Engineering biosynthesis of the anticancer alkaloid noscapine in yeast

Noscapine is a potential anticancer drug isolated from the opium poppy Papaver somniferum, and genes encoding enzymes responsible for the synthesis of noscapine have been recently discovered to be clustered on the genome of P. somniferum. Here, we reconstitute the noscapine gene cluster in Saccharomyces cerevisiae to achieve the microbial production of noscapine and related pathway intermediates, complementing and extending previous in planta and in vitro investigations. Our work provides structural validation of the secoberberine intermediates and the description of the narcotoline-4′-O-methyltransferase, suggesting this activity is catalysed by a unique heterodimer. We also reconstitute a 14-step biosynthetic pathway of noscapine from the simple alkaloid norlaudanosoline by engineering a yeast strain expressing 16 heterologous plant enzymes, achieving reconstitution of a complex plant pathway in a microbial host. Other engineered yeasts produce previously inaccessible pathway intermediates and a novel derivative, thereby advancing protoberberine and noscapine related drug discovery.

P lant natural products span a large and diverse class of compounds with critical ecological functions 1 and important therapeutic niches 2,3 . Due to the structural complexity of many of these compounds, chemical synthesis routes have remained challenging, and the supply of plant natural products still largely relies on extraction from plant biomass. Derivatives of target compounds are difficult to obtain from plant-based production platforms because engineering plants to accumulate pathway intermediates and novel functionalized products is challenging 4 . The recent engineering of microbial production platforms for specific plant natural products, such as medicinal opioids, artemisinin and precursors to Taxol and vinblastine [5][6][7][8][9][10][11] , highlights an alternative approach to sourcing these valuable compounds. However, this approach requires knowledge of the corresponding biosynthetic routes in plants.
Currently, in planta gene silencing studies are used to verify gene identity and identify enzyme function within the pathway; and in vitro biochemical studies are used to understand the function and study enzymatic mechanisms 12,13 . However, for full elucidation of complex pathways these methods can encounter limitations. In planta characterization of individual enzyme activities can be impacted by the complex endogenous protein and metabolite background, and in vitro characterization can be limited by the availability of pathway intermediates as substrates and standards, many of which are difficult to synthesize chemically or extract from producing plants. A number of tools have recently been developed that enable the efficient reconstitution of known plant natural product pathways in the budding yeast Saccharomyces cerevisiae [14][15][16][17] . These tools are advancing the reconstitution of complex plant pathways in yeast [18][19][20] , with recent examples comprising up to 13 heterologous enzymatic steps from precursors supplied in yeast 7,8 . Therefore, step-wise reconstitution of plant pathways in yeast -leveraging the integration of recent methodological advances, including inexpensive DNA synthesis, rapid combinatorial pathway assembly strategies [21][22][23] and biological parts tailored to optimizing expression, folding and processing of plant enzymes [14][15][16][17]24 -can be applied to obtain sufficient levels of intermediate metabolites along a pathway for structural verification, and at the same time provide substrate for the subsequent enzymatic steps. This synthetic approach can be used to connect the upstream pathway identification in planta and the downstream in vitro characterization of enzyme mechanism.
A 10-gene cluster involved in the biosynthesis of noscapine, 1 was recently discovered in Papaver somniferum 37 . The noscapine gene cluster, which represents one of the most complex plant gene clusters identified to date, was initially characterized by gene silencing experiments in P. somniferum 37 (Fig. 1). Subsequent series of in vitro characterization experiments were able to correct the ordering of several steps (that is, the enzymatic steps catalysed by CYP82X1, CYP82X2 and PsAT1) and proposed an updated biosynthetic pathway [47][48][49][50] (Fig. 1); however, the sequence of the final steps in the pathway remained unclear and the enzyme catalysing the conversion of narcotoline to noscapine remained unidentified.
To provide an alternative route to the production of noscapine, and to overcome the inaccessibility of many pathway intermediates and potentially interesting derivatives, we reconstitute the noscapine gene cluster in the yeast host S. cerevisiae. Our synthetic approach provides experimental support for the previously proposed reaction matrix of the final three steps of the pathway and suggests that the remaining 4 0 -O-methylation step is catalysed by two different O-methyltransferases together, and to the best of our knowledge, is likely to be the first known O-methylation natively catalysed by an O-methyltransferase heterodimer. Our reconstituted noscapine gene cluster also provides sufficient levels of pathway intermediates to support structural verification, which allows us to validate the secoberberine structures synthesized by CYP82X1. In addition, the 14-step biosynthesis of noscapine from the early benzylisoquinoline alkaloid norlaudanosoline is achieved in yeast by optimizing the expression of 16 plant enzymes, reaching a titre of 1.64±0.38 mM noscapine.
We started the reconstitution of the noscapine biosynthetic pathway from canadine, as the biosynthesis of 2, from norlaudanosoline, 12, has previously been engineered in S. cerevisiae 52,53 . The genes related to the synthesis of noscapine from canadine were introduced into S. cerevisiae through either plasmid or yeast artificial chromosome constructs or chromosomal integration, as indicated. The engineered yeast strains were then cultured in synthetic defined medium (SDM) supplemented with 250 mM of racemic 2, which is assumed to contain equal amounts of each enantiomer. The medium is then directly analysed for pathway metabolites after 72 h of growth by liquid chromatography/mass spectrometry (LC/MS) analysis. Previous studies indicated that using galactose as a carbon source improves alkaloid production in yeast from fed substrates, which may be related to the metabolic or transcriptional responses induced by galactose 54 . In addition, since the GAL1 promoter was used in several constructs for reconstituting the biosynthesis of noscapine, galactose was used as the carbon source in all the in vivo assays in this work. Ring opening Yeast strains were engineered to express PsTNMT, CYP82Y1 and a cytochrome P450 reductase partner (CPR) from Arabidopsis thaliana (AtATR1) that was previously shown to function with diverse plant P450s (refs 52,55). When grown in the presence of 250 mM fed racemic 2 at 30°C, the engineered strains synthesized 4 (m/z þ ¼ 370; Figs 1 and 2a and Supplementary Fig. 1c), in agreement with recent in vitro biochemical characterizations of TNMT and CYP82Y1 (refs 48,56). Nuclear magnetic resonance (NMR) analyses further confirmed the identity of 4 synthesized in the engineered yeast ( Supplementary Fig. 2 and Supplementary  Table 1). However, the synthesis of 4 was inefficient (21.9 ± 5.5 mM, Fig. 3a), limited by the CYP82Y1-catalysed step (B20% conversion from B100 mM 3) when that enzyme is expressed from the GAL1 promoter at 30°C (ref. 57). To ensure that our strains make sufficient levels of 4 for conversion through the remaining downstream pathway steps, we optimized CYP82Y1 activity by testing a combination of different N-terminal tags, promoters and growth temperatures. While most of the N-terminal tags did not improve CYP82Y1 activity, the N-terminus of a related plant cytochrome P450, (S)-cis-Nmethylstylopine-1-hydroxylase (MSH), resulted in slightly enhanced activity from the chimeric protein (26.9 ± 2.0 mM) when expressed downstream of the GAL1 promoter at 30°C ( Fig. 3a and Supplementary Fig. 3a). Both CYP82Y1 and the MSH N-terminus swapped CYP82Y1 (CYP82Y1A) exhibit higher activity at 25°C than 30°C. The enhanced activity in synthesizing 4 at lower temperature (P ¼ 0.0021) and the N-terminus engineering (P ¼ 0.0436) are both significant through a standard two-way analysis of variance (Fig. 3a, Supplementary Fig. 3a and Supplementary Table 2). The highest CYP82Y1 activity (96.5±3.0 mM) was observed when CYP82Y1A was expressed from either the strong glycolytic GPD promoter or the late stage HXT7 promoter 58 at 25°C (Fig. 3a and Supplementary Fig. 3a). Our optimization of CYP82Y1 activity allowed us to achieve B39% conversion from a racemic mixture of 2 to 4.  Reconstituting CYP82X2, PsAT1 and CYP82X1 activities.
In addition, the structure of 7 synthesized by CYP82X1 has not been confirmed due to the lack of authentic standards and limited quantity of the compound achieved in the in vitro investigations 47 . We utilized our optimized 4-producing strain to achieve the microbial production of secoberberines to resolve the sourcing of this group of compounds.
We first engineered the 4-producing strain to express CYP82X1 for the synthesis of 5 0 . When grown in the presence of 2 at 25°C, the engineered strain synthesized 5 0 (m/z þ ¼ 386), which is confirmed by NMR analyses (Fig. 2b Supplementary Fig. 3b). We optimized CYP82X1 by testing its activity under different promoters, and observed the highest activity (22.9 ± 1.5 mM, Supplementary  Fig. 3b) when the enzyme was expressed from a strong constitutive promoter (PGK1) at 25°C ( Supplementary Fig. 3b), compared with expression from the GAL1 promoter (paired t-test, P ¼ 0.00068).
When CYP82X2 is introduced into the 4-producing yeast strain, 5 (m/z þ ¼ 386) (Fig. 2b and Supplementary Fig. 1d) is synthesized, and the identity of 5 was confirmed through NMR analyses (Supplementary Fig. 5 and Supplementary Table 4). We optimized CYP82X2 for the synthesis of 5 by testing its activity under different promoters, and observed the highest activity (42.1±2.2 mM) when the enzyme was expressed from the HXT7 promoter at 25°C (Fig. 3b and Supplementary Fig. 3c). By engineering more copies of the CYP82X2 expression construct into our 5-producing strain, the engineered strain exhibits  significantly higher activity towards the synthesis of 5 (71.0 ± 0.8 mM, paired t-test, P ¼ 0.000027; Fig. 3b). The subsequent addition of PsAT1 in the 5-producing yeast strain results in the synthesis of 6 (m/z þ ¼ 428; Fig. 2c and Supplementary  Fig. 1g). The identity of 6 synthesized from the engineered yeast strain was confirmed through NMR analyses (Supplementary Fig. 6 and Supplementary Table 5). The expression of PsAT1 from the GPD promoter in the optimal 5-producing strain (Fig. 3b, HXT7-2 and Supplementary Data 1) leads to the production of 6 at 40.0±0.5 mM. The expression of CYP82X1 in the 6-producing yeast strain leads to the synthesis of 7 (m/z þ ¼ 444) ( Fig. 2d and Supplementary Fig. 1h). Subsequent NMR analyses explicitly indicate the structure of 7 to be 4 0 -O-desmethyl-3-O-acetylpapaveroxine (Supplementary Fig. 7 and Supplementary Table 6). To ensure that our strains make sufficient 7 for conversion through the remaining pathway steps, we optimized the CYP82X1 for synthesis of 7 by examining its activity under different promoters. The highest activity (5.2 ± 0.4 mM in the strain harbouring only one copy of CYP82X2 regulated by HXT7 promoter) was observed when the enzyme was expressed from relatively weak promoters (ADH1 and CYC1) and late promoter (HXT7) at 25°C ( Fig. 3c and Supplementary  Fig. 3d). When we introduced the expression of CYP82X1 regulated by the HXT7 promoter in the optimal 5-producing strain ( Fig. 8a,c). This peak was also observed in the 5-producing strain expressing CYP82X1 ( Supplementary  Fig. 8a,c). This data implies that 7 may cyclize spontaneously to form 8, and that CYP82X1 functions on 5 to afford the synthesis of 6 0 or cyclized 8. The absence of 1,8-dihydroxy-13-O-acetyl-N-methylcanadine implies that the conversion from 1,8-dihydroxy-13-O-acetyl-Nmethylcanadine to 7 is either highly efficient in yeast, or more likely, that this reaction is facilitated by CYP82X1 ( Supplementary Fig. 8f). In addition, the acetylation is highly specific towards 5, and 6 is the only detectable acetylated product in the 7-producing strain (Fig. 2c). With structures of the products of CYP82X1, 5 0 and 7, explicitly verified by NMR in our study, our results support the previous in vitro biochemical characterization of the synthesis of secoberberine that CYP82X2 catalyses the enzymatic step subsequent to CYP82Y1, followed by PsAT1, which is required for CYP82X1 to function efficiently towards the synthesis of noscapine 47 . We hypothesize that the acetylation occurring before the C-N bond cleavage may enable catalysis via a tetrahedral acyl intermediate rather than simple proton abstraction, resulting in more efficient ring formation subsequent to the synthesis of the unstable aldehyde ( Supplementary Fig. 8f).
Reconstituting PsCXE1 and PsSDR1 activities. According to previous in planta gene silencing studies in P. somniferum 37 , the accumulation of the 4 0 -O-desmethyl version of the final product noscapine, narcotoline, 9, is observed when inactivating PsMT2. In addition, inactivating the carboxylesterase and noscapine synthase encoded in the noscapine gene cluster (PsCXE1 and PsSDR1, respectively) resulted in accumulation of 4 0 -Omethylated 3-O-acetylpapaveroxine, 7 0 , and narcotinehemiacetal, 8 0 , respectively. However, an in vitro study of PsMT2 indicated that PsMT2 has no activity towards narcotoline, but functions as a S9OMT with strict substrate specificity 50 . The recent in vitro biochemical study supports the activity of PsCXE1 in the conversion of 7 0 to 8 0 and PsSDR1 in the conversion of the m/z þ ¼ 402 peak synthesized spontaneously from 7 to narcotoline, 9 (ref. 47). Thus, the role of PsMT2 in the pathway (that is, whether it catalyses the narcotoline-4 0 -O-methylation or (S)-scoulerine-9-O-methylation), the identity of the 4 0 -O-methyltransferase required for the 4 0 -O-methylation and the sequence of the enzymatic steps catalysed by the 4 0 -O-methyltransferase, PsCXE1 and PsSDR1 remain in question.
We first engineered the 7-producing strain to express PsMT2. However, when grown in the presence of 2 at 25°C, the engineered strain did not synthesize 7 0 or any methylated intermediates (Fig. 2g). Then, we engineered the 7-producing strain to express PsCXE1 downstream of a strong promoter (PYK1 and GPD) and observed synthesis of 8 (m/z þ ¼ 402, 7.2 ± 0.5 mM in the optimal 7-producing strain; Supplementary  Fig. 3e) with MS/MS fragmentation that matches the reported spectra of narcotolinehemiacetal ( Fig. 2e and Supplementary  Fig. 1f). However, due to the low efficiency in the synthesis of 8, and the tailing of 8 on the high-performance liquid chromatography (HPLC) column, we were unable to purify 8 and conduct NMR analyses to further confirm the structure. The titre of 8 in the medium was therefore estimated by comparison to the standard curve of structurally similar 9.
We further engineered the optimal 7-producing strain to express PsCXE1 and PsSDR1, and observed the efficient synthesis of 9 from 8 (m/z þ ¼ 400, 17.8±0.1 mM; Fig. 2f and Supplementary Fig. 1i). The identity of 9 synthesized in yeast was confirmed by NMR analyses (Supplementary Fig. 9 and Supplementary Table 7). The higher titre of 9 compared with 8 indicates that PsSDR1 is much more efficient than PsCXE1, and with the addition of PsSDR1, the metabolic flux is driven towards the synthesis of 9. When PsCXE1 is removed from the 9-producing yeast strain, the synthesis of 9 is dramatically decreased but not abolished ( Supplementary Fig. 8b,d). This result, together with the structural verification of 9 synthesized from 8 suggests the structure of 8 as the cyclized narcotolinehemiacetal, and implies that 8 is present in the absence of PsCXE1. Our results indicate that 7 is prone to cyclize and form 8 spontaneously, and the presence of PsCXE1 enhances the rate of this cyclization. As the 5-producing strain expressing CYP82X1 also synthesized trace amounts of the m/z þ ¼ 402 peak at RT ¼ 2.8 min, we further introduced PsSDR1 to test if the cyclized 8 is also present in this strain. The synthesis of 9 was detected in the 5-producing strain expressing both CYP82X1 and PsSDR1 ( Supplementary Fig. 8b,d), which indicates the presence of the cyclized 8 in the m/z þ ¼ 402 peak. Although the data indicates that the m/z þ ¼ 402 peak at RT ¼ 2.8 min contains the cyclized 8, it cannot be excluded that this peak is a mixture of 6 0 and 8. The efficient synthesis of 9 from 7 catalysed by PsCXE1 and PsSDR1 indicates that 7 and 8 are also likely to be the native substrates of PsCXE1 and PsSDR1, respectively; and the 4 0 -O-methylation could be the final step.
PsMT2 and PsMT3 are both required for noscapine production.
To reconstruct the full noscapine biosynthetic pathway, we engineered the 9-producing strain to express PsMT2. However, when grown in the presence of 2 at 25°C, the engineered strain did not synthesize noscapine (Fig. 2i). Our results agree with previous in vitro characterization of PsMT2 (ref. 50), in which no activity was observed when incubating with 9. We hypothesized that the activity of PsMT2 was not correctly identified in the original plant studies 37,50 , and that a different O-methyltransferase enzyme may catalyse the final step in the reaction. Another O-methyltransferase (PsMT3) is encoded in the proposed gene cluster, and shares high sequence similarity with a norcoclaurine 6-O-methyltransferase from P. somniferum (Ps6OMT). Previous in vitro characterization studies demonstrated that PsMT3 showed no activity on narcotoline and proposed that this enzyme acts as a S9OMT, similar to PsMT1 (ref. 50). In this study, we engineered the narcotolineproducing strain to express PsMT3 and did not observe conversion of 9 to noscapine in either of these strains (Fig. 2i).
Plant O-methyltransferases function as dimers in which the substrate binding pocket is formed at the interface of the two monomers 59 . While all known O-methyltransferases are characterized to act as homodimers in their native plant pathways, heterodimers formed between different O-methyltransferases have been shown to act with low efficiencies on unique substrates in a heterologous context 60 . The observation that PsMT3 or PsMT2 do not act individually to catalyse the methylation of 9, led us to hypothesize that the synthesis of noscapine from 9 may require both O-methyltransferases. A 9-producing strain co-expressing PsMT2 and PsMT3 efficiently converted 9 to noscapine (m/z þ ¼ 414, 14.8±0.9 mM in the optimal 9-producing strain; Fig. 2i and Supplementary Fig. 1j), suggesting that these methyltransferase enzymes may act as a heterodimer. The identity of noscapine was confirmed through NMR analyses (Supplementary Fig. 9 and Supplementary Table 7). Due to the high sequence similarity between Ps6OMT and PsMT3, we replaced PsMT3 with Ps6OMT in the noscapine-producing strain, and observed similar noscapine titre from the engineered yeast strains (Fig. 2i and Supplementary Fig. 10a) The PsMT2/PsMT3 enzyme complex was then purified from Escherichia coli co-expressing the hexa-histidine tagged PsMT2 and T7-tagged PsMT3. The conversion of 9 to noscapine was detected when the PsMT2/PsMT3 complex was incubated with 9 and S-adenosyl methionine, while no noscapine was synthesized with purified PsMT2 or PsMT3 alone (Fig. 4a). SDS/(polyacrylamide gel electrophoresis) PAGE, western blotting and size-exclusion chromatography (SEC) analyses further support that the functional 4 0 -O-methyltransferase is a dimer harbouring both PsMT2 and PsMT3 (Fig. 4b,c). The efficient conversion of noscapine from 9 suggests that the 4 0 -O-methylation is likely to be naturally catalysed by the PsMT2-PsMT3 heterodimer ( Supplementary Fig. 10b). We also detected the synthesis of 7 0 (m/z þ ¼ 458) and 8 0 (m/z þ ¼ 416, B5.8 ± 0.7 mM in the optimal 8-producing strain) in these strains (Fig. 2g,h and Supplementary Fig. 10b). Similar to 8, we were unable to purify 8 0 from the medium due to the low synthesis efficiency and tailing on the HPLC column, and the titre reported for 8 0 was estimated by comparison to the standard curve of 1.
Since the observed activity on narcotoline for 6OMT was unexpected, we examined whether the same enzyme from a plant host that does not synthesize noscapine, Thalictrum flavum, would exhibit this activity (Fig. 2i). 9-producing strains co-expressing PsMT2 and Tf6OMT did not synthesize noscapine, indicating that the observed 6OMT activity was unique to the P. somniferum variant. We also purified the PsMT2/Ps6OMT heterodimer from E. coli ( Supplementary Fig. 11a,b). The in vitro conversion of narcotoline to noscapine is lower than that observed from the PsMT2/PsMT3 heterodimer ( Supplementary  Fig. 11c), which is different from the conversion efficiencies observed in yeast ( Supplementary Fig. 10a). Further kinetic analysis and in planta characterization of these two heterodimers is required to fully understand their native functions in P. somniferum.
To further elucidate the ordering of the final steps in the noscapine biosynthetic pathway in yeast, we engineered both 7and 8-producing strains to co-express PsMT2 and either PsMT3 or Ps6OMT (Fig. 2g,h). The narcotoline-4 0 -O-methyltransferase exhibits a much higher efficiency on 8 and 9 compared with 7 ( Supplementary Fig. 10b). The data supports that O-methylation and dehydrogenation are more likely to be the final two steps in the biosynthesis of noscapine, and thus the full biosynthetic pathway of noscapine from 2 is proposed as shown in Fig. 1. We believe that PsMT2/PsMT3 and PsSDR1 catalyse the final two steps in yeast; however, in planta, the reaction matrix may be shifted due to distinct localization patterns in the plant 61-63 , The extracellular titre of noscapine reaches 2.2±0.3 mM when the yeast strain is engineered to express each enzyme on an optimized expression construct and grown in the presence of 250 mM racemic 2 ( Fig. 3d and Supplementary Data 1). The extracellular titre of noscapine was increased over sevenfold to 14.8 ± 0.9 mM (B6% overall conversion from fed 2) by increasing the expression of CYP82X2, as suggested in the aforementioned optimization for the synthesis of 5 ( Fig. 3b and Supplementary Data 1).
We then fed 125 mM (S)-(98% purity) or (R)-(96% purity) canadine to our engineered yeast strains that synthesize the major pathway intermediates to roughly estimate the promiscuity of the pathway enzymes on each enantiomer. Interestingly, although (S)-canadine is the natural substrate of the pathway in opium poppy, (R)-canadine can also be converted through the pathway to noscapine when fed independently, but at much lower efficiencies than (S)-canadine in the early steps of the pathway ( Supplementary Fig. 12, for example, 112.66±16.07 mM versus 10.63 ± 0.34 mM of 4 is synthesized when fed with (S)-canadine and (R)-canadine, respectively). However, we also observed that the ratio between (S)-enantiomer and (R)-enantiomer decreases along the pathway ( Supplementary Fig. 12), which may be attributed to the 4% impurity in (R)-canadine.
Yeast biosynthesis of other protoberberine structures. In addition to supporting a synthesis platform for noscapine and all intermediates along the noscapine biosynthetic pathway, our yeast-based platform also supports the synthesis of some other protoberberine compounds by leveraging promiscuous enzyme activities and combinations of enzymes not found in nature. For example, yeast strains engineered to express CYP82Y1 convert 2 to 1-hydroxycanadine, 10 ( Supplementary Fig. 13), which is a novel protoberberine not found in nature. Yeast strains engineered to express TNMT and CYP82X2 synthesize N-methylophiocarpine, 11 ( Supplementary Fig. 13). The titres of 10 and 11 are estimated by comparing to the standard curves of structurally similar 2 and 4, respectively. Yeast strains engineered to express TNMT, CYP82Y1 and CYP82X1 synthesize 5 0 (Fig. 2b and Supplementary Fig. 1e). These off-pathway enzyme activities were optimal under different promoters from those for the native pathway activities (Supplementary Fig. 3b,f,g).
Yeast biosynthesis of noscapine from norlaudanosoline. Finally, we examined extension of the noscapine biosynthetic pathway in yeast to support biosynthesis from a less expensive fed precursor. The biosynthesis of 2 from an early benzylisoquinoline alkaloid (BIA) intermediate, 12, has recently been reported in S. cerevisiae 53 . Our successful reconstitution and optimization of the biosynthetic pathway from 2 to noscapine enables the connection of these two pathway parts to achieve the microbial biosynthesis of noscapine from 12. Since Ps6OMT can function together with PsMT2 as the narcotoline-4 0 -O-methyltransferase, yeast strains were engineered to express AtATR1, Ps6OMT, 3 0 -hydroxy-Nmethylcoclaurine-4 0 -O-methyltransferase from P. somniferum (Ps4 0 OMT), coclaurine N-methyltransferase from P. somniferum (PsCNMT), berberine bridge enzyme from P. somniferum (PsBBE), S9OMT from P. somniferum (PsS9OMT or PsMT1), CAS from Coptis japonica (CjCAS), PsTNMT, CYP82Y1, CYP82X2, PsAT1, CYP82X1, PsCXE1, PsSDR1 and PsMT2 without or with PsMT3 (CSYN15 without PsMT3, CSYN16 with PsMT3, Supplementary Data 1). When grown in the presence of 2 mM racemic 12, which is assumed to contain equal amounts of each enantiomer, CSYN15 synthesized 0.048 ± 0.011 mM of noscapine, while CSYN16 synthesized a significantly higher level of noscapine (0.49 ± 0.16 mM, paired t-test, P ¼ 0.000083; Fig. 5a,b). CYP82X2 and PsS9OMT are enzymes associated with two bottleneck points in the pathway. By increasing the expression levels of both CYP82X2 and PsS9OMT, the synthesis of noscapine was increased over threefold (1.64 ± 0.38 mM) to that observed from CSYN16 (paired t-test, P ¼ 0.00047; Fig. 5b,c and Supplementary Data 1). To estimate the transport efficiency of the pathway metabolites, intracellular concentrations of major pathway intermediates were estimated and compared with the concentrations measured in the growth media ( Supplementary Fig. 14). Our study indicates that with the exception of 5, all major pathway intermediates are efficiently transported out of yeast cells to the medium, with the export of 6 and 7 relatively lower.

Discussion
The 10-gene cluster encoding enzymes responsible for the conversion of scoulerine to noscapine was first discovered and verified in P. somniferum in 2012, and activities of many of the enzymes were proposed based on virus-induced gene silencing assays in the native plant host 37 . A series of in vitro characterization studies on the biosynthetic enzymes involved in the synthesis of noscapine were recently published [47][48][49][50] , which provide additional insights into the biosynthesis of noscapine in opium poppy. In this work, we reconstituted the noscapine gene cluster in yeast to achieve the microbial production of secoberberine (5 0 and 7) and phthalideisoquinoline (noscapine and 9) alkaloids from fed 2 or 12. The activities of enzymes encoded in the gene cluster reconstituted in yeast agree well with the in planta and in vitro biochemical characterization studies, and provides additional insights into the biosynthesis of noscapine. Our work also enables the supply of many pathway intermediates and derivatives for structural confirmation and further investigation. We primarily fed racemic canadine, 2, to our strains, which is assumed to be composed of equal amounts of each enantiomer. The substrate specificity of the noscapine pathway enzymes towards different enantiomers has not previously been investigated 47,48,56 ; instead, racemic canadine has typically been used to characterize the enzymes along the pathway in vitro. In addition, individual enantiomer standards of pathway intermediates between canadine and noscapine are unavailable. Thus, it is difficult to estimate the ratio between each enantiomer along the noscapine pathway. We fed (S)-and (R)-canadine to our engineered yeast strains, and the data indicated that although our racemic canadine-fed strains mainly produce (S)-enantiomers of the pathway metabolites, (R)-enantiomers may also be present in small quantities. However, considering that the (R)-canadine is of 96% purity, it is difficult to estimate the exact percentage of pure (R)-enantiomers in strains fed with (R)-canadine. In addition, when fed with racemic 2, the conversion of (S)-and (R)-canadine to the pathway intermediates may shift from when they are fed separately to the engineered yeast strains, and whether or not (R)-canadine can be converted when both enantiomers are present is unknown.
Recent advances in the de novo production of reticuline in yeast 64,65 highlight the possibility of constructing yeast strains that can synthesize noscapine and related metabolites from simple carbon and nitrogen sources versus feeding of expensive intermediate alkaloids. Whereas the microbial platforms described here are suitable for smaller scale production of noscapine and related compounds, for example, for drug screening and enzyme activity identification, the de novo production of noscapine and related molecules could provide a cost-efficient process for the production of these valuable molecules. However, no less than eleven enzymes, including four cytochrome P450s, will need to be introduced into the reticuline-producing strain for the de novo synthesis of noscapine. The full de novo pathway to noscapine is thus composed of up to 18 heterologous enzymatic steps from tyrosine and at least 23 different heterologous enzymes, including 5 cytochrome P450s. Our data indicates that as the heterologous pathway introduced into yeast includes more P450s, the conversion efficiency of each P450 will decrease, which may be due to challenges associated with endoplasmic reticulum loading and cofactor regeneration. Therefore, introducing a pathway comprising the expression of at least 5 cytochrome P450s will require further engineering of the yeast background to increase efficiencies. In addition, the reported strains produce relatively low levels of reticuline from tyrosine with titres reported at o100 mg l À 1 (refs 64,65). Thus, further optimization will be required to develop improved reticuline-producing platform strains to ultimately drive the metabolite flux from reticuline to noscapine. To develop a commercially competitive process for noscapine production via microbial fermentation, titres would need to reach at least the range of 10-20 mM (ref. 6). In addition to improving the pathway within a single yeast strain to increase production levels of noscapine 8,64,65 , other engineering strategies such as co-culturing a noscapine-producing strain with a canadine-or norcoclaurineproducing strain may also be a promising strategy. Multiplestrain biosynthetic strategies have recently been described to achieve enhanced microbial production of taxanes and opioids 9,11 .
Throughout our study, we measured the pathway metabolite levels in the growth medium from the engineered yeast strains. Our data indicates that most metabolites were transported out of the yeast cells efficiently (Supplementary Fig. 14). However, the export of 5, 6 and 7 are not as efficient, and thus the titre in the growth medium can be an underestimate of the production of the corresponding molecules. The efficient export of the other pathway intermediates may lead to lower intracellular concentrations, and can be deleterious to the metabolic flux through the pathway especially for slower enzymatic steps. Future engineering efforts directed to optimizing the export of pathway metabolites and noscapine may be critical to further enhancing the production of target molecules (for example, noscapine or target pathway intermediates).
We observed that PsSDR1 or PsMT2/PsMT3 could catalyse the final step in the synthesis of noscapine in yeast (Fig. 1). Previous in planta and in vitro studies implied that the synthesis of 7 to noscapine is likely to be branched into several routes 47 . Our work validates the reaction matrix of the final three steps (that is, PsCXE1, PsSDR1 and PsMT2/PsMT3), and showed the efficiency of each route in yeast. While the noscapine pathway in plants may follow the sequence of the final three steps proposed based on our yeast pathway reconstitution study, it is also possible that the pathway exhibits slightly different patterns in the native plant host. For example, the pathway enzymes have specific cellular localization and spatial organization in the opium poppy that may result in the biosynthetic matrix being routed in different patterns depending on the pathway localization in the plant. For example, alkaloid synthesis in opium poppy involves several cell types and transport steps 61 . Therefore, the pathway organization and sequence is much influenced by the subcellular localization of pathway enzymes. The previous in planta study found that the enzymes catalysing the final three steps of morphine biosynthesis are among the most abundant active proteins in the latex sub-proteome, which indicates that the final three steps are likely to principally occur in the laticifers 61 . Among the noscapine pathway enzymes, only PsSDR1 is enriched in laticifers 49 . Thus, biosynthesis of noscapine may follow similar organization to morphine, with PsSDR1 catalysing the final step in laticifers. Additional investigations in opium poppy are necessary to further understand the organization and sequence of noscapine biosynthesis. More broadly, the microenvironments in yeast, in vitro and in planta in which the pathway enzymes function can be different, and these differences can result in distinct substrate preferences, enzyme activities and product transportation processes. Thus, observations of pathway sequence and efficiency across these different contexts should ultimately be considered with regard to these differences.
Our study also uncovered a unique O-methylation step that is likely to be natively catalysed by an O-methyltransferase heterodimer in opium poppy. While previous studies demonstrated that plant O-methyltransferase heterodimers can exhibit different substrate specificity from the corresponding homodimers in synthetic contexts 59,60 , our results suggest a role for O-methyltransferase heterodimers in native plant pathways.
These findings suggest that some uncharacterized O-methylation steps in the biosynthesis of plant natural products may be catalysed by heterodimers rather than homodimers. In addition, our results suggest that O-methyltransferases can catalyse more than one methylation reaction in their original plant hosts by acting on one substrate as a homodimer and acting on different substrates when forming heterodimers with different O-methyltransferase partners. Further in planta investigations are essential for validating that PsMT2 and PsMT3 form a heterodimer in opium poppy, and studying the native functions of PsMT2, PsMT3, Ps6OMT and their dimers.
Step-wise pathway reconstitution in yeast can be applied to bridge in planta and in vitro characterizations of plant natural product pathways to provide a complementary approach for better understanding and engineering the biosynthetic potential of plants. The yeast-based plant pathway reconstitution also has the advantage of simultaneously building production strains for potentially valuable compounds, including all intermediates along a pathway and novel derivatives thereof. While the end-product noscapine accumulates in plants at low levels, most of the pathway intermediates, which can serve as interesting lead compounds for drug discovery, are in exceedingly low abundance in the producing plants, and thus not accessible. In addition, enzymatic steps can be readily removed or introduced in the yeast platform to enable the synthesis of novel structures such as 10.
Our work demonstrates the potential of synthetic biology as a powerful tool for aiding complex pathway elucidation, while enabling microbial production of valuable natural or unnatural medicinal compounds originated from plants.

Methods
Compounds and strain culture condition. Racemic canadine (98% purity), 2, norlaudanosoline (98% purity), 12, and reticuline were purchased from Finetech Industry Limited, Santa Cruz Biotechnology and Specs, respectively. (R)-and (S)-enantiomers of 2 were purchased from Toronto Research Chemicals. S-Adenosyl methionine (80% purity) and noscapine (97% purity), 1, were purchased from Sigma-Aldrich. Narcotoline (495% purity), 9, was purified from 9-producing yeast strain in this study as described below. For DNA manipulation and amplification, we used chemically competent or electrocompetent E. coli strains TOP10 and ccdB Survival 2 T1 R purchased from Life Technologies. E. coli strains were grown at 37°C in lysogeny broth medium obtained from Fisher Scientific supplemented with 100 mg ml À 1 ampicillin (EMD Chemicals) or 50 mg ml À 1 kanamycin (Fisher Scientific) for plasmid maintenance. All engineered yeast strains described in this work, as listed in Supplementary Data 1, were constructed in a haploid W303a background (MATa leu2-3,112 trp1-1 can1-100 ura3-1 ade2-1 his3-11,15) 66 . Yeast strains were cultured at 25 or 30°C in complex yeast extract peptone dextrose (YPD, all components from BD Diagnostics) medium or SDM containing yeast nitrogen base (YNB) without Amino Acids (BD Diagnostics), ammonium sulfate (Fisher Scientific), 2% dextrose and the appropriate dropout solution for plasmid maintenance. 200 mg l À 1 G418 sulfate (Calbiochem) or 200 mg l À 1 Hygromycin B (Life Technologies) were used in YPD medium for selection. BL21(DE3) and Ni-NTA agorose resin used for E. coli protein expression and hexa-histidine tagged protein purification were kindly provided by Professor Chaitan Khosla, and T7 Tag Affinity Purification Kit used for T7-tagged protein purification was purchased from EMD Millipore. NuPAGE Novex 4-12% Bis-Tris Protein Gels (1.0 mm, 10 well) used for purified protein SDS-PAGE analysis and western blot were purchased from Life Technologies and Color Prestained Protein Standard (Broad Range, 11-245 kDa) used as the protein marker was obtained from New England Biolabs. The Anti-6X His tag antibody (horseradish peroxidase (HRP), ab1269) and Anti-T7 tag antibody (HRP, ab21477) used for western blot of PsMT2/PsMT3 and PsMT2/Ps6OMT heterodimers were purchased from Abcam PLC. Size-exclusion chromatography was performed on a BioSep-SEC-s4000 column (Phenomenex, kindly provided by Professor Chaitan Khosla), and the column was equilibrated with Gel Filtration Markers Kit for Protein Molecular Weights 12,000-200,000 Da purchased from Sigma.
General techniques for DNA manipulation. Plasmid DNA was prepared using the QIAprep Spin Miniprep Kit (QIAGEN) and Econospin columns (Epoch Life Science) according to manufacturer's protocols. PCR reactions were performed with Taq polymerase, PfuUltra Hotstart DNA Polymerase (Agilent Technologies) and Expand High Fidelity PCR System (Roche Life Science) according to manuacturer's protocols. PCR products were purified by Zymoclean Gel DNA Recovery Kit (Zymo Research). All DNA constructs were confirmed through DNA sequencing by Elim Biopharmaceuticals Inc. Restriction enzymes (NEB) and T4 ligase (NEB) were used to digest and ligate the DNA fragments, respectively. BP Clonase II Enzyme Mix, Gateway pDONR221 Vector and LR Clonase II Enzyme Mix (Life Technologies) and the S. cerevisiae Advanced Gateway Destination Vector Kit (Addgene) 67 were used to perform Gateway Cloning. Gibson one-pot, isothermal DNA assembly 68 was conducted at 10 ml scale by incubating T5 exonuclease (NEB), Phusion polymerase (NEB), Taq ligase (NEB) and 50 ng of each DNA fragment at 50°C for 1 h to assemble multiple DNA fragments into one circular plasmid. Yeast strains are constructed through homologous recombination and DNA assembly 23 . Yeast strains and plasmids utilized in this study are listed in Supplementary Data 1 and 2. DNA sequences of genes involved in this work are listed in Supplementary Data 3.
Culture conditions for metabolite production. For all the assays reported in this work, yeast were grown in 500 ml media in 96-well plates (BD falcon) covered with AeraSeal film (Excel Scientific), at 480 r.p.m., 80% humidity in a Kuhner Lab-Therm LX-T plate shaker. For step-wise reconstitution of the noscapine biosynthetic pathway from canadine, yeast strains were first cultured overnight in 500 ml SDM with 2% dextrose. 5-10 ml of the overnight seed culture was inoculated in fresh 500 ml SDM with 250 mM canadine and 2% galactose, and incubated at 25 or 30°C for 72 h before metabolite analysis of the growth medium. For the synthesis of noscapine from canadine and norlaudanosoline, 5-10 ml of the overnight seed culture was inoculated in 500 ml 5 Â concentrated SDM (10% dextrose, 5 Â YNB) and incubated at 25°C for 48 h (ref. 54). The cell were then harvested by centrifugation at 15,627g for 5 min, pellets resuspended in fresh SDM with 2 mM norlaudanosoline and 2% galactose, and incubated at 25°C for 96 h before metabolite analysis 54 .
Analysis of benzylisoquinoline alkaloid production. For all assays, yeast culture medium was analysed by reverse phase LC-MS/MS on an Agilent 6300 Series Ion Trap LC-MS (Agilent ZORBAX SB-Aq 4.6 Â 50 mm, 5 mm) and an Agilent 6420 triple quad LC-MS (Agilent EclipsePlus C18, 2.1 Â 50 mm, 1.8 mm) using positive ionization. On the Agilent 6300 Series Ion Trap LC-MS, metabolites in the medium were separated on a linear gradient of 20% CH 3 OH (v/v in water, 0.1% acetic acid) to 60% CH 3 OH (v/v in water, 0.1% acetic acid) over 7 min with a flow rate of 0.5 ml min À 1 . On the Agilent 6420 triple quad LC-MS, metabolites in the medium were separated on a linear gradient of 10% CH 3 CN (v/v in water, 0.1% formic acid) to 50% CH 3 CN (v/v in water, 0.1% formic acid) over 5 min with a flow rate of 0.4 ml min À 1 for scan mode; or 10% CH 3 CN (v/v in water, 0.1% formic acid) to 40% CH 3 CN (v/v in water, 0.1% formic acid) over 5 min with a flow rate of 0.4 ml min À 1 for multiple reaction monitoring (MRM) mode. The identity of each compound was verified by comparing MS/MS spectra of samples and literature or standards. The conversion of each enzymatic step was estimated by comparing the integrated peak area of extract ion chromatograms (EICs) to a standard curve. The quantities of 3, 4, 5, 5 0 , 6, 7 and 9 were estimated by comparing to the stand curves of the corresponding purified and NMR verified standards. The quantities of 8, 8 0 , 10 and 11 were estimated by comparing to the standard curves of structurally similar 9, 1, 2 and 4, respectively, due to the lack of standards. Noscapine levels were quantified by MRM using the highest characteristic precursor ion/product ion transition (Supplementary Table 8), and comparing the integrated peak area to a standard curve of noscapine.
Estimation of metabolite intracellular concentrations. To estimate the intracellular concentrations of the pathway metabolites, 500 ml overnight seed culture of CSYN16 was first inoculated in 50 ml 5 Â concentrated SDM (10% dextrose and 5 Â YNB) and incubated at 25°C for 48 h (ref. 54). The cells were harvested by centrifugation at 11,305g for 5 min, pellets resuspended in fresh SDM with 2 mM norlaudanosoline and 2% galactose, and incubated at 25°C (ref. 54). Three 1 ml cell cultures were collected each day after the induction until day four. The yeast culture medium of each 1 ml sample was separated from cell pellets by centrifugation at 15,871g for 2 min. The cell pellets were extracted with 500 ml acetone. The organic phase was separated, evaporated to dryness and dissolved in 10% CH 3 CN (v/v in water, 0.1% formic acid) in a volume equal to that of the cell pellet (normally range 30-40 ml). The medium and cell pellet extracts were analysed on an Agilent 6420 triple quad LC-MS (Agilent EclipsePlus C18, 2.1 Â 50 mm, 1.8 mm) using positive ionization. The substrate and pathway metabolites reticuline, scoulerine, tetrahydrocolumbamine, 1-9 and 12 were monitored by MRM using the highest characteristic precursor ion/product ion transition (Supplementary Table 8). The quantities of 1-7, 9 and 12 were estimated by comparing to the standard curves of corresponding purified or purchased standards. Due to the lack of the corresponding authentic standards, the quantities of scoulerine and tetrahydrocolumbamine were estimated using the structurally similar 2 as a standard, based on the fact that similar structures usually exhibit similar ionization efficiency. 8 was not quantified as the concentration of 8 is below the detection limit.
Compound purification and structural elucidation. To obtain sufficient amount of 1, 3, 4, 5, 5 0 , 6, 7 and 9 for structural elucidation by NMR, and for in vitro characterization of the PsMT2/PsMT3 heterodimer, the 500 ml in vivo assay was scaled up to 200-1,000 ml with each 100 ml of medium containing 2% galactose and 12 mg canadine, and cultured in 500 ml shake flask at 25°C for 120 h. Yeast culture medium was separated from cell pellets by centrifugation at 4,416g at 5 min. The supernatant was incubated with 20 g of Amberlite XAD4 (Sigma-Aldrich) overnight. XAD4 resins were then collected through vacuum filtration, and metabolites were recovered from the resin by extracting with 50 ml MeOH for three times. The resultant organic extracts were combined and evaporated to dryness, redissolved in 20% CH 3 CN (v/v in water, 0.1% formic acid). 1, 3, 4, 5 0 , 6 and 7 were purified by reverse-phase HPLC (Varian Pursuit XRs-C18, 5 mm, 250 Â 10 mm) on an isocratic elution with 20% CH 3 CN (v/v) over 15 min and 95% CH 3 CN (v/v) for 5 min in water (0.1% formic acid) at a flow rate of 4 ml min À 1 ; and 5 and 9 were purified by reverse-phase HPLC (Varian Pursuit XRs-C18, 5 mm, 250 Â 10 mm) on an isocratic elution with 15% CH 3 CN (v/v) over 15 min and 95% CH 3 CN (v/v) for 5 min in water (0.1% formic acid) at a flow rate of 4 ml min À 1 . The 1 H, 1 H-1 H correlation (COSY), heteronuclear multiple quantum coherence (HMQC) and heteronuclear multiple bond correlation (HMBC) NMR spectra of 6 and 7, the 1 H-1 H correlation (COSY) NMR spectra of 5 0 were performed on the Varian Inova 600 spectrometer using CD 3  performed on the Varian ui500 spectrometer using CD 3 CN as the solvent. The 1 H, HMQC and HMBC NMR spectra of 5 0 were performed on the 800 MHz Agilent (Varian) VNMRS spectrometer using CD 3 CN as the solvent. The 1 H, HMQC and HMBC NMR spectra of 5 were performed on the Varian Inova 600 spectrometer using CDCl 3 as the solvent. The 1 H-NMR spectra of 3 and 4 were performed on the Varian Inova 600 spectrometer using CD 3 CN as the solvent, and the 1 H-NMR spectra of 1 and 9 were performed on the Varian Inova 600 spectrometer using DMSO-d6 as the solvent. High-resolution mass spectrometry data were obtained on a coupled Agilent 6520 Accurate-Mass Q-TOF ESI mass spectrometer in positive ion mode, and analysed using Mass Hunter Qualitative Analysis software (Agilent).
NMR data of 3 were as follows: 1  O-Methyltransferase expression and purification. pCS3306 and pCS3307 were transformed individually and in combination into E. coli strain BL21(DE3) for PsMT2 or PsMT3 homodimer expression and PsMT2 and PsMT3 co-expression. The cells were cultured at 37°C and 250 r.p.m. in 500 ml of lysogeny broth medium supplemented with 50 mg ml À 1 kanamycin and/or 100 mg ml À 1 ampicillin to a final OD600 between 0.4 and 0.6. The culture was then incubated on ice for 10 min before addition of 0.4 mM isopropylthio-b-D-galactoside to induce protein expression. The cells were further cultured at 16°C for 12-16 h. The cells were harvested by centrifugation (2,164g, 15 min, 4°C), resuspended in 25 ml lysis buffer (20 mM Tris-HCl, pH 7.9, 0.5 M NaCl and 10 mM imidazole), and lysed by sonication on ice. Cellular debris was removed by centrifugation (27,167g, 20 min, 4°C), and appropriate affinity agarose resins were then added to the supernatant (2 ml l À 1 of culture, Ni-NTA agarose resin for PsMT2, and T7 tag antibody agarose resin for PsMT3). The suspension was swirled at 4°C for 2 h and loaded onto a gravity column.
For the purification of PsMT2, the PsMT2-bound resin was washed with 50 column volume of 10 mM imidazole in buffer A (50 mM Tris-HCl, pH 7.9, 2 mM EDTA and 2 mM dithiothreitol (DTT)), followed by 20 column volume of 20 mM imidazole in buffer A. PsMT2 was then eluted with 250 mM imidazole in buffer A. For the purification of PsMT3, the PsMT3-bound resin was washed with 10 column volume of binding buffer (4.29 mM Na 2 HPO 4 , 1.47 mM KH 2 PO 4 , 2.7 mM KCl, 137 mM NaCl, 0.1% Tween-20, 0002% sodium azide and pH 7.3). The T7-tagged protein was eluted with T7 Tag Elute Buffer (100 mM citric acid, pH 2.2), and each 1 ml of elute was neutralized with 150 ml of the T7 Tag Neutralization Buffer (2 M Tris base, pH 10.4). For the purification of PsMT2/ PsMT3 heterodimer, the supernatant of the cell lysate was first incubated with Ni-NTA agarose resin, and the Ni bound protein was isolated following the same procedure as that of PsMT2. The elution from the Ni-NTA agarose resin was concentrated and exchanged into 25 ml binding buffer, and incubated with T7 tag antibody agarose resin. The PsMT2/PsMT3 heterodimer was then purified following the same procedure for the purification of PsMT3. Purified enzymes were concentrated and exchanged into storage buffer (50 mM sodium phosphate, pH 7.5, 100 mM NaCl, 10% glycerol and 2 mM DTT) with the centriprep filters (Amicon), aliquoted and flash frozen using acetone/dry ice. Protein concentration was determined with the Bradford assay 74 using bovine serum albumin as a standard, and purified PsMT2 and PsMT3 heterodimer was analysed on SDS-PAGE gel.
The PsMT2/Ps6OMT heterodimer was purified from E. coli co-transformed with pCS3306 and pCS3534 following a similar procedure as described for the PsMT2/PsMT3 purification process.
Western blot analysis of O-methyltransferases. Prestained protein standard and 0.5-2 mg (0.5 mg for Anti-T7 western blot, and 2 mg for Anti-His western blot) of the purified PsMT2 homodimer, PsMT3 homodimer, PsMT2/PsMT3 heterodimer and PsMT3/Ps6OMT heterodimer were loaded on a NuPage Novex 4-12% Bis-Tris protein gel and run with NuPAGE MOPS SDS running buffer (Life Technologies) at 175 V for 45 min. The protein material was then semi-dry transferred to a nitrocellulose membrane or polyvinylidene difluoride membrane at 15 V for 15 min with transfer buffer (Life Technologies). The membrane was blocked with 5% bovine serum albumin in tris buffered saline and Tween-20 (TBST) buffer (50 mM Trizma base, 150 mM NaCl, 0.05% Tween-20 and pH 8.0) for 1 h, and then probed overnight at 4°C with Anti-6X His tag antibody (ab1269 from Abcam PLC, 1:5000) or Anti-T7 tag antibody (ab21477 from Abcam PLC, 1:5000) conjugated with HRP in TBST buffer for 12-16 h. Chemiluminescence was induced by SuperSignal West Pico substrate (Pierce) and images were acquired by a G:Box Chemi XT4 imaging system (Syngene).
In vitro reconstitution of O-methyltransferases. The in vitro methyltransferase reactions were performed as previously described 50 at 50 ml scale containing 100 mM Gly-NaOH buffer (pH 9.0) in the presence of 25 mM sodium ascorbate, 2 mM DTT, 10% glycerol, 300 mM S-Adenosyl methionine, 10 mM purified 9 and 50 mg purified enzyme or no enzyme as a control. The reactions were performed overnight and quenched with 150 ml of 99.9% MeOH/0.1% formic acid. The quenched reaction mixture was analysed on an Agilent 6420 triple quad LC-MS (Agilent EclipsePlus C18, 2.1 Â 50 mm, 1.8 mm) using positive ionization. Noscapine and narcotoline were detected with MRM using the highest characteristic precursor ion/product ion transition (Supplementary Table 8).
Size-exclusion chromatography analysis. The PsMT2/PsMT3 and PsMT2/Ps6OMT enzyme complex as a dimer were determined by SEC on a normal-phase HPLC (Agilent) using the BioSep-SEC-s4000 column. The column was equilibrated and eluted with 300 mM NaCl in 100 mM sodium phosphate buffer, pH 7.5, at a flow rate of 1 ml min À 1 , and calibrated with the following protein standards: sweet potato b-amylase (200 kDa); yeast alcohol dehydrogenase (150 kDa); bovine serum albumin (66 kDa); bovine erythrocytes carbonic anhydrase (29 kDa) and horse heart cytochrome c (12.4 kDa). Protein elution was detected by absorbance at 280 nm.
Data availability. The authors declare that the data supporting the findings of this study are available within the article and its Supplementary Information files.