Tumor progression is facilitated by regulatory T cells (Treg) and restricted by effector T cells. In this study, we document parallel regulation of CD8+ T cells and Foxp3+ Tregs by programmed death-1 (PD-1, PDCD1). In addition, we identify an additional role of CTL antigen-4 (CTLA-4) inhibitory receptor in further promoting dysfunction of CD8+ T effector cells in tumor models (CT26 colon carcinoma and ID8-VEGF ovarian carcinoma). Two thirds of CD8+ tumor-infiltrating lymphocytes (TIL) expressed PD-1, whereas one third to half of CD8+ TIL coexpressed PD-1 and CTLA-4. Double-positive (PD-1+CTLA-4+) CD8+ TIL had characteristics of more severe dysfunction than single-positive (PD-1+ or CTLA-4+) TIL, including an inability to proliferate and secrete effector cytokines. Blockade of both PD-1 and CTLA-4 resulted in reversal of CD8+ TIL dysfunction and led to tumor rejection in two thirds of mice. Double blockade was associated with increased proliferation of antigen-specific effector CD8+ and CD4+ T cells, antigen-specific cytokine release, inhibition of suppressive functions of Tregs, and upregulation of key signaling molecules critical for T-cell function. When used in combination with GVAX vaccination (consisting of granulocyte macrophage colony-stimulating factor–expressing irradiated tumor cells), inhibitory pathway blockade induced rejection of CT26 tumors in 100% of mice and ID8-VEGF tumors in 75% of mice. Our study indicates that PD-1 signaling in tumors is required for both suppressing effector T cells and maintaining tumor Tregs, and that PD-1/PD-L1 pathway (CD274) blockade augments tumor inhibition by increasing effector T-cell activity, thereby attenuating Treg suppression. Cancer Res; 73(12); 3591–603. ©2013 AACR.
An effective immune response that leads to meaningful antitumor effects requires not only an increase in immune activation but also reduction of suppressive or inhibitory elements of the immune system (1–3). Mechanisms regulating immune activation in cancer progression have been extensively investigated (4–16). There are multiple inhibitory mechanisms that suppress immune responses, including inhibitory receptors, secreted soluble inhibitors [interleukin (IL)-10 and TGF-β], and regulatory T cells (Treg; refs. 3, 16–22). However, how these factors interact with each other in the tumor environment needs further study.
It is well established that tumors use programmed death-1 (PD-1) and CTL antigen-4 (CTLA-4) inhibitory pathways to silence the immune system. Both these pathways are critical for physiologic homeostasis. Although PD-1 is broadly expressed on activated T cells and other hematopoietic cells (23–25), CTLA-4 is expressed on activated T cells including Tregs (26). PD-1 binds 2 ligands, PD-L1 (B7-H1) and PD-L2 (B7-DC). Upregulation of PD-L1 occurs on a wide variety of human tumors suggesting that cancer cells coopt the PD-1/PD-L1 inhibitory pathway to evade the host immune response (4–16). PD-L1 can also interact with B7.1, resulting in inhibition of T-cell activation (24, 25). Despite these similarities as terminators of T-cell activation, the difference in regulatory roles of the PD-1 and CTLA-4 pathways has been recognized (27–29). CD4+CD25+Foxp3+ T cells (Tregs) play critical roles in the control of antitumor immune responses (21, 22, 30–38). Indeed, recent data have shown that blocking of PD-1 and CTLA-4 can modulate Treg functions and enhance antitumor responses (2, 35). However, the direct role and function of PD-1 in Tregs in the cancer environment remains unknown.
Immunoregulatory pathways involving PD-1, CTLA-4, and their ligands are highly complex. Coexpression of more than one receptor by CD8+ T cells has been correlated with a more severe exhausted phenotype (39–42). Antibody blockade of either PD-1 or CTLA-4 can enhance effector T-cell responses and induce T cell–mediated tumor rejection in mouse models (2, 4–15). However, the intricacies of each pathway and cross-talk between them make it unlikely that blockade of PD-1 or CTLA-4 would have identical effects. Hence, to optimally design the next generation of immunotherapeutic strategies, careful delineation of the individual contributions of PD-L1/PD-1, PD-L2/PD-1, CTLA-4/B7-1, and PD-L1/B7-1 interactions in vivo will be crucial.
In this study, we provide evidence that reversal of T-cell dysfunction can be achieved by simultaneously targeting effector T cells and Tregs. First, we show that CTLA-4 is preferentially expressed by PD-1+CD8+ T cells, and coexpression of both PD-1 and CTLA-4 is associated with marked dysfunction of antigen-specific T cells. Second, blockade of PD-1 and CTLA-4 pathways reversed T-cell dysfunction. Blockade therapy with GVAX vaccination further enhanced tumor rejection in mice. Third, adoptive transfer of CD8+CTLA-4+PD-1+ tumor-infiltrating lymphocytes (TIL) that had been pretreated in vitro with αPD-1 and αCTLA-4 antibodies eliminated tumors in vivo. Fourth, blockade of PD-1/PD-L1 pathway in Tregs attenuates their suppressive function. Taken together, our study describes an approach involving vaccination and inhibitory receptor blockade to reverse T-cell exhaustion, which we envision to be a promising anticancer immunotherapy.
Materials and Methods
Mice and cell lines
Six- to 8-week-old female BALB/c or C57BL/6 mice (The Jackson Laboratory) were used. CT26 colon carcinoma cell line was purchased from American Type Culture Collection (43). ID8-VEGF ovarian carcinoma cell line was developed previously from a mouse ovarian epithelial papillary serous adenocarcinoma cell line (ID8) obtained from Dr. Paul F. Terranova (University of Kansas, Lawrence, KS), and used as described previously (44–46).
pMG-Lyt2 vector expressing murine granulocyte macrophage colony-stimulating factor (GM-CSF) was kindly provided by Dr. J.P. Allison (MD Anderson Cancer Center, Houston, TX; ref. 2). Development of CT26 or ID8-VEGF cells expressing murine GM-CSF (CT26/GVAX or ID8-VEGF/GVAX) was based on the methods described previously (2, 47). CT26/GVAX and ID8-VEGF/GVAX cells showed secretion of murine GM-CSF 365 ng/106 cells/48 hours [95% confidence interval (CI), 231–499] and 397 ng/106 cells/48 hours (95% CI, 275–519), respectively, by ELISA (R&D Systems).
Tumor experiments and in vivo blockade
BALB/c or C57BL/6 mice were implanted subcutaneously on the right flank with either 5 × 105 CT26 or 5 × 106 ID8-VEGF tumor cells, respectively. Two hundred microgram of rat α-mouse PD-1 (29F.1A12), PD-L1 (10F.9G2), PD-L2 (3.2; ref. 24), or 100 μg of αCTLA-4 (clone 9D9) were administered intraperitoneally, either 3, 6, and 9 days following CT26 inoculation or 10, 13, and 16 days following ID8-VEGF inoculation. In vaccination experiments, 106 irradiated (150Gy) CT26-GVAX or ID8-VEGF-GVAX were given on the contralateral flank intradermally once either day 3 or day 10 after CT26 or ID8-VEGF inoculations, respectively. CD8+ T cells and CD4+CD25+ T cells from tumor leucocytes were purified either on a MACS column or FACSAria.
Adoptive transfer experiments
CD8+CTLA-4+PD-1+ TILs from CT26 tumor (25 g pooled from 80 mice) were sorted to more than 95% purity by fluorescence-activated cell sorting (FACS). Sorted TILs were treated in vitro with 10 μg/mL of αPD-1- and αCTLA-4–blocking antibodies and cultured in the presence of AH-1 peptide-loaded antigen-presenting cells (APC) for 7 days. The cells expanded an average of 10-fold in vitro in the week of culture. Following expansion, 5 × 106 purified CD8+ T cells per mouse were injected intratumorally 5 times on alternate days starting from day 10 following subcutaneously CT26 tumor inoculation.
In vitro proliferation assay
PD-1+CTLA-4+ and PD1+CTLA-4− CD8+ TIL were sorted to more than 95% purity by FACS and cocultured with splenocytes from naïve CD45.1+ C57BL/6 mice in the presence of AH-1 (CT-26) or folate receptor-α (FR-α; ID8-VEGF) peptides for 3 days (48, 49). Thymidine incorporation was measured after pulsing with 0.5 μCi/well [3H]-thymidine during the last 8 hours.
In vitro Treg suppression and cytokine analysis
Carboxyfluorescein succinimidyl ester (CFSE)-labeled CD8+ T cells (5 × 104) were stimulated with 1.5 × 105 irradiated splenic CD45.1− APCs in the presence of αCD3. CD4+CD25+ Tregs (105) from CT26 TILs, along with αPD-L1, αCTLA-4, or control rat immunoglobulin G (IgG) were added to the culture as indicated. Four days later, in vitro Treg suppression was determined on the basis of CFSE dilution by flow cytometry.
The whole tumor TILs were isolated and adjusted to 2 × 106/mL in 24-well plates and cultured with AH-1 peptide ± 10 μg/mL αPD-1, αPD-L1, αCTLA-4, or control IgG. After 3 days, supernatants were analyzed for secretion of TGF-β, IL-10, and IFN-γ by ELISA and cytokine bead array.
PD-1, CTLA-4, and their ligands are highly expressed in tumor models
To examine the interplay between the PD-1 and CTLA-4 in cancer, we used the murine CT26 colon carcinoma and ID8-VEGF ovarian carcinoma models. In the CD8+ TIL population, two thirds expressed PD-1, and the majority of these (over two thirds) were PD-1hi (Fig. 1A and B). We also found expression of CTLA-4 on TIL, however, CTLA-4 was expressed primarily in a subset of PD-1+CD8+ TIL, whereas PD-1−CD8+ cells were mostly CTLA-4−. About half of PD-1+CD8+ CT26 TIL and one third of PD-1+CD8+ ID8-VEGF TIL were CTLA-4+, whereas the remainder of PD-1+CD8+ TILs were CTLA-4−. CTLA-4 was expressed on PD-1hi as well as PD-1int cells in CT26 TILs, whereas CTLA-4 was expressed on PD-1+ cells in ID8-VEGF TILs. Taken together, among CD8+ TIL, one third coexpressed PD-1 and CTLA-4, one third expressed PD-1 only, and the remainder were negative for either receptor (Fig. 1A and B). Among the CD4+ TIL, only a third of the population expressed either of the receptors and approximately 10% were double-positive (Fig. 1A and B), whereas a quarter of splenic CD4+ T cells expressed CTLA-4, yet there were very few PD-1+ cells (Supplementary Fig. S1). We also found that CT26 and ID8-VEGF tumors upregulate PD-1 and CTLA-4 ligands on tumor and myeloid cells (Supplementary Fig. S2). Thus, CT26 and ID8-VEGF models are suitable for studying PD-1 and CTLA-4.
To begin defining the functional status of TIL, we examined their memory phenotype. Among CD8+ T cells, effector memory (TEM) cells were most prevalent in the double-positive (PD-1hiCTLA-4+) population (∼90% of CT26 and 75% of ID8-VEGF cells) but also accounted for approximately 60% (CT26) and 40% (ID8-VEGF) of the PD-1hiCTLA-4− population and half of the rare PD-1−CTLA-4Int population (CT26). One third of the PD-1Int CT26 CD8+ T cells were TEM independent of their CTLA-4 status. Both single- and double-positive CT26 CD4+ T cells had a TEM phenotype, whereas only PD-1+CTLA-4+ ID8-VEGF CD4+ T cells had a TEM phenotype (Fig. 1A and B). Thus, a significant fraction of TILs upregulated PD-1, and many of these cells were TEM. A proportion of PD-1hi TILs also upregulated CTLA-4, and these double-positive cells were mostly TEM.
Coexpression of PD-1 and CTLA-4 correlates with more severe dysfunction of tumor-specific CD8+ T cells
Given the strong association of PD-1 expression and TEM phenotype, we hypothesized that most tumor-reactive TIL are present in the PD-1+ population. To address this, we used AH-1 (gp70423–431) and H2-Kb folate receptor (FR161–169), CD8+ T-cell specific epitopes expressed by CT26 and ID8-VEGF tumor cells, respectively (45, 46). We hypothesized that the PD-1+CTLA-4+ cells exhibit greater dysfunction in response to cognate antigens than the single-positive (PD-1+) counterparts. To address this, we sorted PD-1+CTLA-4+ and PD-1+CTLA-4− CD8+ TILs, labeled them with CFSE, and cocultured for 3 days with peptide-pulsed APCs. CFSE dilution was observed only in the PD-1+CTLA-4− population but not in the PD-1+CTLA-4+ population (11.39% vs. 2.45% and 32.3% vs. 4.4% CT26 or ID8-VEGF CFSElo cells, respectively), indicating a profound suppressed state of the double-positive cells (Fig. 2A).
Next, we directly compared the functional properties of sorted PD-1+CTLA-4− or PD1+CTLA-4+ CD8 TILs by measuring cytokine secretion in response to peptide stimulation. The frequency of IFN-γ+, TNF-α+, IL-2+, or CD107a/b+ cells was approximately 2-fold higher in the PD-1+CTLA-4− subset compared with the PD1+CTLA-4+ subset (Fig. 2B; P < 0.05). The PD1+CTLA-4+ CD8 T cells displayed higher levels of other inhibitory receptors such as 2B4, LAG-3, and TIM-3, along with lower levels of CD62L and CD127 (Fig. 2B; P < 0.01). These results therefore indicate that dysfunction of CD8+ T cells in tumor was associated with coexpression of PD-1 and CTLA-4.
Next, we tested whether checkpoint blockade could rescue the proliferation of double-positive cells and/or enhance the proliferation of single-positive cells. We isolated PD-1+CTLA-4+ and PD-1+CTLA-4− CD8+ CT26 and ID8-VEGF TIL subsets and cocultured them with AH-1 or folate receptor peptide pulsed splenocytes (APCs) ± αPD-1 and/or αCTLA-4 antibodies. Again, in the absence of any neutralizing antibodies, we observed modest proliferation of the PD-1+CTLA-4− CD8+ TIL and no proliferation of the PD-1+CTLA-4+ CD8+ TIL (Fig. 2C; P = 0.019 and 0.015, respectively). PD-1 blockade alone was able to enhance the proliferation of both PD-1+CTLA-4− and PD-1+CTLA-4+ CD8+ TIL, although it was unable to completely restore the proliferation of double-positive cells. CTLA-4 blockade alone was able to enhance the proliferation only of PD-1+CTLA-4+ CD8+ TIL and had no effect on CTLA-4− cells. However, dual blockade fully restored the proliferation of PD-1+CTLA-4+ CD8+ TIL (Fig. 2C; P = 0.038 and 0.004, respectively). Thus, the double-positive population represents cells with severely dysfunctional status, but their function can be rescued by double blockade.
Combination therapy promotes CT26 and ID8-VEGF tumor rejection
Given that a significant proportion of tumor-reactive CD8+ TILs were found to be PD-1+CTLA-4+, and given that their proliferation in response to cognate tumor antigen was restored by double blockade in vitro, we tested whether simultaneous blockade of PD-1 and CTLA-4 could promote tumor rejection in mice. We found that treatment with αPD-1 or αPD-L1 showed CT26 tumor regression in 25% and 33% of the mice, respectively (Fig. 3A and Table 1). CTLA-4 blockade alone led to tumor regression in 50% of the mice. Furthermore, we noticed that combination of either αPD-1 or αPD-L1 with αCTLA-4 induced tumor regression in 75% of mice (P < 0.01). Similarly, we found that treatment with αPD-1, αPD-L1, or αCTLA-4 showed ID8-VEGF tumor regression in 25%, 37.5%, and 25% of the mice, respectively (Fig. 3B and Table 1). Combined blockade induced tumor regression in 50% of mice. In contrast, αPD-L2 blockade did not induce tumor regression in both models. Furthermore, in vivo CD4+ and CD8+ depletion completely abolished tumor regression (Supplementary Fig. S3).
|Treatment groups .||CT26 .||ID8-VEGF .|
|GVAX + αPD-1||42||37.5|
|GVAX + αPD-L1||50||50|
|GVAX + αCTLA-4||75a||37.5|
|GVAX + αPD-1+ αCTLA-4||100a||75b|
|GVAX + αPD-L1+ αCTLA-4||100a||75b|
|Treatment groups .||CT26 .||ID8-VEGF .|
|GVAX + αPD-1||42||37.5|
|GVAX + αPD-L1||50||50|
|GVAX + αCTLA-4||75a||37.5|
|GVAX + αPD-1+ αCTLA-4||100a||75b|
|GVAX + αPD-L1+ αCTLA-4||100a||75b|
NOTE: P values among different treatments were calculated using the log-rank test.
aP ≤ 0.05.
bP ≤ 0.01.
Hypothesizing that the effects of inhibitory receptor blockade could be enhanced by increasing the frequency of tumor-reactive T cells, we combined our treatment with GM-CSF–transduced whole tumor cells (GVAX; ref. 50). Remarkably, 100% of CT26-bearing mice and 75% of ID8-VEGF–bearing mice in the triple combination groups (GVAX/αPD-1/αCTLA-4) rejected their tumors (Fig. 3A and B and Table 1). Thus, double checkpoint blockade has the potential to induce significant tumor rejection, which could be maximized by boosting antitumor immune response with vaccine.
Adoptive transfer of in vitro pretreated CD8+CTLA-4+PD-1+ TILs with αPD-1 and αCTLA-4 antibodies cause regression of CT26 tumor in mice
Next, we tested whether CD8+CTLA-4+PD-1+ TILs treated in vitro with αPD-1 and αCTLA-4 would kill the tumor cells in vivo. We sorted CD8+CTLA-4+PD-1+ TILs from CT26 tumor and cultured in the presence of αPD-1- and αCTLA-4–blocking antibodies and AH-1 peptide for 7 days. We found that in vitro–treated cells showed profound increase in IFN-γ cytokine production and Ki-67 expression (Fig. 4B, left). These in vitro–treated cells were adoptively transferred into CT26 tumor-bearing mice as described in the Materials and Methods. We found that adoptive therapy resulted in tumor regression in 75% of mice (Fig. 4A). In addition, we found that the adoptively transferred TILs were still functional (IFN-γ+Ki-67+) in the regressing tumors a week after the treatment (Fig. 4B, right).
Combination blockade increases TIL activation and antigen-specific inflammatory cytokine production
Next, we tested whether tumor regression following in vivo blockade was associated with activation of TIL. We resected and analyzed tumors (n = 8) that showed evidence of rejection from 13 to 15 day onward. As expected, Ki-67 expression on TIL was markedly increased following PD-1, PD-L1, or CTLA-4 monotherapy (P = 0.001). Confirming our observation in vitro (Fig. 2), combination of αPD-1 with αCTLA-4 doubled the frequency of Ki-67+CD8+ T cells (Fig. 5A; P = 0.002), whereas αPD-L1 and αCTLA-4 combination increased the frequency of Ki-67+CD8+ T cells by 50% (P = 0.012) compared with single antibody blockade (Fig. 5A). Combined blockade also increased granzyme B expression, suggesting an enhancement in cytolytic potential of effector T cells (Fig. 5B; P = 0.01).
We corroborated the above findings by testing the phosphorylation status of key transcription factors by flow cytometry (51, 52). The ribosomal S6 kinase (S6K) is implicated in IL-2–induced T-cell proliferation (53). Validating the increased expression of Ki-67 by TIL, we observed an increase in the expression of phosphorylated (p)-S6K in CD8+ T cells following single antibody (αPD-1) blockade, and combined PD-1 and CTLA-4 blockade further enhanced pS6K levels (P = 0.01; Fig. 5C). T-bet and Eomes are 2 T-box transcription factors regulating T-helper cell Th1 and cytolytic function of CD8+ cells (54–56). αPD-L1 antibody alone induced high levels of pT-bet, whereas αPD-1 or αCTLA-4 alone or combined showed moderate increases in pT-bet levels (Fig. 5C). Validating the enhanced granzyme B expression, blocking antibodies increased pEomes expression levels (Fig. 5C). Thus, checkpoint blockade induced TIL activation, and combined αPD-1 and αCTLA-4 therapy significantly increased proliferation and cytolytic function of TIL relative to monotherapy.
We next tested whether the functional capacity of tumor-reactive CD8+ TIL increased following single versus double blockade. CT26 tumors from the various treatment groups were dissociated, and mixed cells from these tumors were seeded in primary cocultures with AH-1 peptide. Cultures derived from single antibody (PD-1, PD-L1, or CTLA-4) treated mice showed modestly increased IFN-γ+TNF-α+ or IFN-γ+IL-2+ CD8+ TILs (Fig. 5D; P = 0.05). Cultures from combined αPD-1 and αCTLA-4 treatment, exhibited increased numbers of IFN-γ+TNF-α+ (Fig. 5D; P = 0.05) but not IFN-γ+IL-2+ CD8+ TILs (Fig. 5D; P = 0.12). These data suggest that combination blockade induces polyfunctional activation of tumor antigen-specific CD8+ T cells. As expected, combining GVAX with αPD-1, αPD-L1, or αCTLA-4 inhibitory blockade further increased the frequency of CT26-specific IFN-γ+TNF-α+ CD8+ TIL (Fig. 5E; P = 0.009).
In vivo PD-1 and CTLA-4 blockade enhances infiltration of T cells and reduces Tregs in tumors
As expected on the basis of the above experiments, combined blockade not only significantly increased total inflammatory infiltration and CD8+ cells in tumors, but also CD4+ cell infiltration (Fig. 6A; P = 0.002, 0.025, and 0.001, respectively). Then, we tested whether the effects of inhibitory receptor blockade could be further enhanced by depleting Tregs. Interestingly, we could not find any additional benefit by combining αCD25 antibody with blocking antibodies (Supplementary Fig. S4). This suggested that blocking antibodies could independently affect Treg accumulation in tumors as well as tumor draining lymph nodes (Supplementary Fig. S5).
To test this hypothesis, we analyzed the frequency of Tregs in tumors of mice following single or double blockade. PD-1 or PD-L1 blockade alone produced a moderate reduction in Treg levels in the tumor (Fig. 6B and C; P = 0.024), whereas CTLA-4 blockade did not affect Treg numbers. However, combined blockade profoundly reduced the frequency of Tregs (from 9.06% to 3.24% of CD4+; Fig. 6B and C; P = 0.004). As a result, the CD8/Treg and CD4/Treg ratios increased markedly, especially after combined αPD-L1 and αCTLA-4 treatment.
We asked whether the reduction in Treg frequency was due to apoptosis. None of the blocking antibodies affected the apoptosis rate in Tregs [based on Annexin-V and 7-amino-actinomycin D (7-AAD) staining—data not shown]. Importantly, combined αCTLA-4 and αPD-1 reduced GITR (a functional marker of activated Tregs) expression from 66.7% to 45.8% of Tregs (Fig. 6B; P < 0.001), indicating that double blockade could affect not only the number but also the function of tumor-infiltrating Tregs.
Coexpression of PD-1 and CTLA-4 by Tregs is associated with heightened T-cell dysfunction in tumor
Because we found approximately one fourth of the CD4+ TIL expressing PD-1 and/or CTLA-4 (Fig. 1), we first examined whether these CD4+ TIL are effector cells or Tregs. We found that Tregs were primarily comprised in the CTLA-4+ CD4+ population. Importantly, among the CTLA-4+ cells, most were Foxp3+CD25+ cells, representing activated effector Tregs (30, 35, 37), and were PD-1+CTLA-4+ double-positive cells (Fig. 7A). Notably, splenic Tregs expressed much lower levels of PD-1 and CTLA-4 (Fig. 7A). We also observed not only increased expression of CTLA-4, but also of inducible T-cell costimulator (ICOS) and FR-4 expression on PD-1hi Treg compared with PD-1lo Treg (Fig. 7B; P = 0.001, 0.001, and 0.034, respectively), indicating an activated functional status. Thus, tumor Tregs were mostly CTLA-4+ and among them, the PD-1+ Tregs seemed to be activated Tregs.
To understand the specific contribution of PD-1+ Treg to the suppressive program, CD4+CD25+CTLA-4+ Tregs were preincubated with PD-1–blocking antibody or control IgG. Control Tregs or PD-1–neutralized Tregs were then incubated with CD8+ T cells stimulated by immobilized αCD3 antibody (responder cells) in the presence of splenic APCs (CD11c+ cells) or PD-L1–neutralized tumor-derived APCs. Responder T-cell proliferation was assessed by CFSE dilution. Although control Tregs were able to suppress responder cell proliferation, PD-1 neutralized Tregs as well as PD-L1–neutralized tumor-derived APCs were unable to suppress CD8+ T-cell proliferation (Fig. 7C; P = 0.0014). Thus, the suppressive function of tumor-derived Tregs is critically dependent on PD-1 signaling, compatible with the notion that the PD-1+CTLA-4+ Tregs are responsible for the suppressor function.
Next, we tested whether PD-1 blockade could attenuate the suppressive phenotype of Tregs and their regulatory cytokine programming (Fig. 7E–G). Moreover, we noted that suppression of CD8+ T cells by Tregs was not completely dependant on direct cell contact (Fig. 7D). To address these issues, the whole tumor digests from various in vivo treatment groups were cocultured with AH-1 peptide for 5 hours, and we measured TGF-β, IL-10, and IFN-γ in the supernatants. We detected significant amounts of TGF-β and IL-10 and minimal IFN-γ in leukocyte cocultures from untreated mice, indicating ongoing activity of Tregs and no reactivity of antitumor effector cells. TGF-β secretion was suppressed ex vivo following αPD-1 treatment and more effectively after αCTLA-4 treatment in vivo (Fig. 7E; P = 0.017). In contrast, IL-10 secretion was decreased ex vivo following αPD-1 treatment, but not after αCTLA-4 treatment in vivo (Fig. 7F; P = 0.024). Furthermore, dual αPD-1 and αCTLA-4 blockade elevated AH-1–specific IFN-γ induction (Fig. 7G; P = 0.03). Thus, checkpoint blockade reduces the ability of Tregs to secrete immunoregulatory cytokines.
In this study, we showed a non-redundant regulation of CD8+ T-cell exhaustion by PD-1 and CTLA-4 in CT26 colon carcinoma and ID8-VEGF ovarian carcinoma models. About one fourth of CD8+ TIL expressed both PD-1 and CTLA-4, and this was associated with more severe CD8+ T-cell exhaustion. Blockade of either PD-1 or CTLA-4 pathways enhanced effector T-cell infiltration into the tumor. However, simultaneous blockade of both pathways had synergistic effects in activation of tumor antigen specific CD8+ and CD4+ effector T cells as well as affecting Tregs, ultimately improving long-term survival rates. Adoptive transfer of CD8+CTLA-4+PD-1+ TILs that had been pretreated in vitro with αPD-1 and αCTLA-4 antibodies eliminated CT26 tumors in vivo, further showing the role of the PD-1 and CTLA-4 pathways in the tumor.
These effects were further augmented through combination with a GVAX vaccine. Combining blockade therapy with a single dose of GVAX vaccination resulted in enhanced intratumoral CD45+ and CD8+ T cells and as a result, the CD8/Treg ratios increased markedly. Moreover, mice treated with αPD-1/CTLA-4 combined with GVAX further increased IFN-γ+ TNF-α+ CD8+ TILs (Fig. 5E) resulting in enhanced tumor rejection. Hence our results suggest that the mechanisms underlying the synergistic effects of vaccine could possibly be due to the capacity of vaccine to increase the frequency of tumor-reactive CD8+ TILs, which could be activated by CTLA-4 and PD-1 blockade. The levels of GM-CSF produced by CT26/GVAX and ID8-VEGF/GVAX vaccines are comparable with the previously published preclinical studies using B16/GVAX and Glioma/GVAX (2, 57). The lack of in vivo efficacy of PD-1/PD-L1 blockade with GVAX in B16 melanoma seen by Curran and colleagues (2) may be due to the fact that the B16 tumor line does not express PD-L1, whereas CT26 and ID8-VEGF cells express high levels. Our results are also in line with previous observations that higher dose of GVAX recruited MDSCs into the tumor (2, 57).
The PD-1/PD-L1 pathway mediates T-cell exhaustion by antagonizing activation signaling pathways (28, 29, 58–62). We investigated our therapy's ability to counteract these effects by comparing the levels of key signaling molecules in different treatment groups. First, we looked at T-bet, which is expressed preferentially by type I effector T cells, and Eomes, which is expressed in memory T cells (61, 62). We also monitored the S6K that, when phosphorylated, acts downstream of mTOR to activate cell proliferation, protein translation, and survival (53). In untreated mice, we found that tumor infiltrating T cells showed no or very low expression of pT-bet, pEomes, or pS6K. However, upon single αPD-L1 antibody treatment, we observed a marked increase in pT-bet expression. αPD-L1 combined with αCTLA-4 antibody further increased pEomes, and αPD-1 combined with αCTLA-4 blockade further increased pS6K. That our treatments led to increased levels of these phosphorylated signaling molecules supports the contention that we were activating classical antitumor effector T cells. However, the differential effects of combining αCTLA-4 with αPD-1 versus αPD-L1 could be because αPD-1 blocks the inhibitory contributions of PD-1 binding to PD-L1 and PD-L2 but leaves the PD-L1/B7-1 inhibitory interaction unaffected (24, 25). In contrast, αPD-L1 blocks both the PD-L1/PD-1 and the PD-L1/B7-1 inhibitory pathways but leaves the PD-1/PD-L2 inhibitory interaction unaffected (24, 25).
Our experiments with sorted PD-1+CTLA-4− and PD-1+CTLA-4+ cells showed that CT26 or ID8-VEGF antigen-specific CD8+ T cells were exclusively located in the double-positive population and were highly dysfunctional, highlighting the relevance of these receptors in escape from the antitumor T-cell response. Studies report that PD-1 mediates inhibition by blocking phosphoinositide 3-kinase (PI3K) activation, whereas CTLA-4 functions through binding to the phosphatase PP2A, leading to inhibition of Akt phosphorylation (29, 59, 63, 64). Our study proves that double blockade targets both pathways resulting in an additive response by targeting both PD-1 and CTLA-4 signaling molecules allowing more effective T-cell activation, proliferation, and function of TILs, and further augmentation of therapeutic effect. Taken together, our results indicate that PD-1 and CTLA-4 pathways act non-redundantly to inhibit TIL function. This is an important lesson to bring to the clinic, highlighting why blockade of both pathways may be essential to completely regain T-cell function.
Our therapy's ability to limit the Treg population in vivo is highly desirable. Here, we found simultaneous expression of PD-1 and CTLA-4 on Tregs and distinct effects of blocking each pathway. The importance of the PD-1 pathway for Tregs, however, is not well characterized, although one recent study described PD-L1 expression by Tregs (36). Although PD-1 enhanced in vivo Treg function, CTLA-4 blockade did not impact the suppressive function of Tregs in CT26 in vivo (Fig. 6). However, our findings indicate that complete restoration of CD8+ T-cell proliferation required combined αPD-1 and αCTLA-4 blockade both in vivo as well as in vitro (Figs. 6 and 7). In our models, PD-L1 on Tregs might bind directly to PD-1 on CD8+ T cells (cell autonomous), or it may indirectly activate nearby DCs by reversing suppression through PD-1, allowing the induction of a stronger adaptive immune response (nonautonomous). We investigated which of these mechanisms was occurring. We found that blocking either PD-1 or PD-L1 on Tregs led to an increase in CD8+ T-cell proliferation (Fig. 7C). Also, in the absence of PD-L1–depleted APCs, the Tregs were not able to suppress CD8+ T-cell proliferation (Fig. 7C). This shows that PD-1 expression by effector T cells and Tregs, as well as PD-L1 expression by APCs, are all involved in the suppression of CD8+ TIL responses. We also found that dual blockade reduces production of other immunoinhibitory cytokines including TGF-β and IL-10. This provides further novel insights into the mechanisms of dual blockade.
αCTLA-4 has already been approved by U.S. Food and Drug Administration for treatment of melanoma. The antitumor activity of αPD-1 and αPD-L1 are evident in animal models, and αPD-1 and αPD-L1 antibodies are now under intense clinical evaluation (65–67). It will be critical to understand the direct effects of blockade on effector T cells as well as Tregs. Our work suggests that administration of αPD-1 antagonists may harness the therapeutic potential of effector T cells in vivo, whereas concomitantly suppressing the functions of activated Tregs. Our study provides the scientific basis for a clinical trial that would involve combination of tumor-cell vaccination and simultaneous PD-1 and CTLA-4 blockade for sustained tumor control in patients with cancer.
Disclosure of Potential Conflicts of Interest
G.J. Freeman has ownership interest (including patents) in Bristol-Myers Squibb, Roche/Genentech, Merck, EMD-Serrono, Boehringer-Ingelheim, Amplimmune, and Costim Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.
Conception and design: J. Duraiswamy, G.J. Freeman, G. Coukos
Development of methodology: J. Duraiswamy, G. Coukos
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Duraiswamy
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Duraiswamy
Writing, review, and/or revision of the manuscript: J. Duraiswamy, K.M. Kaluza, G.J. Freeman, G. Coukos
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Duraiswamy
Study supervision: J. Duraiswamy, G. Coukos
The authors thank Drs. James Allison and Michael Curran for providing GM-CSF plasmids.
This work was supported by NIH grant 2P50-CA083638 (G. Coukos), SPORE-Pilot grant (to J. Duraiswamy), and NIH-P01 AI056299 (G.J. Freeman).
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.