Type 1 T Helper Cells Induce the Accumulation of Myeloid-Derived Suppressor Cells in the Inflamed Tgfb1 Knockout Mouse Liver†
Potential conflict of interest: Nothing to report.
Abstract
Immune-mediated liver injury in hepatitis is due to activated T cells producing interferon-γ (IFN-γ). It is important to identify negative feedback immune mechanisms that can regulate T cell activity. In this study, we demonstrate that liver inflammation mediated by type 1 T helper (Th1) cells can induce the accumulation of myeloid-derived suppressor cells (MDSCs), pleiomorphic cells capable of modulating T cell–mediated immunity, that heretofore have been studied almost exclusively in the context of tumor-associated inflammation. Mice deficient in the gene encoding transforming growth factor-β1 (Tgfb1−/− mice) acutely develop liver necroinflammation caused by IFN-γ–producing clusters of differentiation 4–positive (CD4+) T cells. Liver Th1 cell accumulation was accompanied by myeloid cells expressing CD11b and Gr1, phenotypic hallmarks of MDSCs. Isolated Tgfb1−/− liver CD11b+Gr1+ cells were functional MDSCs, readily suppressing T cell proliferation in vitro. Pharmacologic inhibitors of inducible nitric oxide (NO) synthase completely eliminated suppressor function. Suppressor function and the production of NO were dependent on cell–cell contact between MDSCs and T cells, and upon IFN-γ, and were specifically associated with the “monocytic” CD11b+Ly6G− Ly6Chi subset of liver Tgfb1−/− CD11b+ cells. The rapid accumulation of CD11b+Gr1+ cells in Tgfb1−/− liver was abrogated when mice were either depleted of CD4+ T cells or rendered unable to produce IFN-γ, showing that Th1 activity induces MDSC accumulation. Conclusion: Th1 liver inflammation mobilizes an MDSC response that, through the production of NO, can inhibit T cell proliferation. We propose that MDSCs serve an important negative feedback function in liver immune homeostasis, and that insufficient or inappropriate activity of this cell population may contribute to inflammatory liver pathology. (HEPATOLOGY 2010;)
Thymus-derived lymphocytes (T cells) are the proximal agents of parenchymal liver damage in inflammatory liver diseases such as autoimmune hepatitis (AIH) and viral hepatitis. In AIH, clusters of differentiation 4–positive (CD4+) T cells infiltrate liver parenchyma1 and release hepatotoxic cytokines such as interferon gamma (IFN-γ) and tumor necrosis factor alpha (TNF-α).2, 3 IFN-γ expression by ex vivo cultured T cells strongly correlates with disease activity,4 implicating type 1 T cell responses in hepatocellular damage. In hepatitis C virus infection, liver pathology results from the activity of T cells producing IFN-γ within liver parenchyma, because hepatitis C virus is not cytopathic.5-7 IFN-γ is essential for parenchymal damage in mouse models of T cell–mediated liver injury, including concanavalin A–induced liver injury,8 and spontaneous liver injury in BALB/c transforming growth factor beta 1 (TGF-β1) knockout mice.9 A common theme, therefore, in immune-mediated liver injury is pathology associated with activated T cells producing IFN-γ.
Given the potential for liver damage by activated type 1 T helper cells (Th1) cells, it is important to identify mechanisms that regulate their activity. A variety of liver resident cells participate in the regulation of T cells, including regulatory T cells, dendritic cells, Kupffer cells, natural killer cells, natural killer T cells, stellate cells, and liver sinusoidal epithelial cells.10 Whether regulatory immunocytes accumulate in liver in response to activated T cells is not known. Such cells may represent an important negative feedback mechanism mitigating pathology mediated by T cell activation. It is reasonable to postulate that inflammatory pathology in liver is attributable both to aberrant activation of T cells and to a deficit in appropriate counter-regulatory mechanisms.
Studies emerging from the field of tumor immunity show that tumor-associated inflammation induces the development and accumulation of myeloid-lineage cells with immunomodulatory activity. Termed myeloid-derived suppressor cells (MDSCs), these pleiomorphic cells are capable of suppressing T cell proliferation and subjugating T cell–mediated immunity.11, 12 MDSCs comprise a heterogeneous group of myeloid cells, which employ a variety of mechanisms to inhibit T cell responses. Murine MDSCs are operationally defined as CD11b+Gr1+ myeloid cells that suppress T cell proliferation.11, 12 Although MDSCs have been most extensively described in the context of tumors, recent studies show their involvement in inflammatory responses not associated with tumors.13, 14 MDSCs home to liver in tumor-bearing mice,15 and hepatocellular carcinoma, like other solid tumors, exhibits associated populations of MDSCs,16, 17 but little is otherwise known about MDSCs in liver, particularly in inflammatory pathology. Here, we demonstrate in the BALB/c TGF-β1 knockout mouse model that Th1 cells, through release of IFN-γ, drive accumulation in liver of an MDSC population that can effectively inhibit T cell proliferation through a mechanism involving expression of inducible nitric oxide synthase (iNOS) and the production of nitric oxide (NO).
Abbreviations
AIH, autoimmune hepatitis; CCL2, chemokine (C-C motif) ligand 2; CCR2, chemokine (C-C motif) receptor 2; CD, clusters of differentiation; CFSE,5-(and-6)- carboxyfluorescein diacetate, succinimidyl ester; D-NMMA, D-NG- monomethyl arginine citrate; IL, interleukin; iNOS, inducible nitric oxide synthase; L-NIL, N6-(1-iminoethyl)-L-lysine; L-NMMA, L-NG- monomethyl arginine citrate; mAb, monoclonal antibody; MDSC, myeloid-derived suppressor cell; NO, nitric oxide; nor-NOHA, N- hydroxy-nor-arginine; TCR, T cell receptor; Th1, type 1 T helper cell.
Materials and Methods
Mice.
Mice were bred at Dartmouth Medical School according to Association for Assessment and Accreditation of Laboratory Animal Care practices. BALB/c-background Tgfb1−/− mice, Ifng−/− (null for IFN-γ gene) Tgfb1−/− mice, and Rag1−/− (null for recombination activating gene 1) Tgfb1−/− mice were genotyped as described.9, 18, 19 Depletion of CD4+ T cells and Gr1+ cells used intraperitoneal injections of anti-CD4 and anti-Gr1 (Clone RB6-8C5; BioXCell), respectively, initiated at postnatal day 5 as described.18 Flow cytometry at day 11 confirmed depletion efficiencies of >95%.
Liver Cell Subset Isolation.
Following cardiac perfusion with phosphate-buffered saline, livers were aseptically removed and mechanically disrupted between sterile frosted microscope slides. Cell suspensions were passed twice through 70 μm filters before cell isolation. Liver CD11b+ cells were isolated using anti-CD11b magnetic beads and positive selection columns (Miltenyi) per the manufacturer's protocol. Gr1+ cells were isolated using phycoerythrin-tagged anti-Gr1 (RB6-8C5; eBioscience) and positive immunomagnetic separation using a phycoerythrin selection kit (StemCell Technologies, Inc.). CD11b+Ly6GhiLy6Clo cells were isolated via positive selection employing biotinylated anti-Ly6G and anti-biotin magnetic microbeads (Miltenyi). CD11b+Ly6G−Ly6Chi cells were isolated by negative selection of Ly6G− cells followed by positive selection with biotinylated anti-Ly6C and anti-biotin magnetic microbeads (Miltenyi). Flow cytometry verified that all cell isolations yielded >90% pure populations.
Th1 Effector Cell Development.
Bead-isolated CD4+ T cells were cultured for 3 days with plate-bound anti-CD3ε (10.0 μg/mL), soluble anti-CD28 (1.0 μg/mL; BD Biosciences) recombinant interleukin-12 (IL-12) (10 ng/mL; Peprotech) and anti-IL-4 (10 μg/mL; NCI). Th1 effector development was confirmed by intracellular IFN-γ staining.
Flow Cytometry.
Cells were incubated with Fc Block (anti-CD16/CD32; eBioscience) for 20 minutes at 4°C then washed twice. Cells were stained with antibodies to CD4, CD11b, Gr1, F4/80, programmed death ligand 1 (PD-L1; eBioscience, San Diego), Ly6G (Clone 1A8), Ly6C (Clone 1G7.G10), major histocompatibility complex class II (BD Biosciences), or CD14 (Biolegend) and acquired on either BD FACSCalibur (eBiosciences) or Accuri C6. Data analysis was performed with FlowJo, version 8.8.6 (Tree Star) software.
iNOS Detection.
Cells obtained from suppression assay cultures at 48 hours were surface stained as described above, fixed and permeabilized (CytoFix/CytoPerm; BD Biosciences), stained with rabbit anti-mouse iNOS (BD Biosciences), followed by blocking with 10% normal goat serum, and secondary staining with goat anti-rabbit (Jackson ImmunoResearch). Surface-marker-appropriate isotype and intracellular staining with secondary antibody alone served as negative control. RAW 264.7 cells cultured for 24 hours with lipopolysaccharide and IFN-γ served as positive control. Cells were acquired by flow cytometry.
Suppression Assays.
Prior to inclusion in cocultures, bead-isolated CD4+ or CD8+ T cells from wild-type mouse spleens were stained with 5.0 μM 5-(and-6)-carboxyfluorescein diacetate, succinimidyl ester (CFSE; Invitrogen) for 10 minutes and quenched by washing twice in Roswell Park Memorial Institute medium 1640/10% fetal bovine serum. Isolated Tgfb1+/− or Tgfb1−/− CD11b+ cells were added at 3.0 × 105 (“300K”) or 1.0 × 105 (“100K”) cells to cocultures with CFSE-labeled purified CD4+ T cells or CD8+ T cells (@3.0 × 105 cells). Th1 cells were stimulated with plate-bound anti-CD3ε (BD Biosciences) at 10.0 μg/mL, whereas cultures of isolated splenic T cells also included soluble anti-CD28 at 1.0 μg/mL (BD Biosciences). DO11.10 mouse splenocytes (1.0 × 106) were stimulated with 0.3 μM ovalbumin (OVA323-339) peptide. Inhibitors were added at the start of culture as follows: 5.0 mM NG-monomethyl-L-arginine (L-NMMA; Calbiochem), 5.0 mM NG-monomethyl-D-arginine (D-NMMA; Calbiochem), 0.5 mM N6-(1-iminoethyl)-L-lysine (L-NIL; Sigma), 1.0 mM N-hydroxy-nor-arginine (nor-NOHA; Caymen), 0.2 mM 1-methyl-tryptophan (1-MT; Sigma), 1000 U/mL catalase (Sigma), 200 U/mL superoxide dismutase (MP Biomedicals), 10 μg/mL anti-PD-L1 (CD274; Clone 10F.9G2; Biolegend), 10 μg/mL anti-PD-1 (CD279; Clone RMP1-14; Biolegend), 10 μg/mL anti–TGF-β1,2,3 (Clone 1D11; R&D Systems), 10 μg/mL anti-IFN-γ (Clone 37895.11; R&D Systems), 20 μg/mL anti–IL-10 (Clone JES5-2A5), 20 μg/mL anti–IL-10R/CD210 (Clone 1B1.3A). To assess contact dependence, assays used 0.2 μm transwell inserts (Costar), with Gr1+CD11b+ cells and responder T cells separated by membrane. Cells were cultured in standard media for 72 hours and analyzed by flow cytometry for CFSE dilution.
Measurements of Nitrite and IFN-γ.
NO production was determined by measuring nitrite.20 IFN-γ protein levels in plasma and in supernatants were determined by enzyme-linked immunosorbent assay (ELISA; eBiosciences).
Histology.
Liver hematoxylin and eosin staining was as described.9 Isolated CD11b+ cells were analyzed for cell morphology following cytospin centrifugation and Wright-Giemsa staining.
Statistics.
A Student t test was employed using GraphPad Prism, version 4.0. All bar graphs indicate mean ± standard deviation. Statistical significance is defined as P ≤ 0.05.
Results
Myeloid CD11b+Gr1+ Cells Rapidly Accumulate in Tgfb1−/− Mouse Livers.
Tgfb1−/− mice rapidly develop acute liver necroinflammation9 and a liver CD4+ T cell lymphocytosis.18 Liver damage requires CD4+ Th1 cells producing the cytokine IFN-γ.9, 18, 21 CD11b+ myeloid cells also are abundant in Tgfb1−/− liver,18 but have not been further studied at present. Histologic analysis confirmed the presence of cells with myeloid morphology in or apposed to necrotic areas ( Fig. 1A). We assessed the kinetics of accumulation of Gr1+ myeloid cells by flow cytometry. At postnatal days 4 and 7, Gr1+ cell numbers were equivalent between Tgfb1−/− livers and healthy littermate Tgfb1+/− livers. At postnatal day 11, Gr1+ cells were approximately three-fold more numerous in Tgfb1−/− livers (Fig. 1B). The rapid rise in Gr1+ cells closely paralleled the rise in CD4+ T cells (Fig. 1C). Gr1+ cells from 11-day-old Tgfb1−/− liver strongly coexpressed CD11b (Fig. 1D), as did liver resident Gr1+ cells from littermate Tgfb1+/− mice (Fig. 1D). Tgfb1−/− liver CD11b+ cells were heterogeneous, with both granulocytic forms and monocytic forms, and representative of various stages of lineage maturation (Fig. 1E).
Myeloid CD11b+Gr1+ cells accumulate in Tgfb1−/− livers. (A) Abundant myeloid cells (arrows) are observed in liver from day 11 Tgfb1−/− mice (−/−) but not age-matched Tgfb1+/− mice (+/−). Bar = 50 μm. (B,C) CD4+ T cells and Gr1+ cells were counted after isolation from livers from Tgfb1−/− mice and Tgfb1+/− mice at 4, 7, and 11 days of age. n = 3 or more mice per group. (D) Isolated liver Gr1+ cells from Tgfb1−/− mice and Tgfb1+/− mice coexpress CD11b. Filled histogram indicates isotype control. (E) After cytospin, Tgfb1−/− liver CD11b+ cells were stained with Wright-Giemsa. **P < 0.01; ***P < 0.0001.
Tgfb1−/−Liver CD11b+ Gr1+ Cells Potently Inhibit T Cell Proliferation.
We tested the hypothesis that Tgfb1−/− liver CD11b+ Gr1+ cells represent MDSCs by specifically assessing their ability to suppress T cell proliferation. Wild-type splenic CD4+ T cells were CFSE-labeled and stimulated in vitro with anti-CD3/28. Tgfb1−/− liver CD11b+Gr1+ cells suppressed the proliferation of T cells completely when added at either 3 × 105 or 1 × 105 cells per well ( Fig. 2A), and partially when added at 3 × 104 cells per well (data not shown). Control Tgfb1+/− liver CD11b+Gr1+ cells had no effect. Tgfb1−/− liver CD11b+ Gr1+ cells also suppressed proliferation of CD8+ T cells (Fig. 2B), and of effector Th1 cells (Fig. 2C), which is the cell type chiefly responsible for necroinflammation in the Tgfb1−/− mouse. Suppression was also observed with T cell stimulation mediated by cognate antigen, because Tgfb1−/− liver CD11b+Gr1+ cells suppressed antigen-presenting cell/ovalbumin (APC/OVA)-induced proliferation of DO11.10 CD4+ T cells (Fig. 2D). Control Tgfb1+/− liver CD11b+Gr1+ cells had no suppressor effects in any assay.
Tgfb1−/− liver CD11b+Gr1+ cells suppress T cell proliferation. CFSE-labeled T cells were stimulated in vitro. Some cultures included liver CD11b+Gr1+ cells from Tgfb1−/− mice or Tgfb1+/− mice. (A) CD4+ T cells or (B) CD8+ T cells were stimulated with anti-CD3ε/anti-CD28. (C) CFSE-labeled Th1 effector cells were stimulated with anti-CD3ε. (D) DO11.10 splenocytes were stimulated with OVA. Cells harvested at 72 hours were analyzed by flow cytometry. Gates were established using unstimulated CFSE-stained cells. Data represent three or more independent experiments.
Thus, Tgfb1−/− liver CD11b+Gr1+ cells are functional MDSCs that strongly suppress T cell receptor (TCR)-mediated T cell proliferation. The lack of similar activity in control Tgfb1+/− liver CD11b+Gr1+ cells demonstrates that the suppressor function is specific to inflamed liver, and not a general property of liver-resident CD11b+Gr1+ cells. Tgfb1−/− liver CD11b+Gr1+ cells exhibited higher expression of F4/80, CD11c, CD14, major histocompatibility complex class II, and PD-L1 (Supporting Fig. 1), supporting the conclusion that Tgfb1−/− liver CD11b+Gr1+ cells are distinct from control liver-resident CD11b+Gr1+ cells.
Tgfb1−/− Liver MDSC Suppression of T Cell Proliferation Is Dependent on NO, IFN-γ, and Cell-Cell Contact.
To assess the mechanism(s) of suppression, we carried out the suppression assay as before, blocking specific pathways individually. Specific inhibitors of arginase, indoleamine 2,3-dioxygenase, reactive oxygen species, PD-L1/PD-1, TGF-β, and IL-10 had no effect on Tgfb1−/− liver MDSC suppressor function (Table 1; data not shown). L-NMMA, an inhibitor of NO synthases, completely eliminated suppressor function, whereas the inactive enantiomer D-NMMA had no effect ( Fig. 3A; Table 1). Supporting these findings, nitrite levels in culture supernatants were significantly increased when Tgfb1−/− liver MDSCs were cocultured with stimulated T cells, but not when control CD11b+Gr1+ cells were used (Fig. 3B); as expected, nitrite production was suppressible by L-NMMA but not D-NMMA. L-NMMA inhibits all three isoforms of NO synthase (iNOS, neuronal NOS, and endothelial NOS). The iNOS-specific inhibitor L-NIL, similar to L-NMMA, abrogated suppression (Fig. 3C; Table 1). Flow cytometry confirmed iNOS expression in a subset of Tgfb1−/− liver CD11b+ cells, but not in Tgfb1+/− liver CD11b+ cells (Fig. 3D). Suppression was not observed when MDSCs and T cells were physically separated by a transwell membrane, indicating that cell-cell contact is required (Fig. 4A). The monoclonal antibody (mAb) neutralization of IFN-γ in vitro partly inhibited suppression (Fig. 4A). Additional studies clarified that cell-cell contact and IFN-γ are required for NO production, because nitrite was undetectable in the transwell assay, and significantly reduced with anti–IFN-γ (Fig. 4B). ELISA confirmed IFN-γ production in cocultures, albeit lower than in cultures of T cells stimulated alone (Fig. 4C). Because substantial IFN-γ was produced in cocultures of T cells and Tgfb1+/− liver CD11b+Gr1+ cells (Fig. 4C), IFN-γ appears insufficient to confer MDSC activity on liver-resident CD11b+Gr1+ cells. Indeed, when IFN-γ was added exogenously, Tgfb1+/− liver CD11b+Gr1+ cells were unable to inhibit T cell proliferation (Fig. 4D), and NO production was not augmented (Fig. 4B). Thus, IFN-γ is necessary but not sufficient for MDSC activity.
Suppression requires the production of nitric oxide. (A,C) Cocultures of CD4+ T cells and CD11b+ cells were set up as in the legend to Fig. 2; L-NMMA, D-NMMA, or L-NIL was added to some wells at the beginning of the assay. Flow plots represent at least three independent experiments. (B) Nitrite was measured in coculture supernatants of T cells and CD11b+ cells. ***P < 0.001. (D) Intracellular iNOS protein expression was determined by flow cytometry at 48 hours. Lipopolysaccharide/IFN-γ–treated RAW 264.7 cells served as a positive control for iNOS staining.
Suppression requires cell-cell contact and IFN-γ. (A) Cocultures were set up as before. In some wells, T cells were separated from CD11b+ cells by a 0.2 μm transwell membrane; in others, neutralizing anti–IFN-γ was included. (B,C) Nitrite (B) and IFN-γ (C) were measured in culture supernatants. *P < 0.05; **P < 0.01; ***P < 0.001. (D) Cocultures of T cells and Tgfb1+/− liver CD11b+ cells were established; some wells included exogenously added rIFN-γ. Data represent at least three independent experiments.
| Target Enzyme or Molecule | Inhibitor | Concentration | Effect on MDSC Activity |
|---|---|---|---|
| Nitric oxide synthase-1 (NOS-1), NOS-2, NOS-3 | L-NG-monomethyl arginine citrate | 5.0 mM | Blocks |
| NOS-2 | N6-(1-iminoethyl)-L-lysine | 0.5 mM | Blocks |
| Arginase | N-hydroxy-nor-arginine | 1.0 mM | No effect |
| Indoleamine 2,3-dioxygenase | 1-methyl-tryptophan | 0.2 mM | No effect |
| Reactive oxygen species | Catalase | 1000 U/mL | No effect |
| Superoxide dismutase | 200 U/mL | No effect | |
| PD-L1/PD-1 | Blocking anti–PD-L1 mAb | 10 μg/mL | No effect |
| Blocking anti–PD-1 mAb | 10 μg/mL | No effect | |
| TGF-β | Blocking anti–TGF-β-1, anti–TGF-β-2, anti–TGF-β-3 mAb | 10 μg/mL | No effect |
| IL-10/IL-10R | Blocking anti–IL-10 mAb | 20 μg/mL | No effect |
| Blocking anti–IL-10R mAb | 20 μg/mL | No effect |
Suppressive Activity Is Preferentially Found in the Monocytic CD11b+Ly6G−Ly6Chi Subset of Liver Tgfb1−/− CD11b+ Cells.
Anti-Gr1 recognizes two highly related cell surface proteins, Ly6C and Ly6G.22 Expression patterns of these cell surface proteins distinguish two major MDSC subsets, with the Ly6G−Ly6Chi phenotype characteristic of monocyte-like MDSCs and the Ly6GhiLy6Clo phenotype characteristic of granulocyte-like MDSCs.23 Most CD11b+ cells from Tgfb1−/− livers coexpressed Ly6C ( Fig. 5A). Among CD11b+Ly6C+ cells, approximately two-thirds were Ly6GhiLy6Clo, whereas the rest were Ly6G−Ly6Chi (Fig. 5A). After isolation of these subsets (Fig. 5B), we observed that suppressor activity resides exclusively in the “monocytic” CD11b+Ly6G-Ly6Chi cell population, with no activity found in the “granulocytic” CD11b+Ly6GhiLy6Clo cell population (Fig. 5C). NO production tracked with suppressor function (Fig. 5D).
Suppressor activity is in the Ly6G−Ly6Chi monocytic subset. (A) Isolated Tgfb1−/− liver mononuclear cells were stained for CD11b, Ly6C, and Ly6G. Gated CD11b+ cells were analyzed for expression of Ly6C and Ly6G. (B) CD11b+ cells were fractionated into Ly6GhiLy6Clo and Ly6G−Ly6Chi cell populations, and (C) placed in coculture with CFSE-labeled stimulated CD4+ T cells. (D) Nitrite was measured in supernatants. **P < 0.01; n.s., P not significant.
The Accumulation of CD11b+Gr1+ Cells in Tgfb1−/− Liver Is Dependent on CD4+ T Cells and IFN-γ.
The rapid accumulation of MDSCs parallels that of CD4+ T cells (Fig. 1). Therefore, we asked whether one cell type influences the accumulation of the other in vivo, by examining CD11b+Gr1+ cell accumulation at day 11 in livers of Tgfb1−/− mice rendered deficient either in all adaptive lymphocytes (Rag1−/−) or specifically in CD4+ T cells (anti-CD4 mAb). Neither Rag1−/−Tgfb1−/− mice nor anti-CD4–treated Tgfb1−/− mice exhibited an increase in liver CD11b+Gr1+ cells (Fig. 6A). Conversely, CD4+ T cells accumulated to high levels in Tgfb1−/− mice whether CD11b+Gr1+ cells had been depleted (anti-Gr1; Fig. 6B). Thus, CD4+ T cells are required for MDSC accumulation in Tgfb1−/− liver, whereas CD4+ T cells accumulate despite MDSC depletion.
Th1 cells drive MDSC accumulation in liver through IFN-γ. (A-D) Th1 cells drive MDSC accumulation in liver, not vice versa. (A,C) CD11b+Gr1+ cells, (B) CD4+ T cells, and (D) plasma IFN-γ were quantified on day 11 for mice of the genotypes (Tgfb1; Rag1; Ifng) and treatments (anti-CD4; anti-Gr1) indicated. n = 3 or more mice per group. Numbers represent P values for the comparisons indicated by lines. (B,D) Data are for Tgfb1−/− mice that were either untreated (“−”) or treated (“+”) with an anti-Gr1 mAb. (E) Liver CD11b+ cells from Ifng−/− Tgfb1−/− mice do not suppress T cell proliferation. Data aggregate four separate experiments.
We examined the role of IFN-γ in MDSC accumulation. Circulating plasma IFN-γ levels are highly elevated in Tgfb1−/− mice,21 IFN-γ is necessary for hepatocellular damage,9 and CD4+ T cells are the only significant source of IFN-γ in this model.21 The Ifng−/− Tgfb1−/− mice exhibited normal liver CD11b+Gr1+ cell numbers ( Fig. 6C). Conversely, depletion of CD11b+Gr1+ cells had no effect on plasma IFN-γ levels in Tgfb1−/− mice (Fig. 6D), which remained elevated. In addition, Ifng−/− Tgfb1−/− liver CD11b+ cells failed to suppress T cell proliferation in vitro (Fig. 6E). Thus, IFN-γ is essential both for the in vivo accumulation of CD11b+Gr1+ cells and for their in vitro suppressor function. Ifng−/− Tgfb1−/− livers and Ifng+/+ Tgfb1−/− livers exhibit equivalent accumulation of CD4+ T cells,21 indicating that the effects of IFN-γ on MDSC accumulation are not attributable to an indirect effect of this cytokine on CD4+ T cell accumulation.
Discussion
In Tgfb1−/− mice, a murine model of acute Th1-mediated hepatocellular injury, CD11b+Gr1+ cells accumulate in liver in response to the production of IFN-γ from CD4+ T cells. Tgfb1−/− liver CD11b+Gr1+ cells are potent MDSCs in vitro, producing NO to inhibit the proliferation of TCR-activated T cells. The production of IFN-γ is important for the development of the MDSC response at several junctures. First, IFN-γ is required for the accumulation of MDSCs in liver, which does not occur in Ifng−/− Tgfb1−/− mice; second, IFN-γ is required for full MDSC suppressor function, because inclusion of a neutralizing anti–IFN-γ mAb in coculture of MDSCs and T cells partially abrogates suppressor activity. These studies show that IFN-γ is necessary not only for hepatocellular injury but also for the development of the MDSC response. Thus, IFN-γ sits at a critical node of the liver immune response, responsible on one hand for T cell–mediated parenchymal damage and on the other hand for initiating an MDSC-mediated negative feedback pathway that can restrain T cell proliferation.
Murine liver schistosomiasis is a classic model of Th2-mediated inflammation, with granulomata forming around parasite eggs deposited in the liver.24, 25 Myeloid cells restrain granulomatous inflammation and fibrosis through activity of arginase,26 which acts by depleting T cells of the essential amino acid L-arginine. By contrast, inflammation and parenchymal damage in Tgfb1−/− mice is a “pure” Th1 phenomenon, dependent on the Th1 cytokine IFN-γ and independent of the Th2 cytokine IL-4.9 Thus, distinct types of inflammation induce distinct subsets of myeloid suppressor cells that act through subset-specific mechanisms. The association of iNOS with myeloid cells in Th1 responses and arginase with myeloid cells in Th2 responses is a recurring theme in inflammation,27 and the dichotomy appears applicable to liver inflammation as well.
An important aspect of our work is the demonstration that Th1 cells themselves are responsible for the accumulation of MDSCs in liver. Although it has been shown that IFN-γ can activate MDSCs,28, 29 to our knowledge, this is the first demonstration that IFN-γ from CD4+ T cells can drive MDSC accumulation to a site of inflammation. How might IFN-γ effect MDSC accumulation? Although IFN-γ might act directly, it is more likely that IFN-γ acts indirectly, inducing other cells (e.g. hepatocytes, endothelial cells, Kupffer cells) to secrete chemoattractants that in turn recruit MDSCs. Previous work shows that MDSCs accumulate at sites of inflammation in response to a number of inflammatory molecules. MDSCs isolated from hepatocellular carcinoma tumors in B6 mice express the chemokine (C-C motif) receptor 2 (CCR2) and migrate in vitro in response to the chemokine (C-C motif) ligand 2 (CCL2).30 Directly implicating the importance of this pathway, murine CCR2−/− MDSCs exhibit deficiencies in migration into hepatocellular carcinoma tumors,30 and, in another system, into the ovarian tumor microenvironment.31 The IL-1 response axis as well as proteins of the S100 family are important for MDSC accumulation in the tumor microenvironment.13, 32-34 Microarray analyses show that, at the messenger RNA level, in Tgfb1−/− liver, CCR2 and CCL2 are overexpressed ∼10-fold,35 IL-1β is overexpressed 17-fold,35 and various S100-encoding messenger RNAs are overexpressed 2-fold to 11-fold (unpublished data), but we have not yet tested whether any of these pathways is important for MDSC accumulation.
As discussed, unrestrained autoreactive Th1 responses in the liver likely contribute to the pathophysiologic basis of AIH, but the participation of cells of myeloid origin is currently unclear. It is known that populations of CD11b+ myeloid cells infiltrate the livers of patients with AIH,1 but functional analyses of these cells are lacking. Longhi et al.36 recently characterized peripheral blood monocytes from patients with AIH. Although they are surrogates for their intrahepatic counterparts, compared to circulating monocytes from healthy controls, circulating monocytes from patients with AIH are more numerous (with frequency correlating with AST), more spontaneously migratory, and express greater Toll-like receptor 4 and TNF-α.36 The authors suggested that “monocyte involvement in the liver damage [would] perpetuate the autoimmune attack.” However, this study did not examine iNOS expression or the production of NO, and did not test whether blood (or liver) monocytes from patients with AIH are capable of inhibiting T cell proliferation in vitro. Therefore, we offer an alternative possibility, that the activated myeloid/monocytic cell population in patients with AIH represents monocytic MDSCs recruited by activated T cells producing IFN-γ, with the potential, perhaps unrealized or somehow blocked, to inhibit T cell–mediated autoimmunity. Whether and how cells of myeloid origin participate in regulating inflammatory and/or autoimmune processes in the liver, and whether and how MDSCs may fail in their suppressor function, are important research questions in AIH and other inflammatory liver diseases.
Acknowledgements
We thank Drs. Mary Jo Turk, Edward Usherwood, and Jose Conejo-Garcia (all at Dartmouth Medical School) for, respectively, the GK1.5 antibody, the IL-10/IL-10R neutralizing antibodies, and the use of the microscope and related software, and Beverly Gorham and Christine Kretowicz for mouse breeding.
