24Mathematical models can enable a predictive understanding of mechanism in cell biology by quantitatively 25 describing complex networks of interactions, but such models are often poorly constrained by available 26 data. Owing to its relative biochemical simplicity, the core circadian oscillator in Synechococcus elongatus 27 has become a prototypical system for studying how collective dynamics emerge from molecular interac-28 tions. The oscillator consists of only three proteins, KaiA, KaiB, and KaiC, and near-24-h cycles of KaiC phos-29 phorylation can be reconstituted in vitro. Here, we formulate a molecularly-detailed but mechanistically 30 agnostic model of the KaiA-KaiC subsystem and fit it directly to experimental data within a Bayesian pa-31 rameter estimation framework. Analysis of the fits consistently reveals an ultrasensitive response for KaiC 32 phosphorylation as a function of KaiA concentration, which we confirm experimentally. This ultrasensitivity 33 primarily results from the differential affinity of KaiA for competing nucleotide-bound states of KaiC. We ar-34 gue that the ultrasensitive stimulus-response relation is critical to metabolic compensation by suppressing 35 premature phosphorylation at nighttime.
36
Synopsis 37This study takes a data-driven kinetic modeling approach to characterizing the interaction between KaiA and 38 KaiC in the cyanobacterial circadian oscillator and understanding how the oscillator responds to changes in 39 cellular metabolic conditions. 40 • An extensive dataset of KaiC autophosphorylation measurements was gathered and fit to a detailed 41 yet mechanistically agnostic kinetic model within a Bayesian parameter estimation framework. 42 • KaiA concentration tunes the sensitivity of KaiC autophosphorylation and the period of the full oscil-43 lator to %ATP. 44 • The model reveals an ultrasensitive dependence of KaiC phosphorylation on KaiA concentration as a 45 result of differential KaiA binding affinity to ADP-vs. ATP-bound KaiC. 46 • Ultrasensitivity in KaiC phosphorylation contributes to metabolic compensation by suppressing pre-47 Achieving a predictive understanding of biological systems and chemical reaction networks is challenging 50 because complex behavior can emerge from even a small number of interacting components. Classic ex-51 amples include the propagation of action potentials in neurobiology and chemical oscillators such as the 52 Belousov-Zhabotinsky reaction. The collective dynamics in such systems cannot be easily intuited through 53 qualitative reasoning alone, and thus mathematical modeling has long played an important role in summa-54 rizing and interpreting existing observations and formulating testable, quantitative hypotheses. 55 In general, mathematical modeling can be classified as either "forward" or "reverse." In forward mod-56 eling, known interactions are expressed mathematically, which allows a researcher to draw out the logical 57 implications of the model and its underlying assumptions (Gunawardena,...