Peptide-Producing Powerhouses

SOME OF THE biggest cellular assembly lines are found in the smallest of beasts. Nature's largest stand-alone catalysts are found in microscopic bacteria and fungi, which employ the enzymes to produce a variety of chemically complex peptides. These intricate peptides, which are often cyclized and methylated or contain unusual D-amino acids and organic groups such as dihydroxyl benzoate, can keep synthetic chemists tinkering at their benches for years in the quest to build copycats.

Microbes make the quirky peptides using so-called nonribosomal peptide synthetase (NRPS) enzymes. Just one of these mega enzymes can possess as many as 40 catalytic sites that are located in sequential modules.

Many of the peptides produced through NRPS enzymes are potent antibiotics—think penicillin or vancomycin—which microbes use against each other in biological warfare. But researchers hope that if they figure out how NRPSs work, reengineering the enzymes to produce new peptide antibiotics will be possible.

It may come as a surprise that microbes use NRPS enzymes instead of the ribosome—the cell's protein-making workhorse—to make these peptides. Although the ribosome is good at high-throughput production, its assembly machinery can accommodate only some 20 L-amino acids as building blocks, ingredients that are far too mundane for a bacterium's entire chemical to-do list.

The ribosome's 2,500-kilodalton composite of 50 proteins and three RNAs is twice as heavy as the largest NRPS and up to 10 times the size of a typical NRPS. But NRPSs are single proteins. Each enzyme is encoded in a single, enormous gene—we're talking up to 45,000 base pairs, or some 15,000 amino acids.

The largest known NRPS comes from a fungus that produces the antibiotic cyclosporin using 11 catalytic modules.

Enzymes in the NRPS family can use hundreds of different compounds as building blocks, and they do more than just catalyze peptide bond formation. NRPS enzymes also perform heterocyclization, methylation, and addition of sugars and other functional groups to growing peptide chains.

"These multi-subunit enzymes are the molecular equivalents of moving assembly lines," wrote the late Jonathan B. Spencer, a chemist at Cambridge University (Nature 2007, 448, 755). "Growing substrate molecules are handed, bucket-brigade style, from one specialized catalytic site to the next, with each site performing a specific and predictable function."

All NRPSs have a sequence of individual modules that are each responsible for selecting a new monomer, adding it to the growing peptide chain, and sometimes modifying that monomer. To assemble their peptides of choice, microbes use an assortment of different types and quantities of modules, but the overall blueprint of the peptide factory is universal.

Every module in the NRPS assembly line has at least three basic components. An adenylation domain selects and activates an incoming monomer. Next is a condensation domain, which is responsible for catalyzing the addition of the new monomer to the growing peptide chain. Finally, each module has a carrier domain that moves new monomers from the adenylation to the condensation domains and also shuttles the growing peptide to the next module. Sometimes a given module will also contain additional domains that further modify a particular monomer, say by methylation or epimerization.

Because the molecules made by NRPS are often "pharmacologically interesting, and many are on the market," researchers are trying to reverse engineer the enzymes so that they can be harnessed to mass-produce valuable antibiotics, which could sidestep complicated or costly medicinal chemistry, explains Mohamed A. Marahiel, a biochemist at the University of Marburg, in Germany. But even more tantalizing, Marahiel adds, is the prospect of reconfiguring NRPS enzymes to build entirely new, nonnatural peptides. This reverse engineering approach is also being used to reengineer polyketide-building enzyme assembly lines in addition to NRPS's.

"It would be fantastic if we could take a module here and a module there to create millions of new compounds in a stereospecific manner," says Volker Doetsch, a chemist at Germany's University of Frankfurt. "Like using Legos," he adds, "one can imagine putting together modules from different NRPS systems to create new assembly lines for new antibiotics and other medically interesting molecules."

IN 1997, Marahiel published the first X-ray crystallography structure of an NRPS domain. Since then, biochemists have solved the structure of many catalytic portions of NRPS modules, but there has not been a clear picture of how all the domains in a given module work together to increase the peptide chain by one monomer, let alone how all the modules work together to build an entire peptide.

But in August, Marahiel and his colleagues published the first X-ray crystallography structure of an entire module. In particular, they solved the 144-kDa termination module from an NRPS used to make the antibiotic surfactin (Science 2008, 321, 659). The module first catalyzes the last elongation step, which adds the amino acid L-leucine. Then it cyclizes the peptide chain and releases it.

The biggest insight from the full structure was that the adenylation and condensation domains, which are present in every NRPS module, create a catalytic platform, or "workbench," on top of which the carrier arm shuttles intermediates, Marahiel explains. The thioesterase domain, which cyclizes and hydrolyzes the completed peptide, sits atop the workbench.

Shortly after Marahiel's paper was published, a team of chemists led by Gerhard Wagner and Christopher T. Walsh at Harvard Medical School published a nuclear magnetic resonance structure of components from the same type of termination module, one that comes from an organism that makes the antibiotic enterobactin (Nature 2008, 454, 903).

The NMR structure of the 35-kDa component revealed the dynamics between the module's catalytic carrier arm and the domain responsible for cyclizing and releasing the completed peptide.

The two groups' structures represent the first view of how separate catalytic domains are organized within an NRPS enzyme, comments Janet L. Smith, a biochemist at the University of Michigan, Ann Arbor. "Understanding the combined structures allows us to figure out how to engineer NRPS enzymes to do even more varied chemistry than what we have characterized them to do in nature."

A fuller picture of how NRPS catalytic domains function within a module, as well as how different modules work together, is a necessary step in being able to repurpose the enzymes to new chemistry, Marahiel says. Thus far, attempts to mix and match modules to create new NRPS assembly lines haven't made it far from the proof-of-concept stage.

Several groups have managed to splice together modules from different assembly lines, but none of the chimera enzymes produced much more than a sprinkling of product: just enough to be detected but too little to compete with any industrial processes. "The redesigned catalytic machinery did not work well," says Marahiel, who was involved in one of the proof-of-principle examples. "It was like there was sand in the machine."

Part of the problem was that researchers didn't know how the different catalytic domains in a given module are organized with respect to each other. "We didn't have the knowledge of exactly where to cut and paste" the domains together, Marahiel says.

ANOTHER CRITICAL issue is that dynamics may play just as important a role as structure in getting new chimeras operating at full power.

"Thus far, attention has focused mainly on identifying static features that allow protein domains to dock selectively to each other," such as electric charge or overall structure, note Stanford University chemists Shiven Kapur and Chaitan Khosla in a recent commentary. "But there is mounting evidence that protein dynamics have a vital role."

"Imagine you have an assembly line and you know all the components, like in a car factory with all the robots, and you know the structure of all of these robots very well," Doetsch says. "But if each robot is working at its own speed and not considering the other robots sitting to the left and right of it, then it is easy to see that the assembly line can't function."

To make sure that the dynamics of chimera modules and domains also fit to one another, "we have to understand the dynamic process," the speed of conformational change inherent in every single domain during its catalytic cycle, Doetsch adds.

Doetsch and his colleagues found that dynamics are also essential to the proper functioning of an NRPS repair enzyme (Nature 2008, 454, 907). But dynamics aside, there are many other challenges to reengineering NRPS enzymes for new purposes.

For example, most of the products of NRPS factories are antibiotics aimed at killing competing microorganisms. So transplanting a full NRPS system, or parts of one, into a new organism that may be convenient to work with on an industrial scale—say Escherichia coli—could easily just kill the host microbe, Smith notes.

And understanding how one complete module is structured doesn't clarify the whole assembly line, Marahiel says. "I would really like to see a glimpse of the big machine" or even how two independent modules interact with each other, he adds.

The long-term goal of reengineering these mammoth enzymes to produce new peptides, perhaps even new antibiotics, "is at least a few years away," Doetsch says. "But I am convinced we can do it."