To produce the most desirable product on the market, beverage companies use multivariate testing. They test multiple variants of their original product by changing variables such as flavor, packaging, and branding through the use of focus groups and surveys. Using this feedback, these companies can optimize their product and improve its marketability. Bioengineers perform a similar process when it comes to optimizing proteins.
“Anybody can make a lousy enzyme, but what makes a good enzyme? If we knew that then we could rationally design one,” says Frances Arnold, professor of chemical engineering and biochemistry at the California Institute of Technology, where she has been experimenting with directed evolution techniques for over twenty years. “It’s wonderful that we have a method that circumvents that near total ignorance.”
Directed evolution is improving rationally designed proteins to solve practical problems, such as treating a disease or developing a biofuel, while providing more insight into how proteins work.
The marriage of rational design and directed evolution
“One of the opportunities for directed evolution is to optimize computationally designed enzymes or other primitive catalysts,” says Donald Hilvert, professor of organic chemistry and lab head at the ETH Zurich in Switzerland. “Many reactions that one would like to catalyze in an industrial or a medical setting might not be available, but new tools for designing enzymes are improving.”
Rational protein design experiments are limited not by the technical challenges but by a lack of knowledge. Researchers cannot predict how single changes in the amino acid sequence will affect a protein’s performance or how changes that do enhance function will affect the protein’s other characteristics.
Rational design has allowed researchers to get in the ballpark of the desired functionality and design proteins with the potential to behave as they would like. But these designs always need refinement.
“You can either monkey with it rationally for 24 years and likely not get anywhere. Or you can just sit down and rationally optimize through evolution,” says Arnold. “I would prefer to rationally optimize through evolution.”
Using directed evolution, researchers can look at many more samples through a single round of mutagenesis than with step-by-step rational design optimization. By generating random mutations in a high-throughput approach, researchers can cover more sequence space faster and more efficiently.
Rational design and directed evolution are complementary approaches, and they are increasingly used in combination for protein engineering: rational design cannot produce good designs without directed evolution, and directed evolution needs a starting point. “It’s a marriage of these methods that gives us the ability to manipulate biology in a useful way,” says Arnold.
For twenty years, Arnold has been evolving proteins to make better industrial enzymes, exploring the range of traits that can be engineered into enzymes by evolution. Her group was the first to apply sequential rounds of mutation and screening to enzymes. They used error-prone PCR in their first experiments to introduce random mutations, screened to find the beneficial mutations, and repeated that process.
“It turned out to be a remarkably powerful process, and it’s still actually the preferred process,” says Arnold. “It’s not necessarily the most efficient; it’s also not the most inefficient way.”
Arnold is interested in using her experience with directed evolution to solve big industrial enzyme problems.“It’s all very well to write protein engineering papers,” she says. “I do my share of those. But I also try to make proteins that will solve actual problems.”
In 2005, Arnold founded Gevo, Inc, a biofuel company that is developing isobutanol from biomass. This year the company has filed to go public, has purchased its first commercial production facilities, and has a potential commitment from United Airlines for an isobutanol-based jet fuel. A key part of this is to optimize biosynthetic pathways.
“We used rational design to pull together a biomolecular pathway,” explains Arnold. “But we use evolution to optimize them. And that can make the difference between being successful and not.”
Despite millions of dollars of investment into cellulose research, scientists have not yet been able to produce cheap cellulose. One reason is that an entire consortium of enzymes contributes to the chemical reaction that breaks down biomass, and those enzymes are hard to improve.
Arnold decided to take on this challenge. “You can’t just evolve one and hope that that’s going to make a better mixture,” says Arnold. “So how do you evolve a population of enzymes and how do you screen in a 96-well plate for whether those enzymes allow you to make more ethanol in a plant? How do you translate which properties really matter?”
Andreas Plückthun, professor of biochemistry at the University of Zurich, is attempting to find the answers to such questions. For the past twenty years, his team has used directed evolution approaches to develop recombinant antibodies, and then small high-affinity proteins, such as designed ankyrin repeat proteins (DARPins), that bind to overexpressed tumor markers.
In 1997, Plückthun’s laboratory developed the ribosomal display technique, which enabled directed evolution to be done completely in vitro. In ribosomal display, the protein and its encoding mRNA can be optimized while attached to the ribosome. The technique frees directed evolution from cell-based systems, allowing the process to occur more quickly and with more generations (1).
“It’s really a method that is happening completely in vitro and can be interfaced with random mutagenesis very easily,” says Plückthun. “In the ribosome display, you are making mutants while you go along and you evolve the population.”
In 2004, Plückthun co-founded Molecular Partners AG (Zurich, Switzerland) to develop therapies using the DARPins. Earlier this year, the company’s first DARPin product began Phase I clinical trials.
Understanding unstable GPCRs
While directed evolution can optimize rationally designed proteins, it can also help build our knowledge about protein structure and function.
“You can do the directed evolution experiment to find the range of possible solutions, and thereby map out the binding sites, map out the stability determinants, and so forth,” says Plückthun. “This is information that couldn’t have been gained otherwise.”
Plückthun’s group has recently been focused on understanding the properties of G protein–coupled receptors (GPCRs), a family of transmembrane receptors that sense molecules outside the cell and activate inside signal transduction pathways. These receptors are involved in many diseases and are a target for many modern drugs. Structural studies of GPCR have been difficult because of the molecule’s instability; it tends to fall apart quickly when immobilized in detergent.
To study the structure of GPCRs, Plückthun’s lab has turned again to directed evolution. But unlike their earlier work improving the binding of tumor-targeting proteins, here they use directed evolution not to change the function of the molecule but to increase its stability. Mutated GPCRs still bind to the same ligands, but they are stable enough that structural biology experiments can be done on them.
“We do this by making a library of [mutants of] a given GPCR that are then expressed in the inner membrane of Escherichia coli, where they are functional,” says Plückthun. “We can pull out those mutants that are expressed at high levels and are more stable.”
With this technique, Plückthun’s team has been able to stabilize several GPCRs, and he is very hopeful that significant structural information can be obtained from these more robust versions of the receptor. “That has been surprisingly successful and turned out some very interesting things,” says Plückthun, who is preparing several papers detailing his group’s findings.
At the Laboratory of Organic Chemistry at the ETH Zurich, group leader Peter Kast is also using directed evolution to elucidate enzyme structure and function. Kast’s group is interested in chorismate mutase (CM), an enzyme that accelerates the reaction of chorismate to prephenate by million fold. CM is necessary in fungi, bacteria, and plant cells to balance the aromatic amino acids in the shikimate pathway.
To understand how CM functions in the cell, Kast’s group randomizes specific CM-coding sequences that they think are interesting based on homology or available structural information (2). They then select the variants that can grow in knockout strains, and sequence the gene to develop patterns compatible with the enzyme’s function.
“From these patterns, we can deduce essential features of the catalytic mechanisms of these enzymes,” explains Kast.
Kast’s group has discovered a surprising CM variant that is secreted by many pathogens. “It doesn’t really make sense to secrete the CM by the bacteria into the host tissue,” he says. “If the host is mammal, this host has no use for the prephenate that is being produced in the CM reaction, nor does it have the substrate chorismate mutase. This is still a puzzle.”
Limiting the variants
While some techniques within the directed evolution approach have matured, such as the creation of variant gene libraries and randomization, technical challenges still remain.
“The hard part of directed evolution is screening: it’s finding what you want from the zillions of enzymes that you could possibly make,” says Arnold. Because the sequence space is so large, it’s impossible to create all possible mutants, even for small 100–amino acid proteins. So techniques to reduce the set of residues to randomize would help.
“A big challenge is the selection of which parts of the gene to randomize,” says Kast. “We have had an approach to use binary patterning to reduce the diversity in specific regions of the protein. Another way to do that would be to recognize patterns of useful combinations of residues to randomize to restrict the sequence space that you must look at.“
Another issue for Kast has been the tuning of the selection process. Since he is interested in metabolic functions of proteins, his selection process can be coupled to cell growth. But once the protein has evolved to the point where it is sufficient for cell survival, the selection process becomes more difficult. “It is very hard to go beyond that point,” says Kast. “If you would like to evolve that function further, you would need to increase the stringency of the system.” His team has been working on a number of ways to fine-tune their selection system by lowering the concentration of the catalyst to be evolved or by introducing a competing enzyme to make the catalyst work harder.
But the technique itself may not be the major question in these studies anymore. The real question is, what protein properties are desired from a screen? “It’s very hard to do that. It’s very hard to screen for whether a protein cures a disease or whether an enzyme improves profits,” says Arnold. “That’s what matters, whether your celluloses are going to let you make more ethanol. We’re not sure how to screen for that.”
While there are certain characteristics that researchers can look for, it has always been a challenge to make meaningful assessments after multiple rounds of screening on which protein is better in the end.
“I think that the methods are there that allow one to explore sequence space reasonably efficiently,” says Hilvert, who thinks this brings up new questions for directed evolution studies. He wonders whether we are exploring the right parts of sequence space.
“The question is, how amenable are these starting points to future evolution?” he says. “Will they ever reach the levels of true enzymes? That’s unknown.”
As rational design continues to improve and provide more insights into protein function, directed evolution is allowing researchers to discover beneficial enzymes right now.
“We’ll never be able to type up the sequence of a new enzyme that catalyzes a reaction in industrially useful ways,” says Arnold. “It’s a huge goal, but it’s not going to happen any time soon because details matter. And we don’t understand the details.”
- Hanes, J. and A. Plückthun. 1997. In vitro selection and evolution of functional proteins by using ribosome display. Proc. Natl. Acad. Sci. 94: 4937–42.
- Roderer, K. and P. Kast. 2009. Evolutionary cycles for pericyclic reactions – or why we keep mutating mutases. CHIMIA 63:313–317.