Tumor-induced T-cell tolerance is a major mechanism that facilitates tumor progression and limits the efficacy of immune therapeutic interventions. Regulatory T cells (Treg) play a central role in the induction of tolerance to tumor antigens, yet the precise mechanisms regulating its induction in vivo remain to be elucidated. Using the A20 B-cell lymphoma model, here we identify myeloid-derived suppressor cells (MDSC) as the tolerogenic antigen presenting cells capable of antigen uptake and presentation to tumor-specific Tregs. MDSC-mediated Treg induction requires arginase but is transforming growth factor-β independent. In vitro and in vivo inhibition of MDSC function, respectively, with NOHA or sildenafil abrogates Treg proliferation and tumor-induced tolerance in antigen-specific T cells. These findings establish a role for MDSCs in antigen-specific tolerance induction through preferential antigen uptake mediating the recruitment and expansion of Tregs. Furthermore, therapeutic interventions, such as in vivo phosphodiesterase 5–inhibition, which effectively abrogate the immunosuppressive role of MDSCs and reduce Treg numbers, may play a critical role in delaying and/or reversing tolerance induction. [Cancer Res 2008;68(13):5439–49]
- myeloid-derived suppressor cells
- regulatory T cells
- tumor-induced tolerance
The clonotypic idiotype of malignant B cells was the first tumor-specific antigen identified as capable of eliciting a T cell–mediated immune response ( 1, 2). Despite the existence of this tumor-antigen and the well-documented expression of MHC class I and class II molecules ( 3), as well as costimulatory molecules by a majority of human B-cell malignancies, no spontaneous clinically significant immune response in these diseases has been documented ( 4). A number of studies have shown that malignant B cells can process and present antigen to T cells in vitro, including the presentation of epitopes derived from their own unique immunoglobulin idiotype to CD4+ and CD8+ T cells ( 5– 7). However, despite of their intrinsic antigen-presenting capabilities, B-cell tumors fail to be eliminated and tumor-specific T cells are tolerized rather than activated in vivo ( 8). An understanding of why a T cell–mediated immune response is not elicited but suppressed in vivo is essential for the development of new therapeutic strategies aimed at optimizing the clinical efficacy of immunotherapy in these diseases.
In attempting to address these issues, we previously showed that a murine B-cell lymphoma (A20) transfected to express the model antigen influenza hemagglutinin (HA) activates HA-specific CD4+ T cells from T-cell receptor (TCR) transgenic mice in vitro. However, with the adoptive transfer of HA-specific T cells into A20-HA bearing mice, the transgenic T cells are antigen experienced but anergized ( 8, 9). A more accurate analysis of the phenotype of these anergic CD4+ T cells revealed two subpopulations. The first population is CD25−/FOXP3− and does not proliferate, whereas the second CD25+/FOXP3+ population proliferates and inhibits the effector function of peptide stimulated naive T cells ( 10, 11).
More recently, we showed that in vivo disruption of host cross-presentation removes the tolerogenic mechanisms induced by the tumor and unmasks the intrinsic ability of malignant B cells to directly present tumor antigens ( 12, 13). Despite these findings, the true nature of the tolerogenic antigen presenting cell (APC) remains elusive.
Myeloid-derived suppressor cells (MDSCs) have recently been recognized as critical mediators of tumor progression in numerous solid tumors through their inhibition of tumor-specific immune responses ( 14). This monocyte/macrophage population is characterized by the expression of CD11b ( 15), F4/80 ( 16), and IL4Rα ( 17), variable expression of Gr1, and low expression of CD11c, MHC class I, and MHC class II ( 18). Whereas the number of MDSCs may not increase in certain models ( 19), their suppressive function clearly parallels increases in tumor burden ( 20). These cells blunt antitumor CTL responses through the expression of arginase and/or nitric oxide synthase (NOS; ref. 21) or the secretion of transforming growth factor-β (TGF-β; ref. 19). The activation of all these suppressive pathways seems to be regulated by IL4Rα, because genetic ablation or pharmacologic down-regulation of this receptor on MDSCs restores tumor-specific T-cell responsiveness and immunosurveillance ( 17, 22).
Recently, the administration of progenipoietin-1 (a synthetic granulocyte colony-stimulating factor/Flt-3 ligand molecule) to donors in an allogeneic bone marrow transplantation model generated MDSCs that upon transfer suppressed the initiation of graft-versus-host disease (GVHD) in recipients by inducing a population of MHC class II–restricted, interleukin 10 (IL-10)–producing regulatory T cells (Treg; ref. 23). Similarly, in a colon carcinoma murine model, MDSCs either generated or expanded the pool of CD4+CD25+FOXP3+ Tregs ( 24).
Here, we show a role for MDSCs during lymphoma progression. Specifically, with an increasing tumor burden, MDSCs up-regulate IL4Rα expression, increase their suppressive activities, uptake and process tumor-associated antigens (TAA), and, importantly, by expanding naturally occurring tumor-specific Tregs, induce T-cell tolerance.
Materials and Methods
Mice. BALB/c (Thy1.2+/+CD45.2+/+) mice, 6 to 8 wk old, were purchased from the National Cancer Institute. TCR transgenic mice (6.5 Tg mice) on a BALB/c background expressing an αβ TCR specific for amino acids 110-120 from HA were a gift from H. von Boehmer (Harvard Medical School, Dana-Farber Cancer Institute). The 6.5 Tg mice on Thy1.1+/+ or Thy1.1+/1.2+ background were used in the experiments as specified. Clone 4 mice transgenic for the H-2Kd–restricted TCR recognizing the influenza virus, HA peptide (HAp512-520) were a kind gift of L.A. Sherman (Scripps Research Institute). CD45.1+/+ BALB/c mice were a gift of H. Levitsky (Johns Hopkins University). Experiments using mice were conducted in accordance with protocols approved by the Animal Care and Use Committee of the Johns Hopkins University School of Medicine.
Antibodies and flow cytometry. The following antibodies were used for flow cytometry analysis: antimouse CD45.2 [peridinin chlorophyll protein (PerCP)], antimouse CD11b [phycoerythrin (PE) or APC], antimouse B220 (PE), antimouse CD124 (PE), antimouse CD80 (PE), antimouse CD86 (PE) antimouse Gr-1 (PE), antimouse IAd (PE), antimouse H2d (PE), antimouse CD11c (APC), antimouse F4/80 (APC), antimouse Thy1.2 (APC or FITC), antimouse Thy1.1 (PerCP or PE), antimouse CD25 (PE), antimouse CD4 (APC or PerCP), antimouse FOXP3 (FITC, APC, or PE; e-Biosciences), antimouse IgG2a (streptavidin). Chloromethylfluorescein diacetate succinimidyl ester (CFSE) labeling of cells (Molecular Probes) was previously described ( 22). All antibodies were purchased from BD Biosciences unless otherwise specified. All fluorescence-activated cell sorting (FACS) analysis was of surface expression except for FOXP3 for which cells were permeabilized. A total of 50,000 gated events were collected on a FACSCalibur (Becton Dickinson) unless otherwise specified and analyzed using FCS3 express software (De novo software Thornhill).
Tumor cells and adoptive T-cell transfer. A20-HA and A20WT cells were previously described ( 8). A20-HAGFP (a kind gift of H. Levitsky, Johns Hopkins University) was generated by stable transfection of the A20-HA tumor with the enhanced green fluorescent protein and selected following in vivo passages and maintained in RPMI 10% FCS supplemented with G418 (400 μg/mL) and hygromycin (400 μg/mL). A20-HAGFP tumor progression is similar to the parental cell lines A20-HA and A20WT (data not shown). Tumor cells (1 × 106) were injected via tail vein. For adoptive transfer using whole CD4+ T cells, single-cell suspensions obtained from lymph nodes and spleens of 6.5 Tg donors were enriched for CD4+ cells via negative selection using the CD4 isolation kit (Miltenyi Biotech). The percentage of lymphocytes positive for CD4 and the clonotypic TCR (monoclonal antibody 6.5) was determined by flow cytometry. A total of 2.5 × 106 CD4+ 6.5 TCR+ T cells were injected i.v. into each BALB/c recipient. Vaccination was performed by i.p. injection of 107 plaque-forming units of HA-encoding recombinant Vaccinia virus 25 d after tumor challenge.
Cells purification. Splenic CD11b+ cells were purified as previously described ( 22). CD4+CD25+ and CD4+CD25− T cell were magnetically enriched using the CD4/CD25 isolation kit (Miltenyi Biotech) from the spleens and lymph nodes of 6.5 Tg mice. For experiments requiring the purification of CFSElow clonotypic cells, T cell were first magnetically enriched for CD4, labeled with Thy1.1-PE and sorted using a cell sorter (FACSVantage SE; BD Biosciences). To maximize MDSC and dendritic cell (DC) purity in some experiments, cells were sorted using a FACSARIA (BD Biosciences) after magnetic preenrichment.
Coculture experiments. Sorted MDSCs or DCs (105) were cultured for 5 d with 105 purified clonotypic Thy1.1+ T cells and with 106 Thy1.1−/− BALB/c splenocytes in 96–flat-bottomed well plates. Proliferation or phenotype was evaluated by flow cytometry.
Drugs and inhibitors. Sildenafil (Pfizer) was dissolved in the drinking water (20 mg/kg/24 h); NOHA and l-NMMA (Calbiochem) were used at 500 μmol/L in vitro anti–TGF-β neutralizing antibody clone 2G7 (a kind gift of C. Drake, Johns Hopkins University) at 50 μg/mL.
Enzymatic assays. The arginase assay was performed as previously described ( 25) on purified CD11b+ cells. NOS activity was measured as nitrate/nitrite production on purified CD11b+ cultured for 24 h in DMEM, 10% FCS supplemented with 50 mmol/L of l-arginine, using the nitrate nitrite colorimetric assay kit (Cayman Chemical) according to the manufacturer's instructions. Results were normalized to 106 cells. Data are from triplicate wells. TGF-β was determined by ELISA (R&D Systems) on the supernatant of purified CD11b+ cells cultured for 24 h in AIM-V (Invitrogen). The supernatants were activated through acidification/neutralization before being tested. No TGF-β was detected without activation.
Preferential antigen uptake by MDSCs. We previously showed that tumor-induced tolerance requires cross-presentation from bone marrow–derived APCs ( 12). While the precise cell population responsible remains unclear, the requirements for cross-presentation by APCs are well defined and include (a) antigen uptake, (b) processing, and (c) presentation through their MHC molecules. To evaluate antigen uptake by APCs in our model, we used the CD45.2+ A20-HA-GFP tumor, a B-cell lymphoma transfected to coexpress the HA antigen, a model TAA, and the GFP molecule. These modifications do not interfere with normal tumor growth (data not shown) and permit identification of the malignant cells in vivo through expression of GFP, as well as the congenic marker CD45.2 in the CD45.1 BALB/c background.
A20-HA-GFP (CD45.2+/+) was injected i.v. into CD45.1+/+ BALB/c mice. Analysis on day 28 revealed the presence of a CD45.2−/GFP+ population consistent with antigen capture ( Fig. 1A ). Further analysis of this population showed that a majority of the cells (65.7%) are characterized by a CD11bhigh/CD11clow phenotype ( Fig. 1B), whereas only few (6.4%) possessed a CD11blow/CD11chigh phenotype characteristic of myeloid DCs. As expected, few cells within the tumor gate possessed markers for either MDSCs or myeloid DCs. Further analysis of the CD45.2−GFP+ CD11bhigh population shows low expression of MHC class I and class II molecules and of the costimulatory molecules B7.1 and B7.2 and high expression of granulocyte marker Gr-1. Moreover, this population expresses IL4Rα and F4/80 but is negative for B220 ( Fig. 1C). Taken together, this phenotype is consistent with that of MDSCs described in many murine solid tumors (reviewed in ref. 14) and excludes the possibility of contamination with residual tumor (B220+). The fact that the majority of host GFP+ cells share a phenotype consistent with MDSCs suggests that they may play an important role in antigen uptake and tolerance induction in this model.
Properties of CD11b+ MDSCs isolated from lymphoma bearing mice. Although the cells identified in Fig. 1 are phenotypically similar to the MDSCs described in solid tumors, further functional analyses are required to determine if these CD11b+ cells are indeed MDSCs. In solid tumors, MDSCs strongly suppress antigen-driven CD8 T-cell proliferation yet have minimal effect on CD4 expansion.
It is well established that impairment of T-cell effector function is mediated by mechanisms that require arginase, NOS, or both enzymes. TGF-β production by MDSCs has also been shown to suppress T-cell function ( 14). To evaluate if these functional properties also existed in the B-cell lymphoma model, we isolated CD11b+ cells from mice injected with A20-HA (gray columns) or PBS (naive mice, black columns) 28 days earlier. Purified splenic CD11b+ cells were admixed with naive BALB/c splenocytes and either HA-specific (clone 4) CD8+ ( Fig. 2A, left ) or HA-specific (6.5) CD4+ T cells ( Fig. 2A, right). As expected, whereas the addition of CD11b+ cells from naive mice did not alter antigen specific T-cell proliferation, tumor-derived CD11b+cells inhibited the proliferation of HA-specific CD8+ cells stimulated with the relevant peptide ( Fig. 2A, left). In contrast, HA-specific CD4+ T-cell function was not impaired by tumor-derived MDSCs (right). Interestingly, in the absence of exogenous peptide, but in the presence of CD11b+ cells isolated from A20-HA bearing mice, proliferation of CD4+ T cells could suggest endogenous uptake, processing, and presentation by CD11b+ cells of tumor-derived antigens.
As previously shown in solid tumors ( 14), CD11b+ cells from lymphoma-bearing mice also showed pronounced arginase activity and nitric oxide (NO) production, but failed to show increases in TGF-β secretion ( Fig. 2B). To further evaluate which mechanism(s) were responsible for CD8 inhibition, we repeated the suppression assay shown in Fig. 2A with the addition of NOHA (an arginase inhibitor), l-NMMA (an NOS2 specific inhibitor), and anti–TGF-β neutralizing antibody (clone 2G7) alone or in combination ( Fig. 2C). Whereas the addition of anti–TGF-β neutralizing antibody failed to revert MDSC suppression on CD8+ T cells, the addition of either NOHA or l-NMMA partially restored T-cell proliferation. Moreover, as previously reported in solid tumors ( 25), combined inhibition of both NOHA and l-NMMA fully restored T-cell proliferation.
Taken together, these data indicate that MDSCs isolated from A20-HA–bearing mice share many functional properties with MDSCs found in solid tumors. Interestingly, because HA-specific CD4+ T-cell proliferation occurred even in the absence of exogenous cognate peptide, these findings suggest that tumor-derived MDSCs are capable of antigen cross-presentation and likely mediate tolerance induction.
In vitro MDSCs preferentially induce proliferation of natural Tregs. CD4+ T cells encompass various populations with different cellular and immunologic functions (e.g., naive, effector, regulatory, etc.). We, therefore, examined the role of MDSCs on these populations. Splenic CD11b+ MDSCs were purified from BALB/c mice injected with either PBS or A20-HA 28 days earlier. They were then cocultured with HA-specific, Thy1.1+, CD4 purified, CFSE-labeled T cells and Thy1.2 BALB/c splenocytes used as feeder cells ( Fig. 3 ). In Fig. 2, we showed that the addition of tumor-derived CD11b+ cells failed to reduce HA-specific CD4+ T-cell proliferation. However, the phenotype of this proliferating clonotypic CD4+ population is significantly altered by tumor-derived CD11b+ cells ( Fig. 3A, left). In fact, analysis of the CFSElow population revealed that the control cocultures (no CD11b+ cells or CD11b+ cells from naive mice) pulsed with the HA-peptide contained FOXP3+ cells representing 13.7 ± 0.8% of the proliferating population (CFSElow). However, this percentage significantly increased with CD11b+ cells purified from A20-HA–bearing mice (23.6 ± 2%; Fig. 3A, right). Most interesting, however, is the selective expansion of FOXP3+ cells (77.3± 3.2% of CFSElow cells) seen in the absence of exogenous peptide when cocultured with A20-HA–derived CD11b+ cells. This finding underscores the ability of tumor-derived MDSCs to selectively expand Tregs and seems to suggest that exogenous peptide can influence the expansion of only FOXP3− cells.
We, thus, evaluated the effect of different amounts of exogenous peptide in vitro on CD4+ proliferation (Supplementary Fig. S1). In the absence of MDSCs, both FOXP3+ and FOXP3− T-cell proliferation correlates with the amount of exogenous peptide added (Supplementary Fig. S1, left). Interestingly, in the presence of A20-HA–derived MDSCs, Treg proliferation is elevated and independent of the amount of exogenous peptide added (Supplementary Fig. S1, top). In contrast, proliferation of the effector FOXP3− T cells is similar to the control and unaltered by the presence of A20-HA–derived MDSCs (Supplementary Fig. S1, compare bottom panels).
Because splenocytes are used as feeder cells in these experiments, it is possible that different APC populations may be responsible for Treg or effector T-cell proliferation. To evaluate this hypothesis, highly purified MDSCs (CD11b+, CD11Clow, MHC class IIlow) or DCs (CD11chigh, MHC class IIhigh, and CD11blow/−) were sorted from spleens of mice challenged with A20-HA or A20WT 4 weeks prior. Sorted cells were cocultured with CFSE-labeled Thy1.1+/− 6.5+/− HA-specific CD4+ T cells in the presence of Thy1.2+ BALB/c splenocytes for 48 hours ( Fig. 3B). The relevant peptide was added where indicated. In the absence of exogenous peptide, A20-HA–derived MDSCs significantly stimulated Treg proliferation, whereas no effect was observed on FOXP3− effector T cells ( Fig. 3B). In contrast, A20-HA–derived DCs induced a small but significant proliferation of effector CD4+ T cells with no measurable effect on Treg proliferation ( Fig. 3B). The addition of exogenous peptide again minimized the observed differences. Taken together, these data support the hypothesis that different APC populations selectively expand specific CD4+ subsets.
To evaluate if the same mechanisms involved in MDSC-mediated CD8+ T-cell suppression are involved in the differential proliferation of regulatory FOXP3+ compared with effector T cells, we repeated the coculture experiments described in Fig. 2C using clonotypic CD4+ T cells. As shown in Fig. 3C, l-NMMA and the anti–TGF-β neutralizing antibody showed only modest decreases in FOXP3+ T-cell proliferation. In contrast, the addition of the arginase inhibitor, NOHA, completely reduced MDSC-induced Treg proliferation to levels comparable with those seen in the control group (CD11b cells obtained from non–tumor-bearing mice).
Taken together, these data strongly suggest that MDSCs can alter the homeostatic equilibrium between regulatory and effector T cells. The preferential expansion of Tregs in the culture seems to be determined by both the presence of tumor-derived MDSC and their arginase activity.
MDSCs expand Tregs from a preexisting population of natural Tregs. The experiments described above show that MDSCs can expand the pool of Tregs. However, they do not establish whether this FOXP3+/CFSElow population is derived from the conversion of FOXP3− effector T cells or from the selective expansion of a preexisting population of FOXP3+ Tregs, nor whether MDSC-mediated tumor antigen processing and presentation are required. To answer these questions, HA-specific CD4+/CD25+ Tregs were purified from Thy1.1+/+/Thy1.2−/− 6.5+ or Thy1.1+/−/Thy1.2−/− 6.5+ tumor-free mice and admixed with CD25-depleted, HA-specific CD4+ effector T cells from Thy1.1+/−/Thy1.2+/− 6.5+ naive mice at a 1:10 ratio, respectively. This mixture, in which almost 90% of the FOXP3+ cells are negative for the Thy1.2 marker (Supplementary Fig. S2), was cocultured in the absence of exogenous peptide with purified CD11b+ cells obtained from (a) naive, (b) A20-WT, or (c) A20-HA mice. Two days later, the cells were analyzed for expression of FOXP3, Thy1.1, and Thy1.2. If MDSC-induced Tregs originate from a preexisting regulatory population (Thy1.1+/+/Thy1.2−/−), most of the FOXP3+ cells on analysis would be Thy1.2−. On the contrary, if the generation of Tregs result from the conversion of effector T cells (Thy1.1+/−/Thy1.2+/−), the FOXP3+ T cells would be Thy1.2+.
As expected, CD11b+ cells from naive mice did not increase the Treg pool. Similar results were also obtained with the A20WT-derived MDSCs. In contrast, A20-HA–derived CD11b+ cells expanded FOXP3+ CD4+ T cells ( Fig. 4A, top and Supplementary Fig. S2). The fact that Treg expansion can be induced only by CD11b+ cells isolated from A20-HA–bearing mice and not A20WT invokes the role of cross-presentation of tumor antigens by MDSCs as the critical requirement for Treg proliferation. Moreover, because most newly generated FOXP3+ cells are Thy1.2−, this result strongly suggests that MDSCs mediate Treg expansion from a preexisting natural Treg population and not by conversion of naive/effector T cells. This finding is supported by (a) the absence of any significant increase in the percentage of FOXP3+ cells when Tregs are depleted ( Fig. 4A, bottom) and (b) the CFSE analysis demonstrating the selective proliferation of the preexisting Thy1.2− regulatory population ( Fig. 4B).
For in vivo confirmation of the above-mentioned in vitro findings, HA-specific, CD4+/CD25+ Tregs (Thy1.1+/+Thy1.2−/−) purified from naive mice were admixed at a 1:10 ratio with CD25-depleted, HA-specific, CD4+ effector T cells (Thy1.1+/−/Thy1.2+/−) as previously described. This mixture was CFSE-labeled and injected into Thy1.2+/+ BALB/c mice alone or with naive or A20-HA–derived CD11b+ cells ( Fig. 4C, left). In this system, adoptively transferred effector clonotypic T cells coexpress both the Thy1.1 and Thy1.2 congenic markers. Regulatory clonotypic T cells are only positive for the Thy1.1 marker and the host T cells are negative for Thy1.1 expression ( Fig. 4C, right). Splenocytes were harvested 60 hours later, and the CFSE dilution was analyzed by gating on either the adoptively transferred Thy1.1+/Thy1.2− Tregs ( Fig. 4C, top left quadrant) or Thy1.1+/Thy1.2+ effector T cells ( Fig. 4C, top right quadrant). The control groups (no CD11b or CD11b cells isolated from naive mice) showed virtually no proliferation of the adoptively transferred T cells. In contrast, coinjection of T cells with tumor-derived CD11b+ cells preferentially expanded the Thy1.1+/Thy1.2− Treg population but not the effector T cells ( Fig. 4D). These data confirm the in vitro findings by showing that MDSCs induce the proliferation of tumor-specific Tregs.
These results strongly suggest that MDSCs play an important role in inducing Treg expansion but not in the conversion of regulatory FOXP3+ T cells from FOXP3− effector T cells by a mechanism that requires tumor-associated antigen capture, processing and presentation by MDSCs.
IL4Rα expression on MDSCs correlates with tumor progression and can be inhibited by sildenafil. Recently, we showed that IL4Rα expression plays an important role in MDSC-mediated T-cell suppression in solid tumors ( 17). Genetic ablation of this marker on macrophages and granulocytes is, in fact, sufficient to restore not only antitumor T-cell responsiveness but also efficacy of adoptively transferred, tumor-primed, CD8+ T cells ( 17). Similar results can be obtained by the pharmacologic inhibition of phosphodiesterase-5 (PDE5) using sildenafil (Viagra; ref. 22). Because IL4Rα is also expressed on lymphoma-derived MDSCs ( Fig. 1), we asked whether sildenafil could alter its expression on MDSCs during lymphoma progression. In splenocytes from A20-HA–bearing mice, we examined whether the presence of MDSCs correlated with tumor burden at various time points after tumor challenge. MDSCs defined by the classic markers CD11b and Gr1 do not accumulate during tumor progression ( Fig. 5B ), and their numbers in the spleen do not correlate with tumor burden (Supplementary Fig. S3). In contrast, IL4Rα up-regulation on MDSCs significantly correlates with tumor progression (Spearman P < 0.001; Fig. 5A and C). This finding supports the notion that the growing tumor burden is associated with increases in the suppressive phenotype of MDSCs but not with their accumulation in the secondary lymphoid organs in the A20 model. Although interesting, this phenomenon is not unique to lymphoid malignancies; similar findings were also seen in the 15-12RM fibrosarcoma model ( 19).
Based on our data showing that sildenafil can down-regulate IL4Rα expression on MDSCs in solid tumors ( 22), we examined its role during B-cell lymphoma progression. As expected, sildenafil effectively down-regulates IL4Rα expression on lymphoma-derived MDSCs ( Fig. 5C). However, in contrast to its measurable antitumor effect in solid tumors ( 22), it does not significantly reduce A20-HA tumor outgrowth (data not shown). This contradictory result can possibly be explained in several ways: (a) immune-mediated eradication is primarily a CD4-dependent, and not CD8-dependent, process ( 26) and sildenafil has less effect on activating CD4 compared with CD8 effector T cells ( 22) or (b) the possibility that other immunosuppressive pathways are also present in the A20 system, such as indoleamine 2,3-dioxygenase (IDO), the expression of which can convert effector T cells into Tregs and thus regulate the immune response during lymphoma progression ( 27).
Taken together, these data show that (a) IL4Rα expression on MDSCs correlates with A20-HA lymphoma progression and (b) in vivo sildenafil administration can down-regulate IL4Rα expression.
Sildenafil reduces lymphoma-induced T-cell anergy and Treg expansion. The data described above underscores the ability of lymphoma-induced MDSCs to cross-present tumor antigens to Tregs ( Figs. 1, 3, and 4) through an arginase-dependent mechanism ( Fig. 3B). Because most MDSC suppressive pathways can be inhibited by down-regulating IL4Rα ( 17, 22, 28, 29), we asked if sildenafil could also reverse tumor-induced T-cell anergy and Treg proliferation. To this end, mice were either challenged with A20-HA or left tumor-free and given purified Thy1.1+, HA-specific, CD4+ T cells 10 days later. Sildenafil was added to the drinking water of half the animals at the time of T-cell transfer. Two weeks later, half the mice in each group were primed with VaccHA ( Fig. 6A ). Clonal expansion ( Fig. 6B), as well as FOXP3 expression ( Fig. 6C), was evaluated 5 days after VaccHA priming. Whereas VaccHA immunization induced a robust expansion of clonotypic T cells in naive mice, antigen-driven proliferation was strongly inhibited in tumor-bearing mice ( Fig. 6B). Interestingly, sildenafil was sufficient to restore the proliferative capacity of otherwise anergic tumor-specific T cells in tumor-bearing mice. As previously reported ( 10, 11), T-cell anergy correlated with FOXP3 expression. In fact, more then 40% of the adoptively transferred clonotypic T cells were FOXP3+ in the tumor-bearing mice. In the absence of tumor, Tregs represented only 5-10% of the clonotypic T cells. Interestingly, sildenafil administration to A20-HA–bearing mice reduced FOXP3+ T-cell expansion in tumor-specific T cells to ∼14% ( Fig. 6C).
In this tumor model, the majority of CFSElow clonotypic T cells harvested on day 30 expressed high levels of GITR and FOXP3 and suppressed the effector function of naive T cells consistent with a Treg phenotype ( 10, 11). To confirm these previous findings and to evaluate the effect of sildenafil on Treg expansion, 30 days after transfer, HA-specific CFSElow CD4+ Thy1.1+/+ T cells were isolated from A20-HA–bearing mice ( Fig. 6A). CFSElow purified T cells obtained from (a) tumor-free, (b) tumor-bearing, or (c) tumor-bearing, sildenafil-treated mice were admixed with naive HA-specific, CD4+ T cells (CFSElow/naive T-cell 1:3 ratio) in the presence of the cognate peptide ( Fig. 6D). IFN-γ production from naive T cells was significantly impaired in the presence of CFSElow cells from A20-HA–bearing mice but not with CFSElow cells from non–tumor-bearing donors. Sildenafil nearly completely eliminated the Treg phenotype of the A20-HA CFSElow cells.
These data provide evidence that PDE5 inhibition can profoundly abrogate Treg expansion and overcome T-cell anergy in this model. Moreover, considering that the primary cellular target of sildenafil is likely the MDSC and not T cells ( 22), these data again point to the myeloid suppressor population as the mediator of tumor-induced T-cell anergy.
The early events in tumor-specific T-cell anergy involve interactions in an increasingly complex immunosuppressive network through mechanisms, including cross-presentation by functionally impaired host APCs ( 12, 13, 30, 31) and expansion of Tregs ( 10, 11). Both expansion of a preexisting regulatory cell pool and the de novo generation of Tregs from FOXP3− cells can contribute to tumor-specific T-cell tolerance ( 11). In the A20 lymphoma model, the conversion of CD4+CD25− cells into CD4+CD25+ Tregs has been recently associated with tumor-associated IDO expression ( 27). Here, we showed that the expansion of a preexisting pool of Tregs can be mediated by MDSCs.
Whereas considerable evidence exists pointing to an important role of MDSCs in the generation of cancer-induced immune suppression, their exact role within this network remains unclear. Using the A20 lymphoma model, we show that MDSCs are the tolerogenic APCs that induce antigen-specific T-cell tolerance. This occurs preferentially through cross-presentation of tumor antigens resulting in the subsequent expansion of the preexisting natural Tregs and not the generation of de novo Tregs. Although extensively studied, the role of MDSCs has thus far been only described in solid tumors ( 14) or in subcutaneous lymphoma models ( 32). In those models, MDSCs generally expand in secondary lymphoid organs and in the tumor bed where they suppress the tumor-specific T-cell response by mechanisms requiring TGF-β secretion ( 19), arginase induction ( 25, 33), and/or NO production ( 21).
Here, for the first time, we describe a role for MDSCs in a systemic hematopoietic malignancy, A20 B-cell lymphoma. In our model, MDSCs fail to expand during tumor progression ( Fig. 5B). Interestingly, however, they acquire IL4Rα expression during tumor progression—a critical requirement for MDSC-mediated suppression ( 17). IL4Rα up-regulation on MDSCs during lymphoma progression strongly suggests that systemic tumor outgrowth does not affect MDSC recruitment but rather their differentiation toward a suppressive phenotype. This is supported by our in vitro data showing that MDSCs isolated from lymphoma-bearing mice acquire arginase activity ( Fig. 2B), the ability to secrete NO ( Fig. 2B), and the capacity to suppress CD8 T-cell proliferation ( Fig. 2A). Underscoring the critical role of these cells is the recent demonstration that antigen-specific CD8 tolerance is induced by MDSC-mediated nitrosylation of tyrosine residues on the CD8 TCR ( 34). In fact, the genetic disruption in macrophages and neutrophils ( 17) or pharmacologic inhibition ( 22, 28) of this marker on MDSC is sufficient to revert their suppressive activities and to reestablish tumor immune surveillance.
Considerable interest exists in establishing links between the various immunosuppressive pathways such as MDSCs and Tregs. MDSCs, in fact, share many features with immature DCs (e.g., low expression of MHC class II, expression of CD80, the ability for antigen uptake, etc.) that are often associated with either T-cell tolerance or Treg expansion ( 30). In an allogeneic bone marrow transplant setting, the transfer of MDSCs into recipient mice suppressed the initiation of GVHD through Treg induction ( 23). Moreover, using a murine melanoma model and a rat colon carcinoma model, Ghiringhelli and colleagues elegantly showed that TGF-β–producing CD11b+/CD11c+/MHC class IIlow cells were responsible for Treg expansion, both intratumorally and in the draining lymph nodes ( 35).
Our data not only confirm previous results in solid tumors but also extend these observations to a hematologic malignancy and provide evidence as to the putative mechanisms involved. MDSCs isolated from A20-HA–bearing mice expand the preexisting Treg pool even in the absence of exogenous peptide ( Fig. 3).
Furthermore, by analyzing the effect of A20-HA–derived MDSCs and DCs, we provide new data supporting the hypothesis that different APC populations significantly influence the resultant immune response ( 36). In fact, while CD11chighMHC class IIhigh DCs from A20-HA tumor-bearing mice stimulate effector CD4+ T-cell proliferation, MDSCs promote Treg expansion ( Fig. 3B). These findings suggest that the relative number of each population can determine the immunologic outcome. Accordingly, in the spleens of mice challenged with A20-HA-GFP 28 days prior, MDSCs loaded with the antigen (Supplementary Fig. S4) greatly outnumbered their DC counterparts (GFP+CD11bhigh = 2.93% versus GFP+CD11chigh = 0.02%).
Contrary to previous reports, various pieces of evidence seem to exclude a predominant role of TGF-β in Treg expansion in our model: (a) the presence of tumor does not significantly increase TGF-β production by MDSCs ( Fig. 2B), (b) the addition of an anti–TGF-β neutralizing antibody to the culture fails to inhibit FOXP3+ T-cell expansion ( Fig. 3C), and (c) in vivo anti–TGF-β administration fails to reverse either T-cell anergy or Treg expansion (Supplementary Fig. S5). Although these differences can be attributable to intrinsic differences in the tumor type (solid versus hematologic), the compartment (lymph nodes versus spleen), or the background of the mice (C57Bl/6 versus BALB/c), these data clearly indicate that MDSCs can regulate Treg homeostasis through TGF-β–independent mechanisms. The fact that the addition of NOHA reverses MDSC-mediated Treg expansion ( Fig. 3B) suggests that this process is arginase dependent. Arginase expression plays an important role in MDSC-mediated suppression of CD8+ T-cell proliferation, either in concert with NOS to generate peroxynitrite ( 25) or by depleting the microenvironment of the semiessential amino acid l-arginine ( 33). Our data suggests a new scenario whereby modulation of Treg expansion through arginase metabolism controls effector T-cell function.
Interestingly, differences in the mechanisms used by MDSCs to suppress CD8+ effector function and to induce CD4+ Treg proliferation emerged. While arginase-1 and NO production are both required to suppress CD8+ proliferation, CD4+ Treg proliferation seems to only require arginase activity. Several hypotheses can explain these results: (a) CD4+ and CD8+ cells are differentially susceptible to l-arginine deprivation ( 37). l-Arginine has, in fact, been shown to be important for CD8+ T-cell proliferation, whereas l-arginine depletion does not inhibit CD4 proliferation. Similarly, CD4+ proliferation is unaltered in the presence or absence of NOS2, whereas CD8+ expansion is greatly inhibited ( 38). (b) Through l-ornithine, arginase can generate polyamines with different outcomes on CD4+ and CD8+ T-cell signaling. For example, putresceine can alter the Ca2+ influx in concanavalin A–stimulated CD4+ T cells, whereas no effect is observed in CD8+ T cells ( 39). (c) Finally, arginase can induce MDSCs to produce superoxide that may have different effects on effector or Tregs. Superoxide can anergize activated effector cells by lowering the affinity of the TCR ( 34). Furthermore, by down-regulating PTEN expression ( 40), it can possibly restore the ability of Tregs to respond to IL-2 and to proliferate ( 41). However, further studies are required to understand which, if any, of these pathways regulate Treg proliferation.
Although the connection between MDSCs and Tregs has been previously suggested ( 24), it is still unclear whether MDSCs increase Treg numbers through the conversion of naive T cells or the expansion of a preexisting antigen-specific Treg population. Using a system whereby the regulatory and effector populations are distinguished by the presence of the congenic markers Thy1.1 and Thy1.2, tumor-derived MDSCs expand the preexisting Treg population rather than convert naive T cells into regulatory ones in vitro ( Fig. 4B and C). Moreover, we extend these findings to our in vivo model ( Fig. 4D). By cotransferring A20-HA–derived MDSCs and regulatory Thy1.1+/Thy1.2− FOXP3+ and effector Thy1.1+/Thy1.2+ CD25−/FOXP3− T cells into naive mice, we were able to show the expansion of the preexisting Tregs. Considering that these adoptive transfer experiments were subsequently performed in non–tumor-bearing mice and thus removed from their tumor-induced immunosuppressive microenvironment, we show that A20-HA–derived MDSCs are sufficient per se to induce the proliferation of HA-specific Tregs.
We previously showed that PDE5 inhibition with sildenafil (Viagra) can restore tumor immunity by reversing the MDSC-mediated suppressive pathways in solid tumors ( 22). Here, we extend these findings to lymphoma. Sildenafil treatment down-regulated IL4Rα on MDSCs ( Fig. 5), reduced the number of tumor-specific Tregs, and reverted tumor-induced T-cell anergy ( Fig. 6). These data indicate that sildenafil can effectively reverse the immunosuppressive state in hematologic malignancies and provides additional confirmation that MDSCs play a central role in Treg induction and T-cell anergy. Whereas we cannot fully exclude that sildenafil acts solely on MDSCs, our data would suggest that it does not directly augment T-cell function. However, PDE5 inhibition may interfere with other immunosuppressive pathways in vivo that ultimately reduce T-cell tolerance and improve effector T-cell function. This hypothesis, however, requires further investigation.
Finally, it must be pointed out that despite sildenafil treatment, 14% of the clonotypic T cells remain FOXP3+ compared with the 5% to 10% present in the non–tumor-bearing mice. Although this can be attributable to the incomplete pharmacologic inhibition of PDE5, an intriguing hypothesis is that this increase derived from A20-HA–mediated IDO activity converted effector cells into regulatory ones ( 27).
In conclusion, our findings describe a tight interrelationship between host tumor-derived MDSCs, Treg induction ( 10, 11), and antigen-specific T-cell anergy ( 8, 9, 42) responsible for the immunosuppressive state associated with an increasing tumor burden in lymphoma-bearing mice. Furthermore, by demonstrating the ability to pharmacologically reverse this tolerogenic process through the functional blockade of tumor-derived MDSCs, we point to a critical role of MDSCs in mediating immune suppression and provide new hope for the immunologic treatment of hematologic malignancies.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: NIH grant 1 R01 CA124996-01A1.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Dr. Diana Lopez for her critical reading of the manuscript.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received December 11, 2007.
- Revision received April 11, 2008.
- Accepted April 16, 2008.
- ©2008 American Association for Cancer Research.