Probing natural killer cell education by Ly49 receptor expression analysis and computational modelling in single MHC class I mice

Abstract

Murine natural killer (NK) cells express inhibitory Ly49 receptors for MHC class I molecules, which allows for "missing self" recognition of cells that downregulate MHC class I expression. During murine NK cell development, host MHC class I molecules impose an "educating impact" on the NK cell pool. As a result, mice with different MHC class I expression display different frequency distributions of Ly49 receptor combinations on NK cells. Two models have been put forward to explain this impact. The two-step selection model proposes a stochastic Ly49 receptor expression followed by selection for NK cells expressing appropriate receptor combinations. The sequential model, on the other hand, proposes that each NK cell sequentially expresses Ly49 receptors until an interaction of sufficient magnitude with self-class I MHC is reached for the NK cell to mature. With the aim to clarify which one of these models is most likely to reflect the actual biological process, we simulated the two educational schemes by mathematical modelling, and fitted the results to Ly49 expression patterns, which were analyzed in mice expressing single MHC class I molecules. Our results favour the two-step selection model over the sequential model. Furthermore, the MHC class I environment favoured maturation of NK cells expressing one or a few self receptors, suggesting a possible step of positive selection in NK cell education. Based on the predicted Ly49 binding preferences revealed by the model, we also propose, that Ly49 receptors are more promiscuous than previously thought in their interactions with MHC class I molecules, which was supported by functional studies of NK cell subsets expressing individual Ly49 receptors.

Figure 1. Description of the models.

Flow chart of the sequential model simulation (A), and the two-step selection program (B). In each run, the parameters of the simulation are fixed throughout the run. The simulations were each run for a large number of parameter sets (1,700,000 sets for the two-step selection simulation and 110,000 sets for the sequential model simulation), for each MHC background and staining. See text for details.

Figure 2. Cell surface expression frequencies (in percentages) of receptor pairs and triplets on NK cells (NK1.1⁺CD3⁻) in single-MHC mice.

The expression frequencies shown are averages of measurements in three to six animals. Each panel shows results from a staining by a different trio of antibodies; from top to bottom: (A/G2/I), (A/C/G2), (A/C/I), (C/G2/I). The frequencies are exclusive, that is, e.g., the expression frequency of “A” in the (A/G2/I) staining (top panel) includes only cells that were found to be Ly49A⁺G2⁻I⁻ in that experiment, etc.. Error bars denote the data standard deviations, and asterisks denote that the difference from the frequency of the same receptor combination in MHC⁻ mice is statistically significant (using Student's T test).

Figure 3. Functional capacity of NK cells from single MHC class I mice expressing only one inhibitory receptor suggest functional outcome of previously unrecognized Ly49-MHC class I interactions.

The figures in each plot show percent CD107a⁺ NK cells, expressing only the indicated inhibitory receptor, after stimulation with 50 µg/ml PK136, recognizing the NK1.1 antigen, in vitro. One representative experiment out of 3.