NEDDylation Regulates CD8⁺ T-cell Metabolism and Antitumor Immunity Open Access

Posttranslational modifications (PTM) are central regulators of immune responses in the tumor microenvironment (1). The most extensively studied PTMs include those that trigger signaling pathways [e.g., phosphorylation cascades triggered by T-cell receptor (TCR) engagement] and those that maintain protein homeostasis, such as ubiquitination (2, 3). The latter refers to the conjugation of ubiquitin units to substrate proteins, whereas similar PTMs, including NEDDylation and SUMOylation, utilize other substrates called ubiquitin-like proteins to modify target proteins. Given that PTMs are recognized as key regulators of T-cell function (4), we aimed to study the role of NEDDylation in CD8⁺ T cell–mediated antitumor immunity.

NEDDylation involves the covalent attachment of the ubiquitin-like protein NEDD8 to target proteins (5). This reaction is mediated by the formation of an isopeptide bond between the carboxy-terminal glycine of NEDD8 (Gly76) and a lysine residue on the target protein (6). NEDDylation is regulated by NEDD8-activating enzyme (NAE), which is comprised of two subunits: an E1 catalytic subunit referred to as UBA3 and an E1 regulatory subunit identified as NAE1. NAE is responsible for activating NEDD8, which can then be transferred to an E2 conjugating enzyme, such as NEDD8-conjugating enzyme UBC12 (UBE2M) or NEDD8-conjugating enzyme UBE2F, which load NEDD8 into an E3 ligase complex (7). The most described targets of NEDD8 are cullin–RING ligases (CRL), whose primary role is to tag ubiquitin to specific proteins for degradation by the proteasome. This process is regulated by NEDDylation, because NEDD8 ligation to CRLs leads to a conformational change that activates the CRL complex (8, 9). Finally, deNEDDylases, such as SENP8 (also called NEDP1), are responsible for removing NEDD8 from NEDDylated substrates in a dynamic and tightly regulated cycle (10).

In addition to CRLs, NEDDylation can act on alternative protein substrates. Notably, NEDDylation can reduce the activity of the tumor suppressor P53 and promote the degradation of EGFR (11, 12). Conversely, NEDD8 attachment can increase the stability of certain proteins, such as TGF-β receptor 2 (13). Therefore, NEDDylation effects are variable and substrate-dependent (14), including protein stabilization or degradation, intracellular cell trafficking, modulation of enzymatic activity, or regulation of DNA damage response and metabolism (15–18). As such, uncovering additional NEDDylation targets beyond CRLs is critical if we are to fully understand its role in health and disease.

In the context of cancer, the NEDDylation pathway has been described to be upregulated in several tumor types (19–22), and it is considered a poor prognostic marker (23). The dysregulation of NEDDylation in cancer cells presents an opportunity for the development of cancer therapies (24). MLN4924 (pevonedistat) is a first-in-class small-molecule inhibitor of the E1-activating enzyme NAE, which blocks NEDDylation, disrupting protein turnover, reducing angiogenesis and proliferation, and inducing apoptotic death of tumor cells (25, 26). NAE is an adequate target for therapies that only aim to regulate NEDDylation because it exclusively interacts with NEDD8, whereas other enzymes of the NEDDylation cycle can also use ubiquitin as substrate. Currently, MLN4924 is being evaluated in clinical trials for the treatment of various cancer types (27, 28). However, the phase III clinical trial evaluating the efficacy of MLN4924 for the treatment of myelodysplastic syndromes or acute myeloid leukemia failed in 2022 to meet the primary endpoint (29). Similarly, the NAE inhibitor TAS446 also failed to complete phase I clinical trial testing against hematologic cancers (ClinicalTrials.gov identifier: NCT02978235). Given the relevance of the immune system in these pathologies (30–32), elucidating the role of NEDDylation on immune cell function is of relevance, but few studies have looked into it (33, 34).

Within current knowledge, NEDDylation regulates macrophage function in the context of responses to checkpoint receptor blockade (35). Inhibition of NEDDylation suppresses the function of dendritic cells, limiting inflammatory responses (36), and reduces the activity of CD4⁺ T cells against blood-stage Plasmodium infection. The suppression of NEDDylation is also being explored as a method to alleviate COVID-19–associated inflammation (37) and multiple sclerosis (38). These insights reveal that modulating NEDDylation could be key in devising new treatments for immune-mediated conditions.

The precise role of NEDDylation on regulating the function of CD8⁺ T-cell subsets, such as effector or exhausted T cells, is not fully understood. In this study, we report that the use of a combination of in vitro, in vivo, and computational approaches showed that NEDDylation is necessary for CD8⁺ T-cell proliferation and differentiation into effector cells. Furthermore, abrogating NEDDylation by both genetic deletion and pharmacologic inhibition of NAE1, as well as increasing NEDDylation by the deletion of deNEDDylase SENP8 in CD8⁺ T cells, uncovered its critical role in regulating CD8⁺ T-cell fate, metabolism, and antitumor function.

C57/BL6 female mice were purchased from Charles River Laboratories, whereas dLCK-Cre mice (C57/BL6 background) were purchased from The Jackson Laboratory (strain #:012837). NAE1^fl/fl mice (C57/BL6 background) were kindly provided by Dr. Ashwin Whoodoo (University of Santiago de Compostela, Spain; ref. 39). dLCK-Cre and NAE1^fl/fl mice were crossed to obtain NAE1 knockout (KO) mice. Cre negative littermates were used as controls. OT-I mice were obtained from The Jackson Laboratory (strain #:003831, C57/BL6 background) and were crossed with NAE1-KO mice to generate NAE1-KO-OT-I mice, which were used for antigen-specific assays. Animal procedures were performed following the ethical guidelines established by the Biosafety and Welfare Committee at CIC bioGUNE and the recommendations from Association for Assessment and Accreditation of Laboratory Animal Care International. These experiments have been approved by the Committee on Bioethics and Animal Welfare (CBBA) and by the government of Bizkaia (Spain) under license numbers “P-CBG-CBBA-0319” and “P-CBG-CBBA-521.”

For the isolation of mouse CD8⁺ T cells, individual spleens from NAE1-KO and NAE1-KO-OT-I mice were homogenized using a syringe plunger in 20 mL of RPMI 1640 Glutamax medium (Gibco, cat. #: 11875093), and a 70 µm strainer was used for further dissociation. The single-cell suspension was centrifuged (5 minutes at 600 g, 4°C). Red blood cells were lysed by resuspending the pellet in 3 mL of ammonium-chloride-potassium (ACK) lysis buffer (Thermo Fisher Scientific, cat. #: A1049201). After 3 minutes, 10 mL of RPMI 1640 Glutamax medium were added, and the cell suspension was centrifuged (5 minutes at 600 g, at 4°C). Splenocytes were resuspended in RoboSep buffer (Stemcell Technologies, cat. #: 20104) and counted. CD8⁺ T cells were magnetically isolated using EasySep Mouse CD8⁺ T cell isolation Kit (Stemcell Technologies, cat. #: 19853A). Isolated CD8⁺ T cells were activated using plate-bound anti-CD3 (5 μg/mL, Thermo Fisher Scientific, cat. #: 16-0031-85) and soluble anti-CD28 (1 μg/mL, Thermo Fisher Scientific, cat. #: 16-0281-85). Cell culture medium consisted of 1640 RPMI Glutamax medium supplemented with β-mercaptoethanol (50 μmol/L, Thermo Fisher Scientific, cat. #: 31350010) and HEPES (25 mmol/L, Thermo Fisher Scientific, cat. #: 15630056). CD8⁺ T-cell culture density was 1.5 × 10⁶ cells per well (24-well plates) or 0.2 × 10⁶ per well (96-well plates) at a concentration of 10⁶ cells/mL. DMSO was added to control wells. After 3 days of culture, CD8⁺ T cells were washed three times and reseeded with cell culture medium supplemented with IL2 (20 ng/mL, BioLegend cat. #: 575404). When indicated, NEDDylation inhibitor MLN4924 (MedChemExpress, cat. #: HY-70062) was added at 0.2 μmol/L or 0.4 μmol/L concentrations. When mentioned, galactose (10 mmol/L, Sigma-Aldrich, cat. #: 5388-100G) was added to no-glucose RPMI medium (Gibco cat. #: 11879020). For the analysis of thymocytes, thymi were harvested and homogenized with a syringe plunger in RPMI 1640 Glutamax medium. For the analysis of inguinal lymph nodes, these were homogenized with a syringe plunger in RPMI 1640 Glutamax medium. Cells were directly stained and analyzed by flow cytometry.

For the isolation of human CD8⁺ T cells, buffy coats were provided by the Basque Biobank (www.biobancovasco.org) and processed following standard operation procedures with appropriate approval of ethical and scientific committees (code CEIC E19-75). Human peripheral blood mononuclear cells were isolated from buffy coats using Ficoll-Paque Plus (Cytiva, cat. #: 17144003) density centrifugation at 1,200 g for 10 minutes at room temperature (RT) using SepMat-50 tubes (Stemcell Technologies, cat. #: 85450). CD8⁺ T cells were magnetically isolated using EasySep Human CD8⁺ T Cell Isolation Kit (Stemcell Technologies, cat. #: 17953RF) in an automated RoboSep cell separator (Stemcell Technologies) according to the manufacturer’s indications. Human cD8⁺ T cells were stimulated on culture plates coated with anti-CD3 (5 μg/mL, BD Biosciences, cat. #: 567118) and soluble anti-CD28 (1 μg/mL, Thermo Fisher Scientific, cat. #: 16-0289-81) in OpTmizer CTS medium (Thermo Fisher Scientific, cat. #: A1048501) supplemented with CTS Immune Cell Serum Replacement (Thermo Fisher Scientific, cat. #: A2596101). Cells were also stimulated with phorbol 12-myristate-13-acetate (PMA, 200 ng/mL, Sigma-Aldrich, cat. #: P8139) and ionomycin (200 ng/mL, Sigma-Aldrich, cat. #: I0634). Cells were seeded at a density of 10⁶ cells/mL of medium. DMSO was added in control wells. After 3 days, CD8⁺ T cells were washed three times and refreshed with new media. When indicated, MLN4924 was added at 0.2 or 0.4 μmol/L concentrations.

LLC (cat. #: CRL-1642; 2019) and B16 cells (cat. #: CRL-6475; 2019) were purchased from the ATCC, cultured in DMEM (Gibco cat. #: 41966029), and passaged every 2 days. The B16-OVA cell line was a gift from Dr. Ignacio Melero (Centro de Investigación Médica Aplicada, University of Navarra, Spain) and passaged every 2 days. The rest of the ovalbumin (OVA) and GFP luciferase-expressing cell lines were generated through lentiviral infection. HEK 293 cells (Takara Bio Inc., cat. #: 632180; 2020) were cultured in DMEM (Gibco, cat. #: 41966029) and passaged every 2 days. YUMMER 1.7-H2B-GFP5 cells were purchased from Sigma-Aldrich (cat. #: SCC245; 2024), cultured in DMEM-F12 (Gibco cat. #: 11320033) plus nonessential amino acids (Gibco, cat. #: 11140035), and passaged every 3 days. RAMOS cells were generously provided by Dr. Francisco Borrego (Biobizkaia Health Research Institute, Spain), cultured in RPMI 1640 Glutamax medium, and passaged every 2 days. Jurkat (clone E6-1) cells (cat. #: TIB-152, 2019) were purchased from the ATCC, cultured in RPMI 1640 Glutamax medium, and passaged every 2 days. The JE6.1 triple parameter reporter (Jurkat TPR) cell line was generously provided by Dr. Peter Steinberger (Medical University of Vienna, Austria), cultured in RPMI 1640 Glutamax medium, and passaged every 2 days. Jurkat TPR cells are genetically modified to encode fluorescent proteins under the control of response elements for transcription factors like NF-κB (eCFP proteins), NFAT (eGFP proteins), or AP-1 (mCherry proteins). All cell lines were cultured in media supplemented with 10% FBS (Thermo Fisher Scientific, cat. #: 10270106) and 1% penicillin–streptomycin (Thermo Fisher Scientific, cat. #: 15140122) and tested for Mycoplasma using MycoAlert Mycoplasma Detection Kit (Lonza, cat. #: LT27-221) following the manufacturer’s instructions. Microscopic examination to check cell shape, size, and growth patterns was performed daily. Cells were maintained in an incubator at 37°C in a 5% CO₂ atmosphere and cultured for a maximum of 10 passages.

Plasmid encoding OVA was a gift from Maria Castro (University of Michigan, United States, Addgene plasmid # 25097; http://n2t.net/addgene:25097; RRID: Addgene_25097), and the OVA region was cloned into MSCV-pLV lentiviral plasmid by GenScript Biotech (resulting plasmid: cOVA-puro-MSCV-pLV).

For CAR T-cell generation, we used CAR constructs which included the CD8α signaling peptide (UniProt entry: P01732), the anti-CD19 scFv derived from the FMC63 mAb (40), the CD8α hinge/transmembrane domain (UniProt entry: P01732), the 4-1BB costimulatory domain (UniProt entry: Q07011), the CD3ζ intracellular domain (UniProt entry: P20963-3), and a truncated form of the nerve growth factor receptor (NGFRt) as a selection marker (UnipPot entry: P08138). NGFRt was expressed separately to the CAR via a T2A self-cleaving peptide. This construct was cloned into the third-generation lentiviral vector pCCL (41), under the control of the EF1a constitutive promoter and synthesized by Genscript, resulting in the plasmid FMC63-BBz-NGFRt pCCL (item no. U2706FF040-10). For SENP8-KO plasmid production, plasmid encoding “lentiCRISPR v2” was a gift from Feng Zhang (Broad Institute, United States; Addgene plasmid cat. #: 52961). The guide RNAs against human AAVS1 (safe harbor gene used as control: GGG​GCC​ACT​AGG​GAC​AGG​AT) and SENP8 (AAC​TCA​GTT​CAC​GCA​AAG​C; ref. 42) were cloned into this vector by GenScript Biotech, resulting in plasmids “hAAVS1 KO_(ctrl)_LentiCRISPRv2” and “NEDP1_2_GFP_PURO_Lenticrisprv2,” respectively. GFP as a visualization marker and the puromycin resistance gene as a selection marker were also included in the plasmid.

The heat shock method was used to incorporate plasmids into 15 μL of One Shot TOP10 (Thermo Fisher Scientific, cat. #: C404003) bacteria. Briefly, 0.3 μg of plasmid was added to 15 μL of bacteria and left 20 minutes on ice. After this, cells were heated to 42°C for 1.5 minutes and then incubated on ice for an additional 5 minutes. Bacteria were grown overnight for 16 hours in agar plates with ampicillin (100 μg/mL, Sigma-Aldrich, cat. #: A9518) as the selection antibiotic. The selected colony was used to inoculate a starter culture, which was grown overnight for 16 hours. Subsequently, the starter culture was used to inoculate a larger culture for maxiprep during additional 16 hours. Then, DNA was isolated using HiPure PureLink Maxiprep Isolation Kit (Thermo Fisher Scientific, cat. #: K210007).

To generate lentiviral particles, 5 × 10⁶ HEK 293T cells were seeded in 10-cm plates and transfected on the next day using jetPEI Kit (Polyplus Transfection, cat. #: 101-10N) using packaging plasmids psPAX2 (4 μg, a gift from Didier Trono, École Polytechnique Fédérale de Lausanne, Switzerland, Addgene plasmid cat. #: 12260), and VSV-G (1.5 μg, a gift from Tannishtha Reya, Columbia University, United States, Addgene plasmid cat. #: 14888). For the rest of the plasmids, 5 μg per transfection plate was used. Viral particles were harvested from the supernatant 48 hours after transfection, centrifuged, and filtered through a 0.45-μm filter (VWR cat. #: 514-0063).

Virus used for T-cell transductions was concentrated using a Lenti-X concentrator (Takara Bio, cat. #: 631232) and titrated on Jurkat cells. Briefly, 5 × 10⁴ Jurkat cells were seeded in 96-well plates with 8 μg/mL of polybrene (Sigma-Aldrich, cat. #: TR-1003). A serial dilution of the used virus was added to the wells. Each dilution was assayed in duplicate. After 2 days, the efficiency of infection was analyzed by flow cytometry. Only the range of 5% to 20% of infection was used to quantify the amount of active viruses per μL.

For the generation of stable cell lines, LLC, B16, and B16-OVA cells were seeded in a 6-well plate, and 24 hours later, nonconcentrated lentiviral particles were added in the presence of polybrene at 5 µg/mL for LLC and 10 µg/mL for B16 and B16-OVA. After 2 days, the expression of the desired antigen was checked by flow cytometry. For luciferase-expressing cells (B16-GFP-luciferase and B16-GFP-luciferase-OVA), transduced cells were sorted using a BD FACSAria Fusion sorter (BD Biosciences, purity >95%). For OVA-expressing LLC cells (LLC-OVA), puromycin selection was performed (2 µg/mL, Thermo Fisher Scientific, cat. #: A1113803).

Human CD8⁺ T cells isolated from healthy blood donors were activated with anti-CD3 and anti-CD28 Dynabeads (1:1 ratio, Thermo Fisher Scientific, cat. #: 11132D) at 10⁶ cells/mL with the mentioned human T-cell medium supplemented with IL2 (100 IU/mL, Miltenyi Biotech, cat. #: 130-097-746). On day 1 of activation, 0.5 × 10⁶ cells were seeded in 24-well plates and transduced with concentrated lentivirus. First, hAAVS1-KO (control) or SENP8-KO viral infection was performed at a multiplicity of infection of 6. The medium mentioned above contained polybrene at 8 μg/mL. Spinoculation was performed (2,000 RPM for 1 hour at 32°C). At day 3 of activation, transduced cells were selected with puromycin (1 μg/mL) for 3 days. Dynabeads were removed after 6 days. Every 2 days, IL2 was replenished with fresh media and cells were left at 0.5 × 10⁶ cells/mL. After 15 days of culture, cells were reactivated with anti-CD3 and anti-CD28 Dynabeads following the same protocol as stated before. After 1 day of activation, CD8⁺ T cells were infected with anti-CD19 CAR virus at a multiplicity of infection of 3. Cells were expanded for 15 days, after which they were purified using the NGFR positive selection kit (Stemcell Technologies, cat. #: 17849), obtaining purities greater than 80%. After 2 days, SENP8-KO anti-CD19 CAR T cells were used for the cytotoxicity assay.

For the mouse OT-I–based cytotoxicity assay, OT-I CD8⁺ T cells from control or NAE1-KO OT-I mice were magnetically isolated from the spleen and were activated in vitro for 3 days. At day 2, 2 × 10⁴ B16-luciferase-OVA cells were seeded in a white 96-well plate (PerkinElmer, cat. #: 6005680). At day 3, activated OT-I CD8⁺ T cells were washed and added to the culture at the indicated effector cell:T cell ratios for 24 hours.

For the human SENP8-KO anti-CD19 CAR T cell cytotoxicity assay, 2.5 × 10⁴ RAMOS-luciferase cells were seeded in the same plates as the previous experiments. On the same day, hAAVS1 or SENP8-KO anti-CD19 CAR T cells were added to each well at the indicated effector cell:T cell ratios for 24 hours.

In both assays, at endpoint 50 μL of D-Luciferin (3 mg/mL, PerkinElmer, cat. #: 122799) were added to each well, and the luminescence signal corresponding to alive cells was measured using a multimode plate reader (PerkinElmer Victor Nivo).

For the in vitro T-cell exhaustion model, we followed the protocol previously established by Scharping and colleagues (43). Briefly, splenic CD8⁺ T cells were isolated from control and NAE1-KO mice. Cells were activated at 1 × 10⁶ cells/mL using anti-CD3 and anti-CD28 Dynabeads in a 1:1 ratio, in RPMI 1640 Glutamax medium supplemented with 10% FBS, 1% penicillin–streptomycin, 25 ng/mL IL2, and 10 ng/mL IL12 (Miltenyi Biotech, cat. #: 130-096-708). After 24 hours (day 1), beads were magnetically removed, and the cells were divided into two conditions: acute activation (without Dynabeads) and chronic activation (by readding Dynabeads). Fresh media supplemented with 25 ng/mL of IL2 was added. On days 3 and 5, the cells were split in half with fresh IL2 supplemented media. Finally, on day 6, the cells were analyzed.

IFNγ levels in cell culture supernatants were quantified after all the stimulations of the T-cell exhaustion protocol using a sandwich ELISA kit according to the manufacturer’s protocol (R&D Systems, cat. #: DY485).

The activity of lactate dehydrogenase (LDH) in cell culture supernatants was quantified on day 3 of activation according to the manufacturer’s colorimetric protocol (Sigma-Aldrich, cat. #: MAK066).

Single-cell suspensions were seeded in V-shaped 96-well plates (Thermo Fisher Scientific, cat. #: 249570). If cells were to be fixed, they were first stained with fixable LIVE/DEAD viability dyes for 30 minutes at 4°C protected from light. Then, cells were incubated with TruStain FcX blocker for 10 minutes at RT protected from light. After this, cells were stained with fluorochrome-conjugated antibodies in staining buffer (Thermo Fisher Scientific, cat. #: 00-4222-26) for 30 minutes at 4°C in the dark. After staining, if cells were to be fixed, they were resuspended in paraformaldehyde at 0.5% and acquired no more than 5 days later in a BD FACSymphony A3 flow cytometer. If cells were not fixed, 4′,6-diamidino-2-phenylindole (DAPI,1:10,000 dilution, Thermo Fisher Scientific, cat. #: D1306) was added as a viability marker. Analysis was performed using FlowJo v10 or Cytobank Premium software. All centrifugation steps were performed at 450 g for 5 minutes at 4°C. The antibodies used are listed in Supplementary Table S1. Specific staining procedures are detailed in the corresponding sections of “Materials and Methods.”

Absolute cell numbers of mouse or human CD8⁺ T cells were obtained by using CountBright absolute counting beads (Thermo Fisher Scientific, cat. #: C36995) after acquisition in a BD FACSymphony A3 flow cytometer.

For the carboxyfluorescein diacetate succinimidyl ester (CFSE) dilution assay of mouse CD8⁺ T cells, 10⁶ CD8⁺ T cells were resuspended in RPMI medium supplemented with FBS and P/S and were incubated with 5 μmol/L CFSE (Thermo Fisher Scientific, cat. #: C34554) for 5 minutes at RT in the dark. Cells were washed three times and 3.5 × 10⁵ CD8⁺ T cells were activated with plate-bound anti-CD3 and soluble anti-CD28 in round-shaped 96-well plates. At day 3, cells were acquired in a BD FACSymphony A3 flow cytometer.

For the CFSE dilution assay of human CD8⁺ T cells, 10⁶ CD8⁺ T cells were resuspended in PBS and incubated with 5 μmol/L CFSE for 7 minutes at 37°C in the dark. Five mL of ice-cold RPMI were added to the cells, and the cells were incubated for 5 minutes and washed twice. Then, 1.75 × 10⁵ CD8⁺ T cells were activated with plate-bound anti-CD3 and soluble anti-CD28 in a round-shaped 96-well plate and acquired in a BD FACSymphony A3 flow cytometer at day 4 of activation.

For the annexin V viability assay, activated mouse CD8⁺ T cells were harvested after 3 days of culture and stained with 3 µL of annexin V-BV421 in 100 µL of binding buffer (BD Biosciences, cat. #: 556454). After incubating for 5 minutes in the dark, 2 µL of 7-aminoactinomycin D (7AAD) in 100 µL of binding buffer were added to samples, which were kept in ice until acquisition in a BD FACSymphony A3 flow cytometer. For human CD8⁺ T cells, DAPI-negative cells were considered alive cells. Data analysis was performed with FlowJo v10 or Cytobank software programs.

Total cell lysates were collected using lysis buffer (50 mmol/L Tris, pH 8.0, 150 mmol/L NaCl, 1% Triton X-100, and 2 mmol/L EDTA) supplemented with protease inhibitors (cOmplete EDTA-free tablets, EASYPack, Roche, cat. #: 04693132001), phosphatase inhibitors (PhosSTOP EASYpack, Roche, cat. #: 04906837001), phenylmethylsulfonyl fluoride (Cell Signaling Technologies, cat. #: 8553), and deNEDDylase inhibitor 2-iodoacetamide (2 mg/mL, Sigma-Aldrich, cat. # 8047440025). Protein quantification was performed using Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, cat. #: 23227). Samples were mixed with LDS sample buffer (Thermo Fisher Scientific, cat. #: NP0007) containing ditiothreitol (DTT; Melford, cat. #: MB1015), heated for 10 minutes at 95°C, and resolved in a 7.5% or 4% to 15% Mini-PROTEAN TGX precast protein gels (Bio-Rad, cat. #: 4561023 and cat. #: 4561085, respectively) with 1X Tris/glycine/SDS electrophoresis buffer (Bio-Rad, cat. #: 1704156) and PageRuler Plus Prestained Protein Ladder (Thermo Fisher Scientific, cat. #: 26619). Protein transfer to 0.2 µm PVDF membranes (Bio-Rad, cat. #: 1704156) was carried out in a Trans-Blot Turbo transfer system (Bio-Rad). Membranes were blocked for 5 minutes in EveryBlot Blocking Buffer (Bio-Rad, cat. #:12010020), washed in T-PBS (0.5% Tween-20, Sigma-Aldrich, cat. #: P2287), and incubated overnight with primary antibodies (Supplementary Table S1). The following day, membranes were washed and incubated with the corresponding secondary HRP-conjugated antibodies for 1 hour (Supplementary Table S1). Chemiluminescence detection was performed using Clarity Max Western ECL Substrate (Bio-Rad, cat. #: 170506) in an iBright CL1500 system (Thermo Fisher Scientific). Densitometric quantifications were performed using ImageJ–Fiji software. When a membrane was reprobed with an additional antibody, it was incubated overnight and developed immediately after probing with the specific antibody, following the same methodology as previously described.

Total RNA was isolated from CD8⁺ T cells using NucleoSpin RNA Kit (Macherey-Nagel, cat. #: 740955.250). A total of 325 ng of RNA were used to synthesize the cDNA with the M-MLV reverse transcriptase (Thermo Fisher Scientific, cat. #: 28025-013) and random oligos (Thermo Fisher Scientific, cat. #: 58875). qPCR reactions were performed in triplicate in a ViiA 7 Real-Time PCR system (Thermo Fisher Scientific) from 1 μL of cDNA using PerfeCTa SYBR Green SuperMix Reagent (Quantabio cat. #: 95056-500) and gene-specific primers (Supplementary Table S1). The 2^−ΔΔCt method was used to calculate the relative gene expression considering the 18S or Rplp0 genes as housekeeping genes for human and mouse species, respectively.

For immunofluorescence, subcutaneous LLC tumors were harvested and flash-frozen in optimal-cutting-temperature medium at −80°C. Sections of 8 μm were made with a Leica CM 1860 UV cryostat and fixed with 4% paraformaldehyde (Thermo Fisher Scientific, cat. #: J61899.AP) for 10 minutes. Slides were washed twice with PBS and permeabilized with 0.1% Triton X-100 (Sigma-Aldrich, cat. #: T9284) for 10 minutes. After washing, tissue slides were incubated with 2.5% normal goat blocking serum (Vector Laboratories, cat. #: S-1012-50) for 30 minutes at RT to reduce nonspecific staining. Serum was washed, and samples were incubated at 4°C overnight with primary antibodies against NEDD8, CD3, or Ki-67 (Supplementary Table S1). The next day, slides were washed five times with PBS and stained for 45 minutes with secondary antibodies at RT (goat anti-rat Alexa Fluor 488 and goat anti-rabbit Alexa Fluor 647; Supplementary Table S1). Slides were washed again, stained with DAPI for 10 minutes at RT, and mounted with Prolong Gold Antifade reagent (Thermo Fisher Scientific, cat. #: P36930). Images were obtained using a confocal microscope (Leica SP8 Lightning, HC PL APO objective lens with 100X magnification) and analyzed using LasX software.

Human CD8⁺ T cells were activated and cultured with MLN4924 at the indicated doses as stated in the Cell Culture section. Mouse control and NAE1-KO mouse CD8⁺ T cells were activated and cultured as stated in the “Culture of mouse and human primary CD8⁺ T cells” section of this article. At day 3, cells were washed and resuspended in Seahorse XF RPMI medium (Agilent, cat. #: 103576-100) supplemented with 10 mmol/L glucose (Agilent, cat. #: 103577-100), 1 mmol/L sodium pyruvate (Agilent, cat. #: 103578-100), 2 mmol/L L-glutamine (Gibco, cat. # 25030-024), and MLN4924 at the indicated doses. Seahorse XF24 Cell Culture Microplates were coated with poly-L-lysine (50 μg/mL, Sigma-Aldrich, cat. #: P8920) for 2 hours. A total of 10⁵ or 2 × 10⁵ CD8⁺ T cells were seeded per well for the Seahorse XF Glycolytic Rate Assay (Agilent, cat. #: 103344-100) and Seahorse XF Cell Mito Stress Test, respectively. After centrifuging at 400 g for 5 minutes, cells were left 15 minutes resting at 37°C in a non-CO₂ incubator. A measure of 400 μL of supplemented medium were very carefully added to each well. For the Seahorse XF Glycolytic Rate Assay, rotenone and antimycin A (0.5 μmol/L) and 2-deoxy-D-glucose (2-DG, 500 mmol/L) were used. Basal and compensatory glycolysis calculations were made through the Seahorse Analytics software. For the Seahorse XF Cell Mito Stress Test, the following drugs were used: oligomycin (1.5 μmol/L, Sigma-Aldrich, cat. #: O4876), BAM15 (2.5 μmol/L, MedChemExpress, cat. #: HY-110284), and rotenone (0.5 μmol/L, Merck, cat. #: 557368) and antimycin A (0.5 μmol/L, Sigma-Aldrich, cat. #: A8674). Basal, maximal respiration and spare respiratory capacity were calculated through the Seahorse Analytics software. Both assays were run in the Seahorse XF24 Extracellular Flux Analyzer (Agilent Technologies).

For each immunoprecipitation (IP) reaction, 1.5 mg of protein lysate and 10 μg of antibody were used (anti-NEDD8 or IgG Isotype control, listed in Supplementary Table S1). Human CD8⁺ T cells isolated from buffy coats from healthy donors were activated in vitro. Various healthy donors were used in each IP assay, and no donors were repeated in each assay. Activated CD8⁺ T cells were lysed according to the protocol stated in the “Western blot” section of this article. Antibodies were coupled to Protein G magnetic Sepharose beads (50 μL, Cytiva, cat. #: 28951379) for 1 hour at 4°C. After this, antibodies were cross-linked to the beads with BS3 compound (Thermo Fisher Scientific, cat. #: 21580) according to the manufacturer’s protocol. Protein lysates were precleaned with Protein G magnetic Sepharose beads coupled to 5 μg of IgG antibody (not cross-linked) during 1 hour at 4°C. Following this, lysates were incubated with the cross-linked bead–antibody complex overnight at 4°C. Subsequently, beads were washed six times, eluted in 30 μL of LDS sample buffer (Thermo Fisher Scientific, cat. #: NP0007) containing DTT (Melford, cat. #: MB1015), and boiled at 95°C for 10 minutes. Eluates were analyzed by Western blotting or by mass spectrometry (MS). In the case of MS analysis, lysis buffer did not include 2-Iodoacetamide.

Immunoprecipitated protein samples were boiled in a buffer containing 50 mmol/L Tris, pH 6.8, 5% glycerol, 1.67% β-mercaptoethanol, 1.67% SDS, and 0.0062% bromophenol blue for 5 minutes and resolved in 4% to 12% gradient acrylamide gels (NuPAGE Bis-Tris gels 1.0 × 12 well, Thermo Fisher Scientific) using an XCell SureLock Mini-Cell Electrophoresis System (Thermo Fisher Scientific). A constant voltage of 150 V was applied for 45 minutes. Gels were fixed in a solution containing 10% acetic acid and 40% ethanol for 30 minutes and stained overnight in SYPRO Ruby (Bio-Rad). Gels were then washed in a solution containing 10% ethanol and 7% acetic acid for 30 minutes, and the image was acquired using a Chemidoc gel imager system (Bio-Rad). Each lane was subjected to tryptic digestion. Gel bands were washed in Milli-Q water. Reduction and alkylation were performed using 10 mmol/L DTT in 50 mmol/L ammonium bicarbonate at 56°C for 20 minutes, followed by iodoacetamide (50 mmol/L iodoacetamide in 50 mmol/L ammonium bicarbonate) for another 20 minutes in the dark. Gel pieces were dried and incubated with trypsin (12.5 μg/mL in 50 mmol/L ammonium bicarbonate) for 20 minutes on ice. After rehydration, the trypsin supernatant was discarded. Gel pieces were hydrated with 50 mmol/L ammonium bicarbonate and incubated overnight at 37°C. After digestion, acidic peptides were cleaned with TFA 0.1% and dried out in a RVC2 25 speedvac concentrator (Christ). Peptides were resuspended in 10 μL 0.1% FA and sonicated for 5 minutes prior to LC-MS/MS analysis.

Activated mouse CD8⁺ T cells were sorted to isolate alive cells using a BD FACS Aria fusion sorter. Total protein from activated alive mouse CD8⁺ T cells was extracted by incubating the extracts in a buffer containing 7 mol/L urea, 2 mol/L thiourea, and 4% CHAPS. Samples were incubated in this buffer for 30 minutes at RT with agitation and digested following the FASP protocol described by Wisniewski and colleagues (44) in 2009 with minor modifications. Trypsin was added in 50 mmol/L ammonium bicarbonate to a trypsin:protein ratio of 1:10, and the mixture was incubated for overnight at 37°C. Peptides were dried out in an RVC2 25 speedvac concentrator (Christ) and resuspended in 0.1% FA. Peptides were desalted and resuspended in 0.1% FA using C18 stage tips (Millipore) prior to acquisition.

Samples were analyzed in a timsTOF Pro with PASEF (Bruker Daltonics) coupled online to an Evosep ONE liquid chromatograph (Evosep). timsTOF was operated in data-dependent acquisition mode, and the data were acquired using the default data-dependent acquisition PASEF-standard_1.1sec_cycletime method provided by Bruker. Data were acquired in positive mode. Scan range was established between 100 and 1700 m/z, and 1/K0 was between 0.60 and 1.60 V·s/cm³, with a total ramp and accumulation time of 100 ms (100% duty cycle). Collison energies applied were 20 and 59 eV for the lower and upper limits of the mobility range, respectively. Charge states between 0 and 5 were specifically selected for analysis. Total PASEF ramps were 10, to a final cycle time of 1.16 seconds.

A total of 200 ng were directly loaded onto the Evosep ONE and resolved using the default 30 samples-per-day protocol (44 minutes gradient flow and 48 minutes total cycle time). The analytic column was equilibrated at 1,500 nL/minute, and the gradient flow was 500 nL/minute, increased to 1,500 nL/minute for washing. Solvents A and B were water with 0.1% formic acid (FA) and acetonitrile with 0.1% FA, respectively.

Protein identification and quantification was carried out using PEAKS X software (Bioinformatics solutions) or MaxQuant. Searches were carried out against a database consisting of Mus musuculus entries from UniProt Swiss-Prot. Precursor and fragment tolerances of 20 ppm and 0.05 Da were considered for the searches, respectively. Only proteins identified with at least one peptide at FDR < 1% were considered for further analysis. Total protein extracts were compared by means of a Student t test. Protein abundances were compared between control and immunoprecipitated samples. Immunoprecipitated proteins with at least a twofold difference were kept for further analyses. For Gene Ontology Biological Processes selection, the list of immunoprecipitated proteins was examined using the website software (https://geneontology.org/), and the most enriched processes with at least 10 proteins associated with each process were selected.

Ingenuity pathway analysis (IPA, QIAGEN) was used for a characterization of the molecular events lying behind the differential protein patterns under analysis. The calculated P values for the different analyses performed determine the probability that the association between proteins in the dataset and a given process, pathway, or upstream regulator is explained by chance alone, based on a Fisher exact test (P value < 0.05 being considered significant). The activation z-score represents the bias in gene regulation that predicts whether the upstream regulator exists in an activated (positive values) or inactivated (negative values) state, based on the knowledge of the relation between the effectors and their target molecules. For published figures based on these data, the list of proteins associated with each pathway was provided by IPA software.

For the immunophenotyping of tumor infiltrates, 5 × 10⁵ LLC cells or 10⁶ YUMMER 1.7-GFP cells were subcutaneously injected in 100 μL of PBS or saline solution, respectively, in the right flank of 6- to 10-week-old control and NAE1-KO mice. Tumor growth was monitored every other day using a digital caliper. All mice were sacrificed when the tumor diameter of one mouse reached 15 mm, and single-cell suspensions from spleens, tumors, and tumor-draining lymph nodes were prepared for flow cytometry staining and analysis. Spleens and lymph nodes were dissociated using a syringe, whereas mouse tumor samples were finely minced with a scalpel blade in a Petri dish. Tumors were incubated for 30 minutes at 37°C with type I collagenase (0.5 mg/mL, Sigma-Aldrich, cat. #: C0130) in DMEM. Cell suspensions were filtered through a 70-μm cell strainer. Red blood cell lysis of spleen and tumor samples was performed using ACK buffer (Thermo Fisher Scientific, cat. #: A1049201) for 3 minutes at RT. After counting, 2 × 10⁶ cells from each tumor and spleen were stained with a LIVE/DEAD fixable blue or green dead cell stain kit for 30 minutes at 4°C protected from light. Then, cells were washed in PBS and incubated with TruStain FcX blocker for 10 minutes at RT, protected from light. After this, cells were stained with fluorochrome-conjugated antibodies in staining buffer for 30 minutes at 4°C in the dark. The antibodies used are listed in Supplementary Table S1. After staining, cells were washed and fixed with paraformaldehyde 0.5% and acquired using a BD FACSymphony A3 flow cytometer. Data analysis was performed with FlowJo v10 or Cytobank Premium software programs. Absolute cell numbers were calculated using CountBright absolute counting beads. All centrifugation steps were performed at 450 g for 5 minutes at 4°C.

For the measurement of transcription factors, tumor cell suspensions were stained following the manufacturer’s protocol of the intracellular transcription factor staining kit (Thermo Fisher Scientific, cat. #: 00-5523-00). For cytokine and granzyme B production measurement, prior to the mentioned intracellular staining, tumor cell suspensions were stimulated with PMA and ionomycin, plus protein transport inhibitors (Thermo Fisher Scientific, cat. #:00-4975-93), or as negative control, only protein transport inhibitors (Thermo Fisher Scientific, cat. #: 00-4980-93) for 4 hours.

For antigen-specific experiments, 5 × 10⁵ LLC-OVA cells were subcutaneously injected in the right flank of 6- to 8-weeks-old C57/BL6 females. Nine days after tumor injection, OT-I CD8⁺ T cells from the spleens of donor control or NAE1-KO mice were sorted to enrich viable cells in a BD FACS Aria Fusion sorter. A measure of 9 × 10⁵ OT-I CD8⁺ T cells were injected intravenously in the tail vein of the recipient mice. Donor OT-I CD8⁺ T cells had been previously activated in vitro for 3 days with anti-CD3 and anti-CD28. Mice bearing tumors bigger than 35 mm³ at the day of injection were included in the study and randomized according to their size in the different experimental conditions. Mice were monitored for tumor growth every other day using a digital caliper. Tumor volumes were calculated using the following formula: V (mm³) = (π × L × W2)/6 (L is tumor length and W is tumor width). Mice were sacrificed when tumor length reached 15 mm.

We first examined NEDD8 expression in tumor-associated immune and nonimmune cell populations using a publicly available single-cell lung cancer atlas, which includes 185 human primary non–small cell lung cancer samples. To complement this analysis, we assessed three additional datasets: 58 lung adenocarcinoma samples from 44 patients (GSE131907), 31 melanoma tumors (GSE115978), and 12 hepatocellular carcinoma samples (GSE125449). In these datasets, we specifically investigated NEDD8 expression in naïve and effector-like CD8⁺ T cells. Briefly, all datasets underwent quality control filtering based on the number of detected features and the percentage of mitochondrial gene expression. Data were then normalized, and pure CD8⁺ T cells were identified using scGate. Naïve and effector-like CD8⁺ T cells were subsequently annotated based on the human CD8⁺ tumor-infiltrating lymphocyte (TIL) atlas using ProjecTILs. All computational analyses were performed in R (version 4.4.1) using the following packages: Seurat (v5.1.0), UCell (v2.7.7), ProjecTILs (v3.3.0), scGate (v1.6.2), Biostrings (v2.70.3), dplyr (v1.1.4), stringr (v1.5.1), patchwork (v1.2.0), tidyverse (v2.0.0), MetBrewer (v0.2.0), RColorBrewer (v1.1-3), Matrix (v1.7-0), and biomaRt (v2.58.2).

The input data processed in the single-cell RNA sequencing (scRNA-seq) reanalysis comprised three datasets acquired from the NCBI Gene Expression Omnibus: GSE131907 (lung cancer), GSE115978 (melanoma), and GSE125449 (hepatocarcinoma). Additionally, the integrated lung cancer atlas by Prazanowska and Lim (45) was downloaded from https://figshare.com/articles/dataset/NSCLC_Final_dataset/22114682?backTo=/collections/An_integrated_single-cell_transcriptomic_dataset_for1_non-small_cell_lung_cancer/6222221.

Regarding our proteomic analysis of activated mouse CD8⁺ T cells, the MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (46) partner repository with the dataset identifier PXD050529 (https://www.ebi.ac.uk/pride/archive?keyword=PXD050529&sortDirection=DESC&page=0&pageSize=20). For the mouse NEDD8 IP data, the MS data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD050527 and 10.6019/PXD050527 (https://www.ebi.ac.uk/pride/archive?keyword=PXD050527&sortDirection=DESC&page=0&pageSize=20). Any remaining information can be obtained from the corresponding author upon request.

The R code used for single-cell data reanalysis can be found at https://github.com/eprieto012/NEDD8_APLab.

To explore the NEDDylation state of immune cells within tumors, we first carried out an in silico approach to analyze publicly available scRNA-seq data from patients with lung cancer. We first examined the expression of NEDD8 in tumor-associated immune and nonimmune populations using the scRNA-seq lung cancer atlas generated by Prazanowska and Lim (45), and we found expression of NEDD8 in all tested populations (Fig. 1A). We then compared NEDD8 expression in CD8⁺ TIL subpopulations, classified according to their differentiation status using ProjectTILs (47). Higher expression of NEDD8 was observed in effector memory compared with naïve-like CD8⁺ T cells in three representative datasets from different cancer types: lung cancer (48), melanoma (49), and hepatocarcinoma (Fig. 1B; ref. 50). Based on these findings, we assessed expression of NEDD8 in LLC mouse subcutaneous tumors. Immunofluorescence images in Fig. 1C showed that NEDD8 was expressed in the intracellular compartment of CD3⁺ TILs. Subsequently, we carried out in vitro experiments to confirm the expression of NEDD8 in murine and human CD8⁺ T cells, isolated from the spleen and blood of healthy donors, respectively. We found that activation of both mouse (Fig. 1D) and human (Fig. 1E) CD8⁺ T cells promoted the expression of NEDD8 transcripts. Activated CD8⁺ T cells also upregulated the expression of UBE2M, a key enzyme of the NEDDylation cycle, whereas NAE1 levels remained constitutive. This transcriptional regulation was accompanied by an enrichment of NEDDylation at the protein level, as evidenced by the accumulation of NEDDylated cullins and free NEDD8 in both mouse and human activated CD8⁺ T cells (Fig. 1F and G). The protein levels of NAE1 and UBC12 (UBE2M) were also increased upon activation (Fig. 1F and G; Supplementary Fig. S1A and S1B). In addition, stimulation of TCR signaling with an anti-CD3 was sufficient to upregulate NEDDylated proteins and NAE1 (Supplementary Fig. S1C–S1E).

To assess the role of NEDDylation in CD8⁺ T-cell function, we generated a transgenic mouse in which NAE1 was conditionally deleted by the Cre recombinase under the control of the distal Lck promoter (NAE1-KO; Fig. 2A). This led to the specific deletion of NAE1 in T cells after their thymic maturation. We confirmed that the thymus and the spleen of NAE1-KO mice contained normal immune populations (Supplementary Fig. S2A–S2C). However, we did observe a modest reduction in the percentage of CD8⁺ T cells present in lymph nodes (Supplementary Fig. S2D), suggesting a potential role of NEDDylation in mature and antigen-experienced CD8⁺ T cells.

We confirmed that the deletion of NAE1 in CD8⁺ T cells resulted in a reduction of NEDDylated cullins upon activation (Fig. 2B). Similarly, treatment with the pharmacologic inhibitor of NAE, MLN4924, also reduced the amount of NEDDylated cullins in both mouse (Fig. 2B) and human (Fig. 2C) CD8⁺ T cells. Thereafter, we measured the viability and proliferation of mouse CD8⁺ T cells with defective NEDDylation, either isolated from NAE1-KO mice or treated with MLN4924. Both conditions resulted in a decrease in viability and an increase in late apoptotic cells, as measured by 7AAD and annexin V staining (Fig. 2D; Supplementary Fig. S3A). Similar reductions in viability were observed in human CD8⁺ T cells treated with MLN4924 (Fig. 2E; Supplementary Fig. S3B). Mouse CD8⁺ T cells with deficient NEDDylation had a defect in proliferation, measured by cell counting (Fig. 2F) and by CFSE dilution (Fig. 2G and H), potentially resulting from the accumulation of cell-cycle regulators such us WEE-1, P21, and P27 (Fig. 2I; Supplementary Fig. S3C–S3E), as previously described for other cell types (51). A similar proliferation impairment was observed when human CD8⁺ T cells were treated with MLN4924 (Fig. 2J and K).

The reduced viability and proliferation of NEDDylation-deficient CD8⁺ T cells may result from attenuated T-cell activation, as indicated by reduced CD69 expression in NAE1-KO CD8⁺ T cells (Fig. 2L) and in human CD8⁺ T cells treated with increasing doses of MLN4924 (Fig. 2M). We also measured the activity of the NF-κB and NFAT signaling pathways in Jurkat TPR cells (52–54). NF-κB and NFAT are important transcriptional regulators for CD8⁺ T-cell survival and proliferation (55–57), and their activity was reduced when Jurkat TPR cells were treated with MLN4924 (Fig. 2N and O).

We next sought to validate whether these defects had functional consequences in vivo. Because the role of NEDDylation in T cell–mediated antitumor immunity is poorly described, we decided to investigate the impact of NEDDylation in the ability of T cells to fight against solid tumors. To this end, we subcutaneously implanted LLC cells in control or NAE-1-KO mice. We observed that dysfunctional NEDDylation in T cells reduced their capacity to control tumor growth (Fig. 3A and B). In this context, the amount of CD4⁺ (Fig. 3C) and CD8⁺ T cells (Fig. 3D) found per milligram of tumor was severely reduced in NAE1-KO mice. A decrease in the amount of NK cells (NK-1.1⁺) per milligram of tumor was also observed, whereas the amounts of intratumoral B lymphocytes (CD19⁺) and myeloid cells (CD11b⁺) were not affected (Supplementary Fig. S4A). In addition, NAE1-KO CD8⁺ TILs showed a less differentiated phenotype, as measured by the expression of CD44 and CD62L (Supplementary Fig. S4B) and lower levels of LAG-3, PD-1, CD137, and TIM-3 (Supplementary Fig. S4C and S4D). Furthermore, the percentage of intratumoral Ki-67⁺ T cells was reduced in NAE1-KO mice (Supplementary Fig. S4E and S4F). Immunophenotyping of secondary lymphoid organs revealed no differences in immune-cell composition in the spleen (Supplementary Fig. S4G) and a reduction in the proportion of CD8⁺ T cells in the tumor-draining lymph nodes (Supplementary Fig. S4H).

We performed additional in vivo experiments with the melanoma cell line YUMMER 1.7-GFP, which is more immunogenic than the LLC cell line (58, 59). Again, tumor growth was accelerated in NAE1-KO mice (Fig. 3E; Supplementary Fig. S4I), and the amount of CD4⁺ and CD8⁺ T cells was drastically reduced per milligram of tumor (Supplementary Fig. S4J and S4K). Analyses of the composition of the NAE1-KO TILs revealed a reduction in the percentage of T-bet⁺CD4⁺ T cells (Fig. 3F) and CD44⁺CD62L^-CD8⁺ T cells (Fig. 3G), suggesting an impairment of the CD4⁺ Th1 phenotype and CD8⁺ T-cell effector differentiation, respectively. Within the CD8⁺ T-cell compartment, NAE1 deletion resulted in a decrease in the percentage of PD-1⁺TIM-3⁺TIGIT⁺ triple-positive T cells (Fig. 3H), CD39⁺TIM-3⁺ double-positive T cells (Fig. 3I), and CD44^highPD-1⁺ T cells (Fig. 3J). Whereas control cells expressed higher quantities of these exhaustion markers, TCF7 expression was increased in CD44^highPD-1⁺ NAE1-KO CD8⁺ T cells, indicating a higher generation of progenitor exhausted cells (Fig. 3K). Overall, the ability of NAE1-KO CD8⁺ T cells to produce granzyme B was significantly impaired (Fig. 3L), contributing to their diminished antitumor functionality, which parallels their reduced differentiation capability. Next, we evaluated the role of NEDDylation in T-cell exhaustion using the protocol described by Scharping and colleagues (43), which involves repeated in vitro stimulations to induce exhaustion in CD8⁺ T cells. The levels of NEDDylated proteins were significantly elevated in exhausted CD8⁺ T cells (Supplementary Fig. S5A and S5B). Furthermore, whereas the percentage of exhausted cells was comparable between control and NAE1-KO CD8⁺ T cells (Supplementary Fig. S5C), the functional capability of NAE1-KO cells was diminished, as indicated by a reduction in the production of IFNγ (Supplementary Fig. S5D).

To further explore this phenotype, we took advantage of the OT-I antigen–specific model, in which the role of CD8⁺ T cells is critical. For this purpose, we crossed NAE1-KO mice with OT-I mice to generate control-OT-I mice and NAE1-KO-OT-I mice. We then transferred activated control-OT-I or NAE1-KO-OT-I CD8⁺ T cells into LLC-OVA tumor-bearing mice and analyzed tumor progression and mice survival. OT-I cells were sorted to isolate and inject only alive cells. Whereas transfer of control-OT-I CD8⁺ T cells resulted in a delay in tumor growth, transfer of NAE1-KO-OT-I cells showed a diminished therapeutic activity and less ability to control tumor progression (Fig. 3M–O). Consequently, the survival rate of mice infused with NAE1-KO-OT-I cells was significantly reduced in comparison with mice infused with control-OT-I cells (Fig. 3P). We quantified the levels of OT-I CD8⁺ T-cell infiltration taking advantage of the congenic marker CD45.1⁺, finding that CD45.1⁺ NAE1-KO-OT-I cells infiltrated the tumors of CD45.2⁺ recipient mice in similar quantities to the CD45.1⁺ control-OT-I cells (Supplementary Fig. S6A). At this point, tumor weight was significantly greater in mice receiving NAE1-KO-OT-I cells (Supplementary Fig. S6B). We also found that the in vitro cytotoxic function of NAE1-KO-OT-I cells was reduced by coculturing control-OT-I and NAE1-KO-OT-I cells with B16-luciferase-OVA target cells (Supplementary Fig. S6C and S6D).

Given the consistent observation that NAE1-KO CD8⁺ T cells displayed a reduced antitumor function, we next explored whether upregulating NEDDylation could have the opposite effect. For this, we developed an approach to enhance NEDDylation in CD8⁺ T cells based on knocking out the deNEDDylase SENP8, which specifically removes NEDD8 from non-cullin substrates (60, 61). As a result, the amount of NEDDylated proteins was increased in human CD8⁺ T cells (Fig. 3Q). To explore its antitumor effects, we generated SENP8-KO anti-CD19 CAR CD8⁺ T cells and cocultured them with CD19-expressing RAMOS target cells. Knocking out SENP8 in CAR T cells enhanced their cytotoxic capability (Fig. 3R). In summary, these findings show the importance of NEDDylation in the antitumor function of CD8⁺ T cells.

Our in vivo assays revealed that NEDDylation has an important effect on CD8⁺ T-cell function. To obtain a detailed perspective of the effects of NEDD8 in CD8⁺ T cells, we performed a proteomic analysis in NAE1-KO CD8⁺ T cells and MLN4924-treated CD8⁺ T cells, compared with control cells. We identified many significantly dysregulated proteins (Supplementary Table S2). These were analyzed by IPA, revealing several dysregulated pathways upon inhibition of NEDDylation (Fig. 4A; Supplementary Tables S2 and S3), including glycolysis, the HIF-1α pathway, and oxidative phosphorylation (Fig. 4B). Additionally, other identified pathways were the TCR, migration, and MYC signaling pathways (Supplementary Fig. S7A–S7C). To functionally validate these findings, we evaluated the metabolic activity of NAE1-KO CD8⁺ T cells by measuring the extracellular acidification rate (Fig. 4C) and oxygen consumption rate (Fig. 4D). NAE1-KO CD8⁺ T cells had lower basal glycolysis (Fig. 4E), compensatory glycolysis (Fig. 4F), and basal respiration (Fig. 4G). Spare respiratory capacity and maximal respiration were also impaired (Supplementary Fig. S7D and S7E). A similar phenotype was observed in MLN4924-treated human CD8⁺ T cells (Fig. 4H and I), leading to lower basal glycolysis (Fig. 4J), compensatory glycolysis (Fig. 4K), and basal respiration (Fig. 4L). To gain further insight into the link between NEDDylation, metabolism, and cytotoxic capability, we assessed granzyme B production in control and NAE1-KO CD8⁺ T cells that had been treated with galactose to restrict their glycolysis. A second strategy to inhibit glycolysis was treatment with 2-DG, a stronger inhibitor of the glycolytic pathway (bioRxiv.2020.02.05.935627). Under standard culture conditions, NAE1-KO CD8⁺ T cells exhibited a diminished ability to produce granzyme B (Supplementary Fig. S7F). Glycolysis inhibition with galactose decreased granzyme B production in control cells but had no effect on NAE1-KO CD8⁺ T cells. Consistently, 2-DG treatment compromised granzyme B production in control cells to the levels observed in NAE1-KO cells (Supplementary Fig. S7F), suggesting that the observed reduction of granzyme B production in NAE1-KO T cells could be induced by their metabolic impairment.

The HIF response can directly influence metabolic T-cell adaptations in certain conditions (62). We confirmed that Hif1α mRNA levels were similar in both NAE1-KO and MLN4924-treated CD8⁺ T cells compared with control cells (Supplementary Fig. S7G). At the protein level, HIF-1α accumulated in both models of NEDDylation deficiency (Supplementary Fig. S7H), an effect previously documented in other cell lines (63, 64). However, this accumulation did not correlate with higher HIF-1α transcriptional activity. In fact, NAE1-KO CD8⁺ T cells exhibited reduced expression of HIF-1α target genes (Ldha, Glut1, Pdk1, and Pgk1; Supplementary Fig. S7I; ref. 65), confirming the impaired HIF-1α activity in NAE1-KO CD8⁺ T cells shown in Fig. 4A.

We next focused on the identification of NEDDylated proteins that could mediate the observed CD8⁺ T-cell phenotype. To this end, we performed an IP of NEDD8 in activated mouse CD8⁺ T cells, followed by MS analysis (Fig. 5A). Numerous proteins were identified as NEDD8 targets (Supplementary Table S4) and analyzed using the Gene Ontology database (66). A variety of cellular metabolic processes were found to be enriched (Fig. 5B). Glycolysis was one of the top scoring pathways, and key glycolytic enzymes were found in the list of NEDD8 targets, such as LDHA, α-enolase (ENO1), and hexokinase 1 (HK-1; ref. 67). Given the central role of glycolysis in controlling T-cell function, we carried out an IP assay in activated human CD8⁺ T cells, in which we validated LDHA, ENO1, and HK-1 as targets of NEDD8 (Fig. 5C). Furthermore, LDHA protein expression was downregulated in a dose-dependent manner (Fig. 5D) in MLN4924-treated human CD8⁺ T cells, whereas the levels of ENO1 and HK-1 remained unchanged. This reduction in LDHA expression was also observed in NAE1-KO CD8⁺ T cells and in MLN4924-treated mouse CD8⁺ T cells (Fig. 5E). Densitometric quantifications are shown in Supplementary Fig. S8A and S8B. We observed that inhibition of NEDDylation led to a transcriptional impairment of LDHA, as evidenced by reduced transcript levels in both human (Fig. 5F) and mouse (Fig. 5G) NEDDylation-deficient CD8⁺ T cells. This transcriptional downregulation was accompanied by a decrease in LDHA activity (Fig. 5H), suggesting a link between defective NEDDylation and glycolytic dysfunction.

Our study elucidates the pivotal role of NEDDylation, mediated by NAE1, in the regulation of CD8⁺ T-cell metabolism and antitumor immunity, highlighting its significance beyond the well-documented impact on tumor cell proliferation and survival (68). We have identified that TILs upregulate the NEDDylation pathway as they differentiate into effector memory cells, favoring an adaptive metabolic shift that is essential for their antitumor function. Specifically, we show that both genetic and pharmacologic inhibition of NEDDylation severely impair CD8⁺ T-cell viability, proliferation, and metabolic pathways, resulting in a reduction of their effector function.

Our findings have implications for the current therapeutic landscape, in which NEDDylation inhibitors, such as MLN4924, are under clinical investigation for cancer treatment. In vivo, MLN4924 is effective in controlling tumor growth and promoting antitumor immunity (33, 34). However, the direct effects of inhibiting NEDDylation in T cells require further understanding. Recent clinical trial shortcomings of MLN4924 against several tumor types have tempered expectations, highlighting the need for further research of this pathway, including its immune modulation properties. For this purpose, the NAE1-KO model provides a tool to elucidate the roles of NEDDylation specifically in CD8⁺ T cells.

Beyond NAE inhibitors, emerging antitumor approaches include targeting NEDDylation E2 conjugating enzymes such as UBE2M (69). However, inhibiting UBE2M can also compromise CD4⁺ T-cell function (70) and could similarly impact CD8⁺ T-cell function given that UBE2M expression is upregulated by activated CD8⁺ T cells.

In this study, we describe the detrimental effects of NEDDylation inhibition on CD8⁺ T-cell function, demonstrating a defect in their ability to combat tumor cells. These findings can help to explain the limited efficacy of NEDDylation inhibitors in the clinic. This paradox underscores the complexity of targeting cellular processes that are pivotal for both tumor progression and immune defense mechanisms.

In this context, we show that NEDDylation favors the acquisition of a glycolytic phenotype by T cells. Specifically, we demonstrate that NEDDylation-deficient CD8⁺ T cells are characterized by reduced LDHA transcript levels, protein expression, and enzymatic activity. This finding aligns with the dysregulation of MYC and HIF1-α signaling pathways, known regulators of LDHA and the associated metabolic reprogramming that drives T-cell activation (65, 71). In fact, metabolic reprogramming is a central feature of the robustness of immune responses (72). In the tumor microenvironment, effector T cells rely on LDHA for their differentiation and function (73), a finding consistent with our observation of impaired differentiation in vivo during tumor-challenge assays.

This mechanism can offer opportunities for therapeutic intervention by modulating the NEDDylation of these glycolytic enzymes and metabolic pathways to enhance the metabolic fitness and antitumor capabilities of T cells. We identified the glycolytic enzymes LDHA, HK-1, and ENO1 as NEDD8 targets in activated CD8⁺ T cells, suggesting another layer of regulation of metabolism by NEDDylation, similar to that observed with other cell types (74, 75). PTMs are known to regulate multiple members of the same signaling pathway (61), suggesting that NEDDylation could control glycolysis acting on several enzymes of this pathway. Determining whether these enzymes are regulated posttranslationally by NEDD8 is a critical area of investigation that requires further elucidation.

The specific regulation of these NEDD8 targets is of particular interest, and the substrate specificity of E3 ligases can aid in this purpose (76). Identifying the E3 ligases involved in the specific NEDDylation of these substrates can avoid the negative outcomes of overall NEDDylation inhibition and unlock new ways to manipulate protein–protein interactions, such as using proteolysis-targeting chimeras. In a similar manner, the extension of PTM strategies into the field of adoptive cell therapies is a novel approach. For example, engineering of CAR T cells to circumvent CAR ubiquitination enhances their antitumor effectiveness and endurance within the host (77, 78). In this study, we propose an approach wherein knocking out the deNEDDylase SENP8 confers superior antitumor characteristics to CD8⁺ CAR T cells. This strategy involves upregulating NEDDylation in CD8⁺ T cells, serving as a mechanism to optimize a variety of adoptive cell therapies, including engineered CAR and TCR approaches. An alternative approach was recently proposed by Liao and colleagues (79), based on the inhibition of cullin 5 in CAR T cells. Notably, adoptive cell therapies provide an opportunity to specifically treat engineered T cells without requiring systemic administration of the tested agent, which can be useful to regulate NEDDylation to keep its beneficial outcomes while avoiding its deleterious effects.

In conclusion, our study highlights NEDDylation as a critical regulatory mechanism in CD8⁺ T-cell metabolism and antitumor immunity, delineating the dual role of NEDDylation in tumor cells and immune cells, and exploring molecular pathways supporting potential therapeutic strategies that harness this pathway for cancer immunotherapy.

B. Jiménez-Lasheras reports a patent for EP21382780 pending. A. Palazon reports a patent for EP21382780 pending. No disclosures were reported by the other authors.

B. Jiménez-Lasheras: Conceptualization, resources, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. P. Velasco-Beltrán: Conceptualization, resources, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. L. Egia-Mendikute: Resources, formal analysis, supervision, validation, investigation, visualization, methodology, writing–review and editing. L. Pérez-Gutiérrez: Resources, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. S.Y. Lee: Resources, formal analysis, validation, investigation, visualization, methodology. A. de Blas: Resources, formal analysis, validation, investigation, visualization, methodology. A. García-del Río: Resources, formal analysis, validation, investigation, visualization, methodology. S.R. Zanetti: Formal analysis, validation, investigation, visualization, methodology. A. Antoñana-Vildosola: Resources, formal analysis, validation, investigation, visualization, methodology. A. Barreira-Manrique: Resources, formal analysis, validation, visualization, methodology. A. Bosch: Resources, Formal analysis, validation, investigation, visualization, methodology. J. Etxaniz-Díaz de Durana: Resources, formal analysis, validation, investigation, visualization, methodology. E. Prieto-Fernández: Resources, data curation, software, formal analysis, validation, investigation, visualization, methodology. M. Serrano-Maciá: Resources, formal analysis, validation, investigation, visualization, methodology. N. Goikoetxea-Usandizaga: Resources, formal analysis, methodology. M. Azkargorta: Resources, data curation, software, formal analysis, methodology. F. Elortza: Resources, data curation, software, formal analysis, methodology. T. Gruber: Resources, formal analysis, validation, investigation, visualization, methodology. S. Peer: Resources, formal analysis, validation, investigation, visualization, methodology. G. Baier: Resources, supervision, funding acquisition. A. Woodhoo: Resources, formal analysis, validation. M.L. Martínez-Chantar: Conceptualization, resources, formal analysis, supervision, validation, investigation, visualization, methodology. A. Palazon: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

This research was supported by the following: Ministerio de Ciencia, Innovación y Cultura (MICIU)/Agencia Española de Investigación (AEI; 10.13039/501100011033) financed “PID2019-107956RA-I00” project (A.P); MICIU/AEI/10.13039/501100011033 and European Union Next Generation EU/PRTR financed “PDC2022-133300-I00” (A. Palazon), “TED2021-129433B-C21” (A. Palazon) and “JDC2022-048612-I” (L. Pérez-Gutiérrez) projects; MICIU/AEI (10.13039/501100011033) and “Fondo Europeo de Desarrollo Regional (FEDER)”/UR financed “PID2022-139344OB-I00” project (A. Palazon); MICIU/AEI/10.13039/501100011033 and “El Fondo Social Europeo (FSE) invierte en tu futuro” financed “RYC2018-024183-I” (A. Palazon) and “PRE2020-092342” (P. Velasco-Beltrán) grants; AEI financed “RYC2021-031213-I” grant (E. Prieto-Fernández), FSE⁺ financed “PRE2022-104817” (J. Etxaniz-Díaz de Durana) predoctoral grant; Ministerio de Ciencia, Innovación y Universidades MICINN: PID2023-146933OB-100 funded by MCIU /AEI /10.13039/501100011033/FEDER, UE, as part of Plan Estatal de Investigación Científica y Técnica e Innovación (M.L. Martínez-Chantar); MCIN/AEI/10.13039/501100011033 financed the grant CEX2021-001136-S (M.L. Martínez-Chantar); the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program financed grant agreement “no 804236” (A. Palazon); the ERC awarded to G. Baier with “ERC-2018-ADG 786462-HOPE” grant; the Provincial Section of Bizkaia and of the Scientific Foundation of the Spanish Association against cancer (AECC) financed “PRDVZ21640DEBL” (A. de Blas) and “PRDVZ1900BOSCH” (A. Bosch) predoctoral grants; the Scientific Foundation of AECC financed “LABAE211744PALA” project (A. Palazon); “La Caixa” Foundation financed the “HR21-00925” (A. Palazon); “La Caixa” Foundation under the program of INPhINIT fellowships financed Doctoral INPhINIT Fellowship – Retaining under the code “LCF/BQ/DR20/11790022” (A. Antoñana-Vildosola); “XVI” BECA FERO (A. Palazon); the Basque Government financed “PRE_2019_1_0320” predoctoral grant (B. Jiménez-Lasheras) through the “Programa Predoctoral de Formación de Personal Investigador No doctor del Departamento de Educación,” “2023333027” project (A. Palazon) through the “Programa Ayudas a proyectos de investigación y desarrollo en salud del Departamento de Salud,” and “AF-W1-2019-00012” through Programa Bikaintek (A. García-del Río).