Marionette strains aim to make refining metabolic pathways faster and easier.

The Voigt lab at The Massachusetts Institute of Technology recently reported the optimization of 12 transcription factor sensors and their integration into a series of E. coli strains, collectively referred to as the Marionette strains (1). While transcription factors that can be controlled with small molecules are not new (the lactose and tetracycline inducible systems have been around for decades), synthetic biologists are limited to using only a few simultaneously in a cell due to issues such as unintended activation by cognate or non-cognate molecules, cross-interactions with the other sensors, and ‘leakiness’, or low-level transcription in the absence of the activating molecule. Previously, the maximum number of transcription factor sensors used simultaneously in a cell was four. The Marionette strains could make optimizing heterologous metabolic pathways faster and easier by expanding the design space for complex genetic circuits. The Voigt lab used directed evolution to find variants of sensors with high specificity, low leakiness and greater activation ranges (in other words the difference between the ‘off’ and ‘on’ states). A dual selection system was designed: a negativeselection marker (the promiscuous PheS allele) would kill cells with sensor variants that had leaky expression or were reactive with unintended small molecules, while positive selection to find highly sensitive and specific variants was achieved by the expression of a DNA polymerase (DNAP) and subsequent emulsified polymerase chain reaction. The emulsification isolates the sensors, and those responsible for the highest levels of expression are amplified more strongly. In addition, the DNAP used in the positive selection could be alternated between a stringent or error-prone polymerase in order to produce and control the diversity on which the selections would act, and these mutagenized sequences would be incorporated into the next round of selections. To demonstrate the utility of a Marionette strain in tuning the expression of a metabolic pathway, the lycopene synthesis pathway was used as a model (2). The five genes in the lycopene synthesis pathway were placed under the control of five different sensors, and three levels (zero, maximum and 50% maximum) of each inducer were added to the culture resulting in 243 combinations. Lycopene concentrations were measured for all 243 combinations, and new minima, maxima and midpoints were derived for each of the five transcription factors. The experiment was repeated this way for a total of four iterations, resulting in a maximum lycopene titer of 90 mg/l. To equal the design space explored in the Marionette lycopene optimization example (243 combinations screened four times), you would have to synthesize 972 constructs. Synthesizing these 972 constructs (7 Mb total), would cost $700 000 (assuming $0.10/base), not to mention the enzyme and labor costs to clone all of those variants, the lost time due to human error, unforeseen issues with the sequences, etc. In contrast, using the Marionette strain requires the production of only one genetic construct and the purchase of the small molecule inducers, meaning that previously infeasible explorations of design space are now open to investigators. Advancements in technology precede scientific breakthroughs, as they provide new avenues of inquiry previously unavailable to scientists. With the ability to act as a ‘metabolic breadboard’, Marionette strains can rapidly and inexpensively tune heterologous pathways in vivo providing synthetic biologists the capability to refine and optimize genetic circuits installed in organisms for a variety of purposes.