High-Yield Resveratrol Production in Engineered ...homepages.rpi.edu/~koffam/papers/2011_Lim_Fowler_Hueller.pdf · High-Yield Resveratrol Production in Engineered Escherichia coli!

APPLIED AND ENVIRONMENTAL MICROBIOLOGY, May 2011, p. 3451–3460 Vol. 77, No. 100099-2240/11/$12.00 doi:10.1128/AEM.02186-10Copyright © 2011, American Society for Microbiology. All Rights Reserved.

High-Yield Resveratrol Production in Engineered Escherichia coli!†Chin Giaw Lim,1 Zachary L. Fowler,1 Thomas Hueller,2 Steffen Schaffer,2 and Mattheos A. G. Koffas1*Department of Chemical and Biological Engineering, University at Buffalo, The State University of New York, Buffalo, New York 14260,1

and Evonik Degussa GmbH, Creavis Technologies & Innovation, Paul-Baumann-Strasse 1, D-45772 Marl, Germany2

Received 14 September 2010/Accepted 13 March 2011

Plant polyphenols have been the subject of several recent scientific investigations since many of themolecules in this class have been found to be highly active in the human body, with a plethora of health-promoting activities against a variety of diseases, including heart disease, diabetes, and cancer, and with eventhe potential to slow aging. Further development of these potent natural therapeutics hinges on the formationof robust industrial production platforms designed using specifically selected as well as engineered proteinsources along with the construction of optimal expression platforms. In this work, we first report the inves-tigation of various stilbene synthases from an array of plant species considering structure-activity relation-ships, their expression efficiency in microorganisms, and their ability to synthesize resveratrol. Second, welooked into the construct environment of recombinantly expressed stilbene synthases, including differentpromoters, construct designs, and host strains, to create an Escherichia coli strain capable of producingsuperior resveratrol titers sufficient for commercial usage. Further improvement of metabolic capabilities ofthe recombinant strain aimed at improving the intracellular malonyl-coenzyme A pool, a resveratrol precursor,resulted in a final improved titer of 2.3 g/liter resveratrol.

Resveratrol (3,5,4!-trihydroxy-trans-stilbene) is a polypheno-lic compound that belongs to the stilbene class and is com-monly found in red wine (up to 6.8 mg/liter), bushberries,peanuts, cranberries, other vine plants, and even trees andferns. Resveratrol is speculated to be responsible for a de-creased risk of heart disease and diabetes. Often called the“French paradox,” the high intake of saturated fat, affected bylevels of radical scavengers such as resveratrol, has been shownto correlate with a low mortality rate (8, 28). Yet, resveratrolmay have additional health benefits, as numerous studies havefound that its biological activities include antioxidative, anti-inflammatory, anticancer, and chemopreventive abilities (6, 10,14). However, the most interesting activity attributed to res-veratrol may be its potential to slow the aging process andprolong life spans, as seen in a variety of evolutionarily distantspecies, including Saccharomyces cerevisiae, Caenorhabditis el-egans, Drosophila melanogaster, mice, and vertebrate fish (2, 13,30, 35). Nevertheless, there are contradicting results statingthat the life span elongation is not due to resveratrol itself butdue to a synergistic action of resveratrol with a fluorophore-signaling molecule (5, 24).

The universal polyphenol biosynthetic pathway begins withthe formation of phenylpropanoic acids from the aromaticamino acid phenylalanine or tyrosine or both. The phenylpro-panoate is then activated by ligation to coenzyme A (CoA) bycoumarate:CoA ligase, often acting only in the presence of a4!-hydroxy group on the phenyl ring and thus termed 4-coumaroyl:CoA ligase (4CL). Polyphenols are then diversi-fied by the type III polyketide synthases (PKSs), such as stil-

bene synthase (STS) for stilbene (i.e., resveratrol) synthesisand chalcone synthase (CHS) for flavonoid synthesis. In thepathway-committing step, PKSs act to condense successiveunits of malonyl-CoA with coumaroyl-CoA, forming a linearpolyketide molecule before a cyclization reaction is carried outby the same PKS. For STS, 3 units of malonyl-CoA are used inthe chain lengthening and a C2 to C7 aldol cyclization is usedto form stilbenes (Fig. 1), while flavonoids are formed via a C6to C1 Claisen reaction by CHS.

Two significant aspects of natural resveratrol formationhamper its application as a widespread nutraceutical. First,even in plants possessing the most abundant levels of resvera-trol, this compound is produced in only trace amounts. Forinstance, peanuts and grapes contain no more than 4 "g/g ofdry plant matter and red wines contain an average of 2 mg/liter,yet the antiaging effects seen in mice are achievable only withbiologically active dosages nearly 50 times that found in wine(3, 27). Second, the biologically active form of resveratrol is thetrans isomer. However, isolation from plants, as for many nat-ural products, yields a mixture of multiple isomers and substi-tuted forms, including cis, trans, and various less-active glyco-sylated forms (3, 18). While recently transgenic stilbeneproduction has been achieved in commonly consumed plantspecies (7, 12), titers have remained low, making the biosyn-thesis of resveratrol an ideal target for microbial fermentation.This is because the method is easily scaled to commercialapplication, unlike methods using plant cell cultures, whichhave general limitations in cell development control, tissuespecificity, uniform inducibility, and promoter strength. Fur-thermore, bacterial fermentations overcome many of the chal-lenges associated with traditional chemical synthesis, specifi-cally, the use of a toxic catalyst and the significant formation ofby-products and various isomers at each synthetic step.

Previous efforts to engineer microorganisms for stilbene bio-synthesis have resulted in relatively low production titers (17,25, 34), and little effort has been made to “tune” expression

* Corresponding author. Mailing address: Department of Chemicaland Biological Engineering, University at Buffalo, The State Universityof New York, 904 Furnas Hall, Buffalo, NY 14260. Phone: (716)645-1198. Fax: (716) 645-3822. E-mail: [email protected].

† Supplemental material for this article may be found at http://aem.asm.org/.

! Published ahead of print on 25 March 2011.

3451

constructs or explore different STS proteins to improve pro-duction levels. In this study, we examine multiple constructs forresveratrol production by varying the “construct environment”through the use of different Escherichia coli strains (BL21 Starand BW27784), different promoter systems (specifically thestrong T7 promoter and the constitutive promoter of the glyc-eraldehyde-3-phosphate dehydrogenase [GAPDH] gene), dif-ferent gene expression combinations, and a library of isolated4CLs and STSs. Additionally, we describe a more detailedinvestigation of stilbene biosynthesis in microorganisms byconducting both sequence analysis and protein-structure anal-ysis as well as biochemical comparisons using a selection ofstilbene synthases to identify highly active recombinant en-zymes. With these optimization efforts, high production titersof resveratrol were achieved at grams per liter scale fromp-coumaric acid.

MATERIALS AND METHODS

Strains and media. E. coli BL21 Star (Invitrogen) and strain BW27784 (E. coliGenetic Stock Center, New Haven, CT) were used for plasmid cloning andrecombinant molecule production. DNA manipulations were performed accord-ing to standard recombinant DNA procedures (29). Restriction enzymes and T4DNA ligase were purchased from New England BioLabs. All PCR amplificationand cloning reactions were performed using Accuzyme DNA polymerase (Bio-line). Quantitative reverse transcription-PCR (qRT-PCR) was performed usingthe iScript with the SYBR green RT-PCR kit from Bio-Rad. Plasmid DNA wasprepared from stock strains using a Zyppy plasmid miniprep kit, while fragmentDNA was isolated by gel extraction using a Zymoclean gel DNA recovery kit(Zymo Research). Protein concentrations were measured using a bicinchoninicacid (BCA) assay kit (Pierce Scientific) with bovine serum albumin (BSA) as thestandard. Plasmid-bearing cultures were grown in media containing, whenneeded, ampicillin (70 "g/ml), kanamycin (40 "g/ml), chloramphenicol (20 "g/ml), and/or streptomycin (40 "g/ml). p-Coumaric acid, malonyl-CoA, coenzymeA, ATP, and the resveratrol standard were all purchased from Sigma-Aldrich,

while cerulenin was purchased from Cayman Chemical. Coumaroyl-CoA wasprepared according to the previously published method (32).

Sequence analysis. A BLAST search of the National Center of BiotechnologyInformation (NCBI) nucleotide data bank identified numerous sequences pre-dicted to encode stilbene synthases. From this search, the initial phylogenetictrees were constructed using MEGA4 to identify groupings of stilbene synthases.For both mRNA and protein sequences, minimum evolution trees were usedwith 500 bootstrap replicates and gamma distributed rates of change for each site(gamma parameter # 1.0). For mRNA, a Tamura 3-parameter model was usedwith heterologous rates of mutation, while protein trees used a Poisson modeland homogenous rates of change. A sample of stilbene synthases across the treewas then sent to be codon optimized and synthesized (Epoch Biolabs). Theaccession numbers for those sequences are as follows: Pinus strobus STS (PsSTS),Z46915; Pinus massoniana (PmSTS), DQ647829, Pinus densiflora STS (PdSTS),AB030140; Psilotum nudum STS (PnSTS), AB022685; Polygonum cuspidatumSTSs, DQ459350 (Pcu1STS) and EF090266 (Pcu3STS). In addition, the mRNAsfor the STS of Vitis vinifera (VvSTS) and Arachis hypogaea (AhSTS) were isolatedfrom seedlings using a Qiagen RNeasy isolation kit, converted to cDNA byRT-PCR using Taq polymerase (New England BioLabs), and then purified. ForAhSTS and VvSTS, the accession numbers for primer designs are A00769 andDQ459351, respectively. Homology modeling was performed using the Swiss-PdbViewer (Swiss Institute of Bioinformatics) (11), with images and furtheranalysis done using MacPyMol (DeLano Scientific).

Plasmid construction. Plasmids pACYCDuet-1 and pCDFDuet-1 were pur-chased from Novagen, while plasmid pUC18 containing the Arabidopsis thaliana4CL (At4CL) and VvSTS genes downstream of the gap promoter (pGAP) wasprovided by Evonik Degussa. Plasmids pA-ACC and pCO-BirA were providedby M. A. G. Koffas laboratory stocks. The construction of pCDF was done by firstcloning the Petroselinum crispum 4CL (Pc4CL) gene into the first multiple clon-ing site via EcoRI/SalI. Subsequently, STS genes were amplified and cloned viaeither BglII/PacI or NdeI/PacI. For amplification of the resveratrol biosynthesispathway in pUC18, STS genes were cloned between either XbaI/NotI (V. vinif-era), XbaI/PstI (A hypogaea), XbaI/NdeI (P. strobus), or XbaI/StuI (P. densifloraand P. massoniana) sites, while 4CL genes were cloned between two NotI sites.Only two STS genes were combined with the Pc4CL gene on one vector. Forconstruct pUCo-AhSTS-Pc4CL, the AhSTS gene was cloned between XbaI/PstIsites, while the Pc4CL gene was cloned between PstI/NotI sites. For constructingpUCo-VvSTS-Pc4CL, the VvSTS gene was cloned between XbaI/NotI sites, while

FIG. 1. Metabolic pathways involved in resveratrol production. Resveratrol biosynthesis and connecting metabolic pathways affecting resvera-trol formation, including malonyl-CoA biosynthesis by acetyl-CoA carboxylase (ACC) and biotin ligase (BirA) as well as fatty acid biosynthesis.The stepwise decarboxylative condensation with 3 acetyl units, each derived from malonyl-CoA, to form the linear tetraketide intermediate andthe inhibition of fabB-fabF by cerulenin are shown.

3452 LIM ET AL. APPL. ENVIRON. MICROBIOL.

the Pc4CL gene was cloned between two NotI sites. Primers for all cloningreactions are available in the supplemental material.

For the isolation of purified proteins, N-terminal 6-histidine-tagged proteinfusions were made for each STS in pACYCDuet-1. This was done by PCRamplification of each STS gene using restriction tail primers and cloning betweenAcsI and SalI restriction sites downstream of the T7 promoter (pT7). Constructswere confirmed by both restriction digest screening and sequencing before trans-formation into E. coli BL21 Star for expression. All plasmids used in this studyare listed in Table 1.

Protein purification. To isolate purified STS proteins, E. coli BL21 Star cul-tures harboring the STS His-tagged constructs were grown in LB broth at 37°Cuntil the optical density (OD) reached 0.6. Cultures were then induced with 1mM isopropyl-$-D-thiogalactopyranoside (IPTG) and grown at 30°C for another4 h. Cells were then harvested by centrifugation, washed with ice-cold bindingbuffer (20 mM sodium phosphate, 500 mM sodium chloride), and reduced involume 50-fold. Cells were then sonicated at 4°C for a total of 1 min (15-s pulses)and then centrifuged for 20 min at 14,000 rpm on a precooled benchtop micro-centrifuge. Next, samples were pooled and passed through a 0.22-"m syringefilter as crude extract. After the HisTrap column was washed (1 ml; GE Health-care) with sterile deionized (DI) water and equilibrated with 5 column volumes(CV) of binding buffer, the crude soluble proteins were applied to the column ata flow rate of 1 CV per minute. Elution was carried out using 5 CV for each stepin a three-step profile with increasing concentrations of imidazole (30 mM, 100mM, and 250 mM), collecting 1.0-ml fractions during the elution. The columnwas then stripped and regenerated according to the manufacturer’s instructions.Each elution fraction was tested for the presence of proteins using a BCA assay(Pierce), with positive fractions pooled. Purified proteins were then passedthrough a Microcon YM10 column (Millipore), and buffer was changed to anassay buffer (20 mM HEPES, 5 mM EDTA). These purified proteins were thenremeasured to verify purified protein concentration. All His-STS proteins werefound in the first three fractions of the 250 mM elution.

In vitro assays. For crude enzyme assays, all substrates were first dissolved in0.1 M Tris-HCl at a pH of 4.0. The final concentrations of substrates in eachreaction were 0.5 mM malonyl-CoA, 0.25 mM p-coumaric acid, 1.0 mM coen-zyme A, and 20 mM ATP, with 200 "g of total crude protein. Purified-proteinassay mixtures contained 200 pmol of purified protein with excess coumaroyl-CoA (%0.2 mM). To determine the Michaelis-Menten constants for each en-zyme, a time course of samples over 30 min was assayed with various concen-trations of malonyl-CoA and acetyl-CoA. All reaction mixtures were incubatedat 30°C before being extracted with equal volumes of 1% HCl in ethyl acetateand quantified by high-pressure liquid chromatography (HPLC) analysis. Prod-uct concentrations were plotted on Lineweaver-Burk plots to obtain traditionalMichaelis-Menten kinetic parameters.

Transcription analysis. Primers used for qRT-PCR runs can be found in thesupplemental material. To have a direct comparison between the constitutivepGAP expression and the inducible pT7 expression, all strains harboring the pT7construct with 4CL and STS coding sequences in one bicistronic operon weregrown with IPTG at all stages. To isolate RNA from all strains, preinoculumswere grown overnight, then diluted to OD of 0.05 in 25 ml of fresh LB culture.After incubation at 37°C until the OD reached 0.6, cells were harvested for RNAisolation using the RNeasy isolation kit (Qiagen) according to the manufacturer’sinstructions.

Resveratrol fermentations. For E. coli strain BL21 Star, harboring the genesencoding the stilbene biosynthetic pathway in the pCDF plasmid, preinoculumswere grown overnight in LB broth, then diluted to OD of 0.1 in 25 ml of fresh LB.Cells were incubated at 37°C until the OD reached 0.8 before they were inducedwith 1 mM IPTG and grown for another 3 h at 30°C for protein production. Next,cells were collected by centrifugation and resuspended in M9 medium containingthe necessary antibiotics, 1 mM p-coumaric acid, 1 mM IPTG, and/or 0.05 mMcerulenin. Incubation continued at 30°C for 32 h prior to analysis of the heter-ologous products. For strains expressing acetyl-CoA carboxylase and biotin li-gase, 6 "M biotin was added. For HPLC analysis, 500 "l of fermentation samples

TABLE 1. Plasmids and strains used in the present study

Plasmid or E. coli strain Relevant properties or genetic marker Source or reference

PlasmidspACYCDuet P15A ori (pACYC184), Cmr NovagenpCOLADuet ColA ori Kmr NovagenpCDFDuet CloDF13 ori Strr NovagenpUC18 ColE1 Ampr EvonikpC-Pc4cl2-Ahsts pCDFDuet & 4cl-2 from P. crispum, sts from A. hypogaea This studypC-Pc4cl2-Vvsts pCDFDuet & 4cl-2 from P. crispum, sts from V. vinifera This studypC-Pc4cl2-Pssts pCDFDuet & 4cl-2 from P. crispum, sts from P. strobus This studypC-Pc4cl2-Pmsts pCDFDuet & 4cl-2 from P. crispum, sts from P. massoniana This studypC-Pc4cl2-Pnsts pCDFDuet & 4cl-2 from P. crispum, sts from P. nudum This studypC-Pc4cl2-Pdsts pCDFDuet & 4cl-2 from P. crispum, sts from P. densiflora This studypC-Pc4cl2-Pcu1sts pCDFDuet & 4cl-2 from P. crispum, sts from P. cuspidatum 1 This studypC-Pc4cl2-Pcu3sts pCDFDuet & 4cl-2 from P. crispum, sts from P. cuspidatum 3 This studypC-At4cl1-Ahsts pCDFDuet & 4cl-1 from A. thaliana, sts from A. hypogaea This studypC-At4cl1-Vvsts pCDFDuet & 4cl-1 from A. thaliana, sts from V. vinifera This studypCo-Pc4cl2-Ahsts pCDFDuet & sts from A. hypogaea, 4cl-2 from P. crispum in an operon This studypCo-Pc4cl2-Vvsts pCDFDuet & sts from V. vinifera, 4cl-2 from P. crispum in an operon This studypCo-At4cl1-Ahsts pCDFDuet & sts from A. hypogaea, 4cl-1 from A. thaliana in an operon This studypCo-At4cl1-Vvsts pCDFDuet & sts from V. vinifera, 4cl-1 from A. thaliana in an operon This studypUC-Ahsts-At4cl1 pUC18 & sts from A. hypogaea, 4cl-1 from A. thaliana This studypUC-Vvsts-At4cl1 pUC18 & sts from V. vinifera, 4cl-1 from A. thaliana This studypUCo-Ahsts-At4cl1 pUC18 & sts from A. hypogaea, 4cl-1 from A. thaliana in an operon This studypUCo-Vvsts-At4cl1 pUC18 & sts from V. vinifera, 4cl-1 from A. thaliana in an operon This studypUCo-Pssts-At4cl1 pUC18 & sts from P. strobus, 4cl-1 from A. thaliana in an operon This studypUCo-Pmsts-At4cl1 pUC18 & sts from P. massoniana, 4cl-1 from A. thaliana in an operon This studypUCo-Ahsts-Pc4cl2 pUC18 & sts from A. hypogaea, 4cl-2 from P. crispum in an operon This studypUCo-Vvsts-Pc4cl2 pUC18 & sts from V. vinifera, 4cl-2 from P. crispum in an operon This studypA-Placc pACYCDuet & accABCD from P. luminescens 19pCO-PlbirA pCOLADuet & birA from P. luminescens 19

StrainsBL21 Star F' ompT hsdSB (rB

' mB') gal dcm rne131 (DE3) Invitrogen

BW27784 F' ((araD-araB)567 (lacZ4787(::rrnB-3) )' ((araH-araF)570(::FRT)(araEp-532::FRT *Pcp18araE533 ((rhaD-rhaB)568 hsdR514

CGSCa

a CGSC, E. coli Genomic Stock Center, Yale University.

VOL. 77, 2011 COMMERCIAL RESVERATROL TITERS IN ENGINEERED E. COLI 3453

was centrifuged and the supernatant was injected into the HPLC to quantifyresveratrol levels.

For strain BW27784, carrying the 4CL and STS genes in the pUC construct,the culture medium consisted of yeast extract M9 medium (YM9) adjusted to pH7 (1+ M9 salts, 10 g/liter yeast extract, 3% glycerol, and 42 g/liter MOPS[morpholinepropanesulfonic acid] [%99.5% purity]) along with required antibi-otics. Preinoculums were grown at 37°C for 6 h before being diluted to OD of 0.1in 25 ml of fresh medium and grown overnight, again at 37°C. The next day, cellswere collected by centrifugation and resuspended in fresh YM9 additionallysupplemented with 15 mM p-coumaric acid, 10 g/liter of polyethylene glycol,and/or cerulenin. Incubation continued at 37°C for 24 h, and 100-"l samples werecollected at 6 and 24 h for analysis. A 100-"l or sometimes 200-"l (with highconcentration of resveratrol) sample of ethyl acetate containing 1% HCl wasadded to each sample. The mixture was then vortexed for 30 s and separatedthrough centrifugation before being injected into the HPLC.

HPLC analysis. For the pCDF system, cell-free culture medium from fermen-tations (100 "l) was injected into the HPLC using a solvent profile with 0.1%(vol/vol) formic acid in acetonitrile (buffer A) and 0.1% (vol/vol) formic acid inwater (buffer B) as the mobile phases and a flow rate of 1.0 ml per minute. Themobile-phase composition profile was fixed at 65% buffer B and 35% buffer A for4 min, with 30 s of postrun time allowed for column equilibration. The resvera-trol peak was found to elute at 3.6 min under this solvent profile. For productionin BW27784, as well as in vitro assays, 2 "l of the ethyl acetate fraction wasinjected into the HPLC. Ethyl acetate extraction using twice the volume ofculture sampled was done prior to the HPLC injection to prevent polyethyleneglycol from the culture medium or protein in the in vitro assay from clotting upthe guard column in the HPLC.

RESULTS

An initial resveratrol construct. A strain design similar tothat from work done previously by our group for flavonoidbiosynthesis was used as the initial resveratrol production plat-form (9, 20). This consisted of E. coli strain BL21 Star harbor-ing plasmid pCDFDuet-1, encoding 4CL from Petroselinum

crispum (parsley; Pc4CL) and STS from Arachis hypogaea (pea-nut; AhSTS). Previous studies revealed that Pc4CL had a morediverse substrate specificity than other 4CL enzymes (21),while the STS from peanuts is well studied and known to beactive when produced recombinantly in E. coli with broadsubstrate specificity (23, 31, 34). p-Coumaric acid was supple-mented as the precursor for the biotransformation rather thanthe basic amino acids for several reasons. First, both phenyl-alanine ammonia lyase (PAL) and tyrosine ammonia lyase(TAL) suffer from low turnover numbers. Another P450 en-zyme, cinnamate-4-hydroxylase (C4H), catalyzing the conver-sion of cinnamic acid to p-coumaric acid, fails to be effectivelyexpressed in E. coli. Moreover, p-coumaric acid is the secondmost abundant compound derived from lignin and is thereforerelatively cheap. As a result, batch fermentation was able toachieve up to 33 mg/liter of resveratrol production in shakeflask cultures.

Screening for new STS via bioinformatics and homologymodeling. To further expand the production potential of thisconstruct, it was hypothesized that a more active STS wouldincrease production rates and levels. Therefore, an array ofSTS proteins was identified for testing. As a first step, an initialphylogenetic tree built from a BLAST search led to the selec-tion of seven new STS sequences based on different tree group-ings. Those selected include Vitis vinifera (wine grape; VvSTS),2 sequences from Polygonum cuspidatum (Japanese knotweed;Pcu1STS and Pcu3STS), Psilotum nudum (whisk fern; PnSTS),Pinus massoniana (Chinese red pine; PmSTS), Pinus strobus(eastern white pine; PsSTS), and Pinus densiflora (Japanese

FIG. 2. Stilbene synthase bioinformatic analysis. (A) Phylogenetic tree constructed for the eight selected STS sequences. (B) SDS-PAGE (top)and Western blots (bottom) of isolated STS proteins AhSTS (lane 1), VvSTS (lane 2), PsSTS (lane 3), codon-optimized PsSTS (lane 4), andcodon-optimized PmSTS (lane 5) with molecular weight ladders (M). Proteins were produced under T7 promoter control in BL21 Star for in vitroanalysis using crude cell lysates. (C) ClustalW alignment of the protein sequences of the active-site regions of all eight selected STS proteins(AhSTS residue numbering). Active-site residues are indicated by asterisks, and nonhomologous regions are in red. Columns below the alignmentrepresent percentages of conservation of the corresponding amino acids in those positions.

3454 LIM ET AL. APPL. ENVIRON. MICROBIOL.

red pine; PdSTS). After being realigned using their proteinsequences, all STS enzymes used in this study were used toconstruct a new phylogenetic tree (Fig. 2A). Inspection of theprotein alignment revealed a highly conserved active site withgreater than 90% positional amino acid sequence identity(data not shown). On the other hand, regions downstream andadjacent to the residues involved in the hydrogen-binding do-main show a large variance in conservation (Fig. 2C).

To investigate this further, homology models were built foreach of the STS proteins selected by mapping the translatedamino acid sequence to the previously published peanut STS,a known highly active STS, crystal structure (PDB identifica-tion number 1Z1F). After homology models were built, a con-siderable variance in the hydrogen-binding domain comparedto that of peanut STS was seen in VvSTS, PsSTS, PmSTS,PdSTS, and PnSTS (Fig. 3). Specifically, Met98 (AhSTS num-bering) has been changed to a variety of residues of bothincreasing and decreasing sizes as well as increased electronwithdrawing abilities. In contrast, a significant difference isseen in the supporting residues for both Polygonum STSs com-pared to AhSTS (Fig. 3). Phe265 is lost in Pcu3STS completely,while in Pcu1STS the Gly256 is changed to a Leu, which ex-tends into the catalytic pocket.

Investigating the kinetics of the STS and 4CL library. TheVvSTS gene was isolated from plant material, while the sixremaining STS gene sequences were synthesized followingcodon optimization. Similar to that of AhSTS, other STS geneswere simultaneously cloned into the second multiple cloningsite of the pCDFDuet-1 vector, with the first being occupied bythe Pc4CL gene. Functional expression of each STS gene wasverified by fermentation. Out of all STSs, only two STSs inaddition to the AhSTS produced resveratrol above detectablelimits: VvSTS (0.7 mg/liter) and PsSTS (2.0 mg/liter) (Table 2).As a second validation of functional expression, crude E. coliextract, after expression of each STS gene under pT7 control inthe pACYC vector, was separately mixed with another crudeextract harboring an active Pc4CL (data not shown). This ledto the identification of a new active STS from P. massoniana inaddition to those previously verified. Other STS enzymes,while produced in significant amounts as verified by SDS-PAGE, failed to form resveratrol above detectable limits andwere not investigated further.

The four active proteins were further characterized in vitrousing Ni& affinity column purification, a method previouslyused for STS protein purification (26). Interestingly, variousproduction levels were found for each STS in crude extracts, asseen in SDS-PAGE and Western blot analysis (Fig. 2B). Fol-lowing purification and a buffer exchange to remove excessimidazole, in vitro assays established the apparent kinetic pa-rameters of each STS (Table 3). Corresponding to the levels ofresveratrol in the respective host plant material, the kcat/Km

TABLE 2. In vivo resveratrol production with E. coli strainsharboring STS- and 4CL-encoding genes in mono- or

bicistronic transcriptional units

Strain and plasmiddescription 4CL gene STS gene Product titera

(mg/liter)

BL21 Star & pCDF(2 promoters)

Pc4CL PmSTS —b

Pc4CL VvSTS 0.74 , 0.04Pc4CL PsSTS 2.01 , 0.29Pc4CL AhSTS 33.0 , 1.70

BW27784 & pUC(operon)

At4CL PmSTS 23.7 , 0.7At4CL PsSTS 61.5 , 7.0At4CL AhSTS 404 , 70At4CL VvSTS 1,380 , 183

a All values are the averages of three trials with standard deviations.b —, below limit of detection.

FIG. 3. Homology modeling. Homology models clearly show dif-ferences in the hydrogen-binding network (A) and in the supportdomain (B) of the STS active site. Resveratrol (blue) with an electrondot cloud and the catalytic residues (red) common to all STSs areshown. Differences are shown by species for residues M98 (A. hypo-gaea, green; P. densiflora and P. massoniana, pink; P. strobus, purple; V.vinifera, cyan; P. nudumn, orange) and G255, G256, and F265 (A.hypogaea, yellow; P. cuspidatum, brown [Pcu1STS] and tan [Pcu3STS]).

TABLE 3. Michaelis-Menten kinetic constants of variousstilbene synthases

ConstantValuea for:

A. hypogaea V. vinifera P. strobus P. massoniana

kcat (min'1) 0.10 , 0.03 0.18 , 0.07 0.03 , 0.01 0.14 , 0.03Km ("M) 4.43 , 0.25 8.56 , 0.50 2.87 , 0.2 11.85 , 0.55kcat/Km

(M'1 s'1)314 210 160 170

Ki ("M)b 520 , 30 460 , 60 970 , 70 787 , 50Ki! ("M)c 70 , 20 20 , 7 150 , 30 40 , 10

a All values are the average of three trials with standard deviations.b Apparent competitive inhibition constant for acetyl-CoA.c Apparent uncompetitive inhibition constant for acetyl-CoA.

VOL. 77, 2011 COMMERCIAL RESVERATROL TITERS IN ENGINEERED E. COLI 3455

ratios for AhSTS and VvSTS were significantly higher thanthose for PsSTS and PmSTS. Additionally, in vitro assays usingpurified STS were performed while varying the acetyl-CoAconcentration to investigate any inhibitory effect. Significantreductions in production were seen for all STS enzymes testedwhen concentrations were above 1 mM, indicating a degree ofinhibitory activity well above normal cellular levels of acetyl-CoA. Due to the complexity of the ping-pong kinetic system

(substrate binding, three successive substrate condensations,and product cyclization and release), only the apparent inhib-itory constants for competitive and noncompetitive inhibitionare reported.

Based on the enzyme kinetics, both VvSTS and AhSTS ap-peared to have higher turnover numbers than the other STSs,especially in the case of VvSTS, with catalytic turnover ofalmost twice as that of AhSTS. Surprisingly, this observationwas not reflected in vivo with host strain BL21 Star, in whichPc4CL coupled with AhSTS (33 mg/liter) performed muchbetter than when used with VvSTS (0.7 mg/liter). Anothercloning approach, in which both 4CL and STS genes werecloned into the same operon under the control of a single pT7promoter, was attempted (Fig. 4A). In this case, partial activityof VvSTS can be recovered and resveratrol production (15.5mg/liter) was greater than with AhSTS (0.1 mg/liter) (Table 4).Apart from variable STS proteins, 4CL diversification was alsoaddressed. 4CL from Arabidopsis thaliana (At4CL1) was em-ployed in addition to Pc4CL because At4CL1 was also shownto be functional in recombinant E. coli for flavonoid biosyn-thesis (34). In the pT7 system, both 4CL proteins gave similarproduction levels, indicating that 4CL is able to efficientlyconvert the p-coumaric acid into its coumaroyl-CoA ester. Thisresult implies that either the activity (or expression) of STS isnot sufficiently high to consume the available ester or that theintracellular pool of malonyl-CoA is limited, thereby slowingthe generation of resveratrol.

High-production resveratrol strains by changing the expres-sion construct. The work of Watts et al. achieved significantlyhigher titers than our original construct using a high-copy-number vector and an E. coli strain that is deficient in arabi-nose transport (34). Since STS and 4CL gene expression fromthe pCDF backbone, and thus resveratrol biosynthesis, arelimited by the activity of the T7 RNA polymerase encoded onthe chromosome of BL21 Star, a new expression constructusing a constitutive promoter based on the expected metabolicstate of the cells during the production phase was proposed. Incontrast to the IPTG-inducible promoter (pT7) in the initialexpression system, the constitutive gap promoter (pGAP) wasselected as it is actively induced during fermentation on glu-cose as the carbon source, thus allowing the continuous tran-scription of both 4CL and STS genes. Additionally, the alter-native strain BW27784 was used to improve the level ofbiomass during fermentations in an effort to enhance resvera-trol titers. Two approaches, one in which both genes wereexpressed under the control of separate gap promoters and theother in which both were expressed in an operon, were again

FIG. 4. Resveratrol production constructs and recombinant geneexpression in E. coli. (A) Plasmid constructs with either a bicistronicoperon or two monocistronic transcriptional units, gap, or T7 promot-ers in vectors pCDF and pUC18. (B) Transcription intensities for both4CL (white) and STS (black) genes under the control of either the gappromoter on pUC derivatives in strain BW27784 or the T7 promoteron pCDF derivatives in strain BL21 Star. Transcription intensities forall the strains are normalized to expression of the genes encodingVvSTS and At4CL in strain BW27784, the most efficient resveratrol-producing strain.

TABLE 4. In vivo resveratrol production with different combinations of STS- and 4CL-encoding genes in different strain backgrounds

Recombinant 4CLgene source

Recombinant STSgene source

Product titera (mg/liter) for:

BL21 Star & pCDF(2 promoters)

BL21 Star & pCDF(operon)

BW27784 & pUC(2 promoters)

BW27784 & pUC(operon)

A. thaliana A. hypogaea —b 0.10 , 0.01 172 , 3.3 404 , 70A. thaliana V. vinifera — 20.0 , 0.66 348 , 9.4 1,380 , 183P. crispum A. hypogaea 33.0 , 1.70 0.08 , 0.01 — 142 , 13P. crispum V. vinifera 0.74 , 0.04 15.5 , 0.19 — 610 , 25

a All values are the averages of three trials with standard deviations.b —, not determined.

3456 LIM ET AL. APPL. ENVIRON. MICROBIOL.

used to introduce the genes into new strain BW27784 in high-copy-number plasmid pUC18 (Fig. 4A). Fermentation titersrevealed the inefficiency of a two-promoter construct in thisnew expression system compared to the one with a bicistronicoperon (Table 4). The production of resveratrol in the pUCexpression system using either Pc4CL or At4CL was muchhigher than that by the pCDF system. Interestingly, Pc4CLlimited resveratrol production compared to At4CL in all of thepUC constructs (Table 4). PmSTS, PsSTS, AhSTS, and VvSTSwere all found to be active in shake flask fermentations whenexpression of the respective genes was coupled with At4CLgene expression in an operon, resulting in final volumetricproduction levels of 23.7 mg/liter, 61.5 mg/liter, 404 mg/liter,and 1,380 mg/liter, respectively (Table 2). The last is the high-est resveratrol titer reported to date and represents an as-tounding increase of more than 40-fold compared to that withthe starting pC-Pc4CL-AhSTS construct in BL21 Star ofmerely 33 mg/liter. These results also confirmed the accuracyof the in vitro assay with purified proteins, in which the catalyticefficiency of VvSTS is higher than that of AhSTS and both oftheir kcat/Km values were significantly higher than those ofPsSTS and PmSTS.

In order to address the production discrepancy between thetwo systems and also among different 4CL-STS combinations,quantitative RT-PCR was performed with the use of the com-mon housekeeping gene encoding GAPDH as a referencegene. Expression of either the 4CL or STS gene in each con-struct was compared to the transcription of the respective genein the construct yielding highest resveratrol production, pUCo-VvSTS-At4CL. As seen in Fig. 4B, the transcription of the 4CLgene relies heavily on the STS gene, to which it is coupled.Similar trends for the At4CL and Pc4CL genes can be observedin both the pCDF and pUC constructs. While the transcriptionof the STS genes is independent of the 4CL genes, the expres-sion of the AhSTS gene in all the constructs is uniform andsignificantly lower than the expression of the VvSTS gene.

High-titer resveratrol strains by increasing the precursorsupply. Additional modifications to the fermentative strategyand construct design were also attempted to further elevateresveratrol titers by improving the availability of malonyl-CoAaccording to prior results (19, 20). As a simple means to in-crease the intracellular malonyl-CoA level, cerulenin, a specificinhibitor of the fabB-fabF gene products (Fig. 1), was intro-duced to the fermentation postinduction. However, the addi-tion of cerulenin is cost prohibitive for a production scalefermentation process; therefore, another metabolic engineer-ing strategy was attempted, one that used the simultaneousoverexpression of acetyl-CoA carboxylase (ACC) and biotin

ligase (BirA) from Photorhabdus luminescens, allowing acetyl-CoA to be more effectively converted to malonyl-CoA.

In the pCDF system, both strategies were able to improvethe overall resveratrol production of pC-Pc4CL-AhSTS (Table5). Cerulenin was able to increase production up to 2-fold to 65mg/liter, while when ACC and BirA were overexpressed, theproduction increased to 46 mg/liter. For the pUC system, ad-ditional coumaric acid supplementation was needed (increasedfrom 6 mM to 15 mM) to maintain a substrate driving force forthe formation of resveratrol. Production as a result of thecerulenin treatment was improved to 2,340 mg/liter in the pUCsystem. Overexpression of ACC/BirA was not attempted in thepUC system, mainly due to the host strain, BW27784, whichlacks )DE3 prophage for T7 RNA polymerase expression.

Cell dynamics during resveratrol production. The engi-neered strain BW27784 harboring the STS gene from V. vinif-era and the 4CL gene from A. thaliana on a pUC backbone wascultured in YM9 to monitor the cell growth rate, availability ofsubstrate, and resveratrol production over the course of 24 h(Fig. 5). A substrate concentration of 15 mM (2,462 mg/liter)was used due to the effective conversion of the substrate. Es-pecially in the cerulenin-supplemented culture, only 200 mg/liter, or less than 10% of the initial substrate concentration,remained at the end of the fermentation period. High levels ofresveratrol production persisted over the 24-hour period forthe strain, both with and without cerulenin, reaching maxi-mums of 2,390 and 1,260 mg/liter with linear rates in the first5 h of 250 mg/liter/h and 180 mg/liter/h, respectively.

DISCUSSION

Developing efficient production platforms for natural prod-ucts requires the engineering of both proficient catalytic pro-teins and stable expression constructs in addition to properhost selection. In this work we explored different strategies thatwould lead to the construction of an effective resveratrol pro-duction platform. These included the identification of alterna-tive stilbene synthases for recombinant production of resvera-trol and the comparison of different strain environments anddifferent expression constructs. Previous characterization ef-forts have revealed significant conservation of the active-sitecavity between CHS and STS. Namely, there are three catalyticresidues (His303, Asn336, and Cys164) responsible for trans-ferring the carboxyl group and an adjacent proline residue(Pro375) critical for carrying out the cyclization of thepolyketide (15, 16, 33). Recently, studies of the peanut andpine STS crystal structures as well as an STS-like mutant CHS

TABLE 5. In vivo resveratrol production with E. coli strains overproducing acetyl-CoA carboxylase/biotin ligaseor in the presence of cerulenin

Strain PlasmidSource of: Product titera (mg/liter) with:

4CL gene STS gene ACC & BirA Cerulenin

BL21 Star pCDF (2 promoters) P. crispum A. hypogaea 45.9 , 1.7 65.3 , 0.1BW27784 pUC (operon) A. thaliana V. vinifera —b 2,340 , 90

a All values are the average of three trials with standard deviations.b —, not determined.

VOL. 77, 2011 COMMERCIAL RESVERATROL TITERS IN ENGINEERED E. COLI 3457

have identified three residues (Thr131, Thr132, and Met98)critical in mediating stilbene-forming aldol cyclization by con-necting an active-site water molecule to a buried glutamate(Glu192) (1, 31). It has been suggested that variation of thisregion may shift the water molecule away from Cys164, result-ing in a loss of catalytic activity (31). Conserved glycine resi-dues (Gly255 and Gly256), by which substrates enter the cat-alytic space, have been shown to affect the chain length of thegrowing polyketide via steric hindrance (15). Considering theseresidues, we compared and contrasted the isolated STS se-quences we obtained, followed by an investigation of theirimpact on both catalytic turnover and functionality.

Phylogenetic analysis suggested an evolutionarily close rela-tionship between the STS proteins considered in this study.Protein structure analysis revealed that, while cross-speciesconservation is seen in the catalytic triad and cyclization resi-dues, STS proteins varied significantly in the hydrogen-bond-ing domain used to transfer electrons from a water molecule toa buried glutamate residue. We hypothesize that these changeshave altered the catalytic activities of the different STS pro-teins. For example, the lack of activity in PdSTS is most likelycaused by the loss of a residue near Arg307, as this residue has

been implicated as a proton acceptor (26). Similarly, PnSTScontains three arginine residues, in place of only one in AhSTS,directly after Met98 (Fig. 3), a residue greatly influenced byside chain charges. While enzyme activity of PmSTS shows akcat and Km similar to those of VvSTS, meriting its selection foruse in a stilbene production construct, Western blot analysisshowed PmSTS to be produced at a considerably lower levelthan both VvSTS and AhSTS (Fig. 2B). Due to the low kcat andKm determined for PsSTS during the in vitro trials and the lackor limited production for the remaining STS proteins screened,high-yield resveratrol production appears to be feasible onlywith the use of VvSTS or AhSTS. Recently, it was shown thatPcu3STS is actually a bifunctional enzyme with both chalconesynthase and benzalacetone synthase activity, with preferencefor the latter (22).

Exploration of various expression constructs led to an effi-cient platform using a constitutive promoter and unique genecombination in E. coli. The a priori selection of an efficientstrain background for a given expression construct continues tobe a challenging task. In this study a large disparity betweenthe T7 promoter system in E. coli BL21 Star and the gappromoter system in E. coli BW27784 was seen. One advantageof strain BW27784 was its ability to achieve and maintain highcell density (Fig. 5A) without the need of concentrating thebiomass prior to the actual biotransformation, as done previ-ously for the production of other flavonoid compounds (17).

In the T7 promoter system, only three active STSs werefound: AhSTS, VvSTS, and PsSTS. However, PmSTS wasshown to be active in vitro using crude extract and pure protein.A possible explanation could be that production of PmSTSin vivo was below the limit detectable by HPLC. When thestrain background was moved to BW27784, the pGAP expres-sion increased more than 40-fold over that of the best constructin BL21 Star, indicating an inherent preference for theBW27784 genotype. The four STS proteins had a productionrange from 23.7 to 1,380 mg/liter. Production differences be-tween AhSTS and VvSTS using the same construct are mostlikely due to a combination of the lower observed catalyticactivity for AhSTS (Table 3) and the higher transcriptionalefficiency for VvSTS (Fig. 4B). In the pCDF system, resultsshowed that resveratrol production using VvSTS was enhancedwhen VvSTS and Pc4CL were expressed in an operon whileproduction associated with AhSTS was enhanced by a system inwhich each gene was expressed under the control of a separatepromoter. However, when either At4CL or Pc4CL was coupledwith VvSTS and expressed under the control of a separate T7promoter, higher transcription of the VvSTS gene resulted inthe reduction of the overall yield.

On the other hand, for the pUC system, the constitutivepromoter pGAP was utilized and both 4CL and STS geneswere continuously transcribed. One of the expected advantagesof using an operon-based expression platform is that bothgenes can be transcribed together as a polycistronic mRNAstrand and translated together in the cytoplasm to form aprotein complex. By coupling both the more efficient VvSTSand At4CL, resveratrol production at 1.4 g/liter was attained.However, when both the genes were expressed by separatepromoters, production was reduced 5-fold to 348 mg/liter, pos-sibly due to unbalanced expression, which could result in anincreased accumulation of the coumaroyl-CoA intermediate,

FIG. 5. Dynamics of the production of resveratrol with recombi-nant E. coli strains. OD600 (A), p-coumaric acid consumption (B), andresveratrol production (C) during biotransformation with engineeredE. coli BW27784 strains expressing STS and 4CL are shown. E,BW27784 pUCo-VvSTSAt4CL; ‚, BW27784 pUCo-VvSTSAt4CL with0.05 mM cerulenin.

3458 LIM ET AL. APPL. ENVIRON. MICROBIOL.

which could either inhibit STS or prove detrimental to cellviability. A similar effect was observed in the pCDF system, inwhich the VvSTS gene coupled with the At4CL gene workedmore efficiently in an operon than with separate promoters.Production by AhSTS in the pUC construct was second to thatby VvSTS, while the production achieved using the remainingSTS enzymes was significantly lower. This result was somewhatrelated to the enzyme efficiency data obtained from the in vitroassays.

In the case of 4CL, no significant productivity differenceswere seen in the pCDF system when Pc4CL was replaced withAt4CL. However, in the pUC system both VvSTS and AhSTScoupled with At4CL were considerably more efficient in pro-ducing resveratrol than either STS when coupled with Pc4CL(Table 4). This can be explained by the role of the expressionsystem in each construct. We postulated that in the pCDFsystem, which suffered from low resveratrol production, therewas a buildup of the coumaroyl-CoA intermediate, which wasnot efficiently converted to the end product by the STS. There-fore, the efficiency of the 4CL was not reflected in the overallproduction; hence, the catalytic conversion rates of both 4CLscould not be compared. On the other hand, with the pUCsystem, STSs actively converted the intermediate to the endproduct, driving the need for both malonyl-CoA and the inter-mediate, which required the 4CL to be fully functional. Par-ticularly in the case of VvSTS, with twice the catalytic turnoverefficiency of AhSTS, an efficient 4CL is required for resveratrolproduction, with At4CL favored over Pc4CL.

Resveratrol production platform efficiency was further im-proved through the inhibition of fatty acid biosynthesis bycerulenin and by the overproduction of ACC/BirA (Table 5).Both approaches were able to improve the production of res-veratrol with BL21 Star pC-Pc4CL2-AhSTS by up to 2- and1.4-fold, respectively. Even with the pUC system, in whichresveratrol production was very efficient, malonyl-CoA re-mained a limiting factor. By the addition of 0.05 mM ceruleninto the cultivation broth, the resveratrol titer was further in-creased to 2.3 g/liter (1.4 g/liter without adding cerulenin),another astounding 64% improvement. Similarly, an increasein volumetric resveratrol productivity was observed (Fig. 5C).Within the first 5 h, when product concentration increasesalmost linearly with time, cerulenin was able to improve volu-metric productivity at a rate of approximately 70 mg/h.

Previous efforts to engineer recombinant microbes for stil-bene biosynthesis focused on the overexpression of pathwayenzymes and diversification of the chemistry of stilbene com-pounds (4, 17, 25, 34). However, these studies lack sufficientunderstanding of resveratrol biosynthesis on the molecularlevel, for example, the catalytic turnover of the pathway en-zymes as well as their transcription and translation efficiencies,largely determined by the employed gene expression system.Our study demonstrated the construction of an optimized res-veratrol production system in E. coli by screening genes alongthe stilbene biosynthetic pathway from various plant sourcesfor highest enzyme activity. A few enzyme combinations of4CL and STS were subsequently selected for further optimi-zation, including the gene expression system, strain back-ground, and malonyl-CoA availability. Production discrepan-cies among different enzyme combinations and expressionsystems were resolved by quantitative RT-PCR and Western

blotting. Various metabolic engineering approaches presentedin this paper are critical to achieve significant production ofresveratrol, and such strategies are applicable not only to res-veratrol but also other highly valued nutraceuticals and sec-ondary metabolites.

ACKNOWLEDGMENTS

We acknowledge the financial support from Evonik Industries toM. A. G. Koffas. The work was funded by the State of North-RhineWestphalia (Germany) and cofinanced by the European Union.

We thank the Buffalo and Erie County Botanical Gardens for pro-viding plant sources for RNA isolation. We acknowledge the assistanceof Effendi Leonard and Kok Hong Lim for creating the initial con-struct pC-AhSTS-Pc4CL. The assistance of Lynn Wong is gratefullyacknowledged.

REFERENCES

1. Austin, M. B., M. E. Bowman, J. L. Ferrer, J. Schroder, and J. P. Noel. 2004.An aldol switch discovered in stilbene synthases mediates cyclization speci-ficity of type III polyketide synthases. Chem. Biol. 11:1179–1194.

2. Baur, J. A., et al. 2006. Resveratrol improves health and survival of mice ona high-calorie diet. Nature 444:337–342.

3. Baur, J. A., and D. A. Sinclair. 2006. Therapeutic potential of resveratrol: thein vivo evidence. Nat. Rev. Drug Discov. 5:493–506.

4. Beekwilder, J., et al. 2006. Production of resveratrol in recombinant micro-organisms. Appl. Environ. Microbiol. 72:5670–5672.

5. Beher, D., et al. 2009. Resveratrol is not a direct activator of SIRT1 enzymeactivity. Chem. Biol. Drug Des. 74:619–624.

6. Bhat, K. P. L., J. W. Kosmeder II, and J. M. Pezzuto. 2001. Biological effectsof resveratrol. Antioxid. Redox Signal. 3:1041–1064.

7. D’Introno, A., et al. 2009. Antioxidant and anti-inflammatory properties oftomato fruits synthesizing different amounts of stilbenes. Plant Biotechnol. J.7:422–429.

8. Fauconneau, B., et al. 1997. Comparative study of radical scavenger andantioxidant properties of phenolic compounds from Vitis vinifera cell cul-tures using in vitro tests. Life Sci. 61:2103–2110.

9. Fowler, Z. L., W. W. Gikandi, and M. A. Koffas. 2009. Increased malonylcoenzyme A biosynthesis by tuning the Escherichia coli metabolic networkand its application to flavanone production. Appl. Environ. Microbiol. 75:5831–5839.

10. Fremont, L., L. Belguendouz, and S. Delpal. 1999. Antioxidant activity ofresveratrol and alcohol-free wine polyphenols related to LDL oxidation andpolyunsaturated fatty acids. Life Sci. 64:2511–2521.

11. Guex, N., and M. C. Peitsch. 1997. SWISS-MODEL and the Swiss-Pdb-Viewer: an environment for comparative protein modeling. Electrophoresis18:2714–2723.

12. Hanhineva, K., et al. 2009. Stilbene synthase gene transfer caused alterationsin the phenylpropanoid metabolism of transgenic strawberry (Fragaria xananassa). J. Exp. Bot. 60:2093–2106.

13. Howitz, K. T., et al. 2003. Small molecule activators of sirtuins extendSaccharomyces cerevisiae lifespan. Nature 425:191–196.

14. Jang, M., et al. 1997. Cancer chemopreventive activity of resveratrol, anatural product derived from grapes. Science 275:218–220.

15. Jez, J. M., M. E. Bowman, and J. P. Noel. 2001. Structure-guided program-ming of polyketide chain-length determination in chalcone synthase. Bio-chemistry 40:14829–14838.

16. Jez, J. M., and J. P. Noel. 2000. Mechanism of chalcone synthase. pKa of thecatalytic cysteine and the role of the conserved histidine in a plant polyketidesynthase. J. Biol. Chem. 275:39640–39646.

17. Katsuyama, Y., N. Funa, I. Miyahisa, and S. Horinouchi. 2007. Synthesis ofunnatural flavonoids and stilbenes by exploiting the plant biosynthetic path-way in Escherichia coli. Chem. Biol. 14:613–621.

18. King, R. E., J. A. Bomser, and D. B. Min. 2006. Bioactivity of resveratrol.Comp. Rev. Food Sci. Food Safety 5:65–70.

19. Leonard, E., K. H. Lim, P. N. Saw, and M. A. Koffas. 2007. Engineeringcentral metabolic pathways for high-level flavonoid production in Esche-richia coli. Appl. Environ. Microbiol. 73:3877–3886.

20. Leonard, E., et al. 2008. Strain improvement of recombinant Escherichia colifor efficient production of plant flavonoids. Mol. Pharm. 5:257–265.

21. Lozoya, E., et al. 1988. Primary structures and catalytic properties of isoen-zymes encoded by the two 4-coumarate: CoA ligase genes in parsley. Eur.J. Biochem. 176:661–667.

22. Ma, L. Q., et al. 2009. Identification of a Polygonum cuspidatum three-introngene encoding a type III polyketide synthase producing both naringenin andp-hydroxybenzalacetone. Planta 229:1077–1086.

23. Morita, H., H. Noguchi, J. Schroder, and I. Abe. 2001. Novel polyketidessynthesized with a higher plant stilbene synthase. Eur. J. Biochem. 268:3759–3766.

VOL. 77, 2011 COMMERCIAL RESVERATROL TITERS IN ENGINEERED E. COLI 3459

24. Pacholec, M., et al. 2010. SRT1720, SRT2183, SRT1460, and resveratrol arenot direct activators of SIRT1. J. Biol. Chem. 285:8340–8351.

25. Park, S. R., et al. 2009. Engineering of plant-specific phenylpropanoidsbiosynthesis in Streptomyces venezuelae. J. Biotechnol. 141:181–188.

26. Raiber, S., G. Schroder, and J. Schroder. 1995. Molecular and enzymaticcharacterization of two stilbene synthases from Eastern white pine (Pinusstrobus). A single Arg/His difference determines the activity and the pHdependence of the enzymes. FEBS Lett. 361:299–302.

27. Reagan-Shaw, S., M. Nihal, and N. Ahmad. 2008. Dose translation fromanimal to human studies revisited. FASEB J. 22:659–661.

28. Renaud, S., and M. de Lorgeril. 1992. Wine, alcohol, platelets, and theFrench paradox for coronary heart disease. Lancet 339:1523–1526.

29. Sambrook, J., E. F. Fritsch, and T. Maniatis. 1989. Molecular cloning: alaboratory manual, 2nd ed. Cold Spring Harbor Laboratory Press, ColdSpring Harbor, NY.

30. Sengottuvelan, M., P. Viswanathan, and N. Nalini. 2006. Chemopreventive

effect of trans-resveratrol—a phytoalexin against colonic aberrant crypt fociand cell proliferation in 1,2-dimethylhydrazine induced colon carcinogenesis.Carcinogenesis 27:1038–1046.

31. Shomura, Y., et al. 2005. Crystal structure of stilbene synthase from Arachishypogaea. Proteins 60:803–806.

32. Stockigt, J., and M. H. Zenk. 1975. Chemical syntheses and properties ofhydroxycinnamoyl coenzyme A derivatives. Z. Naturforsch. C 30:352–358.

33. Suh, D. Y., et al. 2000. Identification of amino acid residues important in thecyclization reactions of chalcone and stilbene synthases. Biochem. J. 350(Pt.1):229–235.

34. Watts, K. T., P. C. Lee, and C. Schmidt-Dannert. 2006. Biosynthesis ofplant-specific stilbene polyketides in metabolically engineered Escherichiacoli. BMC Biotechnol. 6:22.

35. Wood, J. G., et al. 2004. Sirtuin activators mimic caloric restriction and delayageing in metazoans. Nature 430:686–689.

3460 LIM ET AL. APPL. ENVIRON. MICROBIOL.