Blockade of N-Glycosylation Promotes Antitumor Immune Response of T Cells

Adoptive cellular therapy is an approach in which T cells from a patient undergo population expansion in vitro and are then transferred back into the patient with the goal of improving their immune response to cancer (1–3). Transfer of effector cells, such as αβ T cells, γδ T cells, NK cells, and NKT cells, into tumor-bearing recipients has been demonstrated to have therapeutic potential. In a derivative approach, chimeric AgR T cell therapy, T cells are engineered to express a single-chain Fv region fused to the TCR signaling domain (4, 5). Upon recognizing ligands expressed on tumor cells, the transferred lymphocytes are activated and then promote immune reactions that eliminate ligand-expressing cells. For cell-based immunotherapy to succeed, three major problems must be solved: 1) generation of a robust and stable population of T cells, 2) the identification of a tumor-specific Ag that is exclusively expressed on tumor cells but not on healthy cells, and 3) the generation of tumor-infiltrating T cells, which escape the immunosuppressive effect of the surrounding tumor cells.

To be effective, immune cells must undergo correct metabolic reprogramming. Cells produce ATP via two major energy-producing pathways: glycolysis and oxidative phosphorylation (6). Hyperactive glycolysis in T cells can accelerate terminal differentiation, whereas inhibition of glycolysis leads to development of more stable CD8⁺ T cell memory (7). Experiments using the glucose analogue 2-deoxy-d-glucose (2DG) have shown that limiting glycolysis in CD8⁺ T cells favors the establishment of immunological memory (8). In contrast, 2DG regulates N-glycosylation in addition to phosphorylation of glucose (9, 10). Specifically, 2DG is converted to 2-deoxy-d-mannose, which inhibits N-glycan biosynthesis when it is incorporated into the lipid-linked oligosaccharide precursor (11). This raises the question of whether N-glycan modified with 2DG affects T cell immunity.

Glycosylation is an enzymatic process in which glycosidic linkages are formed between saccharides and proteins, lipids, or other saccharides, producing a diverse repertoire of glycans (12, 13). Glycans function as ligands for lectins and also as steric elements that alter molecular interactions at the cell surface and between extracellular compartments. During maturation, the AgR of T cells are posttranslationally modified with N- and O-glycan chains, and the addition of different N-glycan linkages to the TCR controls the adaptive immune response by changing the threshold of TCR activation (12–14). The N-glycan branch is a ligand of various lectins of the galectin family (15). Galectin-3 is cross-linked with the TCR on T cells, and disruption of this interaction promotes TCR clustering with CD8 while decreasing the threshold for activation (13, 14, 16). Furthermore, in postactivated mature cytotoxic T cells, weaker effector activity is associated with reduced colocalization of the TCR and CD8 during anergy, but this phenotype can be reversed by ex vivo treatment with inhibitors of galectin-3 (17). Thus, N-glycosylation can have profound effects on the immune responses of T cells.

In this study, we examined the diverse effects of 2DG on T cell biology. In contrast to a previous report showing that 2DG promotes memory T cell differentiation through its inhibitory effect on glycolysis (8), we found that 2DG treatment augmented the antitumor activity of human T cells against tumor cells expressing high levels of NKG2D ligands. In addition, we found that 2DG diminished the immunosuppression of T cells by tumor-releasing galectins. Collectively, our results suggest a novel metabolic reprogramming strategy involving administration of 2DG during ex vivo expansion of T cells. We propose that 2DG-treated T cells could be suitable for T cell–based immunotherapies against cancer.

For flow cytometry and immunoblotting, the following primary Abs were used: anti–CD3-FITC (A07746; Beckman Coulter), anti–CD8-PE (A07757; Beckman Coulter), anti-CD56- allophycocyanin (362504; BioLegend), anti–CD244(2B4)-PE/Cy7 (329519; BioLegend), anti–CD94-FITC (305504; BioLegend), anti–thymocyte selection-associated high-mobility group box protein (TOX) (682601; BioLegend), anti-Eomesoderm (EOMES) (662001; BioLegend), anti–T-bet (644801; BioLegend), anti–CD107a(LAMP1)-allophycocyanin (328619; BioLegend), anti–CD8-PC5 (A07758; Beckman Coulter), anti–CD314(NKG2D)-PE (A08934; Beckman Coulter), anti–MHC class I chain-related protein (MIC) A/B (320910; BioLegend), anti-ULBP1 (MAB1380; R&D Systems), anti-ULBP2/5/6 (MAB1298; R&D Systems), anti-ULBP3 (MAB1517; R&D Systems), mouse IgG1 (400124; BioLegend), mouse IgG2a (400202; BioLegend), goat anti-mouse IgG-FITC (115-096-146; Jackson Immunoresearch), anti–CD197(CCR7)-FITC (353215; BioLegend), anti-CD45RO-PE/Cy5 (304208; BioLegend), anti-Galectin3 (LS-C140105; LifeSpan BioSciences), anti–TCRαβ-PC7 (306720; BioLegend), anti–TCRγδ-FITC (331208; BioLegend), anti–CD45RA-FITC (A07786; Beckman Coulter), and anti-CD62L (IM2114; Beckman Coulter).

The following reagents were used: α-galactosylceramide (αGalCer) (67576; Sigma-Aldrich) 2DG (D8735; Sigma-Aldrich), Human Cytokine G1 27-plex Panel (M500KCAFOY; Bio-Rad laboratories), Human Perforin ELISA Kit (ab46068; Abcam), Human Granzyme B ELISA Kit (ab46142; Abcam), CellstainR–calcein-AM solution (341-07901; Dojindo), CytoRed solution (C410-10; Dojindo), sodium oxamate (O2751; Sigma-Aldrich), bromopyruvic acid (Sigma-Aldrich), D-mannose (M2069; Sigma-Aldrich), L-mannose (M1308; Tokyo Kasei Kogyou), IntraPrep Permeabilization Reagent (A07803; Beckman Coulter), IL-2 (87890; Nipro), Human IFN-γ ELISA Kit (Ab100537; Abcam), Recombinant Galectin 3 (P734-1; BBI Solutions), lactose (L3750; Sigma-Aldrich), Annexin V–FITC Apoptosis Detection Kit (15342-54; Nacalai Tesque), EasySep Human CD8 Positive Selection Kit II (17953; STEMCELL Technologies), and EasySep Human CD4 Positive Selection Kit II (17952; STEMCELL Technologies).

Blood samples were collected after approval by the institutional review boards of all collaborating institutions. For culture of T cells, PBMCs were isolated from the blood of donors who did not have cancer. PBMCs were cultured for 2 wk with an immobilized anti-hCD3 mAb (5 μg/ml, clone OKT3; Janssen Pharmaceutical K.K., Tokyo, Japan) in ALyS505N-175 (Cell Science and Technology Institute, Tokyo, Japan) supplemented with autologous plasma. For culture of CD8⁺ T cells, CD8⁺ cells were isolated from PBMCs by immunomagnetic positive selection using the EasySep Human CD8 Positive Selection Kit II. The desired cells were targeted with Ab complexes recognizing CD8 and magnetic particles. Labeled cells are separated using an EasySep magnet and suspended in culture medium. For culture of CD4⁺ T cells, CD4⁺ cells were isolated from PBMCs by immunomagnetic positive selection using the EasySep Human CD4 Positive Selection Kit II. For culture of NK cells, NK cells were expanded from PBMCs, as previously described (18).

Flow cytometry was carried out on a FACSCanto II (BD Biosciences). Cells were washed in PBS, dissociated using TrypLE Select (Invitrogen), and centrifuged at 300 × g for 5 min. Cell pellets were resuspended in PBS containing 1% FBS, filtered through a cell strainer to remove aggregates, and incubated with primary conjugated Abs for 1 h on ice. Cells were then washed twice, suspended in PBS with 1% FBS, and analyzed. Unlabeled cells and isotype controls were used to set gates for negative populations. Data analysis was performed using the Kaluza Software (Beckman Coulter).

Microarray analysis was performed by Agilent Technologies. Total RNA was extracted from cell-sorted CD8⁺ IL-2R⁺ T cells treated with or without 2DG. The normalized data were analyzed to identify genes whose expression was upregulated by an arbitrary cutoff of 10-fold.

Total cellular RNA was extracted using the TRIzol reagent and reversed transcribed using SuperScript III First-Strand Synthesis System (Invitrogen). Real-time quantitative PCR was performed using the Applied Biosystem Power SYBR Green PCR Master Mix and analyzed on the company’s 7300 Real-Time PCR System. The following primers (forward and reverse, respectively) were used: human TOX2, 5′-AGTCGGAAGTGCATTTCAAGAT-3′, 5′-GGCCTGAGTGTCTCTGAAGA-3′; human KLRD1 (CD94), 5′-GTGAACAGAAAACTTGGAACGAAA-3′, 5′-AGGCGGTGTGCTCCTCACT-3′; human EOMES, 5′- AAGGCATGGGAGGGTATTAT-3′, 5′-AAACACCACCAAGTCCATCT-3′; human T-box transcription factor (TBX21), 5′-CAGAATGCCGAGATTACTCAG-3′,5′-GGTTGGGTAGGAGAGGAGAG-3′; human IL-2R, 5′-ACGGGAAGACAAGGTGGAC-3′, 5′-TGCCTGAGGCTTCTCTTCAC-3′; human perforin, 5′-AGGAGCTGGGCAGAAGGACAAGA-3′, 5′-CACCATAGAGGGCACAAGGGAAGG-3′; human granzyme B, 5′-GCGGTGGCTTCCTGATACAAG-3′, 5′-CCCCCAAGGTGACATTTATGG-3′; and human ACTB, 5′-CTGGAACGGTGAAGGTGACA-3′, 5′-AAGGGACTTCCTGTAACAATGCA-3′.

HOS cells cultured on glass-bottom plates were stained with CytoRed and T cells with calcein-AM, for 30 min. Labeled HOS cells were incubated with labeled T cells at appropriate E:T ratios (25:1) in reaction medium without IL-2. Plates were mounted with Prolong Gold Antifade (P-36930; Life Technologies), and fluorescence images were obtained randomly 1 and 3 h after the start of coculture with an LSM 700 microscope (Carl Zeiss, Oberkochen, Germany).

Cells (5 × 10⁵) were incubated for 30 min at 4°C with various amounts of IL-2–biotin (R&D Systems, Minneapolis, MN) (0–100 nM) in PBS–BSA (3 mg/ml) and then with streptavidin–Alexa Fluor 633 (SA633; 0.5 mg/ml; Invitrogen) for 15 min at 4°C. Cells were washed two times with PBS. Flow cytometry data were acquired on a FACSCanto II flow cytometer and analyzed with the Kaluza Software.

Cells (5 × 10⁵) were incubated at 4°C for 30 min with IL-2–biotin (50 nM). The cells were washed three times with cold HBSS, resuspended in culture medium, and shifted to 37°C. At the indicated times, cells were pelleted by centrifugation, resuspended in 0.5 ml of buffer containing 0.01 M sodium citrate and 0.5 M NaCl (pH 2), incubated for 5 min at 4°C, then neutralized with buffer containing 150 mM NaCl, 5 mM KCl, 1 mM CaCl₂, 1 mM MgCl₂, and 20 mM HEPES (pH 7.4). The cell pellets were lysed in NaCl-Tris-EDTA buffer, incubated with streptavidin–FITC (Funakoshi, Tokyo, Japan), and measured on an Arvo ×40 microplate reader (PerkinElmer, Waltham, MA). The intensity that remained associated with the cells after incubation with low-pH buffer represented internalized IL-2. In parallel, cell surface IL-2–biotin was measured without incubation in low-pH buffer.

Cells (5 × 10⁵) were incubated for 30 min at 4°C with recombinant galectin-3 (10 μg/ml) with or without lactose (0.1 M). Cells were washed twice with PBS and then stained with anti–galectin-3 Ab and FITC-conjugated goat anti-mouse IgG to detect galectin binding. Flow cytometry data were acquired on a FACSCanto II flow cytometer and analyzed with the Kaluza Software.

Analysis of total cellular N-glycans was performed by Sumitomo Bakelite, as previously described (19). For extraction of N-glycans, total proteins were extracted and precipitated from 5 × 10⁸ T cells treated with IL-2 or IL-2 plus 2DG; the proteins were reduced, alkylated, and digested with trypsin. N-glycans were released from trypsin-digested glycopeptides by incubating the samples with peptide-N-glycosidase F (PNGase F). An aliquot of digestive mixture was mixed with N-glyBlotGlyco H beads (Sumitomo Bakelite). After washing, N-glycans were washed, eluted, and purified. For mass spectrometry, purified N-glycans were crystallized, and analytes were subjected to MALDI-TOF mass spectrometry. N-glycan spectral peaks were picked and quantified by normalizing the corresponding peak of each glycan against an internal standard. Compositions of N-glycan spectral peaks were estimated using the GlycoMod database.

Fluorochromasia cell–mediated cytotoxicity assay was performed using Terascan VPC (Minerva Tech, Tokyo, Japan). Target tumor cells, including K562, Daudi, DLD1, and HOS cells (1 × 10⁶), were washed and dissociated, and the cell suspensions were labeled for 30 min with calcein-AM (10 μl) (Dojindo). Labeled target cells were incubated for 4 h with primary T cells at appropriate E:T ratios (25:1) in reaction medium without IL-2. For blocking experiments, T cells were preincubated for 60 min with anti-human NKG2D blocking mAb (MAB139; R&D Systems) or anti-IgG mAb. Cytotoxic activity was evaluated by release of fluorochromasia into the medium; 100% of activity was estimated by release of fluorochromasia from target cells treated with 1% NP-40. Viability of target cells prior to the assay was >90%.

T cells were incubated with tumor cells at a ratio of 1:25 in the presence of anti-CD107b Abs and brefeldin A (10 mg/ml). After 4 h of culture, the cells were washed and stained with allophycocyanin-coupled anti-CD8 mAb. Flow cytometry data were acquired on a FACSCanto II and analyzed using the Kaluza Software.

T cells (2 × 10⁵) in RPMI 1640 (200 μl) were stimulated with recombinant galectin-3 (50 μg/ml) or a control protein, with or without 0.1 M β-lactose for the indicated times. Treated cells were washed twice with PBS and then stained with propidium iodide and FITC–annexin V. Stained cells were acquired on a FACSCanto II and analyzed using the Kaluza Software.

Oxygen consumption rate (OCR) was measured at 37°C using an XF96 Extracellular Flux Analyzers (Seahorse Bioscience). Cells (1 × 10⁴ per well) seeded in 96-well plates were loaded into the machine for oxygen concentration determination. Cells were exposed to oligomycin (1 μM), carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP; 300 nM), and rotenone (100 nM) plus actinomycin (100 nM). After each injection, OCR was measured for 5 min. Representative traces are shown in Fig. 1A. Every point represents an average of four different wells. Basal OCR was calculated as the difference in OCR before and after oligomycin. Maximum OCR was calculated as the difference between OCR measured after FCCP and OCR measured after exposure to rotenone plus actinomycin.

All animal experiments were performed according to the Guide for the Care and Use of Laboratory Animals published by the National Institutes of Health. The institutional Ethics Review Committee for Animal Experimentation approved all experimental protocols.

NOD/Shi-scid,IL-2Rγ knockout Jic mice (NOG mice) (6–8 wk old) were obtained from InvivoGen and housed under specific pathogen–free conditions according to the guidelines of the Animal Care Committee at Osaka University. On day 0, 1 × 10⁶ luciferase-expressing K562 cells (K562-Luc) human leukemic cells were i.v. transplanted into the mice. Subsequently, αβ T cells treated with or without 2DG were injected i.v. once a week at a dose of 1 × 10⁷ cells per mouse. Transplanted mice underwent in vivo bioluminescence imaging at the indicated times. Luciferase-based bioluminescence imaging was performed on an IVIS 200 imaging system equipped with a camera box and warming stage. Following i.p. injection of 150 mg/kg D-luciferin, mice were imaged with 5 min exposure. Images were captured, and bioluminescence intensity was quantitated using the Living Image 3.2 acquisition and analysis software (Caliper Life Sciences, Waltham, MA). Total values were determined by drawing regions of interest of identical sizes over each mouse and are presented as photons per second.

Data are presented as means with SE. Comparisons between two groups were evaluated using the JMP 11.0.0 software (SAS Institute, Cary, NC). Wilcoxon rank-sum test was used to analyze the data. Comparisons among three groups in the NOG mouse xenograft experiment were evaluated using the Wilcoxon rank-sum test. A p value < 0.05 was considered to indicate statistical significance.

To better understand the effects of 2DG on human T cell development, we cultured human PBLs (PBMCs) in the presence of IL-2 and mAb against CD3, a conventional protocol for promoting T cell proliferation. The cells were cultured for 14 d of culture with various concentrations of 2DG (0, 0.5, 1, 2, 3, 4 mM). Treatment with 2DG decreased total cell counts in a dose-dependent manner (Supplemental Fig. 1A). Populations of αβ T cells (CD8^high, TCRαβ^high), γδ T cells (CD8^high, TCRγδ^high), and NK cells (CD3^low, CD56^high) among PBMCs and populations of CD8⁺ and CD4⁺ T cells among CD3⁺ T cells, were measured in cultures grown in control and 2 mM 2DG-containing medium (Supplemental Fig. 1B). Each population was similar between 2DG-containing and control medium (Supplemental Fig. 1B). Measurement of metabolic changes in the extracellular acidification rate (ECAR), an indicator for overall glycolytic flux, and the OCR, an indicator for mitochondrial respiration (Supplemental Fig. 1C), revealed that human PBMCs treated with 2DG had a higher OCR/ECAR ratio. This indicated that 2DG switches metabolic reprogramming from glycolysis to oxidative phosphorylation, a metabolic state that favors memory T cell differentiation (7). Central memory T cells (CD45RA^low, CD62L^high or CCR7^high, CD45RO^high) were detected by flow cytometry (Supplemental Fig. 1D). In line with the observed metabolic remodeling, flow cytometry detected a slight increase in the number of central memory T cells in the 2DG-treated samples.

T cells cultured without glucose or with reduced glycolytic activity express low levels of cytotoxic molecules such as perforin and granzyme (8, 20, 21). Unexpectedly, however, we found that human 2DG-treated T cells secreted higher levels of cytokines, including IFN-γ, TNF-α, IL-17, MIP1a, and MIP1b (Fig. 1A, Supplemental Fig. 2). 2DG-treated T cells also expressed higher levels of cytotoxic effector molecules, such as perforin and granzyme B, than control T cells, but lower levels than NK cells (Fig. 1B, 1C). Therefore, we suspected that human 2DG-treated T cells had acquired NK cell properties. To explore this possibility, we used flow cytometry to examine the expression pattern of NKR on 2DG-treated T cells (Fig. 1D, 1E). 2DG-treated T cells expressed early NK cell markers CD56, 2B4 (CD244), and CD94 at higher levels than control T cells but at lower levels than NK cells. Thus, human T cells acquired NK cell properties following 2DG treatment.

A unique subset of T cells, designated NKT cells, express typical NK receptors and produce substantial amounts of cytokines, especially IL-4, upon stimulation through the TCR (22). To determine whether 2DG-treated T cells would differentiate into NKT cells, we again performed flow cytometry to monitor the expression of Vα24 TCR, which is exclusively present in NKT cells (Supplemental Fig. 3A). NKT cells are CD1d-restricted T cells that are activated by glycolipid Ags such as αGalCer (22). In the absence of αGalCer, NKT cells (CD3^high, TCRVα24^high) were similar in control and 2DG-containing medium (0.96 versus 0.85%) (Supplemental Fig. 3A). In the presence of αGalCer, NKT cells proliferated in control and 2DG-containing medium (11.23 versus 5.66%) (Supplemental Fig. 3A). These results indicated that 2DG treatment did not promote differentiation of NKT cells. Thus, 2DG induced an NK-like T cell lineage distinct from NKT cells.

Next, we conducted DNA microarray experiments to identify genes that were differentially expressed in CD8⁺ IL-2R⁺ T cells derived from PBMCs treated with or without 2DG. After normalization and filtering of signals, this analysis revealed that 301 probes were at least 10-fold upregulated in 2DG-treated T cells relative to control T cells (Supplemental Fig. 4). In particular, various types of chemokine genes were significantly upregulated. In addition, TOX2, killer cell lectin-like receptor subfamily D, member 1 (KLRD1; also known as CD94), TBX21, and EOMES, all of which are transcription factors involved in induction of NK cell differentiation (23, 24), were highly expressed in 2DG-treated T cells (Supplemental Fig. 4). The microarray data regarding TOX2, CD94, TBX21, and EOMES were confirmed by real-time quantitative PCR (RT-qPCR) and immunoblotting (Fig. 1F, 1G). Of note, the acquisition of NK cell properties occurred within 7 d after the initiation of 2DG treatment (Fig. 1F). Thus, microarray, RT-qPCR, and immunoblot data supported the idea that 2DG-treated T cells acquired NK cell properties via activation of an NK cell–specific transcriptional regulation.

Based on the elevated expression of cytotoxic molecules in 2DG-treated T cells, we performed an in vitro cytotoxicity assay using fluorescence-labeled tumor cells with cultured T cells. Owing to the suppressive effect of 2DG on cell growth, especially above 3 mM (Supplemental Fig. 1A), T cells cultured with three concentrations of 2DG (0.5, 1, 2 mM) were subjected to a cytotoxicity assay (Supplemental Fig. 3B). 2DG-treated T cells exhibited increased cytotoxicity against K562 and HOS in a dose-dependent manner. Of the four kinds of tumor cells used as targets, 2DG (2 mM)-treated T cells yielded higher cytotoxicity against K562, HOS, and DLD-1 but not Daudi cells than T cells not treated with 2DG (Fig. 2A). Enhanced cytotoxicity was confirmed by examining degranulation of T cells, as estimated by surface expression of CD107a, a prerequisite for cytolysis. At rest, CD107a resides in cytolytic granules; upon activation, it is mobilized to the cell surface by exocytosis. The enhanced cytotoxicity was paralleled by an increase in the percentage of degranulating T cells (Fig. 2B). These results indicated that 2DG enhanced the cytotoxicity of T cells against tumor cells. When CD8⁺ T cells isolated from cultured T cells were used in the cytotoxicity assay, CD8⁺ T cells treated with 2DG (2 mM) yielded higher cytotoxicity against K562, HOS, and DLD-1 tumor cell lines but not Daudi cells than CD8⁺ T cells not treated with 2DG (Supplemental Fig. 3C). By contrast, when CD4⁺ T cells isolated from cultured T cells were used in the cytotoxicity assay, the CD4⁺ cells were much less cytotoxic to tumor cell lines, irrespective of 2DG treatment, than CD8⁺ T cells treated with 2DG (Supplemental Fig. 3D). Thus, CD8⁺ T cells are a major subset of T cells responsible for the enhanced cytotoxicity of 2DG-treated T cells.

T cell–mediated killing of tumor cells depends on activation of the receptor for tumor-specific Ag (25). NKG2D, a receptor for NKG2D ligands, is a C-type, lectin-like, type II transmembrane glycoprotein expressed in human NK and CD8⁺ T cells (26). Flow cytometry of T cells revealed that 2DG-treated T cells had levels of NKG2D on the cell surface similar to those of control T cells (Fig. 2C). Two families of NKG2D ligands have been identified in human cells: MICA and B and HCMV16-binding protein (ULBP1–6) (25). Flow cytometry of tumor cell lines revealed that K562, HOS, and DLD1 cells contained high levels of MICA/B and ULBP1–6 on the cell surface, whereas lines such as Daudi had lower cell surface levels of these ligands (Fig. 2D). This finding paralleled the results of the killing assay, which showed that 2DG treatment augmented cytotoxicity against tumor cell lines with high NKG2D ligand expression (e.g., K562, HOS, and DLD1) but not against those with low expression (e.g., Daudi) (Fig. 2A, 2D). The cell–cell interaction was visualized by fluorescence images using coculture of living HOS cells stained with CytoRed and living T cells stained with calcein-AM (Fig. 2E). One hour after the start of coculture, fluorescence images revealed that HOS cells were contacted by 2DG-treated T cells more than control T cells. Thus, the cytotoxicity of T cells against tumor cells could be caused by direct cell–cell interaction. These results suggested that 2DG-mediated enhancement of cytotoxicity may be due to increased interaction between NKG2D on T cells and NKG2D ligands on targeted tumor cells.

To investigate the NKG2D-dependent cytotoxicity of 2DG-treated T cells, we preincubated T cells with anti-NKG2D Abs and then subjected them to the killing assay. Anti-NKG2D Abs decreased the 2DG-mediated enhancement of cytotoxicity against tumor cells expressing high levels of NKG2D ligand (K562, HOS, and DLD1) but did not alter toxicity against low-expressing cells (Daudi) (Fig. 2F). Thus, 2DG enhanced T cell cytotoxicity in an NKG2D-dependent manner.

To determine whether inhibition of glycolysis mediates the 2DG-mediated enhancement of cytotoxicity in T cells, we treated T cells with potent glycolytic inhibitors, oxamate (1 mM) and bromopyruvate (50 μM) (27), which block lactate dehydrogenase and hexokinase II, respectively (Fig. 3A). Flow cytometry of T cells treated with each of these molecules revealed an increase in the population of memory T cells (CCR7^high, CD45RO^high) (Fig. 3B). Moreover, T cells treated with each of these molecules underwent metabolic reprogramming from glycolysis to oxidative phosphorylation (Fig. 3C). However, neither of these compounds enhanced T cell cytotoxicity in comparison with control T cells (Fig. 3D). These results suggested that a mechanism other than inhibition of glycolytic activity contributes to the enhancement of cytotoxicity by 2DG.

In addition to glucose metabolism, 2DG also interferes with N-glycosylation in the endoplasmic reticulum (9, 10) (Fig. 3A). Because of its structural similarity to mannose, 2DG inhibits N-glycosylation via competition with mannose metabolism and misincorporation into dolichol-pyrophosphate (lipid)–linked oligosaccharides, the precursors for N-glycosylation. Because glycosylation shapes the profiles of individual T cell subsets by controlling T cell activation, differentiation, and survival (13, 14), we hypothesized that 2DG modifies T cell properties by blocking N-glycosylation of surface proteins. To determine whether 2DG affected N-glycosylation of surface proteins on T cells, we performed glycoblotting-based quantitative glycomic analysis (19). Total levels of high-mannose–type N-glycans were almost the same in both T cells. However, striking differences were apparent in the composition of these glycans (Fig. 4A, Table I). N-glycans, consisting of (Hex)_0–2 (HexNAc)_1–4 (Sulph)_0–1 (Man)₃ (GluNAc)₂, were major components in T cells cultured in control medium (peaks 1, 3, 5, 7), whereas the levels of these N-glycans were decreased 0.2–0.5–fold in T cells cultured with 2DG-containing medium. The levels of deoxyhexose-containing N-glycans, consisting of (Hex)_1–2 (HexNAc)_1–2 (Deoxyhexose)₁ (Sulph)_0–1 (Man)₃ (GluNAc)₂ (peaks 2, 4, 6), were increased 10–15–fold in 2DG-treated T cells. This result indicated that the primary effect of 2DG on N-glycosylation in T cells was to promote aberrant incorporation of deoxyhexose into mature branching N-glycans rather than blocking synthesis of (Hex)₃ (Man)₉ (GluNAc)₂, a precursor for N-glycans, in the endoplasmic reticulum. Therefore, N-glycosylated surface proteins on 2DG-treated T cells may have different affinities for various ligands.

To determine whether 2DG enhances cytotoxicity against tumor cells by modifying N-glycosylation of proteins in T cells, we took advantage of the finding that inhibition of N-glycosylation by 2DG can be effectively reversed by addition of exogenous D-mannose (10). In particular, we cultured T cells with 2DG in combination with 1 mM D-mannose or 1 mM L-mannose, a biologically inactive stereoisomer, and subjected them to the cytotoxicity assay (Fig. 4B). The addition of D-mannose potently reversed the enhancement of cytotoxicity against tumor cells, whereas addition of L-mannose had no effect. Furthermore, the addition of D-mannose potently reversed the increase in perforin and granzyme B content, whereas, again, L-mannose had no effect (Fig. 4C). Because the ability of 2DG to inhibit glycolysis was not affected by D-mannose, as measured by ECAR and OCR (data not shown), these results supported the notion that 2DG enhances the cytotoxicity of T cells against tumor cells via modification of N-glycosylation.

Exocytosis of cytotoxic molecules, such as perforin and granzyme, is the main mechanism by which cytotoxic T cells kill cancer cells. Perforin expression is upregulated primarily by IL-2R signaling (28). IL-2Rα, which along with IL-2Rβ and γC plays a critical role in T cell function, is highly glycosylated (29). Accordingly, we expected that the enhanced secretion of cytotoxic molecules was due to the enhanced IL-2R signaling in 2DG-treated T cells. Binding of IL-2–biotin at the cell surface of 2DG-treated T cells was analyzed by flow cytometry (Fig. 5A). From an IL-2–biotin saturation curve, the concentration for half-maximal binding (K_0.5) was calculated to be 4–32 nM in both control and 2DG-treated T cells and did not significantly differ between the two. Thus, the affinity of IL-2 binding was the same in control and 2DG-treated T cells, consistent with previous observations that ligand affinity is not altered by variations in N-glycan processing (30, 31). In contrast, saturation was reached at concentrations above 10 nM IL-2–biotin, with a plateau up to 100 nM, and maximum IL-2–biotin binding per cell was increased 4-fold in 2DG-treated T cells relative to control T cells. Thus, the increase in IL-2–biotin binding per cell in 2DG-treated T cells was due to increased surface IL-2R rather than the higher affinity of IL-2R for IL-2. Receptor density at the cell surface is influenced by rates of de novo production, endocytosis, recycling, and degeneration (32). Flow cytometry of 2DG-treated T cells revealed a 2-fold increase in surface IL-2R expression (Fig. 5B). To delineate whether surface IL-2R was retained in 2DG-treated T cells, we analyzed endocytosis of IL-2R (Fig. 5C, 5D). Internalization of IL-2 by T cells was measured by first pretreating the cells with IL-2–biotin in the cold and then shifting the cells to 37°C to assess the fate of cell-bound IL-2. IL-2–biotin internalization was slower in 2DG-treated T cells than in control T cells (Fig. 5C). In accord with the internalization, 2DG-treated T cells had more cell surface–bound IL-2 than control T cells (Fig. 5D). The reduced rate of internalization of IL-2 in 2DG-treated T cells was reversed by D-mannose but not L-mannose. Thus, the modified N-glycan with 2DG slowed endocytosis of IL-2R in T cells. Furthermore, the IL-2R mRNA level in T cells was also increased by 2DG (Fig. 5E). The IL-2–IL-2R signaling induces the transcription of IL-2Rα via STAT5 phosphorylation (33). Together, these data indicate that 2DG-modified N-glycosylation caused T cells to be hypersensitive to IL-2 by increasing the level of IL-2R on the cell surface.

Consistent with the higher sensitivity of 2DG-treated T cells to IL-2, IFN-γ secretion induced by IL-2 in 2DG-treated T cells was augmented in a dose-dependent manner relative to control T cells (Fig. 5F). Thus, 2DG enhanced IL-2R expression on T cells, leading to the IL-2–mediated expression of cytotoxic molecules. Addition of D-mannose but not L-mannose, suppressed IL-2–mediated IFN-γ expression in 2DG-treated T cells (Fig. 5F). These findings indicated that posttranslational modifications, such as N-glycosylation, are important for increasing IL-2R expression and IL-2R downstream signaling in 2DG-treated T cells.

Recent studies showed that tumor-associated galectin contributes to tumor immune evasion by inhibiting the functions of tumor-reactive T cells (13, 14, 16, 34, 35). Because galectin is a carbohydrate-binding protein, we hypothesized that 2DG can block the interaction between galectin and N-glycosylated surface proteins on T cells. Flow cytometry of T cells revealed that T cells bound to FITC-labeled galectin-3, and that this binding was inhibited by β-lactose (0.1 M), a competitive inhibitor of galectin-3 binding (Fig. 6A), indicating that galectin-3 binds to glycan ligands on the T cell surface. By contrast, 2DG-treated T cells bound FITC-labeled galectin to a lesser extent than control T cells. Binding of galectin-3 to T cells was blocked by 2DG in a dose-dependent manner (Fig. 6B). Thus, 2DG blocked the interaction between galectin and N-glycosylated surface proteins on T cells.

Several members of the galectin protein family induce T cell apoptosis in a carbohydrate-dependent manner (13, 14, 17, 36). Hence, we investigated whether 2DG has the ability to block galectin-mediated T cell apoptosis. In the presence of 1.5 μM galectin-3, T cells not treated with 2DG underwent apoptosis to a greater extent than 2DG-treated T cells (Fig. 6C). Thus, 2DG inhibited galectin-3–mediated T cell apoptosis, probably by blocking the interaction between tumor-derived galectin and surface proteins on T cells. Collectively, these observations indicate that the observed enhancement in T cell cytotoxicity is due to the fact that galectin-3 binds more weakly to 2DG-treated T cells than control T cells.

The goal of adoptive cellular therapy is to eliminate tumor cells through the transfer of ex vivo expanded and activated immune cells. We injected NOG mice with K562-Luc (1 × 10⁶) via the tail vein in combination with T cells treated with or without 2DG (10 × 10⁶ cell per mouse per wk). Four weeks after the start of K562-Luc in vivo, we performed bioluminescence imaging (Fig. 6D). NOG mice injected with 2DG-treated T cells exhibited significantly lower bioluminescence than NOG mice treated with PBS. Moreover, injection of 2DG-treated T cells significantly prolonged survival of NOG mice relative to mice treated with control T cells (Fig. 6E). Thus, 2DG promoted the antitumor cytotoxicity of T cells in vivo.

The major finding of this study is that treatment with 2DG allowed T cells to overcome three major barriers against adoptive cellular therapy by undergoing phenotypic changes. First, human 2DG-treated T cells acquired NK cell properties, including surface marker expression, increased expression of cytotoxic molecules, such as perforin and granzyme B, and exhibited enhanced cytotoxicity to tumor cells. Second, 2DG-treated T cells enhanced NKG2D-specific tumor cytotoxicity. Third, treatment with 2DG prevented T cells from binding to galectin-3, a potent tumor Ag associated with T cell anergy. Furthermore, quantitative glycomics and functional assays using D-mannose revealed that 2DG-modified N-glycan plays a role in 2DG-mediated enhancement of T cell–mediated antitumor immunity.

Limiting glycolysis and promoting mitochondrial metabolism during T cell priming allow more cells to enter the memory T cell pool, and the resultant cells maintain antitumor function and persistence after challenge (37, 38). 2DG effectively promotes differentiation into memory-like T cells in mouse (8), but our findings in this study indicate that this process is inefficient in humans. Instead, we unexpectedly found that 2DG caused T cells to acquire NK cell properties. Previous studies using a genetically engineered mouse showed that the IL-15/STAT5 and TOX2/T-bet/EOMES axes act in separate pathways to promote early NK cell development (23, 39, 40). In this study, analysis by microarray, RT-qPCR, and immunoblotting revealed that 2DG-treated T cells expressed very high levels of TOX2/T-bet and EOMES, indicating that 2DG-treated T cells acquired NK cell properties at the transcriptional level. It was of interest to determine the subset to which 2DG-treated T cells belonged. A unique subset of T cells, NKT cells, express NK receptors and rapidly produce substantial amounts of cytokines, especially IL-4, upon stimulation through their CD1d-restricted TCR (22). Despite phenotypic similarities to NKT cells, proliferation in response to αGalCer, a glycolipid Ag that activates NKT cells, was almost identical between control and 2DG-treated T cells. Alternatively, a small subpopulation of human PBMCs are CD8⁺ T cells with NK cell markers (NK-like T cells) (41–43). These NK-like T cells, which are characterized by CD56 and/or CD57 surface expression, exhibit enhanced IFN-γ production and antitumor cytotoxicity, including perforin and granzyme production, after stimulation with IL-2. Although we did not perform a transcriptional analysis or global cytokine analysis of NK-like T cells, we speculate that 2DG-treated T cells may be homologous to NK-like T cells.

Our data on enhanced cytokine secretion from 2DG-treated T cells contrasted with previous studies reporting reduced secretion of some cytokines in these cells (20, 44–46). The main methodological difference in our study was the duration of 2DG incubation in our T cell cultures. The effects of hypoglycolysis on T cells are achieved by 2DG after at most 72 h, during which changes in memory T cell or regulatory T cell differentiation occur (8, 45, 46), whereas in this study we observed NK-like T cell proliferation at least 7 d after the start of culture with 2DG. Therefore, it is likely that NK-like T cell differentiation by 2DG is promoted by other mechanisms, such as blockage of N-glycosylation, rather than hypo-glycolysis, a well-known effect of 2DG.

T cells activated by 2DG preferentially targeted NKG2D ligand-expressing tumor cells. NKG2D ligands exclusively expressed on tumor cells represent useful targets for immunotherapeutic approaches to cancer treatment (25, 47, 48). By binding to NKG2D ligands, NKG2D on T cells activates STAT5-mediated transcription. NKG2D also acts as a costimulatory receptor that augments TCR-induced responses in T cells (49–51). In NK cells, NKG2D acts as a costimulatory molecule, able to induce cytolytic activity, when cotriggered with other activating receptors, such as NKp46 or 2B4 (52). Although NKG2D expression in 2DG-treated T cells was similar to that in control T cells, the levels of many stimulatory cytokines and activating receptor 2B4 were elevated (Fig. 1D, Supplemental Fig. 2). By analogy to the role of NKG2D in NK cells, it is likely that cytokine-mediated priming of NKG2D responsiveness accounts for the enhanced T cell–mediated killing of NKG2D ligand-expressing tumor cells by 2DG-treated T cells. Another advantage stems from the fact that NKG2D-based T cell immunity recognizes target ligands independently of the MHC complex. MHC-independent recognition of target cells allows the transferred effector cells to respond to tumor cells that evade immune detection through downregulation of MHC, enabling these therapies to be used irrespective of MHC haplotype.

N-glycans directly regulate T cell function (13, 14). Glycosylation influences the threshold of TCR activation and has a profound effect on the adaptive immunity. Glycosylated TCR and its coreceptor CD8, expressed on cytotoxic T cells, cooperate in the recognition of a complex formed by antigenic peptide and class I MHC (53, 54). The affinity of the TCR–CD8 complex for the MHC–peptide complex is ∼10-fold higher than that of TCR alone (54). The association between TCR and CD8 depends on reduced glycosylation of TCR. Indeed, mannoside acetylglucosaminyl transferase 5 (Mgt5)–deficient mice, which lack N-glycosylation branch formation, exhibit increased TCR clustering with CD8, resulting in hyperimmune responses and autoimmunity (16). Therefore, the modified glycosylation of surface receptors by 2DG could result in a hyperimmune response, including antitumor cytotoxicity. Furthermore, secretion of cytotoxic molecules, such as perforin and granzyme, is the main mechanism used by cytotoxic T cells to kill target cancer cells, and perforin expression is primarily upregulated by N-glycosylated IL-2R (28). Consistent with this, our results showed that 2DG-treated T cells expressed higher levels of surface IL-2R in parallel with higher responsiveness to IL-2. Thus, 2DG-modified N-glycosylation of IL-2R could augment the surface residency of the receptor, thereby enhancing IL-2Rresponsive T cell immunity.

Galectin, a glycan-binding protein, binds to various glycosylated T cell surface receptors (13, 14, 16). Interaction of galectin-3 with N-glycan impairs TCR clustering and decreases T cell activation by restraining lateral TCR movement (16, 17). Recently accumulated evidence emphasizes that the presence of galectins on tumor cell surfaces contributes to tumor-infiltrating lymphocyte dysfunction (13, 14, 17). Consequently, cross-linking glycoproteins at the T cell surface can induce deactivation and apoptosis of activated T cells (36). Based on our finding that 2DG-treated T cells exhibited reduced binding to galectin-3 and underwent galectin-3–induced apoptosis to a lesser extent than control T cells, it is plausible that infiltrating 2DG-treated T cells in tumor could escape from tumor-mediated anergy.

In conclusion, our results demonstrated that modified N-glycans on tumor-reactive T cells treated with 2DG can improve the efficacy of T cell–based immunotherapies against cancer.

We thank Kumiko Yamashita for technical assistance.

The authors have no financial conflicts of interest.