Engagement of the ICOS pathway markedly enhances efficacy of CTLA-4 blockade in cancer immunotherapy

Similar to what we previously observed in cancer patients but even more dramatically, ICOS was up-regulated on CD8 and CD4 Foxp3⁻ T_eff cells in mouse B16/F10 melanoma after treatment with CTLA-4 blockade. We used irradiated parental B16 tumor cells as a control vaccination approach, which did not affect ICOS expression on any T cell compartment (unpublished data). In this situation, a very small fraction of CD8 T cells in the tumor expressed ICOS, whereas about half of CD4 T_eff cells and the majority of CD4 Foxp3⁺ regulatory T cells (T_reg cells) were ICOS positive (Fig. 1 A). Blockade of CTLA-4 in addition to the vaccination released the limit on T cell activation and generally increased ICOS expression in all of the T cell compartments, but the most significant change was observed on CD8 T cells, with a six- to eightfold increase in frequency. This trend led to a much greater presence of CD8 T cells, but much less presence of T_reg cells in the total ICOS-positive pool inside the tumor (Fig. 1 B). These data further support the rationale of activating the ICOS pathway as immunotherapy, as it would be more likely to benefit the antitumor CD8 T cells rather than immunosuppressive T_reg cells.

In light of the significant up-regulation of ICOS on intratumoral CD8 T cells, we developed a strategy to activate the ICOS pathway by transducing tumor cells with the cognate ligand, ICOSL (Yoshinaga et al., 1999), and using the irradiated ICOSL-positive tumor cells as a vaccine (IVAX) to treat tumor-bearing mice. B16/F10 melanoma cells were engineered to express ICOSL on the cell surface and tested for stable expression throughout the treatment process both in vitro and ex vivo (Fig. 2 A). ELISA of tissue culture supernatant from these cells failed to show the presence of soluble/shed ICOS (detection range >0.1 ng/ml). Mice were given an intradermal (i.d.) tumor challenge with parental (ICOSL negative) B16/F10 cells and subsequently treated with anti–CTLA-4 plus a vaccine comprised of irradiated ICOSL-negative B16/F10 cells or ICOSL-positive B16/F10 cells (IVAX). Neither IVAX nor the control vaccine of irradiated untransduced B16/F10 cells had any protective effect against tumor growth in the absence of anti–CTLA-4, probably because of poor activation status and low frequency of ICOS expression on CD8 and CD4 T_eff cells (Fig. 1 A). In this model, anti–CTLA-4 treatment alone also failed to elicit tumor rejection. The combination of anti–CTLA-4 with control vaccine resulted in tumor rejection in a minority of mice, whereas anti–CTLA-4 in combination with IVAX led to tumor rejection in >80% of mice, a fourfold increase in efficacy (Fig. 2, B–D). The increase in efficacy was specific to ICOS engagement, as the effect was completely lost in ICOS-deficient hosts (Fig. 2 E). To further test the potency of this combination therapy in a more stringent and clinically relevant situation, we doubled the initial dose of tumor challenge or delayed the onset of therapy. The combination of IVAX and CTLA-4 blockade still generated robust protection against a high dose of B16/F10 challenge (Fig. 3, A–C) or a more established tumor (Fig. 3, D–F).

The combination of IVAX and CTLA-4 blockade also improved memory response against secondary challenge in the tumor survivors (Fig. 4). We pooled the mice surviving from either the combination of control vaccine and CTLA-4 blockade or the IVAX and anti–CTLA-4 combination. At least 4 mo after their initial tumor rejection, they were challenged with a very high dose (four times higher than regular) of the same B16/F10 tumor without any further intervention. About half the mice from the control combination group obtained enough memory response against the secondary challenge, but all of the mice that had undergone IVAX and anti–CTLA-4 treatment rejected the second tumor without need of additional therapy. This is especially promising because one of the major advantages of tumor immunotherapy is immune memory.

These results suggest that the elevated expression of ICOS on T cells in anti–CTLA-4–treated tumor-bearing hosts is not just a marker for T cell activation, but ICOS can actively participate in further enhancing immune responses against tumors. Thus, in the context of CTLA-4 blockade, which leads to significant up-regulation of ICOS on CD8 and CD4 T_eff cells, a cellular vaccine expressing ICOSL can trigger the ICOS pathway to synergize with CTLA-4 blockade to provide potent tumor protection.

We next sought to dissect the basis for the enhanced efficacy of anti–CTLA-4/IVAX combination. We have previously shown that anti–CTLA-4, when combined with GM-CSF–secreting cellular vaccine (GVAX), increases the ratio of CD8 T cells to T_reg cells in the tumor (Quezada et al., 2006). To evaluate the impact of IVAX and CTLA-4 blockade on the intratumoral cell composition, we counted the absolute numbers of CD8, CD4 Foxp3⁻, and CD4 Foxp3⁺ T cells in B16/F10 tumors on day 14 and normalized these numbers by the tumor weight. IVAX alone did not change the composition of T cells in the tumor but, when administered together with anti–CTLA-4, dramatically increased the density of CD8 (approximately fivefold) and CD4 T_eff cells (approximately threefold) compared with the combination of control vaccine and anti–CTLA-4 (Fig. 5 A, left and middle). It is notable that the density of CD4 Foxp3⁺ T_reg cells was unaffected (Fig. 5 A, right). The enrichment of CD4 and CD8 T_eff cells was primarily observed at the tumor site but not in draining lymph nodes or spleen (unpublished data). Because the density of T_reg cells remained unchanged and CD8 or CD4 T_eff cells increased several fold, the combination of IVAX and anti–CTLA-4 blockade raised the intratumoral CD8/T_reg cell ratio almost sixfold and doubled the CD4 T_eff/T_reg cell ratio as compared with values in mice treated with control vaccine and anti–CTLA-4 (Fig. 5 B). The enhanced ratio of effector to regulatory T cells marked the shift from an immunosuppressive to immunostimulatory tumor microenvironment and provides one possible explanation for the potent antitumor efficacy observed with the combination therapy of anti–CTLA-4 and IVAX.

The higher intratumoral CD8/T_reg and CD4 T_eff/T_reg cell ratios could quantitatively contribute to the tumor protection effect of IVAX and CTLA-4 blockade. Furthermore, there were also profound qualitative changes in the immune functions of these enriched CD8 and CD4 T_eff cells. We isolated tumor-infiltrating lymphocytes from day 14 tumors, briefly restimulated them ex vivo with DCs loaded with B16 lysate and assayed their cytokine production and cytotoxic activity. It has been reported that ICOS signaling in human T cells induced Th17 polarization (Paulos et al., 2010). However, in our models, we were unable to detect IL-17A expression in CD4 T_eff cells isolated from the tumors or the vaccination sites (unpublished data). Although mice treated with cellular vaccine alone (either control vaccine or IVAX) or a combination of control vaccine and anti–CTLA-4 had quite few CD4 T_eff cells producing IFN-γ, there was a distinct population of polyfunctional CD4 T_eff cells secreting both IFN-γ and TNF, with an increase of ∼25-fold in the frequency of IFN-γ⁺ TNF⁺ cells after treatment with anti–CTLA-4 plus IVAX (Fig. 6 A). This cytokine profile indicated that these cells were potent tumor antigen–specific Th1 cells. We also observed that the frequency of either IFN-γ⁺ or TNF⁺ CD4 T_eff cells was significantly higher in tumors treated with IVAX and anti–CTLA-4 blockade as compared with the combination of control vaccine plus anti–CTLA-4 (Fig. 6 B). These data suggested that the synergy between IVAX and anti–CTLA-4 was capable of inducing Th1 polarization in vivo and thus providing strong help to the antitumor cytotoxic CD8 T cells. The CD8 T cells in the treated tumors did not produce as much proinflammatory cytokines (IFN-γ or TNF) as CD4 T_eff cells, although the combination of IVAX and CTLA-4 blockade still induced higher IFN-γ and TNF production than the combination of control vaccine and anti–CTLA-4 (unpublished data). However, the cytolytic activity of tumor-infiltrating CD8 T cells was considerably enhanced by IVAX and anti–CTLA-4, as measured by coexpression of granzyme B and CD107a (LAMP-1; Fig. 6, C and D), which is one of the most striking features of this novel combinatorial approach. Thus the combination of IVAX and anti–CTLA-4 not only enriched CD8 and CD4 T_eff cells in the tumors, but also enhanced their antitumor functions, including secretion of proinflammatory cytokines and cytolytic activity at the tumor site. Calculating the effect of these changes altogether, the numbers of CD4 T_eff cells producing IFN-γ and TNF and degranulating cytotoxic CD8 T cells were increased by 70- and 8-fold, respectively (Fig. 6 E). Overall, the combination of anti–CTLA-4 and IVAX could increase the density of tumor-reactive helper and killer T cells in the tumor to significantly higher levels, which might be the leading cause for tumor rejection.

The expanded pool of CD8 and CD4 T_eff cells were critical in the IVAX-driven tumor rejection. We repeated our tumor protection experiments with additional treatment with depleting antibody or with genetically deficient hosts. Mice receiving CD8 depleting antibody or lacking MHC class II molecules suffered greatly diminished therapeutic efficacy (Fig. 7). Of note, although both the CD8 and CD4 population were required for the maximum protection, other cell populations also play a role in the absence of either one, suggesting that the tumor rejection caused by this treatment regimen involves multiple helper and effector cell populations. However, the Th1 cytokine IFN-γ was indispensable for any tumor protection as IFN-γR KO mice completely lost the survival benefit (Fig. 7).

To show that the protection efficacy of IVAX combination therapy is not tumor model specific, we also used the same protocol to generate ICOSL-expressing tumor cell vaccines from another mouse melanoma cell line, B16/BL6, as well as the mouse prostate cancer cell line TRAMP C2. As we observed with the B16/F10 model, the combination of IVAX and anti–CTLA-4 can also cure TRAMP prostate cancer. All of the mice rejected TRAMP C2 tumors after treatment with IVAX plus anti–CTLA-4, as compared with the 50% protection rate in the group of control vaccine plus anti–CTLA-4 (Fig. 8 A). The control TRAMP C2 vaccine did not synergize with anti–CTLA-4, as the tumor growth kinetics is very similar between the group of anti–CTLA-4 alone and the one of control vaccine plus anti–CTLA-4 (Fig. 8 B). In contrast, there was statistically significant survival benefit with the combination of IVAX plus anti–CTLA-4 (Fig. 8 C). These data were also reproducible in a third tumor model, B16/BL6 (unpublished data). Thus, the synergy between IVAX and CTLA-4 blockade was observed across different tumor models, and in each model, the combination therapy significantly improved tumor rejection and overall survival. These results suggest that the underlying principle of this combination therapy could potentially be applied to multiple types of cancer in clinical application.

One aspect that could potentially expand the clinical application of IVAX therapy is whether the ICOS signal can be delivered independently of the cognate tumor antigen. This, if true, would lead to easier development of off-the-shelf ICOS agonist that can enhance T cell immunity against any target antigen. We tested this hypothesis by treating mice with B16 tumors with TRAMP-based IVAX or a 1:1 mixture of irradiated wild-type ICOSL-negative B16 and TRAMP-based IVAX. In this setting, the primary TCR signal (B16 tumor antigen) and the secondary ICOS signal were presented in trans. This strategy was unable to generate the same degree of tumor protection as with B16-based IVAX where the two signals were presented in cis (Fig. 9). The cognate TCR signal is required in order for ICOS signal to take effect, as TRAMP-based IVAX alone was no more effective than irradiated wild-type B16. The ICOS signal in trans did provide some additional survival benefit when compared with mice treated with irradiated B16 control vaccine alone, but the difference was not significant. This result suggests that cognate TCR signal and ICOS stimulation should be incorporated on the same vehicle for optimal therapeutic effect.

With the FDA approval of PROVENGE and more recently ipilimumab, the effectiveness of immunotherapy in the treatment of cancer is firmly established. Ipilimumab has quickly become a standard-of-care agent for the treatment of late-stage melanoma, and its application will possibly expand as results are reported from ongoing phase III trials in prostate and other tumor types. As with previous standard-of-care therapies, it will be necessary to develop combination strategies to improve clinical benefit. Here, we demonstrate that the efficacy of anti–CTLA-4 therapy is greatly enhanced by targeting the ICOS/ICOSL pathway with a cellular vaccine (IVAX).

The synergy of IVAX with anti–CTLA-4 results in a dramatic enhancement of tumor rejection. However, in the absence of ICOS up-regulation in CD8 and CD4 T_eff cells as a result of CTLA-4 blockade, IVAX monotherapy has minimal effects. This finding is consistent with previous reports of minimal efficacy when ICOS was targeted as monotherapy. For example, it has been shown that ectopic expression of ICOSL could elicit tumor-specific T cell response, but antitumor responses could only be generated in a prophylactic and not a therapeutic setting (Liu et al., 2001; Wallin et al., 2001; Zuberek et al., 2003). Similarly, engagement of the ICOS pathway with an ICOSL-Ig fusion protein alone failed to induce rejection of poorly immunogenic tumors such as B16 melanoma (Ara et al., 2003; Zuberek et al., 2003). This was probably caused by the fact that in the absence of CTLA-4 blockade, ICOS was expressed by only a few CD8 and/or CD4 T_eff cells but expressed highly by the majority of T_reg cells (Fig. 1), so that engagement of the ICOS pathway primarily stimulated the regulatory rather than the effector population. This hypothesis was supported by the fact that pretreatment of mice with cyclophosphamide, which preferentially depleted T_reg cells, could help improve the efficacy of ICOSL-Ig against poorly immunogenic tumors (Ara et al., 2003). A recent study showed that human melanoma cells expressing high levels of ICOSL facilitated expansion and IL-10 production in the T_reg cell population in the setting of high-dose IL-2 treatment (Martin-Orozco et al., 2010). In our study, however, there was no expansion of T_reg cells in mice treated with IVAX alone or with anti–CTLA-4, but the balance between effector and regulatory T cells in the tumor was enhanced considerably by the combination of IVAX and CTLA-4 blockade as the result of selective accumulation of effector CD4 and CD8 cells.

CTLA-4 blockade has been reported to selectively reduce the frequency and number of T_reg cells in B16 tumors but not in the peripheral lymphoid organs (Quezada et al., 2006; Simpson et al., 2013). The depletion is strongest in the combination of GVAX and anti–CTLA-4 on adoptively transferred TCR-transgenic Trp1 CD4 cells when compared with GVAX monotherapy. This effect is potentially driven by heavy infiltration of macrophages in the tumor (Simpson et al., 2013), which is a hallmark of GVAX therapy but not observed in the IVAX model. We did verify that CTLA-4 blockade reduced the frequency of T_reg cells in the tumor, but that was primarily because of relatively greater expansion of CD8 and CD4 T_eff cells rather than a decrease in the number of T_reg cells per milligram of tumor. In contrast, this also opens up new avenues of combination therapy by incorporating T_reg cell depletion regimen into IVAX therapy to further improve the efficacy.

The underlying mechanism of IVAX therapy is likely to be distinct from GVAX and Flt3L-secreting vaccine (FVAX; Curran and Allison, 2009). An optimal cancer immunotherapy entails ample tumor antigen presentation, co-stimulation, and/or removal of co-inhibition on T cells. Both GVAX and FVAX enhance the differentiation, maturation, and chemoattraction of DCs, which synergizes very well with CTLA-4 blockade. Although IVAX and anti–CTLA-4 combination strategy focuses more on the two facets of T cell activation, adding positive signals and blocking negative ones, using irradiated tumor cell vaccine as the vehicle to carry the positive ICOS signal still provides some help to tumor antigen presentation, but we reason that it would provide further benefit to combine the strength of IVAX and other strategies like GVAX or FVAX.

Our working hypothesis is that CTLA-4 blockade leads to enhanced activation of tumor-reactive T cells with concomitant up-regulation of ICOS, thereby enabling their responses to be enhanced by ICOS engagement. Thus, in the context of CTLA-4 blockade, IVAX triggers the ICOS pathway to enhance the proliferation, survival, and/or migration of effector cells into the tumor, which led to a higher density of T_eff cells inside the tumor, as indicated by an increase in the T_eff/T_reg cell ratio and marked increases in both Th1 CD4 T_eff and cytolytic CD8 T cells. These results clearly demonstrate a proof-of-concept that antitumor responses enhanced by CTLA-4 blockade can be greatly improved by targeting the ICOS/ICOSL pathway with ICOSL-expressing tumor cell vaccines. We are currently exploring other strategies that might be more suitable for clinical application for providing agonistic signals through ICOS locally and systemically in combination with CTLA-4 blockade. We anticipate that these combination strategies will translate to the clinic to increase the number of patients who derive benefit from anti–CTLA-4 therapy.

6-wk-old C57BL/6 and ICOS^−/− mice were purchased from the Jackson Laboratory. Mice were housed in specific pathogen–free conditions in accordance with institutional guidelines. All animal experiments were approved by the Memorial Sloan-Kettering Cancer Center Institutional Animal Care and Use Committee.

The poorly immunogenic mouse melanoma cell lines B16/F10 and B16/BL6 were obtained from I. Fidler (The University of Texas MD Anderson Cancer Center, Houston, TX) and described previously (van Elsas et al., 1999). The prostate cancer cell line TRAMP C2 was maintained as described previously (Foster et al., 1997). B16/F10, B16/BL6, and TRAMP C2 were transduced with retrovirus to express full-length mouse ICOSL on the cell membrane and tested for stable expression throughout the treatment process.

HEK 293T cells were transfected with vectors encoding full-length mouse ICOSL (provided by W. Sha, University of California, Berkeley, Berkeley, CA), envelope glycoprotein from the vesicular stomatitis virus (VSV-G), and Gag-Pol using FuGENE HD (Roche). Supernatant containing packaged virus was collected 48 and 72 h later, filtered with 0.45-µm microfilters, and applied to cultured B16/F10, B16/BL6, or TRAMP C2 cells. 5 µg/ml polybrene was also added to the virus solution. Target cells were spun at 2,600 rpm and 32°C for 2 h before being transferred into 37°C incubators. Tumor cells stably expressing ICOSL were selected by surface staining with anti-ICOSL.

Anti–CTLA-4 (9H10) was purchased from Bio X Cell and administered i.p. Antibodies for flow cytometry were purchased from eBioscience and BD.

Mice were challenged i.d. on the right flank with 5 × 10⁴ B16/F10, 2 × 10⁴ B16/BL6, or 7.5 × 10⁵ TRAMP C2 tumor cells on day 0. In experiments in which mice would be sacrificed on day 14, initial B16/F10 challenge was 2 × 10⁵. Mice were then treated with i.p. injection of 100 µg anti–CTLA-4 (clone 9H10) and i.d. vaccination on the left flank with 10⁶ irradiated (150 Gy) ICOSL-expressing tumor vaccine (IVAX) on days 3, 6, 9, and 12. The dose of anti–CTLA-4 was doubled on day 3. The mice were then followed for tumor growth or sacrificed on day 14 for dissection of lymphoid organs and tumors.

Mice used for functional experiments were sacrificed on day 14 after tumor challenge, and tumor draining lymph nodes, vaccine draining lymph nodes, and tumors were isolated. Tumors were digested in Liberase TL (Roche) and DNase I (Roche) at 37°C for 30 min, filtered, and centrifuged over Histopaque-1119 (Sigma-Aldrich). Tumor-infiltrating T cells were restimulated for 4 h at 37°C with 5 × 10⁴ DCs loaded with B16 lysate, in the presence of Golgi-Plug (BD). When cytolytic activity was measured, tumor infiltrates was incubated with anti-CD107a at 37°C for 2 h before staining with other antibodies, in the presence of monensin (BD).

Samples were stained with anti–CD4-APC–eFluor 780 (L3T4), anti–CD8-PerCP-Cy5.5 (53-6.7), and anti–ICOS-PE (17G9), fixed and permeabilized (eBioscience) according to the manufacturer’s instructions, and stained with anti-Foxp3–Alexa Fluor 700 (FJK-16s), anti–IFN-γ–Alexa Fluor 488 (XMG1.2), anti-TNF–Pacific Blue (MP6-XT22), anti–granzyme B–allophycocyanin (GB11), and anti–CD107a-PE (1D4B). Flow cytometry reference beads (PeakFlow blue; Invitrogen) were added to the samples before analysis for quantification of T cells in each tumor. The absolute number of a subset of T cells per milligram of tumor was calculated as the following example shows: density of CD8 cells = (number of beads added to each sample × count of CD8 cells/count of beads)/tumor weight.

Data were analyzed with Prism 5.0 (GraphPad Software). Experiments were repeated two to three times. Statistical significance was determined by one-way ANOVA and Bonferroni’s multiple comparisons test. Tumor survival data were analyzed with the Kaplan-Meier method. The log-rank test was used to compare survival curves for different groups on univariate analyses. P < 0.05 was considered statistically significant.

We would like to thank Dr. W. Sha for providing the retroviral vector encoding mouse ICOSL. We would also like to thank J. Geddes, J. Lu, W. Montalvo, and A. Trumble for technical assistance and M. Curran and R. Waitz for helpful discussion of the project.

This work was supported in part by a Department of Defense Idea Development Award (W81XWH-10-1-0073 01), a Cancer Prevention Research Institute Individual Investigator Award (RP120108), and a National Institutes of Health NCI R01 grant (R01CA163793-01), all to P. Sharma; a Prostate Cancer Foundation Challenge Grant in Immunology to P. Sharma and J.P. Allison; and a Cancer Research Institute Tumor Immunology Predoctoral Fellowship to X. Fan. S.A. Quezada is funded by a Cancer Research UK Career Development Fellowship and a Cancer Research Institute Investigator Award.

J.P. Allison is an inventor of intellectual property licensed by the University of California to Medarex and is a consultant for Medarex and Bristol Meyers Squibb. The authors declare no further competing financial interests.