CD226^hiCD8⁺ T Cells Are a Prerequisite for Anti-TIGIT Immunotherapy

Therapeutic strategies for blocking the interaction between coinhibitory receptors on tumor-infiltrating T cells and their ligands expressed on tumor cells and/or antigen-presenting cells have proven successful in multiple types of cancer (1, 2). To date, six immune checkpoint inhibitors (ICI) targeting cytotoxic T lymphocyte-associated antigen 4 (CTLA-4), programmed death 1 (PD-1), and programmed death-ligand 1 (PD-L1) have been approved by the FDA. However, ICI therapy typically yields an overall response rate of 20% to 40% in patients with solid tumors (3). The identification of additional immune checkpoint receptors on tumor-infiltrating T cells suggests that combining immune checkpoint blockers and activators may improve both the limited response rates and durability of responses in patients with cancer. A number of ICIs including anti–Lag-3 (lymphocyte activation gene 3; ref. 4), anti–TIM-3 (T-cell immunoglobulin and mucin-domain containing-3; ref.5), and anti-TIGIT (T-cell immunoreceptor with immunoglobulin and ITIM domains; ref. 6) are undergoing clinical trials for various types of cancer. The mechanisms by which ICIs deliver therapeutic effects need to be further investigated to inform strategies for ICI combination therapy.

Tumor-infiltrating CD8⁺ T cells become dysfunctional, or exhausted, upon binding with their cognate antigens (7). Exhausted CD8⁺ T cells exhibit impaired proliferation, altered metabolism, decreased cytokine production, and high expression of inhibitory receptors, including PD-1, Lag-3, Tim-3, and TIGIT (8, 9). ICI treatment may confer antitumor effects by reinvigorating tumor-specific exhausted T cells. A subset of exhausted CD8⁺ T cells are responsive to anti–PD-1/PD-L1 therapy (10–12); this cell population has stem cell–like properties, higher expression of costimulatory receptors, and lower expression of coinhibitory receptors (13, 14). Identification of T-cell subsets responsive to ICI therapies would improve clinical efficacy in patients with cancer.

TIGIT is a coinhibitory receptor that binds PVR (poliovirus receptor, CD155) and Nectin-2 (CD112; refs. 15–17). TIGIT is upregulated and coexpressed with PD-1 on exhausted T cells, and antibody-mediated blockade of TIGIT restores the antitumor immunity of CD8⁺ T cells (18, 19). Combined treatment of anti-TIGIT with ICIs such as PD-1, PD-L1, and Tim-3 resulted in enhanced antitumor immunity in mouse tumor models (18, 20). PVR and Nectin-2 also interact with CD226 (DNAM1, DNAX Accessory Molecule-1), which results in enhanced natural killer (NK)–cell and T-cell activity; in contrast, the interaction of TIGIT with the same ligands counteracts CD226-dependent immune cell activation (16, 21). CD226 regulates antitumor immune responses, as demonstrated by accelerated tumor growth in CD226 knockout mice (22, 23). Blocking CD226 abrogates the effect of anti-TIGIT plus anti–PD-L1 combination therapy (18, 24), and tumor-infiltrating CD8⁺ T cells upregulate TIGIT and downregulate CD226, suggesting that the effect of TIGIT blockade is likely related to CD226 expression (20, 25–28).

Here, we identify the subsets of CD8⁺ T cells that are compartmentalized by their level of CD226 expression. CD226^hi and CD226^loCD8⁺ tumor-infiltrating lymphocytes (TIL) represent phenotypically and functionally distinct states that respond differently to TIGIT blockade. The abrogation of TIGIT binding to PVR induces tyrosine phosphorylation of CD226 that potentiates anti-TIGIT therapy. This CD226 expression–dependent effect of anti-TIGIT therapy is confirmed in patients with pancreatic ductal adenocarcinoma (PDAC) treated with modified FOLFIRINOX (5-fluorouracil, leucovorin, irinotecan, and oxaliplatin). Our study provides a mechanism-based rationale for combining TIGIT blockade with other cancer therapies.

Tumor and matched adjacent normal tissue samples were collected from patients with cancer undergoing surgery after obtaining written consent from each patient. Blood samples were obtained from patients enrolled in a phase II clinical trial of mFOLFIRINOX, who had locally advanced PDAC before and after receiving mFOLFIRINOX (29). Peripheral blood mononuclear cells (PBMC) were isolated from blood samples from patients and healthy donors by Percoll density gradient centrifugation.

Blood samples from healthy donors and patients with PDAC were obtained according to the protocols approved by the Institutional Review Board at Asan Medical Center (#2016-0010 and #2018-1525). Tumor and TIL samples were collected according to a separately approved protocol (#2019-0155). The study was conducted under the guidelines of the Helsinki declaration.

By using Gibson assembly, full-length human CD226 (aa 1–336), TIGIT (aa 1–244), and PVR (aa 1–372) genes were synthesized and cloned into the pSBbi-RB vector (Addgene plasmid #60522, a gift from Eric Kowarz, Goethe-University, Frankfurt/Main, Germany). CD226 was N-terminally tagged with FLAG. The point mutants of CD226 (Y322A, S329A, and Y322A/S329A) and TIGIT (Y225A/Y231A) were generated by PCR-based site-directed mutagenesis and verified by DNA sequencing (21, 30). The gene encoding anti-human CD3 single-chain fragment variable (scFv) was synthesized and cloned into the pLV-EF1α-IRES-Puro vector (Addgene plasmid #85132; Addgene; ref. 31). The firefly luciferase (FLuc) cDNA was PCR-amplified from the pGL2-Luc vector (Promega) and subcloned into the pSBbi-RB vector. The Renilla luciferase cDNA was PCR-amplified from the pRL vector (Promega) and subcloned into the pSBbi-BP vector (Addgene plasmid #60512).

Jurkat, Raji, B16F10, 4T1, CT26, EL4, and E.G7-OVA cells were obtained from the American Type Culture Collection (TIB-152; CCL-86; CRL-6475; CRL-2539; CRL-2638; TIB-39; CRL-2113). MC38 was purchased from Kerafast, Inc. (ENH204-FP). The Platinum-E retroviral packaging cell line (Plat-E) was a kind gift of Dr. Yun-cai Liu (La Jolla Institute for Immunology, La Jolla, CA; ref. 32). CHO-S cells were purchased from Invitrogen and maintained in serum-free CD CHO medium (Thermo Fisher). All cells were maintained as recommended by the suppliers and kept in culture for less than 3 consecutive months for any given experiment. All cells were not reauthenticated in the past year. The cells for in vivo mouse experiments were tested and found negative for Mycoplasma contamination before transplantation.

Jurkat cells were transfected using the Neon electroporation system (Invitrogen) according to the manufacturer's protocol. The pSBbi-RB–expressing CD226 (WT, Y322A, S329A, or Y322A/S329A) or TIGIT (WT or Y225A/Y231A) were transfected into Jurkat cells together with pCMV (CAT)T7-SB100 (Addgene plasmid #34879). After transfection, positive cells were selected in the presence of blasticidin for 2 weeks and FACS sorted based on their coexpression of RFP and the transfected receptor. For the generation of T-cell stimulator cells, CHO (Chinese hamster ovary) cells were infected with a lentivirus encoding OKT3 scFv. After 2 weeks of Puromycin selection, OKT3 scFv expression was verified by FACS using an anti-mouse IgG. To generate CHO-OKT3 cells expressing PVR, CHO-OKT3 cells were transfected with a plasmid expressing PVR and selected using blasticidin. To measure antigen-dependent T-cell activation, Raji B cells stably expressing PVR were generated in the same manner as described above. The PVR expression in CHO-OKT3-PVR and Raji B-PVR cells was confirmed with anti-human PVR FACS staining.

Human CD8⁺ T cells were purified from PBMCs using a magnetic isolation kit (STEMCELL Technologies). For activation, CD8⁺ T cells were stimulated with anti–CD3/anti–CD28-coated Dynabeads (Invitrogen) at a cell-to-bead ratio of 2:1. For experiments where T-cell proliferation was measured, purified CD8⁺ T cells were labeled with CellTrace Violet (CTV; Invitrogen) or carboxyfluorescein diacetate succinimidyl ester (CFSE; Invitrogen) prior to cell culture. The proliferation of CTV-labeled CD8⁺ T cells was evaluated by quantification of CTV dilution. To measure antigen-specific memory T-cell response, CD8⁺ T cells were cocultured with CEF [cytomegalovirus (CMV), Epstein–Barr virus (EBV), and Flu] peptide pool (JPT Peptide Technologies)–loaded autologous dendritic cells (DC) at a ratio of 4:1 (T:DC). To generate monocyte-derived DCs, CD14⁺ monocytes were isolated from PBMCs using the MojoSort Human Pan Monocyte Isolation Kit (BioLegend) and cultured in complete RPMI1640 medium containing GM-CSF (2 ng/mL) and IL4 (10 ng/mL). DC maturation was carried out by culture with LPS (1 μg/mL) for 16 hours. IFNγ secretion in culture medium was analyzed by ELISA (BioLegend). To assess the effects of CD226 mutants on primary T-cell function, human primary CD8⁺ T cells were electroporated with plasmids expressing CD226 WT or mutants and then FACS sorted based on CD226 and FLAG staining. The FACS-sorted CD8⁺ T cells were stimulated with CHO-OKT3 or CHO-OKT3-PVR cells for the indicated times at a ratio of 10:1 (T:CHO). For antigen stimulation, Jurkat cells were cocultured with staphylococcal enterotoxin E (SEE, Toxin Technologies)–pulsed Raji B cells for 24 hours at a ratio of 10:1 (Jurkat:Raji B).

Female C57BL/6 and BALB/c mice at 6 to 8 weeks of age were obtained from Joongah Bio (Suwon, Korea). OT-I T-cell receptor (TCR)–transgenic mice [C57BL/6-Tg(TcraTcrb)1100Mjb/J] were purchased from The Jackson Laboratory. All mice were kept in pathogen-free conditions in the animal facility of Asan Medical Center. All animal experiments were approved and supervised by the Asan Medical Center Institutional Animal Care and Use Committee.

For tumor challenge, mouse tumor cells were cultured and passaged 2 times prior to inoculation. B16F10 or MC38 cells were resuspended in Hank's Balanced Salt Solution (HBSS) without phenol red (Welgene) and s.c. injected into the left flank of C57BL/6 mice. 4T1 and CT26 cells were injected into BALB/c mice as described above. Tumor volumes were calculated as (length × width²)/2. Tumors were analyzed by flow cytometry at 14 days after implantation.

Tumor tissues were dissected and digested with collagenase IV (250 unit/mL, Worthington Biochemical Cooperation) and DNase I (100 μg/mL, Roche) in HBSS using a gentleMACS dissociator (Miltenyi Biotec). TILs were enriched using Lymphopure (BioLegend), and single cells were recovered for analysis by flow cytometry and in vitro assay.

Mouse and human single cells were blocked with anti-mouse CD16/32 (Cat #101302, BioLegend) and human TruStain FcX (Cat #422302, BioLegend), respectively, and then stained with indicated antibodies in FACS buffer (PBS containing 2% FBS and 0.1% sodium azide) for 20 minutes at 4°C. Dead cells were excluded using the Zombie NIR Fixable viability dye (BioLegend). For intracellular staining, cells were stained with antibodies to surface markers and then fixed and permeabilized with Cytofix/Cytoperm buffer (BD Biosciences) or Foxp3/Transcription Factor Staining Buffer Set (Invitrogen), followed by staining with indicated antibodies diluted in Perm/Wash buffer (BD Biosciences). The following fluorescent dye–conjugated anti-human antibodies were used: anti-CD8 (SK1), anti-CCR7 (G043H7), anti-CD45RO (UCHL1), anti-CD226 (11A8), anti-TIGIT (MBSA43), anti–PD-1 (EH12.2H7), anti-Lag3 (11C3C65), anti-Tim3 (F38-2E2), anti–CTLA-4 (L3D10), anti–4-1BB (4B4-1), anti-IFNγ (4S.B3), and anti-TNFα (MAb11). Antibodies against mouse proteins were as follows: anti-CD45 (30-F11), anti-CD3 (17A2), anti-CD8 (53–6.7), anti-CD226 (10E5), anti-TIGIT (1G9), anti–PD-1 (29F.1A12), anti-Tim3 (B8.2C12), anti–Lag-3 (C9B7W), anti-CD38 (90), anti-CD127 (A7R34), anti-Slamf6 (330-AJ), anti-TNFα (MP6-XT22), anti-IFNγ (XMG1.2), anti–T-bet (4B10), anti–Ki-67 (11F6), and anti-FLAG (L5; all purchased from BioLegend). Anti-Eomes (Dan11mag) was purchased from Invitrogen. Anti-CD101 (Igsf2) was purchased from BD Biosciences. Data acquisition was performed on a CytoFLEX flow cytometer (Beckman Coulter) and analyzed with the FlowJo software (v.10.5.3, TreeStar). The automated analysis of multiparameter flow cytometry data was subjected to the t-Stochastic Neighbor Embedding (t-SNE) algorithm or the FlowSOM algorithm (33, 34).

Monoclonal antibodies against human PD-1 (MK-3475, Selleckchem) and TIGIT (MBSA43, eBioscience) or isotype controls were used for human ex vivo assays. Anti-mouse PD-1 (J43) and anti-mouse TIGIT (1G9) antibodies were purchased from BioXCell. Anti-human CD226 agonistic antibody (NewE1) was purchased from MilliporeSigma. Anti-human CD226 antagonistic antibody (DX11) was purchased from Abcam. Anti-phospho-CD226 (Tyr322) was generated by injecting a KLH-conjugated Tyr322-phosphorylated peptide into rabbits and affinity-purified on a phosphopeptide column.

To construct a plasmid encoding mouse CD226–specific shRNA, oligonucleotides were cloned in the modified LMP vector, which expresses the fluorescent protein mAmetrine1.1 (35). Oligonucleotide sequences are as follows: mouse CD226 shRNA, 5′-TGCTGTTGACAGTGAGCGACACTAGGATCTACATTGATAATAGTGAAGCCACAGATGTATTATCAATGTAGATCCTAGTGGTGCCTACTGCCTCGGA-3′. Plat-E packaging cells were used to generate shRNA-expressing retrovirus. Splenocytes isolated from OT-I mice (OVA-specific TCR Tg mice) were stimulated with SIINFEKL peptide (2 μg/mL) in the presence of IL2 (2 ng/mL). After 1-day stimulation, the splenocytes were infected with shRNA-expressing retrovirus together with 8 μg/mL polybrene (MilliporeSigma) by spin inoculation. Two days after infection, CD8⁺ T cells were sorted based on the expression of mAmetrine1.1 and CD226. E.G7-OVA cells stably expressing FLuc and EL4 cells stably expressing Renilla luciferase (RLuc) were generated and used for an in vitro OT-I T-cell–mediated killing assay. E.G7-OVA-FLuc and EL4-RLuc cells (1:1) were coincubated with FACS-sorted OT-I T cells in 96-well plates at the designated effector:target (E:T) ratios. The plates were incubated for 48 hours at 37°C and 5% CO₂. The relative luciferase activity was measured by the Dual-Glo luciferase assay system (Promega) according to the manufacturer's instructions.

Jurkat cells stably expressing CD226 WT or mutants were stimulated with CHO-OKT3-PVR at a ratio of 5:1 (Jurkat:CHO). The cells were lysed in SDS sample buffer containing 60 mmol/L Tris HCl, 2% SDS, 10% glycerol, and 100 mmol/L dithiothreitol. Cell lysates were resolved by SDS-PAGE and immunoblotted with the indicated antibodies (32). For immunoblotting, primary antibodies against phospho-ERK1/2 (#4370), phospho-p38 (#4511), phospho-AKT (#4060), and β-actin (#8457) were purchased from Cell Signaling Technology. Anti-FLAG M2 was from MilliporeSigma.

Gene expression data of patients with colon adenocarcinoma (normalized RNAseq FPKM-UQ) were retrieved from The Cancer Genome Atlas (TCGA) database using UCSC Xena (https://xena.ucsc.edu/). Microsatellite instability (MSI) status for TCGA colon adenocarcinoma cohort was obtained at https://tcia.at/ (36). The available data for 449 patients were analyzed without exclusion. Data visualization and statistical tests were performed using the R software (R Foundation for Statistical Computing; https://www.r-project.org/).

Statistical analysis was performed using GraphPad Prism 8.0. Statistical analyses were performed using the appropriate statistical comparison, including the two-tailed paired or unpaired Student t test, or one-way ANOVA with Holm–Sidak multiple comparisons. P values < 0.05 were considered statistically significant (*, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001).

We first examined the role of the costimulatory receptor CD226 in TIGIT blockade–mediated T-cell activation. A pool of CMV, EBV, and Flu CD8⁺ epitopes (CEF) was used to measure antigen-specific responses in PBMCs. Coblockade of TIGIT and CD226 abrogated the increased CEF antigen–specific recall response of CD8⁺ memory T cells compared with TIGIT blockade alone (Supplementary Fig. S1). We then assessed CD226 expression in mouse 4T1 mammary carcinoma, B16F10 melanoma, CT26 colon carcinoma, and MC38 colon adenocarcinoma to determine if CD8⁺ TILs exhibit altered CD226 expression compared with splenic CD8⁺ T cells. Naïve T cells show a different expression profile of CD226 compared with memory T cells (Supplementary Fig. S2). Although the majority of splenic CD8⁺ memory T cells showed high expression of CD226 (CD226^hi), the proportion of the CD226^lo population within the CD8⁺ TILs was higher in mice bearing 4T1, B16F10, and CT26 tumors. The CD226^lo population in CD8⁺ TILs isolated from MC38 tumors, which are known to be highly immunogenic, was not increased (Fig. 1A). Expression of CD226 was lower in CD8⁺ TILs than in the spleens, especially in B16F10 tumors (Fig. 1B). In patients with renal cell carcinoma (RCC) or colorectal cancer, CD226^loCD8⁺ memory T cells accumulated at the tumor site more than in matched normal tissues. Although patients with non–small cell lung carcinoma (NSCLC) showed no differences of CD226^loCD8⁺ memory T-cell frequencies between tumor tissues and normal tissues (Fig. 1C), expression of CD226 was lower in CD8⁺ TILs than in normal tissues from patients with RCC, colorectal cancer, and NSCLC (Fig. 1D). TCGA database analysis showed that CD226 mRNA expression was higher in colorectal cancers with high MSI than in colorectal cancers with stable or low-instability microsatellites, suggesting a correlation between CD226 expression and local cytotoxic immune response (ref. 37; Supplementary Fig. S3). Thus, accumulation of CD226^hiCD8⁺ T cells may be associated with tumor immunogenicity.

We assessed CD226 expression in memory subsets of peripheral CD8⁺ T cells to determine whether CD226 downregulation is associated with differentiation of CD8⁺ T cells. In healthy donors, CD226 expression was inversely proportional to memory differentiation (Supplementary Fig. S4A–S4C). Because CD226 expression stratifies memory subsets into CD226^hi and CD226^lo populations under steady-state conditions, we profiled the phenotypic properties of peripheral CD8⁺ memory T cells according to CD226 expression. Compared with CD226^hiCD8⁺ effector memory T (T_em) cells, CD226^loCD8⁺ T_em cells expressed more TIGIT and PD-1 and less of the early memory marker CD127 (Supplementary Fig. S4D), indicating that CD226^loCD8⁺ T_em cells are a more differentiated subset than CD226^hiCD8⁺ T_em cells.

The inverse correlation between CD226 and TIGIT expression was previously reported in HIV-specific T cells that are dysfunctional and likely exhausted (38, 39). We analyzed CD8⁺ TILs from patients with RCC, colorectal cancer, and NSCLC by flow cytometry. The CD226^lo population expressed more of the coinhibitory receptors (TIGIT, PD-1, Lag-3, and Tim-3) than did CD226^hiCD8⁺ TILs, similar to the phenotype of HIV-specific CD8⁺ T cells (Fig. 2A). To determine the identity of CD226^loCD8⁺ T cells, we analyzed CD8⁺ TILs in mice bearing 4T1 tumors with markers that correlate with different states of T-cell exhaustion: TIGIT, PD-1, Tim-3, Lag-3, CD101, CD38, CD127, Slamf6, T-bet, and Eomes. CD226^hi and CD226^lo populations of CD8⁺ TILs differed in marker expression when visualized using t-SNE maps (Fig. 2B; Supplementary Fig. S5). CD226^loCD8⁺ TILs showed an exhausted phenotype with significant upregulation of TIGIT, PD-1, Tim-3, Lag-3, CD101, CD38, and Eomes, and reduced expression of CD127, Slamf6, and T-bet (Fig. 2C and D). Tim-3 and Slamf6/Tcf7 identify progenitor and terminally exhausted T cells, respectively (12). Although the majority of CD226^loCD44⁺PD-1⁺CD8⁺ TILs were Tim-3⁺Slamf6⁻ (terminally exhausted) CD8⁺ TILs, some Tim-3⁻Slamf6⁺ (progenitor exhausted) CD8⁺ TILs were among the CD226^loCD44⁺PD-1⁺CD8⁺ TILs from mice bearing 4T1 and MC38 tumors (Fig. 2E). Such heterogeneity among CD226^loCD8⁺ TILs implies that CD226 is a nonredundant marker of T-cell exhaustion.

Given their phenotypic differences, CD226^hi and CD226^loCD8⁺ T cells may differ in function. Compared with CD226^loCD8⁺ TILs, CD226^hi CD8⁺ TILs in mice bearing 4T1 or MC38 tumors expressed more Ki-67 and effector cytokines IFNγ and TNFα (Fig. 3A and B; Supplementary Fig. S6). CD226^hiCD8⁺ TILs also displayed increased polyfunctionality as measured by coexpression of IFNγ and TNFα (Fig. 3C). We examined expression of Granzyme B (GrzB) and cytokines in CD8⁺ T cells under steady-state conditions to determine whether CD226 downregulation is responsible for the reduced functionality of CD226^loCD8⁺ T cells. The proportions of GrzB-, IFNγ-, or TNFα-expressing populations were higher in CD226^hi than in CD226^loCD8⁺ T cells from healthy donor–derived PBMCs upon anti-CD3/CD28 stimulation (Fig. 3D). To assess the responsiveness of CD226^hi and CD226^loCD8⁺ T cells in an antigen-specific manner, we measured the recall response of CD8⁺ memory T cells in response to CEF antigen. CD226^hi and CD226^loCD8⁺ T_em cells were FACS sorted from healthy donor PBMCs and cocultured with CEF antigen–loaded autologous DCs (Fig. 3E). Although CD226^hiCD8⁺ T_em cells were reactivated by CEF antigen stimulation, CD226^loCD8⁺ T_em cells showed less proliferation and IFNγ secretion (Fig. 3F and G). Thus, CD226 expression defines CD8⁺ T-cell subsets that are phenotypically and functionally distinct.

The suppressive activity of TIGIT on CD8⁺ T-cell responses is dependent on CD226 (18, 40). However, the interplay between TIGIT and CD226 is still unknown in the context of TIGIT blockade. We hypothesized that CD226^hiCD8⁺ T cells would respond better to TIGIT blockade. CD226^hi and CD226^lo CD8⁺ T_em cells were FACS sorted from healthy donor PBMCs and stimulated with the CEF antigen in the presence or absence of anti-TIGIT. TIGIT blockade enhanced both proliferation and IFNγ production of CD226^hiCD8⁺ T_em cells upon CEF antigen stimulation, but no effect was observed in CD226^loCD8⁺ T_em cells. PD-1 blockade also showed a skewed effect in CD226^hiCD8⁺ T_em cells, which is consistent with the known PD-1–mediated regulation of CD226 activation (41). A synergistic effect of TIGIT and PD-1 coblockade was only observed in CD226^hiCD8⁺ T_em cells (Fig. 4A and B).

To determine whether CD226 downregulation affects TIGIT blockade activity, we utilized shRNA-mediated knockdown of CD226 in OT-I T cells. CD226^hi and CD226^lo/KD OT-I T_em cells were FACS sorted (Fig. 4C; Supplementary Fig. S7A) and cocultured with E.G7-OVA cells expressing firefly luciferase (E.G7-Fluc) and EL4 cells expressing Renilla luciferase (EL4-Rluc) that expressed similar amounts of PVR (Supplementary Fig. S7B). The OVA-specific cytotoxicity of OT-I T cells was measured by dual-luciferase assays, where FLuc signal changes are normalized by Renilla luciferase signal. CD226^hi OT-I T cells showed a higher cytotoxic activity than did CD226^lo/KD OT-I T cells, and the difference was greater at a lower E:T ratio (1:20; Fig. 4D). When anti-TIGIT or anti–PD-1 was added, TIGIT blockade enhanced the cytotoxic activity of CD226^hi OT-I T cells against E.G7-Fluc but did not affect CD226^lo/KD OT-I T cells at E:T ratio of 1:10 (Fig. 4E). Likewise, PD-1 blockade only showed significant effects in CD226^hi OT-I T cells. These data suggest that CD226 expression is a prerequisite for the effects of TIGIT blockade.

CD226 promotes NK-cell activation through ITT (an immunoreceptor tyrosine tail)-like motif-mediated signal transduction (42). The molecular mechanism by which intracellular activation of CD226 modulates TIGIT, PVR, and TCR signaling modulation remains unclear. To determine which phosphorylation site of CD226 propagates TCR signaling, we generated Jurkat cells expressing CD226 WT or mutants that carry tyrosine 322 to alanine (Y322A) mutation, serine 329 to alanine (S329A) mutation, or both (Supplementary Fig. S8A). Upon stimulation with CHO-OKT3-PVR cells (CHO cells expressing both OKT3 scFv and human PVR; Supplementary Fig. S8B and S8C), phosphorylation of downstream TCR signaling molecules (ERK1/2, p38, and AKT) was reduced in both Jurkat^Y322A and Jurkat^Y322A/S329A cells but not in Jurkat^S329A cells, which is consistent with studies that identified the role of serine and tyrosine phosphorylation of CD226 in NK-cell signaling and activation (ref. 43; Fig. 5A).

We exogenously expressed CD226 WT and mutants in human primary CD8⁺ T cells to determine if the impaired tyrosine phosphorylation of CD226 affects TIGIT blockade activity. CD226 WT, Y322A, S329A, or Y322A/S329-expressing CD8⁺ T cells were FACS sorted (Supplementary Fig. S9) and stimulated with CHO-OKT3 or CHO-OKT3-PVR cells. Mutation at tyrosine 322, not serine 329, of CD226 affected T-cell proliferation and IFNγ secretion in response to PVR costimulation by CHO-OKT3-PVR cells (Fig. 5B and C). The impaired activation of CD226^Y322A CD8⁺ T cells was not restored by the addition of anti-TIGIT, whereas CD226^WT CD8⁺ T cells showed increased effector responses (Fig. 5D). These results show that tyrosine phosphorylation of CD226 is required for the PVR costimulation-mediated TIGIT blockade activity.

Based on these data, we reasoned that TIGIT blockade would induce tyrosine phosphorylation of CD226. As there are no commercially available antibodies detecting the tyrosine-phosphorylated human CD226, we generated an antibody against Tyr322-phosphorylated CD226 (pCD226^Y322). PVR-induced CD226 phosphorylation at tyrosine 322 was detected in both Jurkat^WT and Jurkat^S329A cells by anti-pCD226^Y322 (Fig. 5E). TIGIT is phosphorylated at Y225 after its ligation with PVR and transmits inhibitory signals in NK cells (44). Because Jurkat cells do not express TIGIT, we generated Jurkat cells that stably express TIGIT WT or phosphorylation mutants (Y225A/Y231A; Supplementary Fig. S10A). Expression of TIGIT WT in Jurkat cells abrogated the PVR-induced CD226 phosphorylation at Tyr 322, whereas TIGIT Y225A/Y231A-mutant expression had no effect on CD226 phosphorylation. Treatment with anti-TIGIT restored the impaired tyrosine phosphorylation of CD226 in Jurkat^{TIGIT WT} cells to a similar extent to that in Jurkat^{TIGIT Y255A/231A} cells (Fig. 5F). In accordance with the deleterious effect of TIGIT phosphorylation on CD266 activation/phosphorylation, TIGIT WT expression resulted in impaired T-cell activation upon stimulation with SEE peptide–loaded Raji cells expressing PVR (Supplementary Fig. S10B). TIGIT blockade repaired this defect in Jurkat^{TIGIT WT} cells, but had minimal effect on Jurkat^{TIGIT Y225A/Y231A} cells with similar CD69 expression and TCR signaling activation as WT Jurkat cells (Fig. 5G; Supplementary Fig. S11). To confirm the tyrosine phosphorylation–mediated CD226 activation of TIGIT blockade activity, we utilized an anti-CD226 agonist that activates CD226 signaling without triggering TIGIT signaling. The engagement of CD226 by this agonist antibody promoted tyrosine phosphorylation of CD226 and TCR signaling (Fig. 5H; Supplementary Fig. S12A), which led to a synergistic effect with anti-TIGIT in CD8⁺ memory T-cell reactivation by CEF antigen (Supplementary Fig. S12B). CD226 agonism rendered CD226^loCD8⁺ T_em cells responsive to CEF antigen and TIGIT blockade (Fig. 5I). Thus, TIGIT blockade activity depends on CD226 tyrosine phosphorylation.

We then used clinical samples to determine whether CD226 upregulation in CD8⁺ memory T cells predicts the response to TIGIT blockade. Given the role of CD226 in TIGIT blockade, we hypothesized that therapies that increase the proportion of CD226^hiCD8⁺ T cells enhance response to TIGIT blockade. Therefore, we performed immune monitoring of peripheral blood CD8⁺ T cells in patients with PDAC before and after mFOLFIRINOX therapy, a standard chemotherapy agent for metastatic PDAC (45). Flow cytometry using immune checkpoint markers (CD45RO, CCR7, CD226, TIGIT, PD-1, Lag-3, Tim-3, CTLA-4, and 4-1BB) and t-SNE analysis revealed that the CD8⁺ T cells acquired a more activated phenotype after mFOLFIRINOX therapy (Fig. 6A). To gain a comprehensive understanding of the effect of mFOLFIRINOX therapy on the phenotype of peripheral blood CD8⁺ T cells, we employed an unsupervised Self-Organizing Map (FlowSOM) algorithm to generate clusters of CD8⁺ T cells based on marker expression. The minimum spanning tree of eight CD8⁺ T-cell clusters was generated, and each cluster was assigned to metaclusters (Fig. 6B). By comparing the frequency of the same cluster between before and after mFOLFIRINOX therapy, we observed decreases in the frequencies of clusters 1 and 5 after mFOLFIRINOX therapy, which likely represented terminally differentiated CD8⁺ memory T cells with CD226^loTIGIT^hiPD-1^hi4-1BB^lo/med expressions. In contrast, mFOLFIRINOX treatment resulted in increased frequencies of clusters 2 and 3 with a profile of CD226^hiTIGIT^medPD-1^med4-1BB^med/hi expressions (Fig. 6C). The ratio between naïve and memory CD8⁺ T cells was unaltered. CD226 upregulation in CD8⁺ memory T cells was confirmed (Fig. 6D and E). To determine whether the altered phenotype/CD226 upregulation after mFOLFIRINOX therapy leads to enhanced responses of CD8⁺ memory T cells, we assessed their responsiveness to CEF peptide pool stimulation. Compared with those prior to treatment, CD8⁺ memory T cells after mFOLFIRINOX treatment proliferated more and secreted more IFNγ (Fig. 6F and G). We next evaluated if mFOLFIRINOX-induced CD226 upregulation potentiated the effect of TIGIT or PD-1 blockade. The proliferation rate of CD8⁺ memory T cells obtained from patients after, but not before, mFOLFIRINOX therapy was enhanced by the addition of anti-TIGIT or anti–PD-1 (Fig. 6H). Although the change of IFNγ secretion in response to anti-TIGIT or anti–PD-1 was 1.2 and 1.9, compared with control hIgG1 antibody, respectively, in CD8⁺ memory T cells obtained from patients before mFOLFIRINOX therapy, TIGIT or PD-1 blockade increased the fold change of IFNγ secretion to 1.7 and 2.5, respectively, in CD8⁺ memory T cells after mFOLFIRINOX treatment (Fig. 6I). Thus, mFOLFIRINOX therapy renders CD8⁺ T cells more responsive to TIGIT or PD-1 blockade. The CD226^hiCD8⁺ T-cell frequency has functional and potential predictive value for defining effective anti-TIGIT therapy alone and in combination with other cancer therapies.

Therapy with anti-TIGIT, an ICI, can restore T-cell immune activity. Response is improved when such therapy is combined with PD-1/PD-L1 blockade. Clinical success depends on stratification of patients according to their likelihood of responding to anti-TIGIT therapy.

We found that CD226^loCD8⁺ memory T cells accumulate in both mouse and human tumors compared with spleen or normal tissues, indicating that CD226 downregulation may be due to tumor antigen exposure and related to T-cell exhaustion. Indeed, CD226^loCD8⁺ TILs exhibited a more differentiated and exhausted phenotype, whereas CD226^hi CD8⁺ TILs retained expression of an early memory marker. PD-1⁺-exhausted CD8⁺ TILs can be further subdivided into progenitor and terminally exhausted cells based on their Tim-3 and Slamf6 surface expression (12). CD226^loCD8⁺ TILs did not fit neatly into either progenitor or terminally exhausted subsets, as they contained both Tim-3⁺ Slamf6^– terminally exhausted CD8⁺ TILs and Tim-3⁻Slamf6⁺ progenitor–exhausted CD8⁺ TILs. CD226^loCD8⁺ TILs were dysfunctional, unlike the terminally exhausted CD8⁺ TILs that possess superior cytotoxicity. These discrepancies demonstrate that low CD226 expression indicates CD8⁺ T-cell dysfunction. The CD226^loCD8⁺ memory T-cell subset resembles aged dysfunctional CD8⁺ effector memory RA T (T_emra) cells that express high amounts of TIGIT and little CD226 (46). As exhausted and aged T cells share characteristics (47), subpopulations within CD226^loCD8⁺ memory T cells need better definition. Nevertheless, the dysfunctional phenotype provides an explanation for the lack of accumulation of CD226^lo CD8⁺ memory T cells in immunogenic tumors, including mouse MC38 tumors and clinical samples of NSCLC.

Given the imbalance between CD226 and TIGIT expression in dysfunctional CD8⁺ T cells of patients with HIV or cancer (19, 38, 39), downregulation of CD226 may be involved in TIGIT-mediated inhibition of CD8⁺ T-cell responses. Indeed, our study reveals that TIGIT blockade affects CD226^hiCD8⁺ T cells but not CD226^loCD8⁺ T cells, showing that TIGIT blockade is CD226-dependent. The greater functionality of CD226^hiCD8⁺ T cells may not be solely due to CD226 expression, so we reasoned that the CD226^hiCD8⁺ T-cell subset would also predict the efficacy of other immune checkpoint blockades. Although PD-1 blockade also enhanced the responsiveness of the CD226^hiCD8⁺ T cells, anti-TIGIT activity showed a greater dependency on CD226. PD-1 inhibition promotes CD8⁺ T-cell effector responsiveness in a CD226-dependent manner by preventing PD-1-SHP2–mediated CD226 dephosphorylation (41). PD-1 and CD226 may be colocalized and coregulatory in the central supramolecular activation cluster of the immunologic synapse following T-cell antigen recognition (30, 48, 49), which could explain the PD-1 blockade effect in CD226^hiCD8⁺ T cells. However, how the PD-1–SHP2 axis integrates into the PVR-CD226 signaling pathway remains unclear. We showed evidence of intracellular TIGIT-mediated regulation of CD226 activation: impaired phosphorylation in tyrosine, not serine, of CD226 rendered CD8⁺ T cells less responsive to both PVR binding and TIGIT blockade. We also detected increased PVR-induced CD226 phosphorylation upon TIGIT blockade and showed that PVR-induced TIGIT phosphorylation inhibits T-cell responses by promoting CD226 dephosphorylation. Thus, we showed that TIGIT, previously known only as a decoy receptor for CD226, affects intracellular regulation of CD226 activation upon PVR binding.

Insight into the molecular mechanism of anti-TIGIT therapy may help optimize clinical strategies for both monotherapy and combination therapies. Because of the CD226-dependent activity of TIGIT blockade, we suggest that CD226 agonism may provide a synergistic benefit to anti-TIGIT immunotherapies. Indeed, cotreatment with TIGIT blockade and agonistic anti-CD226 enhanced antigen-specific CD8⁺ T-cell responses. CD226 agonism also showed a lesser synergistic effect with PD-1 blockade. The functional defect of CD226^loCD8⁺ T cells was partially reversed by CD226 signaling activation. Thus, CD226 may be a therapeutic target for improving immune checkpoint therapies.

Conventional chemotherapeutics and targeted anticancer agents can succeed by eliciting immune responses (50). Preclinical studies have shown that various chemotherapeutic modalities confer stronger therapeutic effects in immunocompetent hosts than in immunodeficient hosts (51). Our results showed that CD226^hiCD8⁺ memory T cells were increased in patients who received mFOLFIRINOX therapy, which was associated with a heightened response to anti-TIGIT and anti–PD-1 blockades. These results suggest that measuring CD226^hiCD8⁺ T cells in the peripheral blood of patients with cancer helps predict response to anti-TIGIT and anti–PD-1 therapies. Indeed, mFOLFIRINOX treatment may synergize with anti-TIGIT and anti–PD-1 therapies in patients with PDAC.

In summary, our study clarifies the CD226-dependent mechanism of anti-TIGIT therapy. Such mechanism-based understanding may guide design of cancer immunotherapies and offer criteria for patient stratification.

No potential conflicts of interest were disclosed.

Conception and design: H.-s. Jin, C. Yoo, Y. Park

Development of methodology: H.-s. Jin, M. Ko, J.H. Kim, C. Yoo, Y. Park

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.-s. Jin, D.-s. Choi, H.J. Lee, K.-p. Kim, C. Yoo, Y. Park

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.-s. Jin, M. Ko, J.H. Kim, D.-h. Lee, E.K. Choi, C. Yoo, Y. Park

Writing, review, and/or revision of the manuscript: H.-s. Jin, D.-h. Lee, H.J. Lee, C. Yoo, Y. Park

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.-H. Kang, I. Kim, C. Yoo

Study supervision: H.-s. Jin, Y. Park

This study was supported by the National Research Foundation of Korea (NRF-2016M3A9E8941331 and NRF-2018R1D1A1B07050715), Korean Health Technology R&D Project, Ministry for Health and Welfare, Republic of Korea (HI15C0972), Asan Institute for Life Sciences, Asan Medical Center (2019-763), and KIST institutional program. Biospecimens were provided by Bio-Resource Center of Asan Medical Center. The authors thank the core facilities of Genetically Engineered Animal Core and flow cytometry at the ConveRgence mEDIcine research cenTer (CREDIT), Asan Medical Center. The authors also thank Dr. Joon Seo Lim from the Scientific Publications Team at Asan Medical Center for his editorial assistance in preparing this article.

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.