Microenvironment actuated CAR T cells improve solid tumor efficacy without toxicity

Abstract

A major limiting factor in the success of chimeric antigen receptor (CAR) T cell therapy for the treatment of solid tumors is targeting tumor antigens also found on normal tissues. CAR T cells against GD2 induced rapid, fatal neurotoxicity because of CAR recognition of GD2⁺ normal mouse brain tissue. To improve the selectivity of the CAR T cell, we engineered a synthetic Notch receptor that selectively expresses the CAR upon binding to P-selectin, a cell adhesion protein overexpressed in tumor neovasculature. These tumor microenvironment actuated T (MEAT) cells ameliorated T cell infiltration in the brain, preventing fatal neurotoxicity while maintaining antitumor efficacy. We found that conditional CAR expression improved the persistence of tumor-infiltrating lymphocytes because of enhanced metabolic fitness of MEAT cells and the infusion of a less differentiated product. This approach increases the repertoire of targetable solid tumor antigens by restricting CAR expression and subsequent killing to cancer cells only and provides a proof-of-concept model for other targets.

Fig. 1.. GD2 CAR T cells cause fatal neurotoxicity in immunodeficient mice.

(A) Design of lentiviral αGD2CAR construct and αGD2CAR AkaLuc reporter. (B) Representative image of CD45 histological staining of NSG mouse brains treated with primary human αGD2CAR T cells. Left-hand panels show zoom of white box region of interest. Comparative analysis of CD4- and CD8-positive brain-infiltrating populations of primary human GD2CAR T cells. (C) Quantification of brain-infiltrating cells (n = 3 mice). (D) Time course of GD2 CAR T cell expansion via BLI of AkaLuc expressing primary T cells (n = 4 mice) or control-injected mice (Cntrl, mock T cell not expressing AkaLuc). All mice are in indicated order as in the top-left panel. (E) Ex vivo imaging of mouse brains on day 14 end point by BLI. (F) Quantification of brain-specific CAR T cell expansion. Quantification of mouse weight (G), grip strength (H), and overall survival (I) for a dose titration of GD2 CAR T cells (n = 3 mice per group). mil, million. Data are means with individual data points (C) or means ± SD [(F) to (I)].

Fig. 2.. P-selectin is an actuator target in the TME.

(A) Representative IF stain of human neuroblastoma patient tissue for P-selectin (green), CD31 (red), and their colocalization (orange). Scale bars, 200 μm. (B) Quantification of P-selectin/CD31 expression in human neuroblastoma and normal peripheral nerve tissue via tissue microarray (n = 27 tumor, n = 9 normal). (C) MEAT cell system. MEAT cells contain an anti-human P-selectin synNotch, which induces expression of either CD19 or GD2 CAR upon engagement to P-selectin. (D) αhSELP synNotch constructs actuating either the GD2 CAR or CD19 CAR. (E) Transduction efficiency of αhSELP synNotch → GD2 CAR MEAT primary T cells. (F) Jurkat MEAT cell actuation by plate-bound recombinant human P-selectin. FSC-A, forward scatter area. Data represented by the violin plot are individual data points, and statistics were calculated using unpaired t test (B). **P ≤ 0.01.

Fig. 3.. MEAT cells are specific and cytotoxic.

(A) Actuation of primary αhSELP synNotch → CD19 CAR MEAT cells by soluble and plate-bound recombinant human P-selectin (n = 3 technical replicates). (B) On kinetics of CD19 CAR in αhSELP synNotch → CD19 CAR MEAT primary T cells incubated with recombinant human P-selectin (n = 3 technical replicates). h, hours. (C) Time course of antigen-specific activation in αhSELP synNotch → GD2 CAR MEAT cells cocultured with GD2+ target cells as measured by CD69 expression. (D) Live-cell killing coculture schematic of effector (E), actuator (A), and target (T) cells. Live-cell killing of (E) αhSELP synNotch → CD19 CAR MEAT cells and (F) CD19 CAR T cells at varying effector cell concentrations. (G) Fluorescence imaging of actuator (green) and target (red) cells cocultured with either synNotch CD19 CAR, CD19 CAR, or mock T cells (unlabeled) for 72 hours (representative images of three technical triplicates). Live-cell killing of (H) αhSELP synNotch → GD2 CAR MEAT cells and (I) GD2 CAR T cells at varying actuator cell concentrations. The gray dotted line represents 50% killing of target cells. (J) Top five cytokines secreted after 72-hour coculture of effector cells with target cells expressing both the actuator and antigen (P-selectin⁺ and CD19⁺) or only one present (P-selectin⁺ or CD19⁺) as assayed by the Luminex 25-plex human cytokine panel. Data are shown as means ± SD, and statistics were calculated using one-way ANOVA (A) or two-way ANOVA with Tukey’s post hoc test (J) where ns > 0.05, ****P ≤ 0.001.

Fig. 4.. MEAT cells are less differentiated and more metabolically fit.

(A to C) Resting (unstimulated) CD19 CAR, synNotch, or mock T cells assayed for surface expression of (A) Tim3, Lag3, and PD1 and (B) memory phenotype by flow cytometry. (C) SK-N-BE(2) target cell killing of unstimulated effector cells (E:T of 1:1). (D to I) CAR T cells were chronically activated for 15 days by coculture with 20% actuator (P-selectin⁺) and 80% target (CD19⁺) cells and with fresh actuator and target cells added on days 4, 8, and 12. Chronically activated CAR T cell (D) Tim3, Lag3, and PD1 surface expression and (E) memory phenotype were assessed by flow cytometry. (F) SK-N-BE(2) target cell killing of chronically activated effector cells (E:T of 1:1). [(G) to (I)] Metabolic analysis of effector cells following 15-day stimulation as measured by Seahorse mitochondrial stress test. (G) OCR, (H) SRC, and (I) ECAR. Flow data are shown as means ± SEM (n = 3 biological replicates). Cytotoxicity and Seahorse data are shown as means ± SEM (n = 3 technical replicates, n = 2 biological replicates). T_N (naïve), T_SCM (stem cell memory), T_CM (central memory), T_TM (transitional memory), and T_EM (effector memory). Data are shown as means ± SD, and statistics were calculated using one-way ANOVA with Dunnett’s post hoc test [(A) and (B)], unpaired t test [(B) and (E)], or one-way ANOVA with Tukey’s post hoc test to correct for multiple comparisons (H) where ns > 0.05, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001.

Fig. 5.. MEAT cells are safe and well tolerated in mice.

(A) Weight change in mice treated with 3 × 10⁶ constitutive GD2 CAR T cells, αhSELP synNotch → GD2 CAR MEAT cells, or mock T cells (control). (B) Quantification of CD8+ cells in murine brain tissue on days 1, 5, 9, and 12 after T cell injection. (C) Representative IF staining of CD8⁺ cells on day 12 after T cell injection from (B) (scale bars, 100 μm). (D) Spleen weight on days 1, 5, 9 and 12 after T cell injection. Ex vivo flow of CAR T cells at the end point (day 12) in the spleen (E and F) and brain (G). MFI, mean fluorescence intensity. Data are shown as means ± SD (n = 3 mice per group), and statistics were calculated using one-way ANOVA with Dunnett’s post hoc test (D) or unpaired t test [(E) to (G)] where *P ≤ 0.05 and ****P ≤ 0.001.

Fig. 6.. P-selectin–gated GD2 CAR T cells ameliorate toxicity.

(A) Experimental scheme to assess the efficacy of MEAT cells in a solid tumor model of neuroblastoma. Mice were treated with 3 × 10⁶ constitutive GD2 CAR T cells, αhSELP synNotch → GD2 CAR MEAT cells, or control mock T cells 14 days after tumor inoculation. iv, intravenously; sc, subcutaneously. Tumor growth kinetics from SK-N-BE(2) tumor–bearing mice with (B) or without (C) P-selectin. (D) Top: Representative IF staining of CD8+ and CD4+ cells in murine brain at the survival end point (>20% weight loss or >2000-mm³ tumor size). Scale bars, 50 μm. Bottom: Quantification of T cell infiltration. Weight change in SK-N-BE(2) tumor–bearing mice with (E) or without (F) P-selectin. Kaplan-Meier survival curves from SK-N-BE(2) tumor–bearing mice with (G) or without (H) P-selectin. Data are shown as means ± SEM, and statistics were calculated using unpaired t test [(B) and (C)], one-way ANOVA with Tukey’s post hoc test (D), or Kaplan-Meier survival analysis [(G) to (H)] where ns > 0.05, *P ≤ 0.05 and ****P ≤ 0.001. n = 5 mice per group.

Fig. 7.. P-selectin–gated CD19 CAR T cells improve T cell persistence and antitumor efficacy.

Mice were treated with 3 × 10⁶ constitutive CD19 CAR T cells, αhSELP synNotch → CD19 CAR MEAT cells, or control mock T cells 14 days after tumor inoculation. Tumor growth kinetics from SK-N-BE(2) tumor–bearing mice with (A) or without (B) P-selectin. Kaplan-Meier survival curves from SK-N-BE(2) tumor–bearing mice with (C) or without (D) P-selectin. Ex vivo flow cytometric analysis of (E) the spleen and lung and (F) tumor-infiltrating T cells. (G) Mixed antigen model to assess biodistribution of MEAT cells in a trans system. (H) Ex vivo flow cytometric analysis of tumor-infiltrating T cells 17 days after treatment in P-selectin⁺ tumors or P-selectin⁻ tumors. Data are shown as means ± SEM, and statistics were calculated using unpaired t test [(A), (B), (E), and (F)], Kaplan-Meier survival analysis [(C) and (D)], or one-way ANOVA with Tukey’s post hoc test (H) where ns > 0.05, *P ≤ 0.05, **P ≤ 0.01, and ****P ≤ 0.001.