Psilocybin: crystal structure solutions enable phase analysis of prior art and recently patented examples

Psilocybin is a zwitterionic tryptamine natural product found in numerous mushroom species and has been revisited as a clinical candidate for several psychiatric indications. Earlier literature and process-scale work reported multiple solid forms of synthetic psilocybin, including solvent solvates, a hydrate and an anhydrous form, and more recent crystallisation studies identified three reproducible crystalline phases referred to here as Hydrate A, Polymorph A and Polymorph B. A patent application described related powder X-ray diffraction (PXRD) observations but used the ambiguous terminology 'isostructural variant' and raised the possibility that minor differences in diffractograms from large-scale production reflected a distinct single phase rather than mixtures or processing artifacts. Sherwood and colleagues aimed to solve the crystal structures of the two anhydrous forms (Polymorphs A and B) using powder diffraction and computational methods, and then to apply Rietveld-method (RM) quantitative phase analysis (QPA) to a set of historical and contemporary bulk psilocybin PXRD data. The study sought to determine whether all published diffraction patterns could be explained by the three characterised forms, to assess correlations between crystal form, processing and storage, and to test whether the phases invoked in recent patent claims were in fact single phases or mixtures arising from drying and scale effects.

The investigators combined single-crystal and powder diffraction, structure solution from powder, Rietveld refinement and density functional theory (DFT) optimisation. Hydrate A had been solved previously by single-crystal X-ray diffraction and served as a structural reference. Polymorphs A and B were insoluble to single-crystal methods as isolated, so their structures were determined by indexing powder patterns (laboratory and synchrotron) and solving the structures using Monte Carlo simulated annealing implemented in EXPO2014, followed by Rietveld refinement in GSAS-II and DFT geometry optimisation (VASP and CRYSTAL17). Synchrotron data for Polymorph A were collected at beamline 11-BM of the Advanced Photon Source (wavelength 0.458162 Å) with high-resolution analyzers; samples were rotated in 1.5 mm Kapton capillaries. Rietveld refinements included restraints on bond distances and angles using Mogul-derived targets, background modelling with a shifted Chebyshev polynomial, a generalized microstrain peak profile and spherical harmonics to model preferred orientation. For Polymorph B, similar strategies were used though isotropic microstrain and different refinement parameterisations were applied. DFT optimisations used VASP (GGA-PBE, 400 eV cutoff, k-point spacing ~0.5 Å^-1) with fixed experimental cells and additional single-point population analyses with CRYSTAL17/B3LYP. For phase analysis, the authors assembled PXRD data for 24 unique psilocybin samples produced between 1963 and 2021 from multiple sources (archived Sandoz material, literature PXRD, USP authentic substance, clinical-trial lots, process-development batches and material associated with patent filings). Older literature peak lists were converted to pseudo-raw data, and scanned diffractograms were digitised where necessary. Fixed structural models for Hydrate A, Polymorph A and Polymorph B were imported into GSAS-II and RM-based QPA was performed on each pattern, modelling preferred orientation with the minimum spherical-harmonic orders required to achieve acceptable fits.

The authors report successful structure solutions and refinements for the two anhydrous polymorphs based on powder data. Polymorph A was indexed and refined against high-resolution synchrotron data; the final Rietveld refinement for Form A used 85 variables, 28,045 observations and 49 restraints and produced Rwp = 0.0740 and GOF = 1.36. Polymorph B was solved using two psilocybin molecules as fragments, with a final Rwp = 0.0749 and GOF = 1.23. DFT geometry optimisation and population analyses were carried out to support the models. The hydrate (OKOKAD) crystallises in orthorhombic Pbca as a trihydrate and was used as the starting molecular fragment. Validation metrics included r.m.s. Cartesian displacements between Rietveld-refined and DFT-optimised non-hydrogen atoms: Form A 0.053 Å, Form B molecules 0.160 and 0.121 Å, and the trihydrate OKOKAD 0.480 Å (the latter noted as outside typical single-crystal ranges). A VASP energy comparison indicated Form B is 11.8 kJ mol^-1 lower in energy than Form A, but the authors cautioned this difference is near the expected uncertainty of such calculations and thus the two anhydrates are similar in energy. Morphologically, Polymorph A tended to form needle-like crystals while Polymorph B and Hydrate A were expected to be platy; strong preferred orientation was observed for many powder samples and was modelled during refinement. Hydrogen-bonding analysis showed consistent motifs across forms: the phosphate P–O–H group forms strong O–H...O hydrogen bonds creating R2 2(8) rings, and the ring N–H and protonated side-chain N+ participate in N–H...O hydrogen bonding to phosphate oxygens; in the hydrate water molecules engage in water-phosphate and water-water hydrogen bonds. Force-field and DFT-derived analyses indicated intramolecular deformation energy was dominated by angle distortion terms and intermolecular lattice energy had substantial electrostatic (hydrogen-bond) contributions. Using the three solved structures, RM-based QPA was applied to 24 PXRD data sets. Every sample was found to contain one or more of Hydrate A, Polymorph A or Polymorph B. Representative results: the 1963 Sandoz sample (Sample 1) showed only Hydrate A; an early literature PXRD converted for analysis (Folen, Sample 2) was interpreted predominantly as Polymorph A with detectable Polymorph B and Hydrate A (quantification less precise due to low-resolution source data). The Johns Hopkins clinical-trial lot (Sample 4) was primarily Polymorph A, with perhaps a trace of Hydrate A; this lot had supported clinical trials from 2015–2018. Samples originating from the same laboratory (Samples 5–7) exhibited variability: Sample 5 contained appreciable amounts of all three forms with ~12% Polymorph B, Sample 6 was essentially Hydrate A, and Sample 7 was mainly Polymorph A with an insignificant hydrate trace. Process-development and commercial batches (Samples 12–18) were largely Polymorph A with minor hydrate in some cases; Sample 19 (1.2 kg scale) was single-phase Polymorph A. Single-phase reference patterns were obtained for Hydrate A (Sample 20) and Polymorph B (Sample 21). Crucially, the authors examined materials and patent-associated diffractograms that had been described as a distinct 'Polymorph A' or an 'isostructural variant' with a weak reflection at ca. 17.5° 2θ. Rietveld analysis demonstrated that patterns showing this weak feature were consistent with mixtures of Polymorph A and Polymorph B rather than a novel single-phase variant; in the patent example RM yielded an approximate composition of 81% Polymorph A and 19% Polymorph B. Controlled dehydration experiments showed that rapid vacuum dehydration at lower temperatures favoured formation of single-phase Polymorph A, whereas slower dehydration or drying at higher temperatures could introduce Polymorph B alongside Polymorph A. These findings provide a mechanistic explanation for scale-dependent observations related to oven/drying conditions and crystal-bed depth. Overall, no additional crystalline forms beyond Hydrate A, Polymorph A and Polymorph B were required to account for the historical and recent diffraction data examined.

Sherwood and colleagues interpret their combined structural, computational and phase-quantification work as demonstrating that three crystalline forms—Hydrate A (trihydrate), Polymorph A (anhydrate) and Polymorph B (anhydrate)—are sufficient to explain published PXRD patterns of bulk psilocybin from 1963 through 2021. They position these results against both early literature descriptions and recent patent claims, arguing that ambiguity in terminology such as 'isostructural variant' in patent filings obscured the more parsimonious explanation that minor reflections arise from phase mixtures produced by processing conditions rather than novel single phases. The authors highlight several process–structure relationships: recrystallisation from aqueous or water-containing organic solvents tends to give Hydrate A, whereas controlled rapid vacuum drying of Hydrate A at moderate temperatures produces Polymorph A; slower dehydration at elevated temperatures or inadequate vacuum/large crystal-bed depths can produce mixtures containing Polymorph B or convert material to Polymorph B. Preferred orientation in PXRD sample preparation strongly affects apparent intensities of certain diagnostic peaks (for example a reflection near 10.1° 2θ for Polymorph A), which the authors note can complicate phase assignments if not properly modelled. Limitations acknowledged in the report include the intrinsic uncertainty in some analyses arising from low-resolution or digitised historical PXRD data, the sensitivity of certain diffraction intensities to sample texture and preparation, and the proximity of computed energy differences between Forms A and B to the expected error bounds of DFT methods—such that relative stabilities should be considered with caution. The hydrate single-crystal model exhibited a larger r.m.s. displacement in comparisons with DFT-optimised structures, which the authors noted in assessing model quality. Practical implications discussed by the authors include the value of using solved crystal structures combined with RM-based QPA to assess phase composition of API material, the need for controlled dehydration conditions (crystal-bed depth, temperature and pressure) to reproducibly obtain single-phase Polymorph A at scale, and the relevance of hygroscopicity and storage humidity to the appearance of Hydrate A. The authors further contend that their findings undermine claims of novelty for the crystalline forms asserted in recent patent documents, since the same phases were produced historically and are reproducible by routine methods. They indicate that more systematic studies of drying conditions at scale will follow to refine processing recommendations.

Structure solutions from laboratory and synchrotron PXRD, supported by DFT, were achieved for two anhydrous psilocybin polymorphs encountered in API development, adding to the previously solved trihydrate form. Using these structural models, Rietveld-method whole-pattern fitting permitted quantitative phase analysis of 24 psilocybin samples produced between 1963 and 2021 and showed that every sample contained one or more of Hydrate A, Polymorph A or Polymorph B. Recent and historical data are thus consistent with repeated production and observation of these three pharmaceutically relevant forms. The work further links processing and storage conditions to phase composition—rapid vacuum dehydration at controlled temperatures yields single-phase Polymorph A, while slower or higher-temperature dehydration can introduce Polymorph B—and demonstrates that minor diffraction features previously described as novel variants are better explained as phase mixtures. Recommendations include careful control of crystal-bed depth, temperature and pressure during dehydration to obtain large-scale batches of pure Polymorph A.