Metabolic engineering of yeasts for green and sustainable production of bioactive ginsenosides F2 and 3β,20S-Di-O-Glc-DM.

Both natural ginsenoside F2 and unnatural ginsenoside 3β,20S-Di-O-Glc-DM were reported to exhibit anti-tumor activity. Traditional approaches for producing them rely on direct extraction from Panax ginseng, enzymatic catalysis or chemical synthesis, all of which result in low yield and high cost. Metabolic engineering of microbes has been recognized as a green and sustainable biotechnology to produce natural and unnatural products. Hence we engineered the complete biosynthetic pathways of F2 and 3β,20S-Di-O-Glc-DM in Saccharomyces cerevisiae via the CRISPR/Cas9 system. The titers of F2 and 3β,20S-Di-O-Glc-DM were increased from 1.2 to 21.0 mg/L and from 82.0 to 346.1 mg/L at shake flask level, respectively, by multistep metabolic engineering strategies. Additionally, pharmacological evaluation showed that both F2 and 3β,20S-Di-O-Glc-DM exhibited anti-pancreatic cancer activity and the activity of 3β,20S-Di-O-Glc-DM was even better. Furthermore, the titer of 3β,20S-Di-O-Glc-DM reached 2.6 g/L by fed-batch fermentation in a 3 L bioreactor. To our knowledge, this is the first report on demonstrating the anti-pancreatic cancer activity of F2 and 3β,20S-Di-O-Glc-DM, and achieving their de novo biosynthesis by the engineered yeasts. Our work presents an alternative approach to produce F2 and 3β,20S-Di-O-Glc-DM from renewable biomass, which lays a foundation for drug research and development.

Introduction

Panax ginseng, known as the king of herbs, is famous for its excellent performance in the medicines and dietary supplements. Ginsenosides are the major active components of P. ginseng, which possess a wide range of pharmacological activities, including but not limited to anti-tumor, anti-aging, immunomodulation and neuroprotection,. Ginsenoside F2 exhibits anti-glioma, anti-breast cancer and anti-gastric cancer activities, but it is a trace component whose content is less than 0.01% in ginseng. Unnatural ginsenoside 3β,20S-Di-O-Glc-DM displays cytotoxicity against cancer cell lines, but until now it can only be produced by in vitro enzymatic reaction with a UDP-glycosyltransferase UGT109A1 from Bacillus subtilis. Traditional methods for producing ginsenosides usually rely on direct extraction from P. ginseng, chemical synthesis and in vitro enzymatic catalysis. However, the scarcity of natural resources, long cultivation period (5‒7 years) required for the growth of qualified roots, the continuous cropping obstacle of P. ginseng, low content and complicated extraction-separation processes of ginsenosides limit their direct extraction from P. ginseng. Although ginsenosides can be chemically synthesized, their complex structures and especial difficulty in stereoselective/stereospecific glycosylation make their chemical synthesis commercially infeasible. Furthermore, the approach of in vitro enzymatic catalysis still requires expensive substrates to produce target ginsenosides. Therefore, an economical approach for the large-scale production of F2 and 3β,20S-Di-O-Glc-DM is highly expected.

To date, the biosynthetic pathways of many medicinal natural products have been introduced into Saccharomyces cerevisiae, and the resultant engineered strains have been used as yeast cell factories for producing these products,. This approach can not only ensure the continuous acquisition of natural products from renewable biomass, but also protect natural resources and environment. Hence it has been acknowledged as a green and sustainable method for cost-effective production of natural products. Many natural medicines or their precursors have been produced with high yields in microbial cell factories, such as artemisinic acid, miltiradiene, taxadiene and so on10, 11, 12, 13, 14, 15, 16, 17. Especially, the identification of genes involved in the biosynthetic pathways of ginsenosides, along with the rapid progress in synthetic biology tools, has contributed to the construction of a growing number of yeast cell factories for producing ginsenosides. Compound K is the first ginsenoside biosynthesized in the engineered yeast. The titer of compound K reached 5.0 g/L in a 5 L bioreactor by engineering yeast subcellular compartments. The highest compound K titer of 5.74 g/L in a 1.3 L bioreactor was achieved by optimizing UGTPg1 expression and improving UDP-glucose (UDPG) supply. A yeast cell factory producing 2.25 g/L ginsenoside Rh2 in a 10 L bioreactor was built by improving the C3‒OH glycosylation efficiency and optimizing cytochrome P450 (CYP450) expression level. The yeast strain producing ginsenoside Rg1 was constructed and further optimized by systematic engineering strategies and the titer of Rg1 reached 1.95 g/L in a 1.3 L bioreactor.

The construction of microbial cell factories can provide an alternative approach for the mass production of F2 and 3β,20S-Di-O-Glc-DM, but it has not been reported until now. In this study, we first constructed the biosynthetic pathways of F2 and 3β,20S-Di-O-Glc-DM in S. cerevisiae. Subsequently, we applied multistep metabolic engineering strategies to increase the titers of F2 and 3β,20S-Di-O-Glc-DM, including: (1) tuning the promoters of protopanaxadiol synthase (PPDS) and Arabidopsis thaliana NADPH-cytochrome P450 reductase 2 (ATR2) to make more carbon flux flow to protopanaxadiol (PPD), (2) screening high-efficiency C3-glycosyltransferase, (3) enhancing the supply of glycosyl receptors and overexpressing the transcriptional activators, (4) overexpressing the key genes involved in the glycosyl donor UDPG biosynthesis. These efforts led to the titers of 21.0 mg/L F2 and 346.1 mg/L 3β,20S-Di-O-Glc-DM at shake flask level. We further demonstrated that both F2 and 3β,20S-Di-O-Glc-DM exhibited anti-pancreatic cancer activity and the activity of 3β,20S-Di-O-Glc-DM was even better. Finally, the titer of 3β,20S-Di-O-Glc-DM reached 2.6 g/L by fed-batch fermentation in a 3 L bioreactor. This work provides a green and sustainable approach for producing F2 and 3β,20S-Di-O-Glc-DM from renewable biomass.

Materials and methods

Chemicals

Standard dammarenediol-II (DM), PPD, ginsenosides F2, compound K and Rh2 were purchased from Sigma‒Aldrich Co. Ltd. (USA). 3β,20S-Di-O-Glc-DM, 3β-O-Glc-DM and 20S-O-Glc-DM were obtained by in vitro enzymatic reactions of DM catalyzed by UGT109A1, PgUGT74AE2 and UGTPg1, respectively,.

Construction of plasmids

The gene expression cassettes of DS-GFP, PgUGT74AE2, UGTPg1, tHMG1, IDI1, ERG20, ERG9, ERG1, ERG7- and HAC1 were constructed in our previous work. The genes (INO2, PGM1, PGM2, UGP1), the promoters (ENO2, TEF2) and the terminators (FBA1, IDP1) from S. cerevisiae were all synthesized by GenScript (Nanjing, China). The genes M7 (mutant of UGT74AC1) from Siraitia grosvenorii, UGTPn50-HV (mutant of UGTPn50) from Panax notoginseng, M7-1 (mutant of UGT51) from S. cerevisiae and UGT109A1 from B. subtilis were all codon-optimized and synthesized. The ergosteryl-β-glucosidase (EGH1) gene was amplified from the genomic DNA of S. cerevisiae YPH499. Construction of integration modules were performed as our previous methods. The plasmids harboring the integration modules are listed in Supporting Information Table S1.

In vitro UDP-glycosyltransferase activity assays

Heterologous expression of PgUGT74AE2 and UGTPg1 in Escherichia coli BL21 (DE3), enzymatic assays and identification of the target products were carried out as described previously.

Construction and optimization of the engineered yeasts

To construct ginsenoside F2-producing strain, the gene expression cassettes of tHMG1, DS-GFP, PPDS, ATR2, PgUGT74AE2, UGTPg1 and HIS (Supporting Information Fig. S1) together with plasmid p-Cas9-gRNA were co-transformed into the δ1 site of Y-ΔHXK2. A total of 10 single colonies were randomly picked from SD-HIS-URA plates and cultivated in YPD medium for 3 days to compare their titers of F2. The strain with the highest titer of F2 was named as YF2 (Table 1). Then, the promoters of PPDS and ATR2 were adjusted. In the strain YF2, the expression cassettes of PPDS and ATR2 were P-PPDS-T and P-ATR2-T. They were replaced by the expression cassettes P-PPDS-T and P-ATR2-T, P-PPDS-T and P-ATR2-T to construct the strains YF2-B and YF2-C, respectively (Table 1). YF2, YF2-B and YF2-C were cultivated in triplicate to compare their titers of F2. The strain YF2-C with the highest titer of F2 was selected for further engineering.

Table 1

Strains constructed in this study.

Strain	Description
YF2	PTEF1-DS-GFP-TCYC1, PPGK1-tHMG1-TADH1, PPYK1-PPDS-TADH2, PTDH3-ATR2-TTPI1, PTEF1-UGTPg1-TADH2, PPGK1-UGT74AE2-TCYC1 and HIS marker gene were integrated into the δ1 site of Y-ΔHXK2 with p-Cas9-gRNA
YF2-B	PTEF1-DS-GFP-TCYC1, PPGK1-tHMG1-TADH1, PTDH3-PPDS-TADH2, PPYK1-ATR2-TTPI1, PTEF1-UGTPg1-TADH2, PPGK1-UGT74AE2-TCYC1 and HIS marker gene were integrated into the δ1 site of Y-ΔHXK2 with p-Cas9-gRNA
YF2-C	PTEF1-DS-GFP-TCYC1, PPGK1-tHMG1-TADH1, PTDH3-PPDS-TADH2, PTEF2-ATR2-TTPI1, PTEF1-UGTPg1-TADH2, PPGK1-UGT74AE2-TCYC1 and HIS marker gene were integrated into the δ1 site of Y-ΔHXK2 with p-Cas9-gRNA
YDP	PTEF1-DS-GFP-TCYC1, PPGK1-tHMG1-TADH1, PTDH3-UGTPg1-TADH2, PTEF1-UGT74AE2-TCYC1 and HIS marker gene were integrated into the δ1 site of Y-ΔHXK2 with p-Cas9-gRNA
YDS	PTEF1-DS-GFP-TCYC1, PPGK1-tHMG1-TADH1, PTDH3-UGTPg1-TADH2, PTEF1-M7-TCYC1 and HIS marker gene were integrated into the δ1 site of Y-ΔHXK2 with p-Cas9-gRNA
YDN	PTEF1-DS-GFP-TCYC1, PPGK1-tHMG1-TADH1, PTDH3-UGTPg1-TADH2, PTEF1-UGTPn50-HV-TCYC1 and HIS marker gene were integrated into the δ1 site of Y-ΔHXK2 with p-Cas9-gRNA
YDC	PTEF1-DS-GFP-TCYC1, PPGK1-tHMG1-TADH1, PTDH3-UGTPg1-TADH2, PTEF1-M7-1-TCYC1 and HIS marker gene were integrated into the δ1 site of Y-ΔHXK2 with p-Cas9-gRNA
YDB	PTEF1-DS-GFP-TCYC1, PPGK1-tHMG1-TADH1, PTDH3-UGT109A1-TADH2 and HIS marker gene were integrated into the δ1 site of Y-ΔHXK2 with p-Cas9-gRNA
YFC1	PTDH3-IDI1-TTPI1, PPGK1-ERG20-TADH1, PTEF1-ERG9-TCYC1, PPGK1-ERG1-TADH1, PTEF1-ERG7--TCYC1 and LEU marker gene were integrated into the δ4 site of YF2-C
YFC2	PTDH3-IDI1-TTPI1, PPGK1-ERG20-TADH1, PTEF1-ERG9-TCYC1, PPGK1-ERG1-TADH1, PTEF1-ERG7--TIDP1, PENO2-HAC1-TFBA1 and LEU marker gene were integrated into the δ4 site of YF2-C
YFC3	PTDH3-IDI1-TTPI1, PPGK1-ERG20-TADH1, PTEF1-ERG9-TCYC1, PPGK1-ERG1-TADH1, PTEF1-ERG7--TIDP1, PTEF2-INO2-TTDH2 and LEU marker gene were integrated into the δ4 site of YF2-C
YFC4	PTDH3-IDI1-TTPI1, PPGK1-ERG20-TADH1, PTEF1-ERG9-TCYC1, PPGK1-ERG1-TADH1, PTEF1-ERG7--TIDP1, PENO2-HAC1-TFBA1, PTEF2-INO2-TTDH2 and LEU marker gene were integrated into the δ4 site of YF2-C
YS1	PTDH3-IDI1-TTPI1, PPGK1-ERG20-TADH1, PTEF1-ERG9-TCYC1, PPGK1-ERG1-TADH1, PTEF1-ERG7--TCYC1 and LEU marker gene were integrated into the δ4 site of YDS
YS2	PTDH3-IDI1-TTPI1, PPGK1-ERG20-TADH1, PTEF1-ERG9-TCYC1, PPGK1-ERG1-TADH1, PTEF1-ERG7--TIDP1, PENO2-HAC1-TFBA1 and LEU marker gene were integrated into the δ4 site of YDS
YS3	PTDH3-IDI1-TTPI1, PPGK1-ERG20-TADH1, PTEF1-ERG9-TCYC1, PPGK1-ERG1-TADH1, PTEF1-ERG7--TIDP1, PTEF2-INO2-TTDH2 and LEU marker gene were integrated into the δ4 site of YDS
YS4	PTDH3-IDI1-TTPI1, PPGK1-ERG20-TADH1, PTEF1-ERG9-TCYC1, PPGK1-ERG1-TADH1, PTEF1-ERG7--TIDP1, PENO2-HAC1-TFBA1, PTEF2-INO2-TTDH2 and LEU marker gene were integrated into the δ4 site of YDS
YFR	PTEF1-PGM1-TCYC1, PPGK1-PGM2-TADH1, PTDH3-UGP1-TADH2 and TRP marker gene were integrated into the rDNA site of YFC2
YFE	PTEF1-PGM1-TCYC1, PPGK1-PGM2-TADH1, PTDH3-UGP1-TADH2 and TRP marker gene were integrated into the EGH1 site of YFC2
YSR	PTEF1-PGM1-TCYC1, PPGK1-PGM2-TADH1, PTDH3-UGP1-TADH2 and TRP marker gene were integrated into the rDNA site of YS3
YSE	PTEF1-PGM1-TCYC1, PPGK1-PGM2-TADH1, PTDH3-UGP1-TADH2 and TRP marker gene were integrated into the EGH1 site of YS3

Similarly, the 3β,20S-Di-O-Glc-DM-producing strain was constructed by transforming the expression cassettes of tHMG1, DS-GFP, PgUGT74AE2, UGTPg1 and HIS (Supporting Information Fig. S2) together with plasmid p-Cas9-gRNA into the δ1 site of Y-ΔHXK2. Among 10 single colonies picked from SD-HIS-URA plates, the one with the highest titer of 3β,20S-Di-O-Glc-DM was named as YDP (Table 1). Then PgUGT74AE2 was replaced by other C3-glycosyltransferase genes encoding M7 from S. grosvenorii, UGTPn50-HV from P. notoginseng, M7-1 from S. cerevisiae and UGT109A1 from B. subtilis (Fig. S2). The resultant strains with the highest titer of 3β,20S-Di-O-Glc-DM in each group were named as YDS, YDN, YDC and YDB, respectively (Table 1). The strain YDB did not harbor UGTPg1 gene because UGT109A1 can transfer a glucose moiety to not only C3‒OH but also C20‒OH of DM6. The above four strains and YDP were cultivated in YPD medium for 3 days in triplicate to compare their titers of 3β,20S-Di-O-Glc-DM. The strain YDS with the highest titer of 3β,20S-Di-O-Glc-DM was retained for further engineering.

The integration module containing the upstream pathway genes IDI1, ERG20, ERG9, ERG1, antisense ERG7 and LEU (Supporting Information Fig. S3) was transformed into the δ4 site of YF2-C and YDS, resulting in the strains YFC1 and YS1, respectively. The integration module containing the aforementioned genes and the transcriptional activator gene HAC1 (Fig. S3) was transformed into the δ4 site of YF2-C and YDS, resulting in the strains YFC2 and YS2, respectively. The integration module containing the aforementioned genes and the transcriptional activator gene INO2 (Fig. S3) was transformed into the δ4 site of YF2-C and YDS, resulting in the strains YFC3 and YS3, respectively. The integration module containing the aforementioned genes and the transcriptional activator genes HAC1 and INO2 (Fig. S3) was transformed into the δ4 site of YF2-C and YDS, resulting in the strains YFC4 and YS4, respectively. A total of 10 single colonies were randomly picked from SD-HIS-URA-LEU plates, respectively, and cultivated in YPD medium for 3 days to compare the titers of their respective target products. The strains YFC2 and YS3 with the highest titers of their respective target products were used for further engineering.

The integration module containing the gene expression cassettes of PGM1, PGM2, UGP1 together with TRP (Supporting Information Fig. S4) was transformed into the rDNA site of YFC2 and YS3, resulting in the strains YFR and YSR. Similarly, the same integration module was transformed into the EGH1 site of YFC2 and YS3, resulting in the strains YFE and YSE. A total of 10 single colonies were selected randomly from SD-HIS-URA-LEU-TRP plates, respectively, and cultivated in YPD medium for 3 days to compare the titers of their respective target products.

The strains constructed in this study are listed in Table 1.

Yeast cultivation and fed-batch fermentation

The strains were first inoculated into 10 mL SD medium and grown at 30 °C and 220 rpm for 18 h, and then transferred to 150 mL YPD medium in 500 mL shake flask with an initial OD600 of 0.2 and grown at 30 °C and 220 rpm for 6 days. 5 mL fed solution described previously was fed into the medium at 48, 72, 96 and 120 h, respectively.

Fed-batch fermentation of the strain YSR was conducted as described previously. About 100 mL seed cultures were inoculated into 1 L YPD medium in a 3 L bioreactor. Fermentation was carried out at 30 °C. The pH was controlled at 5.5 by addition of ammonium hydroxide. Dissolved O2 was controlled at approximately 30% by the agitation cascade and air flow rate.

Metabolite extraction

Ginsenosides F2, 3β,20S-Di-O-Glc-DM, 3β-O-Glc-DM, 20S-O-Glc-DM, compound K and Rh2 were extracted from the cell pellets and the supernatant with n-butanol, while aglycones DM and PPD were extracted with n-hexane.

HPLC, LC‒MS and NMR analysis

All the chemicals were analyzed by HPLC (Agilent 1260 Infinity II). Chromatographic separation was performed at 28 °C with a Cosmosil 5C18-MS-II column (5 μm, 4.6 mm × 150 mm). The flow rate was 1 mL/min and the detecting wavelength was 203 nm. The mobile phase consisted of water and acetonitrile (ACN). The separation of ginsenosides (F2, 3β,20S-Di-O-Glc-DM, 3β-O-Glc-DM, 20S-O-Glc-DM, compound K and Rh2) was conducted using gradient procedure as follow: 0‒20 min (20%‒85% ACN), 20‒30 min (85%‒100% ACN), 30‒40 min (100% ACN), 40‒45 min (20% ACN). The separation of ginsenoside aglycones (DM and PPD) was conducted using the isocratic procedure as follow: 0‒30 min (85% ACN). Semi-preparative HPLC and LC‒MS were performed as described previously. NMR spectra were recorded on an INOVA 500 NMR spectrometer (Varian, CA, USA) in CD3OD with TMS as an internal standard.

In vivo evaluation of the anti-pancreatic cancer activities of F2 and 3β,20S-Di-O-Glc-DM

C57BL/6J mice (14‒18 g, male) were purchased from NIFDC (Beijing, China). PAN02 tumors inoculated in the axillae of mice were stripped under aseptic conditions and made into the tissue cell suspension. Then the cells were diluted to 2 × 107 cells/mL. 0.2 mL cell suspension was subcutaneously implanted in the left flank of each mouse. The mice were randomly divided into eight groups after being inoculated for 24 h. Model control was administered with 30% PEG400, positive control was administered with 10.0 mg/kg Rg3, and the other six groups were continuously administered with 5.0 mg/kg, 10.0 mg/kg and 20.0 mg/kg of F2 or 3β,20S-Di-O-Glc-DM (dissolved in 30% PEG400) for 11 days, once a day (i.g.). The tumors were excised and weighed to calculate the inhibition rate (IR) of tumor growth after the mice were euthanized.

Results and discussion

Production of ginsenosides F2 and 3β,20S-Di-O-Glc-DM by enzymatic catalysis with PgUGT74AE2 and UGTPg1

In previous studies, it was demonstrated that PgUGT74AE2 from P. ginseng could transfer a glucose moiety to C3-OH of PPD and DM to produce ginsenosides Rh2 and 3β-O-Glc-DM, respectively, while UGTPg1 from P. ginseng could transfer a glucose moiety to C20‒OH of PPD and DM to produce compound K and 20S-O-Glc-DM, respectively,. However, it has not been reported so far that PgUGT74AE2 and UGTPg1 can transfer a glucose moiety to C3‒OH and C20‒OH of PPD and DM simultaneously to produce F2 and 3β,20S-Di-O-Glc-DM, respectively. In order to demonstrate this, recombinant PgUGT74AE2 and UGTPg1 were incubated simultaneously with UDPG as the glycosyl donor, PPD and DM as the glycosyl acceptors, respectively. The production of F2 and 3β,20S-Di-O-Glc-DM was confirmed by LC‒MS, 1H and 13C NMR (Fig. 1, Supporting Information Figs. S5 and S6). The results indicated that both F2 and 3β,20S-Di-O-Glc-DM could be produced by in vitro enzymatic catalysis with PgUGT74AE2 and UGTPg1, which also meant that PgUGT74AE2 and UGTPg1 could be used as basic functional elements to construct the metabolic pathways of F2 and 3β,20S-Di-O-Glc-DM in microbes. In vitro enzymatic catalysis for the production of F2 and 3β,20S-Di-O-Glc-DM requires expensive UDPG as the glycosyl donor, PPD and DM as the glycosyl receptors, thereby limiting its practical application. Therefore, we aimed to construct the yeast cell factories for the de novo production of F2 and 3β,20S-Di-O-Glc-DM from cheap glucose.

Figure 1

Identification of the products by in vitro enzymatic catalysis. (A) HPLC analysis of the product of PPD catalyzed by PgUGT74AE2 and UGTPg1. Blue, authentic samples; red, the product of enzymatic catalysis. (B) The MS spectrum of the product of PPD catalyzed by PgUGT74AE2 and UGTPg1. (C) HPLC analysis of the product of DM catalyzed by PgUGT74AE2 and UGTPg1. Blue, authentic samples; red, the product of enzymatic catalysis. (D) The MS spectrum of the product of DM catalyzed by PgUGT74AE2 and UGTPg1.

Constructing yeast cell factories for the de novo biosynthesis of ginsenosides F2 and 3β,20S-Di-O-Glc-DM

Besides enzymatic catalysis, metabolic engineering of S. cerevisiae has been widely used as a green and sustainable biotechnology to produce natural and unnatural products. It was reported that deletion of HXK2 gene could promote the biosynthesis of exogenous terpenoids in S. cerevisiae probably because metabolic flux change in the mutant led to reinforcement of the glycolysis flux to the MVA pathway. Previously, we knocked out HXK2 gene to boost the production of ginsenosides in S. cerevisiae, resulting in the optimized chassis strain Y-ΔHXK2. To construct the biosynthetic pathway of F2 in S. cerevisiae, the gene expression module containing DS, PPDS, PgUGT74AE2, UGTPg1 from P. ginseng, ATR2 from A. thaliana and tHMG1 from S. cerevisiae was integrated into the δ1 site of Y-ΔHXK2 via the CRISPR/Cas9 system (Fig. 2 and Fig. S1). Likewise, the biosynthetic pathway of 3β,20S-Di-O-Glc-DM was constructed by introducing the gene expression module containing DS, UGTPg1, PgUGT74AE2 and tHMG1 into the δ1 site of Y-ΔHXK2 (Fig. 2 and Fig. S2). A total of 10 transformants from each group were randomly picked and cultivated to identify the products and compare their titers. As expected, LC‒MS analysis of intracellular and extracellular extracts confirmed the production of F2 (Fig. 3A and B) or 3β,20S-Di-O-Glc-DM (Fig. 3C and D) in these transformants. The transformants with the highest titers of F2 and 3β,20S-Di-O-Glc-DM were named as YF2 and YDP, respectively (Fig. 4). F2 and 3β,20S-Di-O-Glc-DM were isolated from YF2 and YDP, respectively, purified by semi-preparative HPLC, and identified by comparing their 1H and 13C NMR spectroscopic data with F2 and 3β,20S-Di-O-Glc-DM produced by the above enzymatic catalysis.

Figure 2

The biosynthetic pathways of ginsenosides F2 and 3β,20S-Di-O-Glc-DM in the engineered yeasts.

Figure 3

Identification of the fermentation products of the engineered yeasts. (A) HPLC analysis of the product of the strain YF2. (B) The MS spectrum of the product of the strain YF2. (C) HPLC analysis of the product of the strain YDP. (D) The MS spectrum of the product of the strain YDP.

Figure 4

Increasing the production of ginsenosides F2 and 3β,20S-Di-O-Glc-DM in the engineered yeasts by multistep metabolic engineering strategies. The strains were cultivated in YPD medium in shake flasks for 3 days. The genes in blue boxes are involved in the upstream biosynthetic pathway. The genes in orange boxes are involved in the downstream biosynthetic pathway. The genes in green boxes are involved in the UDPG biosynthetic pathway. The genes in purple boxes are the transcriptional activator genes.

Tuning the promoters of PPDS and ATR2 genes to increase the production of ginsenoside F2

The balance between CYP450 enzymes and their reductases has a great effect on the biosynthesis of natural products,. To boost the production of F2, we tuned the promoters of PPDS and ATR2 genes. In the strain YF2, the promoters of PPDS and ATR2 genes were the weak promoter PYK1 and the strong promoter TDH3, respectively. The strain YF2-B was constructed with PPDS gene under the strong promoter TDH3 and ATR2 gene under the weak promoter PYK1, while the strain YF2-C was constructed with PPDS and ATR2 genes under the strong promoters TDH3 and TEF2, respectively. As shown in Table 2, after being cultivated for 6 days, the highest titer of F2 reached 8.8 mg/L in YF2-C, which was 7.3-fold higher than that of YF2 (1.2 mg/L). Meanwhile, the titer of intermediate Rh2 in YF2-C (2.4 mg/L) was 8.0-fold higher than that of YF2 (0.3 mg/L). In addition, the titer of PPD in YF2-C (51.0 mg/L) was also higher than that of YF2 (37.7 mg/L). These results indicated that the simultaneous utilization of the strong promoters for the expression of PPDS and ATR2 genes could significantly improve the production of F2. The reason for this was probably that the overexpression of PPDS and ATR2 genes could promote the conversion from DM to PPD, which was the bottleneck of F2 biosynthesis.

Table 2

The production of ginsenosides F2, 3β,20S-Di-O-Glc-DM and their intermediates in the engineered yeasts by fed-batch fermentation in shake flasks for 6 days.

Strain	Biomass (g/L)	F2 (mg/L)	Rh2 (mg/L)	Compound K (mg/L)	3β,20S-O-Glc-DM (mg/L)	3β-O-Glc-DM (mg/L)	20S-O-Glc-DM (mg/L)	PPD (mg/L)	DM (mg/L)
YF2	13.9	1.2	0.3	0.0	0.8	0.7	0.6	37.7	14.5
YF2-C	13.7	8.8	2.4	0.0	3.0	0.6	0.2	51.0	31.4
YFC2	16.3	16.8	2.3	0.0	5.8	0.7	0.1	50.2	29.1
YFR	14.8	21.0	4.3	0.0	18.1	0.2	0.0	31.1	33.8
YDP	12.9	‒	‒	‒	82.0	2.2	1.6	‒	21.6
YDS	14.6	‒	‒	‒	90.5	3.5	0.3	‒	23.3
YS3	14.2	‒	‒	‒	175.1	10.5	1.9	‒	69.3
YSR	15.0	‒	‒	‒	346.1	14.2	2.1	‒	49.2

The MS spectra of the intermediates (Rh2, compound K, 3β-O-Glc-DM, 20S-O-Glc-DM, PPD and DM) are listed in Supporting Information Fig. S7. ‒Not applicable.

Using the high-efficiency C3-glycosyltransferase to increase the production of 3β,20S-Di-O-Glc-DM

To improve the production of 3β,20S-Di-O-Glc-DM, we screened the genes of high-efficiency C3-glycosyltransferases from different species, including M7 (mutant of UGT74AC1) from S. grosvenorii, UGTPn50-HV (mutant of UGTPn50) from P. notoginseng, M7-1 (mutant of UGT51) from S. cerevisiae and UGT109A1 from B. subtilis. The biosynthetic pathway of 3β,20S-Di-O-Glc-DM was constructed in Y-ΔHXK2 by replacing the PgUGT74AE2 gene with the codon-optimized M7, UGTPn50-HV, M7-1 and UGT109A1 genes, resulting in the strains YDS, YDN, YDC and YDB, respectively. After being cultivated for 3 days, the titers of 3β,20S-Di-O-Glc-DM in both YDC and YDB strains were found to be much lower than that of YDP, and YDS had the highest titer of 3β,20S-Di-O-Glc-DM among the above five strains (Fig. 4). It suggested that C3-glycotransferases from plants were more efficient for the production of 3β,20S-Di-O-Glc-DM in the engineered yeasts than those from microbes. After being cultivated for 6 days, the titer of 3β,20S-Di-O-Glc-DM in YDS reached 90.5 mg/L, 1.1-fold higher than that of YDP (82.0 mg/L) (Table 2). Besides, the titer of the intermediate 3β-O-Glc-DM in YDS (3.5 mg/L) was higher than that of YDP (2.2 mg/L), whereas the titer of the intermediate 20S-O-Glc-DM in YDS (0.3 mg/L) was much lower than that of YDP (1.6 mg/L). We speculated that the high efficiency of C3-glycosyltransferase M7 led to more DM towards 3β-O-Glc-DM.

Improving the precursor supply and overexpressing the transcriptional activators to increase the production of ginsenosides F2 and 3β,20S-Di-O-Glc-DM

Previous studies have shown that the improved precursor supply is conducive to increasing the production of the target products in the engineered yeasts,31, 32, 33. Therefore, we have employed the strategies of overexpressing the key enzymes involved in the upstream biosynthetic pathway and tuning down the competitive ergosterol biosynthetic pathway to enhance the yields of ginsenosides in the engineered yeasts,. To increase the production of F2 and 3β,20S-Di-O-Glc-DM, we constructed the expression module containing the expression cassettes of IDI1, ERG20, ERG9, ERG1 and antisense ERG7 genes, and integrated it into the δ4 site of the F2-producing strain YF2-C and the 3β,20S-Di-O-Glc-DM-producing strain YDS, resulting in the strains YFC1 and YS1, respectively. After being cultivated for 3 days, the titers of F2 in YFC1 and 3β,20S-Di-O-Glc-DM in YS1 were 1.1- and 1.6-fold higher than those of YF2-C and YDS, respectively (Fig. 4). Optimizing the upstream biosynthetic pathway improved the production of 3β,20S-Di-O-Glc-DM but did not apparently improve the production of F2, suggesting that the conversion from DM to PPD was a rate-limiting step in F2 biosynthesis.

The expression levels of the heterologous enzymes in S. cerevisiae significantly affect the production of the target products. The transcriptional activator HAC1 is a major regulator of the unfolded protein response in many eukaryotes, which can up-regulate the expression of the heterologous enzymes. We previously showed that the titers of the unnatural ginsenosides 3β-O-Glc-DM and 20S-O-Glc-DM in the engineered yeasts were increased by 1.6- and 1.5-fold, respectively, by overexpressing HAC1. INO2 is a key endoplasmic reticulum (ER) size regulator, which can enlarge ER. By overexpressing INO2, the production of squalene and PPD in the engineered yeasts were increased by 71- and 8-fold, respectively. In view of this, the expression module containing IDI1, ERG20, ERG9, ERG1, antisense ERG7 genes together with the transcriptional activator HAC1 and/or INO2 genes, was transformed into the strains YF2-C and YDS, respectively, resulting in the strains YFC2, YFC3, YFC4, YS2, YS3 and YS4 correspondingly (Table 1). As a result, in F2-producing yeasts, up-regulating the upstream pathway and simultaneously overexpressing HAC1 could maximize the production of F2 (Fig. 4). After being cultivated for 6 days, the titer of F2 in YFC2 reached 16.8 mg/L, which was 1.9-fold higher than that of YF2-C (8.8 mg/L, Table 2). In 3β,20S-Di-O-Glc-DM-producing yeasts, up-regulating the upstream pathway and simultaneously overexpressing INO2 could maximize the production of 3β,20S-Di-O-Glc-DM (Fig. 4). The titer of 3β,20S-Di-O-Glc-DM in YS3 reached 175.1 mg/L, which was 1.9-fold higher than that of YDS (90.5 mg/L, Table 2). These results demonstrated that improving the precursor supply together with overexpressing transcriptional activators led to the increased production of F2 and 3β,20S-Di-O-Glc-DM. Moreover, overexpressing HAC1 was more effective for the production of F2, while overexpressing INO2 was more effective for the production of 3β,20S-Di-O-Glc-DM.

Improving the UDPG supply to increase the production of ginsenosides F2 and 3β,20S-Di-O-Glc-DM

In the previous section, we increased the titers of F2 and 3β,20S-Di-O-Glc-DM by improving the supply of the ginsenoside aglycones (glycosyl receptors). Meanwhile, the supply of the glycosyl donors is also important for the production of ginsenosides. It was reported that enhancing the supply of UDPG significantly improved the yields of ginsenoside Rh2 and compound K in the engineered yeasts,. PGM1, PGM2 and UGP1 are the key enzymes involved in the UDPG biosynthesis. In our present study, we overexpressed the three enzymes for higher titers of F2 and 3β,20S-Di-O-Glc-DM. We constructed the expression cassettes of PGM1, PGM2 and UGP1 genes under the strong promoters of TEF1, PGK1 and TDH3, respectively, to maximize the expression levels of these enzymes. Then we integrated them into the multi-copy rDNA site of the strains YFC2 and YS3, respectively, resulting in the strains YFR and YSR. In addition, EGH1 in S. cerevisiae is a glucosidase with broad specificity, which can remove the C3-glycosyl of Rh2. Because of the structural similarity among F2, 3β,20S-Di-O-Glc-DM and Rh2, we speculated that EGH1 might also hydrolyze F2 and 3β,20S-Di-O-Glc-DM. Therefore, the expression module containing PGM1, PGM2 and UGP1 was integrated into the EGH1 site of YFC2 and YS3 to overexpress the UDPG biosynthetic genes and knock out EGH1 simultaneously, resulting in the strains YFE and YSE. The titer of F2 in YFR was higher than that of YFE (Fig. 4) and reached 21.0 mg/L after being cultivated for 6 days, which was 1.3-fold higher than that of YFC2 (16.8 mg/L, Table 2). The titer of 3β,20S-Di-O-Glc-DM in YSR was higher than that of YSE (Fig. 4) and reached 346.1 mg/L after being cultivated for 6 days, which was 2.0-fold higher than that of YS3 (175.1 mg/L, Table 2). Simultaneously, the titer of the precursor PPD of F2 in YFR (31.1 mg/L) was lower than that of YFC2 (50.2 mg/L), and the titer of the precursor DM of 3β,20S-Di-O-Glc-DM in YSR (49.2 mg/L) was also lower than that of YS3 (69.3 mg/L, Table 2). These results demonstrated that in both F2-producing and 3β,20S-Di-O-Glc-DM-producing yeasts, multi-copy integration of the UDPG biosynthetic genes was more effective for the high production of target products than single-copy integration of them together with deletion of EGH1 gene. Furthermore, the titers of the precursor PPD of F2 in YFR and the precursor DM of 3β,20S-Di-O-Glc-DM in YSR were both lower than those of their respective controls, indicating that increasing UDPG supply could improve the conversion from precursors to target products.

Taken together, we were able to increase the titers of F2 and 3β,20S-Di-O-Glc-DM in the engineered yeasts by 17.5-fold (from 1.2 to 21.0 mg/L) and 4.2-fold (from 82.0 to 346.1 mg/L), respectively, by multistep metabolic engineering strategies (Table 2, Fig. 5A and B). However, the titer of F2 was much lower than that of 3β,20S-Di-O-Glc-DM probably because the biosynthesis of F2 has one more rate-limiting step from DM to PPD than that of 3β,20S-Di-O-Glc-DM. Surprisingly, F2 was mainly detected extracellularly, but 3β,20S-Di-O-Glc-DM was mainly detected intracellularly. The reason for this phenomenon needs to be further investigated. Moreover, little compound K was detected in F2-producing strains. It seems likely that compound K produced in F2-producing cells could be rapidly converted to F2. Lots of aglycones PPD and DM still accumulated in the engineered strains, probably due to the low catalytic efficiency of the UDP-glycosyltransferases. Therefore, to increase both the catalytic efficiency of UDP-glycosyltransferases by protein engineering and their gene copy numbers in the engineered yeasts is expected to further boost the titers of F2 and 3β,20S-Di-O-Glc-DM in the future. In addition, during the optimization process of the engineered yeasts, the cell biomass after fermentation did not decrease, but rather increased slightly, suggesting that the enhanced production of F2 and 3β,20S-Di-O-Glc-DM did not interfere with cell growth of the engineered yeasts (Table 2).

Figure 5

Production of ginsenosides F2 and 3β,20S-Di-O-Glc-DM in the engineered yeasts. (A) Production of F2 in shake flasks. The strains were cultivated in YPD medium by fed-batch fermentation for 6 days. (B) Production of 3β,20S-Di-O-Glc-DM in shake flasks. The strains were cultivated in YPD medium by fed-batch fermentation for 6 days. (C) Production of 3β,20S-Di-O-Glc-DM in YSR by fed-batch fermentation in a 3 L bioreactor.

Evaluating the anti-pancreatic cancer activity of ginsenosides F2 and 3β,20S-Di-O-Glc-DM

Several previous studies reported that F2 exhibited anti-glioma, anti-breast cancer and anti-gastric cancer activities. However, the anti-pancreatic cancer activity of F2 and 3β,20S-Di-O-Glc-DM has never been reported before. Accordingly, we evaluated them in the PAN02 model. Rg3 was used as the positive control because it has been developed to be an anti-tumor drug (Shenyi capsules) for twenty years and has a good performance in the clinic application. The results are shown in Table 3. In the group treated with 10 mg/kg F2, the inhibition rate was 42.9%, which was slightly lower than that of 10 mg/kg Rg3 (44.5%). In the group treated with 5 mg/kg 3β,20S-Di-O-Glc-DM, the inhibition rate was 47.0%, which was higher than that of 10 mg/kg Rg3. The results demonstrated that both F2 and 3β,20S-Di-O-Glc-DM have anti-pancreatic cancer activity. Especially, it was inspiring that 3β,20S-Di-O-Glc-DM displayed better activity even at a lower concentration than both Rg3 and F2.

Table 3

In vivo evaluation of anti-tumor activity of ginsenosides F2 and 3β,20S-Di-O-Glc-DM in the PAN02 model.

Group	Dosage	Body weight (g)		Tumor weight	Inhibition rate
	(mg/kg)	Begin	End	(g)	(%)
Negative control	‒	18.1 ± 0.5	20.7 ± 1.1	2.45 ± 0.54	‒
Rg3	10.0	17.4 ± 0.4	20.5 ± 1.1	1.36 ± 0.57∗∗	44.5
F2	5.0	17.5 ± 0.7	21.0 ± 0.9	1.60 ± 0.55∗	34.7
	10.0	17.3 ± 0.3	20.7 ± 1.2	1.40 ± 0.33∗∗∗	42.9
	20.0	16.9 ± 0.7	20.4 ± 2.2	2.03 ± 0.55	17.0
3β,20S-Di-O-Glc-DM	5.0	17.6 ± 0.5	21.0 ± 0.9	1.30 ± 0.23∗∗∗	47.0
	10.0	17.7 ± 0.8	20.3 ± 1.9	1.74 ± 0.58∗	29.0
	20.0	17.2 ± 0.5	20.8 ± 1.2	1.60 ± 0.71∗	34.6

∗P < 0.05, ∗∗P < 0.01 and ∗∗∗P < 0.001, compared with the vehicle control. ‒Not applicable.

Conducting fed-batch fermentation to increase the production of 3β,20S-Di-O-Glc-DM

Fed-batch fermentation can significantly improve the cell density, and then improve the production of target products, which has been used to produce artemisinin acid, miltiradiene, resveratrol and ginsenosides. After optimizing the engineered yeasts at the gene level, we achieved notable increase in the production of F2 and 3β,20S-Di-O-Glc-DM. Considering the higher production and the better pharmacological activity of 3β,20S-Di-O-Glc-DM than F2, we chose the engineered strain YSR for fed-batch fermentation in a 3 L bioreactor to further improve the titer of 3β,20S-Di-O-Glc-DM. The cell biomass continued to increase at the early stage of the fermentation and reached the maximum OD600 of 1356 at 144 h (Fig. 5C). The production of 3β,20S-Di-O-Glc-DM continued to increase until the highest titer reached 2.6 g/L at 96 h. After that, the titer of 3β,20S-Di-O-Glc-DM began to decrease slightly. At the end of the fermentation, the titer of 3β,20S-Di-O-Glc-DM was 2.1 g/L at 192 h (Fig. 5C). Ultimately, the titer of 3β,20S-Di-O-Glc-DM was remarkably increased by 7.5-fold (from 346.1 mg/L to 2.6 g/L) by fed-batch fermentation. The result provides a favorable condition for industrial production of 3β,20S-Di-O-Glc-DM.

Conclusions

In this study, we first identified that PgUGT74AE2 and UGTPg1 from P. ginseng could catalyze the glycosylation of PPD and DM at C3-OH and C20-OH simultaneously to produce natural ginsenoside F2 and unnatural ginsenoside 3β,20S-Di-O-Glc-DM, respectively. Then we used PgUGT74AE2 and UGTPg1 as basic functional elements to construct the complete biosynthetic pathways of F2 and 3β,20S-Di-O-Glc-DM in S. cerevisiae. The titers of F2 and 3β,20S-Di-O-Glc-DM were increased to 21.0 and 346.1 mg/L at shake flask level, respectively, by multistep metabolic engineering strategies. The titer of 3β,20S-Di-O-Glc-DM was much higher than that of F2, and most excitingly, 3β,20S-Di-O-Glc-DM exhibited better anti-pancreatic cancer activity than F2. Hence, we further improved the titer of 3β,20S-Di-O-Glc-DM to 2.6 g/L by fed-batch fermentation in a 3 L bioreactor. As far as we know, this is the first report on demonstrating the anti-pancreatic cancer activity of F2 and 3β,20S-Di-O-Glc-DM, and achieving their de novo biosynthesis from glucose in yeast cell factories. Compared with the traditional approaches, the construction of yeast cell factories for producing F2 and 3β,20S-Di-O-Glc-DM can avoid many disadvantages such as low yield and high cost. Therefore, the engineered yeasts constructed in this study can serve as renewable resource for the green and sustainable production of F2 and 3β,20S-Di-O-Glc-DM, which accelerates their industrial application for drug research and development.

Acknowledgments

We thank Prof. Jungui Dai (Institute of Materia Medica, Chinese Academy of Medical Sciences & Peking Union Medical College) for his constructive suggestions on our research work. This research was financially supported by the grants of CAMS Innovation Fund for Medical Sciences (CIFMS) (2021-I2M-1-029, China), the (7212158, China), the (81673341), Disciplinary Development of Synthetic Biology (201920100801, China).

Author contributions

Jinling Yang and Ping Zhu designed the experiments; Fenglin Jiang, Chen Zhou, Yan Li and Haidong Deng performed the experiments; Fenglin Jiang, Jinling Yang, Ping Zhu, Ting Gong, Jingjing Chen and Tianjiao Chen analyzed the data; Fenglin Jiang and Jinling Yang wrote the paper; and all authors reviewed the manuscript.

Conflicts of interest

The authors declare no conflicts of interest.