A yeast platform for high-level synthesis of tetrahydroisoquinoline alkaloids

The study addresses the challenge of producing tetrahydroisoquinoline (THIQ) alkaloids—particularly benzylisoquinoline alkaloids (BIAs) like morphine and codeine—at industrially relevant titers using microbial hosts. Plant production is often low and chemical synthesis is difficult due to structural complexity and stereocontrol issues. Prior microbial reconstructions of BIA pathways in yeast yielded very low titers (<2 mg L−1), with (S)-reticuline formation being a key bottleneck. Although E. coli can accumulate (S)-reticuline at higher levels, it lacks the membrane machinery for P450-dependent downstream steps, complicating total biosynthesis. The research aims to create a high-flux yeast platform for THIQ production by overcoming redox shunting of aldehyde intermediates in the Ehrlich pathway, boosting precursor supply, and exploiting the broad substrate tolerance of norcoclaurine synthase (NCS) to access both canonical and non-canonical THIQ scaffolds. The purpose is to enable scalable biosynthesis of BIAs and expand accessible THIQ chemical space for pharmaceuticals.

Previous successes in microbial biomanufacturing (e.g., artemisinic acid and β-farnesene) demonstrate industrial potential. For BIAs, multiple pathways (noscapine, sanguinarine, morphine, codeine, hydrocodone) have been reconstituted in yeast; however, titers remained very low largely due to inefficient formation of (S)-reticuline. E. coli reached 160 mg L−1 (S)-reticuline but required multi-strain compartmentalization for morphinan biosynthesis due to P450 limitations. The Pictet–Spengler condensation catalyzed by NCS underlies >3000 THIQs, yet natural diversity mainly derives from a few aldehydes; in vitro work shows NCS can accept many aldehydes and even ketones, suggesting untapped chemical space. The yeast Ehrlich pathway produces fusel alcohols/acids from amino acids via aldehydes, but the responsible oxidoreductases have been poorly defined with redundancy complicating assignments. Prior attributions focused on ADH/ALD families without definitive links for specific substrates like 4-HPAA. This study builds on these insights to identify key reductases/dehydrogenases, stabilize aldehyde pools, and exploit NCS promiscuity in vivo.

Host and construction: S. cerevisiae BY4741 was the base strain. Dopamine production pathway was introduced, and extensive strain engineering was performed through iterative CRISPR-Cas9-mediated genomic integrations and in vivo DNA assembly. gRNAs were designed using CCTop and a CRISPRi design tool; multiple genomic landing pads and characterized integration sites were used to assemble multi-copy and multi-gene pathways. Key pathway engineering: - Canonical BIA initiation: Enhanced condensation of 4-hydroxyphenylacetaldehyde (4-HPAA) and dopamine via norcoclaurine synthase (NCS). Initial NCS was NdNCSΔN20; a superior Coptis japonica NCS truncation (CjNCSΔN35) was identified and implemented, providing higher activity. - Precursor supply: Upregulated shikimate and L-tyrosine pathways via overexpression of ARO2 (chorismate synthase), TYR1 (prephenate dehydrogenase), ARO10 (phenylpyruvate decarboxylase), and feedback-resistant ARO4FBR and ARO7FBR; deleted PHA2 and TRP3 to channel flux to L-tyrosine in early strains, later restored for prototrophy. - Dopamine supply: Improved tyrosine hydroxylation by introducing CYP76AD5 in addition to (then deleting) a less efficient engineered CYP76AD1W13L F309L; balanced precursor supply based on supplementation assays (L-DOPA vs L-tyrosine). - Eliminating aldehyde loss to fusel products: Systematic deletions identified and removed NADPH-dependent oxidoreductases responsible for converting 4-HPAA to tyrosol (alcohol) or 4-HPAC (acid). Deletions included ARI1, ADH6, YPR1, YDR541C, AAD3, GRE2 (major reductase), and HFD1 (major dehydrogenase to 4-HPAC). ALD4 was deleted early to reduce acid formation and reintroduced later to support ethanol utilization in fed-batch. - Extension to (S)-reticuline: Added Ps6OMT and PsCNMT (Papaver somniferum) for THIQ methylations to (S)-coclaurine and (S)-N-methylcoclaurine; EcNMCH/CYP80B1 (Eschscholzia californica) for benzylic hydroxylation; AtATR2 (A. thaliana CPR) for P450 electron transfer; Ps4′OMT2 to methylate to (S)-reticuline, with a second copy installed to relieve a bottleneck. - Strain fitness improvements: Removed CYP76AD1 variant to avoid side reactions; repaired mitochondrial stability-associated alleles (SAL1G403L, MIP1A661T, CAT5191M); restored PHA2 and TRP3 for amino acid prototrophy in production strain LP507. Cultivation and assays: - Microtiter plate cultivations in sucrose media to screen titers and fusel products; growth curves via OD measurements. - Fed-batch fermentations (3 L) in simple mineral medium with pulsed sucrose feeding; monitored OD, metabolites by HPLC-UV and LC-MS. - THIQ synthesis assay from supplemented amino acids: Grew reticuline-producing strain (LP501 with multiple oxidoreductase deletions) on media where single amino acids served as main nitrogen sources to channel Ehrlich-derived aldehydes into NCS-catalyzed THIQs. Analytics: - HPLC-FT-MS and QTOF-MS for detection/identification of dopamine, BIAs, and substituted THIQs; targeted MS/MS fragmentation; chiral LC-MS for norcoclaurine enantiomeric analysis confirming (S)-enantiomer; HPLC-UV quantification of key metabolites (tyrosol, 4-HPAC, dopamine, (S)-norcoclaurine, (S)-reticuline). Statistical analyses via two-tailed Student’s t-tests (P<0.05).

- High titers: Achieved 1.6 g L−1 (S)-norcoclaurine in fed-batch (strain LP478) and up to 4.6 g L−1 (S)-reticuline with 0.6–0.7 g L−1 (S)-norcoclaurine in fed-batch (strain LP507/LP501 lineage), representing a 57,000-fold improvement over the first-generation strain. - Rate-limiting steps resolved: Identified and deleted key NADPH-dependent oxidoreductases diverting 4-HPAA to tyrosol and 4-HPAC. GRE2 deletion reduced tyrosol by 83% in plates; combined gre2Δ hfd1Δ reduced fusel byproducts drastically in fed-batch (tyrosol and 4-HPAC each <0.3 g L−1), increasing flux to BIAs. - Enzyme improvements: CjNCSΔN35 outperformed NdNCSΔN20 (~5-fold higher in early comparisons). CYP76AD5 improved dopamine supply; adding a second Ps4′OMT2 copy alleviated (S)-3′-hydroxy-N-methylcoclaurine accumulation and raised reticuline by ~45% (to 492 mg L−1 in plates). - Chiral control: Norcoclaurine produced in high-titer strains was exclusively the (S)-enantiomer. - Non-canonical THIQs de novo: Detected THIQs formed via NCS from endogenous Ehrlich-pathway aldehydes: from L-phenylalanine (via PAA), L-tryptophan (via IAA), and L-leucine (via 3-methylbutanal); salsolinol formed spontaneously from acetaldehyde and dopamine in dopamine-producing strains. - Fed amino acids expand diversity: Feeding amino acids as nitrogen sources yielded >10 substituted THIQs, including norlaudanosoline (validated by MS/MS vs authentic standard), sulfur-containing THIQ from methional, and alkyl-substituted THIQs (ethyl, propyl, butyl, pentyl). BIA methyltransferases (Ps6OMT, PsCNMT) modified these scaffolds to produce derivatives such as N-methylisosalsoline and lophocerine (from leucine-derived scaffold). - Growth and fitness: Despite extensive engineering, key production strains maintained acceptable growth; LP478 had ~38% reduced µmax vs BY4741 in microtiter assays but supported high fed-batch titers; mitochondrial stability allele repairs improved robustness.

The study demonstrates that resolving aldehyde redox shunting in the yeast Ehrlich pathway unlocks high-flux THIQ biosynthesis. By identifying GRE2 as a major 4-HPAA reductase and HFD1 as the predominant dehydrogenase to 4-HPAC, and deleting a redundant set of oxidoreductases, the authors stabilized the 4-HPAA pool and redirected carbon to the BIA pathway. Coupled with optimized precursor supply (shikimate/tyrosine pathway upregulation), improved NCS and tyrosine hydroxylase activities, and removal of non-productive reactions, yeast attained gram per liter titers of (S)-reticuline in a simple mineral medium. This directly addresses the historical bottleneck of low (S)-reticuline titers that impeded downstream opioid and BIA biosynthesis in a single yeast host. The platform also exploits NCS promiscuity to generate diverse non-canonical THIQs both de novo from endogenous amino acid catabolism and from fed amino acids, and shows that canonical BIA tailoring enzymes can derivatize these scaffolds. These findings extend the accessible THIQ chemical space and suggest a viable route to microbially synthesize both natural and novel THIQ-derived therapeutics. The work further revises understanding of fusel metabolism control in yeast, with implications for flavor compound management and aldehyde-based bioproduction.

This work establishes a robust yeast platform for THIQ biosynthesis, achieving 4.6 g L−1 (S)-reticuline and demonstrating exclusive (S)-norcoclaurine formation at high titers. By systematically eliminating key oxidoreductases (including GRE2 and HFD1) to prevent aldehyde loss and by optimizing pathway enzymes (CjNCSΔN35, CYP76AD5, Ps6OMT/PsCNMT/EcNMCH, and Ps4′OMT2), the authors overcame the major bottleneck in BIA production. The platform synthesizes diverse non-canonical THIQ scaffolds from endogenous and supplemented amino acids, and these scaffolds can be tailored by BIA methyltransferases to yield natural and novel alkaloids (e.g., lophocerine). This provides a blueprint for scalable microbial production of BIAs, including downstream opioids, and broadens access to THIQ chemical diversity. Future work should focus on integrating and optimizing downstream morphinan pathways (e.g., leveraging newly identified thebaine and neopinone biosynthetic enzymes), confirming structures of additional non-canonical THIQs, assessing pharmacological properties, improving strain fitness and productivity at high cell density, and applying the aldehyde-stabilizing strategy to other aldehyde-derived products.

- Structural confirmation: Except for norlaudanosoline (validated against an authentic standard), identities of many substituted THIQs are based on accurate mass, retention time shifts, and MS/MS fragmentation consistent with expected structures but lack full structural confirmation. - Downstream pathway uncertainty: Efficiency and compatibility of downstream opioid biosynthetic enzymes in this high-titer yeast background remain to be demonstrated. - Fitness and growth: Engineered strains exhibited reduced growth rate (e.g., ~38% reduction in µmax for LP478 vs BY4741 in microtiter), which may impact industrial performance and may require further optimization. - Scope of oxidoreductase redundancy: While key enzymes were identified and deleted, additional redundant activities may exist and could affect other aldehyde pools under different conditions. - Reliance on supplemented amino acids for some analogs: Production of certain non-canonical THIQs currently requires feeding specific amino acids; endogenous pathway engineering would be needed for fully de novo production at scale.