Preemptive Donor Apoptotic Cell Infusions Induce IFN-γ–Producing Myeloid-Derived Suppressor Cells for Cardiac Allograft Protection

In clinical transplantation, lifelong immunosuppression is often necessary to control graft rejection but could be associated with life-threatening complications, including opportunistic infections, malignancies, and cardiovascular diseases (1). Therefore, to establish and maintain donor-specific transplant tolerance is highly desirable and remains an unmet need. Previous work in our laboratory has demonstrated that infusions of donor splenocytes (SPs) treated with the chemical cross-linker 1-ethyl-3-(3′-dimethylaminopropyl)-carbodiimide (ECDI) induce long-term donor-specific tolerance to islet allografts (2–4), and, when combined with a short course of rapamycin or anti-CD20, also to heart allografts and islet xenografts (5), respectively. A first in-human clinical trial based on the same principle using peptide-coupled autologous cells in patients with multiple sclerosis was recently published and established the clinical feasibility, tolerability, and safety of this novel tolerance strategy (6). Ongoing preclinical studies in nonhuman primate models of allogeneic and xenogeneic transplantation are attempting to establish the efficacy of this tolerance strategy for clinical transplantation. Mechanistically, treatment of splenocytes with ECDI induces their apoptosis and phagocytosis by recipient splenic phagocytes (3, 7), and the subsequent graft protection appears to be dependent on the presence of Gr1⁺ cells and the enzyme IDO (8). However, the direct effect of donor ECDI-SPs on the Gr1⁺ cell populations and the source of IDO in this model remain unclear.

Myeloid-derived suppressor cells (MDSCs) bearing the cell surface markers CD11b and Gr1 have been described in cancer and inflammation (9). Although a barrier to promoting antitumor responses, they have been shown to play an important role in establishing allograft tolerance (10, 11). Murine CD11b⁺Gr1⁺ cells can be subdivided into monocytic or granulocytic MDSCs according to their surface expression of the components of the Gr1 Ag, Ly6C and Ly6G, and are defined as CD11b⁺Ly6C^highLy6G⁻ and CD11b⁺Ly6C^INTLy6G^high cells, respectively (12). A number of conditions, including inflammation, infection, and tumors, can induce MDSCs [reviewed in (13)]. They suppress T cell proliferation in both Ag-dependent and independent manners through a variety of effector mechanisms, including NO, arginase, and reactive oxygen species (12, 14, 15); and they promote regulatory T cell (Treg) induction through production of IL-10, TGF-β, and IDO (16, 17).

There is also evidence that most MDSC subsets require IFN-γ, both for their induction and effector function, through IFN-γ–inducible genes such as inducible NO synthase (iNOS) (18). Studies in transplantation have demonstrated that IFN-γ can be protective via a variety of mechanisms, including mediating the suppressive function of MDSCs (10, 19–21). The potential sources of such IFN-γ are thought to be macrophages, T, NK, and/or NKT cells and can also result from phagocytosis of apoptotic cells in the spleen (22). More recent studies have suggested that immature myeloid cells expressing Gr1 can themselves produce IFN-γ after pathogen infection for host protection (23).

In this article, we show that infusions of apoptotic donor ECDI-SPs induce an increase in two cell populations that resemble monocytic and granulocytic MDSCs. These cells are distinct in their phenotype and function and protect cardiac allografts in part through IFN-γ production themselves and subsequent IFN-γ–mediated pathways of T cell suppression. Our study thus describes a novel therapeutic approach for the manipulation of these suppressor cell populations to facilitate the development of transplant tolerance.

Eight- to 10-wk-old male BALB/c (H-2^d), C57BL/6 (H-2^b), and SJL/J (H-2^s) mice were purchased from The Jackson Laboratory and Harlan. All mice were housed under specific pathogen-free conditions at Northwestern University, and protocols were approved by Northwestern University Institutional Animal Care and Use Committee.

Donor (BALB/c) splenocytes were treated with ECDI, as previously described (2). Briefly, spleens were processed into single-cell suspensions, and erythrocytes were lysed. Splenocytes were incubated with ECDI (Calbiochem; 150 mg/ml per 3.2 × 10⁸ cells) on ice for 1 h, followed by washing. A total of 10⁸ cells was injected i.v. into recipient C57BL/6 (B6) mice on days −7 and +1 with reference to transplantation on day 0.

Abdominal heart transplantation was performed, as previously described (24). BALB/c (donor, H-2^d) to C57BL/6 (recipient, H-2^b) combination was used. In some cases, SJL/J mice (H-2^s) were used as donor for third-party heart allografts. Briefly, the donor heart was excised en bloc via median sternotomy. The ascending aorta and pulmonary artery of the donor were anastomosed end to side to the recipient abdominal aorta and inferior vena cava, respectively. Direct abdominal palpation of the heart beating was used to assess graft survival. Rejection is defined by loss of palpable cardiac impulses.

In experiments in which cardiac recipient B6 mice were treated with the anti-Gr1 Ab, the Ab (clone RB6-8C5; BioXCell) was injected i.p. at a dose of 200 μg on day −8, followed by 100 μg on days −7, −3, −1, and +1 with reference to transplantation on day 0 (10, 25–27).

For isolating cells from the spleens and the cardiac allografts, animals were sacrificed, and spleens and cardiac grafts were excised and digested at 37°C with collagenase type IV (Worthington) for 25 min, followed by lysis of erythrocytes. For graft cells, leukocytes were further enriched by density centrifugation (Lympholyte; Cedarlane) to remove the myocytes. For functional assays, Gr1⁺ cells were enriched by negative selection using the following biotinylated Abs (clones): B220 (RA3-6B2), CD4 (GK1.5), CD8α (53-6.7), Ter¹¹⁹ (erythroid cells), CD49b (DX5), and CD25 (7D4), and streptavidin-conjugated magnetic beads, according to manufacturer’s instructions (Miltenyi Biotec). Cell purity was determined by flow cytometry and ranged from 70 to 90%. For indicated assays, Gr1^high and Ly6C^high cells were further purified by FACS to a purity of >98%. All cell cultures were performed in RPMI 1640 (Life Technologies) supplemented with 10% FCS (Life Technologies), 1% PenStrep, 1% L-glutamine (Life Technologies), 1% HEPES (Lonza), 0.5% gentamicin (Life Technologies), and 0.1% 2-ME (Life Technologies). Splenic CD4⁺ or CD8⁺ T cells were isolated by negative selection first using the following biotinylated Abs (clones): CD11b (M1/70), Gr1 (RB6-8C5), Ter¹¹⁹, CD49b (DX5), and B220 (RA3-8B2), followed by positive selection for CD4 or CD8.

Splenic CD8⁺ T cells were purified, as described above, labeled with 0.5 mM CFSE (Life Technologies), and plated at 1 × 10⁵ per well as responder cells for in vitro stimulation. T cells were activated by either anti-CD3/28 Dynabeads per manufacturer’s instructions (Invitrogen), or 5 × 10⁵ irradiated donor (BALB/c) APCs. FACS-sorted suppressor cells were added at a 1:1 ratio with the CFSE-labeled T responder cells and incubated at 37°C for 96 h. T cell proliferation was measured by CFSE dilution. For indicated studies, FACS-sorted suppressor cells were either pretreated at room temperature for 30 min with 10 μg/ml anti–IFN-γ (clone XMG1.2; BioXCell) prior to addition to the proliferation assays, or added to the proliferation assays in the presence of 5 mM L-NMMA or D-NMMA (Cayman Chemical,) or 2 mM 1-methyl-DL-tryptophan (1-MT; Sigma-Aldrich) or vehicle (2% carboxymethylcellulose). For suppression assays by graft Gr1^high and Ly6C^high cells, CFSE-labeled responder CD8⁺ T cells were plated at 1 × 10⁴ per well and cocultured with 1 × 10⁴ anti-CD3/28 Dynabeads or 5 × 10⁴ BALB/c APCs and 1 × 10⁴ FACS-sorted suppressor cells from the graft. T cell proliferation was determined by CFSE dilution after 96 h.

Cells were stained with fluorochrome-conjugated Abs for 30 min on ice, washed, read on the Canto II (BD Biosciences), and analyzed using FlowJo v6.4.7 (Tree Star). For intracellular staining, cells were also fixed and permeabilized after surface staining using Cytofix/Cytoperm buffers, according to manufacturer’s instructions (BD Biosciences), and stained with fluorochrome-conjugated Abs for cytokine detection. The following Abs (clones) were used: Gr1-PE (RB6-8C5), CD11c-allophycocyanin (HL3), and CD80-FITC (16-10A1), all from BD Biosciences; Ly6C-eFluor450 (HK1.4), CD11b-eFluor780 (M1/70), F4/80-PerCPCy5.5 (BM8), MHCII-PeCy7 (MS/114.15.2), IL-12-PerCPCy5.5 (C17.8), IL-10-FITC (Jes5-16E3), IFN-γ-PeCy7 (XMG1.2), CD4-eFluor450 (GK1.5), and CD8-PerCPCy5.5 (53-6.7), all from eBioscience; Ly6G-PeCy7 (1A8) from BioLegend; and CCR2- allophycocyanin (475301) from R&D Systems. For annexin V staining, cells were incubated with allophycocyanin-conjugated annexin V (1:20, eBioscience) for 10 min at room temperature, followed by immediate analysis by flow cytometry.

Tissue cytokines were analyzed by 32-Plex multiplex assays (Millipore). Tissues were homogenized to obtain cell lysates and centrifuged at 13,000 rpm for 2 min, and the soluble portion was collected and analyzed by the multiplex assays per manufacturer’s instructions. Results were normalized to the amount of total protein as measured by the Bradford assay (Pierce Biotechnology).

Total RNA was extracted using the RNeasy kit (Qiagen), according to manufacturer’s instructions. Total RNA was reverse transcribed to cDNA using the High Capacity RNA-to-cDNA kit (Applied Biosystems). RT-PCR amplifications were performed using TaqMan Universal Master Mix II and TaqMan gene expression assays (Applied Biosystems). The reactions were run at 50°C for 2 min, followed by 95°C for 10 min and 40 cycles of 95°C for 15 s, and 60°C for 1 min. Reactions were run on the 7500 Real Time PCR System, and data were analyzed using 7500 v2.0.1 software (Applied Biosystems). The Δ cycle threshold values for each duplicate sample were calculated with reference to 18S.

Grafts were snap frozen in OCT compound with liquid nitrogen. All sections were 8 μm thick. Frozen sections were blocked with avidin/biotin blocking kit (Vector Laboratories), followed by staining with anti-mouse Foxp3 mAb (1:400, rat IgG2a, κ clone FJK-16s; eBioscience) or anti-mouse CD8 (1:250, rat IgG2a, κ clone 53-6.7; BD Biosciences). Samples were then stained with biotinylated goat anti-rat Ig for Foxp3 (1:200, goat Ig clone polyclonal; BD Biosciences) or biotin-SP-AffiniPure donkey anti-rat Ig for CD8 (1:250; Jackson ImmunoResearch Laboratories). Visualization of Foxp3 and CD8 was performed with Vectastain ABC kit (Vector Laboratories) and diaminobenzidine substrate kit (BD Biosciences).

Significance between groups was calculated by Student t tests, one-way ANOVA, or the Wilcoxon rank sum test, as appropriate. A p value <0.05 was considered significant. All statistical analysis was performed using Prism v5.0 (GraphPad).

To determine whether allogeneic ECDI-SP infusions could expand cells bearing features of MDSCs, naive B6 mice were injected with either one or two doses of BALB/c ECDI-SPs, and the spleens were harvested for phenotypic analysis for CD11b⁺Gr1⁺ cells on the indicated day (Fig. 1A). Clone RB6-8C5 was used to stain for Gr1. As cells expressing the Gr1 Ag can be subdivided according to their relative expression of Ly6C, a component of the Gr1 Ag (28), clone HK1.4, was used to further stain for Ly6C. As shown in Fig. 1B and 1C, one dose of BALB/c ECDI-SPs induced a significant increase in the percentage as well as the total numbers of the CD11b⁺Gr1^INTLy6C^high (hereafter referred to as Ly6C^high) cells and the CD11b⁺Gr1^highLy6C^INT (hereafter referred to as Gr1^high) cells. The total number of the Gr1^high cells was further increased after an additional dose of BALB/c ECDI-SPs (Fig. 1C). In contrast, injection of syngeneic B6 ECDI-SPs did not significantly increase either population (Fig. 1B, 1C). Phenotypic analysis (Fig. 1D) revealed that the Ly6C^high cells expressed CCR2 and F4/80, a marker profile consistent with that of monocytic-like MDSCs, whereas the Gr1^high cells expressed Ly6G (stained with clone 1A8) and exhibited a high side scatter resembling granulocytic-like MDSCs (Fig. 1D). Other markers associated with the MDSC phenotype such as CD124 and CD115 were also measured, but were not expressed by either population (data not shown). Comparison of cells from naive untreated versus BALB/c ECDI-SP–treated animals revealed that markers of these two populations were not changed by a single injection of BALB/c ECDI-SPs (data not shown) or two injections of BALB/c ECDI-SPs (Fig. 1D). Collectively, these data suggest that treatment with allogeneic ECDI-SPs expands two cell populations in the spleen that resemble monocytic- and granulocytic-like MDSCs.

To further characterize ECDI-SP–expanded Ly6C^high and Gr1^high cells and to determine their similarities to MDSC populations, in vitro T cell suppression assays were performed. In this study, we primarily focused on CD8⁺ T cells, as they are the dominant effector cells in the graft rejection in our model (8). As shown in Fig. 2A, both subsets significantly inhibited CD8⁺ T cell proliferation in response to anti-CD3/28 stimulation, with the Gr1^high cells suppressed more efficiently. The unexpanded Ly6C^high and Gr1^high cells from naive untreated mice were also tested and showed a similar pattern of suppression of CD8⁺ T cell proliferation (Fig. 2A, right bar graph), albeit slightly less potently than the expanded Ly6C^high and Gr1^high cells from ECDI-SP–treated mice. MDSCs are known to exert their effects in part through cytokine production (29). We next examined the cytokine profiles of the ECDI-SP–expanded Ly6C^high and Gr1^high cells after two infusions of ECDI-SPs. Neither IL-12 nor IL-10 was produced by either population (data not shown). Interestingly, IFN-γ was produced by the Gr1^high cells (Fig. 2B), and they were the predominant source of IFN-γ within the CD11b⁺ fraction (Fig. 2C). Furthermore, this IFN-γ production was not detected in Gr1^high cells from naive untreated mice (Fig. 2B, 2D), was expressed at a very low level after one dose of ECDI-SPs, but increased significantly after the second dose of ECDI-SP (Fig. 2D).

We have previously demonstrated that pharmacological inhibition of IDO resulted in the loss of protection of cardiac allografts provided by infusions of donor ECDI-SPs, and that depletion of Gr1⁺ cells in the same model was concomitant with a reduction of IDO⁺ cells in the cardiac allografts (8). However, the exact source of IDO remained unclear. IDO is known to be an IFN-γ–inducible gene in other cell populations such as dendritic cells (30). Given that ECDI-SPs induce Gr1^high cells to produce IFN-γ, we next sought to determine whether IDO could be produced by these Gr1^high IFN-γ⁺ cells. CD11b⁺Gr1^high cells from untreated mice or mice treated with ECDI-SPs were FACS sorted to a purity of >98%. Quantitative RT-PCR analysis of this cell population revealed that the Gr1^high cells from untreated animals did not express ido, but they did modestly increase their ido expression after incubation with rIFN-γ (p = 0.008). However, after treatment with two doses of ECDI-SPs, not only was the baseline ido expression significantly higher than that of the control cells, there was a further significant increase after IFN-γ stimulation (p = 0.039) (Fig. 3A). iNOS is another IFN-γ–inducible gene and a known effector molecule of MDSCs. As shown in Fig. 3B, in response to IFN-γ stimulation, the Gr1^high cells from ECDI-SP–treated mice showed a significant increase in the expression of nos2, the murine equivalent of inos (p = 0.033). This increase in response to IFN-γ was not observed in the Gr1^high cells from untreated animals (Fig. 3B). Expression of arginase 1 and heme oxygenase-1, two additional molecules associated with MDSC function, was similarly examined but did not appear to be induced by treatment with either ECDI-SPs or IFN-γ stimulation (data not shown). Collectively, these data indicate that, compared with Gr1^high cells from naive untreated mice, ECDI-SP–expanded Gr1^high cells are induced to produce IFN-γ themselves and have a further enhanced capacity to upregulate the expression of IFN-γ–inducible genes. These characteristics may therefore form a positive feedback loop to enhance their suppressive function in vivo in addition to their increase in numbers by the ECDI-SP treatment.

To determine whether IDO, iNOS, or IFN-γ played a role in the suppression mediated by the Gr1^high expanded by ECDI-SP treatment, we performed similar suppression assays, as described in Fig. 2. Addition of pharmacological inhibitors to IDO (1-MT) or iNOS (L-NMMA) to the suppression assays partially reversed Gr1^high cell–mediated suppression (Fig. 3C). Inhibition of arginase 1 did not affect T cell suppression (data not shown), consistent with the lack of its expression, as determined by qRT-PCR. Most strikingly, pretreatment of Gr1^high cells with an anti–IFN-γ–neutralizing Ab completely abrogated their suppressive function (p < 0.0001) (Fig. 3D). Collectively, these data indicate that ECDI-SP–expanded Gr1^high cells critically depend on the presence of IFN-γ and its downstream effector molecules to exert their suppressive function.

To determine how the Ly6C^high and Gr1^high cells behave after cardiac transplantation in recipients with or without ECDI-SP infusions, allogeneic heterotopic cardiac transplants were performed. Spleens and grafts from untreated control or ECDI-SP–treated animals were harvested 7 and 21 d after transplantation. In the spleen (Fig. 4A), Ly6C^high cells significantly increased in percentages and total numbers at day 7 in transplant recipients compared with untransplanted recipients shown in Fig. 1B and 1C. This increase was similar in both untreated control (white bars) and ECDI-SP–treated recipients (black bars) and sustained at day 21. Similarly, Gr1^high cells also significantly increased in percentages and numbers at day 7 compared with untransplanted recipients shown in Fig. 1B and 1C, but this increase appeared to be more profound in untreated recipients (white bars) compared with ECDI-SP–treated recipients (black bars) at day 7 (p = 0.059) and subsided in both to baseline at day 21 (Fig. 4A). Interestingly, the splenic Gr1^high cells from ECDI-SP–treated recipients were no longer positive for IFN-γ expression (Fig. 4A, right histograms).

In the cardiac allografts, however, a very different picture emerged (Fig. 4B), as follows: Ly6C^high cells were found in considerably higher percentages and total numbers in the protected grafts from ECDI-SP–treated recipients compared with grafts from untreated recipients at day 7 (Fig. 4B). In contrast, the percentages and total numbers of Gr1^high cells were significantly lower in the protected grafts in ECDI-SP–treated recipients compared with rejecting/rejected grafts from control untreated recipients at day 7. Importantly, the Gr1^high cells from allografts of ECDI-SP–treated recipients continued to express IFN-γ at both day 7 and day 21 (Fig. 4B, right histograms), contrasting to the Gr1^high cells from allografts of untreated recipients that were IFN-γ⁻. By day 21, the already rejected cardiac grafts from control untreated recipients were progressing to fibrosis, and considerably fewer graft-infiltrating cells, including Ly6C^high cells and Gr1^high cells, were recovered compared with the protected grafts from ECDI-SP–treated recipients.

We next measured cytokine and chemokine production of the graft tissue to determine whether there was a correlation between cell infiltration, cytokine/chemokine production, and graft outcome. Tissue lysates were prepared from day 7 and day 21 heart allografts. Multiplex analysis of heart graft tissue lysates revealed that IL-1β and M-CSF, factors important for MDSC induction, were significantly increased in the protected grafts compared with control grafts at day 21. IL-10, a factor known to be produced by MDSCs, was also significantly higher in protected grafts at day 21 compared with control grafts. CCL4, another soluble factor implicated in the recruitment of Tregs, was similarly increased in protected grafts at days 7 and 21 compared with control grafts. In contrast, factors associated with neutrophil accumulation, including G-CSF, CXCL1, and CXCL2, were significantly higher in control grafts at day 21 (Fig. 4C).

Collectively, these data indicate that the Gr1^high cells induced by ECDI-SP infusions and those induced by allogeneic cardiac transplantation may be distinct from each other, and that cardiac allografts from recipients treated with ECDI-SPs show enhanced accumulation of Gr1^highIFN-γ⁺ cells, as well as Ly6C^high cells, and a graft environment consistent with their local induction and accumulation.

To determine the functional significance of the Ly6C^high and Gr1^high cells from the protected cardiac allografts in recipients treated with donor ECDI-SPs, we first examined their ability to suppress T cell proliferation. We chose to examine this at day 21 posttransplant, a time point when the fate of the graft in ECDI-SP–treated recipients was readily distinguishable from that of the graft in control untreated recipients. As shown in Fig. 5A, left panel, neither population isolated from the cardiac allografts on day 21 suppressed in vitro CD8⁺ T cell proliferation stimulated by anti-CD3/28. MDSCs have been reported to also have the ability to exert Ag-specific suppression of T cell proliferation (12, 31). We next tested whether these cells suppressed CD8⁺ T cell proliferation stimulated by donor APCs. As shown in Fig. 5A, right panel, they also did not substantially suppress donor APC-stimulated CD8⁺ T cell proliferation. As a control, in vitro suppression assays using splenic Ly6C^high and Gr1^high cells from the protected recipients on day 21 were also set up and showed similar inhibition of proliferation, as observed in Fig. 2A (Supplemental Fig. 1). Interestingly, when we examined cytokine production in the supernatants from the suppression cultures by multiplex analysis, we found that coculture with allograft, but not splenic, Ly6C^high and Gr1^high cells led to a significantly higher production of IL-10 and CCL4, two soluble mediators implicated in the induction and homing of Tregs (Fig. 5B). Supporting this possibility, cardiac allografts retrieved from donor ECDI-SP–treated recipients showed a progressive increase of the number of Foxp3⁺ cells compared with those from control untreated recipients (Fig. 5C). This characteristic was unique to graft Ly6C^high and Gr1^high cells, as spleen Ly6C^high and Gr1^high cells from the same recipient mice in the same cocultures did not lead to an increased production of IL-10 or CCL4 (Fig. 5B). Collectively, these data indicate that posttransplant the graft, but not the splenic, Ly6C^high and Gr1^high cells may function by establishing an environment conducive to local Treg induction and homing, pointing to the functional difference between the graft and the splenic Ly6C^high and Gr1^high cells.

To test the in vivo protective role of the Ly6C^high and Gr1^high cells, we treated mice with the Gr1-depleting Ab (clone RB6-8C5). As shown in Fig. 6A, this Ab depleted both Ly6C^high cells and Gr1^high cells in vivo. In this study, the Gr1^high cells were shown as SCC^highLy6C^INT cells (as shown in Fig. 1B, 1D) to avoid the use of the same clone (clone RB6-8C5) for cell surface staining. Given that these cells suppress CD8⁺ T cells ex vivo (Fig. 2A), we sought to confirm these findings in vivo. As schematically outlined in Fig. 6B, B6 mice were treated with BALB/c ECDI-SPs and anti-Gr1 Ab, and BALB/c cardiac transplants were performed. Consistent with our previous findings (8), in sharp contrast to recipients treated with ECDI-SPs alone, all recipients additionally treated with the anti-Gr1 Ab promptly rejected the cardiac allograft by day 10 (Fig. 6B). Of note, injection of BALB/c ECDI-SPs did not protect third-party SJL cardiac allografts (Fig. 6B), consistent with our previous demonstrations of the Ag specificity of this regimen (2, 5, 8). At day 10, grafts were harvested and graft-infiltrating cells were analyzed by histology and by flow cytometry. It was observed that allografts from recipients treated with ECDI-SP infusions only had preserved graft architecture, overall few graft-infiltrating cells (Fig. 6C, top image), and specifically few CD8⁺ T cells (Fig. 6D, left image), consistent with graft protection. However, with concomitant anti-Gr1 Ab treatment, there was a discernable loss of graft integrity with a significant increase in overall graft-infiltrating cells (Fig. 6C, bottom image) and specifically graft-infiltrating CD8⁺ T cells (p = 0.043) (Fig. 6D, right image). These data suggest that the presence of Ly6C^high and Gr1^high cells is critical for the protection of the cardiac allograft, possibly through their control of graft-infitrating CD8⁺ T cells.

In the current study, we demonstrate that infusions of apoptotic donor cells in the form of ECDI-SPs expand two myeloid suppressor populations (Ly6C^high and Gr1^high) that function to inhibit T cell proliferation. Moreover, treatment with ECDI-SPs induces the Gr1^high population to produce a high level of IFN-γ and enhances the responsiveness of this population to IFN-γ to produce downstream effector molecules, including IDO and iNOS, which mediate their ability to suppress T cell proliferation. Finally, these cells traffic to and protect cardiac allografts through inhibiting graft CD8 T cell accumulation, and potentially also through inducing a Treg-rich environment.

Phenotypic analysis of Ly6C^high cells expanded in our model revealed similarities with monocytic MDSCs (M-MDSCs); however, this population was distinct in its absence of expression of the IL-4R (CD124) and the M-CSFR (CD115) (data not shown). Cell subset heterogeneity is a hallmark of MDSCs, with cell surface phenotype dependent on the model, stimulus, and anatomic location. M-MDSCs can be further classified as M1 or M2, dependent on the polarizing conditions. M1 M-MDSCs are defined by their production of iNOS and TNF-α and expression of IFNγR, whereas M2 M-MDSCs produce arginase and IL-10 and express CCR2 (32). As shown in Fig. 1D, ECDI-SP–expanded M-MDSCs appear to exhibit a M2-like phenocyte by their expressing CCR2. The function of these Ly6C^high cells also appears to correspond to that of M2 M-MDSCs in their ability to suppress CD8 effector T cells and produce factors for Treg induction (Figs. 2A, 5B) (32).

ECDI-SP–expanded Gr1^high cells are Ly6G^high, Ly6C^INT, SSC^high, and CCR2⁻ (Fig. 1B, 1D). This phenotype is indicative of neutrophil-like cells or granulocytic MDSCs (G-MDSCs). Several recent studies have attempted to phenotypically and functionally differentiate between neutrophils, G-MDSCs, and a further population of neutrophil-like cells termed tumor-associated neutrophils (TAN) (33, 34). G-MDSCs can be subcategorized according to their phenotype, with G2 G-MDSC expressing arginase, IL-10, and CCL2 (32). Graft-infiltrating Gr1^high cells in the ECDI-SP–treated cardiac transplant recipients appeared to have an enhanced ability to induce IL-10 production in comparison with their splenic counterpart (Fig. 5B), indicating that the graft environment may skew the Gr1^high phenotype to resemble that of G2 G-MDSCs, which mirrors that of TAN from tumor-bearing mice (35). Interestingly, the TAN phenotype has been shown to be dependent on tumor TGF-β, which may also perpetuate Treg development in conjunction with IL-10; and depletion of TAN resulted in a significant increase in local CD8⁺ T cell proliferation (35). In contrast, bone marrow neutrophils did not exhibit these characteristics (34, 35). Collectively, our data indicate that ECDI-SP–expanded Gr1^high cells resemble G-MDSC or TAN rather than bone marrow neutrophils, a phenotype that may be further perpetuated by the graft microenvironment.

Flow cytometric analysis of Gr1^high cells revealed that these cells produced a high level of IFN-γ after repeated infusions of ECDI-SPs (Fig. 2C, 2D). Although not yet reported in transplant models, CD11b⁺Gr1⁺ cells producing IFN-γ have been previously demonstrated in models of infection. In 2002, two groups demonstrated that the immune response to Nocardia asteroides and Salmonella infections induced CD11b⁺Gr1⁺ cells to produce IFN-γ (36, 37). This was corroborated recently in a Streptococcus infection model, in which phenotypic analysis demonstrated that these cells were CD11b⁺Gr1⁺, and Ly6G⁺, Ly6C^low, CCR2⁻, and CX3CR1⁺ (23). Interestingly, this expression profile closely resembles that of ECDI-SP–expanded Gr1^high cells. In each of these studies, the IFN-γ–producing cells were defined either as neutrophils (36, 37) or as immature myeloid cells based on their induced differentiation with addition of growth factors (23). However, only the immature myeloid cells could suppress T cell responses (23). Therefore, although more than one subtype of CD11b⁺Gr1⁺ cells may be capable of producing IFN-γ it is clear that not all may be suppressive.

It is interesting to note that our previous attempts to induce transplant tolerance to islet allografts using donor ECDI-SPs were unsuccessful in IFN-γ–deficient mice (3), indicating an obligatory role of IFN-γ in mediating tolerance induced by the ECDI-SP–based regimen. We have previously shown that IFN-γ contributes to the contraction and deletion of allospecific T cells (3). Our data in this study provide an additional role of IFN-γ in graft protection by promoting and perpetuating the function of MDSCs, in part via IDO and iNOS-mediated suppression (12, 38). These findings indicate the complex role that IFN-γ may play in transplantation tolerance, consistent with studies by others demonstrating a protective role of IFN-γ in transplant models via a variety of mechanisms. For example, exogenous IFN-γ prevented graft-versus-host disease in a murine bone marrow transplantation model through posttransplant inhibition of T cell proliferation (19, 20). IFN-γ stimulation of dendritic cells resulted in an upregulation of IDO production, which mediated allograft protection in a rat liver transplantation model (21). Most relevant to our findings, a recent study by Garcia et al. (10) in a tolerant heart transplant model demonstrated that a monocytic suppressive cell population bearing the phenotype CD11b⁺Gr1⁺CD115⁺ mediated their suppression via signaling responses to IFN-γ, as Ifngr^−/− Gr1⁺ monocytes could not provide graft protection as wild-type Gr1⁺ cells did. Interestingly, the source of the IFN-γ was not the Gr1⁺ monocyte themselves, as Ifng^−/− Gr1⁺ monocytes were equally efficacious in graft protection as wild-type Gr1⁺ cells (10). Therefore, to the best of our knowledge, ours is the first report of IFN-γ production by MDSCs themselves potentially mediating the induction of transplant tolerance.

It is unclear at the moment what signals are transmitted to these cells by the injection of ECDI-SPs to allow the induction of IFN-γ in the Gr1^high cells, as these cells themselves do not directly interact and phagocyse the injected ECDI-SPs (J. Bryant and X. Luo, unpublished observations). On a broader base, determining the exact mechanism through which ECDI-SPs induce the expansion of both the Ly6C^high and Gr1^high cells is of clinical relevance, as it will potentially allow for a more direct and effective manipulation of this response therapeutically. The pathways through which MDSCs are differentiated and expanded have been extensively studied in other model systems. Inflammatory mediators arising from infections and tumors have been implicated by many studies, including IL-1β, PGE₂, cyclooxygenase-2, GM-CSF, IL-6, vascular endothelial growth factor, S100A8/9, and IFN-γ (13). Interestingly, the process of apoptotic cell uptake (efferocytosis) also induces PGE₂ production (39). Apoptotic ECDI-SPs are phagocytosed in the recipient spleens by subsets of dendritic cells (3), which may generate subsequent mediators associated with MDSC differentiation or expansion. This may provide a link between efferocytosis and the expansion of Ly6C^high and Gr1^high cells in our model. Further studies are ongoing to delineate this relationship.

Previous work in our laboratory demonstrated that IDO was critical for cardiac allograft tolerance after ECDI-SP infusions (8). In the current study, we demonstrate that this molecule can be produced by the Gr1^high cells and it mediates the T cell–suppressive function of the Gr1^high cells. Although the role of IDO in positive allograft outcome has been extensively studied, its production by MDSCs has not been fully investigated. Two recent studies in human tumor MDSCs showed that IDO was directly produced by these MDSCs, and that it played a role in T cell suppression and Treg induction by this cell population (40, 41). Additionally, in allogeneic stem cell transplant recipients, monocytic MDSCs were also shown to produce IDO, playing a role in T cell suppression (42). Our studies demonstrate that infusions of ECDI-SPs induce a MDSC population with a heightened ability for IDO production in response to IFN-γ, resulting in a positive graft outcome via both mediators through CD8⁺ T cell suppression and potentially Treg induction. Collectively, ours and published data, demonstrating IDO production by MDSCs provides a link between T cell suppression and Treg induction by the same cell population, therefore point to an attractive target for therapeutic exploitation in transplantation.

Characterizing the ECDI-SP–expanded MDSCs before (Fig. 1C) and after (Fig. 4A, 4B) cardiac transplantation suggests that, prior to transplantation, these cells are present in the spleen and function to suppress proliferation of T cells via IFN-γ–dependent mechanisms, whereas, after transplantation, they migrate to the cardiac allograft and protect the graft via possible induction and homing of Tregs. Migration of MDSCs has been best examined in tumors and is thought to be mediated by multiple tumor-secreted factors, including IL-1β (43) and M-CSF (17), two factors that we have also observed in our protected cardiac allografts (Fig. 4C). Furthermore, it has been well recognized that functional differences exist between the MDSCs isolated from the spleen and the MDSCs isolated from the site of tumors (44–46). Similarly, migration of MDSCs to transplanted allografts has also been described (10), although there the expression of the ligands for P- and E-selectins on the MDSCs was found to be crucial for the homing. Thus, findings of the spleen versus the graft MDSCs in our model are consistent with these notions. However, the exact signals necessary for their trafficking to the allograft and the basis for their functional evolution warrant further investigation.

In summary, we demonstrate that ECDI-SPs expand MDSCs that function to protect transplant allografts paritally through their own production of IFN-γ and its downstream effector molecules. These studies thus expand the recent advances in the understanding of regulatory myeloid cells in transplantation (47) and provide the basis for future therapeutic manipulations of the population size and functionality of MDSCs for the purpose of transplant tolerance induction.

We acknowledge the Northwestern University Interdepartmental ImmunoBiology Flow Cytometry Core Facility for support of this work.

The authors have no financial conflicts of interest.