You may have heard the proclamation before: The next generation of biofuels will be derived from cellulosic plant material. And, in theory, this makes sense. Whereas ethanol can be produced via the fermentation of simple sugars in food crops such as corn or sugarcane, it would be more economical to make fuel from nonfood sources that are cheaper and more abundant—such as switchgrass, Miscanthus or wood chips. The problem? Today's methods for breaking down cellulose, the fibrous complex sugar that is the main structural component in green plants, are too expensive.

To degrade the tough plant material, engineers use enzymes isolated from organisms, such as termites, that rely on the molecular machines to convert their cellulose-rich meals into simpler, digestible sugars. But the enzymes currently available are not efficient enough to make cellulose-to-fuel conversion worthwhile. "If the industry is going to move forward, it's going to need new enzymes," says Eddy Rubin, the director of the U.S. Department of Energy's Joint Genome Institute. Rubin and 16 colleagues report in the January 28 issue of Science how they discovered nearly 30,000 new enzyme candidates by analyzing DNA collected from a cow's rumen—the first compartment in the animal's four-section stomach and home to a vast population of microbes equipped with potent enzymes that help digest the grasses their bovine host consumes.

The researchers employed a cow with a surgically placed tube, called a fistula, which allowed them direct access to the rumen. As Rubin explains, "Over millions of years, in exchange for housing in the cows, these organisms have gotten good at paying their rent by providing the host with broken down cellulose—sugars the cow can use as an energy substrate."

To gather rumen microbes so as to analyze their genetic material, the group placed nylon bags filled with switchgrass, the much-hyped next-generation biofuel feedstock, into the cow's rumen via the fistula. Plant-digesting organisms then "glommed on" to the switchgrass, and after 72 hours "we would pull the whole bag of material out and extract the DNA that was adherent to it," Rubin explains. This experimental setup was innovative compared with traditional in vitro methods for isolating enzymes in which microbes are grown in a laboratory incubator—a process that wouldn't have worked for these microbes, Rubin notes. "They're quite happy in the belly of the cow but they are not so happy living in your incubator."

The group then isolated the collected DNA, which (because it came from hundreds of separate organisms) was like "a whole bunch of jigsaw puzzles," Rubin notes. Next, using high-throughput sequencing technology, the researchers "were able to assemble big pieces" from those puzzles.

In all, the project generated almost 270 billion DNA bases—nearly a hundredfold more than are contained in the entire human genome. With the help of extremely powerful computers, the group analyzed this massive pile of data by essentially scanning the puzzle pieces for sequences that resembled genes that code for previously documented "carbohydrate-active" enzymes. Using this analysis, the researchers identified 27,755 genes that were good enough matches to be viable candidates for application toward cellulosic biofuel production.

The researchers then selected 90 of the candidate genes, expressed them to produce the enzymes for which they code, and then applied these molecular machines to cellulosic biofuel feedstocks Miscanthus and switchgrass. More than half of this subset showed the capacity to degrade at least one of the feedstocks, which the study authors say suggests the larger pool of candidates is "highly enriched" with enzymes whose activity could be useful in biofuel production.

Finally, the group checked their computational results against the genome of an actual single-celled organism isolated from the cow's rumen. Although they could not culture the cell in the lab, they could sequence its genome, and the fact that it almost entirely matched up with genomic puzzle pieces they had previously put together was "sort of a proof that the individual genomes we were assembling were authentic," Rubin says.

Rubin says this study should prove valuable to the biofuel industry because it is such an immense addition to the library of known enzymes. "In this one study we've probably doubled the number of [documented] enzymes" relevant to cellulose degradation, he says. Besides the novel enzymes, the database also features whole genomes of previously unstudied organisms that break down plant material but have resisted prior attempts to culture them. Biofuel researchers can now use this information in a similar way that biomedical researchers reference the human genome to accelerate studies of human disease, he says.

The study also shows the usefulness this kind of direct, high-powered analysis of genetic material in samples taken from the environment. Rubin notes that the method, called metagenomics, netted lots of potentially useful information besides that which is relevant to biofuel production. "In this paper we focused on enzymes involved in the breakdown of plant material, but there are all kinds of different activities in this data set," he says. "So it really is a different way of looking at the diversity of functions that exist in nature."