Science News

Structural basis for psilocybin biosynthesis

March 22, 2025

The PsiD proenzyme undergoes self-cleavage

To uncover the catalytic mechanism of PsiD, we expressed the full-length PsiD (1–439) protein from Psilocybe cubensis and determined its crystal structure using the molecular replacement method with the model predicted by AlphaFold2 (Fig. 1b, Supplementary Table 1, Supplementary Fig. 1a)^24,31. There are six PsiD molecules in the asymmetric unit that have similar conformations, with an RMSD of about 0.2–0.3 Å over equivalent backbone atoms (Supplementary Fig. 1b). However, the density of the N-terminal ~53–57 residues could not be modeled in all these molecules (Supplementary Fig. 1a–h); therefore, we selected one of the PsiD molecules containing residues 53–438 (hereafter named apo-PsiD) for further analysis (Supplementary Fig. 1c).

Consistent with previous observations in studies of the E. coli phosphatidylserine decarboxylases (PSDs)^32,33,34, which indicate that the proenzyme undergoes a self-cleavage reaction and can be divided at a conserved “LGST” motif by a canonical serine protease mechanism^35,36, the PsiD enzyme exhibited self-cleavage between a “GGSS” (G401-G402-S403-S404) motif, leaving a pyruvoyl prosthetic group within the N-terminal of the α-chain, which was then named Pvl403, and a β-chain containing a Gly402 terminus (Fig. 1c, d). The production of the active Pvl403 group was confirmed by our mass spectra results (Supplementary Fig. 2a). The self-cleavage reaction mechanism of the PsiD protein can be deduced from previous studies of the PSDs, the biochemical characterization and the in-silico modeling structure of P. cubensis PsiD^30,34. Firstly, the peptide bond between Gly402 and Ser403 undergoes serinolysis attacked by the side chain hydroxyl group of Ser403, forming an ester bond between the two amino acids. Then, an α, β-elimination reaction will break the linkage between the two residues, and the β-subunit will be released, leaving a dehydroalanine at the N-terminus of the α-subunit; a water molecule will further attack the double bond of the dehydroalanine intermediate to yield the α-hydroxyl alanyl residue, which will subsequently form the mature pyruvoyl moiety by releasing ammonia^34,37,38 (Fig. 1e).

The overall structure of PsiD

The α and β chains of each PsiD protomer in the asymmetric unit form a globular fold that consists of 16 α-helices and 13 β-strands (Fig. 1d, Supplementary Fig. 2b). The Gly402 at the C-terminus of the β chain is positioned ~11 Å away from the pyruvoyl group, as predicted by the published AlphaFold2 model (Fig. 1f)³⁰. This conformational change results in the exposure of the pyruvoyl group for further catalytic reactions. The putative active site is surrounded by four structural elements mainly composed of hydrophobic residues, including the Loop 1 region (aa. 108–120), the Loop 2 region (aa. 290–297), the Loop 3 region (aa. 324–341), and the Loop 4 region (aa. 371–377) (Fig. 1g, h).

Further analysis revealed that the PsiD molecules might form a homodimer that adopts central symmetry with an extensive dimer interface (Supplementary Fig. 3a). Half of the dimer interface was then selected to show these contacts (Supplementary Fig. 3b). The Asp251, Glu281, Gln253, Ser254, Lys261, and Asp256 from one molecule formed various hydrogen bonds with Asp71, Arg137, Arg135, Gln136, Asn139, and Thr134 from another molecule, either directly or mediated by water molecules. Meanwhile, Thr257, Val259, and Phe260 from one molecule had hydrophobic contacts with Thr134, Ile348, Phe130, Pro321, and Tyr344 from another molecule. The residues involved in dimer formation are all located within the β-chain (Supplementary Fig. 3c). Following our observations, Arg135 and Gln136 were mutated (R135A/Q136A) to assess the in-solution status of PsiD through gel filtration, with wildtype PsiD and S403A serving as controls (Supplementary Fig. 3d). However, our evaluation indicated that both wildtype PsiD and the PsiD mutants predominantly exist as monomers, with only a minor fraction displaying the homodimer form. Additionally, the R135A/Q136A mutation did not significantly alter the in-solution status of PsiD. These results suggested that the PsiD protein mainly exists as a monomer in the solution state, while some also form a homodimer state.

Structure of the Schiff-base intermediate of PsiD with L-tryptophan

To further investigate substrate recognition and the catalytic mechanism, we determined the structure of PsiD complexed with the substrate L-tryptophan through co-crystallization (Supplementary Table 1). Hereafter, we refer to this structure as PsiD-Trp. There are also six PsiD molecules in the crystallographic asymmetric unit, showing minor differences (Supplementary Fig. 4a–g). Thus, we selected one of them to compare with the apo-PsiD structure (Fig. 2a, Supplementary Fig. 4b). Structural comparisons between the PsiD-Trp complex and the apo-PsiD revealed no significant conformational changes in the PsiD structure upon substrate binding, as demonstrated by the RMSD of about 0.24 Å between these two structures over equivalent backbone atoms (Fig. 2a).

**Fig. 2: The substrate-binding properties of PsiD.**

In accordance with previous observations that the pyruvoyl prosthetic group forms a Schiff-base intermediate with the primary amine of the substrate during catalysis^32,33,34, the Schiff-base intermediate between Pvl403 and the amino group of L-tryptophan was captured in our PsiD-Trp structure (Fig. 2b). Meanwhile, the carboxyl group of L-tryptophan could not be modeled, indicating that the structure of PsiD in complex with L-tryptophan presents a covalently bound intermediate state with tryptamine. In this structure of the PsiD-Trp complex, the methyl group of Pvl403 was hydrophobically anchored by the side chains of Phe398, His296, and Leu290. The covalent amide was hydrogen-bonded to the main chain of Leu290. The only hydrophilic group in the indole ring of tryptamine forms hydrogen bonds with the side chain of Thr374 and the main chain of Met373. The indole ring was enclosed by Phe289, Val292, Leu339, Leu112, Tyr338, Tyr117, Pro114, Met373, Thr374, and Ile376, thereby stabilizing the intermediate product (Fig. 2b, c). These residues were almost all within the pocket-composing loops (Fig. 2d). Structure superposition suggested minor deviations in the catalytic pocket between the apo-PsiD and the PsiD-Trp complex. Only two residues had slight conformational differences between these structures: His296 and Val292, which are closer to the indole ring when PsiD binds to the intermediate product (Fig. 2e).

Decarboxylase activity assay

To examine the functional roles of the residues involved in substrate binding, we made several PsiD mutants and measured their decarboxylase activities using the thin-layer chromatography (TLC) method. Before doing that, we first tested the catalytic activity of the wildtype PsiD using a TLC experiment and validated the catalytic product through mass spectrometry analysis (Fig. 2f, g, Supplementary Fig. 5a). Based on these results, we then performed the TLC assays to test the activity of the PsiD mutants (Fig. 2h, i, Supplementary Fig. 5b). In vitro decarboxylation activity assays revealed that the L112A and F289A mutations did not decrease the decarboxylase activity of PsiD but enhanced its catalytic activity. These observations may be attributed to weak hydrophobic interactions with the substrate; in contrast, Y117A reduced the activity. On the other hand, L290A almost completely lost its decarboxylase activity for L-tryptophan, indicating the key role of Leu290 in producing and stabilizing the pyruvoyl group. The Tyr338, necessary for hydrophobic contact with the substrate, also shows slightly increased activity when mutated to alanine. The Thr374, which forms a hydrogen bond with the L-tryptophan molecule, did not dramatically impact the catalytic ability of PsiD; as a negative control, the mutants of His296 and Ser403, which are indicated to be essential for the self-cleavage reactions^30,39, abolished both self-cleavage and L-tryptophan decarboxylase activity when replaced with alanine (H296A and S403A) (Fig. 2g–j, Supplementary Fig. 5b), while the Gly402 mutant (G402A), located at the terminus of the β-chain and not involved in covalent bond formation, showed similar activity to the L-tryptophan substrate as the wildtype did (Fig. 2g–j, Supplementary Fig. 5b). The results mentioned above validated the roles of the residues we identified in the catalytic pocket for decarboxylase activity. Unfortunately, although we tried to purify the other mutations involved in substrate binding and pyruvoyl group formation, such as F398A, M373A, and P114G, these mutants were not stable enough to be purified.

The N-terminal of PsiD revealed a self-inhibition mechanism

Based on the above enzymatic assays, a confusing observation was that the wildtype PsiD protein exhibited insufficient catalytic activity toward the substrate, leaving a certain amount of L-tryptophan not fully reacted (Fig. 2f, Supplementary Fig. 5a, c). To figure out why the recombinant full-length PsiD did not have strong enough activity, we performed the decarboxylase assay using TLC methods with gradient pH values (Tris pH 7.4-9.0) according to our protein purification and crystallization conditions. However, the different pH values showed a minor impact on the catalytic activity of the PsiD protein under our experimental conditions (Supplementary Fig. 5d). In contrast, previous studies suggested that the optimal condition for PsiD activity occurred in sodium phosphate buffer (pH 6.6) and between 33 and 36 °C³⁰. These distinctions may explain the capture of the covalent-bound tryptamine product state in our structure of the PsiD-Trp complex.

To explore whether other factors exist that cause the insufficient catalytic activity of PsiD, we reanalyzed its N-terminal region (aa. 1–55), which remained invisible in both the structures of apo-PsiD and the PsiD-Trp complex. A structure superposition between the AlphaFold2-predicted full-length PsiD and our PsiD-Trp structures showed that the N-terminal region might not impact the dimerization of the PsiD protein (Supplementary Fig. 6a, b). However, the N-terminal region contains three alpha-helices that lie on the surface of the solved part of the PsiD structure. Moreover, the first helix covers the substrate-binding pocket of PsiD, with a tryptophan (Trp27) residue inserted into the substrate-binding pocket, sharing the same set of residues for substrate recognition (Fig. 3a, b). These analyses strongly suggest a cryptic self-inhibition mechanism in the PsiD protein that inhibits the enzyme’s activity. Therefore, we chose to construct two additional truncations, PsiD (aa. 50–439) and PsiD (aa. 30–439), as well as a W27A mutant of the full-length PsiD to test their catalytic activity. The S403A was selected as a negative control (Fig. 3c–f, Supplementary Fig. 7a–e). The TLC results revealed that the PsiD (1–439) W27A mutant, as well as the PsiD (30–439), showed significantly higher catalytic activities for the L-tryptophan substrate than the wildtype PsiD enzyme (Fig. 3c–e, Supplementary Fig. 7a–c). Furthermore, deleting the N-terminal 50 residues dramatically increased the catalytic activity of the PsiD protein for the L-tryptophan substrate (Fig. 3c, f, Supplementary Fig. 7d). Incubating these different constructs with L-tryptophan also validated these results; all the L-tryptophan substrate was converted to tryptamine in PsiD (1–439) W27A, PsiD (30–439), and PsiD (50–439), but not in the wildtype PsiD under the same conditions (Fig. 3g, Supplementary Fig. 7f, g). Accordingly, none of these truncations or the W27A impact the self-cleavage activity of PsiD (Fig. 3h).

**Fig. 3: The self-inhibition mechanism of PsiD.**

We then co-crystallized PsiD (50–439) with L-tryptophan and successfully obtained the structure at a high resolution of 1.6 Å (Supplementary Table 1, Supplementary Fig. 8a). Structural comparison of the PsiD (50–439)-Trp complex revealed no significant differences from the PsiD-Trp complex (Supplementary Fig. 8b). However, in contrast to the tryptamine in the latter complex, which is presented in a covalently bound state, the former complex shows a pre-release state when superimposing the catalytic centers of these two structures (Fig. 3i, j). These results demonstrate that the PsiD enzyme has a self-inhibition mechanism through its N-terminal region, particularly the Trp27 residue, thereby regulating its activity. Deleting the N-terminal region will dramatically improve decarboxylation efficiency.

The working model and reaction scheme of PsiD

Searching for the structural homolog of PsiD using the Dali online server identified only one homologous protein, PSD, from E. coli⁴⁰. Therefore, we compared the structures of EcPSD and PsiD to elucidate their distinctions and similarities in substrate recognition. The detailed descriptions of the comparison results are provided in Supplementary Note 1 (Supplementary Figs. 9 and 10). Drawing from our structural studies, functional assays, and biochemical characterization from earlier research³⁰, along with the reaction scheme of EcPSD³⁸, we proposed a working model and catalytic mechanism for the biosynthesis of tryptamine catalyzed by PsiD (Fig. 3k): (a) In the catalytically inactive state, the N-terminus of PsiD covers the surface region above the catalytic pocket with Trp27 inserted, leading to the self-inhibition of the PsiD enzyme; (b) Then, the PsiD proenzyme undergoes self-cleavage between Gly402 and Ser403, producing the mature enzyme with a catalytically active pyruvoyl prosthetic group (Pvl403) at the N-terminal of the α-chain; (c) During the catalytic process, the L-tryptophan substrate triggers conformational changes in the N-terminal, at least competing with the Trp27 residue to enter the catalytic center. The detailed catalytic reaction scheme can be deduced from the PSDs^34,38: (i) First, the hydrophobic indole ring of L-tryptophan is attracted into the pocket and binds to the active site via a Schiff-base formed between the primary amine of L-tryptophan and the α-carbonyl carbon of Pvl403; (ii) The electron rearrangement favors decarboxylation, and the formation of an azomethine intermediate; (iii) Subsequent protonation generates another intermediate product in Schiff-base linkage with the enzyme; (iv) The addition of a water molecule across the Schiff-base regenerates the pyruvoyl prosthetic group and liberates tryptamine from the active site; (d) Finally, the enzyme may regenerate to catalyze a new round of reactions.

The overall structure of PsiK

Next, to uncover the phosphorylation mechanism of PsiK, we solved the structure of the PsiK complex with the cofactor ADP and tryptamine, which were used to capture the substrate-bound state of PsiK (Fig. 1a, Supplementary Table 1)²⁴. In addition, two Mg²⁺ ions derived from the crystallization conditions were observed alongside the phosphate groups of the ADP molecule (Fig. 4a). The PsiK molecule contains a β-strand-rich smaller N-lobe (aa. 1–118) and a larger helix-rich C-lobe (aa. 126–362), with a total of nine β-strands and twelve α-helices (Fig. 4a). The cofactor ADP and the substrate were located in a long cleft beneath the N-lobe, with two metal ions situated between the cofactor and substrate (Fig. 4b).

**Fig. 4: Structure of 4-hydroxytryptamine kinase PsiK.**

Searching the structure of homologous proteins of PsiK revealed that the PsiK enzyme structurally resembles protein kinase-like phosphotransferases⁴⁰, especially the 5-Methylthioribose (MTR) kinases from A. thaliana and B. subtilis, with Z-scores of 32.4 and 31.3, and C_α RMSD values of 2.8 Å and 3.1 Å, respectively^41,42 (Supplementary Fig. 11a, b). Nevertheless, PsiK shares only ~16% sequence identity with these two MTR kinases. Previous studies have uncovered four loop regions in these MTR kinases that play critical roles in ADP binding, divalent cation chelation, and substrate recognition, including a glycine-rich loop (G-loop) between β1 and β2, an HGD catalytic loop, an Mg²⁺ binding DXE motif, and a semi-conserved W-loop^41,42 (Supplementary Fig. 12). In our PsiK structure, there are four loop regions and a linker that correspond to these regions in the MTR kinases, playing a role in substrate and cofactor binding (Fig. 4c). Based on these analyses, we further inspected the cofactor binding, divalent cation chelation, and substrate recognition properties of the PsiK protein.

Cofactor binding and substrate recognition properties of PsiK

During the phosphorylation process, the co-substrate ATP donates the γ-phosphate group to the substrate and produces the ADP molecule. The ADP molecule in our structural complex is clamped in a cleft between two β-sheet regions composed of β1-β5 and β7-β8 from the two lobes of PsiK (Fig. 4d). The adenine ring of the ADP points to the linker region and resides in a relatively hydrophobic area. The N¹ and N⁶ nitrogen atoms of the ADP are hydrogen-bonded to the amide nitrogen of Val120 and the main-chain carbonyl of Gln118, respectively (Fig. 4d). The N³ position of adenine forms hydrogen bonds with Arg41 via a water molecule. In addition to these hydrogen interactions, the ADP base is enclosed by several hydrophobic residues, including Val120, Leu231, Pro103, Leu248, Met117, Leu31, and Ile55. The conserved Lys57 residue interacts with the α- and β-phosphoryl groups of ADP, and the main chain of Gly228 contacts both the α-phosphoryl group and the O³’ position of the ribose group of ADP (Fig. 4e, Supplementary Fig. 12). Asn37 and Thr39 form hydrogen bonds to the β-phosphoryl group to anchor and coordinate the ADP, either directly or via a water molecule (Fig. 4e).

The two Mg²⁺ ions (M1 and M2), which assist in the binding of ADP and are essential for kinase activity, were observed forming hexa-coordination structures with water molecules, PsiK residues, and ADP phosphoryl groups (Fig. 4e). The M1 ion hexa-coordinates with the oxygen atoms of the α- and β-phosphoryl groups of ADP, the side chains of Asp249 and Asn229, and two water molecules named w1 and w2; M2 has a hexa-coordinate status with the side chains of Glu251 and Asp249, the oxygen atom from the β-phosphoryl group of ADP, and two water molecules named w3 and w4. Asp249 and Asp251 are part of the strictly conserved “DXE” motif on the loop between strands β8 and β9, while Asn229 is located in the Loop 3 region (Supplementary Fig. 12).

The tryptamine we used to capture the substrate-binding state is stabilized by a channel full of hydrophobic residues, including Phe35, Met63, Val36, Phe69, Ile71, Thr180, Leu252, Leu184, Trp316, and Phe319 (Fig. 4f). Among them, two phenylalanine residues, Phe35 and Phe319, adopt a bolt-like conformation to lock in the tryptamine (Fig. 4f, g). The indole ring of tryptamine is inserted into the pocket toward the Mg²⁺ ions, leaving the terminal amino group pointing toward the pocket entry (Fig. 4f–h). The C⁴ position of the indole ring points to Asp224, with about 5 Å of distance separated by two water molecules, w5 and w6 (Fig. 4h). Modeling 4-hydroxytryptamine in the pocket revealed that it is somewhat far from the catalytic active residue Asp224 (Fig. 4i). Therefore, the captured tryptamine represents an inactive state of the enzyme. We speculate that when the phosphorylation reaction of PsiK occurs, the 4-hydroxytryptamine substrate moves closer to Asp224 and occupies the position of w5.

We then compared the differences and similarities between the PsiK protein and the plant MTK enzyme in cofactor binding, Mg²⁺ chelation, and substrate recognition to deduce the catalytic scheme of PsiK (Supplementary Fig. 13a, b). The detailed description can be found in Supplementary Note 2. The PsiK employs a general acid-base mechanism to achieve the phosphorylation process (Fig. 4j): when the substrate binds, Asp224 functions as a base, accepting a proton from the substrate’s 4-hydroxyl group. This action initiates the in-line nucleophilic attack of the negatively charged oxygen on the γ-phosphate, resulting in a transient complex. Subsequently, the substrate’s 4-hydroxyl group cleaves the γ-phosphate, which is stabilized by M2, causing the intermediate to split into the phosphorylated products and ADP. The product is then released from the catalytic center, while ADP exhibits a lower affinity for the binding pocket⁴³.

Conformational changes of PsiM during the stepwise methylation reaction

To illustrate the methylation steps from norbaeocystin to psilocybin, we have also solved six structures of PsiM alone and in various complexes with SAH, SAM, SAH-norbaeocystin, SAH-baeocystin, and SAH-psilocybin (Fig. 5a–f, Supplementary Figs. 14–19). However, our structures align with the recently published findings⁴⁴. Therefore, detailed descriptions of our results are provided in Supplementary Notes 3. Even so, our results provide a clearer picture of the sequential conformational changes of PsiM during the methylation steps (Fig. 5g, h): (i) In the apo state, the region between Pro185 and Gly222 was dynamic and unstable in the absence of the cofactor and substrate; (ii) When SAM entered the cofactor binding site, Arg75 and Leu68 changed their conformations to stabilize SAM along with other residues. The region between Ala199 and Ser211 remained flexible without the substrate, while the regions of Pro185-Ala198 and Gly212-Gly222 were observed in the presence of the cofactor; (iii) When norbaeocystin entered, Arg281 rotated to a new conformation to anchor the phosphate group of norbaeocystin; the terminal amine group of norbaeocystin pointed toward the methyl group of SAM, and the methylation reaction occurred; (iv) With the first catalytic step finished, baeocystin would slightly shift along the indole ring and away from the cofactor. Then, the methyl group of baeocystin rotates in the other direction, leaving the amine group toward the cofactor; (v) The new cofactor SAM transfers the methyl group to the baeocystin, producing psilocybin, which will further shift along the indole ring to accommodate the two methyl groups; (vi) Finally, the psilocybin substrate will be released from the catalytic pocket with the rotation of Arg281 to the free state without substrate binding.

**Fig. 5: The methylation processes of psilocybin production.**

In vivo activity of the biosynthetic intermediates of psilocybin

Previous studies have reported that psilocybin produces antidepressant effects in mice^45,46. However, the methylation step produces two unfinished catalytic intermediates structurally quite similar to psilocybin; it is still unclear whether these intermediates have similar biological effects. Therefore, we adopted a 6-day sub-chronic variable stress (SCVS) paradigm in female mice to induce depressive-like behaviors, followed by a single intraperitoneal injection of saline, psilocybin, norbaeocystin, and baeocystin (1 mg/kg) one day after the completion of SCVS, to evaluate the biological effects of these intermediates. The behavioral tests, including the sucrose splash test (ST), social interaction test (SIT), novelty-suppressed feeding (NSF), forced swim test (FST), sucrose preference test (SPT), and the open-field test (OFT), were performed to examine the motivated behaviors and anxiety levels of the mice (Fig. 6a). The behavioral results showed that the SCVS successfully induced weight loss and depression-like behaviors in the female mice (Supplementary Fig. 20a–i).

**Fig. 6: Antidepressant effects of psilocybin and its biosynthetic intermediates.**

Compared to the vehicle-injected SCVS mice, the psilocybin- and norbaeocystin-injected SCVS mice showed a significant decrease in their immobility time in the forced swim test and an increased in social interaction time (Fig. 6b, c), indicating that both psilocybin and norbaeocystin alleviated stress-induced depression-like phenotypes and increased motivated behaviors. Notably, the SCVS mice treated with norbaeocystin also showed a reduced latency to eat in the novelty-suppressed feeding test and increased intake of sucrose water in the sucrose preference test, suggesting the anti-anhedonic effects of norbaeocystin (Fig. 6d, e). Meanwhile, no differences were observed among all groups of SCVS mice regarding grooming time in the splash test (Fig. 6f). In the open-field test that measured the anxiety level of the animals, the time spent and distance traveled in the central area of the open field remained unaltered in the SCVS mice treated with psilocybin and its two biosynthesis intermediates (Fig. 6g).

Changes in neuronal activity of the medial prefrontal cortex (mPFC) have been reported in stress-induced depression-like behaviors in mice^47,48. Our results showed that, compared with the vehicle-injected SCVS mice, norbaeocystin-injected SCVS mice exhibited a significant elevation in c-FOS expression, an indicator of neuronal activity, in the infralimbic PFC (IL-PFC) of SCVS mice (Fig. 6h–l). These findings in SCVS female mice suggest the therapeutic potential of the biosynthetic intermediates of psilocybin, specifically norbaeocystin, in alleviating clinical symptoms of depression.

Source link