Rapid, modular, on-demand, cell-free protein synthesis. The process of mixing cellular lysates with substrates for transcription and translation to make proteins in a test tube. This is the technology we had when we first set out to develop an in vitro platform for studying biosynthetic enzymes to speed up research and development cycles for industrial biotechnology. Now, in vitro systems to study enzyme activity have been around for more than a century. But, almost all involve protein expression in a host organism prior to cell lysis and often purification which is time-consuming. Not to mention selecting productive enzymes using existing single-enzyme kinetic data has limited applicability in multi-enzyme pathways and consequently requires more engineering iterations.

We had a vision early on that we could use cell-free protein synthesis, a tool the Jewett Lab at Northwestern University had extensive expertise in, to produce enzymes directly in vitro, bypassing the time and effort to construct and test enzymatic pathways in cells. The key idea is that this new pace of research—from weeks to days—would allow us to test significantly more enzymes and variants thereof as well as pathway designs toward a desired biochemical product that we might not otherwise test using traditional methods.

Our first proof-of-concept demonstrated that we could indeed use cell-free protein synthesis to construct a biosynthetic pathway from linear DNA templates. In one-pot we expressed five enzymes to enable E. coli-based lysates to produce butanol from glucose. From there we’ve also been able to show this approach can be used for a number of other biochemicals.

It was this proof-of-concept along with the close-proximity of our campuses in the greater Chicago area that we became interested in a partnership with LanzaTech. LanzaTech’s unique approach to industrial biotechnology using the anaerobic acetogen, Clostridium autoethanogenum, allows for almost any C1 carbon waste product to be efficiently converted to valuable product. The problem at the forefront was that creating prototypes of new biosynthetic routes in this organism is time-consuming, taking months to construct a handful of designs. This was the precise problem we wanted to address.

Only through our conversations with LanzaTech were we able to improve our proof-of-concept to what is presented in our most recent publication in Nature Chemical Biology. Our in vitro Prototyping and Rapid Optimization of Biosynthetic Enzymes (iPROBE) approach is the process of designing hypotheses for a biosynthesis, building those hypotheses through a modular, two-pot cell-free construction, testing those prototypes, and applying or informing better design in cells. However, we struggled with figuring out how to bridge the gap between cell-free results and cellular performance, especially when we layer on the fact that the cell-free system is based on E. coli not C. autoethanogenum.

In our manuscript we step through the development of iPROBE, how we wanted to establish a starting point for the comparison of cell-free and cellular systems that did not exist before. We show an approximation of how genetic architectures for cellular plasmid design could be decided upon by observing protein levels and results in vitro. Though, a key question we had to address before starting any extensive prototyping efforts was do we calculate end-point titers or reaction productivities in vitro for comparison with in vivo results?

We didn’t have an answer to this question because we didn’t know if either would correlate with our in vivo results. The reality is the systems are different and we needed to account for this uncertainty in some way. Thus, the TREE score was developed, a single-quantitative metric that combined the titer, rate, and enzyme expression through multiplication. Rather than extensively test, introduce weighting parameters, or perfect this metric we moved forward knowing that this was an important first step in establishing cell-free to cell correlations.

We used this score to prototype pathways for 3-hydroxybutyrate and butanol production. We showed that the TREE score could be used to rank cell-free constructed pathways and that in C. autoethanogenum high-ranking pathways performed well and low-ranking pathways performed poorly. This was an important validation that our approach was working. We could in fact use iPROBE to inform our designs for C.autoethanogenum without having to test every design in the living organism. The pathways selected by iPROBE actually performed so well that we decided to scale these pathways up and optimize fermentation conditions to show that C. autoethanogenum could make more 3-HB than had ever been reported.

Deciding to use iPROBE on longer pathways like for butanol production meant there would be more enzymes that could be tested and exponentially more combinations of them. We had to limit what we selected. We had started having conversations with Lockheed Martin on how we might be able to couple their computational approaches to materials design and characterization to synthetic biology applications. This was a perfect opportunity to explore machine-learning approaches for the prediction and selection of enzyme combinations to test in vitro from the ocean of possible combinations. We used a neural network-based model to predict pathway combinations in lieu of biochemical information a priori. We used these computational predictions to test a total of 205 biosynthetic pathways improving cell-free production over four-fold.

This work represents an important step for synthetic biology in accelerating research and development pipelines. We observed a high correlation between cell-free results and cellular performance for over 20 designs. While we would love to have more data points here, it is important to remember that transformation idiosyncrasies limit the field and genetic tools are still developing for non-model organisms.

Cell-free systems are becoming evermore prevalent and many are exploring how they can be utilized to understand biological systems and improve the world around us. We anticipate that iPROBE will make a lasting impact on the community as it is already impacting our workflows at LanzaTech and transforming research at the Northwestern Center for Synthetic Biology. We are also excited to see future developments by us and others to improve the capabilities of cell-free systems to transform industrial biotechnology.

Read the full article at Nature Chemical Biology:

Karim, A.S., Dudley, Q.M., Juminaga, A. et al. In vitro prototyping and rapid optimization of biosynthetic enzymes for cell design. Nat Chem Biol (2020). https://doi.org/10.1038/s41589-020-0559-0