Recurrent episodes of inflammation underlie numerous pathologies, notably those of inflammatory bowel diseases. In this study, we describe a population of macrophages in a novel state of activation that mitigates colitis in mice. The cells responsible for this effect, called IFN-γ-stimulated monocyte-derived cells (IFNγ-MdC), derive from mouse spleen, blood, and bone marrow monocytes and are distinguished from known macrophage populations by mode of generation, cell surface phenotype, and function. IFNγ-MdC only arise when macrophages are cultivated in the presence of CD40L-expressing CD4+ T cells, M-CSF, and IFN-γ. IFNγ-MdC express markers including F4/80, CD11b/c, CD86, and CD274; they are negative for CD4, CD8, Gr1, CD19, CD80, and CD207. Functionally, IFNγ-MdC are defined by their capacity to enrich cocultured T cell populations for CD4+CD25+Foxp3+ regulatory cells; this enrichment, constituting up to 60% or more of residual lymphocytes, is attributed to an expansion, but also to a cell contact and caspase-dependent depletion of activated T cells. In mice, IFNγ-MdC delivered i.v. traffic to gut-associated peripheral lymphoid tissues, including the mesenteric lymph nodes, Peyer’s patches, and colonic mucosa, and promote the clinical and histological resolution of chronic colitis. We conclude that IFNγ-MdC represent macrophages in a novel state of activation, possessing multiple T cell-suppressive effects with therapeutic potential for the treatment of autoimmune inflammation.
Macrophages exhibit great phenotypic variation and engage in diverse, often antagonistic processes. For instance, both immunogenic and immunosuppressive functions are ascribed to macrophages, and macrophages participate in both tissue-destructive and reparative processes (1). The extent to which this heterogeneity reflects the development of distinct, differentiated macrophage lineages as opposed to inducible, interchangeable states of macrophage activation is unresolved (2). Nevertheless, it is convenient to categorize macrophage subsets according to static functional and phenotypic criteria (3). In particular, classically activated (M1-polarized) macrophages are monocytes treated with IFN-γ alone or in combination with microbial components, which elaborate IL-12 and IL-23, promoting Th1-type T cell responses. By contrast, alternatively activated (M2-polarized) macrophages form when monocytes are exposed to IL-4 or IL-13 (1, 4, 5); alternatively activated macrophages produce IL-10, but little IL-12 and IL-23, and express high levels of scavenger and mannose receptors, and favor Th2-polarized T cell responses.
Owing, perhaps, to a historical emphasis on dendritic cell biology and the lack of appreciation for the subtle heterogeneity of macrophage populations, reports of T cell-suppressive activities mediated by macrophages have not received due attention. Descriptions of T cell-suppressive macrophage populations in the literature are often conflicting, indicating that immunosuppressive activity is not a property of a unique macrophage subset, much as “tolerogenicity” is not the property of any single dendritic cell population (6). Hence, subsets of both classically and alternatively activated macrophages have been attributed tolerogenic capabilities (4, 5, 7). Especially well characterized is the human M-CSF and CD40 ligation-induced, IDO-expressing macrophage first described by Munn et al. (8, 9). These classically activated macrophages influence T cell function through IDO-dependent mechanisms and, although there is no definite evidence that these macrophages represent a physiological entity, it has been speculated that they are arbiters of peripheral tolerance. Currently, there is a renewed interest in studying the suppressive activities of specific macrophage populations and better classifying these populations for scientific and therapeutic purposes.
Recently, work undertaken in our laboratories has led to the identification of a hitherto unrecognized macrophage derivative, the IFN-γ-stimulated monocyte-derived cell (IFNγ-MdC),3 which suppresses T cell proliferation in vitro (10) and dampens acute rejection responses against allogeneic, ectopic heart transplants in mice (S. Inoue et al., abstract no. 1042, as described for the American Transplant Congress 2006, Boston, MA). In our study, an initial description of these cells is given, confirming their separate identity from other immunosuppressive macrophage populations. Because autoimmune chronic colitis has been a long-standing interest of our research group, and with the observation that IFNγ-MdC are capable of migrating to gut-associated lymphoid tissues, the effect of IFNγ-MdC on the course of CD4+CD62L+ T cell and dextran sulfate sodium (DSS)-induced colitis in mice was tested. IFNγ-MdC exert a profoundly protective effect in these models, leading to rapid disease resolution in most cases. Furthermore, using in vitro coculture experiments, we show that IFNγ-MdC profoundly delete intestinal lymphocytes derived from mice with colitis. Most intriguing is the finding that lymphocytes surviving in IFNγ-MdC cocultures are highly enriched for CD4+CD25+Foxp3+ T cells with regulatory activity. Additionally, we show that signaling via the IFN-γ receptor and CD40 on IFNγ-MdC is necessary for the enrichment of regulatory T cells by IFNγ-MdC, and that IFNγ-MdC require the presence of CD4+ T cells to gain this functional capacity. Contrary to expectation, IDO is not a mediator of the immunosuppressive activities of IFNγ-MdC. Therefore, IFNγ-MdC represent a distinct macrophage subtype with multiple T cell-suppressive effects in vitro and in vivo.
Materials and Methods
Preparation of IFNγ-MdC and other macrophage populations
Heparinized blood, bone marrow, and spleen were obtained from female BALB/c mice (Charles River Breeding Laboratories). Tissues were dispersed and mononuclear cells isolated by Ficoll gradient separation. Cells were cultured for 5 days in RPMI 1640 medium containing 10% FCS and 5 ng/ml mouse recombinant M-CSF (R&D Systems). After 1 and 3 days, cultures were gently washed to select for adherent cells and fresh medium was added to the adherent cell layer. On day 4, 25 ng/ml mouse recombinant IFN-γ (Chemicon International) was added to the cultures for 16–20 h. The next day adherent cells were harvested with a cell scraper and washed before use. These cells are the IFNγ-MdC; no further separation procedure is performed on these cells before their use.
Classically activated (LPS and IFN-γ-stimulated; M1) macrophages were prepared from bone marrow tissue, as previously described (11). Resting macrophages were also obtained from bone marrow, according to previously described procedures (12), whereby these macrophages were produced by M-CSF stimulation of bone marrow cells for a total of 9 days.
Abs and flow cytometry
The following FITC, PE, allophycocyanin, Alexa Fluor 647, and Alexa Fluor 488 conjugated mAbs (from BD Biosciences) were used for flow cytometric analyses: anti-CD1d, anti-CD3, anti-CD4, anti-CD8, anti-CD11a (LFA-1, α-chain), anti-CD16, anti-CD19, anti-CD25, anti-CD38, anti-CD40, anti-CD45, anti-CD49b, anti-CD54 (ICAM-1), anti-CD62 ligand (anti-CD62L), anti-CD64, anti-CD80, anti-CD86, anti-CD103, anti-CD152 (CTLA-4), anti-CD154, anti-Gr-1, anti-I-A/I-E (MHC class II), and anti-LPAM-1 (integrin α4β7 complex). Anti-CD14 and anti-CD28 were obtained from Caltag Laboratories. Additionally, anti-CD205 was obtained from Serotec; anti-PD-L1, anti-Foxp3, anti-CD123-biotin, anti-CD207, anti-CD11b, anti-CD11c, and anti-F4/80 were obtained from eBioscience. The biotinylated mAb was visualized with streptavidin-PE. Ab specificity was controlled using matched isotype mAbs. Flow cytometry was performed using FACSCalibur instrumentation (BD Biosciences).
IFNγ-MdC were used at 7.5 × 105/well in 12-well plates for Transwell and flow cytometry experiments. Unless stated otherwise, equal numbers of cocultured lymphocytes were added to these cultures. For all other coculturing experiments, a specified number of IFNγ-MdC or control cell populations was mixed with 3 × 105 lymphocyte targets in 24-well plates. “Control cells” consist of the starting bone marrow, spleen, and blood cell population used to make IFNγ-MdC; also, fresh monocytes purified from these tissues with CD11b-labeled magnetic beads (CD11b MicroBeads; Miltenyi Biotec) were compared for control purposes to IFNγ-MdC activities. For experiments involving “colitis lymphocytes,” target cells were obtained from mesenteric lymph nodes (LNs) of mice with DSS-induced colitis; otherwise, fresh splenocytes were used as a source of cocultured lymphocytes. In some experiments, target lymphocytes were washed and fluorescently labeled with CFSE using the Vybrant CFDA SE Cell Tracer kit (Molecular Probes). In specified cocultures, intestinal Ag was provided in the form of sterile fecal extract (200 μg of protein/ml) (13). Apoptosis pathways were blocked with a general caspase inhibitor (20 μM, Z-VAD-FMK; Promega); an inactive peptide was used as a control for caspase inhibition (20 μM, Z-FA-FMK). Unlabeled and CFSE-labeled lymphocytes were counted by light and fluorescent microscopy, respectively, using a Neubauer chamber.
T cell proliferation was measured by stimulating CD4+ cells (purified with Miltenyi CD4 MicroBeads) with CD3 and CD28 plate-bound mAbs (10 μg/ml each for coating). Purified CD4+ T cells were CFSE-labeled and combined with an equal number (5 × 105) of either CD4+CD25− cells or CD4+CD25+ cells (sorted from IFNγ-MdC cocultures). These cell populations were sorted first by a negative selection of CD4+ cells using the CD4+ T cell isolation kit (Miltenyi Biotec), and then by either negative or positive selection using CD25-PE mAb and anti-PE MicroBeads (Miltenyi Biotec). Flow cytometry analysis was performed on the CFSE cells after 3 days.
BALB/c and C57BL/6 mice (Charles River Breeding Laboratories) were used for the in vitro experiments. For experiments testing the role of IFN-γ, we purchased mice homozygous for a targeted mutation in the IFN-γ receptor 1 gene on C57BL/6 background (B6.129S7-Ifngrtm1Agt; The Jackson Laboratory). Similarly, knockout (KO) mice lacking expression of CD40 and IDO (B6.129P2-CD40tm1Kik/J and B6.129-Indotm1Alm/J mice on C57BL/6 background, respectively; The Jackson Laboratory) were also used. These KO mouse strains are referred to as IFN-γR−/−, CD40−/−, and IDO−/−.
In specified experiments, the starting population of mixed mononuclear cells used to generate IFNγ-MdC was depleted of CD4, CD8, or B cells. Miltenyi anti-CD4, anti-CD8, and anti-CD45R/B220 MicroBeads were used to deplete the various lymphocyte subpopulations. Flow cytometric analysis of the resultant population with respective mAbs showed a >90% depletion in all cases. In some experiments, CD11b+ cells were depleted from the generated IFNγ-MdC population using Miltenyi CD11b MicroBeads. Populations of purified CD4+ and CD11b+ cells were used to generate IFNγ-MdC in other cases; these methods have been mentioned previously. Where CD4+ and CD11b+ cells were mixed to produce IFNγ-MdC, 50 million CD11b+ cells were combined in a 100-mm petri dish with 12.5, 25, or 50 million CD4+ cells to achieve final ratios of 4:1, 2:1, and 1:1, respectively. Finally, the conditions were also altered in a set of coculture experiments, whereby soluble CD40L (Acris Ab GmbH), known to form biologically active homotrimers (14), was added to cultures at a concentration (500 ng/ml) previously shown to activate monocytes/macrophages when combined with IFN-γ (8). The chemicals 1-methyl-tryptophan (1-MT), l-NIL (N6-(iminoethyl)-l-lysine), an inducible NO synthase (iNOS) inhibitor, and concanavalin A (Con A) were purchased from Sigma-Aldrich.
In vivo tracking experiments
IFNγ-MdC were labeled using the PKH67 cell linker dye kit (Sigma-Aldrich), and 2.5 × 106 labeled cells were injected via the tail vein into BALB/c mice. Animals were sacrificed 2 or 7 days after injection, and the tissue distribution of IFNγ-MdC was recorded with fluorescent ex vivo microscopy using a modified Axiotech Vario II Microscope (Zeiss) equipped with a modified SONY 3CCD color video camera (AVT Horn). In a separate experiment, mice were injected with PKH67-labeled IFNγ-MdC, but cells were lethally irradiated (30 Gy) just before injection. PKH67-labeled IFNγ-MdC were verified as dead before injection by trypan blue staining. This experiment was performed to show whether labeled debris from dead IFNγ-MdC was being taken up by resident phagocytic cells producing false-positive artifacts.
Mouse colitis models
For colitis induction, female BALB/c mice (Charles River Breeding Laboratories) intermittently received 2.5% DSS (m.w. 40,000; ICN Pharmaceuticals) in the drinking water. Specifically, mice received four cycles of DSS treatment for 7 days, followed each time by a 10-day interval with normal drinking water, as we have previously reported (15).
Female 8- to 10-wk-old SCID mice (C.B17/lcrCrl-scid, congenic to BALB/c; Charles River Breeding Laboratories) were used for experiments involving the “cell-transfer colitis” model. As previously described (16, 17), colitis was induced in SCID mice by transfer of naive T cells. T cells were isolated from spleens of immunocompetent female BALB/c donors by MACS (Miltenyi MicroBeads). First, the cells were selected to be CD4+, and then CD62L+ cells were selected from the CD4+ population. For colitis induction, 5 × 105 CD4+CD62L+ cells were injected i.p. into SCID mice. Animal procedures were approved by the regional authorities.
Animals in both models were weighed and monitored for their general condition on a daily basis. At the specified experimental endpoint, mice were sacrificed and colon tissues were removed and paraffin-embedded; tissues were sectioned and stained with H&E. For histological evidence of colitis, a scoring system, described in detail elsewhere (18), was used by two blinded investigators to quantitate intestinal inflammation (average score used). Assessments were made using a 0–4 scale, where a score of 4 represents the most severe colitis.
Data from in vitro experiments are expressed as the mean ± SD. Animal weights from in vivo experiments are shown as the mean ± SEM from multiple animals. Statistical analyses for in vitro assays were performed using a standard t test. Histologic scoring comparisons between groups were made with the Mann-Whitney U test. In Figs. 5–10, where quadrant plots of flow cytometry results are given, the percentage of cells in each quadrant is shown in the respective quadrant.
IFNγ-MdC have a macrophage-like phenotype
IFNγ-MdC were generated from the bone marrow, spleen, and blood of BALB/c mice by M-CSF stimulation for 5 days, at which time an adherent population of large, macrophage-like cells became predominant in the cultures. On the fourth day of culture, this adherent population of cells was stimulated with IFN-γ. Although the number of IFNγ-MdC harvested from each separate tissue were generally different (bone marrow > spleen > blood), the morphologic characteristics of the cells were essentially indistinguishable (Fig. 1 A).
Flow cytometry on the large granular IFNγ-MdC revealed features typical of the macrophage lineage. IFNγ-MdC were negative for classical lineage markers of T cells (CD4 and CD8), B cells (CD19), NK cells (CD49b), and granulocytes (Gr1), and did not express CD152 or CD154. Consistent with their origin, IFNγ-MdC expressed CD1d, CD11b, CD11c, CD14, CD16/32, CD38, CD45, CD64, CD68, and CD123 (Fig. 1 B). In addition, IFNγ-MdC expressed intermediate levels of MHC class II Ags, CD40, and CD86, but not CD80. IFN-γ treatment induced expression of CD274 (programmed death receptor ligand 1 (PD-L1)). Expression of CD62L, LFA-1, and ICAM-1, but not integrin α4β7, was detected. IFNγ-MdC can be confidently said not to be a subpopulation of dendritic cells, owing to expression of F4/80 Ag, and the absence of dendritic cell-restricted markers including CD205, CD207, and CD209.
By their mode of derivation, IFNγ-MdC were expected to more closely resemble classically activated (M1-polarized) macrophages rather than any M2-polarized macrophage subset. Hence, we compared the expression of a panel of cell surface markers in IFNγ-MdC with resting and M1 macrophages (Fig. 2). Resting macrophages expressed similar levels of CD11b, CD14, CD86, and F4/80 to levels in IFNγ-MdC, but relatively higher levels of CD11c, CD40, CD274, and MHC class II. Using the same set of markers, M1 macrophages were found to express less CD11c than IFNγ-MdC, and relatively more CD40. IFNγ-MdC expressed lower levels of MHC class II and CD86 than M1 macrophages, which is consistent with the notion that these cells exist in a state of “partial maturation” (6). Although IFNγ-MdC share markers of both M1 and resting macrophages, they can be readily distinguished from cells in these activation states. Likewise, the cell surface phenotype of IFNγ-MdC does not exactly correspond to that of other macrophage subsets described in the literature. Therefore, we consider IFNγ-MdC to be macrophages in a unique, novel state of activation.
IFNγ-MdC eliminate lymphocytes from inflamed mesenteric LNs
An initial indication that IFNγ-MdC mediate deletion of T cells came from the observation that during IFNγ-MdC culture, the number of contaminating lymphocytes dramatically declines over time. To further investigate this effect, lymphocytes from mesenteric LNs of mice suffering from DSS-induced colitis (colitis lymphocytes) were isolated and labeled with CFSE, before being cocultured with IFNγ-MdC from healthy syngenic mice. After 48 h, the remaining lymphocytes were quantified by fluorescence microscopy, revealing a significant reduction of CFSE-labeled cells in the presence of IFNγ-MdC, but not in the presence of freshly isolated control cells in equal number (Fig. 3,A); control cells are the starting bone marrow, spleen, and blood cell population used to make IFNγ-MdC. Depletion of lymphocytes was observed at IFNγ-MdC:colitis lymphocyte titers as low as 1:30. Time course experiments showed a near-linear reduction in lymphocyte numbers over time, reaching completion at ∼48 h (Fig. 3,B). Furthermore, IFNγ-MdC generated from mice with colitis have the same capacity to kill colitis lymphocytes as those generated from healthy donors (Fig. 3,C). Regardless of the source of IFNγ-MdC, inspection of cocultures showed IFNγ-MdC containing fluorescent cell debris from the CFSE-labeled lymphocytes, suggesting that IFNγ-MdC are actively phagocytic (Fig. 3 D).
To determine whether IFNγ-MdC preferentially delete lymphocytes from animals with colitis, the efficacy of IFNγ-MdC in eliminating lymphocytes derived from mice with DSS-induced colitis was compared with the deletion of lymphocytes from healthy controls. IFNγ-MdC were found to have similar activities against both lymphocyte populations (Fig. 3,E). Addition of sterilized fecal extract to the cultures enhanced, but was not essential for, the killing of lymphocytes from both healthy and colitic donors, suggesting that cell activation may be an important factor for T cell elimination. Consistent with this idea, lymphocytes cocultured with IFNγ-MdC in the presence of the T cell activation inhibitor cyclosporine were not eliminated (Fig. 3 F). Additionally, flow cytometric analysis of cells in cocultures of IFNγ-MdC with lymphocytes revealed a high proportion of activated T cells, judged by physical characteristics and expression of CD25; furthermore, Con A-stimulated lymphocytes cocultured with IFNγ-MdC were, as expected, more readily deleted (data not shown). It appears, therefore, that IFNγ-MdC delete reactive T cells independently of their antigenic specificity, but only following activation.
Lymphocyte killing by IFNγ-MdC is cell contact and caspase-dependent
Although T cell activation appears to be a prerequisite for subsequent deletion by IFNγ-MdC, it is difficult to reconcile this effect with conventional, T cell-inherent activation-induced cell death (19, 20). Firstly, it has previously been established that freshly isolated resting T cells are relatively resistant to autonomous apoptotic death due to expression of FLICE-like inhibitory protein, which prevents Fas-initiated activation-induced cell death (20). Secondly, we observed no reduction in T cell elimination when Fas-Fas ligand (FasL) interactions were blocked with mAb against FasL, Fas chimeric protein, or with a combination of both (data not shown). Thirdly, as we demonstrate below, T cell depletion by IFNγ-MdC cocultures is a strictly contact-dependent process. Together, these observations led us to the hypothesis that the elimination of T cells by IFNγ-MdC is a receptor-mediated activity and should be regarded as “killing” in the true sense.
To delineate the basic mechanism of the IFNγ-MdC killing effect, we attempted to dissociate T cell activation, a necessarily contact-dependent event, from killing using the Transwell system (Fig. 3,G). In these experiments, highly activated Con A-stimulated (5 μg/ml) lymphocytes were separated from IFNγ-MdC in a Transwell chamber, and cell counts were performed after 48 h. Con A-stimulated lymphocytes in direct coculture with IFNγ-MdC were efficiently eliminated (77 ± 7% elimination, vs Con A-stimulated cells alone); however, without this direct contact, using Transwell systems, Con A-stimulated lymphocytes were not depleted and, in fact, were present in 28 ± 10% greater numbers than Con A-stimulated lymphocytes grown in the complete absence of IFNγ-MdC (results are the mean ± SD of quadruplicate determinations from one of two independent experiments). These data lend weight to the argument that IFNγ-MdC cytotoxicity is a receptor-receptor interaction-mediated activity and that we are not observing T cell death through activation-induced cell death or activated T cell autonomous death. Next, we asked whether this killing of colitis lymphocytes was mediated through apoptotic signaling pathways. In this respect, introduction of a general caspase inhibitor (Z-VAD-FMK) into the culture system blocked any detectable killing of colitis lymphocytes (Fig. 3 H). Attempts to identify apoptotic T cells from IFNγ-MdC cocultures with Annexin V for flow cytometric analysis were unsuccessful (data not shown); we attribute this to the rapid phagocytosis of apoptotic T cells by IFNγ-MdC.
Efforts have been made to identify possible factors mediating the killing effect of IFNγ-MdC. Besides considering the possible role of Fas-FasL (not apparently involved according to our earlier experiments), we considered the possible involvement of the programmed death receptor 1 (PD-1) and its ligand PD-L1. PD-L1, which induces lymphocyte apoptosis (21, 22), is highly expressed by IFNγ-MdC following IFN-γ treatment. However, using PD-L1 blocking Abs (data not shown) and PD-1 KO mice (Fig. 4 A), we were unable to demonstrate a role for this costimulatory pathway in IFNγ-MdC function.
In a series of elegant papers, Munn et al. (8, 9) demonstrated that IFN-γ and soluble CD40L-induced macrophages were capable of inducing T cell apoptosis through IDO-mediated tryptophan depletion. Somewhat unexpectedly, chemical inhibition of IDO with 0.5 mM 1-MT did not diminish the cytotoxic effect of IFNγ-MdC (Fig. 4,B). Furthermore, IFNγ-MdC derived from IDO KO mice were fully able to delete lymphocyte targets (Fig. 4,C), indicating a lack of IDO involvement. Likewise, we were unable to establish the involvement of iNOS in IFNγ-MdC function, as inhibition of this enzyme with 100-1000 μM l-NIL had no effect on T cell killing (Fig. 4 D). In summary, our experiments to date indicate that the killing effect of IFNγ-MdC is mediated through apoptotic mechanisms requiring target cell contact, which likely involve T cell activation. At present, however, the mechanism responsible for IFNγ-MdC-induced T cell death remains unknown.
Lymphocytes surviving IFNγ-MdC coculture are enriched for regulatory T cells
Regardless of the initial ratio of IFNγ-MdC to T cells in coculture, a proportion of residual T cells survived. To better characterize these surviving T cells, IFNγ-MdC were cocultured with unlabeled mouse mesenteric LN cells for 3 days, after which the lymphocyte population was examined by flow cytometry. Consistent with our previous findings, we observed an expansion of the activated CD4+ pool identified by high CD25 expression (Fig. 5,B), whereas control lymphocyte populations, grown in the absence of IFNγ-MdC (Fig. 5,A) or in the presence of control cells (Fig. 5,C) or the presence of purified monocytes (Fig. 5,D), contained few CD4 CD25 double positive cells. In terms of absolute numbers of CD4+CD25+ cells, there were 2.9 × 104 cells present in the added lymphocyte population, with a final number of 2.6 × 105 cells after 3 days of coculture with IFNγ-MdCs, indicating a true numerical expansion of regulatory T cells. Further characterization of the CD4+CD25high T cell subset demonstrated cytoplasmic CTLA-4 (CD152) expression (Fig. 5,E). It was also observed that half of the CD4+CD25high population expressed CD103 (Fig. 5,F), which is a molecule previously shown to define a unique subpopulation of regulatory T cells able to control inflammatory bowel disease (23, 24). RT-PCR analysis of MACS-separated T cells showed greater expression of Foxp3 mRNA in the CD4+CD25+ T cell population than in CD4+CD25− cells sorted from the same IFNγ-MdC cocultures (Fig. 5,G); Foxp3 expression in the CD4+CD25high T cell subset was subsequently confirmed by flow cytometry (see Figs. 6–10). To test the functional activity of CD4+CD25+ cells from IFNγ-MdC cocultures, MACS-sorted CD4+CD25+ cells were added into CD3/CD28 mAb-stimulated T cell (CFSE-labeled) cultures. T cell proliferation after 3 days was suppressed in the presence of CD4+CD25+ cells from IFNγ-MdC cocultures, but not with CD4+CD25− cells, formally demonstrating that the CD4+CD25+ population contains regulatory T cells (Fig. 5,H). Quantitative PCR from MACS-sorted CD4+CD25+ cells revealed far higher expression of IL-10 mRNA than in the CD4+CD25− subset; notably, expression levels of TGF-β and IFN-γ mRNA were not different (Fig. 5 G).
Two mechanisms could underlie enrichment of regulatory T cells, namely a relative expansion from naive T cells or pre-existing regulatory T cells, or selective survival in the face of IFNγ-MdC-mediated cytotoxicity. Addressing this issue, when purified CD4+CD25− cells were added to IFNγ-MdC cultures, CD4+CD25+Foxp3+ cells continued to be enriched (Fig. 5,I). In contrast, purified CD4+CD25+ cells from naive mice were not particularly resistant to the IFNγ-MdC cytotoxicity, with only a slight survival advantage apparent when IFNγ-MdC were present in low relative numbers (Fig. 5 J). Therefore, it is concluded that IFNγ-MdC enrich for CD4+CD25+ cells primarily by inducing or expanding T cells with a regulatory phenotype; however, it does not strictly follow from these data that passive enrichment is not a contributory mechanism for regulatory T cell enrichment in our system.
CD4+CD25+ cells are not enriched in resting or M1 macrophage cocultures
To begin to determine whether there is a functional distinction between IFNγ-MdC and other well-characterized populations of macrophages, we compared the enrichment of regulatory T cells in cocultures with resting or M1 macrophages to cocultures with IFNγ-MdC. In these experiments, normal splenocytes were cocultured with the different populations of macrophages and CD4+CD25+Foxp3+ T cells were quantified after three days. Results show that only low relative and absolute numbers of CD4+CD25+ T cells were measurable in cultures containing either resting or M1 macrophages, compared with the high proportion and absolute numbers of Foxp3+, CD4+CD25+ T cells in IFNγ-MdC cocultures (Fig. 6). It should be noted that the coculturing of splenocytes with M1 macrophages did result in a high rate of lymphocyte depletion that was not significantly different from that observed with IFNγ-MdC (data not shown). Nonetheless, the striking difference in the ability of IFNγ-MdC to highly enrich for CD4+CD25+Foxp3+ T cells lends to the distinctive nature of these macrophages, in comparison to resting or M1 macrophages.
CD11b+ MdCs mediate IFNγ-MdC killing and regulatory T cell activities
Strictly, the effects so far attributed to the IFNγ-MdC population might have been activities of minor contaminating cell populations. To exclude this possibility, CD11b+ cells were removed (MACS, <5%) from the cell mixture obtained on day 5 of IFNγ-MdC generation. The CD11b-depleted fraction of the IFNγ-MdC cultures was then tested for the ability to kill fresh lymphocyte targets (1:1 ratio) and to enrich for CD4+CD25+Foxp3+ T cells. Results from these experiments show that the killing effect was completely abrogated (Fig. 7,A) and that CD4+CD25+Foxp3+ T cells were not enriched in cocultures when the CD11b+ cells were removed (Fig. 7, B–D). Thus, the functional activities of IFNγ-MdC cultures are a property of the adherent CD11b+ IFNγ-MdC population, and not of other cell types in the preparation.
IFNγ receptor and CD40 function are necessary for enrichment of CD4+CD25+Foxp3+ cells
Earlier previous work had suggested that addition of IFN-γ to IFNγ-MdC culture was important for their ability to suppress allogeneic T cell responses in vivo (F. Fändrich, unpublished data), although the cellular and molecular basis of this effect was not known. To investigate the role of IFN-γ stimulation in IFNγ-MdC generation, we produced IFNγ-MdC from IFN-γ receptor KO (IFN-γR−/−) mice. Compared with IFNγ-MdC from wild-type (WT) C57BL/6 mice, the number of target lymphocytes surviving coculture with IFNγ-MdC derived from IFN-γR−/− mice was not altered (Fig. 8,A). However, the number of CD4+CD25+Foxp3+ T cells was markedly reduced in cocultures with IFN-γR−/−-derived IFNγ-MdC (Fig. 8, B–D). Although we are conscious of the difficulties in interpreting experiments using KO mice, it proved technically infeasible to block the activity of endogenously produced IFN-γ in IFNγ-MdC cultures, and cocultures, with neutralizing Abs because IFNγ-MdC are Fc receptor-bearing and actively phagocytic. Using IFNγ-MdC from IFN-γR−/− mice, but responder T cells from WT mice, also provided additional information because it allowed us to demonstrate that IFN-γ must act directly on IFNγ-MdC and that its effects are independent of IFN-γ stimulation of cocultured cells.
It has been previously shown that IFN-γ and CD40 act together to induce macrophages into a suppressive phenotype that functions via IDO (8). CD40 is expressed by IFNγ-MdC (Fig. 1,B) and the T cells in coculture with IFNγ-MdC express CD40L. Using IFNγ-MdC derived from CD40−/− KO mice we showed that, although killing of lymphocyte targets was not diminished in cocultures, development of CD4+CD25+Foxp3+ T cells was severely impaired (Fig. 8, E–H), indicating a key role for CD40. Interestingly, although IFN-γ and CD40 involvement is apparent, we have found no evidence that IDO plays a crucial role in IFNγ-MdC function. Besides our earlier experiments showing that IDO does not appear to influence T cell survival (Fig. 4, B and C), chemical inhibition with 1-MT had no effect on CD4+CD25+Foxp3+ T cell generation (Fig. 9,A); moreover, IFNγ-MdC produced from IDO−/− mice showed no reduction in CD4+CD25+Foxp3+ T cell generation in coculture (Fig. 9 B).
Essential conditions for the generation of IFNγ-MdC
IFNγ-MdC development depends upon signaling through CD40 and the IFN-γ receptor, and we hypothesized that other auxiliary T cell signals were important in this process. To test this contention, we produced IFNγ-MdC from splenic mononuclear cells depleted of specific lymphocyte populations, namely CD4+ or CD8+ T cells or B cells. From these experiments, a clear requirement for CD4+ T cells in the starting cell mixture was evident, as IFNγ-MdC produced in the absence of CD4+ T cells failed to gain regulatory T cell-inducing capacity (Fig. 10,A). Depletion of CD8+ cells or B cells had no effect on IFNγ-MdC functional development (Fig. 10,A). To strictly confirm this finding, and to exclude the role of other possible cells in the optimal production of IFNγ-MdC, we demonstrated that fully competent IFNγ-MdC were generated from mixtures of purified CD11b+ and purified CD4+ cells. Thus, the generation of IFNγ-MdC can be shown to be a CD4+ dose-dependent effect: using regulatory T cell-inducing capacity in coculture as a measure of IFNγ-MdC function, it is clear that very high activity is achieved when, in the initial generation of IFNγ-MdC, monocytes and CD4+ T cells are present in a 1:1 ratio. Such “optimal” IFNγ-MdC cultures induced up to 87% CD4+CD25+ cells in the subsequent coculture step (Fig. 10 B).
Because CD4+ T cells are necessary for IFNγ-MdC generation, and as we have shown, signaling through CD40 and IFN-γ receptor are indispensable in this process, we next investigated whether these signals were sufficient to substitute for the action of T cells in IFNγ-MdC development. When IFNγ-MdC were produced from purified monocytes (CD11b+) in the presence of CD4+ cells at a 1:1 ratio, a high degree of CD4+CD25+Foxp3+ T cell enrichment was observed; when purified monocytes were subjected to the same culture process in the absence of CD4+ T cells, no regulatory T cell enrichment effect was evident and the addition of recombinant soluble CD40L to purified monocytes cultured under the same conditions had no discernible effect (Fig. 10 C). Importantly, the absolute requirement for CD4+ in IFNγ-MdC development distinguishes them from the macrophage type described by Munn et al. (8, 9), which arises from soluble CD40L-stimulated, highly purified, macrophage populations in the complete absence of T cells.
IFNγ-MdC therapy reduces inflammation in mice with chronic colitis
Finally, we investigated whether IFNγ-MdC could be used to treat inflammatory colitis. To ascertain that IFNγ-MdC from BALB/c mice are capable of migrating to intestinal tissues, 2.5 × 106 fluorescently labeled IFNγ-MdC were injected into the tail vein of syngenic animals. Two days later, IFNγ-MdC could be observed in the spleen, Peyer’s patches, mesenteric LNs, and colonic mucosa (n = 3 mice). These infiltrates of IFNγ-MdC were still present after 7 days (Fig. 11 A). Additionally, IFNγ-MdC were also detected in the lung, liver, bone marrow, and cervical LN; no cells were found in the thymus (data not shown). To demonstrate that the tissue-resident fluorescent cells were genuinely IFNγ-MdC, and not resident macrophages that had phagocytized particles originating from the labeled IFNγ-MdC, IFNγ-MdC were prepared as usual, but killed by irradiation before injection into the mice. In this case, no green fluorescent cells were observed in any of the tissues of the mice (n = 3) when examined 2 days after injection. We conclude, therefore, that the fluorescent cells observed after viable cell injection were IFNγ-MdC, and that these cells migrate to locations where they might encounter colitis-associated Ags and disease-causing T cells.
Because in vitro experiments had implicated the generation of regulatory T cells as a mechanism by which IFNγ-MdC operate, we first tested for the therapeutic potential of IFNγ-MdC in a mouse model whereby purified naive T cells (CD4+CD62L+), essentially depleted of regulatory T cell populations, induce chronic colitis in SCID mice. As shown in Fig. 11, B and C, this regimen resulted in clinically apparent disease with weight loss and histologically evident colonic inflammation in control mice. In contrast, administration of 5 × 106 IFNγ-MdC to these animals led to a steady gain of body weight (Fig. 11,B), and less colonic inflammation (Fig. 11 C), compared with control mice. In a separate experiment, where we tested the effects of freshly purified monocytes, weight changes and histologic changes indicated the development of colitis with an even greater severity compared with the no treatment group (data not shown).
As a rigorous test for the therapeutic potential of IFNγ-MdC, fully immunocompetent BALB/c mice were induced to develop chronic DSS colitis. One day after the last DSS treatment, mice were either left untreated, were given 5 × 106 control cells, or a single i.v. injection of 5 × 106 syngenic IFNγ-MdC; 3 wk later, harvested colon tissue was histologically scored (Fig. 11, D and E). A minority of mice (3/14) treated with IFNγ-MdC showed almost complete histological resolution of their colitis, with the remaining majority (8/14) showing low-to-intermediate grades of colitis (score ≤2). By contrast, DSS-treated control mice fared considerably worse, where all mice (31/31) that received either control cells or no injection of cells showed gross signs of colitis (Fig. 11 F) and had a histological score of ≥2, with a majority showing the most severe grade of colitis (score ≥ 3.5).
In this study, we give the first description of a macrophage subset, IFNγ-MdC, which are defined by a unique constellation of phenotypic markers and functional properties that overlap with, but do not exactly conform to, previously described macrophage populations. IFNγ-MdC profoundly deplete cocultured T cells through a presently unidentified mechanism and, most remarkably, T cells surviving otherwise lethal engagement with IFNγ-MdC are highly enriched for CD4+CD25highFoxp3+ regulatory T cells. Acquisition of regulatory T cell-enriching capacity crucially depends upon interaction with CD4+ T cells and we have shown that signaling through the IFN-γ receptor and CD40 are necessary, but not sufficient, for this to happen. Finally, we have demonstrated that IFNγ-MdC are effective in the treatment of established colitis in mice, suggesting that IFNγ-MdC may eventually find clinical application in the treatment of patients with inflammatory bowel disease or other autoimmune illnesses.
By their mode of derivation and cell surface phenotype, IFNγ-MdC correspond most closely with those macrophages classified by Gordon and Taylor (1) as “deactivated” macrophages; in particular, IFNγ-MdC are F4/80+, CD11b+, and CD62L+ and express intermediate or low levels of MHC class II. IFNγ-MdC are not DC-like in morphology or cell surface phenotype. Because IFNγ-MdC can be generated from bone marrow, blood, and spleen, it seems most probable that they ultimately derive from the circulating CCR2+ “inflammatory” monocyte subset, although this hypothesis has not yet been tested. Surprisingly, perhaps, because the conditions under which IFNγ-MdC are generated might be expected to elicit classically activated (M1-polarized) macrophages, this expectation is not the result. It is also clear that IFNγ-MdC are not alternatively activated (M2-polarized) macrophages, not having been exposed to IL-4 or IL-13 and because IFNγ-MdC have a CD16/32+CD64+ phenotype (1, 4). IFNγ-MdC expressed intermediate levels of CD86 and were CD80−, with low or intermediate MHC class II expression, but were nevertheless capable of inducing resting T cells to express IL-2R α-chain. Therefore, we consider IFNγ-MdC to be a more mature form than the “resting” macrophage. Our direct phenotypic comparisons of IFNγ-MdC to resting macrophages revealed differences consistent with this idea, and more importantly, resting macrophages did not demonstrate an ability to enrich for regulatory T cells. It should also be noted that we find IFNγ-MdC cultures to be a relatively consistent mixture of cells and we have not identified contaminating populations of dendritic cells that might account for the functions ascribed to IFNγ-MdC. Furthermore, when comparing the functional capacities of IFNγ-MdC to classically activated macrophages, only IFNγ-MdC in our hands showed a capacity to enrich regulatory T cells. Together, our results suggest in vitro generated IFNγ-MdC show significant differences to previously described subsets of macrophages and dendritic cells.
It is clear that IFNγ-MdC have functional similarities to the population of IFN-γ-induced human macrophages previously described by Munn et al. (8, 9), although we do not consider IFNγ-MdC to be directly equivalent to these cells. Crucially, as we discuss in greater detail, the capacity of macrophages described by these studies to eliminate activated T cells has been principally attributed to IDO-dependent mechanisms, whereas the action of IDO makes no substantial contribution to the T cell-depleting or regulatory T cell-enriching activities of IFNγ-MdC. Moreover, Munn’s group was able to generate suppressive macrophages from highly purified, M-CSF-treated monocyte preparations using recombinant soluble CD40L and IFN-γ; by contrast, we have shown an absolute requirement for CD4+ T cells in IFNγ-MdC development. The role of CD4+ T cells in the development of the IFNγ-MdC is totally consistent with our view that IFNγ-MdC are part of the heterogeneous group of macrophages previously categorized as deactivated. It seems increasingly likely that multiple inhibitory receptor interactions cooperate in delivering the tolerizing impetus to IFNγ-MdC precursors. For the reasons given, we contend that IFNγ-MdC represent a previously unrecognized macrophage derivative that is not an alternatively activated macrophage, classically activated macrophage nor a previously described tolerogenic form, but a cell in a novel state of activation.
The capacity to delete cocultured lymphocytes is not unique to IFNγ-MdC, as evidenced by the fact that cells derived from CD40−/− and IFN-γR−/− animals were competent in this regard. Nevertheless, removal of autoaggressive T cells may contribute to the ability of IFNγ-MdC to control colitis in mice. We have established that IFNγ-MdC-induced T cell death is a contact-dependent process, in which T cell activation preceding elimination may be an absolute requirement. This contention is supported by the fact that CD25+ activated T cells are observed in IFNγ-MdC cocultures and that cyclosporine prevents T cell depletion. Experiments using a Transwell apparatus support the assertion that IFNγ-MdC cytotoxicity is a receptor-receptor interaction-mediated activity, rather than an activation-induced T cell autonomous process or an activity mediated by diffusible factors. Although we have been unable to identify the molecular mechanism of IFNγ-MdC killing, we have excluded the involvement of several prominent candidate interactions, notably Fas-FasL and PD-L1-PD1.
In addition to deleting T cells, IFNγ-MdC are capable of substantially increasing the absolute number of bona fide regulatory T cells in cocultures of unactivated T cells with IFNγ-MdC. This activity depends upon activation of T cells by the IFNγ-MdC and contact-dependent induction of regulatory function through an, as yet, unidentified interaction. Moreover, this activity is unique to cells stimulated with IFN-γ, as M-CSF-derived macrophages from IFN-γR−/− animals were incapable of enriching regulatory T cells. IFN-γ has paradoxical activities with regard to controlling the immune response: recent publications have suggested that one of its critical roles may be to down-regulate immune reactions. With regard to autoimmunity, for example, regulatory T cells are unable to prevent collagen-induced arthritis in IFN-γR−/− mice (25), and IFN-γ protects against the development of experimental allergic encephalitis and experimental autoimmune uveitis (26, 27).
We attribute the enrichment of CD4+CD25highFoxp3+ regulatory cells in IFNγ-MdC cocultures to an absolute increase in their number and not to a selective survival advantage in the face of IFNγ-MdC-mediated killing. The relative contribution of uncommitted T cells and regulatory T cells, present both in the mixed T cell population added in coculture phase and as a minor contaminant of the original IFNγ-MdC cultures, to the total number of regulatory T cells observed at the end of the coculture phase is not fully resolved. It is acknowledged that regulatory T cells may derive from each of these sources and that, depending on the time point at which the cell mixture is analyzed, the proportion of cells originating from each potential source may vary. The analysis of this system is complex because the enrichment of regulatory T cells from any particular parent population may be influenced by the presence of other T cell populations in the culture. Work to determine the origins of the regulatory T cells in IFNγ-MdC cultures is ongoing.
The immunological basis of inflammatory bowel diseases, ulcerative colitis and Crohn’s disease is incompletely understood, but in broad terms, is best considered as an exaggerated response to normal intestinal luminal constituents. In part the response is precipitated by defective mucosal barrier function resulting in repeated antigenic exposure to normal gut flora. This mechanism of human disease is simulated in the DSS-induced colitis mouse model, in which ingested DSS facilitates the passage of microbial components across the intestinal mucosa causing a very severe colitis (28). Our second mouse model, in which colitis is induced by i.p. injection of purified CD4+CD62L+ T cells from healthy donors, underscores the etiological significance of an imbalance between effector T and regulatory T cell functions in the development of inflammatory bowel disease (16, 29). In the present study, we show that IFNγ-MdC have properties capable of restoring the immunologic balance, exhibited by amelioration of symptoms and pathology associated with colitis in mice. Delineating the mechanisms of the therapeutic effects of IFNγ-MdC in vivo was beyond the scope of this work; indeed, it is very difficult to establish that IFNγ-MdC delete disease-causing T cell clones in vivo. However, it was observed that IFNγ-MdC migrated to the inflamed intestinal mucosa and mesenteric LNs, and that IFNγ-MdC-treated animals had less severe mucosal lymphocytic infiltrates.
We conclude with a clinical perspective. The use of human IFNγ-MdC in patients with inflammatory bowel disease is an entirely feasible proposition. In the setting of human renal transplantation, Phase I clinical trials with GMP-produced IFNγ-MdC are nearing completion, without apparent toxicities or other adverse effects (F. Fändrich, unpublished data). In the case of patients with ulcerative colitis or Crohn’s disease, we envisage that autologous cells will be used, reducing the risks of transfusion-related pathology, infection, and failure of the therapy owing to rejection responses against allogeneic IFNγ-MdC. Importantly, we have shown that IFNγ-MdC derived from mice with colitis are equally as effective in suppressing experimentally induced colitis as those from healthy donors. Understanding the full immunoregulatory potential of IFNγ-MdC and testing their clinical application, we believe, will lead to important advances in the treatment of inflammatory bowel disease.
We thank Dr. Gabriel Glockzin, Bärbel Schell, and Erika Frank for excellent technical expertise. Finally, we also thank Professor Ian V. Hutchinson for making valuable comments on the manuscript.
Fred Fändrich is a member of a company (Blasticon GmBH), which controls intellectual property claims of the cells described in this paper. Edward K. Geissler has received grant money from Fresenius GmbH to in-part fund this study.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This study was supported by Grant GE 1188/1-1 from the Deutsche Forschungsgemeinschaft, by the University of Regensburg Medical School Research Emphasis Program (ReFoRM C Program), and by Fresenius Biotech.
Abbreviations used in this paper: IFNγ-MdC, IFN-γ-stimulated monocyte-derived cell; FasL, Fas ligand; DSS, dextran sulfate sodium; CD62L, CD62 ligand; Con A, concanavalin A; iNOS, inducible NO synthase; KO, knockout; LN, lymph node; l-NIL, N6-(iminoethyl)-l-lysine; 1-MT, 1-methyl-tryptophan; PD-1, programmed death receptor 1; PD-L1, programmed death receptor ligand 1; WT, wild type.