A synthetic-natural hybrid oscillator in human cells

Abstract

Recent studies have shown that many cell-signaling networks contain interactions and feedback loops that give rise to complex dynamics. Synthetic biology has allowed researchers to construct and analyze well-defined signaling circuits exhibiting behavior that can be predicted and quantitatively understood. Combining these approaches--wiring natural network components together with engineered interactions--has the potential to precisely modulate the dynamics of endogenous signaling processes and control the cell decisions they influence. Here, we focus on the p53 signaling pathway as a template for constructing a tunable oscillator comprised of both natural and synthetic components in mammalian cells. We find that a reduced p53 circuit implementing a single feedback loop preserves some features of the full network's dynamics, exhibiting pulses of p53 with tightly controlled timing. However, in contrast to the full natural p53 network, these pulses are damped in individual cells, with amplitude that depends on the input strength. Guided by a computational model of the reduced circuit, we constructed and analyzed circuit variants supplemented with synthetic positive and negative feedback loops and subjected to chemical perturbation. Our work demonstrates that three important features of oscillator dynamics--amplitude, period, and the rate of damping--can be controlled by manipulating stimulus level, interaction strength, and feedback topology. The approaches taken here may be useful for the rational design of synthetic networks with defined dynamics, and for identifying perturbations that control dynamics in natural biological circuits for research or therapeutic purposes.

Fig. 1.

A reduced p53–Mdm2 circuit generates graded damped oscillations of fixed frequency. (A) Feedbacks regulating p53 dynamics after IR. IR activates ATM and Chk2 phosphorylation, which leads to p53 stabilization. Negative feedbacks act through the Mdm2 ubiquitin ligase to degrade p53, and through the Wip1 phosphatase to inactivate the upstream kinases. (B) A reduced p53 circuit based on transcriptional activation and the p53–Mdm2 feedback loop. Zinc stimulates p53–CFP transcription from a metallothionein (MT) promoter, bypassing IR-induced activation through the ATM/Chk2 kinase cascade and Wip1 negative feedback loop. Induced p53 activates Mdm2 transcription, which negatively regulates p53 stability. (C) Protein levels of p53 and its regulators in response to transcriptional activation of p53. Cells expressing p53–CFP under the MT promoter were treated with 50 μM zinc or 10 Gy IR, and samples were taken at the indicated time points. Protein levels were analyzed by Western blot, with actin as a loading control. Zinc led to p53 induction followed by Mdm2 induction. Zinc did not affect the level of Chk2 phosphorylation in contrast to activation of the full network by IR. (D and E) Fluorescence intensities of p53–CFP (blue curve) and Mdm2–YFP (green curve) from single cells following IR (D) and transcriptional stimulation by zinc (E). IR (full network) leads to undamped p53 oscillations and zinc induction (p53–Mdm2 core feedback) induces damped oscillations. (F and G) Heat maps of 25 representative cells after 50 μM zinc treatment. Each row represents a single cell. The p53–CFP (F) and Mdm2–YFP (G) levels were normalized to the maximum amplitude for each cell. (H) p53–CFP pulse timing and (I) fluorescence are shown for each pulse after stimulation with Zinc or IR (mean + SEM computed from at least 50 cells per condition).

Fig. 2.

Mathematical modeling of the core p53–Mdm2 negative feedback circuit. (A–C) A fitted model reproducing p53/Mdm2 dynamics. The p53 (blue) and Mdm2 (green) amplitudes (A), frequency of p53 oscillation (B), and p53 damping coefficient (C) at varying zinc concentrations are shown. The mean amplitude is plotted for each pulse; mean frequency and damping were taken across all pulses. Points represent experimental data from the zinc-stimulated cells shown in Fig. 1 H and I (mean + SEM); curves show model results. (D) Model dynamics in the presence of transcriptional noise. The lower panel shows modeled p53 (blue trace) and Mdm2 (green trace) levels after 25 μM zinc stimulation in the presence of noise in p53 and Mdm2 production. The upper trace shows the corresponding transcriptional noise applied to p53 [ξ_p(t)] and Mdm2 [ξ_m(t)] (SI Appendix, “Noise simulations”). The p53 timing (E) and amplitude (F) for 100 modeled cells are plotted for each pulse after stimulation (mean + SEM).

Fig. 3.

Model predictions and experimental perturbations for modulating oscillations damping rate. Model predictions for damping rate in a circuit with positive and negative feedbacks (NPF, A) or two negative feedbacks (2NF, B). The x and y axes represent variation of two parameters, γ_f0 and β_p0, across two orders of magnitude, representing feedback protein production delay and its effect on p53 transcription. Color bars show the rate of damping at each point, where white indicates the value from the model without additional feedback. Color scales on each plot range from 50 to 150% of the unperturbed value, permitting relative comparisons between plots. (C and D) Schematics of experimental synthetic feedback systems: (C) A combination of a negative and positive feedbacks on p53; the negative feedback is the natural feedback through Mdm2 (red arrows) and the positive feedback is a synthetic feedback through MTF1 (orange arrows). (D) Two negative feedbacks on p53; one through Mdm2 (red arrows) and the second through MTF1-KRAB (orange arrows). The p53 amplitude (E) and timing (F) measured in 1NF (green), NPF (orange), and 2NF (blue) cells after stimulation with 50 μM zinc.

Fig. 4.

Model predictions and experimental perturbations for modulating oscillation frequency. (A) Predicted oscillation frequencies in the p53–Mdm2 circuit are plotted against Nutlin3A dose, measured in multiples of its IC50. (B) Representative single-cell trajectories showing p53–CFP (blue trace) and Mdm2–YFP (green trace) after treatment with 50 μM zinc in the presence of 2.5 μM Nutlin3A. (C) Oscillation frequency in nutlin-pretreated cells compared with control cells. (D) Distribution of oscillation frequencies in response to transcriptional induction of p53 by zinc in Nutlin3A-pretreated and control cells.

Fig. 5.

Synthetic–natural hybrid systems can provide insights into the full network. (A) A synthetic circuit (Left) is constructed from a small number of components (red nodes) in an isolated context. Although easy to manipulate, such circuits are often too distinct from natural biological networks and therefore might provide limited insights. A complete natural system (Right) is highly interconnected and complex. A synthetic–natural hybrid system (Center) includes a reduced number of interactions (faded blue nodes/arrows are not activated) in their natural context (blue background) supplemented by artificially engineered feedbacks (yellow nodes). Such systems are less complicated than full networks, but can be perturbed in a controlled way and are thus useful for developing insights and predictions about the behavior of the full natural network. (B) Oscillation frequency in cells irradiated with 10 Gy IR, and in 10 Gy IR-irradiated cells pretreated with 2.5 μM Nutlin3A. Perturbation of the p53/Mdm2 interaction by Nutlin3A reduces the frequency of p53 oscillations in response to DNA damage.