NK Receptor Signaling Lowers TCR Activation Threshold, Enhancing Selective Recognition of Cancer Cells by TAA-Specific CTLs

Abstract

CTL recognition of non-mutated tumor-associated antigens (TAA), present on cancer cells but also in healthy tissues, is an important element of cancer immunity, but the mechanism of its selectivity for cancer cells and opportunities for its enhancement remain elusive. In this study, we found that CTL expression of the NK receptors (NKR) DNAM-1 and NKG2D was associated with the effector status of CD8⁺ tumor-infiltrating lymphocytes (TIL) and long-term survival of melanoma patients. Using MART-1 and NY-ESO-1 as model TAAs, we demonstrated that DNAM-1 and NKG2D regulate T-cell receptor (TCR) functional avidity and set the threshold for TCR activation of human TAA-specific CTLs. Superior costimulatory effects of DNAM-1 over CD28 involved enhanced TCR signaling, CTL killer function and polyfunctionality. Double transduction of human CTLs with TAA-specific TCR and NKRs resulted in strongly enhanced antigen sensitivity, without a reduction in the antigen specificity and selectivity of killer function. In addition, the elevation of NKR-Ligand expression on cancer cells by chemotherapy also increased CTL recognition of cancer cells expressing low levels of TAA. Our data help to explain the ability of self-antigens to mediate tumor rejection in the absence of autoimmunity and support the development of dual-targeting adoptive T cell therapies that use NKRs to enhance the potency and selectivity of recognition of TAA-expressing cancer cells.

Synopsis:

The authors show that NKR signals enhance TCR signaling in human T lymphocytes, allowing them to selectively recognize low levels of tumor-associated antigens on “distressed” cancer cells, providing tools to develop novel immunotherapies with enhanced potency and selectivity.

INTRODUCTION

Cytotoxic CD8⁺ T lymphocytes (CTL) recognizing unmutated tumor-associated antigens (TAA), preferentially expressed by cancer cells, but also present in healthy tissues, constitute an important component of spontaneous and immunotherapy-induced cancer immunity (1–4). However, the mechanisms of their selectivity in killing cancer cells are only partially understood, limiting the effectiveness and safety of their therapeutic targeting (1–3).

Prototypal TAAs, such as MART-1/Melan-A or NY-ESO-1, are abundant in cancer-bearing individuals and involved in control of highly heterogenous solid tumors (1–7), raising the questions about how TAA-specific CTLs avoid eliminating healthy tissues. This paradox is partially explained by the generally higher expression of TAAs on cancer cells in combination with the low affinity of TAA-specific TCRs (1–3), but the recently-recognized high promiscuity of individual TAA-specific T cells for multiple T-cell receptor (TCR) ligands (1), indicates that additional mechanisms may be involved in the discrimination between TAAs expressed on different target cells.

The “self/non-self” discrimination at the stage of the initiation of adaptive immunity is explained by thymic elimination of CTL precursors bearing high-affinity TCRs able to recognize own (self) antigens and by the “two-signal” paradigm of T-cell activation (8). Resting T cells avoid responses against TCR-binding antigens (signal 1) expressed by healthy cells that do not provide costimulatory signal 2, only responding to signal 1 provided jointly with the CD28-mediated signal 2 by professional antigen-presenting cells (APC), such as dendritic cells (DC) (9). Upon binding to its cognate ligands B7.1 (CD80) and B7.2 (CD86) on activated DCs, CD28 mediates multiple costimulatory effects, reducing the threshold for cognate TCR triggering, promoting T-cell cytokine production, survival and clonal expansion (9, 10). However, the above processes do not prevent multiple low-, medium- and even high-affinity CTLs against TAAs from persisting and expanding in cancer patients, without overt autoimmunity.

In contrast to the requirement for CD28-mediated costimulation during the (cross)priming of naïve CD8⁺ T cells, the role of signal 2 in the reactivation of CTLs is less clear, especially within the context of anticancer responses. Expression of “classical signal 2” ligands is restricted to professional APCs, with non-hematologic (solid) cancers failing to express these molecules. Meanwhile, activated CTLs upregulate expression of CTLA-4, a much higher affinity receptor for B7.1/B7.2, which has an inhibitory function (9). This raises the question of whether effector CTLs can receive an alternative signal 2 to replace classical costimulatory signals during the recognition and killing of TAA-expressing cancer cells.

The NK receptors (NKR) DNAM-1 and NKG2D were first demonstrated to mediate cytotoxicity of NK cells, but have been also shown to regulate CTL function and to mediate TCR-independent “NK-like” activation of cytokine-activated CTLs (11–17), raising the question of whether they have a role in the interplay with TCR during the recognition and killing of cancer cells. Unlike CD28 ligands, the ligands for DNAM-1 and NKG2D are widely expressed by stressed, infected and mutated cells, but are not expressed by most differentiated healthy tissues (18, 19), making them suitable candidates as cancer-specific sources of costimulation for CTLs.

Here we report that DNAM-1, and to a lesser extend NKG2D, regulate TCR functional avidity of human TAA-specific CTLs, allowing effective recognition and killing of low-TAA-expressing cancer cells in a TCR-restricted manner. Our data indicate that the ability of self-antigens to mediate tumor rejection in the absence of autoimmunity benefits from the interplay between NKR- and TCR-mediated cancer recognition. These data may help in the development of new therapies that target NKRs to enhance the potency and selectivity of TAA-specific T cells.

MATERIALS AND METHODS

Human samples

Human melanoma tumor-infiltrating lymphocytes (TILs) from 6 patients were isolated from melanoma tissues collected during surgery using protocols approved by the Institutional Review Board of the University of Pittsburgh (IRB: CR19080226–009), between 2011 and 2016, and stored in LN2 vapor phase until processing. Human peripheral blood cones (byproduct of platelet collection) were obtained from 35 healthy adult volunteers under the Roswell Park Comprehensive Cancer Center IRB-approved protocol 163222, between 2018 and 2023. Additional human peripheral blood leukopaks from 4 healthy donors were purchased from StemCell Technologies Inc, Vancouver, Canada in 2020. Mononuclear cells were isolated from blood cones and leukopaks and stored in LN2 vapor phase until further processing. Informed written consent was obtained from each subject. The studies were conducted in accordance with U.S. Common Rule.

Cell lines, media, and reagents

Human HLA-A*02:01⁺, MART-1⁺ melanoma cell lines Mel526 (RRID:CVCL_8051) and Mel624 (RRID:CVCL_8054) were kind gifts from Dr. Steven Rosenberg (NCI; (6)). Human melanoma cell line 2183-Her4 (RRID:CVCL_L268) was provided by Dr. Marc Ernstoff (Roswell Park) in 2020. Human HLA-A*02:01⁺ cancer cell lines SW620 (Cat# CCL-227, RRID:CVCL_0547), Caco-2 (Cat# HTB-37, RRID:CVCL_0025) and OVCAR3 (Cat# HTB-161, RRID:CVCL_0465) were purchased from ATCC from 2018 to 2022. MART-1–specific clone (clone 40) was established from the expanded metastatic tumor-infiltrated lymph node cells of a melanoma cancer patient at the University of Lausanne, as previously described (22). All cell lines were authenticated and tested negative for mycoplasma contamination, respectively by IDEXX BioAnalytics (2021) and MycoAlert™ mycoplasma detection kit (Lonza; repeated monthly). All cell lines were used within 10 passages. Mel526 and Mel624 were cultured in Dulbecco’s Modified Eagle Medium (Gibco) supplemented with 10% fetal bovine serum (FBS; Gibco). SW620 were cultured in Leibovitz’s L-15 medium (Gibco) supplemented with 10% FBS. Caco-2 were cultured in Minimum Essential Medium (Gibco) supplemented with 20% FBS. 2183-Her4 and OVCAR3 were cultured in RPMI1640 (Gibco) supplemented with 10% FBS. SW620 were incubated at 37°C without CO₂. All other cells were incubated at 37°C with 5% CO₂. CellGenix DC medium (Sartorius CellGenix) was used to generate monocyte-derived dendritic cells. AIM-V medium (Gibco) supplemented with 5% human serum (GeminiBio) was used as the base T-cell culture medium. The following cytokines and reagents were used to generate immature DCs and induce DC maturation, or to activate and culture T cells: GM-CSF (Leukine sargramostim) was purchased from Partner Therapeutics; IL-4, IL-1β, TNFα, and IFN-γ were purchased from Miltenyi; IL-6 was purchased from R&D Systems; IFN-α (Intron A- IFN-α−2b) and IL-2 (Proleukin aldesleukin) were purchased from McKesson; PGE2 and poly-I:C were purchased from Sigma-Aldrich; IL-7 and IL-12p70 were purchased from Peprotech; Staphylococcus Enterotoxin B (SEB) was purchased from List Labs; Dynabeads Human T-Activator CD3/CD28 was purchased from Gibco; purified Streptavidin was purchased from Biolegend to crosslink biotinylated-antibodies for short-term CTL stimulation. Oxaliplatin and cis-Diammineplatinum(II) Dichloride (Cisplatin) were purchased from Sigma-Aldrich.

Generation of DCs

Peripheral blood mononuclear cells (PBMCs) were obtained from blood cones and leukopaks using gradient centrifugation with Ficoll Paque Plus (Sigma-Aldrich). Fractions of monocytes and lymphocytes were further separated using density gradients made with Percoll (Sigma-Aldrich). Monocytes were purified by plastic adherence and cultured for 6 days in 24-well plates in CellGenix DC medium supplemented with 1000 IU/ml GM-CSF and 1000 IU/ml IL-4. At day 6, DCs were exposed to the following combinations of stimuli for 18 hours to induce standard mature DCs (sDCs; induced by 25 ng/ml IL-1β, 50 ng/ml TNFα, 1000 IU/ml IL-6, and 1 μM PGE₂) or α-type-1 polarized DCs (αDC1s; induced by 25 ng/ml IL-1β, 50 ng/ml TNFα, 3000 IU/ml IFN-α, 1000 IU/ml IFN-γ, and 20 μg/ml poly-I:C). These DCs were then used for the induction of MART-1-specific CTLs and the SEB-based polyclonal T-cell activation, as previously described (20, 21).

DC induction of MART-1–specific CTLs in in vitro sensitization

To induce MART-1–specific CTLs, bulk CD8⁺ T cells were isolated from the lymphocyte fraction of PBMCs from HLA-A*02:01⁺ donors by magnetic cell separation using CD8 MicroBeads (Miltenyi). T cells were cocultured with 1 μg/ml MART-1 (ELAGIGILTV; AnaSpec)-loaded autologous αDC1s in a 10:1 (T:DC) ratio. The cultures were supplemented with 50 IU/ml IL-2 and 10 ng/ml IL-7 at day 3 and every 2–3 days afterwards. MART-1–specific CTLs were sorted by flow cytometry based on the staining of MART-1 Dextramer (Immudex Cat# WB2162-PE) on day 8±1. Sorted CTLs were incubated for at least 2 days before using them in functional assays. MART-1–specific clone (clone 40) was restimulated in vitro by MART-1 peptide–loaded αDC1s using the same method.

DC-induced polyclonal T-cell activation and restimulation

To induce polyclonal activation of naïve CD8⁺ T cells by DCs, naïve CD8⁺ T cells were isolated from the lymphocyte fraction of PBMCs by magnetic cell separation using EasySep™ Human Naïve CD8⁺ T Cell Isolation Kit II (StemCell), labeled with CFSE (Invitrogen), and cocultured with SEB (1 ng/ml)-loaded sDCs (generated from autologous monocytes) in a 5:1 (T:DC) ratio. When indicated, prior to coculture, T cells were incubated with anti-human NKG2D (BioLegend Cat# 320814, RRID:AB_2810480) or anti-human DNAM-1 (Abcam Cat# ab33397, RRID:AB_726268) (both 10 μg/ml) and DCs were incubated with recombinant CTLA-4-Ig (Bio X Cell Cat# BE0099, RRID:AB_10949064) (50 μg/ml) for 15 minutes at 37°C, to block NKG2D, DNAM-1, and B7 molecules, respectively. The cultures were supplemented with medium containing the same blocking antibodies along with 50 IU/ml IL-2 and 10 ng/ml IL-7 at day 2. T-cell proliferation was analyzed based on the CFSE dilution using flow cytometry at day 5. When testing the restimulation of CTLs by DCs, Dynabeads-induced CTLs were cocultured with SEB (0.1 ng/ml) loaded autologous sDCs in a 5:1 (T:DC) ratio for 6 hours. When indicated, blocking Abs were applied using the same method as described above. Brefeldin A (Invitrogen) was added to the coculture 4 hours before the end of incubation. CTL restimulation was analyzed based on intracellular IFN-γ staining using flow cytometry.

CTL induction by Dynabeads

Naïve CD8⁺ T cells were isolated from PBMCs or the lymphocyte fraction of PBMCs by magnetic cell separation using EasySep™ Human Naïve CD8⁺ T Cell Isolation Kit II (StemCell). Cells were then activated at 8×10⁴ cells per well in 96-well round-bottomed plates with an equivalent number of washed CD3/CD28-coated Dynabeads, 50 IU/ml IL-2, 10 ng/ml IL-7, and 10 ng/ml IL-12 in 200 μl of AIM-V medium supplemented with 5% human serum (starting point = day 0). Cells were activated for 48 hours, then the beads were magnetically removed, and the cells were incubated in 24-well plates with 50 IU/ml IL-2 and 10 ng/ml IL-7. Cultures were split and replenished with fresh medium supplemented with IL-2 and IL-7 every 2–3 days. Effector CTLs were harvested for functional assays between days 7 and 14.

T-cell activation by immobilized antibodies

To activate naïve CD8⁺ T cells, 1μg/ml solution of anti-human CD3 (OKT3) (BioLegend Cat# 317326, RRID:AB_11150592) was prepared in sterile PBS and dispensed to 96-well flat bottom plate. The plate was sealed and incubated at 4 °C overnight. When indicated, OKT3-coated plates were washed and coated with the following antibody solutions at 37 °C for 2 hours: anti-human NKG2D (BioLegend Cat# 320814, RRID:AB_2810480), anti-human DNAM-1 (Abcam Cat# ab33397, RRID:AB_726268), or mouse IgG1, κ (BioLegend Cat# 400165, RRID:AB_11150399); all were used at 10 μg/ml. CFSE-labeled naïve CD8⁺ T cells were aliquoted into microwells of antibody-coated plates at 2×10⁴ cells per well in 200 μl of culture medium supplemented with 50 IU/ml IL-2 and 10 ng/ml IL-7. To activate CD28, 5 μg/ml anti-human CD28 (BioLegend Cat# 302934, RRID:AB_11148949) was added into cell suspension. On day 3, for each condition, 100 μl supernatant was carefully harvested for IFN-γ ELISA and replenished by 100 μl fresh medium supplemented with IL-2 and IL-7. T-cell proliferation was analyzed based on the CFSE dilution using flow cytometry at day 6. To stimulate effector CTLs, OKT3 was coated at different concentrations as indicated in figure legends, and other antibody stimuli were given in the same method as described above. After 4 hours of stimulation, supernatant was harvested for IFN-γ ELISA, and cells were harvested for cytokine and gene expression profiling. For examining CTL activation by intracellular IFN-γ, brefeldin A was added after 2 hours of stimulation and cells were cultured for an additional 4 hours before intracellular staining.

T-cell transduction

The retroviral vector production and transduction of NY-ESO-1–specific TCR genes 19305DP and CD8SP were performed as previously described (23). RetroNectin (catalog #T1008; TaKaRa) was used for coating of retrovirus. Transduction efficiency was determined by flow cytometry using NY-ESO-1_157–165 (SLLMWITQC) tetramer (MBL Cat#TB-M011–1). To generate a NKG2D/DNAM-1 co-expressing vector, human full-length CD314 and CD226 coding sequences were fused via P2A-skipping site and cloned into the previously described MSCV-based retroviral vector (24). To generate TCR/NKR double-transgenic T cells, HLA-A*02:01⁺ PBMCs were preactivated using 50 ng/ml OKT3 and 300 IU/ml IL-2 (starting point = day 0). On day 2, cells were added to the 19305DP retrovirus–coated plate with T-cell culture medium containing 300 IU/ml IL-2, spun at 1,000×g at 32 °C for 10 minutes and incubated at 37 °C and 5% CO2. The same process was repeated after 8 hours of incubation. On day 3, cells were transduced with NKG2D/DNAM-1 co-expressing retrovirus using the same protocol. On day 4, double-transgenic T cells were harvested and cultured in T-cell culture medium containing 300 IU/ml IL-2. CD8⁺tetramer⁺ T cells were flow sorted on day 7 and cultured for at least 2 days in T-cell culture medium containing 50 IU/ml IL-2 and 10 ng/ml IL-7 before use in functional assays.

Loading cancer cells with tumor-associated antigens (TAAs)

To model cancer cells expressing different levels of TAAs as target cells for TAA-specific CTL recognition and killing, HLA-A*02:01⁺ cancer cells (SW620, Caco-2, or OVCAR3) were suspended at 1×10⁶ cells/ml in T-cell culture medium mixed with MART-1 (ELAGIGILTV; AnaSpec) or NY-ESO-1 (SLLMWITQC; MBL International) at indicated concentrations. Cell suspensions were incubated at 37 °C for 2 hours with mixing by vortex every 30 minutes, followed by washing 3 times using T-cell culture medium to remove residual free peptides.

Antibody blockade of CTL–cancer cell interactions

To block target molecules prior to coculture, CTLs and cancer cells were incubated with indicated antibodies for 15 minutes at 37 °C. For CTLs, the antibodies were anti-human NKG2D (BioLegend Cat# 320814, RRID:AB_2810480) and anti-human DNAM-1 (Abcam Cat# ab33397, RRID:AB_726268) were used at 10 μg/ml. Mouse IgG1, κ (BioLegend Cat# 400165, RRID:AB_11150399) was used as the isotype control for anti-NKG2D and DNAM-1. For cancer cells, anti-human HLA-ABC (BioLegend Cat# 311412, RRID:AB_493132) was used at 5 μg/ml as a suboptimal concentration to partially block MHC class I.

IFN-γ ELISA

IFN-γ in the supernatant of T-cell cultures was measured using a human IFN-γ ELISA kit according to the manufacturer’s protocol (catalog #DY285B, R&D Systems). Plate washing was performed using a BioTek 405LS Microplate Washer. Plate reading was performed using BioTek Epoch Microplate Spectrophotometer with Gen5 software.

IFN-γ ELISPOT assay

To perform IFN-γ ELISPOT assays, 96-well MultiScreen Filter Plates (Millipore) were coated with 10 μg/ml anti-human IFN-γ (mAb 1-D1K) (MABTECH Cat# 3420–3-1000, RRID:AB_907282) at 4 °C overnight, washed with PBS, and blocked with T-cell culture medium at 37 °C for 1 hour before use. 2×10⁴ cells/well TAA-loaded cancer cells or melanoma cancer cells were cocultured with effector cells (numbers are indicated in figure legends) in 100 μl T-cell culture medium at 37°C with 5% CO₂ for 24 hours. Plates were rinsed with PBS containing 0.05% Tween-20 and coated with 10 μg/ml anti-human IFN-γ (7-B6–1–Biotin) (MABTECH Cat# 3420–6-1000, RRID:AB_907272) at 4 °C overnight. Immunospots were developed using VECTASTAIN Elite ABC-HRP Kit, Peroxidase (Standard) and AEC Substrate Kit, Peroxidase (HRP), (3-amino-9-ethylcarbazole) according to the manufacturer’s protocol (catalog #PK-6100 and SK4200, Vector Laboratories, Inc.). Spots were imaged and enumerated using CTL ImmunoSpot S6 Core Analyzer (Cellular Technology Ltd).

Cytotoxicity assays

For the LDH cytotoxicity assay, 1×10⁴ MART-1–specific CTLs were preincubated with indicated blocking antibodies as described above and cocultured with 2×10⁴ MART-1–loaded SW620 in 200 μl T-cell culture medium in a 96-well plate at 37 °C with 5% CO₂ for 24 hours. LDH activity of each sample was determined using a CyQUANT LDH Cytotoxicity Assay Kit according to the manufacturer’s protocol (catalog #C20300, Thermo Fisher Scientific). For the apoptosis assay, 1×10⁵ T cells were cocultured with 2×10⁵ peptide-loaded SW620 in 500 μl T-cell culture medium in 24-well ultra-low attachment plate at 37 °C with 5% CO₂ for 24 hours. Cells were then harvested and stained using Alexa Fluor 488 annexin V/Dead Cell Apoptosis Kit (catalog #V13245, Invitrogen) with an adapted protocol. In brief, cells were washed in cold PBS, and resuspended in 100 μl 1× annexin-binding buffer containing Alexa Fluor 488 annexin V and BV786 mouse anti-human CD8 (BD Biosciences Cat# 563823, RRID:AB_2687487). Samples were incubated at room temperature for 15 minutes and 400 μl 1× annexin-binding buffer containing 1 μM DAPI (Sigma-Aldrich) were added. Samples were then kept on ice and acquired on a flow cytometer immediately.

Intracellular Ca²⁺ flux assay

Cell labeling, stimulation and detection were performed in HBSS (1X) with calcium chloride, magnesium chloride (Gibco) supplemented with 2% FBS. CTLs were labeled with Fluo-4, AM using a Fluo-4 Calcium Imaging Kit according to the manufacturer’s protocol (catalog #F10489, Thermo Fisher Scientific). Cells were then incubated with the following antibodies either alone or in combination at 37 °C for 15 minutes: biotin anti-human CD3 (BioLegend Cat# 317320, RRID:AB_10916519), biotin anti-human NKG2D (BioLegend Cat# 320804, RRID:AB_492958), biotin anti-human DNAM-1 (BioLegend Cat# 338326, RRID:AB_2721495), and biotin anti-human CD28 (BioLegend Cat# 302904, RRID:AB_314306). After incubation, cells were collected by centrifugation and resuspended in assay buffer containing 1 μM DAPI. Intracellular Ca²⁺ levels over time were detected by flow cytometry. For each sample, baseline fluorescent signal was recorded for 30 seconds, then streptavidin was added at a final concentration of 20 μg/ml for crosslinking and the fluorescent signal was followed for an additional 9 minutes.

Flow cytometry

Surface staining was performed in PBS containing 2% BSA, 1 mM EDTA and 0.02% NaN₃. Antibodies used for surface staining are as follows: APC anti-NKG2D (BD Biosciences Cat# 558071, RRID:AB_398654), BV510 anti-NKG2D (BioLegend Cat# 320815, RRID:AB_2562746), FITC anti-DNAM-1 (BD Biosciences Cat# 559788, RRID:AB_397329), BV510 anti-DNAM-1 (BioLegend Cat# 338330, RRID:AB_2728300), BV786 anti-CD8 (BD Biosciences Cat# 563823, RRID:AB_2687487), PerCP/Cyanine5.5 anti-TIGIT (BioLegend Cat# 372717, RRID:AB_2632932), APC anti-CD112R (BioLegend Cat# 301505, RRID:AB_2876586), BV421 anti-CD96 (BioLegend Cat# 338417, RRID:AB_2629536), BV785 anti-PD-1 (BioLegend Cat# 329929, RRID:AB_11218984), BV605 anti-TIM-3 (BioLegend Cat# 345018, RRID:AB_2563859), PE anti-PVR (BioLegend Cat# 337610, RRID:AB_2174019), PE anti-Nectin-2 (BioLegend Cat# 337410, RRID:AB_2269088), PE anti-MICA/MICB (BioLegend Cat# 320906, RRID:AB_493193), PE anti-ULBP1 (R and D Systems Cat# FAB1380P, RRID:AB_2687471), PE anti-ULBP3 (R and D Systems Cat# FAB1517P, RRID:AB_10719122), PE anti-ULBP4 (R and D Systems Cat# FAB6285P, RRID:AB_3083729), and PE anti-ULBP2/5/6 (R and D Systems Cat# FAB1298P, RRID:AB_2214693). For IFN-γ intracellular staining, T cells were stimulated in the presence of brefeldin A as described above. Cells were fixed and permeabilized using the BD Fixation/Permeabilization Kit according to the manufacturer’s protocol (catalog #554714, BD Biosciences), and stained with APC mouse anti-human IFN-γ (BD Biosciences Cat# 554702, RRID:AB_398580) or BV711 mouse anti-human IFN-γ (BD Biosciences Cat# 564039, RRID:AB_2738557). Intracellular Melan A was stained by Alexa Fluor 647 anti-MelanA (Abcam Cat# ab225500, RRID:AB_2868593). For the degranulation assay, 2×10⁵ MART-1–specific CTLs were cocultured with 8×10⁵ MART-1–loaded SW620 in T-cell culture medium for 6 hours at 37 °C, in the presence of PE mouse anti-human CD107a (BD Biosciences Cat# 555801, RRID:AB_396135) and BD GolgiStop protein transport inhibitor (containing monensin) (catalog #554724, BD Biosciences). Cells were then stained with BV786 mouse anti-human CD8 (BD Biosciences Cat# 563823, RRID:AB_2687487) and resuspended in staining buffer containing 1 μM DAPI. When CD8⁺ TILs and peripheral blood cells were used, cells were rested in T-cell culture medium supplemented with 50 IU/ml IL-2 and 10 ng/ml IL-7 overnight, before being stimulated by immobilized antibodies as described above. For cell sorting, sample preparation was performed in PBS containing 2% BSA with 1% penicillin/streptomycin (Gibco). Cells were stained with MART-1 Dextramer (Immudex Cat# WB2162-PE), or NY-ESO-1 Tetramer (MBL Cat#TB-M011–1). To detect CTL–cancer cell conjugates, MART-1–specific CTLs were labelled with CM-Dil according to the manufacturer’s protocol (Invitrogen Cat#C7001), incubated with indicated blocking antibodies, and cocultured with CFSE-labelled MART-1–loaded SW620 in 100 μl T-cell culture medium at E:T ratio of 1:1. Samples were incubated at 37 °C incubator for 10 minutes, then resuspended by adding 400 μl PBS containing 2% BSA and mild pipetting, and immediately acquired on a flow cytometer. Flow cytometry and cell sorting were performed using a BD LSRFortessa Cell Analyzer and a BD FACSAria II Cell Sorter (BD Biosciences). Data was analyzed using FlowJo 10.10.0 (RRID:SCR_008520).

Analysis of synapse formation by ImageStream

DC-primed MART-1–specific CTLs were incubated with CFSE-labelled MART-1–loaded SW620 at an E:T ratio of 1:1 in a 37°C incubator for 15 minutes. Samples were fixed with 4% paraformaldehyde and permeabilized using 0.3% Triton X-100, both at room temperature for 10 minutes. Samples were stained with PE anti-human CD3 (BD Biosciences Cat# 555340, RRID:AB_395746), PerCP/Cyanine5.5 anti-human CD11a/CD18 (BioLegend Cat# 363413, RRID:AB_2721710), BV510 anti-human NKG2D (BioLegend Cat# 320815, RRID:AB_2562747) or BV510 anti-human DNAM-1 (BioLegend Cat# 338330, RRID:AB_2728299), and Alexa Fluor 647 Phalloidin (Invitrogen Cat#A22287) at room temperature for 30 minutes. After staining, samples were washed and immediately acquired on an ImageStream^X MKII (Amnis). Data analysis was performed using IDEAS software version 6.2 (Amnis).

Western blot

To prepare samples for western blot, 3×10⁶ CTLs were incubated with 0.5 μg/ml biotin anti-human CD3 (BioLegend Cat# 317320, RRID:AB_10916519) with or without 10 μg/ml biotin anti-human DNAM-1 (BioLegend Cat# 338326, RRID:AB_2721495) in T-cell culture medium at 37°C for 15 minutes. After incubation, cells were collected by centrifugation and resuspended in 200 μl T-cell culture medium. Samples were mixed with 200 μl HBSS (with CaCl₂, MgCl₂) containing 40 μg/ml streptavidin and incubated at 37°C in a water bath for indicated time length. Cells were collected by centrifugation at 4°C and 1,250×g for 2 minutes and washed with ice-cold PBS. Proteins were extracted with lysis buffer containing 1× HALT Protease and Phosphatase Inhibitor Cocktail and quantitated using Pierce BCA Protein Assay Kit, with both reagents purchased from Thermo Fisher Scientific. SDS PAGE was performed using 4–15% Mini-PROTEAN TGX Precast Protein Gels with Precision Plus Protein Dual Color Standards. Wet transfer was performed using PVDF Membrane. All were purchased from Bio-Rad Life Science. After protein transfer, membranes were blocked with 5% BSA for 30 minutes at room temperature and then incubated with primary antibodies at 4°C overnight: Akt (pan) Rabbit mAb (Cell Signaling Technology Cat# 4691, RRID:AB_915783), phospho-Akt (Ser473) Rabbit mAb (Cell Signaling Technology Cat# 4060, RRID:AB_2315049), p44/42 MAPK (Erk1/2) Rabbit mAb (Cell Signaling Technology Cat# 4695, RRID:AB_390779), Phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) Rabbit mAb (Cell Signaling Technology Cat# 4370, RRID:AB_2315112), and mouse anti-β-Actin (Sigma-Aldrich Cat# A5441, RRID:AB_476744). Membranes were then washed and incubated with fluorescent-conjugated secondary antibodies for 1 hour at room temperature: IRDye 680LT Goat anti-Rabbit IgG (LI-COR Biosciences Cat# 925–68021, RRID:AB_2713919) or IRDye 800CW Goat anti-Mouse IgG (LI-COR Biosciences Cat# 925–32210, RRID:AB_2687825). The protein bands were imaged using Odyssey Fc Imager with Image Studio Lite software version 5.5 (LI-COR Biosciences).

Single cell multiplex cytokine profiling

CTLs were stimulated by immobilized antibodies for 4 hours as described above. Cells were harvested for cytokine profiling using Single-Cell Adaptive Immune Chip and Panel according to the manufacturer’s protocol (catalog #ISOCODE-1001–4 and PANEL-1001–4; IsoPlexis). In brief, cells were stained with Stain Cell Membrane 405 (catalog #STAIN-1001–1; IsoPlexis) and approximately 30,000 cells were loaded onto the chip containing 12,000 chambers prepatterned with an array of 32 cytokine capture antibodies. Chips were incubated in the IsoLight system at 37°C with 5% CO₂ for an additional 16 hours and labelled with detection antibodies. The fluorescent signals of each cytokine at single-cell level were detected and analyzed by IsoLight system with IsoSpeak software version 2.9.0 (IsoPlexis). The polyfunctional strength index (PSI) was computed as the percentage of polyfunctional cells, multiplied by the sum of the mean fluorescence intensity of the proteins secreted by those cells.

RNA sequencing

Dynabead-induced CTLs from 3 donors were stimulated by anti-human CD3 either alone or in combination with anti-human DNAM-1, NKG2D, or CD28 for 4 hours, as described above. Total RNA was prepared using the RNeasy Mini Kit and RNase-Free DNase Set according to the manufacturer’s protocol (catalog #74104 and 79254; Qiagen). Paired-end sequencing was performed by the Roswell Park Genomics Shared Resource on an Illumina NovaSeq 6000. Reads were aligned to the human genome (GRCh38) using STAR (RRID:SCR_004463) (version 2.7.9a) and transcripts were quantified by featureCounts (RRID:SCR_012919) from the Subread package (version 2.0.1) (25, 26). Normalization and differential expression analysis was completed with DESeq2 (RRID:SCR_015687) while modeling sample donor as a covariate (27). Gene Set Enrichment Analysis (GSEA) was performed with the fgsea (RRID:SCR_020938) (version 1.18.0) using rank ordered differential expression and gene sets derived from the Molecular Signatures Database (RRID:SCR_016863) (28).

qPCR

RNA was isolated as described above. Reverse transcription was performed using qScript cDNA Synthesis Kit (QuantaBio) according to the manufacturer’s protocol with a T100 Thermal Cycler (Bio-Rad). 250ng RNA per sample was used to make cDNA in a 20 μl reaction volume. The cDNA product was diluted 5 times. Real-Time PCR assay was performed in CFX96 Real-Time PCR System (Bio-Rad) with 40 replication cycles, using 4 μl cDNA template per reaction and iTaq Universal Probes Supermix (Bio-Rad) according to the manufacturer’s protocol. HPRT1 was used as endogenous control (catalog #4325801; Life technologies). The individual gene expression levels were determined as the relative expression to HPRT1 using 2^- ΔΔCT method. Primers were purchased from Thermo Fisher Scientific, including: PVR (Hs00197846_m1), NECTIN2 (Hs01071562_m1), MICA (Hs00792195_m1), MICB (Hs00792952_m1), ULBP1 (Hs00360941_m1), ULBP2 (Hs01127964_m1), ULBP3 (Hs00225909_m1), RAET1E (ULBP4) (Hs01026643_g1), RAET1G (ULBP5) (Hs01584111_mH), RAET1L (ULBP6) (Hs00867544_gH), IFNG (Hs00989291_m1), GZMB (Hs00188051_m1), and PRF1 (Hs00169473_m1).

Analysis of TCGA-SKCM metastatic melanoma cancer cohort

TCGA-SKCM expression and clinical annotations (352 samples) were obtained from the Genomic Data Commons data portal and processed via TCGAbiolinks package in R using TCGAWorkflow guided practices (29). Associations between normalized DNAM-1 (CD226) and NKG2D (KLRK1) expression and typical CD8⁺ T Cell (CD3G, CD8A, CD8B) and NK cell (NCR1, NCR2, NCAM1) lineage and functional (IFNG, GZMK, GZMB, PRF1) markers was performed via Spearman correlation analysis. Overall survival analysis was conducted by Kaplan-Meier curve and log-rank test using the survival package in R. High and low subsets of indicated genes were defined using median expression or scaled z-scores. Hazard ratios for overall survival were calculated using individual log-transformed gene expression. The functional hotness score for combined expression of CD8A, DNAM-1, and NKG2D was calculated as previously described (30).

Statistical analysis

All statistical analyses were performed using GraphPad Prism 9 (RRID:SCR_002798). Data from replicate cultures were presented as mean ± SD. Data from multiple donors were presented as mean ± SEM. The numbers of replicates and donors are provided in the figure legends. A Student’s t-test or paired t-test was used to compare two independent or matched groups. P-values < 0.05 were considered to be significant (∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001). Due to the character of the study (in vitro experiments), no power analysis, sample size determination or randomization were needed. Unless indicated otherwise, each experiment was performed in triplicate per each condition. All experiments were repeated independently using blood from at least 3 different donors. The relevant numbers of donors and replicates for all experiments are described in the figure legends and were sufficient to perform statistical analyses. All data were recorded and analyzed by software with objective readouts, thus blinding was not relevant to this study. No data were excluded from the analysis.

Data availability:

All data generated in this study are available in the main text or the supplementary materials. TCGA gene expression and clinical annotations were obtained from the Genomic Data Commons data portal. Bulk RNA-seq data reported here are available as gene level expression counts from the NCBI Gene Expression Omnibus (GEO) (RRID:SCR_005012) under accession GSE245936. Raw RNA-sequencing read files are not currently provided to ensure confidentiality and protect participant’s privacy as indicated under New York State law. Researchers interested in additional analyses may request access through the Institutional Review Board (IRB) of Roswell Park Comprehensive Center, subject to approval and compliance with ethical guidelines.

RESULTS

Intratumoral CD8⁺ T cell–associated DNAM-1 and NKG2D predict clinical outcomes in melanoma patients

To gain insights into the role of CTL-expressed DNAM-1 and NKG2D, we performed TCGA analysis of markers of CTLs and NK cells in tumor samples from metastatic melanoma patients. We observed that DNAM-1 (CD226) and NKG2D (KLRK1) expression levels were associated with improved overall survival and higher CTL effector markers IFNG, GZMK, GZMB, and PRF1 (Supplementary Fig. S1).

Both DNAM-1 and NKG2D were most strongly correlated with intratumoral CTL markers CD3G, CD8A, and CD8B, rather than NK-cell markers (Fig. 1A). Moreover, only the patients with both elevated CD8A and elevated DNAM-1 and NKG2D, showed survival superior to CD8A-low patients, while the patients with high CD8A expression alone but low expression of DNAM-1 or NKG2D showed no survival advantage compared to patients with low CD8A expression, and showed significantly worse survival than patients with high expression of CD8A, DNAM-1, and NKG2D, suggesting that CD8⁺ T cells lacking DNAM-1 and NKG2D are not effective in tumor control (Fig. 1B). Analysis of hazard ratios further indicated that combined expression of CD8A, DNAM-1, and NKG2D predicted prognosis better than each single gene expression (Fig. 1C).

Figure 1. CTL-associated DNAM-1 and NKG2D are associated with the effector status of tumor-infiltrating CD8⁺ T cells and predict clinical outcomes of melanoma patients.

(A) The correlation between DNAM-1/NKG2D gene expression and CTL/NK cell lineage markers. Size and color of the squares represent correlation coefficient and p values, respectively. (B) The 10-year overall survival probability of metastatic melanoma (SKCM) patients with different expression of CD8A, DNAM-1, and NKG2D. (C) Hazard ratios of CD8A, DNAM-1, NKG2D and their combination to estimate overall survival (means and upper/lower limits). (D) Expression of DNAM-1 and NKG2D on effector and exhausted CD8⁺ TILs from melanoma patients. Left: A representative dot plot showing effector (Teff) and exhausted (Texh) identified by TIM-3 and PD-1 expression. Right: Expression of DNAM-1 and NKG2D on Teff vs. Texh CD8⁺ TILs from a representative donor are shown in histograms and their mean florescent index (MFI) are compared (n=6 donors, each pair of dots represents TILs from an individual donor). (E) Expression of DNAM-1 and NKG2D on effector vs. exhausted MART-1–specific melanoma TILs. (F) Expression of DNAM-1 and NKG2D on different CD8⁺ T-cell populations in the PBMCs of healthy donors. Populations of naive (Tn), central memory (Tcm), effector memory (Tem), and terminally differentiated effector memory (Temra), were gated based on their expression of CD62L and CD45RA. Expression of DNAM-1 (left) and NKG2D (right) on CD8⁺ cells from a representative donor are shown in histograms and their MFI are shown in bar graphs (n=11 donors, mean ± SEM). Data in (A) were modeled by simple linear regression and analyzed by Spearman correlation. Kaplan-Meier survival curves in (B) were analyzed by log-rank test. Data in (D and F) were analyzed by two-tailed ratio paired t test. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001, not significant (ns): P > 0.05.

The expression of DNAM-1 and NKG2D reflected the functional status of tumor-infiltrating CTLs and circulating CD8⁺ T cells. Within the CD8⁺ TILs of melanoma patients, DNAM-1 and NKG2D were exclusively expressed by TILs at the effector stage (Teff), both within total TILs and the MART-1–specific CD8⁺ TIL population (Fig. 1D-E). In contrast, TIM-3⁺PD-1⁺ exhausted TILs (Texh) lacked surface expression of DNAM-1 and NKG2D. Compared to TIM-3^–PD-1^– effector TILs, exhausted TILs further expressed significantly higher levels of TIGIT and lower levels of CD96 (Supplementary Fig. S2A), which are known to act as competitive inhibitors of the PVR–DNAM-1 interaction and mediate inhibitory signals (19). In line with the exhausted phenotype, CD8⁺ TILs lacking DNAM-1 were not able to degranulate in response to anti-CD3 stimulation (Supplementary Fig. S2B). In analogy to TILs, DNAM-1 and NKG2D were preferentially expressed on peripheral blood central memory and effector memory CD8⁺ T cells, compared to their naïve or terminally differentiated counterparts (Fig. 1F).

DNAM-1 and NKG2D are involved in TCR recognition of melanoma cells by MART-1–specific CTLs

To determine the contribution of DNAM-1 and NKG2D to CTL recognition of cancer cells, we used MART-1 (Melan A), a differentiation antigen expressed by melanoma and normal melanocytes, as a model TAA (5–7). MART-1–specific CTLs were induced by in vitro sensitization (IVS) using autologous MART-1 peptide–loaded DCs and CD8⁺ T cells isolated from healthy normal HLA-A*02:01⁺ donors (Supplementary Fig. S3A). DC-primed MART-1–specific CTLs transiently upregulated surface expression of DNAM-1 and NKG2D, compared to naïve and MART-1–nonspecific CD8⁺ T cells, but lost their expression of these NKRs at later stages of activation (Supplementary Fig. S3B and S3C). Ligands of DNAM-1 (PVR and Nectin2) and NKG2D (MICA, MICB and ULBP1–6) are commonly expressed on human cancer cell lines of diverse tissue histology, such as colon cancer (SW620), ovarian cancer (OVCAR3) and melanoma (2183-Her4, Mel526, Mel624), while none of the cell lines expressed CD28 ligands (Supplementary Fig. S3D).

The CTL recognition of cancer cells expressing different levels of endogenous MART-1/Melan A was evaluated by IFN-γ ELISPOT. Blockade of DNAM-1 and NKG2D significantly inhibited CTL recognition of weakly-immunogenic Mel624 melanoma cells, but not the more immunogenic 2183-Her4 or Mel526 melanoma cells (Fig. 2A and B), indicating a negative correlation between the overall strength of effector response and the inhibitory effect of NKR blockade across multiple experiments (Fig. 2C).

Figure 2. DNAM-1 and NKG2D set the threshold for TCR-dependent recognition and killing of cancer cells presenting low levels of MHC I/TAA peptide complexes.

(A) Different levels of endogenous MART-1/Melan A in melanoma cells are associated with their different levels of (re)activation of DC-primed MART-1-specific CTLs. Left: Expression of MART-1/Melan A in cancer cells are shown in histograms with MFI. Right: IFN-γ ELISPOT (n=4 donors, mean ± SEM). (B) IFN-γ secretion by DC-primed MART-1–specific CTLs against melanoma cells expressing different levels of MART-1/Melan A in the absence or presence of NKG2D/DNAM-1 blockade (n=4 donors, each pair of dots represents means of paired triplicate cultures from each individual donor). (C) Correlation between the inhibitory effect of NKG2D/DNAM-1 blockade and the strength of effector response in the recognition of melanoma cells by DC-primed MART-1-specific CTLs. (D and E) IFN-γ secretion by (D) DC-primed MART-1–specific CTLs and (E) a patient-derived MART-1–specific clone in response to cancer cells (SW620) loaded with high- or low-dose MART-1 peptides (1 or 0.01 μg/ml), in the absence or presence of NKR blockers. (D) Representative images of ELISPOT (triplicate wells per condition) and quantification of spots (n=5 donors, mean ± SEM). (E) Data with MART-1-specific clone from a representative experiment performed in triplicate cultures per condition. (F) IFN-γ secretion by NY-ESO-1–specific TCR-transduced CD8⁺ T cells (19305DP and CD8SP) against cancer cells loaded with decreasing doses of NY-ESO-1 peptides, with or without NKR blockade (triplicate cultures per condition, mean ± SD). (G) Representative brightfield pictures of CTL–cancer cell conjugates, and fluorescent pictures showing CFSE labelled cancer cells, CD3, LFA-1, NKG2D or DNAM-1, F-actin and their colocalization. (H) Degranulation of CTLs under indicated conditions was monitored by surface expression of CD107a on CTLs and presented as histograms with mean fluorescent index (MFI) (left) or the ratios of CD107a MFI in each of the indicated conditions to the MFI of unloaded control (right, n=3 donors). (I) Killing of target cells by DC-primed MART-1–specific CTLs under the indicated conditions was analyzed by LDH cytotoxicity assay (n=3 donors). Each pair of dots represents means of paired triplicate cultures from each individual donor. Data were analyzed by two-tailed ratio paired t test (A, B, D, H, and I), or two-tailed unpaired t test (E and F). Data in (C) were modeled by simple linear regression and analyzed by Pearson correlation. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001, not significant (ns): P > 0.05. ND: not detected.

DNAM-1 and NKG2D set the threshold of TCR-mediated recognition of MHC I/TAA complexes

To test whether the NKR-assisted recognition of cancer cells by DC-primed MART-1–specific CTLs is TCR dependent, MART-1–negative (but HLA-A*02:01⁺) SW620 colorectal cancer cells were loaded with increasing concentrations of MART-1 peptide as target cells. Despite the expression of multiple ligands for DNAM-1 and NKG2D by SW620 cells (Supplementary Fig. S3D), DC-primed CTLs exclusively recognized and killed only MART-1–loaded cancer cells, and fully ignored MART-1–negative cells (Fig. 2A, Supplementary Fig. S3E and S3F). These results demonstrate that the activation of DC-primed CTLs is fully dependent on TCR-delivered signal 1, in contrast to LAK/CIK-type activation, which is TCR independent (13, 15).

The selective role of NKRs in the recognition of cancer cells with lower antigenicity suggested that NKR-mediated costimulation may be particularly needed to assist in T-cell activation in the presence of low levels of TCR–pMHC-I–delivered signals. To test this hypothesis, SW620 were loaded with different concentrations of MART-1 peptide to mimic cancer cells presenting low- and high-levels of TAAs. As shown in Fig. 2D, blockade of NKG2D and DNAM-1, either individually or in combination, had no detectable inhibitory effect on CTL recognition of SW620 cells loaded with high-dose MART-1. In contrast, blockade of DNAM-1 inhibited CTL recognition of SW620 cells pulsed with low doses of the MART-1 peptide, with the maximal inhibition observed upon coordinate blockade of both DNAM-1 and NKG2D. The results were confirmed with additional MART-1–negative cancer cells loaded with increasing doses of exogenous MART-1 (Supplementary Fig. S4A). Similar to in vitro DC–sensitized polyclonal MART-1–specific CTLs from healthy donors, a MART-1–specific CTL clone established from tumor-infiltrated lymph node cells from an HLA-A*02:01⁺ melanoma patient (22) also required NKG2D and DNAM-1 for optimal recognition of HLA-matched tumor cells loaded with low, but not high, doses of the MART-1 peptide (Fig. 2E). Moreover, partial blockade of MHC-I on cancer cells loaded with high-dose MART-1 peptide revealed the same requirement of DNAM-1 and NKG2D in cancer-cell recognition by DC-activated CTLs (Supplementary Fig. S4B).

The same pattern was observed using NY-ESO-1, a TAA shared by normal testis and multiple tumors, as an alternative model TAA, where NY-ESO-1–loaded cancer cells were targeted by NY-ESO-1–specific TCR-transduced CD8⁺ T cells [19305DP and CD8SP (23)]. NKR blockade prevented the CTL recognition of cancer cells presenting low levels of peptides, showing a progressively lesser role in cancer cells expressing higher levels of MHC I/TAA peptide complexes (Fig. 2F).

We next tested the involvement of DNAM-1 and NKG2D in the sequential steps of CTL-mediated cytolysis: cell-conjugate formation, CTL cytoplasmic rearrangement, and degranulation. Coculture of CM-Dil labelled MART-1–specific CTLs with CFSE-labelled MART-1–loaded SW620 showed that blockade of NKRs strongly inhibited the development of CTL conjugation with low-dose MART-1–loaded cancer cells, without affecting conjugation with high-dose MART-1–loaded cancer cells (Supplementary Fig. S4C). ImageStream analyses further revealed that both DNAM-1 and NKG2D were polarized at the CTL–cancer cell contact zone, and colocalized with markers of the immune synapse including CD3, LFA-1, and F-Actin (Fig. 2G). Consistently, CTL degranulation (measured as CD107a translocation to plasma membrane) in response to low-dose MART-1–loaded cancer cells was critically dependent on NKRs (Fig. 2H). Furthermore, LDH cytotoxicity assay confirmed that DNAM-1 and NKG2D blockade inhibited the T-cell killing of low-dose, but not high-dose MART-1–loaded cancer cells (Fig. 2I).

DNAM-1 and NKG2D costimulation enhance CTL polyfunctionality

The costimulatory role of NKR signals was further validated using Dynabead-induced CTLs and freshly isolated peripheral blood CD8⁺ T cells, which all showed a similar dependence on DNAM-1, and to a lesser extent NKG2D, in response to weak TCR triggering by low-dose anti-CD3 (OKT3) (Fig. 3A, Supplementary Fig. S5).

Figure 3. DNAM-1, and to a lesser extend NKG2D, enhance CTL polyfunctionality.

Dynabead-induced CTLs were activated by immobilized antibodies. (A) Intracellular IFN-γ levels in CTLs activated by increasing concentrations of OKT3 in the absence or presence of NKR costimulation. Left: Representative flow cytometry histograms showing intracellular IFN-γ levels of differentially stimulated CTLs. Right: Percentages of IFN-γ–producing CTLs activated by low-dose OKT3 in the absence or presence of NKR costimulation (n=5 donors; mean ± SEM). (B to E) Single-cell secretome of CTLs activated under the indicated conditions were analyzed using Adaptive Immune ISOCODE chips and isoLight system (IsoPlexis), showing summary results of 2 separate experiments using different donors. (B) 3D-UMAP projection showing clusters of CTLs activated by OKT3 alone and CTLs activated by OKT3 plus NKR-costimulation. (C) Expression of effector cytokines was overlayed on t-SNE projections, showing intensity of cytokines in CTLs activated by low-dose OKT3 either alone or in combination with the indicated NKR-costimulation. (D) Polyfunctionality was calculated as the percentages of activated CTLs secreting ≥ 2 types of proteins. (E) Polyfunctional Strength Index (PSI) was computed as the percentage of polyfunctional cells, multiplied by the sum of the mean fluorescence intensity of the proteins secreted by those cells. Proteins were grouped and color-coded based on their functions: Effector: Granzyme B, IFN-γ, Perforin, TNFα; Stimulatory: GM-CSF, IL-2, IL-12, IL-15; Chemo-attractive: CCL4, CCL5; Modulatory: IL4, IL10, sCD40L. Data in (A) were analyzed by two-tailed ratio paired t test. ∗∗P < 0.01.

The DC/target cell–free model of Dynabead-induced CTLs allowed us to perform single-cell secretome analysis of the impact of NKR costimulation on CTL functions. The 3D-UMAP projection of 32 mediators of adaptive immunity showed profound differences between the control (CD3-only activated) CTLs and the NKG2D- or DNAM-1-costimulated CTLs (Fig. 3B). DNAM-1, and to a lesser extent NKG2D, enhanced T-cell secretion of effector function-associated factors including Granzyme B, IFN-γ, Perforin, and TNFα (Fig. 3C), and increased the number of individual CTLs secreting multiple factors (Fig. 3D). The polyfunctional strength index (PSI) reflects the ability of a T cell to carry out multiple functions and has been shown to predict the efficacy of immune therapies (31–33). As shown in Fig. 3E, NKR-mediated costimulation of CTLs increased the PSI score and the ability of CTLs to secrete cytokines across all functional categories.

DNAM-1 costimulation lowers TCR activation threshold and enhances TCR signaling

NKG2D signaling in human NK cells and T cells is known to involve the DAP10 adaptor protein, which uses a similar YXNM motif as CD28 to mediate signal transduction (34, 35). However, the optimal responsiveness of CD8⁺ T cells to NKG2D activation requires IL-15 or high dose IL-2 to induce DAP10 expression (15, 16). Consistently, our results showed only a weak NKG2D-mediated costimulatory effect on CTLs induced in absence of high dose IL-2 or IL-15. In contrast, our observations indicated that DNAM-1 was dominant in assisting TCR-driven activation. In the case of NK cells, DNAM-1 transduces signals through a SLP76/VAV1-PLCγ2 pathway, leading to the activation of transcriptional factors AP-1, NFAT, and NF-κB (19, 36), known to be also involved in TCR signaling.

Using RNA sequencing, we compared the gene expression profiles of early CTL activation induced by anti-CD3 (OKT3) alone or in the presence of DNAM-1, NKG2D, or CD28 costimulation. Principle component analysis indicated a profound separation between DNAM-1 costimulated CTLs and CTLs activated under other conditions (Fig. 4A). Differential gene expression analysis, accounting for donor source as a covariate, revealed DNAM-1 costimulation induced significantly more differentially expressed genes compared to NKG2D and CD28 costimulation (log2FC > 1.0, p adj. < 0.05, Fig. 4B), including key biomarkers of initial effector T-cell responses (37), PRF1, IFNG, GZMB, IL2RA, and IL2 (Fig. 4C, Supplementary Fig. S6A). Gene set enrichment analyses showed strong upregulation of TCR downstream transcriptional factors by DNAM-1 signaling, including those involved in the Myc pathway, IL2-STAT5 pathway and TNFα signaling, as well as genes known to be regulated by CD28 costimulation, such as the AKT–mTORc1 and glycolytic pathways (Fig. 4D and 4E). These data demonstrate that DNAM-1 signaling is superior to NKG2D and even CD28 in providing costimulatory effects to assist early effector phase of TCR-driven CTL reactivation.

Figure 4. DNAM-1 provides superior costimulation of effector CTLs, lowering the TCR activation threshold, enhancing metabolic reprograming, and all major TCR signaling pathways.

(A-E) RNA sequencing: Gene expression profiles of CTLs from 3 donors stimulated by OKT3 either alone (control) or in combination with DNAM-1, NKG2D, or CD28 costimulation. (A) Principal component analysis (PC3 and PC4) comparing patterns of gene expression in CTLs activated under the indicated conditions. (B) Volcano plot showing the differentially expressed genes (DEGs) in CTLs activated with the indicated costimulation versus control CTLs. DEGs with log2 fold change > 1 and adjusted P value< 0.05 were defined as significant. (C) Heatmap showing DNAM-1-unique DEGs which are associated with CTL effector and regulatory functions. (D) Top enriched pathways in DNAM-1-costimulated CTLs compared to control CTLs from Gene Set Enrichment Analysis (GSEA). (E) Enrichment of mTORC1 signaling and glycolysis in DNAM-1-costimulated CTLs compared to control CTLs. (F and G) CTLs were stimulated by streptavidin crosslinking of biotinylated αCD3 (OKT3) with or without biotinylated αDNAM-1. (F) Induction of calcium flux by OKT3 with or without DNAM-1 costimulation. Left: Intracellular calcium levels over time in CTLs stimulated with OKT3 and different concentrations of αDNAM-1 (the arrow indicated the addition of streptavidin). Right: Results from 3 independent experiments were normalized by Fluo-4 Peak MFI Ratio calculated as Fluo-4 Peak MFI divided by baseline MFI (n=3 donors; mean ± SEM). (G) Western blot of phosphorylated AKT and ERK in CTLs stimulated with or without DNAM-1 costimulation at different timepoints. Data in (F) were analyzed by multiple paired t test. ∗P < 0.05, not significant (ns): P > 0.05.

To further validate these results, we tested whether DNAM-1 engagement decreased the threshold of TCR triggering. Dynabead-induced CTLs were labelled with the calcium indicator Fluo-4 and pre-incubated with different doses of biotinylated anti-CD3 and anti-DNAM-1. TCR and DNAM-1 signals were then triggered by cross-linking antibodies with streptavidin. We observed that activation of DNAM-1 increased the TCR-triggered intracellular calcium flux in a DNAM-1 dose-dependent manner. Moreover, DNAM-1 engagement allowed CTLs to respond to low-level CD3 stimulation (which was insufficient to induce calcium flux by itself). In contrast, calcium flux in CTLs induced by high-level CD3 engagement was minimally affected by DNAM-1 engagement (Fig. 4F). In the absence of TCR stimulation, even maximal DNAM-1 cross-linking induced only a low level of delayed calcium signaling in T cells (Supplementary Fig. S6B). In addition to the calcium flux (as a marker of calcineurin–NFAT pathway activation), we observed strong increases in phosphorylated AKT and ERK in DNAM-1–costimulated CTLs, reflecting the respective activation of the AKT/mTOR and Ras/MAPK pathways (Fig. 4G).

Effector versus naïve CD8⁺ T cells rely on DNAM-1 versus CD28 costimulatory pathways

Consistent with RNA sequencing results, calcium flux demonstrated that DNAM-1 is superior to CD28 in supporting CTL activation (Fig. 5A), suggesting that CD8⁺ T cells at different stages of activation preferentially use different costimulatory pathways. To test this hypothesis, naïve CD8⁺ T cells and effector CTLs were stimulated by immobilized anti-CD3 (OKT3), alone or in combination with agonistic antibodies to DNAM-1 or CD28. CD28-costimulation was key to the activation of naïve CD8⁺ T cells, based on both proliferation (Fig. 5B) and IFN-γ secretion (Fig. 5C left). However, CTL reactivation required DNAM-1 signaling, rather than CD28, for the optimal costimulation, with CD28 showing only modest costimulatory effects (Fig. 5C right). These observations were confirmed in a model where naïve CD8⁺ T cells and effector CTLs were activated by SEB-pulsed DCs that expressed both DNAM-1 and CD28 ligands (Fig. 5D). Blockade of CD28 engagement by CTLA4-Ig strongly inhibited the proliferation of naïve CD8⁺ T cells by DCs, while blockade of DNAM-1 showed only weak inhibitory effects (Fig. 5E). In contrast, in preactivated CTLs, blockade of DNAM-1, rather than CD28, showed the dominant inhibitory effect (Fig. 5F).

Figure 5. Effector versus naïve CD8⁺ T cells rely on DNAM-1 versus CD28 as the dominant costimulatory pathways.

(A) Intracellular calcium levels over time in CTLs stimulated with OKT3, OKT3 and αDNAM-1, or OKT3 and αCD28. The arrow indicated the addition of streptavidin for crosslinking of biotinylated antibodies. (B) Effects of DNAM-1 versus CD28 costimulation in activation of naïve CD8⁺ T cells. Left: CFSE dilution of naïve CD8⁺ T cells activated by immobilized OKT3 with different costimulatory signals in a representative donor. Right: Percentages of CFSE^low activated cells (n=3 donors; mean ± SEM). (C) IFN-γ secretion by naïve CD8⁺ T cells (left, n=3 donors) or Dynabead-induced CTLs (right, n=4 donors) activated by immobilized OKT3 with the indicated costimulatory signals (mean ± SEM). (D) Expression of DNAM-1 ligands (PVR and Nectin2) and CD28 ligands (B7.1 and B7.2) on DCs. (E and F) DCs were pulsed with Staphylococcus Enterotoxin (SEB) to induce polyclonal T cell activation. %inhibition = [%Activated T cells (control) - %Activated T cells (blockade)] / [%Activated T cells (control) - %Activated T cells (no SEB)] × 100. (E) Activation of naïve CD8⁺ T cells by SEB-pulsed DCs in the presence of blocking antibodies against DNAM-1 or CD28. Left: Representative flow cytometry histograms of T cell CFSE dilution. Right: Summary data of triplicate cultures from 3 donors showing inhibitory effect of each blocker on naïve cell activation. (F) Re-activation of Dynabead-induced CTLs by SEB-pulsed DCs in the presence of blocking antibodies against DNAM-1 or CD28. Left: Representative contour plots of side scatter and IFN-γ expression. Right: Summary data of triplicate cultures from 3 donors showing inhibitory effect of each blocker on CTL re-activation. (G) A model of NKR-mediated “alternative signal 2” supporting the specificity of CTL anti-cancer function. Data in (B and C) were analyzed by two-tailed ratio paired t test. ∗P < 0.05, ∗∗P < 0.01. ND: not detected.

These activation stage–dependent differences indicate that effector CTLs acquire NKRs, which recognize their ligands that are highly expressed on cancer cells, as a source of an “alternative signal 2.” This alternative signal 2 replaces CD28 costimulation to allow selective recognition and killing of cancer cells expressing TAAs, while healthy cells are ignored due to lack of NKR engagement (Fig. 5G).

Application of NKR-costimulation to adoptive T cell therapies and chemo/immuno-therapies

We next evaluated the relevance of our findings to cancer therapy and tested if the modulation of the levels of DNAM-1 and NKG2D on CTLs can be used to enhance their antitumor activity. We employed retroviral vectors encoding NKG2D/DNAM-1 to engineer overexpression of these NKRs by HLA-A*02:01-restricted NY-ESO-1 TCR-transgenic CD8⁺ T cells (Fig. 6A). As shown in Fig. 6B, the TCR/NKR “double-transduced” T cells demonstrated strongly elevated cytotoxic activity against peptide-loaded SW620 tumor cells when compared to the TCR-only “single-transduced” T cells, especially against low-TAA–expressing targets. Furthermore, the “double-transduced” T cells did not show any increase in nonspecific killing of NY-ESO-1–unloaded cancer cells. These results demonstrate the potential for manipulating the NKR levels on the T cells to enhance the effectiveness of T-cell recognition and killing of weakly immunogenic cancer cells.

Figure 6. Enhancing NKR-mediated recognition improves CTL functional avidity and cancer cell elimination.

(A) Overexpression of NKG2D and DNAM-1 on TCR transduced CD8⁺ T cells. Upper: Design of the NKG2D/DNAM-1 co-expressing vector (IC: intracellular domain; TM: transmembrane domain; EC: extracellular domain; SS: spacer sequence). Lower: Contour plots comparing the expression of NKG2D and DNAM-1 on blood-isolated CD8⁺ T cells transduced with the TCR construct (single-transduced) or with the TCR construct and the additional NKG2D/DNAM-1 co-expressing vector (double-transduced). (B) Survival of SW620 cell loaded with increasing concentrations of NY-ESO-1 (0, 0.01, and 0.1 μg/ml) after 24-hour co-culture with single-transduced T cells or double-transduced T cells. Data show representative contour plots of AnnexinV and DAPI (left) and summary results of independent experiments quantifying the percentage of AnnexinV^–DAPI^– surviving cancer cells (right, n=3 donors; mean ± SEM). (C) Comparison of expression of NKG2D and DNAM-1 ligands on untreated or oxaliplatin-treated (100 μM; 72 hours) SW620 cells. (D) IFN-γ secretion by DC-primed MART-1–specific CTLs against untreated or oxaliplatin-treated SW620 cells loaded with MART-1 peptide (low-dose: 0.01 μg/ml, high-dose: 1 μg/ml). Data show a representative image of ELISPOT performed in triplicate wells per condition and quantification of spot number from the same experiment (mean ± SD). (E) IFN-γ secretion by DC-primed MART-1–specific CTLs against SW620 untreated or pre-treated with oxaliplatin for 72 hours and loaded with 0.001 μg/ml MART-1 peptide. Data show summary result of 5 independent experiments performed in triplicate wells per condition (means and individual data points). (F) Schematic depiction of sensitizing chemo-resistant cancer cells to immune recognition through upregulation of NKR ligands. Data in were analyzed by multiple paired t test (B), or two-tailed unpaired t test (D). ∗P < 0.05, ∗∗P < 0.01, not significant (ns): P > 0.05.

Since the DNA damage response is known to result in elevated expression of NKG2D and DNAM-1 ligands (18, 19, 38), we tested if chemotherapeutic agents could upregulate theses ligands and thus facilitate cancer cell recognition by DC-primed TAA-specific CTLs. Oxaliplatin and cisplatin, frequently used to treat colorectal and ovarian cancers, enhanced the expression of NKR ligands (NKR-L) on SW620 (colorectal) and SKOV3 (ovarian) cells, but not Caco-2 (colorectal). The upregulation of NKR-Ls was achieved with low-dose oxaliplatin and cisplatin, which were insufficient for direct cytotoxic effects (Supplementary Fig. S7). Moreover, SW620 cells surviving prolonged high-dose oxaliplatin exposure exhibited elevated NKR-L expression (Fig. 6C). Oxaliplatin-treated SW620 triggered a significantly stronger response from MART-1–specific CTLs, compared to untreated SW620, especially when loaded with a low-dose of MART-1 peptide (Fig. 6D). NKR blockade counteracted the enhanced CTL recognition of oxaliplatin-treated cancer cells, indicating a key role for elevated NKR-L expression in the immuno-sensitizing effects of oxaliplatin (Fig. 6E). While traditional approaches to combine chemo- and immunotherapy have focused on the induction of immunogenic cell death and depletion of Treg/MDSC by chemotherapy (39), our data highlight the potential synergy at the effector stage of antitumor T-cell responses by sensitizing cancer cells to immune attack, suggesting that the proper timing of chemotherapy and immunotherapy can help eliminate chemo-resistant and weakly immunogenic cancer cell variants (Fig. 6F).

DISCUSSION

We show that DNAM-1, and to a lesser extend NKG2D, enhance CTL functional avidity, allowing TCR-restricted activation of DC-primed CTL by low-level MHC-I/peptide complexes or suboptimal levels of TCR stimulation. Our data help to reconcile the controversies regarding the physiologic role of NKRs on human CD8⁺ T cells, highlight their importance in the CTL effector response against weakly immunogenic cancer cells, and suggest new ways of enhancing the effectiveness of adoptive T cell therapies and other cancer treatments.

Lack of classical signal 2 delivery by cancer cells has been recognized in the field of CAR T-cell therapy as a factor limiting its efficacy. Synthetic CD28 signaling domains (or alternatively, 4–1BB or OX40) have been shown necessary to provide “artificial signal 2”, assuring persistence and therapeutic efficacy of modern CAR T cell–based immunotherapies (40). CAR-NK cells with CAR-linked DNAM-1 intracellular domain have been recently shown to have higher cytotoxic abilities vs. CAR-NK cells integrating CD28 signaling domain (41). These observations, and our current results showing a particularly strong benefit of DNAM-1 costimulation in CTL responses against cancer cells expressing low levels of MHC I/TAA peptide complexes, provide a strong rationale to integrate DNAM-1 as a costimulatory component in the engineering of improved TCR- transgenic T cell and CAR T-cell products for therapeutic intervention.

Effective targeting of TAAs in cancer immunotherapy is limited by the expression of overlapping levels of TAAs between cancer and healthy cells and associated risks of immune toxicities. Current designs of CAR T-cell constructs include intracellular domains providing signal 1 and 2 with the same antigen/ligand binder. While this enhances the potency of CARs, it does not enhance the discrimination between cancer and healthy cells expressing the same target antigens, introducing the risk of autoimmunity and T-cell hyperactivation. In sharp contrast, our data demonstrate that the TAA-specific killing of cancer cells by CTLs can be achieved by double-transduction of CTLs with separate TCR and NKRs constructs, which bind to their separate ligands (pMHC I and NKR-Ls). This supports the feasibility of more selective “dual-recognition” systems involving the delivery of a) signal 1 by TCRs or CARs recognizing tumor antigens and b) delivery of signal 2 by modified NKRs or CARs recognizing NKR-Ls on cancer cells. Such “dual-recognition” may allow CTLs to selectively receive two separate cancer-specific signals to achieve higher functional avidity and enhanced recognition of cancer cells vs. healthy cells, even under conditions when both cell types expressing a comparable level of TAAs.

The heterogenous expression of multiple NKR-Ls, at baseline or in response to stressors such as chemo-, radio- or targeted therapies, suggests the general applicability of targeting NKR-Ls in the context of chemo/immunotherapy of diverse forms of solid cancer. Previous studies have also shown that therapeutic agents and irradiation that induce DNA damage responses can upregulate NKR-Ls in different types of cancers and enhance NK cell–mediated antitumor responses (42, 43). Paradoxically, a high expression level of NKR-Ls has shown to be correlated with poor prognosis of cancer patients, which may result from the ability of soluble NKR-Ls released by cancer cells to block immune cell–expressed NKRs or induce their internalization (44–46). These studies support the need for further in-depth analyses of the impact of different forms of chemo-, radio- and targeted therapy on the induction of cell surface versus soluble NKR-Ls in cancer cells.

Our current studies tested the role of DNAM-1–mediated recognition during the priming stage of naïve CD8⁺ T-cell activation and early effector phase of CTL activation. Our data raises the question of whether there is a role for DNAM-1 at other activation stages of CD8⁺ T cells, including their exhaustion, and whether enhanced delivery of NKR signals may affect antitumor activity of other forms of cancer therapies such as immune checkpoint blockade. Our results raise the possibility that the DNAM-1 signal may override the TCR-modulating suppressive signals provided by inhibitory receptors such as CTLA4, PD1 or other inhibitory checkpoint molecules present on exhausted T cells and repetitively stimulated CAR T cells (47), which may help them to survive and retain effector polyfunctionality and prolong cytotoxic activity against large tumor masses. Moreover, the DNAM-1/TIGIT axis is analogous to the CD28/CTLA4 axis, as DNAM-1 and TIGIT compete for PVR in providing stimulatory versus inhibitory signals to CTLs (48, 49). Our data may also help to understand the immune deficit in DNAM-1^– CD8⁺ T cells, which have been identified as a dysfunctional T-cell subpopulation of TILs associated with disease progression (50, 51). Meanwhile, the frequency of DNAM-1^highCD8⁺ T cells appears to hold positive predictive value in patients receiving anti-TIGIT therapy (52). These findings, together with our data showing that overexpression of NKRs benefits CTL function, suggest that enhancing the bioavailability of PVR to DNAM-1 may interfere with TIGIT-mediated suppression and potentially improve the efficacy of TIGIT blockade therapy.

Due to key differences of DNAM-1 and NKG2D signaling and ligand utilization between human and mouse systems, our preclinical studies of adoptive transfer of NKR-assisted TCR- and CAR-transgenic T cells and their combinations will prioritize patient-derived xenograft models and humanized mouse models to validate the antitumor efficacy of enhanced NKR-mediated recognition against tumor targets with heterogenous TAA expression.