Methyl transfer in psilocybin biosynthesis

Consistent with predictions^18,19, the structure of PsiM (Fig. 2a, b) comprises a Rossmann fold characteristic of class I SAM-dependent methyltransferases²³. This fold consists of 6 parallel β-strands, interlaced with α-helices and followed by a single antiparallel strand. The resulting 7-stranded β-sheet exhibits the typical 3-2-1-4-5-7-6 topology (Fig. 2b). Other hallmarks of class I methyltransferases include a glycine-rich β-turn (residues 107–109, sequence GTG) that wraps around the coenzyme²⁴ and an acidic residue (Glu131) engaged in hydrogen bonds with both hydroxyl groups of the ribose moiety²⁵. In addition to the Rossmann fold, PsiM features an N-terminal jaw-like domain also seen in some other methyltransferases²⁶, as well as unique N- and C-terminal extensions.

Inspection of the active site and the contacts between the canonical NPPF motif ²³, SAH and norbaeocystin (Fig. 2c) provides immediate insight into key aspects of the catalytic mechanism. The amino group of norbaeocystin is held in place by two hydrogen bonds, one involving the backbone carbonyl of Pro184 (N-O distance: 2.8 Å) and the other with the side-chain oxygen of Asn183 (N-O distance: 2.9 Å). The geometry of these interactions matches the tetrahedral (sp³) coordination of the nitrogen and directs its lone pair towards the space where the methyl group of SAM would be located. Together with the polarisation caused by the hydrogen bonds, this arrangement poises the nitrogen atom for its nucleophilic attack on the methyl group. Consistent with a key role of the Asn183 side chain in the process, the N183A mutation renders PsiM catalytically inactive¹⁸.

Consistent with earlier predictions¹⁸, the region following strand β4 (residues 189–221), which lacks homology to other methyltransferases, acts as a substrate recognition loop (SRL). The SRL forms a highly contorted structure, which is stabilised by the interaction with norbaeocystin and by intramolecular hydrogen bonds that facilitate turns. For instance, the side chain of Ser196, which is required for enzymatic activity¹⁸, stabilises the conformation of the main chain by forming a hydrogen bond with the backbone amide of Ala198. The SRL not only assists in recognising the substrate, but also forms a barrier that physically closes off the binding pocket, locking the substrate in the interior of the enzyme. Consequently, part of the polypeptide chain must be able to make way to allow the substrate to enter the active site, as well as the reaction product to leave. To gain insight into this process, we crystallised the PsiM-SAH complex in the absence of substrate. A tetragonal crystal form was obtained that contained two PsiM molecules per ASU and diffracted to 2.5 Å (Supplementary Fig. 2 and Supplementary Table 1). Although structural differences with respect to the ternary complex are minimal, part of the SRL (residues 198–207 and 197–209 in chains A and B, respectively) is no longer observed in the electron density, suggesting disorder in the absence of the substrate. The result is a wide opening, sufficiently large to render the active site accessible to small organic molecules.

To investigate the enzymatic reaction cycle in more detail we made use of sinefungin, a general methyltransferase inhibitor that mimics SAM. The 0.89 Å structure of the PsiM-sinefungin-norbaeocystin complex (Fig. 5b) provides insight into the state of the active site immediately preceding the first methyl transfer. In the absence of a nucleophilic attack, the amino groups of norbaeocystin and sinefungin — the latter occupying the position of the methyl of SAM — are engaged in a hydrogen bond (N-N distance: 2.9 Å). In order to accommodate the length of the hydrogen bond, which exceeds that of a covalent sulfur-carbon bond by approximately 1.3 Å, sinefungin is forced into an energetically unfavourable conformer that is not observed with SAH. An alternative state (36% occupancy) is observed where sinefungin adopts a less constrained conformation, very similar to the one seen for SAH. This, however, forces the norbaeocystin amino group away from the active site, towards a different position that allows it to form an equivalent hydrogen bond with the inhibitor. The hydrogen bonds with Asn183 and Pro184 on the other hand are lost. In view of the similarity between sinefungin and SAM it is conceivable that an analogous dynamic equilibrium (involving CH-N instead of NH-N hydrogen bonds) exists in the PsiM-SAM-norbaeocystin complex, immediately after substrate binding and prior to the actual nucleophilic attack or the formation of a pre-transition state CH-N tetrel bond²⁷.

To characterise the state of the active site directly after the first methyl transfer, we cocrystallised PsiM with SAH and the monomethylated product, baeocystin. A well-defined 0.93 Å structure was obtained that revealed unambiguous electron density for both ligands (Fig. 5c). Interestingly, the monomethylated amino group of baeocystin appears to immediately retract from the transition state position to occupy a pair of alternative positions (with 60% and 40% occupancies). One of the alternative positions involves a movement of the indole ring away from the coenzyme, resulting in a C-S distance of 3.4 Å. This movement may constitute a first step towards product release.

The transient state immediately preceding the second methyl transfer was investigated by analysing a PsiM-sinefungin-baeocystin complex (Fig. 5d). In the resulting 0.92 Å structure, the methyl group of baeocystin points towards the side chain of Asn183, which creates the necessary space by rotating away from the catalytic site. This leads to a 1.3 Å displacement of the Asn183 oxygen atom compared to the complex with sinefungin and norbaeocystin, where the oxygen is engaged in a hydrogen bond with the substrate amine. The only conceivable alternative, a conformation with the methyl group pointing towards the carbonyl of Pro185, is not observed at all, presumably because the constrained backbone of the NPPF motif cannot move aside. Intriguingly, we find that substrate occupancy is only 61% in this complex. A large portion of the SRL (residues 198–208), Arg281, as well as numerous associated water molecules also show partial occupancy, consistent with substrate-dependent folding. As all ternary complexes were crystallised in the exact same manner, these observations strongly suggest that, in the presence of the SAM analogue sinefungin, the affinity of PsiM for baeocystin is lower than that for norbaeocystin. This difference in affinity was confirmed by ITC experiments (Supplementary Fig. 3). Furthermore, chromatographic analyses of typical in vitro assays demonstrate that the monomethylated intermediate baeocystin accumulates and becomes the dominant species before psilocybin formation sets in (Fig. 4a). Results of quantitative experiments (Fig. 4b) can be explained by Michaelis-Menten kinetics and indicate that the first methyl transfer is approximately twice as fast as the second, with k_cat values of 0.11 ± 0.01 min⁻¹ and 0.06 ± 0.01 min⁻¹. The Michaelis constants (K_m) for norbaeocystin and baeocystin are 575 ± 100 µm and 492 ± 154 µm. These values translate into catalytic efficiencies of 0.191 and 0.122 nM min⁻¹ for norbaeocystin and baeocystin, respectively. Taken together, these observations demonstrate that PsiM is a comparatively poor dimethyltransferase.

We also crystallised PsiM in the presence of the final reaction products, SAH and psilocybin. The corresponding 0.94 Å structure reveals two conformations of the hallucinogen, which only differ in the orientation of the dimethylated amine (Fig. 5e). One conformation (46% occupancy) likely reflects the situation following the second transfer reaction, with the methyl carbon at 3.1 Å from the sulfur of the coenzyme and the nitrogen maintaining its hydrogen bond to the carbonyl of Pro184. In the other conformation (54% occupancy), the dimethylated amine is rotated by approximately 180°, allowing the nitrogen, now facing away from the carbonyl, to form an intramolecular hydrogen bond with the phosphate of psilocybin. This state represents a likely first step towards the release of psilocybin from the enzyme.

Superposition of the five atomic-resolution crystal structures reveals that the protein models are nearly identical, with pairwise C_α-RMSD values between 0.10 Å and 0.32 Å (Fig. 5f). In contrast, the positioning of the ligand with respect to the substrate-binding cavity varies considerably. By comparing all of the crystal structures and alternative conformers, two principal binding modes can be distinguished, which are related by a rotation of the ligand around the phosphate moiety, roughly within the plane of the indole ring (Fig. 5f). In one of the binding modes, the ethylamide nitrogen is positioned at the catalytic centre and hydrogen bonded to the Pro184 carbonyl and, in the case of norbaeocystin, to the side chain of Asn183. The second binding mode corresponds to a disengaged state with the indole ring positioned further from the catalytic centre, providing space for the ethylamine chain to adopt a variety of conformations depending on available contacts. The second binding mode is observed in both post-reaction complexes (i.e. in the presence of methylated products and SAH) and in the complex containing SAH and norbaeocystin. We propose that the disengaged state provides an exit route for product molecules, as well as an initial substrate binding mode that favours deprotonation of the ethylamide before the transition state is formed.

Results of a sequence-based phylogenetic analysis are shown in Fig. 6b and Supplementary Fig. 4. Consistent with the similarity that was detected at the structural level, the methyltransferases most closely related to PsiM belong to the METTL16 family, which itself branches off from the more ancient ribosomal 23 S m⁶A-methyltransferases. These findings seem highly surprising, given that the ancestral transferases are regulatory proteins, invariably act on RNA and, perhaps most intriguingly, only transfer a single methyl group to their target amine, whereas PsiM dimethylates its substrate. A likely explanation for the unexpected evolutionary relationship between PsiM and METTL16 is provided in Fig. 6c, which shows a superposition of the substrate-binding pockets of the PsiM-SAH-norbaeocystin and METTL16-mRNA complexes. The interaction of PsiM with its phosphorylated substrate norbaeocystin shows a striking resemblance to the binding of METTL16 to the target RNA nucleotide, with the phosphate groups occupying nearly identical positions and the indole ring mimicking the adenine base. Thus, substrate mimicry presumably explains how PsiM arose from m⁶A-monomethyltransferases through a process of evolutionary tinkering³⁶.

Various authors describe aeruginascin (Fig. 1), the quaternary ammonium derivative of psilocybin, as a mushroom metabolite^20–22,37. However, an inconsistent picture has emerged as to whether aeruginascin is a product of the psilocybin biosynthetic pathway. While the trimethylated compound was detected in transgenic psilocybin-producing yeast¹⁶, it was not present in engineered E. coli and A. nidulans strains expressing the psi genes^14,15. In vitro assays did not yield aeruginascin either¹⁰. We sought to confirm PsiM's apparent incapacity for a third N-methyltransfer by in vitro biochemical assays. First, optimal reaction conditions were established. Using the first methylation reaction as readout, optimum turnover was found in a window between 20–30 °C and at a pH 8.4, which may favour a third methylation, analogous to the trimethylating l-lysine methyltransferase of Neurospora crassa³⁸. Subsequently, we performed comparative in vitro assays to monitor methylation of norbaeocystin, baeocystin, and psilocybin by PsiM (Supplementary Fig. 5). When norbaeocystin or baeocystin were added as acceptor substrates to PsiM, psilocybin formation was observed as expected. In contrast, none of the reactions led to detectable amounts of aeruginascin. The inability of PsiM to perform trimethylation most likely indicates that the proline-rich NPPF motif lacks sufficient flexibility to accommodate a dimethylated substrate, which would require main-chain conformational changes and a shift of the Pro184 carbonyl.

The psiM gene of P. cubensis (GenBank entry ASU62238.1) was amplified with primers oJF105 and oJF47 using the plasmid pFB13 as a template¹⁰. The PCR product was cloned into pET-28a (Novagen) via the NheI and XhoI restriction sites. The resulting expression vector, pJF45, which was verified by sequencing, encodes full-length PsiM preceded by a hexahistidine tag and recognition sites for both thrombin and TEV protease. E. coli BL21 (DE3) was transformed with this vector and precultures grown overnight from single colonies at 37 °C in LB medium supplemented with 50 µg/ml kanamycin and 1% glucose. Expression cultures, prepared by diluting precultures 100-fold in fresh LB medium containing 50 µg/ml kanamycin and 1% glucose, were incubated at 37 °C until they reached an OD (600 nm) of 0.45. IPTG was added to a final concentration of 1 mm, followed by a further 3 h incubation at 37 °C. Upon centrifugation (60 min at 10,000 × g, 4 °C), cells from 1.2 l of culture were resuspended in 70 ml lysis buffer (20 mm Tris/HCl pH 8.0, 400 mm NaCl, 0.5 mm DTT) and disrupted by sonication on ice using a Sonopuls HD 2070 (Bandelin Electronic GmbH, Berlin, Germany). Insoluble cell debris was removed by centrifugation for 1 h at 40,000 × g and 4 °C. The cleared lysate was passed through a 0.45 µm filter (Whatman) and applied to a 5 ml metal affinity chromatography Ni²⁺-cartridge (GE Healthcare). Following extensive washing, PsiM was eluted using a 0–300 mm imidazole gradient in lysis buffer and further purified by gel filtration (Superdex 200, GE Healthcare GmbH, Munich, Germany) in a buffer containing 20 mm Tris/HCl pH 8.0, 400 mm NaCl and 1 mm DTT. The protein appeared as a single band on Coomassie-stained SDS-PAGE gels at this stage. To proteolytically remove the hexahistidine-tag, the protein solution (10 ml total volume, 10 mg/ml) was dialysed for 12 h at 4 °C in a G2 Slide-A-Lizer (10 kDa MWCO, ThermoFisher Scientific) against 2l digestion buffer (20 mm Tris/HCl pH 8.4, 150 mm NaCl, 2.5 mm CaCl₂), followed by the addition of 5 units of thrombin (Sigma) directly into the Slide-A-Lizer and continued dialysis for 12 h at 22 °C against 2 l of fresh digestion buffer. The protein was then dialysed for 12 h at 4 °C against 2 l protein storage buffer (20 mm Tris/HCl pH 8.0, 1 mm DTT) and concentrated to 31 mg/ml (0.87 mm) using a Vivaspin 20 ultrafiltration unit (10 kDa MWCO, Sartorius Lab Instruments GmbH, Göttingen, Germany). Protein concentration was determined spectroscopically, using a sequence-based estimate of the molar absorption coefficient at 280 nm (39400 m⁻¹ cm⁻¹). A PsiM-SAH complex was formed by adding a 2-fold molar excess of the coenzyme (Sigma). PsiM-SAH-(nor)baeocystin and PsiM-sinefungin-(nor)baeocystin complexes were prepared by adding a 2-fold molar excess of either the coenzyme or the inhibitor and a 5-fold molar excess of substrate (Usona Institute, Madison, WI, USA). Aliquots of the protein and the protein-ligand mixtures were flash-frozen in liquid nitrogen and stored at −80 °C until further use.

The PsiM-SAH-norbaeocystin complex was crystallised at 4 °C using the hanging drop method. Drops were produced by mixing equal volumes of the protein solution and the reservoir buffer, which contained 100 mm Tris/HCl pH 8.5, 20% PEG 8000 and 200 mm MgCl₂. After a week, morphologically distinct orthorhombic (P2₁2₁2₁) and monoclinic (C2) crystals began to appear, frequently within the same drop. The PsiM-SAH-baeocystin, PsiM-sinefungin-(nor)baeocystin and PsiM-SAH-psilocybin complexes were produced using the same approach, but cross-seeding was used to ensure growth of the better diffracting orthorhombic form. The PsiM-SAH complex was crystallised using the sitting drop method and a reservoir solution containing 100 mm bis-Tris propane pH 7.0, 30% PEG 300 and 15% PEG 1000. Tetragonal (P4₃) crystals of this complex appeared after several weeks.

Crystals were directly flash-cooled in liquid nitrogen, without the addition of further cryoprotectants. For X-ray diffraction experiments, a temperature of 100 K was used. All datasets were collected at ESRF ID23-1⁶⁶ and processed using xia2⁶⁷ with DIALS^68,69, with the exception of the data for the PsiM-SAH complex, which were collected at ESRF ID23-2⁷⁰ and processed using XDS^71,72. Structures were solved by molecular replacement using Phaser⁷³, in combination with the CCP4 suite⁷⁴. A poly-Ala model generated from PDB entry 6B92 was used as the search model to solve the orthorhombic crystal form of the PsiM-SAH-norbaeocystin complex. The refined structure was subsequently used as a search model to solve all other crystal forms. Coot⁷⁵ and Phenix⁷⁶ were used for iterative rounds of model building and refinement. Data collection and refinement statistics can be found in Supplementary Table 1.

All molecular graphics illustrations were produced with PyMOL 0.99rc6 (DeLano Scientific LLC, Palo Alto, CA, USA). The schematic overview of protein-ligand interactions shown in Fig. 3b was prepared using LIGPLOT⁷⁷. Structural comparisons were performed using the SSM superposition method implemented in Coot⁷⁵.

To produce hexahistidine-tagged PsiM without thrombin and TEV protease sites, E. coli KRX (Promega) was transformed with pFB13¹⁰. A 5 ml culture was grown overnight at 37 °C in LB medium containing 50 µg/ml kanamycin and then used to inoculate 500 ml of 2 × YT medium with 50 µg/ml kanamycin. This culture was incubated at 37 °C until an OD (600 nm) of 0.6 was reached. Upon addition of l-rhamnose to a final concentration of 0.1%, the culture was transferred to 16 °C and incubated for another 20 h. The biomass was collected by centrifugation at 4000 × g and 4 °C, resuspended in 7.5 ml lysis buffer (50 mm phosphate buffer, 500 mm NaCl, 20 mm imidazole, pH 8) and lysed using a Sonopuls GM 3200 sonicator (Bandelin Electronic GmbH, Berlin, Germany). To remove cell debris, the lysate was centrifuged (10,000 × g, 20 min, 4 °C). The cleared supernatant was passed through a 0.45 µm filter (Whatman) and purified using a 1 ml HisTrap HP column (GE-Healthcare). Following extensive washing, PsiM was eluted using a 0–250 mm imidazole gradient in lysis buffer. Purification was verified by SDS-PAGE (Supplementary Fig. 6) and the protein quantified photometrically at 280 nm. For subsequent use in activity assays, the enzyme was subjected to buffer exchange on a PD-10 column (Cytiva) eluted with 50 mm Tris (pH 8.4) or the respective buffers used to determine the pH optimum (below).

Using pJF45 as the template, a gene encoding the PsiM N247M mutant was generated by overlap extension PCR⁷⁸ and inserted into pET-28a (Novagen) by means of standard cloning techniques. The resulting construct was verified by sequencing. Expression and purification followed the same protocol as described above for PsiM.

Norbaeocystin, baeocystin and aeruginascin were obtained from the Usona Institute (Madison, WI, USA). Psilocybin was synthesised as described¹³. Methyltransferase activity assays were set up as triplicates in 50 mm Tris buffer (pH 8.4) in a total volume of 50 µl. The reactions were composed of 600 µm SAM and 200 µm acceptor substrate (norbaeocystin, baeocystin, or psilocybin). The reaction was started by adding PsiM (1 µm final concentration) and incubated at 28 °C (or between 14 and 50 °C to record the temperature optimum). Negative controls were run with heat-inactived enzyme. The assays were stopped after 2 and 24 h (60 min for temperature optimum) in liquid nitrogen. To record pH optima, the reactions were set up in the following buffers (each 50 mm): MES (pH 6), phosphate (pH 7) Tris (pH 8–8.4), glycylglycine (pH 8.9), CAPSO (pH 9.5–11). To record kinetic parameters, the reaction tubes were prewarmed to 28 °C before the prewarmed reaction mixture was added. The reactions for kinetics were set up in Tris buffer (pH 8.4) as well, but contained 100 µm – 1 mm acceptor substrate, 1 mm SAM, and 500 nm PsiM. Reactions were stopped in liquid nitrogen after 60, 90, 120, 180, 300, and 600 s. The samples were lyophilised, resuspended in 50 µl methanol and subsequently placed in an ultrasonic bath for 3 min, followed by centrifugation.

Chromatographic analysis was performed on an Infinity II UHPLC-MS instrument (Agilent, Waldbronn, Germany) interfaced with a 6130 single quadrupole mass detector and electrospray ionisation source and equipped with an Ascentis Express 90 Å F5 column (100 × 2.1 mm, 2.7 µm particle size, kept at a constant temperature of 50 °C). Solvent A was 0.1% formic acid in water, solvent B was methanol. For elution, the following gradient was applied at a flow of 0.8 ml/min (0–2.5 min) and 0.6 ml/min (2.5–3.5 min): hold at 2% B (0–2 min), linearly increase to 20% B over the next 0.5 min, hold at this percentage for another 1 min.

ITC was performed at 25 °C in Tris buffer (50 mm, pH 8.4) using a Malvern Microcal PEAQ-ITC microcalorimeter. Titrants and protein concentration were: (i) 750 µm (norbaeocystin, baeocystin, SAH) to 50 µm PsiM, or (ii) 750 µm (norbaeocystin, baeocystin) to 1 mm sinefungin + 50 µm PsiM, or (iii) 750 µm (norbaeocystin, baeocystin) to 2 mm SAH + 50 µm PsiM, or (iv) Tris 50 mm buffer pH 8.4 to 50 µm PsiM. Results were analysed using the MicroCal PEAQ-ITC Analysis Software, assuming a single-site binding model.

Database searches to identify PsiM and METTL16/RlmF homologues were performed using the BLAST³² server at NCBI³³. The proteomes of representative species in the non-redundant protein sequence (nr) database were searched for sequences similar to P. cubensis PsiM and H. sapiens METTL16, respectively. A multiple sequence alignment was then produced with the help of MAFFT^79,80. Phylogenetic analysis was carried out using the maximum-likelihood (ML) approach and JTT matrix-based model⁸¹ in MEGA11⁸². Initial trees for the heuristic search were obtained automatically by applying the Neighbour-Join and BioNJ algorithms to a matrix of pairwise distances estimated using the JTT model, and by subsequently selecting the topology with optimal log likelihood value. A discrete Gamma distribution was used to model evolutionary rate differences among sites, allowing for 5 categories. All positions with less than 80% site coverage were eliminated, i.e., fewer than 20% alignment gaps were allowed at any position (partial deletion option). There were a total of 280 positions in the final dataset. Bootstrap analysis was performed with 1000 repeats.

A list of reagents and oligonucleotide primers is provided in Supplementary Table . 2

Further information on research design is available in the linked to this article. Nature Portfolio Reporting Summary

Supplementary Information Peer Review File Reporting Summary

Source Data