Rational design of memory in eukaryotic cells

Abstract

The ability to logically engineer novel cellular functions promises a deeper understanding of biological systems. Here we demonstrate the rational design of cellular memory in yeast that employs autoregulatory transcriptional positive feedback. We built a set of transcriptional activators and quantitatively characterized their effects on gene expression in living cells. Modeling in conjunction with the quantitative characterization of the activator-promoter pairs accurately predicts the behavior of the memory network. This study demonstrates the power of taking advantage of components with measured quantitative parameters to specify eukaryotic regulatory networks with desired properties.

Figure 1.

Synthetic transcriptional activators in yeast. (A) Schematic diagram of an activator cascade composed of an activator gene and a reporter gene. (B) DIC and fluorescence images of live cells containing the ZifH activator cascade in the absence (−galactose) and presence (+galactose) of the inducer or only the ZifH reporter gene in the presence of the inducer. (C) Dot plots of YFP versus RFP fluorescence intensity for single cells grown as in B.

Figure 2.

Steady-state and dynamic behavior of activation cascades. The mean steady-state reporter versus activator concentrations for the ZifH (A) and LexA (B) activator–reporter strains. The solid lines show best fits to the data. Only s and β could be robustly determined in B. (C) Time-lapsed RFP and YFP fluorescence images of cells containing the ZifH activation cascade 0, 75, and 150 min after induction. (D) Rate of change of reporter concentration as a function of activator concentration for individual cells (red, blue, green, and light-blue lines).

Figure 3.

A memory device in yeast. (A) Schematic diagram of the sensor and autofeedback genes of the memory device. (B) The predicted rate of change in the autofeedback activator concentration as a function of the total activator concentration for the LexA memory device. The decay rate (orange line) was calculated using Equation 2 and τ = 240 min, while the production rate (black line) was calculated using Equation 1; the values of s, n, and K for the LexA activator–promoter pair (Table 1, bold); and the estimated value of β for the LexA autofeedback activator. The high stable state (memory) and the unstable steady state (switch) of the autofeedback loop are indicated. (C) DIC and fluorescence images of live cells harboring the LexA memory device growing in the absence of stimulus (raffinose), in the presence of stimulus (galactose), and in the absence of stimulus after transient induction (raffinose after galactose). (D) The frequency distribution of RFP and YFP fluorescence intensities from single cells as in C.

Figure 4.

Regulating bistability of the memory device. (A) The calculated rate of change of the autofeedback activator concentration versus the total activator concentration for the LexA memory device. Two decay rates, calculated using Equation 2 with τ = 90 min (blue line) and 120 min (orange line), predict the effect of decreasing the growth rate on bistability of the autofeedback loop. The production rate versus activator concentration (black line) is as in Figure 3B. (B) DIC and fluorescence images of cells carrying the LexA memory device that were grown in inducer, then to steady state in glucose with a doubling time of 90 or 120 min. (C) The predicted rate of change of autofeedback activator concentration as a function of the total activator concentration for the LexA memory device and the dual autofeedback memory device. The production rates were calculated using Equation 1, the values of n and K for the LexA activator–reporter strain, and either β (black line) or 2*β (orange line) for the LexA autofeedback activator corresponding to one and two copies of the autofeedback gene, respectively. The decay rate (blue line) corresponds to a doubling time of 90 min. (D) The frequency distribution of RFP and YFP intensities from the LexA memory device strain with one and two copies of the autofeedback gene grown in glucose after induction.