Machine learning prediction
Support vector machine (SVM) Enzyme-models were built from enzyme amino acid sequence information using scikit-learn version 0.21.38, and the resulting computer code was made publically available [https://github.com/nwatanbe/SVM_E_model]. Aromatic amino acid decarboxylase (AAAD), aromatic acetaldehyde synthase (AAS, previously referred to as aromatic aldehyde synthase) and phenylpyruvate decarboxylase (PPDC) prediction models were trained with vectors generated by PROFEAT54. AAAD positive training sequences include L-DOPA decarboxylase (DDC) and other typical PLP-dependent carboxy-lyases that decarboxylate aromatic amino acids (Supplementary Data 1). The AAAD positive examples all contain a catalytic histidine, corresponding to H181 of PpDDC (Fig. 2b). Characterized PsTyDC922 is included as a positive AAAD training sequence to ensure there is no bias towards AAS prediction. For AAS models, the positive training examples consist of sequences with homology to known plant-type and insect-type AAS enzymes, including Petroselinum crispum 4HPAAS (Pc4HPAAS) and insect DHPAAS (Supplementary Data 2). Insect-type AAS sequences are classified based on the presence of N192 (insect DHPAAS numbering), and plant-type AAS enzymes are classified based on the presence of F346 or V346 (Pc4HPAAS numbering).
For PPDC prediction models, positive training vectors included sequences annotated as PPDC and indolepyruvate decarboxylase (Supplementary Data 3 and Supplementary Data 4). Since all current database sequences annotated as phenylpyruvate decarboxylase are from bacteria and fungi (plus 1 from Archaea), typical pyruvate decarboxylase (PDC) sequences also had to be included in the first prediction model (Supplementary Table 3, upper table). After discovering PsPDC1, a rose PPDC sequence was found from continuous literature searches, although its protein accession (BAU70033.1 [https://www.ncbi.nlm.nih.gov/protein/BAU70033.1]) is annotated as ‘pyruvate decarboxylase’55. A second PPDC specific SVM model was therefore built by training with 19 homologous plant sequences in the same phylogenetic clade as rose PPDC as positive training sequences and 3 negative training sequences which were curated as plant PDC, as suggested by the results of a previous report56.
Positive training sequences from AAS and PPDC models were included as negative training sequences for the AAAD model; positive training sequences from AAAD and PPDC models were included as negative training sequences for the AAS model; and positive training sequences from AAS and AAAD models were included as negative training sequences for PPDC models. For all models, general negative training sequences included E. coli, S. cerevisiae and A. thaliana enzymes, excluding sequences classified in the positive training group.
Cytochrome P450 (CYP450) prediction models were trained with vectors generated by ProtVec57. To clarify potential N-methylcoclaurine 3-hydroxylase (NMCH) activities of CYP450 monooxygenases, SVM models were trained with CYP80B sequences as positive examples (Supplementary Data 5). CYP450 reductase (CPR) prediction models were trained using sequences listed in Supplementary Data 6. To clarify potential tyrosine 3-monooxygenase activities, SVM models were trained with sequences related to CYP76AD, CYP98A3 and CYP199A2 as positive examples (Supplementary Data 7 and Supplementary Data 8). CYP76AD, CYP98A3 and CYP199A2 enzymes are reported to mediate aromatic hydroxylation of tyrosine and the similarly sized substrate coumaric acid33,34,35.
Prediction models were first built with high-dimensional vectors. Cross validation of all high-dimensional SVM models resulted in F-scores above 0.96. Candidate sequences were selected based on high-dimensional scores. Two-dimensional and three-dimensional plots were used for visual representation of data in Figures. For two-dimensional plots, high-dimensional vectors were compressed to 2 dimensions using principal component analysis (PCA). 2-dimensional SVM models were then built derived from the PCA compressed vectors. SVM and PCA from the scikit-learn library were used58. The three-dimensional SVM plot in Fig. 4d was adopted from an SVM illustration by Dr. Saptashwa Bhattacharyya [https://towardsdatascience.com/visualizing-support-vector-machine-decision-boundary-69e7591dacea]. Compressed two-dimensional decision scores from the combined model (Supplementary Table 3, upper table) are used as the third dimension of Fig. 4d.
Random forests E-models were built from enzyme amino acid sequence information using scikit-learn version 0.21.38, with the same datasets and feature extractions as that of the corresponding SVM models. As an additional benchmark, machine learning differentiation of AAS versus AAAD sequences, and PPDC versus PDC sequences, was compared to differentiation based on homology to consensus sequences. To do this, consensus sequences were generated for each group of training sequences (AAS, AAAD, PPDC and PDC), by selecting the amino acid of maximum frequency at each position. If a training sequence has higher sequence identity to the consensus sequence of its correct group, compared to that of its related group, then it was counted as a correct prediction by homology.
Training sequences, cross-validation F-scores and additional parameters for high-dimensional models are available in the Supplementary Data files.
Protein structural modeling and docking analysis
Homology models were built with Modeller59 run in UCSF Chimera (candidate version 1.15), using template structures of highly similar proteins from the Protein Data Bank [https://www.rcsb.org]7. Multimeric structures and ligands were first prepared in PyMOL version 1.8.7.0. Structures were refined and prepared for docking analysis using Molecular Operating Environment (MOE) version 2020.09017.
Materials and reagents
KOD -Plus- and Ex-Taq HS DNA polymerases were purchased from Toyobo (Tokyo, Japan) and Takara (Tokyo, Japan), respectively. A-attachment Mix was purchased from Toyobo. Primers were ordered from Eurofins Genomics (Tokyo, Japan). A DNA ligation kit and JM109 chemical competent cells were purchased from Takara. The QIAprep Spin Miniprep Kit was obtained from Qiagen (Hilden, DE). BL21(DE3) and BL21-AI competent cells were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Rosetta gami 2 cells were purchased from Sigma-Aldrich (St. Louis, MO, USA). All restriction endonucleases were purchased from New England Biolabs (NEB, Ipswich, MA, USA). Antibiotics were purchased from Nacalai Tesque (Kyoto, Japan), Sigma-Aldrich and FUJIFILM Wako Pure Chemical (Osaka, Japan). Growth medium components were purchased from BD (Franklin Lakes, NJ, USA) and Nacalai Tesque. The IMPACT system, with pTXB1 and pTYB21 vectors, and chitin resin, was obtained from NEB. Amicon Ultra centrifugal filters were obtained from Merck-Millipore (Darmstadt, Germany). The Fluorimetric Hydrogen Peroxide Assay Kit was from Sigma-Aldrich. Amplex Red (10-acetyl-3,7-dihydroxyphenoxazine) peroxidase substrate was from Thermo Fisher (Waltham, MA, USA). L-DOPA and dopamine were purchased from Tokyo Chemical Industry (TCI, Tokyo, Japan). 4-hydroxyphenylpyruvic acid was from Sigma-Aldrich. L-Tyrosine and L-ascorbic acid sodium salt were obtained from Nacalai Tesque. Analytical standards and isotopes were purchased from Santa Cruz Biotechnology (Dallas, TX, USA), Toronto Research Chemicals (New York, ON, Canada), ALB Technology (Kuala Lumpur, Malaysia), Sigma-Aldrich and Cambridge Isotope Laboratories (Tewksbury, MA, USA). N,O-Bis(trimethylsilyl)trifluoroacetamide (BSTFA) and N-methyl-N-(trimethylsilyl)trifluoroacetamide (MSTFA) were obtained from GL Sciences (Tokyo, Japan). 1,4-dithiothreitol (DTT) and pyridine were obtained from FUJIFILM Wako Pure Chemical. Chlorotrimethylsilane (TMS-Cl) was from Alfa Aesar (Haverhill, MA, USA) and methoxyamine hydrochloride was from MP Biomedicals (Irvine, CA, USA).
Preparation of plasmids
Constructed plasmids (Supplementary Table 7) were transformed into JM109 chemically competent E. coli (Takara). Transformants were grown on LB-agar plates supplemented with the appropriate antibiotics at 30–37 °C. Positive clones were screened using colony PCR and target plasmids were purified using a QIAprep Miniprep Kit (Qiagen). Plasmids were then sequenced using primers listed in Supplementary Table 8, by Eurofins Genomics, or by using a BigDye Terminator v3.1 cycle-sequencing kit and a 3500xL Genetic Analyzer from Applied Biosystems (Foster City, CA, USA).
Preparation of predicted candidate genes
Full-length P. somniferum PsTyDC1 native coding sequence was synthesized by Integrated DNA Technologies (IDT, Coralville, IA, USA). Codon optimization of PsONCS3 and TfNCS nucleotide sequences60 for expression in E. coli was assisted by Codon Optimization OnLine (COOL)61, resulting in the coding sequences shown in Supplementary Table 9, and the selected sequences were synthesized by IDT. The native sequence of full-length P. somniferum NMCH isoform 1 (PsNMCH-I1) was also synthesized by IDT.
Native coding sequences of full-length PsPDC1, full-length Ps2HCLL, and N-terminal truncated PsPDC2 were synthesized and cloned into pBAD-DEST49 (LifeSensors Inc., Malvern, PA, USA) via the Gateway cloning system by GeneArt (Invitrogen, Waltham, Massachusetts, USA). Native coding sequences of full-length EcNMCH, AtATR2, and P. somniferum CPR-like (PsCPR-L) were synthesized and subcloned into the pMA vector by GeneArt (Invitrogen). Native coding sequences of full-length PsTyDC6 and N-terminal truncated PsPDC1-IX1 were synthesized and cloned into pTYB21 (NEB) by GenScript (Piscataway, NJ, USA).
Construction of pACYC-3CjMTs-DDC vectors
The pACYC184-derived vectors containing Coptis japonica 4OMT, CNMT, 6OMT (pACYC184-Cj4OMT-CjCNMT-Cj6OMT), and PpDDC (pACYC184-Cj4OMT-CjCNMT-PpDDC-Cj6OMT) were obtained from the laboratory of Professor Hiromichi Minami at Ishikawa Prefectural University13,14. Active site mutations were introduced into PpDDC in pACYC184, by way of site-directed mutagenesis using PCR with primers shown in Supplementary Table 8.
Construction of subcloning vectors and mutations
To construct subcloning vectors for synthetic genes (PsONCS3, TfNCS, PsTyDC1, CjNCS, PsNMCH-I1, EcNMCH, PsCPR-L, AtATR2), and PCR amplified PpDDC and ARO10 (amplified from pGK424-ARO1062), 3′ end A-protrusions were added to each DNA fragment using A-attachment Mix (Toyobo).
Gene mutations were generated using site-directed mutagenesis by PCR with primers listed in Supplementary Table 8. PsTyDC1 mutations (L205H and Y98F-F99Y-L205N) were generated in subcloning vectors by PCR. PpDDC mutations (H181L, H181L-G344S, Y79F-F80Y-H181N, Y79F-F80Y-H181N-G344S) were generated by PCR. The EcNMCH mutation (Y202H) and PsNMCH-I1 mutation (H203Y) were generated in subcloning vectors by PCR. pBad-PsPDC1-His, pBad-PsPDC2-His and pBad-Ps2HCLL-His were generated by removal of a stop codon with PCR.
Construction of alkaloid production vectors
A PsONCS360 containing DNA fragment was obtained from NcoI and BamHI digestion of the PsONCS3 subcloning vector, and then cloned into pCDFDuet-1 via the NcoI and BamHI restriction sites to produce pCDFD-PsONCS3. A TfNCS containing DNA fragment was obtained from NcoI and BamHI digestion of the TfNCS subcloning vector, and then cloned into pCDFDuet-1 via the NcoI and BamHI restriction sites to produce pCDFD-TfNCS.
DNA fragments of PsTyDC1 were obtained from NdeI and XhoI digestion of PsTyDC1 subcloning vectors, and then cloned into pCDFDuet-1-PsONCS3 via NdeI and XhoI restriction sites to produce pCDFD-PsONCS3-PsTyDC1. The PsTyDC1 containing gene fragments were also cloned into pCDFDuet-1-TfNCS via NdeI and XhoI sites to produce pCDFD-TfNCS-PsTyDC1. Digestion of the PsTyDC1 subcloning vector with NdeI and SapI was used to clone into pTXB1 via NdeI and SapI, resulting in pTXB1-PsTyDC1. To produce pTYB21-PsTyDC1, pTYB21-PsPDC1, pTYB21-PsPDC2 and pTYB21-Ps2HCLL, PsTyDC1, PsPDC1, Ps2HCLL, and N-terminal truncated PsPDC2 were PCR amplified and cloned into pTYB21 digested with SapI and BamHI via Gibson assembly (NEB)63.
EcNMCH and EcNMCH-Y202H gene fragments were digested with SalI and NotI in subcloning vectors and then cloned into pCOLADuet-1 via the SalI and NotI restriction sites. AtATR2 and PsCPR-L fragments were next digested from the subcloning vectors using NdeI and XhoI, and then cloned into pCOLAD-EcNMCH and pCOLAD-EcNMCH-Y202H via the NdeI and XhoI restriction sites to produce pCOLAD-EcNMCH-AtATR2, pCOLAD-EcNMCH-Y202H-AtATR2, pCOLAD-EcNMCH-PsCPR-L and pCOLAD-EcNMCH-Y202H-PsCPR-L.
The DNA fragment encoding PsNMCH-I1 with a truncated N-terminal, was digested by NotI and XhoI from the subcloning vector and then cloned into a pACYC184 derived vector containing C. japonica 4OMT, CNMT, and 6OMT via Not I and Xho I restriction sites to produce pACYC-3CjMTs-PsNMCH. Truncated PsNMCH-I1 and truncated PsNMCH-H203Y gene fragments were PCR amplified from subcloning vectors and then cloned into pCOLAD-EcNMCH-PsCPR-L digested with BamHI and NotI via Gibson assembly to produce pCOLAD-PsNMCH-PsCPR-L and pCOLAD-PsNMCH-H203Y-PsCPR-L. Truncated PsNMCH-I1 and truncated PsNMCH-H203Y gene fragments were also digested with NcoI and NotI and cloned into pCOLAD-EcNMCH-AtATR2 digested with NcoI and NotI to produce pCOLAD-PsNMCH-AtATR2 and pCOLAD-PsNMCH-H203Y-AtATR2.
DNA fragments of PpDDC-H181L, PpDDC-H181L-G344S, PpDDC-Y79F-F80Y-H181N and PpDDC-Y79F-F80Y-H181N-G344S were PCR amplified from subcloning vectors and then cloned into pCDFDuet-1 digested with NcoI and BamHI via Gibson assembly. To produce pTYB21-PpDDC-S, PpDDC-H181L was PCR amplified and cloned into pTYB21 digested with SapI and BamHI via Gibson assembly. A CjNCS DNA fragment was obtained from NdeI and XhoI digestion of the CjNCS subcloning vector, and then cloned into pCDFDuet-1-PpDDC vectors via NdeI and XhoI sites to produce pCDFD-CjNCS-PpDDC. A S. cerevisiae ARO10 gene fragment was digested with NcoI and NotI in the ARO10 subcloning vector and then cloned into pCDFDuet-1 via NcoI and NotI restriction sites. A CjNCS gene fragment was next digested from the subcloning vector using NdeI and XhoI, and then cloned into pCDFDuet-1-ARO10 via the NdeI and XhoI restriction sites to produce pCDFD-CjNCS-ARO10. E. coli HpaBC containing gene fragments were PCR amplified from E. coli using the Gibson assembly primers shown in Supplementary Table 8. The PCR product was cleaned using a conventional column-based kit, and then cloned into XhoI-digested pET23 via Gibson assembly to produce pET23-EcHpaBC.
In vivo production of BIA
BL21(DE3) and BL21-AI competent E. coli cells were transformed with various combinations of plasmids from Supplementary Table 7, resulting in the strains shown in Supplementary Table 2. Strains were tested in M9, LB or TB, supplemented with various substrates according to Supplementary Table 2. Expression of recombinant genes in expression vectors containing the T7 promoter system was induced by the addition of 0.5–1.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) to BL21(DE3) cultures. When using BL21-AI cells 0.08–0.4% arabinose was included. Expression of PsPDC1, PsPDC2, and Ps2HCLL in pBAD-DEST49 was also induced by the addition of 0.08–0.4% arabinose.
For quantification of aromatic products, A1-01-DE3 (3 CjMTs, PsNMCH, CjNCS and ARO10), P1-02-AI (3CjMTs, PpDDC, PsONCS3 and PsPDC1), P1-04-AI (3CjMTs, PpDDC, PsONCS3, PsTyDC1 and PsPDC1), P1-06-DE3 (3CjMTs, PpDDC-Y79F-F80Y-H181N, PsONCS3 and PsPDC1), P1-07-AI (3CjMTs, PpDDC-Y79F-F80Y-H181N, PsONCS3 and PsPDC1), A1-06-AI (3 CjMTs, PpDDC-Y79F-F80Y-H181N, CjNCS, ARO10 and PsTyDC1) and T1-10-DE3 (3CjMTs, PpDDC, PsONCS3 and PsTyDC1) (Figs. 6–8, Supplementary Table 2, and Supplementary Fig. 5b, c) were grown using 3.5 mL teriffic broth (TB) supplemented with sodium ascorbate and appropriate antibiotics, in plastic culture tubes at 34–37 °C with shaking at 180–190 rpm. After reaching late log phase, inducing agent (IPTG or arabinose) and substrates (>8 mM tyrosine, >8 mM L-DOPA, >9 mM tyrosine-13C, >3 mM tyrosine-d4, >11 mM L-DOPA-d3) were added. When tyrosine was used as a substrate, sometimes dopamine was included as indicated in Supplementary Table 2 and Supplementary Fig 5b (4.7–7.5 mM dopamine, 10.3 mM dopamine-d2). The addition of dopamine together with L-DOPA was also tested with strain A1-01-DE3 as indicated in Supplementary Table 2 and Supplementary Fig. 5c (17.3 mM dopamine, 7.9 mM dopamine-d2). Cultures were then incubated at 25 °C with shaking at 180–200 rpm.
DT-01-DE3 (3CjMTs, PpDDC-Y79F-F80Y-H181N and TfNCS), DS-02-DE3 (3CjMTs, PsNMCH, CjNCS and PpDDC-H181L), DD-01-DE3 (3CjMTs, PsNMCH, CjNCS and PpDDC-H181L-G344S), DQ-01-DE3 (3CjMTs, PsNMCH, CjNCS and PpDDC-Y79F-F80Y-H181N-G344S), DT-02-DE3 (3CjMTs, PpDDC-Y79F-F80Y-H181N and PsONCS3), DT-03-DE3 (3CjMTs, PsNMCH, CjNCS and PpDDC-Y79F-F80Y-H181N) and A1-03-DE3 (3CjMTs, PpDDC, CjNCS, ARO10 and EcHpaBC) (Fig. 7a, Supplementary Table 2, and Supplementary Fig. 5a, d) were tested in 3–4.8 mL M9 supplemented with ascorbate and appropriate antibiotics. After reaching log phase in plastic culture tubes at 36–37 °C, IPTG and substrates (>4.5 mM tyrosine, >2 mM L-DOPA, >5 mM tyrosine-13C, >4 mM L-DOPA-d3) were added. When tyrosine was used as a substrate, sometimes dopamine was included as indicated in Supplementary Table 2 (1.2–1.4 mM dopamine). Cultures were then incubated at 20–25 °C with shaking at 180 rpm. Additional ascorbate was added as needed to prevent oxidative degradation of target compounds and melanization.
Conversion of norcoclaurine to reticuline was mediated by NMCH and CPR containing strains N1-01-DE3, N1-02-DE3, N1-03-DE3, N1-04-DE3, N2-01-DE3, N2-02-DE3, N2-03-DE3 and N2-04-DE3 (Supplementary Table 2). Here, strains first grown in LB medium were used to inoculate TB medium to a starting OD600 of 0.02 in 3 mL, with appropriate antibiotics. After four hours at 37 °C with shaking at 200 rpm, recombinant protein expression was induced with 0.68 mM IPTG and the temperature was lowered to 20 °C. After 5.5 h, cells were spun down and re-suspended in 1.5 mL TB supplemented with 1.2 mM norcoclaurine, 5.1 mM sodium ascorbate, and 0.2 mM IPTG. After 1.5 days at 25 °C with shaking at 200 rpm, BIA titers were measured with LC-MS.
Additional bioproduction conditions are given in the legends of Fig. 2e, Fig. 4e, Fig. 8a, Supplementary Fig. 2, Supplementary Fig. 3 and Supplementary Fig. 5. Bioproduction times are based on the addition of substrate.
Quantitative analysis of BIA pathway intermediates with LC-MS, CE-MS, and GC-MS
The culture medium was filtered with Amicon Ultra 0.5 mL centrifugal filters with a molecular weight cut-off of 3000 Da. Filtrates were kept on ice and immediately processed for analysis, or stored at −30 °C or −80 °C before use.
For LC-MS analysis, filtered culture medium was diluted in a solution of camphor sulfonic acid, and then loaded onto a Shimadzu LCMS-8050 system (Shimadzu, Kyoto, Japan) operated in multiple reaction monitoring (MRM) mode7. The electrospray ionization (ESI) ion source was connected to a Shimadzu Nexera X2 UHPLC system where separation was performed on a Discovery HS F5-3 column (3 μm, 2.1 mm × 150 mm, Sigma–Aldrich). Shimadzu LabSolutions LCMS version 5.99 SP2 was used for data collection and analysis. DHPAA [151.30 > 123.15(−)], tyramine [138.00 > 121.15(+)], dopamine [154.10 > 91.05(+)], norcoclaurine [272.00 > 106.95(+)], norlaudanosoline [288.05 > 164.15(+)] and reticuline [330.10 > 192.00(+)] were identified using the MRM transitions listed in brackets, and confirmed by running authentic standards. Over 100 metabolites could be monitored with MRM detection.
For CE-MS analysis, filtered samples were diluted in a methionine sulfone solution when using positive ion mode, or in a piperazine-N,N’-bis(2-ethanesulfonic acid) solution for negative ion mode. CE-MS analysis was performed using an Agilent G7100 CE system with an Agilent G6224AA LC/MSD TOF (Agilent Technologies, Palo Alto, CA)36,37. Agilent MassHunter Workstation versions 10.1 and B.06.00 were used for data acquisition and analysis, respectively. Quantification of isotopes in Fig. 8 and Supplementary Fig. 5 was based on standard curves of non-labeled compounds. CE-MS peak areas in relation to internal standard peak areas were used to quantify all compounds except for 4HPP (Supplementary Fig. 5a), which was quantified based on its own peak intensity.
For GC–MS analysis of in vivo products, filtered samples were dried under vacuum and then derivatized with BSTFA and TMS-Cl. The derivatized aromatic compounds were analyzed on a GCMS-QP2010 Plus (Shimadzu) with a DB-5 capillary column (Agilent). Shimadzu LabSolutions version 2.72 was used for GC-MS data collection and analysis. TMS-derivatized tyrosol and norcoclaurine were identified using the most intense product ions m/z 179.1 and m/z 308.1, respectively, and confirmed by running authentic standards.
In vitro characterizations of PsTyDC6
PsTyDC6 was expressed in Rosetta-gami 2 cells transformed with pTYB21-PsTyDC6 (Supplementary Table 7). After reaching log phase, the cells were induced with 0.15 mM IPTG and grown overnight at 15.5 °C. PsTyDC6 was purified on a chitin column followed by on-column cleavage of the chitin-binding domain and intein fusion via the addition of 50 mM DTT to the column. PsTyDC6 was then eluted into Amicon Ultra centrifugal filters and the buffer was changed to PBS (pH 7.0).
For detection of in vitro produced 4HPAA, purified PsTyDC6 and digested PsTyDC6 cell extract were mixed with 5 mM and 4 mM tyrosine, respectively. PsTyDC6 reactions containing 100 μM PLP were started together with control reactions containing 100 μM PLP and 4 mM tyrosine, followed by incubation at 30 °C for 3.5 h. Samples were lyophilized and then derivatized by treatment with a pyridine and methoxyamine solution followed by treatment with MSTFA. Derivatized compounds were analyzed by GC-MS. TMS- and methoxyamine-derivatized 4HPAA was identified based on product ions m/z 190.1 and m/z 205.1, and confirmed by running an authentic 4HPAA standard after derivatization using the same method. To detect in vitro production of H2O2 by PsTyDC6, a horseradish peroxidase-based fluorescent assay7 was performed with the fluorescent substrate Amplex Red (10-acetyl-3,7-dihydroxyphenoxazine) together with other components of a Fluorimetric Hydrogen Peroxide Assay Kit (Sigma-Aldrich). For the peroxidase-based assay, PsTyDC6 was prepared in PBS (pH 7.0) with 1 μM PLP. Baseline fluorescence from the control with matching PsTyDC6 and PLP, but with no tyrosine, was subtracted from each tested condition containing tyrosine. Initial rates of fluorescence production were plotted against final tyrosine concentration using the Michaelis-Menten function of Prism 7 version 7.0d.
LC-MS operated in MRM mode was applied to detect in vitro produced DHPAA, dopamine, tyramine, norcoclaurine and norlaudanosoline. For DHPAA, dopamine and norlaudanosoline production, purified PsTyDC6 was mixed with 5 mM L-DOPA. For tyramine and norcoclaurine production, purified PsTyDC6 was mixed with 1.25 mM tyrosine and 2.5 mM L-DOPA. In vitro samples were incubated at 30 °C for 80 min. to analyze DHPAA, and for 8 h to analyze norcoclaurine and norlaudanosoline.
Extraction of aromatic compounds for GC-MS quantification
A solution of ammonium carbonate was added to culture samples, followed by addition of ethyl acetate. After vortexing, the organic layer was removed and evaporated under vacuum. The dried extracts were then derivatized in a mixture of BSTFA, TMS-Cl, and ethyl acetate. Quantitative standard curves were produced by extracting alkaloid standards prepared in TB medium, followed by TMS-derivatization in equivalent volumes. The TMS-derivatized samples were analyzed with GC-MS as described above.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
