In vivo production of psilocybin in E. coli

The investigators used Escherichia coli BL21 star (DE3) as the production host and E. coli DH5α for plasmid propagation. Heterologous genes encoding L‑tryptophan decarboxylase (PsiD), a kinase (PsiK), and an S‑adenosyl‑L‑methionine (SAM)‑dependent N‑methyltransferase (PsiM) were obtained from Psilocybe cubensis and cloned into a set of modified ePathBrick/ePathOptimize vectors. Three parallel genetic optimisation strategies were applied: (1) a defined three‑level copy‑number library in which each gene was placed on low, medium or high copy plasmids (27 combinations), (2) a random basic operon library in which all three genes were expressed from a single high‑copy plasmid with randomized T7 promoter variants, and (3) a pseudooperon library with different mutant promoters upstream of individual genes. Libraries were screened in medium throughput using 2 mL cultures in 48‑well plates under standard screening conditions. Standard screening conditions were AMM (Andrew’s Magic Media) supplemented with serine (1 g/L) and 4‑hydroxyindole (350 mg/L), incubation at 37 °C, induction with IPTG (1 mM) four hours after inoculation, and sampling at 24 hours for HPLC analysis. The defined copy‑number library intentionally maintained the same antibiotic burden across strains to control selection pressure. The basic operon library used five mutant T7 promoters (G6, H9, H10, C4, Consensus) generated by standard ePathOptimize methods. For fermentation optimisation, the study varied induction timing (1–6 hours post‑inoculation), IPTG concentration (0.1–1.0 mM), growth/production temperatures (evaluated 30–42 °C with a protocol that initially grew at 37 °C then shifted), media and carbon source identity, and targeted supplements (4‑hydroxyindole, serine, methionine) across several concentrations. Scale‑up was performed as a fed‑batch bioreactor study in which 4‑hydroxyindole was supplied initially at a low concentration and then continuously fed via a syringe pump; feed rates were adjusted using frequent HPLC measurements of pathway intermediates to inform a feedback strategy. The extraction does not fully report detailed bioreactor parameters (agitation, dissolved oxygen set points, pH control and exact feed rates are said to be in supplementary figures). Analytical methods comprised HPLC with diode array detection and refractive index detection for quantification, and high‑resolution LC‑MS/LC‑MS‑MS on an Orbitrap instrument for compound identification. Authentic standards were used where available; for some intermediates expensive standards were substituted with calibration to structurally related compounds. Sample preparation involved simple extraction and direct injection of supernatant for analysis.

Initial heterologous expression of psiD, psiK and psiM under a T7 system in BL21 star (DE3) produced 2.19 ± 0.02 mg/L psilocybin upon IPTG induction. High‑accuracy LC‑MS and MS/MS confirmed the presence of psilocybin and pathway intermediates (4‑hydroxytryptophan, 4‑hydroxytryptamine, norbaeocystin and baeocystin) with mass accuracy better than 5 ppm and expected retention times. Screening the defined copy‑number library yielded only modest improvements over the original ‘‘all‑high’’ construct; the best combination produced 4.0 ± 0.2 mg/L. The basic operon promoter library produced substantially better results: after screening and recloning to exclude plasmid artifacts, operon clone #16 (designated pPsilo16), containing the H10 mutant promoter (a medium‑strength T7 variant), was selected for further study. The top operon mutants showed about a 17‑fold improvement in titre relative to the best clones from the copy‑number library. Further optimisation of fermentation conditions for pPsilo16 raised small‑scale production to 139 ± 2.7 mg/L psilocybin under the identified optimal screening conditions. Key process findings were: maximal production with induction around 3–4 hours post‑inoculation and low sensitivity to induction timing; improved performance with higher IPTG (0.5–1.0 mM versus 0.1 mM); a clear preference for an isothermal 37 °C process rather than lower or higher temperatures; strong sensitivity to media composition and carbon source (rich undefined media such as LB produced low psilocybin and an insoluble coloured product); toxicity and growth inhibition at high concentrations of 4‑hydroxyindole; minimal effect of serine supplementation alone; and a significant enhancement of psilocybin titre when methionine was added in the presence of >350 mg/L 4‑hydroxyindole (p < 0.05). The pseudooperon library mostly underperformed: roughly 95% of mutants showed little or no psilocybin production. The authors report some false positives during screening, which were addressed by recloning and rescreening. Overexpression of native TrpAB was attempted but native levels were judged sufficient, as supported by consistent buildup of 4‑hydroxytryptophan in many fermentations. No significant intracellular accumulation of target metabolites was detected and transmembrane transport was assumed passive but not experimentally investigated. Applying the fermentation knowledge in a fed‑batch bioreactor with an HPLC‑informed feed strategy for 4‑hydroxyindole produced a final psilocybin titre of 1.16 g/L (reported as 1.16 g/L elsewhere in the text). Maximum and final molar yields from the supplied 4‑hydroxyindole substrate were reported as 0.60 and 0.38 mol/mol, respectively. The authors state this is the highest psilocybin titre reported from a recombinant host to date.

Adams and colleagues interpret their findings as proof that a prokaryotic host can produce psilocybin at titres that rival and, upon scale‑up, exceed those reported for recombinant fungal systems. They attribute the large gains to the combined strategy of parallel genetic library screening to find a transcriptionally favourable construct (pPsilo16) and iterative fermentation optimisation, culminating in a fed‑batch process that used HPLC feedback to manage feeding of the toxic substrate 4‑hydroxyindole. The authors place their results relative to prior work by noting that small batch titres were similar to those previously obtained in Aspergillus nidulans, but that the fed‑batch scale‑up achieved roughly a 10‑fold enhancement over the published fungal host titres. They highlight practical insights from the genetic screens: the defined copy‑number approach was laborious and yielded limited benefit (likely due to metabolic burden of multiple plasmids), the pseudooperon library largely failed for unknown reasons, and the simple basic‑operon design with a medium‑strength promoter gave the best outcome, emphasising the current limitations of predictive pathway design and the need for screening. Limitations acknowledged in the paper include the reliance on chemical supplementation (4‑hydroxyindole, serine and methionine) rather than de novo biosynthesis from simple carbon sources; the assumption of passive transport across the membrane without direct verification; and several procedural or numerical details relegated to supplementary figures (for example precise bioreactor feed rates and some fermentation parameters are not fully reported in the extracted text). The authors also note the occurrence of false positives during library screening and the unexplained poor performance of most pseudooperon variants. In terms of implications, the study suggests that further metabolic engineering to enable de novo synthesis (removing the need for substrate supplementation) and to improve SAM availability could make biologically produced psilocybin more competitive with chemical synthesis. The authors indicate that such engineering steps are feasible but beyond the scope of the present study, and propose that their platform and the fed‑batch, HPLC‑informed feeding approach provide a foundation for future strain and process development aimed at industrial production.