A systematic comparison of triterpenoid biosynthetic enzymes for the production of oleanolic acid in Saccharomyces cerevisiae

Triterpenoids are high-value plant metabolites with numerous applications in medicine, agriculture, food, and home and personal care products. However, plants produce triterpenoids in low abundance, and their complex structures make their chemical synthesis prohibitively expensive and often impossible. As such, the yeast Saccharomyces cerevisiae has been explored as an alternative means of production. An important triterpenoid is oleanolic acid because it is the precursor to many bioactive triterpenoids of commercial interest, such as QS-21 which is being evaluated as a vaccine adjuvant in clinical trials against HIV and malaria. Oleanolic acid is derived from 2,3-oxidosqualene (natively produced by yeast) via a cyclisation and a multi-step oxidation reaction, catalysed by a β-amyrin synthase and a cytochrome P450 of the CYP716A subfamily, respectively. Although many homologues have been characterised, previous studies have used arbitrarily chosen β-amyrin synthases and CYP716As to produce oleanolic acid and its derivatives in yeast. This study presents the first comprehensive comparison of β-amyrin synthase and CYP716A enzyme activities in yeast. Strains expressing different homologues are compared for production, revealing 6.3- and 4.5-fold differences in β-amyrin and oleanolic acid productivities and varying CYP716A product profiles, which are important to consider when engineering strains for the production of bioactive oleanolic acid derivatives.


Introduction
Oleanolic acid is a plant specialized metabolite of considerable commercial interest that belongs to the oleanane class of triterpenoids. The oleanane scaffold is composed of 30 carbons arranged in a pentacyclic structure that is decorated with various functional groups to obtain vast structural and functional diversity. Several plant species produce oleanolic acid [1], which can accumulate as itself or as oxidized and glycosylated derivatives called saponins. Oleanolic acid is a particularly important triterpenoid, being the precursor to a range of bioactive triterpenoids such as the vaccine adjuvant QS-21, which is used in a vaccine against feline leukaemia virus and is being investigated in human vaccines against HIV and malaria [2]. Semi-synthetic derivatives of oleanolic acid have been investigated in the treatment of various diseases, including leukaemia, chronic kidney disease, type II diabetes, osteoporosis and Friedreich's ataxia [3,4], while glycosylated derivatives of oleanolic acid have been shown to act as feeding deterrents to common crop pests [5]. As such, there is considerable interest in the production of oleanolic acid derivatives in industrially relevant quantities. Plants typically produce triterpenoids in low amounts in response to stress, and the complex stereochemistry of triterpenoids makes their chemical synthesis economically prohibitive and often impossible. Therefore, use of the yeast Saccharomyces cerevisiae, which is genetically tractable and widely used in industrial fermentations, has been explored for triterpenoid production [6][7][8][9]. Oleanolic acid biosynthesis occurs through the cyclisation of 2,3-oxidosqualene (natively produced by yeast) to produce β-amyrin, followed by the oxidation of carbon-28 (C-28) methyl group of β-amyrin to a carboxyl group. This oxidation reaction is predominantly catalysed by cytochrome P450 monooxygenases (P450s) of the CYP716A subfamily in conjunction with a cytochrome P450 reductase (CPR) partner, and proceeds through three steps, producing the intermediates erythrodiol and oleanolic aldehyde, which have a hydroxyl and aldehyde group at C-28, respectively ( Fig 1A). Oleanolic acid can thus be produced in yeast by the co-expression of a β-amyrin synthase (BAS), a CYP716A and a CPR, with the expression of additional P450s and UDP-dependent glycosyltransferases allowing the production of bioactive oleanolic acid derivatives.
The choice of enzyme homologues can significantly impact productivity, as exemplified by β-amyrin titres of 19 and 36 mg/L when BASs from Medicago truncatula (MtBAS) and Glycyrrhiza glabra (GgBAS) were expressed in yeast [6]. In addition, titres of 107 and 1.9 mg/L βamyrin have been reported for GgBAS and PgBAS1 (from Panax ginseng) [7]. Despite these broad-ranging yields, no systematic comparison of BAS and CYP716A productivities in yeast have been reported. The advantage of such comparisons has previously been demonstrated for the CYP93E subfamily of P450s, which catalyse the three-step oxidation of the β-amyrin C-24 methyl group to a carboxyl group. A 52-fold difference in accumulation of 24-hydroxy-βamyrin was reported between strains expressing eight different CYP93E variants [10]. As observed for the CYP93E subfamily, a comparative analysis of CYP716A enzymes would be particularly relevant due to the multi-step nature of the reaction catalysed, because the ratios of accumulated intermediates and product could differ between homologues. We therefore decided to conduct a comprehensive comparison of the productivities of BAS and CYP716A homologues in yeast, to identify the most productive variants for oleanolic acid production. Such variants could then be used to engineer high yields of oleanolic acid and its derivatives from yeast.

Media
The synthetic defined medium was composed of 6.9 g/L yeast nitrogen base without amino acids (Sigma) and 0.79 g/L complete supplement mixture or 0.77 g/L single drop-out of uracil (Formedium) with carbon source (either 2% D-glucose [Fisher Chemical], or 2% D-(+)-galactose [Acros Organics]).

Plasmid construction
All genes except AsBAS were codon-optimised for S. cerevisiae, synthesised by GeneArt and assembled into low-copy number plasmids expressing the URA3 auxotrophic marker [11]. AsBAS could not be synthesised and the native Avena strigosa sequence was a gift from the Osbourn lab (John Innes Centre, UK).
Plasmid backbones were initially constructed using parts provided in the MoClo Yeast Toolkit (MoClo-YTK) [11]. Plasmids pCPS1ULA, pCP12ULA and pCP2EULA each consisted of a CEN6/ARS4 low-copy origin, URA3 selection marker, ColE1 origin and Amicillin resistance gene for maintanence in E. coli, as well as a cloning site for either a BAS, CYP716A or ATR2 transcriptional unit (consisting of a promoter, ORF and terminator), respectively. Plasmid pMGSEULS instead had an E. coli Spectinomycin resistance gene and a site into which three transcriptional units (BAS, CYP716A and ATR2) can be cloned to form an oleanolic acid biosynthetic pathway.
BAS, CYP716A and ATR2 transcriptional units, and oleanolic pathways, were subsequently assembled into these vectors using the MoClo-YTK system [11]. BAS genes were cloned into pCPS1ULA with the P GAL1 promoter and TDH1 terminator, CYP716A genes into pCP12ULA with P GAL10 and the ENO2 terminator, and ATR2 into pCP2EULA with P GAL3 and the PGK1 terminator. The resulting transcriptional units were then cloned into pMGSEULS to form oleanolic acid pathways, in which BAS, CYP716A and ATR2 are co-expressed.

Generation and culturing of strains
Yeast strains were generated by transforming plasmids into S. cerevisiae BY4741 (β-amyrin production strains by transforming pCPS1ULA-BAS plasmids, and oleanolic acid strains by transforming pMGSEULS-OA plasmids). Empty pCPS1ULA vector was used as a control. All strains were grown at 30˚C in synthetic defined medium lacking uracil and containing either 2% -glucose (for selection of transformants, maintenance of strains, or precultures prior to inoculation in culturing medium) or 2% -(+)-galactose (for quantifying triterpenoid production). For production of triterpenoids, strains were precultured for 24 h, inoculated to a starting OD 600 of 0.2 in galactose-containing medium, and grown for 96 h in glass vials.

Gas chromatography-Mass spectrometry analysis
The triterpene profiles were analysed on an Agilent 7890B gas chromatogram (GC) coupled to an Agilent 7200B quadrupole time-of-flight mass spectrometer (QTOF-MS) with GERSTEL multipurpose sampler (MPS) robotics (Anatune). Trimethysilylated samples (1 μL) were injected at a split ratio of 10:1, with a split flow of 10 mL/min into a DB-5ms 40 m × 250 μm × 0.25 μm GC column (Agilent Technologies). Helium was used as the carrier gas at a flow rate of 1 mL/min, the inlet was set to 250˚C and the GC oven was programmed to 100˚C for 2 min, followed by ramping at 25˚C/min to 300˚C, where it was held for 25 min. The ion source were set to 230˚C, 35 μA filament current, 70 eV electron energy and the mass range of 60-900 m/z was scanned at an acquisition rate of 4 spectra/s with a solvent delay of 13 min. Total ion chromatograms and mass spectra were analysed using the Agilent MassHunter Qualitative Analysis B.07.00 software, and peak areas were calculated using the Agile 2 integrator method. GC/Q-TOF MS peak areas were used to obtain standard curves for β-amyrin, erythrodiol and oleanolic acid. The concentration range of 0 to 120 μg/mL, with increments of 10 μg/mL, was used and the linear region of the curves allowed quantifying concentration of the triterpenes.

Characterization of new BAS and CYP716A enzymes
We first identified new BAS and CYP716A homologues. BLASTp searches on NCBI using GgBAS from G. glabra [12] and CYP716A12 from M. truncatula [13] as queries identified hits with > 70% amino acid sequence identity. GhBAS from Gossypium hirsutum (cotton), CYP716A48 from Olea europaea (olive) and CYP716A147 from Theobroma cacao (cocoa tree) were selected for further analysis (S1 Table). While G. hirsutum and T. cacao produce triterpenoids [14,15], genes encoding oleanane biosynthetic enzymes have not been previously characterised. Similarly, O. europaea accumulates oleanolic acid [1], but a P450 involved in its biosynthesis has not yet been described.
In silico analysis of the GhBAS sequence (S1 Fig) identified the highly conserved DCTAE motif involved in initiation of the cyclization reaction [16,17], the MWCYCR motif known to confer product specificity [18], and the QW motifs that are speculated to be involved in BAS protein stability [19][20][21]. Sequence alignment of CYP716A147 and CYP716A48 with functionally characterised CYP716A proteins (S2 Fig) identified the proton transfer groove, ExxR motif, PERF motif and haem-binding loop, along with the N-terminal transmembrane anchor.
The selected genes were functionally characterised in yeast. GhBAS was expressed alone, while CYP716A48 and CYP716A147 were co-expressed with GgBAS and ATR2, a CPR from Arabidopsis thaliana. While yeast encodes its own native CPR (NCP1), it is often unable to support the catalytic activity of heterologously expressed plant P450s [22]. As such, a plant CPR gene is usually co-expressed with plant P450s, typically one of the A. thaliana CPRs, ATR1 or ATR2 [7,8,23], of which ATR2 has been reported to have a higher catalytic activity [24,25].
Compared to a strain carrying an empty vector (MD-N1), the strain expressing GhBAS produced β-amyrin, with an identical retention time and fragmentation pattern as an authentic standard (Fig 1B and 1C). Similarly, peaks with retention times and fragmentation patterns matching standards of erythrodiol and oleanolic acid were observed from both CYP716Aexpressing strains (Fig 1D-1H). An additional peak not present in MD-N1 was observed in these strains (Fig 1D), with mass spectra matching that reported for an oleanolic aldehyde standard (Fig 1I and 1J) [26]. Thus, GhBAS is a monofunctional BAS that cyclises 2,3-oxidosqualene to β-amyrin, while CYP716A147 and CYP716A48 are C-28 oxidases that oxidise βamyrin to oleanolic acid via erythrodiol and oleanolic aldehyde intermediates.

Strains expressing AaBAS or CqBAS1 produce the highest titre of βamyrin
We screened the amount of β-amyrin produced in yeast by 12 BAS proteins from different plant species (S1 Table). These included the BAS variants most commonly used in engineering studies (GgBAS, MtBAS and PgBAS1), as well as AsBAS, BvBAS, AaBAS, CqBAS1, EtBAS, LjBAS, PtBAS, SlBAS and the newly characterised GhBAS.
Interestingly, in addition to β-amyrin, a second peak was observed in the total ion chromatogram of the BvBAS strain (S3A Fig). This peak appeared to be a combination of two peaks, suggesting the presence of two additional compounds, and was unique to this strain, not appearing in the other BAS strains or the MD-N1 control strain. Comparison of its mass spectrum with that of an authentic lupeol standard indicated that one of the additional compounds contained within the peak most likely corresponds to this triterpene (S3B Fig).

Strain expressing CYP716AL1 produces the highest titre of oleanolic acid
We next compared the activities of 16 CYP716A family proteins from a range of plant species (S1 Table) for their ability to oxidise β-amyrin to oleanolic acid. These included CYP716A12, which is commonly used in engineering studies, the newly identified CYP716A147 and CYP716A48, as well as CYP716A15, CYP716AL1, CYP716A52v2, CYP716A44, CYP716A75, CYP716A79, CYP716A80, CYP716A83, CYP716A110, CYP716A140, CYP716A179, CYP716A244 and CYP716A1.
All strains showed comparable growth to the control strain expressing AaBAS alone ( Fig  2D) but displayed marked differences in triterpene production (Fig 2E), with oleanolic acid titre (mg/L) and productivity (mg/L/OD 600 ) differing 4.8-and 4.5-fold, respectively, between strains ( Fig 2E). Oxidation of β-amyrin was observed in all strains except that expressing CYP716A140, with oleanolic acid produced in 11 of the 16 strains (cyan bars, Fig 2E). The highest and lowest titres of 14.3 ± 1.6 and 3.0 ± 0.0 mg/L oleanolic acid were produced by strains expressing CYP716AL1 and CYP716A52v2, respectively. CYP716AL1 also showed the highest oleanolic acid productivity (2.6 mg/L/OD 600 ) and CYP716A52v2 the least (0.6 mg/L/ OD 600 ) (Fig 2G). Notably, all strains (except that expressing CYP716A140) accumulated varying amounts of erythrodiol (yellow bars, Fig 2E) and oleanolic aldehyde (green bars, Fig 2F; standard unavailable for quantification). Interestingly, CYP716A140 has previously been shown to oxidise β-amyrin to oleanolic acid, in a study using the M. truncatula CPR MTR1 and a non-codon optimised CYP716A140 sequence [27]. These variances in the expression assay may explain the differences in activity observed for CYP716A140 in our study. The CYP716A75 strain produced the highest erythrodiol titre (24.6 ± 3.9 mg/L), with trace amounts of oleanolic aldehyde and no oleanolic acid. Strains expressing CYP716A79, CYP716A110 and CYP716A1 accumulated 1.2 ± 0.1, 1.2 ± 0.1 and 4.5 ± 0.7 mg/L erythrodiol, respectively, with no oleanolic aldehyde or acid.

Combining optimal enzymes to boost oleanolic acid production
To assess the impact of enzyme variant selection on oleanolic acid titres in yeast, we generated strains expressing SlBAS (the lowest β-amyrin titre) with CYP716AL1 (the highest oleanolic

PLOS ONE
Comparison of triterpenoid biosynthetic enzymes for the production of oleanolic acid in S. cerevisiae acid titre) or CYP716A52v2 (the lowest oleanolic acid titre) genes. Growth and oleanolic acid production was compared to strains expressing AaBAS (the highest β-amyrin titre) with CYP716AL1 or CYP716A52v2. All strains showed comparable growth after 96 h of culturing (Fig 3A). The expression of SlBAS with CYP716AL1 or CYP716A52v2 resulted in 3.9 ± 0.2 and 2.8 ± 0.0 mg/L oleanolic acid, respectively (Fig 3B). The combination of AaBAS with CYP716AL1 showed the highest titre (8.5 ± 0.2 mg/L) and productivity (1.40 ± 0.0 mg/L/ OD 600 ) of oleanolic acid (Fig 3B and 3C).

Discussion
In this study, a comprehensive comparison of 12 BAS and 16 CYP716A variants for the production of oleanolic acid in yeast was performed. These included enzymes commonly used in engineering studies (GgBAS, MtBAS, PgBAS1 and CYP716A12), as well as one BAS and two CYP716As newly characterised in this study.
These enzymes displayed a range of activities and product specificities, and impacted growth rates to different extents, when expressed in yeast. Notably, AaBAS and CqBAS1 produced higher titres of β-amyrin (10.8 mg/L), and CYP716AL1 produced higher titres of oleanolic acid (14.3 mg/L), than the BAS and CYP716A enzymes commonly used in engineering studies (GgBAS, MtBAS, PgBAS1 and CYP716A12). In addition, CYP716A75 predominantly produced the triterpenoid erythrodiol, suggesting an inherent ability of CYP716A75 to mainly catalyse the hydroxylation of β-amyrin to erythrodiol, as previously reported [28]. By contrast, CYP716AL1 and CYP716A12 produced substantially less erythrodiol, favouring the production of oleanolic acid. The use of P450s with different product specificities, such as those characterised here, will enable the production of a range of triterpenoids with specific modifications to the triterpene scaffold.
Interestingly, the strain expressing AsBAS was the only strain to show similar growth to the control strain MD-N1 carrying an empty vector. All other BAS strains displayed reduced growth compared to MD-N1. While experiments were not undertaken to determine the cause of this, it is notable that AsBAS was the only BAS gene that wasn't codon-optimised for expression in yeast, which could have led to lower production of the AsBAS enzyme compared to the other BASs, possibly resulting in reduced burden on the cells. Protein sequence alignment also showed that AsBAS differed significantly in sequence to the other BAS proteins, having 47-50% sequence identity to them; by contrast, the other BAS proteins had 69-93% sequence identity to one another. While the expression of BAS genes reduced growth by differing amounts, the additional expression of CYP716A and ATR2 genes appeared to have much less impact on cell growth, with the growth of oleanolic acid-producing strains varying far less than the BAS strains. The observed reductions in growth may result from (i) amino acid burden caused by expression of the recombinant proteins, (ii) aberrant activities of the transgenic enzymes, (iii) accumulation of triterpenoid product and/or (iv) depletion of native metabolites (e.g. redirection of metabolic flux from sterol to triterpenoid biosynthesis) [3,29]. Most likely, a combination of these factors and others could explain the differences in growth observed between the strains.
The strains used in this study were derived from a BY4741 parent strain transformed with plasmids expressing the triterpenoid biosynthetic gene(s). These therefore represent unoptimised strains and numerous strain engineering approaches could be taken to improve titres. In an early study, overexpression of a truncated version of the native tHMG1 (a rate-limiting enzyme of the mevalonate pathway required for the production of 2,3-oxidosqualene), and downregulation of the native ERG7 (lanosterol synthase which uses 2,3-oxidosqualene as substrate), increased β-amyrin titres in yeast expressing AaBAS by 1.5-fold (to 6 mg/L), compared with expression of AaBAS alone [30]. Higher titres of 50 mg/L β-amyrin were later achieved by expressing GgBAS in a strain overexpressing tHMG1 and with repression of the ERG7 gene [6]. The strain was cultured in the presence of methyl-β-cyclodextrin, which improved titres, possibly by sequestering β-amyrin in the culture media and preventing toxic build-up within

PLOS ONE
Comparison of triterpenoid biosynthetic enzymes for the production of oleanolic acid in S. cerevisiae the cells. Meanwhile, another study reported oleanolic acid titres of 71 mg/L [7]. Two copies of GgBAS were co-expressed with CYP716A12 and AtATR1, together with overexpression of tHMG1 and the native genes ERG9 and ERG1, which produce squalene from farnesyl pyrophosphate (FPP, a product of the mevalonate pathway) and convert it to 2,3-oxidosqualene, respectively. In addition, the strains were grown in rich medium (YPD), which may have facilitated growth and the production of triterpenoids. An organelle engineering approach increased titres of β-amyrin and its derivative 28-O-glucose-medicagenic acid by 8-and 16-fold compared to an unengineered strain [8]. Knock-out of the PAH1 gene, encoding phosphatidic acid phosphatase, resulted in proliferation of the ER and a corresponding increase in the intracellular levels of triterpenoid biosynthetic enzymes, boosting strain productivity. The use of fermenter apparatus has also resulted in increased triterpenoid production. For example, oleanolic acid titres of 607 mg/L were achieved in a fermenter, compared with 186 mg/L in shake flasks using the same strain [31].
This study compares BAS and CYP716A enzmye productivities in yeast, reports a range of product specificities between CYP716A enzymes, and identifies that expression of AsBAS has minimal impact on yeast growth while producing relatively high titres of β-amyrin. This work should therefore serve as an important resource for the production of industrially relevant triterpenoids in yeast. (expressing no BAS). While both AaBAS and BvBAS have a peak corresponding to β-amyrin, the BvBAS strain has a second unique peak ( � ) that is potentially a combination of two peaks and which is not present in either the AaBAS or control strains. (B) Mass spectrum at retention time marked by asterix ( � ) (top). Comparison with a mass spectrum of an authentic lupeol standard (bottom) suggests one of the compounds contained within the peak corresponds to lupeol. (DOCX) S1 Table