Blockade of CTLA-4 and PD-1 Enhances Adoptive T-cell Therapy Efficacy in an ICOS-Mediated Manner

Adoptive transfer of autologous ex vivo–expanded T cells and genetically engineered tumor-reactive T cells [chimeric antigen receptor T cells (CAR-T cells); refs. 1–5], collectively known as adoptive T-cell therapy (ACT), is being used to eradicate tumor cells in patients. ACT with autologous ex vivo–expanded T cells has shown modest clinical benefits in solid malignancies including metastatic melanoma (6, 7). Treatment failures with this therapy are primarily due to inadequate migration of adoptively transferred cells into the tumor microenvironment and poor function and persistence of transferred T cells. Investigations are under way to evaluate novel strategies to improve clinical outcomes.

Previous studies have demonstrated that adoptively transferred T cells express immune-checkpoint receptors such as PD-1. PD-1^hi cells produce less effector cytokines such as interferon gamma (IFNγ) compared with PD-1^low adoptively transferred cells (8). Furthermore, blockade of PD-1 increases migration of adoptively transferred T cells to the tumor site in a murine model of melanoma (9). However, the survival benefit of combining with ACT with anti–PD-1 therapy has not been documented. Similarly, ACT has been combined with anti–CTLA-4 therapy (10, 11); however, this study (11) used a genetically engineered B16F10 melanoma model expressing the lymphocytic choriomeningitis virus peptide GP33 (B16F10-GP33), which is a foreign antigen GP33 and might not reflect the immune responses against endogenous tumor antigens.

In our study, we hypothesized that the combination of anti–CTLA-4 plus anti–PD-1 would improve the efficacy of ACT and provide a survival benefit. To test our hypothesis, we used the PMEL-1 transgenic murine model, which consists of CD8⁺ T cells bearing a specific T-cell receptor for the gp100 antigen (12). In this model, we noted that the gp100 peptide induces expression of CTLA-4, PD-1, and ICOS on transgenic CD8 T cells. We demonstrated that the combination of anti–CTLA-4 plus anti–PD-1 therapy with ACT consisting of ICOS⁺PMEL-1 CD8⁺ T cells, compared with monotherapy using either anti–CTLA–4 or anti–PD-1 or ACT alone, led to improved antitumor responses and durable survival benefit. To determine whether ICOS played an important role in PMEL-1 Tg⁺ CD8 T cells, we generated ICOS^−/−PMEL-1 mice. Deletion of ICOS largely abolished the therapeutic benefits of ACT combined with immune-checkpoint blockade (ICB), which coupled with the selective downregulation of Eomesodermin (Eomes), IFNγ, and perforin in PMEL-1 Tg⁺ CD8 T cells. Collectively, our study showed the critical role of the ICOS signaling in sustaining effector function of tumor antigen–specific PMEL-1 Tg⁺ CD8 T cells. In conclusion, our preclinical studies provide a strong rationale for combining anti–CTLA-4, anti–PD-1, and ACT to treat solid malignancies.

Tumor biopsy samples were collected from patients with metastatic melanoma after appropriate informed consent was obtained on MD Anderson IRB-approved protocol PA13-0291. Patients with localized, high-risk prostatic adenocarcinoma who were candidates for radical prostatectomy were consented on an MD Anderson Cancer Center IRB-approved protocol (MDACC 2009-0135; NCT01194271, N = 19) to receive one intramuscular injection of ADT (leuprolide acetate; 22.5 mg) on week 0 and ipilimumab at 10 mg per kg body weight per dose on weeks 1 and 4 (13). Peripheral blood mononuclear cell and tumor biopsy samples were collected for immune-monitoring studies after appropriate informed consent was obtained on MD Anderson IRB-approved protocol PA13-0291. All the studies were conducted in accordance with ethical standard of Declaration of Helsinki.

C57BL/6 (B6) mice were purchased from The Jackson Laboratory and the NCI. ICOS^−/−, ICOSL^−/−, Rag-1^−/−, CD45.1, and human gp100-TCR transgenic mice (PMEL-1 mice, congenically marked with Thy1.1), all on the B6 background, were obtained from The Jackson Laboratory. PMEL-1ICOS^−/− mice were generated by crossing PMEL-1 mice with ICOS^−/− mice, and ICOS deficiency in those mice was confirmed by PCR. ICOS-YF mice were generated as previously described (14). These mice possess a tyrosine-to-phenylalanine mutation at amino acid residue 181 in the cytoplasmic tail of ICOS, which abrogates ICOS-mediated PI3K recruitment and subsequent signaling cascade. All mice were kept in specific pathogen-free conditions in the Animal Resource Center at The University of Texas MD Anderson Cancer Center. Animal protocols were approved by the Institutional Animal Care and Use Committee at MD Anderson.

The mouse melanoma cell line B16-F10 was obtained from I. Fidler in 2006 (The University of Texas MD Anderson Cancer Center, Houston, TX) and described previously (15). These cells were not authenticated in the past year. Freshly thawed B16-F10 tumor cells were cultured for 3 days to reach 60% to 80% confluency, passaged once and cultured for an additional 2 days, and confirmed to be Mycoplasma free prior to tumor inoculation. Mouse and human melanoma differential antigen-derived peptide, H-2D^b-restricted hgp100_25-33 (KVPRNQDWL; ref. 12), and control LCMV gp_33-41 or LCMV_33-41 (KAVYNFATC) peptide were synthesized by Invitrogen with a purity >95%. Mouse CTLA-4 (clone 9H10) and PD-1 (RMP1-14) blockade antibodies were purchased from Bio X Cell.

Isolated splenocytes or lymph node cells (2 million/mL) from PMEL-1 or PMEL-1ICOS^−/− mice were stimulated with titrated hgp100_25-33 (0.05, 0.2, and 1 μg/mL; Invitrogen) or control peptides in complete culture medium (Click's medium supplemented with 10% FBS, 100 units/mL penicillin, 10 mg/mL streptomycin, 2 mmol/L L-glutamine, and 50 μmol/L 2-mercaptoethanol [2-ME], all from Invitrogen) at 37°C, in 5% CO₂ for 48 and 72 hours. In vitro–activated cells were either further incubated for an additional 4 hours in the presence of 3 μmol/L monensin (BD Biosciences) for intracellular cytokine staining or were directly examined for expression of ICOS or transcription factors, as described below.

Harvested splenocytes and total lymph node cells from PMEL-1 or PMEL-1ICOS^−/− mice were briefly activated with 1 μg/mL hgp100 overnight. Then, CD8⁺ bead–purified cytotoxic T lymphocytes (CTL; 6 million) were transferred into recipient Rag-1^−/− mice or into sublethally irradiated CD45.1 and ICOSL^−/− mice (400 rad/mouse using a sealed irradiator with a Cs¹³⁷ source, MARK I irradiator from J.L Shepherd & Associates) that had been allowed to rest for more than 4 hours before the intravenous infusion of CTLs, followed by immunization with 50 μg of hgp100_25-33 emulsified in CFA (Complete Freund's Adjuvant) per mouse on the next day. Transferred Thy1.1⁺ CTLs collected from spleen and lymph nodes were then analyzed for intracellular cytokine production after brief stimulation with 50 ng/mL phorbol-12-myristate 13-acetate (PMA; Sigma) plus 1 μmol/L ionomycin (Sigma) in the presence of monensin or were directly examined for expression of transcriptional factors, as described below.

Harvested splenocytes and total lymph node cells from PMEL-1 or PMEL-1ICOS^−/− mice were briefly activated with 1 μg/mL hgp100 overnight in Click's medium supplemented with 10% (vol/vol) FBS and 1% (vol/vol) penicillin–streptomycin. Retroviruses (RV) were prepared as described (16) by using constructs linked with a GFP marker. In brief, ICOS and Eomes cDNAs were inserted into the polylinker site of the empty retroviral vector (a generous gift from Dr. Hongbo Chi lab at St. Jude Children's Research Hospital). For retroviral transduction, the above-activated CTLs were transduced with RV encoding either ICOS or Eomes by “spin inoculation” (650 g for 1 hour), as described previously (16). Cells were cultured for an additional 4 or 5 days to analyze IFNγ production by intracellular cytokine staining (17) and Eomes expression after permeabilization and fixation of cells using the transcription factor fixation/permeabilization kit from eBioscience. For PMEL-1 and PMEL-1ICOS^−/− CD8 T cells transduced with ICOS RV, GFP⁺ cells were sorted using the BD FACSAria II sorter, and 1 million sorted cells were injected into B16-F10 tumor–bearing mice as ACT, as described below.

Draining lymph node (DLN) cells from immunized (CFA/hgp100) PMEL-1 and PMEL-1ICOS^−/− mice were further stimulated ex vivo with 1 μg/mL hgp100_25-33 peptide for 3 days. CTLs were then harvested from cultured DLN cells to set up coculture with tumor cells to examine their cytotoxic capacity, as described previously (18). Briefly, B16/F10 cells were washed once with D-PBS and resuspended at 5 × 10⁶/mL in labeling buffer containing 0.6 μmol/L DDAO-SE (Invitrogen) in D-PBS and incubated at 37°C for 15 minutes with periodic mixing. The cells were washed in culture medium and resuspended in culture medium at 1 × 10⁶ cell/mL before they were mixed thoroughly with the effector CTLs at designated effector-to-target ratios (E:T ratios). The cell mixtures were then centrifuged at low speed (200 rpm) for 1 minute and incubated in a humidified incubator at 37 °C, 5% CO₂ for 3 hours. After culture, DDAO-SE⁺ tumor cells from the cocultures were stained for Annexin V and 7-AAD to detect apoptotic cells (Annexin V⁺7-AAD⁺). CTLs from the cocultures were briefly stimulated with PMA and ionomycin and then were subjected to intracellular cytokine staining for IFNγ and IL2, as well as surface expression of ICOS.

For analysis of surface markers, cells were stained in PBS containing 2% (wt/vol) BSA, with anti–CD4-PB (RM4-5, BioLegend, 100531, 1:200), anti–CD8α-BV786 (53-6.7, BD Horizon, 563332, 1:200), anti–TCRβ-APC Cy7 (H57-597, BD Pharmingen, 560656, 1:200), anti–Thy1.1-AF700 (OX-7, BioLegend, 202528, 1:200), and anti–ICOS-BV510 (C398.4A, BioLegend, 313525, 1:200) on ice for 30 minutes. Intracellular anti–Blimp-1-AF647 (5E.7, BD Pharmingen, 563643, 1:100), anti–T-bet-BV711 (4B10, BioLegend, 644819, 1:100), anti–Eomes-PE (Dan11mag, eBioscience, 12-4875-82, 1:100), anti–IFNγ-PE-eF610 (XMG1.2, eBioscience, 61-7311-82, 1:200), and anti–IL-2-PE-Cy7 (JES6-SH4, eBioscience, 25-7021-82, 1:100), anti–TNFα-BV650 (MP6-XT22, BD Pharmingen, 563943, 1:100), and anti-Perforin (eBioOMAK-D, eBioscience, 12-9392-82, 1:100) were analyzed by flow cytometry according to the manufacturer's instructions. Dead cells were excluded from analysis using a fixable live stain (LIVE/DEAD fixable blue dead cell stain, freshly prepared from Invitrogen, L23105, 1:300). Flow cytometry data were acquired on LSRII or BD FACS Canton II (BD Biosciences), and singlets based on SSW and FSA were analyzed by using FlowJo software (Tree Star).

Rag-1^−/− or CD45.1 mice (5–10 mice/group) were inoculated intradermally in the right flank with 6 × 10⁵ B16-F10 melanoma cells on day 0. On day 4, CD45.1 mice with palpable tumors were sublethally irradiated (400 Rad), and 4 hours later, 5 million activated PMEL-1 or PMEL-1ICOS^−/− CTLs were adoptively transferred via tail-vein injection (ACT), followed by subcutaneous immunization with CFA/hgp100_25-33 (50 μg/mouse) on the next day. Anti–CTLA-4 and anti–PD-1 antibodies (200 μg/mouse) were intraperitoneally injected on day 6 after tumor inoculation, and in some experiments, a booster dose of 100 μg/mouse was given on day 9. Rag-1^−/− mice with palpable tumors were directly used without sublethal irradiation. Tumor growth measurements were taken 2 or 3 times per week, starting from day 6 after tumor inoculation. Tumor size was calculated as length × width² × 0.52 (mm³). On days 12 to 14, isolated tumor-infiltrating T cells (TIL) were stimulated for 4 to 5 hours with phorbol-12-myristate 13-acetate (PMA) plus ionomycin in the presence of monensin or hgp100_25-33 peptide for cytokine analysis, or directly used for analysis of transcriptional factor expression. Mice were euthanized when tumors reached 2.0 cm in diameter, when tumor ulceration occurred, or when mice became moribund (all of these were recorded as death for the survival curve).

Collected blood samples were analyzed by flow cytometry for CD4, CD8, and ICOS expression using the BD LSRII flow cytometer and FlowJo software. Pre- and post-anti–CTLA-4-treated tumor tissues were fixed by immersion in 10% (vol/vol) neutral buffered formalin solution, which were then subject to immunofluorescence or IHC staining for CD8, CD4, and ICOS. Fixed tissues were embedded in paraffin and transversely sectioned. Sections of 4 μm were subjected to multiplex immunofluorescence staining following the Opal protocol staining method (17). Slides were deparaffinized in xylene and rehydrated in graded ethanols. Antigen retrieval was performed in citrate buffer (citrate pH 6.0) using microwave heating (MWT) 15 minutes per cycle at 10% power. The slides were incubated in antibody diluent for CD8 (1:200, Mouse, C8/144B, cat. # M7103, Dako) with visualization using FITC (1:50, cat. # T20948, Life Technology) and ICOS (1:100, Rabit, cat. # M3980, Spring Bioscience) with visualization using AF-647 (1:50, cat. # T20951, Life Technology). Nuclei were visualized with DAPI (1:2,000, cat. # FP1490, PerkinElmer). All sections were coverslipped with use of Vectashield Hardset 895 mounting media. The slides were scanned using the Vectra slide scanner (PerkinElmer). Nuance spectral analysis software (PerkinElmer) was used for the multispectral and colocaliztion analysis, and each of the individually stained sections were utilized to establish the spectral library of the fluorophores. Five random areas on each sample were analyzed blindly by at 20×.

All experiments were repeated two or three times. Results were expressed as mean ± SEM. Data were analyzed by using a two-sided Student t test or one-way ANOVA after confirming their normal distribution. The log-rank test was used to analyze data from the survival experiments. All analyses were performed with use of Prism 5.0 (GraphPad Software, Inc.), and P < 0.05 was considered statistically significant.

ACT with PMEL-1 Tg⁺ CD8⁺ T cells and vaccination with hgp100_25-33 transiently suppressed tumor growth, but all of the mice eventually developed tumors and succumbed to excessive tumor burden (Supplementary Fig. S1A). Activation of PMEL-1 Tg⁺ CD8⁺ T cells with hgp100_25-33 increased PD-1 and CTLA-4 expression on these cells (Fig. 1A). Dual blockade of PD-1 and CTLA-4 increased infiltration of adoptively transferred PMEL-1 Tg⁺ CD8 T cells into the tumor microenvironment (Fig. 1B), and increased IFNγ secretion by PMEL-1 Tg⁺ CD8 T cells (Fig. 1C) in comparison with the ACT monotherapy or ACT combined with either ICB. Dual blockade of PD-1 and CTLA-4 with ACT induced profound tumor growth delay and improved survival compared with ACT alone or ACT with anti–PD-1 or ACT with anti–CTLA-4, whereas anti–CTLA-4 or anti–PD-1with ACT did not have any survival advantage over ACT alone (Fig. 1D; Supplementary Fig. S1B).

Costimulatory molecules have an essential role in driving the efficacy of both immune-checkpoint therapy and ACT (9–11, 19–21). We previously reported an increase in the frequency of CD4⁺ICOS⁺ T cells, after ipilimumab therapy in patients with bladder cancer (22), prostate tumors (23), and metastatic melanoma (24), with enhanced CD4⁺ T-cell effector functions as a result of ICOS expression (25). Here, we showed an increased expression of ICOS on CD8⁺ T cells following ipilimumab therapy (Supplementary Fig. S2A–S2C).

To determine whether tumor antigen stimulation could activate ICOS signaling in PMEL-1 CD8⁺ T cells, PMEL-1 CD8⁺ T cells were stimulated in vitro with hgp100_25-33 or an unrelated viral peptide LCMV_33-41. ICOS expression was increased on CD8⁺ T cells as early as 24 hours (Fig. 2A) and with a low dose of hgp100_25-33 (0.05 μg/mL; Fig. 2B), indicating that ICOS is an early and sensitive biomarker for CD8⁺ T-cell activation. Next, purified PMEL-1 CD8 T cells (Thy1.1⁺) were adoptively transferred to Rag-1^−/− mice, followed by immunization with hgp100_25-33 or control LCMV_33-41 peptide. Both the frequency and absolute number of ICOS⁺ PMEL-1 CD8⁺ T cells from the DLN were increased and peaked at ∼6 days after hgp100_25-33 immunization but not after immunization with unrelated LCMV_33-41 (Fig. 2C), confirming that this is a tumor antigen–specific effect. Adoptive transfer of ICOS^lo PMEL-1 CD8⁺ T cells into tumor-bearing mice could not suppress melanoma growth as effectively as ICOS^hi counterparts (Supplementary Fig. S2D).

To determine whether ICOS signaling in CD8⁺ T cells orchestrated the therapeutic effects of ACT, we generated PMEL-1ICOS^−/− mice by crossing PMEL-1 Tg⁺ mice with ICOS^−/− mice, which was confirmed by the lack of ICOS upregulation in PMEL-1ICOS^−/− CD8⁺ T cells stimulated with hgp100_25-33 (Fig. 3A).Upon transfer into tumor-bearing mice, PMEL-1ICOS^−/− CD8 T cells exhibited impaired ability to restrain tumor growth (Fig. 3B). ICOS-defective effector CD8⁺ T cells displayed impaired capacity to induce apoptosis of cocultured B16F10 tumor cells (7-AAD⁺Annexin V⁺; Fig. 3C). Next, we wanted to determine whether ICOS signaling in CD8⁺ T cells contributes to the efficacy of ACT plus dual blockade of CTLA-4 and PD-1. We observed that 2 of 10 tumors were eradicated in mice treated with ACT plus dual blockade of anti–CTLA-4 and anti–PD-1, compared with zero of 10 tumors in mice that received the ICOS^−/− ACT plus dual ICB (Supplementary Fig. S3; Fig. 3D). Together, our data provide direct evidence that ICOS signaling in CD8⁺ T cells plays an essential role in orchestrating the antitumor immunity elicited by ACT and long-term protection induced by ACT plus dual blockade of CTLA-4 and PD-1.

Having demonstrated that ICOS is required for antitumor responses in the setting of ACT, we next evaluated how ICOS signaling modulates the effector function of tumor antigen–specific CD8⁺ T cells. First, we stimulated total splenocytes with different concentrations of hgp100 in vitro. ICOS^−/− in PMEL-1 CD8⁺ T cells resulted in reduced production of IFNγ (Supplementary Fig. S4A; Fig. 4A) and perforin (Fig. 4B) but not TNFα (Fig. 4C) and IL2 (Fig. 4D), suggesting a selective role of ICOS in mediating production of effector cytokines. Second, we sorted CD8⁺ T cells (Thy1.1⁺) cells from PMEL-1 and PMEL-1ICOS^−/− mice and injected them into CD45.1 recipient mice that were sublethally irradiated (400 rad), followed by immunization with hgp100. Consistent with in vitro results, a selective reduction of IFNγ and perforin but not TNFα (Fig. 4E) was observed. Third, we utilized Rag-1^−/− mice that lack mature T cells as the recipients. In this system, transferred PMEL-1 or PMEL-1ICOS^−/− CD8⁺ T cells are the only T cells, eliminating any potential interfering effects from endogenous T cells. We noted upregulation of IFNγ but not TNFα (Supplementary Fig. S5A). Collectively, these results confirmed that ICOS signaling in CD8⁺ T cell is selectively required for IFNγ and perforin but not for TNFα and IL2 production upon tumor antigen stimulation.

To directly assess whether ICOS signaling in CD8⁺ T cells drives IFNγ production, we ectopically knocked-in ICOS into PMEL-1ICOS^−/− CD8 T cells by using a retroviral vector encoding ICOS. This rescued the reduced IFNγ production in PMEL-1ICOS^−/− CD8⁺ T cells (Fig. 4F), suggesting that enforced ICOS expression in ICOS-defective CD8⁺ T cells is sufficient to increase IFNγ production. We also examined the selective requirement of ICOS signaling for IFNγ production in CD8⁺ T cells using ICOSL^−/− mice. Because ICOSL is the only known ligand for ICOS (26, 27), we adoptively transferred sorted PMEL-1 or PMEL-1ICOS^−/− CD8⁺ T cells into wild-type (WT) or ICOSL^−/− mice that were sublethally irradiated, followed by immunization with hgp100_25-33 on the next day to analyze cytokine production. Consistent with the above results, we found reduced production of IFNγ and perforin but not of TNFα and IL2 in PMEL-1ICOS^−/− CD8⁺ T cells when transferred into WT recipients but not into ICOSL^−/− recipients (Fig. 4G).

We previously demonstrated that in CD4⁺ T cells, ICOS signaling modulates IFNγ production by primarily regulating T-bet expression (14, 25). T-bet and Eomes are 2 transcription factors known to be important for CD8⁺ T cell–mediated antitumor immune responses (28); therefore, we analyzed T-bet and Eomes expression in PMEL-1ICOS^−/− CD8⁺ T cells by using complementary in vitro and in vivo systems. Reduced expression of Eomes in PMEL-1ICOS^−/− CD8⁺ T cells following stimulation with hgp100_25-33 in vitro (Fig. 4H), transferred into CD45.1 recipients (Fig. 4I) or into Rag-1^−/− recipients (Fig. 4J). Interestingly, no notable changes were seen in T-bet expression in our models (Supplementary Fig. S4B–S4D). Importantly, Eomes expression was largely restored in PMEL-1ICOS^−/− cells with ICOS reexpression (Supplementary Fig. S4E). Human ICOS^hi CD8⁺ T cells have higher expression of IFNγ and Eomes compared with the ICOS^lo CD8⁺ T-cell counterparts (Supplementary Fig. S4F and S4G).

Dual blockade of PD-1 and CTLA-4 increased the infiltration of ICOS⁺Eomes⁺ PMEL-1 Tg⁺ CD8⁺ T cells into the tumor microenvironment (Supplementary Fig. S4H). Next, we assessed the ICOS-mediated downstream signaling pathways that regulate Eomes expression in CD8⁺ T cells. We used ICOS-YF mice that possess a mutation in the cytoplasmic tail of ICOS, abrogating ICOS-mediated PI3K recruitment. ICOS-YF B16 tumor–bearing mice had fewer CD8⁺ICOS⁺Eomes⁺ cells in the tumor-DLNs compared with WT counterparts following anti–CTLA-4 treatment (Supplementary Fig. S4I). These data suggested that ICOS-mediated PI3K signaling is required for Eomes expression during an antitumor response elicited by immune-checkpoint therapy. Collectively, ICOS signaling drives the effector function of CD8⁺ T cells by selectively modulating IFNγ, perforin, and Eomes expression in CD8⁺ T cells.

Two major breakthroughs in the field of immunotherapy have been ICB and ACT. ACT provides antigen specificity but lacks the clinical durability seen with ICB, especially in treating solid tumors (29). The lack of clinical benefit with ACT is primarily due to inadequate migration of adoptively transferred cells into the tumor microenvironment and poor function of the transferred T cells (9). We observed that tumor antigen stimulation increases expression of ICOS, CTLA-4, and PD-1 on adoptively transferred CD8⁺ T cells. PD-1 blockade increases the migration of adoptively transferred T cells (9). The combination of ACT and ICB controls refractory melanoma (10) and induces durable regression of a tumor in a patient with chemorefractory hormone-positive breast cancer (21). Dual CTLA-4 and PD-1 blockade potently revamps the function of tumor-infiltrating lymphocytes (30, 31). Therefore, we investigated whether the efficacy of ACT can be enhanced with concomitant administration of anti–CTLA-4 and anti–PD-1 and demonstrated that the combination indeed improves long-term durable antitumor responses and the efficacy of ACT.

ICOS correlates with anti–CTLA-4–mediated enhanced CD4⁺ T-cell effector functions (24, 25). Developments in ACT indicate that the inclusion of CD28 and ICOS costimulatory signaling can improve survival and clinical response of ACT (1, 32). CD28 is constitutively expressed on T cells and its ligands (B7.1 and B7.2) are more restrictively distributed on antigen-presenting cells. On the other hand, ICOS is induced upon T-cell activation and its ligand, ICOSL, has more widespread expression in nonhematopoietic tissues, including tumors. CD28 costimulation is crucial in the early stage of T-cell activation and ICOS costimulation is more important in the late effector phase (33), suggesting both CD28 and ICOS are probably needed for a complete antitumor immune response. This is corroborated by the fact that CD28 and ICOS double knockout mice have pronounced defects in T-cell proliferation and cytokine production compared with CD28 single knockout (34). Although extensive studies on how ICOS signaling regulates CD4⁺ T-cell functions have been reported (35–39), its role in mediating CD8⁺ T-cell functions is poorly defined, particularly in a tumor setting. The importance of ICOS signaling in CD8⁺ T cells in antitumor response has been suggested in early studies that used ICOSL-overexpressing tumor models (40), IVAX vaccine (15), soluble ICOSL (41), or locoregional application oncolytic virus encoding ICOSL (42). Nevertheless, direct evidence of how ICOS signaling affects the function of adoptively transferred T cells is not understood.

Our results demonstrated that CD8⁺ T-cell–intrinsic ICOS expression is required for antitumor responses and long-term survival of tumor-bearing mice. Impaired tumor suppression by ACT with ICOS^−/−PMEL-1 CD8⁺ T cells correlates with reduced expression of Eomes and decreased IFNγ production by tumor-infiltrating CD8⁺ T cells. Therefore, we propose that the ICOS–Eomes–IFNγ axis in CD8⁺ T cells plays an essential role in driving optimal antitumor response. Unlike CD4⁺ T cells, ICOS deficiency does not affect T-bet expression in PMEL-1 CD8⁺ T cells. Instead, a selective downregulation of Eomes was observed. Using ICOS-YF mice, ICOS-mediated PI3K signaling is required for Eomes expression in CD8 T cells. Interestingly, ICOS signaling in CD8⁺ T cells does not affect production of TNFα and IL2, 2 cytokines closely linked to naïve T cells (43), or early-stage T-cell activation (44), in agreement with a more prominent role of ICOS in the late stage of antitumor response. The specific mechanisms involved in ICOS signaling in CD8 T cells that differentially mediates IFNγ and TNFα production await further exploration.

In our preclinical model, we did not observe any immune-related toxicities and we only gave 2 doses of ICB. Thus, our results suggest that combining ACT with limited use of ICB represents an effective way to mitigate toxicity without sacrificing therapeutic efficacy. No significant toxicities were noted in our murine model; however, we acknowledge the limitation of the PMEL model, which might not mimic the human immune system. A pilot clinical trial with a dose-escalation design will be required to evaluate the safety of this treatment strategy, with plans for additional trials to evaluate the best-tolerated regimen for the efficacy of ACT combined with anti–CTLA-4 plus anti–PD-1 in human. Overall, our study provides a solid foundation of testing combinatorial therapy of ACT consisting of CD8⁺ T cells plus dual blockade of CTLA-4 and PD-1 in clinical studies.

J.P. Allison has ownership interest (including patents) in Amgen, Apricity Health, Marker Therapeutics, Tvardi, BioAtla, Jounce Therapeutics, Neon Therapeutics, Forty Seven, Polaris, Bristol-Myers Squibb, Merck, and Codiak and is a consultant/advisory board member for Amgen, Apricity Health, BioAtla, Jounce Therapeutics, Neon Therapeutics, Forty Seven, Polaris, Codiak, Marker Therapeutics, and Tvardi. J. Gao is a consultant/advisory board member for AstraZeneca, CRISPR Therapeutics, Jounce Therapeutics, Polaris, Pfizer, and Symphogen. P. Sharma has ownership interest (including patents) in Apricity Health, Constellation, Hummingbird, Marker Therapeutics, Imaginab, Jounce Therapeutics, Polaris, Neon Therapeutics, BioAtla, Forty Seven, Dragonfly, Oncolytics, and Codiak and is a consultant/advisory board member for Neon Therapeutics, Jounce Therapeutics, Codiak, Forty Seven, Merck, Bristol-Myers Squibb, Dragonfly, Imaginab, Polaris, Oncolytics, Hummingbird, Marker Therapeutics, Pieris, and BioAtla. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L.Z. Shi, S. Goswami, T. Fu, P. Sharma

Development of methodology: L.Z. Shi, S. Goswami, T. Fu, P. Sharma

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L.Z. Shi, S. Goswami, T. Fu, B. Guan, J. Chen, L. Xiong, X. Zhang, L. Vence, R. Collazo, J. Gao, P. Sharma

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L.Z. Shi, S. Goswami, T. Fu, L. Vence, J. Blando, R. Collazo, J. Gao, P. Sharma

Writing, review, and/or revision of the manuscript: L.Z. Shi, S. Goswami, T. Fu, J. Blando, J.P. Allison, J. Gao, P. Sharma

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L.Z. Shi, B. Guan, J. Zhang, D. Ng Tang

Study supervision: L.Z. Shi, S. Goswami, P. Sharma

The authors thank other members in the Sharma lab for constructive discussions and the Immunotherapy Platform at The University of Texas MD Anderson Cancer Center for assistance with obtaining patient samples. P. Sharma and J.P. Allison are members of the Parker Institute for Cancer Immunotherapy at The University of Texas MD Anderson Cancer Center. This study was supported by a Stand Up To Cancer–Cancer Research Institute Cancer Immunology Dream Team Translational Research Grant (SU2C-AACR-DT1012; to P. Sharma and J.P. Allison). Stand Up To Cancer is a division of the Entertainment Industry Foundation. Research grants are administered by the American Association for Cancer Research, the scientific partner of SU2C.

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.