[OrthoMetaPara]

Does that article have a second page? I read ten paragraphs of preamble followed by two paragraphs of topic.

Anyway, it is about using a class of bacteriocidal polypeptides, called bacteriocins, to treat bacterial infections in humans.

Because we have the technology to create custom polypeptide sequences, the allure is that we could design proteinaceous binding domains that are specific for some nasty pathogen. Through combinatorial chemical methods, you can generate millions of different of different peptides, and then select for the one that binds your bacterial target. Most current antibiotics are small molecules, and although you can screen chemical libraries for some antimicrobial property, you can't make millions or billions of derivatives of some candidate small molecule like you can with a peptide polymer. This could be a breakthrough method for the rapid development of antimicrobial agents.

However, there's a downside, and that is that large molecules have a more difficult time travelling through the body, specifically through tissues that have tight junctions between cells. It seems like these bacteriocins would have to be introduced intravenously (or maybe through the respiratory system, like ricin can be) as most peptides are hydrolyzed in the stomach. Even then, peptides don't easily diffuse through the blood brain barrier. I suppose they could be applied topically to treat MRSA.

[rcthompson]

I agree the article is light on substance, but for a layperson reading it, much of that preamble is probably necessary.

Anyway, I don't think the idea is to generate completely random polypeptides. The analogy with the lego bricks sounds more like they're mixing and matching known protein domains. Proteins are made of subunits called domains, and these are relatively modular, in the sense that you can concatenate the sequences for multiple domains into a single polypeptide and reasonably often end up with a functional protein consisting of the desired domains tethered together. These are called fusion proteins. In this case, the lego analogy suggests that there is a range of domains that bind to different structures on the external surfaces of bacterial cells, and another range of domains that kill bacterial cells by different mechanisms.

So, if you want to kill species X, you first find a binding domain A that binds X cells, and then find an enzymatic domain B that can kill X cells, and create a fusion. This fusion protein AB will now attach to X cells and then kill them. If species X evolves to change its shape so that A no longer binds, you find a new binding domain C that does, and create a CB fusion, which can go on killing the evolved X cells via the same mechanism. If X evolves immunity to B's mode of action, then you find a new enzymatic domain D with another mode of action and create fusion AD. Repeat as necessary.

Obviously, the success of this method depends on how many binding and enzymatic domains are effective against a particular bacterial strain. I have no idea how many bactercins are known, but presumably there must be enough to give hope to the author of this article.

[FungalRaincloud]

So, assuming I'm parsing you correctly, would you say that at this time, the benefit is that we can avoid the bacteria becoming resistant by creating a large range of things that it would have to become resistant to, and the down-side is that since these polypeptides are larger, they will not be as able to be administered where needed? Could this also mean that we end up treating an infection in a specific spot, but allowing it to continue to colonize a different part of the body? For instance, say, in the case of a UTI, where the infection could have moved further up the GI tract, and treatment with polypeptides would work for the urethra, but possibly not further up?

[OrthoMetaPara]

If a particular bacterial strain were to become resistant to a particular bacteriocin, we could use directed evolution methods to modify the bacteriocin.

For example, say the pathogen develops a mutation so that the bacteriocin's binding domain no longer recognizes its bacterial target. Because the bacteriocin is a class of peptide, and not a small molecule, its (theoretically) rather simple to run a directed evolution program in order to discover a variant of the original bacteriocin that will bind the mutant strain. It would be far more difficult to do the same thing with some small molecule antibiotic, and you'd never be able to generate the number of derivatives of the small molecule that you could with a peptide.

Similarly, if the bacteria develops a mutation so that the active domain of the bacteriocin no longer works properly, you can run the same sort of directed evolution experiment.

Or, you could conceivably use a different binding domain altogether as was suggested in the article and in this thread.

As for the delivery of the antibiotic, it needs some way to come in contact with the pathogen. A topical wound that is infected with some form of staphylococcus, for example, could be treated with some sort of ointment. But relying on the body's circulatory system to deliver the drug, well that would seemingly be more difficult than just taking a pill.

[DonaldFisk]

Bacteriophage already evolve in the way you suggest, and are typically administered topically rather than systemically.

[mirimir]

Maybe one could synthesize analogs that don't get hydrolyzed in the stomach.

[ksenzee]

That's harder than it sounds. Insulin is a prime candidate for that - just synthesize an insulin analog that doesn't get hydrolyzed in the stomach, and you have oral insulin. It would be an instant blockbuster drug. There have been billions of dollars spent on research trying to do just that, with no success yet.