Disruption of TGF-β Signaling Prevents the Generation of Tumor-Sensitized Regulatory T Cells and Facilitates Therapeutic Antitumor Immunity¹

It is evident from preclinical studies that tumor cells can be recognized by the immune system (1). Current research has focused on improving tumor vaccine strategies to elicit more potent antitumor immune responses (2). However, the promising results from the preclinical data have not translated into significant clinical benefit for cancer patients. It appears that multimodal therapy concepts have to be developed and tested in vivo. Recently, the combination of CTLA-4 blockade, GM-CSF-stimulated vaccination, and adoptive T cell transfer into lymphopenic hosts was shown to overcome tumor-induced immune suppression (3). Therefore, it is of great interest to elucidate the molecular and cellular mechanisms that are in place to block effective tumor vaccine responses to adjust and improve multimodal antitumor vaccine strategies.

Regulatory T (Treg)³ cells are implicated as part of the tumor-induced immune suppressive network (4). Regulatory T cells are critical mediators of tolerance against self and foreign Ags (5, 6). An increased frequency of Treg cells has been observed in patients with various types of cancer (7, 8, 9), suggesting a role for Treg cells in the development and/or progression of human malignancies. Treg cells can be divided into natural and adaptive subpopulations (10). Natural Treg cells are phenotypically characterized as CD4⁺CD25⁺Foxp3⁺ T cells (11) and acquire this phenotype in the thymus (12). In contrast, adaptive Treg cells acquire the same phenotype in the periphery (13, 14, 15). Evidence for the generation of adaptive Foxp3⁺ Treg cells outside the thymus was derived from the observation that chronic and low-dose Ag presentation leads to the generation of adaptive Foxp3⁺ Treg cells (16, 17, 18). Interestingly, in a model system using cancer cells expressing hemagglutinin, peptide vaccination against hemagglutinin led to increased vaccine-specific Treg cells (19). These data suggest that vaccination against self Ags may induce tolerance rather than productive immune responses. Investigators have tried to minimize the immune-suppressive effects of Treg cells by attempting to eliminate them. Recently, in a preclinical model the depletion of Treg cells before adoptive transfer into a lymphopenic host resulted in statistically improved tumor rejection (20). Already in a clinical study, the administration of a recombinant IL-2 diphtheria toxin conjugate before vaccination with tumor RNA-transfected dendritic cell vaccines resulted in significantly improved priming of tumor-specific T cell responses and reduced absolute numbers of Treg cells in renal cell cancer patients (21).

Multiple investigators have shown that TGF-β is a crucial mediator of tumor-induced immune suppression (reviewed in Ref. 22). TGF-β is a pleiotropic immune modulator, which influences the function of different cells of the immune system, including APCs, NK cells, and B and T cells (23). TGF-β can prevent the development of Th1 (24) and Th2 (25) responses by blocking T-bet and GATA-3, respectively. TGF-β is also known to regulate the maintenance and induction of Treg cells (26). It was shown in TGF-β1 knockout mice that TGF-β was not required for the generation of natural Treg cells in the thymus, but it was necessary for their maintenance in the periphery (27). TGF-β has also been shown to be a key cytokine for the induction of adaptive Treg cells (13). CD4⁺CD25⁻ T cells can be induced ex vivo to become Foxp3⁺ T cells with in vivo and in vitro suppressive capacity when they are stimulated in the presence of TGF-β (13, 28, 29). Two animal models have demonstrated a crucial role for TGF-β in the Treg cell-mediated prevention of autoimmunity. In these models, adoptive transfer of Treg cells could prevent autoimmunity only if they were cotransferred with T cells that had an intact intracellular TGF-β signaling pathway (30, 31).

Other models suggest that Treg cells mediate suppression by both TGF-β-dependent and -independent mechanisms. In vitro suppression of proliferation can be achieved in T cells insensitive to TGF-β (32), indicating the dispensability of TGF-β. However, there are also data from both in vitro and in vivo experiments that show that T cells insensitive to TGF-β cannot be suppressed by Treg cells (31, 33).

Work by Gorelik and Flavell, using mice whose CD4⁺ and CD8⁺ cells had a genetically impaired TGF-β receptor, demonstrated that blocking TGF-β allowed immune-mediated tumor rejection (34). However, it is unclear whether TGF-β is used by Treg cells to mediate suppression and/or is important for the generation of tumor-sensitized Treg cells that mediate suppression (35). Already antisense oligonucleotides are being employed in clinical trials to block TGF-β either at the tumor vaccine site (36, 37) or in the tumor microenvironment (38). Here, we have attempted to further define the role of TGF-β in both the generation of tumor-sensitized Treg cells and the mediation of immune suppression. We used a preclinical model where the generation of tumor-sensitized Treg cells is separated from the priming of tumor-specific effector T cells and the in vivo evaluation of therapeutic activity. The effect of TGF-β was modeled by using transgenic mice that express dominant-negative TGF-βR type II (dnTGFβRII) (39) in CD8⁺ and CD4⁺ cells (40). In these transgenic mice CD4⁺ and CD8⁺ cells have only a minimal sensitivity to TGF-β, because the receptor complex formed by the TGFβRI and the dnTGFβRII impairs TGF-β signaling due to the truncated intracellular signaling domain of the dnTGFβRII. Results of these studies document that TGF-β is critical for the generation of tumor-sensitized adaptive Treg cells that are efficient suppressors of effective T cell priming. Furthermore, we show that tumor-sensitized Treg cells can mediate effective suppression of a therapeutic immune response through a mechanism that is independent of the effect of TGF-β on T cells.

Female C57BL/6J (B6) mice were purchased from Charles River Laboratories. Female dnTGFβRII mice (40) were provided by Dr. R. Flavell (Department of Immunobiology, Yale School of Medicine, New Haven, CT) and were bred with C57BL/6J mice. Offspring were typed by PCR for the integration of the dnTGFβRII. Foxp3^GFP/GFP mice (11) were provided by Dr. A. Rudensky (Department of Immunology, University of Washington School of Medicine, Seattle, WA) and crossed to dnTGFβRII mice to generate male dnTGFβRII Foxp3^GFP/0 and female dnTGFβRII Foxp3^GFP/− mice. Recognized principles of laboratory animal care were followed (Guide for the Care and Use of Laboratory Animals, National Research Council, 1996).

D5 is a poorly immunogenic subclone of the spontaneously arising B16BL6 melanoma. D5 cells secrete 1.5 ng of TGF-β/ml/10⁶ cells/24 h. D5-G6 is a stable clone of D5 that was transduced with a murine GM-CSF retroviral MFG vector (41). D5-G6 cells secrete 40–60 ng of GM-CSF/ml/10⁶ cells/24 h. All tumor cells were cultured as described previously (42). In brief, we used complete medium (CM), which consisted of RPMI 1640 (BioWhittaker) supplemented with β-ME (Sigma-Aldrich), 10% FBS (Invitrogen), nonessential amino acids, sodium pyruvate, and l-glutamine. Cell lines were maintained in T-75 or T-150 culture flasks in a 5% CO₂ incubator at 37°C.

Mice were depleted of NK cells by i.p. injection of 500 μg of NK1.1 Ab purified from the hybridoma HB-191 (American Type Culture Collection) 24 h before injection of D5 cells. Pulmonary metastases were established in wild-type (WT) and dnTGFβRII mice by tail vein injections with 0.15 × 10⁶ and 0.5 × 10⁶ D5 tumor cells, respectively. Fifteen days after i.v. injection, tumor-bearing mice (TBM) were sacrificed by CO₂ narcosis and their spleen cells were used to reconstitute lymphopenic recipients.

Lymphopenia was induced by sublethal irradiation of mice with 500 rad (Gammacell 3000; MDS Nordion). The reconstitution with spleen cells, vaccination, and in vitro activation and expansion were performed as described previously (43). Briefly, 20 million spleen cells from female mice were used to reconstitute irradiated female mice. These reconstituted lymphopenic mice (RLM) were vaccinated with D5-G6 tumor cells. Four aliquots of 1 × 10⁶ tumor cells each were injected into both the fore and hind flanks. Eight days after vaccination, two enlarged inguinal, axillary, and scapular tumor vaccine draining lymph nodes (TVDLN) were collected, and single-cell suspensions were prepared. The TVDLN cells were cultured at 1 × 10⁶ cells/ml in CM in 24-well plates with 5 μg/ml 2c11 Ab (anti-CD3) and 2.5 μg/ml anti-CD28 mAb. After 2 days of activation, the T cells were harvested and subsequently expanded in CM containing 60 IU/ml IL-2 (Chiron) in tissue culture bags or 6-well plates for 2 or 3 additional days. These in vitro activated and expanded cells are referred to as effector T cells. The specific modifications for the individual experiments are described in the figure legends and are depicted in Figs. 1 A, 3A, and 5A.

To induce expression of Foxp3 in Foxp3⁻ cells, spleen cells from FoxP3^GFP/GFP mice were sorted into a GFP⁻ population (Foxp3⁻GFP⁻) and cultured at 4 × 10⁶ cells/8 ml/well in CM in 6-well plates with 5 μg/ml 2c11 Ab (anti-CD3), 2.5 μg/ml anti-CD28 mAb, and 120 IU/ml IL-2 and various concentrations of recombinant human TGF-β1 (eBioscience) for 48 h.

Effector T cells were washed twice in HBSS and injected i.v. in C57BL/6 female mice, in which pulmonary metastases were established 3 days earlier by tail vein injection of 0.3 × 10⁶ D5 tumor cells. Starting on the day of T cell infusion, mice received 90,000 IU of IL-2 i.p. once per day for 4 days. Animals were sacrificed 13 days after tumor injection by CO₂ narcosis. Lungs were removed and fixed in Fekete’s solution. The number of macroscopic pulmonary metastases was counted blinded, and metastases that were too numerous to count accurately were assigned a value of 250.

After in vitro activation and expansion, effector T cells were washed, resuspended in CM and IL-2 (60 IU/ml), and seeded at 1 × 10⁶/2 ml/well in a 48-well plate. The cells were either cultured without further stimulation (negative control) or stimulated with 1 × 10⁵ D5 tumor cells, MCA-310 tumor cells (syngeneic fibrosarcoma, specificity control), or immobilized anti-CD3 (positive control). Supernatants were harvested after 24 h and assayed for the release of IFN-γ by ELISA using commercially available reagents (IFN-γ; BD Pharmingen). The concentration of cytokines in the supernatant was determined by regression analysis.

All samples were acquired using Summit 4.2 software on a Dako Cyan ADP flow cytometer equipped with three diode lasers (488, 635, and 407 nm) and modified with optimal bandpass and dichroic filters. FITC-anti-CD3, PE-anti-Foxp3, allophycocyanin-Alexa 750-anti-CD4, and allophycocyanin-anti-CD25 mAbs were purchased from eBioscience. GFP⁺ and GFP⁻ cells were sorted from single-cell suspensions of spleen cells in HBSS without calcium or magnesium and 0.05% FBS using a MoFlo instrument (Dako). The purity of the sorted cells was determined immediately after sorting and was usually 95–98% for the cell populations of interest. CD25⁺ cells were purified from spleen cell suspensions in HBSS without calcium or magnesium. Allophycocyanin-anti-CD25 mAb (eBioscience) was used to stain CD25⁺ cells, followed by a second magnetic bead-linked anti-APC Ab (Miltenyi Biotec). Cell purification was performed in accordance with the manufacturer’s instructions.

We have previously reported that vaccination of lymphopenic mice reconstituted with naive WT spleen cells (RLM) augments priming of tumor-specific T cells that secrete IFN-γ when restimulated with specific tumor in vitro and can mediate significantly increased therapeutic activity (43, 44). We have also observed that this RLM strategy is ineffective when cells from 8- to 15-day TBM are used to reconstitute lymphopenic mice (62). Importantly, this suppressive effect could be eliminated by removing the CD25⁺ population from TBM spleen cells. Furthermore, adding back CD25⁺ cells from TBM spleen blocked the generation of tumor-specific effector T cells by naive T cells (C. Poehlein et al., manuscript submitted).

Given the significant role of TGF-β described in the context of Treg cells, we posited that TGF-β played a critical role in mediating the immune suppression observed in vaccinated RLM that were reconstituted with spleen cells from TBM. To test this hypothesis we examined whether T cell-insensitive to TGF-β would exhibit the immune suppressive effects that WT mice exhibited when exposed to systemic tumor burden. Spleen cells from 15-day tumor-bearing dnTGFβRII or WT mice were transferred into lymphopenic recipients that were vaccinated on the same day with D5-G6, and 8 days later TVDLN were harvested (Fig. 1,A). Consistent with observations of Poehlein et al., effector T cells generated from mice reconstituted with TBM spleen cells failed to secrete substantial amounts of tumor-specific IFN-γ (Fig. 1,B). In contrast, TGF-β-insensitive effector T cells, generated from RLM reconstituted with dnTGFβRII TBM spleen cells, secreted statistically significantly higher amounts of tumor-specific IFN-γ than did effector T cells generated from RLM reconstituted with WT TBM splenocytes. The difference between the amounts of IFN-γ secreted by effector T cells generated from RLM reconstituted with dnTGFβRII TBM spleen cells and effector T cells generated from RLM reconstituted with naive WT spleen cells was not statistically different. Evaluation of therapeutic efficacy revealed a similar profile. Adoptive transfer of effector T cells generated from WT TBM failed to mediate significant regression of established pulmonary metastases compared with mice receiving IL-2 alone (Fig. 1 C). In contrast, effector T cells generated from RLM receiving dnTGFβRII TBM spleen cells mediated a statistically significant reduction of pulmonary metastases.

Since our previous data demonstrated that CD25⁺ cells were responsible for suppression mediated by WT TBM splenocytes, we questioned if dnTGFβRII TBM spleen cells lacked CD3⁺CD4⁺CD25⁺FoxP3⁺ Treg cells. Evaluation of the frequency of Treg cells in the spleen of WT and dnTGFβRII mice revealed that both naive and TBM spleen cell populations contained comparable numbers of Treg cells (Fig. 2,C). To exclude the possibility that there was skewing of the CD3⁺ T cell compartment we measured CD3⁺CD4⁺ T cells. No significant difference was observed between these groups (Fig. 2,A). Additionally, there was no significant difference in activated T cells between dnTGFβRII and WT mice by measuring CD3⁺CD4⁺CD25⁺ cells (Fig. 2 B).

Since TGF-β-insensitive spleen cells from TBM could be primed to become effector T cells with therapeutic activity, and the frequency of total Foxp3⁺ T cells was not altered by tumor burden and insensitivity to TGF-β, we questioned whether Treg cells from TBM were unable to suppress the priming of tumor-specific dnTGFβRII T cells in the RLM model. To test this hypothesis we attempted to prime naive dnTGFβRII spleen cells in the presence of Foxp3⁺ Treg cells from WT TBM. If TGF-β was the primary mechanism of suppression in this model, the tumor-sensitized Treg cells from WT mice would be unable to suppress the generation of tumor-specific effector T cells from dnTGFβRII mice. Lymphopenic mice were reconstituted with either naive WT or naive dnTGFβRII spleen cells together with sorted GFP⁺ Treg cells from spleens of female WT Foxp3^GFP/− TBM (Fig. 3,A). Effector T cells generated from RLM reconstituted with WT naive spleen cells could be primed to become tumor-specific T cells, and the addition of GFP⁺Foxp3⁺ Treg cells from WT TBM to the naive spleen cells used in the RLM model effectively eliminated the tumor-specific IFN-γ response (Fig. 3,B). Similarly, effector T cells generated from RLM reconstituted with naive dnTGFβRII spleen cells also showed a tumor-specific IFN-γ response that, while lower, was not statistically different than the response of effector T cells generated from RLM reconstituted with WT naive spleen cells. Unexpectedly, the addition of GFP⁺Foxp3⁺ T cells from WT TBM to naive dnTGFβRII spleen cells at the time of reconstitution eliminated the tumor-specific IFN-γ response of the naive dnTGFβRII T cells (Fig. 3 B). Thus, we conclude that TGF-β signaling does not play an essential role in mediating the suppression of priming and/or effector T cell generation in this model and that there are redundant mechanisms, besides TGF-β, that can suppress the priming of tumor-specific effector T cells in RLM.

Since dnTGFβRII T cells could be suppressed by tumor-sensitized WT Treg cells, we next asked whether the absence of suppression observed in Fig. 1,C was because dnTGFβRII mice failed to generate tumor-induced Treg cells. Furthermore, we reasoned that since the insensitivity to TGF-β did not alter the frequency of the Foxp3⁺ T cell compartment that this compartment consists primarily of natural Treg cells. We also questioned whether we missed a difference between Treg cells from WT and dnTGFβRII mice by measuring total Foxp3⁺ cells. Recently it was published that naive Foxp3⁻ T cells converted into Foxp3⁺ cells in the presence of TGF-β (45, 46). Based on this observation we hypothesized that Foxp3⁻ cells from dnTGFβRII mice could not convert into Foxp3⁺ adaptive Treg cells. To examine this hypothesis naive male WT or dnTGFβRII Foxp3^GFP0/+ spleen cells were sorted for GFP⁻(FoxP3⁻) cells and cultured with anti-CD3 mAb, anti-CD28 mAb, IL-2, and 1 ng/ml recombinant human TGF-β. Sort purity was 98% for GFP⁻ spleen cells (data not shown). As hypothesized, the frequency of CD4⁺Foxp3⁺ cells increased significantly when WT spleen cells were cultured in the presence of TGF-β (Fig. 4,A). Furthermore, the de novo induction of Foxp3 expression by TGF-β in WT cells was dose dependent (Fig. 4 B). In contrast, dnTGFβRII T cells failed to express Foxp3 even when cultured with the highest concentration of TGF-β.

Since the in vitro data demonstrated that FoxP3⁻ cells from dnTGFβRII mice could not be induced to express FoxP3, we hypothesized that Treg cells from dnTGFβRII TBM would contain only natural Treg cells and no tumor-sensitized adaptive Treg cells. Since there are no phenotypic markers that can separate natural from adaptive Treg cells, one way is to identify them by their functional suppression of tumor-specific T cells in the RLM model. To test whether dnTGFβRII mice generate tumor-sensitized adaptive Treg cells that can suppress the generation of tumor-specific effector T cells we reconstituted lymphopenic mice with naive spleen cells together with CD25⁺ cells from either WT TBM or dnTGFβRII TBM mice (Fig. 5 A).

As shown previously, effector T cells generated from RLM reconstituted with WT naive spleen cells could be primed to become tumor-specific T cells as demonstrated by their ability to secrete IFN-γ when restimulated with tumor (Fig. 5,B), whereas the addition of CD25⁺ cells from WT TBM essentially eliminated the tumor-specific IFN-γ response. In line with that observation therapeutic efficacy was significantly reduced by the addition of CD25⁺ cells from WT TBM to the naive WT spleen cells used in vaccinated RLM (Fig. 5,C). In contrast, the addition of CD25⁺ cells from TBM dnTGFβRII at the time of reconstitution did not significantly reduce the tumor-specific IFN-γ response (Fig. 5,B) or the therapeutic efficacy of effector T cells generated from these mice (Fig. 5 C). Thus, we conclude that spleen cells from TBM that are insensitive to TGF-β do not contain tumor-sensitized adaptive Treg cells, because in vitro these dnTGFβRII cells cannot be induced to express Foxp3, and phenotypical Treg cells from tumor-bearing dnTGFβRII mice fail to suppress the generation of tumor-specific effector T cells in the RLM model.

Treg cells are considered to be an important component of the immunosuppressive environment caused by cancer (47, 48). Therefore, it is important to understand mechanisms by which Treg cells are induced and how they block antitumor immunity to develop multimodal therapy strategies that dampen the immune suppressive effects of Treg cells. Our findings, as well as observations by others (reviewed in Ref. 49), support the notion that TGF-β is a crucial molecule involved in the suppression of antitumor immunity. Secretion of TGF-β is not limited to our mouse model, as multiple studies have demonstrated secretion of TGF-β by human tumors, including tumors of the breast (50, 51), colon (52, 53), and pancreas (54). Furthermore, TGF-β secretion is associated with metastatic spread and disease progression indicating its important role in human carcinogenesis (51, 52, 53). Our studies identified that TGF-β plays an essential role in the generation of tumor-sensitized Treg cells, but it is not required as the suppressive molecule that blocks tumor-specific priming in the RLM model. We show that Treg cells (Foxp3⁺ and CD25⁺ cells) obtained from WT TBM mediated immune suppression, while Treg cells from dnTGFβRII bearing equivalent tumor burden were not suppressive. Furthermore, the suppressive mechanism used by tumor-sensitized Treg cells was independent of TGF-β acting directly on the responding T cells. Therefore, we conclude that the lack of immune suppression observed in dnTGFβRII TBM is due to a deficiency of tumor-sensitized adaptive Treg cells or the inability of tumor-sensitized natural Treg cells to be activated by TGF-β in the tumor bearing host. We consider these findings to be clinically important, as we are investigating this RLM strategy in clinical trials with cancer patients (55). Our current protocol for reconstituting patients made lymphopenic using chemotherapy is to reinfuse an apheresis product depleted of Treg cells. Given the observation that TGF-β is necessary for the de novo induction of tumor-sensitized Treg cells, the blockade of TGF-β could be used to prevent the generation of new tumor-sensitized Treg cells in reconstituted patients.

It could be argued that dnTGFβRII CD25+ cells from tumor-bearing mice contain effector T cells that could mediate the regression of pulmonary metastases upon adoptive transfer. This would be in accordance to what has been shown by Fahlen and colleagues in a colitis model (33). However, note that the ability to prime effectors from dnTGFβRII mice is lost if WT Treg cells are cotransferred into RLM (Fig. 3). Thus, the susceptibility of dnTGFβRII T cells to Treg cell-mediated immune suppression is evident, and it suggests that the critical element for priming dnTGFβRII tumor-specific T cells in the TBM is the lack of functional tumor-sensitized Treg cells. In the same colitis model Fahlen and colleagues demonstrated that TGF-β is not the direct suppressor molecule used by Treg cells. In that model Treg cells from TGFβ-1^−/− mice were suppressive. They concluded that there were other sources of TGF-β besides Treg cells, because systemic blockade of TGF-β with an Ab resulted in the abrogation of the immune suppression. Our data are in agreement with the findings of Fahlen and suggest that mechanisms other than TGF-β secretion by Treg cells can suppress the priming of tumor-specific T cells in vaccinated RLM. In line with our observation, Piccirillo et al. showed in vitro that CD4⁺CD25⁻ dnTGFβRII T cells could be suppressed by WT CD25⁺ T cells and that WT CD4⁺ T cells could be suppressed by CD4⁺CD25⁺ TGF-β^−/− T cells, indicating the dispensability of TGF-β for direct suppression of T cell priming (32). There are other possible mechanisms by which Treg cells can mediate their suppressor function. For example, CTLA-4 expressed on Treg cells can indirectly prevent activation of T cells by dampening the costimulatory signals provided by APCs (56), and IL-10 secreted by Treg cells can mediate immune suppression (57, 58). Moreover, it has been reported that Treg cells dampen the immune response by killing effector T cells through granzyme B in a contact-dependent manner (59). Additionally, IL-35 has been introduced as a potential Treg cell-specific suppressor molecule (60).

The observation that insensitivity to TGF-β did not alter the frequencies of CD3⁺CD4⁺CD25⁺Foxp3⁺ T cells in naive mice is in accordance with published data showing that naive dnTGFβRII mice had the same level of Foxp3 RNA expression in CD4⁺CD25⁺ T cells as did their WT littermates (33). Fahlen’s and our observation suggest that WT and dnTGFβRII mice generate the same amount of FoxP3⁺ Treg cells. Tumor burden also did not lead to a measurable increase in the frequency of total Treg cells after 15 days. This is in contrast to what has been reported in other clinical (7, 8, 9) and preclinical models (61). However, T cells from WT TBM spleens were suppressed from becoming tumor-specific effector T cells, and depletion of CD25⁺ cells from WT TBM spleen cells could reverse this suppression (C. Poehlein et al., manuscript in preparation). Additionally, this suppressive effect was mediated by the CD25⁺ fraction of TBM spleen cells, as the addition of small numbers of CD25⁺ cells from the spleen of a TBM abolished the priming of effector T cells with therapeutic efficacy in the RLM model. Additionally, we show herein that sorted FoxP3⁺GFP⁺ cells from WT TBM were also potent suppressors of effector T cell priming. Taken together, these observations argue for the generation of tumor-sensitized Treg cells with strong immune suppressive capacity. However, phenotypically, there is no increase in the frequency or absolute number of FoxP3⁺ T cells in these TBM with highly functional suppressors. We have speculated that this may reflect a homeostatic mechanism that controls the total number of Treg cells at a certain level. Such a mechanism, while controlling the total number of Treg cells, would allow for fluctuations in the Treg cell repertoire. Furthermore, we postulate that tumors sensitize or induce Treg cells from either natural existing Treg cells or from the FoxP3⁻ population (adaptive Treg cells). Therefore, the difference between WT and dnTGFβRII mice seems to be the ability to generate a small population of adaptive tumor-sensitized Treg cells, which are indistinguishable from natural Treg cells. In an attempt to address this question we examined the ability of FoxP3⁻ (GFP⁻) WT and dnTGFβRII CD4⁺ T cells to convert to FoxP3⁺ Treg cells. T cells were activated in vitro in the presence of TGF-β to induce FoxP3⁺ T cells, as shown by others (45, 46). While WT Foxp3⁻ T cells converted to FoxP3⁺ cells in the presence of TGF-β, FoxP3⁻ dnTGFβRII T cells failed to express FoxP3.

These results indicate a T cell-intrinsic mechanism by which activated T cells turn on Foxp3 in the presence of TGF-β. These in vitro data suggest that dnTGFβRII T cells are unlikely to generate adaptive Treg cells. Since dnTGFβRII mice maintain levels of FoxP3⁺ T cells comparable to WT mice, we postulate that FoxP3⁺ populations in dnTGFβRII mice are primarily natural Treg cells that are not dependent on TGF-β for their generation. Given that they do not generate the potent suppressors of tumor-specific T cells, we consider that they lack tumor-sensitized adaptive Treg cells. An alternative explanation is the possibility that TGF-β produced in the tumor-bearing host activates WT natural Treg cells to exert their immune-suppressive capacity.

Taken together, these data show that T cells insensitive to TGF-β can still be suppressed by WT tumor-sensitized Treg cells, but that the generation of tumor-sensitized Treg cells is dependent on TGF-β. If, as we predict, tumor-sensitized Treg cells are adaptive Treg cells, our results argue that blockade of TGF-β before vaccination might reduce the highly suppressive tumor-sensitized Treg cells that may be induced in the tumor-bearing host. This strategy may prevent induction of new tumor-sensitized Treg cells as well as eliminate the role of TGF-β in maintaining adaptive Treg cells in the periphery (27). Collectively, these data suggest a rationale for TGF-β blockade as an adjunct to cancer vaccines in patients enrolled on clinical immunotherapy trials.

The authors have no financial conflicts of interest.