U87 cells were obtained from the American Type Culture Collection (ATCC, USA). U251 and A172 cells were kindly provided by the Chinese Academy of Sciences’ Cell Bank/Stem Cell Bank. Cancer cells were cultured in DMEM medium supplemented with 10% fetal bovine serum (Gibco, USA), 100 μg/ml streptomycin, and 100 U/ml penicillin (Hyclone, USA). All the cells were incubated at 37°C with 5% CO₂.

Sodium valproate was obtained from Solarbio Life Science (China). DMEM media, Opti-MEM I Reduced Serum Medium, and Fetal Bovine Serum (FBS) were obtained from Gibco (USA). APC anti-human CD314 (NKG2D) antibody(Cat# 320807), APC anti-human MICA/B antibody (Cat# 320907), APC anti-human CD314 (NKG2D) antibody(Cat# 320807), APC mouse IgG2a isotype control antibody (Cat# 400219), APC Goat anti-mouse IgG antibody (Cat# 405308), APC anti-CD3 antibody (Cat# 300411) and PE anti IFN-γ (Cat# 502508) were obtained from BioLegend (USA). Anti-human ULBP1 antibody (Cat# MAB1380-SP), anti-human ULBP2/5/6 antibody (Cat# MAB1298), anti-human ULBP3 antibody (Cat# MAB1517), and mouse IgG2a isotype control (Cat# MAB003) were obtained from R&D Systems (USA). Anti-human ULBP4 antibody (Cat# sc-53133) was obtained from Santa Cruz Biotechnology (USA). Human Th1/Th2 Cytokine Cytometric Bead Array (CBA) Kit (Cat#550749), and Cytofix/Cytoperm fixation and permeabilization kit (Cat#554714) were obtained from BD biosciences (USA).

The extracellular domain of human NKG2D was synthesized (Genewiz) and fused to a CAR backbone comprising a human CD8a hinge spacer and transmembrane domain, 4-1BB costimulatory domain, and CD3ζ. The entire encoding sequence of the CAR expression molecule was cloned into the lentiviral vector pWPXLd (Addgene). For the lentiviral package, the lentiviral plasmids were co-transfected into HEK293T cells with the packaging plasmids psPAX2 and pMD2.G (Addgene) at a ratio of 5:3:2. Lentivirus was harvested 48h after transfection.

The CAR-expressing lentiviral vectors were transduced into human primary T cells as previously described, with some modifications (15). T cells were isolated from peripheral blood mononuclear cells (PBMC) obtained from healthy donors and activated with Dynabeads™ CD3/CD28 (Thermo Fisher Scientific) at a 1:1 ratio and then cultured in Corning^® KBM 581 serum free medium containing IL-2 (50 U/mL, Novoprotein). After 24 h, activated T cells were infected with the lentiviral particles. Seven days after stimulation, the beads were removed. The percentage of NKG2D⁺ cells was used to calculate transduction efficiency using flow cytometry. All the studies involving human subjects were approved by the Institutional Review Board at Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences (approved ID: SIAT-IRB-190715-H0363) with written informed consent obtained from participants and conducted in accordance with the international ethical guidelines for biomedical research involving human subjects.

Cells were harvested, washed twice with 1xPBS, and resuspended at a density of ~1×10⁶ cells/mL in cold PBS containing 0.5% BSA. Subsequently, labeled primary antibodies were added to the cell suspension according to the manufacturer’s instructions and incubated at 4°C in the dark for 0.5 h. All flow cytometry measurements were taken with the Beckman CytoFLEX flow cytometer system and analyzed with the FlowJo software.

The cytotoxicity of VPA on glioblastoma cell lines U251, U87, and A172 was assessed using the Cell Counting Kit-8 (CCK-8) assay. Glioblastoma cells were seeded at a density of 5×10⁴ cells/mL in 96-well plates. Cells were incubated for 48 h with increasing concentrations of VPA after 24 h of attachment (0, 0.4, 0.8, 1.6, 3.2, 6.4 and 12.8 mM). Cell viability was determined using a CCK-8 kit (Shanghai Yeasen Biotech) as described by the manufacturer. For flow cytometric analysis, Glioblastoma cells were seeded in 12-well plates at a cell density of 1×10⁵ cells/mL. Cells were incubated with 3.2 mM VPA for 48 h after 24 h of attachment. NKG2D ligands were detected using antibody-based flow cytometry, as previously described. For the LY294002 and PD98059 experiments, the pathway inhibitors were combined with VPA and applied to the target cells.

Cytotoxic activity of CAR-T cells in vitro was determined using the xCELLigence RTCA SP instrument (ACEA Biosciences) with minor modification (16). Target cells were seeded at a density of 5×10³ cells/well into the plate and cultured for ~24 h. Data recording was paused at the time of initiating treatments, and CAR-T or untransduced (UTD) cells were added into the target cells at an indicated effector: target (E: T) ratio in triplicate. The target cell-only control was included. The cytotoxic activity based on the viability of attached target cells was monitored for at least 24 h, as reflected by cell index (CI) values. CI was normalized at the end of the experiment to remove any well-well variation using the RTCA software Pro (version 2.3.0). Percentage cytotoxicity was calculated at the end of the experiment as:

% cytotoxicity = ((CI no effector - CI effector)/(CI no effector)) × 100.

In the experiments involving VPA treatment, 3.2 μM of VPA or PBS vehicle control was used to pretreat target tumor cells (U251, U81, and A172) for 48 h, and media was changed to complete media. NKG2D CAR-T cells were added directly to the pretreated cells at different E: T ratios.

In the 3D spheroid model, U251 cells (5×10³/well) were seeded in Corning^® spheroid 96-well microplates and incubated for 72 h under physiological conditions to form the required U251 spheroids. After 3.2 mM VPA treatment for 72 h, then NKG2D CAR-T cells were added at an E: T ratio of 2 with 0.5% propidium iodide solution (BioLegend). The red fluorescence intensity values in each well were collected and analyzed by Cytation 5 Cell Imaging Multi-Mode Reader (BioTek, USA) over 24 h co-culture.

After co-culturing CAR-T cells and glioblastoma cell lines for 24 h, the culture supernatant was collected through centrifugation at 300 g for 5 minutes and transferred to new flat-bottom 96-well plates. The cytokine levels Interleukin-2/4/6/10, interferon-gamma (IFN-γ) and tumor necrosis factor-alpha (TNF-α) were quantified by Human Th1/Th2 Cytokine Cytometric Bead Array (CBA) Kit (BD Biosciences) according to the manufacturer’s instructions. Data were analyzed using FlowJo software.

T cells (5 × 10⁵) were co-cultured with target cells treated with VPA or not at a ratio of 1:1 for 24 h followed by incubation with Brefeldin A (BD Biosciences) for 4 h. Cytofix/Cytoperm fixation and permeabilization kit (BD Biosciences) was used to permeabilize cell membrane and perform intracellular staining according to the manufacturer’s instruction. APC anti-CD3 antibody and PE anti-IFN-γ antibody were used for immunostaining. IFN-γ expression was detected via antibody-based flow cytometry as described previously.

Total RNA was extracted from cells using RNAeasy™ Animal RNA Isolation Kit with Spin Column (Beyotime Biotechnology) and reverse transcribed into cDNA using the EasyScript™ First-Strand cDNA Synthesis SuperMix (Shanghai Yeasen Biotech). The expressions of MICA, MICB, ULBP1, ULBP2, ULBP3, ULBP4, ULBP5, ULBP6, and GAPDH were quantified using the quantitative SYBR Green PCR kit (Shanghai Yeasen Biotech) according to the manufacturer’s protocol. The primers used for qRT-PCR are shown in Supplementary Table S1 .

Mouse experiments were performed in accordance with relevant guidelines and regulations and were approved by the Institutional Animal Care and Use Committee at Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences. Mice were maintained under specific-pathogen-free conditions with daily cycles of 12-h light–12-h darkness and health monitoring was carried out on a regular basis. Six-week-old female B-NDG mice used in this study were purchased from Biocytogen (Beijing, China). The mice were subcutaneously injected with 2 × 10⁶ U251 tumor cells. Tumor growth was monitored every 3 days. When the tumor volume reached around 100mm³ (calculated as volume = (width)² x length/2), the mice were treated by VPA at a dosage of 100mg/kg, or a vehicle control intravenously every two days. After VPA was administered 4 times, 5.0×10⁶ NKG2D CAR-T or untransduced T (UTD) cells were administered intravenously to the mice. Throughout the experiment, tumor volume and body weight of mice were measured and monitored. The study was concluded on day 51, and all animals were euthanized. Tumor tissues and peripheral blood samples were collected from each mouse for further analysis.

The results are presented as the mean ± SD. Normality was assessed using the Shapiro-Wilk test, and homogeneity of variances was evaluated with the Levene test. For comparisons between two groups, a two-tailed t-test was applied when normality was not rejected, while one-way ANOVA with a Tukey’s multiple-comparisons test was used for comparisons for multiple groups. When normality was rejected, the Mann-Whitney test was used for two-group comparisons, and the Kruskal-Wallis test for multiple groups comparisons. Statistical significance was indicated by an asterisk (*) for p < 0.05 between the specified groups. Experimental sample numbers (n) are indicated in figure legends.

Three representative glioblastoma cell lines (U251, U87, A172) were selected for assessing cell surface expression of NKG2DL (MICA/B, ULBP1, ULBP2/5/6, ULBP3, and ULBP4) by flow cytometry. As expected, all the tested glioblastoma cell lines expressed high levels of NKG2DL, particularly ULBP2/5/6 ( Figure 1A ). The expression of NKG2DLs in glioblastoma cells suggested that NKG2D CAR-T cells could be used as potential therapy for glioblastoma. Therefore, we constructed an NKG2D-CAR lentiviral vector with the NKG2D extracellular region, 4-1BB, and CD3 signaling domains ( Figure 1B ).

The NKG2D-CAR lentivirus was used to transduce T cells from healthy donors, and the expression of NKG2D CAR was assessed by flow cytometry 5 d post transduction. As shown in Figure 1C , a relatively high percentage of T cells (approximately 80%) were positive for NKG2D CAR expression. To assess the cytotoxicity of NKG2D CAR-T cells against glioblastoma cells, we incubated CAR-T and untransduced control T (UTD) cells with target cells (U251, U87, and A172) at various E: T ratios (10:1, 5:1, 2.5:1, and 1.25:1). When compared to UTD cells, NKG2D CAR-T cells lysed the glioblastoma cells more efficiently ( Figure 1D ).

Additionally, since cytokine secretion by CAR-T cells targeting cancer cells indicates T cell activation and specific cytotoxicity, the levels of the classic cytokines (IFN-γ, IL-2, and TNF-α) were measured when CAR-T cells were incubated with glioblastoma cells to assess the cytokine profile. In comparison to control cells, the concentrations of tested cytokines were significantly higher in the supernatant of the NKG2D CAR-T co-culture system (P < 0.001, Figure 1E ).

These findings suggested that glioblastoma cell lines overexpress NKG2D ligands and NKG2D CAR-T can efficiently lyse glioblastoma cells. As a result, NKG2D CAR-T cells could be a promising therapeutic approach for patients with glioblastoma.

The safe and effective clearance of cancer cells by CAR T cells necessitates both selective and consistent cell surface expression of the target antigen. All components of an antitumor CAR T cell response can be affected by the density of antigen expression (17). We analyzed the expression of NKG2D ligands in different types of human cancers using The Human Protein Atlas Database (http://www.proteinatlas.org). Glioblastoma exhibited moderate expression of the NKG2D ligands when compared to other types of cancer. Furthermore, glioblastoma has been shown to evade NKG2D-mediated immune detection by shedding soluble NKG2D ligands from the cell surface (11). The level of expression of the target antigen on the surface of tumor cells has a direct impact on the therapeutic effect of CAR-T cells (18). As a result, we hypothesized that pharmacologically increasing NKG2D ligand expression would improve glioblastoma cell recognition and susceptibility to NKG2D CAR-mediated immune response.

VPA ( Figure 2A ) is an antiepileptic drug that is a histone deacetylase (HDAC) inhibitor and has been shown to increase the surface expression of NKG2DL on several cancer cells (12–14). However, whether VPA increases the surface expression of NKG2D ligands on glioblastoma cancer cells has not been known. Considering the ability of VPA to cross the blood-brain barrier, we tested VPA’s ability to sensitize glioblastoma cancer cells to NKG2D CAR-T cytotoxicity.

To investigate this, we first looked for an in vitro VPA dose at which most glioblastoma cancer cells remained viable. The viability of the three glioblastoma cell lines was assessed after 48 h of treatment with increasing doses of VPA (0, 0.8, 1.6, 3.2, 6.4, or 12.8 mM). VPA at the concentration of 3.2 mM and lower had negligible cytotoxicity on these cell lines, with > 95% cancer cell viability ( Figure 2B ). We then used flow cytometry to examine the cell surface expression levels of NKG2DL on cancer cells after 48 h of exposure to 3.2 mM VPA (the subtoxic concentration). When compared to the respective untreated controls, all three glioblastoma cell lines treated by VPA showed an increase in NKG2DL expression ( Figure 2C ).

Specifically, all cell lines showed increased ULBP2/5/6 expression. MICA/B expression was also significantly increased on U87 and A172 cells, and ULBP3 expression was increased on U251 cells. On the other hand, VPA treatment had no effect on ULBP4 ligand expression in the glioblastoma cells tested.

To investigate whether the enhanced expression levels of NKG2DL on the tumor cell surface are associated with their improved recognition and cytotoxicity by NKG2D CAR-T cells, we measured the cytotoxicity of NKG2D CAR-T cells against the tumor cells after co-culture with a subtoxic concentration of VPA or the untreated glioblastoma cancer cells. As expected, VPA-induced NKG2DL upregulation improved antigen-specific recognition and cytotoxicity of NKG2D CAR T cells ( Figure 3A ).

The cytotoxicity of U251 and U87 was significantly increased at all E: T ratios, whereas A172 was significantly increased at E: T ratios of 1 and 0.5 (P < 0.05). Additionally, the ability of NKG2D CAR-T cells to produce IFN-γ (assessed via intracellular cytokine staining) was used to assess their activation after co-culture with VPA-treated or untreated U251 and U87 cells. The results indicated that VPA-treated glioblastoma cancer cells can stimulate CAR-T to secrete a higher IFN-γ response compared to the untreated groups ( Figures 3B, C ), further confirming that VPA pretreatment can enhance glioblastoma cell recognition to NKG2D CAR-T mediated immune response.

Furthermore, 3D tumor spheroids have been shown to mimic in vivo solid tumor biology more accurately. Spheroids have been shown in studies to be more resistant to chemotherapy than 2D monolayer cell cultures. 3D spheroids develop hypoxic cores and have a drug diffusion profile similar to tumors (19). Based on the above premise, we created a 3D tumor spheroids model with U251 cells and tested the cytotoxicity of NKG2D CAR-T cells after VPA-treated or untreated 3D tumor spheroids. Consistent with the results of the 2D cell model experiments, the 3D tumor spheroids treated with VPA for 48 h were more susceptible to NKG2D CAR-T cells than untreated spheroids ( Figures 3D–F ). Collectively, these findings suggest that VPA treatment increased the susceptibility of glioblastoma cancer cell lines to the antitumor immune response elicited by NKG2D CAR-T cells.

The expression of NKG2D ligands is linked to various signaling pathways in different cells, including the PI3K/Akt, ERK, HER2/HER3, and P53 pathways (20). VPA has been shown to increase the expression of NKG2D ligands in myeloma cells via an ERK-dependent mechanism (21) and to increase the expression of MICA/B in pancreatic cancer cells via a PI3K/Akt signal pathway (22).

To determine the specific signaling pathway involved in VPA-induced upregulation of NKG2D ligand in glioblastoma cancer cells, we exposed U251, U87, and A172 cells to 3.2 mM VPA for 48 h in the presence or absence of 10 μM of the PI3K inhibitor LY294002, or 20 μM of the ERK inhibitor PD98059, followed by flow cytometry analysis.

Flow cytometric data showed that LY294002 significantly inhibited VPA’s ability to upregulate the expression of NKG2D ligands (P < 0.05, Figures 4A, B ), whereas PD98059 exposure had no effect on NKG2D ligand expression (data not shown). These findings suggest that the PI3K/Akt signaling pathway, but not the ERK signaling pathway, is involved in VPA-induced upregulation of MICA/B and ULBPs in glioblastoma cancer cells. To support this observation, RT-qPCR analysis revealed that LY294002 significantly reduced the expression of NKG2D ligands upregulated by VPA treatment at the mRNA level (P < 0.05, Figure 4C ). Overall, the data show that VPA probably increases NKG2D ligands expression in glioblastoma cells via the PI3K/Akt signaling pathway.

To assess the anti-tumor efficacy of NKG2D CAR-T in combination with VPA against glioblastoma in vivo, a xenograft mouse model bearing U251 cells was established. To determine the dosage of VPA for the animal experiment, a dosage of 100mg/kg were administered intravenously to the mice model every two days, based on the published literature (23). The results showed that tumor growth was not inhibited after 4 times of VPA treatment ( Supplementary Figures S1A, B ). Therefore, the same treatment regime of VPA was used in the following combinational therapy with NKG2D CAR-T cells in vivo ( Figure 5A ).

The results revealed that NKG2D CAR-T treatment significantly inhibit tumor growth in B-NDG mice as indicated by reduced tumor volume and tumor weight; and notably, pretreatment of mice with VPA further showed significant tumor growth inhibition ( Figure 5B ; Supplementary Figures S1C, D ). In addition, no obvious difference in body weight across the different treatment groups was observed ( Figure 5C ), indicating the safety of the combinational therapy. Moreover, analysis of harvested tumor tissues showed that combination of VPA and NKG2D CAR-T treatment remarkably enhanced T cells infiltration into tumor sites as shown by increased CD3⁺ T cells (P < 0.005) in tumor tissues ( Figure 5D ; Supplementary Figure S1E ). Collectively, our findings suggest that VPA potentiates the efficacy of NKG2D CAR-T cells against glioblastoma.