|Home | About | Journals | Submit | Contact Us | Français|
Vanillin is one of the world's most important flavor compounds, with a global market of 180 million dollars. Natural vanillin is derived from the cured seed pods of the vanilla orchid (Vanilla planifolia), but most of the world's vanillin is synthesized from petrochemicals or wood pulp lignins. We have established a true de novo biosynthetic pathway for vanillin production from glucose in Schizosaccharomyces pombe, also known as fission yeast or African beer yeast, as well as in baker's yeast, Saccharomyces cerevisiae. Productivities were 65 and 45 mg/liter, after introduction of three and four heterologous genes, respectively. The engineered pathways involve incorporation of 3-dehydroshikimate dehydratase from the dung mold Podospora pauciseta, an aromatic carboxylic acid reductase (ACAR) from a bacterium of the Nocardia genus, and an O-methyltransferase from Homo sapiens. In S. cerevisiae, the ACAR enzyme required activation by phosphopantetheinylation, and this was achieved by coexpression of a Corynebacterium glutamicum phosphopantetheinyl transferase. Prevention of reduction of vanillin to vanillyl alcohol was achieved by knockout of the host alcohol dehydrogenase ADH6. In S. pombe, the biosynthesis was further improved by introduction of an Arabidopsis thaliana family 1 UDP-glycosyltransferase, converting vanillin into vanillin β-d-glucoside, which is not toxic to the yeast cells and thus may be accumulated in larger amounts. These de novo pathways represent the first examples of one-cell microbial generation of these valuable compounds from glucose. S. pombe yeast has not previously been metabolically engineered to produce any valuable, industrially scalable, white biotech commodity.
In 2007, the global market for flavor and fragrance compounds was an impressive $20 billion, with an annual growth of 11 to 12%. The isolation and naming of vanillin (3-methoxy-4-hydroxybenzaldehyde) as the main component of vanilla flavor in 1859 (8), and the ensuing chemical synthesis in 1874 (41), in many ways marked the true birth of this industry, and this compound remains the global leader in aroma compounds. The original source of vanillin is the seed pod of the vanilla orchid (Vanilla planifolia), which was grown by the Aztecs in Mexico and brought to Europe by the Spaniards in 1520. Production of natural vanillin from the vanilla pod is a laborious and slow process, which requires hand pollination of the flowers and a 1- to 6-month curing process of the harvested green vanilla pods (37). Production of 1 kg of vanillin requires approximately 500 kg of vanilla pods, corresponding to the pollination of approximately 40,000 flowers. Today, only about 0.25% (40 tons out of 16,000) of vanillin sold annually originates from vanilla pods, while most of the remainder is synthesized chemically from lignin or fossil hydrocarbons, in particular guaiacol. Synthetically produced vanillin is sold for approximately $15 per kg, compared to prices of $1,200 to $4,000 per kg for natural vanillin (46).
An attractive alternative is bioconversion or de novo biosynthesis of vanillin; for example, vanillin produced by microbial conversion of the plant constituent ferulic acid is marketed at $700 per kilogram under the trade name Rhovanil Natural (produced by Rhodia Organics). Ferulic acid and eugenol are the most attractive plant secondary metabolites amenable for bioconversion into vanillin, since they can be produced at relatively low costs: around $5 per kilogram (37). For the bioconversion of eugenol or ferulic acid into vanillin, several microbial species have been tested, including gram-negative bacteria of the Pseudomonas genus, actinomycetes of the genera Amycolatopsis and Streptomyces, and the basidiomycete fungus Pycnoporus cinnabarinus (19, 23, 25, 27, 31, 34, 35, 36, 45, 48). In experiments where the vanillin produced was absorbed on resins, Streptomyces cultures afforded very high vanillin yields (up to 19.2 g/liter) and conversion rates as high as 55% were obtained (15). Genes for the responsible enzymes from some of these organisms were isolated and expressed in Escherichia coli, and up to 2.9 g/liter of vanillin were obtained by conversion of eugenol or ferulic acid (1, 3, 32, 49).
Compared to bioconversion, de novo biosynthesis of vanillin from a primary metabolite like glucose is much more attractive, since glucose costs less than $0.30/kilogram (42). One route for microbial production of vanillin from glucose was devised by Frost and coworker Li (6, 20), combining de novo biosynthesis of vanillic acid in E. coli with enzymatic in vitro conversion of vanillic acid to vanillin. 3-Dehydroshikimic acid is an intermediate in the shikimate pathway for biosynthesis of aromatic amino acids, and the recombinant E. coli was engineered to dehydrate this compound to form protocatechuic acid (3,4-dihydroxybenzoic acid) and methylate this to form vanillic acid. The vanillic acid was subsequently converted into vanillin in vitro using carboxylic acid reductase isolated from Neurospora crassa. The main products of the in vivo step were protocatechuic acid, vanillic acid, and isovanillic acid in an approximate ratio of 9:4:1, indicating a bottleneck at the methylation reaction and nonspecificity of the OMT (O-methyltransferase) enzyme for the meta-hydroxyl group of protocatechuic acid. Serious drawbacks of this scheme are the lack of an in vivo step for the enzymatic reduction of vanillic acid, demanding the addition of isolated carboxylic acid reductase and costly cofactors such as ATP, NADPH, and Mg2+, and the generation of isovanillin as a contaminating side product.
In this study, we have genetically engineered single-recombination microorganisms to synthesize vanillin from glucose, according to the metabolic route depicted in Fig. Fig.1.1. To avoid the synthesis of isovanillin as an undesired side product, a large array of OMTs was screened for the desired high substrate specificity, and an appropriate enzyme was identified. A synthetic version of an aromatic carboxylic acid reductase (ACAR) gene, optimized for yeast codon usage, was introduced to achieve the reduction step. The vanillin pathway was introduced into both Saccharomyces cerevisiae and Schizosaccharomyces pombe yeast, and significant levels of vanillin production were obtained in both organisms. Vanillin β-d-glucoside is the form in which vanillin accumulates and is stored in the fresh pod of the vanilla orchid (Vanilla planifolia). During the “curing” process of the pod, β-glucosidases are liberated and facilitate a partial conversion of the vanillin β-d-glucoside into vanillin. Upon consumption or application, the conversion of vanillin β-d-glucoside into free vanillin by enzymes in the saliva or in the skin microflora can provide for a slow-release effect that prolongs and augments the sensory event, as is the case for other flavor glycosides investigated, such as menthol glucoside (14, 16). In addition to the increased value of vanillin β-d-glucoside as an aroma or flavor compound, production of the glucoside in yeast may offer several advantages. Vanillin β-d-glucoside is more water soluble than vanillin, but most importantly, compounds such as vanillin in high concentrations are toxic to many living cells (4). It has been shown that glucosides of toxic compounds are less toxic to yeasts (24). We found this to be the case with vanillin and S. cerevisiae yeast as well. Thus, to facilitate storage and accumulation of higher vanillin yields, we introduced a step for vanillin glucosylation in S. pombe.
The 1,104-bp gene sequence of the Podospora pauciseta 3-dehydroshikimate dehydratase (3DSD) gene has no introns and was PCR amplified from genomic P. pauciseta DNA with flanking XbaI and BamHI restriction sites. The isolated PCR product was subcloned into the pCR-Blunt II-TOPO vector (Invitrogen Corp.), and the sequence-verified gene was inserted in pJH606, a proprietary S. pombe expression vector containing the S. pombe leu1+ selection marker and the adh1+ promoter. The resulting plasmid was named pJH643. The Nocardia sp. ACAR gene was synthesized with S. pombe codon optimization (to match as closely as possible the average codon usage as defined by all S. pombe sequences present in the NCBI-GenBank database) and flanking XbaI and BamHI sites (GENEART GmbH, Germany) and was inserted in the proprietary S. pombe expression vector pSP-Ex-Kan. This vector contains the KanMX selection marker, conferring resistance to the drug G418, and the S. pombe adh1+ promoter for gene expression. The resulting plasmid was named pJH573. The Ms-OMT and Hs-OMT genes were synthesized with S. pombe codon optimization (to match as closely as possible the average codon usage as defined by all S. pombe sequences present in the NCBI-GenBank database) and flanking XbaI and BamHI sites (GENEART GmbH, Germany). All other OMT-encoding genes were amplified by PCR from cDNA libraries (Stratagene Inc.) or cDNA clones (Cc-OMT, courtesy of Mary O'Connell, and Fa-OMT, courtesy of Stefan Lunkenbein) using primers containing flanking XbaI and BamHI sites. After being cloned into pCR-Blunt II-TOPO and sequence verification, the genes were transferred with XbaI and BamHI restriction sites into the proprietary S. pombe expression vector pJH609. This vector contains the HphMX selection marker, conferring resistance to hygromycin B, and the S. pombe adh1+ promoter for gene expression. The expression plasmids constructed were named as follows: pJH620 (Hs-OMT), pJH622 (Ms-OMT), pJH623 (Cc-OMT), pJH624 (At-OMT), pJH625 (Nt-OMT-a1), pJH627 (Nt-OMT-b1), and pJH628 (Fa-OMT). UGT71C2, UGT72B1, and UGT72E2 were all PCR amplified from proprietary Arabidopsis thaliana clones (C. Kristensen, E. H. Hansen, T. H. Andersen, G. Kock, F. T. Okkels, B. L. Møller, and J. Hansen, unpublished data) with appropriate flanking restriction sites for insertion in the proprietary S. pombe expression vector pJH610 (identical to pJH606 except the leu1+ marker is exchanged with a NatMX [nourseothricin resistance] marker). The resulting plasmids were pJH632 (UGT71C2), pJH633 (UGT72B1), and pJH665 (UGT71E2). For expression in S. cerevisiae, the 3DSD gene was inserted with XbaI-BamHI in a proprietary derivate of plasmid pYC070 (12), containing the strong constitutive S. cerevisiae TPI1 promoter and terminator and the AurC-R (aureobasidin A resistance) selection marker. This resulted in plasmid pJH500. The Hs-OMT gene was likewise inserted with XbaI-BamHI into a similar expression vector derived from pYC050 (12) (containing the NatMX selection marker), resulting in plasmid pJH543. The ACAR gene was inserted with XbaI-BamHI into a similar derivative of plasmid pYC040 (12) (containing the HpHMX selection marker), resulting in plasmid pJH674. Finally, most PPTase genes were obtained by PCR amplification of genomic DNA from E. coli, Bacillus subtilis, Mycobacterium bovis, and Corynebacterium glutamicum, while the Nocardia farcinica gene was obtained as a synthetic gene construct optimized for S. cerevisiae codon usage (GENEART GmbH, Germany). In all cases, the genes contained flanking XbaI-BamHI or XbaI-BglII (E. coli acpS) and were inserted in the XbaI-BamHI sites of the proprietary yeast shuttle (CEN-ARS replication) expression vector pJH259 containing the TPI1 promoter and terminator and the URA3 selection marker. This resulted in plasmids pJH587 to pJH596 and pJH701. All PCRs were performed using a Peltier thermal cycler DNA engine DYAD PCR machine, with an initial preheating at 94°C for 2 min and a final 7-min elongation step at the selected elongation temperature. Pwo polymerase (Roche Biochemicals) was used for all reactions. All plasmids used or constructed are listed in Table Table11.
The 3DSD gene expression cassette was transformed into S. pombe strain SP887 as a linearized plasmid, pJH643, with integration directed to the leu1+ locus. A leucine prototrophic transformant was isolated and denoted strain FSC264, and after confirmation of its ability to produce protocatechuic acid, it was kept as strain VAN264. The ACAR gene expression cassette was transformed into strain VAN264 as linearized plasmid pJH573, with integration directed to the adh1+ promoter region. Eight G418-resistant transformants were selected, and the one with the highest total production of protocatechuic acid and aldehyde was kept as strain VAN244. All plasmids containing expression cassettes for OMTs were transformed into strain VAN244 after linearization to direct integration to the adh1+ promoter region. Two hygromycin B-resistant transformants of each type were tested for production of vanillin pathway metabolites, and the one with the highest vanillin production was kept as strain VAN294 (Hs-OMT), VAN298 (At-OMT), or VAN302 (Fa-OMT). The UDP-glycosyltransferase (UGT)-containing plasmids pJH632 (UGT71C2), pJH633 (UGT72B1), and pJH665 (UGT71E2) were all linearized in order to direct integration to the adh1+ promoter region, and strain VAN294 was transformed with the plasmid preparations. One stable nourseothricin-resistant transformant of each type was kept as strains VAN512 (UGT71C2), VAN513 (UGT72B1), and VAN515 (UGT72E2). Plasmid pJH500 was linearized with Bsu36I in order to direct integration to the TPI1 promoter region, and S. cerevisiae strain VAN100 (adh6 bgl2) was transformed with the plasmid preparation. One PCR-confirmed, aureobasidin A-resistant transformant was kept as strain VAN265. This strain was transformed with Bsu36I-linearized plasmid pJH543. One PCR-reconfirmed, nourseothricin-resistant transformant was kept as strain VAN277. Strain VAN277 was transformed with Bsu36I-linearized plasmid pJH674, and one PCR-reconfirmed, hygromycin B-resistant transformant was kept as strain VAN286. All yeast strains used or created in this study are listed in Table Table2.2. S. pombe and S. cerevisiae were transformed with plasmid DNA using the respective lithium acetate methods for these two organisms (7, 29), and the proper insertion of all expression cassettes at the desired genomic location was confirmed by analytical PCR on genomic material from the various yeast strains.
Yeast strains were in all cases grown at 25°C with 170-rpm shaking in appropriate growth media (synthetic complete [SC] or yeast extract-peptone-dextrose [YPD] for S. cerevisiae strains, yeast extract with supplements [YES] for S. pombe strains), after inoculation from precultures grown under the same conditions. No precautions were taken to avoid the presence of aromatic amino acids in these growth media, which potentially could limit dehydroshikimic acid biosynthesis. Growth media were in all cases obtained from Q-BioGene, Montreal, Canada. To analyze the rate of turnover of vanillin and vanillin β-d-glucoside in the yeast cultures, these compounds were supplied in final concentrations of 1 mM from 1 M stock solutions in ethanol (the yeast strains analyzed are listed in Table Table2).2). For metabolite analysis, the ferment of growth culture samples was separated from the yeast cells by centrifugation. Ferment (500 μl) was then combined with 500 μl of 100% methanol and centrifuged (16,100 × g, 12 min) to precipitate macromolecules. Aliquots (25 μl) were analyzed by high-performance liquid chromatography (HPLC) as described below.
A preculture of S. cerevisiae strain VAN100 in SC medium (optical density at 600 nm [OD600] of approximately 2.5) was diluted into five equal batches (100 ml) of the same medium (OD600 of 0.4, 250-ml Ehrlenmeyer flasks). Vanillin was added to the flasks in final concentrations of 0.5, 1.0, and 5 g/liter. Vanillin β-d-glucoside was added in a final concentration of 25 g/liter to a fourth culture, while a fifth culture to which neither vanillin nor vanillin β-d-glucoside was added was used as a control. The cultures were grown at 25°C with 170-rpm shaking, and the OD600 was measured after 5.5, 9.5, and 23 h.
Vanillin was extracted from supernatants of large-scale yeast cultures using CH2Cl2 in three serial extractions (333 ml per 1 liter of supernatant). The extract was concentrated in a rotary evaporator, the residue was resuspended in toluene, and the suspension was applied to a silica gel column, which was eluted with 30% ethyl acetate in pentane. The fractions containing vanillin, as monitored by thin-layer chromatography, and their UV fluorescence were combined and concentrated by drying in a rotary evaporator.
Intermediates in vanillin biosynthesis and vanillin catabolites were analyzed using an Agilent 1100 series HPLC system using a Zorbax SB-C18 column (4.6 by 150 mm, 3.5-μm particle size). The elution buffer was a gradient of acetonitrile (MeCN) and H2O (adjusted to pH 2.3 with H2SO4) composed as follows: 0% to 40% MeCN for 3 min, 40% MeCN for 1 min, 40% to 80% MeCN for 2 min, and 80% to 90% MeCN for 1 min. The temperature of the solvent was thermostated at 30°C, and a diode array detector was used to detect eluted compounds by their UV fluorescence at 210 nm and 250 nm. Vanillin, protocatechuic acid, protocatechuic aldehyde, vanillic acid, and vanillyl alcohol standards were obtained from Merck Chemical Co. Vanillin β-d-glucoside was obtained from Apin Chemicals Ltd., United Kingdom.
Nuclear magnetic resonance (NMR) spectra were recorded in deuterated chloroform on a Bruker Avance 400 instrument using tetramethyl silane as an internal standard. The 1H spectrum exhibited the following signals: 9.82 ppm (CHO), multiplets at 7.43 (2H) and 7.04 (1H) (aromatic protons), and 3.95 ppm (CH3O). The 13C spectrum showed signals at 191.0 (CHO), 151.8, 147.3, 129.9, 127.5, 114.5, and 108.9 (aromatic carbons) and 56.1 ppm (CH3O). The 1H and 13C spectra were identical to those of authentic vanillin and clearly different from those of isovanillin, which among other signals had 13C signals at 124.5 and 110.2 ppm and a multiplet at 6.98 (1H) in the 1H spectrum.
The production organism was chosen based on the evaluation of several parameters: (i) GRAS (“generally regarded as safe”) recognition, (ii) proven suitability in at least one established production system, (iii) reasonably well developed genetic tools available, and (iv) inherent vanillin metabolism that is as low as possible. From a genetic point of view, the most obvious candidates were strains of baker's yeast (Saccharomyces cerevisiae) and Escherichia coli. These are GRAS organisms and constitute well-known production systems, their genome sequences are available, and genetic manipulation is relatively straightforward. From a consumer acceptance point of view, S. cerevisiae would appear to be the best choice. However, a growing culture of S. cerevisiae (laboratory strain X2180-1A) quantitatively reduced externally added vanillin (1 mM) to vanillyl alcohol within 48 h (data not shown). This prompted us to test a range of different yeast species of the genus Saccharomyces, along with strains of Zygosaccharomyces fermentatii, Zygosaccharomyces bisporus, Debaromyces occidentalis, Torulaspora delbrueckii, Kluyveromyces lactis, Pichia pastoris, and Schizosaccharomyces pombe (Table (Table2).2). Schizosaccharomyces pombe was by far the most satisfactory, since after 48 h it had reduced less than 50% of the vanillin provided and oxidized none (data not shown), whereas all other strains tested converted all vanillin to either vanillyl alcohol or vanillic acid within the same period of time. In a similar manner, we tested hydrolysis of vanillin β-d-glucoside by S. pombe and S. cerevisiae. While S. pombe left vanillin β-d-glucoside intact even after prolonged incubation, S. cerevisiae hydrolyzed all vanillin β-d-glucoside within 24 h (data not shown). This in turn prompted us to test S. cerevisiae mutants of known β-glucosidase genes (ALF2, BGL1, BGL2, DSE2, DSE4, EXG2, KRE6, SCW10, SCW11, SCW4, SKN1, SPR1, SUN4, and the homologous gene YOL155C; mutants were obtained from the Euroscarf collection). One mutant, the bgl1 strain, hydrolyzed less than 5% of the vanillin β-d-glucoside present, while all other mutants had the same activity as the wild-type yeast (data not shown). Finally, we tested whether S. cerevisiae mutants in any of the 29 known or hypothesized alcohol dehydrogenases, aryl-alcohol dehydrogenases, or the related aldose reductases (AAD3, AAD4, AAD6, AAD10, AAD14, ADH1, ADH2, ADH3, ADH4, ADH5, ADH6, ADH7, ARA1, ARA2, BDH1, BDH2, GCY1, GRE3, SFA1, XYL2, YPR1, ZTA1, YCR102c, YDL124w, YJR096w, YLR460c, YNL134c, YPL088w, and YPR127w; mutants were obtained from the Euroscarf collection) had a reduced ability to convert vanillin into vanillyl alcohol. The screen identified ADH6 as the most important gene encoding a vanillin reductase (data not shown). Consequently, we bred an adh6 mutant of S. cerevisiae bgl1 strain Y05210 (Euroscarf), strain VAN100 (Table (Table2).2). This strain grew normally under all circumstances tested, hydrolyzed vanillin β-d-glucoside to only a very limited extent, and showed a 50%-decreased ability to reduce vanillin to vanillyl alcohol. Thus, we decided to test vanillin biosynthesis in a wild-type S. pombe yeast and in the bgl1 adh6 mutant of S. cerevisiae.
3DSD catalyzes the conversion of 3-dehydroshikimic acid to protocatechuic acid. This enzyme activity is known from filamentous fungi (40), so we isolated the gene encoding this enzyme from the dung mold Podospora pauciseta. The gene was PCR isolated from genomic DNA and transformed into S. pombe strain SP887 on the linearized pJH643 S. pombe expression plasmid. One transformant, denoted strain VAN264 (Table (Table2),2), was isolated and tested for its ability to produce protocatechuic acid by growing a batch culture (5 ml) for 48 h, after which the supernatant was analyzed by HPLC. A new compound eluting at 5.4 min was identified as protocatechuic acid based on its coelution with authentic protochatecuic acid and an identical absorption spectrum. The production of protocatechuic acid reached more than 360 mg/liter (Table (Table33).
ACARs (EC 18.104.22.168) catalyze the ATP-driven reduction of protocatechuic acid to protocatechuic aldehyde. Bacteria of the Nocardia genus as well as filamentous and ligninolytic fungi are known to possess this enzyme activity (9, 11, 21), and a method to reduce vanillic acid to vanillin using purified Nocardia ACAR enzyme was devised by Rosazza and Li (39). The corresponding 3.5-kb ACAR gene has been isolated, and a recombinant E. coli strain expressing the enzyme bioconverts vanillic acid to vanillin (13). The codon GC content in the Nocardia genus is around 70%, while it is a mere 40% in S. pombe. To optimize expression, a synthetic version of the gene was built based on S. pombe codon usage and transformed into S. pombe strain VAN264 on the linearized expression plasmid pJH573. Eight transformants were grown in batch cultures (5 ml) for 48 h, the cells were removed by centrifugation, and the supernatant was analyzed by HPLC. In addition to protocatechuic acid, a new constituent was found to elute at 5.8 min and was identified as protocatechuic aldehyde, based on coelution with an authentic standard and spectral analysis. The transformant with the highest total production of protocatechuic acid plus protocatechuic aldehyde afforded 300 mg/liter and was kept as strain VAN244 (Table (Table2).2). This strain converted 53% of the formed protocatechuic acid into protocatechuic aldehyde (Table (Table33).
Two OMTs, from alfalfa (Medicago sativa) and strawberry (Fragaria × ananassa) (Ms-OMT and Fa-OMT) (5, 47), were reported to catalyze 3′-OH position-specific methylation of protocatechuic aldehyde. Based on the sequence information for these genes, similar OMT genes from Capsicum chinense (Cc-OMT), Arabidopsis thaliana (At-OMT), and Nicotiana tabacum (Nt-OMT-a1 and -b1) were isolated. All of the genes encoding these proteins are approximately 1,100 bp. A different class of methyltransferase-encoding genes of approximately 700 bp, widespread in animals, is annotated as catechol methyltransferase. For comparative purposes, we expanded the screen with a human (Homo sapiens) catechol methyltransferase (Hs-OMT) (18). The OMT-encoding genes were PCR amplified from cDNA or synthesized with S. pombe codon optimization (Ms-OMT and Hs-OMT) and transformed into S. pombe strain VAN244 as linearized plasmids pJH620 (Hs-OMT), pJH622 (Ms-OMT), pJH623 (Cc-OMT), pJH624 (At-OMT), pJH625 (Nt-OMT-a1), pJH627 (Nt-OMT-b1), and pJH628 (Fa-OMT) (Table (Table1).1). Two of each type of transformant were grown in batch cultures (5 ml) for 48 h, and the supernatants were analyzed by HPLC. Only expression of Hs-OMT, At-OMT, and Fa-OMT resulted in in vivo methylation, measured as the accumulation of vanillic acid (elution time, 5.9 min) and/or vanillin (elution time, 6.6 min) and confirmed by comparison of the elution profile and absorbance with authentic standards. One strain expressing each of these OMTs was kept: VAN294 (Hs-OMT), VAN298 (At-OMT), and VAN302 (Fa-OMT). The three OMTs afforded quite different product profiles (Fig. (Fig.22 and Table Table3).3). VAN298, carrying At-OMT, produced the smallest amount of vanillin, despite the fact that the level of the precursors protocatechuic acid and protocatechuic aldehyde were the highest in this strain. VAN302, carrying Fa-OMT, produced nearly twice as much vanillin. VAN294, expressing the human catechol methyltransferase (Hs-OMT), was by far the most efficient enzyme and more than tripled the amount of vanillin made by VAN298. VAN294 also produced vanillyl alcohol (elution time, 5.5 min) and vanillic acid (elution time, 6.2 min). Because of the singularly high vanillin formation in the VAN294 strain, harboring expression cassettes for 3DSD, ACAR, and Hs-OMT, this strain was chosen for vanillin production.
Small-scale vanillin production was performed using strain VAN294. Cultures (four at 3 liters each) were started from precultures (OD600 of 0.04) in rich medium and allowed to grow for 48 h (vanillin production ceased after 45 h). Vanillin content in the four culture flasks varied between 21 mg/liter and 31 mg/liter, corresponding to a total production of approximately 300 mg of vanillin in the 12 liters of culture. Invariably, a reduced yield of vanillin was observed when the culture volume was increased. Currently, the reduced yield cannot be related to specific growth parameters. Extraction of the cleared culture supernatant with CH2Cl2 (as described in Materials and Methods) afforded approximately 200 mg of vanillin as white powder. The isolated vanillin showed an HPLC elution time and UV spectrum indistinguishable from those of a vanillin standard and an NMR spectrum identical to that of authentic vanillin (NMR signals reported in Materials and Methods). The NMR analysis documented that no isovanillin (3-hydroxy-4-methoxybenzaldehyde) was present.
The successful design of a de novo pathway for vanillin biosynthesis in S. pombe prompted us to investigate the possibility of converting the vanillin formed into vanillin β-d-glucoside. This set of experiments was further accentuated by the observation that the glucosylated form of vanillin was less toxic to yeast than vanillin. The growth-inhibitory effects of the two compounds were tested using the S. cerevisiae strain VAN100 (Fig. (Fig.3).3). Whereas vanillin was toxic at a concentration of less than 0.5 g/liter, as monitored by growth inhibition, vanillin β-d-glucoside was nontoxic even at 25 g/liter. The reduced toxicity of vanillin β-d-glucoside in comparison to vanillin was not caused by an inability of the yeast cells to take up vanillin β-d-glucoside, as demonstrated by analysis of the intracellular content of vanillin β-d-glucoside after 48 h of growth in the presence of 10 or 25 g/liter. In both experiments, the intracellular concentration of vanillin β-d-glucoside was approximately twice that found in the growth supernatant (data not shown). Accordingly, we conclude that vanillin β-d-glucoside is truly nontoxic to S. cerevisiae even at high concentrations. Plant family 1 glycosyltransferases are involved in the glycosylation of bioactive plant natural products. They belong to a group of glycosyltransferases often referred to as the UGTs, because they transfer sugar moieties (most often glucose) from UDP-bound sugars to low-molecular-mass aglycons (30, 33). To provide a platform for glycosylation of bioactive aglycons, we cloned and heterologously expressed 98 UGT enzymes from the plant Arabidopsis thaliana along with a few from other plant sources (Kristensen et al., unpublished). Following expression in the yeast Pichia pastoris, we tested crude enzyme preparations for their ability to catalyze in vitro glucosylation of vanillin. Seven UGTs were identified as possessing particularly high in vitro catalytic activity toward vanillin, namely, UGT71C2, UGT72B1, UGT72E2, UGT84A2, and UGT89B1 from A. thaliana, UGT85B1 from Sorghum bicolor (17), and arbutin synthase from Rauwolfia serpentina (2). Of these seven enzymes, the first three exhibited the highest affinity for vanillin. The genes encoding these UGTs were inserted into the S. pombe vanillin producer to examine their in vivo functions. The UGT-encoding genes were combined with the TPI1 promoter in the expression plasmids pJH632 (UGT71C2), pJH633 (UGT72B1), and pJH665 (UGT72E2) (Table (Table1),1), and each was integrated into the adh1+ locus of strain VAN294. The three resulting strains (VAN512, VAN513, and VAN515) were tested by growing them for 48 h in 100 ml of YES medium in Ehrlenmeyer flasks. Strain VAN515, harboring UGT72E2, was by far the most efficient in vivo vanillin glucosyltransferase. Figure Figure44 shows the results of the ensuing HPLC analyses of the ferment from growth of the VAN515 strain and the control strain VAN294. Production of vanillin β-d-glucoside was verified by the elution time of 5.3 min and NMR and UV/visible light spectral identity with a vanillin β-d-glucoside standard. With strain VAN515, a total of 56% of the total vanillin potential (i.e., the sum of formed vanillin and its precursors protocatechuic acid, protocatechuic aldehyde, and vanillic acid) was transformed into vanillin glucoside. Interestingly, while about 80% of the vanillin was glucosylated, only half of the vanillyl alcohol was, confirming a much higher affinity of UGT72E2 for vanillin than for vanillyl alcohol.
The vanillin engineering studies reported above were accomplished with Schizosaccharomyces pombe as the host. Because S. cerevisiae is the more commonly used “workhorse” for metabolic engineering and production, a parallel study was performed in an attempt to construct an S. cerevisiae strain that would also produce vanillin. As in the studies with S. pombe, the P. pauciseta 3DSD, the synthetic Nocardia ACAR, and the synthetic human Hs-OMT genes were all inserted in proprietary S. cerevisiae expression cassettes, in all cases making use of the strong glycolytic TPI1 gene promoter (resulting in integration plasmids pJH500, pJH543, and pJH674 [Table [Table1]).1]). The expression cassettes were sequentially inserted into the endogenous TPI1 locus of strain VAN100, directing insertion by the linearization of plasmids in the TPI1 promoter sequence, resulting in S. cerevisiae strain VAN286 (Table (Table2).2). After growth of this strain in batch cultures (5 ml) with SC medium for 48 h, the clarified medium was found to contain the vanillin precursors protocatechuic acid and vanillic acid. However, none of the corresponding aldehydes, including vanillin, was detected. This indicated that the ACAR enzyme was not expressed or not functional in S. cerevisiae. ACARs as well as the related nonribosomal peptide synthetases, fatty acid synthetases, and polyketide synthetases require specific phosphopantetheinylation for functionality (13, 44). Obviously, an endogenous activity mediating phosphopantetheinylation of ACAR proceeded in the S. pombe strain, whereas this activity was absent in S. cerevisiae. Consequently, we cloned phosphopantetheine transferases from Bacillus subtilis (acpS and sfp), E. coli (acpS, acpT, entD, and a homologue, PPTec-1), Mycobacterium bovis (acpS and a pptT homologue), and Corynebacterium glutamicum (acpS and PPTcg-1), as well as a homologue from Nocardia farcinica (PPTnf-1, a synthetic gene optimized for S. cerevisiae codon usage), and expressed these in strain VAN286 from low-copy-number-replicating plasmids (CEN-ARS) and the yeast TPI1 promoter. The M. bovis genes were included because Mycobacterium is a genus closely related to Nocardia, the source of the ACAR gene. Expression of three of the genes, the E. coli entD, the C. glutamicum PPTcg-1, and the N. farcinica PPTnf-1 gene in strain VAN286 (thus harboring either plasmid pJH589, pJH592, or pJH701), resulted in a functional ACAR enzyme and the identification of protocatechuic aldehyde as well as vanillin in the clarified fermentation broth. PPTcg-1 was the most efficient PPTase for activation of the ACAR gene and resulted in formation of 45 mg/liter of vanillin after 48 h of growth in SC medium (Table (Table3).3). Thus, the three-step biosynthesis pathway for de novo vanillin biosynthesis already established in S. pombe is just as efficient in S. cerevisiae, but in contrast to the situation in S. pombe, a heterologous PPTase enzyme is needed for activation, by phosphopantetheinylation, of the ACAR gene in S. cerevisiae.
In this study, we demonstrate complete de novo vanillin production outside the Vanilla planifolia seed pod or other plants. This represents the first example of one-cell microbial generation of this valuable compound from glucose, at a production level scalable to industrial needs. The capability for vanillin biosynthesis was introduced into two common yeast species, Schizosaccharomyces pombe and Saccharomyces cerevisiae. The heterologous pathway for vanillin biosynthesis was engineered in both organisms by the expression of three genes, one from a mold, one from a bacterium, and one of human origin, and in the case of S. cerevisiae, one additional bacterial gene. We obtained a vanillin production of 65 and 45 mg/liter in S. pombe and S. cerevisiae, respectively, free of contaminating isomers, without any specific optimization of media and growth conditions. Although vanillin biosynthesis was less efficient in S. cerevisiae than in S. pombe, our data actually indicate a higher vanillin production potential in S. cerevisiae, since the combined production of vanillin and its precursors and metabolites was almost twice as high with S. cerevisiae as with S. pombe (Table (Table3).3). The accumulated levels of the various metabolites indicate that more dehydroshikimic acid is converted to protecatechuic acid in our S. cerevisiae experiment but also that about the same proportion of this (70% for S. cerevisiae, 75% for S. pombe) is reduced by the introduced ACAR enzyme. The reason for the lower production of vanillin in S. cerevisiae is a higher ability of this organism to reduce vanillin to its corresponding alcohol. This undesired property of S. cerevisiae became obvious at the beginning of the project and was addressed by inactivation of the ADH6-encoded alcohol dehydrogenase. In the set of experiments undertaken to identify the importance of different alcohol dehydrogenases in vanillin reduction, a modest effect of inactivation of several other genes (e.g., ADH7) was registered, and it is likely that inactivation of additional alcohol dehydrogenases in the S. cerevisiae vanillin producer would result in a significant increase in vanillin production.
The observation that nearly identical proportions of the biosynthesized protocatechuic acid were reduced by both yeast strains demonstrates that introduction of the C. glutamicum PPTase gene in our S. cerevisiae vanillin producer resulted in an activation of the ACAR enzyme to the same level as that seen in S. pombe. It is indeed puzzling that bacterial ACAR can be activated by inherent enzymes in one yeast but not in another. Enzymes requiring phosphopantetheinylation for activation are not abundant in these yeast species, but one well-known example present in both is α-aminoadipate reductase. Both species carry a known PPTase activity taking care of this (Lys5p in S. cerevisiae, Lys7p in S. pombe), and these are obvious candidates for heterologous ACAR activation (though another could be the PPTase activating mitochondrial fatty acid synthase). A plausible explanation for the differences in PPTase activity in the two yeasts is derived from the following observations (10). Whereas S. pombe α-aminoadipate synthase can be activated by PPTases present in E. coli, this is not the case for α-aminoadipate synthase from Candida albicans. The C. albicans enzyme is much more closely related to the S. cerevisiae enzyme than to the S. pombe enzyme. Turning the argument around, this may imply that S. pombe (via its lys7+-encoded PPTase), but not S. cerevisiae, has the inherent ability to activate the bacterial ACAR enzyme. Not surprisingly, a PPTase from Corynebacterium glutamicum, a high-GC, gram-positive bacterium related to Nocardia sp., turned out to be the most efficient in ACAR activation.
As previously outlined, vanillin β-d-glucoside is the storage form of vanillin found in the Vanilla pod. It is nontoxic to most organisms, including yeast, and has a higher solubility in water than does vanillin. In addition, the formation of vanillin β-d-glucoside most likely pulls the biosynthesis further in the direction of vanillin production. The Arabidopsis thaliana UDP-glucose glycosyltransferase UGT72E2 exhibited high substrate specificity toward vanillin. In concordance with this observation, its expression in the vanillin-producing S. pombe strain resulted in almost all vanillin being converted into vanillin β-d-glucoside. The ability to turn vanillin into vanillin β-d-glucoside in vivo is very important, because microbial production of nonglucosylated vanillin beyond the 0.5- to 1-g/liter scale would be hampered by the toxicity of free vanillin. Glucosylation would serve to circumvent the inhibitory effect. Although glucosylation did not give rise to a major increase in vanillin production, the content of nonmethylated intermediates (protocatechuic acid and aldehyde) was reduced by more than 50% (Fig. (Fig.4).4). This indicates that glucosylation does indeed drive production of methylated vanillin equivalents, but that only a certain amount of dehydroshikimic acid is available during the period of time when our introduced vanillin pathway is active. There could be many reasons for this and we are currently studying several possibilities.
“Sustainable” and “renewable” biological production systems are attracting a lot of attention these days, due to the global warming issue and associated interest in developing a chemical industry that is independent of fossil fuel starting materials; thus, “white biotechnology” is having a tremendous comeback. S. cerevisiae is a very attractive production organism in white biotechnology, because this yeast species is well characterized, is easy to manipulate and grow, and has gained GRAS status. Metabolic engineering of S. cerevisiae has resulted in very high yields of certain primary yeast metabolites, e.g., 153 g/liter of pyruvate (43), but de novo productivities of novel metabolites have usually been quite modest, ranging from 153 mg/liter (the terpenoid amorphadiene ) to only just detectable amounts (e.g., the polyketide precursor methylmalonyl-coenzyme A ) (reviewed in reference 28). To our knowledge, our study is the first in which aromatic amino acid biosynthesis intermediates are used for production of a novel compound, and in that perspective, we find our initial productivity of 45 mg/liter satisfactory. We are aware, however, that even though the market prices for “natural” vanillin and for vanillin-β-d-glucoside are high, the biological production system presented here needs to be improved significantly to offer a truly sustainable alternative. It was recently shown that simple genetic modifications may increase the metabolic flux through the S. cerevisiae aromatic amino acid biosynthesis pathway 4.5-fold and the extracellular concentration of shikimic acid (the direct metabolite of dehydroshikimic acid) more than 200-fold (22). This provides obvious opportunities for significant future increases in vanillin production using yeasts as production organisms.
M. O'Connell (New Mexico State University), S. Lunkenbein (Technical University München), and M. Ruppert (Johannes Gutenberg University, Mainz) are sincerely thanked for providing cDNA clones of methyltransferase genes. Anders Bøgh Jensen and Thomas Hvid Andersen are thanked for valuable discussions, and Christine Tachibana is acknowledged for reading of the manuscript.
Published ahead of print on 13 March 2009.