A microbial supply chain for production of the anti-cancer drug vinblastine

Chemical standards

All chemical standards (Supplementary Table 2) had a purity of 95% or higher.

Gene discovery

Sequence for Vmi8HGO-A was obtained from V. minor transcriptome data⁴¹ using BLASTn⁴² and the C. roseus 8HGO-A sequence (GenBank accession no. AHK60836.1) as the query. Open reading frames were predicted using TransDecoder (http://transdecoder.github.io).

Plasmid construction

All plasmids were constructed by USER cloning⁴³ or restriction enzyme cloning, followed by chemical transformation into E. coli DH5α competent cells, and plated on Luria–Bertani agar containing 100 μg ml⁻¹ ampicillin and grown at 37 °C overnight (16–20 h). E. coli transformants were genotyped by colony PCR, and colonies with correct size of insertion were grown in 5 ml of Luria–Bertani medium containing 100 μg ml⁻¹ ampicillin at 37 °C for 16–20 h, followed by miniprep plasmid extraction and validated by Sanger sequencing. All biosynthetic genes (Supplementary Table 1) were codon optimized for yeast expression and synthesized by either IDT or Twist Biosciences. All genes were assembled as expression cassettes in either centromeric (pRS series) or 2-μm (pESC series) plasmids by USER cloning. Single guide-RNA plasmids were constructed by USER cloning a double-stranded oligo containing the 20-nucleotide guide-RNA sequence into a 2-μm vector with either the LEU2 or URA3 selection marker. Several guide-RNA plasmids were constructed by USER assembly of all guide-RNA expression cassettes amplified from single guide-RNA plasmids. All plasmids (Supplementary Table 4) were verified by Sanger sequencing of the insertion.

Yeast strain construction

All yeast strains (Supplementary Table 2) were constructed on the basis of the wild-type strain CEN.PK2-1C using the lithium acetate/single-stranded carrier DNA/PEG protocol⁴⁴, and have the same auxotrophy for uracil, histidine, leucine and tryptophan. Genome editing (integration, deletion or swapping) was performed using the CasEMBLR method⁴⁵. All DNA parts were amplified from precloned plasmids using Phusion High-Fidelity PCR Master Mix with HF Buffer (Thermo Fisher Scientific) using primers containing 25–30 bp overhangs (IDT). For strains MIA-AU and MIA-AW-2, SpyCas9 was expressed from a centromeric plasmid with the TRP1 selection marker. For all other MIA strains, CEN.PK2-1C wild-type strain was first transformed with a plasmid (TRP1) harbouring the SpyCas9 cassette, followed by another transformation to integrate the SpyCas9 cassette into EasyClone site XII-1 for stable expression⁴⁶. For all editing were transformed with a plasmid expressing a single or multiple-targeting guide-RNA fragments, as well as DNA parts as repair templates Transformants were plated on SC agar lacking the appropriate amino acid(s) (uracil, leucine, histidine or tryptophan) for selection of the guide-RNA plasmid, and grown for 3–4 days at 30 °C. All strains with genomic integration were confirmed by genotyping PCR and Sanger sequencing of the integrated gene cassettes.

For all strictosidine strains except MIA-AU and MIA-AW-2, a unique guide-RNA sequence was included at the beginning of all gene cassettes (beginning of the promoter sequence) for easy swapping of the promoter or deletion of whole gene cassettes (Extended Data Fig. 3a). To swap a promoter/gene in an expression cassette, yeast cells were transformed with a plasmid expressing the guide-RNA, together with a new promoter/gene and parts overlapping with sequences up- and downstream of this promoter/gene (Extended Data Fig. 3b). Similarly, to delete a whole cassette, yeast cells were transformed with a plasmid expressing the specific guide-RNA and two parts (with a 50–60 bp overlapping sequence) that are homologous to the up- and downstream of the locus to be deleted (Extended Data Fig. 3c). Transformants were plated on SC agar lacking either leucine or uracil for selection of the guide-RNA plasmid and confirmed by colony PCR and sequencing.

Media and yeast cultivation

Yeast cells were cultivated in either SC defined medium or YPD rich medium, both containing 2% glucose as the carbon source (except for the screening of the iridoid synthase + cyclase and 8HGO variants, which used glucose and galactose). For batch-fed cultivations, 3× SC medium (2% glucose) supplemented with 3 mM tryptophan was used to reach a high biomass.

For all small-scale cultivations, precultures were typically made by growing yeast colonies (biological replicates) in 150 µl of SC medium (lacking appropriate amino acid for auxotrophic selection) or YPD medium on a 96-well microtitre plate and incubated at 30 °C shaking at 300 r.p.m. overnight (roughly 16–20 h). The maximum specific growth rate for all yeast strains were characterized using the Growth Profiler 960 system (Enzyscreen) (Supplementary Fig. 9). All cultures were grown in 300 μl of SC medium in 96-well square-shaped microtitre plates for at least 3 days. Cultures were grown at 30 °C with shaking speed of 250 r.p.m. and the cell density was measured every 30 min. Data from the exponential growth phase were used to determine the maximum specific growth rate by linear regression. Small-scale production assays were performed using either 96-well (with 0.5 ml of medium) or 24-well (with 2 ml of medium) deep-well plates. After 1 day (typically reaching the stationary phase), 10 μl of precultures were transferred to either SC lacking appropriate amino acid(s) or YPD medium supplemented with one or more precursors. For testing strictosidine production, yeast strains were fed with 0.1–1 mM tryptamine and one of the precursors including geraniol, 8-hydroygeraniol, 8-oxogeranial, cis-trans-nepetalactol, loganic acid, loganin and secologanin. For testing tabersonine and catharanthine production, cells were fed with 1 mM tryptamine and 50–250 μM secologanin. For testing vindoline or vinblastine production, cells were fed with 100 μM tabersonine or 100 μM vindoline and 100 μM catharanthine, respectively. All compounds used in the feeding experiments are listed in Supplementary Table 5.

Small-scale fed-batch cultivations were performed in BioLector Pro systems using flower-shaped plates. The fermentation started with inoculating 1 ml 3× SC (2% glucose) medium supplemented with 3 mM tryptophan to an initial OD₆₀₀ of 0.5. Continuous exponential feeding with the same medium except containing 36% glucose started at 20 h using the following equation:

where t is the feeding time (h).

Temperature was kept at 30 °C throughout the whole cultivation. The pH was controlled at 5.5 using 10% NH₄OH solution. The plate was shaken at speed 1,000 r.p.m. The relative humidity in the growth chamber was maintained at 85% using distilled water to minimize evaporation of the media. An automated liquid handler was used to take 0.1 ml culture samples on days 3, 5 and 7 for metabolite analysis.

Batch-fed fermentations were performed on an ambr 250 system using single-use microbial vessels. The fermentation started by inoculating 100 ml of 3× SC medium (2% glucose) supplemented with 3 mM tryptophan to an initial OD₆₀₀ of 5.0. The same medium, except containing 36% glucose was fed continuously into the bioreactor between 20 and 96 h using the following equation:

Between days 4 and 10, 3 g l⁻¹ of galactose was added into the fermentation every day to induce the expression of catharanthine synthase. The temperature was kept at 30 °C throughout the whole fermentation. The pH of all bioreactors was controlled at 5.0 using 10% NH₄OH solution. The air flow rate was initially at 1 volume of air per unit of medium per unit of time (vvm), and dissolved oxygen (dO₂) was maintained above 40% by increasing the agitation speed. CO₂ in the exhaust gas was monitored by a gas analyser. An automated liquid handler was used to take 1.0 ml broth samples every 24 h for metabolite analysis. At the end of the cultivation, all (150–200 ml) broth was used for crude extraction of MIA compounds including catharanthine and vindoline.

MIA compounds extraction and purification

In some cases, MIA metabolites were extracted from fermentation broths using the Oasis HLB (hydrophilic–lipophilic balance) cartridges and 96-well plates following the manufacturer’s manual (Waters). Solid phase extraction (SPE) of all samples (0.5–2 ml) from 96- or 24-well deep plates were performed on Oasis HLB 96-Well Plates (SKU 186000679) and eluted with 0.5 ml of pure methanol. Samples taken from the ambr 250 bioreactors were extracted using dichloromethane and the resulting crude extract was further purified by preparative thin-layer chromatography followed by preparative HPLC (Supplementary Methods). The final samples were dissolved with appropriate amounts of water and used for chemical coupling to produce vinblastine.

Synthesis of vinblastine from vindoline and catharanthine

Photo-chemical coupling of vindoline and catharanthine was performed in 0.1 M Tris-HCl buffer pH 7.0 containing 1 mM MnCl₂ (ref. ³⁸). The reaction was initiated by adding 100 μM catharanthine, 10 μM vindoline and 50 μM flavin mononucleotide in a 96-well plate. The reaction (200 μl) was then placed under UV-A light (peak wavelength at 365 nm) for 5 min. After the exposure, 20 μl of 100 mM NADH solution was added (9.1 mM final concentration) and incubated at 0 °C for 5 h. Chemical coupling of vindoline and catharanthine was performed in 75 mM FeCl₃ and 0.05 M HCl containing 10% (v/v) 2,2,2-trifluoroethanol and incubated at 25 °C for 15 min, followed by a reduction using 10 mM NaBH₄ at 0 °C for 4 h (ref. ²). All assay samples were kept at −20 °C until SPE using an Oasis HLB 96-well plate (Waters) following the manufacturer’s protocol.

Metabolites sample preparation and analysis

After 3 or 6 days, 200 μl of cultures were mixed with 20 μl of 10 mg l⁻¹ caffeine solution (as the internal standard for normalization) and filtered through a filter plate (PALL, AcroPrep Advance, 0.2 μm Supor membrane for media/water) by centrifugation at 2,200g for 1 min. For tabersonine and vindoline analysis, strains MIA-CR-A and MIA-CW-1 were also cultivated in 4 ml of YPD medium on a 24-well deep-well plate for 6 days, and the whole spent medium was used for SPE using Oasis HLB 96-well plate (Waters) following manufacturer’s protocol. Metabolites were eluted with 0.8 ml of pure methanol, evaporated to dryness at room temperature and reconstituted in 100 µl of water and mixed with 10 μl of 10 mg l⁻¹ caffeine solution before filtration. A series of standard mixtures (Supplementary Table 5) were prepared in the same way as analytical samples. Typically, 1 μl samples were injected for analysis. For de novo tabersonine, catharanthine and vindoline strains, 3 μl samples were injected to produce a higher signal. Targeted metabolite analysis of loganic acid, loganin, secologanin, strictosidine, tabersonine, catharanthine, vindoline and vinblastine was conducted using an Advance ultra-HPLC (UHPLC) system (Bruker Daltonics) coupled to an EVOQ Elite triple quadrupole mass spectrometer (Supplementary Table 6) (Bruker Daltonics). Accurate masses of some analytes (Supplementary Methods) were analysed on a Dionex UltiMate 3000 UHPLC (Fisher Scientific) connected to an Orbitrap Fusion Mass Spectrometer (Supplementary Methods) (Thermo Fisher Scientific).

For the analysis of non-polar metabolites (geraniol and 8-hydroxygeraniol, and by-product citronellol), cells were grown in 2 ml of medium using 24-well deep plates. Geraniol and 8-hydroxygeraniol were extracted by mixing 1.6 ml samples with 400 µl of ethyl acetate (containing 50 µg ml⁻¹ nerol as the internal standard), shaken at 4 °C for 30 min. The aqueous and organic phases were separated by centrifugation at 13,000g for 1 min, and 250 µl of the upper (ethyl acetate) phase was transferred to a gas chromatography vial with glass insert and analysed on a gas chromatography–flame ionization detector. Details for the gas chromatography and LC–MS programmes are described in the Supplementary Methods.

Fluorescence microscopy

The yEGFP was fused to the N or C terminus of target proteins depending on their reported or predicted structural information. All yEGFP proteins were integrated into the wild-type yeast strains and confirmed with colony PCR. Localization markers for nucleus (Addgene plasmid no. 133648), endoplasmic reticulum membrane (Addgene plasmid no. 133647), mitochondria (Addgene plasmid no. 133655) and vacuole (Addgene plasmid no. 133654), containing mCherry or DuDre, were ordered from Addgene and integrated into the strains with yEGFP fusion protein as described previously⁴⁷. Strains were cultivated in SC-Trp medium overnight (16–20 h) and imaged in the morning, or overnight culture was diluted 1:5 and cultivated for 4 h before analysis. Then 3 µl of culture was applied to an objective glass, covered with a cover glass and imaged immediately. Confocal images were acquired with a laser-scanning upright confocal microscope LSM 700, Axio Imager 2 (Carl Zeiss, Inc.), equipped with four diode lasers (405, 488, 555, 639 nm), using a C-Apochromat ×63/1.2 W Korr M27 water objective and ×4 scan zoom. Pinhole size was set to 1 Airy unit. Samples were illuminated with 488 and 555 nm lasers at 0.4–4 mW. 512 × 512 pixel images were acquired using photomultiplier tube detectors. yeGFP acquisition parameters were: 300–578 nm (emission wavelength range), 0.39 µs (line time), 600–1,000 (gain) and 0 (offset). mCherry/DuDre acquisition parameters were: 578–800 nm (emission wavelength range), 0.39 µs (line time), 600–1.000 (gain) and 0 (offset). A line average of eight was applied to both channels. Images were processed with ZEN v.3.2 (blue edition) (Carl Zeiss Microscopy GmbH).

Analysis of proteins

All yeast strains were grown in 2 ml of SC or YPD media in three biological replicates at 30 °C 300 r.p.m. for 2 days. Cells (roughly 50 OD × ml units) were pelleted by centrifugation at 5,000g for 3 min and kept at −80 °C until lysis and sample preparation. Yeast cells were lysed and protein extracted as described previously⁴⁸. Briefly, it consists of a bead-beating step, cell lysis, protein precipitation, protein resuspension, protein quantification and normalization of protein concentration followed by standard bottom-up proteomic procedures of reducing and blocking cysteine residues and tryptic digestion. Peptide samples were analysed on an Agilent 1290 UHPLC system coupled to an Agilent 6460QQQ MS (Agilent Technologies) as described previously⁴⁹. Briefly, peptide samples were loaded onto an Ascentis ES-C18 Column (Sigma–Aldrich) and introduced to the MS using a Jet Stream source (Agilent Technologies) operating in positive-ion mode. The data were acquired with Agilent MassHunter Workstation Software v.B.08.02. LC–MS Data Acquisition operating in dynamicMRM mode. Multiple reaction monitoring (MRM) transitions for the targeted proteins were generated by Skyline software v.20.2 (MacCoss Laboratory Software) and selection criteria excluded peptides with Met/Cys residues, tryptic peptides followed by further cut sites (KK/RR) and peptides with proline adjacent to K/R cut sites. The data and Skyline methods are available on Panoramaweb (https://panoramaweb.org/microbial-synthesis-of-vinblastine.url). Proteins from the MIA-DJ (tabersonine-vinblastine double module) strain were analysed using the Cap‐LC system equipped with a C18 easy spray column (Thermo Fisher Scientific), coupled to Orbitrap Q Exactive HF‐X mass spectrometer (Thermo Fisher Scientific)⁵⁰. The resulting data were analysed using Proteome Discover v.2.3 (Thermo Fisher Scientific) by searching against the S. cerevisiae proteome data (Uniprot ID UP000002311) combined with all heterologous protein sequences. The abundance of each protein is reported as the relative intensity to the total intensity of all identified peptides.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.