CAR Binders Affect CAR T-cell Tonic Signaling, Durability, and Sensitivity to Target

Abstract

Patients can develop human anti-mouse immune responses against CD19-specific chimeric antigen receptor (CAR) T cells due to the use of a murine CD19-specific single-chain variable fragment to redirect T cells. We screened a yeast display library to identify an array of fully human CD19 single-chain variable fragment binders and performed a series of studies to select the most promising fully human CAR. We observed significant differences in the ability of CARs employing these CD19 binders to be expressed on the cell surface, induce tonic signaling, redirect T-cell function, mediate tumor killing, recognize lower levels of CD19 antigen, and maintain function upon continuous antigen exposure. From this initial analysis, CAR T cells using two binders (42 and 52) were selected for additional studies. Although CAR T cells using both binders controlled tumor growth well in vivo, we advanced a CAR construct using binder 42 for more advanced preclinical testing because of its greater similarity to binders based on the antibody FMC63, which is the murine antibody underlying four FDA-approved CD19-specific CAR T-cell therapies, and ability to robustly respond to tumors expressing lower levels of CD19. We found that this binder uniquely bound CD19 using distinct contact residues than FMC63 and with ∼40-fold lower affinity. CARs using binder 42 were non-inferior to those using the FMC63 binder in a mouse model of acute lymphoblastic leukemia, indicating that CAR T cells using binder 42 should be considered for clinical use.

Figure 1:. Identification of fully human CD19 binders that can be used to redirect T cells:

a) Schematic representation of the discovery process for the identification of CD19-specific scFvs. b) Primary human CD4 and CD8 T cells were stimulated using anti-CD3/28 beads and transduced with lentiviral vectors encoding CARs that employed fully human CD19 binders. Cells were stained with FITC-labeled human CD19 protein on day 6 after transduction and evaluated for CAR expression (gated on CD4+ T cells) by flow cytometry. c) Summary bar graph illustrating the CAR MFI on T cells over the designated time course for CAR binders. Data represent the mean ± SEM of three independent experiments with unique donors. P values (*, P < 0.05, **, P < 0.01, ***, P < 0.001, and ****, P < 0.0001) were determined using two-way ANOVA. d) CAR transduced T cells were co-cultured with Nalm-6 cell line for 5hr and stained for intracellular cytokines IL-2 and TNF-α. e-f) Summary bar graph illustrating the IL-2 and TNF-α cytokine response in T cells, respectively. Data represent the mean ± SEM of two independent experiments (n=2) with unique donors. P values (*, P < 0.05, and **, P < 0.01) were determined using one-way ANOVA.

Figure 2:. Binding domain influences the degree of tonic signaling:

a) Jurkat-NFAT-GFP reporter cell line was transduced with each of the CD19 CARs and NFAT activity was measured by GFP 7 days after transduction. b-c) Primary human T cells stimulated using anti-CD3/28 beads and the following day were transduced with lentiviral vectors encoding the indicated CARs specific for CD19 or left not transduced (NTD). The beads were removed after 5 days of culture and T cells were evaluated for expansion and cell size. Data is representative of 3 independent experiments using T cells from different donors. d) HLA-DR expression was measured after 11 days of culture by flow cytometry. Data represent the mean ± SEM of two independent experiments with unique donors. P value (*, P < 0.05) was determined using one-way ANOVA.

Figure 3:. CARTs employing scFv 42 or 52 demonstrate durable control of tumors:

a) CARTs using unique CD19 binders were co-cultured with Nalm-6-IRF720+ target cells at various ratios. A real-time Incucyte-based killing assay was conducted for 48 hours. b) Experimental timeline for continuous antigen exposure (CAE) assay. c) On the indicated days, CountBrightTM Absolute Counting Beads were used to calculate target cells and T cells counts in cell suspensions. After calculations, CD19 binder T cells were seeded with fresh Nalm-6-IRF720+ cells at a ratio 1:4 (CD45+: Nalm-6-IRF720+). This process was repeated every 3–4 days for total 6 rounds of stimulation. d) Experimental timeline for in vivo study. e-f) On day (−7) NSG mice were injected i.v. with Jeko-1-CBG-GFP cell line. On day (−1) mice were distributed into 3 groups of 5 animals based on the bioluminescence values. Mice were treated with 1×10⁶ CARTs or non-transduced T cells on day 0 and tumor growth was monitored by whole body bioluminescence every week. N=5 mice per group were used for this study.

Figure 4:. CARTs using binder 42 can respond to tumors with low CD19 expression:

a) K562 cells expressing low, medium, or high CD19 expression were stained with an anti-CD19 Ab and analyzed by flow cytometry. b) The indicated CARTs were co-cultured with different levels of CD19 Ag expressed K562 cells for 5hr and stained for intracellular cytokines IL-2 and TNF-α. c-d) Summary bar graph depicting the IL-2 and TNF-α cytokine response in CD4+ T cells respectively. Data represent the mean ± SEM of three independent experiments using unique donors. P values (*, P < 0.05, **, P < 0.01, and ***, P < 0.001) were determined using two-way ANOVA.

Figure 5:. CARTs employing binder 42 provide more durable control and are more sensitive than CARTs using FMC63:

a) On the indicated days, CountBrightTM Absolute Counting Beads were used to calculate target cells and T cells counts in cell suspensions. After calculations, CD19 binder T cells were seeded with fresh Nalm-6-GFP+ cells at a ratio 1:4 (CD4+CD8+: Nalm-6-GFP+). This process was repeated every 3–4 days for total 7 rounds of stimulation. b-c) The indicated CARTs were co-cultured with different levels of CD19 Ag expressed K562 cells for 5hr and stained for intracellular cytokines IL-2 and TNF-α. Summary bar graph illustrating the IL-2 and TNF-α cytokine response in CD4+ T cells respectively. Data represent the mean ± SEM of two independent experiments (n=2 donor). P values (*, P < 0.05, and **, P < 0.01) were determined using two-way ANOVA.

Figure 6:. Binder 42 binds a distinct region of CD19 from FMC63 with lower affinity:

a) Visualization of critical residues for Ab 42 binding. Critical residues (red spheres) were visualized on a cryo-EM structure of CD19 (PDB ID: 7JIC). Also shown are secondary residues (blue spheres) that have a lower contribution to binding. b-c) scFv based on FMC63 (b) or binder 42 (c) binding kinetics and affinity sensorgrams depicting the binding kinetics in soluble scFv-Fc format to recombinant human CD19 protein. FMC63 binding to human CD19 was measured at 25, 50, 100, 200, 300 and 400 nM and 42 binding to human CD19 was measured at 50, 100, 200, 400 and 800 nM. d) Plasmids containing cDNA clones of ~6,000 membrane proteins targets were each transfected into HEK-293T cells (18,000 cells/well) in 384-well cell-culture plates. After incubation for 36 hours, the cells were placed into a two-dimensional matrix in a new 384-well plate by rows and columns. 42 scFv-Fc was added to membrane proteome array matrix plate and binding was determined by flow cytometry. Each target is assessed for binding in duplicate. Any potential binding interactions identified on the MPA that did not validate in titration experiments were removed from the graph. Dotted line represents 3 SD of the calculated background. e) In the validation titration experiment, the test article was screened for binding against targets identified on the MPA screen as well as the positive controls (Protein A, known binder CD19), and the negative control (pUC empty Vector) expressed in HEK-293T cells. The 42 scFv-Fc showed strong binding to the positive controls Protein A and CD19. The 42 scFv-Fc was screened at four different concentrations, as mentioned at x-axis.

Figure 7:. CARTs using binder 42 are non-inferior to FMC63-based CARTs in vivo:

a) On day (−7) NSG mice were injected i.v. with Nalm-6-CBG-GFP cell line. On day (−1) mice were distributed into 3 groups of 8 animals based on the bioluminescence values. Mice were treated with 1×10⁶ CARTs or non-transduced T cells on day 0 and tumor growth was monitored by whole body bioluminescence every week. b) Survival curve of mice after CART cell treatment. c-f) Peripheral blood was collected from humanized mice on the indicated days after CART treatment and stained with anti-human CD45, CD3, CD8, CD4 antibodies, and CD19 protein. As indicated, mice were re-challenged with 1M Nalm-6-CBG-GFP 49 days after CART infusion. N=8 mice per group were used for this study.