Antitumor immunity. A shed NKG2D ligand that promotes natural killer cell activation and tumor rejection

Abstract

Immune cells, including natural killer (NK) cells, recognize transformed cells and eliminate them in a process termed immunosurveillance. It is thought that tumor cells evade immunosurveillance by shedding membrane ligands that bind to the NKG2D-activating receptor on NK cells and/or T cells, and desensitize these cells. In contrast, we show that in mice, a shed form of MULT1, a high-affinity NKG2D ligand, causes NK cell activation and tumor rejection. Recombinant soluble MULT1 stimulated tumor rejection in mice. Soluble MULT1 functions, at least in part, by competitively reversing a global desensitization of NK cells imposed by engagement of membrane NKG2D ligands on tumor-associated cells, such as myeloid cells. The results overturn conventional wisdom that soluble ligands are always inhibitory and suggest a new approach for cancer immunotherapy.

Figure 1. NK cells promote the rejection of tumors that shed MULT1

(A) ELISA detection of soluble MULT1 in sera from tumor bearing Eμ-Myc mice, nontransgenic littermates, and diseased Apoe−/− mice fed a Western diet (n=6–8). Each point represents a different mouse. (B) Comparison of growth of 2 × 10⁴ subcutaneously transferred B16 melanoma tumor cells transduced with secMULT1, full length MULT1 or empty vector, in WT B6 mice (n=4 mice). Rejection was usually partial but was complete in some animals in some experiments. (C) Subcutaneous growth of B16-secMULT1 tumors in B6 mice (2 × 10⁴ cells were inoculated) treated with control IgG, NK1.1 antibody or CD8 antibody (n=13 mice). (D) After inoculation of 2 × 10⁴ B16 cells transduced with pFG12-secMULT1, mice were treated or not with doxycycline starting from the time of tumor implantation (n=6 mice). (E) Mice (n=6) received 2 × 10⁴ B16 cells alone, or 2 × 10⁴ B16 cells mixed with 2 × 10³ B16-secMULT1 cells. Panels show representative examples of ≥3 (panels B and E) or 2 (panel D) experiments performed, whereas panel C includes combined data from 3 experiments. Tumor volumes ± SE are shown. Panel A was analyzed with a Mann-Whitney test, and panels B-E were analyzed by 2 way ANOVA with Bonferroni multiple comparison tests. *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001.

Figure 2. Soluble MULT1 amplifies NK cell responses and causes tumor rejection

(A–C) B6 mice were injected i.p. with 5 × 10⁶ irradiated B16 or B16-secMULT1 cells, or PBS. Peritoneal wash cells (pooled from 5 mice) were recovered three days later and tested for killing of YAC-1 target cells (A) or tested for intracellular IFNγ after stimulation with YAC-1 cells (B) or immobilized NKp46 antibody (C); control responses to PBS are depicted by white segments of the bars. (D) B6 mice were injected subcutaneously with 3–5 × 10⁵ B16 or B16-secMULT1 cells in 100 μl matrigel. The tumors were dissociated 7 days later, and gated NK cells from individual mice (n=5) were tested for responses to immobilized NKp46 or NKRP1C Abs. (E–F) Subcutaneous tumors were established with 3–5 × 10⁵ B16 cells in matrigel. The tumor cells in one group were mixed with 1 μg of recombinant MULT1 (rMULT1). After 4 days, an additional 1 μg of rMULT1 (or PBS for control mice) was injected into each matrigel/tumor for that group. On day 7, tumors were extracted, weighed (E), dissociated, and the tumor cells were counted (E). The immune cells within the tumors were stimulated with immobilized NKp46 and NKRP1C Abs, and the IFNγ responses of gated NK cells were determined (F). Panels show representative examples of 2 (panel A) or ≥3 (panels B–F) experiments performed. Panels A–D and F were analyzed by 2-way ANOVA with Bonferroni multiple comparison tests, panel E was analyzed by Mann-Whitney test. *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001.

Figure 3. Mechanisms of immune activation by soluble MULT1

(A, B) Membrane NKG2D staining after exposure of NK cells to secMULT1 in intraperitoneal (A) or subcutaneous (B) tumors. (C) Model of secMULT1 action. Persistent NKG2D engagement by endogenous NKG2D ligand-expressing cells associated with the tumor desensitizes NK cells. Soluble MULT1 competitively blocks the NKG2D receptor, preventing NK cell desensitization and therefore augmenting tumor rejection mediated through distinct NK activating receptors. (D) Expression of NKG2D ligand RAE-1 by gated CD11b⁺F480⁺ peritoneal myeloid cells in mice injected i.p. 3 days before with PBS or 5 × 10⁶ irradiated B16 or B16-secMULT1 tumor cells. Cells were stained with biotin-pan-RAE-1 Ab (blue). The staining was specific as it could be blocked by including an excess of unconjungated pan-RAE-1 antibody in the reaction (red). Grey shows isotype control staining. (E) Expression of RAE-1 by gated CD11b⁺F480⁺ intratumoral myeloid cells in mice injected SC with 2 × 10⁴ B16 or B16-secMULT1 tumor cells 20 days before. (F) rMULT1 and NKG2D antibody (MI6 clone, in F(ab′)₂ form) block RAE-1 binding to NKG2D on NK cells. The MFI of RAE1ɛ-Fc staining of NK cells was used to calculate % inhibition. Panel A is combined data from 14 experiments. Panels B, D–F show representative examples of ≥3 experiments performed. Panel A was analyzed by one-way ANOVA Kruskal-Wallis test, panel B was analyzed by 2-way ANOVA with Bonferroni multiple comparison tests, ns indicates P>0.05, *P < 0.05, and ***P < 0.001.

Figure 4. Augmented NK cell responses in RAE-1-deficient and NKG2D-deficient mice

(A–C) Peritoneal NK cells from Raet1d^−/−Raet1e^−/− mice (KO) exhibited increased amounts of cell surface NKG2D (A) and increased functional responses ex vivo to YAC-1 tumor cells (B) or NKp46 antibody stimulation (C); control responses to PBS are depicted by white segments of the bars. The effects were larger when the mice were injected 3 days earlier with irradiated B16 tumor cells, but smaller when they were injected with B16-secMULT1 tumor cells. (D) NKG2D-deficient (Klrk1^−/−) NK cells exhibited increased functional activity. Splenic NK cells from Rag^−/− and Rag^−/− Klrk1^−/− mice were stimulated ex vivo with immobilized NKp46 or NKRP1C Abs, and the IFN-γ responses of gated NK cells were determined. (E) B6 mice were injected i.p. with 50 μg MI6 (anti-NKG2D) F(ab′)₂ or F(ab′)₂ of rat IgG on days 0, 3 and 6. On Day 8, peritoneal NK cells were stimulated ex vivo with immobilized NKp46 Abs, and the IFNγ responses of gated NK cells were determined. (F, G) Increased tumor rejection responses in NKG2D-deficient mice. Growth of B16-secMULT1 (F) or B16 (G) tumor cells in Rag2^−/− or Rag2^−/−Klrk1^−/− mice (n=5). Panels F and G are from separate experiments. Separate, direct comparisons showed retarded growth of B16-secMULT1 vs B16 tumors in Rag2^−/− mice. All experiments show representative examples of ≥3 experiments performed. Tumor volumes ± SE are shown. Figure 4A–G were analyzed by 2-way ANOVA with Bonferroni multiple comparison tests. *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001.