Potent antitumor efficacy of anti-GD2 CAR T cells in H3-K27M+ diffuse midline gliomas

Abstract

Diffuse intrinsic pontine glioma (DIPG) and other diffuse midline gliomas (DMGs) with mutated histone H3 K27M (H3-K27M)^1-5 are aggressive and universally fatal pediatric brain cancers ⁶ . Chimeric antigen receptor (CAR)-expressing T cells have mediated impressive clinical activity in B cell malignancies^7-10, and recent results suggest benefit in central nervous system malignancies^11-13. Here, we report that patient-derived H3-K27M-mutant glioma cell cultures exhibit uniform, high expression of the disialoganglioside GD2. Anti-GD2 CAR T cells incorporating a 4-1BBz costimulatory domain ¹⁴ demonstrated robust antigen-dependent cytokine generation and killing of DMG cells in vitro. In five independent patient-derived H3-K27M⁺ DMG orthotopic xenograft models, systemic administration of GD2-targeted CAR T cells cleared engrafted tumors except for a small number of residual GD2^lo glioma cells. To date, GD2-targeted CAR T cells have been well tolerated in clinical trials^15-17. Although GD2-targeted CAR T cell administration was tolerated in the majority of mice bearing orthotopic xenografts, peritumoral neuroinflammation during the acute phase of antitumor activity resulted in hydrocephalus that was lethal in a fraction of animals. Given the precarious neuroanatomical location of midline gliomas, careful monitoring and aggressive neurointensive care management will be required for human translation. With a cautious multidisciplinary clinical approach, GD2-targeted CAR T cell therapy for H3-K27M⁺ diffuse gliomas of pons, thalamus and spinal cord could prove transformative for these lethal childhood cancers.

Figure 1. GD2 is an immunotherapy target in DIPG

(a) Top 68 cell surface antigens expressed on DIPG as determined using flow cytometry screening of a monoclonal antibody panel in patient-derived DIPG cell cultures (complete data available in Supplementary Table 1). (b) Assessment of hit overlap between screened cultures identified a total of 36 hits present at a mean fluorescence intensity (MFI) of at least 10 times isotype control in all screened cultures. (c) Flow cytometry staining of histone 3 K27M DIPGs reveals high, generally homogeneous GD2 expression in contrast to histone 3 WT pediatric high-grade glioma cultures VUMC-DIPG10, diagnosed as a DIPG, and SU-pcGBM2, which arose in cortex. (d) Double immunohistochemistry of primary DIPG tumor specimens utilizing an antibody against mutant H3K27M (brown) to identify tumor cells and the anti-GD2 mAb 14g2a (blue) reveals extensive local GD2 expression in primary DIPG (scale bar = 100 microns). (e) Schematic of the GD2.4-1BB.z-CAR utilized in functional experiments. (f/g) GD2-CAR, but not CD19-CAR T-cells, mediate potent lysis (f) and produce high levels of IFN-gamma and IL-2 (g) following co-culture with GD2^hi H3K27M DIPG cells, but not GD2^lo/neg H3WT tumor cells. (h) GD2-CAR T-cells do not produce substantial levels of IFN-gamma or IL-2 following co-culture with H3K27M GD2^neg line generated using CRISPR/Cas9 to knockout GD2 synthase compared with unmodified control cells or Cas9 targeting the control AAVS1 locus. Data as shown are mean±SEM, n=3 for in vitro cytokine and cell lysis experiments. In (f–h), n=3 independent samples; experiments in (c–d) were repeated twice.

Figure 2. GD2-CAR T-cells mediate potent and lasting antitumor response in DIPG orthotopic xenografts

(a) Bioluminescence imaging of NSG mice xenografted with luciferase-expressing SU-DIPG6 into the pons (color map for all images: radiance, min = 5E4, max = 5E6) and infused intravenously with 1×10⁷ GD2-CAR or CD19-CAR T-cells as designated. Each column represents one mouse; each row represents time point of imaging. Antitumor response was observed between 14 and 28 days post treatment (DPT) in GD2-CAR T-cell treated mice. (b) Tumor burden over time expressed as fold change in flux. (c) Quantification of H3K27M+ tumor cell density within infiltrated brainstem regions of SU-DIPG6 GD2-CAR (n=5) vs. CD19-CAR (n=3) T-cell treated mice. Within GD2-CAR T-cell-treated SU-DIPG6 xenografts, we identified approximately 36 H3K27M+ cells remaining per mouse in the sampled volume, compared with approximately 18,596 cells per mouse in the sampled volume of CD19-CAR T-cell treated controls. (d) Representative immunofluorescence confocal microscopy of CD19-CAR and GD2-CAR treated SU-DIPG6 tumors staining for the mutant histone H3K27M (green). (e–i) GD2-CAR activity in a second patient-derived orthotopic xenograft model of DIPG, SU-DIPG13FL. (e–f) Bioluminescent imaging over time as above. (g) Representative immunofluorescent confocal microscopy of SU-DIPG13FL xenografts treated with CD19- or GD2-CAR T-cells reveals clearance of H3K27M+ tumor cells. (h) Tiled immunofluorescence images across engrafted regions. (i) Quantification of H3K27M+ tumor cell density within infiltrated brainstem regions of SU-DIPG13FL. In SU-DIPG13FL xenografts, approximately 32 total H3K27M+ cells remained in the sampled volume of each GD2-CAR T-cell treated mouse, compared to approximately 31,953 cells per mouse in the sampled volume of CD19-CAR T-cell treated controls. Data as shown are mean±SEM. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05 by unpaired 2-tailed Student’s t-test with Holm-Sidak correction for multiple comparisons applied for bioluminescence imaging data. Scale bars = 100 microns. Experiments were replicated in two independent cohorts of mice. In all panels, n values indicate independent animals.

Figure 3. GD2-CAR T-cell therapy improves survival in DIPG orthotopic xenografts

(a) Survival analysis of GD2-CAR T-cell treated orthotopic xenografts in SU-DIPG-13P*, a particularly aggressive patient-derived xenograft model of DIPG that is lethal within one month of engraftment, reveals a robust survival improvement in GD2-CAR T-cell treated animals (p<0.0001 Log-rank (Mantel-Cox) test, n=22 CD19-CAR and 23 GD2 CAR across 3 independent cohorts (Supplementary Figure 6)). While CD19-CAR T-cell treated xenografts were universally lethal by study endpoint, all GD2-CAR T-cell-treated animals that survived the acute toxicity of therapy survived to study endpoint at which time they manifested GVHD-like symptoms (Supplementary Figure 7). (b) Hematoxylin-eosin staining of SU-DIPG13P* xenografts at DPT50 demonstrate clearance by GD2-CAR T-cells of highly-infiltrative parenchymal tumor observed throughout the brain in CD19-CAR T-cell-treated controls and normal gross tissue architecture. (c) Hematoxylin-eosin staining of SU-DIPG6 GD2-CAR T-cell-treated xenograft analyzed at DPT14 demonstrates ventriculomegaly but histologically normal-appearing neurons in cortex, hippocampus, and brainstem (inset images). (d) Fluorescence microscopy of DPT7 SU-DIPG13FL xenografts reveals intravenously-administered GD2-CAR-mCherry T-cells infiltrating the engrafted tumor. (e) Representative image of infiltrating GD2-CAR-mCherry T-cells at DPT14 in a SU-DIPG13FL xenografted medulla demonstrates spatial association with Iba1+ macrophages. (f) Representative image of GD2-CAR-mCherry T-cell-mediated tumor cell killing with apoptosis of GFP+ tumor cells as evidenced by co-localization with cleaved caspase 3+. (g) Tumoricidal activity occurs in proximity to non-apoptotic NeuN+ neurons in the xenografted pons, shown here at DPT7 (Supplementary Figure 9). (h) Representative images of GD2-CAR-mCherry T-cells infiltrating the parenchyma of SU-DIPG13FL xenografts during the period of acute antitumor activity (Supplementary Figure 8). Experiments in (b–h) were repeated twice.

Figure 4. GD2 CAR T-cell therapy effectively clears other midline H3K27M mutant pediatric diffuse midline gliomas but is associated with toxicity in thalamic xenografts

(a) Anatomic sites of origin of H3K27M+ DMGs. (b,c)Patient-derived culture models of H3K27M mutant tumors that arose in the thalamus (QCTB-R059) and spinal cord (SU-pSCG1) highly and uniformly express GD2 as assessed by flow cytometry (b) and induce antigen-dependent secretion of IFNγ and IL-2 when incubated with GD2 or CD19-CAR T-cells in vitro (c). (d,e) SU-pSCG1 cells stably transduced to express GFP and luciferase were engrafted into the medulla of NSG mice and treated with intravenous infusion of 1×10⁷ GD2-CAR T-cells (n=10) or CD19-CAR T-cells (n=9), and substantial clearance of engrafted tumor was observed by DPT14. Each row represents one mouse over time. (f) Quantification of H3K27M+ cells remaining in SU-pSCG1 xenografts at study endpoint revealed near complete clearance of engrafted tumor in GD2-CAR T-cell treated (n=3) animals compared to CD19-CAR T-cell controls (n=3). (g) Tiled immunofluorescence images across affected regions (GFP = green; H3K27M = red, DAPI = white). (h–i) The H3K27M mutant patient-derived cell culture QCTB-R059 was orthotopically engrafted into the thalamus of NSG mice and treated by systemic administration of GD2 or CD19-CAR T-cells as described for SU-pSCG1. Tumor burden over time as determined by bioluminescence imaging illustrates marked reduction by DPT 14, and histological clearance in surviving animals at DPT 30 (Supplementary Figure 10). (j) Diagram showing the risk for 3^rd ventricular compression and herniation through the tentorium cerebelli (red) accompanying inflammation in the thalamus. (k) GD2-CAR T-cell therapy-associated deaths in mice with thalamic xenografts observed by DPT14 highlight the hazards of immunotherapy for midline tumors. Data as shown are mean±SEM. ***p<0.001, *p<0.05 by unpaired 2-tailed Student’s t-test with Holm-Sidak correction for multiple comparisons, n=3 independent samples for in vitro cytokine experiments. Experiments in (b,c,e,f,g) were performed twice.