Engineering Cell-based Factories

Some experts predict that biotechnology will produce 20% of industrial chemicals by 2010. Advances in basic knowledge and technology are moving biotechnology in this direction; however, the challenges of commercialization are complex and formidable.

The world's most ubiquitous and useful commodities spring from limited--and limiting--wells. More than 90% of the chemicals used in industry are made from petroleum products. Numerous drugs are derived from rare or endangered plants or organisms. Some chemicals are generated by processes so hazardous that the compounds aren't even made in the U.S.

It's long been presumed that the future of the chemical and pharmaceutical industries lies largely with microbes. With help from biotechnology, common bugs like the bacterium Escherichia coli and the yeast Saccharomyces cerevisiae can perform complicated syntheses, producing drug precursors from glucose, ethanol from waste biomass, or industrially important chemicals from "green" starting materials.

At the outset, this strategy might appear to involve mostly garden-variety genetic engineering: Overexpress a few key genes in a cell or snip out genes from a separate organism and insert them into a cell, which then begins churning out antibiotics or food flavorings. The actual picture is far more complex. Boosting production of one compound or enzyme in a cell may overwhelm other aspects of its metabolism and shut down production. Or the cell may respond by introducing a new, unwanted pathway. And some products are toxic to the cell itself.

Pulling industrially useful quantities of products out of cells, while keeping them happy, requires navigating, altering, and understanding the complicated relationships between the cells' metabolic pathways and their products. In just the past decade, the area now known as metabolic engineering has evolved into a distinct field that uses a panoply of modern biotechnological tools: genetic manipulation, high-throughput methods, directed evolution, mathematical modeling, and analysis of transcriptomes--that is, the levels of all of the mRNA products in a cell. Some compounds, like certain amino acids, are already produced industrially by fermentation; researchers hope to add many more products to their ranks, from supercommodities like ethanol to fine chemicals.

Some facilities using metabolically engineered bugs are already poised to go into large-scale production. For example, DuPont, along with fermentation experts Tate & Lyle, is building a plant that uses engineered E. coli to produce 1,3-propanediol, a key component of DuPont's Sonora brand fabric polymer. Working with BC International, Japan-based TSK-Marubeni is constructing a facility that will use an E. coli variant to produce ethanol from waste wood chips.

WHAT'S IN A NAME?
One Person's Synthetic Biology Is Another's Metabolic Engineering
In the past several years, a new biotech buzz phrase has entered mainstream consciousness: synthetic biology. It's a nascent field that has undertaken the mission of designing, learning to control, and building organisms to do new tasks and produce new products.In a lot of ways, it sounds remarkably like metabolic engineering.For example, Jay D. Keasling is a chemical engineering professor at the University of California, Berkeley, who recently made headlines when he engineered Escherichia coli to produce a precursor to an antimalarial drug. At the same time, he heads a new synthetic biology department at Lawrence Berkeley National Laboratory. Such involvement in what apparently are distinct--but in fact are very similar--areas of scientific inquiry is causing consternation among some scientists who maintain that "synthetic biology" is a slick, provocative new name slapped onto an already established field."So as far as I can tell, synthetic biology is very similar, if not identical, to what we call metabolic engineering. It adds a layer of confusion," says Douglas C. Cameron, director of biotechnology at Cargill Research in Minneapolis."If people want to call it that, that's fine," says John W. Frost, chemistry professor at Michigan State University. Consider, however, Louis Pasteur's manipulation of yeast to make acetic acid back in the 1800s, he points out. "Is that synthetic biology? Yes, but it's not new."There is a difference between the two fields, says Drew Endy, a young Massachusetts Institute of Technology biological engineering professor who is considered the father of synthetic biology. "At the moment, most of the folks working in synthetic biology are heavily focused on building the basic infrastructure that would make the engineering of biological systems simpler, more predictable, and more widely accessible," he says. His group is using engineering and computer techniques to design and precisely regulate cell behavior.By contrast, metabolic engineering "has typically required an ad hoc research process, with significant uncertainty in the cost, timeline, and chance of project success," Endy says. "I'm in both camps," says James C. Liao, a chemical engineering professor at the University of California, Los Angeles, whose research includes the design of cell control circuits. Echoing Endy, he notes that metabolic engineering is perhaps more concerned about the practical application, while synthetic biology focuses more on fundamental science. Ultimately, he says, "the spirits are pretty similar."

YET ROUTINE industrial chemical production from metabolically engineered organisms is still a ways off, partly because of the monetary commitment required to introduce new production technology. "The biggest hurdle is financing the first-of-a-kind plant," notes Lonnie O. Ingram, microbiology and cell science professor and director of the Florida Center for Renewable Chemicals & Fuels at the University of Florida, Gainesville. "Lots of people want to build plants two through 100, but nobody wants to build the first one."

Along with improvements in biotech methods and sequencing of microbial genomes are coming a wealth of new advances. Scientists can now begin to generate pictures of the cogs in a cell's machinery, and models can provide practical solutions to real-world problems.

Despite the futuristic ring of "metabolic engineering," its roots are in the age-old practice of fermentation, harking back thousands of years to when humans brewed their first beers. Louis Pasteur's discovery in the 1800s that Acetobacter bacteria convert alcohol to acetic acid marked the start of the knowing, deliberate efforts to influence microbes to produce substances for us.

In the modern industrial age, the chemistry of petroleum has driven the production of synthetic organic compounds. Some say the long-pondered power of microbes is finally poised to take center stage.

"Almost from the day we started cloning, people started thinking about what we now call metabolic engineering," says Douglas C. Cameron, director of the Biotechnology Development Center at Cargill Research in Minneapolis. Citing early metabolic engineering efforts by companies like Amgen, Genentech, and Cetus, he notes that, during those years, scientists were forced to rely largely on speculation and trial and error to optimize genetic modifications. "It was a very qualitative effort," he says.

At around the same time, chemical engineers and mathematicians were developing esoteric models of cells and computational programs to study metabolic fluxes. With the rise of genomic data, those models could be put to the test, and what were two relatively separate arenas--practical and theoretical--are now a blended discipline encompassing everything from mathematical flux analysis to gene shuffling, a perspective echoed last September in Lake Tahoe, Calif., where scientists and engineers gathered for the field's fifth biennial metabolic engineering conference. "In the last five years, that convergence has really become the way the field operates," says Cameron, who co-organized the meeting with James C. Liao, a chemical engineering professor at the University of California, Los Angeles, and Joseph J. Heijnen, a professor in the bioprocess technology department of the Technical University of Delft, in the Netherlands.

"The ultimate goal in academia is to predict cell behavior," Liao says. "Because of advances in genetic manipulation, detailed mathematical analysis has become feasible and important."

THE TASKS AHEAD to advance both theory and practice are formidable. Far from a simple connection of gene expression to enzyme to product, metabolic pathways are intimately interconnected through, for example, use of common cofactors such as ATP and NADH. As Jens Nielsen, director of the Center for Microbial Biotechnology at the Technical University of Denmark, in Lyngby, points out, 86 metabolites are involved in more than 10 reactions of S. cerevisiae.

It's not enough simply to add genes for the product you want toward the end of the metabolic line. When a new pathway is imposed on a cell, the cell suddenly experiences an unnatural drain on precursors, Nielsen explains, likening the effect to struggling with a knotted rope. "If you try to unwind it by pulling in one place, then you make one of the knots even tighter," he says.

Many groups are focusing on the best way to analyze, model, and predict the fluxes of products and their precursors that make up metabolic pathways. In the past few years, high-throughput studies of whole-genome sequences have made possible the mapping of transcriptomes. Because transcriptome levels vary with levels of gene expression, some people have presumed that transcriptomes provide a map of what genes are activated and, thus, the fluxes of metabolites.

That has turned out not to be the case, Nielsen argues. "Relatively rarely is there a [straightforward] correlation between transcription data and fluxes," he says. To get at the relationships hidden in the data, Nielsen's group is developing algorithms that combine gene expression data with metabolic models of whole genomes, allowing them to spot potentially important clusters of key "reporter" metabolites.

Large amounts of genomic data are key to improving models, Liao also noted at the meeting. "Currently, we don't have a unified theory to explain complex biology. Either you model a cell with lots of parameters, but it's not very predictive; or you use fewer parameters that everybody agrees on, and most of the predicted results are pretty trivial. It's a dilemma." Liao's strategy, called network component analysis, involves coupling computational techniques to DNA microarray data to deduce the activity of each transcription factor, and from that information, to tease out regulatory information.

The well-explored techniques of genetic manipulation aren't quite so straightforward either. For example, it might seem that directed evolution, the popular high-throughput technology in which enzymes are iteratively mutated to evolve strains that perform new or improved tasks, might be a dominant technique. By contrast, a cellular pathway is not as simple to evolve, explains Gregory Stephanopoulos, a chemical engineering professor at Massachusetts Institute of Technology.

Although many parallels exist between directed evolution in proteins and directed evolution in pathway modification, cells subjected to successive mutations in hopes of improving on an already-promising change might not continue down that path. "The landscape is very nonlinear, with sharp peaks and valleys," Stephanopoulos says. "And perhaps an initial mutation seems to be beneficial, but it eventually does not lead to a global maximum. That's a lot more likely to happen with cells than with proteins."

This effect highlights the need to examine whole systems.

In addition to internal cues, cells also pay attention to the cues they receive from their neighbors, which can have profound effects on what products cells eventually form. As Liao explains, in some cases a cell doesn't produce anything unless it gets to some common stage, and then all cells turn on their product formation pathways and simultaneously convert substrates to products. It's a strategy used by pathogenic bacteria, which often sit quietly until the environment is favorable and then launch a communal attack.

Liao and his group members reason that they could control product formation by creating an artificial cell communication system. For example, they engineered E. coli to produce the light-emitting protein GFP in the presence of acetate, a low-level signal that the bacteria constantly secrete. Bacteria emitting constant light signaled that the process was working, illustrating the potential for using this method.

In the past decade, examples of successful metabolic engineering have been piling up. Ingram's early work with ethanol-producing E. coli in the 1980s was one of the field's first major successes. Although production of ethanol from yeast fermentation is a millennium-old technique, the kinds of sugars yeast requires to do that are limited. On the other hand, E. coli excels at breaking down five- and six-carbon sugars, as well as agricultural waste such as hemicellulose, into pyruvic acid, an intermediate. E. coli, however, turns pyruvic acid into products that aren't that useful--succinic acid, lactic acid, acetic acid, and very little ethanol. Meanwhile, the bacterium Zymomonas mobilis excels at turning pyruvate into ethanol. So Ingram's group inserted key genes from the Z. mobilis pyruvate pathway into E. coli, resulting in a bug that eats agricultural waste sugar products such as leaves and even wood chips and produces ethanol for fuel.

Other groups, including those led by Nancy Ho, a senior research scientist and the leader of the molecular genetics group in the Laboratory of Renewable Resources Engineering at Purdue University and at the National Renewable Energy Laboratory, have engineered ethanol-producing E. coli as well. The potential, Ingram says, is huge. "If we took cobs and leaves and stems, we could double the amount of ethanol--4 billion gal in the U.S. per year--from the same plant," Ingram says.

METABOLIC ENGINEERING can also provide options to natural products that are scarce or difficult to obtain, particularly those from plant sources. An example is shikimic acid, a key precursor to the antiflu drug Tamiflu. Shikimic acid is naturally isolated from the poison star anise, a plant found in China. Shikimic acid is also an intermediate in a common pathway of aromatic amino acid synthesis in E. coli. In a widely publicized effort, the group of Michigan State University chemistry professor John W. Frost figured out how to channel a "huge amount" of carbon flow into the pathway and direct the organism to produce and export shikimic acid, Frost says. With this method of making shikimic acid, "you're no longer tied to a growing season or a drought or deluge in China," he says.

Frost's prolific lab has also developed an engineered microbe that makes gallic acid, a potent antioxidant traditionally isolated from gall nuts or the carapace of certain wasp species. Often, the end product is toxic to the cell. One solution, Frost elaborates, is to synthesize the key intermediates inside the cell and then do simple organic synthesis in the lab to get the final product. The Frost lab has demonstrated this strategy with catechol, an industrially important compound usually produced from benzene. Although E. coli doesn't do well in a bath of catechol, Frost's lab has engineered the microbe to produce the harmless intermediate protocatechuate from glucose; the isolated protocatechuate is then readily decarboxylated in water, producing catechol. Frost's group has also engineered E. coli to produce the rocket-fuel precursor 1,2,4-butanetriol (C&EN, May 31, 2004, page 31). The precursor is currently made only in China and India from sodium borohydrate and strong acids.

The powerful potential of metabolic engineering has turned the heads of wealthy grant givers. The Bill & Melinda Gates Foundation recently awarded $42.6 million to the Institute for OneWorld Health, which will work in part with Jay D. Keasling, a chemical engineering professor at the University of California, Berkeley, and head of Lawrence Berkeley National Laboratory's synthetic biology department, to develop an organism that produces the antimalarial drug artemisinin (C&EN, Dec. 20, 2004, page 14). Keasling's group has focused on isoprenoid molecules, including terpenoid drugs such as artemisinin. The usual process of extracting artemisinin from the sweet wormwood tree is hazardous and difficult. Now, Keasling's group has created an entirely new pathway in E. coli by inserting genes from yeast that synthesize a key precursor, amorphadiene. The group's goal is to perform the complete artimisinin synthesis in E. coli.

Metabolic engineering can also benefit already-established industrial fermentation, such as that used to produce lysine, noted Oskar Zelder, a scientist at BASF's biotechnology research center in Ludwigshafen, Germany, at the meeting. In the 1990s, scientists sequenced the genome for Corynebacterium glutamicum, which produces lysine. Since then, Zelder's group has worked to improve lysine production, experimenting with different sugar substrates such as sucrose, fructose, and glucose, and using the genomic data to study the cell's metabolic fluxes. The results are improved strains with a 12% increase in lysine yield.

The carotenoids, such as lycopene, are highly sought-after antioxidants that are beneficial to human health. Although carotenoids usually are extracted from plants, Stephanopolous' group is developing carotenoid production in E. coli, using stoichiometric models to identify key genes, along with random mutation and high-throughput searches for cells producing high amounts of lycopene. The result, which he presented at the Tahoe meeting, is a lycopene-producing strain of E. coli that produces 6,800 ppm, a large improvement over the wild type's production level of 500 ppm. Ultimately, they were able to construct E. coli mutants capable of amassing over 20,000 ppm of lycopene in feed batch cultivation--an important step toward industrial feasibility.

Only a handful of organisms have been exhaustively characterized, and metabolic engineers rely heavily on two of these workhorses, E. coli and S. cerevisiae. E. coli, in particular, is a predictable microbe, Ingram says. "It's the best known and best understood living organism." They each have different strengths, depending on the product of interest. For example, yeast is more tolerant of low-pH environments and so might be a good choice for production of organic acids.

Other, less well-known organisms may be an even better choice for some applications, Liao notes. For example, Aspergillus niger is good for making enzymes. That leads to a current question in metabolic engineering: "Do you start with a handful of cells that are well-understood and modify them to make a desired product, or do you start with less well-understood cells that make what you want?" Liao muses. "That's a big dilemma in industry."

Fortunately, as the whole genomes of more and more organisms become available, those organisms become possible metabolic engineering platforms. For example, the recent sequencing of Rhodopseudomonas palustris, a member of the extremely metabolically versatile nonsulfur, purple, photosynthetic bacteria, could lead to numerous new applications, notes F. Robert Tabita, microbiology professor and director of the plant microbe genomics facility at Ohio State University. His group is using this microbe as a model for studying processes such as hydrogen production and organic-carbon degradation.

As the new tricks being taught to microbes accumulate, the question ultimately becomes whether these strategies can be scaled up for industry: They have to be better and cheaper than already-existing processes, scientists said at the meeting. That standard can be a deterrent for changing some well-established processes, they said. To be worthwhile, biotech processes need to reduce cost by 30%, making it difficult to introduce a new technology for existing industrial problems.

Cameron agrees, "The absolute major hurdle is doing it faster and cheaper."