Avoidance of On-Target Off-Tumor Activation Using a Co-stimulation-Only Chimeric Antigen Receptor

Abstract

Chimeric antigen receptors (CARs) combine T cell activation with antibody-mediated tumor antigen specificity, bypassing the need for T cell receptor (TCR) ligation. A limitation of CAR technology is on-target off-tumor toxicity caused by target antigen expression on normal cells. Using GD2 as a model cancer antigen, we hypothesized that this could be minimized by using T cells expressing Vγ9Vδ2 TCR, which recognizes transformed cells in a major histocompatibility complex (MHC)-unrestricted manner, in combination with a co-stimulatory CAR that would function independently of the TCR. An anti-GD2 CAR containing a solitary endodomain derived from the NKG2D adaptor DAP10 was expressed in Vγ9Vδ2⁺ T cells. Differential ligation of the CAR and/or TCR using antibody-coated beads showed that pro-inflammatory cytokine response depended on activation of both receptors. Moreover, in killing assays, GD2-expressing neuroblastoma cells that engaged the Vγ9Vδ2 TCR were efficiently lysed, whereas cells that expressed GD2 equivalently but did not engage the Vγ9Vδ2 TCR were untouched. Differentiation between X-on tumor and X-off tumor offers potential for safer immunotherapy and broader target selection.

Keywords: chimeric antigen receptor; toxicity; γδT cell.

Figure 1

Chimeric Antigen Receptor Expressed in Expanded γδT Lymphocytes Enhances Antigen-Specific Cytotoxicity (A) Second-generation anti-GD2 chimeric antigen receptor (CAR) containing the CD3ζ and CD28 endodomains allows activation of T cells in the absence of TCR engagement. (B and C) ⁵¹Cr release assay demonstrating cytotoxicity of Vδ2⁺ γδT cells expressing an anti-GD2-28ζ CAR against GD2⁺ and GD2⁻ neuroblastoma cell lines. In (B), antibody-dependent cytotoxicity at an E:T ratio of 10:1 was measured after coating the target cells with ch14.18/CHO anti-GD2 antibody or isotype control antibody (three independent experiments). In (C), statistical differences are shown for 10:1 E:T ratio. Error bars indicate mean values ± SEM; **p < 0.01 and ****p < 0.0001 after Bonferroni post hoc analysis. The bar chart on the right shows 10:1 E:T ratio with analysis by unpaired t test; n = 3 independent donors for GD2-28-ζ and untransduced αβT cells, n = 6 for GD2-28-ζ CAR⁺ γδT cells, and n = 14 for untransduced γδT cells. Cytotoxicity assays were performed at day 6 after T cell transduction (9 days of culture for αβT cells and 11 days of culture for γδT).

Figure 2

GD2-DAP10 CARs Expressed in Vδ2⁺ γδT Cells (A) Vδ2⁺ γδT cells are activated in an MHC-independent manner by indirectly sensing increased phosphoantigen levels on tumors through engagement of the TCR with cognate ligand. Tumors evade γδT cell cytotoxicity by suppressing the co-stimulatory signals required. CARs containing co-stimulatory domains can bypass this evasion while retaining tumor specificity. Healthy tissues do not produce increased phosphoantigen levels so that the TCR is not engaged and the CAR alone is insufficient to provide full activation. (B) The GD2-DAP10 and GD2-28-ζ CARs have an Fc-derived linker and are co-expressed with a marker gene containing a minimal CD34 epitope. Transduced cells are therefore Fc⁺/CD34⁺. The gating strategy for determining transduction efficiency is shown. Transduced cells have much greater median fluorescence intensity (MFI) for CD34-APC than non-transduced cells (4,242 ± 1,215 versus 46.6 ± 15.4; n = 6). (C) Transduction efficiency of Vδ2⁺ cells and αβT cells was consistent, though there is a large degree of donor-donor variation in Vδ2⁺ cell percentages after transduction. Transduction with either CAR had no significant effect on Vδ2⁺ percentage. (D) After 10 days of expansion, we determined the total number of CAR-expressing Vδ2⁺ γδT cells and related this to the volume of blood drawn to generate this many cells (n = 8 independent donors). Error bars denote SEM.

Figure 3

CAR-Mediated Co-stimulation Is Essential to Allow Functional Activation of γδT Cells upon TCR Engagement (A) Schematic representation of the experimental design. Anti-CD3 antibodies coupled to the surface of beads were used to activate TCR (signal 1); anti-CAR (anti-Fc) antibodies activate the CAR (signal 2) through binding the CH2CH3 linker of the CAR. (B) Representative dot plots of γδT cell analysis by flow cytometry after incubation with beads labeled with anti-CD3 and/or anti-CAR antibodies. The binding and crosslinking of a second-generation CAR is sufficient to allow full T cell activation, since both signals 1 and 2 are provided by the CAR, i.e., by the CD3ζ chain and the CD28 endodomain, respectively. According to the co-stimulation-only CAR concept, the engagement of a CAR providing only signal 2, e.g., by a DAP10 endodomain, does not lead to γδT cell activation. Only if signal 1, e.g., due to TCR activation, is provided in parallel do γδT cells become activated and respond by functional changes, e.g., IFN-γ production. (C) The engagement of a second-generation CAR, i.e., GD2-28-ζ, is sufficient to induce a profound production of pro-inflammatory IFN-γ by γδT cells. In contrast, γδT cells expressing a co-stimulation-only CAR (GD2-DAP10) produce IFN-γ only upon both TCR and CAR engagement. (D) No differences in TNF-α production were found for respective anti-CD3 and/or anti-CAR antibody-stimulated γδ T cells expressing the GD2-DAP10 or the GD2-28-ζ CAR. For (C) and (D), we performed Bonferroni post hoc analysis after one-way ANOVA. Graphs display merged results from two independent experiments (n = 4–6 independent donors). Error bars represent mean values ± SD.

Figure 4

Cytokine Secretion by GD2-DAP10 Vδ2⁺ Cells Is Dependent on CD3 and CAR Engagement Cytokine production by Vδ2⁺ γδT cells expressing GD2-DAP10 or GD2-28-ζ CARs following 23 hr stimulation with antibody-coated beads engaging CD3, the CAR, or both. Non-transduced γδT cells (NT) are included for comparison, and CD3/CD28 bead stimulation was included as a positive control. Both the CAR and CD3 must be stimulated in order for GD2-DAP10 Vδ2⁺ γδT cells to produce cytokine, whereas CAR stimulation alone is sufficient to generate cytokine production by GD2-28-ζ Vδ2⁺ cells. Error bars indicate SEM of three independent donors.

Figure 5

Vδ2⁺ γδT Cells Expressing GD2-DAP10 CARs Kill GD2⁺ Tumor Cells (A and B) Specific killing of (A) neuroblastoma line LAN-1 (n = 14 independent donors for untransduced and (B) GD2⁺ Ewing sarcoma cell line TC-71 (n = 3 independent donors) by GD2-DAP10 CAR⁺. (C) This effect is not seen against the GD2⁻ neuroblastoma cell line SK-N-SH (n = 3 independent donors). (D) The GD2-DAP10 CAR does not function in αβT cells where signal 1 is absent, indicating that the signal is restricted to the co-stimulatory pathway (n = 3 independent donors for each subgroup). Frequencies are shown as mean ± SEM; significance was determined by Bonferroni post hoc correction after two-way ANOVA, with only E:T ratio 10:1 annotated for clarity. Cytotoxicity assays were carried out at day 6 after transduction of γδT cells.

Figure 6

Target Cells Not Engaging the Vγ9Vδ2 Are Not Killed by GD2-DAP10 CAR Vδ2⁺ γδT Cells (A) The murine colonic carcinoma cell line CT26 was transduced to express human GD2 and to derive the cell line CT26-GD2. Expression of GD2 was as homogeneous as that detected for LAN-1. (B) ⁵¹Cr release assay demonstrating the ability of αβT cells expressing the GD2-28-ζ CAR to kill CT26-GD2 cells in an antigen-specific manner. Killing of wild-type CT26 and background killing by untransduced αβT cells are minimal (n = 3 independent donors for each subgroup). (C) ⁵¹Cr release assay demonstrating that Vδ2⁺ γδT cells expressing the GD2-28-ζ CAR kill CT26-GD2 cells in an antigen-specific manner, even though CT26 does not engage the Vδ2⁺ γδTCR, whereas Vδ2⁺ γδT cells expressing the GD2-DAP10 CAR do not (n = 7 independent donors for each subgroup). A difference for killing of CT26 target cells was also determined at an E:T ratio of 10:1 using GD2-28-ζ CAR⁺ cells compared to GD2-DAP10 CAR⁺ cells (n = 7 independent donors for each subgroup). Frequencies are shown as mean ± SEM, with significance by Bonferroni post hoc correction after two-way ANOVA; only E:T ratio 10:1 is annotated for clarity.

Figure 7

Expression of GD2-DAP10 CAR in Vδ2⁺ γδT Cells Leads to Less Exhaustion and Activation Marker Expression Than GD2-28-ζ CAR (A) Vδ2⁺ T cells expressing the GD2-DAP10 CAR were transduced and then maintained in medium containing 100 IU/mL IL-2 for 3 weeks. They were then washed thoroughly and co-cultured with irradiated GD2⁺ LAN-1 or GD2⁻ SK-N-SH neuroblastoma cells in the absence of exogenous IL-2. A combination of live cell counting by trypan blue exclusion and fluorescence-activated cell sorting (FACS) analysis for CAR⁺/Vδ2⁺ cells was used to determine fold change over a 3-week co-culture. Only in the presence of GD2⁺ targets did the CAR⁺ cells mount a secondary expansion (n = 3 independent donors, fold change compared by paired t test at day 14 and day 20). (B) Gating to determine phenotype of CAR transduced γδT and αβT cells over prolonged culture in medium containing 100 IU/mL IL-2. Transduction efficiencies as well as the expression of TIM-3 and PD-1 were assessed by flow cytometry at each time point. Note that analysis of TIM-3 and PD-1 expression for conventional αβT cells was based on gating on Vδ2⁻ events. Cells expressing TIM-3 and/or PD-1 were considered to have an exhausted phenotype. (C) After 10 days of culture (5 days after transduction), no significant differences were found by comparing the frequencies of exhausted events in all subgroups. Exhausted cells were defined as those expressing at least one of the exhaustion markers TIM-3 or PD-1. (D) At day 16 of culture (10 days after transduction), αβT or γδT cells expressing GD2-28-ζ CARs were more exhausted than either untransduced αβT or γδT cells. Furthermore, γδT cells expressing the GD2-28-ζ CAR were significantly more exhausted than γδT cells expressing the GD2-DAP10 CAR. As in (C), for the purposes of this analysis, cells were considered to be exhausted if they expressed TIM-3, PD-1, or both. (C) and (D) display combined results from two independent experiments (n = 5 for γδT cells and n = 6 for αβT cells) subjected to Bonferroni post hoc analysis after one-way ANOVA. Columns represent mean values ± SD.