DUSP6 ablation restores CAR T-cell fitness impaired by tumor CD58 loss through invigoration of AP-1 signaling

TL;DR

Researchers found that blocking DUSP6 can restore the effectiveness of CAR T-cells weakened by tumor CD58 loss, improving their function and survival.

Contribution

The study identifies DUSP6 as a novel target to enhance CAR T-cell fitness by restoring AP-1 signaling and mitochondrial function.

Findings

DUSP6 blockade restores AP-1 signaling and mitochondrial fitness in CAR T-cells impaired by tumor CD58 loss.

DUSP6 ablation reduces CAR T-cell apoptosis and enhances long-term cytotoxicity and proliferation.

DUSP6 downregulation correlates with positive patient outcomes in T-cell-based immunotherapy.

Abstract

Primary resistance to chimeric antigen receptor (CAR) T-cell therapies has limited their widespread application. Our prior genome-wide CRISPR/Cas9 screening revealed that the loss of CD58, a crucial intrinsic resistance factor in tumors, resulted in insufficient immune synapse formation and impaired CAR T-cell activation and cytotoxicity. However, the specific signaling pathway and transcriptional changes associated with CAR T-cell dysfunction have not been addressed. Here, we revealed that AP-1-mediated activation was attenuated in CAR T cells impaired by tumor CD58 loss, driving a decrease in mitochondrial biogenesis, metabolic kinetic impairment, mitochondrial membrane potential loss and ROS accumulation. Moreover, this AP-1 attenuation triggered death receptor-independent apoptosis through the intrinsic mitochondrial pathway. In seeking therapeutic strategies, we pharmacologically and genetically blocked three distinct inhibitory phosphatases positioned upstream of AP-1 signaling. Multifaceted validation has demonstrated that dual specificity phosphatase 6 (DUSP6) blockade is an effective approach to supplement AP-1 signaling while notably reducing CAR T-apoptosis and enhancing mitochondrial fitness, proliferation and long-term cytotoxicity. The transcriptomic profiles of DUSP6-ablated CAR T cells revealed markedly upregulated T-cell activation signatures and enriched metabolic pathways. Clinically, bulk and single-cell RNA-seq analyses revealed that DUSP6 was downregulated in patients who responded to T-cell-based immunotherapy, implying its relevance to patient outcomes. Our findings repositioned CD58 not merely as an immune synapse component but also a metabolic checkpoint in CAR T-cell biology, the loss of which triggers AP-1-dependent mitochondrial derangement and creates a permissive landscape for intrinsic apoptosis, which can be ameliorated by ablation of the inhibitory phosphatase DUSP6. Crucially, DUSP6 ablation represents a promising engineering target to potentiate CAR T-cell efficacy in broader applications.

Figures

a). Nevertheless, in NY-ESO-1 TCR-Jurkat reporter cell lines, tumor CD58 loss triggered altered activation dynamics in all three signaling pathways (Supplementary Fig. 2b). Additionally, time-dependent attenuation of c-Jun and c-Fos phosphorylation, key subunits that form the functional AP-1 transcription factor complex, was detected in CAR T cells stimulated with CD58^KO^ tumor cells, suggesting compromised AP-1 complex assembly kinetics, whereas minimal alterations in NFAT and NF-κB p65 phosphorylation were observed (Fig. [1d](#Fig1)). This finding was further corroborated by phosphoprotein

Modulation of AP-1 signaling influences CAR T-cell efficacy and mitochondrial fitness in CD58-deficient tumor environments. a–c Changes in FACS-measured activation levels in AP-1, NFAT or NF-κB reporter CD19 CAR-Jurkat cells in response to T-5224 at 60 μM for 12 h compared with those in WT Nalm6 cells at an E:T ratio of 1:1. d FACS analysis of the impact of T-5224 on the mitochondrial mass in CD19 CAR T cells treated with WT tumor cells compared with that in CD58^KO^ tumor cells. The mitochondrial fold change was calculated on the basis of the mitochondrial quantification of untreated CD19 CAR

DUSP6 ablation mitigates CD58 loss-mediated resistance to CAR T-cell therapy by increasing functional activation, mitochondrial fitness, and resistance to apoptosis. a Schematic of electroporation-based CRISPR-Cas9 gene editing in CAR T cells. b Western blot validation of the protein-level knockout efficiency of SHP1, PTPN7 and DUSP6. c FACS-based measurement of Ki-67 expression in CD19 CAR T cells 4 days post-electroporation. d Schematic of the repeated tumor challenge model used to assess the impact of SHP1, PTPN7 and DUSP6 ablation on CD19 CAR T-cell expansion and increased mitochondrial ma

a). Notably, CAR T cells cocultured with CD58^KO^ tumor cells exhibited a robust increase in apoptosis (Fig. [3a](#Fig3)). At the transcriptomic level, we also observed a marked concomitant increase in the expression of mitochondrial apoptosis-related genes, suggesting the unanticipated initiation of mitochondrial-dependent cell death circuitry upon tumor CD58 loss (Fig. [3b](#Fig3)). We then inspected the expression of mitochondrial apoptosis-related markers at the protein level. A notable decrease in the antiapoptotic protein Bcl-XL and an increase in the proapoptotic protein Bim were detect

Enriched pathways related to cell activation and energy metabolism were found in DUSP6^KO^ CAR T cells. a Gene ontology enrichment analysis of the transcriptomic profiles of Ctrl or DUSP6^KO^ CAR T cells after CD58^KO^ Nalm6 challenge. Ctrl or DUSP6^KO^ CAR T cells were sorted by flow cytometry after stimulation with CD58^KO^ Nalm6 cells at a 1:1 E:T ratio. b GSEA highlighting key signaling pathway alterations in DUSP6^KO^ CAR T cells versus Ctrl CAR T cells when stimulated with CD58^KO^ Nalm6 cells. c Six distinct transcriptional trajectories identified by the Mfuzz series test of cluster ana

DUSP6 could be negatively correlated with T-cell immunotherapy outcome. a Proportion of PD and CR patients in the low- and high-DUSP6 subgroups of CD8^+^ and CD4^+^ CAR T cells collected 10 days post-infusion from our tandem CD19/20 CAR T-cell-treated R/R diffuse large B-cell lymphoma cohort ([GSE223655](GSE223655)). The classification of high and low DUSP6 expression in patient samples was determined by the median expression value within each sample group. b Proportion of PD patients and CR patients in the low- and high-DUSP6 subgroups among premanufactured T cells from the same cohort. c UMA

Full text

Introduction

CAR T-cell therapy has extensively reshaped the treatment landscape of oncology and immune-related diseases, especially hematological malignancies.^1–3^ In B-cell malignancies, CD19 CAR T cells have achieved unprecedented efficacy. However, primary or acquired resistance to CAR T-cell therapy still occurs and is the main reason for treatment failure.^1,4–6^ The efficacy of CAR T cells is determined not only by the binding affinity between CAR molecules and target antigens as well as the expression levels of these antigens, but also by the ligands of CAR-activating costimulatory molecules expressed on tumor cells. In recent years, CAR T-cell therapy resistance induced by the deficiency of tumor-derived costimulatory ligands has attracted increasing attention. Our prior genome-wide CRISPR/Cas9 screening in a Nalm6/CD19 CAR T-cell coculture model highlighted the loss of tumor CD58 as a critical intrinsic resistance factor.^7^ However, the specific underlying mechanisms by which tumor CD58 deficiency induces CAR T cell dysfunction remain incompletely elucidated.

CD58 is a broadly expressed costimulatory receptor that engages its natural ligand CD2 to mediate immune synapse (IS) formation in T and natural killer (NK) cells.^8^ The CD58-CD2 interaction dynamically enhances adhesion at the effector-target cell interface during antigen-specific recognition. This interaction is essential for T/NK cells to mediate target recognition, functional activation, and interface stabilization. CD58 gene aberrations are prevalent across various hematological malignancies, with reported frequencies of 13.7% in diffuse large B-cell lymphomas (DLBCL), 47.6% in follicular lymphomas, 82.7% in mantle cell lymphomas, 76.8% in marginal zone lymphomas, and 81.3% in small lymphocytic lymphomas.^9^ Loss of CD58 also occurs in 50% of primary mediastinal large B-cell lymphomas, lymphoplasmacytic lymphomas, Burkitt lymphomas, and plasmablastic lymphomas and is even more pronounced in T- and NK-cell lymphomas (73.2–95.8% downregulation).^9^ CD58 alterations are associated with poor prognosis, promoting immune evasion by impairing tumor-infiltrating lymphocyte- and NK cell-mediated killing.^10,11^ Studies in Burkitt lymphoma, Hodgkin lymphoma, and a Chinese DLBCL cohort (n = 151) consistently link CD58 loss to adverse outcomes.^12–14^ In response to cellular immunotherapy, CD58 alterations are more common in CAR T-cell nonresponders, as CD58 facilitates effector cell migration and aggregation.^15^ Recent data reported by Majzner et al. and Carturan et al. revealed that patients with higher CD58 expression before CD19 CAR T-cell therapy had significantly prolonged survival.^16,17^ Additionally, comparisons of CD58 expression in patient biopsies before and after CAR T-cell therapy revealed that CD58 expression scores decreased significantly after treatment.^17^ Notably, multiple studies, including our own, have identified CD58 loss as a key resistance mechanism in CAR T cells and other immunotherapies, underscoring its role as a critical immune evasion biomarker.^10,15,18,19^ However, the precise mechanisms underlying tumor CD58 loss remain underexplored.

We confirmed in our previous study that disruption of CD58 in tumor cells impaired the IS formation of CAR T cells and affected CAR T-cell proximal signaling, leading to impaired proliferation, cytotoxicity, and cytokine secretion.^7^ Nevertheless, beyond phenotypic observations, the underlying mechanisms via which tumor CD58 deficiency impacts the downstream signal transduction and functional homeostasis of CAR T cells remain largely uncharacterized at a systematic level. Reported studies on tumor CD58 deficiency-mediated CAR T cell therapy resistance have mainly focused on the role of CD58 as a physical component of immune synapses, whereas its function in modulating tumor immune signaling within CAR T cells remains largely underexplored. Although previous studies have investigated the potential role of CD2 in assisting T-cell receptor (TCR) activation,^8,20–24^ CAR T cells are endowed with a unique activation paradigm.^25^ On the one hand, CAR design enables direct recognition of surface antigens on target cells without MHC presentation, triggering faster and more sustained activation than the classic TCR signal does. On the other hand, variations in structure and costimulatory patterns across different CARs can result in unique activation characteristics. Owing to these differences, certain research conclusions derived from TCR studies may not be directly extrapolated to CAR T-cell dynamic models. In this context, the role of the CD58-CD2 interaction in regulating CAR T-cell downstream signaling (notably transcription factor activity) and its mechanistic contributions to sustaining cytotoxicity, metabolic fitness, and persistence remain largely unexplored. Deciphering how the CD58-CD2 axis modulates transcriptional networks in CAR T cells could reveal novel targets for enhancing metabolic plasticity and long-term antitumor activity, providing a rational framework for optimizing next-generation T-cell immunotherapies.

In this study, we found that tumor CD58 deficiency preferentially disrupts AP-1 signaling over the NFAT and NF-κB pathways in CAR T cells. This selective impairment triggered a metabolic crisis characterized by decreased mitochondrial biogenesis, mitochondrial membrane potential (ΔΨm) loss and reactive oxygen species (ROS) overaccumulation, culminating in intrinsic apoptosis via a mitochondria dependent, death receptor-independent cascade. When DUSP6 was pharmacologically or genetically ablated, AP-1 activation was significantly complemented with enhanced proliferation, cytotoxicity and persistence in CAR T cells cocultured with CD58^KO^ tumor cells, resulting in significantly superior mitochondrial function and a lower level of apoptosis. Collectively, these findings reveal that tumor CD58 loss undermines CAR T-cell function through AP-1-mediated mitochondrial dysfunction and apoptosis and identify DUSP6 inhibition as a potential strategy to alleviate such CD58 deficiency-induced impairment while increasing overall CAR T-cell efficacy.

Results

CD58 deficiency in tumors causes attenuated CAR T-cell activation, mainly via the AP-1 pathway

To explore the transcriptomic differences in CAR T cells upon tumor CD58 loss, we performed bulk RNA sequencing of CD19–4-1BBζ CAR T cells (hereafter referred to as CD19 CAR T cells) cocultured with wild-type (WT) or CD58-knockout (CD58^KO)^ tumor cells. The perturbation of three key transcription factors—NF-κB, NFAT, and AP-1—was observed, with AP-1 complex-associated transcriptional signatures exhibiting the most significant alterations and demonstrating greater consistency across different donors (Fig. 1a). To further elucidate the transcriptional regulatory dynamics, we constructed reporter cell lines based on CD19 CAR-Jurkat cell lines and NY-ESO-1 TCR-Jurkat cell lines by incorporating response elements for NF-κB, NFAT, and AP-1 that drive activation-induced fluorescent expression to enable real-time monitoring of pathway activation (Fig. 1b). Intriguingly, in CAR-engineered systems, reporter cells revealed largely unaltered NFAT and NF-κB activation upon CD58^KO^ tumor cell stimulation, while AP-1 signaling exhibited marked attenuation, with markedly elevated AP-1 transcriptional activity in the WT tumor-challenged system compared with that in the CD58^KO^ counterparts at the 6^th^ hour of coculture (Fig. 1c, Supplementary Fig. 2a). Nevertheless, in NY-ESO-1 TCR-Jurkat reporter cell lines, tumor CD58 loss triggered altered activation dynamics in all three signaling pathways (Supplementary Fig. 2b). Additionally, time-dependent attenuation of c-Jun and c-Fos phosphorylation, key subunits that form the functional AP-1 transcription factor complex, was detected in CAR T cells stimulated with CD58^KO^ tumor cells, suggesting compromised AP-1 complex assembly kinetics, whereas minimal alterations in NFAT and NF-κB p65 phosphorylation were observed (Fig. 1d). This finding was further corroborated by phosphoprotein array analysis, which revealed more pronounced phosphorylation of AP-1 signaling components in CAR T cells stimulated with WT tumor cells (Fig. 1e). Specifically, we observed marked phosphorylation of upstream phosphosites within the Ras-ERK-AP-1 signaling axis upon WT tumor stimulation. The reduced AP-1 transcriptional activity was further confirmed by the expression levels of its target genes/proteins (Supplementary Fig. 3). To further clarify whether the attenuation of AP-1 signaling is directly associated with tumor CD58 loss, we reconstituted CD58 expression in CD58^KO^ tumor cells (Fig. 1f). Coculturing CD19 CAR-Jurkat AP-1 reporter cells with WT, CD58^KO^ or CD58-reexpressing Nalm6 cells revealed that CD58 re-expression could effectively restore the AP-1 activation impaired by CD58 loss in tumor cells (Fig. 1g, Supplementary Fig. 4). These findings indicate that in the context of tumor-intrinsic CD58 loss, while multiple key signaling pathways in CAR T cells are affected, the attenuation of AP-1 transcriptional activation is consistently prominent across different experimental systems. This observation suggests that AP-1 signaling may serve as the central mechanistic node linking the tumor CD58 status to the functional competence of CAR T cells.Fig. 1. Loss of tumor CD58 causes attenuated CAR T-cell activation, mainly through AP-1 signaling. a Heatmap demonstrating differentially expressed genes (DEGs) related to NF-κB, NFAT and AP-1 activity in CD19 CAR T cells in response to tumor cell stimulation. CD19 CAR T cells were sorted by flow cytometry 3 days after coculture with WT or CD58^KO^ Nalm6 cells at an E:T ratio of 1:1 (n = 2 different PBMC donors). b Schematic of the generation of NF-κB, NFAT and AP-1 activity reporter cell lines from CD19 CAR-Jurkat cells or NY-ESO-1 TCR-Jurkat cells. c Fold change in the activation level of the CD19 CAR Jurkat reporter cell lines in the WT group relative to that in the CD58^KO^ group. “sgCD58-1” and “sgCD58-2” represent two independent single-guide RNAs targeting nonoverlapping regions of the CD58 gene, which were separately used to generate two stable CD58^KO^ tumor cell lines from the same parental wild-type tumor cell line. Reporter cell lines were stimulated with WT or CD58^KO^ Nalm6 cells at a 1:1 E:T ratio. The activation of the reporter cell lines was measured via FACS at 6, 12 and 24 h after the stimulation of the tumor cells (n = 3). The dashed line at a fold change = 1 indicates no difference in activation level between the WT and CD58^KO^ groups. d FACS-based measurement of phospho-c-Jun, phospho-c-Fos, phospho-NF-κB p65, NFAT1, c-Jun, c-Fos and NF-κB p65 expression in CD19 CAR T cells stimulated with WT or CD58^KO^ Nalm6 cells (n = 3). e Phosphorylation protein array data showing the pathway enrichment of significantly altered pathways enriched in CAR T cells stimulated with WT tumor cells compared with those enriched in CAR T cells stimulated with CD58^KO^ tumor cells. Heatmap showing the changes in the phosphorylation of differential phosphosites in the AP-1 signaling pathway. The color gradient indicates the signal intensity. f Efficacy of CD58 knockout and re-expression in Nalm6 cells. An mCherry-only control was used to exclude fluorescence interference. g Activation levels of CD19 CAR-Jurkat AP-1 reporter cells after 6 h of coculture with WT, CD58^KO^ or CD58-reexpressing Nalm6 cells at an E:T ratio of 1:1 (n = 3). Statistical comparisons of the data in panel d were performed via two-way ANOVA. Significance was assessed via a 2-tailed unpaired t test in panel g. *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant (P > 0.05). FACS, fluorescence-activated cell sorting

Loss of tumor CD58 results in impaired mitochondrial quality and impaired function of CAR T cells

Emerging evidence indicates that AP-1 is a critical regulator of mitochondrial metabolism and energy production.^26–29^ Given this, we hypothesized that tumor CD58 deficiency would induce a metabolically compromised CAR T-cell state via AP-1 signaling axis attenuation. To decipher the specific metabolic changes in CAR T cells impaired by CD58^KO^ tumor cells, we established a repeated stimulation model in which CD19 CAR T cells were cocultured with three rounds of WT or CD58^KO^ tumor cells at an E:T ratio of 1:1 (Fig. 2a) to amplify the metabolic differences in CAR T cells between the WT and CD58^KO^ tumor groups. Repeated stimulation by CD58^KO^ tumor cells markedly suppressed the expression of PGC-1α, which positively regulates mitochondrial biogenesis (Fig. 2b). Correspondingly, mitochondrial mass expansion was significantly repressed in CD19 CAR T cells repetitively stimulated with CD58^KO^ tumor cells (Fig. 2c), with similar reductions observed in CD19.28z CAR T cells, tandem CD19/CD20 CAR T cells and CD20 CAR T cells (Fig. 2d–f), suggesting uniformity among various CAR designs. Interestingly, given the low CD20 expression in Nalm6 cells, the mitochondrial impairment in CD20 CAR T cells might reflect a more fundamental effect than one restricted to strong antigen stimulation (Supplementary Fig. 5a). The hindered mitochondrial mass expansion was also reflected in the significantly lower relative mtDNA content in the CD58^KO^ Nalm6-stimulated CAR T cells than in their WT Nalm6-stimulated counterparts (Supplementary Fig. 5b). Moreover, the decrease in CAR T-cell metabolic fitness was directly reflected in the inferior maximal oxygen consumption rates (OCR) and the markedly dropped maximal respiratory of the CAR T cells stimulated with CD58^KO^ tumor cells (Fig. 2g, h). In addition, a reduction in spare respiratory capacity indicated impaired primal metabolic reserve and potential response capacity to cell energy demand upon tumor stimulation (Fig. 2i). A decreased extracellular acidification rate (ECAR) was also observed in CAR T cells repetitively stimulated with CD58^KO^ tumor cells, accompanied by profound suppression of glycolytic activity and capacity (Fig. 2j–l). Concomitantly, ROS accumulation was detected in CAR T cells cocultured with CD58^KO^ tumor cells (Fig. 2m), which was further confirmed by significant transcriptional alterations in ROS detoxification-associated genes (Fig. 3b), suggesting increased cellular oxidative stress as well as impaired mitochondrial function and self-renewal.^30^ Critically, significantly reduced ΔΨm was noted in CAR T cells stimulated with CD58^KO^ tumor cells (Fig. 2n), suggesting compromised mitochondrial function and early-stage apoptosis. Importantly, by re-expressing CD58 in CD58^KO^ tumor cells, both the mitochondrial mass and the membrane potential of CAR T cells could be markedly restored (Supplementary Fig. 5c, d). These findings mechanistically revealed that tumor CD58 loss triggered mitochondrial dysfunction in CAR T cells.Fig. 2. Impact of CD58-deficient tumor cells on mitochondrial quantity and metabolic function in CAR T cells. a Schematic demonstration of the protocol for the repeated stimulation assay. b WB analysis of NT (N-terminal)-PGC-1α expression in CAR T cells that were untreated or treated with WT or CD58^KO^ tumor cells. Three independent experiments were performed, and representative results are shown. FACS analysis of mitochondrial expansion in (c) CD19.BBz CAR T cells, (d) CD19.28z CAR T cells, (e) CD19/20 CAR T cells, and (f) CD20 CAR T cells after three rounds of stimulation with WT or CD58^KO^ tumor cells (n = 3). Mitochondrial expansion was expressed as the fold change compared with that of untreated CAR T cells. Mitochondria quantity was represented by the MFI of MitoTracker^TM^ Deep Red. g Mitochondrial stress assay showing the OCR of CAR T cells repeatedly stimulated with 3 rounds of WT or CD58^KO^ tumor cells, with untreated CAR T cells serving as the control group. h Quantification of maximum respiration. i Quantification of the SRC, calculated as the maximum respiration minus the basal respiration. j Glycolytic rate assay showing the ECAR of CAR T cells repeatedly stimulated with 3 rounds of WT or CD58^KO^ tumor cells, with untreated CAR T cells serving as the control group. k Quantification of glycolysis. l Quantification of glycolytic capacity. m Representative FACS plot and quantification of ROS levels in CAR T cells repeatedly stimulated with WT or CD58^KO^ tumor cells (n = 3). The CellROX® Deep Red reagent was used in conjunction with SYTOX^®^ Blue Dead Cell stain to differentiate live stressed cells from dead cells. n Representative FACS plot and quantification bar plot of JC-1 monomers and aggregates in CAR T cells repeatedly stimulated with WT or CD58^KO^ tumor cells (n = 3). Statistical comparisons were performed via a 2-tailed unpaired t test. The values are presented as the means ± SDs. *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant (P > 0.05). WB, western blot; MFI, mean fluorescence intensity; SRC, spare respiratory capacityFig. 3CD58 deficiency in tumors induces intrinsic apoptosis in CAR T cells via the mitochondria-dependent apoptotic pathway. a FACS was used to measure the percentage of Annexin V^+^ CD19 CAR T cells after three rounds of coculture with WT or CD58^KO^ Nalm6 cells at an E:T ratio of 1:1 (n = 3). b Heatmap demonstrating DEGs related to mitochondrial apoptosis, mitochondrial dynamics and ROS detoxification in CD19 CAR T cells treated with WT or CD58^KO^ Nalm6 cells at an E:T ratio of 1:1 (n = 2 different PBMC donors). c WB analysis of mitochondrial apoptosis-related proteins in CD19 CAR T cells treated with WT or CD58^KO^ tumor cells at a 1:1 E:T ratio. d WB analysis of Bax/Bak translocation in CAR T cells stimulated with WT or CD58^KO^ tumor cells. WB assays were independently repeated three times, and representative blots are shown. Statistical comparisons were performed via a 2-tailed unpaired t test. The values are presented as the means ± SDs. *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant (P > 0.05)

Loss of tumor CD58 induces intrinsic apoptosis in CAR T cells via the mitochondria-dependent apoptotic pathway

Due to the mitochondrial membrane depolarization and ROS accumulation that could lead to apoptosis, we subsequently examined the apoptotic status of CAR T cells sorted after coculture with tumor cells (Supplementary Fig. 6a). Notably, CAR T cells cocultured with CD58^KO^ tumor cells exhibited a robust increase in apoptosis (Fig. 3a). At the transcriptomic level, we also observed a marked concomitant increase in the expression of mitochondrial apoptosis-related genes, suggesting the unanticipated initiation of mitochondrial-dependent cell death circuitry upon tumor CD58 loss (Fig. 3b). We then inspected the expression of mitochondrial apoptosis-related markers at the protein level. A notable decrease in the antiapoptotic protein Bcl-XL and an increase in the proapoptotic protein Bim were detected in CAR T cells stimulated with CD58^KO^ tumor cells, which mobilized Bak and Bax, which mediated mitochondrial apoptosis through cytochrome c release, caspase-9 cleavage, and downstream caspase activation, ultimately resulting in mitochondrial apoptosis (Fig. 3c). Moreover, a significant increase in mitochondrion-localized Bax and Bak, along with a corresponding reduction in their cytosolic counterparts, was observed in CAR T cells stimulated with CD58^KO^ tumor cells, which further confirmed the increase in apoptosis (Fig. 3d). Consistent with these findings, immunofluorescence-based colocalization analysis of cytochrome c and mitochondria revealed increased cytochrome c release in these CAR T cells, further supporting increased apoptotic activity (Supplementary Fig. 6b). In peripheral T cells, apoptosis primarily occurs via extrinsic and intrinsic pathways. To further elucidate the apoptotic mechanism, we investigated the expression levels of the death receptor Fas and Fas-associated death domain (FADD) in CAR T cells treated with WT or CD58^KO^ tumor cells, and no substantial differences were detected in these two key components of the death receptor pathway, indicating death receptor-independent apoptosis (Supplementary Fig. 6c). These findings suggest that tumor CD58 loss creates a permissive landscape for intrinsic apoptosis via the mitochondria-dependent pathway.

Restoration of AP-1 activation reinstates the fitness of CAR T cells

To further validate whether pharmacological blockade of AP-1 signaling could simulate the CAR T-cell mitochondrial dysregulation elicited by CD58-deficient tumor engagement, we blocked AP-1 signaling in CAR T cells with T-5224, which specifically inhibits AP-1 without affecting NFAT and NF-κB signaling (Fig. 4a–c). Similar to CAR T cells treated with CD58^KO^ tumor cells, CAR T cells treated with WT tumor cells in the presence of T-5224 presented decreased mitochondrial quantification and mitochondrial membrane potential loss (Fig. 4d, e). Given that AP-1 is the pathway most significantly affected upon tumor CD58 loss and that pharmacological inhibition of AP-1 in the WT group could simulate the mitochondrial changes in the CD58^KO^ group to a great extent, we next explored strategies to restore CAR T-cell function by targeting the AP-1 pathway.Fig. 4. Modulation of AP-1 signaling influences CAR T-cell efficacy and mitochondrial fitness in CD58-deficient tumor environments. **a–**c Changes in FACS-measured activation levels in AP-1, NFAT or NF-κB reporter CD19 CAR-Jurkat cells in response to T-5224 at 60 μM for 12 h compared with those in WT Nalm6 cells at an E:T ratio of 1:1. d FACS analysis of the impact of T-5224 on the mitochondrial mass in CD19 CAR T cells treated with WT tumor cells compared with that in CD58^KO^ tumor cells. The mitochondrial fold change was calculated on the basis of the mitochondrial quantification of untreated CD19 CAR T cells. e FACS analysis of the impact of T-5224 on the ΔΨm in CD19 CAR T cells treated with WT tumor cells compared with that in CD58^KO^ tumor cells. f Transcriptomic profiling revealed the upregulation of inhibitory phosphatases upstream of AP-1, specifically SHP1, PTPN7, and DUSP6, in CD19 CAR T cells cocultured with CD58^KO^ tumor cells compared with those cocultured with WT tumor cells (n = 2 different PBMC donors). The arrows indicate selected candidate inhibitory phosphatases. g WB analysis of SHP1, PTPN7, and DUSP6 expression in sorted CD19 CAR T cells. CD19 CAR T cells were treated with WT or CD58^KO^ tumor cells at a 1:1 E:T ratio for 24 h. h Impact of SHP1, PTPN7, and DUSP6 inhibition on AP-1 activation evaluated in AP-1 reporter CD19 CAR Jurkat cells stimulated with CD58^KO^ tumor cells at a 1:1 E:T ratio for 12 h (n = 3). The dashed line represents the activation level of AP-1 reporter CD19 CAR Jurkat cells when cocultured with CD58^KO^ tumor cells without phosphatase inhibition. i Phosphorylation of c-Jun in CAR T cells upon inhibitory phosphatase blockade. CD19 CAR T cells were incubated with different concentrations and types of phosphatase inhibitors in CD3-precoated 6-well plates for 6 h. j Impact of SHP1, PTPN7, and DUSP6 inhibition on the CD19 CAR T-cell-mediated cytotoxicity of CD58^KO^ tumor cells at a 1:2 E:T ratio for 6 h. The dashed line represents CAR T-cell-mediated cytotoxicity of CD58^KO^ tumor cells without phosphatase inhibition. k Impact of SHP1, PTPN7, and DUSP6 inhibition on the number of mitochondria in CD19 CAR T cells treated with CD58^KO^ tumor cells at a 1:1 E:T ratio for 24 h. The dashed line represents the quantity of mitochondria in CAR T cells cocultured with CD58^KO^ tumor cells without phosphatase inhibition. Three independent experiments were performed, and representative results are shown. Statistical comparisons were performed via a 2-tailed unpaired t test in panels a-e. Significance was assessed via one-way ANOVA followed by Dunnett’s test in panels h, j, k, with the CD58^KO^ group without inhibitors designated the reference control. The values are presented as the means ± SDs. *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant (P > 0.05). DUSP6i, DUSP6 inhibitor; SHP1i, SHP1 inhibitor; PTPN7i, PTPN7 inhibitor

To clarify whether supplementation with AP-1 signaling contributes to the recovery of CAR T-cell function, we analyzed the transcription levels of key genes in AP-1 and its upstream pathways via bulk sequencing data to screen for possible AP-1 inhibitory factors and found that the loss of tumor CD58 significantly led to the upregulation of a group of inhibitory phosphatases, among which the protein tyrosine phosphatase Src homology 2 domain-containing protein tyrosine phosphatase 1 (SHP1), protein tyrosine phosphatase nonreceptor type 7 (PTPN7) and DUSP6, which target Lck or ERK, showed the most pronounced differences and better concordance across donors (Fig. 4f). Changes in these phosphatases were then verified at the protein level (Fig. 4g), where a marked increase in DUSP6 and marginal increase in SHP1 and PTPN7 were noted. Given this distinction, we selectively blocked these three inhibitory phosphatases via BCI hydrochloride, TPI-1 and MurA-IN-1, which specifically targeted DUSP6, SHP1 and PTPN7, respectively. Blockade of SHP1 and DUSP6 led to significantly increased AP-1 activation in CAR-Jurkat AP-1 reporter cells stimulated with CD58^KO^ tumor cells, whereas PTPN7 blockade, although not significantly, resulted in an increasing trend (Fig. 4h). Moreover, blockade of these three inhibitory phosphatases did not significantly affect NFAT or NF-κB signaling (Supplementary Fig. 7). The phosphorylation of c-Jun was increased in CAR T cells after inhibitory phosphatase blockade, with the most obvious increase triggered by DUSP6 blockade (Fig. 4i). Consistent with the changes in the AP-1 activation level, a concentration-dependent increase in CAR T-cell cytotoxicity to CD58^KO^ tumor cells was observed (Fig. 4j). Concomitantly, the mitochondria expansion was also increased by the application of an inhibitory phosphatase blockade (Fig. 4k), suggesting the amelioration of metabolic dysfunction. The above results further suggest a potential mechanistic link between tumor CD58 loss-induced AP-1 attenuation and the compromised metabolic fitness and effector functions of CAR T cells. In addition, blocking inhibitory phosphatases that target AP-1 signaling might be a feasible approach to invigorate CAR T cells. However, a point to be noted is that the DUSP6 inhibitor BCI did not improve CAR T-cell-mediated tumor lysis at lower concentrations. Prior studies have suggested that DUSP6 inhibition in tumors could reduce oxidative stress and increase survival.^31,32^ Since the inhibitor acts on both CAR T cells and tumors in coculture, tumor-side DUSP6 blockade may partially offset CAR T-cell functional gains from DUSP6 inhibition, which prompted us to generate inhibitory phosphatase-knockout CAR T cells to more accurately evaluate the individual roles of the three inhibitory phosphatases in restoring CAR T-cell function.

Genetic ablation of DUSP6 in CAR T cells renders them resistant to impairment caused by tumor CD58 loss

To further determine whether ablation of specific inhibitory phosphatases could rescue CAR T-cell dysfunction caused by tumor CD58 loss, we knocked out PTPN7, SHP1 and DUSP6 via the CRISPR/Cas9 system and detected the effects of their ablation on CAR T-cell functions (Fig. 5a). Successful ablation of each phosphatase was confirmed at the protein level (Fig. 5b). Elevated Ki-67 expression observed on day 4 following electroporation implicated augmented proliferative capacity in CAR T cells with DUSP6 ablation (Fig. 5c). Next, we used a repeated tumor stimulation model to assess the effects of SHP1, PTPN7, and DUSP6 ablation. Remarkably, DUSP6^KO^ CAR T cells exhibited robust expansion when cocultured with CD58^KO^ tumor cells (Fig. 5e). Correspondingly, DUSP6^KO^ CAR T cells also manifested a notable advantage in alleviating the impairment of mitochondrial expansion induced by CD58 deficiency in tumor cells, indicating increased metabolic fitness (Fig. 5f). After repeated tumor challenge, we sorted and rechallenged control and phosphatase knockout CAR T cells with WT tumor cells at a lower E:T ratio to assess their cytotoxicity. Compared with control CAR T cells, all phosphatase knockout CAR T cells preexposed to CD58^KO^ tumors maintained superior killing of WT Nalm6 cells (Supplementary Fig. 8). DUSP6^KO^ and SHP1^KO^ also increased the CD107a level in CAR T cells stimulated with CD58^KO^ tumor cells (Supplementary Fig. 9). Although this increase has not yet reached that of the CAR T cells in the WT tumor group, it still indicates an improvement in degranulation capacity.Fig. 5DUSP6 ablation mitigates CD58 loss-mediated resistance to CAR T-cell therapy by increasing functional activation, mitochondrial fitness, and resistance to apoptosis. a Schematic of electroporation-based CRISPR-Cas9 gene editing in CAR T cells. b Western blot validation of the protein-level knockout efficiency of SHP1, PTPN7 and DUSP6. c FACS-based measurement of Ki-67 expression in CD19 CAR T cells 4 days post-electroporation. d Schematic of the repeated tumor challenge model used to assess the impact of SHP1, PTPN7 and DUSP6 ablation on CD19 CAR T-cell expansion and increased mitochondrial mass. Ctrl, SHP1,^KO^ PTPN7^KO^ and DUSP6^KO^ CD19 CAR T cells were primed with 3 rounds of WT or CD58^KO^ Nalm6 cells at a fixed E:T ratio of 1:1. e CD19 CAR T-cell expansion was measured after each round of tumor challenge. Statistical analysis was performed via repeated-measures two-way ANOVA. f FACS analysis of the impact of SHP1, PTPN7 and DUSP6 ablation on mitochondrial expansion in CD19 CAR T cells treated with 3 rounds of CD58^KO^ Nalm6 cells. The mitochondrial fold change was calculated on the basis of the mitochondrial quantification of untreated CD19 CAR T cells. The auxiliary line indicates the mitochondrial fold change in CD58^KO^ tumor cells repeatedly stimulated with Ctrl CAR T cells. g Schematic of the multiround cytotoxicity assay with a fixed number of target cells. h Ratio of tumor cells to CAR T cells measured after three rounds of tumor stimulation (n = 3). i FACS analysis of the impact of DUSP6 ablation on the activation phenotype of CD19 CAR T cells treated with WT or CD58^KO^ Nalm6 cells at a 1:1 E:T ratio for 24 h (n = 3). j Representative FACS plot and quantification bar plot of the degree of apoptosis (annexin V/7-AAD staining) in Ctrl and DUSP6^KO^ CAR T cells repeatedly stimulated with WT or CD58^KO^ tumor cells for 3 rounds at a 1:1 E:T ratio (n = 3). k FACS-based measurement of JC-1 monomers in Ctrl and DUSP6^KO^ CAR T cells repeatedly stimulated with WT or CD58^KO^ Nalm6 cells for 3 rounds at a 1:1 E:T ratio (n = 3). l Western blot analysis of proapoptotic Bim and antiapoptotic Bcl-XL expression. Ctrl or DUSP6^KO^ CAR T cells were sorted by flow cytometry after coculture with WT or CD58^KO^ tumor cells. Three independent experiments were performed, and representative results are shown. Statistical comparisons were performed via a 2-tailed unpaired t-test. n = 3 donors in triplicate. The values are presented as the means ± SDs. *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant (P > 0.05). RNP, CRISPR-Cas9 ribonucleoprotein; 7-AAD, 7-Aminoactinomycin D

On the basis of the above results, we focused on DUSP6^KO^ CAR T cells for further functional investigations. To provide a more holistic assessment of cytotoxic potential, we then performed a multiround cytotoxicity assay with a fixed number of target cells to account for both CAR T-cell expansion capacity and acute cytotoxic capacity (Fig. 5g). Following three rounds of stimulation with CD58^KO^ tumor cells, the tumor-to-CAR T-cell ratio in the Ctrl CAR T-cell group was nearly threefold greater than that in the DUSP6^KO^ group, demonstrating the cytotoxicity advantage of DUSP6 ablation (Fig. 5h). While this effect was slightly attenuated with WT tumor cells, DUSP6^KO^ CAR T cells still showed significantly increased killing capacity compared with that of the controls. Notably, the enhanced tumor clearance capacity was accompanied by improvements in cellular activation, with DUSP6 ablation significantly augmenting CAR T-cell activation upon tumor cell stimulation (Fig. 5i). Specifically, AP-1 activation was improved in DUSP6^KO^ CD19 CAR-Jurkat reporter cells when they were stimulated with CD58^KO^ tumor cells (Supplementary Fig. 10a, b). Improved granzyme B, IFN-γ, TNF-α, and IL-2 levels were also observed in DUSP6^KO^ CAR T cells cocultured with CD58^KO^ tumor cells (Supplementary Fig. 10c, d). Furthermore, DUSP6^KO^ CAR T cells exhibited reduced apoptosis (Fig. 5j), and importantly, DUSP6 ablation potently alleviated the ΔΨm damage induced by tumor CD58 loss, indicating improved mitochondrial function and metabolic fitness (Fig. 5k), which was also verified in different tumor cell lines (Supplementary Fig. 11). The improved mitochondrial function of DUSP6^KO^ CAR T cells was also reflected in the increased maximal OCR and SRC following repeated stimulation with CD58^KO^ tumor cells (Supplementary Fig. 10e–g). Moreover, elevated expression of the antiapoptotic protein Bcl-XL and decreased expression of the proapoptotic protein Bim further demonstrated increased mitochondrial fitness in DUSP6^KO^ CAR T cells (Fig. 5l). Collectively, these findings underscore that DUSP6 ablation is an effective approach to complement AP-1 activation signaling, improve mitochondrial function and thus mitigate CD58 loss-induced CAR T dysfunction.

DUSP6 ablation mitigates CD58 loss-induced CAR T-cell therapy resistance in vivo

To further examine whether DUSP6^KO^ CAR T cells could restore the functions of CAR T cells dampened by tumor CD58 loss in vivo, we evaluated the tumor-suppressive ability of equal amounts of DUSP6^KO^ or Ctrl CAR T cells in mice transplanted with WT or CD58^KO^ tumor cells (Fig. 6a). In CD58^KO^ tumor-bearing mice, the tumor clearance ability of DUSP6^KO^ CAR T cells was superior to that of Ctrl CAR T cells in WT tumor-bearing mice (Fig. 6b, c). In addition, DUSP6^KO^ CAR T cells prolonged survival in CD58^KO^ tumor-bearing mice (Fig. 6d). An increase in survival was also observed in WT tumor-bearing mice that received DUSP6^KO^ CAR T cells, indicating a potential role of DUSP6 ablation in enhancing the therapeutic efficacy of CAR T cells at a broader level. Consistently, DUSP6^KO^ CAR T cells proliferated significantly better than control CAR T cells did in both tumor models (Fig. 6e). Moreover, ameliorated cell exhaustion, as measured by the coexpression of TIM3 and LAG3, was found in DUSP6^KO^ CAR T cells in the absence of tumor CD58 (Fig. 6f). Elevated IFN-γ release was also observed in DUSP6^KO^ CAR T cells in the CD58^KO^ tumor group (Fig. 6g). No significant weight loss or body temperature change was observed (Supplementary Fig. 12). To further delineate the role of DUSP6 in the context of tumor relapse, we administered a higher dose of CAR T cells to CD58^KO^ tumor-bearing mice (Supplementary Fig. 13a). On day 21 postinfusion, mice in the Ctrl group exhibited prominent tumor recurrence compared with those in the DUSP6^KO^ group (Supplementary Fig. 13b). Survival analysis via the Kaplan‒Meier method revealed that CD58^KO^ tumor-bearing mice in the DUSP6^KO^ group had a significantly prolonged survival duration relative to those in the Ctrl group (58 days vs. 42 days), with 40% (2/5) of the mice remaining alive (Supplementary Fig. 13c). Additionally, on day 21 post infusion, the DUSP6^KO^ group showed superior CAR T-cell expansion efficiency and attenuated T-cell exhaustion compared with the Ctrl group (Supplementary Fig. 13d, e). Taken together, these findings indicate that DUSP6 ablation in CAR T cells not only renders them resistant to impairment caused by tumor CD58 loss in vivo but also endows CAR T cells with improved cytotoxicity and persistence, providing a promising approach to overcome resistance to CAR T-cell treatment.Fig. 6. Genetic ablation of DUSP6 enhances CAR T-cell resistance to tumor CD58 loss in vivo. a Schematic of the xenograft model for evaluating CAR T-cell efficacy. NPG mice were intravenously engrafted with 10⁵ WT or CD58^KO^ Nalm6 cells. Seven days after tumor implantation, the mice received 10⁶ Ctrl or DUSP6^KO^ CD19 CAR T cells intravenously (n = 5 per group). BLI was performed weekly to monitor the tumor burden. b BLI images of tumor progression at the indicated time points. c Quantification of the tumor burden according to the average radiance over time. d Survival analysis of mice treated with NT cells, Ctrl CAR T cells or DUSP6^KO^ CAR T cells. e FACS-based measurement of CAR T-cell expansion in peripheral blood on day 7 postinfusion. f FACS-based measurement of TIM3⁺ LAG3⁺ CAR T cells on day 7 after CAR T-cell infusion. g IFN-γ levels in the peripheral blood on day 7 after CAR T-cell infusion were measured via enzyme-linked immunosorbent assays (n = 5). Statistical comparisons were performed via a 2-tailed unpaired t-test; survival between groups was analyzed via the log-rank test. The values are presented as the means ± SDs. *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant (P > 0.05). BLI, bioluminescence imaging

DUSP6 ablation reinvigorates CAR T-cell fitness through augmented activation and metabolic reprogramming

To decipher the molecular mechanism underlying the improved function of CAR T cells caused by DUSP6 knockout, we performed protein‒protein interaction (PPI) network analysis via the STRING database (version 12.0). PPI analysis revealed that DUSP6 was strongly correlated with MAPK1 and MAPK3 and with Jun and Fos (Supplementary Fig. 14). On the basis of these preliminary results, we sorted tumor-stimulated Ctrl and DUSP6^KO^ CAR T cells for bulk RNA sequencing. Enrichment analysis revealed a succession of altered biological pathways related to the regulation of cell activation, energy metabolism, cytokine-related signaling, cell adhesion, cell division and mitochondrial organization (Fig. 7a, Supplementary Fig. 15). Notably, significant enrichment of positive regulation of the MAPK cascade was observed in DUSP6^KO^ CAR T cells, suggesting overall upregulation of CAR T-cell activation and mechanistically linking DUSP6 ablation to increased AP-1 transactivation through this signaling axis. (Fig. 7b). Critically, DUSP6 ablation promoted robust upregulation of antiapoptotic signatures in CAR T cells, indicating improved persistence. These results position DUSP6 as a promising molecular orchestrator that modulates CAR T-cell activation, metabolic fitness and persistence.Fig. 7. Enriched pathways related to cell activation and energy metabolism were found in DUSP6^KO^ CAR T cells. a Gene ontology enrichment analysis of the transcriptomic profiles of Ctrl or DUSP6^KO^ CAR T cells after CD58^KO^ Nalm6 challenge. Ctrl or DUSP6^KO^ CAR T cells were sorted by flow cytometry after stimulation with CD58^KO^ Nalm6 cells at a 1:1 E:T ratio. b GSEA highlighting key signaling pathway alterations in DUSP6^KO^ CAR T cells versus Ctrl CAR T cells when stimulated with CD58^KO^ Nalm6 cells. c Six distinct transcriptional trajectories identified by the Mfuzz series test of cluster analysis across three experimental conditions (Ctrl CAR_WT, Ctrl CAR_CD58,^KO^ and DUSP6^KO^ CAR_CD58.^Ko)^ d Bubble plot displaying the KEGG enrichment results for Cluster 4 and Cluster 6. The circle diameter corresponded to the gene count (number of genes mapped to the pathway), whereas the color intensity reflected the FDR-adjusted P value (Benjamini‒Hochberg method). The rich factor was calculated as the ratio of the number of differentially expressed genes (from Cluster 4 or 6) annotated to a specific KEGG pathway to the total number of genes. e KEGG enrichment analysis of the transcriptomic profiles of DUSP6^KO^ or Ctrl CAR T cells after WT Nalm6 challenge. Ctrl or DUSP6^KO^ CAR T cells were sorted by flow cytometry after stimulation with WT Nalm6 cells at a 1:1 E:T ratio. f GSEA highlighting key signaling pathway alterations in DUSP6^KO^ CAR T cells compared with Ctrl CAR T cells when stimulated with WT Nalm6 cells. n = 3 PBMC donors. GSEA, gene set enrichment analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; FDR, false discovery rate

We then performed an Mfuzz series test of clusters to dissect the transcriptional landscape across three cellular states (Ctrl CAR_WT, Ctrl CAR_CD58,^KO^ and DUSP6^KO^ CAR_CD58^KO^). This approach leverages fuzzy C-means clustering to systematically identify the gene modules exhibiting temporal expression synchronism, thereby elucidating the molecular programs underlying DUSP6-mediated functional augmentation in CAR T-cell immunotherapy. Six distinct transcriptional trajectories were identified across the three experimental conditions, with Clusters 4 and 6 representing genes that were simultaneously downregulated and upregulated in the Ctrl CAR_WT and DUSP6^KO^ CAR_CD58^KO^ groups, respectively (Fig. 6c). Enrichment analysis revealed suppressed catabolic processes concurrent with upregulated T-cell activation signatures and enhanced metabolic pathways (Fig. 7d). These coordinated transcriptomic shifts mechanistically explain how DUSP6 ablation rescues CAR T-cell functional impairment.

Additionally, compared with their control counterparts, DUSP6^KO^ CAR T cells also demonstrated persistent activation and metabolic kinetic signatures in CD58-sufficient tumor microenvironments, with concomitant enrichment of signaling pathways regulating pluripotency stem cells, suggesting that long-term tumor control is maintained (Fig. 7e, f). Taken together, these findings indicate that DUSP6 ablation reinvigorated CAR T cells by increasing the activation of functional signaling pathways and increasing overall metabolic kinetics.

Clinical relevance of DUSP6 transcriptomic levels to T-cell-based immunotherapy

To assess the clinical relevance of DUSP6 in CAR T-cell therapy, we analyzed the transcriptomic profiles of premanufactured T cells and CAR T cells from our tandem CD19/20 CAR T-cell-treated R/R DLBCL cohort.^33^ The DUSP6^low^ subgroup of CD8^+^ CAR T cells presented a remarkably greater proportion of patients with a complete response (CR) of 80% than did the DUSP6^high^ subgroup, whereas little difference in the CR rate was detected between the two subgroups of CD4^+^ CAR T cells (Fig. 8a). Consistent with CD8^+^ CAR T cells, substantially greater CR rates were also observed in the DUSP6^low^ subgroup of both CD8^+^ and CD4^+^ non-naïve premanufactured T cells (Fig. 8b).Fig. 8DUSP6 could be negatively correlated with T-cell immunotherapy outcome. a Proportion of PD and CR patients in the low- and high-DUSP6 subgroups of CD8^+^ and CD4^+^ CAR T cells collected 10 days post-infusion from our tandem CD19/20 CAR T-cell-treated R/R diffuse large B-cell lymphoma cohort (GSE223655). The classification of high and low DUSP6 expression in patient samples was determined by the median expression value within each sample group. b Proportion of PD patients and CR patients in the low- and high-DUSP6 subgroups among premanufactured T cells from the same cohort. c UMAP visualization of T cells from an anti-PD-1-treated skin cancer cohort (GSE123813). The cells are color coded according to inferred cell types. d UMAP visualization of T cells color coded according to treatment response. e Expression of DUSP6 overlaid onto the UMAP plot. f Heatmap showing average DUSP6 expression in different T-cell clusters from patients who were resistant or responsive to treatment. g Violin plots showing DUSP6 expression (normalized counts) in exhausted CD8 + T cells from patients who were resistant or responsive to treatment. h Summary of findings. BioRender (https://www.biorender.com) was used to generate part of the figure. The statistical significance of (a-b) was calculated via Fisher’s exact test; the statistical significance between two unpaired sample groups was calculated via an unpaired two-tailed Wilcoxon‒Mann‒Whitney U test. ***P < 0.001, ****P < 0.0001; ns, not significant (P > 0.05). PD, progressive disease; CR, complete response

Building on these findings in CAR T-cell therapy, we sought to determine whether the role of DUSP6 extends to other immunotherapeutic contexts. Interrogating single-cell RNA-seq data from an anti-PD-1-treated skin cancer cohort,^34^ we performed clustering and visualization via uniform manifold approximation and projection on scRNA-seq data stratified by cell subset, treatment response and DUSP6 expression (Fig. 8c–e). Notably, a pronounced overlap between the exhausted CD8^+^ T-cell cluster, PD-1 therapy-resistant cell cluster, and DUSP6^high^ cluster was observed, suggesting that DUSP6 is upregulated in exhausted CD8^+^ T cells from patients resistant to anti-PD-1 treatment. Subsequent statistical analyses confirmed that the DUSP6 expression level in exhausted CD8^+^ T cells was significantly lower in responders than in nonresponders (Fig. 8f, g). These cross-modal clinical datasets not only reinforce DUSP6 as a potential negative prognostic biomarker of T-cell effector competence across immunotherapy modalities but also strengthen its candidacy as a therapeutic target for improving treatment durability.

Discussion

Despite the established role of CD58 in immune synapse formation and membrane signaling,^7,15,35^ its regulatory hierarchy over CAR T-cell effector functions, specifically through transcription factor networks to sustain cytotoxicity, metabolic plasticity, and long-term persistence, remains inadequately characterized. In this study, we addressed a pivotal but yet unknown mechanism by which tumor CD58 preserves metabolic fitness and long-term antitumor ability in CAR T cells by regulating the AP-1 transcriptomic pathway. For solutions, mechanistic validation has revealed that DUSP6 ablation is a salvage approach to restore AP-1 signaling and rescue CAR T-cell cytotoxicity. Crucially, DUSP6 ablation also enhanced CAR T-cell proliferation, metabolic fitness and long-term antitumor activity against wild-type tumors, indicating a dual-purpose therapeutic strategy. These findings are summarized in Fig. 8h.

The dual role of the CD58-CD2 interaction in adhesion and costimulation is vital for IS formation and T-cell activation, orchestrating intricate signaling networks that modulate immune responses. CD58-CD2 signaling is known to assist in TCR activation. Upon binding to CD58, intracellular kinases such as Lck and Fyn are recruited, which assist in the phosphorylation of the T-cell receptor complex and downstream of ZAP70, LAT, and SLP76.^20,36^ However, the MHC-independent activation mode and structural specificities of CAR T cells endow them with a unique activation paradigm. In this study, we extended our previous findings by delineating the specific transcriptional networks impacted by tumor CD58 that modulate metabolic plasticity and long-term antitumor activity in CAR T cells. Tumor CD58 loss caused a significant loss of AP-1 signaling rather than NFAT or NF-κB signaling, and this selective defect in AP-1 signaling led to metabolic failure and long-term damage. Moreover, tumor CD58 loss creates a permissive landscape for intrinsic apoptosis, providing new insights into the current understanding of CAR T-cell death.

The AP-1 transcription factor family is implicated in a spectrum of cellular processes.^37^ As a heterogeneous mixture of dimers from multiple distinct families, AP-1 can act in a context-dependent way and function as a bidirectional nuclear decision-making signal, dictating cell fate toward survival or death in response to extracellular stimuli, although the regulatory mechanisms governing this signaling balance remain partially understood.^27^ In T cells, AP-1 collaborates with NFAT signaling to drive the expression of effector-related genes following immune activation, whereas AP-1 deficiency abrogates T-cell activation and proliferation, culminating in T-cell anergy.^38,39^ Recent studies have illuminated the impact of AP-1 family members on CAR T-cell function. Lynn et al. demonstrated that dysregulation of AP-1-associated bZIP/IRF signaling in CAR T cells impairs Fos/Jun heterodimer function, thereby driving exhaustion programs; conversely, c-Jun overexpression enhances long-term antitumor activity and prevents exhaustion in multiple tumor models.^40^ While the overexpression of BATF, another AP-1 family member, has been shown to enhance CAR T-cell persistence and reduce exhaustion,^41^ the genetic deletion of BATF was reported to confer similar benefits by suppressing exhaustion markers in a distinct experimental model.^42^ Collectively, these divergent findings highlight the complexity of AP-1 in CAR T-cell antitumor responses. Notably, our study revealed that the loss of tumor CD58 results in profound attenuation of AP-1 transcriptional activity in CAR T cells. This discovery uncovered a previously unreported source of AP-1 signaling during CAR T-cell activation, expanding our understanding of AP-1 regulatory mechanisms. While our study proposed restoring AP-1 signaling to recover CAR T-cell function under CD58 deficiency, another interesting study by Wang et al. revealed that MEK1/2 inhibitors reduce c-Fos/JunB to inhibit exhaustion.^43^ In our work, AP-1 is enhanced to supplement deficient signals, whereas Wang et al. focused on aberrant AP-1 overactivation in 28z CAR T cells from sustained stimuli. These seemingly opposing conclusions are actually not contradictory but collectively illustrate the dual role of AP-1, as it varies in activation intensity, duration, and the composition of its downstream dimeric complexes—rather than being a unidirectional “profunction” or “proexhaustion” signal. These studies have also emphasized the common objective of the precision regulation strategy of AP-1 activity, which is to support effector function while avoiding exhaustion.

The increasing recognition of the ability of CD58 to confer resistance to T-cell-based immunotherapies has driven the development of precision strategies aimed at counteracting its functional loss. Focusing on reconstituting CD58 expression, Otsuka et al. suggested the use of EZH2 inhibitors to reverse epigenetically silenced CD58 in lymphoma cells.^44^ A recent study employed recombinant CD58-Fc chimeric protein with immune checkpoint inhibitors to reverse T-cell dysfunctions.^12^ Although they may effectively compensate for CD58 expression, these strategies may risk eliciting unintended systemic adverse effects due to the ectopic activation of proinflammatory pathways or aberrant activation of bystander T cells. Other investigative efforts have targeted effector cell engineering to compensate for deficient CD58-CD2 immune synaptic communication, as exemplified by the integration of the CD2 costimulatory domain within the CAR construct.^16^ However, this design may restrict signal transduction efficiency due to steric hindrance. Moreover, the integration of single-molecule signals may lead to T-cell exhaustion or an inability to adapt to dynamic antigenic stimulation in the complex in vivo microenvironment. A recent study of our group demonstrated that CD2 overexpression could enhance the persistence of CAR T cells.^45^ However, this strategy did not augment T cell activation signaling in the context of CD58-deficient tumors. Another interesting study developed a PD-1-CD2 switch receptor to bypass CD58 loss, as the extracellular PD-1 domain binds to tumor-expressed PD-L1, whereas the intracellular CD2 domain triggers CD2 signaling upon CAR engagement.^17^ The switch receptor enhances the stability of the IS and cytokine secretion, yet its efficacy is dependent on PD-L1 expression in tumor cells. In this study, we propose that DUSP6 ablation in CAR T cells is an effective strategy. DUSP6 is a member of the DUSP family with dephosphorylation activity that targets serine/threonine and tyrosine residues and plays a key role in regulating the MAPK signaling pathway, including the activities of ERK, JNK, and p38.^46^ A previous study reported increased IFN-γ secretion and mitochondrial respiration in T cells from DUSP6^-/-^ mice.^47^ Moreover, increased DUSP6 activity was proposed as a signature of age-related T-cell senescence,^48^ with DUSP6 repression improving the T-cell response.^49^ These findings are in accordance with our findings of increased activation and proliferation in DUSP6^KO^ CAR T cells. We also established a link between T-cell DUSP6 expression and immunotherapy outcome. High levels of DUSP6 are observed in patients with resistance, suggesting that DUSP6 is a promising target for the therapeutic rejuvenation of impaired T cells. Collectively, these findings underscore the potential relevance of DUSP6 to T-cell functionality and the efficacy of T-cell-based immunotherapies.

Despite these findings, there are still limitations to acknowledge. While T-cell activation generally leads to the upregulation of DUSP6 to prevent overactivation, we observed marked DUSP6 elevation in CAR T cells upon tumor CD58 loss, where overall CAR T-cell activation was notably reduced. This finding suggests that CD58 might provide a crucial signal in regulating DUSP6 expression, potentially indicating a mechanism that operates independently of the well-established negative feedback loop. However, the underlying mechanisms of this regulation remain unclear. Importantly, the current results indicate that the improvement in CAR T-cell function induced by DUSP6 ablation is slightly more pronounced in CD58^KO^ tumors than in WT tumors. We speculate that this may be partially because, in WT tumors, although DUSP6 ablation still enhances CAR T-cell activation and function, the AP-1 signaling pathway might quickly reach a plateau, thus preventing further improvement in CAR T-cell function. Therefore, DUSP6 ablation could rescue the downregulation of AP-1 transcriptional activity caused by CD58 deficiency, but it is unclear whether it can compensate for other potential signal inhibition resulting from CD58 loss. Furthermore, since DUSP6 knockout could enhance the sustained antitumor efficacy of CAR T cells, our future studies will focus on validating this sensitization strategy in CAR T cells that target solid tumors to establish its broader applicability.

Overall, our study underscores the vital but relatively understudied role of tumor CD58, which not only affects proximal activation signals by impairing IS formation but also profoundly affects CAR T-cell fitness and persistence by attenuating AP-1-dominated transcription signaling. This impaired signaling significantly disrupted metabolic fitness and initiated mitochondria-dependent intrinsic apoptosis in CAR T cells. Furthermore, we propose that DUSP6 ablation in CAR T cells is an effective approach to alleviate tumor CD58 loss-induced CAR T-cell dysfunction and, most importantly, to invigorate CAR T cells in a general context.

Materials and methods

Cells and ethics

The cell lines were purchased from ATCC (Manassas, VA) and cultured in RPMI-1640 (Gibco) supplemented with 10% fetal bovine serum (Gibco) and 100 U/ml penicillin/streptomycin (Gibco). Cell line identity was confirmed via short tandem repeat profiling at the time of acquisition, and mycoplasma contamination was routinely tested via PCR-based assays. Peripheral blood mononuclear cells (PBMCs) were obtained from healthy donors, and written informed consent was obtained. All procedures were conducted in accordance with the Declaration of Helsinki and were approved by the Research Ethics Board of Chinese PLA General Hospital.

CAR T-cell generation

The generation of CAR constructs has been described in our previous reports.^50,51^ The CAR constructs used in this study are displayed in Supplementary Fig. 1. Unless specifically stated otherwise, CD19 CAR T cells in this study refer to CD19–4-1BBζ CAR T cells. The CAR construct was subsequently cloned and inserted into a pRRLSIN lentiviral vector under the control of the EF-1α promoter. Lentiviral particles were produced in 293 T cells via psPAX2/pMD2G packaging plasmids. T cells were activated with 1 mg/mL OKT3 in IL-2-supplemented X-VIVO 15 medium (Lonza) for 2 days and then transduced with viral supernatant in RetroNectin (Takara)-coated plates with 4 μg/mL polybrene (Sigma–Aldrich, USA) at 2000 × g for 2 h. Mock-transduced T cells served as controls. CAR T cells were labeled with biotinylated anti-mouse IgG F(ab)’2 (Jackson ImmunoResearch) and PE-streptavidin (BioLegend) and sorted with magnetic beads (Miltenyi) following the manufacturer’s instructions. Functional assays were conducted 8–10 days post-transduction. All CAR T cells used in this study were purified with CAR efficiency higher than 90% (Supplementary Fig. 1).

CRISPR/Cas9-mediated knockout of genes

CRISPR/Cas9-mediated gene editing was performed by electroporation of Cas9/gRNA ribonucleoprotein (RNP) complexes. Two different sgRNA (referred to as sgRNA-1 and sgRNA-2) were designed and tested for SHP1, PTPN7, DUSP6, respectively. For each target gene, the two corresponding sgRNAs were pooled and used simultaneously during RNP complex formation to improve knockout efficiency. The RNP complex was prepared by mixing 0.9 μL of 62 μM Cas9 protein (Integrated Device Technology) with each 0.3 μL of 100 μM sgRNA-1 and sgRNA-2 into Opti-MEM medium (Gibco) to create a final mixture of 20 μL. The mixture was incubated at room temperature for 20 min for RNP complex formation, then immediately transfer the mixture onto ice until electroporation. A total of 10^6^ activated T cells were washed twice in 3 mL of Opti-MEM medium and the supernatant was carefully removed. The cell pellet was resuspended in 80 μL of Opti-MEM medium, then kept on ice for at least 5 min before electroporation. The cell suspension and RNA complex were immediately transferred into a pre-cooled (4 °C) 2 mm cuvette (Harvard Apparatus BTX), then electroporated with a BTX Gemini System (Harvard Apparatus BTX) at 250 V with a pulse length of 5 ms. Electroporated T cells were immediately and gently transferred into 2 mL of prewarmed medium and cultured at 37 °C. CAR transduction was performed 24 h after electroporation. Gene knockout efficiency was confirmed by western blot analysis. The gRNA sequences targeting SHP1, PTPN7, DUSP6 and negative control gRNA sequence (used to generate control-edited T cells) were listed in Supplementary Table 1.

Generation of reporter cell lines for NF-κB, NFAT, or AP-1 signaling

The responsive promoter sequences of NFAT, NF-κB, and AP-1 were derived from Addgene plasmids #118031, #118094, and #118095 (deposited by Peter Steinberger; Addgene, Watertown, MA, USA), respectively. mCherry fragments were inserted into these sequences to generate NFAT-mCherry, NF-κB-mCherry, and AP-1-mCherry fluorescent reporter constructs. The PGK-GFP fragment was excised from the lentiviral backbone vector pRRLSIN.cPPT.PGK-GFP.WPRE using BstXI and SalI, and replaced with above reporter constructs to generate pRRLSIN.cPPT.NFAT-mCherry.WPRE, pRRLSIN.cPPT.NF-κB-mCherry.WPRE, and pRRLSIN.cPPT.AP-1-mCherry.WPRE, respectively. Lentiviruses were packaged in 293 T cells by co-transfecting the recombinant plasmids with psPAX2 and pMD2.G at a mass ratio of 6:2:1 using Lipofectamine3000 (Invitrogen), followed by collection and filtration of viral supernatants 48 h post-transfection. For cell transduction, 6-well plates were pre-coated with RetroNectin solution, and 1.5 mL of viral stock was added to each well, followed by centrifugation at 2000 g for 2 h at 32 °C. Without removing the viral supernatant, 2 × 10⁶ Jurkat cells or CD19 CAR-Jurkat cells resuspended in 1.5 mL RPMI 1640 complete medium containing 3 μL polybrene (8 μg/mL) were gently added to each well, centrifuged at 1000 rpm for 10 min at 32 °C, and then cultured in a 37 °C incubator with 5% CO₂, with fresh medium replaced after 20 h. Cell stimulation cocktail containing phorbol 12-myristate 13-acetate and ionomycin (Invitrogen) was used to confirm signaling responsiveness.

Jurkat reporter cell lines were transduced with the NY-ESO-1 specific TCR (clone 1G4) to establish the TCR reporter systems. Transduced cells were subsequently purified via flow cytometric sorting based on TCR expression. For target cell preparation, OVCAR3 tumor cells were pulsed with 1 µM SLLMWITQC peptide overnight at 37 °C prior to coculture. To compare TCR reporter cell activation by CD58-deficient versus WT tumor cells, peptide-pulsed OVCAR3 cells were incubated with an isotype control mAb or anti-CD58 blocking mAb (Biolegend) at 8 μg/mL for 1 h in the dark. After washing with PBS, flow cytometry using a detection antibody of the same clone confirmed complete and sustained blockade of membrane CD58 throughout the duration of subsequent coculture experiments.

Generation of CD58 Re-expression Cell Line

To generate the CD58 re-expression cell line, a lentiviral expression plasmid encoding human CD58 together with an mCherry fluorescent reporter was constructed for CD58 overexpression (OE). Lentiviruses were packaged by co-transfecting the recombinant CD58-OE-mCherry plasmid with psPAX2 and pMD2.G packaging plasmids into 293 T cells using Lipofectamine 3000 (Invitrogen) according to the manufacturer’s protocol. Viral supernatants were collected at 48 post-transfection and filtered through a 0.45 μm membrane. The transduction protocol applied to CD58^KO^ Nalm6 cells was identical to that utilized for the establishment of reporter cell lines in the preceding section. At 72 h post-infection, mCherry-positive cells were enriched by flow cytometric sorting (BD Biosciences) with gating on mCherry fluorescence. The sorted population was expanded, and successful CD58 re-expression was verified by flow cytometry using an anti-CD58 antibody (BioLegend). This validated cell line was designated as Nalm6-CD58^KO^-CD58^OE^-mCherry.

Quantification of mtDNA

CAR T cells repeatedly stimulated by WT or CD58^KO^ Nalm6 cells were sorted and had total DNA extracted by TIANamp Genomic DNA Kit (TIANGEN), according to the manufacturer’s protocol. The relative abundance of mtDNA was quantified by real-time PCR. Reactions were set up in triplicate using eight-tube strips. Each PCR reaction (final volume 20 μL) contained 200 ng DNA, 10 μL of Taq Pro Universal SYBR qPCR Master Mix (Vazyme) and 0.8 μM of each forward and reverse primer. mtDNA was quantified using two primer sets specific for MT-ND1 and MT-CO2 gene, and primers sets for reference gene ACTB were used for normalization. Primers were as follows: MT-ND1-F (CCCTAAAACCCGCCACATCT), MT-ND1-R (GAGCGATGGTGAGAGCTAAGGT), MT-CO2-F (CGTCTGAACTATCC TGCCCG), MT-CO2-R (TGGTAAGGGAGGGATCGTTG), ACTB-F (TTGCCGACAGGATGCAG), and ACTB-R (AGGTGGACAGCGAGGCC).

Flow cytometry analysis

To evaluate apoptosis in CAR T cells, Annexin V/7-AAD staining was conducted per the manufacturer’s protocol (BD Biosciences apoptosis detection kit).

To evaluate CAR T cell degranulation, CAR T cells were co-cultured with tumor cells (E:T = 1:1) in 1640 medium containing anti-human CD107a-APC (BioLegend) for 1 h, followed by 3 h incubation with BD GolgiPlug protein transport inhibitor (BD Biosciences).

To measure the proliferation of CAR T cells, an amount of 10⁵ CAR T cells were seeded into 12-well plates. Fresh 1640 medium was replenished every 48 h, followed by trypan blue staining at indicated time points to quantify viable cells.

To evaluate tumor cell lysis, 5 × 10⁴ tumor and CAR T cells were co-cultured in 96-well plates (200 μl/well) at diverse E:T ratios. Post-incubation, 10 μL of 2 × D-luciferin (300 μg/ml) was injected, and luminescence signals (Varioskan™ LUX) were recorded after 2–5 min. Lysis rate = 1-([sample]–[negative])/([positive]–[negative]).

ROS level was measured using the CellROX^®^ Deep Red Flow Cytometry Assay Kit (Invitrogen) following the manufacturer’s instructions. CAR T cells were collected and resuspended at a density of 5 × 10^5^ cells/mL in complete medium, with TBHP and NAC from the kit used for positive and negative controls, respectively. CellROX^®^ Deep Red reagent was applied at a final concentration of 500 nM and incubated for 30 min at 37 °C in dark. SYTOX^®^ Blue Dead Cell Stain was introduced during the last 15 min of staining at a concentration of 1 μM, followed by immediate flow cytometry analysis.

The protocol recommended by the manufacturer guided all experimental procedures. Multicolor flow cytometry analysis of all samples was performed on a DxFLEX flow cytometer (Beckman Coulter), and subsequent data analysis was conducted using FlowJo software v 10.0 (FlowJo LLC). Antibodies used were described in Supplementary Table 2.

Mitochondrial quantification and membrane potential analysis

Mitochondrial quantification was performed via MitoTracker^®^ Deep Red FM (Invitrogen) according to the manufacturer’s instructions. CAR T cells were resuspended in prewarmed PBS supplemented with 25 nM MitoTracker^®^ Deep Red FM, incubated for 15 min at 37 °C in the dark, rinsed twice with prewarmed PBS and immediately analyzed via flow cytometry.

The mitochondrial membrane potential in CAR T cells was analyzed via an Enhanced Mitochondrial Membrane Potential Assay Kit with JC-1 (Beyotime) following the manufacturer’s instructions. The cells were incubated with JC-1 working solution (37 °C, 20 min), washed with ice-cold assay buffer, immediately resuspended and analyzed by flow cytometry. ΔΨm loss was quantified by a green/red fluorescence shift.

Phosphorylation analysis

CD19 CAR T cells were stimulated with WT or CD58^KO^ Nalm6 tumor cells at a 1:1 E:T ratio for different co-culture time, then collected for phosphorylation analysis of c-Jun, c-Fos, NFAT and NF-κB p65. For the phosphorylation analysis, cells were immediately fixed with 4% formaldehyde for 15 min in room temperature, followed by permeabilization with ice-cold 100% methanol for 10 min on ice to enable intracellular epitope accessibility. After fixation and permeabilization, cells were washed and incubated with primary antibodies specific to the phosphorylated targets, diluted in an antibody buffer containing 0.5% BSA in PBS. Following a 1 h incubation at room temperature, cells were washed and stained with fluorochrome-conjugated anti-rabbit secondary antibodies for 30 min, protected from light. Following a final wash, and cells were resuspended in PBS for flow cytometry analysis.

Western blot analysis

CD19 CAR T cells were treated with WT or CD58^KO^ Nalm6 cells under specified conditions in this study, then harvested by FACS. Cell lysates were prepared by lysing harvested cells in RIPA buffer (YangGuangBio) supplemented with protease and phosphatase inhibitors (cOmplete Mini and PhosSTOP, Roche). For detection of Bax translocation to mitochondria, a mito/cytosol fractionation experiment was performed using a mitochondria separation kit (Proteintech), and the lysates were prepared according to instructions. Concentrations of extracted proteins were determined using BCA Protein Assay (YangGuangBio). Prepared proteins were loaded and separated by 10% or 15% precast Bis-Tris gels (Genscript) and transferred onto 0.45 µm PVDF membranes (Millipore). Membranes were gently blocked with 5% non-fat milk in TBS-T buffer (YangGuangBio) for 1 h at room temperature and incubated with primary antibodies overnight at 4 °C. After washing, membranes were incubated with HRP-conjugated secondary antibodies for 1 h at room temperature. Protein bands were visualized using enhanced chemiluminescence (ECL) reagents (YangGuangBio) and imaged with a chemiluminescence detection system. For the western blot evaluation of the impact of inhibitory phosphatase blockade on the activation of AP-1 signaling, CD19 CAR T cells were cultured in anti-CD3 monoclonal antibody (OKT3, 50 ng/mL, Takara) precoated plates with RetroNectin (10 μg/mL, Takara), in X-VIVO 15 medium (Lonza) supplemented with 300 U/mL recombinant human IL-2 (PeproTech) and different concentration of TPI-1, MurA-IN-1 and BCI hydrochloride (MCE) for 6 h, then harvested and processed as described in above procedures. Antibodies used in all western blot assays were described in Supplementary Table 2. Original Western blot data are presented in Supplementary Fig. 16–18.

Seahorse assays

Mitochondrial respiration and glycolytic function were analyzed using the Seahorse XFe96 Extracellular Flux Analyzer (Agilent Technologies). Calibrated sensor cartridge from XFe96 FluxPak (Agilent Technologies) was hydrated overnight in XF Calibrant Solution (Agilent Technologies). To enhance cell adherence, XFe96 Cell Culture Microplates were precoated with Poly-ʟ-Lysine Hydrobromide (Sigma-Aldrich). CD19 CAR T cells that had undergone three rounds of stimulation with WT or CD58^KO^ Nalm6 cells were sorted and resuspended in assay medium and seeded into precoated microplates at 10⁵ cells/well. For mitochondrial stress testing, cells were equilibrated in XF assay medium (pH 7.4) supplemented with 10 mM glucose, 1 mM pyruvate, and 2 mM glutamine (Sigma-Aldrich) for 1 h prior to analysis. Sequential injections delivered final concentrations of: 1.5 μM oligomycin, 1.0 μM FCCP, and 0.5 μM rotenone/antimycin A. For glycolysis stress testing, cells were equilibrated in XF assay medium (pH 7.4) supplemented with 2 mM glutamine (Sigma-Aldrich) for 1 h prior to analysis. Glycolysis stress testing was performed using the Glycolysis Stress Test Kit with sequential injection of 10 mM glucose, 1.0 μM oligomycin, and 50 mM 2-DG (2-deoxy-D-glucose). Extracellular acidification rate (ECAR) values were recorded under basal conditions and following pharmacological perturbations. Measurements were conducted using a cycle of 3 min mixing, 3 min waiting, and 3 min measurement intervals. All parameters were normalized to the cell number per well by Seahorse Wave Desktop Software (v2.6.3, Agilent Technologies).

Cytochrome c release detection by immunofluorescence

Cytochrome c release was detected by immunofluorescence. CD19 CAR T cells were stimulated by three rounds of WT or CD58^KO^ tumor cells. After the third round of stimulation, CAR T cells were sorted from the coculture system. Unstimulated CAR T cells served as the control. For staining, coverslips were pre-coated overnight at 4 °C with 0.025% poly-L-lysine. Sorted cells were allowed to adhere onto the coverslips, then fixed with 4% paraformaldehyde for 10 min at room temperature (RT), permeabilized with 0.1% Triton X-100 for 15 min (RT), and blocked with 5% BSA for 30 min (RT). Cells were incubated overnight at 4 °C with primary antibodies against cytochrome c and TOMM20 (both 1:500, Proteintech). After washing, samples were incubated for 1 h at RT in the dark with fluorophore-conjugated secondary antibodies: CoraLite^®^488-conjugated goat anti-rabbit IgG and CoraLite^®^594-conjugated goat anti-mouse IgG (both 1:500, Proteintech). Images were acquired using a confocal laser scanning microscope (Leica).

Mice

A total of 1 × 10^5^ luciferase-expressing WT or CD58^KO^ Nalm6 cells were transplanted intravenously into female NOD-Prkdc-scid-Il2rg-deficient mice (NPG/Vst strain, Vitalstar) aged 4–6 weeks. Animals received purified CAR T cells or control T cells intravenously 7 days after tumor engraftment at the indicated amount. The tumor burden was monitored once per week through bioluminescence imaging using an in vivo imaging system (Perkin Elmer). For immunophenotyping, peripheral blood samples were collected by retro-orbital bleeding. Erythrocytes were lysed with ammonium-chloride-potassium buffer (Thermo Fisher Scientific), and leukocytes were stained with fluorochrome-conjugated antibodies for flow cytometric analysis. The experimental protocols received ethical approval from the Laboratory Animal Ethics Committee of Vitalstar Biotechnology (Beijing, China).

Phospho-antibody array

Phosphoprotein profiling was conducted using the cancer-signaling phospho-antibody microarray PCS248 (Full Moon Biosystems, Inc., Sunnyvale, CA). CD19 CAR T cells were treated with WT or CD58^KO^ Nalm6 cells at a 1:1 E:T ratio for 1 h, sorted by flow cytometry, and lysed under non-denaturing conditions. Protein lysates were biotinylated and purified according to the manufacturer’s protocol. Microarray slides were blocked for 30 min, rinsed thoroughly, and dried. Biotinylated protein samples were incubated with the slides for 2 h at RT. After washing, bound proteins were detected with Cy3-conjugated streptavidin and visualized using a GenePix 4000 scanner (Molecular Devices). Fluorescence signals were quantified using GenePix Pro 6.0, with phosphorylation ratios calculated as phospho-value/non-phospho-value and normalized to β-actin for total proteome standardization.

RNA sequencing and analysis

Ctrl CAR T cells or DUSP6^KO^ CAR T cells co-cultured with WT or CD58^KO^ cells at an E:T ratio of 1:1 for 3 days were sorted by magnetic beads (Miltenyi Biotec). Total RNA was extracted from samples using Trizol reagent (Thermo Fisher) following the manufacturer’s protocol. RNA quality was assessed using a Bioanalyzer 2100 with the RNA 6000 Nano LabChip Kit (Agilent), and only high-quality RNA samples (RIN > 7.0) were selected for library construction. mRNA was isolated from 5 µg of total RNA using oligo(dT) magnetic beads with two rounds of poly(A) selection to ensure high purity. The purified mRNA was fragmented using divalent cations (Magnesium RNA Fragmentation Module, NEB) at 94 °C for 5–7 min. First-strand cDNA synthesis was performed with SuperScript™ II Reverse Transcriptase (Invitrogen), followed by second-strand synthesis using E. coli DNA polymerase I (NEB), RNase H (NEB), and dUTP Solution (Thermo Fisher) to generate U-labeled double-stranded cDNA. After A-tailing and adapter ligation (dual-indexed adapters), size selection was performed with AMPureXP beads. UDG enzyme (NEB) treatment removed U-labeled strands, and PCR amplification (8 cycles) generated the final cDNA libraries with an average insert size of 300 ± 50 bp. Paired-end sequencing (2 × 150 bp) was conducted on an Illumina Novaseq™ 6000 platform (LC-Bio Technology). For bioinformatics analysis, raw sequencing reads were processed to remove adapters, polyA/G sequences, low-quality bases (Q < 20), and reads with >5% unknown nucleotides using Cutadapt. Quality-filtered “clean reads” were assessed using FastQC to evaluate Q20, Q30, and GC content. Clean reads were aligned to the reference genome using HISAT2 (v2.2.1), allowing up to two mismatches and multiple alignments. Transcript assembly and quantification were performed with StringTie (v2.1.6), followed by merging transcriptomes across samples using gffcompare. Gene expression levels were calculated as FPKM values. Differential gene expression analysis was performed using both DESeq2 and edgeR to ensure robustness. Genes with an adjusted P value (FDR) < 0.05 and an absolute fold change ≥ 2 were considered significantly differentially expressed. Sample relationships were evaluated via PCA analysis in R. Functional enrichment of differentially expressed genes was analyzed using GO and KEGG pathway databases, with hypergeometric tests (P < 0.05). GSEA (v4.1.0) was applied to identify enriched pathways (|NES | > 1, FDR < 0.25). Alternative splicing events were detected using rMATS (v4.1.1) with FDR < 0.05. Single-nucleotide polymorphisms were called using Samtools (v0.1.19) and annotated with ANNOVAR. Raw sequencing data were deposited in the NCBI Gene Expression Omnibus under accession number GSE292279.

Statistical analysis

Statistical analyses were conducted via GraphPad Prism 10.4.0 (GraphPad Software Inc., La Jolla, CA, USA). Comparisons between groups were performed via two-tailed Student’s t tests or two-way ANOVA. Survival data were analyzed via the log-rank test. P < 0.05 was considered significant. The group data are presented as means ± SDs unless otherwise noted.

Supplementary information

SUPPLEMENTAL MATERIAL