Accelerating metabolic engineering with designer biosensors

Published: April 19, 2016

Harvard scientists have developed a method for creating custom-designed biosensors that can dramatically accelerate the metabolic engineering of organisms for high-yield speciality chemical production. Dr Cynthia Challener reports

Metabolic engineering has led to the development of many different high-producing yeast and bacteria strains that are or will be applied to the commercial production of speciality chemicals. With traditional approaches, however, that process takes years to determine the necessary mutations that provide the desired performance. Researchers at Harvard’s Wyss Institute for Biologically Inspired Engineering and Harvard Medical School (HMS) have tackled this problem with the development of new biosensor technology that can be used to screen millions of strains rapidly for target compounds to identify the most productive variants.1

The biosensor technology is based on the use of allosteric transcription factors (aTFs), regulatory proteins in bacteria that bind specific target molecules and then undergo conformational changes that alter their affinity for DNA sequences which influence metabolic activity, thus effecting environmental responses. In synthetic biology, for instance, aTFs are used as gene expression switches.

The challenge with using aTFs for the sensor technology was the need to modify their binding preference for other target molecules without disrupting their response behaviour, or allostery (the process by which the proteins transmit the effect of binding at one site to another, typically to allow for regulation of activity), according to Srivatsan Raman, formerly a post-doctoral fellow at the Wyss Institute and HMS and currently assistant professor of biochemistry at University of Wisconsin-Madison.

“In order to function as a biosensor, the protein must be able to bind the target compound and respond in some way, but in general it is very difficult to maintain allostery when modifying the ligand binding interactions of aTFs,” he explains.

Figure 1 – Structures of the natural ligand (IPTG) and target molecules for LacI

For their initial work, the researchers chose LacI, a natural regulatory protein from E. coli that binds allolactose, but also responds to isopropyl β-D-1-thiogalactopyranoside (IPTG). Four alternative target molecules that differ in size (slightly smaller and larger, and twice the size) and with other functional groups (nucleophilic chlorine substituents) were selected as test compounds: fucose, gentiobiose, lactitol and sucralose, respectively, none of which can be metabolised by natural E. coli (Figure 1).

The first step towards development of the biosensors involved computational protein design. At this stage, only changes to the binding site were considered and the impact of the predicted binding interactions on allostery was not factored in. As a result, even the highest-scoring predicted mutants showed broken allostery, so no response behaviour could be determined.

At this point, a high-throughput method was employed to build a variant library of potential new biosensor designs based on computationally-designed LacI proteins as well as single-amino-acid saturation mutagenesis via multiplex gene assembly.

A high-throughput in vivo selection-screening method was designed to rapidly detect aTF variants that are both allosterically functional and bind to the target compound, according to Noah Taylor, a recent PhD graduate of Harvard and currently sensor engineering lead for Enevolv, a new company co-founded by senior researcher George Church, a Wyss Institute core faculty member.

To identify the variants with the most specific responses to the four target molecules of interest, the team engineered groups of E. coli bacteria to express green fluorescent protein (GFP) when the desired molecule was detected. High-throughput in vivo screening allowed identification of the most effective variants, based on their high fluorescence.

DARPA funded the sensor technology project

These proteins were filtered out and genetically sequenced to reveal their mutations. Whilst the proteins did bind to the four target molecules, they were not selective for them; many also retained the ability to bind to IPTG, which would make them ineffective as biosensors for the target compounds.

“At this point we developed a ‘maturation step’ to increase the selectivity of the variants that exhibited binding for the four target saccharides. This step involved combining the various mutations that were found to result in weak to strong binding for each different molecule,” Raman observes. The result was engineered proteins with high selectivities for the target sugars.

The process was developed over three years. The researchers now believe the entire process can be done in approximately two months: two weeks to complete the initial calculations, a couple of weeks for delivery of the necessary DNA oligonucleotides from a vendor, a few days for cloning and the remainder for the flow cytometry study to identify hit variants and incorporate the multiple mutations to obtain a highly selective biosensor.

Because  biosensors not only indicate the presence of a specific compound, but also its quantity, they can be used to accelerate metabolic engineering dramatically, according to Taylor.“Custom-designed aTFs can be introduced into metabolically engineered microbes to monitor the production of the desired chemical, which could be a pharmaceutical or agrochemical intermediate, other fine or specialty chemicals or biofuels. The proteins can even be engineered to lead to cell death if a specific compound is lacking, which eliminates the undesirable strains from the culture,” he notes.

Figure 2 – Enevolv sensor engineering platform

Enevolv is commercialising the technology (Figure 2), which it refers to as intracellular sensors, to rapidly screen billions of strain variants generated using multiplex automated genome engineering (MAGE). MAGE enables massively parallel whole genome editing to take place directly in living cells through the introduction of diverse combinations of specific modifications, rapidly building vast numbers of genomic designs.

“As a result, we are able to accelerate the genome engineering process by orders of magnitude,” Taylor says. Enevolv has received funding from the US Department of Defense Advanced Research Projects Agency (DARPA) to accelerate development of the sensor technology and is in discussions with numerous companies that are interested in having custom-designed biosensors or strains developed.

Importantly, the LacI protein used in the initial development effort is only one of thousands of different allosteric transcription factors that exist in nature. “Given that the LacI protein was engineered using only sequence and structure information, the researchers expect this approach to be applicable to the engineering of other aTFs for the creation of hundreds if not thousands of biosensors for specific target molecules,” he adds.

Indeed, the researchers have already demonstrated that biosensors based on TetR aTFs from the TetOn-Off system, which are used in a number of expression systems for chemical production, can be rapidly engineered. “The TetR protein family consists of a large number of proteins with significant variation, large numbers of which have been well-characterised. With the ability to engineer these proteins, we have the potential to choose a much wider variety of target compounds,” states Raman, who is also a scientific advisor to Enevolv.

For instance, he notes that TetR controls the membrane pumps in cells and regulates the expression of antibiotics. His group is currently investigating the ability to engineer one specific TetR protein that targets erythromycin to detect not only similar polycyclic aromatic compounds, but compounds with very different structures, including straight-chain alkanes.

Figure 3 – Pathway engineering using biosensors

“The idea is to push the envelope as far as we can to determine the limits to protein design – how much change can the protein tolerate and still have not only binding but also allostery in order to provide the desired response,” Raman observes. He welcomes any suggestions for target fine and speciality chemicals with commercial relevance that would be desirable to produce via fermentation.

One unexpected finding has been the malleability of allosteric transcription factors. “The proteins we have tested to date have shown exceptional amenability to mutations, which is a huge advantage, because it suggests that we may be able to go fairly far from the chemical space of the natural target,” Taylor comments.

“At Enevolv, we’ve engineered TetR to sense and respond to several molecules vastly different than its cognate ligand, well beyond the difficulty of what we achieved with LacI and new sugar derivatives.” As a reason for this malleability, he points to the fact that in cells, aTFs often evolve new molecular responses as conditions change, such as the appearance of a new molecule.

Biosensors have also been developed using allosteric transcription factors from the LysR-type transcriptional regulator family. “This advance represents a powerful new way for us to access the chemical diversity of the biosphere by mining for new pathways to make useful molecules,” says Raman. “These results suggest that it should be possible to engineer optimised aTF sensors for practically any molecule, opening new doors in synthetic biology by putting allosteric proteins in the control of genetic engineers.”

In addition to accelerating metabolic engineering, potential applications of the biosensors include environmental monitoring, medical diagnostics, bioremediation and precision gene therapies. The biosensors may also make it possible for scientists to better investigate the metabolic state in individual cells, a task that has been challenging to date. Custom-engineered aTFs inserted into a live cell could detect the presence and quantities of different metabolites as a function of time, providing near real-time information on its metabolic state, according to Taylor.



1. N.D. Taylor et al., Nature Methods 2016, 13, 177. DOI: 10.1038/nmeth.3696.



Enevolv, Inc.

Noah Taylor

Sensor Engineering Lead

Tel: +1 617 817 5315

E-mail: n.taylor@enevolv.com

Website: www.enevolv.com


University of Wisconson-Madison

Srivatsan Raman

Assistant Professor of Biochemistry and Bacteriology

Tel: +1 608 890 1036

E-mail: sraman4@wisc.edu

Website: www.ramanlab.org



Copyright © 2017 Mack Brooks Speciality Publishing.