Transforming Growth Factor-β Receptor Blockade Augments the Effectiveness of Adoptive T-Cell Therapy of Established Solid Cancers

Adoptive T-cell therapy of solid tumors has been a goal for decades (1–3). Although significant progress has been made in animal models and in early clinical trials, this approach remains challenging, with clinical successes primarily restricted to small numbers of patients with melanomas (4–6). Steady progress has been made over the past 20 years in developing more successful strategies for adoptive transfer of T cells to treat cancer (1, 6). There have been advancements in the generation of more effective T cells for transfer by developing better ways to expand T cells, by using transgenic T-cell receptors, T-cell clones, or chimeric T-cell receptors (6–8), and progress in developing ways to increase the proliferation and persistence of transferred T cells (i.e., lymphodepletion, cyclophosphamide, total body irradiation, and concurrent vaccination with antigen; refs. 4, 9–13).

Generating antigen-specific cytotoxic T cells is not enough, however. For example, despite the presence of large numbers of antigen-specific T cells in the blood, spleen, and lymph nodes in vaccine-treated patients, tumor progression can still occur (14, 15). This suggests that additional manipulations aimed at increasing trafficking of effector T cells into tumors and/or preventing these T cells from either undergoing apoptosis or loss of effector function within the lymphoid system, the blood, and/or the tumor microenvironment could be very valuable (7).

It has been increasingly realized that the tumor microenvironment is extremely immunosuppresive due to the effects of inhibitory leukocytes (such as CD4⁺/CD25⁺ regulatory T cells, type 1 regulatory T cells, B cells, and immature myeloid and dendritic cells), down-regulation of activating chemokines and cytokines, and the presence of immunosuppressive agents, such as vascular endothelial cell growth factor, interleukin-10 (IL-10), prostaglandin E2, arginase, reactive oxygen species, indoleamine 2,3-dioxygenase, soluble Fas, soluble Fas ligand, etc. (3, 16–18). One of the most important of these T-cell immunosuppressive factors is transforming growth factor-β (TGF-β), a cytokine produced by tumor cells and nearly every immunologic cell type, including T and B cells, macrophages, dendritic cells, and platelets (19–22).

TGF-β affects both the tumor microenvironment and T cells directly. TGF-β has been shown to modulate tumor-associated macrophages and shift them from a more cytotoxic M1 phenotype to a more tumor-supportive M2 phenotype (21). An especially important consequence of this change may be the ability of TGF-β to stimulate the production of arginase by M2 macrophages (23, 24). Arginase production in the tumor microenvironment by myeloid cells inhibits T-cell receptor expression and antigen-specific T-cell responses (25). TGF-β also serves to inhibit leukocyte migration into tumors (26).

In addition, there are a number of direct effects on T cells. TGF-β affects the expression of key transcription factors and cytokines involved in T-cell development, differentiation, and activation (reviewed in ref. 21). TGF-β inhibits antigen presentation (27), shifts the T-cell repertoire to a primarily TH2 phenotype, and exerts inhibitory effects on T-cell proliferation by modulating the expression and signaling function of IL-2 and IL-2R (21). Suppression of Tbet and GATA-3 mRNA expression levels by TGF-β hinders CD4 and CD8 T-cell differentiation (20). Additionally, TGF-β supports the maintenance of FOXP3 expression, regulatory function, and homeostasis in peripheral CD4⁺CD25⁺ regulatory T cells, including those found in the tumor microenvironment (28, 29). Perhaps most importantly for adoptive transfer, TGF-β induces apoptosis in activated T cells, attenuates the acquisition and expression of T-cell effector function, and directly acts on CTLs to inhibit expression of cytolytic gene targets (such as perforin granzymes and IFN-γ; refs. 30–32).

Accordingly, we hypothesized that globally blocking TGF-β function would augment the antitumor responses produced by adoptive transfer of T cells. T cells with specificity for the immunodominant peptide of chicken ovalbumin were generated from OT-1 transgenic mice and adoptively transferred into mice bearing large EG7 thymomas expressing the cognate ligand (chicken ovalbumin) in the presence or absence of the small, orally available TGF-β type-1 receptor kinase (ALK5) inhibitor, SM16 (33). We have recently shown that SM16 is highly effective in blocking the phosphorylation of SMAD2/SMAD3 within tumors (34).

Our results show that TGF-β receptor (TGF-βR) blockade markedly augmented the efficacy of adoptive transfer, resulting in cures of most animals. Potential mechanisms included increased persistence of adoptively transferred T cells in the spleen and lymph nodes, increased numbers of adoptively transferred T cells within tumors, increased activation of these infiltrating T cells, and altered tumor microenvironment with significant increases in tumor necrosis factor-α and decreases in arginase mRNA expression.

EG7 is a derivative of a parental murine thymoma cell line (EL4) that was transfected with an ovalbumin cDNA construct. EG7 cells were cultured and maintained in media consisting of RPMI 1640 (Life Technologies) supplemented with 100 mg/mL G418, 10% fetal bovine serum, 2 mmol/L glutamine, 100 units/mL penicillin, 100 μg/mL streptomycin, 100 mmol/L sodium pyruvate, and 2.5% β-mercaptoethanol. EL4 cells were cultured in similar media without G418. The cell lines were regularly tested and maintained negative for Mycoplasma contamination. The cells grow in mice with a C57B6 background (35).

Pathogen-free female C57/B6 mice (6-8 wk old) were purchased from Taconic Laboratories. OT-1 mice express a transgenic MHC-1 restricted T-cell receptor specific for the ovalbumin peptide (SIINFEKL; ref. 36). For adoptive transfer/T-cell tracking studies, a double transgenic mouse strain in which green fluorescent protein (GFP) is expressed by all T cells (DPE^GFP) was crossed to the OT-1 strain (OT-1xgfp mice; ref. 37). The mice are on the C57/B6 background, backcrossed for more than seven generations.

Animals were housed in the animal facility at the Wistar Institute. The Animal Use Committees of the Wistar Institute approved all protocols in compliance with the care and use of animals.

OT-1 or OT-1xgfp mice were anesthetized, and spleens were harvested and strained through a 70-μm filter with PBS with 10% fetal bovine serum. RBCs were removed with lysing buffer (BD PharMingen) for 15 min at room temperature. After washing, cells were counted and placed in tissue culture at a final concentration of 1 × 10⁶/mL. OT-1 splenocytes were activated with media containing the SIINFEKL peptide (2 μg/mL; Bachem) on day 0. After 2 h, cells were spun down and placed in media without peptide. Cells were stimulated with murine IL-2 (20 ng/mL; R&D Systems) on days 2, 4, and 6, harvested, and analyzed with flow cytometry or used for adoptive transfer and T-cell tracking studies on day 7.

The following antibodies were obtained from BD Biosciences PharMingen and used for flow cytometry: anti–CD8-PE, CD8-APC, CD8-FITC, CD69-PE, CD44-FITC, CD25-PE, CD62L-APC, V-α2-PE, and CD16/32.

Phenotypic profile of T cells generated ex vivo. Activated cells were harvested and 1 million cells per sample were Fc blocked with antimouse unconjugated CD16/32 antibody for 15 min at 4°C, washed, and then incubated for 50 min at 4°C with antibodies. Data acquisition was done on a FACSCalibur (BD Biosciences), and data analysis was completed using FloJo software (Tree Star, Inc.).

Chromium-51 release cytotoxicity assays were used to determine the lytic activity of OT-1 effector cells in vitro, as previously described (38).

SM16, a 430MW ALK4/ALK5 kinase inhibitor produced by Biogen Idec, has recently been described in detail (including its chemical structure; ref. 33). This small molecule can be given i.p. or formulated in mouse chow, which allows for daily oral administration (34). We used the chow formulation for all studies. We have previously shown that SM16 chow at a dose of 0.45 g/kg of chow is well tolerated by the animals, results in therapeutic drug levels, and effectively blocks SMAD2 phosphorylation within tumor cells (34).

Tumors were established with s.c. flank injections of 1 × 10⁶ cells suspended in 100 μL PBS. Tumors were measured twice weekly, and volumes were estimated using the formula 3.14 × [largest diameter × (perpendicular diameter)²] / 6. Treatment was given when tumors were ∼200 mm³ in size, and mice followed for tumor growth. Mice were sacrificed when the tumors became >10% body weight or the mice showed signs of distress.

When the tumors reached a minimal volume of 200 mm³, one group of mice was started on SM16 chow at a dose of 0.45 g/kg chow. Three days later, 10 × 10⁶ OT-1 cells that had been stimulated for 7 d in vitro were adoptively transferred via tail vein injection. The average tumor volumes of the combination treatment group over time was plotted and compared with untreated control, SM16 chow only and adoptive transfer only groups. Each experiment had five to eight mice per group, and the study was repeated four times.

EG7 tumors were established, and mice were started on control or SM16 chow. Three days after starting chow, 20 × 10⁶ OT-1xgfp T cells activated for 7 d were adoptively transferred via tail vein injection in control or SM16 chow-fed mice. Spleens, lymph nodes, and tumors were harvested 3 d after adoptive transfer. Tissues were harvested from three groups: (a) control chow, no treatment; (b) control chow + adoptive transfer; and (c) SM16 chow + adoptive transfer. Spleens and lymph nodes were processed as described above. Tumors were chopped, and single-cell suspensions were obtained with collagenase/DNase digestion at 50°C for 1 h. Cells were run on a Ficoll gradient, filtered, washed, and counted. All cells were then subjected to fluorescence-activated cell sorting (FACS) using anti-CD8 antibody and/or gated appropriately to determine the percentage of GFP⁺ adoptively transferred cells.

Intracellular IFN-γ staining was done with a fixation/permeabilization solution kit (BD Biosciences PharMingen). Briefly, cells were isolated as detailed above, and 1.5 × 10⁶ cells were stimulated in a 96-well plate for 5 h at 37°C with either (a) 1 μg/mL of SIINFEKL peptide plus 50 units/mL of IL-2 (Roche), (b) 50 units/mL of IL-2, or (c) 50 ng/mL phorbol 12-myristate 13-acetate (PMA) plus 1 μg/mL of ionomycin. GolgiPlug (1 μL/mL) was present for the entire incubation. After several washes, cells were stained with the desired cell surface markers as detailed above, washed, fixed, and permeabilized with Cytofix/Cytoperm solution for 20 min at 4°C. After two washes with PermWash, cells were stained with a PE-labeled IFN-γ antibody for 30 min at 4°C, washed, and analyzed on Becton Dickinson FACSCalibur flow cytometer. In one set of controls, cells were first treated with an unlabeled anti–IFN-γ antibody before the labeled anti-IFN antibody to quantify nonspecific binding and set highly stringent gates.

Quantitation of EG7 tumor mRNA levels was done as previously described (39). EG7 tumor-bearing mice were started on control or SM16 chow. Three days later, tumors from these mice were harvested and total RNA isolated using Trizol reagent (Invitrogen). Semiquantitative analysis of gene expression was done using a Cepheid Smart Cycler following the manufacturer's protocol for SYBR Green kit supplied by Roche. cDNA concentrations from each gene pool were normalized using β-actin as a control gene. Relative levels of expression of each of the selected genes (fold change versus saline control) were determined. Each sample was run in triplicate or quadruplicate.

Data are presented as mean ± SE. For tumor size studies, data from multiple experiments (four studies) were normalized (by dividing each experimental value by the average tumor volume at the time of adoptive transfer) and combined. Data comparing difference between two groups were assessed using unpaired Student's t tests. Comparisons with more than two groups were done using ANOVA and confirmed with the Kruskal-Wallis test. ANOVA analyses were adjusted for multiplicity of analyses by the Bonferroni method, correcting for the number of time points where an analysis was done. Pairwise comparisons of the means of treatment groups within time points were analyzed using the Tukey method. Difference were considered significant when P < 0.05.

Generation of OT-1 effector CD8⁺T cells for adoptive transfer. To evaluate the combination therapy of adoptive transfer and TGF-β blockade, we used a well-established tumor model with a known tumor antigen: EG7 thymoma cells expressing the neoantigen chicken ovalbumin (35). T cells for transfer were derived from OT-1 transgenic mice, in which the T cells express an MHC class I–restricted T-cell receptor that recognizes the ovalbumin peptide SIINFEKL (36). As previously reported (38, 40), after splenocytes were activated by peptide and IL-2, the cells were predominately CD8⁺ (>95%). Their phenotype was CD44^HI, CD69^HI, CD25^HI, and CD62L^LO (Supplementary Fig. S1A), consistent with activated effector T cells, and they were very active in an in vitro chromium cytotoxicity assay (Supplementary Fig. S1B). EG7 cells were killed by activated OT-1 CD8⁺ T cells (>50% specific lysis) even at low E/T ratios (3:1). EL4 cells, the parental line of EG7 lacking ovalbumin expression, were not killed by activated OT-1 cells (Supplementary Fig. S1B).

TGF-βR1 blockade enhances the efficacy of adoptive transfer. To test the system under stringent conditions, we transferred activated T cells to mice with large (>200 mm³) flank tumors. An example of one such study is shown in Fig. 1A (n = 8 mice in each group). Injection of EG7 cells into the flanks of C57/B6 mice led to rapid tumor growth (Fig. 1A, diamonds). I.v. injection of 10 million activated OT-1 cells on day 12 (200 mm³ sized tumors) led to a slowing of tumor growth but no tumor regression (Fig. 1A, squares). This dose of cells is similar to that used by other investigators (38). We also treated established EG7 tumors on day 9 (>150 mm³ sized tumors) with the TGF-β inhibitor, SM16, formulated in chow at a dose of 0.45 g/kg chow. This dose was previously shown to have minimal toxicity and effectively block SMAD2/SMAD3 within tumors (34). As shown in Fig. 1A (triangles), SM16 induced a slowing of tumor growth but no tumor regression. In contrast, animals treated with SM16 chow on day 9, followed by adoptive transfer on day 12, showed complete tumor regression (Fig. 1A, crosses).

Figure 1B shows data combined from the four different independent experiments (n = 26 in each group). To control for minor differences in tumor size at the beginning of each experiment, the average tumor volume at each time point was normalized to the volume at the start of treatment. Day 0 represents the day SM16 chow was started, and adoptive transfer was done 3 d later (arrow). Again, these data show that (a) untreated tumors grew rapidly, (b) tumors treated with either SM16 chow or adoptive transfer alone experienced a delay in tumor growth, and (c) tumors treated with combined therapy had striking decreases in size. ANOVA analyses with a Bonferroni correction showed significant overall differences between treatment groups beginning on day 9 (9 d after beginning chow, 6 d after adoptive transfer). From days 9 to 16, tumors treated with the combination regimen had significantly more (P < 0.05) tumor regression than either treatment alone, an average of 80% decrease in tumor volumes, with complete regression in the majority of the tumors (Fig. 1B).

These data show that the addition of TGF-βR blockade markedly augmented adoptive transfer therapy.

Mechanisms. Given this augmented effect of combination treatment, we explored a number of possible mechanisms. We took advantage of the availability of a double transgenic mouse (OT-1.gfp), in which the OT-1 T cells are expressing GFP. Splenocytes from OT-1xgfp mice were activated for 7 days as described above. On day 7, cells underwent flow cytometric analysis to confirm the presence of both the GFP (96.9%) and the transgenic T-cell receptor indicated by the V-α2 chain (93.5%) on CD8+ T cells at the time of transfer (Supplementary Fig. S1A).

TGF-βR1 blockade increases the number of antigen-specific T cells in the spleens and lymph nodes. One mechanism by which TGF-βR1 blockade might augment adoptive transfer could be to increase the proliferation or persistence of the transferred T cells. To evaluate this possibility, we injected activated GFP-expressing OT-1 cells into tumor-bearing animals on control or SM16 chow. Three days after transfer, spleen, and lymph nodes (and tumors; see below) from treated mice were isolated and subjected to FACS analysis to identify transferred CD8⁺/GFP⁺ T cells.

FACS analysis of isolated spleens and lymph nodes from EG7 tumor-bearing mice showed a clearly detectable subset of CD8⁺/GFP⁺ adoptively transferred cells in the spleen (8-14% of the CD8+ cells; Fig. 2A) and lymph nodes (1.5-2.5% of cells; Fig. 2B). In both spleen and lymph node, there were significantly more GFP⁺ (P < 0.05) cells in SM16-treated animals versus control (Fig. 2A and B).

TGF-βR1 blockade increases the number of tumor-associated T cells. Another way by which TGF-β blockade might augment adoptive transfer could be by increasing the number of adoptively transferred cells within tumors. To assess this, FACS analysis of tumors in mice adoptively transferred with or without SM16 chow was done 3 days after cell transfer. After adoptive transfer, a small, but easily detectable, population (estimated to be ∼110,000 cells) was identified (Fig. 3A versus B ). In animals treated with SM16 chow and given OT-1 cells, however, there was a significant (P < 0.01) 3-fold to 4-fold increase in the number of transferred cells in tumors (430,000 cells; Fig. 3C). The average values for three independent experiments are shown (Fig. 3, bottom).

TGF-βR1 blockade augments the effector function of transferred CD8⁺T cells. TGF-β can inhibit the activation of effector T cells (19, 21, 31, 32). We therefore did FACS on T cells isolated from either spleens or tumors of control and treated animals and assessed the ability of these cells to produce IFN-γ in response to nonspecific activating signals (IL-2 and PMA/ionomycin) and to the specific activation signal of the OT-1 peptide (SIINFEKL) plus IL-2.

The effect of systemic TGF-βR1 blockade was determined in isolated splenocytes from adoptively transferred tumor-bearing animals. As expected, the percentages of GFP⁺ cells that expressed IFN-γ increased after stimulation with IL-2 alone (19-22%), PMA (34-42%), and peptide plus IL-2 (59-78%). However, there were no significant differences between control and SM16-treated animals in their ability to induce IFN-γ after stimulation. These results suggest that TGF-β inhibition does not increase the effector function of the transferred T cells present in spleen (Supplementary Fig. S2).

The effector activity of the tumor-associated GFP+ (transferred T) cells were also investigated. Due to the very small numbers of T cells within tumors, we pooled cells from four tumors and, after Ficoll enrichment, stimulated the cells with IL-2, PMA/ionomycin, or IL-2 plus SIINFEKL peptide and then did FACS to detect intracellular IFN-γ staining. Figure 4 shows two independent experiments. As expected, the amount of IFN-γ produced by tumor-infiltrating lymphocytes was much lower than was seen in splenocytes (41, 42). However, CD8⁺ T cells from the SM16-treated animals showed enhanced IFN-γ staining (2-fold to 3-fold increase) after nonspecific (IL-2 or PMA) or specific (IL-2 plus SIINFEKL peptide) stimulation compared with tumor-infiltrating lymphocytes from animals treated with control chow. Thus, blockade of TGF-βR signaling seems to enhance responsiveness of intratumoral T cells, but not splenic T cells after adoptive transfer.

TGF-β blockade alters the tumor microenvironment. Although it is likely that many of these effects on T cells were direct (i.e., due to blockade of TGF-β signaling on T cells themselves), TGF-βR blockade could also have effects on the tumor microenvironment, generating conditions that might favor enhanced T-cell trafficking or cytotoxic T-cell activity. To evaluate the effect of TGF-β inhibition on the tumor microenvironment in this model, EG7 tumor-bearing mice were treated with SM16 chow for 3 days, tumors were harvested, and real-time reverse transcription–PCR was done to determine the expression levels of key cytokines, chemokines, and immunomodulatory enzymes. The results in Table 1 show that tumors from animals in which the TGF-βR was blocked with SM16 displayed significant (P < 0.01) increases in the mRNA for IP-10 (CXCL10; 2-fold), MIG (CXCL9; 2.5-fold), tumor necrosis factor-α (6.4-fold), and IL-12 (2.7-fold). Up-regulation of T-cell chemoattractants and cytokine expression would be consistent with increased trafficking of CD8⁺ T cells into the tumors. In addition, the message for arginase was significantly decreased (P < 0.001) by 60%. There were no significant changes in the expression of RANTES (CCL5) or TGF-β. These results suggest that inhibition of TGF-β signaling alters the tumor microenvironment by significantly changing intratumoral expression of key immunomodulatory cytokines, chemokines, and enzymes.

As discussed in detail above (and recently reviewed in ref. 22), there are many potential mechanisms by which TGF-β in the tumor microenvironment may function to directly and indirectly inhibit the function of antitumor T cells. We thus hypothesized that blockade of TGF-β function might augment cancer immunotherapy induced by adoptive transfer of T cells.

To inhibit TGF-β–mediated effects, we used SM16, a small, highly specific, orally available ALK4/5 kinase inhibitor that effectively blocks TGF-β or activin-induced SMAD2/SMAD3 phosphorylation in vitro and in tumors (33, 34). It should be noted that any inhibitor may have off-target effects; however, SM16 seems to be quite specific. In HepG2 cells, SM16 inhibits TGF-β–induced plasminogen activator inhibitor–luciferase activity (IC₅₀, 64 nmol/L) and TGF-β–induced or activin-induced Smad2 phosphorylation at concentrations between 100 and 620 nmol/L. SM16 was tested against >60 related and unrelated kinases and showed moderate off-target activity only against Raf (IC₅₀, 1 μmol/L) and p38/SAPKa (IC₅₀, 0.8 μmol/L). SM16 exhibited no inhibitory activity against ALK family members, ALK1 and ALK6 (33).

To test our hypothesis, we used a well-established murine system of adoptive transfer, the EG7/OT1 model (35–37). There are advantages and limitations to this approach. The target antigen is not a natural tumor antigen and was highly expressed. However, although this system does not mimic most human tumors in these respects, it does allow for the generation of large numbers of uniformly activated T cells (Supplementary Fig. S1A) that are highly cytotoxic and directed against a defined antigen (Supplementary Fig. S1B). These T cells can be tracked in vivo using an antibody against the V-α2 chain of the T-cell receptor, tetramer staining, or by GFP staining (in double transgenic mice; ref. 37). To give our system the most stringent test possible, we used adoptive transfer in large (200 mm³) established flank tumors and did not supplement treatment with total body irradiation, lymphodepletion, or vaccination. Another limitation is that we only evaluated the effect of TGF-β blockade in one system. It will be important to validate this approach in other adoptive transfer models, and these studies are ongoing. However, despite these limitations, our model is very similar to animal and clinical studies using T cells with transgenic T-cell receptors or chimeric T bodies (6) and thus provides a valuable “proof of principal” for this approach.

Consistent with most published studies (see ref. 38), transfer of 10 to 20 million activated T cells under these conditions only slowed tumor growth but did not induce remission (Fig. 1). As we have seen in other models, systemic blockade of the TGF-βRI had some antitumor activity that resulted in a slowing of tumor growth (34). However, when we combined adoptive transfer with blockade of the TGF-βRI, we saw markedly augmented effectiveness with consistent tumor regressions, most often leading to complete disappearance of the tumors (Fig. 1). Given the size of the tumors at treatment, the rapid growth rate of these tumors, and the lack of any conditioning or radiation, we feel these results reveal a very potent antitumor effect.

This augmentation of efficacy seems to be the result of multiple mechanisms, as might be expected from an approach that targets a cytokine that has effects in both the T cells and the host. Similar to the studies using T cells expressing a dominant negative TGF-βRII (43–48), we observed increased numbers of the transferred T cells in the spleens and lymph nodes of the SM16-treated animals (Fig. 2A and B). We did not determine whether this was due to increased proliferation and/or decreased cell death, but this finding is consistent with the observation that TGF-β is both an inhibitor of CD8⁺ T-cell growth and an inducer of T-cell apoptosis (21, 30, 45). We postulate increased persistence of adoptively transferred cells was one factor that contributed to enhanced efficacy; however, this conclusion is tentative because measuring proliferation in this short-term model is not feasible due to the fact that the vigorous in vitro T-cell stimulation before transfer commits the cells to multiple rounds of replication (49).

Accompanying the increased numbers of adoptively transferred cells observed in the lymphoid tissue, we found increased numbers of transferred T cells within the tumor (Fig. 3). Although the actual number of labeled T cells within the tumors is relatively low, similar to observations in other studies (50, 51), treatment with SM16 increased this population by ∼4-fold. Our experimental design did not allow us to determine whether this was due to increased trafficking of cells, enhanced proliferation of T cells within the tumors, and/or decreased cell death within the tumor, but current studies are planned to address this issue.

Although we observed increased numbers of T cells within the tumors, it is well established that the tumor microenvironment can inactivate the functional capacity of these CTLs to kill tumors (16, 41, 42). An important mechanism by which this inactivation occurs is the presence of TGF-β (22, 41, 42). We thus also evaluated the “quality” of the intratumoral T cells by assessing their ability to undergo activation. As has been reported, there was an overall reduced ability of the intratumoral CD8⁺ T cells to produce IFN-γ compared with the splenic CD8⁺ T cells, consistent with local tumor microenvironmental inactivation of these cells (41, 42). However, our results showed that intratumoral T cells from the SM16-treated mice produced more intracellular IFN-γ after nonspecific (IL-2 or PMA/ionomycin) stimulation and antigen-specific (antigenic peptide) stimulation than did T cells from control-treated animals (Fig. 4). Thus, systemic TGF-βRI blockade with SM16 resulted in increased numbers of intratumoral T cells that have higher levels of effector function. We postulate that these changes account for much of the enhanced therapeutic effects of the adoptively transferred cells.

Given that SM16 is a systemic agent that affects other cell types (i.e., tumor cells, macrophages, etc.), in addition to direct effect on T cells, we also evaluated the effect of this agent on the tumor microenvironment using real-time reverse transcription–PCR of tumor samples (Table 1). Blockade of TGF-βRI signaling resulted in increased mRNA levels of a number of proinflammatory cytokines and chemokines, such as MIG (CXCL-9) and IP-10 (CXCL-10), IL-12, and tumor necrosis factor-α, that would be expected to augment trafficking of activated T cells (52). The increased levels of inducible nitric oxide synthase, IL-12, and tumor necrosis factor-α suggest that TGF-β blockade produces a more TH-1–like environment and likely affects the tumor-associated macrophage phenotype favoring a more M1 (cytotoxic) versus M2 (alternatively activated) phenotype (23, 24, 53). This observation is consistent with an earlier report showing improvement of macrophage dysfunction by administration of anti–TGF-β antibody in EL4-bearing hosts (54). Furthermore, a decrease in tumor arginase levels would also be expected to augment T-cell persistence and function (25). Thus, SM16 seems to induce a less immunosuppressive tumor microenvironment that would encourage the migration of T cells into the tumor and further enhance the ability of CD8⁺ T cells to kill tumor cells directly and indirectly (i.e., through cytokine-mediated activation of tumor-associated macrophages).

Our data are consistent with the direct effects of TGF-β inhibition on T cells, as studied by introducing a dominant-negative TGF-βR into mouse and human T cells using transgenic approaches or by transfection (43–48). This strategy has consistently led to enhanced T-cell proliferation, higher numbers of T cells within tumors, and better antitumor effects. However, there may be some limitations to this approach if applied to patients. T-cell manipulation and reinfusion in the clinical setting is technically challenging and expensive. There are concerns that malignancy could be induced by insertional mutagenesis or by the introduction of genes (like TGF-β) that could inhibit apoptosis or function as a potential tumor-suppressor gene. In support of this possibility, Lucas et al. reported that the reduced TGF-β signaling associated with the transgenic introduction of a dominant-negative TGF-βR in mice led to rapid expansion of a CD8⁺ memory T-cell population and subsequent transformation to leukemia/lymphoma (45). This side effect was not observed, however, by Lacuesta et al. (47) using multiple infusions of adoptively transferred, retrovirally transduced mouse T cells encoding a dominant-negative TGF-βR.

In summary, our data show that systemic blockade of TGF-β function using the TGF-βRI blocker SM16 markedly augmented the efficacy of adoptively transferred T cells. This augmentation is likely due to multiple mechanisms that include an altered tumor microenvironment, increased numbers of CTLs in the spleen and lymph nodes, increased numbers of CTLs within the tumors, and increased CTL effector function (as measured by IFN-γ release after stimulation) in intratumoral T cells. Although we detected only small numbers of CD4⁺/CD25⁺ regulatory T cells in this model (data not shown), inhibition of this pathway by a TGF-β function blocker may be important in other models (28, 29).

A number of TGF-β blocking agents are in active clinical development (55, 56). If these drugs show acceptable toxicity, our data indicate that systemic blockade of TGF-β function in conjunction with adoptive transfer may be useful. Even if long-term administration of these agents proves to be limited by side effects, our study suggests that one important use for these agents could be short-term administration in combination with adoptive T-cell strategies.

L.E. Ling is an employee of Biogen Idec.