Toward Peptide Analog Libraries

Researchers have been competing recently to be able to create enormous collections of varied peptide analogs with attached identity tags, subject the analogs to iterative modification and screening to find those with useful properties, and develop the best ones as bioactive agents. If all goes as planned, the efforts could lead to a plethora of peptidomimetics for a wide range of diagnostic, therapeutic, and catalytic uses--and, not incidentally, to a bountiful harvest of profits for those who make this work commercially.

The work is reminiscent of the early 1980s, when syntheses of the first large libraries of natural peptides marked the beginnings of the field that is today called combinatorial chemistry. Natural peptides are potentially useful because they can bind to and interact with a range of biological receptors very selectively. But they are attacked by proteolytic enzymes in the body, greatly abbreviating their longevity of action. And they aren't orally active because they're digested by the gastrointestinal system and generally cannot reach the bloodstream in intact form.

Peptide analogs--often called nonnatural or unnatural peptides--sidestep these problems because proteolytic enzymes don't recognize them. Recently, a growing number of scientific teams have been honing novel synthetic techniques that could make it possible to create such peptide variants in collections of previously inaccessible size, select and evolve them, and easily identify promising peptidomimetics from those libraries.

Many of the new techniques involve cell-free (in vitro) translation systems for making peptide analogs. In these systems,mRNA is used as a template, and the in vitro synthesis of peptide variants is <br > directed by ribosomes, the cellular organelles that mediate conventional in vivo peptide and protein synthesis.

"The in vitro evolution of extremely diverse libraries of peptidomimetics, and small molecules in general, has great potential, yet very few are aware of the field," notes Chuck Merryman, a postdoc in assistant professor Rachel Green's group in the department of molecular biology and genetics at Johns Hopkins University School of Medicine. "To me, the whole field is amazing. I really believe that if we can apply the principles of evolution to small molecules, we can make Paul Ehrlich-style magic bullets for just about anything and make them just about overnight."

In the late 19th century, German physician Paul Ehrlich coined the term "magic bullets" for compounds that would zero in on disease-causing microorganisms in a very selective manner.

Efforts to produce nonnatural peptides by in vitro translation trace their origin to 1962, when a collaborative group "showed that a purposefully misacylated tRNA could incorporate the misacylated amino acid into discrete positions in a single protein," says chemistry professor Sidney M. Hecht of the University of Virginia, Charlottesville. The group included Fritz Lipmann of Rockefeller University, Günter von Ehrenstein of Johns Hopkins University, and Seymour Benzer of Purdue University. The misacylated amino acid was not a nonnatural one--it was an alanine attached to a cysteine tRNA--"but the principle was established unequivocally," Hecht says, "and decades before any other lab."

In 1971, biophysics professor Alexander Rich of Massachusetts Institute of Technology and coworker Stephen Fahnestock reported that ribosomes could be used to polymerize not just peptides and proteins but polyesters as well. The key was a technique similar to the one that Lipmann, von Ehrenstein, Benzer, and coworkers had used--modifying the tRNAs that normally bring amino acids to the ribosome to instead deliver amino acid analogs capable of forming nonbiological polymers.

In the later 1970s and '80s, several groups, including Hecht's, extended ribosome-based synthesis to permit the incorporation of modified amino acids. In 1983, after working out methods to chemically charge tRNAs with nonnatural amino acids, Hecht and coworkers used ribosomes to produce and characterize peptides that each contained a single nonnatural amino acid residue. Hecht's work "provided important impetus for the idea that you could use the ribosome to synthesize peptidomimetics," comments assistant professor of chemistry Virginia W. Cornish of Columbia University.

In 1989, a team led by chemistry professor Peter G. Schultz of the University of California, Berkeley (now at Scripps Research Institute), developed a "nonsense suppression" technique for making nonnatural peptides. In this technique, a modified tRNA recognizes an mRNA nonsense codon (stop codon)--which normally serves as a translation termination signal--and inserts a nonnatural amino acid into a protein at that point during translation. Chemistry professor A. Richard Chamberlin of the University of California, Irvine, reported a similar nonsense suppression approach a few months later. "This strategy is actually found in nature," Chamberlin notes. "The noncoded amino acid selenocysteine is incorporated into proteins by suppression of a stop codon."

In 1994, Larry C. Mattheakis of Affymax Research Institute in Palo Alto, Calif. (now at Cytokinetics, South San Francisco), and coworkers reported the first use of ribosome display, a technique for synthesizing peptides that remain linked to their encoding mRNA. A major advantage of this technique is that it makes it easy to identify peptides of interest from a peptide library by reading their attached mRNA.

And in 1997, two independent teams developed an mRNA-peptide fusion technique for hitching ribosome-synthesized peptides to their encoding mRNAs, with the antibiotic puromycin used as an intermediary. The groups were molecular biology professor Hiroshi Yanagawa, now at Keio University, Yokohama, Japan, and coworkers; and molecular biology professor Jack W. Szostak of Massachusetts General Hospital, Boston, and coworker Richard W. Roberts, now assistant professor of chemistry at California Institute of Technology.

With either ribosome display or mRNA-peptide fusion, researchers could envision carrying out in vitro selection (iterative selection, amplification, and modification) on libraries of nonnatural peptides much larger than could be used previously--because identifying active peptidomimetics from such libraries could now be more easily accomplished than before. The two techniques suggested an opportunity to produce nonnatural libraries of virtually infinite complexity and screen or "evolve" them to identify novel peptide variants.

In 2002, Roberts and coworkers demonstrated an initial step toward that goal. They combined nonsense suppression and in vitro selection by creating a small mRNA-linked library of peptides and selecting for those containing one nonnatural amino acid.

FOR TECHNICAL REASONS, nonnatural peptides created up to last year were limited to one or two different nonnatural amino acid replacements per peptide. But researchers had begun thinking about how they might do multiple specific replacements well beyond just two.

Last spring, a collaborative team reported in vitro ribosome translations with purified components to create peptides with three different adjacent nonnatural amino acids or five identical adjacent nonnatural residues, flanked on the ends, in either case, by two natural residues [Proc. Natl. Acad. Sci. USA, 100, 6353 (2003)]. The team--which included instructor in pathology Anthony C. Forster and associate professor of pathology Stephen C. Blacklow at Harvard Medical School, and Cornish and grad student Zhongping Tan at Columbia--achieved this using a sense-decoding technique developed in collaboration with Herbert Weissbach, biology research professor at Florida Atlantic University, Boca Raton.

Sense decoding refers to the use of sense codons instead of nonsense codons as sites for exclusive incorporation of nonnatural amino acids. With sense decoding, one could potentially use any of the 61 sense codons, instead of just the three stop codons, to encode multiple different amino acid analogs, so the technique is much more versatile and could aid multiple replacements.

At about the same time, professor of chemistry and chemical engineering David A. Tirrell and coworkers at Caltech demonstrated that a natural codon could be used to encode nonnatural amino acids at multiple sites by using a translation system inside living cells as well [J. Am. Chem. Soc., 125, 7512 (2003)]. In the genetic code, most amino acids are encoded by two different mRNA triplets, a phenomenon called degeneracy. For example, phenylalanine is encoded by both UUC and UUU (U = uridine and C = cytosine). UUC binds strongly to the GAA anticodon on phenylalanyl tRNA, but UUU doesn't bind it quite as well. Tirrell and coworkers introduced into bacteria a genetically engineered tRNA containing an AAA anticodon--an exact match for UUU--and an aminoacyl-tRNA synthetase that charged the engineered tRNA with a nonnatural amino acid. The result: efficient replacement of phenylalanine by the nonnatural amino acid at UUU sites.

However, Roberts notes that most researchers have been using in vitro instead of in vivo translation because "it's actually quite complicated to add extra residues to the genetic code in vivo, and you really can't rewrite the genetic code wholesale because you would kill the cell. The other 99.9% of proteins in the cell wouldn't work."

Last June, Hecht and coworkers reported reengineering of a key functional site on the bacterial ribosome, making the modified ribosome capable of incorporating d-amino acids into proteins with much greater facility [J. Am. Chem. Soc., 125,6616 (2003)]. And in a complementary study, Roberts and coworkers incorporated both D- and -amino acid substrates with natural rabbit ribosomes [J. Am. Chem. Soc., 125, 8090 (2003)]. Together, the studies represented an important advance for the incorporation into protein of stereochemically modified amino acids. "This has the wherewithal to dramatically increase the diversity of ribosomally encoded libraries of peptides and proteins," Hecht says.

Associate professor of chemistry Hiroaki Suga at the State University of New York, Buffalo, and coworkers recently developed a resin-immobilized ribozyme called Flexiresin that provides an alternative to existing methods for charging engineered tRNAs with amino acid analogs [Chem. Biol., 10, 1077 (2003)]. The researchers used the reusable resin in an in vitro translation system to incorporate different phenylalanine analogs at one or two sites in a protein.

And Roberts' group extended the mRNA-peptide fusion strategy by linking nonnatural peptides, instead of natural ones, to puromycin-mRNA [Chem. Biol., 10, 1043 (2003)]. In the study, the group used in vitro translation to make a chain of N-methylated amino acids. N-Methylated peptides are extremely stable to proteolysis and could therefore be orally bioavailable if used as drugs. The researchers call their mRNA-linked nonnatural peptides "encodamers."

"It's a big technical achievement to use a natural codon to code a synthetic amino acid and get it to work in the puromycin display format," Cornish comments.

This nonnatural fusion technique could help lead to the in vitro synthesis of large mRNA-based peptidomimetic display libraries. Roberts and coworkers haven't demonstrated that yet. However, "we think we should be able to make libraries that have literally trillions of compounds in them and use them in selections and screens, and ultimately they would have the properties you want built into them," Roberts says.

"The real problem as we see it is not making individually encoded weird polymers," Merryman says. "It's actually getting to the point where you can make the library itself." In the past, the chemical preparation of tRNAs in sufficient quantities to construct such libraries has been an extremely labor-intensive process. However, Green and Merryman have "a patent in now, and a paper submitted, where we can make about a gram of 20 different ones in an afternoon in a quite simple chemical transformation step," Merryman says.

BIOMOLECULAR ANALOGS that aren't synthesized ribosomally are being pursued as well. For example, associate professor of chemical engineering Annelise E. Barron of Northwestern University and coworkers specialize in peptoids (N-substituted glycines), protease-resistant achiral peptide analogs that were developed initially in 1990 and continue to be of strong interest for a range of potential biological applications.

And chemistry and chemical biology professor David R. Liu of Harvard University and coworkers have developed the use of DNA sequences as templates to synthesize a range of peptidic and nonpeptidic small molecules and polymers, which remain attached to their DNA templates [J. Am. Chem. Soc., 125,13924 (2003)]. Large libraries can be made this way, subjected to in vitro selection, and iteratively modified and amplified to identify small molecules that bind targets or catalyze reactions. A similar technology called Chemetics is being developed by Nuevolution of Copenhagen, Denmark.

"The special advantage of making nonnatural peptide libraries by translation--instead of by solid-phase library synthesis, which is arguably much easier and was accomplished decades ago--is that the resulting libraries can be subjected to selection and amplification, powerful processes that greatly increase the number of compounds that can be evaluated and greatly decrease the amount of material that must be made for the discovery process to work," Liu comments. "Efforts to translate libraries of molecules including, but also extending beyond, those that can be made by the ribosome allow these concepts to be applied to even more diverse classes of molecules."

Meanwhile, in the field of ribosome-based peptidomimetic synthesis per se, "I think the field is really exciting because the promise is there," Cornish says. "What we will explore next is how far can the chemistry of backbone analogs depart from that of a-amino acids, and what are the limitations to cross-reactivity when you start using multiple different codons?"

"The possibilities could be endless," Howard's Blacklow adds. "What we'd dearly love to be able to do would be to display large libraries and then use the power of large numbers to pull out all kinds of different things, where you're only restricted by the ability of the ribosome to carry out bond-forming reactions--and by your imagination."