Genetic engineering may keep one of the richest drug gold mines from being played out. BY JOHN K. BORCHARDT The rule of thumb may be that 1 compound in 5000 becomes a drug, but among the compounds called polyketides, drug prospecting odds have been much better—around 1 in 100. Polyketides are small, cyclized molecules. Between 5000 and 10,000 are known, and about 1% of them possess drug activity. Sales of the more than 40 polyketide drugs—including antibiotics, immunosuppressants, cholesterol-lowering agents, antifungals, and cancer chemotherapeutics—exceed $15 billion a year (Table 1). So it isn’t surprising that pharmaceutical companies attach enormous value to the polyketide compound libraries that they have gathered from bacteria, fungi, and plants. But there is concern that the drug discovery potential remaining in existing libraries is diminishing. The obvious, but hardly easy, thing to do is to find new polyketides. One interesting way of doing that, which can potentially create unlimited numbers of new polyketides, uses a combinatorial chemistry technique called combinatorial biosynthesis. Combinatorial biosynthesis manipulates genes in natural-product biosynthesis pathways as a way of producing natural-product analogs. Mutation or substitution of a gene in a biosynthesis pathway may result in synthesis of a previously unknown analog of a natural product (1). Production of new polyketides is probably the most advanced application of combinatorial biosynthesis; genetically altering their biosynthesis pathways can potentially generate a nearly inexhaustible number of new polyketides for drug assays and further development. In current practice, combinatorial biosynthesis is generally used to produce relatively small libraries of focused compounds rather than extremely large libraries of drug candidates. A particular benefit of the technology is that atoms in a natural product can often be altered in ways that are difficult or impossible to achieve by the methods of medicinal chemistry. The polyketide assembly line It’s usually the case that a class of compounds is named after a shared chemical structure. Structurally, however, polyketides often have little resemblance to each other. They get their name by sharing biosynthetic pathways that produce a common ketone structure among their synthesis intermediates. Exploiting combinatorial biosynthesis to make new polyketides depends both on understanding the process of making polyketides and how the genes in their synthesis pathway are organized. A polyketide is built from a linear polymer of carbon atoms created by sequential reactions of enzyme complexes called polyketide synthases (PKSs) (2). Anywhere from 5 to 50 PKS reactions are needed to make a polyketide (the antibiotic erythromycin requires 25). Synthesis begins when the first monomer in the polymer chain is “loaded” (covalently tethered as a thiol ester) on the first PKS in the reaction sequence (Figure 1). After loading, an extension reaction lengthens the chain by two carbon atoms. Following extension, the chain is tethered to a new carrier protein, then another two-carbon extension is made, and so on. The tethered, growing chain passes through reactions as if rolling along an assembly line, with new parts being added at steps along the way. Every other carbon in the chain becomes a ketone (hence, polyketide). At the end of the line (i.e, as the final PKS reaction), the chain is untethered, and then it is cyclized by non-PKS enzymes. Additional enzymatic reactions add final trimming (carbohydrates, methyl groups, etc.). A PKS is not a single protein; rather, it is comprised of several polypeptides having, at minimum, loading, chain-extension, and chain-releasing activities. Between extension and untethering, a PKS sometimes sandwiches in one or more ketone-modifying reactions. Extension and modifier polypeptides are together called “modules”. A PKS may contain one module, or several. In the way that DNA encodes proteins, polyketides are said to be “encoded” by the sequence of modules used to synthesize them. Once it was understood that polyketides were encoded by modules, it was a quick leap to the idea of making new polyketides by altering “codes”; that is, by changing enzymes within modules. Changes would be made by altering genes for PKS proteins. With thousands of polyketides and many more thousands of PKS genes, it was obvious to researchers that the number of possibilities for recoding pathways to create new polyketides was astronomical. Standing in the way of polyketide recoding, however, would seem to be the practical matter of isolating the genes belonging to their pathways. Yet this is not as difficult as might be expected, because genes in a polyketide pathway are always found clustered together in a contiguous DNA sequence (2). Thanks to this clustered gene organization, when one gene in a PKS pathway is isolated, finding the others is easy—their genes are located nearby. Such gene clustering is quite unusual: Genes of biochemical pathways are usually dispersed to different chromosomal locations, and that usually means they must be isolated the hard way—one at a time. Making unnatural polyketides A decade ago, Hopwood and co-workers showed that recombining similar polyketide synthesis genes could generate “hybrid antibiotics” (3). They arbitrarily mixed genes for the biosynthesis of actinorhodin with those for granaticin and medermycin. By transferring partial or complete biosynthetic genes between different polyketide-producing strains, the new antibiotics mederrhodin A and dihydrogranatirhodin were produced. Generation of the molecules involved enzyme activity in the later steps of the biosynthetic pathways rather than the earlier PKS-catalyzed steps. Unfortunately, in follow-up experiments, only minor structural changes were achieved, and the initial intense excitement surrounding this approach faded. In 1994, Chaitan Khosla and a team at Stanford University developed a technique called chemobiosynthesis (4), which was used successfully in altering PKS genes in order to make new polyketides (Figure 2). Their success depended on overcoming the preference of PKS enzymes to process compounds produced by the host. A team member, John R. Jacobsen, solved the preference problem by altering a gene so that the extension step in the first module was blocked. This prevented production of second-step precursors, so natural biosynthesis would never get started. The Stanford team then fed the PKS enzymes with synthetic precursors designed to load at the second module. Loading was successful, and the precursors were processed by the rest of the PKS system. “We expected to find other obstacles once we had this initial problem solved, but we haven’t seen any that are serious,” Khosla said. “We’ve been pleasantly surprised by how tolerant the entire system appears to be toward unnatural substrates. The overall process seems to work remarkably well.” Of the first 11 polyketides that the team produced by using chemobiosynthesis, only 2 were previously known. “We appeared to be opening a door to interesting molecules that have not been found in nature,” Khosla said. “You are no longer limited to the confines of cellular metabolism in terms of what molecules are taken up by the assemblies and what molecules are spit out by these assemblies.” At Kosan Biosciences (Hayward, CA), scientists led by Khosla have used chemobiosynthesis to make analogs of erythromycin. Natural synthesis intermediates in the erythromycin pathway can be replaced with carefully altered compounds, which are accepted at the appropriate pathway points and processed by the remaining erythromycin PKS enzymes. Several chemobiosynthesized analogs exhibited antibacterial potency comparable to that of erythromycin in laboratory tests. The company hopes to find analogs with greater potency, broader spectrum of action, and the ability to evade resistance by bacteria. Combinatorial polyketides Kosan has also used genetic engineering to produce polyketides in a random manner, similar to the way combinatorial chemistry is used to produce new compounds. This is done by recombining cloned PKS genes from different biosynthetic pathways to create new pathways. In one method, the process starts with a PKS gene cluster cloned into E. coli. Selected genes are then be replaced by PKS genes from different clusters. New gene clusters can also be created by ligating together polymerase chain reaction (PCR) DNA of PKS genes from different parental clusters. Gene clusters made by either method are then transferred to bacteria, which are then plated on bacterial culture plates. The result is a bacterial library of combinatorial polyketides. Extracts of bacterial colonies producing the new compounds can then be assayed for drug activity (Figure 3). Combinatorial biosynthesis can be taken a step further once a bacterial colony with a desirable new polyketide has been found. PKS genes responsible for the new polyketide can be subjected to random mutation and used to generate more polyketides. By this process, combinatorial biosynthesis can produce a focused library of lead polyketides. Before 1998, polyketide production was limited to natural PKS hosts and Streptomyces bacteria (many different polyketides come from Streptomyces species, which is why these have been the primary laboratory bacteria for this field). Then Kosan researchers reported the first expression of a polyketide in E. coli and yeast, two organisms readily cultivated on an industrial scale (5). Production in these organisms is still limited by the restricted availability of suitable polyketide metabolic precursors, but the problem is not seen as insurmountable (6). Other compound classes for combinatorial biosynthesis Like polyketides, nonribosomal peptides are natural products built from simple monomers—in this instance, amino acids linked and altered by nonribosomal enzymes. The immunosuppressant cyclosporin A and the penicillin-class antibiotics are examples of nonribosomal peptides. Like polyketide PKSs, nonribosomal peptide synthetases contain repeated, coordinated, active sites that catalyze polypeptide chain elongation and functional group modification. Researchers have demonstrated that the number, content, and order of these active sites can be genetically altered to modify nonribosomal peptide structures (6). Advances in understanding the biosynthesis and genetics of deoxysugars of antibiotics could make combinatorial biosynthesis of libraries of these sugars important for discovery of better antibiotics (7). The deoxysugar moiety is part of the chemical structure of some antibiotics (e.g., the tetracenomycin class). Researchers at Transcell Technologies, Inc. (Cranbury, NJ), have discovered techniques for generating carbohydrate-based combinatorial libraries. Their technology includes the formation of glycosidic linkages on a solid support and formation of multiple glycosidic linkages in a single process step. Other combinatorial biotechnologies Scientists at TerraGen Discovery, Inc. (Vancouver, BC), use a somewhat different approach to combinatorial biosynthesis. In what they term “combinatorial biology”, they isolate DNA from microbes that are difficult to cultivate and clone the DNA into E. coli and yeast (Figure 4). The company has gathered microbial populations (in pools of as many as 1000 species) from lichens, deep-sea thermal vents, and rain forest soil samples in a search for novel genes. Yeast and E. coli libraries become permanent archives containing the biosynthetic potential of entire microbial communities. Enzymes expressed from the archived genes can be used to produce libraries of novel compounds. Omniscience, Inc. (Tempe, AZ), uses combinatorial genetics as the basis of a drug discovery program focusing on novel anti-infectives and problems of antibiotic drug resistance. Using newly discovered species of marine organisms, actinomycetes, and myxobacteria as DNA donor sources, the company intends to clone genes for entire biosynthetic pathways. The goal is production of novel and diverse lead compound libraries without culturing or maintaining the source microorganisms in the laboratory. The company’s approach includes recombining DNA from its libraries to create “hybrid” DNA. This is similar in spirit to Kosan’s combinatorial biosynthesis technique, and it has the same goal: permitting synthesis of “unnatural” products not known in nature. A technique called combinatorial biocatalysis, developed by EnzyMed, Inc. (Iowa City, IA), uses enzymes and microbes for the iterative derivatization of small-molecule lead compounds. EnzyMed’s technology combines iterative enzymatic and microbial reactions with organic synthesis and screening expertise. The company claims that its technology can optimize lead compounds regardless of structural complexity. The technology uses solution-phase reactions performed under mild conditions and is said to provide controllable regio- and enantioselectivity (8). Conclusions Commercial biocombinatorial research activity is increasing rapidly. Schering-Plough Research Institute (Kenilworth, NJ) will use TerraGen’s combinatorial biosynthesis technology to discover new anti-infective compounds. After TerraGen generates compound libraries and screens them using Schering-Plough’s antimicrobial assays, Schering-Plough will be responsible for lead compound development. In a similar move, Kosan will work with Ortho-McNeil Pharmaceutical and R. W. Johnson Pharmaceutical Research Institute (both of Raritan, NJ), using combinatorial biosynthesis to produce antibiotics. Amylin Pharmaceuticals, Inc. (San Diego, CA), is screening candidates from EnzyMed’s combinatorial catalysis librar-ies in an effort to develop new drugs to treat obesity and diabetes. High-throughput screening methods increase the need for intelligently designed, diverse compound libraries. More than 70% of drugs in clinical use are derived from natural products, so screening synthetic compound libraries based on natural products seems to be a worthwhile way to search for new drugs. Combinatorial biosynthesis and its variants offer innovative ways to produce natural-product-based libraries for finding agents to treat serious health problems. References (1) Hutchison, C. R. Combinatorial biosynthesis for new drug discovery. Current Opin. Microbiol. 1998, 1, 319–329. (2) McDaniel, R., et al. Rational design of aromatic polyketide natural products by recombinant assembly of enzymatic subunits. Nature 1995, 375, 549–554. (3) Hopwood, D. A., et al. Production of “hybrid” antibiotics by genetic engineering. Nature 1985, 314, 642–644. (4) McDaniel, R., et al. Engineered biosynthesis of novel polyketides. Science 1993, 262, 1546–1550. (5) Kealey, J. T., et al. Production of a polyketide natural product in nonpolyketide-producing prokaryotic and eukaryotic host. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 505–509. (6) Cane, D. E.; Walsh, C. T.; Khosla, C. Harnessing the biosynthetic code: Combinations, permutations, and mutations. Science 1998, 282, 63–68. (7) Johnson, D. A.; Liu, H. Mechanisms and pathways from recent deoxysugar biosynthesis research. Curr. Opin. Chem. Biol. 1998, 2, 642–649. (8) Michels, P. C., et al. Combinatorial biocatalysis: A natural approach to drug discovery. Trends Biotechnol. 1998, 16, 210–215. John K. Borchardt is a science writer living in Houston, TX. Comments and questions for the author may be e-mailed to mdd@acs.org, faxed to 202-776-8166, or mailed to Modern Drug Discovery, 1155 16th St. NW, Washington, DC 20036. SEE OTHER HOT ARTICLE FROM THE JULY/AUGUST ISSUE: The Mysterious Placebo Effect