Programmable Attenuation of Antigenic Sensitivity for a Nanobody-Based EGFR Chimeric Antigen Receptor Through Hinge Domain Truncation

Abstract

Epidermal growth factor family receptor (EGFR) is commonly overexpressed in many solid tumors and an attractive target for chimeric antigen receptor (CAR)-T therapy, but as EGFR is also expressed at lower levels in healthy tissues a therapeutic strategy must balance antigenic responsiveness against the risk of on-target off-tumor toxicity. Herein, we identify several camelid single-domain antibodies (also known as nanobodies) that are effective EGFR targeting moieties for CARs (EGFR-sdCARs) with very strong reactivity to EGFR-high and EGFR-low target cells. As a strategy to attenuate their potent antigenic sensitivity, we performed progressive truncation of the human CD8 hinge commonly used as a spacer domain in many CAR constructs. Single amino acid hinge-domain truncation progressively decreased both EGFR-sdCAR-Jurkat cell binding to EGFR-expressing targets and expression of the CD69 activation marker. Attenuated signaling in hinge-truncated EGFR-sdCAR constructs increased selectivity for antigen-dense EGFR-overexpressing cells over an EGFR-low tumor cell line or healthy donor derived EGFR-positive fibroblasts. We also provide evidence that epitope location is critical for determining hinge-domain requirement for CARs, as hinge truncation similarly decreased antigenic sensitivity of a membrane-proximal epitope targeting HER2-CAR but not a membrane-distal EGFRvIII-specific CAR. Hinge-modified EGFR-sdCAR cells showed clear functional attenuation in Jurkat-CAR-T cells and primary human CAR-T cells from multiple donors in vitro and in vivo. Overall, these results indicate that hinge length tuning provides a programmable strategy for throttling antigenic sensitivity in CARs targeting membrane-proximal epitopes, and could be employed for CAR-optimization and improved tumor selectivity.

Keywords: CAR optimization; CAR-T; EGFR; cancer selectivity; cell therapy; cellular immunotherapy; hinge domain.

Figure 1

Identification of EGFR-specific sdAbs with CAR functionality. (A) The structural elements of human EGFR and anti EGFR-sdAb CAR tested in this study are shown. Three EGFR-specific sdAbs were cloned into a modular CAR backbone with 41BB and CD3z signaling domain via golden gate cloning. Jurkat cells were then electroporated with the resulting constructs. Control cells with no plasmid or with a CD19 (FMC63) scFv CAR were also tested here. Jurkat cells (30 000/well) transiently expressing various CAR plasmids as shown were co-cultured with varying doses of (B, C) EGFR-high H292 or SKOV3 cells, (D) EGFR-low MCF7 cells, (E, F) EGFR-negative Raji cells or Ramos Cells and examined for activation via staining with APC-labelled anti-human CD69 antibody. CAR-J results show the mean +/- SEM from a three independent experiments performed in duplicate. Primary human T cells were then transduced with lentivirus encoding the sdAb021-EGFR-sdCAR construct and tested for activity against mKate2-expressing target cells with varying EGFR expression. (G) Automated fluorescent counting was used to examine CAR-T mediated target cell killing and (H) expansion of EGFP-labelled CAR-T cells. Primary CAR-T results show the mean of 3 experiments performed in duplicate. NSG mice were injected with 6x10⁶ H292 human lung cancer cells subcutaneously, then treated with 5x10⁶ sdAb021-BBz CAR-T or untransduced control T cells intravenously (n=5 mice/group). (I) Growth of H292 tumors via regular caliper measurements is shown. (J) Probability of survival throughout the experiment is shown (P values are derived from a Mantel-Cox comparison of treatment groups).

Figure 2

Hinge truncation decreases target response for EGFR-sdAb CAR constructs. (A) Structure of single domain antibody based CARs targeting human EGFR were generated with hinge domains of varying length [full length human CD8-hinge (45CD8h), truncated CD8-hinge (34CD8h or 22CD8h), or no hinge element. (B, C) Jurkat cells were electroporated with varying CAR plasmid constructs before co-incubation with no target cells (1:0 E:T), EGFR-high SKOV3 cells, or EGFR-low MCF7 cells. 30 000 CAR-Jurkat cells were incubated overnight with varying numbers of target cells before CAR/GFP+ cells were examined for activation via staining with APC-labelled anti-human CD69 antibody. (D) Shows the relative CAR-J activation of various constructs in response to SKOV3 or MCF7. Results show the mean +/- SEM from three experiments performed in duplicate.

Figure 3

Single amino acid truncations can throttle CAR antigenic sensitivity for membrane proximal epitope targeting CARs. (A) A range of sdCAR constructs were produced containing EGFR-specific sdAb021 and N-terminally truncated versions of the linker-extended human CD8 hinge domains ranging between 1 and 62 residues in length. Jurkat cells were electroporated with the resulting constructs and then co-incubated at an effector to target ratio of 1:1 with EGFR-high SKOV3 cells or EGFR-low MCF7 cells. After overnight incubation, CAR/GFP+ cells were examined for activation via staining with APC-labelled anti-human CD69 antibody. (B) Similarly, HER2-specific trastuzumab-derivative scFv CAR constructs were generated with hinge domains of varying lengths. CAR-expressing Jurkat cells were co-incubated with HER2-high SKOV3 cells or HER2-low MCF7 cells and then CAR-T activation was assessed. (C) Similarly, EGFRvIII-specific CAR constructs were generated with hinge domains of varying lengths. CAR-expressing Jurkat cells were co-incubated with EGFRvIII-overexpressing U87vIII cells or EGFRvIII-negative U87wt cells and then CAR-T activation was assessed. All results show means +/- SEMs of three separate experiments.

Figure 4

Hinge truncation progressively diminishes CAR-cell binding with target cells. Jurkat cells with stable expression of sdAb021 EGFRsdCAR were generated through lentiviral transduction and cell sorted for similar surface expression. EGFR-sdCAR-Jurkat cells were then mixed at varying doses with mKate2-expressing target cells and co-incubated for 30 minutes at 37°C. (A) Flow cytometry was used to assess the number of CAR-Jurkat cells, target cells, and CAR-Jurkat/target doublets. (B) Examining the size/granularity parameter shows that doublet cells (blue) appear similar in size to the larger target cell population. (C) Examining the height/width parameter indicates that identifies the doublet population as wider than the target cells alone (blue vs red). The proportion of CAR-Jurkat cells engaged in doublet formation with (D) EGFR-high SKOV3 cells or (E) EGFR-negative Ramos cells were quantified using gating as described in the main text. (F) The target cell binding across varying hinge constructs at a fixed effector:target ratio of 1:5 is shown. Graphs show the mean +/- SEM from three experiments performed in duplicate.

Figure 5

Hinge truncation progressively diminishes tumor cell killing and expansion of primary sdCAR-T cells in response to EGFR expressing target cells. Concentrated lentiviral particles encoding hinge-modified EGFR-specific sdAb021 CARs as well as a GFP marker were generated. Peripheral blood T cells were isolated from 3 independent healthy human donors before polyclonal expansion and lentiviral transduction. Varying doses of sdCAR-T cells or mock transduced cells (empty CAR backbone lentivirus) were placed at an E:T ratio of 1:1 in low density co-culture (2000 sdCAR-T cells and 2000 target cells). Co-cultures were examined over 7 days via live fluorescence microscopy (Incucyte) to differentiate red-fluorescent (NLS-mKate2) target cell counts or total area of green-fluorescent (NLS-NanoGreen) CAR-T cells. (A, B) Depicts the response to EGFR-high SKOV3 targets, (C, D) depicts the response to EGFR-low MCF7 targets, (E, F) depicts the response to EGFR-high healthy donor human dermal fibroblast cells. Each graph depicts automated cell counts or fluorescent areas from a single independent experiment. (G) Day 5 mean fold change in various target cell growth at varying E:T ratio for hinge-variant EGFR-sdCAR-T cells derived from 3 donors is shown +/- SEM, P values are derived from a two-way ANOVA comparison of response curves for untransduced T cell co-cultures with CAR-T cells expressing hinge-truncated constructs. (H) Similarly the mean EGFR-sdCAR-T fold expansion at day 5 of hinge-variant CAR co-cultures from 3 donors is shown +/-SEM is shown. P values are derived from a two-way ANOVA comparison on 45CD8h-hinge containing constructs with other constructs tested in parallel.

Figure 6

Hinge truncated CAR-T cells maintain enhanced tumor selectivity in triple co-cultures with both healthy donor cells and SKOV3 tumor cells. Hinge variant EGFR sdCAR cells were placed in co-culture with (A) equal number of mKate2-expressing SKOV3 cells or (B) equal numbers of SKOV3 and unmodified human healthy donor derived dermal fibroblasts (HDF). Graphs depict the average fold change in growth of red fluorescent tumor cells over time from CAR-T cells derived from 3 independent donors. (C) Pictures show the state of triple co-cultures after 6 days of incubation. In a similar experiment, triple co-cultures of hinge variant EGFR-sdCAR cells, HDF, and SKOV3 cells were incubated overnight before examining (D) HDF or SKOV3 viability and (E) T-cell activation marker CD69 expression via flow cytometry. Graphs depict the mean results for CAR-T or control T cells from 3 independent blood donors. P values indicate a student T-test comparison for variance between differing hinge constructs, an ANOVA comparison of survival curves is also shown where marked.

Figure 7

Hinge truncated EGFR sdCAR-T cells show progressively diminished response to human lung cancer xenografts in vivo. NOD/SCID/IL2γ-chain-null (NSG) mice were injected subcutaneously with 2×10⁶ SKOV3 cells stably expressing mKate2. Mice (N=5 mice/group) were injected with 5×10⁶ total T cells (~1x10⁶ CAR-T cells) intravenously. (A) Tumor volume was assessed using caliper measurements and (B) time to defined humane endpoint (tumor volume 2000 mm³) was assessed. P values are derived from Mantel-Cox test of survival curves in comparison with untreated mice. Δ Note that the experiment was ended early due to non-experimental related animal facility shutdown, and thus the final tumor measurement of day 112 is shown in (C). (D) Fluorescence imaging was performed at varying timepoints after tumor cell injection. (E) Blood was also collected at selected timepoints after tumor challenge to quantify the proportion of human CD45+ cells that were GFP/CAR+. (F) Staining for human CCR7 and CD45RA was used to assess the differentiation status of CAR/GFP+ cells or total T cells for mice treated with mock-transduced CAR-T cells. See inset for gating strategy used to define effector, naïve/SCM, central memory, or effector memory-RA+ (EMRA) cells.

Figure 8

Hinge truncation results in similar progressively diminished EGFR-specific sdCAR response to human glioblastoma xenografts in vivo. (A) NSG mice were injected (N= 5 mice/group) subcutaneously with 1×10⁶ U87vIII cells stably expressing mKate2. At 7 and 14 days post-tumor challenge mice were injected intratumorally with 1×10⁷ total T cells (approx. 2.5×10⁶ hinge-modified sdCAR-T cells) or with untransduced control T cells (mock). Tumor growth was monitored via caliper measurements. N=5 mice per group. (B) Mice were sacrificed at pre-determined endpoints based on animal condition or tumor volume >2000 mm³. P values are derived from Mantel-Cox test of survival curves in comparison with untreated mice. (C) In vivo imaging was performed to examine the mKate2 fluorescent signal associated with tumor cells. (D) Blood was drawn from challenged mice and examined for the proportion of sdCAR-transduced cells (GFP+) within the hCD45+ lymphocyte fraction at the final timepoint where all experimental mice were alive (day 21). (E) sdCAR-T cells, or total hCD45+ cells for mock T cell treated mice, were examined for differentiation status using antibody staining for human CCR7 and CD45RA. See inset for gating strategy used to delineate effector, naïve/SCM, central memory, or effector memory-RA+ (EMRA) cells.