Engineering a microbial biosynthesis platform for de novo production of tropane alkaloids

Tropane alkaloids (TAs) are a class of phytochemicals produced by plants of the nightshade family used for treating diverse neurological disorders. Here, we demonstrate de novo production of tropine, a key intermediate in the biosynthetic pathway of medicinal TAs such as scopolamine, from simple carbon and nitrogen sources in yeast (Saccharomyces cerevisiae). Our engineered strain incorporates 15 additional genes, including 11 derived from diverse plants and bacteria, and 7 disruptions to yeast regulatory or biosynthetic proteins to produce tropine at titers of 6 mg/L. We also demonstrate the utility of our engineered yeast platform for the discovery of TA derivatives by combining biosynthetic modules from distant plant lineages to achieve de novo production of cinnamoyltropine, a non-canonical TA. Our engineered strain constitutes a starting point for future optimization efforts towards realizing industrial fermentation of medicinal TAs and a platform for the synthesis of TA derivatives with enhanced bioactivities.


Supplementary Figure 1. Design of genomic integrations for pathway construction in yeast.
Block arrows represent gene expression cassettes with unique promoter and terminator for each locus. Genus and species sources for heterologous genes are indicated by two letters preceding the gene symbol. Superscript annotations on gene symbols indicate N-or Cterminal modifications. Refer to Supplementary Table 1 for gene sources and Supplementary Table 2 for strain genotypes.

Supplementary Figure 2. Functional validation of agmatine/putrescine biosynthetic pathway genes in yeast.
Wildtype yeast strain CEN.PK2 was transformed with three low-copy plasmids to co-express between zero (negative control) and three of the indicated biosynthetic genes. Plasmids expressing blue fluorescent protein (BFP) were used as negative controls for each of the three auxotrophic selection markers URA3, TRP1, and LEU2 (pCS4208, 4212, 4213). Transformed strains were cultured in selective media (YNB-DO) with 2% dextrose at 30 o C for 48 h prior to LC-MS/MS analysis of metabolite production. All data show titers as measured by LC-MS/MS peak area relative to the negative control (CEN.PK2). Data represent the mean of n = 3 biologically independent samples (open circles) and error bars show standard deviation. Source data are provided as a source data file. Wild-type strain CEN.PK2 or meu1 disruption strain CSY1229 were co-transformed with low-copy plasmids expressing SPE1, AsADC, and speB (pCS4239) and AbPMT1 (pCS4193). Data indicate mean NMP titer relative to CEN.PK2 control as quantified by LC-MS/MS peak area for n = 4 biologically independent samples (open circles) after 48 hours of growth at 30 o C in selective media (YNB-DO) with 2% dextrose. Error bars show standard deviation. Student's two-tailed t-test: *P < 0.05, ** P < 0.01, *** P < 0.001. Source data are provided as a source data file.  Figure 11. Effect of reconstituting ALD6 activity on metabolite flux through NMPy towards tropine in CSY1249. (a) Effect of reconstituting functional ALD4 or ALD6 genes on the growth of the NMPy-producing yeast strain (CSY1246) with and without acetate supplementation. ALD4 and ALD6 were expressed from low-copy plasmids. 'WT' indicates CSY1246 with control (BFP) plasmid. Adjacent columns show ten-fold dilutions. (b) Production of 4MAB acid and hygrine side products with reconstituted acetate metabolism in engineered yeast strain CSY1246. Plus and minus symbols indicate presence or absence of fed metabolite (acetate) or ALD4 and ALD6 genes expressed from low-copy plasmids. Side product abundances were measured by LC-MS/MS analysis of the extracellular medium after 48 hours of growth in selective media supplemented with or without 0.1% w/v potassium acetate at 25 o C. (c) Production of intermediates between NMPy and tropinone in engineered strains with and without functional ALD6.  Supplementary Figure 19. Representative standard curves for metabolite titer measurements in culture media. All standard curves were prepared in either non-selective (YNB(A)-SC) or selective (YNB(A)-DO) media, depending on the corresponding culture conditions. Dotted line indicates linear regression fit; regression equation and R 2 are indicated. Source data are provided as a source data file.

Supplementary Note 1. Optimization of MPO activity via sub-cellular localization and truncation studies.
Prior studies have shown that while NtPMT is expressed in the cytosol of tobacco cells, NtMPO1 localizes to the peroxisome lumen 5 . We performed in silico prediction of enzyme subcellular localization using the SherLoc2 utility for signal peptide detection 6 , which revealed that NtMPO1 harbors a strong yeast consensus peroxisome-targeting sequence (PTS) at its C-terminus (Ala-Lys-Leu, denoted PTS1). This observation suggested that NtMPO1 may localize to peroxisomes when expressed heterologously in yeast (Supplementary Figure 5a). We confirmed this prediction via fluorescence microscopy of wild-type yeast cells expressing either N-or C-terminal GFP-tagged AbPMT1 (pCS4209, 4210) and NtMPO1 (pCS4214, 4215) from low-copy plasmids, which indicated that while AbPMT1 is found primarily in the cytosol, localization of NtMPO1 to peroxisomes is contingent on an exposed C-terminal PTS (Supplementary Figure   5b, Fig. 3b). Given that yeast peroxisomes likely do not possess the same transport machinery for exchanging NMP, 4MAB, and NMPy as plant peroxisomes, we hypothesized that the poor apparent activity of NtMPO1 was caused by its sequestration within peroxisomes. However, cytosolic expression of NtMPO1 achieved by masking the C-terminal PTS with a GFP fusion did not significantly impact extracellular 4MAB or NMPy levels (Supplementary Figure 5c), indicating that its peroxisomal localization was likely not the primary factor limiting 4MAB and NMPy production.
We next examined whether orthologs of NtMPO1 from TA-producing species exhibit greater activity in yeast. We performed a tBLASTn search of the transcriptomes of A. belladonna and Datura metel in the 1000 Plants Project database 7 using the amino acid sequence of NtMPO1 as a query and an E-value threshold of 10 -150 . We identified two fulllength ortholog sequences, denoted AbMPO1 and DmMPO1, which each shared 91% sequence identity with NtMPO1 (Supplementary Figure 6a). We obtained yeast codon-optimized sequences for the two orthologs and cloned them into low-copy expression plasmids. To evaluate their activity, we co-expressed each of the three MPO variants (pCS4218, 4227, 4228) with AbPMT1 (pCS4193) from low-copy plasmids in our putrescine-overproducing strain (CSY1235), and measured 4MAB and NMPy accumulation in the medium by LC-MS/MS following 48 hours of growth. DmMPO1 showed comparable levels of 4MAB and NMPy production to the original NtMPO1 variant, whereas AbMPO1 exhibited significantly lower activity, despite differing by less than 10% in amino acid sequence (Supplementary Figure 6b). We next examined whether the differences in activity between the three orthologs could be partially attributed to structural differences in their active sites. We constructed template-based homology models of NtMPO1, AbMPO1, and DmMPO1 based on the crystal structure of a Pisum sativum copper-containing amino oxidase (PDB: 1KSI) using the RaptorX web server 8 . Although no clear conclusions could be drawn regarding relationships between differences in active site or substrate-binding pocket and observed in vivo activity, the homology models indicated that the orthologs possess long, unstructured N-and C-terminal tail regions (Supplementary Figure 6c). Although such unstructured regions are frequently caused by failure of the homology modeling software to adequately model regions of the polypeptide chain with poor homology to an existing template, we verified whether these unstructured regions were modeling artifacts or genuine structural features by testing truncations of the two active orthologs, NtMPO1 and DmMPO1. N-terminal truncations removed the first 84 and 81 residues of the two orthologs, respectively; whereas C-terminal truncations removed the last 21 residues. Since the importance of the PTS for MPO activity was unclear, we also constructed Cterminal truncations wherein the unstructured tail was removed but the PTS was retained (denoted ΔC-PTS1 ). We coexpressed each of the MPO truncations (pCS4233-4238) with AbPMT1 (pCS4193) from low-copy plasmids in the putrescine-overproducing strain CSY1235, and quantified 4MAB and NMPy accumulation in the media after 48 hours of growth by LC-MS/MS. No significant differences in either 4MAB or NMPy titers were observed between the NtMPO1 truncations ( Supplementary Figure 6d, Fig. 3c). However, removal of the C-terminal unstructured region from DmMPO1 while retaining the C-terminal PTS tripeptide resulted in a 19% increase in extracellular 4MAB levels relative to the wildtype DmMPO1 enzyme and a 55% increase relative to wild-type NtMPO1. Although only an 11% increase in NMPy titer was observed between these variants, we suspected that the additional 4MAB accumulation would improve flux towards downstream intermediates upon reconstitution of subsequent biosynthetic steps in the pathway. As such, all further optimization of the MPO step was performed using this truncated variant, denoted DmMPO1 ΔC-PTS1 .

Supplementary Note 2. Elucidation of co-substrate for spontaneous condensation with NMPy to form hygrine.
To elucidate the co-substrate that condenses with NMPy to produce hygrine, we performed experiments in which we assayed combinations of NMPy-producing and -non-producing yeast strains in media supplemented with acetate or acetoacetate for hygrine accumulation by LC-MS/MS (Supplementary Figure 10). No hygrine accumulated in cultures of wild-type yeast (CEN.PK2) fed either acetate or acetoacetate following 48 hours of growth. In cultures of NMPyproducing strains, hygrine accumulated with acetate or acetoacetate supplementation (CSY1235 expressing AbPMT1 and DmMPO1 ΔC-PTS1 : 38 μg/L hygrine with acetate, 1.9 mg/L with acetoacetate; CSY1243: 108 μg/L with acetate, 3.5 mg/L with acetoacetate), but not in the absence of either. Hygrine accumulation was detected (to 17 mg/L) when 100 mg/L NMPy was incubated in media with acetoacetate in the absence of cells, but not when NMPy was incubated with acetate alone. These data suggest that acetate supplementation may increase levels of an endogenous keto-metabolite, such as acetoacetate or acetoacetyl-CoA, which condenses with NMPy to form hygrine.
Supplementary Note 3. Elimination of acetate auxotrophy to decrease spontaneous hygrine production.
We attempted to reduce hygrine production resulting from spontaneous condensation with an endogenous ketometabolite derived from acetate. We examined the impact of removing fed acetate in the NMPy-producing strain (CSY1246), which lacks the enzymes for MPOB and tropinone biosynthesis, to isolate the acetate-dependent hygrine production mechanism from the MPOB-dependent one (Fig. 4a). We rescued acetate auxotrophy in CSY1246 by expressing ALD4 and ALD6 on low-copy plasmids (pCS4248, 4249), and evaluated accumulation of hygrine and 4MAB acid via LC-MS/MS analysis after 48 hours. While reconstitution of ALD4 or ALD6 enabled CSY1246 to grow in the absence of fed acetate (Supplementary Figure 11a), addition of ALD4 caused a five-fold increase in the accumulation of 4MAB acid while ALD6 did not produce a significant increase (Supplementary Figure 11b). Elimination of acetate feeding with ALD4 or ALD6 resulted in 38% and 59% decreases in hygrine accumulation, respectively, confirming that condensation with an acetate-derived metabolite contributes substantially to hygrine production. As acetate is an essential metabolite, it may not be feasible to completely eliminate this route for hygrine production. We re-integrated ALD6 into our tropine-producing strain (CSY1248) to make strain CSY1249, and measured the accumulation of metabolites between NMPy and tropine via LC-MS/MS analysis after 48 hours of growth. Restoration of acetate metabolism in CSY1249 resulted in a 2.7-fold increase in tropine titers (1.5 mg/L) relative to CSY1248 (565 μg/L) and a 1.6-fold increase in hygrine accumulation (Fig. 4d). ALD6 expression resulted in increases in NMPy and tropinone production and MPOB consumption (Supplementary Figure 11c), suggesting that elimination of acetate auxotrophy may improve metabolite flux through the entire pathway despite increasing the hygrine side reaction.

Supplementary Note 4. Optimization of media composition for de novo tropine biosynthesis in CSY1251.
We examined the impact of media composition on tropine production. We cultured CSY1251 in defined (yeast nitrogen base; YNB) or complex (yeast extract and peptone; YP) media supplemented with 0-5× amino acids, 2-4% dextrose as a carbon source for growth, and 2-5% of other carbon sources, and measured tropine titers in the media by LC-MS/MS after 72 hours (Fig. 4e). Increasing amino acid concentration to 5× or 2.5× resulted in 25% and 18% decreases in tropine titer in defined and complex media, respectively. Doubling the dextrose concentration from 2% to 4% decreased tropine titers by 67% in defined media, but increased titers by 55% in complex media. Supplementation of defined media with 2% dextrose and 2% galactose, raffinose, trehalose, arabinose, or sorbitol resulted in 13-60% increases in tropine titer relative to 4% dextrose, whereas all carbon sources except glycerol resulted in 48-68% decreases in tropine titer relative to 2% dextrose only. Supplementation of dextrose and glycerol increased tropine production relative to dextrose alone; under the best media condition (YNB + 1X amino acids + 2% dextrose + 5% glycerol; denoted YNB-G), tropine titers reached 4.9 mg/L, 35% greater than with 2% dextrose alone.
We observed that synthetic defined media outperformed richer YP media, contrary to previous reports of heterologous biosynthesis in yeast 9,10 . Similarly, higher starting concentrations of amino acids, dextrose, and other fermentable sugars resulted in decreased tropine production, potentially due to higher protein synthesis and growth rates leading to increased enzyme misfolding 10 , as well as repression of late-stage promoters and endogenous pathways for biosynthesis and transport of basic amino acids 11,12 . Among alternate carbon sources, only glycerol supplementation increased tropine production, possibly due to its stabilization of cellular lipid membranes, improved folding and stability of heterologous proteins, and its role in the regeneration of the NADPH cofactor required for the activity of AbCYP82M3 and DsTR1 10 .

Supplementary Note 5. Substrate feeding experiments to elucidate the stereochemistry of the EcCS reaction.
Given that a prior report of recombinant EcCS detected acyltransferase activity only on β-tropane isomers, such as pseudotropine and methylecgonine, and not on α-tropanes 13 , we performed feeding experiments to support the observed EcCS activity on α-tropine. We cultured wild-type yeast (CEN.PK2) or the tropine-producing strain (CSY1251) transformed with plasmid-based AtPAL1 (pCS4252), At4CL5 and EcCS (pCS4207) in media supplemented with pure αtropine and/or trans-cinnamic acid for 72 hours and analyzed the medium for cinnamoyltropine accumulation by LC-MS/MS (Fig. 5b-i). In the transformed CEN.PK2 strain, cinnamoyltropine was detected with expression of AtPAL1, At4CL5, and EcCS and supplementation of 0.5 mM α-tropine, but not in the absence of any of these components, validating that EcCS expressed in yeast exhibits activity on the α isomer of tropine. In the transformed tropine-producing strain (CSY1251), cinnamoyltropine was detected with expression of At4CL5 and EcCS and addition of 0.2 mM cinnamic acid to the media, as well as with expression of AtPAL1 in place of cinnamate supplementation, but not in the absence of a cinnamate source, verifying that the detected tropane ester is derived from trans-cinnamic acid.

Supplementary Discussion
Our analysis of intermediate accumulation between NMPy and tropine initially identified AbCYP82M3 as a pathway bottleneck, consistent with a prior study indicating that this enzyme is rate-limiting for tropinone biosynthesis in A. belladonna 2 . We improved stability of P450 expression from genomic integration compared to plasmid-based expression, consistent with previous reports of ER stress caused by plasmid-based overexpression of P450 enzymes in yeast 14 . However, the impact of this alteration on MPOB accumulation remains unclear. We observed three distinct peaks in the MRM (186 → 84) spectrum for MPOB produced in yeast (Supplementary Figure 9a, S11d), in contrast to a single peak previously observed for chemically synthesized MPOB 2 , suggesting that additional isomers of MPOB may be Substrate feeding experiments may elucidate the identity of the three MPOB peaks and the kinetic preferences of AbCYP82M3 for each, which could inform structure-guided AbPYKS engineering to improve flux through the desired MPOB substrate.