A subset of natural killer cells achieves self-tolerance without expressing inhibitory receptors specific for self-MHC molecules

Abstract

It is widely believed that self-tolerance of natural killer (NK) cells occurs because each NK cell expresses at least one inhibitory receptor specific for a host major histocompatibility complex (MHC) class I molecule. Here we report that some NK cells lack all known self-MHC-specific inhibitory receptors, yet are nevertheless self-tolerant. These NK cells exhibit a normal cell surface phenotype and some functional activity. However, they respond poorly to class I-deficient normal cells, tumor cells, or cross-linking of stimulatory receptors, suggesting that self-tolerance is established by dampening stimulatory signaling. Thus, self-tolerance of NK cells in normal animals can occur independently of MHC-mediated inhibition, and hyporesponsiveness plays a role in self-tolerance of NK cells, as also proposed for B and T cells.

Figure 1.

A subset of NK cells lacking known inhibitory receptors specific for self-MHC molecules. Freshly isolated B6 splenocytes were depleted of T cells, including NK1.1⁺ T cells, and stained with mAbs specific for NK1.1, Ly49C, Ly49I, and NKG2. Control staining was performed with control rat IgG (Ig ctrl.) instead of anti-NKG2 mAb and streptavidin-PE alone instead of anti-Ly49C and anti-Ly49I mAbs (ctrl.). Dot plots are gated on NK1.1⁺ cells. Numbers in the quadrants represent mean percentages ± SEM (n = 3 mice). An average of 11.4% of the NK cells lacked Ly49C, Ly49I, and CD94/NKG2 (CI/NKG2^- NK cells).

Figure 2.

CI/NKG2^- NK cells in B6 mice exhibit reduced responsiveness to NK-sensitive target cells. Splenocytes from poly(I:C)-treated β2m^-/-, B6, or B10.M mice were stimulated for 5 hours in the presence of brefeldin A with Con A–activated lymphoblasts from the designated mice or with tumor cell lines before staining the cells with mAbs specific for cell-surface markers and intracellular IFN-γ. Results are depicted as the percentage ± SEM (n = 3 mice) of IFN-γ⁺ cells in each subset (CI/NKG2⁺ or CI/NKG2^-). (A) CI/NKG2^- NK cells in B6 mice respond poorly to β2m^-/- lymphoblasts. Similar results were obtained in 5 independent experiments (P = .006 for the % IFN-γ⁺ cells in CI/NKG2⁺ versus CI/NKG2^-). (B) CI/NKG2^- NK cells in B10.M mice respond as well as CI/NKG2⁺ NK cells to β2m^-/- lymphoblasts or YAC-1 tumor cells. A repetition of the experiment yielded similar results. (C) CI/NKG2^- NK cells in B6 mice respond poorly to YAC-1 tumor cells. Results are representative of 3 experiments (P = .003 for CI/NKG2⁺ versus CI/NKG2^-). (D) CI/NKG2^- NK cells in B6 mice respond poorly to RMA tumor cell transductants expressing NKG2D ligands Rae1 or H60. Results are representative of 3 experiments (P < .001 for CI/NKG2⁺ versus CI/NKG2^-).

Figure 3.

Cytotoxicity of CI/NKG2^- NK cells. B6 mice were depleted of CI/NKG2⁺ NK cells by treatment with a mixture of the SW5E6 and 16a11 mAbs. The SW5E6 and 16a11 mAbs routinely deplete 95% or 80%, respectively, of the corresponding NK-cell subsets (data not shown). Control NK-cell populations were depleted of irrelevant Ly49A⁺ and Ly49G2⁺ NK cells by treatment with A1 and 4D11 mAbs, or were from untreated mice. The NK-cell populations were tested for lysis of β2m^-/- Con A blasts (left) or B6 Con A blasts (middle), or, in a separate experiment, YAC-1 tumor cells (right). The effector-target (E/T) ratio was calculated according to the number of NK1.1⁺CD3^- cells in the population. Each experiment was repeated with comparable results. Splenocytes from CI/NKG2^--depleted mice were significantly less cytotoxic than both Ly49A/Ly49G2-depleted or control mice against β2m^-/- Con A blasts (P ≤ .008 at 6:1 or 3:1 E/T ratios) or YAC-1 targets (P ≤ .03 for all E/T).

Figure 4.

CI/NKG2^- NK cells in B6 mice respond poorly to cross-linking of stimulatory receptors. (A-C) Splenocytes from poly(I:C)-treated β2m^-/- or B6 mice were cultured for 5 hours on plates coated with increasing concentrations of the anti-NKG2D mAb MI-6 (A), the anti–NKR-P1C mAb PK136 (B), the anti-Ly49D mAb SED85 (C), or control rat IgG (A-C) in the presence of brefeldin A before staining and analysis. NK cells were gated as NK1.1⁺CD3^- cells except in panel B, where they were gated as DX5⁺CD3^- cells. Results are depicted as percentage ± SEM (n = 3 mice) of IFN-γ⁺ cells in each subset. In panels B and C, data were normalized based on percentage of NKR-P1C⁺ or Ly49D⁺ NK cells, respectively, in each subset, as determined before stimulation. CI/NKG2^- NK cells produced significantly less IFN-γ than their CI/NKG2⁺ counterparts after anti-NKG2D (P ≤ .03 for all doses), anti–NKR-P1C (P ≤ .04 for all but the lowest dose), or anti-Ly49D stimulation (P ≤ .03 for all but the lowest dose). (D) NK cells were stimulated with the indicated dilutions of a mixture of ionomycin and PMA for 5 hours in the presence of brefeldin A before analysis. The highest concentrations (dilution factor = 1) were 2.5 μg/mL ionomycin and 250 ng/mL PMA.

Figure 5.

NK-subset activation in B6 mice infected with Listeria monocytogenes. At the indicated time points, splenocytes from Listeria-infected B6 mice were harvested and immediately stained with mAbs. The results are depicted as the percentage ± SEM of cells in each subset (gated also for NK1.1⁺CD3^- phenotype) with intracellular IFN-γ (n = 3 mice). A repetition of the experiment yielded similar results.

Figure 6.

Cell-surface phenotype of CI/NKG2⁺ and CI/NKG2^- NK subsets. Freshly isolated splenocytes from B6 mice (A) or β2m^-/- mice (B) were stained with mAbs for the 3 self-specific receptors (Ly49C, Ly49I, and NKG2), NK1.1, CD3, and the other markers indicated. Expression of various markers was compared on gated CI/NKG2⁺NK1.1⁺CD3^- cells (solid line) and CI/NKG2^-NK1.1⁺CD3^- cells (dashed line), except the NK1.1 staining, which was gated on CD3^-NKG2D⁺ cells. The first histogram represents the negative control for all the samples with the exception of NKG2D; the negative control for NKG2D staining is the shaded histogram within that panel.

Figure 7.

Depletion of CI/NKG2⁺ NK cells prevents rejection of bone marrow grafts from class I–deficient mice. Recipient mice (β2m^-/- or B6) were treated with mAbs to deplete the indicated NK-cell subset, followed by lethal irradiation. At day 0, a mixture of CFSE-labeled bone marrow cells from β2m^-/- Ly5.1⁺ and B6 (Ly5.2) mice was injected intravenously. After 3 days, splenocytes were stained with PE-conjugated anti-Ly5.1 Ab. The results are represented as graft acceptance, calculated as the ratio of β2m^-/- (Ly5.1⁺) bone marrow cells to B6 (Ly5.1^-) bone marrow cells among gated CFSE-positive cells. The B6 cells serve as a nonrejected internal reference population. (A) Representative histograms depicting results in which the test graft was rejected (in untreated B6 mice, left) or accepted (B6 mice from which NK1.1⁺ cells were depleted, middle). The effect of depleting Ly49C/I⁺ and NKG2⁺ NK cells is shown in the right panel. The percentage of positive cells (marked by a bar) are indicated in each panel. Note that many of the nonstaining cells in all the panels are in the lowest fluorescence channel along the left axis due to the compensation settings of the flow cytometer. (B) Acceptance of β2m^-/- bone marrow grafts by control mice or B6 mice depleted of NKG2A⁺ NK cells, Ly49C/I⁺ NK cells, or both subsets. Data represent the mean ± SEM (n = 3 mice). Results were reproduced in a replicate experiment. (C) Acceptance of β2m^-/- bone marrow grafts by B6 mice treated with mAbs to deplete the indicated cell populations, or by control β2m^-/- mice. Data represent the mean ± SEM (n = 2 mice). Results were reproduced in 2 replicate experiments. The Ly49C/I- and NKG2A-depleted group differed significantly from the Ly49A/G2-depleted group (P ≤ .04) but not from the NK1.1-depleted group (P = .5).