Inference and adaptation in innate immunity

Cells such as natural killer cells and macrophages can recognize and eliminate targets with aberrant surface ligand expression without antigen specificity. This innate mechanism of activation must be tightly regulated to prevent autoimmunity. We describe a quantitative model of the regulation of nonspecific activation that is grounded in Bayesian inference. Our model captures known behaviors of innate immune cells, including adaptation to changing environments and the development of hyposensitivity after prolonged exposure to activating signals. Our analysis reveals a tradeoff between precision and flexible adaptation to different environments. Maintaining the ability to adapt naturally leads to heterogeneous responses, even for hypothetical populations of immune cells and targets that have identical surface receptor and ligand expression. Collectively, our results describe an adaptive algorithm for self/nonself discrimination that functions even in the absence of antigen restriction and supports biological observations of single-cell heterogeneity in response to cell-cell interactions. The same model could also apply more broadly to the adaptive regulation of activation for other immune cell types.