Methyl transfer in psilocybin biosynthesis

CONCLUSION

In the present study, we describe the reaction mechanism behind the biosynthesis of psilocybin and present the highest-resolution crystal structures for any methyltransferase to date. The structures are consistent with the "proximity and desolvation" model, which states that the active site of a methyltransferase forces the reactive groups into close proximity while excluding any interfering solvent molecules from their environment. Formation of so-called near-attack Fig.| Consecutive stages of the reaction cycle. The NPPF motif of PsiM is shown as a stick model (beige). SAH (yellow), sinefungin (light/dark green) and (nor) baeocystin/psilocybin (pink/purple) are represented as ball-and-stick models, with the darker colour representing the main conformation. Corresponding OMIT maps are shown in grey (2mF o -DF c with both ligands omitted from the model, maps contoured at 1.0 σ). Key hydrogen bonds are shown as dashed lines. Each model is rotated slightly differently so as to clearly reveal its relevant properties. a The PsiM-SAH-norbaeocystin complex (PDB 8PB4, 0.91 Å), revealing an alternative substrate conformation potentially involved in amine deprotonation. b The PsiM-sinefunginnorbaeocystin complex (PDB 8PB5, 0.89 Å), a model for the state of the complex prior to the first methyl transfer. c The PsiM-SAH-baeocystin complex (PDB 8PB6, 0.93 Å), revealing the state of the active site immediately after the first methyl transfer. d The PsiM-sinefungin-baeocystin complex (PDB 8PB7, 0.92 Å), approximating the situation immediately before the second methyl transfer. e The PsiM-SAH-psilocybin complex (PDB 8QXQ, 0.94 Å). f Superposition of the structures shown in panels a-e. Interactions shown as dashed lines belong to the main conformation of the PsiM-SAH-norbaeocystin complex (a). Two distinct ligand positions relative to the binding site have been emphasised by means of the colours magenta and cyan. conformers (NACs), electronic pre-organisation, cratic effectsand compactionhave often been described as likely contributing factors, whereas more recent studies highlight the role of pre-transition state tetrel bond formation, chalcogen bondingand CH hydrogen bonding 52 in the process. Although the (pre-)transition state cannot be directly observed in structural studiesthe ternary complexes under investigation do not contain SAM as it would allow the reaction to proceedour atomic-resolution data prompt a detailed analysis of the active site geometry in the presence of two closely related coenzyme analogues. Modelling SAM on the basis of the experimentally determined coordinates of SAH or sinefungin places the transferable methyl group at 1.7-1.8 Å from the fully engaged nucleophilic nitrogen, i.e. equidistant from the nitrogen and the sulfur, with S-C-N angles of 167°and 173°, respectively (Fig.). Thus, the ligand binding mode observed in our structures would be expected to stabilise the transition state and favour shortening of the initial tetrel bond between the nucleophile and the methyl carbon. Consistent with observations made for other methyltransferases, several CHhydrogen bonds link the coenzyme to PsiM (Fig.). On the other hand, we do not observe any chalcogen bondinginteractions between PsiM and the sulfur atom of the coenzyme. Intriguingly, the atomic-resolution structures reveal a number of alternative substrate and product binding modes. Roughly, two types of ligand positioning with respect to the binding site can be distinguished. In the first, the substrate is fully engaged and the nucleophilic nitrogen held in place by hydrogen bonds. In the second, the indole ring is further away from the catalytic site, providing space for the ethylamide to change its conformation and engage in a variety of alternative interactions. This disengaged state is incompatible with methyl transfer, but likely to play a role during initial substrate capture, subsequent deprotonation of the amine and, following the reaction, as a first step towards product release. Our study also sheds a new light on the evolution of SAMdependent methyltransferases and particularly on their ability to adapt to novel substrates. Consistent with the idea that highly variable loop regions inserted into the core Rossmann fold play a crucial role in this process, we find that the SRL of PsiMoriginally a motif interacting with stem-loop structures in RNAdeveloped into a highly specialised lid structure that recognises the small-molecule substrate and sequesters it by closing off the binding site (Fig.). In contrast, the Rossmann fold proper, the substrate-binding pocket and the catalytic centre remained virtually unchanged in the course of evolution. Emergence of PsiM from m 6 A-methyltransferases appears to have been driven by substrate mimicry, ultimately resulting in a moderately efficient small-molecule dimethyltransferase. Although the presence of N247 in PsiM, instead of M274 as in METTL16, enables PsiM to catalyse dimethylation, the second methyl transfer remains less efficient than the first. Moreover, exchange of SAH for SAM requires release of the intermediate product, baeocystin, further hampering the second round of methylation and favouring accumulation of the monomethylated species. This may indicate that the enzyme is caught in local optimum, an "evolutionary cul-de-sac", as a consequence of the inherent limitations of its original scaffold. Consistent with this idea, small-molecule methyltransferases that are known to efficiently di-or trimethylate amines, such as the histidine methyltransferase EgtD, the N,N-8-demethyl-8-amino-D-riboflavin methyltransferase RosAand the phosphoethanolamine methyltransferase PfPMT, possess an altogether different active site architecture without an NPPF motif and are capable of exchanging the coenzyme while the substrate remains bound. With the notable exception of PsiM, the NPPF motif has only been found in N-methyltransferases acting on planar substrates where the target amine is part of a conjugated system. It thus seems plausible that the rigid geometry of the proline-rich sequence helps imposing sp 3 hybridisation, rather than sp 2 , while guiding the lone pair towards the methyl group of SAM. The fact that the amines in the psilocybin precursors norbaeocystin and baeocystin are not part of a conjugated system and, in principle, should not benefit from such a mechanism, is a further indication pointing at the evolutionary origins of PsiM. Reactive oligomers resulting from oxidation of the phenolic hydroxyl group of psilocin, the product of spontaneous dephosphorylation of psilocybin, precipitate proteins and cause intracellular damage. Hallucinogenic mushrooms protect themselves by means of another enzyme of the psilocybin biosynthetic pathway, the promiscuous kinase PsiK, which re-phosphorylates psilocin to psilocybin. Formation of the tertiary amine is advantageous to the fungus as it enhances intramolecular hydrogen bonding between the phenolic hydroxyl group of the dephosphorylated species and the ethylamine. This effect markedly slows down oxidation, enabling PsiK to keep pace with the dephosphorylation process. In the light of these considerations, PsiM's incapacity to perform a third methylation seems consistent with the absence of protective hydrogen bond formation in aeruginascin's dephosphorylated degradation product, 4-hydroxy-N,N,N-trimethyltryptamine. The METTL16 family of methyltransferases was recently discovered and shown to represent a novel class of epitranscriptomic "writer" proteins, which place regulatory mA marks on mRNA. METTL16 was also found to modify various kinds of non-coding RNA, including U6 snRNA. In humans, aberrant RNA modification patterns are frequently associated with disease, leading to significant interest in the emerging field of epitranscriptomics. PsiM and the methyltransferase domain of human METTL16 are closely related and retain 37% sequence identity. Consequently, our results are likely to provide direct insight into the catalytic mechanism of METTL16. While the latter could be crystallised in its apo-formand in complexes containing either SAHor RNA, cocrystals with both ligands remain elusive. Our structures of PsiM can be used as templates to model the corresponding ternary complexes of METTL16. Moreover, in view of the resemblance between the substrate-binding modes of the two enzymes, norbaeocystin and baeocystin may turn out to be effective lead compounds for the design of therapeutic METTL16 inhibitors, or indeed of novel general-purpose nucleotide analogues. Despite the remarkable level of similarity between PsiM and METTL16, several intriguing differences are also apparent. In the first place, the enzymatic activity of METTL16 is tightly controlled, consistent with its role in epitranscriptomic gene regulation. Residues Lys163 and Met167 of the so-called K-loop, which play essential roles in the regulatory mechanism, have not been conserved in PsiM (Supplementary Fig.). Furthermore, all METTL16 proteins carry variable and partially disordered C-terminal extensions, comprising a pair of structured "vertebrate conserved region" (VCR) domains that influence RNAbinding. The VCR domains were also reported to mediate dimerisation, although this has been contested more recently. Finally, METTL16 was shown to bind RNA before it recruits SAM, whereas our results indicate that PsiM must recruit its cofactor before it is able to interact with substrates. It is conceivable that initial binding of RNA to the METTL16 surface alonetriggers SAM recruitment. The presence of the coenzyme might then stabilise the active site and enable it to sequester the target nucleotide. Accurate three-dimensional protein structures are key assets in bioengineering, enabling rational modification of enzymes and their substrates. Inspection of the substrate-binding pocket of PsiM reveals that moderately sized substituents at positions 6 and 7 of the indole ring are likely to be tolerated by the wild-type enzyme. Substitutions at the remaining ring positions, on the other hand, are likely to require substantial remodelling of the active site through mutagenesis. The phosphate moiety seems especially important for substrate recognition and presumably cannot be deleted or moved to a different position on the ring without severely affecting binding affinity. These structure-based findings are confirmed by recent experiments involving alternative substrates in psilocybin-producing E. coli strains. As the relevance of psilocybin as a future medication against a growing number of common mental health conditions is becoming clear, we expect that the atomic-resolution structures of PsiM will constitute important tools in bioengineering efforts aimed at producing novel analogues with improved therapeutic properties.