Memory-like Differentiation, Tumor-Targeting mAbs, and Chimeric Antigen Receptors Enhance Natural Killer Cell Responses to Head and Neck Cancer

Abstract

Purpose: Head and neck squamous cell carcinoma (HNSCC) is an aggressive tumor with low response rates to frontline PD-1 blockade. Natural killer (NK) cells are a promising cellular therapy for T cell therapy-refractory cancers, but are frequently dysfunctional in patients with HNSCC. Strategies are needed to enhance NK cell responses against HNSCC. We hypothesized that memory-like (ML) NK cell differentiation, tumor targeting with cetuximab, and engineering with an anti-EphA2 (Erythropoietin-producing hepatocellular receptor A2) chimeric antigen receptor (CAR) enhance NK cell responses against HNSCC.

Experimental design: We generated ML NK and conventional (c)NK cells from healthy donors, then evaluated their ability to produce IFNγ, TNF, degranulate, and kill HNSCC cell lines and primary HNSCC cells, alone or in combination with cetuximab, in vitro and in vivo using xenograft models. ML and cNK cells were engineered to express anti-EphA2 CAR-CD8A-41BB-CD3z, and functional responses were assessed in vitro against HNSCC cell lines and primary HNSCC tumor cells.

Results: Human ML NK cells displayed enhanced IFNγ and TNF production and both short- and long-term killing of HNSCC cell lines and primary targets, compared with cNK cells. These enhanced responses were further improved by cetuximab. Compared with controls, ML NK cells expressing anti-EphA2 CAR had increased IFNγ and cytotoxicity in response to EphA2+ cell lines and primary HNSCC targets.

Conclusions: These preclinical findings demonstrate that ML differentiation alone or coupled with either cetuximab-directed targeting or EphA2 CAR engineering were effective against HNSCCs and provide the rationale for investigating these combination approaches in early phase clinical trials for patients with HNSCC.

Figure 1.. ML NK cells from normal donors exhibit improved ability to control HPV+ and HPV- HNSCC compared to cNK cells.

(A) PB-derived NK cells from HD were activated with IL-12/IL-15/IL-18 or in low dose (LD) IL-15 as control for 16–18 hours (1). Activated NK cells were differentiated into ML NK cells in vitro for 7 days in the presence of low dose IL-15 (2) and restimulated (3). (B) Representative flow cytometry and gating strategy to evaluate cytokine secretion and degranulation (CD107a+) of cNK and ML NK cells stimulated with HNSCC cell line UM-SCC1. Numbers indicate the percentage of positive cells. Summary data of (C) IFN-γ, (D) degranulation (CD107a), and (E) TNF of control (blue) and ML (green) NK cells re-stimulated 6 hours in vitro with UM-SCC1, UM-SCC9, and UM-SCC47 cells at a 5:1 effector to target ratio, n=12; 4 independent experiments. (F) 4-hour ⁵¹Cr release assay and (G) 120-hour IncuCyte® assay at 2.5:1 E: T ratio of ML NK cells: UM-SCC1. (H) NSG mice were injected intraperitoneal (i.p) with 2.5 × 10⁵ UM-SCC1 expressing luciferase 3 days before NK-cell injection. On day 3, tumor-bearing mice were injected with 5 × 10⁶ control or ML NK cells (i.p) and the tumor burden was assessed by bioluminescent imaging (BLI) weekly. (I) BLI summary data of tumor burden and (J) survival. n=10–11 mice per group from two independent experiments, black arrow indicates NK cell injection i.p.. Bars represent mean SEM. Statistical significance was determined by Two-way ANOVA test or paired t-test. For (F), n=3; two independent experiments. (G) shows a representative experiment out of three independent experiments.

Figure 2.. Enhanced ML NK cell responses are partially dependent on DNAM-1, NKG2D and CD2.

(A) Expression of DNAM-1 ligands (CD155, CD112), NKG2D ligand (MICA/B) and CD2 ligand (CD58) on primary HNSCC tumors. (B-E) cNK and ML NK cells derived from healthy donors (HD) were stimulated for 6 hours with head and neck cell lines or tumor cells with or without αNKG2D (5 μg/ml), αCD2 (5 μg/ml) and αDNAM-1 (5 μg/ml) blocking antibodies or isotype controls. B, Representative flow cytometry dot plots and summary data showing (C) IFN-γ, (D) TNF, and (E) degranulation (CD107a+) by control and ML NK cells after DNAM-1, NKG2D and CD2 blockade. (F) Growth of primary HNSCC tumor cells co-cultured with HD control and ML NK cells at 2.5:1 NK cell: Target cell ratio with or without blocking antibodies measured using IncuCyte®. Error bars represent mean ± SEM. Statistical significance was determined by Two- way ANOVA test. For C-E, n=6; three independent experiments. For F, n=3; two independent experiments.

Figure 3.. Cetuximab enhances the ability of ML NK and cNK cells to control HPV+ and HPV- HNSCC cell lines in vitro.

HD cNK and ML NK cells were stimulated with HNSCC cell lines pre-incubated with IgG1 isotype control or cetuximab antibody. (A) Summary data of IFN-γ, degranulation (CD107a+) and TNF, of cNK (blue) and ML (green) NK cells re-stimulated 6 hours in vitro with UM-SCC1, UM-SCC9, and UM-SCC47 cells with either IgG1 isotype control (-, open) or cetuximab (+, filled). Cell lines were used at a 5:1 E:T ratio. N = 7–12, 4 independent experiments. (B) Short-term killing in 4 hour ⁵¹Cr release assay of both cNK and ML NK cells against the UM-SCC47 cell line. N =3, two independent experiments (C) IncuCyte® killing assay of cNK and ML NK cells with IgG1 isotype control or Cetuximab against UM-SCC1; representative experiment out of 3 independent experiments. (D) NK cells derived from HD were pre-activated as indicated in Fig. 1A. Head and neck squamous cell carcinoma cell lines generated from primary tumor tissue were used as target cells for cytokine production, degranulation, and long-term killing assessment. (E) Summary data of IFN-γ, degranulation (CD107a+) and TNF, of control (blue) and ML (green) NK cells re-stimulated 6 hours in vitro with primary HSNCC cells with either IgG1 isotype control (-, open) or cetuximab (+, filled) at a 5:1 E:T ratio. (F-G) IncuCyte® assay at (F)1:1 E:T and (G) 0.5:1 E:T ratio. F and G show a representative experiment out of 3 independent experiments. Error bars represent mean ± SEM. Statistical significance was determined by Two-way ANOVA test.

Figure 4.. Cetuximab and human ML NK cells effectively control head and neck targets in a xenograft model in NSG mice.

(A) NSG mice were injected with 2.5 × 10⁵ UM-SCC1 HNSCC cells expressing luciferase i.p.. On day 3, tumor-bearing mice were injected i.p. with cetuximab or rituximab (Isotype control) and 1 × 10⁶ cNK or ML NK cells i.p. the tumor burden was assessed using BLI weekly. (B) Representative BLI images. (C) BLI Summary data of tumor burden. n=7–10 mice per group, 2 independent experiments (rituximab only = 10 mice, cetuximab only = 8 mice, cNK cells + cetuximab = 7 mice, and ML NK cells + cetuximab = 8 mice). Bars represent mean ± SEM. Statistical significance calculated by Two-way ANOVA -mixed-effect model with Tukey post-test.

Figure 5.. EphA2-CAR ML NK cells display enhanced functional responses and antigen (ephA2)-specific response against HNSCC tumor cell lines.

(A) Schematic representation of EphA2 CAR construct. P2A indicates the ribosomal P2A skip site. Transmembrane (TM)/hinge: CD8a; costimulatory domain (D1): CD137; and stimulatory domain (D2): CD3z. ITAMs indicated in light blue. (B) Schema of in vitro experiments. Purified NK cells were activated with IL-12, IL-15, and IL-18 or were control treated for 16 hours, washed, and transduced with CAR lentivirus for 2 days. After differentiating for 1 week, CAR ML NK cell functionality was assessed. (C) Representative flow plots of ephA2 CAR (GFP+) cNK or ML NK cells stimulated with UM-SCC9 targets (total NK:Tumor, 5:1) depicting IFN-γ (top) and degranulation (CD107a; bottom). Summary IFN-γ and degranulation (CD107a+) from stimulation with (D) HPV- cell line UM-SCC9 and (E) HPV+ cell line UM-SCC47, cNK cells (blue) EphA2-CAR cNK cells (black), ML NK cells (green), EphA2-CAR-ML NK cells (brown). n=9; 5 independent experiments. (F) IncuCyte® assay at 1:1 E: T of ML NK cells co-cultured with EphA2 expressing UM-SCC9 cells (G-H) EphA2-CAR-ML NK cells were incubated with wild type UM-SCC9 or UM-SCC9 EphA2 knock-out target cells for 6 hours at a 5:1 total NK/Target ratio. Summary data show percentage of (G) IFN-γ and (H) CD107a positive cells. n=8; four independent experiments. Purified NK cells were used for all assays. Statistical significance was calculated by Two-way ANOVA. F shows a representative experiment out of three independent experiments.

Figure 6.. EphA2-CAR ML NK cells are effective at controlling primary HNSCC cells in vitro.

(A) Expression of the tumor antigen EphA2 on primary HNSCC tumors. ML NK cells (green) or EphA2-CAR-ML NK cells (brown) from normal donors were incubated with primary HNSCC tumor cells for 6 hours at a 5:1 total NK/Target ratio. (B-C) Summary data of IFN-γ and degranulation (CD107a+) for ML NK cells (green) compared to EphA2-CAR-ML NK cells (brown). n=10; 3 independent experiments. (D) Summary data of IFN-γ and degranulation for EphA2-CAR ML NK cells compared to EphA2-CAR with truncated intracellular signaling domain (EphA2-CAR^trunc ML NK cells). n=5; 3 independent experiments. (E) EphA2-CAR-ML NK cells control tumor growth at low E:T ratio of 0.5:1 in IncuCyte® assay. Error bars represent SEM. Statistical significance calculated by Two-way ANOVA Test. E shows a representative experiment out of three independent experiments.