Table 1.
Hypothetical Interactions for the A. thaliana CC Network.
Table 2.
Experimental Interactions for the A. thaliana CC Network and their Evidence.
Fig 1.
Regulatory network of the A. thaliana CC.
The network topology depicts the proteins included in the model as well as the relationship among them. Nodes are proteins or complexes of proteins and edges stand for the existing types of relationships among nodes. The trapezoid nodes are transcription factors, the circles are cyclins, the squares are CDKs, the triangle represent stoichiometric CDK inhibitor, the hexagons are E3-ubiquitin ligase complexes and the octagon is a negative regulator of E2F proteins. Edges with arrow heads are positive regulations and edges with flat ends illustrate negative regulations. The red edges indicate regulation by phosphorylation while blue ones indicate ubiquitination, the green ones show physical protein-protein interactions and the black edges transcriptional regulation. Only CDK-cyclin interactions are not represented with a line. Interactions to or from rhombuses stand for interactions that involve the CDK as well as the cyclin. A solid line indicates that there is experimental evidence to support such interaction and dotted lines represent proposed interactions grounded on evidence from other organisms or in silico analysis.
Fig 2.
Attractor corresponding to a dynamic network of CC in A. thaliana.
100% of the whole set of network configurations converges to a unique attractor composed by 11 configurations. Each column is a network configuration (state of each network component) and the rows represent the state of each node during CC progression. The squares in green indicate components that are in an “ON” state and the ones in red are nodes in an “OFF” state.
Fig 3.
Attractor robustness analysis.
Random networks with similar structure to A. thaliana CC GRN were less tolerant to perturbations than original CC GRN. The frequency of perturbations that recovered the original attractor after a perturbation in the Boolean functions, is shown in: (A), where the red line indicates that A. thaliana CC GRN recovers its original attractor in 68% of perturbations (the median of random networks was 18.55% and mean 19.12% ± 13.86 SD). When transitions between network configurations are perturbed (B), A. thaliana CC GRN recovers its original attractor in 88% (vertical red line) of perturbations, while the median of random networks that recover the original attractor was 24.2% (mean 24.6% ± 18.2 SD). Vertical blue line indicates the 95% quantile. 1000 random networks were analyzed.
Fig 4.
Continuous version of the A. thaliana CC Boolean model.
In this graph we show the activity of the CDKA;1-CYCD3;1 and the CDKA;1-CYCB1;1 complexes as a function of the amount of cyclins, and KRP1 inhibitor. The CDK-cyclin activity is the limiting factor to pass the G1/S and the G2/M checkpoints. A little more than two complete CC are shown (upper horizontal axis) to confirm that oscillations are maintained.
Table 3.
Phenotypes of gain-of-function mutations in CC components.
Table 4.
Phenotypes of loss-of-function mutations in CC components.
Fig 5.
Attractors recovered by simulations of loss- or gain-of-function mutants of four CC components.
(A) The simulation of loss of CDKB1;1 function produced only one cyclic attractor with period 7 that resembles G1 → S → G2 → G1 cycle, whereas in (B) with simulation of loss of KRP1 function, one cyclic attractor was attained, which has period 11 and comprises 100% of the initial conditions. This attractor is almost identical to WT phenotype but without KRP1. With the simulation of APC/C gain-of-function, a single attractor with period 7 was recovered, which is shown in (C) and is consistent with an endoreduplication cycle. Attractors obtained with the simulation of E2Fa overexpression are shown in (D). Two attractors were found, one of them has period 10 and the 40.48% of the initial conditions converge to that cycle that is closely similar to the WT CC attractor. The second attractor that correspond to E2Fa overexpression has period 8 and it is very similar to the endoreduplication attractor of loss of CDKB1;1 function, which comprises 59.52% of possible network configurations.
Fig 6.
Dynamical behavior of E2Fc and KRP1 according to the continuous model.
These nodes were chosen by their peculiar pattern of expression, which was qualitatively recovered by the Boolean and continuous models.