Unnaturally Productive

It’s human nature to never be satisfied with what nature provides. And chemists seeking to improve or repurpose proteins have adopted this attitude in their work. They’ve developed technologies to introduce unnatural amino acids into a peptide sequence.

This year, results have emerged from the commercial ventures inspired by these chemists’ efforts. In human clinical trials, proteins from two companies, each containing an unnatural amino acid, appear to hold promise as longer-lasting treatments for multiple sclerosis (MS) and growth hormone deficiency. Other drugmakers have complementary unnatural amino acid technologies in the works.

Attaching small molecules to proteins is an established practice in the pharmaceutical industry. For example, linking cancer-killing drugs to antibodies helps target the drugs to tumors. However, “the native 20 amino acids that comprise proteins limit one’s ability to do precise chemical modifications,” says Ho Sung Cho, chief technology officer of ­Ambrx, a La Jolla, Calif.-based biotechnology company specializing in developing drugs containing unnatural amino acids. Only a few amino acids have reactive side chains, and unless only one copy of that reactive amino acid is present in a protein of interest, chemical modifications tend to result in heterogeneous protein mixtures, Cho says.

Ambrx’ technology is based on cofounder Peter G. Shultz’s science at Scripps Research Institute. His team’s most advanced technique for inserting unnatural amino acids into proteins borrows two pieces of the protein translation apparatus from a single-celled microorganism called Methanococcus jannaschii—a transfer RNA and an enzyme called an aminoacyl tRNA synthetase that loads that tRNA with an amino acid. By customizing these borrowed parts and inserting them into Escherichia coli bacteria, biochemists can insert an unnatural amino acid at any point in a protein sequence (C&EN, May 7, 2001, page 57).

Ambrx used that technique to make a longer-lasting version of human growth hormone, a drug that’s normally given in a once-daily injection (Proc. Natl. Acad. Sci. USA, DOI: 10.1073/pnas.1100387108). The modified hormone, called ARX201, is in Phase II clinical trials. This project was well suited to an unnatural amino acid approach, Cho says. The company wanted to attach a molecule of polyethylene glycol (PEG) to the hormone, an established technique drugmakers use to enhance a protein’s half-life in the body. But location was important: The PEG had to be attached at a place that wouldn’t interfere with the hormone’s normal activity and that would boost half-life as much as possible.

Attaching the PEG through one of nature’s 20 amino acids wasn’t a viable option, Cho says. For example, when Ambrx scientists placed a cysteine residue, which could easily be modified with PEG, at a promising attachment point, they obtained disulfide-bonded aggregates of growth hormone, he explains.

With the unnatural amino acid p-acetylphenylalanine, however, Ambrx could try multiple PEG attachment points on the hormone without worrying about aggregation. Their attachment strategy produces a modified hormone that, in human clinical trials, shows the potential to be injected once a week instead of once a day.

Cho thinks what’s most interesting about Ambrx’ advance from a pure chemistry standpoint is the bond the company used to attach the PEG to the growth hormone—an oxime, formed by a reaction between the acetyl group of p-acetylphenylalanine and an alkoxyamine-functionalized PEG. “This is conjugation chemistry that hadn’t really been studied in humans,” he says. Oxime bond formations have a reputation for being reversible, “but when we transitioned the PEGylated protein to physiological pH, a study we did for a year, that bond does not reverse at all,” he says.

At Seattle-based biotechnology company Allozyne, researchers are also working with unnatural amino acids and trying alternative conjugation chemistries, including the palladium-catalyzed Heck and Suzuki reactions. But the Allozyne drug candidate that’s furthest along in development relies on conjugation chemistry with a longer track record—copper-catalyzed click chemistry between an azide and an alkyne.

Like Ambrx’ ARX201, Allozyne’s candidate, called AZ01, is a long-acting version of a protein drug already on the market, but one that’s intended to treat MS. The protein’s unnatural amino acid is inserted in a different fashion as well.

Instead of borrowing translation machinery to make AZ01, Allozyne harnesses the innate promiscuity of translation in E. coli. Scientists deprive E. coli of methionine, one of the 20 amino acids it needs to survive, and introduce an unnatural amino acid, such as azidohomoalanine, into the cell culture medium instead. The unnatural amino acid resembles methionine enough that the bacterium uses it to replace all the methionines in a protein sequence. This “residue specific” substitution technology originated in Allozyne cofounder David A. Tirrell’s lab at California Institute of Technology. Allozyne calls the method CAESAR, after a cryptography technique said to have been used by the Roman emperor himself. Fittingly, the code substitutes one letter of the alphabet for another.

“We’ve followed the basic technology premise from the Tirrell lab, but we’ve improved on it,” explains Kenneth Grabstein, Allozyne’s chief scientific officer. For instance, because substitution happens residue-selectively, “the method requires that we engineer out natural methionine residues,” leaving only one methionine for substitution, he says. Engineering out methionines has worked for Allozyne on a dozen proteins so far, he adds.

When developing AZ01, Allozyne’s scientists left one methionine at the N-terminus of interferon β, an established MS treatment, for substitution and eventual PEG attachment. They also learned how to prevent the unnatural amino acid from being clipped away by modifying other residues in the interferon protein (ChemBioChem, DOI: 10.1002/cbic.200700379). In a Phase Ia clinical trial, Allozyne’s PEGylated interferon lasted long enough in patients’ bloodstreams to suggest that it might be possible to administer it once a month, instead of the current standard of once to three times per week. A Phase Ib trial to further evaluate the drug’s safety and dosing is ongoing.

This year’s results from Ambrx and Allozyne, as well as a 2008 study from Genentech that used engineered cysteines instead of unnatural amino acids (Nat. Biotechnol., DOI: 10.1038/nbt.1480), underscore the advantages of making site-selectively-modified proteins instead of heterogeneous mixtures, says David Rabuka, cofounder and chief scientific officer of Redwood Bioscience, a company developing its own chemistry for site-selectively-modified protein drugs. “At one point people thought that site specificity was nice to have but not a must-have,” he says. “But the data are turning out to show that it is a must-have—you can get something significantly more efficacious than a random mixture.”

Redwood’s protein-modification chemistry is based on aldehyde chemistry developed by cofounder Carolyn Bertozzi’s group at the University of California, Berkeley. The process works by modifying a protein’s genetic code so that it will instruct an enzyme to selectively oxidize a particular cysteine residue to a formylglycine, an unnatural amino acid with an aldehyde side chain. Then, any of myriad small molecules, including PEG chains and fluorophores, can be attached to the aldehyde through oxime or other linkages (Curr. Opin. Chem. Biol., DOI: 10.1016/j.cbpa.2010.09.020). Redwood is using its platform to create combinatorial libraries of conjugates. The collection can be rapidly screened to identify the best sites of attachment for antibody-drug combinations or long-acting protein therapies.

Although the unnatural amino acid products nearest to market are made in E. coli, South San Francisco-based Sutro Biopharma is developing a manufacturing process that avoids intact cells altogether. Chief Scientific Officer Trevor Hallam says Sutro’s technology, licensed from James R. Swartz’s lab at Stanford University, will speed up the process of picking the best spot on a protein for placing an unnatural amino acid. The procedure uses unnatural amino acids attached to tRNAs, fastened either with a synthetase or through chemical synthesis. To make proteins that contain unnatural amino acids, Sutro’s scientists add that loaded tRNA to an extract made from specialized E. coli cells that have been broken open. The company has optimized the recipe to make it easier to scale, and it has generated a library of unnatural amino acids and antibody fragments for screens to rapidly test how tweaks to unnatural amino acid structure or placement affect a protein’s function or pharmacokinetic properties (Biotechnol. Bioeng., DOI: 10.1002/bit.23103).

Hallam thinks the method will come in most handy before scale-up, at early stages of protein drug discovery. “Because we can produce proteins in a few hours, we can make hundreds to thousands of variants in parallel in 96-well plates and have real data on the functions of those proteins in about 12 days,” he says.

If protein drugs containing unnatural amino acids are approved by the Food & Drug Administration, they may raise the bar for protein therapeutics in general, Cho notes. It makes sense that FDA would have stringent standards for small-molecule drug mixtures because for the most part chemists can keep things homogeneous, he says.

But when it comes to approving protein drugs that are mixtures, “regulatory bodies have been willing to give some additional leeway,” Cho says, because of the limitations of the 20 native amino acids. “But when they start to see more homogeneous [protein] drug substances being developed, I think regulatory standards are going to be reevaluated.”