UCSF

In this dissertation, we study the control network that governs the cell division cycle of budding yeast Saccharomyces cerevisiae. The goal is to examine the relationship between the network's topology, its function and its robustness against perturbations. We describe its design principles that give rise to the above properties.

In the study described in Chapter 2, we examine how the functional requirement for a network can constrain its topology and robustness. We investigate how the function performed by the simplified budding yeast cell cycle network [Li et al. 2004] can shape the control network itself. We found that the control network must contain negative feedback and positive feed-forward interactions. We then compared networks that can perform the same function with random networks and found that functional networks are much more robust against different types of perturbations. In addition, the yeast cell cycle network is even more robust compared to most of the functional networks. These results show that the topology, function and robustness of a network are intricately connected.

In the study described in Chapter 3, we examine how the switch-like transition from G1 to S phase of the budding yeast cell cycle is ensured by the underlying control network. We found that the double negative feedback loop (DNFBL) between the B-type cyclins (Clb5/6)-Cdk1 and its stoichiometric inhibitor Sic1 is important to the switch-like degradation of Sic1, a major step in G1/S transition. We show that the DNFBL guards against noise in and perturbations to the system. We also show the division of labor between G1/S cyclins (Cln1/2)-Cdk1 and Clb5/6-Cdk1 in controlling and maintaining the robustness of the onset and the dynamic of Sic1 degradation.

In the study described in Chapter 4, we further examine the multisite phosphorylation of Sic1 and the DNFBL using theoretical and modeling approach. We found that while multisite phosphorylation of Sic1 sets a threshold in CDK activity, whether it can be a switch depends on other factors such as cooperativity and feedback loop. We also show that the DNFBL between Sic1 and Clb5/6-Cdk1 can buffer variations in the CDK activation profile but not high frequency noise.