Envirotune-CAR-T: a hypoxia-responsive and glutamine-enhanced CAR-T cell therapy for overcoming tumor microenvironment-mediated suppression

Abstract

Background

Chimeric antigen receptor (CAR)-T cell therapy has demonstrated remarkable success in hematologic malignancies; however, its efficacy in solid tumors remains limited. A major barrier is the immunosuppressive tumor microenvironment (TME), which is characterized by hypoxia and nutrient deprivation, leading to impaired CAR-T cell proliferation, persistence, and cytotoxic function. To address these barriers, we designed a dual-regulatory CAR-T strategy that integrates hypoxia-responsive control with metabolic enhancement to improve therapeutic efficacy in solid tumors.

Methods

To overcome these barriers, we developed a next-generation CAR-T platform with dual adaptations targeting the metabolic and transcriptional constraints of the TME. Specifically, we engineered hypoxia-responsive regulatory elements derived from VEGF to drive sustained CAR expression under hypoxic conditions. Concurrently, we overexpressed the glutamine transporter SLC38A2 to enhance glutamine uptake and metabolic fitness in nutrient-deprived environments.

Results

Compared with conventional CAR-T cells, our engineered CAR-T cells exhibited superior antitumor activity under hypoxia and nutrient stress, with enhanced proliferation, elevated memory phenotype, and reduced exhaustion markers. Mechanistically, quantitative PCR demonstrated upregulation of glutamine metabolic and glycolytic pathways, while Seahorse assays confirmed enhanced oxidative phosphorylation and glycolysis. SLC38A2 knockout reversed these enhancements, highlighting its role in sustaining CAR-T metabolic fitness.

Conclusion

Our findings establish SLC38A2 as a critical metabolic regulator that enhances CAR-T antitumor efficacy, providing a promising strategy to improve the durability and efficacy of CAR-T cell therapies in TME.

WHAT IS ALREADY KNOWN ON THIS TOPIC.

WHAT THIS STUDY ADDS

• This study introduces a dual-regulatory ENV-CAR-T platform that combines hypoxia-responsive element-driven CAR expression with SLC38A2-mediated glutamine uptake, enabling enhanced metabolic fitness, antitumor efficacy, and TME adaptability under hypoxic and nutrient-deprived conditions.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• These findings provide a mechanistic framework for developing next-generation CAR-T therapies optimized for hostile TMEs, potentially improving clinical outcomes in solid tumor immunotherapy and guiding future CAR design strategies.

Introduction

Chimeric antigen receptor (CAR)-T cell therapy has revolutionized the treatment of hematologic malignancies, achieving unprecedented clinical responses in relapsed/refractory B-cell malignancies.¹ However, its efficacy in solid tumors remains constrained by the dual challenges of immunosuppressive signals and metabolic dysregulation within the tumor microenvironment (TME).^{2 3} Solid tumors establish a pathophysiological niche characterized by fluctuating hypoxia gradients,⁴ nutrient deprivation,⁵ and aberrant metabolite accumulation conditions that collectively impair T cell receptor signaling, mitochondrial fitness,⁶ and epigenetic programming,⁷ ultimately subverting CAR-T cell infiltration,^{8 9} persistence, and effector functions.^{10 11} Among these barriers, hypoxia emerges as a master regulator that not only stabilizes hypoxia-inducible factors (HIF-1α/2α) to drive immunosuppressive angiogenesis but also rewires cellular metabolism through transcriptional repression of oxidative phosphorylation (OXPHOS).12,14 Hypoxic stress imposes a tripartite liability: it cripples mitochondrial respiration by destabilizing electron transport chain complexes,¹⁵ restricts clonal expansion through mechanistic target of rapamycin complex 1 (mTORC1) signaling dampening,¹⁶ and directly induces T cell exhaustion via sustained programmed death protein 1/programmed death-ligand 1 (PD-1/PD-L1) axis activation-mechanistic cascades that collectively erode therapeutic efficacy.17,19 Hypoxic stress triggers a conserved transcriptional program mediated by HIF-1α binding to hypoxia response elements (HREs),²⁰ upregulating genes critical for vascular endothelial growth factor (VEGF), Erythropoietin (EPO),²¹ and metabolic adaptation (lactate dehydrogenase A, Glucose-6-phosphate isomerase (GPI)).²² This oxygen-sensing pathway has been strategically leveraged in CAR-T design, with HRE-driven constructs demonstrating enhanced targeting of solid tumor antigens including ErbB2/Her2 in hypoxic niches.^{22 23}

The metabolic battlefield within the TME imposes an additional layer of complexity.²⁴ Tumor cells and infiltrating immune cells engage in fierce competition for glucose, amino acids, and other critical metabolites.²⁵ Glutamine, a conditionally essential amino acid, serves as a nodal point in this conflict—it fuels the tricarboxylic acid cycle for energy production,²⁶ supports redox homeostasis through glutathione synthesis,²⁷ and provides nitrogen groups for nucleotide biosynthesis.²⁸ Recent studies reveal that glutamine scarcity in the TME correlates with impaired CAR-T cell expansion and accelerated exhaustion phenotypes,²⁹ suggesting that enhancing glutamine utilization capacity could break this metabolic stalemate.³⁰ Notably, solute carrier family 38 member 2 (SLC38A2), a high-affinity glutamine transporter upregulated in activated T cells,³¹ may serve as a gatekeeper for glutamine-dependent metabolic reprogramming.^{32 33} Its expression dynamics under hypoxic stress, however, remain unexplored.

Current strategies to overcome TME-mediated suppression often focus on single-axis interventions, yet fail to address the spatial-temporal heterogeneity of metabolic and hypoxic stresses. We hypothesize that engineering CAR-T cells with dual environmental sensors could enable context-dependent adaptation: (1) SLC38A2 overexpression to dominate glutamine uptake under nutrient competition, and (2) HREs to activate protective pathways when oxygen tension drops below critical thresholds. This orthogonal regulation strategy aims to decouple metabolic fitness from hypoxic stress—a paradigm distinct from previous approaches that inadvertently link nutrient utilization to oxygen-dependent pathways.

Mesothelin (MSLN), a glycosylphosphatidylinositol-anchored glycoprotein, is overexpressed in aggressive malignancies including ovarian and breast cancers,³⁴ while exhibiting restricted expression in normal tissues. While prior clinical efforts have established MSLN as a compelling CAR-T target in ovarian cancer,³⁵ therapeutic efficacy remains constrained by TME suppression and antigen heterogeneity.³⁶ To transcend these limitations, we engineered ENV-CAR-T cells through orthogonal integration of two adaptive modules. This dual-input architecture enables metabolic-state switching—preserving integrated hypoxia-sensing genetic circuits while sustaining CAR-T cells biosynthetic capacity through accelerated SLC38A2-mediated glutamine flux. Preclinical validation across in vitro hypoxia chambers and in vivo desmoplastic tumor models demonstrates superior efficacy and safety over conventional MSLN-CAR-T designs. By reframing the TME’s pathognomonic features as dual activation triggers, our work provides a novel paradigm for engineering next-generation CAR-T therapies tailored to the metabolic constraints of the TME.

Methods

Cells and culture conditions

HCC1806 (human triple-negative breast cancer cell line), CAPAN2 (human pancreatic adenocarcinoma cell line), OVCAR3 (human ovarian adenocarcinoma cell line), SKOV3 (human ovarian adenocarcinoma cell line), 293T (human embryonic kidney 293 cells, expressing the SV40 large T antigen) were originally purchased from American Type Culture Collection (ATCC) and were reauthenticated for this study by ATCC, and individually grown in NIH:OVCAR-3 (Procell, China), McCoy’s 5A (Gibco, USA), RPMI 1640 (Gibco, USA), Dulbecco’s Modified Eagle Medium (Gibco, USA) supplemented with 10% fetal bovine serum (FBS; Gibco, USA). Cell lines were confirmed to be free of mycoplasma for this study. Cells were maintained under normoxic conditions (37°C, 21% O₂, 5% CO₂, balanced with N₂) in a standard incubator, or hypoxic conditions (37°C, 0.5%, 1% or 2% O₂, 5% CO₂, balanced with N₂) using a tri-gas hypoxia chamber (Thermo, USA). All tumor cell lines were transfected with Luc-2A-GFP using lentiviral to constitutively express firefly luciferase and GFP for functional evaluation of CAR-T cells.

Mice

Female C-NKG (NOD-scid IL2Rγ^null) mice (6–8 weeks old, ~22 g) were purchased from Cyagen (Suzhou) Biosciences Inc. Mice were maintained under specific pathogen-free conditions in individually ventilated cages at the Experimental Animal Center of Nanjing Normal University (IACUC-2024234), under a 12 hours light/dark cycle, constant temperature (∼22°C) and 55% relative humidity, with free access to food and water. For HCC1806 and OVCAR3 subcutaneous xenograft models, tumor cells were injected subcutaneously into the right flank of each mouse. Tumor size was measured with calipers, and volume calculated as (length×width²) × 0.5. Survival was defined as the time point when tumor volume reached 2000 mm³.

A total of nine mice per group were included in the study, with three mice designated as satellite animals for additional endpoint analyses (eg, pharmacokinetics (PK) or biomarker validation). The remaining six mice per group were used for primary efficacy evaluation. Mice were randomly allocated to experimental and control groups, ensuring unbiased group distribution. Animals were excluded if they showed signs of severe illness (eg, rapid weight loss, visible discomfort) or were euthanized before the experiment was completed for ethical reasons. All animals that completed the study were included in the analysis.

T cell isolation and lentivirus transduction

Peripheral blood mononuclear cells (PBMCs, MiaoShun Biotech, China) from healthy donors were isolated by Lymphoprep (STEMCELL, Canada), and then T cells were isolated from the cells by negative selection using EasySep Human T Cell Isolation Kit (STEMCELL, Canada). Then activated using anti-human CD3 and CD28 beads (Thermo, USA) at a 3:1 bead to cell ratio on day 0. Purified T cells were cultured in 5% FBS X-VIVO 15 Serum-free Hematopoietic Cell Medium (Lonza, Switzerland) supplemented with recombinant human IL-2 (300 IU/mL, Peprotech, USA). Detection of the CAR-T cell positive rate and detection of cell phenotype was performed following lentivirus infection and continuous culture for 48 hours after T cell isolation.

Construction of the vectors of MSLN-targeted CAR-T cells

The lentivirus vector, Lv-EF-1 alpha-MSLN was used as the basal framework for construction. The CAR sequences were inserted downstream of the EF-1 alpha promoter to create the conventional EF-1α-MSLN-CAR-T, then replaced the EF-1 alpha promoter with a hypoxia-responsive promoter (5H1P, 5E-HRE, 5V1-HRE, 5V2-HRE, and 5V3-HRE) consisting of different sources of HRE and Cytomegalovirus (CMV) or a hybrid of CMV enhancer and chicken β-actin promoter (CAG) or EF-1α promoter to create the hypoxia responsive CAR-Ts. The CAR sequences consisted of MSLN single-chain variable fragment (scFv), as well as the CD8α hinge and transmembrane region, a CD28 costimulatory domain, and the intracellular CD3ζ domains. The sequences for hypoxia-responsive promoters were presented in the online supplemental information. Modification of the 3’ end of the base CAR expression cassette was performed to include the full coding sequence (CDS) of either human SLC38A2. The resulting codon-optimized CDSs were synthesized and subcloned downstream of and within the open reading frame of the base CAR expression cassettes (GenScript), with CDS separated by a self-cleaving P2A peptide. All the sequences were validated by restriction enzyme analysis and DNA sequencing.

Plasmid construction

The SLC38A2-overexpressing MSLN-CAR lentiviral vector was constructed by inserting human SLC38A2 cDNA upstream of a P2A-linked MSLN-CAR cassette in the pCDH-EF1α backbone. The MSLN-CAR comprised MSLN scFv, CD8α hinge/transmembrane domain, CD28, and CD3ζ domains. The correct assembly was verified by restriction digestion and Sanger sequencing (GENERAL BIOL, China).

Flow cytometry

The antibodies used in this study, including anti-MSLN-PE, CD45-PerCP, CD45RA-APC, CCR7-PE, PD-1-APC, TIM 3-PE, CD11b-APC and HLA-DR-FITC were purchased from BioLegend (San Diego, California, USA). Isotype-matched control monoclonal antibodies were applied in all the procedures. All flow cytometry staining procedures were kept out of the dark at room temperature. Cells were washed with phosphate-buffered saline (PBS) three times after incubation for 30 min at 4°C and then all samples were acquired on a CytoFLEX (Beckman Coulter, Indianapolis, Indiana, USA), and data were analyzed using FlowJo software (BD, USA).

Cell proliferation assay

Different CAR-T cells were seeded in 6-well plates at 1×10⁶ cells per well on day 4, and were fluorescently labeled by carboxy-fluorescein succinimidyl ester (Thermo, USA). All cells were cultured in an atmosphere of 1% oxygen, then detected the fluorescence values by flow cytometry on day 9.

Impedance-based kinetics cell lysis assay

Real-time impedance analysis of cell lysis kinetics was evaluated over 50 hours. HCC1806, CAPAN2, OVCAR3, SKOV3 and 293 T cells were plated in a 96-well resistor bottom plate at 2.0×10⁴ cells per well. After being cultured for 24 hours, effector T cells were added into the unit at different effector/target cell (E:T) ratios (1:1, 1:4). Impedance was measured at 28 s intervals. The impedance-based cell index for each well and time point was normalized with the cell index before adding T or CAR-T cells. The kinetics of cell lysis was evaluated as the change in normalized cell index over time.

Luciferase-based cytotoxicity assay for CAR-T cell evaluation

To quantify the toxicity of CAR-T cells, the cell lysis rate was measured using the One-Lite Luciferase Assay kit (Vazyme, China). Add One-Lite assay reagent in a volume equal to that of the cell culture supernatant to be assayed and equilibrate to room temperature. Leave at room temperature for at least 3 min to allow the fluorescent signal to stabilize, then measure the fluorescence value using a microplate reader.

Measurement of cytokines

We performed cytokine analysis both in vitro and in vivo using the BD Human Th1/Th2 Cytokine Cytometric Bead Array (CBA) kit (BD, USA). For standard in vitro assays, CAR-T cells were co-cultured with target tumor cells, and supernatants were collected to assess tumor necrosis factor alpha (TNF-α), interferon-gamma (IFN-γ), interleukin (IL)-6, and IL-10 secretion. For cytokine release syndrome (CRS) evaluation in the presence of macrophages, human monocytes were isolated from PBMCs by adherence and differentiated into macrophages with 10 ng/mL granulocyte-macrophage colony-stimulating factor for 7 days. Macrophage differentiation was confirmed by flow cytometry (CD11b+HLA-DR+). HCC1806 tumor cells, MSLN-CAR-T (or ENV-CAR-T) cells, and macrophages were co-cultured at a 1:2:1 ratio for 4 hours, and supernatants were analyzed for IL-6 and IL-10.

Measurement of glutamine content

Glutamine levels were determined using a glutamine assay kit (Grace Biotechnology, China), CAR-T cells were ultrasound lysed using the lysis buffer in the corresponding kit at 4°C. Cell homogenates were centrifuged at 12,000×g for 30 min at 4°C. Finally, glutamine content was normalized according to cell number. All reagents were thawed to room temperature (25°C) and added sequentially to a 96-well plate. After thorough mixing, the plate was incubated at 37°C in the dark for 40 min, and absorbance was measured at 450 nm.

Apoptosis assay

To measure the apoptotic rate, CAR-T cells were seeded into six-well plates. Annexin V-PE/7-AAD Apoptosis Detection Kit (Vazyme, China) was used based on the manufacturer’s instruction. In brief, cells were washed twice in ice-cold PBS, and incubated with 100 µL 1×Binding Buffer supplemented with 5 µL Annexin V-PE and 5 µL 7-AAD Staining Solution. 10 min later, another 400 µL 1×Binding Buffer was added. Apoptotic rate was measured by flow cytometry.

Mitochondrial oxygen consumption rate and extracellular acidification rate assay

The oxygen consumption rate (OCR) of CAR-T cells was measured in Agilent Seahorse XF media (Agilent, USA) containing 10 mM glucose, 1 mM sodium pyruvate, and 2 mM L-glutamine (pH 7.4) using Agilent Seahorse XF Cell Mito Stress Test Kit (Agilent, USA) with the XF-24 Extracellular Flux Analyzer (Agilent, USA) following the manufacturer’s instructions. The extracellular acidification rate (ECAR) of CAR-T cells was measured in XF media containing 2 mM L-glutamine (pH 7.4) using XF Glycolysis Stress Test Kit (Agilent, USA) with the XF-24 Extracellular Flux Analyzer following the manufacturer’s instructions.

ROS generation detection

For reactive oxygen species (ROS) detection, CAR-T cells were stimulated by HCC1806 cells (E:T=2:1) as described above. After 24 hours incubation, cells were stained with 2,7-dichlorodihydrofluorescein diacetate (Yeasen Biotechnology, China) and incubated at 37°C for 15 min, followed by flow cytometry analysis.

ATP assay

Extracellular ATP concentration in CAR-T cells was analyzed using an ATP Chemiluminescence Assay Kit (Elabscience, China) according to the manufacturer’s instructions.

Generation of CRISPR–Cas9 knockout CAR-T cells

T cells were transduced with lentivirus of pLenti-MSLN. MSLN+cells were sorted, and expression of MSLN was confirmed by flow cytometry. CAR-T cells were then transduced with control single-guide RNA (sgRNA) (sgNTC: ATGACACTTACGGTACTCGT) or sgRNA targeting SLC38A2 (sgSLC38A2: GGAGTAGTTGAAGTCGCTGT). After sorting of MSLN+cells, cells were expanded, and deletion of SLC38A2 was verified by immunoblot analysis.

Measurement of intracellular L-Lactate levels

L-Lactate concentrations were quantified using the L-Lactate Assay Kit (Beyotime, China) as per the manufacturer’s guidelines.

Measurement of cellular glucose uptake

Glucose uptake was determined using the Glucose Uptake Assay Kit (Beyotime, China) following the provided protocol.

Measurement of NAD⁺/NADH levels

The NAD⁺/NADH ratio was assessed using the NAD⁺/NADH Assay Kit (Beyotime, China) according to the standard procedures.

Measurement of α-Ketoglutarate levels

α-Ketoglutarate (α-KG) levels were measured using the Amplex Red Alpha-Ketoglutarate Assay Kit (Beyotime, China) according to the manufacturer’s instructions.

Gene expression analysis by SYBR Green qPCR

Quantitative PCR (qPCR) was performed using SYBR Green chemistry following RNA extraction and reverse transcription. Reactions were run on a QuantStudio 6 Flex system. Gene expression levels were normalized to ACTB as an internal control and calculated using the 2^−ΔΔCt method. Primer sequences are listed in the online supplemental table.

TaqMan probe-based qPCR analysis of MSLN-CAR-T DNA copy numbers

Real-time fluorescent qPCR was applied to determine the copy numbers of MSLN-CAR-T DNA in DNA extracted from mouse plasma. Genomic DNA was extracted from mice PBMCs using a QIAamp DNA Blood Mini Kit (QIAGEN, Germany) after CAR T-cell infusion. The forward primer 5’-ACCTGGTCGACAATCAACC-3’, reverse primer 5’-AAGCAGCGTATCCACATAGC-3’, and probe primer 5’-FAM-CAAAATTTGTGAAAGATTGACTGGT-TAMRA-3’ were used in the qRT-PCR assay.

Western blot analysis

Proteins of cells were extracted with RIPA buffer and the concentration was measured using BCA Protein Assay Kit (Thermo Scientific, USA). Protein samples were separated in 8–12% Sodium Dodecyl Sulfate–Polyacrylamide Gel Electrophoresis (SDS-PAGE) gels and then transferred onto Polyvinylidene Fluoride (PVDF) membranes. Afterward, membranes were blocked for 1 hour and incubated with primary antibodies overnight. After the incubation with the corresponding species-specific secondary antibodies for 1 hour, bands were detected by chemiluminescence using an electrochemiluminescence system. Primary antibodies used in this study are as follows: anti-Tubulin (1:1000, Beyotime, China) and anti-SLC38A2 (1:1000, Cusabio, China).

IHC, H&E and IF staining

All staining procedures were performed by Servicebio Biotech (Wuhan, China). Mouse tissues were fixed in 4% paraformaldehyde (Sigma-Aldrich, USA), paraffin-embedded, and sectioned. Immunohistochemistry (IHC) was performed using anti-CD45 antibody, anti-SLC38A2 antibody and HRP-conjugated secondary antibody, visualized with 3,3'-Diaminobenzidine (DAB) and counterstained with hematoxylin. H&E staining followed standard protocols. For immunofluorescence (IF), sections were incubated with anti-CD45, anti-HIF1α, anti-CD3, anti-CD8, or anti-CD69, followed by Alexa Fluor-conjugated secondary antibody and 4′,6-Diamidino-2-Phenylindole (DAPI) nuclear staining. Images were acquired using the Leica system and analyzed with CaseViewer.

Statistical analysis

All statistical analyses were performed using GraphPad Prism V.8 (GraphPad Software). Comparisons between two groups were conducted using unpaired or paired two-tailed Student’s t-test, as appropriate. For multiple group comparisons, one-way or two-way analysis of variance was applied, followed by Tukey’s or Bonferroni’s post hoc test. Survival data were analyzed using the log-rank (Mantel-Cox) test. Data are presented as mean±SD unless otherwise indicated. P values<0.05 were considered statistically significant. Significance was denoted as follows: p<0.05 (*), p<0.01 (**), p<0.001 (***).

Results

SLC38A2 overexpression enhances CAR-T cell function in vitro but is limited under hypoxia

To explore the role of glutamine metabolism in CAR-T cell efficacy, we analyzed clinical samples from patients with ovarian cancer treated with MSLN-CAR-T therapy. Responders exhibited significantly higher glutamine levels in peripheral blood compared with non-responders, suggesting a potential link between glutamine availability and CAR-T function (online supplemental figure S1A). Further in vitro assays demonstrated that glutamine supplementation (2 mM) upregulated the expression of the glutamine transporter SLC38A2 in MSLN-CAR-T cells (online supplemental figure S1B), particularly after antigen stimulation (online supplemental figure S1C). Further analysis revealed that SLC38A2 expression was low in human T cells but markedly higher in all tumor cell lines used in this study (online supplemental figure S1D).

To exploit this metabolic adaptation, we generated SLC38A2-overexpressing MSLN-CAR-T cells (SLC-CAR-T) using a P2A-linked construct (figure 1A). Flow cytometry and western blot analysis confirmed increased SLC38A2 expression in SLC-CAR-T cells compared with wild-type (WT) CAR-T cells, while CAR expression remained comparable between the two groups (figure 1B–D). Lentiviral titer quantification showed that the viral titers of the two constructs were similar, ensuring comparable transduction efficiency (online supplemental figure S1E,F). Functionally, SLC38A2 overexpression promoted CAR-T proliferation under glutamine-replete conditions, as evidenced by CellTrace Violet dilution assays across varying glutamine concentrations (figure 1E and online supplemental figure S1G).

Figure 1. Overexpression of SLC38A2 enhances CAR-T cells’ antitumor activity in vitro, but is limited under hypoxia. (A) Schematic representation of MSLN-CAR-T and SLC-CAR-T constructs. MSLN-CAR-T contains anti-mesothelin scFv, CD8α hinge/TM, CD28 costimulation, and CD3ζ domains. SLC-CAR-T adds SLC38A2 via P2A peptide for coexpression. (B–C) Flow cytometry (B) and western blot (C) analyses confirm SLC38A2 expression in MSLN-CAR-T and SLC-CAR-T cells. (D) CAR expression levels determined by flow cytometry in 2 mM glutamine medium. (E) Proliferation of MSLN-CAR-T and SLC-CAR-T cells cultured for 5 days under varying glutamine concentrations (0.1 mM, 1 mM, 5 mM, and 10 mM), assessed by CellTrace Violet (CTV) dilution. (F) Intracellular glutamine levels were quantified using a colorimetric assay. MSLN-CAR-T and SLC-CAR-T cells were cultured for 2 days in media supplemented with 0.1 mM, 1 mM, 5 mM, or 10 mM glutamine. (G) Flow cytometry analysis of MSLN antigen expression in target cells (HCC1806, OVCAR3, SKOV3) and non-target cells (293T). (H) Flow cytometry analysis of TNF-α and IFN-γ secretion by MSLN-CAR-T and SLC-CAR-T cells following co-culture with tumor cells (HCC1806, OVCAR3, SKOV3 and 293T). Data represent mean±SD from three independent experiments. (I) Real-time cytotoxicity assay measuring CAR-T-mediated lysis of MSLN+ and MSLN− cells. (J–K) Sustained cytotoxicity of CAR-T cells. Schematic of CAR-T cells undergoing repetitive antigen stimulation (J) and their serial killing capacity against tumor cell lines (HCC1806, OVCAR3, SKOV3-luc) in 24-hour interval assays at E:T=2:1 (K). (L–P) In vitro stimulation model and phenotypic characterization of CAR-T cells. Schematic of MSLN tumor cell-mediated T cell stimulation in vitro (L). Flow cytometry analysis of (M–N) memory T cell subsets (CD45RA-CCR7+) and (O–P) exhaustion markers (PD-1+TIM 3+) in MSLN-CAR-T cells following antigen stimulation. (Q) Apoptosis assessment of untreated and antigen-stimulated MSLN-CAR-T and SLC-CAR-T cells using 7AAD and Annexin V staining. (R–W) Functional characterization of CAR-T cells under hypoxic conditions. (R) Schematic of the hypoxic culture system. (S) Proliferation kinetics of MSLN-CAR-T versus SLC-CAR-T cells. (T) Glutamine uptake capacity under hypoxia. (U) Apoptosis analysis of T-cell subsets. (V–W) Cytotoxic Activity of MSLN-CAR-T and SLC-CAR-T cells against target cells. Data are presented as mean±SD (n=3). Statistical significance was determined by t-test or two-way ANOVA tests. *p<0.05; **p<0.01; ***p<0.001; ns, not significant. ANOVA, analysis of variance; CAR, chimeric antigen receptor; E:T, effector/target cell; FITC, Fluorescein Isothiocyanate; HCC1806, human triple-negative breast cancer cell line; IFN-γ, interferon-gamma; MSLN, mesothelin; OVCAR3, human ovarian adenocarcinoma cell line; PD-1, programmed death protein 1; scFv, Single-Chain Variable Fragment; SKOV3, human ovarian adenocarcinoma cell line; SLC38A2, solute carrier family 38 member 2; 293T, human embryonic kidney 293 cells, expressing the SV40 large T antigen; TCM, central memory T cells; TEFF, effector T cells; TEM, effector memory T cells; TSCM, T Stem Cell Memory (T cells); TM, transmembrane; TNF-α, tumor necrosis factor alpha.

Intracellular glutamine quantification revealed that SLC-CAR-T cells maintained higher glutamine uptake compared with WT CAR-T cells across multiple glutamine concentrations (figure 1F). To assess the functional impact of SLC38A2 overexpression, we next evaluated the cytotoxic potential of SLC-CAR-T cells. Flow cytometry confirmed MSLN expression on target tumor cells, including HCC1806, OVCAR3, and SKOV3-luc, but not on the antigen-negative control cell line 293T (figure 1G). To further investigate the impact of glutamine availability on CAR-T cell function, we conducted real-time cytotoxicity assays across a range of glutamine concentrations. Notably, SLC-CAR-T cells exhibited consistently superior antitumor activity against HCC1806 cells compared with MSLN-CAR-T cells, with the advantage being most pronounced under low-glutamine conditions (0.1–0.5 mM) (online supplemental figure S1H–K). Short-term luciferase-based cytotoxicity assays showed that SLC-CAR-T cells exhibited significantly enhanced tumor-killing activity against MSLN-positive tumor cells compared with WT CAR-T cells (online supplemental figure S1L–O). This was further supported by elevated cytokine production (IFN-γ and TNF-α) following short-term co-culture with target cells (figure 1H). Importantly, neither CAR-T construct showed off-target effects against MSLN-negative 293 T cells and no significant increase in IL-6 or IL-10 levels (online supplemental figure S1P–S), suggesting minimal risk of cytokine-associated toxicity. We then performed extended stimulation assays to assess cytotoxic persistence. Real-time killing assays revealed that SLC-CAR-T cells maintained superior tumor cell lysis compared with WT CAR-T cells over time (figure 1I), confirming improved sustained cytotoxic capacity. To further validate the durability of cytotoxicity, we performed multiround tumor stimulation assays, which showed that SLC-CAR-T cells retained superior killing efficiency over multiple rounds of antigen exposure, in contrast to the gradual functional decline observed in WT CAR-T cells (figure 1J–1K).

In addition to improved cytotoxicity, SLC-CAR-T cells displayed an enhanced memory phenotype (CD45RA-CCR7+subset) and reduced exhaustion markers (PD-1+, TIM 3+) compared with WT CAR-T cells (figure 1L–1P). Moreover, apoptosis assays revealed that SLC38A2 overexpression provided a survival advantage to CAR-T cells (figure 1Q).

Hypoxia attenuates SLC38A2-mediated metabolic advantages

Given the immunosuppressive nature of the TME, we further assessed the function of SLC-CAR-T cells under hypoxic conditions (figure 1R). Hypoxia significantly impaired CAR-T cell proliferation, with both SLC-CAR-T and WT CAR-T cells showing a substantial reduction in expansion compared with normoxia (online supplemental figure S1S,T). Glutamine uptake and apoptosis levels were also elevated under hypoxia, suggesting metabolic stress in hypoxia conditions (figure 1T–1U). Consistent with this, SLC-CAR-T cells exhibited only modest cytotoxic improvement over WT CAR-T cells in hypoxic conditions, indicating that the benefit of SLC38A2 overexpression was limited under metabolic stress (figure 1V–1W). Consistent findings were observed in hypoxic three-dimensional tumor model (online supplemental figure S1U,V).

Overall, these findings highlight that while SLC38A2 overexpression enhances CAR-T proliferation, persistence, and cytotoxicity in vitro, its efficacy is diminished under hypoxic conditions, underscoring the need for additional metabolic strategies to optimize CAR-T function in the solid TME.

5V2-HRE enhances CAR-T cell persistence and function in hypoxic tumor microenvironments

The characteristic hypoxic TME (<2% O₂), arising from malignant proliferation and dysfunctional vasculature, establishes a sharp oxygen gradient compared with healthy tissues (5–10% O₂).³⁷ To exploit this therapeutic window, we engineered tumor-selective CAR-T cells by incorporating HREs derived from EPO/VEGF genes as enhancer elements into EF-1α/CMV/CAG promoter systems.38,41 This genetic modification created a hypoxia-inducible expression platform that enhances CAR-T cell activation specifically within the TME while maintaining basal expression in normoxic conditions.

We engineered five hypoxia-responsive enhancer variants (5H1P, 5E-HRE, 5V1-HRE, 5V2-HRE, 5V3-HRE) upstream of the CAR transgene,^{23 42} with 5H1P as reference control⁴ (figure 2A and online supplemental figure S2A–E), based on established evidence that five copies of 35 bp VEGF HRE fragments with a CMV minimal promoter confer optimal hypoxia-inducible expression.⁴³ Thus, we selected five copies to balance transcriptional efficiency with genomic stability, minimizing risks associated with repetitive sequence elements. Under normoxia, CAR surface expression remained comparable across most constructs. However, after prolonged hypoxic exposure (ranging from 0.5% to 2% O₂, 48 hours), conventional MSLN-CAR-T cells exhibited substantial CAR downregulation, whereas 5V2-HRE CAR-Ts maintained robust expression levels of CAR surface expression across the entire hypoxic range (figure 2B,C and online supplemental figure S1F,G). These results demonstrate that the HRE regulatory system mediates oxygen gradient–responsive transcriptional control, enabling stable CAR expression in physiologically relevant hypoxic tumor environments. Phenotypic characterization revealed that 5V2-HRE CAR-T cells preferentially retained a central memory phenotype even after antigen stimulation in hypoxia (figure 2D–F). In cytotoxicity assays, 5V2-HRE CAR-Ts exhibited superior tumor lysis across multiple MSLN+cancer cell lines, correlating with sustained CAR expression under hypoxic stress (figure 2G–I). The cytotoxicity under normoxic conditions showed no significant difference compared with MSLN-CAR-T cells (p>0.05), indicating that CAR expression is regulated by oxygen levels (figure 2G–I). 5V2-HRE CAR-Ts exhibited significantly elevated levels of cytotoxic cytokines (TNF-α, IFN-γ) (figure 2J–L), while the levels of IL-6 and IL-10 released in co-culture with tumor cells remained low (figure 2M–O). Repeated antigen stimulation further highlighted the persistence of 5V2-HRE CAR-Ts, which maintained effective cytotoxicity over multiple rounds of tumor challenge, in contrast to the gradual functional decline observed in other hypoxia-responsive variants (figure 2P–R). To evaluate infiltration and cytotoxicity in a three-dimensional tumor model, GFP-labeled tumor spheroids were used. 5V2-HRE CAR-T cells effectively infiltrated and disrupted the hypoxic tumor core, whereas other variants exhibited limited penetration and weaker cytolytic activity (Figure 2S–W).

Figure 2. Enhanced hypoxic functionality of 5V2-HRE-engineered CAR-T cells. (A) Schematic of lentiviral vectors and CAR constructs. MSLN-CAR-T (EF-1α), 5H1P (VEGF-1+CMV), 5E-HRE (EPO+CMV), 5V1-HRE (VEGF-2+CMV), 5V2-HRE (VEGF-2+EF-1α), 5V3-HRE (VEGF-2+CAG). (B–C) CAR surface expression under normoxia (21% O₂) and hypoxia (1% O₂, 48 h). (D) Representative flow cytometry plots of memory T-cell subsets (TSCM: CD45RA+CCR7+; TCM: CD45RA-CCR7+; TEM: CD45RA-CCR7−; TEFF: CD45RA+CCR7−) in MSLN-CAR-T cells before and after antigen stimulation under hypoxia (1% O₂). (E–F) Quantitative analysis of memory subset distribution from flow cytometry data. (G–I) Short-term cytotoxicity (6 hours co-culture) against ovarian tumor lines (HCC1806, OVCAR3, SKOV3-luc) under hypoxia (1% O₂) versus normoxia (21% O₂). (J–L) Flow cytometry analysis of TNF-α and IFN-γ secretion by engineered CAR-T cells following co-culture with tumor cells (HCC1806, OVCAR3, SKOV3 and 293T) under hypoxia. (M–O) Safety cytokine profile of IL-6 and IL-10 secretion by engineered CAR-T cells following co-culture with tumor cells (HCC1806, OVCAR3, SKOV3 and 293T) under hypoxia. (P–R) Serial killing assay: Four-round repeated stimulation (E:T=2:1, 24 hours intervals) under hypoxia to assess durable cytotoxicity. (S–W) Tumor microsphere cleavage assay: GFP-labeled tumor spheroids co-cultured with CAR-T cells; representative fluorescence images (I, scale bar: 100 µm) and quantitative cleavage efficiency (J) are shown. Data represent mean±SD (n=3); Statistical significance was determined by two-way ANOVA tests. *p<0.05, **p<0.01, ***p<0.001. ANOVA, analysis of variance; CAG, CMV early enhancer/Chicken β-Actin promoter; CAR, chimeric antigen receptor; CMV, Cytomegalovirus; EPO, Erythropoietin; E:T, effector/target cell; HCC1806, human triple-negative breast cancer cell line; HRE, hypoxia response elements; IFN-γ, interferon-gamma; IL, interleukin; MFI, Mean Fluorescence Intensity; MSLN, mesothelin; ns, not significant; OVCAR3, human ovarian adenocarcinoma cell line; scFv, Single-Chain Variable Fragment; SKOV3, human ovarian adenocarcinoma cell line; 293T, human embryonic kidney 293 cells, expressing the SV40 large T antigen; TCM, central memory T cells; TEFF, effector T cells; TEM, effector memory T cells; TM, transmembrane; TNF-α, tumor necrosis factor alpha; TSCM, T Stem Cell Memory (T cells); VEGF, vascular endothelial growth factor.

Together, these findings establish 5V2-HRE as a potent hypoxia-adaptive regulatory element that enhances CAR-T cell persistence, preserves functional memory, and improves tumor infiltration within metabolically hostile microenvironments.

Dual-regulated ENV-CAR-T exhibits enhanced metabolic fitness and antitumor activity under hypoxia

Building on the hypoxia-adapted 5V2-HRE platform, we engineered a dual-regulated CAR-T variant (ENV-CAR-T) by co-expressing SLC38A2 to simultaneously enhance CAR stability and metabolic fitness. This approach aimed to integrate hypoxia-adaptive CAR regulation with metabolic reprogramming to address both oxygen and nutrient limitations within the TME. The ENV-CAR-T construct incorporated the 5V2-HRE-driven CAR and SLC38A2 in a single vector via a P2A self-cleaving peptide, ensuring synchronized expression in response to hypoxic stress (figure 3A and online supplemental figure S3A). Flow cytometry and western blot confirmed that ENV-CAR-T cells expressed higher levels of SLC38A2 compared with conventional MSLN-CAR-T cells (figure 3B and online supplemental figure S3B). Furthermore, CAR expression was significantly upregulated under hypoxia in ENV-CAR-T, whereas it declined in control cells (online supplemental figure S3C), suggesting improved CAR stability under stress conditions.

Figure 3. ENV-CAR-T exhibits enhanced antitumor activity and metabolic adaptation in an in vitro hypoxia model. (A) Schematic representation of MSLN-CAR-T and ENV-CAR-T constructs. In ENV-CAR-T, SLC38A2 is coexpressed with CAR under the control of 5V2-HRE sequences, separated by a P2A self-cleaving peptide. (B) Flow cytometry analysis of SLC38A2 expression in MSLN-CAR-T and ENV-CAR-T cells. (C) Proliferation of MSLN-CAR-T and ENV-CAR-T cells labeled with CellTrace Violet (CTV) after 2 days of culture with 2 mM glutamine, analyzed by flow cytometry. (D–E) Metabolic adaptation of CAR-T cells under hypoxia: (D) intracellular glutamine levels and (E) ATP content measured after 2 days in hypoxic conditions with 2 mM glutamine. (F) Apoptosis of MSLN-CAR-T and ENV-CAR-T cells under hypoxia before and after antigen stimulation. (G) Reactive oxygen species (ROS) accumulation in MSLN-CAR-T and ENV-CAR-T cells 48 hours after stimulation with HCC1806 cells. (H–I) Mitochondrial function and glycolysis: (H) Real-time measurement of oxygen consumption rates (OCRs) in basal and stimulated conditions; (I) Seahorse glycolytic stress test showing glycolytic capacity of MSLN-CAR-T and ENV-CAR-T cells. (J) Schematic representation of intracellular glutamine metabolic pathways and key regulatory genes. (K) Significantly upregulated glutamine metabolism and glycolytic enzymes (HK2, GPI, GLS and GOT1 etc) in ENV-CAR-T versus MALN-CAR-T. (L–M) Memory differentiation and exhaustion status of MSLN CAR+T cells. Representative flow cytometry plots of CD45RA and CCR7 expression (L), illustrating the memory phenotype distribution of MSLN-CAR-T and ENV-CAR-T cells. Representative flow cytometry plots of PD-1 and TIM-3 expression (M), assessing T cell exhaustion in MSLN-CAR-T and ENV-CAR-T cells. (N–O) Cytotoxic function of CAR-T cells under hypoxia. Cytotoxic activity of MSLN-CAR-T and ENV-CAR-T cells after 6 hours of co-culture with HCC1806, CAPAN2, OVCAR3 and SKOV3-luc tumor cells (N). Measurement of cytokine release following co-culture under hypoxia (O). (P) The real-time cytotoxicity assay is used to evaluate the lysis of CAR-T on MSLN− and MSLN+ cells. (Q) Quantitative cleavage efficiency of GFP-labeled tumor spheroids co-cultured with CAR-T cells. (R) Long-term cytotoxicity of MSLN-CAR-T and ENV-CAR-T cells after four rounds of antigen stimulation (co-culture with HCC1806, CAPAN2, OVCAR3 and SKOV3-luc under hypoxia at E:T=2:1 every 24 hours). Data are presented as mean±SD (n=3). Statistical significance was determined by t-test or two-way ANOVA (*p<0.05, **p<0.01, ***p<0.001). ANOVA, analysis of variance; CAPAN2, human pancreatic adenocarcinoma cell line; CAR, chimeric antigen receptor; E:T, effector/target cell; FITC, Fluorescein Isothiocyanate; HCC1806, human triple-negative breast cancer cell line; IFN-γ, interferon-gamma; α-KG, α-Ketoglutarate; MFI, Mean Fluorescence Intensity; MSLN, mesothelin; ns, not significant; OVCAR3, human ovarian adenocarcinoma cell line; PD-1, programmed death protein 1; SKOV3, human ovarian adenocarcinoma cell line; SLC38A2, solute carrier family 38 member 2; 293T, human embryonic kidney 293 cells, expressing the SV40 large T antigen; TNF-α, tumor necrosis factor alpha.

To assess functional benefits, we next examined metabolic performance. Under hypoxic and glutamine-limited conditions, ENV-CAR-T cells exhibited superior metabolic adaptation compared with conventional CAR-T cells, as evidenced by enhanced proliferation (figure 3C, online supplemental figure S3D), increased intracellular glutamine levels and ATP production (figure 3D,E), and reduced apoptosis and ROS accumulation following antigen stimulation (figure 3F,G and online supplemental figure S3E,F). This suggested superior metabolic fitness. To assess whether ENV-CAR-T cells reshaped the metabolic state, we evaluated mitochondrial OXPHOS by measuring the OCR. ENV-CAR-T cells exhibited significantly enhanced mitochondrial respiration compared with conventional CAR-T cells, with mitochondrial ATP production accounting for 60.75% of total ATP (vs 27.63% in controls, p<0.001) (figure 3H and online supplemental figure S3G–J). Meanwhile, quantitative analysis of glycolytic activity via ECAR revealed a pronounced increase in glycolysis in ENV-CAR-T cells (figure 3I and online supplemental figure S3K–N). Analysis of the basal OCR/ECAR ratio indicated that ENV-CAR-T engineering preserved the balance between OXPHOS and glycolysis (online supplemental figure S3O). Notably, ENV-CAR-T cells achieved markedly higher total cellular ATP levels under hypoxic conditions (1.55-fold increase, p=0.0375) (online supplemental figure S3P), primarily sustained by enhanced OXPHOS capacity. These results collectively demonstrate that ENV-CAR-T cells exhibit enhanced metabolic flexibility in hypoxia, with coordinated upregulation of both OXPHOS and glycolysis sustaining energy production.⁴⁴ To further dissect the metabolic advantage of ENV-CAR-T cells, we found that despite elevated ECAR, glucose uptake and lactate production were unchanged (online supplemental figure S3Q,R), suggesting that the increase in ECAR may reflect enhanced proton efflux capacity and mitochondrial flexibility rather than excessive glycolytic flux. In contrast, α-KG levels and NAD⁺/NADH ratios were significantly elevated (online supplemental figure S3S,T), indicating selective enhancement of OXPHOS via glutamine metabolism.

The metabolic flexibility observed in ENV-CAR-T cells prompted us to examine underlying transcriptional changes. QPCR analysis revealed coordinated upregulation of glycolysis-related (HK2, +2.1 fold; GPI, +2.7 fold) and glutaminolysis-related genes (GLS, +2.7 fold; GOT1, +6.1 fold; figure 3J,K and online supplemental figure S3U), corroborating their bioenergetic adaptability. Notably, ENV-CAR-T cells exhibited hypoxia-sensing capability through HIF1A upregulation (3.3-fold), coupled with MTOR activation (2.2-fold)—a dual regulatory axis known to balance glycolytic flux and mitochondrial biogenesis.45,48 This aligns with their transcriptional profile (HK2↑, GLS↑) and functional metabolic plasticity, suggesting coordinated regulation of energy metabolism and hypoxia sensing. Phenotypic characterization revealed that ENV-CAR-T maintained a greater proportion of central memory T cells following antigen stimulation (figure 3L and online supplemental figure S3V), alongside a substantial reduction in exhaustion marker expression (figure 3M and online supplemental figure S3W). Notably, even in the absence of stimulation, ENV-CAR-T preserved its memory phenotype and exhibited lower basal exhaustion levels, indicating an intrinsic resistance to hypoxia-induced dysfunction. Functional assays further supported the metabolic advantage of ENV-CAR-T, as these cells exhibited greater persistence and cytotoxic capacity under repeated antigen stimulation. In acute tumor lysis assays performed under hypoxia, ENV-CAR-T showed a consistently enhanced killing efficiency across multiple MSLN+cancer cell lines (HCC1806, CAPAN2, OVCAR3, SKOV3), which positively correlated with their respective levels of target antigen expression (figure 3N and online supplemental figure S3X).

Cytokine profiling indicated a favorable safety-efficacy balance for ENV-CAR-T. While proinflammatory cytokines associated with robust antitumor activity (eg, IFN-γ and TNF-α) were upregulated (figure 3O), co-culture experiments in the presence of macrophages showed that IL-6 and IL-10 levels were comparable between MSLN-CAR-T and ENV-CAR-T cells, with no significant differences observed (online supplemental figure S3Y). These results suggest that ENV-CAR-T does not increase the risk of CRS relative to conventional CAR-T cells under these conditions. Real-time killing assays and three-dimensional tumor spheroid models corroborated these results, demonstrating superior tumor penetration and sustained effector function under hypoxia (figure 3P,Q and online supplemental figure S3Z). Repeated tumor stimulation assays further underscored the resilience of ENV-CAR-T, as these cells maintained effective cytotoxicity even after multiple rounds of tumor challenge, whereas unmodified CAR-Ts exhibited progressive functional decline (figure 3R).

To further verify the mechanistic contribution of SLC38A2 to ENV-CAR-T cell function, we performed CRISPR/Cas9-mediated knockout of SLC38A2 in ENV-CAR-T cells and systematically evaluated its impact on phenotype and effector function. Western blot confirmed efficient knockout of SLC38A2 in CRISPR-Cas9-edited cells (sgSLC38A2) compared with non-targeting control (sgNTC) (online supplemental figure S4A). Phenotypic analysis revealed a marked reduction in the proportion of central memory T cells (CD45RA-CCR7+) and increased expression of exhaustion markers PD-1 and TIM-3 in sgSLC38A2 CAR-T cells (online supplemental figure S4B,C). Functionally, SLC38A2-deficient cells exhibited impaired cytotoxicity against tumor targets in real-time killing assays (online supplemental figure S4D), along with significantly reduced glutamine uptake (online supplemental figure S4E). Consistent with a disrupted metabolic state, SLC38A2 knockout led to reduced intracellular ATP levels and elevated ROS accumulation (online supplemental figure S4F–H), indicating increased metabolic stress. Key metabolic intermediates including α-KG and NAD⁺/NADH ratios were also significantly diminished (online supplemental figure S4I,J), further implicating SLC38A2 in sustaining mitochondrial metabolism. Moreover, qPCR analysis showed that deletion of SLC38A2 led to the downregulation of key genes involved in glutaminolysis and glycolysis (online supplemental figure S4K–X), suggesting that SLC38A2 function supports the transcriptional activation of these metabolic pathways. Finally, tumor-induced cytokine release (IFN-γ, TNF-α) was markedly attenuated in SLC38A2-knockout cells (online supplemental figure S4Y), indicating functional impairment.

Collectively, these data demonstrate that SLC38A2 overexpression directly contributes to the enhanced metabolic fitness and effector function of ENV-CAR-T cells. Its deletion not only compromises energy homeostasis and metabolic signaling but also impairs antitumor responses, confirming SLC38A2 as a key functional component of this dual-regulatory CAR-T strategy.

Superior in vivo efficacy of ENV-CAR-T is associated with enhanced persistence and reduced exhaustion

To further validate the superior antitumor efficacy observed in vitro, we established an HCC1806 triple-negative breast cancer cell line-derived xenograft (CDX) model and monitored tumor progression following CAR-T cell treatment (figure 4A). Consistent with the in vitro findings, ENV-CAR-T treatment resulted in the most pronounced tumor growth inhibition, with significantly smaller tumor volumes compared with PBS, MSLN-CAR-T, 5V2-HRE and SLC-CAR-T groups from day 8 post-treatment onward (figure 4B, online supplemental figure S5A–D and H, p<0.05). Importantly, body weight remained stable across all groups, and no significant weight loss was observed in ENV-CAR-T treated mice, indicating a favorable safety profile (figure 4C).

Figure 4. SLC38A2 overexpression or hypoxia-responsive modification improves CAR-T cell therapeutic efficacy in a TNBC tumor model. (A) Schematic of the experimental protocol. NKG mice were subcutaneously injected with 2×10⁶ HCC1806 cells to establish a TNBC model. On day 12 postinoculation, mice received intravenous infusion of 1.5×10⁶ CAR-T cells or PBS (control). (B–C) Tumor volume and weight were monitored every 2 days after CAR-T cell infusion (n=6 mice per group). (D–E) Frequency of CAR-T cells in peripheral blood was evaluated by flow cytometry on day 8 (D) and longitudinally on days 4, 8, 12, 16, and 20 postinfusion (E) (n=3 mice per group). (F–G) Memory phenotype (CD45RA+CCR7+) and exhaustion markers (PD-1+TIM 3+) of CAR-T cells were quantified by flow cytometry in peripheral blood on days 4, 8, 12, 16, and 20 (n=3 mice per group). (H) MSLN CAR copy numbers in peripheral blood were quantified by qPCR targeting the WPRE region on days 4, 8, 12, 16, and 20 postinfusion (n=3 mice per group). (I–L) Serum cytokine levels (IFN-γ, TNF-α, IL-6, IL-10) were measured using cytometric bead array (CBA) on days 4, 8, 12, 16, and 20 (n=3 mice per group). (M) Representative immunofluorescence images of tumor sections on day 8 postinfusion, stained for CD45 (red), HIF1α (green), and DAPI (blue). Scale bar, 50 µm. (N–O) Quantification of CD45+ cell infiltration in tumor regions by immunofluorescence (N) and immunohistochemistry (O). Representative IHC images are shown in online supplemental figure S5I. (P) Representative H&E staining of major non-target organs (heart, liver, spleen, lung, kidney, brain, stomach and intestines) collected at the experimental endpoint (day 28). Scale bar, 50 µm. Data were analyzed by the two-tailed unpaired Student’s t-test (B), one-way ANOVA with Tukey’s multiple (C) and two-way ANOVA with Tukey’s multiple (N–O). *p<0.05, **p<0.01, ***p<0.001. ANOVA, analysis of variance; BW, Body Weight; CAR, chimeric antigen receptor; DAPI, 4′,6-Diamidino-2-Phenylindole; HCC1806, human triple-negative breast cancer cell line; HRE, hypoxia response elements; IFN-γ, interferon-gamma; IL, interleukin; i.v., intravenous; IOD, Integrated Optical Density; MSLN, mesothelin; ns, not significant; PBS, phosphate-buffered saline; PD, Pharmacodynamics; PD-1, programmed death protein 1; PK, pharmacokinetics; qPCR, quantitative PCR; TCM, central memory T cells; TNBC, triple-negative breast cancer; TNF-α, tumor necrosis factor alpha TSCM, T Stem Cell Memory (T cells); WPRE, Woodchuck Hepatitis Virus Posttranscriptional Regulatory Element.

A higher proportion of CD45RA-CCR7+central memory-like T cells was detected in the ENV-CAR-T group, implying improved long-term antitumor potential (figure 4F). This observation aligns with the increased CAR-T cell numbers quantified in peripheral blood (figure 4D,E and H). Additionally, peripheral blood analysis revealed that ENV-CAR-T treated mice exhibited significantly reduced expression of exhaustion markers PD-1 and TIM-3 on circulating CAR-T cells, suggesting enhanced persistence relative to other CAR-T groups (figure 4G). Serum levels of cytokines illustrated the enhanced antitumor activity and safety of engineered CAR-T (figure 4I–4L).

Histological and IHC analyses provided further insights into efficacy and safety. IF staining of tumor tissues showed increased CD45+T cell infiltration following ENV-CAR-T treatment, indicating enhanced immune cell recruitment to hypoxic tumor regions (HIF1α+) (figure 4M). Quantitative analysis confirmed this increase in immune infiltration (figure 4N). Furthermore, IHC staining revealed significant tumor infiltration by CAR-T cells, correlating with the observed tumor regression (figure 4O and S5I). In contrast, H&E staining of major non-target organs, including the heart, liver, spleen, lung, kidney, brain, stomach and intestines demonstrated no significant pathological alterations at both the pharmacokinetic peak (day 8) (online supplemental figure S5J) and study endpoint (day 28) (figure 4P), supporting the safety of ENV-CAR-T therapy.

Long-lasting antitumor efficacy of ENV-CAR-T cells in MSLN-positive CDX models at low dose

To further evaluate the efficacy of ENV-CAR-T cells at a lower dose, we established two additional MSLN-positive CDX models using HCC1806 and OVCAR3 xenografts (figure 5A and N). In both models, ENV-CAR-T treatment significantly suppressed tumor growth compared with control groups, as evidenced by reduced tumor volume and weight over time (figure 5B,C and O,P and online supplemental figure S5E–H and S6A–C). PK analysis confirmed the persistence of CAR-T cells in circulation (figure 5D and R). Flow cytometry analysis revealed that a higher proportion of CD45RA-CCR7+central memory-like T cells was maintained post-infusion (figure 5E and S), suggesting long-term durability of the infused CAR-T cells. Additionally, ENV-CAR-T-treated mice exhibited reduced expression of exhaustion markers PD-1 and TIM-3 on CAR-T cells compared with control groups, further supporting their sustained functionality (figure 5F and T). Representative IF images on day 8 postinfusion showed markedly higher activation (coexpression of CD3, CD8, and CD69) in tumor-infiltrating ENVCART cells than in MSLNCART controls (online supplemental figure S6D,E). Furthermore, IHC staining confirmed in situ overexpression of SLC38A2 in ENV-CAR-T treated tumors (online supplemental figure S6F,G), validating transgene expression in the TME.

Figure 5. Low-dose ENV-CAR-T cells exhibit durable antitumor responses and survival benefit in multiple MSLN+ tumor models. (A) Schematic of experimental design. NKG mice were subcutaneously injected with 1×10⁶ HCC1806 cells and treated intravenously with PBS or CAR T cells (8×10⁵) on day 12 postinoculation. (B–C) Tumor volume and tumor weight were measured every 2 days (n=6). (D) Frequency of CAR-T cells in peripheral blood at indicated time points (n=3). (E–F) Flow cytometric analysis of memory (CD45RA+CCR7+) and exhausted (PD-1+TIM 3+) CAR-T cells in peripheral blood on days 4, 8, 12, 16, and 20 postinfusion (n=3). (G) CAR copy number in peripheral blood was quantified by qPCR targeting WPRE on days 4–20 (n=3). (H–K) Serum cytokine levels (IFN-γ, TNF-α, IL-6, IL-10) were measured by cytometric bead array (CBA) on days 4–20 (n=3). (L) Representative IHC images of tumor sections stained for CD45 (brown, DAB) on day 8. Scale bar: 50 µm. (M) Quantification of CD45+cell infiltration by IOD analysis (n=3 mice/group, 3 fields per section, ImageJ). (N) Schematic of experimental design. NKG mice were subcutaneously injected with 5×10⁶ OVCAR3 cells and treated intravenously with PBS or CAR T cells (8×10⁵) on day 16 post-inoculation. (O–P) Tumor volume and tumor weight were measured every 2 days (n=6). (Q) Kaplan-Meier survival curves of treated mice. (R) Frequency of CAR T cells in peripheral blood at indicated time points (n=3). (S–T) Flow cytometric analysis of memory and exhausted CAR T cells in peripheral blood (n=3). (U) CAR copy number in peripheral blood was assessed by qPCR targeting WPRE (n=3). (V–Y) Cytokine levels (IL-6, IL-10, TNF-α, IFN-γ) in serum were quantified by CBA (n=3). (Z) Representative IHC images and IOD quantification of CD45+ cells in tumors on day 8. Scale bar: 50 µm (n=3 mice/group, analyzed with ImageJ). Data are presented as mean±SD. Statistical significance was assessed using unpaired two-tailed Student’s t-tests (B, M, O, Z), one-way ANOVA followed by Tukey’s multiple comparisons test (C, F, P, T), as appropriate. The log-rank (Mantel-Cox) test was used for survival analysis (Q). *p<0.05, **p<0.01, ***p<0.001. ANOVA, analysis of variance; BW, Body Weight; CAR, chimeric antigen receptor; DAB, 3,3'-Diaminobenzidine; HCC1806, human triple-negative breast cancer cell line; IFN-γ, interferon-gamma; IHC, immunohistochemistry; IL, interleukin; IOD, Integrated Optical Density; i.v., intravenous; MSLN, mesothelin; ns, not significant; OVCAR3, human ovarian adenocarcinoma cell line; PBS, phosphate-buffered saline; PD, Pharmacodynamics; PD-1, programmed death protein 1; PK, pharmacokinetics; qPCR, quantitative PCR; s.c., subcutaneous; TCM, central memory T cells; WPRE, Woodchuck Hepatitis Virus Posttranscriptional Regulatory Element.

Molecular analysis showed that MSLN+CAR copies were detectable in peripheral blood via qPCR (figure 5G and U), and cytokine levels, as measured by CBA, indicated an active immune response and a favorable safety profile (figure 5H–5K and V–Y). IHC staining demonstrated robust CD45+T cell infiltration within tumor tissues, which correlated with the observed tumor regression (figure 5L,M and Z, online supplemental figure S5K and S6H). Notably, ENV-CAR-T treatment significantly prolonged overall survival in the OVCAR3 model, as shown by the survival curve (figure 5Q).

Regarding safety, histopathological evaluation of major organs revealed no significant toxicity (online supplemental figure S5K and S6H). In addition, as detailed in online supplemental table S1, full blood counts at day 28 posttreatment showed no significant differences in hematology parameters among PBS, MSLNCART, and ENVCART groups. Similarly, liver and kidney function tests (alanine aminotransferase, creatinine, blood urea nitrogen) in online supplemental table S2 remained within normal ranges across all groups, confirming that SLC38A2 overexpression does not cause metabolic toxicity. These findings collectively indicate that ENV-CAR-T cells retain potent and durable antitumor activity even at reduced doses, maintain robust intratumoral activation and transgene expression, and preserve a favorable safety profile.

Discussion

The immunosuppressive and metabolically hostile nature of solid tumors has long hindered the efficacy of CAR-T cell therapy,⁴⁹ necessitating innovative strategies to enhance both specificity and functional persistence. Two key features of the TME—hypoxia and nutrient deprivation—profoundly impair T cell metabolism, cytokine secretion, and persistence. Hypoxia-driven metabolic reprogramming, such as enhanced glycolysis and lactate accumulation, can lead to local acidosis, further exacerbating T cell dysfunction. These conditions compromise CAR-T efficacy by disrupting glycolytic flux, mTOR signaling, and effector differentiation. Our study demonstrates that integrating hypoxia-responsive CAR activation with glutamine metabolic reprogramming and memory phenotype engineering significantly improves tumor-targeted cytotoxicity and long-term T cell survival. These findings provide a strategy to address two key challenges in solid tumor immunotherapy: spatial control of CAR activation and metabolic adaptation to nutrient-deprived microenvironments.

A critical innovation of this work lies in exploiting tumor hypoxia—a near-ubiquitous feature of solid malignancies—as a biological gatekeeper to regulate CAR activity.⁵⁰ By coupling CAR expression to hypoxia-inducible enhancer, CAR surface density increases under hypoxic conditions (<1% O₂) compared with normoxia. Importantly, the activation threshold (0.5–1% O₂) aligns with the median oxygenation levels in human solid tumors, suggesting broad applicability across carcinoma subtypes.

Beyond spatial specificity, the metabolic remodeling of CAR-T cells emerges as a pivotal determinant of their functionality in nutrient-scarce TMEs.^{51 52} Solid tumors often exhibit glutamine depletion due to competitive consumption by cancer cells, a challenge circumvented here by overexpressing glutamine transporters (SLC38A2). This engineering strategy enables CAR-T cells to sustain OXPHOS even under glucose-limited conditions, as evidenced by increased basal OCRs, elevated intracellular α-KG levels, and a higher NAD⁺/NADH ratio. Notably, glutamine-driven OXPHOS not only preserves energy homeostasis but also promotes a stem-like memory phenotype—the CD45RA+CCR7+ subset expands in CAR-T cells. These observations align with recent evidence that OXPHOS supports T cell memory differentiation,⁵³ whereas glycolytic metabolism accelerates terminal exhaustion.⁵⁴

ENV CAR-T cells exhibit marked downregulation of exhaustion markers and upregulation of memory-associated transcription factors compared with conventional counterparts. Such phenotypic shifts likely arise from synergistic interactions between hypoxia-primed metabolic adaptations and reduced chronic activation, as unrestrained CAR signaling is known to drive terminal differentiation.⁵⁴ In murine models, this translates to sustained tumor suppression (>60 days postinfusion) underscoring the translational potential of this approach.

Nevertheless, several challenges warrant consideration. First, intratumoral heterogeneity in oxygen and nutrient gradients may necessitate combination therapies with vascular normalizing agents to enhance CAR-T cell infiltration. Second, constitutive overexpression of metabolic enzymes could theoretically predispose cells to genomic instability—a risk mitigated in future designs through hypoxia-regulated CRISPRa systems for dynamic pathway control. Finally, while murine models demonstrate robust efficacy, interspecies differences in TME physiology may limit direct clinical extrapolation, mandating validation in patient-derived xenografts or primed immune-humanized models. Moreover, preliminary results indicate that our single-vector platform is amenable to efficient manufacturing; however, further optimization and GMP validation are required prior to clinical application.

Collectively, this study establishes a paradigm shift in CAR-T cell engineering, where synthetic biology principles are harnessed to align T cell functionality with the physicochemical constraints of solid tumors. By decoupling target recognition from environmental licensing, we not only expand the repertoire of targetable antigens but also redefine the metabolic rules governing T cell survival. Future efforts should explore multiplexed sensing of additional TME cues (eg, pH, extracellular ATP) to achieve multilayered control over therapeutic activity, ultimately bridging the gap between CAR-T promise and clinical reality in solid malignancies.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.