Efficient production of vindoline from tabersonine by metabolically engineered Saccharomyces cerevisiae

Vindoline is a plant derived monoterpene indole alkaloid (MIA) with potential therapeutic applications and more importantly serves as the precursor to vinblastine and vincristine. To obtain a yeast strain for high yield production of vindoline from tabersonine, multiple metabolic engineering strategies were employed via the CRISPR/Cas9 mediated multiplex genome integration technology in the present study. Through increasing and tuning the copy numbers of the pathway genes, pairing cytochrome P450 enzymes (CYPs) with appropriate cytochrome P450 reductases (CPRs), engineering the microenvironment for functional expression of CYPs, enhancing cofactor supply, and optimizing fermentation conditions, the production of vindoline was increased to a final titer as high as ∼16.5 mg/L, which is more than 3,800,000-fold higher than the parent strain and the highest tabersonine to vindoline conversion yield ever reported. This work represents a key step of the engineering efforts to establish de novo biosynthetic pathways for vindoline, vinblastine, and vincristine.

V indoline is a monoterpene indole alkaloid (MIA) derived from Catharanthus roseus, one of the most extensively investigated medicinal plants. Vindoline serves as both the biosynthetic and synthetic precursor of the therapeutically important anticancer drugs, vinblastine, and vincristine 1,2 . However, the low yield of these MIAs in plant as well as the difficulty in chemical synthesis result in high market demands with exorbitant price 3 . To overcome these drawbacks, several studies have shown that metabolically engineered yeasts are promising cell factories for the production of plant-derived alkaloids, such as sanguinarine 4 , strictosidine 5 , noscapine 6 , tropine 7 , and scopolamine 8 . For example, a combination of enzyme engineering, pathway and strain engineering, and fermentation optimization led to the production of noscapine at a titer of ∼2.2 mg/L in shake flasks 6 . In another study, enhancing the supply of precursors, removing competing pathways, and increasing the copy numbers of rate-limiting enzyme encoding genes were employed to achieve de novo production of strictosidine, with a final titer as high as ∼0.5 mg/L 5 .
In the present study, Saccharomyces cerevisiae was employed as a host for the reconstitution of the biosynthetic pathway of vindoline from tabersonine. Multiple metabolic engineering efforts were made to achieve high-yield biosynthesis of vindoline. Initially, the reconstitution of the whole biosynthetic pathway (T16H2, 16OMT, T3O, T3R, NMT, D4H, DAT, CPR, and CYB5 from C. roseus) with one chromosomal copy for each gene resulted in marginal production of vindoline in yeast. Followed by increasing the expression levels of the rate-limiting enzyme encoding genes, pairing CYPs with appropriate CPRs, modifying the microenvironment for CYPs functional expression, enhancing cofactor supply, and optimizing fermentation conditions (Fig. 1b), the production of vindoline was increased for more than~3,800,000fold, with a final titer as high as ∼16.5 mg/L. The vindolineproducing yeast strain constructed in this study not only serves as a classic example of metabolic engineering for producing valueadded secondary metabolites (i.e., plant natural products), but also takes a critical step in establishing de novo biosynthetic pathways to produce vindoline, vinblastine, and vincristine.
Improving vindoline production via biosynthetic pathway optimization. Considering that higher copy numbers of T16H2 and 16OMT were required for vindoline production and the plasmid system was not suitable for practical applications, multiplex and multi-copy genome integration of the biosynthetic pathway genes was employed using the CRISPR/Cas9-mediated genome editing tool ( Supplementary Fig. S1). As shown in Fig. 3, the integration of another copy of T16H2 and 16OMT expression cassettes into strain VSY006 increased the production of vindoline to 130.3 μg/L (VSY008), which was also significantly higher than that of the plasmid system (strain VSY007), indicating the advantage of genome integration for cell factory development. Further integration of another copy of the other pathway genes into VSY008 resulted in the construction of VSY009, which harbored two copies for each gene of the vindoline biosynthesis pathway. The production of vindoline in strain VSY009 was increased~1.71-fold to 221.9 μg/L, when compared with strain VSY008. However, vindorosine was still accumulated to high levels, indicating that T16H2 and 16OMT remained the bottlenecks for high-yield conversion of tabersonine to vindoline.
To further enhance 16-hydroxytabersonine and 16methyoxytabersonine synthesis and accordingly direct the metabolic flow toward vindoline production, additional copies of T16H2 and 16OMT were further integrated into VSY009, resulting in the construction of VSY014 (three copies of T16H2 and 16OMT), VSY015 (four copies of T16H2 and 16OMT), and VSY016 (five copies of T16H2 and 16OMT). VSY015 harboring four copies of T16H2 and 16OMT and two copies of the remaining pathway genes resulted in the highest production of vindoline, which was 3.2-and 1.88-fold higher than that of VSY008 and VSY009, respectively. Integration of extra copies of T16H2 and 16OMT failed to further increase the conversion of tabersonine into vindoline (Fig. 3). Therefore, the biosynthetic pathway with four copies of T16H2 and 16OMT and two copies of T3O, T3R, NMT, D4H, and DAT were optimal for vindoline conversion, and VSY015 was chosen for subsequent optimization.
Engineering CYPs microenvironment and increasing cosubstrate availability for increased vindoline production. As mentioned above, two plant CYPs (T16H2 and T3O) were involved in vindoline biosynthesis. Considering the pivotal role of CPRs in NADPH-mediated electron transfer and the requirement of endoplasmic reticulum (ER) membrane localization for CYPs activities, the following strategies were employed to improve the microenvironment of CYPs for optimal performance: (1) CYP-CPR pairing; (2) ER expansion; (3) NADPH supply enhancement.
Previous studies demonstrated that the source of CPRs could affect the activities of CYPs and the native CPRs were not necessarily the best partner for CYPs in a heterologous host [18][19][20] . Therefore, CYPs should be appropriately paired with CPRs for optimal function and accordingly vindoline biosynthesis. In yeast strain VSY015, the native CPR from C. roseus (CrCPR) was  integrated. To investigate the effect of CYPs-CPRs pairing on vindoline biosynthesis, four different CPR-encoding genes including AtCPR1 (Arabidopsis thaliana), GuCPR1 (Glycyrrhiza uralensis), GlCPR (Ganoderma lucidum), and MTR2 (Medicago) were chosen to replace CrCPR in VSY015, resulting in the construction of VSY017, VSY018, VSY019, and VSY020, respectively. The replacement of CrCPR with AtCPR1 resulted in significant improvements in vindoline production (∼1.8-fold, 662.4 μg/L), while a comparable titer was observed for all the remaining CPR replacement (Fig. 4). Interestingly, the integration of an additional copy of CrCPR into VSY015 and VSY017 resulted in decreased vindoline production, 253.4 μg/L in VSY021 and 455.8 μg/L in VSY022, respectively.
As AtCPR1 demonstrated the best compatibility with the two C. roseus CYPs (T16H2 and T3O), an additional copy of AtCPR1 was integrated into VSY017. Unfortunately, the production of vindoline was decreased in VSY017-2 ( Supplementary Fig. S2), indicating the benefits of overexpression of AtCPR1 was counteracted by the endogenous CPR (ScCPR), encoded by NCP1 21 . ScCPR has been found to possess low compatibility with plant CYPs and might interfere with the electron transfer between AtCPR1 and C. roseus CYPs (T16H2 and T3O). Therefore, to further explore the CYP-CPR interaction for optimal catalytic activity of CYPs, NCP1 was deleted and the effect on vindoline production was investigated in the present study. VSY017 was further engineered by inserting CPRs and CrCYB5 expression cassettes into the NCP1 locus, resulting in the construction of VSY017-3, VSY017-4, VSY017-5, VSY017-6, and VSY017-7, respectively. As expected, the integration of an additional copy of CPRs into the NCP1 locus of VSY017 increased vindoline production, with the highest production achieved in VSY017-3 (1264.2 μg/L), representing a two-fold increase over the parental strain VSY017 (Supplementary Fig. S2). In other words, the deletion of NCP1 in the yeast strain harboring two copies of AtCPR1-CrCYB5 increased vindoline production by about 2.6fold. Based on these results, two genomic copies of AtCPR1 together with NCP1 deletion contributed to the best electron transfer compatibility with CYPs of the vindoline biosynthetic pathway.
As plant CYPs and CPRs are generally ER-localized membrane proteins 22 , the expansion of ER should enable higher enzymatic activities by providing more space for the folding of CYPs and CPRs 23 . Previous investigations have demonstrated that ER expansion could be achieved by the deletion of PAH1 23 , the deletion of OPI1, and the overexpression of INO2 24 . Although attempted multiple times, the deletion of PAH1 was not successful in the vindoline-producing strain. The discrepancy with the previous study might result from the use of yeast strains with different genetic background. Therefore, the overexpression of INO2 and the deletion of OPI1 were employed in the present study. In addition, NADPH functions as the cofactor for CPRs mediated electron transfer and the increased NADPH availability has been shown to enhance CYPs activities 25 . The introduction of GAPN 26 (NADP + -dependent glyceraldehyde-3-phosphate dehydrogenase from Streptococcus mutans) and overexpression of ZWF1 were commonly employed strategies to enhance NADPH supply. Enhancing the supply of S-Adenosyl-Methionine (SAM) via the overexpression of SAM2 has been demonstrated to improve the methyltransferase activity and accordingly the biosynthesis of secondary metabolites in yeast as well 5 . Therefore, VSY017 was further engineered by inserting the GAPN and ZWF1 expression cassettes into the OPI1 locus (VSY023), followed by the integration of SAM2 and INO2 expression cassettes (VSY024). The production of vindoline in VSY023 and VSY024 was increased to 1339.0 μg/L and 1662.7 μg/L, respectively (Fig. 5). Overall, the production of vindoline was increased for about fourfold (from~417.1 μg/L in VSY015 to~1662.7 μg/L in VSY024) via engineering of the CYP microenvironment, including CYPs-CPRs pairing, ER expansion, and NADPH supply enhancement.
Optimization of fermentation conditions for the biosynthesis of vindoline. To further increase vindoline production, the fermentation conditions were briefly optimized in shaker flasks. First, SC and YP medium were investigated with different concentrations of tabersonine (10, 25, 50, 75, 100, 125 mg/L). Generally, more vindoline could be produced if a higher concentration of tabersonine was supplemented. As shown in Fig. 6a, compared with the synthetic medium, the rich medium YP led to a pronounced increase in vindoline titers (as high as 5.8 mg/L), particularly under low tabersonine concentration. The better performance might result from improved enzyme expression levels and better cell growth in the rich medium. Nevertheless, in the YP medium, the conversion yield of   5 Optimization of vindoline production through ER expansion and cofactor supply enhancement (NADPH and SAM). With VSY017 as the parent strain, OPI1 was deleted and INO2 was overexpressed for ER expansion, ZWF1 and GAPN were overexpressed to enhance NADPH supply, and SAM2 was overexpressed to increase SAM availability. All strains were cultured in SC with 2% galactose in the presence of 50 mg/L tabersonine. Error bars represent SD of biological triplicates (n = 3).
To explore the possible reasons, the accumulation of the intermediate metabolites was analyzed when different concentrations of tabersonine were supplemented into YP medium. As shown in Fig. 6b, the percentage of 16-OH-tabersonine (light blue box) and 3-OH-16-MOH-2,3-2H-tabersonine (light green box) were increased with higher substrate supplementation, indicating that vindoline production was limited by the methyltransferases (16OMT and NMT). In addition, the percentage of 3-OH-2,3-2H-tabersonine (light orange box), an intermediate metabolite of the vindorosine pathway, was increased with higher substrate supplementation, implying that the promiscuity of T3O/T3R was another reason for low yield vindoline biosynthesis.
To address the dilemma between titer and yield, a low concentration of tabersonine (15 mg/L) was fed into the fermentation broth every 24 h after galactose induction in VSY024 and VSY025. With a similar amount of tabersonine supplemented (roughly 100 mg/L), the intermittent supply of the substrate at a lower concentration resulted in improved both titer and yield of vindoline, whose maximal titer reached 11.7 mg/L and 16.5 mg/L, respectively (Fig. 7a). In addition, by keeping tabersonine at a low concentration, the accumulation of vindorosine and its biosynthetic intermediates were largely decreased, while the vindoline biosynthetic intermediates (such as 3-OH-16-MOH-2,3-2H-tabersonine and desacetoxyvindoline) were still accumulated to high levels (Fig. 7b). In other words, the concerns with the promiscuity of T3O/T3R could be addressed to some extent by lowering tabersonine concentration, while the biosynthetic pathway should be further optimized to minimize intermediate accumulation.
Finally, to compare the performance of VSY025 with the previously constructed vindoline producing strain by Qu et al. 17 , which harbored the biosynthetic pathway genes on multi-copy plasmids, yeast biotransformation assay was carried out under the same conditions. With the supplementation of~75 mg/L, while the strain constructed by Qu et al. could produce~1.1 mg/L vindoline, VSY025 was able to produce 17.7 mg/L and 29.4 mg/L vindoline in SC medium and YP medium, respectively (Supplementary Fig. S3), indicating the advantages of the genomeintegrated strains in the biosynthesis of natural products. More importantly, tabersonine biotransformation using high concentration of yeast cells not only decreased by-product formation, but also minimized intermediate accumulation, particularly in YP medium (Fig. 7b).

Discussion
Vindoline is suggested to be one of a major alkaloid compound in C. roseus and has been demonstrated to possess antidiabetic activities 27 . More importantly, vindoline is condensed with catharanthine to generate the anticancer drugs vinblastine and vincristine. In the present study, the vindoline biosynthetic pathway from tabersonine was reconstituted and systematically optimized in yeast using the CRISPR/Cas9-mediated multiplex genome editing technology.
CYPs are generally considered as the bottleneck for efficient production of secondary metabolites in microbial cell factories. Vindoline production was significantly increased by improving the expression and activities of eukaryotic P450s in yeast (Fig. 4). Eukaryotic CYPs are a kind of ER located membrane proteins, thus the expansion of ER has been demonstrated to be beneficial to improve the functional expression and activities of CYPs. Based on prior studies, the knockout of OPI1 in conjunction with the overexpression of GAPN, ZWF1, and INO2 enhanced the vindoline titer to as high as 1662.7 μg/L in strain VSY024 (Fig. 5). Notably, the effect of PAH1 deletion on P450 activities has been demonstrated to enhance the production of medicagenic acid and sesquiterpenoid artemisinic acid in S. cerevisiae CEN.PK2. Unfortunately, the knockout of PAH1 was attempted several times without success in the yeast strain used in the present study (BY4741). The discrepancy might result from the differences in the genetic background of these two yeast strains, indicating our limited understanding of the yeast network regulating ER expansion. Considering the complexity of the ER expansion and membrane protein folding mechanisms, very limited targets (i.e. OPI1, INO2, and PAH1) have been identified and employed for metabolic engineering applications. Therefore, genome-scale engineering, such as RNAi-assisted genome evolution 28 and multi-functional genome-wide CRISPR system (MAGIC) 29 , should be performed to identify novel metabolic engineering targets to enhance the functional expression of CYPs.
Although systematic metabolic engineering was carried out to improve the pathway efficiency, the vindoline conversion yield was still not satisfactory. One concern is the accumulation of intermediate metabolites. In particular, 16-OH-tabersonine and 3-OH-16-MOH-2,3-2H-tabersonine were accumulated to high levels when high concentration of tabersonine was supplemented (Fig. 6b), indicating that vindoline production was limited by the methyltransferases (16OMT and NMT). Thus, the enzymatic activities of 16OMT and NMT and/or the supply of co-factors for methyltransferases (such as SAM) should be further enhanced. Recently, the activity of methyltransferases 30 was found to be affected by pH of the culture medium, indicating that fermentation conditions should be optimized to further increase the titer and yield of vindoline. Another concern is the formation of vindorosine as a by-product with higher substrate supplementation, which was resulted from the promiscuity of T3O for the conversion of tabersonine directly. The dilemma of titer and yield to some degree could be addressed by increasing the expression levels of T16H2 and 16OMT in combination of supplying low concentration of tabersonine in an intermittent manner. Nevertheless, a more promising and practical strategy is to engineer the substrate specificity of T3O. Specifically, T3O variants with enhanced activity on 16-methoxyl-tabersonine and minimized activity toward tabersonine are highly demanded for efficient and high-yield production of vindoline. For example, Sun et al. used homology modeling and molecular docking to enhance the activity and substrate specificity of CYP72A63, leading to a highlevel accumulation of 11-oxo-β-amyrin and glycyrrhetinic acid in S. cerevisiae 31 . Due to the lack of structural information for most eukaryotic CYPs, directed evolution based on random mutagenesis and high throughput screening can greatly broaden the research and application of CYPs engineering 31,32 .
Herein, multiple metabolic engineering strategies based on CRISPR/Cas9-mediated facile genome engineering technology were employed to systematically engineer S. cerevisiae to produce vindoline from tabersonine. A combination of pathway optimization, CYPs-CPRs pairing, ER expansion, cofactor supply enhancement, and fermentation optimization led to the construction of a yeast cell factory to produce vindoline from tabersonine with a final titer of 16.5 mg/L, which was more than 3,800,000-fold higher than the parent strain. Therefore, the multiple metabolic engineering strategies can not only be employed to facilitate the production of other value-added secondary metabolites, but also lay a solid foundation for the establishment of de novo biosynthetic pathways for producing vindoline, and even vinblastine and vincristine.

Methods
Strains, media, and reagents. Escherichia coli Trans T1 (Transgen Biotech Co., Ltd, Beijing, China) was used for gene cloning and plasmid amplification. E. coli transformants were selected and cultured in LB medium containing 100 mg/L ampicillin. S. cerevisiae BY4741 strain was used as the host for genome engineering and vindoline production. Yeast strains were routinely cultivated in SCD (5 g/L ammonium sulfate, 1.7 g/L yeast nitrogen base without ammonium and amino acids, 0.6 g/L CSM missing the appropriate nutrients, 20 g/L glucose, and 0.2 g/L methionine) or YPD (10 g/L yeast extract, 20 g/L peptone, and 20 g/L glucose) medium. When necessary, 200 mg/L G418 sulfate (Sangon Bio-tech Co., Ltd, Shanghai, China) was supplemented. All restriction enzymes and T4 DNA ligase were purchased from NEB (Beijing, China). All chemicals were from Sigma-Aldrich (St. Louis, MO, USA) unless specifically mentioned. Tabersonine and vindoline standards were purchased from Chengdu DeSiTe Biological Technology Co., Ltd (Chengdu, China) and Yuanye Bio-tech Co., Ltd (Shanghai, China), respectively.
Plasmid construction. Vindoline biosynthesis pathway genes including CrT16H2, Cr16OMT, CrT3O, CrT3R, CrNMT, CrD4H, CrDAT, CrCYB5, and CrCPR were kindly provided by Prof. Vincenzo De Luca (Brock University) 17 , AtCPR1 from A. thaliana, GlCPR from G. lucidum, and MTR2 from Medicago were kindly provided by Dr. Han Xiao 33 (Shanghai Jiao Tong University) and GuCPR1 from G. uralensis was provided by Prof. Chun Li 18 (TsingHua University). The coding sequences of all the genes used in this study were listed in Supplementary Data 2 and cloned into the multiple cloning sites of the pESC series vectors (pESC-URA, pESC-LEU, pESC-LEU2d, and pESC-HIS) by digestion/ligation (MCS1: BamHI/XhoI; MCS2: NotI) or Gibson Assembly. Oligonucleotides were synthesized by Tsingke Biotech Co., Ltd (Hangzhou, China). KOD-Plus-Neo DNA Polymerase (TOYOBO Biotech Co., Ltd, Tokyo, Japan) was used for gene amplification and PCR products were purified by the Gene JET PCR Purification Kit (Thermo fisher Scientific, Shanghai, China). Plasmids were extracted from E. coli using the AxyPrep Plasmid Miniprep Kit (Axygen, Shanghai, China) according to the manufacturer's instructions. DNA sequencing was performed by Tsingke Biotech Co., Ltd (Hangzhou, China). All the plasmids used in this study were listed in Table 1 and the corresponding primers were provided in Supplementary Table S1.
Strain construction. All genetic modifications in yeast were carried out via the CRISPR/Cas9-mediated genome editing method 34,35 and the schematic overview was briefly demonstrated in Supplementary Fig. S1. Genome integration loci and the corresponding gRNA plasmids (Supplementary Table S2) were constructed in our previous studies 29, [36][37][38] . For the construction of strains VSY001-VSY024, target gene expression cassettes were amplified by PCR with 40 bp homology arms to the target chromosomal loci and co-transformed with the corresponding gRNA plasmids to the Cas9 expressing yeast strains using the LiAc/SS carrier DNA/PEG method 39 . All the strains constructed in this study were listed in Table1.
Fermentation conditions. For vindoline and other intermediates quantification, a single colony was picked from YPD or SCD agar plates and subcultured at 30°C and 250 rpm for 24 h. Then, 100 μL seed culture was inoculated into a 250 mL flask containing 10 mL fresh YPD or SCD medium. After culturing at 30°C for 24 h, the yeast cells were harvested and resuspended in SC or YP medium supplemented with 2 % galactose (SCG or YPG) and 10-150 mg/L tabersonine. The induced yeast cells were cultivated in a 250-mL shake flask with 10 mL volume at 30°C for additional 168 h. Intermittent feeding of tabersonine was performed for the best performing S. cerevisiae strain VSY024, following a similar protocol, except that 15 mg/L tabersonine was supplemented every 24 h after induction by galactose. All the experiments were performed in biological triplicates.
Yeast biotransformation assays. A single colony of VSY025 was inoculated in 2 mL SCD or YPD medium and cultured at 30°C overnight. Yeast cells were harvested and inoculated into fresh 2 mL SCG or YPG medium and induced at 30°C for 24 h. The induced yeast cells were harvested and suspended in 1 mL biotransformation buffer (10 mM Tris-HCl at pH 7.5 and 1 mM EDTA) containing 20 μM tabersonine and incubated at 30°C for 12 h. After the biotransformation assays, the production of vindoline and other intermediates were quantified by LC-MS.
Analytical methods. The culture broth was centrifuged at 11,000 × g for 10 min. For the strains with low vindoline production (VSY006-VSY009), 20 mL supernatant was freeze-dried and then reconstituted in 1 mL water. The resultant mixture was extracted by 1 mL ethyl acetate twice. After being dried by a rotary evaporator, the samples were resuspended in 150 μL methanol and passed through a 0.22 µm membrane filter for LC-MS analysis. For the high-producing strains (VSY010-VSY024), 400 μL supernatant was extracted by 1 mL ethyl acetate twice, and the organic phase was passed through a 0.22 µm membrane filter and subject to LC-MS analysis.
Statistics and reproducibility. General data analysis (means and standard deviation) was performed primarily by OriginPro 2021 V.9.8.0.200. All experiments were performed with biological triplicates and values were expressed as means ± standard errors.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.