Pathway elucidation of bioactive rhamnosylated ginsenosides in Panax ginseng and their de novo high-level production by engineered Saccharomyces…

Discovery of the missing UDP-glycosyltransferase for ginsenoside Rg2 and Re biosynthesis

Previously, we systematically characterized a series of UGTs involved in the biosynthesis of ginsenosides and completely resolved the biosynthetic pathway of Rh1 and Rg121. The downstream pathway from Rh1 and Rg1 to Rg2 and Re is speculated to be catalyzed by an unknown UGT enzyme (Fig.1). Because all previous characterized UGTs responsible for the sugar elongation of ginsenosides belong to the UGT94 family30, we thus focused our efforts on screening UGT candidates belong to this family.

Bluish green arrows represent glycosylation steps using UDP-glucose as a sugar donor, blue arrows represent the biosynthetic pathway of UDP-rhamnose and glycosylation steps using UDP-rhamnose as a sugar donor. Multi-step conversions were presented as multi arrows. Bluish green marked genes represented previous reported genes and thebule marked gene represented the identified onein this study. PPD protopanaxadiol, PPT protopanaxatriol.

We identified 665 UGTs with >350 amino acids residues (typical plant UGTs length) from the P. ginseng transcriptome that were predicted to have a conserved Plant Secondary Products Glycosyltransferase (PSPG) box. These could be clustered into 187 OTUs with a 95% cutoff. Twenty-two OTUs and their representative UGTs which possessed >40% amino acid identity to the previous identified UGT94 family UGT member, PgUGT94Q230 were recognized as belong to the UGT94 family. Through gene co-expression analysis of these UGT candidates, previously characterized P450s, and UGTs involved in PPT-type ginsenoside biosynthesis from P. ginseng transcriptome data, a UGT94 gene named PgURT94 was identified. The expression pattern of PgURT94 was strongly correlated with PgDDS, CYP716A47, CYP716A53v2, PgUGT71A53, and PgUGT71A54, which are involved in Rh1 and Rg1 biosynthesis (Fig.2a). Besides, PgURT94 was highly expressed in the root hairs of P. ginseng, which is consistent with the distribution of Re in this tissue (Supplementary Table1). Thus, we speculated that PgURT94 is most likely to be involved in the biosynthesis of Rg2 and Re.

a Heat-map analysis of the relative abundance of PgURT94 expression, along with PgDDS, CYP716A47, CYP716A53v2, and PgUGT71A53 in different parts of P. ginseng. b HPLC analysis of the in vitro reaction products catalyzed by PgURT94 crude enzyme using Rh1 as sugar acceptor and UDP-rhamnose as sugar donor. c HPLC analysis of the in vitro reaction products catalyzed by PgURT94 crude enzyme using Rg1 as sugar acceptor and UDP-rhamnose as sugar donor. Crude enzymes of E. coli strain harboring pET28a empty vector were used as a negative control for above assays and authentic ginsenoside samples Rh1, Rg1, Rg1, and Re were monitored as standards.

To verify this hypothesis, we firstly cloned this gene from callus of P. ginseng. The PgURT94 gene has an open reading frame (ORF) of 1404bp, encoding a protein of 467 amino acids. PgURT94 protein has a sequence identify of 46.1%, 47.1%, and 44.8% with PgUGT94Q2 (UGT catalyzing C3-O-Glc glucosylation of PPD-type ginsenosides), PgUGT94Q3 (UGT catalyzing C6-O-Glc glucosylation of PPT-type ginsenosides), and PgUGT94Q6 (UGT catalyzing C20-O-Glc glucosylation of PPD and PPT-type ginsenosides), respectively30. For the enzymatic activity test, PgURT94 was initially expressed in E. coli (Supplementary Fig.1) and the crude enzymes from E. coli expressing PgURT94 were incubated with Rh1 and Rg1 as substrates, and UDP-Rha as a sugar donor. The reaction products were subjected to HPLC analysis and results indicated a product was generated in the reaction extract from PgURT94 and Rh1 incubations, which had the same retention time as the Rg2 standard. This compound was not detected in the control reaction with Rh1 and crude enzyme of E. coli strain harboring empty pET28a vector (Fig.2b). A product was also observed in the reaction extract from PgURT94 using Rg1 as a substrate, and was monitored along with the Re standard (Fig.2c). The structures of these two newly produced compounds were confirmed to be Rg2 and Re, respectively, by HPLC/electrospray ionization mass spectrometry (ESIMS) (Supplementary Fig.2) and NMR (Supplementary Fig.3).

To test the sugar donor specificity of PgURT94, in vitro enzymatic assays were performed by using UDP-glucose as a sugar donor and incubating PgURT94 with Rh1 and Rg1, respectively. To ensure accuracy of the assay, a previously reported UGT (PgUGT94Q3) which could catalyze the glycosylation modification of Rh1 and Rg1 using UDP-glucose as a sugar donor30, was used as a positive control. TLC and HPLC analyses of the reaction extracts revealed that while production of glycosylated products Rf and C20-O-Glc-Rf could be detected by PgUGT94Q3 as expected, no products were detected by PgURT94, indicating that PgURT94 could not use UDP-glucose as a sugar donor (Supplementary Fig.4). These results demonstrated that PgURT94 is a specific rhamnosylation UGT.

Through functional characterization of PgURT94, the complete biosynthetic pathway of Rg2 became clear: PgUGT71A54 catalyzes the C6-OH glycosylation of PPT to form Rh1, and PgURT94 then transfers a rhamnose moiety to the C6-O-Glc of Rh1 to produce Rg2 (Fig.1). To achieve de novo biosynthesis of Rg2 in yeast, codon-optimized PgUGT71A54 and PgURT94 (hereafter referred to as synPgURT94), under the control of two strong constitutive promoters respectively, were introduced into the chromosome of strain PPT-10, a PPT-producing chassis constructed in our previous work23. Since S. cerevisiae lacks the native UDP-Rha biosynthetic pathway, AtRHM2 from A. thaliana, which catalyze the formation of UDP-Rha from UDP-glucose, was also expressed in PPT-1031,32. The resulting strain, called Rg2-01, produced 36.8mg/L Rg2 according to the analysis of metabolites in subsequent flask fermentations (Fig.3a, b).

a HPLC analyses of Rg2, Rh1, PPT and PPD production in yeast strains Rg2-01, Rg2-02, Rg2-03, and Rg2-04. The PPT chassis strain PPT-10 was used as a control. Mixed samples of Rf, Rh1, Rg2, PPT and PPD were monitored as standards. b Quantitative analysis of Rg2 and its related intermediates Rh1, PPT, and PPD in yeast strains Rg2-01, Rg2-02, Rg2-03, and Rg2-04. Genetic modification of each strain was drawn under the column, + represent the strain possess the corresponding engineering, while represent the corresponding engineering are missing the strain. All data represent the mean of n=3 biologically independent samples and error bars show standard deviation.

The total triterpenoid production (PPD+PPT+Rh1+Rg2) in strain Rg2-01 decreased sharply compared to that of the parent strain PPT-10 (Fig.3b). Since the triterpenoid biosynthetic pathway is a highly NADPH-consuming pathway and the biosynthesis of UDP-Rha from UDP-Glc by AtRHM2 in the Rg2-producing strain Rg2-01 is also an NADPH-dependent pathway33, we thus focused our attention on the NADPH consumption. For PPT biosynthesis, the precursor pathway from acetyl-CoA to 2, 3-oxidosuqalene requires three NADPH molecules (the formation of mevalonate from 3-hydroxy-3-methylglutaryl-CoA need two NADPH molecules34 and the formation of 2, 3-oxidosqalene from squalene need one NADPH molecule35). Besides, two P450s that catalyze the formation of PPT from DM also consume two NADPH molecules12,13 (Supplementary Fig.5). Therefore, the NADPH supply is of great importance for PPT production. We speculate that the introduced of NADPH-dependent UDP-Rha synthase AtRHM2 may further increase NADPH consumption and caused a reduction in total triterpenoid production in strain Rg2-01 (Supplementary Table6).

To address this issue, we utilized a NADPH-independent UDP-Rha synthase to alleviate the NADPH limitation. Several studies demonstrated that an engineered RHM enzyme (VvRHM-NRS) formed by fusing a bifunctional UDP-4-keto-6-deoxy-d-glucose 3,5-epimerase (NRS)/UDP-4-keto-rhamnose 4-keto-reductase (ER) from A. thaliana to the N-terminal of a Vitis vinifera UDP-Rha synthase VvRHM can be a self-sufficient NADPH-independent enzyme for UDP-Rha synthesis36,37. VvRHM-NRS and synPgURT94 were then introduced into PPT-10 to construct strain Rg2-02. Metabolite analysis of strain Rg2-02 indicated that Rg2 production increase to 66.4mg/L, which is approximately 1.8-fold the amount of Rg2-01 (Fig.3b). No significant reduction in total triterpenoid production was observed compared to the parent strain, PPT-10. These results clearly demonstrate that utilization of a NADPH-independent enzyme for UDP-Rha synthesis could rescue the triterpenoid reduction resulting from insufficient NADPH supply.

During screening the single clones of the construction of strain Rg2-02, a clone, hereafter designated as strain Rg2-03, was found to produce Rg2 with a production of 107.5mg/L, which is significantly higher than Rg2-02. The synPgURT94 gene of Rg2-03 was then amplified and sequenced; a missense T-to-A mutation of the 163th nucleotide was found, which resulted in a leucine to methionine mutation at the 55th amino acid. To explore whether this amino acid mutation contributed to the enhanced production of Rg2 in strain Rg2-03, we expressed the synPgURT94 mutant (synPgURT94m1) in E. coli and performed an in vitro enzymatic activity assay. Using Rh1 as a substrate, the conversion ratio of Rh1 to Rg2 by synPgURT94 and synPgURT94m1 was 70.6% and 92.4%, respectively, which demonstrated a great improvement in catalysis efficiency by synPgURT94m1 (Fig.4a, c, Supplementary Table7). The improved catalytic performance of synPgURT94m1 may explain the increased Rg2 production of strain Rg2-03. The catalytic activity of synPgURT94m1 towards Rg1, to produce Re, was also assessed; significant enhancement of catalytic activity was also observed (Fig.4b, d, Supplementary Table7). Therefore, this mutant was used for Re-cell factory construction.

a, b HPLC analyses of the in vitro reaction products catalyzed by synPgURT94 and synPgURT94m1 crude enzymes using Rh1 (a) and Rg1 (b) as thesugar acceptor and UDP-rhamnose as thesugar donor. Crude enzymes of E. coli strain harboring pET28a empty vector were used as a negative control and authentic ginsenoside samples Rh1, Rg1, Rg1 and Re were monitored as standards. c, d Quantitative analysis of the conversion ratio catalyzed by synPgURT94 and synPgURT94m1, from Rh1 to Rg2 (c) and from Rg1 to Re (d). All data represent the mean of n=3 biologically independent samples and error bars show standard deviation.

Since Rh1 did not accumulate in any of the engineered Rg2-producing strains, the conversion of PPT to Rh1 may be a limiting step in Rg2 production. To address this, we introduced an additional copy of PgUGT71A54 into strain Rg2-03 to create Rg2-04. As expected, the production of Rg2 by Rg2-04 reached 147.1mg/L, representing a 40% improvement compared to Rg2-03 (Fig.3, Supplementary Table6). By combining all engineering strategies, the production of Rg2 increased 4.0-fold from Rg2-01 to Rg2-04. However, in our final strain, there was still more than 263.3mg/L of PPT accumulated and no Rh1 was detected (Supplementary Table6). Therefore, the conversion of PPT into Rh1 remains a major bottleneck for Rg2 production. We believe that the production of Rg2 could be further improved by addressing this limiting step in the future.

A Rg1-producing yeast cell factory, Rg1-02, was constructed previously by inserting PgUGT71A53 and PgUGT71A54 respectively into the single-copy YORW22 and multi-copy delta DNA sites. The production level of Rg1 was 111.45mg/L in shake flasks and 1.95g/L in fed-batch fermentations23. For Re production in yeast, VvRHM-NRS and synPgURT94m1, under the control of strong constitutive promoters, were introduced into Rg1-02 to generate strain Re-01 (Fig.1). Subsequent shaken flask fermentation tests detected Re at a production level of 215mg/L (Fig.5b, Supplementary Fig.6).

a HPLC analyses of Rg1, Re, F1, PPT, CK, DMG, and PPD produced in yeast strain Re-01. Mixed samples of ginsenosides were monitored as standards. The Rg1 chassis strain Rg1-02 was used as a negative control. b Quantitative analysis of Re and its related intermediate Rg1 in engineered yeast strains. All data represent the mean of n=3 biologically independent samples and error bars show standard deviation.

Re production in strain Re-01 was much higher than Rg2 production in strain Rg2-04, despite the fact that Re has a more complicated biosynthetic pathway. The Re content is also much higher than that of Rg2 in Panax plants, including P. ginseng and P. notoginseng1,2. To test whether this phenomenon is determined by some intrinsic factors or just a coincidence, we then examined the difference between the two yeast strains. Re-01 and Rg2-04 share the same PPT-producing background strain and downstream rhamnosylation pathway (Fig.1), thus PgUGT71A53 and PgUGT71A54 may contribute to different production between the two strains.

Previous enzyme kinetics assays indicated that the kcat/Km value of PgUGT71A53 for PPT is higher than that of PgUGT71A54 by more than 400-fold23. Thus, the poor PPT-catalyzing efficiency of PgUGT71A54 in strain Rg2-04 may severely limit the formation of Rh1 and lead to the lower production of Rg2. Accordingly, the high efficiency of PgUGT71A53 towards PPT ensured the rapid conversion of PPT into F1 which can be subsequently converted into Rg1 by PgUGT71A54. To assess the catalytic performance of PgUGT71A54 towards F1, enzyme kinetics assays were performed. Although the kcat/Km value of PgUGT71A54 towards PPT is low (2.58102mM1s1), its kcat/Km value towards F1 is much higher (7.87101mM1s1) (Supplementary Table8). We also observed that PgUGT71A53 could also catalyze the conversion of F1 to Rg1, with a kcat/Km value of 7.40102mM1s1. Thus, the elevated activity of PgUGT71A54 and PgUGT71A53 towards F1 enabled highly efficient conversion of F1 into Rg1 in the Re-01 strain. These results demonstrated that the elaborate coordination of PgUGT71A53 and PgUGT71A54 in the pathway precisely regulated the production of downstream products. The enzyme characteristic of UGTs might determine the contents of Re and Rg2 in Panax plants, as well as that ofour engineered yeast strains.

With the above engineering efforts, we obtained two yeast strains (Rg2-04 and Re-01) that could in turn produce ginsenosides Rg2 and Re directly from glucose, respectively. However, the production titer in shaken flasks remained at relatively low levels. We previously reported a series of successful examples of the promotion of ginsenosides through fed-batch fermentations in bioreactors, including Rh2, CK, Rg1, NgR1, and NgR2, all of which reached the gram-per-liter scale after optimization of the fed-batch fermentation conditions (Supplementary Table9). To achieve higher production of Rg2 and Re, we performed high-density fed-batch fermentation of Rg2-04 and Re-01 using a 1.3-L parallel bioreactor system. Since Rg2-04 and Re-01 have the same strain background (strain PPT-10) to previous constructed NgR1- and NgR2-producing strains, the fermentation control parameters were set as reported previously23. Fresh medium was fed into the fermenter at approximately 24h and the cell biomass of both strains continuously increased after feeding. For strain Rg2-04, the OD600 continuously increased to a maximum of 330.4 at 108h and remained unchanged until 120h. Re-01 reached a growth plateau at 96h (OD600=456.8) and exhibited a slight decrease at 108h (Fig.6 and Supplementary Table6). Unexpectedly, the final cell biomass of Re-01 is more than 37.2% higher than that of Rg2-04, although the mechanism underlying this phenomenon is unclear.

a Time course analysis of cell growth and triterpenoid production of strain Rg2-04. b Time course analysis of cell growth and triterpenoid production of strain Re-01. PPD protopanaxadiol, PPT protopanaxatriol, DMG 20SO--(D-glucosyl)-dammarenediol-II.

The final production amount of Rg2 by Rg2-04 was 1.3g/L, which represents an 8.9-fold increase over shaken flask production. Other important ginsenosides precursors, including PPD and PPT, could also be detected in the fermentation broth, with titers of 2.2 and 3.0g/L, respectively (Fig.6, Supplementary Table6). The final production level of Re by Re-01 reached 3.6g/L, which represents a 16.6-fold increase over shaken flasks. Many important ginsenoside precursors, such as PPD, DMG (20SO--(D-glucosyl)-dammarenediol-II), CK, PPT, F1, and Rg1, could also be detected in the fermentation broth, with titers of 1.0, 0.5, 0.7, 0.2, 0.1, and 0.2g/L, respectively (Fig.6 and Supplementary Table6). The accumulation of numerous triterpenoid precursors in both strains indicated that some rate-limiting steps remained in the engineered strains, and there is great potential for even higher Rg2 and Re production levels if these obstacles can be overcome. Since the physicochemical properties of these ginsenoside precursors vary significantly, it will not be difficult to separate and purify them from the fermentation broth. Thus, engineered strains Rg2-04 and Re-01 may also be useful for the preparation of these valuable ginsenosides.

Read the rest here:
Pathway elucidation of bioactive rhamnosylated ginsenosides in Panax ginseng and their de novo high-level production by engineered Saccharomyces...