Stress-Mediated Attenuation of Translation Undermines T-cell Activity in Cancer

Activation of protein synthesis is a requirement for T-cell growth and effector function (1). Eukaryotic translation initiation factor 2 (eIF2) controls cap-dependent protein translation efficiency by bridging Met-tRNA_i and the ribosomal subunit (2). However, endoplasmic reticulum (ER) stress, catalyzed by accrual of unfolded proteins in the ER lumen, undermines the competency of the process. In response to ER stress, the unfolded protein response (UPR) is initiated via phosphorylation of the α subunit of eIF2 causing translation attenuation as a means to restore proteostasis (3). The tumor microenvironment (TME) is replete with metabolic stressors known to activate the UPR (4–7). We and others have shown that PKR ER-like kinase (PERK), a stress sensor responsible for eIF2α phosphorylation (8, 9), undermines T-cell antitumor efficacy (10, 11). Although these studies implicate the UPR-mediated translational machinery as a potential molecular checkpoint prompted by TME stress, the extent to which translational regulation influences outcomes in the context of antitumor immunity is unknown.

Glycolysis is the critical energy requirement for T cells to undergo protein translation (12). However, upon entering the TME, CD8 T cells encounter competition for exogenous glucose, resulting in a significant reduction in effector function (5). In contrast, T cells that depend on metabolic pathways apt for cell survival in nutrient stress, such as gluconeogenesis (13) or fatty acid oxidation (14), demonstrate heightened tumor control (15, 16). Metabolic reprograming through cytokine conditioning (17) or chronic glucose deprivation (15) generates T cells enriched for such pathways that fuel sustenance in nutrient-deplete settings. Although we previously demonstrated that metabolic reprograming through cytokine conditioning generates T cells capable of sustaining protein translation in solid tumors (18), the underlying mechanism supporting this phenomenon remains unknown.

The proteasome is a proteolytic complex responsible for the degradation of ubiquitinated proteins (19). Proteasome inhibition exacerbates ER stress and promotes the UPR, rendering tumor cells susceptible to apoptosis (20). The relationship between protein translation and degradation is symbiotic, as effective protein catabolism precludes activation of the UPR. Memory T cells, the T-cell subset with heightened antitumor efficacy, are enriched for proteasome subunits and exhibit accelerated protein degradation (21). Activation of the proteasome promotes memory T-cell lineage development, linking protein degradation to memory fate (22). Indeed, memory-like T cells are capable of sustaining protein synthesis under TME stress; however, the synergy between sustained translation, protein degradation, mitigating the UPR, and optimal antitumor immunity has not been studied.

In the present study, we investigated the contribution of stress-mediated attenuation of translation to inhibit antitumor T-cell efficacy. We found that the TME induces eIF2α phosphorylation, driven by nutrient deprivation, to restrict protein synthesis and tumor control in T cells. Metabolic reprograming was able to alleviate accumulation p-eIF2α, promoting continued translation in the TME. Sustained protein synthesis was dependent on robust proteasome activation, protecting T cells from p-eIF2α–mediated translation attenuation, proving critical for tumor control. These data indicate that protein degradation underlies antitumor metabolism and translation that can be harnessed to amplify T-cell tumor immunity.

Patients undergoing surgical removal of tumors granted written informed consent via MUSC Biorepository surgical consent forms. All study participants had not recently undergone chemo- or irradiation therapy. This work was determined by MUSC Institutional Review Board to be exempt under protocol Pro00050181. Tissue samples were de-identified. Studies were conducted in accordance with the Declaration of Helsinki, International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS), Belmont Report, or U.S. Common Rule. Blood (8 mL) was collected in EDTA-coated tubes and PBMCs were isolated via Histopaque-1077 centrifugation. Tumor tissue was collected on ice, cutoff into 2-mm³ pieces then dissociated to single-cell suspensions using the human tumor dissociation kit and gentleMACS dissociator (Miltenyi Biotech) according to the manufacturer's protocol. For normal human donor experiments, PBMCs and ImmunoCult Human CD3/CD28 T-cell activators were obtained from Stemcell Technologies (RRID: AB_2827806).

OT-1 (C57BL/6-Tg(TcraTcrb)1100Mjb/J, RRID: IMSR_JAX:003831), eIF2α^S51A± (B6;129-Eif2s1^tm1Rjk/J, RRID: IMSR_JAX:017601), PERK^f/f (Eif2ak3^tm1.2Drc/J), LCK-cre (B6.Cg-Tg(Lck-icre)3779Nik/J, RRID: IMSR_JAX:012837), CD45.1 (B6.SJL-Ptprc^a Pepc^b/BoyJ, RRID: IMSR_JAX:002014), pmel (B6.Cg-Thy1^a/Cy Tg(TcraTcrb)8Rest/J, RRID: IMSR_JAX:005023), and C57BL/6J (RRID: IMSR_JAX:000664) mice were obtained from the Jackson Laboratory. All animal experiments were approved by both the Medical University of South Carolina (MUSC) Institutional Animal Care and Use Committee and the University of North Carolina at Chapel Hill (UNC) Division of Comparative Medicine. Mice were maintained by the Division of Laboratory Animal Resources at MUSC and Division of Comparative Medicine at UNC.

B16F1 (ATCC, RRID: CVCL_0158), B16F1-OVA (kind gift of Dr. Mark Rubinstein), and MCA-205 (Millipore-Sigma, RRID: CVCL_VR90) tumor lines and OT-1 T cells were maintained in RPMI supplemented with 10% FBS, 300 mg/L l-glutamine, 100 U/mL penicillin, 100 μg/mL streptomycin, 1 mmol/L sodium pyruvate, 100 μmol/L NEAA, 1 mmol/L HEPES, 55 μmol/L 2-mercaptoethanol, and 0.2% Plasmocin Mycoplasma prophylactic. 0.8 mg/mL Geneticin selective antibiotic was added to media of B16F1-OVA cells for multiple passages then cells were passaged once in the Geneticin-free media before tumor implantation. MC-38 (Kerafast, RRID: CVCL_B288) and 293-T cells (ATCC, RRID: CVCL_0063) were maintained in DMEM supplemented with 10% FBS, 100 U/mL penicillin, 100 μg/mL streptomycin, and 0.2% Plasmocin Mycoplasma prophylactic. All tumor lines were determined to be Mycoplasma-free in May 2021. For OT-1 T-cell activation and expansion, whole splenocytes from OT-1 mice were activated with 1 μg/mL OVA 257–264 peptide and expanded for 3 days with 200 U/mL rhIL2 (NCI). In some experiments, T cells were split on day 3 and expanded in rhIL2, IL15 (50 ng/mL), or Cyclosporine A (2.5μmol/L) through day 7, or split to normal or low glucose (1 mmol/L) for 16 hours. Human PBMCs were activated with Immunocult CD3/CD28 T-cell activators (STEMCELL Technologies, anti-human CD3 and anti-human CD28 monospecific antibody complexes) and expanded in 500U rhIL2 (NCI), and cultures were maintained in RPMI supplemented with 10% FBS, 300 mg/L l-glutamine, 2 mmol/L GlutaMAX, 100 U/mL penicillin, 100 μg/mL streptomycin, 50 μg/mL gentamycin, 25 mmol/L HEPES, 55 μmol/L 2-mercaptoethanol, and 0.2% Plasmocin Mycoplasma prophylactic.

For endogenous TIL analysis, B16F1 were established by injecting 2.5×10⁵ cells subcutaneously into the right flank of female C57BL/6 mice. Tumor-draining lymph nodes (tDLN), spleens, and tumors were harvested after 6 days of tumor growth for measurement of protein synthesis and cytokine production. For adoptive cellular therapy experiments, B16F1-OVA melanomas were established subcutaneously by injecting 2.5×10⁵ cells into the right flank of female C57BL/6 mice and tumor-bearing hosts were irradiated with 5 Gy 24 hours before T-cell transfer. After 7 days of tumor growth, 5×10⁵ OT-1 WT, PERK KO, or eIF2α^S51A± T cells conditioned with IL2, IL15, vehicle, or Cyclosporin A were infused in 100 μL PBS via tail vein into mice. For proteasome inhibition, vehicle or MG132 (5 μmol/L) was added to T cells 4 hours before infusion. For adoptive cell therapy tumor harvest experiments, 1×10⁶ CD45.2⁺ OT-1 or Thy1.1⁺ pmel T cells treated with vehicle or Cyclosporine A were infused into C57BL/6 CD45.1⁺ mice bearing B16-F1-OVA tumors or C57BL/6 mice bearing B16-F1 tumors, respectively. Tumors were harvested 5 days after T-cell infusion using the mouse tumor dissociation kit gentleMACS dissociator (Miltenyi Biotech) according to the manufacturer's protocol. Tumor growth was measured every other day with calipers, and survival was monitored with an experimental endpoint of tumor growth ≥300 mm².

A total of 2×10⁴, 5×10⁴, 1×10⁵, 2×10⁵, or 4×10⁵ B16F1, MC-38, or MCA-205 tumor cells or 293-T human embryonic kidney cells were seeded into 6- or 12-well companion plates. After 24 hours, OT-1 T cells or PBMCs were introduced into Transwells inserts in complete T-cell media supplemented with 200 U/mL rhIL2 (NCI) or IL15 (50 ng/mL) and harvested 36 hours later. Mouse T cells were added to Transwells at peptide activation, and in other instances added after 3 days of expansion in 200 U/mL rhIL2 (NCI). Human T cells were added to Transwells at soluble CD3/28 activation, and in other instances added after 1 week of expansion. For metabolic remodeling, high glucose (25 mmol/L), 2DG (1 mmol/L, chronic), IL15 (50 nmol/L), or Cyclosporine A (2.5 μmol/L) were introduced at the time of T-cell addition to Transwells. For acute treatment, 2DG (1 mmol/L, acute), or MG132 (5 μmol/L) were added 2 and 4 hours before bioenergetic analysis, respectively. Glucose concentration in Transwell assay media with increasing tumor density was determined using the B61000 Blood Glucose System on the day of Transwell assay harvest.

T cells were lysed in RIPA Buffer (Sigma) supplemented with Protease Inhibitor Cocktail (Cell Signaling Technology) and Phosphatase Inhibitors I and II (Sigma). Protein concentrations were normalized using Pierce BCA Kit (Thermo Fisher Scientific) and loaded to 4%–10% agarose gels (Bio-Rad). P-eIF2α, eIF2α, PERK, CHOP, ATF4, β-actin, and HRP-linked anti-rabbit and mouse secondaries were obtained from Cell Signaling Technology, ERO1α was obtained from Santa Cruz Biotechnologies. Phospho protein was developed with Pierce ECL Plus Western Blotting Substrate (Thermo Fisher Scientific).

Homopropargylglycine (HPG) protein synthesis was measured using the Click-iT HPG Alexa Fluor 488 Protein Synthesis Assay Kit from Thermo Fisher Scientific. Cells were incubated in methionine-free media and controls were treated with cycloheximide. Cells were stained extracellularly using CD8-Alexa Fluor 647 (BioLegend, RRID:AB_389326) and subsequently incubated in 50 μmol/L L-HPG with added Live-or-Dye Fixable Viability Stain. Cells were fixed using Fixation Buffer (BioLegend) and permeabilized with Triton X-100 solution followed by staining with Alexa Fluor 488 Azide to label active protein synthesis. OPP protein synthesis was measured using the O-Propargyl-puromycin (OPP) Protein Synthesis Assay Kit (Cayman Chemical). T cells were incubated in complete T-cell media, and control cells were treated with cycloheximide (CHX). Cells were incubated in cell-permeable OPP, and Live-or-Dye Fixable Viability Stain was added. Cells were fixed using formaldehyde and subsequently stained with 5 FAM-Azide to label translating polypeptide chains. Cells were stained extracellularly using CD8-Alexa Fluor 647. Samples were run directly on a BD Accuri C6 flow cytometer and analysis was performed with FlowJo software (BD Biosciences, RRID: SCR_008520). For measurements of protein translation rates, each T-cell condition was treated ± CHX, and CHX-treated wells were used to normalize protein synthesis rates in various in vivo and in vitro conditions defined as log₂(T-cell protein synthesis rate/T-cell protein synthesis rate in the presence of CHX).

For proteasome activity analysis, Proteasome Activity Probe (R&D Systems) was stained on cells of interest at 2.5 μmol/L for 2 hours at 37 C in PBS. Samples were run directly on a BD Accuri C6 flow cytometer. For in vitro T-cell phenotyping of vehicle, acute, or chronic 2DG (2-deoxy-glucose)-treated, WT or eIF2α^S51A+/−, and vehicle or Cyclosporine A–treated OT-1 or pmel T cells, samples were stained with Zombie NIR viability dye (BioLegend) in PBS at 4C for 15 minutes. After wash, cells were stained with a varied combination of CD8-Spark Blue 550 (BioLegend, RRID: AB_2819773), CD8-spark NIR 685 (BioLegend, RRID: AB_2819775), CD44-Brilliant Violet 785 (BioLegend, RRID: AB_2571953), CD62L-BV421 (BD Biosciences, RRID: AB_2737885), PD-1-BB700 (BD Biosciences, RRID: AB_2869777), TIM-3-Brilliant Violet 711 (BioLegend, RRID: AB_2716208), LAG-3-APC-eFluor 780 (Invitrogen, RRID:AB_ 2637323), ICOS-Super Bright 436 (Invitrogen, RRID: AB_2744818), CD69-PE/Cyanine5 (BioLegend, RRID: AB_313113), CD95-BV480 (BD Biosciences, RRID:AB_2744016), GITR-BV650 (BD Biosciences, RRID:AB_2740316), and CD27-BV750 (BD Biosciences, RRID:AB_2872091) for 30 minutes at 4C. Cells were washed and fixed overnight with using the Foxp3/Transcription Factor Fixation/Permeabilization Concentrate and Diluent kit and Permeabilization Buffer (eBioscience). Cells were then washed with 1X permeabilization buffer and stained with CTLA-4-PE-Dazzle 594 (BioLegend, RRID:AB_2564496), Ki67-PerCP-eFluor 710 (Invitrogen, RRID:AB_11040981), TOX-PE (Miltenyi Biotechnology), TCF1/7-PE-Cy7 (Cell Signaling Technology), Granzyme B-Alexa Fluor 700 (BioLegend, RRID:AB_2728389), BCL-2-Alexa Fluor 647 (BioLegend, RRID:AB_2274702) for 3 hours at room temperature. Samples were collected on a Cytek Northern Lights and analyzed via Cytek SpectraFlo software.

For cytokine restimulated human and mouse T cells, in some instances cells grown in normal media or tumor supernatant were treated with cycloheximide for 30 minutes before being restimulated for 4 hours with Cell Stimulation Cocktail and GolgiPlug Brefeldin A (eBioscience). For tumor infiltrating T-cell analysis, single-cell suspensions were counted and plated at 5×10⁵ cells per well and stimulated with Cell Stimulation Cocktail and GolgiPlug Brefeldin A for 3 hours at 37°C. Samples were then stained with Zombie NIR viability dye (BioLegend) followed CD45.2-BV510 (BioLegend, RRID:AB_2561393) and CD8-Alexa Fluor 647. IFNγ-PE (Invitrogen, RRID:AB_466192) and TNFα-FITC (Invitrogen, RRID:AB_465418) intracellular FACS staining was performed using the Foxp3/Transcription Factor Fixation/Permeabilization Concentrate and Diluent kit and Permeabilization Buffer (eBioscience). Samples were run on a BD Accuri C6 flow cytometer or a Cytek Northern Lights and analysis was performed with FlowJo software.

Seahorse XF Real-Time ATP Rate assays were performed using the Seahorse XFe96 analyzer. 96-well plates were coated with CellTak (Corning), washed, and air dried. T cells were plated in Seahorse XF DMEM Medium (Agilent) supplemented with 1% FBS and centrifuged for adherence. Next, 1 μmol/L oligomycin and 2 μmol/L rotenone/1 μmol/L antimycin A were injected sequentially, and the oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured. The mitochondrial ATP production rate was quantified on the basis of the decrease in the OCR. The glycolytic ATP production rate was calculated as the increase in the ECAR combined with total proton efflux fate. Seahorse XF Cell Mito Stress Test assays were performed as above and 1 μmol/L oligomycin, 1.5 μmol/L FCCP, and 2 μmol/L rotenone/1 μmol/L antimycin A were injected sequentially, and the OCR was measured. Spare respiratory capacity was calculated as the difference between the basal and maximal OCR readings after addition of FCCP.

RNA as isolated with RNeasy Mini Kit (Qiagen, 74104) and concentration was measured using the SpectraDrop Micro-Volume Microplate (Molecular Devices). Single-strand cDNA was made with 500-ng RNA using the High-Capacity RNA-to-cDNA Kit (Applied Biosystems, 4387406, Thermo Fisher Scientific). TaqMan Gene Expression Array Plates (Thermo Fisher Scientific) were used to perform UPR arrays using the StepOnePlus Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific). Gene expression was normalized using global normalization via the Thermo Fisher Connect Platform analysis software.

Cells were harvested and washed with sodium chloride solution then resuspended in 80% methanol for multiple freeze-thaw cycles at −80°C and stored at −80°C overnight to precipitate the hydrophilic metabolites. Samples were centrifuged at 20,000 × g and methanol was extracted for metabolite analysis. Remaining protein pellets were dissolved with 8 mol/L urea and total protein was quantified using BCA assay. Total protein amount was used for equivalent loading for high performance liquid chromatography and high-resolution mass spectrometry and tandem mass spectrometry (HPLC-MS/MS) analysis.

Samples were dried with a SpeedVac then 50% acetonitrile was added for reconstitution followed by overtaxing for 30 seconds. Sample solutions were centrifuged and supernatant was collected and analyzed by HPLC-MS/MS. The system consists of a Thermo Q-Exactive in line with an electrospray source and an Ultimate3000 (Thermo) series HPLC consisting of a binary pump, degasser, and auto-sampler outfitted with a Xbridge Amide column (Waters; dimensions of 2.3 mm × 100 mm and a 3.5 μm particle size). The mobile phase A contained 95% (vol/vol) water, 5% (vol/vol) acetonitrile, 10 mmol/L ammonium hydroxide, 10 mmol/L ammonium acetate, pH = 9.0; B with 100% Acetonitrile. For gradient: 0 min, 15% A; 2.5 min, 30% A; 7 min, 43% A; 16 min, 62% A; 16.1–18 min, 75% A; 18–25 min, 15% A with flow rate of 150 μL/min. A capillary of the ESI source was set to 275°C, with sheath gas at 35 arbitrary units, auxiliary gas at 5 arbitrary units and the spray voltage at 4.0 kV. In positive/negative polarity switching mode, an m/z scan range from 60 to 900 was chosen and MS1 data were collected at a resolution of 70,000. The automatic gain control target was set at 1 × 10⁶ and the maximum injection time were 200 ms. The top 5 precursor ions were subsequently fragmented, in a data-dependent manner, using the higher energy collisional dissociation cell set to 30% normalized collision energy in MS2 at a resolution power of 17,500. Besides matching m/z, metabolites were identified by matching either retention time with analytical standards and/or MS2 fragmentation pattern. Metabolite acquisition and identification was carried out by Xcalibur 4.1 software (RRID: SCR_014593) and Tracefinder 4.1 software, respectively.

After identification, samples were normalized by taking the peak AUC for each metabolite per sample and dividing by the quotient of the total ion count (TIC) per sample over the lowest TIC in the batch. Subsequent transformation of normalized data was carried out with auto scaling to account for heteroscedasticity (23). Metabolites that were below detection in all samples were removed from analysis; missing values were imputed with 1/5 of the minimum positive value of their corresponding variable. Differential metabolite expression between groups of interest were identified through a combination of fold change >2 and raw P value of <0.1 and visualized using a clustered heatmap. Overrepresentation enrichment analysis using the KEGG database (RRID: SCR_012773) was performed on metabolites meeting this criterion to identify biological processes associated with differential expression (24). All statistical analysis was performed using the MetaboAnalyst 5.0 web server (RRID: SCR_015539; ref. 25).

GraphPad Prism v 9.3.0 (RRID: SCR_002798) was used to calculate P values with one-way ANOVA with a Dunnett or Tukey multiple comparison test where appropriate, unpaired two-sided Student t test or paired Student t test as indicated in the figure legend. For tumor growth curves, longitudinally measured tumor size was analyzed using linear mixed effects regression with factors for experimental condition, time (as a continuous variable), and their interaction, and mouse-specific random effects to account for the correlation among measures obtained from the same animal over time. Analysis was performed on the logarithmic scale to satisfy normality assumptions and to induce linearity over time. Group comparisons at specific time points were conducted using linear contrasts and limited to time points with ≥3 measurements per group. Analyses were performed using SAS 9.4 (RRID: SCR_008567; Supplementary Table S1). Survival was calculated using a log-rank, Mantel–Cox test of survival proportions. The P values of <0.05 were considered significant. Numerical P values are indicated in the figures.

The data generated in this study are available upon request from the corresponding author.

Recent advances in single-cell RNA (26) and ATAC sequencing (27) have allowed for identification of genetic and epigenetic traits associated with enhanced antitumor T-cell function. However, the mechanisms responsible for controlling the translation of these instructions into effector functions have yet to be elucidated in antitumor immunity. We recently published an assay that allows monitoring protein synthesis on a per-cell basis (18). The fluorescent analogue of methionine, L-HPG, is incorporated into new forming polypeptide chains and quantified by flow cytometry through Click-IT chemistry (Fig. 1A; refs. 1, 28–30). Using this approach, we assessed global protein synthesis in endogenous CD8 T cells across multiple organs in B16 melanoma-bearing mice. Rates of translation were determined by normalizing to samples treated with translation inhibitor CHX. Compared with splenic and tDLN, tumor-infiltrating CD8 T cells demonstrated a significant reduction in protein translation (Fig. 1B). We replicated this phenomenon using freshly isolated human tumors from various cancer types and patient matched peripheral blood mononuclear cells (PBMC), suggesting that blunted protein synthesis is a common theme of tumor infiltrating CD8 T cells in both humans and mice (Fig. 1C).

The TME consists of a heterogenous milieu of cell types and biochemical processes that coordinate to suppress T cells in both contact-independent and -dependent manners (31, 32). To resolve the critical aspects restricting protein translation in the TME, we used a coculture assay in which tumor cells were seeded for 24 hours before introducing T cells into Transwell inserts for 36 hours (Fig. 1D; ref. 18). OT-1 T cells (Fig. 1E) or normal human donor PBMC (Fig. 1F) were activated and expanded with cognate OVA antigen or soluble CD3/28 activators, respectively. Thereafter, T cells were introduced into tumor or nontumor seeded Transwells for 36 hours (4, 18) and protein translation rates were measured. In both mouse and human T cells the presence of tumor significantly reduced protein translation rates (Fig. 1E and F). The phenomenon was also apparent when human CD8 T cells were activated with soluble CD3/28 activators and expanded in supernatant from freshly isolated B16 tumors (Supplementary Fig. S1A and S1B). We further validated these observations using a second protein translation assay that incorporates O-propargyl-puromycin into the ribosomal A-site allowing for fluorescent labeling of nascent polypeptide chains (Supplementary Fig. S2). We found that reduced translational capacity in T cells was a general phenomenon specific to tumor cell lines as coculture with both MC-38 Adenocarcinoma or MCA-205 Fibrosarcoma mimicked the B16 data. However, when cocultured with 293-T human embryonic kidney cells, T cells experienced minimal reduction in protein synthesis relative to control (Supplementary Fig. S3). These data indicate that tumor cell–specific reduction in protein translation occurs in part through a contact-independent mechanism.

Given that cytokine synthesis is a product of translation (33), we asked whether the reduction in translation in the TME corresponded with diminished cytotoxic cytokine synthesis. Upon coculture with tumor cells, CD3/28-activated human PBMCs displayed reduced TNFα/IFNγ production (Fig. 1G). Pretreating cells with CHX before restimulation resulted in abrogation of TNFα/IFNγ in control and TME CD8 PBMCs, indicating the requirement of continuous translation for cytokine synthesis. Cytokine production in the presence of tumor cells was also reduced using the OT-1 system, further validating our findings (Supplementary Fig. S4). Our data illustrate that soluble factors or competition with tumor cells induce decreased translation, and more specifically cytokine synthesis in CD8 T cells.

Protein translation at the magnitude seen in effector T-cell activation and expansion is a metabolically intensive process requiring ATP (21, 34). A common stressor affecting T-cell function in solid tumors is competition for exogenous glucose (5). Given that glucose fuels effector cell metabolism, in addition to the contact-independent nature of our above findings, we asked whether glucose competition could drive the reduction in protein synthesis in T cells. Incremental increases in tumor cell numbers resulted in a stepwise reduction in environmental glucose in coculture media (Fig. 2A). This was coupled with an identical graded reduction in protein synthesis, IFNγ, and TNFα production (Fig. 2B–D) highlighting the link between glucose availability and CD8 T-cell functional capacity. Exposure to low glucose associated with TME pressure in the coculture system resulted in a significant reduction in ATP output from glycolysis (glycoATP) and mitochondria-linked ATP (mitoATP) in OT-1 mouse and purified human CD8 T cells as measured by Seahorse Bioanalysis ATP rate assay (Fig. 2E and F). The data indicate that competition for glucose significantly alters T-cell metabolism, thereby limiting T-cell effector function.

We next performed a series of rescue or removal experiments to confirm the necessity of glucose in sustaining protein translation in TME stress. By supplementing the T-cell tumor cell coculture media with exogenous glucose we restored glycolytic ATP to no tumor controls levels (Fig. 2G). In line with this, exogenous glucose rescued protein translation rates in the presence of tumor cells (Fig. 2H). We next inhibited glycolysis in T cells in the Transwell coculture system by adding 2DG 2 hours before measurement of protein translation. 2DG is taken up by cell glucose transporters (35) and competitively inhibits generation of glucose-6-phosphate in the second step of glycolysis. Seahorse bioanalysis indicated a reduction in glycoATP (Fig. 2I) in all 2DG conditions irrespective of tumor cells, validating acute treatment with 2DG blunts glycolysis. We also noted a significant reduction in translation rate in T cells treated with 2DG compared with vehicle control irrespective of tumor presence (Fig. 2J). Combined, these data highlight the importance of exogenous glucose in dictating T-cell metabolic programing, protein translation, and effector function.

Metabolic perturbations such as low-glucose environments lead to disturbances in protein folding in the ER (36). In response, the ER stress sensor PERK can mediate phosphorylation of eIF2α at Ser51 (37, 38) that reduces global translation in favor of maintaining proteostasis (39). We found that OT-1 CD8 T cells exposed to low-glucose conditions in vitro demonstrated a significant increase in p-eIF2α  at Ser51 compared with normal glucose controls (Supplementary Fig. S5). Upon coculture with ascending numbers of tumor cells, we noted a stepwise induction of eIF2α phosphorylation that was significantly correlated with the graded reduction in protein translation (Fig. 3A) that resulted in the depression of cytokine synthesis previously observed (Fig. 2). Along with eIF2α phosphorylation, we observed enrichment of PERK axis proteins ATF4, CHOP, and ERO1α (Fig. 3B), indicating that PERK could be upstream of p-eIF2α  in T cells. Gene expression data generated from a targeted UPR gene array also suggested upregulation of PERK-directed genes Ddit3 (CHOP) and Ero1L (Fig. 3C) in T cells exposed to tumor cell pressure.

To determine the contribution of PERK to eIF2α phosphorylation and translation attenuation in TME stress, we generated a TCR transgenic T-cell–specific PERK knockout mouse by crossing OT-1 mice bearing a cre under the LCK promoter and loxp sites flanking exons 7–9 of the PERK gene. T cells from resulting wild-type (OT-1 LCK-cre(−) PERK^f/f; WT) or PERK KO mice (OT-1 LCK-cre(+) PERK^f/f; PKO) were activated with cognate antigen then introduced into the coculture Transwell system. p-eIF2α was absent in PKO cells harvested from control and tumor-seeded Transwells (Fig. 3D), indicating that PERK is the dominant ER stress sensor to mediate eIF2α phosphorylation in T cells responding to tumor pressure. In line with this, translation attenuation was diminished in PKO T cells exposed to tumor stress (Fig. 3E). Interestingly, at baseline PKO T cells displayed reduced protein synthesis relative to WT cells, indicating that other facets of PERK biology could regulate T-cell translation in non-stressed controls.

Next, we asked whether p-eIF2α was the direct arbiter of translation attenuation in T cells exposed to the TME. We backcrossed TCR transgenic OT-1 mice to mice bearing a single amino acid substitution of serine to alanine at codon 51 (S51A) in the phosphorylation site of the eIF2α protein (3). Animals homozygous for S51A mutation die near birth due to defective gluconeogenesis, thus we studied p-eIF2α regulation in mice bearing T cells heterozygous for the S51A mutation (eIF2α^S51A+/−). Western blotting indicated a reduction in p-eIF2α in eIF2α^S51A+/− T cells exposed to TME stress (Fig. 3F) that correlated with an increase in translational capacity relative to littermate controls (Fig. 3G). Phenotypic analysis by spectral flow cytometry uncovered subtle changes between WT and eIF2α^S51A+/− T cells harvested from control and tumor-seeded Transwells. eIF2α^S51A± T cells showed a reduction in CD62 L in control and tumor-seeded Transwells that was accompanied by TME-specific increases in PD-1 and CTLA-4 as well as a surprising increase in TCF-1 expression, suggesting that p-eIF2α–mediated translation is linked to T-cell activation and stemness in stress (Supplementary Fig. S6).

Given that we have previously reported the ability to sustain translation in tumors propels T-cell tumor control (18), we aimed to test the effect of sustained translation, mediated by loss of PERK or p-eIF2α, on antitumor immunity. Subcutaneous B16-OVA melanomas were established in C57BL/6 mice and WT, PKO, or eIF2α^S51A± OT-1 T cells were infused to mice bearing 7-day established tumors. We observed that PKO and eIF2α^S51A± T cells induced superior tumor control compared with WT-matched controls (Fig. 3H and I). Our findings illustrate that the PERK–p-eIF2α axis reduces translational capacity and tumor control of T cells.

A paradigm exists whereby T cells metabolically reprogramed away from exogenous glucose dependency display heightened tumor control marked by amplified cytotoxic cytokine production (15, 16). We next asked whether relief of glucose stress through metabolic remodeling mitigates translation attenuation as a means to sustain antitumor T-cell function. We administered acute (2 hour) or chronic (36 hour) 2DG to OT-1 T cells seeded in the tumor Transwell coculture. We first assessed the effect of 2DG treatment on T cells through in-depth phenotypic analysis. Although chronic 2DG reduced T-cell expansion in vitro (Supplementary Fig. S7A), this was not a result of apoptosis as Annexin V staining and cell viability were not significantly different across groups (Supplementary Fig. S7B and S7C). High-dimensional spectral flow cytometry analysis revealed that chronic 2DG treatment resulted in a significant increase in memory like T-cell markers such as CD44, CD62L, TCF1, and CD69 coupled with a reduction in effector specific markers, including Granzyme B and Ki67 (Supplementary Fig. S7D–S7I). Furthermore, although IFNγ staining intensity was similar across groups, there was a significant increase in TNFα synthesis, suggesting that chronic 2DG treatment enhanced the polyfunctionality of T cells (Supplementary Fig. S7J and S7K). These findings align with previous data showing chronic 2DG treatment skews T cells toward a memory-like phenotype (16). We next investigated the extent to which the metabolic reprograming strategy altered stress-mediated translation attenuation. Western blotting revealed a pronounced reduction in p-eIF2α in chronic compared with acute 2DG treatment or vehicle control (Fig. 4A). Moreover, chronic 2DG remodeling prompted elevated translation in T cells experiencing TME nutrient deprivation (Fig. 4B and C) relative to nontumor controls. Seahorse bioanalysis indicated an increase in glycolytic and mitochondrial ATP output in chronic 2DG-treated T cells under tumor stress (Fig. 4D and E), suggesting complete metabolic reprograming that supported energy production in nutrient stress.

The possibility exists that T-cell translation was bolstered by an adverse effect of chronic 2DG on tumor cells. As chronic 2DG conditioning promotes stem-like memory T-cell development (Supplementary Fig. S7), we next sought to validate elevated translation in TME stress was a property of metabolic remodeling of T cells to the memory lineage by using the cytokine IL15 to generate memory cells (Supplementary Fig. S8). Similar to chronic 2DG treatment, IL15 memory-like T cells harvested from the tumor Transwell coculture assay displayed a significant reduction in p-eIF2α compared with IL2 effectors (Fig. 4F). Furthermore, when comparing IL15 and IL2-treated T cells exposed to the TME, we saw a substantial downregulation of global UPR genes in IL15-conditioned T cells (Fig. 4G), including those within the PERK axis (Eif2a, Atf4, and Ero1L). Similar to chronic 2DG-treated T cells, IL15 skewing enhanced protein translation rates in both nontumor and tumor conditions (Fig. 4H and I) and produced a marked increase in ATP production in the presence of tumor pressure (Fig. 4J and K). These data demonstrate that metabolic reprograming coupled to memory-like T-cell differentiation alleviates p-eif2α translation attenuation.

We next aimed to elucidate the intracellular mechanism that safeguards memory-like T cells from translation attenuation. Several studies have suggested that accelerated protein degradation is a component of memory cell metabolism (21, 22). Given that optimal protein synthesis requires a delicate balance between protein degradation and translation (40), we asked whether the proteasome could be responsible for mitigating translation attenuation in memory-like T cells exposed to TME stress. We used a probe with fluorescent activity proportional to proteasome subunit activation (22, 41) to assess proteasome activity in T cells. In tumor-seeded cocultures, IL15-primed T cells displayed an increase in the frequency of proteasome activity^high cells relative to IL2 effectors that was extinguished in the presence of proteasome inhibitor MG132 (Fig. 5A). Western blotting of IL15-primed T cells revealed a substantial induction of p-eIF2α upon MG132 treatment (Fig. 5B) that correlated with a robust reduction in translation (Fig. 5C). These data suggest that proteasome function is required to overcome translation attenuation in T cells exposed to nutrient stress.

We next sought to address the mechanism by which the proteasome sustains optimal T-cell function. Given that proteasome-mediated protein degradation supports ATP synthesis in cells experiencing nutrient stress (42), we asked whether the proteasome was involved in metabolic remodeling of IL15 primed cells. Proteasome inhibition induced a reduction in glycolytic and mitochondrial ATP in IL15-conditioned T cells in tumor-seeded Transwells (Fig. 5D–E). Metabolomic profiling of IL15-conditioned cells in tumor stress treated with vehicle or MG132 identified 32 significantly enriched metabolites in IL15 T cells related to proteasome activity (Fig. 5F). Pathway enrichment analysis of these metabolites showed upregulation of processes fundamental for memory T-cell–mediated tumor control such as gluconeogenesis and glutathione metabolism (43). The data also evidenced a significant enrichment in aminoacyl-tRNA biosynthesis, further supporting a role for the proteasome as an integral component of sustained protein translation in the TME (Fig. 5G; Supplementary Fig. S9). Taken together, these data indicate that the proteasome plays a critical role in supporting a metabolic profile necessary for increased T-cell tumor immunity.

Functionally, IL15-conditioned T cells are of interest to the immunotherapy field due to their robust and prolonged levels of tumor control relative to effectors. IL15 conditioning has been shown to enhance CAR T-cell therapy (44) and improve response to checkpoint blockade (45). Thus, we tested whether access to proteasome-mediated protein degradation was a molecular checkpoint required for tumor control in memory-like T cells. Mice bearing B16F1-OVA melanomas were infused with IL15-conditioned T cells treated with vehicle or MG132 for four hours before infusion and tumor growth was assessed. Although IL15-conditioned T cells generated potent tumor control, IL15-conditioned T cells unable to access the proteasome showed reduced tumor control and diminished survival benefits (Fig. 5H and I).

Although MG132 is a potent and selective inhibitor of proteasome function, we sought to conclusively demonstrate the importance of proteasome function through use of pharmacologic stimulators. Cyclosporine A (CsA), most commonly used as an immunosuppressant, has proteasome stimulating properties that drive memory cell development (22). Using the fluorescent activity probe, we validated that CsA treatment increased proteasome function of naïve CD8 splenocytes (Fig. 6A). We next tested whether proteasome stimulation increased protein translation. CsA conditioning produced a marked reduction in p-eIF2α expression (Fig. 6B) and a significant increase in protein synthesis in T cells harvested from the tumor Transwell coculture assay (Fig. 6C). The observation that proteasome activity was heightened in memory-like T cells (Fig. 5) prompted us to assess the phenotype of CsA-conditioned OT-1 T cells. We observed that CsA generated T cells with memory traits such as high CD62 L and Bcl2 expressions coupled with reduced proliferative capacity marked by Ki-67 (Supplementary Fig. S10A–S10C). These findings were replicated using a low-affinity physiologically relevant CD8 TCR-transgenic model, the pmel-gp100 system, indicating that proteasome stimulation could be used to shape T-cell fate across a range of TCR-peptide affinities (Supplementary Fig. S10D–S10F). Combined with our above findings, these data demonstrate that proteasome activity is necessary and sufficient to protect from translation attenuation in T cells responding to TME stress.

We sought to address the viability of proteasome stimulation as an interventional strategy to improve the efficacy of adoptively transferred T cells to control solid tumors in mice. OT-1 T cells were activated with OVA peptide then conditioned in the presence of CsA or vehicle before infusion into B16–OVA-bearing mice. CsA treatment enhanced T-cell tumor control, resulting in extended animal survival (Fig. 6D and E). Next, we performed a similar ACT experiment in which tumors were harvested five days after T-cell infusion and donor T cells in tumors were assessed. CsA treatment significantly enhanced the percentage of live donor OT-1 T cells that accrued in B16-F1–OVA melanomas relative to controls (Fig. 6F). CsA treatment also augmented the frequency of IFNγ–TNFα polyfunctional OT-1 T cells in tumors (Supplementary Fig. S11). Using pmel T cells infused into mice bearing B16-F1 melanomas, we validated that CsA treatment induced a robust increase in percentage of live donor CD8 T cells accruing in tumors responding with lower affinity to a physiologically relevant melanoma antigen (Fig. 6G). Collectively, the data suggest that proteasome activity promotes CD8 TIL survival in tumors, augmenting immunity in TME stress.

Apart from primary translation of mRNA to amino acid chains, protein synthesis also requires successful folding of secondary, tertiary, and quaternary structures (46). Early studies estimated that 70% of the ATP reservoirs are required to sustain translation, making it one of the most bioenergetically taxing cellular processes (47). For T cells, antigenic stimulation activates a dynamic process highlighted by a translation rate of approximately 800,000 proteins per minute (40). Thus, it is not challenging to recognize the substantial metabolic burden on T cells, particularly in a scenario such as the TME where competition for nutrients and global cell stressors are exceptional (5). Despite this, very little is known about the underlying mechanisms required to support translation in the TME. We show here that blunted protein synthesis is a hallmark of tumor-infiltrating CD8 T cells driven by phosphorylation of eIF2α. These findings are a departure from previous studies of ER stress in two ways. First, we demonstrate that the TME represents a biologically relevant model to investigate ER stress in lieu of chemical insult. Second, to our knowledge, this is the first study to establish the antitumor capacity of T cells is directly related to their ability resolve intrinsic protein stress. Thus, these findings implicate T-cell translation attenuation as a previously unappreciated molecular checkpoint that dictates effective immunotherapy.

The metabolic demands of T cells are directly related to their differentiation state and predictive of their effector function (48). Naïve T cells rely on oxidative phosphorylation, activated effector cells are highly glycolytic, and memory T cells employ de novo fatty acid (FA) synthesis and FA β oxidation (49), whereas exhausted tumor infiltrating CD8s display poor mitochondrial health associated with depressed bioenergetic output (48). The dependence on exogenous glucose represents a significant vulnerability to tumor infiltrating T cells, and work has conclusively demonstrated that metabolic reprograming enhances cytokine production and, by extension, antitumor immunity (15, 16). Our results here elaborate on those studies, suggesting that extinguished cytokine synthesis in the TME is a side effect of global translation repression. Given the widespread interest in metabolic reprograming as an interventional approach, the extent to which these strategies relieve ER stress should be further elucidated. Indeed, the concepts uncovered can be extended to recent work highlighting systemic IL10–Fc fusion protein therapy cytokine therapy rehabilitates the terminally exhausted CD8 tumor-infiltrating population through reestablishing metabolic health, as connections between ER stress, the UPR, and exhausted T cells have not been made (50).

Stress-mediated global translation attenuation serves to restore proteostasis by enabling degradation of misfolded proteins via the ubiquitin–proteasome pathway (41, 51). As stress persists the cell compensates by increasing expression of oxidative protein folding machinery, resulting in the unintended consequence of cell reactive oxygen species generation (52, 53). Work from our group and others has illustrated that such oxidative stress is a trait of terminally exhausted T cells (10, 54); however, the concept of a misfolded protein burden in relation to T-cell exhaustion is relatively unexplored concept. Here, we found that metabolic reprograming overcomes TME-mediated translation attenuation in a manner that is dependent on proteasome-mediated protein degradation. Metabolic reprograming strategies have been shown to alleviate oxidative stress in T cells in tumors through invigoration of antioxidant systems (43). Our data indicate for the first time that the potential to protect T cells from accrual of misfolded proteins is intimately tied to engagement of antioxidant metabolism and linked by the proteasome. Although proteasome activation directly mitigates translation attenuation (55), our metabolomic data suggest that proteasome activity also fuels pathways associated with intracellular detoxification such as gluconeogeneic and glutathione metabolism (56), necessary for superior T-cell tumor control (43). Although the direct contribution of the 26S or immunoproteasome to program T-cell antitumor metabolism requires further elucidation, our data strongly suggest proteasome stimulation as a new method for enhancing immunotherapy outcomes.

Although this study illustrates a significant defect in protein translation in T cells during the initial phase of B16 melanoma growth and in multiple human tumor types, there were significant limitations in the study approach that can be rectified in future works. First, the in vitro Transwell coculture assay did not replicate persistent antigen stimulation that leads to T-cell exhaustion. The copper-based fixation required to monitor protein translation with HPG at the single-cell level, whereas innovative, makes it difficult to deeply phenotype protein synthesis in T-cell subsets based on limited availability of resistant fluorochromes. Future work should aim to simultaneously profile translation rates in T-cell subsets, the myeloid compartment, and tumor cells in vivo. These data could elucidate translation as the end-game of the metabolic tug-of-war in the TME for glucose, amino acids, and fats for bioenergetic fuel. A second substantial limitation of our in vitro approach was that it is difficult to test whether glucose (or amino acid) competition as a mechanism that impairs T-cell translation was unique to the T-cell tumor interaction. It is likely that any two proliferative, metabolically active cell types will compete for exogenous nutrients. However, as was a focus of this work, a clear picture has emerged that the battle between T and tumor cells for resources is particularly pertinent in the environment of tumors as it critically defines tumor growth or regression.

L.R. Leddy reports research support from KCI on a project not related to this work and related to wound healing after radiation in human patients. No disclosures were reported by the other authors.

B.P. Riesenberg: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. E.G. Hunt: Data curation, formal analysis, investigation, visualization, methodology. M.D. Tennant: Data curation, formal analysis, validation, investigation, visualization, methodology. K.E. Hurst: Data curation, formal analysis, validation, investigation, methodology. A.M. Andrews: Data curation, formal analysis, validation, investigation, visualization. L.R. Leddy: Resources, methodology. D.M. Neskey: Resources, methodology. E.G. Hill: Resources, formal analysis, methodology. G.O.R. Rivera: Formal analysis, writing–review and editing. C.M. Paulos: Formal analysis, methodology, writing–review and editing. P. Gao: Resources, data curation, visualization. J.E. Thaxton: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

The authors thank Dr. Zihai Li for continued mentorship over the course of development of this manuscript. Metabolomics services were performed by the Metabolomics Core Facility at Robert H. Lurie Comprehensive Cancer Center of Northwestern University. Biostatistics work is supported in part by the Biostatistics Shared Resource, Hollings Cancer Center, Medical University of South Carolina (P30 CA138313 to E.G. Hunt). Financial support from NCI/NIH is as follows: R01CA244361-01A1 and R01CA248359-01 (to J.E. Thaxton), T32 5T32AI132164-04 (to B.P. Riesenberg), T32 DE01755 (to M.D. Tennant), and T32 CA 193201 (to A.M. Andrews). The authors are grateful to Drs. D. Guttridge and A. Baldwin at the Medical University of South Carolina and University of North Carolina at Chapel Hill, respectively, for providing thoughtful guidance on article development and submission.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.