The study was carried out in accordance with the guidelines of the Declaration of Helsinki and was approved by the local ethics committee of the Research Institute of Fundamental and Clinical Immunology (protocol #139 on May 30, 2022). Healthy blood donors (n = 6 males, age 27.33 $\pm$ 1.63 years [mean $\pm$ standard error of the mean (SEM)]) were recruited between June and December 2022. All donors met standard requirements for blood donation, showed no evidence of acute infection or prior blood transfusion, and provided written informed consent.

The human melanoma cell line SK-MEL-37 (RRID: CVCL_3878) and patient-derived cell lines S6 and V9 were provided by Professor H. Shiku (Mie University Graduate School of Medicine, Japan). The local ethics committee of the Research Institute of Fundamental and Clinical Immunology approved the use of patient-derived cell lines S6 and V9 in this study (protocol #139 on May 30, 2022). All cell lines (SK-MEL-37, S6, and V9) tested negative for Mycoplasma (Cat. #MR004, Evrogen, Moscow, Russia), and the SK-MEL-37 cell line was validated by short tandem repeat (STR) profiling. To ensure the identity of the cell lines used, the S6 and V9 patient-derived cell lines were validated using light microscopy to confirm characteristic melanoma morphology (pleomorphism, large nuclei, variable cytoplasm, and the presence/absence of melanin pigment). Cells were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (Cat. #RM10832-500ML, HiMedia, Maharashtra, India), 2 mM L-glutamine (Cat. #1383, Biolot, Saint Petersburg, Russia), 5 $\times$ 10^-5 mM $\beta$-mercaptoethanol (Cat. #M3148, Sigma-Aldrich, Saint-Louis, MO, USA), 25 mM HEPES (Cat. #1261, Biolot, Saint Petersburg, Russia), 80 µg/mL gentamicin (Cat. #A011p, PanEco, Moscow, Russia), and 100 µg/mL ampicillin (Cat. #260317, Sintez, Kurgan, Russia). Cells were seeded into flasks at a concentration of 50–75 $\times$ 10³ cells/mL until they reached the log phase of growth. They were then detached using trypsin-versene solution (Trypsin: Cat. #1225, Biolot, Saint Petersburg, Russia; Versene: Cat. #1232, Biolot, Saint Petersburg, Russia) for experimental use. GD2 expression on SK-MEL-37, S6, and V9 was assessed using flow cytometry with mAb 14.G2a (Cat. #357304, BioLegend, San Diego, CA, USA) according to the manufacturer’s recommendations and analyzed on an Attune NxT flow cytometer (Thermo Fisher Scientific, Waltham, MA, USA) (SK-MEL-37 and S6: GD2-positive; V9: GD2-negative).

A gamma-retroviral vector encoding a GD2-specific CAR was provided by Professor H. Shiku. The structure of the pGD2 transfer Moloney murine leukemia virus (MMLV) plasmid is shown in Fig. 1.

Fig. 1.

Schematic representation showing the structure of the pGD2 transfer MMLV plasmid, and the in vitro and in vivo experiments. The CAR construct includes GD2-specific scFv from a modified mAb 220-51, as well as the human signaling endodomains CD28 (transmembrane and intracellular domain) and $\xi$-chain. GD2.CAR T cells express GITRL, which affects cells through autocrine or paracrine stimulation. CAR, chimeric antigen receptor; GD2, disialoganglioside; MMLV, Moloney murine leukemia virus; IL-2, interleukin-2. Created with Biorender.com.

CAR T cells were generated by isolating PBMCs from healthy adult donors using Ficoll-Urografin (Cat. #17-1440-03, PanEco, Moscow, Russia) density gradient centrifugation and washed twice with PBS (Cat. #1247, Biolot, Saint Petersburg, Russia). Proliferation was stimulated by coating 12-well plates (TPP, Trasadingen, Schaffhausen, Switzerland) with 25 µg/mL RetroNectin (Cat. #T100B, Takara Bio, Kyoto, Japan) and 5 µg/mL anti-CD3 antibodies (Cat. #300465, Biolegend, San Diego, CA, USA). PBMCs were then plated at 0.5 $\times$ 10⁶ cells/mL in 2 mL GT-T551 medium (Cat. #WK551s, Takara Bio, Kyoto, Japan) supplemented with 300 U/mL interleukin-2 (IL-2) (Cat. #011022, Roncoleukin, Biotech, Saint Petersburg, Russia) and 0.6% group AB human serum. Half of the medium was replaced with fresh medium on days 2 and 3 supplemented with 300 U/mL IL-2. For retroviral transduction, retrovirus was diluted in PBS with 2% human serum albumin (Cat. #T160516, Microgen, Tomsk, Russia) and 5% glucose-citrate buffer, then added to RetroNectin-coated 24-well plates (25 µg/mL RetroNectin). Plates were centrifuged at 32 °C and 2000 $\times$g for 2 h (Jouan MR 23i, Saint-Herblain, Nantes, France). After removing supernatant and washing twice with PBS containing 2% albumin, the activated PBMCs (1.5–2 $\times$ 10⁵ cells/well) were transduced with CAR for 10 min under centrifugation at 1000 $\times$g at 32 °C. The next day, these cells were transduced a second time. After 6–8 h, cells were transferred to 6-well plates (TPP, Trasadingen, Schaffhausen, Switzerland) with GT-T551 medium and 300 U/mL IL-2, and cultured at 37 °C and 5% CO₂. Transduced T cells were collected on days 9–10 for analysis (transduction efficiency and phenotyping), and used for in vitro cytotoxicity assays or in vivo experiments on day 11. To validate the generated CAR T cells, the identity of these cells was confirmed by flow cytometry (assessing T lymphocyte markers [CD3, CD4, and CD8] and GD2-specific CAR) and light microscopy (assessing characteristic lymphocyte morphology: small size [7–10 µm in diameter], high nuclear-to-cytoplasmic ratio, round nucleus, scant cytoplasm).

Tumor cells (SK-MEL-37, S6, and V9) were seeded at 4–5 $\times$ 10³ cells/well in 96-well plates (TPP, Trasadingen, Schaffhausen, Switzerland) one day before co-culture with transduced or non-transduced T cells (effector to target ratio of 5:1) in fresh RPMI-1640 medium without exogenous cytokines. After 4–6 h of co-culture, T cells were collected for flow cytometry analysis of antigen-dependent activation and cytotoxicity markers. After 48 h co-culture of transduced or non-transduced T cells with GD2-positive SK-MEL-37 tumor cells, supernatants were collected and frozen for subsequent cytokine measurements. Cytotoxicity was assessed by measuring lactate dehydrogenase (LDH) activity in the conditioned medium of transduced T cells and tumor lines SK-MEL-37, S6, and V9 after 6–8 h.

Flow cytometric analysis was performed using the following monoclonal antibodies (all from Biolegend, San Diego, CA, USA): CD3 (clone HIT3a), CD4 (clone RPA-T4), CD8 (clone SK1), CD45RA (clone HI100), CD62L (clone DREG-56), CD137 (4-1BB) (clone 4B4-1), CD154 (CD40L) (clone 24-31), CD69 (clone FN50), CD178 (FasL) (clone NOK-1), CD107a (clone H4A3), CD366 (TIM-3) (clone F38-2E2), CD279 (PD-1) (clone EH12.2H7), and GD2 (clone 14G2a). These were conjugated with phycoerythrin (PE), PE-Cy7, peridinin-chlorophyll-protein complex (PerCP)-Cy5.5, allophycocyanin (APC)-Cy7, Alexa Fluor (AF) 647, AF488, AF700, Brilliant Violet (BV) 421, BV570, BV647, and BV711. Antibodies were used according to the manufacturer’s recommendations, and FMO controls established negative gates. Transduction efficiency was assessed using biotin-SP-AffiniPure F(ab’)₂-fragment-specific goat anti-mouse IgG (1:1000, Cat. #115-066-072, Jackson ImmunoResearch, West Grove, PA, USA) for the detection of GD2-CAR. Cells were stained by incubating with antibodies (1 to 1000 dilution) for 20 minutes at room temperature in the dark. After washing, streptavidin-PE (1 to 600 dilution) (1:600, Cat. #405203, Biolegend, San Diego, CA, USA), anti-CD3, and Zombie Aqua Dye (1:100, Cat. #423101, Biolegend, San Diego, CA, USA) were added to determine the population of CAR T cells and their viability. Stained cells were washed in PBS with 0.1% NaN₃ and analyzed on an Attune NxT flow cytometer. Non-transduced cells served as controls.

Flow cytometry data analysis was performed using a combination of software tools. Cells were gated using Attune NxT software (version 3.2.1, Thermo Fisher Scientific, Waltham, MA, USA) to exclude debris, doublets, and dead cells, and to identify CD3⁺ and transduced cells (Supplementary Fig. 1). The resulting .fcs files were converted to .csv format using custom Python (version 3.13, Wilmington, DE, USA) code via Jupyter Notebooks (version 6.5.4, Austin, TX, USA). Subsequently, the .csv files were arcsinh-transformed and fdaNorm-normalized using an R script [21] and exported as .fcs files. HSNE dimensionality reduction was performed using the Cytosplore app [22] on cytotoxic and activation markers (CD4, CD8, CD40L, CD69, 107a, 4-1BB, and FasL), and T cell memory and exhaustion markers (CD8, CD45RA, CD62L, CD69, PD-1, and TIM-3). Subsequently, GraphPad Prism (version 9.4.1, GraphPad Prism Software, San Diego, CA, USA) was used to export cell frequencies per cluster and visualize them as box plots.

Cytokine concentrations in supernatant were measured using the LEGENDplex™ Human CD8/NK Panel (13-plex) and Filter Plate Kit (Cat. #740267, Biolegend, San Diego, CA, USA) according to the manufacturer’s recommendations. Cytokines were subsequently classified based on their respective concentration levels.

Cytotoxicity was assessed by measuring LDH release in cell culture supernatants using a CytoTox 96 Non-Radioactive Cytotoxicity Assay (Cat. #G1780, Promega Corporation, Madison, WI, USA) according to the manufacturer’s recommendations. This assay measures LDH activity via a 30-minute enzymatic reaction that converts tetrazolium salt INT into red formazan product. The absorbance of the resulting color was measured using a Varioskan plate reader (Thermo Fisher Scientific, Waltham, MA, USA). Color intensity correlated with the number of lysed cells.

For gene expression analysis, transduced T cells were isolated using a two-step magnetic sorting procedure. First, cells were incubated with F(ab’)₂-fragment-specific goat anti-mouse IgG antibodies for 20 minutes in cold Versene solution containing 0.5% bovine serum albumin, labeled with MojoSort™ Streptavidin Nanobeads (Cat. #480016, Biolegend, San Diego, CA, USA) at a ratio of 10 µL antibodies/beads per 10 $\times$ 10⁶ cells, and then positively selected using a MojoSort™ Magnet (Cat. #480019, Biolegend, San Diego, CA, USA). After resting for 16–17 h in culture medium supplemented with 300 U/mL IL-2, the purified T cells were co-cultured with GD2-positive SK-MEL-37 tumor cells at an effector-to-target ratio of 5:1 for 2 h to activate gene expression of the immune response. Finally, transduced T cells were sorted again by positive selection using MojoSort™ Human CD45 Nanobeads (Cat. #480029, Biolegend, San Diego, CA, USA) to ensure a pure CAR T cell population. Cell viability, assessed by trypan blue staining using a Countess 3 Automated Cell Counter (Thermo Fisher Scientific, Waltham, MA, USA), was consistently $>$92%.

Total RNA was extracted from 3–6 $\times$ 10⁵ cells using the Total RNA Purification Plus Kit (Cat. #48400, Norgen Biotek, Thorold, ON, Canada) according to the manufacturer’s recommendations. RNA concentration and quality were assessed using a Nanodrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Total RNA was frozen at –80 °C until further analysis.

Gene expression analysis was performed using the NanoString nCounter SPRINT Profiler system with 100 ng of total RNA per sample (n = 6). The nCounter Human Immunology v2 panel, comprising 579 genes (Cat. #XT-CSO-HIM2-12, NanoString, Seattle, WA, USA), was used to analyze RNA samples on the NanoString Sprint instrument (NanoString, Seattle, WA, USA). Samples were hybridized according to the manufacturer’s recommendations. Normalization and quality control were performed using nSolver 4, incorporating synthetic positive controls and the 15 housekeeping genes included in the panel. To remove non-expressing genes, a background threshold was applied to the normalized data, defined as: (mean of all NEG controls) + 2 $\times$ (SD of all NEG controls) + (mean of the POS_E controls). Gene Set Enrichment Analysis was performed using GSEApy [23].

To determine the anti-tumor efficacy of transduced T cells in vivo, 8-week-old male or female non-obese diabetic (NOD)/Rag1-null/Il2r$\gamma$-null (NRG) mice (NOD.Cg-Rag^1tm1Mom Il2rg^tm1Wjl/SzIcgn) (Center for Collective Usage SPF-vivarium, Institute of Cytology and Genetics, Siberian Branch of the Russian Academy of Sciences, Novosibirsk, Russia) were injected subcutaneously in the right scapula region with 5 $\times$ 10⁶ SK-MEL-37 cells in 100 µL culture medium. When tumors reached approximately 80–100 mm³, mice were divided into a control group (intratumoral injection of 8 $\times$ 10⁶ activated non-transduced T cells in 100 µL RPMI-1640 medium), an experimental group (intratumoral injection of 8 $\times$ 10⁶ GD2.CAR T cells in 100 µL RPMI-1640 medium), and an untreated group (n = 8 per group). Tumor growth was monitored every 2–3 days using calipers, and tumor volume was calculated as: volume = 0.52 $\times$ width² $\times$ length. Mice were maintained under specific-pathogen-free (SPF) conditions with a 12-h light-dark cycle, and their health monitored regularly. Euthanasia was performed using CO₂ overdose (3 L/min fill rate, resulting in approximately 30% CO₂ concentration) followed by cervical dislocation, as indicated by clinical signs of disease (e.g., not eating, lack of activity, abnormal grooming behavior, hunched posture), excessive weight loss ($>$15% over one week), or tumor size reaching approximately 2000 mm³. The research was conducted at the Center for Genetic Resources of Laboratory Animals, Institute of Cytology and Genetics, Siberian Branch of the Russian Academy of Sciences, and approved by the local ethics committee of the Research Institute of Fundamental and Clinical Immunology (protocol #139 on May 30, 2022) as per the European Community Directive (86/609/EEC) and humane treatment standards.

Data are presented as frequencies, percentages, or mean $\pm$ SEM. The Shapiro–Wilk test was used to assess normality. Statistical significance between transduced and non-transduced cells was determined using GraphPad Prism 9.4.1, with one-way or two-way ANOVA followed by t-test, Dunnett’s test, or Tukey’s test. LDH cytotoxicity was analyzed using one-way ANOVA with Dunnett’s test for multiple comparisons. Differential gene expression and cytokine production were analyzed using multiple t-tests with FDR-adjusted p-values 0.05. Animal experiments were analyzed using two-way ANOVA with Tukey’s test for multiple comparisons. All p-values and n numbers are reported in the figure legends. A p-value 0.05 was considered statistically significant.

Three human melanoma cell lines (SK-MEL-37, S6, and V9) were analyzed for surface expression of GD2 by flow cytometry using mAb 14.G2a. The SK-MEL-37 and S6 cell lines showed high GD2 expression (89.4% and 85.1%, respectively). The V9 cell line was used as a negative control due to its low GD2 expression (2.3%) and antigen density on the cell surface.

The second-generation CAR construct containing GD2-specific scFv from a modified mAb 220-51 [19] and human signaling endodomains CD28 and CD3$\zeta$, was transduced into activated T lymphocytes using retroviral vectors (Fig. 1). Seven days after transduction, we assessed GD2-specific CAR expression by flow cytometry and observed 68.2 $\pm$ 3.5% (n = 6, mean $\pm$ SEM) GD2.CAR⁺ T cells. In order to provide strong co-stimulation and a potential synergistic effect of CD28 signaling in T cell activation [20], T lymphocytes were also genetically engineered to co-express the GITRL with CAR, as demonstrated in a recent study [19].

To further evaluate the signature of GD2.CAR T cell memory subpopulation changes, flow cytometry and HSNE analysis were used to assess marker expression according to T cell memory subsets immediately after transduction and before co-culture with the tumor cells. CAR T cells were classified into four differentiation subsets based on CD45RA and CD62L expression: naive/stem memory T cells (Tn/scm, CD62L⁺CD45RA⁺), central memory T cells (Tcm, CD62L⁺CD45RA^‒), effector memory T cells (Tem, CD62L^‒CD45RA^‒), and late effector memory T cells (Temra, CD62L^‒CD45RA⁺) [24]. HSNE analysis (Fig. 2) revealed the CD8⁺ GD2.CAR T cell population was mainly composed of Tn/scm cells, with only small percentages of Tcm and Temra cells observed (Supplementary Fig. 2).

Fig. 2.

HSNE analysis of the expression of T cell memory and activation/exhaustion markers in CD8⁺ GD2.CAR T cells. Single-cell level HSNE dot plots show the expression of T cell memory (CD62L, CD45RA) and activation/exhaustion (PD-1, TIM-3) markers by CD8⁺ GD2.CAR T cells after transduction. The color scale represents the level of expression for arcsinh transformation phenotypic markers, where violet color indicates low expression and yellow color indicates high expression. The clusters are color-coded based on the heat map (n = 5). HSNE, hierarchical stochastic neighbor embedding; PD-1, programmed cell death protein 1; TIM-3, T-cell immunoglobulin and mucin domain-containing protein 3.

Expression of the cell exhaustion markers programmed cell death protein 1 (PD-1) and T-cell immunoglobulin and mucin domain-containing protein 3 (TIM-3) was also evaluated. In general, the GD2.CAR T cell population did not show signs of exhaustion. However, a correlation was observed between the expression of CD45RA and CD62L markers and the expression of TIM-3 and PD-1 markers. The expression of cell exhaustion markers gradually increased as the cells differentiate into Tcm cells, and decreased in Temra cells (Fig. 2). It was previously reported that the increase in CD69 level is likely a manifestation of the basal level of activation by CAR-mediated tonic signaling [25]. Our transduction protocol resulted in a population of CD69^‒ CAR T cells, indicating the potential absence of tonic signaling. Meanwhile, only a low proportion of CD69⁺ T cells was observed (Fig. 2). Furthermore, the CD69 marker is expressed only when an effector cell recognizes an antigen [26].

To investigate how retroviral transduction affects the expression of genes related to phenotypic signatures (Fig. 3), we used NanoString technology to compare the transcriptome profiles of transduced and non-transduced cells. After transduction, GD2.CAR T cells showed upregulation of genes associated with the CD8 Tn cell differentiation signature, including CD8A, CD8B, SELL, IL7R, CD45RA [27], and IL2RG [28], as well as genes associated with the CD8 Tscm/cm cell differentiation signature, such as FAS and CD45RO [27]. Upregulation of the CD40LG gene, which is canonically expressed on CD4⁺ T cells, was also observed in CAR⁺ cells [29]. In contrast, transduced cells showed downregulation of CD4 Treg-associated genes, including CD4, FOXP3, IL2RA, and CTLA4-TM [27]. It is worth noting that retroviral transduction had an impact on T cell proliferation, whereas anti-CD3-primed non-transduced cells showed upregulation of genes related to T cell activation, such as TNFRSF4 and TNFRSF9 [30]. Furthermore, GD2.CAR T cells showed increased expression of the HAVCR2 gene within the T cell exhaustion signature, whereas non-transduced cells showed increased expression of the LAG3 gene.

Fig. 3.

Differential gene expression in transduced versus non-transduced T cells related to the Tn/scm phenotype. (A) The volcano plot shows significantly up- and down-regulated genes. Violet color indicates up-regulated genes in transduced cells, while green color indicates down-regulated genes in non-transduced cells. q-values 0.005 (dotted line on the y-axis) and log₂ (fold-change) $>$0.847 or log₂ (fold-change) –0.847 (dotted lines on the x-axis) were considered to be significant. (B) The heat map indicates z-score normalized gene expression of transduced and non-transduced T cells. The color scale represents the level of expression for each gene (blue color indicates low expression, yellow color indicates high expression).

Using NanoString technology, we performed differential gene analysis of GD2.CAR T cells before and after co-culture with SK-MEL-37 cells. Because our aim was to examine changes in gene expression linked to the immune response before and after co-culture with tumor cells, we selected the Human Immunology v2 panel for this study. CAR⁺ cells were first magnetically sorted from the general cell population and then stimulated with IL-2 for 16–17 h. GD2.CAR T cells were then cultured with tumor cells for 2 h to activate key genes that regulate the immune response. These cells were then separated from GD2-positive SK-MEL-37 cells using CD45-positive magnetic sorting to remove any contamination with tumor cells.

Although the interaction between GD2.CAR T cells and tumor cells was short, alterations were detected in several genes (Supplementary Table 1) encoding co-signaling molecules, cytokines, and chemokines that affect the immune response (Fig. 4A).

Fig. 4.

Comprehensive analysis of gene expression, signaling, and phenotype in GD2.CAR T cells following co-culture with a melanoma cell line. (A) The volcano plot shows differentially expressed genes in GD2.CAR T cells before and after co-culture with SK-MEL-37 cells. Green color indicates up-regulated genes, while violet color indicates down-regulated genes. q-values 0.005 (dotted line on the y-axis) and log₂ (fold-change) $>$0.847 or log₂ (fold-change) –0.847 (dotted lines on the x-axis) were considered to be significant. (B) A GSEA bubble plots show Gene Ontology Biological Process terms that were significantly enriched in up-regulated genes in the GD2.CAR T cells. The bubble size indicates the percentage of genes involved in the respective biological process or pathway. The color represents the corrected p-value of the enriched pathway (q-values 0.000001). The top pathways involved in biological processes are shown on the x-axis. (C) Single-cell level HSNE dot plots show the expression of activation and cytotoxicity markers (CD69, 4-1BB, CD107a, FasL, and CD40L) by CD8⁺ GD2.CAR T cells after 4–6 h of co-culture with SK-MEL-37 cells. The color scale represents the level of expression for arcsinh transformation cytotoxic markers, where violet color indicates low expression and yellow color indicates high expression. The clusters are color-coded based on the heat map (n = 6).

Signatures for chemokines (CCL2, CCL20, CCL3, CCL4, CXCL1, CXCL10, CXCL11, and CXCL2), cytokines and their receptors (CSF1, CSF2, IL13, IL13RA1, IL1A, IL1B, IL1RAP, IL6, IL8, and IFNG) were found after co-culture with the tumor cell line. In contrast, a decrease in the expression of chemokine receptor genes (CCR1, CCR2, CCR5, CCR7, CCRL2, CX3CR1, CXCR3, CXCR4, and CXCR6) was observed, as well as that of cytokine receptors and their component genes (IL10RA, IL12RB1, IL21R, IL2RG, IL4R, and IL7R). We also observed a decrease in the expression of genes associated with the Tem cell differentiation signature, including CCR7, CD27, CD28, CD45RA, CD45RB, CX3CR1, IL7R, LEF1, and TCF7 [27]. Furthermore, increased expression of the CTNNB1 [31, 32] and BCL6 [33] genes and decreased expression of the LAIR1 [34] and SELPLG [35] genes are also signatures of Tem cell differentiation. Regarding the effector function of T cells, increased expression of genes associated with activation was observed, including TNFRSF9 [30] and MHC class II genes (HLA-DMA, HLA-DMB, HLA-DPA1, HLA-DPB1, HLA-DQA1, HLA-DRA, HLA-DRB1, and HLA-DRB3), as well as genes associated with migration (IL6, IL6ST, SELE, and ICAM1) [36]. We also observed upregulation of TRAF1 gene expression, which is involved in downstream signaling of 4-1BB [37] or GITR [38]. Upregulation of EGR2 gene expression was also observed, which promotes differentiation toward Tem cells and regulation of the exhausted transcriptional state [39], as well as improved survival and proliferation [40]. However, the expression of genes encoding effector molecules (GZMA, GZMK, and pRF1) and signaling molecules (LCK and ZAP70) essential for the initiation of TCR signaling were downregulated [41]. Notably, after co-culture of GD2.CAR T cells with tumor cells, decreases were observed in signatures associated with T cell dysfunction, including expression of the HAVCR2, CISH [42, 43], CD244 and SLAMF7 [44] genes, as well as T cell death genes (BAX, BCL2, CASP1, and CASP8) [45, 46].

The up-regulated genes (Supplementary Table 2) were enriched by gene set enrichment analysis (GSEA) [23] using ontology biological process terms. GSEA analysis revealed that the top three pathways were associated with the cytokine-mediated signaling pathway, cellular response to interferon-gamma, and interferon-gamma-mediated signaling pathways (Fig. 4B). In contrast, genes associated with T cell proliferation and cytokine production were not significantly expressed during early stages of the CAR T cell-tumor interaction.

To investigate the potential application of CAR T cells as effector cells against melanoma, we evaluated the expression of surface activation and cytotoxicity markers (CD69, 4-1BB or CD137, CD107a, FasL, and CD40L) following 4–6 h of co-culture with GD2-positive tumor cell lines. This duration is sufficient to induce expression of the markers. Using a panel of antibodies, we analyzed the flow cytometry data from the GD2.CAR T cells using HSNE dimensionality reduction after arcsinh transformation with automated cofactors. Data were normalized using fdaNorm to correct for batch effects and ensure data integrity. HSNE analysis identified four GD2.CAR T cell subpopulations (Fig. 4C). The predominant population was CD8⁺CD40L⁺CD69^‒CD107a⁺4-1BB⁺FasL⁺ T cells, suggesting cytotoxic function (Supplementary Fig. 3). Smaller populations of CD8⁺CD40L⁺CD69^‒CD107a⁺4-1BB⁺FasL^‒ and CD4⁺CD40L⁺CD69^‒CD107a^‒4-1BB⁺FasL⁺ T cells with similar proportions were also observed, as well as a minor population of CD8⁺CD40L⁺CD69^‒CD107a⁺4-1BB^‒FasL⁺ T cells (Supplementary Fig. 3).

The LDH release assay was employed to assess the cytotoxic activity of GD2.CAR T cells against melanoma tumor cell lines expressing the GD2 antigen. CAR T cells were co-cultured at an effector-to-target ratio of 5:1 for 6–8 h with tumor cells that expressed GD2 (SK-MEL-37 and S6), or did not express GD2 (V9) (Fig. 5). This ratio and co-culture duration were based on the manufacturer’s protocol and optimized for reliable results in the LDH release assay. Significant differences in GD2-specific anti-tumor cytotoxicity, as measured by LDH release, were observed after co-culture with the GD2-expressing SK-MEL-37 and S6 tumor cell lines. Importantly, no significant differences in cytotoxicity were observed between the SK-MEL-37 and S6 cell lines. These results demonstrate the efficacy and specificity of GD2.CAR-T cells in targeting GD2-expressing melanoma cells.

Fig. 5.

GD2.CAR T cell cytotoxicity against melanoma cell lines. Box plots show LDH release, a measure of cytotoxicity, following 6–8 h co-culture of GD2.CAR T cells with GD2-positive SK-MEL-37 (89.4% GD2 expression), S6 (85.1% GD2 expression) and GD2-negative V9 (2.3% GD2 expression) melanoma cell lines at a 5:1 effector-to-target ratio (n = 6). Asterisks indicate significant p-values as follows: ****p 0.0001, one-way ANOVA with Dunnett’s test for multiple comparisons. These results indicate the strong cytotoxic activity of CAR T cells against GD2-positive cell lines. LDH, lactate dehydrogenase; ANOVA, analysis of variance.

Cytokine release from transduced and non-transduced T cells was assessed 48 h after contact with a GD2⁺ target at an effector-to-target ratio of 5:1. The LEGENDplex™ Human CD8/NK panel was used to simultaneously analyze cytokines, including IL-2, IL-6, and IL-17, and effector molecules, including granzyme A, granzyme B, granulysin, perforin, and sFasL, in the cell co-culture supernatants. Compared to non-transduced T cells, GD2.CAR T cells tended to secrete more pro-inflammatory cytokines and effector molecules (Fig. 6). No statistically significant differences between non-transduced cells and GD2.CAR T cells were observed for the levels of perforin, granulysin, IL-2, and IL-6 in the culture medium. However, significant increases were observed in the levels of granzymes A/B, IFN-$\gamma$, and sFasL secreted by GD2.CAR T cells. IL-4, IL-10, and sFas were not detected in the supernatants of GD2.CAR and non-transduced T cells.

Fig. 6.

Cytokine response of GD2.CAR T cells to co-culture with SK-MEL-37. (A) The volcano plot shows fold-changes in cytokine levels in co-culture supernatants of GD2.CAR T cells (green dots indicate significance) with SK-MEL-37 cells (compared to non-transduced T cells). The x-axis values represent log₂ (fold-change), and the y-axis values represent ‒log₁₀ of q-values. The q-values indicate FDR-adjusted p-values for multiple t-tests. The upper right quadrant indicates statistically significant results (q-value 0.05). (B) The heat map shows the average log FCs of cytokine levels (pg/mL) for both transduced and non-transduced cells. The color scale represents cytokine levels (blue color indicates low concentration, yellow color indicates high concentration).

We next evaluated the antitumor efficacy of generated GD2.CAR T cells against GD2-positive tumors in a xenograft model by inoculating SK-MEL-37 cells (5 $\times$ 10⁶ cells per mouse) into NRG mice via subcutaneous injection. When the tumor volume reached 80–100 mm³, tumor-bearing mice were randomized to receive a single intratumoral injection of either 8 $\times$ 10⁶ GD2.CAR T cells or 8 $\times$ 10⁶ anti-CD3-primed non-transduced T cells. Tumor growth was monitored by caliper measurements. Local intratumoral injection was chosen as the route for T cell administration based on the efficacy demonstrated in a preclinical study with locally injected CAR T cells [13]. Three weeks into the experiment, the tumor volume increased rapidly in untreated mice and in mice that received non-transduced T cells (Fig. 7). At 35 days after intratumoral administration of GD2.CAR T cells, a significant reduction in tumor size was observed compared to both the control group and the intact group.

Fig. 7.

Antitumor efficacy of GD2.CAR T cells in vivo. SK-MEL-37 cells were subcutaneously inoculated into NRG mice (n = 24) at a dose of 5 $\times$ 10⁶ cells per mouse. Once the tumor size reached 80–100 mm³, non-transduced T cells were injected into the control group, while GD2.CAR T cells were injected into the experimental group at a dose of 8 $\times$ 10⁶ cells per mouse. Results are presented as the mean tumor volume (mm³) $\pm$ SEM. Asterisks indicate significant differences between GD2.CAR T cell-treated and untreated groups, with p-values indicated as follows: *p 0.05, ***p 0.001, ****p 0.0001; grids indicate significant differences between GD2.CAR T cell and non-transduced T cell-treated groups, with p-values indicated as follows: #p 0.05, ##p 0.0001, two-way ANOVA with Tukey’s test for multiple comparisons. NRG, non-obese diabetic (NOD)/Rag1-null/Il2r$\gamma$-null (NOD.Cg-Rag^1tm1Mom Il2rg^tm1Wjl/SzIcgn); SEM, standard error of the mean.