Effects of dose-response relationships on cellular information transmission

Cell signaling systems can transmit information with high precision, despite relying on complex reaction networks and inherently noisy biochemical reactions. To better understand how they achieve this precision, this work focuses on how cells can tune their quantitative steady-state dose-response relationships to improve signaling fidelity. Using a combination of analytic and numerical approaches, we quantify the maximum information that can be transmitted in single-step and multi-step signaling systems under several biologically realistic noise distributions: Gaussian, gamma, constant coefficient of variation, and binomial. We find that signaling systems transmit more information if their dose-response curves have low baselines, large dynamic ranges, and curve shapes that are well matched to the natural distribution of input signals. For multi-step systems, linear transfer functions typically improve signaling fidelity for all noise levels, consistent with previous work on "dose-response alignment". We also find that switchlike transfer functions transmit a maximum of one bit and are robust to low levels of noise, but fail rapidly with high levels of noise. Simulations of toy signaling systems corroborate these results. Together, these findings provide insight into some cellular design principles that have arisen through evolution and that are widely observed in nature.

Cell signaling systems transmit information about extracellular conditions into the cell, where it can be processed and acted upon. At the steady state, these system responses can be described by dose-response curves that relate system output to input. Because the shapes of these curves can presumably be tuned through evolution, this raises the question of what shape dose-response curves maximize the accuracy of signal transmission. Using a combination of analytical theory and numerical calculations, we find that signaling systems transmit the most information when their dose-response curves are matched to the natural distribution of input signals. In signaling pathways composed of multiple sequential steps, information transmission is typically maximized when downstream signaling steps respond linearly to the outputs of upstream steps. In principle, cells could also use switchlike responses, in which signals are either on or off, but we show that these approaches fail rapidly with high noise levels. These predicted relationships agree with simulation results and are broadly consistent with observations of natural signaling systems.