Minimal gene regulatory circuits for a lysis-lysogeny choice in the presence of noise.

Gene regulatory networks (GRNs) that make reliable decisions should have design features to cope with random fluctuations in the levels or activities of biological molecules. The phage λ GRN makes a lysis-lysogeny decision informed by the number of phages infecting the cell. To analyse the design of decision making GRNs, we generated random in silico GRNs comprised of two or three transcriptional regulators and selected those able to perform a λ-like decision in the presence of noise. Various two-protein networks analogous to the λ CI-Cro GRN worked in noise-less conditions but failed when noise was introduced. Adding a λ CII-like protein significantly improved robustness to noise. CII relieves the CI-like protein of its 'decider' function, allowing CI to be optimized as a decision 'maintainer'. CII's lysogenic decider function was improved by its instability and rapid removal once the decision was taken, preventing its interference with maintenance. A more reliable decision also resulted from simulated co-transcription of the genes for CII and the Cro-like protein, which correlates fluctuations in these opposing decider functions and makes their ratio less noisy. Thus, the λ decision network contains design features for reducing and resisting noise.