A portable expression resource for engineering cross-species genetic circuits and pathways

Genetic circuits and metabolic pathways can be reengineered to allow organisms to process signals and manufacture useful chemicals. However, their functions currently rely on organism-specific regulatory parts, fragmenting synthetic biology and metabolic engineering into host-specific domains. To unify efforts, here we have engineered a cross-species expression resource that enables circuits and pathways to reuse the same genetic parts, while functioning similarly across diverse organisms. Our engineered system combines mixed feedback control loops and cross-species translation signals to autonomously self-regulate expression of an orthogonal polymerase without host-specific promoters, achieving nontoxic and tuneable gene expression in diverse Gram-positive and Gram-negative bacteria. Combining 50 characterized system variants with mechanistic modelling, we show how the cross-species expression resource's dynamics, capacity and toxicity are controlled by the control loops' architecture and feedback strengths. We also demonstrate one application of the resource by reusing the same genetic parts to express a biosynthesis pathway in both model and non-model hosts.


Supplementary Note 6
Parameter values in Supplementary Note 5 were estimated by following a six step fitting process, using fminsearch (MATLAB, Mathworks) to minimize the model's error functions: Step-1  Closed-loop Positive Feedback Loop (PFL+) model was fitted using the time course GFP fluorescence data for T7 RNAP RBSes 328, 1156 and 2244 ( Figure 3C). RBS 9158 was excluded because those cells had GFP fluorescence much higher than basal level and remained in a prolonged lag-phase until 335 min, presumably due to toxicity from the residual pre-chill T7 RNAP (see growth curve below).
 An initial condition of [mRNA T7RNAP =0, T7RNAP=0, mRNA GFP =0, GFP=0] was used for all simulations with some assumptions: (1) most protein production has been stopped by chilling of inoculant cultures, (2) any pre-chill residual levels of mRNA and protein within the inoculant culture will be diluted by cell division, and (3) the amount of the pre-chill residual level of mRNA or protein will be low and will not significantly affect the model fitting. These assumptions seem to hold well for RBS variants 328, 1156 and 2244 where GFP fluorescence at the initial time points is very close to basal level.  Two parameters were manually set (δ T7RNAP and δ GFP ), and ten parameter values were estimated in this step of fitting (33.4% average error)─ CopyN, k init , k b , k f , k elong , PR T7RNAP , PR GFP * Flu PC , primingR T7RNAP , leakyR GFP , and δmRNA. Error function used was |∆x|/x.

RBS-9158
 A sensitivity analysis was carried out, and six of the ten parameters were found to be most sensitive for the model solution─ k init , k f , PR T7RNAP , PR GFP * Flu PC , primingR T7RNAP , and δmRNA. These were re-fitted to the same data, obtaining a best-fit with 16.2% average error.
Step-2  The effect of T7 RNAP toxicity was modelled using Hill Function kinetics, multiplying all protein translation rates by a factor of 1/(1+T7RNAP Ptox ) to account for reduction in global translation. The Positive Feedback Loop (PFL+) model described above was modified to include the toxicity parameter Ptox, and its value was estimated by fitting to steady state GFP fluorescence data and Specific Growth Rates ( Figure 3B, right), using the error function |∆ log x|/log x.
 Four of the five RBSes were fitted at a time. For each fitting, average error of the fitted RBS set as well as that of the excluded RBS were reported for cross-validation. Step-3  The ten parameter values obtained above were plugged into the Open Loop (PFL-) model, and two system-specific parameters (primingR T7RNAP and leakyR GFP ) were estimated by fitting to steady state GFP fluorescence data and Specific Growth Rates ( Figure 3B, left) , using the error function |∆ log x|/log x.  Four of the five RBSes were fitted at a time. For each fitting, average error of the fitted RBS set as well as that of the excluded RBS were reported for cross-validation.