Saccharomyces cerevisiae strains were maintained on Yeast Peptone Dextrose (YPD) medium (20 g L⁻¹ peptone, 10 g L⁻¹ yeast extract, 20 g L⁻¹ glucose; 20 g L⁻¹ agar for plates). Geneticin (200 mg L⁻¹) and nourseothricin (100 mg L⁻¹) were supplemented in the media when needed for selection of Cas9 (KanMX) and gRNA (natMX) plasmids, respectively. Sub-cloning was done in electro-competent Escherichia coli DH5α. Lysogeny Broth (LB) medium (10 g L⁻¹ tryptone, 5 g L⁻¹ yeast extract, 10 g L⁻¹ NaCl; 15 g L⁻¹ agar for plates) with 100 mg L⁻¹ ampicillin was used for E. coli liquid cultivation and selection on solid plates. All strains were stored in 25% (v/v) glycerol at −80°C. Defined mineral medium (Verduyn et al., 1992) was used as basis for the bioconversions, buffered at pH 6.5.

Strains and plasmids used in this work are listed in Table 1. Standard molecular biology methods were used (Sambrook and Russell, 2001). Preparative and diagnostic PCR for genome engineering was performed using Physion and DreamTaq polymerase (Thermo Scientific, US), respectively. All modifications were verified by colony PCR using primers listed in Supplementary Table S1. Primers were designed using SnapGene^® (GSL Biotech LLC. Version 3.03, US) and purchased from Eurofins (Eurofins Genomics, Germany). Agarose (0.8%) gel electrophoresis was used for estimating size of the DNA fragments. All plasmids (Table 1) were purified using GeneJet plasmid MiniPrep Kit (Thermo Scientific, US) after propagation in E. coli. Alanine dehydrogenase gene (Genbank ID 936557) from Bacillus subtilis (BsAlaDH) was codon optimized for expression in S. cerevisiae and obtained in a pUC57 vector (Genescript, US). BsAlaDH gene was amplified with primers flanked with restriction sites SfaAI/XhoI and subsequently cloned into pRP005 between TEF1p promotor and ADH1t terminator flanked by homologous regions for chromosomal integration, generating pNM011. pNM012 was constructed by cloning TEF1p into the backbone plasmid pCfB3034. Correct DNA sequence of constructed plasmids were verified by Sanger sequencing (Eurofins Genomics, Germany). S. cerevisiae strains were engineered by the previously described CRISPR-Cas9 toolkit (Jessop-Fabre et al., 2016). The lithium acetate-based protocol (Gietz and Schiestl, 2007) with the addition of DMSO for improving cell permeability, was used for generating competent yeast cells. TMB4375 strain carrying six copies of Chromobacterium violaceum amine transaminase gene (CvTA) (Weber et al., 2017) was transformed with the pCfB2312 plasmid, carrying Streptococcus pyogenes gene, SpCas9, and used as background strain for the strain engineering. Transformation with a linearized pNM011 (using NotI restriction enzyme) and pCfB3045 (gRNA) plasmid generated strain TMBNM032. Deletion of ADH6 was achieved by transforming TMBNM032 with a PCR product carrying AUR gene cassette for aureobasidin A (AurA) resistance and ∼50bp homologous flanks to the target locus, generating strain TMBNM033. Transformants resistant to AurA were selected in solid medium and correct gene knockout was analysed with diagnostic PCR. Codon optimized genes coding for CaAT (Tyramine N-(hydroxycinnamoyl) transferase from Capsicum annum, and IpfF (Ibuprofen CoA ligase from Sphingomonas sp. Ibu-2)) were integrated using a method developed previously (Muratovska et al., 2022a), generating strain TMBNM034 (Table 1).

Pre-cultures were prepared from a single colony in 5 ml YPD grown overnight. Shake flask with 25 mL defined mineral medium (Verduyn et al., 1992) and glucose (20 g L⁻¹) was inoculated to initial optical density (OD_620nm) ∼0.5 and incubated at 30°C in 180 rpm shaking incubator for 24–26 h to generate biomass. Subsequently, yeast cells were harvested by centrifugation at 3220 g for 10 min (Eppendorf Centrifuge 5810 R, Germany) and re-suspended at OD_620nm ∼5 in 50 mL of reaction medium. The reaction medium was based on defined mineral medium buffered at 6.5 pH, and supplemented with vanillin (5 mM), glucose (10 g L⁻¹) or ethanol (0.5 or 10 g L⁻¹) as co-substrate, and with or without L-alanine (50 mM) and ammonium chloride (200 mM) as nitrogen source. Whole-cell bioconversions were performed under aerobic or anaerobic conditions. For the aerobic conditions, bioconversions were carried out in 500 mL shake flasks containing 50 mL medium and incubated at 30°C in 180 rpm shaking incubator as described above. Anaerobic bioconversions were performed in sealed serum vials containing 50 mL medium and sparged with nitrogen gas for 1 h before the start of the reaction and incubated at 30°C in 180 rpm shaking incubator.

Pre-cultures were prepared from a single colony inoculated in a shake flask with 50 mL defined mineral medium incubated under shaking (180 rpm) at 30°C for around 24 h. The pre-culture was used to inoculate 3 L bench-top bioreactor (Minifors 2. Infors HT, Bottmingen, Switzerland) at OD_620nm = 0.1 in 1 L of defined mineral medium with 50 g L⁻¹ glucose. The pH was kept at 6.5 by automated addition of 3 M KOH, the stirring was constant at 300 rpm, and the temperature was set at 30°C. Yeast cells were first grown under aerobic conditions with air sparging set to 500 ml/min (0.5 vvm) for approximately 20 h, and the whole-cell bioconversions were subsequently started by adding the substrates (vanillin and non-anoic acid) to the media. For the non-sparged condition, aeration was stopped to achieve oxygen-limited conditions after the initial growth phase, and directly prior to the supplementation of substrates.

Samples were collected with a syringe regularly from cell cultures for analysis of yeast cell concentration, yeast cell viability, and extracellular metabolite analysis. Yeast cell concentration was estimated by measuring optical density at 620 nm (OD_620nm) with a spectrophotometer (Ultrospec 2,100 pro UV/Visible spectrophotometer, Amersham Biosciences, Buckinghamshire, United Kingdom). Yeast cell viability was monitored by flow cytometry (Accuri C+, BD, US) using the fluorescent dyes SYBR Green I and propidium iodide, as described previously (Rao et al., 2021). Instrument settings were adjusted to the following: medium flow rate, FSC-H threshold at 80,000, and acquisition of 10,000 events. FCM data analysis was made with FlowJo^® (United States). Gating of yeast cells was based on FSC-H and SSC-H signals to distinguish cells from background noise, and gating of damaged and intact cells were based on SYBR Green I detected with FL1-H (530/30 nm), and PI detected with FL3-H (≥670 nm) signals. Yeast cells treated with 70% ethanol for 30 min were used as control to identify damaged cells and to adjust the gates. For the analysis of glucose, ethanol, glycerol and acetate, a Waters HPLC system equipped with an Aminex HPX-87 H ion-exchange column (7.8 × 300 mm, Bio-Rad, Hercules, United States), and a refractive index detector (Waters 2,414, Milford, United States) were used, as described previously (Weber et al., 2014b). The mobile phase was 5 mM H₂SO₄ with a flow rate of 0.6 mL/min, and the column was kept at 60°C. For reverse-phase HPLC analysis of vanillin, vanillylamine, vanillic alcohol and vanillic acid, a Select C18 column (4.6 × 150 mm) with mobile phase consisting of ultrapure water and 0.1% trifluoroacetic acid (TFA) (A) and acetonitrile (B). The method was isocratic with 65% A and 35% B for 10 min with a flow of 1 ml min⁻¹, at room temperature and the monitored wavelength was at 281 nm, as described previously (Manfrao-Netto et al., 2021). Prior to HPLC analysis of nonivamide, product extraction was performed from 20 ml samples from the bioreactors. Supernatant was collected after centrifugation at 3220 g for 10 min, transferred to an extraction funnel and extracted with one volume of ethyl acetate. The organic phase was collected and evaporated using a rotary vacuum evaporator. The resulting viscous oil was dissolved in 1 mL of methanol and HPLC analysis were performed using the same Select C18 column (4.6 × 150 mm) as previously described (Muratovska et al., 2022a).

The rate ratio is defined as the vanillylamine formation rate (mmol h^-1) divided by the by-product (vanillic alcohol and vanillic acid) formation rate (mmol h^-1) for the initial 6 h. Molar yield is defined as moles of product (vanillylamine) or by-product (vanillic alcohol or vanillic acid) per moles of consumed substrate (vanillin) after 48 h. The specific vanillylamine productivity (µmol OD⁻¹ h⁻¹) is calculated for the initial 6 h.

Yeast carrying six copies of the CvTA gene expressed under control of the strong constitutive TDH3 promoter was evaluated under different reaction conditions for whole-cell conversion of vanillin to vanillylamine (Supplementary Figure S1). Most vanillin reductases are NADPH-dependent, and vanillin dehydrogenase is NAD + -dependent; therefore, the redox balance significantly influence by-product formation. In Crabtree positive yeast, such as Saccharomyces cerevisiae, glucose is readily fermented to ethanol in both aerobic and anaerobic conditions, while ethanol is slowly respired in aerobic condition and stays practically inert in anaerobic condition (De Deken, 1966). The availability of NADPH is limited with the use of ethanol as co-substrate, especially under anaerobic condition resulting in suppression of the activity of NADPH-dependent reductases (but not NADH-dependent reductases) (Kratzer et al., 2008). In contrast, NADPH is efficiently regenerated in the pentose phosphate pathway during glucose assimilation (Parachin et al., 2009). Another important factor is the abundance of pyruvate formed both as central carbon metabolite and from the amine donor L-alanine in yeast (Weber et al., 2014b; Weber et al., 2017). Pyruvate is a strong inhibitor of CvTA and its removal shifts the equilibrium to the amine (Sánchez-Moreno et al., 2012; Weber et al., 2017). Here, four conditions were evaluated: (i) aerobic with glucose as carbon source, (ii) aerobic with ethanol as carbon source, (iii) anaerobic with glucose as carbon source, and (iv) anaerobic with ethanol as carbon source. Pre-cultures were made in the same way and whole-cell bioconversions were performed at relatively high cell density (OD_620nm = 5); therefore, the expression level of endogenous genes was the same at start of the reaction, and any differences in product spectra would be due to the specific applied reaction environment. Significant differences were observed for the different conditions, with the highest vanillin consumption rate (0.11 ± 0.00 mmol OD⁻¹ h⁻¹) under aerobic conditions and with ethanol used as a co-substrate (Supplementary Figure S1). This condition resulted in more production of vanillylamine (3.21 ± 0.56 mM) compared to the rest, but also of the by-products vanillic alcohol (4.11 ± 1.48 mM) and vanillic acid (1.72 ± 0.40 mM) (Supplementary Figure S1). To compare the activity of CvTA and endogenous enzymes acting on vanillin, the rate ratio, defined as vanillylamine formation rate per by-product formation rate, was calculated (Table 2). A higher rate ratio (5-7- fold) was observed with ethanol as co-substrate, both under aerobic and anaerobic condition. The difference in rate ratio under aerobic conditions was mostly due to an increased formation of vanillylamine with the use of ethanol as the co-substrate. A significant amount of vanillic alcohol and vanillic acid was formed for both glucose and ethanol as co-substrate under aerobic conditions, which was expected due to balanced levels of NAD(P)H/NAD(P)+ ratios under respiration (Murray et al., 2011). Under anaerobic conditions, the transamination was less efficient, however, alcohol and acid formation were also significantly lower. Worth noting is the almost complete suppression of vanillic acid formation under anaerobic conditions, possibly due to low availability of NAD⁺, or NADP⁺ (Supplementary Figure S1C, D). In aerobic conditions, the vanillin consumption rate was generally faster, regardless on the choice of co-substrate; however, which end products were formed from vanillin depend on the choice of co-substrate and oxygen availability.

The highest molar yield of vanillylamine was reached under the anaerobic condition with ethanol as co-substrate (0.71 ± 0.44 mol/mol), while the lowest molar yield of vanillylamine per consumed vanillin after 48 h was on glucose in the aerobic condition (0.15 ± 0.05 mol/mol) (Table 2). However, a problem during the anaerobic condition was incomplete conversion of vanillin. A reason for this could be inhibition of the transaminase by pyruvate accumulation, or low limitation in amine donor availability. Alanine was supplied in excess and gets converted to pyruvate by both CvTA, as well as by endogenous transaminases, for example by alanine transaminase (encoded by ALT1) (García-Campusano et al., 2009). Pyruvate could be removed from the reaction system through conversion to ethanol by pyruvate decarboxylase and alcohol dehydrogenase. During respiration, pyruvate is converted to acetyl-CoA by pyruvate dehydrogenase, and further oxidised in the tricarboxylic acid (TCA) cycle to CO₂ and generating NADH (Flikweert et al., 1996). When glucose is present during the bioconversion, there would be higher intracellular pyruvate levels (Weber et al., 2014b; Weber et al., 2017), which could explain why conditions with ethanol gave better transamination yields.

To relieve potential pyruvate inhibition and improve the amine donor supply, an NADH-dependant alanine dehydrogenase from Bacillus subtilis, BsAlaDH, was expressed in strain TMBNM032 under control of the TEF1 promoter. Co-expression of CvTA and BsAlaDH resulted in almost complete conversion of vanillin within 48 h (Supplementary Figure S2), which could indicate that pyruvate accumulation and/or alanine limitation was the issue in the parent strain TMB4375. Vanillin consumption rate for this strain as well as the vanillylamine titres after 48 h was almost double when compared to strain TMB4375 in the anaerobic conditions (Table 2; Supplementary Figures S1C, D vs. Supplementary Figures S2A, B). The rate ratios (vanillylamine formed per by-products formed) were not noticeably different compared to TMB4375 (Table 2), except under anaerobic conditions with ethanol as co-substrate (4.86 rate ratio) Here, only low amounts of vanillic alcohol were detected after 48 h (0.9 mM; Supplementary Figure S2B). Compared to TMB4375 in the same condition, the formation of vanillic alcohol decreased with the expression of AlaDH by around a third, and the vanillylamine yields were increased from 0.71 to 0.83 mol/mol (Table 2).

Strain TMBNM033 was further engineered by expressing CaAT and IpfF, previously shown to be effective in the biosynthesis of nonivamide from non-anoic acid and vanillylamine (Muratovska et al., 2022a). The constructed yeast strain TMBNM034, containing CvTA, BsAlaDH, ΔADH6 and CaAT/IpfF, was characterised in bench scale bioreactors for production of nonivamide from vanillin and non-anoic acid (Figure 3). The yeast cells were first grown under aerobic conditions with glucose as carbon and energy source; subsequently, the substrates vanillin and non-anoic acid were added. At this point, the cell density was high (OD_620nm = 15–17) enabling higher conversion rates. Aeration was found to affect the whole-cell transamination reaction negatively, both in terms of vanillylamine formation, and by-product formation (Table 2). On the other hand, lack of oxygen may result in limited supply of ATP and CoA-SH required for the coenzyme A ligase to generate the acyl-coa for the amide-forming step. To shed light into this possible dilemma, two reaction configurations were compared: (1) air-sparged conditions (Figures 3A, B) and (2) non-sparged oxygen-limited conditions (Figures 3C, D). The supplied vanillin and non-anoic acid were converted to nonivamide in both cases, but slightly more in the presence of oxygen (0.19 vs. 0.12 μmol L⁻¹ OD⁻¹) (Figure 3; Table 3). The results demonstrate that amidation indeed can be combined with reductive amination in one-pot in yeast. However, nonivamide titers were low and a significant accumulation of vanillylamine (3–5 mM) was observed, showing that the amide-forming step is the bottleneck for production from vanillin. It may be speculated that increasing the amount of air in the bioreactor may improve the situation by increasing the supply of ATP and CoA-SH; however, other problems may also be at hand, e.g. CaAT, and IpfF expression levels or the specific activity of the enzymes may be low. Also, the availability of non-anoic acid may be limiting since it is supplemented in low concentration due to toxicity and low solubility (Muratovska et al., 2022a). The experiments indicate that the in vivo transamination reaction is hampered by increased aeration of the reaction broth. The rate ratio of vanillylamine formation per by-product formation was higher under non-sparged conditions (3.5 vs. 2.4). Formation of vanillic alcohol and even more so, of vanillic acid, were lower, resulting in molar yields from consumed vanillin of 0.28 ± 0.03 and 0.03 ± 0.01 at 48 h respectively. This shows that the ADH6 deletion is not sufficient to completely prevent vanillic alcohol formation under the applied conditions. Nevertheless, while ethanol assimilation was not detected under anaerobic condition, it was consumed throughout the bioconversion sparged with air, indicating functional regeneration of cofactors. Cell viability, measured by flow cytometry, remained high throughout the bioconversions with >98% intact cells for both conditions (Supplementary Figure S3 in supplementary), and could not explain the observed differences in activity.