Noise Minimization in Cell-Free Gene Expression

Abstract

Biochemical reactions that involve small numbers of molecules are accompanied by a degree of inherent randomness that results in noisy reaction outcomes. In synthetic biology, the ability to minimize noise particularly during the reconstitution of future synthetic protocells is an outstanding challenge to secure robust and reproducible behavior. Here we show that by encapsulation of a bacterial cell-free gene expression system in water-in-oil droplets, in vitro-synthesized MazF reduces cell-free gene expression noise >2-fold. With stochastic simulations we identify that this noise minimization acts through both increased degradation and the autoregulatory feedback of MazF. Specifically, we find that the expression of MazF enhances the degradation rate of mRNA up to 18-fold in a sequence-dependent manner. This sequence specificity of MazF would allow targeted noise control, making it ideal to integrate into synthetic gene networks. Therefore, including MazF production in synthetic biology can significantly minimize gene expression noise, impacting future design principles of more complex cell-free gene circuits.

Keywords: MazF; cell-free gene expression; gene expression noise; mRNA degradation; microfluidics; transcription and translation.

Figure 1

Overview of the effects of T7p10-MazF on cell-free gene expression. (A) Genetic composition of the noise-minimizing module applied in cell-free gene expression system (left). Adding the MazF-expressing template reduces the deGFP expression noise (middle). To understand the mechanism of noise modulation, stochastic simulations were performed (right). (B) A cell-free gene expression reaction is composed of an in vitro transcription–translation (IVTT) mix and DNA templates (top). deGFP expression yields of 2 nM T7p14-deGFP in the absence and presence of 250 pM T7p10-MazF (bottom). The expressed MazF protein reduces deGFP expression by promoting mRNA degradation in an ACA-sequence-specific manner (bottom right). Each point is the deGFP yield of one replicate.

Figure 2

Effect of overexpressed MazF protein on mRNA degradation. (A) Schematic overview of the experimental workflow. (B) Denaturing agarose gel images of mRNA incubated in IVTT mixtures. Purified wild-type 17ACAdeGFP mRNA (top) or recoded 0ACAdeGFP mRNA (bottom). (C) Half-life of deGFP mRNA with 0, 1, 2, or 17 ACA sites in IVTT expression mixture containing overexpressed MazF. The bar plot displays the half-life values obtained from fitting an exponential decay function to the band intensities of the denaturing agarose gels (Figures 2B and S3), and the error bars represent the standard error of the fit. The inset shows the half-life of each mRNA versus the number of ACA sites in each mRNA (dashed line: half-life = 9.6 – 2.6 ln(# ACA sites)).

Figure 3

Batch expression of a range of T7p14-deGFP and T7p10-MazF template combinations. (A) Schematic overview of experimentally tested conditions. (B) Effect of increasing concentrations of the T7p10-MazF template on deGFP synthesis. The yields after 10 h of expression (highlighted in green) were used in (C). (C) deGFP yields after 10 h of expression for a range of T7p10-MazF and T7p14-deGFP concentrations. The red squares represent the three template combinations used for the droplet experiments. All time courses are shown in Figure S7.

Figure 4

Implementing the noise-minimizing module in picoliter droplets. (A) Schematic overview of the microfluidic droplet production device. (B) Representative microscope images of the droplets immediately after production (left) and at the end of the experiment (right). Scale bars = 100 μm. (C) Single droplet trajectories of deGFP yield for the expression from the 1 nM T7p14-deGFP template. The histogram (right) represents the distribution of droplet intensities after 6.5 h. Mean (μ) and standard deviation (σ) are used to calculate the gene expression noise (Fano factor = σ²/μ). (D) deGFP yield for three tested conditions, with five positions per condition. (E) Average gene expression noise (Fano factor, see (C)) for the three tested conditions, with five positions per condition. The noise values after 6.5 h (highlighted in green) were used in (F). (F) Fano factor for the three tested conditions after 6.5 h. Each point is the Fano factor of a single position.

Figure 5

Stochastic modeling of the effect of MazF on gene expression noise. (A) Schematic representation of the three analyzed models: No MazF (only T7p14-deGFP template present (gray)), No feedback (MazF acts on the degradation of deGFP mRNA but not its own (blue)), and Full system (MazF acts on the degradation of both the deGFP mRNA and its own mRNA (red)). (B) Single trajectories of the simulated deGFP production for the No MazF model; 250 traces were randomly selected from the complete set of 1000 traces. (C) Mean deGFP production of all trajectories over time for all three models. (D) Average Fano factor of the deGFP yield over time for all three models. The noise values after 6.5 h (highlighted in green) were used in (E). (E) Analysis of the Fano factor at the 6.5 h simulated time point. Each point is the Fano factor of a subsampled population of 250 simulated droplets.