Paperwhite narcissus (Narcissus papyraceus) and summer snowflake (Leucojum aestivum) bulbs were purchased from Vesey’s (York, PE, Canada). Bulbs were planted in plastic pots using autoclaved AGRO MIX G6 potting soil (Fafard, Saint-Bonaventure, QC, Canada). The plants were kept at room temperature with exposure to tube lighting in long day (16 h of light/8 h of dark) conditions until flowering. The plants were watered when necessary to keep the soil moist. Different tissues such as bulbs, roots, stems, leaves, and flowers were collected, flash frozen in liquid nitrogen, and stored at −80°C until further used. Nicotiana benthamiana (Goodin et al., 2008) plants were grown in a growth chamber in long day (16 h of light/8 h of dark) conditions at 22°C.

The bacteria used in this study were Escherichia coli DH5α (Invitrogen), E. coli Rosetta (DE3) pLysS (Novagen), and Agrobacterium tumefaciens GV3101 (Holsters et al., 1980). The yeast strain (Saccharomyces cerevisiae) used is Y2HGold (Clontech Laboratories). The bacteria were grown in Luria-Bertani (LB) medium supplemented with the appropriate antibiotics at the following temperatures: E. coli at 37°C; and A. tumefaciens at 28°C. Antibiotics were used at the following concentrations (µg/mL): ampicillin, 100; chloramphenicol, 34; kanamycin, 50; rifampicin, 50; gentamicin, 30. Yeasts were grown at 30°C in selective synthetic complete medium.

RNA was extracted from N. papyraceus (Np) and L. aestivum (La) bulbs using CTAB (cetrimonium bromide) method as described by Singh and Desgagné-Penix (2017). cDNA was synthesized from 1 µg RNA samples using SensiFAST cDNA synthesis kit (Bioline) according to manufacturer’s protocol. The open reading frame (ORF) of full length NpNBS, NpNR, NpTR, LaNBS, LaNR, and LaTR were amplified from bulbs cDNA using PrimeStar GXL premix (TaKaRa Bio) in 50 μL reaction with 0.2 μM forward and reverse primers ( Supplementary Table S2 ). PCR program parameters: 2 min 98°C 1 cycle, 10 s 98°C, 20 s 55°C, 1 min 72°C for 35 cycles, 5 min 72°C 1 cycle, and final infinite hold at 4°C. A classical restriction digestion based cloning approach was used to clone and create all the desired vectors.

For protein expression in E. coli, NpNBS, NpNR, NpTR (tropinone reductase), LaNBS, LaNR, and LaTR were fused to the C-terminus of the maltose binding protein (MBP) in the pMAL-c2x vector (New England Biolabs). Precisely, the full-length ORFs were amplified from cDNA by PCR using primers reported (respective restriction enzyme sites are underlined, Supplementary Table S2 ). PCR products were cleaned using GenepHlow Gel/PCR kit (Geneaid). Purified PCR products of NpNBS, NpNR, LaNBS, and LaNR were digested with BamHI and HindIII, while NpTR and LaTR were digested with BamHI and SalI and ligated into pMAL-c2x vector digested with BamHI/HindIII and BamHI/SalI respectively, using T₄ DNA ligase (New England Biolabs). The recombinant plasmids were transformed into chemically competent E. coli DH5α cells by heat shock transformation and colonies were selected on ampicillin LB agar plates. The positive clones were identified by colony PCR in a 20 μL reaction using Taq DNA polymerase with ThermoPol buffer (New England Biolabs) with PCR parameters: 5 min 95°C 1 cycle, 30 s 95°C, 40 s 55°C, 1 min 68°C for 30 cycles, 5 min 68°C 1 cycle, and final infinite hold at 4°C. The resulting plasmids were verified by DNA sequencing to ensure the correct sequence and exclude undesired mutations.

For split luciferase complementation assays (SLCA) in N. benthamiana leaves, the NpNBS, NpNR, NpTR, LaNBS, LaNR, and LaTR genes were cloned into pCAMBIA1300:CLuc fused to the C-terminal (398-550 amino acids) of firefly luciferase (Cluc), and NpNBS, NpNR, LaNBS, and LaNR into pCAMBIA1300:Nluc fused to the N-terminal (2-416 amino acids) of firefly luciferase (Nluc) and driven by the CaMV 35S promoter (Chen et al., 2008). The full-length ORFs were amplified by PCR using reported primers from cDNA (respective restriction enzyme sites are underlined, Supplemental Table S2 ). After cleaning up using GenepHlow Gel/PCR kit (Geneaid), PCR products were digested with the mentioned restriction enzyme pair and ligated into the corresponding sites of pCAMBIA1300:Cluc or pCAMBIA1300:Nluc vectors using T₄ DNA ligase. The recombinant plasmids were transformed into chemically competent E. coli DH5α cells and colonies were selected on kanamycin LB agar plates. The positive clones were identified by colony PCR. The resulting binary vectors were verified by DNA sequencing.

For yeast two-hybrid assays, genes encoding full-length NpNBS, NpNR, NpTR, LaNBS, LaNR, and LaTR were amplified from N. papyraceus and L. aestivum bulbs cDNA and cloned into the pGBKT7 (bait) or pGADT7 (prey) vectors (Clontech Laboratories) in frame with the GAL4 DNA binding domain (DNA-BD) or GAL4 activation domain (AD). The full-length genes were amplified by PCR using primers reported (respective restriction enzyme sites are underlined, Supplemental Table S2 ). After cleaning up using GenepHlow Gel/PCR kit (Geneaid), PCR products were digested with the mentioned restriction enzyme pair and ligated into the corresponding sites of pGBKT7 (bait) or pGADT7 (prey) vectors using T₄ DNA ligase. The recombinant plasmids were transformed into chemically competent E. coli DH5α cells. Colonies with bait plasmids were selected on kanamycin while colonies with prey plasmids were selected with ampicillin LB agar plates. The positive clones were identified by colony PCR. The resulting plasmids were verified by DNA sequencing.

For subcellular localization, NpNBS, NpNR, LaNBS, and LaNR coding sequences were fused upstream to the gene encoding the yellow fluorescence protein (YFP) in the pBTEX binary vector under the control of the CaMV 35S promoter (Frederick et al., 1998). The full-length ORFs were amplified (respective restriction enzyme sites are underlined, Supplemental Table S2 ). The PCR products were cleaned up using GenepHlow Gel/PCR kit (Geneaid), digested with KpnI/XbaI, and ligated into pBTEX-YFP vector digested with KpnI/XbaI using T₄ DNA ligase. The recombinant plasmids were transformed into chemically competent E. coli DH5α cells and colonies were selected on kanamycin LB agar plates. The positive clones were identified by colony PCR. The resulting plasmids/binary vectors were verified by DNA sequencing.

NpNBS, NpNR, NpTR, LaNBS, LaNR, and LaTR were cloned into the pMAL-c2x vector. The purified plasmids were transformed using heat shock transformation into chemically competent E. coli Rosetta (DE3) pLysS strain for protein expression. Transformed cells were selected on LB plates with ampicillin, and chloramphenicol overnight at 37°C. The positive colonies were screened by colony PCR. A PCR positive single colony was picked and grown overnight at 37°C at 220 rpm in 12.5 mL LB broth containing ampicillin and chloramphenicol. The overnight grown pre-culture was added into fresh 250 mL LB broth containing ampicillin and chloramphenicol and grown at 220 rpm at 37°C to an OD₆₀₀ = 0.5 to 0.6. The cultures were brought to room temperature and Isopropyl-β-D-thiogalactopyranoside (IPTG) was added to a final concentration of 0.25 mM to induce protein expression. The cultures were further incubated for 20 h at 18°C at 150 rpm. Bacterial cultures were pelleted at 10,000 rpm for 15 min at 4°C, resuspended in 25 mL column binding buffer (25 mM Tris-HCl [pH 7.5], 150 mM NaCl, and 1 mM EDTA) and frozen at −80°C overnight. The cultures were thawed on ice water and 1 mM phenylmethylsulfonyl fluoride [PMSF] with 0.1x protease inhibitor cocktail (Cell Signaling Technology) was added. Cultures were then lysed using a sonicator at 41% amplitude for a total of 8 min with 15 s run and 35 s cooling time in ice. The lysates were centrifuged twice at 14,000 rpm for 20 min at 4°C to pellet the cell debris and the clear supernatants were collected. Supernatants were mixed with 500 μL of amylose resin beads (New England Biolabs) (prewashed with column binding buffer and resuspended to 50% slurry) and incubated for 1 h at 4°C with constant rocking. The mixture was passed twice through the filter columns (Thermo Scientific) with gravitational flow to retain the beads with bound proteins in the column matrix. The beads were washed with gravitational flow in the columns three times with 30 mL column binding buffer. The bead slurry was transferred to microcentrifuge tubes and centrifuged at 1000 rpm for 1 min and the supernatant was removed. Finally, the bound proteins from the bead pellet were eluted twice (elution 1 and 2) each time in 500 μL elution buffer (15 mM maltose in column binding buffer) by centrifugation at 1000 rpm for 1 min and supernatant/elute was collected. The protein samples were flash frozen in liquid nitrogen and stored at −80°C. Protein quantification was done using DC protein assay kit (Bio-Rad) according to the manufacturer’s instructions with bovine serum albumin (BSA) as standard, and protein samples were fractionated by 10% (v/v) SDS-PAGE and stained with Coomassie Blue.

The in-silico protein analysis was done by DNAMAN analysis software (Lynnon BioSoft). Protein sequences of NpNBS, and LaNBS were aligned with Narcissus pseudonarcissus ‘King Alfred’ norbelladine synthase (NpKANBS; GenBank: AYV96792.1), and Thalictrum flavum norcoclaurine synthase (TfNCS; GenBank: ACO90248.1). Similarly, NpNR, LaNR, NpTR and LaTR were aligned with N. pseudonarcissus noroxomaritidine/norcraugsodine reductase (NpKANR; KU295569) using CLUSTAL W algorithm in T-Coffee software (Notredame et al., 2000) with default parameters. Sequence alignments were formatted using Boxshade program (https://embnet.vital-it.ch/software/BOX_form.html).

Amino acid sequences corresponding to NpNBS, LaNBS, NpNR, LaNR, NpTR, LaTR candidates were uploaded on Protein Homology/analogY Recognition Engine V 2.0 (Phyre2) (Kelley et al., 2015) website, I-Tasser from Zhang lab (Yang and Zhang, 2015) and MOE 2020.09 software (Chemical Computing Group) to model NBS and NR proteins. Following close analysis of predicted structures and comparison by superimposition with orthologs and homologs, the most consistent models were selected. LaNR from I-Tasser, NpNR and La and NpTR models from Phyre2 were used, while NBS were best modeled by MOE. MOE was further used to analyze the resulting homology model conformation and prepare receptors for docking. First, modeled structures were compared to their template crystal structures in complex with their ligands downloaded from the Protein Data Bank (for NBS: norcoclaurine synthase from Thalictrum flavum in complex with dopamine and hydroxybenzaldehyde 2VQ5 (Ilari et al., 2009), NR: noroxomaritidine/norcraugsodine reductase in complex with NADP+ and tyramine 5FF9 NpKANR (Kilgore et al., 2016b), TR: Tropinone reductase-II complexed with NADP+ and pseudotropine 2AE2 and 5FF9 (Yamashita et al., 1999), aligning amino-acid sequences and then superimposing the structures.

Structure preparation consisted of correcting issues, capping, charging termini, selecting appropriate alternate, and calculate optimal hydrogen position and charges using Protonate 3D. Fixed receptor and tethered active site energy minimization was performed for each modeled protein in presence of template ligands prior to docking. Ready to dock ligands were uploaded from ZINC15 (Sterling and Irwin, 2015) when available, or manually drawn (from SMILES), washed, prepared, and minimized with MOE. The MMFF94× force field was used. Receptors active site was predicted using MOE Site Finder and used as docking site to place ligand using Triangle Matcher as placement method for 200 poses and tethered induced fit as refinement to perform flexible docking. Ten resulting poses were analyzed. The most probable poses based on literature description of templates active site, on comparison with crystalized templates interactions and on scores are presented. For NBS, 3,4-DHBA was docked first, the most consistent pose was further included in the active site used for tyramine docking. Similarly for NR, NADPH was docked first, and the most consistent pose compared to template crystals was further included in the active site used to dock norcraugsodine. PLIP was used to analyze interactions between ligands and receptors (Adasme et al., 2021), and were further processed using PyMOL (Shrödinger).

Norbelladine and norcraugsodine were synthesized as previously described (Singh et al., 2018). Standard solutions of 3,4-DHBA (Fisher Scientific), tyramine (Sigma-Aldrich), and papaverine (Sigma-Aldrich) were prepared at 1000 mg/L in LC-MS grade methanol (Sigma-Aldrich). Standard solutions of norbelladine and norcraugsodine were prepared as previously described (Singh et al., 2018). From these standard solutions, dilutions were performed to obtain working solutions of 100 mg/L in methanol, and 1 mg/L in the mobile phase (ammonium acetate 10 mM (Sigma-Aldrich) (pH 5.0), and acetonitrile (Sigma-Aldrich) [60:40]). Standards and solutions were stored in the dark at −20°C.

Single enzyme assays were performed at 35°C for 2 h. All the enzymatic reactions were terminated by the addition of 10 μL of 20% trichloroacetic acid (TCA). Negative controls were purified MBP-tag protein from E. coli, and reactions without substrate or cofactor. The catalytic activity of purified NBS enzymes were analyzed following the method of Singh et al., 2018 with minor modifications. Reactions were conducted using 80 μg of purified proteins in 100 mM HEPES buffer (pH 6.0), with 10 μM tyramine and 300 μM 3,4-DHBA, in a total volume of 100 μL. Reaction components were equilibrated at 35°C and the reaction was started by the addition of enzyme to the substrate and buffer mixture. NR and TR single enzyme assays were performed as previously reported by Kilgore et al., (2016b) with minor modifications. The assay mix contained 60 μg of purified proteins, 10 μM tyramine, 300 μM 3,4-DHBA, and 1 mM NADPH in 100mM sodium citrate buffer (pH 6.0), in a total volume of 100 μL. The assays with NBS and NR or TR together in a single-step reaction contained 80 μg of purified NBS enzyme, 60 μg of purified NR/TR enzyme, 10 μM tyramine, 300 μM 3,4-DHBA, and 1 mM NADPH in 100mM HEPES buffer (pH 6.0), in a total volume of 100 μL. The assays with NBS and NR/TR in two-step reactions were conducted as follows: 80 μg of purified NBS enzyme in 100 mM HEPES buffer (pH 6.0), with 10 μM tyramine and 300 μM 3,4-DHBA, in a total volume of 100 μL was incubated at 35°C for 2 h. The NBS enzyme was inactivated by boiling at 95°C for 10 min, sample was centrifuged, and the supernatant (100 μL) was used as norcraugsodine solution, and 60 μg of purified NR/TR enzyme and 1 mM NADPH were added and incubated for additional 2 h. Following the reaction termination, papaverine (1000 mg/L) was added to all the reaction serving as an internal standard for the relative quantification of detected compounds. All reactions were performed in triplicates. The reaction samples were diluted 100 fold with mobile phase (ammonium acetate 10 mM [pH 5.0], and acetonitrile [60:40]) and analysis of the enzymatic product (norbelladine) using a high-performance liquid chromatography (HPLC) system coupled with a tandem mass spectrometer (MS/MS) was carried out as described by (Tousignant et al., 2022).

Norcraugsodine (257 m/z) could not be detected by GC-MS without derivatization. For the derivatization, dry enzymatic assay samples were reconstituted in 300 μL HPLC grade acetonitrile and were transferred quantitatively in crimp-seal autosampler vials, without capping the vials. A 150 μL aliquot of BSTFA (with 1% TMCS) was added to each sample along with a magnetic stir bar, then the vials were capped using crimpers. The samples were stirred at room temperature for 60 minutes. For the GC-MS analysis, the derivatized samples were injected into the GC-MS (Agilent Technologies 6890N GC coupled with 5973N inert MSD) in electron ionization mode at 70 eV. The temperature ramp used is described as follows: temperature was set at 100°C for 2 min, followed by 100–180°C at 15°C min⁻¹, 180–300°C at 5°C min⁻¹, and a 10 min hold at 300°C. Injector and detector temperatures were set at 250°C and 280°C, respectively, and the flow rate of carrier gas (He) was 1 mL min⁻¹. A split ratio of 1:10 was applied, and the injection volume was 1 μL. The presence of norcraugsodine in the tested samples was determined by comparison with the GC-MS analysis of a norcraugsodine standard derivatized following the same protocol. Tris-derivatized norcraugsodine (473 m/z) was observed at a retention time of 24.79 minutes and showed characteristic fragment ions (458 m/z and 294 m/z) that we respectively attributed to the loss of a methyl radical and to the loss of a 4-((trimethylsilyl)oxy)benzyl radical. The corresponding bis-derivatized aldehyde (282 m/z) was observed at a retention time of 9.28 minutes and showed characteristic fragment ions (267 m/z and 193 m/z) that we respectively attributed to the loss of a methyl radical and to the loss of a (trimethylsilyl)oxy radical.

The YFP- and LUC-fusion binary vectors were transformed into A. tumefaciens strain GV3101 by electroporation and colonies were selected on LB agar plates with rifampicin, kanamycin, and gentamicin at 28°C. The positive colonies were confirmed by colony PCR using PCR parameters: 10 min 95°C 1 cycle, 30 s 95°C, 40 s 55°C, 1 min 68°C for 30 cycles, 5 min 68°C 1 cycle, and a final infinite hold at 4°C. For transient expression, cultures of A. tumefaciens were grown overnight in 5 mL LB broth with rifampicin, kanamycin, and gentamicin at 28°C. The cultures were pelleted at 8000 rpm for 5 minutes at room temperature, washed three times with 10 mM MgCl₂, resuspended in 5 mL of induction medium (10 mM MgCl₂, 10 mM MES [pH 5.6], and 200 µM acetosyringone), and incubated at 28°C with shaking at 200 rpm for 3−4 h. A. tumefaciens cultures were diluted in the induction media to OD₆₀₀ = 0.25 and infiltrated into young but fully expanded leaves of five-week-old N. benthamiana plants using a 1 mL needleless syringe. After agroinfiltration the plants were incubated in a growth chamber in long day (16 h of light/8 h of dark) conditions at 22°C for 48 h until leaf discs were harvested.

NpNBS, NpNR, NpTR, LaNBS, LaNR, and LaTR genes were cloned in frame to firefly luciferase fragments in the binary vector pCAMBIA1300:NLuc or pCAMBIA1300:CLuc. The obtained binary vectors were transformed into A. tumefaciens. The desired NLuc- and CLuc-fusion A. tumefaciens combinations were mixed at 1:1 ratio (OD₆₀₀ = 0.25) and co-expressed in N. benthamiana leaves. Split luciferase complementation assays were performed as described by (Chen et al., 2008) with minor modifications. Three millimeter-diameter leaf discs were harvested 48 h after agroinfiltration and floated abaxial side up in 100 µL of degassed water on a white 96-well plate. Samples were supplemented with 1 mM D-luciferin (Sigma-Aldrich) and incubated in the dark for 2 min with gentle shaking and an additional 8 min at rest to quench fluorescence. Luminescence was measured using a Synergy H1 Microplate reader (BioTek) with integration time set to 2 s and imaged using a Gel Doc XR system (Bio-Rad).

NpNBS, NpNR, NpTR, LaNBS, LaNR, and LaTR genes were either fused to the GAL4 DNA binding domain (DNA-BD) in the bait vector pGBKT7 or were fused to the GAL4 activation domain (AD) in the prey vector pGADT7 (Clontech Laboratories). The yeast strain Y2H Gold (Clontech Laboratories) was first transformed with the bait vectors (i.e., NBS, NR, and TR in the pGBKT7 vector) and subsequently with prey vectors (i.e., NBS, NR, and TR in the pGADT7 vector). The transformants were selected on synthetically defined (SD) medium lacking leucine and tryptophan (SD-LW). The interactions were verified by testing the activation of the HIS3, ADE2 and AUR1-C reporter genes on selective media plates lacking histidine and adenine (SD-LWHA) or containing the antibiotic Aureobasidin A (AbA), respectively.

NpNBS was amplified and digested with BamHI and HindIII and ligated into a similarly digested pESC-LEU vector (Stratagene) to produce pESC-LEU : NpNBS. Similarly, NpNR and NpTR were amplified and digested with BamHI/HindIII and BamHI/SalI and ligated into a similarly digested pESC-LEU vector to produce pESC-LEU : NpNR and pESC-LEU : NpTR, respectively. To produce pESC-LEU : NpNBS-NpNR, the NpNR ORF was ligated into pESC-LEU : NpNBS plasmid vector digested with SpeI/BglII, with NBS expressed under control of the Gal1 promoter and NR under the Gal10 promoter. Yeast (INVSc1; Invitrogen) was transformed with all the above constructs and selected on SD-LEU plate for 3 d at 28°C. A single colony was used to inoculate 3 mL of SD-LEU glucose medium and incubated with shaking at 28°C for 2 d. A 500-μL aliquot of starter culture was then used to inoculate to 10 mL of SD-LEU galactose medium containing 250 μM tyramine and 250 μM 3,4-DHBA and incubated with shaking at 28°C for 3 d. A culture lacking substrates was used as a control. The cultures were centrifuged, the medium and cell pellet were extracted with 5 mL ethyl acetate, and the extracts were dissolved in mobile phase (100 μL) of (ammonium acetate 10 mM [pH 5.0], and acetonitrile [60:40]). Papaverine (1000 mg/L) was added to all the reaction as an internal standard for the relative quantification. The analysis of the product (norbelladine) was done using a high-performance liquid chromatography (HPLC) system coupled to a tandem mass spectrometer (MS/MS) as described by (Tousignant et al., 2022).

For protein extraction from N. benthamiana leaves, five leaf discs (1 cm diameter) were frozen in liquid nitrogen, homogenized in 300 µL extraction buffer (100 mM Tris [pH 7.5], 1% [v/v] Triton X-100, 1 mM PMSF, and 0.1x protease inhibitor cocktail), and centrifuged at 17,000 g for 30 min at 4°C. The clear supernatant was collected, and protein concentration was determined using DC protein assay kit (Bio-Rad) according to the manufacturer’s instructions.

For protein extraction from yeast, 5 mL overnight-grown cultures were pelleted at 12,000 g for 5 min at 4°C, resuspended in 250 µL ice-cold lysis buffer (4% [v/v] 5 N NaOH and 0.5% [v/v] β-mercaptoethanol), and incubated with 1x SDS sample buffer (30% [v/v] glycerol, 15% [v/v] β-mercaptoethanol, 37.5% [v/v] 500 mM Tris-HCl [pH 6.8], 0.15% [w/v] SDS, and a few grains of Bromophenol Blue) for 10 min at 95°C.

Equal amounts of protein (100 µg) were fractionated by 10% (v/v) SDS-PAGE. Proteins from gels were transferred onto Polyvinylidene difluoride (PVDF) membrane using Trans-Blot Turbo transfer system (Bio-Rad). The membrane was equilibrated with Tris-buffered saline (TBS) buffer (20 mM Tris, 150 mM NaCl pH 7.6) for 15 min, followed by blocking of membrane for 2 h with TBS buffer containing 0.1% tween 20 (TBST), and 5% skim milk. The PVDF membrane was incubated overnight at 4°C in TBST with 5% milk containing 1:1000 dilution of specific primary antibodies. The primary antibodies used in this study are rabbit anti full-length firefly luciferase antibodies (Sigma-Aldrich), which react with both the N-terminal and C-terminal firefly LUC fragments, mouse anti-GFP/CFP/YFP monoclonal antibody (Cedarlane labs), mouse anti-Myc/c-Myc monoclonal antibody (Santa Cruz Biotechnology), and mouse anti-HA-tag monoclonal antibody (GenScript). After primary antibody incubation, the membrane was washed three times each for 5 min in TBST buffer and incubated for 30 min in TBST containing 5% skim milk and goat anti-rabbit horseradish peroxidase (GAR)-HRP or goat anti-mouse horseradish peroxidase (GAM)-HRP conjugate in 1:10,000 dilutions. The immunoblot was washed three times for 5 min each in TBST buffer and developed using clarity Western ECL substrate (Bio-Rad). Finally, the membrane was washed twice with TBST and stained with Ponceau S stain [0.5% (w/v) Ponceau S (Sigma-Aldrich) in 1% (v/v) acetic acid] for 1 min and photographed using Gel Doc XR system (Bio-Rad).

To visualize NpNBS, NpNR, LaNBS, and LaNR subcellular localization, the YFP fusion proteins were expressed via A. tumefaciens in leaves of 5-week-old N. benthamiana plants. Forty-eight hours post infiltration, the abaxial epidermis of leaf discs were placed on a microscopic slide in a water drop, covered by a cover slip, and imaged immediately. Protein localization was visualized by a confocal laser scanning microscope (Leica TCS SP8; Leica Microsystems) with a 40X/1.30 oil immersion objective. Images were first processed with Las AF Lite software (Leica Microsystems). CFP was used as a control for colocalization (Kruse et al., 2010). YFP was excited with an argon laser at 488 nm, while CFP was excited with a diode laser at 405 nm. Emission was detected with a spectral detector set between 500 and 525 nm for YFP and between 420 and 490 nm for CFP. Chlorophyll autofluorescence was observed with an excitation wavelength of 552 nm and the emission of fluorescence signals were detected from 630 to 670 nm. The combined images were generated using the Las X software (Leica Microsystems).

Total RNA was isolated from bulbs, roots, stems, leaves, and flowers using the TRIzol reagent (Invitrogen). Briefly, 100 mg of tissues were frozen in liquid nitrogen, fully ground, and homogenized in 1 mL of TRIzol using a mortar and pestle. The liquid was transferred to a microcentrifuge tube, incubated 5 min at room temperature and extracted with 200 µL chloroform. Following centrifugation at 12,000 g for 15 min at 4°C, the upper phase containing RNA was transferred to a fresh tube. The RNA was precipitated with 500 µL of isopropanol for 10 min at room temperature and centrifuged at 12,000 g for 10 min at 4°C. The RNA pellet was washed twice with 1 mL of 75% ethanol (with DEPC water) and centrifuged at 7500g for 5 min at 4°C. Finally, the RNA pellet was air dried and suspended in 40 µL of DEPC-treated water. The quality and quantity of RNA extracted from different tissues were verified on NanoPhotometer (Implen) and 1.5% (w/v) agarose gel electrophoresis. RNA samples (1 µg) were reverse transcribed using SensiFAST cDNA synthesis kit (Bioline) according to manufacturer’s protocol and subjected to Real-time quantitative PCR (RT-qPCR) using gene-specific primers ( Supplementary Table S2 ). The experiments were performed in triple technical replicates of each plant sample. A total reaction volume of 20 μL containing 1x SensiFAST SYBR Lo-ROX mix (Bioline), 200 μM of each forward and reverse primer, and 2 µL of template cDNA (50 ng/µL) was used for RT-qPCR analysis. RT-qPCR was performed on CFX Connect Real-Time PCR System (Bio-Rad). Amplification conditions were 95°C for 3 min 1 cycle, 95°C for 10 s, and 60°C for 30 s for 40 cycles followed by dissociation step 95°C for 10 s, 65°C for 5 s and 95°C for 5 s. LaGAPDH and NpHISTONE were used as internal reference genes for La and Np, respectively. To verify the specificity of the primers, a melting-curve analysis was also performed. The threshold cycle (C_T) value of each gene was normalized against the C_T value of the reference genes. Mean C_T values calculated from the technical triplicates were used for quantification of relative gene expression involving the comparative C_T method (Pfaffl, 2001). The results were analyzed, and the statistical error was calculated using CFX Maestro software (Bio-Rad).

Sequence data from this article can be found in GenBank under the following accession numbers: N. papyraceus norbelladine synthase (NpNBS; MZ054104), N. papyraceus noroxomaritidine/norcraugsodine reductase (NpNR; MF979872), N. papyraceus histone (NpHistone; MF979875), L. aestivum norbelladine synthase (LaNBS; MW971977), L. aestivum noroxomaritidine/norcraugsodine reductase (LaNR; MW971981), L. aestivum glyceraldehyde-3-phosphate dehydrogenase (LaGAPDH; MW971984), Narcissus pseudonarcissus ‘King Alfred’ norbelladine synthase (NpKANBS; AYV96792), N. pseudonarcissus noroxomaritidine/norcraugsodine reductase (NpKANR; KU295569).

The full-length cDNAs of NpNBS, LaNBS, NpNR, and LaNR candidate genes were obtained from previously reported transcriptome sequencing of N. papyraceus and L. aestivum (Hotchandani et al., 2019; Tousignant et al., 2022). The open reading frame (ORF) of both NpNBS and LaNBS gene is 480 bp and encodes a 159-amino acid protein ( Figure S1 ). In silico protein analysis by DNAMAN analysis software indicated that NpNBS and LaNBS candidates had a predicted molecular weight (MW) of 17.4 kDa and a theoretical isoelectric point (pI) of 5.3 and 5.1, respectively. Multiple amino acid sequence alignments of NpNBS and LaNBS candidates with the already characterized protein NpKANBS (Singh et al., 2018) and norcoclaurine synthase from T. flavum (TfNCS) (Ilari et al., 2009) showed that NpNBS and LaNBS share over 41% of amino acids sequence identity with the ortholog TfNCS ( Figure S1 ). In addition, NpKANBS, NpNBS and LaNBS share 83% identity between each other, while NpNBS and LaNBS share 85% of identity ( Figure S1 ). Domain search using NCBI-conserved-domain-search service revealed the presence of conserved Bet v1 and Pathogenesis-Related (PR-10) protein domains in both NpNBS and LaNBS homologs. Both contained the phosphate-binding loop (P-loop) glycine-rich region ( Figure S1 ), a conserved ligand-binding domain of Bet v1 protein family (Fernandes et al., 2013). As reported previously, the alignment showed that TfNCS catalytic residues Tyr108, Glu110 and Lys122 are well-conserved in NpNBS and LaNBS, corresponding to Tyr68, Glu71, and Lys83, respectively, in their sequences ( Figure S1 ) (Ilari et al., 2009; Singh et al., 2018). The fourth catalytic residue Asp141 from TfNCS is replaced by hydrophobic Ile102 in both NpNBS and LaNBS, as reported previously for NpKANBS (Singh et al., 2018). Homology modeling of the enzymes from both N. papyraceus (Np) and L. aestivum (La) revealed a striking structure similarity with superimposed template crystal of TfNCS, analogous to the overall structure of Bet v1-like proteins family ( Figures 2A ; S2A ; S3A-D ). As TfNCS, both NBS homologs are composed by seven-stranded antiparallel β-sheets, two long C- and N-terminal helices and two short ones, enclosing a cleft with polar residues at its entrance and hydrophobic residues in its core ( Figures 2A, B ; S2A, B ).

The ORF of NpNR and LaNR candidate genes is 810 bp, both encoding a 269-amino acid protein ( Figure S4 ). NpNR and LaNR homologs have a predicted molecular weight (MW) of 29 kDa and theoretical pI of 6.0 and 5.4 respectively. Sequence comparison and domain search confirmed the presence of conserved short-chain dehydrogenases/reductases (SDR) domain in NpNR and LaNR candidates. NpKANR, NpNR and LaNR candidates share over 76% of identity with each other, while NpNR and LaNR share 90% of identity ( Figure S4 ). Like all classical SDRs, NpNR and LaNR contain a TGXXX[AG]XG cofactor binding motif and a YXXXK active site motif, with the Tyr and Lys of the active site serving as critical catalytic residues ( Figure S4 ). Structurally, predicted models of NRs from both species are formed by a seven-stranded parallel β-sheet inserted between a pair of three α-helices ( Figures 2C ; S2C ). A long tunnel is shaped at the C-termini of β-strands partially wrapped by the α-helices and loops that elevate beyond the β-sheets ( Figures 2C, D ; S2C, D ). At the site of ligand interaction, the tunnel expands into a larger pocket where aromatic Phe214 (216 in NpKaNR) is conserved, Ala112 replaces Tyr114, both possibly involved in polycyclic substrate orientation and binding. At the extremity, Glu224 (226), Arg265 (267), Cys162 (164), His170 (172) are preserved. A strong similarity in structure of both NpNR and LaNR homologs with N. pseudonarcissus noroxomaritidine/norcraugsodine reductase (NpKANR, 5FF9) was noted ( Figures S5A-D ).

To identify other reductases that could catalyze similar reduction reactions, we searched for homologs of NR in the transcriptome sequences of N. papyraceus (Hotchandani et al., 2019) and L. aestivum (Tousignant et al., 2022), and identified a candidate tropinone reductase (TR) belonging to the same SDR superfamily in both species. The ORF of the NpTR homolog is 825 bp and encodes a 274-amino acid protein while the LaTR is 816 bp and encodes a 271-amino acid protein ( Figure S4 ). NpTR and LaTR candidates have a predicted molecular weight (MW) of 30 kDa and 29.5 kDa and theoretical pI of 6.9 and 8.4 respectively. NpTR and LaTR contain the TGXXX[AG]XG cofactor binding motif and the YXXXK catalytic active site motif ( Figure S4 ). Gly206 is replaced by a tryptophan at position 203 and 200 in NpTR and LaTR respectively, although this residue was conserved in active SDR/tropinone reductases (Roth et al., 2018). Multiple sequence alignments showed that NpKANR, NpNR, LaNR, NpTR, and LaTR share over 58% of identity between each other. Predicted TRs structures are similar to NRs with some key differences in the ligand active site, including replacement of Phe216 from NpKANR by Arg213 in NpTR and by Leu210 in LaTR, while Tyr114 is replaced by Asn111 in NpTR and Asn108 in LaTR ( Figures 2E, F ; S2E, F ; S6 ; Table S1 ).

To shed light on the reactions involved in norbelladine synthesis ( Figure 1 ), we studied the interaction of NBS and NR candidates with their respective proposed ligands through molecular docking analysis in silico. Tyramine and 3,4-DHBA were docked with scores of -5.03 and -5.23 kCal/mol inside the NBS pocket respectively ( Figures 3 ; S7 ; Table S1 ). Most of the obtained poses displayed the same ligand orientations where 3,4-DHBA and tyramine adopted a stack configuration with their aromatic rings lying on near-to-parallel planes, similarly to dopamine and hydroxybenzaldehyde in crystalized TfNCS (2VQ5) ( Figures 3A, B ; S7A, B ). The carbonyl group of 3,4-DHBA faced the amine group of tyramine. Lys83, whose proposed role is to intercept the carbonyl group of the aldehyde substrate, interacted with 3,4-DHBA carbonyl end through hydrogen bonding ( Figure 3B ; Table S1 ). At the other end, the hydroxyl group of C4 was hydrogen-bonded with the possibly base-acting residue Glu71. All predicted poses implied interaction with Glu71 and Lys83 strengthening the probability of their key role in the catalytic mechanism. Tyramine was hold in place by stacking interaction, and hydrogen bonding between its amine and 3,4-DHBA carbonyl group. PLIP software (Adasme et al., 2021) predicted additional hydrophobic interactions between Phe73, Thr85, Phe104 and Ile143, and 3,4-DHBA, as well as two hydrogen bonds between tyramine and Ser31 and Tyr59 ( Figures 3B ; S7B ; Table S1 ). In general, these predicted interactions and spatial arrangements of the ligands inside NBS are consistent with the reaction proposed by Ilari et al., 2009 that would lead to norcraugsodine biosynthesis.

Following its formation, norcraugsodine would be transferred from NBS to NR active site to be reduced into norbelladine ( Figure 1 ) with NADPH as the electron donor. Upon docking, NADPH positioned into LaNR and NpNR modeled active site following a similar arrangement compared to crystalized reductases such as 5FF9 (NpKANR) with a score of -11.51 kCal/mol ( Figures 3C, D ; S7C, D ; Table S1 ). It interacted with Gly36, Cys53, Arg55, Val109, Ile35, His170, Thr111, Lys177, Asn108, Thr206. PLIP confirmed these interacting residues and additionally predicted hydrophobic interaction with Ile35, hydrogen bonding with Lys33, Cys79, Gly110, Thr208, Val209 and Gln210 along with π-Cation interactions and salt bridges with Arg55 ( Table S1 ). Close to the ligand-binding site, the nicotinamide ring faces Ile158, Pro203, Gly204 on one side, and the substrate-binding pocket on the B-side. As it was the case for docked noroxomaritidine in the active site of noroxomaritidine reductase, docking predicted that norcraugsodine binds to the active site of NR by a combination of polar and non-polar interactions. Docked norcraugsodine displayed two possible conformations in NpNR and LaNR: either bended or diagonal. For both conformations, the amine group of norcraugsodine was positioned close to C4 of NADPH and to Tyr173, obtaining a docking score of -6.36 and -6.15 kCal/mol for NpNR and LaNR respectively ( Figures 3C, D ; S7C, D ; Table S1 ). In both cases, norcraugsodine phenol cycle was located near Phe214, and its dihydroxybenzene group was positioned close to Glu224, interacting with His170 and Arg265. PLIP predicted additional hydrophobic interactions of norcraugsodine with Leu114, Ile205, Phe214 and Leu215, and hydrogen-bonding with Glu224 ( Figures 3C, D ; S7C, D ; Table S1 ). These interactions are consistent with the proposed reduction of the imine functional group of norcraugsodine to yield norbelladine by a mechanism involving NADPH and the catalytic residues Tyr173 and Lys177.

To examine the NBS, NR, and TR candidate protein function from both N. papyraceus and L. aestivum, the full-length ORFs were PCR-amplified from N. papyraceus and L. aestivum bulb cDNA, cloned, and expressed proteins were purified ( Figure S8 ). We were unable to purify the LaTR enzyme in our experimental conditions. We first tested the NBS, and NR purified proteins from N. papyraceus (Np) and L. aestivum (La) separately in assays containing tyramine, 3,4-DHBA, and NADPH. The resulting assay products were subjected to HPLC-MS/MS analysis using positive electrospray ionization mode (ESI+). Before injecting the assays, norbelladine standard was injected at 1 mg/L and predicted parent-ion was observed with a mass-to-charge ratio (m/z) of 260 [M + H]+ at 3.435 min ( Figures 4A ; S9A-R ). Fragmentation of the norbelladine parent-ion yielded to major ion fragments of m/z 121, 123 and 138 using 10 V as collision energy. Multiple reaction monitoring (MRM) transitions of 260 → 138 m/z and 260 → 121 m/z were selected, optimized using MassHunter Optimizer software, and used as quantifier and qualifier ions respectively. We could not observe any parent-ion mass for the norcraugsodine standard, despite repeated trials. We inferred that norcraugsodine was highly unstable in solution and/or thermolabile so the heat used during the ionization process in the HPLC-MS/MS source could lead to its degradation. Enzyme assays containing recombinant NBS candidate protein from Np or La, 3,4-DHBA, tyramine, and with or without NADPH yielded a peak at 3.435 min in MRM acquisition mode on HPLC-MS/MS which was the same retention time as norbelladine standard ( Figures 4A ; S9E, F ). Similarly, for the assays examining the NR candidates, tyramine and 3,4-DHBA were incubated with NR and NADPH, and the resulting product showed a peak with the same retention time as norbelladine standard in the reaction mixture of Np enzyme or La enzyme ( Figures 4A ; S9G, P ). Comparison of the mass spectrum obtained after the fragmentation of authentic norbelladine standard using collision-induced-dissociation (CID) with 10V and the mass spectrum of the products obtained from NBS and NR reactions showed fragmentation patterns to be the same for both, thus confirming the identity of the enzymatic product ( Figure S10 ). In assays lacking substrates or enzyme, no norbelladine was detected ( Figures 4A ; S9B, D ). Similarly, MBP tag alone protein purified from E. coli transformed with empty pMAL-c2X vector showed no activity ( Figures 4A ; S9C ). The papaverine internal standard (1000 mg/L) was used to normalize each signal obtained by LC-MS/MS to have accurate relative quantification for the produced norbelladine. The relative quantification was made by comparing peak area ratios (i.e., the peak area of norbelladine divided by the peak area of papaverine internal standard) of different samples with each other. As expected, NR assays lacking NADPH resulted in no norbelladine production ( Figures 4A ; S9C ). As reported before (Kilgore et al., 2016b; Singh et al., 2018; Tousignant et al., 2022), our results confirm that both NBS and NR homologs alone from N. papyraceus and L. aestivum are able to catalyze the reaction of condensation/reduction of tyramine and 3,4-DHBA to produce a low amount (0.07 for NBS and 0.08 for NR relative peak area ratio)of norbelladine ( Figure 4A ). To examine if NBS and NR work together for the condensation of tyramine and 3,4-DHBA into norcraugsodine followed by its reduction into norbelladine ( Figure 1 ), we tested NBS and NR purified enzymes together in one-step assay and in two-step sequential manner. Our inability to detect the intermediate norcraugsodine through HPLC-MS/MS prompted us to examine the norbelladine production in each assay mixture. We measured the substrates and equal amount of added papaverine (internal standard) in all reaction mixtures for relative quantification of the observed product ( Figures S9A-R ). Norbelladine production was significantly (six-fold, 0.49 relative peak area ratio) higher when both NpNBS and NpNR were present in a single reaction, compared to assays with NpNBS or NpNR separately ( Figures 4A, B ; S9E-J ). Similarly, we observed four-fold higher (0.31 relative peak area ratio) for norbelladine production when LaNBS and LaNR were present in a single reaction, compared to assays with LaNBS or LaNR individually ( Figures 4A, B ; S9O-R ). When the reactions were performed in a two-step manner: NBS first followed by NR, we observed two-fold higher norbelladine level (0.92 relative peak area ratio) than when enzymes were together in a single-step for both Np and La enzymes ( Figures 4A, B ; S9I, J, Q, R ), which was 12 fold (Np) and 8 fold (La) higher in comparison to assays with the enzymes individually ( Figures 4A, B ). In the reverse order, stepwise reaction for both species consisted of NR first followed by NBS and yielded lower levels (0.3 relative peak area ratio) of norbelladine than observed with the original sequence (i.e., NBS first followed by NR), but still produced higher amounts of norbelladine compared to single enzyme reactions ( Figures 4A, B ; S9J, K ). Our results indicate that NBS and NR function together optimally in a sequential manner (NBS first followed by NR) to produce norbelladine. To further confirm the specificity of NR enzyme in norbelladine production, we tested the NpTR purified protein individually and together with NpNBS in assays containing 3,4-DHBA, tyramine, and NADPH. We observed no norbelladine formation in assays containing NpTR enzyme, and assays with both NpNBS and NpTR in one-step or two-step yielded similar level (0.07 relative peak area ratio) of norbelladine compared to assays containing only NpNBS enzyme ( Figures 4A, B ; S9L-N ). These results confirm the role of NBS and NR together to channel the substrates effectively for the condensation/reduction sequence leading to the formation of norbelladine.

Direct analysis of norcraugsodine by HPLC-MS/MS or GC-MS was inconclusive ( Figure S11A ). Therefore, we derivatized the norcraugsodine using BSTFA reagent (N,O-bis(trimethylsilyl)trifluoroacetamide) ( Figure S12 ). Following derivatization, two signals corresponding to tris-TMS-norcraugsodine (24.79 minutes) and the bis-TMS-3,4-dihydroxybenzaldehyde (9.28 minutes) resulting from the hydrolysis of tris-TMS-norcraugsodine were observed by GC-MS ( Figures S11B-D ; S12 ). La and Np NBS, and NR purified proteins were tested separately in enzymatic assays containing tyramine, 3,4-DHBA, and NADPH. The resulting assay products were dried, reconstituted in acetonitrile, derivatized using BSTFA, and injected into the GC-MS using electron ionization (EI) at 70 eV. As expected, in assays lacking substrates or enzymes, tris-TMS-norcraugsodine was not detected ( Figures S13A, C ). Similarly, MBP tag alone protein purified from E. coli transformed with empty pMAL-c2X vector showed no activity ( Figure S13B ). Surprisingly, we did not detect any tris-TMS-norcraugsodine in assays with single NBS or NR enzyme with or without NADPH ( Figures S13D-K ). However, tris-TMS-norcraugsodine was detected in assays containing both enzymes, in a one-step or two-step fashion, from both Np and La species ( Figures S13L-O ). In both cases, the detected tris-TMS-norcraugsodine fragmentation pattern matches with the tris-TMS-norcraugsodine standard ( Figure S14 ). In the assays with the reverse order of enzymes, i.e., NR first followed by NBS, a very small signal for tris-TMS-norcraugsodine was detected ( Figures S13P, Q ), but the fragmentation pattern did not exactly match with the standard due to proximity with the baseline. As expected, the TR enzyme alone or with NBS did not produce any detectable amounts of tris-TMS-norcraugsodine ( Figures S13R-T ). These results confirm that both NBS and NR are specifically required to harness norcraugsodine and for its efficient reduction into norbelladine.

Previous studies suggest that the norcoclaurine synthase (NCS) proteins assemble as dimers to be catalytically active (Samanani and Facchini, 2002; Samanani et al., 2004; Vimolmangkang et al., 2016). Similarly, the NR protein of N. pseudonarcissus was shown to exist as a tetramer through crystal structure of the enzyme (Kilgore et al., 2016b). Based on these observations, we hypothesized that NBS and NR form a dimer which impacts their activity. To explore the ability of using split-luciferase-complementation assay (SLCA) in testing the interactions of NBS and NR in N. benthamiana leaves, we first examined the physical interactions between NpNBS-NpNBS, LaNBS-LaNBS, NpNR-NpNR, and LaNR-LaNR. NpNBS, LaNBS, NpNR, and LaNR were fused to the N-terminal half of the luciferase protein (NLuc) and co-expressed via Agrobacterium in N. benthamiana leaves while NpNBS, LaNBS, NpNR, and LaNR were fused to the C-terminal half of luciferase (CLuc) protein. As negative controls, NpNBS-NLuc, LaNBS-NLuc, NpNR-NLuc, and LaNR-NLuc were co-expressed with CLuc empty vector and CLuc-NpNBS, CLuc-LaNBS, CLuc-NpNR, and CLuc-LaNR were co-expressed with NLuc empty vector ( Figures 5A-D ). Expression of all the fusion proteins was validated by western blot analysis ( Figures S15A, B ). The homodimeric interactions were monitored by measuring luminescence 48 h after agroinfiltration of the tested protein pairs. Co-expression of NpNBS-NLuc with CLuc-NpNBS, LaNBS-NLuc with CLuc-LaNBS, NpNR-NLuc with CLuc-NpNR, and LaNR-NLuc with CLuc-LaNR resulted in emission of significantly higher luminescence compared to the negative controls indicating a physical interaction in planta between NBS-NBS and NR-NR fusion proteins ( Figures 5A-D ). One explanation for the production of norbelladine in reactions containing NBS and NR is that NR functions as an enzyme that acts on norcraugsodine produced from tyramine and 3,4-DHBA by NBS. Alternatively, NR may alter the catalytic properties of NBS through allosteric regulation, which allows NBS to form norbelladine, or vice versa, but it remains formally possible that NBS and NR are present together in a metabolon, NR playing a regulating role by guiding the substrates to the imine intermediate followed by reduction to norbelladine. To test the importance of physical interactions for norbelladine biosynthesis, the interactions of NpNBS and LaNBS with full-length NpNR and LaNR were examined in N. benthamiana leaves by split-luciferase-complementation-assays. NpNBS and LaNBS were fused to the N-terminal half of the luciferase protein (NLuc) and co-expressed via Agrobacterium in N. benthamiana leaves along with NpNR and LaNR fused to the C-terminal half of luciferase (CLuc). As a control, NpNBS-NLuc and LaNBS-NLuc were co-expressed with CLuc-NpTR and CLuc-LaTR, respectively. As negative controls, NpNBS-NLuc and LaNBS-NLuc were co-expressed with CLuc empty vector and CLuc-NpNR and CLuc-LaNR were co-expressed with NLuc empty vector ( Figures 5A-D ). Expression of the examined fusion proteins was confirmed by western blot analysis ( Figures S15A, B ). Protein-protein interactions in planta were quantified by measurements of luminescence 48 h after agroinfiltration. In agreement with the enzymatic assays, co-expression of NpNBS-NLuc with CLuc-NpNR and LaNBS-NLuc with CLuc-LaNR resulted in emission of significantly higher luminescence compared to the negative controls and the control protein CLuc-NpTR and CLuc-LaTR ( Figures 5A-D ), demonstrating a physical interaction in planta between NBS and NR fusion proteins.

To validate the interaction detected in planta and to check if the observed interactions are direct, the interactions between NBS-NBS, NR-NR, NBS-NR, and NBS-TR were then examined by yeast two-hybrid system. NBS, NR and TR from both plant species were fused to both bait and prey plasmids. Expression in yeast of bait and prey proteins was confirmed by western blot analysis ( Figures S15C, D ). Similar interactions between NBS-NBS, NR-NR and NBS-NR were observed ( Figures 5E, F ) when the same protein pairs from both plant species were expressed in yeast as bait and prey proteins in agreement with the observed interaction in planta. As observed in planta, no interaction was found between NBS-TR in yeast ( Figures 5E, F ). Taken together, these results obtained in different experimental systems indicated regulatory interactions between NBS and NR and support the possibility that NBS and NR function as heteromultimeric proteins.

Functional analysis of NBS and NR via RNAi was not possible because Amaryllidaceae transformation has not been achieved. To demonstrate NBS and NR activity in vivo, we reconstituted norbelladine biosynthesis in yeast. Yeast cultures expressing NpNBS, NpNR, and NpTR as single enzymes or NpNBS together with NpNR or NpTR were fed with tyramine and 3,4-DHBA, and norbelladine production was monitored in the yeast cell extracts. We measured the equal amount (1000 mg/L) of added papaverine (internal standard) in all reaction mixtures for relative quantification of the observed norbelladine product. Surprisingly, we did not detect norbelladine in any of the yeast cultures expressing single NBS or NR enzyme ( Figures 6A, B ). However, we detected low levels of norbelladine (0.22 relative peak area ratio) in yeast cultures expressing both NBS and NR ( Figures 6A, B ). This confirms that both enzymes are required for norbelladine production in an in vivo model.

The NBS protein fused with green fluorescent protein (GFP) from both L. aestivum and N. papyraceus was recently shown to localize to the cell cytoplasm and nucleus (Tousignant et al., 2022). Similarly, we found NR lacked any predicted signal peptides. To investigate NBS and NR subcellular localization, the NBS and NR coding regions from both N. papyraceus and L. aestivum plants were fused upstream to the yellow-fluorescent-protein gene (YFP). The NpNBS-YFP, LaNBS-YFP, NpNR-YFP, and LaNR-YFP fusions were transiently expressed in leaves of N. benthamiana plants via A. tumefaciens, and their localization was monitored by confocal fluorescence microscopy. The cyan fluorescent protein (CFP), which localizes to the cytoplasm and nucleus (Kruse et al., 2010), was used as a control. Expression of all the fusion proteins was validated by western blot ( Figure S16 ). As shown in Figure 7A , the NBS-YFP and NR-YFP fusion proteins from both plant species localized in the cell cytoplasm and nucleus like CFP. These results suggest that both NBS and NR are distributed to the same nucleocytoplasm cellular compartment. To further confirm the colocalization pattern of NBS and NR, the NR coding regions from both N. papyraceus and L. aestivum plants were fused upstream to the cyan fluorescent protein (CFP) and co-expressed via Agrobacterium in N. benthamiana leaves while NpNBS and LaNBS were fused to the YFP, and their localization pattern was monitored by confocal microscopy. Expression of all the fusion proteins was validated by western blot ( Figure S16 ). Similar profiles of fluorescence pattern in the cell cytoplasm and nucleus were observed for NBS-YFP and NR-CFP ( Figure 7B ). These results confirm that NBS and NR which lack predicted signal peptides, colocalized to both the cytoplasm and the nucleus of the cell.

The expression profiles of the NBS and NR in different tissues including bulb, root, stem, leaf, and flower were evaluated by using quantitative real-time PCR (qRT-PCR) analysis. The results showed that NBS was ubiquitously expressed in all tissues detected, with the highest expression levels in bulb and root ( Figures 8A, B ). The expression patterns of NpNBS and LaNBS were similar, mRNA accumulated in high amount in bulb and root, and low transcript levels were detected in leaf, stem, and flower ( Figures 8A, B ). The NR expression pattern was different. LaNR and NpNR mRNA specifically accumulated at high levels in bulbs, and in low amount in other tested tissues ( Figures 8C, D ). We further checked the expression pattern of TR in different tissues to compare tissue expression profiles between NR and TR and found that NpTR and LaTR genes were widely expressed in all the examined tissues ( Figures S17A, B ). These results suggest that the highest transcript levels of NR were in bulbs, which parallels with the high expression for NBS transcripts in the bulbs. This high expression of NBS and NR aligns with the higher expression of other AA biosynthetic genes in bulbs of N. papyraceus (Hotchandani et al., 2019) and L. aestivum (Tousignant et al., 2022).