Preclinical efficacy of multi-targeting mRNA-based CAR T cell therapy in resection models of glioblastoma

Abstract

Traditional viral-based chimeric antigen receptor (CAR) T cell therapies have vanquished multiple blood malignancies with decade-long remissions yet struggle against solid tumors. Nonviral engineering of CAR T cells via electroporation or lipid nanoparticle (LNP) delivery of CAR-encoding mRNA results in highly efficient yet transient CAR expression, challenging the adequacy of available preclinical models for mRNA-based CAR T cell evaluation. This study presents a unique three-pronged approach that combines mRNA-based CAR T cells, multi-targeting of glioblastoma (GBM)-associated receptors, and maximal surgical resection as a novel and readily translatable platform for preclinical evaluation of mRNA-based CAR T cells against solid tumors. We performed head-to-head in vitro and in vivo analyses of mRNA-based CAR T cells generated using different expansion conditions, mRNA delivery methods, or combination approaches. Besides potent in vitro cytotoxicity, our findings unveil a therapeutic window of anti-tumor efficacy, as well as robust and durable complete remissions in xenograft mouse models of GBM receiving maximal surgical resection and locoregional injections of multivalent CAR T cells (MVCAR). Such efficacies were significantly better in 5-day expanded versus quiescent T cells. Interestingly, MVCAR T cells were superior to pooled CAR T cells (CARPool) expressing the same CAR scFv combinations in an orthotopic resection model of GBM.

Keywords: MT: Delivery Strategies; chimeric antigen receptor T cells; electroporation; glioblastoma; immunotherapy; lipid nanoparticles; messenger RNA; multiple targeting; surgical resection; tumor-associated antigens.

Graphical abstract

Figure 1

Efficient specific lysis of U87KOpool or U87_HER2+ cells with single, dual, and triple CARPool or MVCAR combinations of selected mRNA-based CAR T cells at 10:1 effector-to-target ratio (A) Study scope and CAR designs. Two distinct multi-targeting approaches were used: (1) pooled CAR T cells (CARPool), generated by electroporating each set of cells with one type of CAR-BBz-encoding mRNA, then pooling cells in equal ratios or (2) multivalent CAR T cells (MVCAR), generated by electroporating each set of cells with two or three CAR-BBz-encoding mRNAs to generate dual or triple MVCAR (nomenclature described in the table). (B–M) Quiescent T cells EP-ed with mRNA encoding for HER2 CAR, IL13Rα2 CAR, EphA2#1 CAR, EphA2#2 CAR, EGFR CAR, or combinations thereof were cocultured with U87_HER2+ or U87KOpool target cells at effector-to-target ratios of 10:1. (B–E) Normalized impedance-based cytolysis analysis of real-time data (mean % cytolysis calculated from n = 2 different experiments ran using T cells from two different healthy donors, with two biological replicates per sample). (F–I) Flow-based cytolysis analysis after 5 days of coculture (mean ± SD of % cytolysis calculated from n = 2 different experiments ran using T cells from two different healthy donors, with two biological replicates per sample). Different healthy donors were used to source CAR T cells for each cytotoxicity assay. (F–I) Ordinary one-way ANOVA, followed by Tukey multiple comparison analysis. ∗: vs. Mock EP; #: vs. IL13Rα2 CAR. ∗ or #: p < 0.05, ∗∗ or ##: p < 0.01, ∗∗∗ or ###: p < 0.001, ∗∗∗∗ or ####: p < 0.0001.

Figure 2

Influence of expansion on single versus multi-targeting mRNA-based CAR T cells T cells obtained from three different donors for each condition (expanded or quiescent) were Mock EP-ed, or EP-ed with mRNA encoding for HER2 CAR, IL13Rα2 CAR, EphA2#2 CAR, or combinations thereof (MVCAR#2). (IL-7/IL-15)-Conditioned medium was used in either condition of T cell expansion. In some instances, single-targeting CAR T cells were combined in equal ratios (CARPool#2). CAR T cells were then cocultured with U87KOpool-CBG-GFP/NLS cells at 8, 2.4, or 0.8 CAR+ cells:target cell ratios, respectively. A set of functional assays were performed: (A and B) mean ± SD luciferase-based cytolysis at 48 h post-coculture (three replicates per donor, three donors), (C–E) mean ± SD flow-based cytolysis, and (F–H) mean ± SD fold T cell proliferation measurements at day 5 post-coculture (three replicates per donor, two donors). (I–N) Mean ± SD fold of Mock EP T cell secretion of IL-2 and IFN-γ measured by ELISA 24 h post-coculture (n = 3 replicates per donor, at least two donors). (O and P) Flow cytometry staining with rhChimera compares staining efficiency and intensity between expansion groups (three different donors). (A–D, O, and P) Ordinary two-way ANOVA, post hoc Tukey analysis. (F, H–L, and N) Ordinary one-way ANOVA, post hoc Tukey. (M) Mixed-effect comparison analysis, post hoc Tukey. Symbols of significance are defined within graphs where necessary. ∗ or #: p < 0.05; ∗∗ or ##: p < 0.01; ∗∗∗ or ###: p < 0.001; ∗∗∗∗ or ####: p < 0.001.

Figure 3

Influence of extent of resection on mRNA-based CAR T cell anti-tumor efficacy in GBM s.c. xenograft mice (A–C) U87KOpool-CBG+ cells were s.c. inoculated into the right flank of NSG mice and randomized into 6 groups (4–6 mice per group) based on BLI measurements before treatment. On day 19, mice received either STR (75% tumor resection) or NTR (90% tumor resection) surgeries and subsequent locoregional injection of Mock EP or mRNA-based multi-targeting CAR T cells, sourced from the same healthy donor (#TMP503). A resection cutoff that maintains minimal residual disease was determined as 4e8 p/s. (C) STR-operated mice had BLI > 4e8 p/s, while NTR-operated ones had BLI ≤ 4e8. (D) Mean ± SD fold change in body weight (two-way ANOVA, not significant) comparing STR- and NTR-operated mice in each group. (E) Kaplan-Meier survival analysis between groups. Log rank test for trend was significant (∗p = 0.0232) between all groups, but no significance was detected in pairwise comparison analyses by Holm-Sidak. (F) Kaplan-Meier survival analysis of pooled STR and NTR mice receiving CAR T cell treatment (log rank [Mantel-Cox] test, p = 0.009, as well as hazard ratios were calculated). (G–M) Longitudinal BLI measurements comparing pooled STR and NTR mice for all CAR T cell-treated mice (G) or STR and NTR mice per group (H–M), mixed effects multiple comparison analysis, post hoc Tukey. ∗: p < 0.05; ∗∗: p < 0.01; ∗∗∗: p < 0.001; ∗∗∗∗: p < 0.0001.

Figure 4

Investigating the efficacy of dual mRNA-based MVCAR T cell injections post-surgery in NSG mice bearing U87KOpool tumors (A–C) Mice bearing U87KOpool-CBG+ cells in the left s.c. flank were subjected, on day 0, to either sham operation or NTR operation and received locoregional DPBS, Mock EP, mRNA-based CAR19-BBz, or MVCAR#2 T cells, sourced from healthy donor #ND625. A second IVIS-assisted local injection was administered on day 5. All T cells were expanded in (IL-7/IL-15)-conditioned medium prior to injections. (A) Model timeline (eight mice per group except for CAR19-BBz, six mice). (B) Selected BLI images of four representative mice per group showing the CAR-dependent therapeutic window of activity of mRNA-based MVCAR T cells in comparison with other groups. (C and D) Flow histograms of CAR staining as well as staining with rhChimera for T cell groups used in both injections. (E) % CR per group on the day of model termination, day 25. Graphs in (F) show mean ± SD of log-transformed total flux data comparing all groups (mixed-effect comparison analysis, post hoc Tukey) as well as total flux of each mouse per group. (G) Side-by-side comparisons of changes in total flux between groups show CAR-dependent and CAR-independent windows of cytotoxicity at selected time points; Kruskal-Wallis ANOVA, post hoc Dunn’s multiple comparison analysis. ∗ or #: p < 0.05; ∗∗: p < 0.01; ∗∗∗: p < 0.001; ∗∗∗∗: p < 0.0001.

Figure 5

In vitro verification of LNP-based mRNA delivery on mRNA-based CAR T cell cytotoxicity The table in (A) provides physical characterization parameters regarding mRNA encapsulation efficiency, diameter, and PDI of single-CAR_LNP versus MVCAR_LNP batches. A representative size (z-average) distribution of MVLNPs is shown in (B), revealing a diameter of approximately 103 nm using dynamic light scattering. (C) T cells from three different donors (ND#1, #2, and #3) were used to investigate the cytotoxicity of LNP-delivered mRNA-based MVCAR#2 (MVCAR#2_LNP) compared with EP-ed MVCAR#2 (MVCAR#2_EP) at three different CAR+ cell-to-target cell ratios (E:T of 8:1, 4:1, and 2:1, respectively) using a flow-based killing assay, 5 days post-coculture with U87KOpool cells (mean ± SD of % cytolysis from n = 3 different donors with three biological replicates each, two-way ANOVA, post hoc Tukey). (D) Flow cytometry analysis of total CAR expression in EP- versus LNP-delivered MVCAR#2 from three different donors. Flow histograms in (E) show the stability of CAR expression in LNP-delivered mRNA-based HER2 CAR or IL13Rα2 CAR from two donors, up to 72 h post-LNP addition. Comparative analysis of rhChimera MFI (F, paired two-tail t test) and % binding (G and H, two-way ANOVA, post hoc Tukey) is provided for MVCAR#2_EP versus MVCAR#2_LNP. ∗: p < 0.05; ∗∗: p < 0.01; ∗∗∗: p < 0.001; ∗∗∗∗: p < 0.0001.

Figure 6

Preclinical evaluation of dual locoregional injections of Mock NTD or LNP-mediated mRNA-based MVCAR or CARPool T cells in NTR-operated mice bearing U87KOpool tumors Mice bearing U87KOpool-CBG+ cells in the left s.c. flank were subjected, on day 0, to NTR operation and received locoregional Mock NTD, CARPool, or MVCAR T cells, all sourced from healthy donor #ND410. A second IVIS-assisted local injection was administered on day 5. All T cells were expanded in IL-7/IL-15-conditioned medium and received Mock NTD or LNP-encapsulated mRNA encoding CAR-BBz. (A) Model timeline. (B) Flow histograms of CAR staining for T cells prepared by LNP delivery of mRNA and used in both injections. (C) Mean ± SD of log-transformed total flux (mixed-effect comparison analysis, post hoc Tukey). (D–H) Individual longitudinal total flux measurements. (I) % CR per group. (J and K) Selected dual comparisons of mean ± SD of BLI measurements on day 40 corresponding to MVCAR#2 versus CARPool#2 (J) or MVCAR#2 versus Mock NTD (K), Mann-Whitney U test. (L) Images of primary and metastatic tumors per mouse, collected during sacrifice. (M) No significant differences were found in weight of tumors collected from primary sites (Kruskal-Wallis ANOVA). (N–Q) Comparison analyses of mean ± SD of human CD4 and CD8 subset distribution and Tim-3 and/or PD-1 expression in T cells collected from the blood (N and O, collected on day 40) and spleen (P and Q, obtained on day of sacrifice) of responders versus non-responders within the pooled CAR T cells-treated mice. (N and P) Mixed-effect multiple comparison analysis, based on uncorrected Fisher’s LSD. (O and Q) Mixed-effect multiple comparison analysis, post hoc Tukey. (R) Images of spleens showing different spleen sizes in responders versus non-responders, as well as a plot of matching spleen weights (unpaired two-tail t test). ∗: p < 0.05; ∗∗: p < 0.01; ∗∗∗∗: p < 0.0001.

Figure 7

Preclinical evaluation of locoregional intracranial injections of Mock EP or EP-mediated mRNA-based MVCAR#2 or CARPool#2 T cells in an intracranial resection model of GBM-bearing patient-derived Ge518 tumors (A) Ge518 cells were phenotypically evaluated for cell surface target expression by staining with antibodies against HER2, IL13Rα2, and EphA2 prior to analysis by flow cytometry. (B–G) A total of 25,000 Ge518 cells was intracranially injected into NSG mice and monitored by IVIS imaging. On day 10, Ge518-bearing mice underwent craniectomy followed by locoregional injection of Mock EP, CARPool#2, or MVCAR#2 T cells that were expanded for 5 days in (IL-7/IL-15)-conditioned medium (9–10 mice per group). (B) Timeline of intracranial inoculation of Ge518 tumors as well as resection and locoregional injection of one dose of CAR T cells. (C) Kaplan-Meier survival analysis, log rank (Mantel-cox) test, followed by Holm-Sidak pairwise analysis. (D–G) BLI measurements plotted as mean ± SD total flux (D) or individually per mouse group (E–G). ∗∗∗: p < 0.001.