As you walk barefoot through your backyard on a hot summer day, luxuriating in the grass between your toes, stop to contemplate the silent war going on below you. Tens of thousands of microorganisms are sharing that little patch of earth, chattering with one another through complex chemical signals, battling for food, and picking one another off using molecular weapons.
The molecules that microbes make to give themselves a competitive edge are the basis of many of the drugs we take. Antibiotics, cancer drugs, and even cholesterol-lowering pills are among the medicines made or derived from microbial sources.
For decades, pharmaceutical companies played with bacteria in petri dishes—raising or lowering the temperature, fiddling with nutrients—to coax them to produce compounds that chemists turned into penicillins, cephalosporins, and carbapenems. But plucking a bug out of its natural environment means that it no longer needs all its molecular defense mechanisms; as a consequence, researchers have managed to find only about 1% of the compounds microbes can make. With productivity waning, drug firms started to abandon natural products research.
A new generation of natural products scientists is now emerging, one that thinks it has figured out how to tap into the other 99%. Rather than rely on the petri dish, scientists are sampling the DNA of thousands of microbes at once in search of gene clusters, sets of genes with the recipes for making families of similar molecules. They then decipher the structures of the molecules and coax organisms into making them. The hope is that those molecules will one day be the basis for drug candidates.
This concept of bacterial “genome mining” is not new. Efforts began in the late 1990s, but they lost traction as it became clear the technology wasn’t ready for prime time. Now, the plummeting cost of gene sequencing, combined with better strategies for finding interesting genes, has scientists feeling more confident about the approach. The high-profile launch last year of Warp Drive Bio, a biotech company dedicated to genome-based natural products drug discovery, has industry watchers taking a second look, although skeptics are waiting for an actual drug to emerge.
When Cambridge, Mass.-based Warp Drive burst onto the scene in early 2012, it captured headlines for the big-name investors, big-name scientists, and big dollar signs attached to the project. Third Rock Ventures, Greylock Partners, and Sanofi agreed to sink up to $125 million into the company’s gene-mining approach. And Sanofi, which allowed Warp Drive to use its vast collection of microbes, agreed to buy the biotech firm if it met certain research goals.
Interest heightened earlier this summer when one of Warp Drive’s founders, renowned Harvard University chemical biology professor Gregory L. Verdine, said he was taking a three-year leave of absence from his academic post to run the company.
Warp Drive is not the only player in the gene-mining game. Novartis, the one big pharma company to maintain a sizable presence in natural products over the years, is also actively identifying gene clusters, using bioinformatics to predict the structures they code for, and applying synthetic biology techniques to turn them into drug candidates, says John Tallarico, head of chemical genetics at the Novartis Institutes for BioMedical Research.
Two nonprofits started after Merck & Co. shut down its natural products activities, Natural Products Discovery Institute (NPDI) and Fundación Medina, are also getting into the act. Although the organizations are steeped in traditional approaches to natural products, they are quickly adapting as they field requests from both academic and biotech scientists who want genomic information from their natural product collections.
“What we’ve seen in just a couple of years is rising interest from companies to mine the rich genome banks that are harbored in bacteria,” says Rubén Henríquez, director of business development at Fundación Medina.
The allure of cracking into the bacterial genome is simple. According to an examination of the origins of drugs approved since 1981, published last year in the Journal of Natural Products, Mother Nature’s molecular building blocks are hard to beat. About 50% of the small-molecule drugs approved in the U.S. between 2000 and 2010 either are natural products or are related to one. In 2010, 10 of the 20 small molecules to win approval from the Food & Drug Administration were natural products, including most of the new cancer drugs.
That stream of new drugs emerged despite dwindling natural products research by the big pharmaceutical companies. Most firms shuttered or significantly downsized their natural products research efforts during the 1990s, leaving Novartis and, to a lesser extent, Pfizer, as the last of the pharma majors to maintain some level of internal activity.
Pharma veterans who spent their careers in the trenches of natural products research point to several reasons behind its industrial demise. Most often cited is the advent of combinatorial chemistry and high-throughput screening, technologies that enable thousands of compounds to be rapidly synthesized and tested.
By comparison, finding and isolating a hit from a natural products discovery campaign is painstaking work: Classical methods mean sloshing fermentation broth over a bacteria-laden petri dish and waiting to see what compounds are made. Even after a promising compound is discovered, it can take up to a year to purify enough of it to conduct the kind of studies that lead to human tests.
By the 1990s, speed had trumped old-fashioned elbow grease. Companies “wanted to have a lead candidate within one-and-a-half to two years,” says William Kinney, director of business development at Doylestown, Pa.-based NPDI, which in 2011 acquired nearly 100,000 plant and microbial samples from Merck. “With that new, accelerated time scale, it was very hard for a natural product to compete.”
Moreover, although natural products R&D had provided a wealth of drugs for decades, companies were having a harder time finding novel compounds. Instead, they kept “discovering” the same crop of molecules that had already been discovered.
The productivity problem was only exacerbated by efforts to adapt natural products to a high-throughput setting, veterans say. Scientists shifted from testing whole fermentation broths—a goopy proposition that didn’t work well with the assays used in high-throughput screening—and began focusing on solvent-specific extracts. But because they tended to choose hydrophobic extracts, the most interesting compounds were often being thrown out with the aqueous layer.
“This was bad for antibacterials because many of the interesting things were polar,” says Lynn Silver, a consultant who worked for years with natural products at Merck.
Meanwhile, drug companies were walking away from developing antibiotics, the area where natural products researchers had the most success. Facing an uphill battle to discover new products, the drug industry shifted its resources to more lucrative therapeutic areas.
And yet, at the tail end of pharma’s retrenchment, scientists were starting to realize that the previous decades of natural products research had unearthed only a tiny sliver of the compounds bacteria were manufacturing. As DNA sequencing technology became more available and less expensive, academic and industry scientists began to ponder how to skip the messy and laborious culturing step altogether.
Just how much untapped molecular diversity is out there? Observers point to the mapping of the genomes of several actinomycetes, a group of soil-dwelling gram-positive bacteria that are the source of many commonly prescribed antibiotics and other drugs.
In 2001, the genome for Streptomyces avermitilis was partially sequenced by a group of scientists led by Satoshi Omura of the Kitasato Institute, in Japan, who found gene clusters for about 25 secondary metabolites, molecules that help with survival. The next year, a group led by David Hopwood at the John Innes Center in Norwich, England, published the sequence for Streptomyces coelicolor, revealing more than 20 gene clusters for secondary metabolites.
Scientists working in petri dishes would expect to see “three, or if you’re lucky, four” compounds from S. coelicolor, explains David Newman, chief of the Natural Products Branch of the National Cancer Institute. And not for lack of trying to find more, he adds. “If I had a penny for every time I fermented that little bugger, I could take you out for a good meal.”
Yet not everyone looked at the abundant gene clusters and saw a sea of drug candidates. The biosynthetic pathways defined by these genes are turned off most of the time. That inactivity caused skeptics to wonder how genome miners could be so sure they carried the recipes for medicinally important molecules.
Researchers pursuing genomics-based natural products say the answer lies in evolution and the environment. “These pathways are huge,” says Gregory L. Challis, a professor of chemical biology at the University of Warwick, in Coventry, England. With secondary metabolites encoded by as many as 150 kilobases of DNA, a bacterium would have to expend enormous amounts of energy to make each one.
Because they use so much energy, these pathways are turned on only when absolutely necessary. Traditional “grind and find” natural products discovery means taking bacteria out of their natural habitat—the complex communities where they communicate and compete for resources—and growing each strain in isolation. In this artificial setting, bacteria have no reason to expend energy to make anything other than what they need to survive.
“I absolutely, firmly believe that these compounds have a strong role to play in the environment in which these organisms live,” says Challis, who also continues to pursue traditional approaches to natural products. “Of course, not all bioactivities will be relevant to human medicine and agriculture, but many of them will be.”
While scientists such as the Japanese researchers were working out the entire genome for specific organisms, others in academia and industry dove right into trying to produce new compounds from environmental DNA, fragments of DNA pulled out of soil samples irrespective of their microbial origins. The first efforts were exploratory. Researchers wanted to find out what was there and hoped they could later clone stretches of genes and plug them into host organisms that would produce them.
The early methods were rudimentary. Ariad Pharmaceuticals, for example, had a small gene-mining effort in partnership with scientists at the University of Wisconsin, Madison. The biotech firm ran a basic experiment to see what a random sampling of soil might yield. “Rather than imposing this really stringent selection of these organisms by asking them to grow at 37 °C, let’s just ask, ‘Who’s there?’ ” recalls Michael Gilman, who was chief scientific officer at Ariad during the inception of those efforts.
To answer that question, they needed soil samples. A particularly enthusiastic Ariad scientist, Marcia S. Osborne, dug up her own backyard in Lexington, Mass., and brought a few buckets of dirt into the office.
It turned out that even when researchers used a very crude DNA extraction method, suburban soil displayed “unbelievable genetic variety,” Gilman recounts. “That dirt—it might have come from Mars, it was so wacky.”
The next question was what to do with all that information. Scientists thought it would be as easy as “cloning DNA from the environment, dump it into a bug, and you have this amazing resource,” says Sean F. Brady, head of Rockefeller University’s Laboratory of Genetically Encoded Small Molecules.
A handful of companies emerged to exploit the commercial potential of gene mining; industry veterans will remember names such as TerraGen and Diversa. But one by one those efforts—at least for the purpose of finding drugs—faded. “It took a long time for people to realize that it’s actually a significantly more complex problem,” Brady says.
The next generation of natural products researchers has spent the past 10 years trying to work through the problems that have held the field back from realizing its commercial potential.
The most progress has been made in understanding how to organize all the information gleaned from both whole bacterial genomes and the fragments of gene sequences found when sampling environmental DNA. Scientists have generated enormous databases of genomic information that can be used as a starting point for drug discovery efforts. For example, if a company wants to find analogs of a specific compound, the database can be mined for DNA that encodes specific structural features of the molecule or biosynthetic operations that make it.
With databases in hand, researchers still need to decide what to search for. Scientists see three avenues, each with an increasing level of difficulty.
The easiest is to look for DNA that encodes for molecules that are closely related to an already marketed drug in hopes of improving its properties. But that approach has commercial limitations. “The drawback is that normally patents are written to cover most of the chemical space around a drug,” Warwick’s Challis notes.
Another avenue is to search for relatives of a natural product that for some reason failed to make it as a drug. This strategy assumes that Mother Nature is the best medicinal chemist and has already come up with the kind of improvements needed to make it viable as a drug.
In both scenarios, abundant information about the synthetic pathway makes a great guide for finding analogs. It’s a direction being pursued by Warp Drive, although Verdine is quick to note that his goal isn’t to find minor tweaks on the original. “We’re looking for analogs in which the evolutionary drift in the structure is large enough that it’s taken the class of compounds to a new target,” he explains.
Warp Drive is trying to improve its chances of quickly finding interesting molecules by narrowing its sights onto compounds made to combat fungi. To be effective, a compound needs to cross the bacteria’s cell wall, survive in the soil, and then get into the fungus. Verdine argues that looking for that ability will improve Warp Drive’s chances of finding molecules that look and act like drugs. “Probably 90% of compounds that people screen for antibacterial activity don’t work because they don’t penetrate the bacterial cell wall,” Verdine says. “Here, we have compounds that have been evolved to get through a very difficult passage—Frodo going to Mordor is nothing—and persist in soil.”
The third avenue, and the one that is the toughest because there’s little guidance about the structure or biology, is to mine the genome for entirely new structures. Although researchers believe this strategy is the farthest from commercial success, they hope it could one day open the door to completely novel classes of biologically active compounds.
Challis’ lab at Warwick was one of the first to show that it was possible to translate a gene sequence to a novel molecule. The group found a gene cluster in the S. coelicolor sequence that coded for a previously unidentified nonribosomal peptide synthetase. Each compound in that enzyme family is responsible for making a single nonribosomal peptide, a class of natural products that includes antibiotics such as vancomycin.
In 2000, the Warwick group predicted that the product of the enzyme would be a tripeptide they called coelichelin. Five years later, to prove the concept was viable, they published work showing they could coax the bacterium into making the product, which turned out to be a tetrapeptide, not a tripeptide.
Challis’ team and others have since identified and made other molecules, and he says the process is getting faster. But when drug companies can use high-throughput screening to test 10,000 compounds with known structures in one fell swoop, the speed and scale of natural products discovery by gene mining looks paltry in comparison.
“A number of people look at this approach and say, ‘It’s kind of one pathway, one molecule,’ ” Challis says. “I think the answer to that question is in emerging DNA synthesis technology and being able to rapidly make clusters and put them under your control rather than the organism’s control.”
Two keys to putting control in the hands of scientists will be the reliable prediction of structures and the ability to persuade organisms to make the compound of interest.
“Our predictions are still not very good,” says Pieter C. Dorrestein, an associate professor at the University of California, San Diego, who specializes in bioanalytical mass spectrometry and proteomics. “There’s always some slight modification of the molecule compared with what was expected.” Dorrestein is applying advanced mass spectrometry techniques to help natural products researchers narrow down their predictions about the structure of a molecule.
Meanwhile, going from a prediction on a computer screen to a real sample in a lab is not trivial. “You can identify clusters to your heart’s content; it’s getting them expressed that becomes a black art,” NCI’s Newman says.
In the case of coelichelin, Challis’ team knew enough about the regulation of the synthetic pathway to choose a set of growth conditions—most important, an iron-deficient medium—that would likely prompt the right genes to be expressed.
But that approach won’t work in every case. “Every microbe presents its own set of challenges,” says Bradley S. Moore, a Scripps Institution of Oceanography biochemist who studies biosynthesis of marine microbial natural products. “It’s not like one method fits all.”
In the absence of clues for how to prompt bacteria to express desired genes, researchers are turning to synthetic biology. The idea is to insert the key stretch of DNA into a microbial host that is more easily manipulated into manufacturing the compound of interest. “We’ve made a very deep commitment to doing engineered overexpression of every single compound we come across,” Warp Drive’s Verdine says. “If there’s a way to engineer it, which almost always there is, we go straight to that.”
Although each step of the gene-mining process has kinks to be worked out, researchers are confident they can turn what has long been an academic pursuit into a commercial one. “I’m really excited about what the next few years are going to bring,” Moore says.
Those watching the field might wonder if the research community’s enthusiasm is grounded in reality. Whenever Verdine waxes too rhapsodically about the science his company is pursuing, he’s brought back to Earth by one of Warp Drive’s board members, Greylock venture capitalist Bill Helman. According to Verdine, he likes to ask, “Yeah, Greg, but where are the drugs?”
That question is one that scientists at the biotech firm pose to one another on a regular basis as a playful reminder of their mission. In its first six months, Warp Drive tested the power of its technology by “rediscovering” through gene mining every known compound, plus a half-dozen new ones, in the drug class it is pursuing, Verdine says. By the end of 2014, Warp Drive expects to have two molecules within striking distance of human studies—that is, with an established mechanism of action and reasonable pharmacological properties.
Although a successful second act for the field of natural products research might depend on Warp Drive meeting those milestones, Verdine is unfazed by the task ahead. “This is a phenomenal opportunity,” he says. “This experiment that we’re in the middle of has never been done before.”