Complete biosynthesis of QS-21 in engineered yeast

Abstract

QS-21 is a potent vaccine adjuvant and remains the only saponin-based adjuvant that has been clinically approved for use in humans^1,2. However, owing to the complex structure of QS-21, its availability is limited. Today, the supply depends on laborious extraction from the Chilean soapbark tree or on low-yielding total chemical synthesis^3,4. Here we demonstrate the complete biosynthesis of QS-21 and its precursors, as well as structural derivatives, in engineered yeast strains. The successful biosynthesis in yeast requires fine-tuning of the host's native pathway fluxes, as well as the functional and balanced expression of 38 heterologous enzymes. The required biosynthetic pathway spans seven enzyme families-a terpene synthase, P450s, nucleotide sugar synthases, glycosyltransferases, a coenzyme A ligase, acyl transferases and polyketide synthases-from six organisms, and mimics in yeast the subcellular compartmentalization of plants from the endoplasmic reticulum membrane to the cytosol. Finally, by taking advantage of the promiscuity of certain pathway enzymes, we produced structural analogues of QS-21 using this biosynthetic platform. This microbial production scheme will allow for the future establishment of a structure-activity relationship, and will thus enable the rational design of potent vaccine adjuvants.

Fig. 1. Complete biosynthetic pathway for the de novo production of QS-21 in yeast from simple sugars.

Native yeast genes and enzymes that have been overexpressed are shown in orange, and heterologous genes and enzymes are shown in black and navy. a, Pathways for the biosynthesis of the QS-21 precursors 2,3-oxidosqualene, UDP-sugars and acyl C9-CoA. DMAPP, dimethylallyl pyrophosphate; GPP, geranyl pyrophosphate; IPP, isopentenyl pyrophosphate; UXE, UDP-xylose epimerase. b, Pathways for the synthesis and oxidation of β-amyrin (1) to QA (6) through six oxidation steps on β-amyrin carried out by three cytochrome P450s. The resulting QA (6) is functionalized with C28 carboxylic acid, C23 aldehyde and C16 hydroxy functional groups.

Fig. 2. Functionalization of QA (6) to yield QS-21-Xyl and QS-21-Api.

a, The C3-OH of 6 is decorated with a branched trisaccharide β-d-Xyl-(1→3)-[β-d-Gal-(1→2)]-β-d-GlcA through three sequential glycosylation steps. The C28-COOH is glycosylated with a linear tetrasaccharide β-d-Xyl-(1→4)-α-l-Rha-(1→2)-β-d-Fuc with a terminal sugar of β-d-Xyl or β-d-Api. b, The fucose ester-linked to C28 is then acylated twice with a nine-carbon branched dihydroxy acid moiety (C9-CoA), before it is α-l-arabinofuranosylated to complete the biosynthesis of QS-21 in yeast. Information for all genes is listed in Supplementary Table 1. Source Data

Fig. 3. Functional expression of cytochrome P450s and pathway engineering for QA.

a, Three cytochrome P450s oxidize β-amyrin (1) at the C28, C23 and C16 positions, resulting in a carboxylic acid, an aldehyde and a hydroxy group, respectively, on QA (6). MW, molecular weight. b, When expressed in yeast, the native sequence of C16 oxidase (CYP716A297) encodes a protein that has both soluble and aggregated forms, whereas the C16 oxidase expressed from the yeast codon-optimized sequence is cytosolic. By fusing the TMD of C28 oxidase to the N terminus of C16, the TMD_C28–C16 fusion protein was correctly anchored to the ER membrane. Images were acquired using a Zeiss LSM 710 confocal microscope. At least three independent experiments were performed. Scale bars, 10 μm. c, Functional expression of TMD_C28–C16 leads to the conversion of gypsogenin (5) to QA (6). d, Metabolic engineering strategies, including the expression of a MBSP, as well as the overexpression of the cytochrome P450s and their redox partners, improved the titre of QA by 60-fold. Data are mean ± s.d.; n = 3 biologically independent samples. Source Data

Fig. 4. Reconstitution of the glycosylation pathway by the functional expression of nucleotide sugar synthases and corresponding GTs.

a, Sequential addition of the C3 branched trisaccharide (GlcA-Gal-Xyl) before a linear tetrasaccharide is added stepwise to the C28 carboxylic acid (Fuc-Rha-Xyl-Xyl or Fuc-Rha-Xyl-Api). b, LC–MS peak area of corresponding products produced in yeast after the expression of the indicated enzymes and the necessary nucleotide sugar synthases. The bars in red and grey indicate the ion abundance of the target molecules and intermediates, respectively. Data are mean ± s.d.; n = 3 biologically independent samples.

Fig. 5. Acylation and terminal glycosylation towards the complete biosynthesis of QS-21.

a, The C9-CoA unit is synthesized by converting 2MB-CoA and two equivalents of malonyl-CoA through the functional expression of type III PKSs and two KRs. 2MB-CoA is not native to yeast but can be obtained by activating 2MB acid through CoA thioesterification, which can be supplemented directly in the culture medium or biosynthesized through an engineered type I PKS LovF-TE. b, Biosynthetic pathway of C9 acylation to the glycosylated 13 and 14 to form a repeating dimeric C18 moiety before the terminal arabinofuranose is added to the 5-OH. c, Ion-extracted chromatograms of the extracted yeast samples in which the engineered strains were grown in the presence of 2MB showed the efficient addition of both C9 units onto the glycosylated molecule substrate (13) to yield acylated 13-C9 (15), and 13-C18 (17). The arabinofuranosylation of 17 led to the biosynthesis of QS-21-Xyl, which co-elutes with the QS-21 chemical standard. The identical isotopic fingerprint patterns further confirm the in vivo production of QS-21. The extracted peak preceding that of QS-21 corresponds to 18+Xyl generated in vitro, possibly owing to the promiscuity of the Araf transferase.

Extended Data Fig. 1. Biosynthesis of β-amyrin in engineered yeast and culture condition optimization.

Full-length β-amyrin synthase (BAS) sourced from Saponaria vaccaria, SvBAS, was integrated into JWY601, a strain with upregulated mevalonate pathway. Production of β-amyrin was sampled over the course of three days after induction. MLY-01 harbouring overexpressed copies of ERG20 and ERG1 led to 899 mg l⁻¹ production of β-amyrin. Data are mean ± s.d.; n = 3 biologically independent samples. *P value < 0.05; ** P < 0.01. Student’s two-tailed t-test. More statistical analysis is available in the source data file. Source Data

Extended Data Fig. 2. Optimization of P450 oxidation efficiency through the expression of a scaffolding MSBP.

a, Phylogenetic tree of Saponaria vaccaria MSBP homologues and related MSBPs (see details in the Supplementary Methods). b, Expression of various MSBP homologues, including 2 newly identified SvMSBPs, led to 2- to 4-fold increase of QA production. Increased of QA production can be attributed to the improved oxidation efficiency. Data are mean ± s.d.; n = 3 biologically independent samples. *P value < 0.05; ** P < 0.01. Student’s two-tailed t-test. More statistical analysis is available in the source data file. Subcellular localization studies show that SvMSBP1 co-localizes with both c, C28 and d, C23 oxidases on the ER membrane of yeast. Images were acquired using a Zeiss LSM 710 confocal microscope (scale bar represents 10 μm; at least three independent experiments were conducted). Source Data

Extended Data Fig. 3. Functional expression of cytochrome P450s and pathway engineering for QA.

a, Introduction of additional copies of pathway enzymes revealed reaction bottlenecks and the increased production of QA. b, Distribution of the terpenoid products in the engineered yeast strains after overexpression of the cytochrome P450s, as well as their redox partners. c, Calculated titres of target products in these yeast strains with a 65-fold increase of QA production in YL-15 compared to YL-4. Data are mean ± s.d.; n = 3 biologically independent samples. Source Data

Extended Data Fig. 4. C3 glycosylation studies.

a, Activity of the two CSL homologues CSLM1 and CSLM2 was investigated by comparing the LC–MS ion intensity of the oxidized intermediates oleanolic acid (3), hederagenin (4), gypsogenin (5), QA (6), as well as their glucuronidated counterparts in the presence of CSLM1 or CSLM2 (shown in grey and orange, respectively). CSLM1 was more specific towards the glucuronidation of 6. Although CSLM2 can also glucuronidate less oxidized intermediates such as 3, 4, and 5, it was 3-fold more active towards 6. *P value < 0.05; ** P < 0.01. Student’s two-tailed t-test. More statistical analysis is available in the source data file. b, Distribution of the glycosylated products and their precursors in the engineered yeast strains. Expression of the downstream C3-GalT effectively increased the conversion of 6 by pushing the equilibrium through the consumption of 7, thus leading to the production of 8. Data are mean ± s.d.; n = 3 biologically independent samples. Source Data

Extended Data Fig. 5. Subcellular localization studies of QS-21 pathway proteins in yeast and tobacco.

C-terminal fluorescent protein fusions of the pathway enzymes are visualized. The first GT CSLM1 was localized in the ER membrane and the downstream enzymes were expressed in the cytosol except for KR2. Images were acquired using a Zeiss LSM 710 confocal microscope (scale bar represents 10 μm).

Extended Data Fig. 6. The C9 acylation of 12 in the presence of exogenously supplemented 2MB acid.

The conversion of 12 to 19 was limited possibly due to the limited availability of intracellular 2MB acid. Increased concentration of 2MB added to the culture medium effectively improved the conversion and the concentration of 2MB supplementation is thus determined to be 500 mg l⁻¹. Data are mean ± s.d.; n = 3 biologically independent samples. *P value < 0.05. Student’s two-tailed t-test. More statistical analysis is available in the source data file. Source Data

Extended Data Fig. 7. Acylated and glycosylated intermediates towards the complete biosynthesis of QS-21.

The C9-CoA unit can be added to both the C28 trisaccharide 12 and tetrasaccharide 13 with the same acyl transferase ACT2. ACT3, which catalyses the second acylation to form a repeating dimeric C18 structural motif, used both 19 and 15 as substrates, before the terminal arabinofuranose was added to the 5-OH of the C18 group. Bottom, ion-extracted chromatograms of the extracted yeast samples showed the efficient addition of the two C9 units onto the glycosylated molecule substrates (12 and 13) to yield fully acylated 12-C18 (20), and 13-C18 (17). The arabinofuranosylation of 20 and 17 led to the biosynthesis of 21 and QS-21-Xyl. 21, a structural derivative of QS-21 accumulated in yeast due to the competitive pathways between glycosylation and acylation.

Extended Data Fig. 8. Characterization of purified QS-21-Xyl from engineered yeast YL-46 compared to that of the QS-21 standard.

QS-21-Xyl production was scaled up according to the protocol in the Supplementary Information to afford purified QS-21-Xyl. a, EIC spectra of QS-21-Xyl extracted from YL-46 cultures show the co-elution with the standard. b, Isotopic fingerprint patterns of the QS-21-Xyl extracted from YL-46 cultures and purified (top) and the QS-21 standard (bottom). c, Structure of QS-21-Xyl with the characteristic mass fragments are indicated. d, Mirror plot comparison of MS2 spectra of QS-21 extracted from YL-46 cultures (top) and of the standard (bottom). A list of the MS2 fragments and their corresponding intensities can be found in Supplementary Table 7. The fragments observed are consistent with those reported in literature. All LC–MS chromatograms were extracted with the theoretical m/z values of the respective compounds of interest.

Extended Data Fig. 9. ¹H NMR spectra of purified QS-21-Xyl from engineered yeast YL-46 compared to that of the QS-21 standard.

a, Full spectral comparison between the QS-21 extracted and purified from engineered yeast YL-46 (top, red) and a QS-21 standard (bottom, black) shows an overall match of all proton peaks. In particular, the peak at chemical shift 9.33 demonstrates the presence of the C23 aldehyde in the QS-21 made in engineered yeast. b, Expanded spectral comparison between the purified QS-21 from YL-46 (top, red) and a QS-21 standard (bottom, black). The well-matched anomeric proton peaks confirm the correct connectivity between the sugar moieties. The absence of apiose proton peaks, highlighted in orange, in the yeast sample further confirms the production of QS-21-Xyl exclusively, owing to the specificity of the GT. Both spectra were recorded in acetonitrile-d₃:D₂O 1:1, 500 MHz.

Extended Data Fig. 10. Reconstitution of the glycosylation pathway of QS-21 with a C3 terminal rhamnose structural analogue.

a, The sequential addition of the C3 branched trisaccharide (GlcA-Gal-Rha) before a linear tetrasaccharide is added stepwise to the C28 carboxylic acid (Fuc-Rha-Xyl) by the functional expression of nucleotide sugar synthases and corresponding GT. The C28 GTs are promiscuous and can also use the Rha trisaccharide derivative. b, The LC–MS peak area of corresponding products produced in yeast after the expression of the indicated enzymes and the necessary nucleotide sugar synthases. Data are mean ± s.d.; n = 3 biologically independent samples. Source Data