Antibody-mediated inhibition of MICA and MICB shedding promotes NK cell-driven tumor immunity

Abstract

MICA and MICB are expressed by many human cancers as a result of cellular stress, and can tag cells for elimination by cytotoxic lymphocytes through natural killer group 2D (NKG2D) receptor activation. However, tumors evade this immune recognition pathway through proteolytic shedding of MICA and MICB proteins. We rationally designed antibodies targeting the MICA α3 domain, the site of proteolytic shedding, and found that these antibodies prevented loss of cell surface MICA and MICB by human cancer cells. These antibodies inhibited tumor growth in multiple fully immunocompetent mouse models and reduced human melanoma metastases in a humanized mouse model. Antitumor immunity was mediated mainly by natural killer (NK) cells through activation of NKG2D and CD16 Fc receptors. This approach prevents the loss of important immunostimulatory ligands by human cancers and reactivates antitumor immunity.

Fig. 1.. MICA and MICB α3 domain–specific antibodies inhibit shedding and stabilize the protein on the surface of human tumor cells for recognition by NK cells.

(A) Illustration of MICA protein bound to a NKG2D homodimer (Protein Data Bank 1HYR). MICA is colored in grayand the NKG2D homodimer in blue and red. The NKG2D dimer binds to the α1 and α2 domains;the α3 domain is the site of proteolytic cleavage. (B) Binding of mAbs to immobilized MICA α3 domain or MICA α1 to α3 domains detected with a fluorescence-based ELISA (one representative of three independent experiments). (C and D) A375 cells were treated for 24 hours with the indicated mAbs. (C) MICA α3 domain–specific mAbs (7C6, 6F11, 1C2) inhibit MICA release into the supernatant, as quantified by sandwich ELISA; mAb 6D4 binds outside the MICA α3 domainand thus does not inhibit shedding. Data show mean ± SD for triplicate measurements from one representative of three independent experiments. (D) MICA α3 domain–specific mAbs stabilize MICA surface expression, as determined by flow cytometry using phycoerythrin (PE)–labeled 6D4 mAb. MFI, mean fluorescence intensity. Data show mean ± SD for triplicate measurements from one representative of three independent experiments. (E) Human NK cells exhibit cytotoxicity against A375 cells in the presence of 7C6-hIgG1 antibody (66.7 nM) but not isotype control antibody. Mean ± SD for quadruplicate measurements. ***P < 0.001 calculated by two-way analysis of variance (ANOVA) and Bonferroni’s post hoc test. Representative of three independent experiments (each experiment was done with different human NK cell donors).

Fig. 2.. Antitumor activity of antibodiesthat inhibit MICA and MICB shedding.

(A to C) C57BL/6 mice were injected intravenously with B16F10-MICA cells and treated with 7C6-mIgG2a, 7C6-mIgG2b, or isotype control antibodies (200 μg per injection) on days 1, 2,7, and 10. (A) Histological analysis (Fontana-Masson staining) of lung tissue demonstrated an apparent reduction in the number and size of metastases in mice that were treated with the MICA antibody (representative of five mice). (B) MICA antibody treatment reduced the number of superficial lung metastases counted by stereomicroscopy on day 14. Data indicate mean ± SD of pooled data from two independent experiments. (C) Serum concentrations ofshed MICA. Data show mean ± SEM for five mice per group and one representative of two independent experiments. (D) 7C6-mIgG2b antibody had activity against CT26-MICA lung metastases (intravenous injection of tumor cells into Balb/c mice; antibody injection on days 1, 2, 7, and 14; lung metastases counted on day 21). Data indicate mean ± SD of pooled data from two independent experiments. (E) Analysis of therapeutic activity of 7C6-mIgG2a antibody against subcutaneous B16F10 tumors that expressed full-length MICA or MICB or a secreted form of MICA, or were transduced with control vector. Tumor cells were inoculated into Igh^−/− mice; treatment with 7C6-mIgG2a or isotype control antibodies was started on day 5 (arrows) and repeated at every tumor measurement. Data show mean ± SEM for10 mice per group pooled from two independent experiments. (F and G) Efficacy of 7C6 antibody treatment required NK cells and their cytotoxic function. Same experimental design as in (A) to (C). (F) Quantification of pulmonary metastases in mice that were CD8 Tcell depleted (α-CD8β) or NK cell depleted (α-NK1.1 and a-asGM1). Data indicate mean ± SD pooled from two independent experiments. (G) Mice deficient for PRF1-mediated cytotoxicity, but not IFN-γ production, were unresponsive to 7C6 mAb treatment. Data indicate mean ± SD pooled from two independent experiments. *P < 0.05, **P < 0.01, and ***P < 0.001, calculated by unpaired Student’s t test (B) and (D) and two-way ANOVA and Bonferroni’s post hoc test (C) and (E) to (G).

Fig. 3.. NK cell activation by dual NKG2D receptor and CD16 engagement enhances antitumor immunity.

(A and B) Single-cell RNA-seq analysis of lung-infiltrating group 1 ILCs. On day 7 after intravenous injection of B16F10-MICA tumor cells, lung-infiltrating group 1 ILCs were isolated on the basis of NK1.1 and NKp46 staining (cells pooled from nine mice for isotype control and eight mice for 7C6-mIgG2a groups). Natural killer T–like cells that expressed both T cell and NK cell markers were also identified (even though TCRβ and CD3ε positive cells had been excluded), likely because the T cell receptor (TCR) is internalized after T cell activation.(A) t-distributed stochastic neighbor embedding (tSNE) plots illustrating identified cell populations in isotype control (left)–and 7C6-mIgG2a (right)–treated mice. Major populations and key markers are indicated. (B) Expression of key genes in group 1 ILCs on pooled data from isotype- and 7C6-mIgG2a–treated groups. FCER1G, Fc epsilon receptor gamma chain. (C) Fluorescence-activated cell sorting analysis of lung-infiltrating activated NK cells (EOMES⁺ GZMA⁺) across indicated time points relative to tumor burden (five mice per group and time point). Data indicate mean ± SEM; days 7 and 14 are representative of two independent experiments. (D) Contribution of NKG2D and Fc receptor activation to NK cell–mediated cytotoxicity. A375 cells were treated for 48 hours with indicated antibodies and then cocultured for 4 hours with human NK cells in a ⁵¹Cr-release assay. The 7C6-hIgG1-DANA mutant lacked binding to activating Fc receptors. NKG2D recognition was blocked with anti-NKG2D mAb 1D11. Data indicate mean ± SD and one representative of three independent experiments. (E) Both NKG2D and CD16 receptors contribute to therapeutic activity of MICA antibody. Rag2^−/− Il2rg^−/− mice were reconstituted with WT NK cells or NK cells mutant for NKG2D (Klrk1) and/or CD16 (Fcgr3a) genes. Mice werethen injected intravenously with B16F10-MICA cells and treated with 7C6-mIgG2b or isotype control antibodies, and lung metastases were quantified on day 14. Data indicate mean ± SD of pooled data from three independent experiments. **P < 0.01 and ***P < 0.001, calculated by two-way ANOVA with Bonferroni’s post hoc test (C) and (D) or multiple two-tailed unpaired Student’s t test (E).

Fig. 4.. MICA antibody shows therapeutic activity in a metastasis model with human tumor cells and human NK cells.

(A to C) NSG mice were reconstituted with IL-2 pretreated human NK cells. NK cell survival was supported by injection of human IL-2 on alternate days until day 8. One day after NK cell transfer, A2058 cells were injected intravenously into all mice. 7C6-hIgG1 or isotype control antibodies were administered on days1 and 2 after tumor inoculation and then once per week. On day 30, mice were euthanized, and metastases were counted in different organs. (A) Outline of the experimental procedure. (B) Quantification of the number of lung metastases by stereomicroscopy. Data are mean ± SD of pooled data from two independent experiments. (C) Quantification of liver metastases by stereomicroscopy. Data are mean ± SD of pooled data from two independent experiments. (D) Liposomal clodronate (or control liposomes) were injected intravenously to deplete liver macrophages (same day as tumor cell inoculation and then once per week). Liver metastases were quantified by stereomicroscopy3 weeks after tumor inoculation. Tumor cell inoculation and antibody treatments were done as shown in (A) but without NK cell reconstitution and with analysis of metastases a week earlier. Data are mean ± SD of pooled data from two independent experiments. (E) Survival analysis of mice reconstituted with human NK cells, inoculated intravenously with human melanoma cells, and treated with isotype or 7C6-hIgG1 antibodies. Same conditions as in (A). Data are pooled from two independent experiments.(F) Cartoon illustrating proposed therapeutic mechanism. *P < 0.05,**P < 0.01, and ***P < 0.001, calculated by two-way ANOVA with Bonferroni’s post hoc test (B) and (C) or multiple two-tailed unpaired Student’s t test(D). In (E), the comparison of survival curves is by Mantel-Cox test.