The art of blowing up protein

In my option, there are many things that make biotin interesting.

Starting with its boat-like (endo envelope) puckered structure: it has two aliphatic fused 5-membered rings, one with a ureido group and the other with a bridging sulfur, the former is involved in catalysis (cf. acetyl-CoA carboxylase mechanism), while the latter is not conjugated with it and acts solely as a steric hinderance.

One of the few enzymes using it, acetyl-CoA carboxylase, make a metabolite needed to make it. Chicken-and-egg paradoxical pathways are not too uncommon and here a simple metabolite (malonate) is made, so can be explained via the Horowitz retrograde hypothesis of pathway formation. 

In urishiol and olivetolic acid (THC precursor) the fatty acid part is an off-the-shelf metabolite that gets used in the usual polyketide way; For biotin, a methylester is smuggled into the fatty acid biosynthesis pathway.

However, the strangest part is that the thiophane is really hard to make. This ring is the last step in biotin biosynthesis and made by the radical SAM enzyme BioB. Radical SAM enzymes have a vast repertoire generally C-H activation via a radical reaction, but generally it centres on methylation or skeletal rearrangement possibly with carbon insertion, but this enzymes inserts the SAM's sulfur itself into desthiobiotin —sulfur being the only part missing.

A further oddity is that the C-H activation part to make the thiophane ring happens on a carbon that was the beta-carbon of an alanine, which was added by BioF via the usual PLP-dependent decarboxylative aldol condensation.

So why does BioF not use cysteine?

If it used cysteine then the ring would need to be closed in a reaction requiring C-H activation, simpler than BioB. Several enzymes do this kind of reaction. Penicillins have a thiophane-like ring, but it's made differently, namely a short non-ribosomal peptide is polymerised at the Cys-Val part via a curious oxidoreductase enzyme, isopenicillin N synthase, which as a αKG-dependent hydroxylase family member uses a non–heam-chelated iron to generate a radical from dioxygen (cytochrome P450 are predominantly eukaryotic). In this enzyme the electron is passed to the sulfur, which attacks the highly substituted beta-carbon of the valine sidechain. For isopenicillin N it's a textbook anti-Markovnikov's rule radical reaction, for the thiophane the strain would probably favour a 5 membered ring. So it is not that.

Cysteine cannot undergo a PLP-carbanion step without undergoing a spontaneous elimination with the leaving of the thiol. During my PhD, I encountered this partially with serine racemase, which led me to look into cysteine racemase, an enzyme reported in the 1970s and then not spoken of, as it is likely impossible as I have blogged about nearly a decade ago. Methionine racemase and methionine transaminase have it easier as they have to fight not beta- but gamma-elemination.

However, thioethers don’t have this problem (or not as badly). Out of curiosity recently I read up on the biosynthesis of the fireworm (not firefly) luciferin not because I had ever heard of Odontosyllis, but because its luciferin sports a cata-fused tricyclic with thiazole and a thiazine rings. These are made by forming two thioethers with levodopa and two cysteines, which are then transaminated triggering a spontaneous dehydrative intramolecular cyclisation (like glutamic semialdehyde cyclisation and several others).

Therefore, whereas it is impossible that nature could start with pimeloyl-CoA and cysteine via BioF, it is concievable that nature could start not with pimeloyl-CoA and alanine, but with beta-enoyl version of pimeloyl-CoA and cysteine via a hypothetical enzyme and a BioF-like enzyme, but operating intramolecularly. The acyl compound (pimelenoyl-CoA) is generated in biosynthesis/degration albeit with a different carrier, but then again pimeloyl-CoA is too. The missing hypothetical enzyme to make the S-conjugated cysteine-pimeloyl-CoA would need to so a thiol-Michael addition (like MetB), so could be an oxidase similarly to fireworm tyrosinase, or could be a PLP-dependent enzyme that acts on an "activated" cysteine such as O-acetylcysteine or O-succinylcysteine or phosphoserine (maybe) as seen in methionine metabolism —if you want to know more Ferla and Patrick 2014 is a nice read! The enzyme that does the operation of BioF but intramolecular with the cysteine conjugate may struggle with the tightness given the boat-like conformation as after all the sulfur-bridge is there to sterically hinder carbonyls binding instead of carbon dioxide. BioA would need to contend with a geometry change, but a transamination or reductive amination is easy for enzymes. If this hypothesis were correct there would be zero need for a BioB as the tetrathydro-thiophene part is already made at the bioF step.

There are several clades that inexplicably lack BioB, so I would bet in its favour. The major counter argument is that there is a clade without it, the Rhizobiales if I recall correctly, whose members secrete a lot of desthiobiotin so this would favour the hypothesis that their BioF accepts cysteine and has a poor yield (unusual for an enzyme but unheard of in secondary metabolism) and for the unwanted elimination the PLP gets unstuck (somehow magically) and for the last step some enzyme radically activates the sulfur for ring closure (either another, simpler, radical SAM or an oxidoreductase working like isopenicillin N synthase).

Getting hypothetical answers is always nice, even if it comes a decade too late and is unimportant. And what better candidate enzyme to possible solve the riddle that a PLP enzyme (I might be bias)!