3 min read

Call me, Ishmael

Making small changes in text to escape plagiarism-detection software is challenging but not really difficult. The relationship between your text and the urtext in the software is precise and syntactic; the software doesn’t know about ideas. Making small changes in a text that change or obscure the meaning is also easy, and is why copyeditors exist. Making small changes in a text that give an interesting new meaning is enormously harder, as in the opening of Peter de Vries’ “The Vale of Laughter”, quoted as the title of this post.

A similar distinction connects two recent science stories.  A Berkeley-Montreal biochemistry collaboration  has made large steps towards GMO yeast that produces morphine. A US-German collaboration discovered a new candidate antibiotic, teixobactin, and claimed it was resistant to antibiotic resistance.

Engineering yeast to produce morphine sounds like it should be easy. After all, morphine is a hugely simpler molecule than insulin, and recombinant insulin is older than “Y.M.C.A” or Space Invaders.  The distinction is that insulin is a protein, and morphine isn’t.

The Central Dogma of Molecular Biology is that DNA is transcribed to RNA and translated to protein, and that (for protein-coding parts of the genome) this is a precise one-to-one copying process. Like all dogmas, this isn’t quite true — splicing matters, as does all the bling attached after translation (formally, post-translational modification). It’s still a good approximation. Making a protein such as insulin is almost as simple as sticking the insulin gene into a cell and running the transcription/translation program on it.

Morphine isn’t a protein. Proteins (enzymes) are the machinery that manufacture morphine. Getting a yeast cell to produce morphine means getting it to produce a whole series of enzymes, but that’s not enough. In order to produce morphine, this set of machinery must get assembled to a properly connected production line that takes in the right raw materials and spits out morphine at the end. That’s not easy.

Most targets of antibiotics are protein or RNA. Penicillins, cephalosporins, and carbapenems target the penicillin-binding proteins that construct the bacterial cell wall. Macrolides, lincosamides, streptogramins, and tetracyclines attack the RNA/protein machinery of the ribosome. Fluoroquinolines such as Cipro block the protein DNA gyrase. Enzymes make good drug targets because the cell produces only small amounts; you don’t need much drug.  Teixobactin is an exception, as are the current last-resort antibiotic vancomycin and teicoplanin. These block the construction of cell walls as penicillin does, but bind to non-protein construction materials rather than protein machinery.  

To evolve resistance to antibiotics with protein targets, bacteria need to make lots of random changes to the proteins until they find a version that does pretty much the same job but doesn’t bind to the antibiotic. That was hard, but there are a lot of bacteria and they reproduce very fast, so we got MRSA.

Resistance to vancomycin took a lot longer to develop. A because a vancomycin-resistant germ has to find mutated versions of enzymes that do a different job. Vancomycin-resistant enterococci make a new enzyme that strips off one end of the target construction material and replaces it with a new piece that still works for cell wall construction but won’t bind to the antibiotic.

Proteins are incredibly complicated molecules. Morphine and vancomycin-resistance are much simpler, but they are new ideas for the cell rather than just edits to the same text.