Irmantis Rokaitis from Biomatter in Lithuania gave a rousing start with his keynote on protein engineering. Biomatter has been working on an enzyme design platform for several years now. Rokaitis put out the tantalizing claim that if you can model the needed transition states for a given reaction, then Biomatter can quickly, by sequence generation and testing, get to a working enzyme.
A very interesting example he sketched is the developing a process to methylate pseudouridine for incorporation into mRNA therapeutics. Current processes are relatively inefficient and expensive, making the N1-methyl pseudouridine a major contributor to the cost of production.
Biological methylation processes typically rely on S-adenosylmethionine aka SAM. If your starting point is a natural enzyme for your engineering campaing, then it will probably use SAM. But SAM is expensive, plus it can degrade to unusable byproducts such as homocysteine. One way to cope with the expense is to have an enzymatic regeneration system, but that is more complexity.
Biomatter decided that since this would be an in vitro reaction, they weren't bound to biological methyl donors. Instead they picked methyl tosylate, an inexpensive synthetic methyl donor (tosylate is one of the favorite leaving groups of chemists). But it is truly synthetic, so there are no natural enzymes which use it. Biomatter successfully designed a methylase which has the catalytic and stability properties to be commercially successful.
An issue that came up in the Q&A is an interesting one: if enzyme design really is that good (or when it is that facile), is there any intellectual property to be had in this space? A common strategy is to write a set of concentric series of claims covering anything 90% identical, anything 80%, anything 70%, etc. If a competitor can take the sequence in your patent and develop a competing design completely outside that envelope, is it better to keep your design a trade secret? Though if one can just go from proposing a mechanism to designing an enzyme, having a patented sequence might just be a bit of a useful crib and not essential.
Codon Devices actually thought about building a whole business around the general concept almost two decades ago. The idea would be to take a patented enzyme and engineer in a flurry of functionally neutral changes that would get it outside the claimed distances. Our last protein engineering project was such an attempt, and for it I wrote a really cool algorithm to maximize the number of intended mutations we could make given the plan for making the library. Alas, the provided assay was absolutely miserable in terms of reliability and then we ran dry in the treasury department.
But I think that was the only purely designed enzyme I heard or saw in the conference - but other presentations were still plenty impressive. For example, Wai Ling Cheung-Li described two of the engineering campaigns that were part of a huge recent publication on how Merck created an economic production process for their macrocyclic oral PCSK9 inhibitor enlicitide. In one case, two big pieces of the molecule are glued together by a repurposed thioesterase lifted from a non-ribosomal peptide synthase. The first enzyme to work reasonably well had extreme batch-to-batch variation - as in some batches worked and some didn't - that was traced to very poor thermostability. Further engineering - assisted with a kitchen sink of structural and biophysical investigations - created a robust enzyme with very favorable catalytic properties.
But in my opinion there's still much to learn from nature.
For example, there have been a flurry of papers around an enzyme that makes specific dinucleotide repeat sequences without any sort of nucleic acid template. Apparently it has a neat little swinging side chain that toggles the specificity of the active site after every nucleotide addition - a molecular flip-flop, flip-flop running reliably ad invinitum. How close are we to being able to take that kind of complexity and just design a library of molecules which will have at least one winner within it>?
Or consider enzymes with multiple interlinked activities. Some DNA polymerases have 5'-->3' polymerase, 5'-->3' exonuclease (or alternatively strand displacement), and 3'->5' exonuclease activities. If one wanted to design those into another polymerase - say an RNA polymerase - could those be designed de novo?
In the absolute last talk I attended, Tadhg Begley of Texas A&M gave a masterclass on dissecting complex catabolic clusters. I had never been exposed before to the fact that while we know how most enzymatic cofactors are biosynthesized, little is known about degradation of these central biomolecules. His subject was riboflavin, and after isolating an organism that could grow with riboflavin as sole carbon source, his group found that it only broke apart the molecule into a ribose which it consumed and a highly insoluble flavin group called lumichrome. Lumichrome is apparently less soluble than blackboard chalk, but somehow escapes the cell.
Begley's group then looked for organisms that could grow on lumichrome as sole carbon source, and found the rest of a degradation pathway - a very complex cluster. I hope to get around to a second post on SIMB which will reference that a bit more. What's relevant to here is that one of the reactions involves free radicals and can be reasonably described as "spooky action at a distance" - the free radical is created at one end of the molecule but the unpaired electron is shepherded by the enzyme so that the big rearrangement takes place far away on the molecule. I'm guessing that sort of chemistry is not yet in the "draw it and we can design the enzyme" category!
But what an amazing time we live in! Designing enzymes becoming easier and easier, with increasingly good success rates. What enzyme do you wish existed that doesn't?
No comments:
Post a Comment