11E4) mAbs were purchased from Serotec. Anti-mouse CD8α (53-6.7), CD11b (M1/70), CD49d (R1-2), CD62L (MEL-14), and ICAM-1 (3E2) mAbs were purchased from BD Bioscience. Anti-mouse MHC class I (28-14-8), MHC class II (M5/114.15.2), and TLR9 (M9.D6) mAbs were purchased from eBioscience. Anti-mouse CD44 (IM7), CD80 (16-10A1), CD86 (GL-1), CD11a (M17/4), VCAM-1 (429), TLR-4 (MTS510), and Foxp3 (150D) mAbs were purchased from Biolegend (San Diego, CA). For intracellular staining, cells were incubated for 20 min at 4°C in Cytofix/Cytoperm solution (BD Bioscience) and were stained with anti-Foxp3, TLR9, and isotype-control mAbs. Splenic cells were evaluated with FACSCalibur and FACSCanto II flow cytometer (BD Biosciences) and analyzed with FlowJo software (Tree Star).

mAbs used for immunohistochemistry (IHC) included anti mouse-F4/80 (Cl:A3-1) from Biolegend (San Diego, CA), anti-CD169 (MOMA-1) from Serotec, and anti-mouse CD3ε (145-2C11), CD11c (HL3), and B220 (RA3-6B2) from BD Bioscience. Goat anti-Syrian hamster IgG-biotin and goat anti-rat IgG-biotin for secondary Abs were purchased from Jackson ImmunoResearch Laboratories. Spleens were embedded in OCT compound (SAKURA), and 10-μm sections were prepared. The sections were fixed in acetone for 7 min, washed three times with PBS, and blocked with 10% normal goat serum for 30 min, followed by avidin/biotin blocking (Nichirei) for 10 min. After washing three times, the sections were stained with primary Abs for 1 h. Endogenous peroxidase was depleted by incubating sections for 5 min in 0.6% H₂O₂ in PBS. Bound Abs were detected with biotin-conjugated goat secondary Abs. Staining was amplified using a Vectastain-ABC kit (Vector Laboratories), according to the manufacturer’s recommendations, followed by diaminobenzidine (DAKO) visualization of peroxidase activity.

Total RNA was isolated using TRIzol reagent (Invitrogen). Reverse transcription was performed on 5 μg total RNA using the Transcriptor First Strand cDNA Synthesis Kit (Roche Diagnostics), according to the manufacturer’s directions. From these cDNA pools, specific targets amplified by PCR were quantified with a Lightcycler FastStart DNA Master SYBR Green I kit (Roche Diagnostics). The results were normalized to G3PDH. The following primers were used: G3PDH, 5′-ACC ACA GTC CAT GCC ATC AC-3′ (sense) and 5′-TCC ACC ACC CTG TTG CTG TA-3′ (antisense); Foxp3, 5′-TTC ATG CAT CAG CTC TCC AC-3′ (sense) and 5′-CTG GAC ACC CAT TCC AGA CT-3′ (antisense); and TGF-β2, 5′-CCA CCT CCC CTC CGA AAA-3′ (sense) and 5′-AGACATCAAAGCGGACGATTCT-3′ (antisense).

For phagocytosis assays, fluorescein-conjugated zymosan A (Saccharomyces cerevisiae) (Molecular Probes) was opsonized by fresh mouse serum at 37°C for 1 h. Isolated MΦs (1 × 10⁵ cells) were plated onto 96-well plates and adhered for 1 h at 37°C. Then, opsonized zymosans were added to wells at a concentration of 20 μg/ml and incubated for 3 h at 37°C, and nonphagocytosed particles were washed out with PBS containing EDTA and trypsin. The cells were then detached and analyzed by FCM (25). Surface-attached zymosans were quenched by trypan blue solution. FITC-labeled zymosan (400 μg) was injected i.v. in normal mice for in vivo phagocytosis experiments. The animals were killed at 27 h following the i.v. injection of FITC-labeled zymosan, and the splenocytes were recovered. Then, the cells were stained with anti-F4/80 and Mac-1 mAbs and analyzed by FCM. The uptake was measured by cellular fluorescence.

Clodronate was a gift of Roche Diagnostics (Mannheim, Germany). C57BL/6 mice received i.v. administration of 0.2 ml clodronate liposomes (clod-lip) and were killed at 1 or 10 d thereafter. Spleens were removed and processed.

Cytokine levels in supernatants of in vitro-cultured cells were measured by ELISA (IL-1α, IL-2, IL-4, IL-6, IL-10, IL-12p40, IL-12p70, IFN-γ, and TGF-β from BD Biosciences). ELISA assays were performed according to the manufacturer’s protocol.

A statistical comparison between the groups was performed with a two-tailed paired Student t test. Data were considered statistically significant at p < 0.05.

It was reported that splenic F4/80^hiMac-1^low cells are MΦs that are located in the red pulp (3, 5, 17–20). We performed FCM analysis to examine the role of F4/80^hiMac-1^low MΦs on T cell responses. Consistent with the previous reports, we detected F4/80^hiMac-1^low MΦs and F4/80⁺Mac-1^hi cells in the spleen (Fig. 1A). It is well known that F4/80⁺Mac-1^hi cells are a splenic M_o (17–19) and MΦ (3, 26) population that can produce inflammatory cytokines and have a strong Ag-presenting ability (3, 26). Freshly isolated splenic F4/80⁺Mac-1^hi MΦs from mice that were infected with Listeria monocytogenes induced strong Ag-specific CD4⁺ and CD8⁺ T cell responses in vitro, indicating that these cells possess a strong Ag-priming activity in vivo (27–29). First, we examined the expression of surface and intracellular molecules on F4/80^hiMac-1^low MΦs. The expression levels of classes I and II MHC molecules, costimulatory molecules (CD80 and CD86), TLRs (TLR4 and TLR9), α4 integrin, and ICAM-1 were very similar between F4/80^hiMac-1^low MΦs and F4/80⁺Mac-1^hi M_o/MΦs. F4/80⁺Mac-1^hiM_o/MΦs, but not F4/80^hiMac-1^low MΦs, expressed LFA-1 and inflammatory chemokine receptor CCR2 at high levels, as previously reported (3). In contrast, a majority of F4/80^hiMac-1^low MΦs expressed VCAM-1 (20), dectin-2 (18), and α9 integrin with autofluorescence, whereas F4/80⁺Mac-1^hi M_o/MΦs did not (Fig. 1B, Supplemental Fig. 1).

Because we detected α9 integrin expression on F4/80^hiMac-1^low MΦs by FCM analysis, we performed IHC using a mAb for α9 integrin to confirm whether α9 integrin can be useful as a marker for F4/80^hiMac-1^low MΦs. As expected, the staining of α9 integrin was restricted to red pulp cells, and the staining pattern of α9 integrin was very similar to that of F4/80, with a typical RPM pattern (Supplemental Fig. 2), indicating that α9 integrin can be useful as a marker for F4/80^hiMac-1^low MΦs in the red pulp.

A previous report demonstrated that F4/80⁺ RPMs are required for clearance of RBCs (6). To examine the activity of phagocytosis by F4/80^hiMac-1^low MΦs in vitro, we purified F4/80^hiMac-1^low MΦs and F4/80⁺Mac-1^hi M_o/MΦs from the spleen using anti-α9 integrin, F4/80, and Mac-1 mAbs and incubated them with opsonized FITC-labeled zymosan. Although F4/80^hiMac-1^low MΦ and F4/80⁺Mac-1^hi M_o/MΦ populations phagocytosed FITC-labeled zymosan, the phagocytic activity of F4/80^hiMac-1^low MΦs was stronger than that of F4/80⁺Mac-1^hi M_o/MΦs, as evidenced by FITC⁺ cells (Supplemental Fig. 3A). Furthermore, to examine the phagocytic activity of F4/80^hiMac-1^low MΦs in vivo, we injected FITC-labeled zymosan into mice. Twenty-seven hours later, FITC-fluorescence was detectable in F4/80^hiMac-1^low MΦs and F4/80⁺Mac-1^hi M_o/MΦs (Supplemental Fig. 3B). These results suggested that F4/80^hiMac-1^low MΦs have a strong phagocytic activity in vitro and in vivo.

Recent evidence points to a central role of APCs in fine-tuning the quality of the immune response by their cytokine secretion (30). Therefore, we evaluated the ability of cytokine expression in F4/80^hiMac-1^low MΦs. Freshly isolated splenic F4/80⁺Mac-1^hi M_o/MΦs spontaneously produced IL-6 and IL-12p40 in vitro, consistent with previous reports (3, 26). In contrast, F4/80^hiMac-1^low MΦs spontaneously produced IL-1α, IL-12p70, and, importantly, TGF-β1 (Fig. 2A). Additionally, constitutive TGF-β2 mRNA expression was detected in F4/80^hiMac-1^low MΦs at high levels (Supplemental Fig. 4).

It was shown that F4/80⁺ RPMs and marginal zone MΦs secreted TGF-β following injection of zymosan, which is a stimulus for TLR2/6 (27). Therefore, we next evaluated the cytokine profiles of purified F4/80^hiMac-1^low MΦs after stimulation with LPS (TLR4 ligand) and CpG-B (TLR9 ligand). F4/80^hiMac-1^low MΦs showed marked production of IL-1α and IL-10 in response to LPS and CpG-B stimulation (Fig. 2B). In contrast, F4/80⁺Mac-1^hi M_o/MΦs produced IL-12p40 in response to CpG-B stimulation and IL-6 in response to LPS and CpG-B stimulation, as expected (3, 26). Importantly, F4/80^hiMac-1^low MΦs constitutively produced TGF-β1, and the production of TGF-β1 was not altered, even after LPS and CpG-B stimulation (Fig. 2B). These data indicated that F4/80^hiMac-1^low MΦs produced suppressive cytokines, such as TGF-β and IL-10.

Because of the data showing that F4/80^hiMac-1^low MΦs had phagocytic activity and produced anti-inflammatory cytokines, such as TGF-β and IL-10, we further investigated whether F4/80^hiMac-1^low MΦs can regulate T cell responses. To address this question, we used naive CD4⁺ T cells from 2D2-TCR transgenic mice, expressing transgenic TCR specific for MOG_35–55 epitope (21). Naive 2D2-CD4⁺ T cells were cultured with F4/80^hiMac-1^low MΦs or F4/80⁺Mac-1^hi M_o/MΦs in the presence of MOG_35–55 peptide; T cell proliferation and IL-2 production were assessed. As previously reported (3), F4/80⁺Mac-1^hi M_o/MΦs induced strong T cell proliferation, which was comparable to that induced by splenic APCs containing dendritic cells, and showed strong IL-2 production (Fig. 3A). 2D2-CD4⁺ T cells activated by F4/80⁺Mac-1^hi M_o/MΦs also produced large amounts of IFN-γ and IL-4 (Fig. 3B). In sharp contrast, F4/80^hiMac-1^low MΦs failed to induce the proliferation of 2D2-CD4⁺ T cells, consistent with small amounts of IL-2 production and undetectable levels of IFN-γ and IL-4 production (Fig. 3A, 3B). It should be pointed out that F4/80^hiMac-1^low MΦs are not simply incapable of presenting Ag, because a majority of 2D2-CD4⁺ T cells cultured with F4/80^hiMac-1^low MΦs showed a CD44^hi phenotype (Fig. 3C). Next, we tried to understand the reason why F4/80^hiMac-1^low MΦs were unable to induce Ag-specific CD4⁺ T cell proliferation. Because F4/80^hiMac-1^low MΦs spontaneously produced a large amount of TGF-β (Fig. 2, Supplemental Fig. 4), we hypothesized that TGF-β may be involved in poor T cell proliferation. Indeed, T cell proliferation induced by F4/80^hiMac-1^low MΦs was partially, but significantly, rescued when TGF-β was neutralized. Meanwhile, T cell proliferation induced by F4/80⁺Mac-1^hi M_o/MΦs was not affected by the anti–TGF-β Ab (Fig. 3D). Because F4/80^hiMac-1^low MΦs also secreted suppressive cytokine IL-10 (Fig. 2), we tested the suppressive role of IL-10. We found that the T cell proliferation was partially, but significantly, rescued by anti–IL-10R mAb treatment (Supplemental Fig. 5), indicating the additional involvement of IL-10 in suppression of the T cell response. We next examined whether F4/80^hiMac-1^low MΦs have a suppressive effect on T cell proliferation induced by F4/80⁺Mac-1^hi M_o/MΦs. F4/80^hiMac-1^low MΦs were capable of inhibiting the proliferation of 2D2-CD4⁺ T cells induced by F4/80⁺Mac-1^hi M_o/MΦs in a cell concentration-dependent manner (Fig. 3E). We then cultured naive 2D2-CD4⁺ T cells with Ag-pulsed F4/80⁺Mac-1^hi M_o/MΦs plus culture supernatant from nonpulsed F4/80^hiMac-1^low MΦs. We observed that the culture supernatant from nonpulsed F4/80^hiMac-1^low MΦs markedly inhibited the 2D2-CD4⁺ T cell proliferation induced by Ag-pulsed F4/80⁺Mac-1^hi M_o/MΦs (Supplemental Fig. 6). These data suggested that F4/80^hiMac-1^low MΦs are capable of suppressing CD4⁺ T cell proliferation by producing soluble suppressive factors, including TGF-β and IL-10, without Ag presentation.

It was demonstrated that, upon Ag stimulation in the presence of TGF-β, peripheral naive CD4⁺ T cells are forced to differentiate into Tregs expressing Foxp3 (31). To test whether F4/80^hiMac-1^low MΦs are capable of inducing the differentiation of naive CD4⁺ T cells into Tregs, we cultured naive 2D2-CD4⁺ T cells with F4/80^hiMac-1^low MΦs pulsed with MOG peptide and analyzed Foxp3 expression in 2D2-CD4⁺ T cells by RT-PCR and FCM. Indeed, F4/80^hiMac-1^low MΦs, but not F4/80⁺Mac-1^hi M_o/MΦs, induced Foxp3 expression in 2D2-CD4⁺ T cells (Fig. 4A, 4B). As shown in Fig. 4C, Foxp3 expression in 2D2-CD4⁺ T cells induced by F4/80^hiMac-1^low MΦs was inhibited when anti–TGF-β Ab was added to the culture. Furthermore, when rTGF-β1 was exogenously added to the cultures, F4/80^hiMac-1^low MΦs enhanced the generation of Foxp3⁺ T cells from naive CD4⁺ T cells. In contrast, the generation of the Foxp3⁺ T cell population by F4/80⁺Mac-1^hi M_o/MΦs was significantly less than that by F4/80^hiMac-1^low MΦs, even after rTGF-β1 was added (Fig. 4D). However, it remains to be clarified whether Foxp3⁺ Tregs generated by F4/80^hiMac-1^low MΦs have suppressive functions. Because live Foxp3⁺ cells cannot be isolated based on Foxp3 expression in the absence of a reporter protein, we could not examine the suppressive function of the Foxp3⁺CD4⁺ T cells generated using the experimental setup described above. Therefore, to test the ability of Foxp3⁺ T cells induced by F4/80^hiMac-1^low MΦs to suppress the proliferation of naive CD4⁺ T cells, EGFP⁻CD4⁺ T cells were isolated from Foxp3-EGFP mice and were activated in the presence of F4/80^hiMac-1^low MΦs and anti-CD3 mAb. F4/80^hiMac-1^low MΦs, but not F4/80⁺Mac-1^hi M_o/MΦs, led to the generation of Foxp3⁺CD4⁺ T cells (Supplemental Fig. 7A). Then, we sorted the Foxp3⁺CD4⁺ population and used it to examine the suppressive activity of Tregs induced by F4/80^hiMac-1^low MΦs (Supplemental Fig. 7B). We found that induced Foxp3⁺CD4⁺ T cells potently inhibited naive CD4⁺ T cell proliferation (Supplemental Fig. 7C). Taken together, F4/80^hiMac-1^low MΦs have the ability to induce the generation of functional Foxp3⁺ Tregs from naive CD4⁺T cells.

It was reported that op/op mice exhibited ∼50% reduction in F4/80⁺ RPMs, as evaluated by IHC (14–16). We used op mutant mice to examine the necessity of CSF-1 for F4/80^hiMac-1^low MΦ development. F4/80^hiMac-1^low MΦ proportion and numbers were substantially reduced in op mutant mice, depending on their genotypes (Fig. 5A). In contrast, F4/80⁺Mac-1^hi M_o/MΦ proportion was unaffected, although their absolute numbers were slightly reduced in op mutant mice, as previously reported (Fig. 5A) (9). These results clearly indicated that the generation and development of splenic F4/80^hiMac-1^low MΦs are dependent on CSF-1. To understand an additional aspect of F4/80^hiMac-1^low MΦ development, we injected clod-lip i.v. into normal mice. As expected, this completely eliminated F4/80⁺ RPMs by day 1, as demonstrated by IHC (Fig. 5B) (16). In addition, by using FCM, we found that F4/80^hiMac-1^low MΦ and F4/80⁺Mac-1^hi M_o/MΦ populations were almost completely depleted compared with those in mice treated with PBS-encapsulated liposomes. At day 10, we detected the substantial, but not complete, repopulation of F4/80⁺ RPMs by using IHC, despite the fact that the F4/80^hiMac-1^low MΦ population still was not detectable by FCM (Fig. 5B). In contrast, the F4/80⁺Mac-1^hi M_o/MΦ population reappeared at day 10 using FCM, again indicating that the appearance of two splenic MΦ and M_o/MΦ populations is differentially regulated. Thus, our present data clearly demonstrated that F4/80^hiMac-1^low MΦs, previously recognized as RPMs (3, 5, 17–20), are a subpopulation of RPMs, and their development is dependent on CSF-1.

To confirm the above notion that generation of F4/80^hiMac-1^low MΦs is CSF-1 dependent, we examined the characteristics of M-MΦs and generated GMDCs as a positive control for Ag presentation (10). The majority of M-MΦs was F4/80^hiMac-1⁺, α9 integrin⁺, but CCR2^low, which is very similar to the phenotype found in splenic F4/80^hiMac-1^low MΦs. In contrast, GMDCs were F4/80⁺Mac-1^hi and expressed CCR2 (10, 32) but not α9 integrin (Fig. 6A). We next analyzed the function of M-MΦs in vitro. M-MΦs failed to elicit a CD4⁺ T cell-proliferative response (Fig. 6B), whereas GMDCs induced a strong proliferative response, as previously reported (10). In addition, M-MΦs inhibited the proliferation of CD4⁺ T cells induced by GMDCs (Fig. 6C). Similar to the findings in splenic F4/80^hiMac-1^low MΦs (Fig. 3C), M-MΦs were able to present Ag to 2D2-CD4⁺ T cells, because CD44 expression was upregulated in 2D2-CD4⁺ T cells (Fig. 6D). Furthermore, M-MΦs, but not GMDCs, induced the expression of Foxp3 in CD4⁺ T cells via a mechanism partially dependent on TGF-β (Fig. 6E, Supplemental Fig. 8), consistent with previous reports on human M_o-derived MΦs induced by CSF-1 (33). Thus, in vitro-generated M-MΦs were phenotypically and functionally similar to splenic F4/80^hiMac-1^low MΦs.

We then decided to use M-MΦs to analyze whether the in vitro suppressive effect of splenic F4/80^hiMac-1^low MΦs on the CD4⁺ T cell response could also operate in vivo. To address this question, normal C57BL/6 recipients were transferred with naive 2D2-CD4⁺ T cells, followed by M-MΦs or GMDCs pulsed with MOG peptide. On day 7 after T cell transfer, the absolute number of 2D2-CD4⁺ T cells in the spleen was assessed (Fig. 6F). We found that Ag-specific T cell proliferation was not induced in recipients treated with M-MΦs, because the number of 2D2-CD4⁺ T cells determined by TCR Vα3.2 and Vβ11 expression was comparable to that in recipient mice treated with PBS (Fig. 6G). The inability of M-MΦs to induce in vivo T cell proliferation was not due to the absence of function for Ag presentation, because a significant proportion of 2D2-CD4⁺ T cells in recipient mice treated with M-MΦs showed the CD44^hi phenotype (Fig. 6H). In contrast, 2D2-CD4⁺ T cells transferred into recipient mice treated with GMDCs showed robust expansion and the CD44^hi phenotype (Fig. 6G, 6H). Furthermore, 2D2-CD4⁺ T cells primed by M-MΦs in vivo expressed Foxp3, but T cells primed by GMDCs did not (Fig. 6I). Finally, we examined whether the difference in the proportion of Foxp3⁺ T cells between M-MΦ– and PBS-treated mice correlated with the in vivo immune-inhibitory activity when rechallenged with GMDCs (Supplemental Fig. 9A). We found that PBS-primed mice, which contained smaller proportions of Foxp3⁺ T cells, exhibited significant expansion of Ag-specific CD4⁺ T cells. In contrast, Ag-pulsed M-MΦ–primed mice, which contained a greater proportion of Foxp3⁺ cells, exhibited little, if any, expansion of Ag-specific T cells (Supplemental Fig. 9B). Collectively, these data strongly supported that splenic F4/80^hiMac-1^low MΦs are able to regulate immune responses, and their generation is regulated by CSF-1.

In the spleen, the blood vascular system is designed to allow most of the blood to flow directly into the red pulp, and RPMs actively phagocytose and remove copious amount of aging and injured RBCs and blood-borne particulates from the blood (6, 34). However, many scattered T cells exist in the red pulp (Supplemental Fig. 10) (7, 8). Therefore, T cell responses against such self-Ags must be regulated to avoid harmful autoimmune responses. In this regard, it was reported that the injection of yeast zymosan with specific Ag into mice resulted in Ag-specific T cell tolerance (30). Although this tolerance induction is partially explained by TGF-β produced by zymosan-stimulated RPMs, detailed analyses of RPMs had not been performed. In this study, we analyzed how splenic F4/80^hiMac-1^low MΦs residing in red pulp regulate T cell responses, and found the potential importance of splenic F4/80^hiMac-1^low MΦs in the regulation of peripheral immune homeostasis.

Although it was reported that splenic F4/80^hiMac-1^low cells are MΦs that are found in red pulp (5, 17, 18, 20), there was no detailed report on the immunological phenotype and function of these cells. In this study, we demonstrated the surface profile for F4/80^hiMac-1^low MΦs; they expressed MHC molecules, costimulatory molecules, and TLRs. F4/80^hiMac-1^low MΦs also expressed a relatively minor integrin (α9 subunit). Splenic F4/80⁺Mac-1^hi M_o/MΦs expressed high levels of chemokine receptor CCR2 (35), which is a dominant receptor for MCP-1 and a key chemokine receptor that regulates migration and infiltration of M_o and MΦs (24). In contrast, F4/80^hiMac-1^low MΦs did not express CCR2, suggesting that following the capture of Ags, they remain inside the red pulp rather than migrating to inflammatory sites and participating in inflammation.

Earlier IHC studies showed that the number of F4/80⁺ RPMs was reduced by half in op/op mice compared with normal littermates, suggesting that RPMs are composed of CSF-1–dependent and -independent populations (14, 15). In this study, we found that the generation and differentiation of F4/80^hiMac-1^low MΦs were critically regulated by CSF-1, and F4/80^hiMac-1^low MΦs are a subpopulation of RPMs. Other resident MΦ populations, including muscle, kidney, and synovium MΦs, are reduced in op/op mice (10, 14), indicating that factors other than CSF-1 contribute to the generation of splenic F4/80^hiMac-1^low MΦs in red pulp, as well as other tissue-resident MΦs. Interestingly, a recent study demonstrated that mice deficient in Spi-C, belonging to the Ets transcription factor family, specifically lacked splenic F4/80⁺ RPMs (20), which indicated that Spi-C might be critical for the development of CSF-1–dependent and -independent RPMs. In addition, M-MΦs and splenic F4/80^hiMac-1^low MΦs possess similar phenotypes and functions. They commonly expressed α9 integrin but not CCR2. Because freshly isolated bone marrow myeloid cells did not express α9 integrin (Supplemental Fig. 11), the expression of α9 integrin by splenic F4/80^hiMac-1^low MΦs may be regulated by CSF-1 during their development. It was reported that Spi-C expression in RPMs induced the expression of VCAM-1 (20), which is a ligand for α9 integrin (36), suggesting the possibility of an interaction between α9 integrin and VCAM-1 on F4/80^hiMac-1^low MΦs (Fig. 1B). However, we do not know whether α9 integrin-mediated signaling plays a role in the generation and function of splenic F4/80^hiMac-1^low MΦs.

In this study, we found that F4/80^hiMac-1^low MΦs are capable of suppressing CD4⁺ T cell responses by the mechanism dependent on suppressive soluble factors, because culture supernatant from F4/80^hiMac-1^low MΦs efficiently suppressed T cell proliferation in vitro. F4/80^hiMac-1^low MΦs produced large amounts of anti-inflammatory cytokines, such as TGF-β and IL-10 (IL-10 production by F4/80^hiMac-1^low MΦs was further augmented following TLR ligation); however, these cytokines had only a partial suppressive effect on the T cell response, suggesting that other suppressive factor(s) might be secreted by F4/80^hiMac-1^low MΦs. The fact that F4/80^hiMac-1^low MΦs suppressed CD4⁺ T cell proliferation induced by F4/80⁺Mac-1^hi M_o/MΦs suggested that there might be cases in which F4/80^hiMac-1^low MΦs interact with other APCs in the red pulp to regulate excessive immune responses.

F4/80^hiMac-1^low MΦs are also capable of suppressing CD4⁺ T cell responses through induction of Foxp3⁺ Tregs. It is known that the differentiation of Tregs from naive CD4⁺ T cells is required for TGF-β and stimulation from TCRs (31). The fact that splenic F4/80^hiMac-1^low MΦs and in vitro-generated M-MΦs could induce the differentiation of naive CD4⁺ T cells into functional Foxp3⁺ Tregs (Figs. 4, 6E, 6I) seems reasonable, because these MΦs have the ability to present Ag and produce a large amount of TGF-β. We also found that Foxp3⁺ Tregs induced by M-MΦs strongly suppressed Ag-specific CD4⁺ T cell proliferation induced by GMDCs in vivo (Supplemental Fig. 9B). It should be pointed out that Tregs, as well as soluble factors produced by M-MΦs, contributed to suppression of T cell proliferation in this experiment.

F4/80^hiMac-1^low MΦs induced the generation of Foxp3⁺ Tregs via a TGF-β–dependent mechanism (Fig. 4C). The contribution of Foxp3⁺ Tregs to suppression of the T cell response seemed minor in our in vitro coculture experiment (Fig. 3D), because the effect of anti–TGF-β Ab treatment was partial. However, it should be noted that it takes >72 h to generate Foxp3⁺ Tregs from naive CD4⁺ T cells following TCR and TGF-β stimulation (37, 38), and the regulatory event induced by Tregs occurs even before the first division of target T cell (39). In our in vitro coculture experiment, we cultured naive T cells and F4/80^hiMac-1^low MΦs for 3 d and analyzed T cell proliferation. Therefore, it is likely that Foxp3⁺ Tregs induced by F4/80^hiMac-1^low MΦs are not capable of contributing to the suppression of T cell proliferation because of the delay in Treg generation.

In summary, our results demonstrated that the F4/80^hiMac-1^low MΦs suppress autoimmune responses through regulating CD4⁺ T cell responses and inducing the generation of Tregs. It seems likely that F4/80^hiMac-1^low MΦ also regulate excessive immune responses by suppressing T cell responses induced by other red pulp professional APCs. The generation of F4/80^hiMac-1^low MΦs is strictly regulated by CSF-1, and in vitro-generated M-MΦs showed suppressive functions in vivo. Thus, our findings provide a new aspect for F4/80^hiMac-1^low MΦs and raise the possibility that CSF-1–induced regulatory MΦs may be useful for the treatment of autoimmune disorders.

We thank K. Danzaki and D. Ohta (Hokkaido University) for preparation of mice and tissue sections.

The authors have no financial conflicts of interest.