Granzyme B-activated IL18 potentiates αβ and γδ CAR T cell immunotherapy in a tumor-dependent manner

Abstract

Interleukin (IL)18 is a potent pro-inflammatory cytokine that is activated upon caspase 1 cleavage of the latent precursor, pro-IL18. Therapeutic T cell armoring with IL18 promotes autocrine stimulation and positive modulation of the tumor microenvironment (TME). However, existing strategies are imperfect since they involve constitutive/poorly regulated activity or fail to modify the TME. Here, we have substituted the caspase 1 cleavage site within pro-IL18 with that preferred by granzyme B, yielding GzB-IL18. We demonstrate that GzB-IL18 is constitutively released but remains functionally latent unless chimeric antigen receptor (CAR) T cells are activated, owing to concomitant granzyme B release. Armoring with GzB-IL18 enhances cytolytic activity, proliferation, interferon (IFN)-γ release, and anti-tumor efficacy by a similar magnitude to constitutively active IL18. We also demonstrate that GzB-IL18 provides a highly effective armoring strategy for γδ CAR T cells, leading to enhanced metabolic fitness and significant potentiation of therapeutic activity. Finally, we show that constitutively active IL18 can unmask CAR T cell-mediated cytokine release syndrome in immunocompetent mice. By contrast, GzB-IL18 promotes anti-tumor activity and myeloid cell re-programming without inducing such toxicity. Using this stringent system, we have tightly coupled the biological activity of IL18 to the activation state of the host CAR T cell, favoring safer clinical implementation of this technology.

Keywords: CAR T-cell; Chimeric antigen receptor; IL18; cancer; granzyme B; immunotherapy; αβ T cell; γδ T cell.

Graphical abstract

Figure 1

Co-expression of IL18 variants with pCAR-H/T (A) The caspase 1 cleavage site in human pro-IL18 (P1–4; shown in bold) is aligned above a mutated variant, GzB-IL18, in which this sequence has been replaced with an optimized human granzyme B (GzB)-cleavage site. (B) Cartoon structure of the MUC1-specific parallel CAR, pCAR-H/T. The HMFG2 single-chain variable fragment (scFv) binds to underglycosylated (tumor-associated) MUC1. The T1E peptide binds to eight of nine ErbB homo- and heterodimers. (C) SFG retroviral vectors that encode for pCAR-H/T alone (designated as + nil) or pCAR-H/T + pro-IL18, + GzB-IL18, + constitutively active (const.) IL18 or + GzB-IL18/GzB. Constitutively active IL18 was generated by placing the mature IL18 sequence downstream of a CD4 signal peptide. In the + GzB-IL18/GzB construct, additional GzB is co-expressed with GzB-IL18 and pCAR-H/T. GzB and caspase 1 cleavage sites are indicated. LTR, long terminal repeat. (D) Representative examples of pCAR expression by T cells transduced with the vectors shown in (C). (E) Transduction efficiency of replicate donors as determined by flow cytometric analysis of surface CCR expression (mean ± SEM; n = 8). All not significant (N/S) by one-way ANOVA. (F) Fold expansion of each CAR T cell population over 10 days (mean ± SEM; n = 4). All N/S by one-way ANOVA. (G) IL18 was measured by ELISA in supernatants harvested from the indicated transduced T cell populations after expansion for 10 days (mean ± SEM, n = 4 donors). ∗p < 0.05; ∗∗p < 0.01 by one-way ANOVA. (H) IFN-γ was measured by ELISA in supernatants harvested from the indicated transduced T cell populations following expansion for 10 days (mean ± SEM, n = 2 donors measured in duplicate). ∗p < 0.05; ∗∗∗p < 0.001; N/S, not significant by one-way ANOVA. (I) T cells were transduced with the indicated retroviral vectors, or untransduced (untrans.) as control. T cells were plated at a density of 5 × 10⁵/mL and cultured alone, co-cultured with MDA-MB-468 cells (at a ratio of 10 to 1), or co-cultured with anti-CD3/CD28 TransAct beads. Supernatants were collected after 24 h and analyzed for IL18 by ELISA (mean ± SEM, n = 3 donors measured in triplicate). ∗∗∗∗p < 0.0001 by two-way ANOVA. (J) Supernatants described in (B) were added to HEK-Blue IL18 reporter cells to assess IL18 biological activity, measured as optical density (OD) at 450 nm (mean ± SEM, n = 3–8 donors measured in triplicate). ∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05 by two-way ANOVA.

Figure 2

GzB-IL18 promotes CAR T cell anti-tumor activity in vitro (A) Cytotoxicity assays were conducted with MDA-MB-468 TNBC (left) or BxPC3 pancreatic cancer cells (right), which were incubated for 72 h with the indicated CAR T cell populations at the specified effector:target (E:T) ratio (mean ± SEM, n = 5). After removal of residual T cells, tumor cell viability was measured using an MTT assay. ∗∗∗∗p < 0.0001, ∗p < 0.05 by two-way ANOVA. (B) IFN-γ concentration was measured in supernatants harvested after 72 h from co-cultures described in (A) (mean ± SEM, n = 5). ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 by two-way ANOVA. (C) CAR T cells were added to MDA-MB-468 (left) or BxPC3 tumor cells (right) at a 1:1 E:T ratio (1 × 10⁴ tumor cells). Tumor viability was determined after 72 h and T cells were transferred to a fresh well containing 1 × 10⁴ tumor cells. T cells were re-stimulated in this manner until they could no longer be retrieved from tumor monolayers. A stimulation cycle was deemed successful if ≥60% of tumor cells were destroyed. The number of successful re-stimulation cycles for each T cell/tumor cell condition is shown (mean ± SEM; n = 4–8 donors measured in triplicate). ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05 by two-way ANOVA. (D) T cell number was determined prior to each re-stimulation cycle, undertaken as described in (C) (mean ± SEM, n = 3–7 donors). ∗∗∗∗p < 0.0001, ∗∗p < 0.01, ∗p < 0.05 by two-way ANOVA.

Figure 3

GzB-IL18 promotes CAR T cell anti-tumor activity in vivo (A) 1 × 10⁶ ffLuc-expressing MDA-MB-468 tumor cells were injected i.p. into female SCID Beige mice. After confirmation of tumor engraftment using BLI, mice were randomly assorted into groups with similar mean tumor burden. Animals received a single dose of 10 × 10⁶ of the indicated CAR T cells i.p. or PBS as control on day 11 (overhead arrow). Plots indicate serial bioluminescence emission from each mouse. ∗∗∗∗p < 0.0001 by two-way ANOVA. Number of tumor-free mice at the end of the experiment is indicated. (B) Survival curve of mice treated as described in (A). Statistical analysis was by log rank (Mantel-Cox) test: ∗∗p < 0.01; ∗p < 0.05; N/S, not significant. (C) Weight of mice treated as described in (A) (mean ± SEM, n = 5–7).

Figure 4

GzB-IL18 promotes γδ CAR T cell function in vitro (A) Flow cytometric analysis of γδ T cells engineered to express the indicated transgenes. γδ TCR expression was determined using a pan-γδ TCR antibody (mean ± SEM, n = 4–5 donors). All N/S by two-way ANOVA. (B) Fold expansion of the indicated γδ T cells over 21 days of culture is shown (mean ± SEM, n = 4 donors). Untrans., untransduced. All N/S by two-way ANOVA. Percentage γδ T cell purity (C) and percentage γδ T cells that co-express pCAR-H/T (D) are shown for replicate cultures on day 21 of expansion (mean ± SEM, n = 8–10 donors). All N/S by one-way ANOVA. (E) γδ T cells were transduced with the indicated retroviral vectors or untransduced as control. γδ T cells were plated at a density of 5 × 10⁵/mL and cultured alone, co-cultured with MDA-MB-468 cells (at a ratio of 10 to 1), or co-cultured with anti-CD3/CD28 TransAct beads. Supernatants were collected after 24 h and analyzed for IL18 by ELISA (mean ± SEM, n = 3 donors measured in triplicate). ∗∗∗∗p < 0.0001 by two-way ANOVA. (F) Supernatants described in (E) were added to HEK-Blue IL18 reporter cells to assess IL18 biological activity, measured as OD at 450 nm (mean ± SEM, n = 2–6 donors measured in triplicate). ∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05 by two-way ANOVA.

Figure 5

GzB-IL18 promotes γδ CAR T cell function in vivo (A) 1 × 10⁶ ffLuc-expressing MDA-MB-468 tumor cells were injected i.p. into SCID Beige mice. After confirmation of tumor engraftment using BLI, mice were randomly assorted into groups with similar mean tumor burden. Animals received a single dose of 10 × 10⁶ of the indicated CAR γδ T cells i.p. or PBS as control on day 11 (overhead arrow). Plots indicate serial bioluminescence emission from each mouse. Number of tumor-free mice at the end of the experiment is indicated. ∗∗∗∗p < 0.0001 by two-way ANOVA. (B) Survival curve of mice treated as described in (A). ∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01 by log rank (Mantel-Cox) test. (C) Weight of mice treated as described in (A) (mean ± SEM, n = 5–11).

Figure 6

Effects of IL18 armoring on CAR T cell metabolism γδ T cells (left) or conventional (mainly αβ) T cells (right) from n = 5 donors were transduced with the indicated constructs and expanded in culture for 3 weeks. GLUT1 (A), CD98 (B), and mitochondrial mass (C) were quantified in CAR T cells by flow cytometry post expansion and post two stimulation cycles with MDA-MB-468 tumor cells. Mean fluorescence intensity of each marker was normalized (norm.) to that of pCAR-H/T + nil T cells, which was set to 1.0 in each experiment. ∗p < 0.05, ∗∗p < 0.01 using one-sample t test, making comparison with pCAR-H/T + nil. (D) Post tumor cell co-culture, CAR T cells were incubated with puromycin for 45 min to determine protein synthesis rate according to the SCENITH method. CAR T cell metabolic pathways were inhibited as described in section “materials and methods” concurrently with puromycin incubation to determine glucose or oxygen dependence. All N/S using one-sample t test.

Figure 7

GzB-IL18 safely potentiates anti-tumor activity of panErbB-specific CAR T cell, whereas const. IL18 induces lethal toxicity in this model (A) The caspase 1 cleavage site in mouse pro-IL18 (P1–4; shown in bold) is aligned above a mutated variant, GzB-IL18, in which this sequence has been replaced with an optimized mouse GzB-cleavage site. (B) Murine T cells were transduced using an SFG retroviral vector to express the m2G-T panErbB CAR alone or together with the indicated murine IL18 variants. (C) T cell transduction was assessed by staining for murine TIE peptide within the CAR ectodomain at day 5 post transduction. (D) Mouse IL18 was measured by ELISA in supernatant collected from the indicated mouse T cell cultures and plated at 1 million cells/mL for 24 h (mean ± SEM, n = 3). ∗∗∗∗p < 0.0001 by one-way ANOVA. (E) Survival curve of tumor-free BALB/c mice treated with 4 × 10⁶ mouse CAR T cells (n = 3 per group). (F) Weight of mice treated as described in (E) (mean ± SEM, n = 3). (G) 1 × 10⁶ B7E3 tumor cells were injected s.c. into female BALB/c mice. After confirmation of tumor engraftment using caliper measurements, mice were randomly assorted into groups with similar mean tumor burden and preconditioned on day 13 with cyclophosphamide (cycloph., first arrow). On day 14 (second arrow), mice received a single dose of 2 × 10⁶ of the indicated mouse CAR T cells i.v., or PBS as control. Plots indicate serial caliper measurements of tumor for each mouse. ∗∗∗∗p < 0.0001, ∗∗p < 0.01 by two-way ANOVA comparing post-treatment tumor burden. (H) Percentage weight change in mice described in (G) (mean ± SEM, n = 8–13).

Figure 8

GzB-IL18 polarizes macrophages toward an M1 phenotype, while const. IL18 induces CRS B7E3 tumor cells were engrafted s.c. in Balb/c mice. On day 13, mice were conditioned with cyclophosphamide 50 mg/kg i.p. followed by i.v. infusion of 4 × 10⁶ of the indicated CAR T cell populations on day 14. Data show toxicity scores at 24 h post CAR T cell treatment (mean ± SEM; A) and body weight pre- and 24 h post CAR T cells (mean ± SEM; B). (C) Since toxicity in m2G-T + (m)const. IL18-treated mice exceeded humane endpoints, all mice were killed and terminal bleeds performed at 24 h post T cell infusion. Indicated cytokines were measured in derived sera (mean ± SEM, n = 11–18 mice). Spleens were also harvested from the mice and subjected to immunophenotyping. (D) M1 macrophages were identified as Lin⁻, CD11b⁺, F4/80⁺, Gr-1⁻, and MHC-II⁺, while (E) dendritic cells (DCs) were identified as CD11b⁺, CD11c⁺, and MHC-II ⁺ (mean ± SEM, n = 2–7). ∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05 by one-way ANOVA.