Endotoxin-Induced Myeloid-Derived Suppressor Cells Inhibit Alloimmune Responses via Heme Oxygenase-1


* Corresponding author: Alain Le Moine, alemoine@ulb.ac.be


Inflammation and cancer are associated with impairment of T-cell responses by a heterogeneous population of myeloid-derived suppressor cells (MDSCs) coexpressing CD11b and GR-1 antigens. MDSCs have been recently implicated in costimulation blockade-induced transplantation tolerance in rats, which was under the control of inducible NO synthase (iNOS). Herein, we describe CD11b+GR-1+MDSC-compatible cells appearing after repetitive injections of lipopolysaccharide (LPS) using a unique mechanism of suppression. These cells suppressed T-cell proliferation and Th1 and Th2 cytokine production in both mixed lymphocyte reaction and polyclonal stimulation assays. Transfer of CD11b+ cells from LPS-treated mice in untreated recipients significantly prolonged skin allograft survival. They produced large amounts of IL-10 and expressed heme oxygenase-1 (HO-1), a stress-responsive enzyme endowed with immunoregulatory and cytoprotective properties not previously associated with MDSC activity. HO-1 inhibition by the specific inhibitor, SnPP, completely abolished T-cell suppression and IL-10 production. In contrast, neither iNOS nor arginase 1 inhibition did affect suppression. Importantly, HO-1 inhibition before CD11b+ cell transfer prevented the delay of allograft rejection revealing a new MDSC-associated suppressor mechanism relevant for transplantation.


T-cell responses can be suppressed during sepsis, viral infection and trauma or in the context of chronic inflammation such as autoimmune diseases, infections cancer and transplantation (1–14). Myeloid-derived suppressor cells (MDSCs) coexpressing CD11b and GR-1 antigens have been shown to be central in the downregulation of T-cell immunity in those contexts (15,16). MDSCs are heterogeneous populations including macrophages, neutrophils and dendritic cells (12,15,16). In cancer patients, this is considered a mechanism by which tumors escape immune surveillance (12). Experimental models indicated that tumor-associated MDSCs mediate T-cell suppression through the activation of two enzymes, the inducible nitric oxide synthase (iNOS) and arginase 1 in an IFN-γ or IL-4/IL-13-dependent manner, respectively (17). Both enzymes catabolize l-arginine leading to the production of nitric oxide (NO) for iNOS and superoxide for arginase 1 (11,18). However, the mediator of suppressor functions by other subtypes of MDSCs such as immature neutrophils still remains unknown (10).

In line with recent observations made by Vaknin et al. (4), repeated injections of LPS were administered to elicit the emergence of MDSCs, using a protocol known to induce endotoxin tolerance (see ‘Material and Methods’). We assessed the capacity of these LPS-induced MDSCs to suppress allogeneic and polyclonal T-cell responses in vitro, and their potential to control skin allograft rejection. In addition, we describe an unsuspected mechanism by which these MDSCs mediate suppression of T-cell responses in vitro and delay allograft rejection.

Materials and Methods


C57BL/6, B6.C-H2.Ab1bm12, C57BL/6-Tg (UBC-GFP) and BALB/c mice were originally purchased from the Jackson Laboratory. All animals were bred and maintained under specific pathogen-free (SPF) condition. All animals received humane care in compliance with the Principles of Laboratory Animal Care formulated by the National Institute of Health (NIH publication no. 86-23, revised 1985), and protocols were approved by the local committee for animal welfare.

Reagents and antibodies

Lipopolysaccharides from E. coli 0111:B4 (LPS) nonspecific inhibitor of NOS nG-methyl-l-arginine acetate, indoleamine 2,3-dioxygenase (IDO) inhibitor 1-methyl-DL-tryptophan were purchased from Sigma-Aldrich (Bornem, Belgium). Arginase 1 inhibitor nw-hydroxy-nor-l-arginine and specific inhibitor of iNOS L-N6-(1-iminoethyl)lysine were purchased from Calbiochem (San Diego, CA). Heme oxygenase-1 (HO-1) inhibitor tin protoporphyrin IX (SnPP) was purchased from Alexis Biochemicals (Lausen, Switzerland). Anti-IL-10 (JES-2A5) was a gift from Muriel Moser (Department of Immunology, Faculty of Sciences, IBMM-Gosselies), anti-IL-10R (1B1.2) was purchased from BioXCell (West Lebanon, NH), anti-TGF-b mAb was obtained from R&D Systems (Abingdon, Oxon, UK), anti-PDL-1 mAb (MIH5) was purchased from Immunosource (Halle-Zoersel, Belgium). Anti-DNP rat IgG LO-DNP was used for in vitro experiments (LO-IMEX, Université Catholique de Louvain, Brussels, Belgium).

Endotoxin treatments

Mice were injected intraperitoneally with LPS dissolved in RPMI-1640. The protocol consisted of three daily intraperitoneal injections (2 μg/g body weight the first 2 days followed by a third dose of 16 μg/g body weight considered as a lethal dose). Blood was taken 4 h after LPS challenge. Spleen cells from LPS-treated mice were used for in vitro assays between day 7 and 9 after the first injection of LPS.

Skin graft and adoptive cell transfer

Mice were anesthetized with a mixture of xylazine (Rompun®) 5% and ketamine (Imalgene) 10% in PBS. Skin grafting was performed as previously described (19). Bandages were removed on day 9, and grafts were considered as rejected when more than 75% of epithelial breakdown had occurred. In the case of adoptive transfer, 5 × 106 CD11b+ spleen cells from female LPS-treated mice were administered intravenously to recipient mice via the tail vein the day prior to transplantation. In some experiments, CD11b+ cells were incubated for 3 h with SnPP in the culture medium and then washed twice with RPMI before injection.

Isolation of CD11b+ cell populations

CD11b+ cells were isolated from spleens using the CD11b cell Isolation Kit (Miltenyi). Purity was routinely more than 93% as assessed by FACS analysis. When specified, CD11b+GR-1 high, CD11b+GR-1 low, CD11b+GR-1 negative, CD11b+Ly6C+Ly6G negative and CD11b+Ly6C+Ly6G+ were isolated from enriched CD11b+ cells by using a Moflo cell sorter (DakoCytomation, Glostrup, Denmark) to obtain pure populations (99% purity as determined by FACS analysis).

Mixed lymphocyte reaction, T-cell stimulation and in vitro suppression assays

CD4+ or CD8+ T cells or total T cells were purified from spleens by CD4, CD8α+ or CD90.2 isolation kits (Miltenyi Biotec, Bergisch Gladbach, Germany). Responder cells were used at the final concentration of 5 × 105/mL. For mixed lymphocyte reactions (MLRs), irradiated spleen cells (1 × 106/mL) were used as stimulators. For polyclonal T-cell stimulation, anti-CD3/CD28 microbeads (Invitrogen, Merelbeke, Belgium) were used. CD11b+ cells were added at a final concentration of 1 × 106/mL or in a dose titration manner in 96-well round-bottom plates in the RPMI 1640 culture medium containing 5% heat-inactivated fetal calf serum. Supernatants were harvested after 72 h. Transwell experiments were performed in flat-bottom 96-well plates. 1 × 105 CD90+ cells stimulated with anti-CD3/CD28 were placed at the bottom and 2 × 105 CD11b+ cells were placed in transwell chambers (Anopore inserts 0.2 μm; Nunc, Roskilde, Denmark).

Flow cytometry

Cell surface staining was performed using the specified mAbs (BD Biosciences, Erembodegem-Aalst, Belgium). FACS analysis was performed on a CyAn-LX cytometer with Summit 4.1 software (DakoCytomation). IFN-γ intracytoplasmic staining was performed after cell incubation (1 × 106/mL) with 50 ng/mL PMA and 500 ng/mL ionomycin for 4 h with brefeldin A (10 μg/mL) in the last 2 hours; then the cells were incubated for 10 min with Fc block (2.4.G2), stained for surface markers, washed with PBS/BSA 0.1%, fixed with CytoFix/CytoPerm (BD Biosciences), permeabilized with Perm/Wash buffer (BD Biosciences) and labeled with IFN−γ FITC mAb (BD Biosciences). When specified, CD90+ cells were labeled with 1 μM carboxyfluorescein diacetate succinimidyl ester (CFSE) (Invitrogen) for 10 min at 37°C before stimulation. Cell division accompanied by CFSE dilution was analyzed by FACS after 72 h of culture.

Enzyme-linked immunosorbent assay

Cytokines were measured by enzyme-linked immunosorbent assay kits (Duoset, R&D Systems) in the serum or in the culture supernatants. The detection limits (pg/mL) were 8 for IL-2, 30 for IFN-γ, 40 for IL-13 and 17 for IL-10.

Purification and quantification of IL-10 and HO-1 mRNA

Cells were harvested from cultures after polyclonal T-cell stimulation in 24-well plates and isolated through magnetic beads as previously described. Messenger RNAs were extracted from 106 cells for each condition using a MagNA Pure LC mRNA Isolation Kit I (Roche Diagnostics, Brussels, Belgium) on the MagNA Pure Instrument (Roche Applied Science). For IL-10 transcripts, reverse transcription and quantitative real-time PCR were performed in a single step using the Lightcycler RNA Master hybridization probes Kit (Roche). Transcription and amplification were carried out on a LightCycler® (Roche Diagnostics). A total of 45 cycles were performed. mRNA levels were expressed in an absolute copy number, normalized against fixed numbers of copies of β-actin mRNA. Copy numbers were calculated for each sample from standard curves constructed from serial dilutions of purified plasmids for IL-10 and β-actin. The sequence of primers and probes for the real-time PCR reactions included β-actin sense 5′-CCGAAGCGGACTACTATGCTA-3′, β-actin antisense 5′-TTTCTCATAGATGGCGTTGTTG-3′, β-actin probe 5′-(6-Fam)ATCGGTGGCTCCATCCTGGC(Tamra)(phosphate)-3′; IL-10 sense 5′-GAAGACCCTCAGGATGCGG-3′, IL-10 antisense 5′-CCTGCTCCACTGCCTT-GCT-3′, IL-10 probe 5-(6-Fam)CGCTGTCATCGATTTCTCCCCTGTGA(Tamra)(phosphate)-3′.

For HO-1 transcripts, reverse transcription quantitative polymerase chain reaction was performed with a GenAmp 7700 sequence detection system (Applied Biosystems, Foster City, CA) using SYBR Green PCR core reagents (Applied Biosystems). Mouse primer sequences were used to target HO-1 and glyceraldehyde 3-phosphate dehydrogenase (GAPDH). The transcript accumulation index (TAI) (arbitrary units, AU) is defined as the fold change in mRNA levels in a given sample (Q) relative to levels in a calibrator (CB)—in this case, CD11b+ cells. The calibrator is the 1× expression of each gene. The TAI is calculated as follows: TAI = 2 Ct = (CtTarget − CtHPRT)Q − (CtTarget − CtHPRT)CB. Specific amplifications were checked by amplicon melting curves.

Western blot analysis

Cell protein extracts were boiled, electrophoresed on a sodium dodecyl sulfate–polyacrylamide gel and blotted. Membranes were blocked and incubated with a rabbit polyclonal anti-HO-1 antibody (Stressgen, Victoria, BC, Canada) or a mouse antitubulin monoclonal antibody (Calbiochem). Membranes were then incubated with horseradish peroxidase-labeled secondary Abs (Jackson ImmunoResearch, West Grove, PA) and detection performed by enhanced chemoluminescence (Amersham Biosciences, Otelfingen, Switzerland).

Immunofluorescence and confocal image analysis

HO-1 protein expression was studied on CD11b+Gr-1+ and CD11b+Gr-1- cells purified from LPS-treated mice. Briefly, 4% PFA-fixed cells were incubated with an anti-HO-1 rabbit polyclonal antibody (Stressgen) followed by incubation with secondary FITC-labeled Abs. Cell nuclei were counterstained with TO-PRO-3 iodide (Invitrogen) and slides were mounted in the ProLong AntiFade reagent (Invitrogen). Slides were analyzed with a Leica confocal microscope (Heidelburg, Germany) and the Leica TCS NT software.

Statistical analyses

Data are expressed as mean ± standard deviation (SD). Statistical analysis of differences between groups was performed using the two-tailed Mann−Whitney nonparametric test. Graft survival curves were compared by the log-rank test (p < 0.05 considered as significant).


LPS-treated mice contain large amounts of CD11b+GR-1+myeloid-derived suppressor cells

We performed multiple injections of LPS (see ‘Material and Methods’). Prior exposure to two nonlethal doses of LPS made animals resistant to a normally lethal dose of LPS, a process referred as LPS desensitization. Indeed, when rechallenged with a lethal dose of LPS, no animals died and the production of inflammatory cytokines (TNF-α, IL12, IFN-γ and IL6) was dramatically reduced whereas IL-10 secretion was upregulated (Figure 1A). We monitored the percentage of CD11b+GR-1+ cells in the spleen, lymph nodes and bone marrow of LPS-treated mice. CD11b+GR-1+ cells increased up to 4 fold in the three localizations and peaked at day 3 in the bone marrow and lymph node and day 7 in the spleen after LPS desensitization. Increased number of CD11b+GR-1+ cells persisted for 15 days after LPS desensitization (Figure 1B).

Figure 1.

LPS desensitization promotes IL-10 secretion and expands CD11b+GR-1+ myeloid cells. (A) Serum TNF-α, IFN-γ, IL-6 and IL-10 were measured 4 h after the injection of 16 μg/g body weight of LPS in either naive C57BL/6 mice (white bars) or LPS-treated C57BL/6 mice (black bars). Results are expressed as mean ± SD. *p < 0.05; **p < 0.01. Two separate experiments with minimum five mice per group showed similar results. (B) Kinetic profile of CD11b+GR-1+ cells after LPS desensitization. The percentages of CD11b+GR-1+ cells were measured at day 0, day 3, day 7 and day 15 after LPS desensitization (black diamonds) in bone marrow, spleen and axillary and inguinal lymph nodes. The percentages of CD11b+GR-1+ cells in naive mice are shown (gray squares). Results are expressed as mean ± SD of three mice per group for each time point.

We then tested whether these CD11b+GR-1+ cells from LPS-treated animals were similar to MDSCs that have been previously described and whether they could modulate a T-cell alloreactive response. As shown in Figure 2, CD11b+ cells extracted from LPS-treated mice suppressed alloreactive CD4+ T-cell proliferation in MLR down to background levels observed in syngeneic conditions. This effect was specific to CD11b+ cells from LPS-treated mice, because neither CD11b-negative cells from LPS-treated animals nor CD11b+ cells from control mice suppressed proliferation in MLRs (Figure 2A). Under these conditions (BALB/c stimulator cells and B6 responder cells), both CD4+ and CD8+ T cells can recognize allogeneic MHC molecules; we then tested the suppressive capacities of CD11b+ cells in MLR in a mixed population of alloreactive CD4+ and CD8+ T cells using CD90+ cells as responders. As shown in Figure 2B–F, CD11b+ cells from LPS-treated animals suppressed T-cell proliferation, IFN-γ, IL-13 and IL-2 productions in a dose-dependent manner. Importantly, the IL-10 detected in MLR supernatants was proportional to the amount of the added CD11b+ cells.

Figure 2.

CD11b+ cells from LPS-treated mice regulate CD4+ and CD8+ T cells in mixed lymphocyte reaction. (A) CD11b+ cells from LPS-treated mice regulate CD4+ T-cell alloreactivity. MLR was performed by using C57BL/6-irradiated spleen cells as stimulators and BALB/c CD4+ T cells as responders. Cell proliferation was measured. From left to the right: syngeneic stimulation and allogeneic stimulation (gray bars), suppression of allogeneic condition by C57BL/6 CD11b+ cells (CD11b+/T-cell ratio = 1) from LPS-treated mice, absence of suppression by C57BL/6 CD11b+ cells from untreated mice referred as CD11b+ LPS− cells or by spleen cells from LPS-treated C57BL/6 mice but depleted of CD11b+ cells (referred as CD11b− LPS+ cells) (black bars). The direct allostimulatory capacities of CD11b+ cells from either LPS-treated or undersensitized mice are compared (white bars). Results are expressed as mean ± SD of triplicates. One experiment representative of three independent experiments is shown. (B–F) CD11b+ cells regulated the MLR in a dose dependent manner. MLR was performed by using irradiated C57BL/6 spleen cells as stimulators and BALB/c CD90+ T cells as responders (containing CD4+ and CD8+ T cells). CD11b+ cells from LPS-treated C57BL/6 mice were added at different ratios. Cell proliferation was measured by thymidine incorporation after 72 h of culture. Cytokines (IFN-γ, IL-13, IL-2 and IL-10) were collected after 72 h of culture. Results are expressed as mean ± SD of triplicates. Minimum three independent experiments showed similar results.

In order to assess the direct suppressive effect of CD11b+ cells on T cells, independently of antigen-presenting cells, T cells were polyclonally stimulated with anti-CD3- and anti-CD28-coated microbeads before the addition of CD11b+ cells. T-cell proliferation was abolished by CD11b+ cells in a dose-dependent manner as shown by CFSE dilution experiments (Figure 3A) and this equally affected CD4+ and CD8+ T cells (Figure 3B). Similarly, CD11b+ cells blocked IFN-γ productions by both CD8+ and CD4+ cells as attested by intracytoplasmic staining (Figure 3C). Again, analysis of culture supernatant revealed a dose-dependent suppression of Th1 and Th2 cytokine productions by CD11b+ cells with an opposite effect on IL-10 secretion (Figure 3D).

Figure 3.

CD11b+ cells from LPS-treated mice suppress polyclonally activated CD4+ and CD8+ T cells. (A) CFSE-labeled CD90+ T cells were stimulated in vitro by anti-CD3/CD28 microbeads and CD11b+ cells from LPS-treated mice were added at different ratios. CFSE intensity of CD3+ T cells is proportional to the number of cell divisions. Stimulated T-cells showed multiple CFSE dilution peaks, which were abolished when CD11b+ cells were added at a ratio of 4/1 (CD11b+/T cells), and the inhibition of CFSE dilution was dose dependent. One experiment representative of three independent experiments is shown. (B) The inhibition of proliferation by CD11b+ cells affected CD3+CD4+ T cells (black bars) and CD3+CD8+ T cells (white bars) uniformly, in a dose-dependent manner. Three independent experiments showed similar results. (C) CD11b+ cells inhibited IFN-γ production by T cells. The percentage of IFN-γ-positive T cells (gated on CD3+ cells) is shown in unstimulated T cells, in anti-CD3/CD28-stimulated T cells with or without CD11b+ cells. Percentages represent a pool of three individual wells. Results are expressed as mean ± SD and are representative of two independent experiments. (D) IFN-γ, IL-13 and IL-10 production is modulated by CD11b+ cells in a dose-dependent manner. CD11b+ cells from LPS-treated mice were added at different ratios to anti-CD3/CD28-stimulated T cells. Cytokines were collected after 72 h of culture. Results are expressed as mean ± SD of triplicates. One experiment representative of three independent experiments is shown.

The degree of GR-1 expression by CD11b cells governs their suppressive potential

Because MDSCs are defined by the coexpression of CD11b and GR-1 antigens, we compared the capacities of CD11b+ cells to suppress T-cell polyclonal activation, according to their degree of GR-1 expression. CD11b+ cells from LPS-treated mice were first gated on GR-1-high, GR-1-intermediate and GR-1-negative cells and then cell-sorted by flow cytometry. As shown in Figure 4, both GR-1-high and GR-1-intermediate cells suppressed T-cell proliferation and IFN-γ and IL-13 production, whereas GR-1 negative cells had virtually no impact on T-cell proliferation and rather increased IFN-γ and IL-13 production. Both GR-1-high and GR-1-intermediate cells, but not GR-1 negative cells, were associated with IL-10 production in a dose-dependent manner. Again, this suggests that the CD11b+GR-1+ cells we observed after LPS challenge belong to the MDSC heterogeneous family. We also compared the suppressive capacities of CD11b+GR-1+ cells from either LPS-treated mice or control mice (not injected with LPS). In addition to the increased number of CD11b+GR-1+ cells after LPS desensitization, repeated titration experiments revealed that CD11b+GR-1+ cells from LPS-treated mice acquire greater suppressive capacities on a per cell basis, compared with CD11b+GR-1+ cells from control mice (data not shown). The GR-1 marker is a mixture of Ly-6G and Ly-6C. Among GR-1 cells, Ly-6G-positive (Ly-6C-intermediate) and Ly-6G-negative (Ly-6C-high) populations have been reported as functionally distinct (6). However, after LPS desensitization, no difference in terms of the regulation of polyclonally stimulated T cells was observed (data not shown).

Figure 4.

Only CD11b+GR-1+ cells mediate suppression. (A) CD11b+ cells from LPS-treated mice were separated according to their degree of GR-1 expression in GR-1 high, GR-1 intermediate and GR-1 negative, and cell-sorted by flow cytometry. (B–E) Either GR-1-high or GR-1-intermediate or GR-1-negative CD11b+ cells were added at different ratio to anti-CD3/CD28-stimulated CD90+ T cells, as indicated on the graph. The respective inhibition of proliferation (B), IFN-γ (C), IL-13 (D) and IL-10 (E) productions are shown. Results are expressed as mean ± SD of triplicates except for GR1-negative cells. One experiment representative of two independent experiments is shown.

MDSCs from LPS-treated mice impair T-cell activation

The antiproliferative effect of classically described MDSCs on T cells is thought to depend on the inhibition of TCR signaling pathway (20), IL-2 production, IL-2R signaling (17) and induction of T-cell apoptosis (21). We first compared the capacity of Tcells to be activated by anti-CD3/CD28 stimulation in the presence or absence of MDSCs from LPS-treated mice. As shown in Figure 5, MDSCs strongly inhibited T-cell activation in a dose-dependent manner as attested by a decreased expression of CD25, CD44 and CD69 on T cells. In contrast, no significant increase in T-cell apoptosis (annexin V) was observed in the presence of MDSCs (data not shown).

Figure 5.

CD11b+ cells inhibit T-cell activation. (A) CD25, CD69 and CD44 expression was measured after 48 h of culture on unstimulated T cells and anti-CD3/CD28-stimulated T cells with or without CD11b+ cells at the ratio of 1 T cell for 2 CD11b+ cells. Two independent experiments showed similar results. (B) Dose-dependent effect of CD11b+ cells: percentage of positive cells for each marker was measured at 24, 48 and 72 h for CD69, CD25 and CD44, respectively (time of maximal expression on stimulated T cells). CD11b+ cells were added to stimulated T cells at the indicated ratio. Results (simplicates) are obtained from pooled culture wells. Two independent experiments showed similar results.

T-cell suppression by MDSCs is cell-to-cell-contact-dependent and requires IL-10 and HO-1 activity

Several studies have shown that suppression of T cells by MDSCs requires a close contact between CD11b+ cells and T cells (11,22). We addressed this question in the context of MDSCs appearing after LPS desensitization. As shown in Figure 6A, when a transwell membrane was placed between anti-CD3/CD28 stimulated T cells and CD11b+ cells, suppression of T-cell proliferation was abolished and IFN-γ production restored. Interestingly, the transwell membrane also prevented the IL-10 production (Figure 6A), suggesting a role for this cytokine in the CD11b+ cell-mediated suppression. To determine the source of IL-10, IL-10 mRNA expression was quantified in T- and CD11b+ cell subsets having been cultured separately or together. As shown in Figure 6B, IL-10 mRNA was already detected in CD11b+ cells cultured alone but not in activated T cells. Importantly, the amount of IL-10 mRNA within CD11b+ cells significantly increased when cultured in the presence of activated T cells, whereas IL-10 mRNA in T cells only marginally increased (Figure 6B). Next, the role of IL-10 production in CD11b+ cell-mediated suppression was addressed by adding anti-IL10 and anti-IL10 receptor mAbs in the culture medium (Figure 6C). Notably, IL-10 neutralization fully restored IFN-γ production but not T-cell proliferation. Therefore, we looked for other effectors responsible for the suppression of both cell proliferation and cytokine productions. Even though IFN-γ is considered as a key factor in suppression mediated by many MDSCs, in our experiments, T-cell suppression remained unaffected by neutralizing anti-IFN-γ mAb (Figure 6C). Arginase 1 and iNOS, two enzymes involved in the metabolism of l-arginine, have been implicated in the inhibitory function of MDSCs in different settings (23,24). However, in our experiments, the inhibition of those enzymes had no effect on the suppressive capacity of CD11b+ cells (Figure 6D) even when added together in the same well (data not shown). Together with their high IL-10 production, this indicates that those cells are most probably distinct from ‘classically’ described MDSCs. The inhibition of other important molecules in immune regulation processes such as IDO, TGF-β and PD-L1 was also unsuccessful (Figure 6C,D). Finally, we considered the cytoprotective enzyme, HO-1 that catabolizes heme groups in the three immunosuppressive byproducts: carbon monoxide, biliverdin and ferritin (25). HO-1 suppressive activity has been reported in several allogeneic circumstances and plays a critical role in the inhibition of allograft rejection through donor-derived mesenchymal stem cells (26–29). Importantly, blocking HO-1 activity through tin-protoporphyrin IX (SnPP) completely abrogated T-cell suppression by CD11b+ cells (Figure 6D). Indeed, SnPP addition to the culture restored T-cell proliferation, IFN-γ production and abolished IL-10 production (Figure 6E), while the addition of SnPP to polyclonally activated T cells in the absence of CD11b+ cells interfered neither with their proliferation nor cytokine production (data not shown). Large amounts of HO-1 mRNA were detected by quantitative PCR only in CD11b+ suppressor cells but not in CD90+ T-cells after cell coculture (Figure 7A). High quantities of HO-1 mRNA were also observed within CD11b+GR-1+ cells purified from LPS-treated animals (data not shown). At the protein level, both western blot and confocal microscopy revealed the HO-1 expression in CD11b+GR-1+ cells. Notably, HO-1 was not detected in CD11b+GR-1-negative cells which are devoid of suppressive capacities (Figure 7B and C).

Figure 6.

MDSC-mediated suppression is cell-to-cell-contact-dependent, and requires heme oxygenase-1 activity and partly IL-10. (A) CD90+ T cells were stimulated in vitro by anti-CD3/CD28 microbeads. CD11b+ cells from LPS-treated mice were added at the ratio of 1 CD90+ T cell for 2 CD11b+ cells. Proliferation and IFN-γ, IL-13 and IL-10 productions were measured after 72 h under the following conditions: unstimulated T cells (white bars), anti-CD3/CD28-stimulated T cells with or without CD11b+ cells (black bars and gray bars, respectively) or in the presence of a transwell membrane between CD90+ T cells and CD11b+ cells (black lined bars). Results are expressed as mean ± SD of triplicates. One experiment representative of three independent experiments is shown. (B) IL-10 mRNA expression was quantified by q-PCR, in anti-CD3/CD28-stimulated CD90+ T cells and CD11b+ cells respectively, cultured either separately (black bars) or together (white bars). CD90+ cells and CD11b+ cells that were cocultured were purified before mRNA extraction. Results represent a pool of two independent experiments with minimum three individual samples per condition and are expressed as mean ± SD. (C) CD90+ T cells were stimulated in vitro by anti-CD3/CD28 microbeads. CD11b+ cells from LPS-treated mice were added at a ratio of 2 CD11b+ cells for 1 CD90+ T cell. Proliferation and IFN-γ production were measured after 72 h, under the following conditions: unstimulated T cells (white bars), stimulated T cells (gray bars) and stimulated T cells in the presence of CD11b+ cells (black bars) with or without the following antibodies (different doses were tested and the highest, 10 μg/mL, is shown): LO−DNP (control isotype), anti-TGF-β, anti-IFN-γ and decreasing doses of IL-10 neutralizing antibodies. Results are expressed as mean ± SD of triplicates. One experiment representative of three independent experiments is shown. (D) Inhibition by CD11b+ cells is mediated through HO-1. CD90+ T-cells were stimulated by anti-CD3/CD28 microbeads and CD11b+ cells from LPS-treated mice were added at a ratio of 2 CD11b+ cells for 1 CD90+ T cell in the presence of enzyme inhibitors or blocking antibodies. Proliferation was measured after 72 h in unstimulated T cells (white bars), stimulated T cells (gray bars) and stimulated T cells in the presence of CD11b+ cells (black bars) with or without the following inhibitors (different concentrations gave similar results and only one is shown): 5 mM of nG-methyl-l-arginine acetate (L-NMMA, a nonspecific inhibitor of NOS), 0.5 mM of L-N6-(1-iminoethyl)lysine (L-NIL, a specific inhibitor of iNOS), 0.5 mM of nw-hydroxy-nor-l-arginine (Nor-NOHA, an inhibitor of arginase-1), 0.2 mM of 1-methyl-DL-tryptophan (1-MT, an inhibitor of IDO), 10 μg/mL of anti-PD-L1 and decreasing concentrations of tin protoporphirin IX (SnPP, an inhibitor of HO-1). One experiment representative of three independent experiments is shown. (C) CD90+ T cells were stimulated by anti-CD3/CD28 microbeads and CD11b+ cells from LPS-treated mice were added at a ratio of 2 CD11b+ cells for 1 T cell in the presence of various doses of the HO-1 inhibitor SnPP. IFN-γ and IL-10 productions were measured after 72 h in unstimulated T cells (white bars), stimulated T cells (gray bars) and stimulated T cells in the presence CD11b+ cells (black bars), and in the presence of decreasing doses of SnPP. Results are expressed as mean ± SD of triplicates. Two independent experiments showed similar results.

Figure 7.

CD11b+GR-1+ cells from LPS-treated mice express HO-1. (A) HO-1/GAPDH mRNA ratios were determined through quantitative RT–PCR and expressed as arbitrary units in purified CD11b+ cells and anti-CD3/CD28-stimulated CD90+ T-cells after 72 h of coculture. (B) HO-1 protein levels were determined by western blot in CD11b+GR-1+ and CD11b+GR-1- cells from LPS-treated mice. In the graphic, quantification of bands from two different experiments is shown as HO-1/tubulin ratio. (C) Immunofluorescence staining for HO-1 (green) in CD11b+GR-1+ and CD11b+GR-1- cells purified from LPS-treated mice. One experiment representative of two is shown in A, B and C.

MDSCs from LPS-treated mice delay allograft rejection in an HO-1-dependent manner

We tested the capacity of LPS-treated mice to reject MHC or minor antigen-disparate skin grafts. Mice received three consecutive injections of LPS (see ‘Material and Methods’) and were grafted 3 days after the last injection. As shown in Figure 8, this protocol induced a significant delay in skin graft rejection in the case of male to female syngeneic combination (minor antigen disparity) and in the case of a single MHC class II disparity (bm12 to C57BL/6). However, this protocol had no impact on the rejection kinetics of a full (major plus minor) mismatched antigen disparity BALB/c to C57BL/6 (data not shown). To test a possible role of IL-10 in the delay of male skin graft rejection, we neutralized IL-10 in LPS-desensitized female recipients, a setting in which high serum IL-10 levels were observed (see Figure 1). Chronic IL-10 neutralization by three weekly injections of 200 μg of anti-IL-10R mAb (1B1) slightly accelerated skin graft rejection compared with control mAb-treated animals, although this did not reach statistical significance (p = 0.11).

Figure 8.

CD11b+ cells delay allograft rejection in an HO-1-dependent manner. (A) LPS desensitization delays allograft rejection. The normal kinetic of male skin graft rejection by female recipients is shown (black circles, n = 11 recipients). In the other curve, both donors and recipient (white circles, n = 17 recipients) were LPS-treated prior to male skin transplantation into females. (B) bm12 MHC-II disparate skin graft survival in C57BL/6 recipients. Normal kinetic of bm12 skin graft rejection by female recipients is shown (black circles). In the other curve, both donors and recipients (white circles) were LPS-treated before transplantation. Results are pooled from two independent experiments showing similar results. (C) IL-10 neutralization after LPS desensitization does not prevent the delay of allograft rejection. Kinetics of male skin graft rejection by untreated female mice is shown (circles). LPS-desensitized mice received three weekly doses of 200 μg of either the anti-IL-10R mAb (1B1) (triangles) or control mAb (LO-DNP) (squares) starting the day of skin graft. Injection of anti-IL-10R mAb in control mice (without LPS pretreatment) did not change rejection kinetics (data not shown). (D) Detection of GFP-transgenic CD11b+ cells after adoptive transfer in recipient spleens or graft draining lymph nodes (day 7 after transfer). Untransferred control mice (CTRL) are shown in the upper dot plots. The numbers of positive events/106 total spleen or lymph node cells are summarized in the right table. Results are expressed as mean ± SD of three individual mice per time point. (E) The adoptive transfer of CD11b+ cells improves graft survival. Male C57BL/6 skin grafts were transplanted onto untreated female C57BL/6 recipients (black circles) or recipients transferred with 5 × 106 CD11b+ cells from LPS-treated female mice 1 day before transplantation (white circle). The role for HO-1 was tested by incubating purified CD11b+ cells from LPS-treated mice with the HO-1 specific inhibitor SnPP during 3 h before cell transfer (triangles). Graft survival rates were compared by the log-rank test.

To specifically address the role of MDSCs in the delay of skin allograft rejection, we adoptively transferred 5.106 CD11b+ cells purified from female LPS-treated mice into naïve C57BL/6 female mice 1 day prior to male skin graft. As shown in Figure 8C, this transfer significantly delayed male skin graft rejection compared to control untransferred recipients. By using CD11b+ cells from green fluorescent protein (GFP)-transgenic donors, we observed that transferred cells preferentially localized to the spleen and to a lesser extent to the draining lymph nodes (Figure 8D). Finally, a role for HO-1 in this MDSC-mediated delay of allograft rejection was tested by incubating purified CD11b+ cells with the HO-1-specific inhibitor SnPP during 3 h before an adoptive transfer in female mice. SnPP treatment of MDSC abrogated the inhibition of allograft rejection (Figure 8C). This demonstrates that HO-1 activity is a dominant effector of in vivo immune suppression mediated by MDSCs.


Our findings are reminiscent of previous observations made in the course of microbial sepsis where the suppression of T-cell responses was shown to depend on a TLR-mediated expansion of immature CD11b+GR1+ myeloid cells (2,4,15,16). Interestingly, such MDSCs have been also shown to regulate autoimmune pathologies (6). Herein, we show that exposure to bacterial LPS is sufficient to induce the emergence of MDSCs able to dampen alloimmune responses. This observation might be relevant to clinical settings in which transplantation is associated with translocation of endotoxins from the intestinal tract. Indeed, it has been suggested that the reduced immunogenicity of liver allografts as compared with other organ transplants involve gut-derived bacterial products circulating in portal blood (30–32).

We identified HO-1 as an additional mechanism by which these MDSCs regulate alloreactive T cells. The suppression of T cells by MDSCs had previously been shown to be mediated through their ability to metabolize l-arginine, leading to l-arginine depletion and the production of nitric oxide, reactive oxygen species and peroxynitrite, inducing T-cell apoptosis and the prevention of proliferation (15–17). In our experiments, neither iNOS nor arginase neutralization was able to reverse suppression. Although an immunoregulatory activity of HO-1 has already been reported in the context of transplantation, to our knowledge, this is the first observation of HO-1-dependent MDSC-mediated alloreactive T-cell suppression. The regulatory properties of HO-1 have been reported in many other settings. HO-1 catabolizes pro-oxidant heme groups into carbon monoxide, biliverdin and ferritin, three metabolites involved in immunoregulatory processes (25,33,34). Indeed, HO-1 expression in allografts or xenografts has been associated with improved graft survival, protection against ischemia reperfusion injury, arteriosclerosis and chronic rejection, and its overexpression by gene transfer in heart transplants facilitates tolerance induction (27,28,33,35–40). Notably, in rats, the transfusion of donor-derived mesenchymal stem cells delays cardiac allograft rejection in an HO-1 and iNOS codependent manner while only HO-1 was required for in vitro T-cell suppression by human adult mesenchymal stem cells (29).

In our experiments, the inhibition of HO-1 by SnPP also inhibited IL-10 production by MDSCs. This is consistent with another report describing the promotion of IL-10 secretion by HO-1 (41). In this study, HO-1 activity prevented dendritic cell maturation and proinflammatory cytokine production while preserving IL-10 secretion. In return, IL-10 can induce HO-1 expression, a process which in part depends on STAT3 (42,43). This might explain why IL-10 neutralization partly restored T-cell response in vitro and only marginally affected skin graft rejection after LPS desensitization. Indeed, in vitro, IL-10 neutralization did not restore T-cell proliferation in the presence of MDSCs, a process that is essential to expand antidonor T-cell precursors up to a sufficient threshold able to trigger skin graft rejection. This could explain the marginal effect observed in vivo.

Other studies have already pointed out the link between MDSCs and transplantation. In a rat model of kidney allograft, anti-CD28 therapy induced long-term survival and was associated with the presence, in tolerated allografts, of MDSCs that operate through iNOS activity (13). However, MDSC transfer in those experiments did not prevent allograft rejection which is in contrast with our results. Importantly, by using the bm12 to B6 MHC class II-disparate skin graft model, Zhang et al. recently showed that the interaction between the inhibitory receptor immunoglobulin-like transcript (ILT)-2 and HLA-G molecule induces the expansion of MDSCs that prolong allograft survival as attested by the MDSC adoptive transfer (14). In a mouse model of corneal allografts, tolerance was also associated with a dense CD11b+ cell infiltration within tolerated grafts, although details about GR-1 expression were not given (44). In a different context, mice genetically deficient for SHIP1 (SH2 [Src homology 2]-containing inositol phosphatase-1), a molecule involved in the homeostasis of certain hematopoietic cell types, have a significant expansion of myeloid suppressor cells in peripheral lymphoid tissues. Importantly, these animals are protected against experimental graft versus host disease (GVHD), and this was associated with a lack of alloreactive T-cell priming compared to wild-type hosts (45).

The results presented here provide evidence that MDSCs, alone, rapidly suppress polyclonally activated T cells in a nonantigen-specific manner and should therefore be considered as a new form of endogenous immunosuppression. As such, their superiority compared to conventional immunosuppression has not been investigated in the present work. However, the rapid effect of MDSCs on whole T-cell populations (including memory and naïve T cells) suggests they might be beneficial in addition to other treatments such as costimulation blockade. Experiments aiming to assess their possible synergistic effect with other immunosuppressive therapies are under current investigation.


We thank Marisa Alegre and Ken Field for helpful discussions and valuable criticisms of the manuscript; Philippe Horlait, Laurent Depret, Gregory Waterlot and Christophe Notte for the animal care; and Frédéric Paulart, Marie-Claude Lalmand and Gaëlle Tilly for technical assistance.

Funding Sources: This work was funded by research grants of the Walloon Region, the FNRS-Belgium and GlaxoSmithKline Biologicals.

Conflict of Interest Statement

VDW is a Research Fellow of the Fonds National de la Recherche Scientifique (FNRS-Belgium). NVR is a Research Fellow of the Fonds Erasme (Erasme Hospital, Belgium). ALM is a Research Associate of the FNRS-Belgium. Authors declare no competing financial interests.