Metabolic Engineering of a Novel Muconic Acid Biosynthesis Pathway via 4-Hydroxybenzoic Acid in Escherichia coli

Abstract

cis,cis-Muconic acid (MA) is a commercially important raw material used in pharmaceuticals, functional resins, and agrochemicals. MA is also a potential platform chemical for the production of adipic acid (AA), terephthalic acid, caprolactam, and 1,6-hexanediol. A strain of Escherichia coli K-12, BW25113, was genetically modified, and a novel nonnative metabolic pathway was introduced for the synthesis of MA from glucose. The proposed pathway converted chorismate from the aromatic amino acid pathway to MA via 4-hydroxybenzoic acid (PHB). Three nonnative genes, pobA, aroY, and catA, coding for 4-hydroxybenzoate hydrolyase, protocatechuate decarboxylase, and catechol 1,2-dioxygenase, respectively, were functionally expressed in E. coli to establish the MA biosynthetic pathway. E. coli native genes ubiC, aroF^FBR, aroE, and aroL were overexpressed and the genes ptsH, ptsI, crr, and pykF were deleted from the E. coli genome in order to increase the precursors of the proposed MA pathway. The final engineered E. coli strain produced nearly 170 mg/liter of MA from simple carbon sources in shake flask experiments. The proposed pathway was proved to be functionally active, and the strategy can be used for future metabolic engineering efforts for production of MA from renewable sugars.

INTRODUCTION

Muconic acid (cis,cis-muconic acid [MA]) is an unsaturated dicarboxylic acid with a molecular mass of 142.11 g/mol. It is used for the production of industrially relevant chemicals such as adipic acid (AA), terephthalic acid, caprolactam, and 1,6-hexanediol as well as food ingredients and pharmaceuticals, with a combined annual market value of more than $8 billion (1). AA is the immediate target chemical to be produced from MA, since MA can be converted to AA by hydrogenation at mild operating conditions with high yield. AA is currently produced in a two-step chemical catalysis process. In the first step, a mixture of cyclohexanone and cyclohexanol, called KA oil, is produced from benzene. Oxidation of KA oil with nitric acid in the second step produces AA. Benzene is a derivative of fossil fuel and is a carcinogen, while the nitrous oxide (N₂O) generated during the oxidation of KA oil is a major contributor to ozone depletion and global warming (2). In contrast to chemical processes, bioprocesses are safe, sustainable, and environment friendly for the production of chemicals and fuels from biomass.

Several biotechnology companies are actively pursuing research for the production of biobased AA. Genomatica Inc. patented a synthetic biological pathway for the fermentative production of AA (M. J. Burk, A. P. Burgard, R. E. Osterhout, and P. Pharkya, U.S. patent application 20,100,317,069). Using Escherichia coli as a model organism, Genomatica reported a theoretical yield of 0.92 mol of AA per mol of glucose through in silico analysis. However, the proposed pathway is a hypothetical one which has not yet been demonstrated experimentally. Rennovia Inc. developed a chemical catalysis process for production of AA from glucose via glucaric acid (T. R. Boussie, E. L. Dias, Z. M. Fresco, V. J. Murphy, J. Shoemaker, R. Archer, and H. Jiang, U.S. patent application 20,100,317,823). The yield of this chemical catalysis is rather low (around 60%), and the process results in a mixture of components. Verdezyne has reported that they were successful in proof of concept for production of AA from a nonfood, plant-based feedstock using Saccharomyces cerevisiae, but the actual pathway details are not available (S. Picataggio, K. A. L. Salmon, and J. M. Laplaza, U.S. patent application PCT/US2010/041618). Other biotechnology companies such as DSM and BioAmber have also shown interest in the production of AA through microbial fermentation.

MA was suggested as an intermediate for the production of AA since MA can be produced from glucose and can be converted to AA by hydrogenation at 50 lb/in² for 3 h at room temperature (3). Several approaches have been proposed for fermentative production of MA from benzoate, toluene, and catechol (4, 5). These aromatic substrates were derived from fossil fuels or effluents from other petrochemical industries. For the first time, Frost and Draths (6) patented a hybrid method for production of AA from glucose. In this hybrid method, MA was produced via the aromatic amino acid pathway in E. coli. Subsequently, MA was hydrogenated to AA using a Pt catalyst. The approach of Frost and Draths required heterologous expression of three nonnative genes in E. coli for the production of MA. The first gene, aroZ, encoded the enzyme 3-dehydroxyshikimic acid (DHS) dehydratase, which catalyzed the conversion of DHS to protocatechuic acid (PCA). The second enzyme, PCA decarboxylase (encoded by aroY), converted PCA to catechol, while the third enzyme, catechol 1,2-dioxygenase (encoded by catA), converted catechol to MA. The metabolic reaction that converted DHS to shikimic acid (SA) was blocked to increase the availability of DHS for MA synthesis. The recombinant E. coli strain was able to produce MA from glucose, with an average yield of 0.22 mol/mol of MA per mol of glucose (the theoretical yield is 0.86 mol/mol) and with a productivity of 0.77 g/liter/h in aerobic fed-batch fermentation. Conversion of MA to AA was achieved at around 97% (7).

However, there were some disadvantages with this scheme. The first was that blocking of SA production in the aromatic amino acid pathway of E. coli reduced the cell viability in minimal medium. To overcome this, medium had to be enriched with the aromatic amino acids phenylalanine (Phe), tyrosine (Tyr), and tryptophan (Trp) and the aromatic vitamins p-aminobenzoic acid, 4-hydroxybenzoic acid (PHB), and 2,3-dihydroxybenzoic acid. The second disadvantage was that the production of MA was not coupled with growth due to the incapability of the cells to grow in minimal medium. This pathway was implemented by Weber et al. (8) and Curran et al. (9) with Saccharomyces cerevisiae, and the yields of MA reported were 1.56 mg/liter and 141 mg/liter, respectively. Considering the above-mentioned disadvantages, efforts have been made to develop engineered strains with different novel pathways for the production of MA. Sun et al. (10) developed a recombinant E. coli strain with a novel pathway for production of MA from anthranilate, the first intermediate in the Trp biosynthetic pathway from chorismate, by expressing anthranilate 1,2-dioxygenase and catechol 1,2-dioxygenase. Several genes from the aromatic amino acid pathway were also overexpressed to increase MA production. Using a modified M9 minimal medium with glycerol and glucose as carbon sources, the recombinant E. coli strain produced 390 mg/liter of MA. Since the Trp biosynthetic pathway was knocked out in the engineered E. coli strain, growth medium had to be supplemented with tryptophan. Lin et al. (11) also demonstrated a new pathway for MA production starting from chorismate, with salicylic acid and catechol as intermediates. After knockout of genes that encoded enzymes which utilized chorismate in the Phe and Trp biosynthetic pathway and overexpression of aromatic pathway genes, MA production reached 1.2 g/liter in the shake flask cultures of E. coli. Since Phe and Trp production was eliminated in this pathway, the growth medium had to be supplemented with yeast extract.

In the present study, we used an alternative pathway for the production of MA from glucose. In the proposed pathway (Fig. 1), chorismate from the aromatic amino acid pathway was converted to MA via PHB. Three nonnative genes, namely, pobA, aroY, and catA, encoding 4-hydroxybenzoate hydrolyase, protocatechuate decarboxylase, and catechol 1,2-dioxygenase, respectively, were functionally expressed in E. coli to produce MA. The plasmid with these three nonnative genes also expressed two native genes, ubiC and aroF^FBR, which encoded chorismate pyruvate-lyase and 3-deoxy-arabino-heptulosonate 7-phosphate (DAHP) synthase, respectively. The aroF^FBR gene was a mutated aroF gene (P148L) which encoded a mutant DAHP synthase which was resistant to feedback inhibition (FBR) (12). In the proposed MA production pathway, chorismate was first cleaved by native chorismate pyruvate-lyase (encoded by ubiC) to PHB, which was converted to PCA by PHB hydrolyase (encoded by pobA), along with the conversion of NADPH to NADP. PCA was then decarboxylated to catechol by protocatechuate decarboxylase (encoded by aroY), and catechol was converted to MA by catechol 1,2-dioxygenase (encoded by catA). E. coli was transformed with the recombinant plasmid for the functional expression of the nonnative pathway for MA production. Deletion of several genes and overexpression of other native genes was also carried out to increase MA production.

FIG 1.

Proposed nonnative pathway for muconic acid synthesis. The proposed nonnative pathway converts the 4-hydroxybenzoic acid (PHB) from the ubiquinone pathway to muconic acid. Abbreviations: PEP, phosphoenolpyruvate; E4P, erythrose-4-phosphate; DAHP, 3-deoxy-arabino-heptulosonate 7-phosphate; DHQ, 3-dehydroquinic acid; DHS, 3-dehydroshikimic acid; SA, shikimic acid; S3P, shikimate-3-phosphate; EPSP, 5-enolpyruvyl-shikimate 3-phosphate; PHE, phenylalanine; TYR, tyrosine; TRYP, tryptophan; PYR, pyruvate; PHB, 4-hydroxybenzoic acid; PCA, protocatechuic acid. The genes shown encode proteins as follows: aroF, DAHP synthase; aroB, DHQ synthase; aroD, DHQ dehydratase; aroE, shikimate dehydrogenase; aroL, shikimate kinase; aroA, 3-phosphoshikimate-1-carboxyvinyltransferase; aroC, chorismate synthase; trpE and trpD, anthranilate synthase components I and II, respectively; tyrA and pheA, TyrA and PheA subunits of chorismate mutase, respectively; pobA, 4-hydroxybenzoic acid hydrolyase; aroY, protocatechuic acid decarboxylase; catA, catechol 1,2-dioxygenase; ubiC, chorismate pyruvate-lyase; ubiA, 4-hydroxybenzoate octaprenyltransferase; and ubiD, 3-octaprenyl-4-hydroxybenzoate carboxy-lyase.

MATERIALS AND METHODS

Construction of recombinant plasmids and transformation.

A nonnative pathway was introduced into E. coli for MA production. Individually, five genes—catA, aroY, pobA, ubiC, and aroF^FBR—were synthesized. The pobA gene was first subcloned into plasmid pTRCHis 2B (Life Technologies, USA) to construct recombinant plasmid pMA1g. All four remaining genes were subcloned subsequently into this plasmid to construct recombinant plasmid pMA5gO. A map of plasmid pMA5gO containing five genes is shown in Fig. 2a. In this plasmid, the expressions of the catA and pobA genes were under the control of a trc promoter and the expression of ubiC was under the control of a native gapA promoter. The expression of aroY and aroF^FBR was controlled by second copy of the trc promoter (Fig. 2a). The nucleotide sequence for the three nonnative genes was obtained from GenBank from different organisms: the pobA gene sequence was obtained from Pseudomonas putida KT2440 (accession number NC_002947), the aroY sequence from Klebsiella pneumoniae (accession number AB479384), and the catA gene sequence from Acinetobacter sp. strain ADP1 (accession number CR543861.1). The remaining two sequences were for native genes ubiC (EcoCyc accession number b4039/ECK4031) and aroF^FBR (GenBank accession number NC_000913), which encoded chorismate pyruvate-lyase and mutant phospho-2-dehydro-3-deoxyheptonate aldolase, respectively. The mutant gene aroF^FBR encodes an amino acid replacement (Pro-148 is replaced with Leu), and this replacement has been proven to be resistant to Tyr amino acid feedback inhibition (12).

FIG 2.

Diagrams of the two plasmids used in this study, pMA5gO (a) and pMA3gEL (b).

Further, to enhance the MA production levels, coexpression of three genes was necessary. These genes are those encoding shikimate dehydrogenase (aroE) (EcoCyc nucleotide sequence accession number EG 10077), shikimate kinase 2 (aroL) (EcoCyc nucleotide sequence accession number EG 10082), and transketolase 1 (tktA) (EcoCyc nucleotide sequence accession number EG 11427). These three genes were synthesized and subcloned into plasmid pEM7/Zeo (Life Technologies) to construct recombinant plasmid pMA3gEL. In pMA3gEL, the aroE and aroL genes were expressed under the control of a tac promoter and tktA was expressed under the control of a second tac promoter, as shown in the Fig. 2b.

The recombinant plasmids were amplified by transformation into electrocompetent E. coli Top 10 cells. Transformation was conducted by mixing 10 μl of the diluted plasmid (1:50 dilution) DNA into 50 μl of electrocompetent cells using the preset pulse program EC2 (2.49 kV and 6.10 ms) in the Gene Pulser electroporation system (Bio-Rad, Hercules, CA). The transformants were regenerated in 1 ml of Luria broth (LB) at 37°C and 120 rpm for 30 min. Portions (20 μl) of the regeneration culture were applied to plates containing LB-ampicillin at 100 mg/liter (pMA5gO) and low-salt LB broth-zeocin at 20 mg/liter (pMA3gEL) and incubated overnight. A few colonies were selected randomly and inoculated into LB-ampicillin and low-salt LB-zeocin and incubated overnight at 37°C and 120 rpm for plasmid isolation. The plasmid DNA was isolated from the overnight stationary-phase bacterial culture using a miniprep kit (Qiagen, Hilden, Germany) and eluted in a final volume of 50 μl. The isolated plasmids were verified for the correct size and transformed into electrocompetent BW25113 cells using the same protocol as described above.

Strains, media, and culture conditions.

Throughout this study, wild-type E. coli K-12 BW25113 and its derivative JW1666-3 [F⁻ Δ(araD-araB)567 ΔlacZ4787(::rrnB-3) λ⁻ ΔpykF751::kan rph-1 Δ(rhaD-rhaB)568 hsdR514] were used as base strains. JW1666-3 was obtained from the Keio collection (Coli Genetic Stock Center, Yale University) (13), and a new derivative called FPTS was created in this study by knocking out the ptsH, ptsI, and crr genes from this strain. ptsH, ptsI, and crr code for three proteins involved in the phosphoenolpyruvate phosphotransferase system (PEP-PTS). The gene knockout experiments were done by using the Quick & Easy E. coli gene deletion kit from Gene Bridges, Germany. First, the kanamycin resistance (Kan^r) gene was removed from the pykF site of JW1666-3 by introducing the FLP expression plasmid (e.g., 707-FLPe, catalog no. A104; Gene Bridges) at 30°C. A temperature upshift to 37°C resulted in the expression of an FLP site-specific recombinase enzyme, which carried out the FLP-mediated recombination. To knock out the ptsH, ptsI, and crr genes, the Kan^r gene was next inserted at the pts region containing these three genes. Gene-specific forward and reverse primers were designed to remove the pts region. A 73-bp forward primer, primer 1 (Table 1), carried the FLP recognition target (FRT) site and a 50-bp sequence, homologous to the chromosomal sequence directly upstream of the pts region. Similarly, a 72-bp reverse primer, primer 2 (Table 1), carried the FRT site and a 50-bp sequence, which was the reverse complement to the chromosomal sequence directly downstream of the pts region. Primer 1 and primer 2 were used to amplify the FRT-PGK-gb2-neo-FRT cassette, carrying the kanamycin antibiotic resistance gene, generating the functional cassette along with the homologous arms and FRT sites. The expected size of the PCR product for the FRT-PGK-gb2-neo-FRT cassette was 1,737 bp, which was verified by gel electrophoresis. The primers were designed so that the excision of the kanamycin resistance cassette with the FLP recombinase would create an in-frame deletion of the respective chromosomal gene in the future.

TABLE 1.

List of primers used in present study

Primer no.	Description	Sequence
1	pts KOa forward primer	5′AGGCTAGACTTTAGTTCCACAACACTAAACCTATAAGTTGGGGAAATACAA ATTAACCCTCACTAAAGGGCG3′
2	pts KO reverse primer	5′AAGCATAAAAAAATGGCGCCGATGGGCGCCATTTTTCACTGCGGCAAGAATAATACGACTCACTATAGGGCTC3′
3	pts KO upstream verification forward primer	5′ CAGGCTAAAGTCGAACCGCC3′
4	pts KO upstream verification reverse primer	5′ CCTCTCCACCCAAGCGGCCG3′
5	pts KO downstream verification forward primer	5′ CTTGGCGGCGAATGGGCTGAC3′
6	pts KO downstream verification reverse primer	5′ TTGATTTTGCCGCCGCTGGCG3′

^aKO, knockout.

The JW1666-3 strain was first transformed with pRedET plasmids which expressed the Red/ET recombination proteins on being induced with 10% l-arabinose. Next, the linear FRT-PGK-gb2-neo-FRT fragments with homologous arms were transformed into the cells and the cultures were incubated at 37°C with shaking for 3 h for genomic recombination to occur. Thus, the pts region was knocked out, generating the FPTS strain. The integration of the kanamycin resistance gene at pts site in the FPTS genome was verified by PCR. Upstream ligation was verified by primer 3 (Table 1), which corresponds to sequence upstream of the pts region, and primer 4 (Table 1) within the FRT-flanked kanamycin resistance cassette. Similarly, downstream ligation was verified by primer 5 (Table 1) within the kanamycin resistance cassette and primer 6 (Table 1), which corresponds to a sequence downstream of the pts region. The expected upstream product was 784 bp and the expected downstream product was 334 bp, which was verified by gel electrophoresis.

pMA3gEL and pMA5gO were next introduced into the FPTS strain. All the wild-type cultures were maintained on LB agar plates, and knockout cultures were recovered on LB agar plates containing 10 mM sodium pyruvate and kanamycin (25 μg/ml). pMA5gO-containing cultures were selected on LB containing ampicillin (100 μg/ml), and pMA3gEL-containing cultures were selected on low-salt LB medium containing zeocin (20 μg/ml). The transformants carrying pMA3gEL and pMA5gO were selected on M9 minimal medium plates containing three antibiotics: ampicillin (100 mg/ml), kanamycin (25 mg/ml), and zeocin (20 mg/ml). All the production studies were done in modified M9 medium containing the following (per liter): NaH₂PO₄·7H₂O, 12.8 g; K₂HPO₄, 3 g; NH₄Cl, 1 g; NaCl, 0.5 g; MgSO₄, 0.5 g; CaCl₂, 0.0275 g; and thiamine chloride, 10 mg, as well as trace elements (0.1% [vol/vol]). The trace elements included ferric citrate (3 g/liter), CoCl₂ (0.065 g/liter), ZnSO₄·7H₂O (0.05 g/liter), MnCl₂·4H₂O (0.015 g/liter), NaMo₄·2H₂O (0.015 g/liter), NiCl₂·6H₂O (0.01 g/liter), and CuSO₄·5H₂O (0.005 g/liter). Glucose (10 g/liter) was the sole carbon source in the medium. A 0.25 mM concentration of isopropyl-β-d-thiogalactopyranoside (IPTG) was added to induce the expression of the nonnative genes. All the cultures were grown at 37°C and 180 rpm in orbital shakers. Preinocula of the transformed strains were grown for 16 h on modified M9 minimal medium with all appropriate antibiotics at 37°C and 160 rpm. The preinocula were used to inoculate three sets of flask, each having 50 ml of modified M9 minimal medium (0.5% [vol/vol]), with appropriate antibiotics, and grown for 24 h. The time of induction also had an influence on MA production. The cultures were induced for the first time around an optical density at 600 nm (OD₆₀₀) of 2 and were subsequently induced every 24 h.

All the synthetic oligonucleotides were obtained from Sigma-Aldrich (Singapore). All DNA manipulations and microbiological experiments were carried out by standard methods (14). Plasmids were isolated from bacteria using a plasmid minikit from Qiagen, Netherlands.

In silico analysis of E. coli metabolic network.

There are several methods for in silico analysis of metabolic networks for identification of candidates for genetic modification, knockout, and insertion of new reactions, which potentially increase the yield of desired products. Most of these methods use optimization algorithms for identification of gene or reaction targets for modification. OptKnock (15) is the first bilevel mixed-integer linear programming (MILP) method to identify gene knockout strategies that increases the yield of the desired chemical while the algorithm internally maximizes the biomass. Though this approach was implemented successfully in a few cases, there is no guarantee that the targets identified by OptKnock increase product yield in vivo due to multiple competing pathways. RobustKnock (16) was later proposed, extending the bilevel concept that maximizes the minimum yield of product while inherently maximizing biomass. Identification of metabolic targets by maximizing the minimum yield of product results in reaction knockout candidates by considering all competing pathways, which increases the in vivo success rate. New algorithms called OptStrain (17) and OptReg (18) have also been developed to identify gene overexpression and new gene insertion candidates. In this study, we have implemented the RobustKnock algorithm for the analysis of the E. coli genome-scale metabolic network for the production of MA.

Detection and measurement of PCA and MA.

PCA is the first component in the proposed nonnative MA pathway. The pobA gene encodes 4-hydroxybenzoate hydrolyase, which catalyzes the conversion of 4-hydroxybenzoic acid to PCA. To our knowledge, this is the first time that the pobA gene from Pseudomonas putida has been expressed in E. coli. To validate the functional expression of the pobA gene, PCA was detected and quantified by measuring the optical density of the ferric complex at 640 nm as previously described by Gross (19). A volume of 500 μl of Tris-maleate buffer (pH 5.2) was added to 750 μl of cell supernatant to maintain the pH, since assays at pH 5.3 to 5.5 gave consistent extinction coefficients at 640 nm. Next, 967 μl of the above-described mixture was added to 33 μl of 1% ferric chloride solution and mixed well. From each sample, 300-μl samples were pipetted into 3 wells in the 96-well plate and the optical density of each sample was measured at 640 nm using the Varioskan Flash spectral scanning multimode reader. The optical density was compared against the calibration curve, which varied linearly with PCA concentration.

MA, PHB, PCA, and catechol were measured by high-performance liquid chromatography (HPLC). The standards were first run to identify their retention times. The retention times for MA, catechol, PCA, and PHB were 21.26, 31.8, 38.2, and 61.96 min, respectively. The metabolites were separated by HPLC (Shimadzu) using an Aminex HPX-87H column (Bio-Rad Laboratories, Inc.). The column was eluted with 12 mM H₂SO₄ as the mobile phase with a flow rate of 0.65 ml/min, and the temperature was maintained at 50°C. For metabolite analysis, the bacterial cells were removed from the samples by centrifugation and the supernatant was filtered using 0.45-μm filters. Glucose, lactic acid, acetic acid, and ethanol were detected by a Shimadzu RID-10A refractive index detector. The retention times for glucose, acetic acid, lactic acid, and ethanol were 8.45, 11.88, 14.14, and 19.98 min, respectively. A Shimadzu Prominence SPD-20AV UV-visible (UV-Vis) detector was used for the detection of MA, PHB, PCA, and catechol (270 nm). For data evaluation, LabSolutions software (Shimadzu) was used.

RESULTS

Functional validation of the first nonnative gene, pobA, leading to PCA production.

The first heterogeneous gene of the nonnative metabolic pathway, pobA, was genetically introduced into BW25113 and JW1666-3 in order to develop a strain which could convert PHB to PCA. The transformed cultures carrying plasmid pMA1g were grown for 2 h at 37°C and 180 rpm. IPTG (0.25 mM) and PHB (0.5 g/liter) were added after 2 h of growth, and the culture was allowed to grow for 24 h. The concentration of PCA in the supernatant was detected spectrophotometrically by measuring the OD at 640 nm by the method described under “Detection and measurement of PCA and MA” above. Transformed strains BW25113 and JW1666-3 produced 364.4 mg/liter and 753.45 mg/liter of PCA, respectively. This indicated that the nonnative pobA gene obtained from Pseudomonas putida was functionally active in BW25113 and that the enzyme encoded by this gene was capable of performing the desired chemical conversion in vivo. Furthermore, strain JW1666-3, in which the pykF gene had been deleted, produced a higher concentration of PCA than did the wild type, BW25113.

Functional validation of the nonnative metabolic pathway.

A nonnative metabolic pathway was genetically introduced into E. coli BW25113 to study the conversion of chorismate to MA via PHB, PCA, and catechol as shown in Fig. 1. Detection of PCA in the strains expressing pobA confirmed the conversion of PHB to PCA and indicated substrate availability for the enzymes expressed in the nonnative pathway for MA production. The transformed cultures carried plasmid pMA5gO, which contained the nonnative pobA, aroY, and catA genes and native ubiC and aroF^FBR genes. Expression of ubiC increased the availability of PHB, which is the precursor for the proposed nonnative pathway. Expression of aroF^FBR increased the concentration of feedback-uninhibited DAHP synthase that catalyzed the first committed step in the aromatic amino acid pathway, leading to the production of more DAHP (Fig. 1). The culture was grown in minimal medium for 4 h at 37°C and 180 rpm, then IPTG (0.25 mM) was added, and the culture was allowed to grow after induction. The samples were collected after 24 h of growth, and MA production was measured using HPLC, as described under “Detection and measurement of PCA and MA” above. BW25113 when transformed with pMA5gO produced only 0.56 mg/liter of MA after 24 h of growth in a shake flask (Table 2). Though this strain produced a very low concentration of MA, the results confirmed that the proposed nonnative pathway is functionally active in E. coli. The challenges in further increasing the MA production were to increase the precursor compounds for both the aromatic amino acid pathway and the nonnative MA pathway, increase the expression of enzymes in the aromatic amino acid pathway, and reduce the accumulation of nonnative pathway intermediates. To identify strategies for addressing these challenges, we analyzed the genome-scale metabolic network of E. coli, followed by experimental verification of the identified strategies; the results are given below.

TABLE 2.

Muconic acid production by the wild-type and transformed knockout strains^a

Strain	OD600	Concn (mg/liter) of acid produced
		Muconic acid	PCA
BW25113	5.6	0.09
BW25113 transformed with pMA5gO	6.2	0.56
JW1666-3	3.5	0.07
JW1666-3 transformed with pMA5gO	4.3	6.26	1.8
FPTS transformed with pMA5gO	8.63	44.55	0.751
FPTS transformed with pMA3gEL and pMA5gO	8.92	120

^aThe results indicate the biomass attained and acid produced after 72 h of growth in M9 medium with glucose (10 g/liter). The experiment was done in triplicates (biological replicates). Standard deviations were within 10% of the values shown. No PHB was detected.

Analysis of the E. coli model for MA production.

The E. coli metabolic network was extensively analyzed to identify genetic targets for developing a muconic acid-producing strain. For this analysis, the nonnative pathway for MA production was included in the E. coli metabolic network model containing 848 reactions and 765 metabolites. Network analysis was performed using the RobustKnock method, and several candidate reactions were identified to be knocked out from the E. coli genome to increase the metabolic flux toward MA. The candidate reactions for knockout studies are shown in Table 3.

TABLE 3.

Reaction knockout candidates identified by in silico analysis of the E. coli metabolic network

Reaction name	Reaction formulaa
PTS	Glc-D[e] + PEP[c] → G6P[c] + PYR[c]
PYK	ADP[c] + h[c] + PEP[c] → ATP[c] + PYR[c]
TALA	G3P[c] + S7P[c] ↔ E4P[c] + F6P[c]
GND	6PGC[c] + NADP[c] → CO2[c] + NADPH[c] + Ru5p-D[c]
AKGDH	AKG[c] + CoA[c] + NAD[c] → CO2[c] + NADH[c] + SucCoA[c]
MDH	MAL-L[c] + NAD[c] ↔ h[c] + NADH[c] + OAA[c]

^aGlc-D, d-glucose; 6PGC, d-gluconate 6-phosphate; AKG, alpha-ketoglutarate; CoA, coenzyme A; SucCoA, succinyl CoA; MAL-L, malate; OAA, oxaloacetate.

As shown in Table 3, the analysis of metabolic network of E. coli identified several candidate reactions to be knocked out to increase MA production. Not surprisingly, some candidate reactions identified by our method matched the literature, while some were novel candidates. For example, the phosphoenolpyruvate phosphotransferase system for glucose transport (GLCpts) and pyruvate kinase (PYK) reactions that are involved in the uptake of glucose by E. coli and production of pyruvate (PYR) from phosphoenolpyruvate (PEP) have already been proved to increase aromatic amino acid production experimentally (20). As these two reactions consume PEP, deletion of these two reactions from E. coli increased the availability of PEP, which is one of the precursor compounds for the aromatic amino acid pathway (Fig. 1). The metabolic network analysis identified other reaction knockout candidates which were novel and to our knowledge not reported in the literature. Two of those reactions are catalyzed by gluconate dehydrogenase (GND) and transaldolase A (TALA) and have an indirect effect of increasing the availability of erythrose-4-phosphate (E4P), which is the second compound required for aromatic amino acid synthesis. We obtained simulated results obtained from the model analysis after in silico deletion of the candidate reactions. Deletion of candidate reactions increases the flux for the TKTA reaction (F6P[c] + G3P[c] ↔ E4P[c] + Xu5P-D[c]) from 0.5116 to 0.7735 mol/mol of glucose, which increases the production of E4P (Fig. 1). The effects of other candidate reactions catalyzed by alpha-ketoglutarate dehydrogenase (AKGDH) and malate dehydrogenase (MDH) are not obvious, but we believe that these reactions are involved in balancing the cofactors or reducing the glucose flux to biomass production. In the model simulations, deletion of the candidate reactions increased the production of MA to 0.46 mol/mol of glucose, while no MA production was produced without these knockouts. The in silico results were validated in vivo by knocking out candidate genes and overexpressing some native genes as discussed below.

Experimental validation of model predictions for MA production.

In silico analysis of the E. coli metabolic network resulted in several gene knockout candidates that could increase muconic acid production. We have experimentally created the knockouts and verified them for the improvement in MA production. The final results of the different knockout experiments after 72 h of growth are given in Table 2. The wild-type E. coli strain, BW25113, produced only 0.56 mg/liter of MA with the plasmid carrying the nonnative MA pathway genes, pMA5gO, when grown for 3 days in shake flasks. Under similar growth conditions, strain JW1666-3 produced 6.26 mg/liter of MA. Deletion of the pykF gene increased MA production 11-fold compared to the production by the wild-type strain with the same nonnative pathway. Knocking out the pts region carrying the ptsH, ptsI, and crr genes further increased the MA production (by 7.1 times), to 44.55 mg/liter, as seen in Table 2. Knockout of the pykF gene and pts region resulted in decreased utilization of PEP by these reactions and potentially increased PEP availability for the aromatic amino acid pathway. This led to increased production of MA, which corresponded with the model-based predictions leading to MA production.

In silico analysis of the E. coli metabolic network predicted two more gene knockouts, those of gnd and talA. Simulated results show that knockout of these two genes increases the flux in the reaction that produces E4P. Hence, these two gene knockouts were replaced by overexpression of the tktA gene (encoding transketolase), which catalyzes a step in the pentose phosphate pathway to increase the E4P concentration in the cell. Further, overexpression of two native genes, aroE and aroL, in E. coli was expected to increase the expression of the genes coding for the enzymes shikimate dehydrogenase and shikimate kinase, respectively, in the aromatic amino acid pathway, thereby helping to increase MA production. Hence, a second plasmid, pMA3gEL, was introduced which carried the aroE and aroL genes under the control of a tac promoter and the tktA gene under the control of a second tac promoter into E. coli. Introduction of the above-mentioned genes increased MA production from 44.55 mg/liter to 120 mg/liter, as seen in Table 2. To check the effect of the induction time on MA production, we conducted production studies by increasing the time of induction from an OD₆₀₀ of 0.6 to an OD₆₀₀ of 2.7. It has been observed that the final concentration of MA increased with increase in the initial biomass (measured as OD₆₀₀) from 80 mg/liter to around 170 mg/liter, as seen in Table 4. The growth, glucose consumption, and MA production of the culture induced at the initial OD₆₀₀ are shown in Fig. 3.

TABLE 4.

Effect of time of induction in FPTS pMA3gEL pMA5gO cultures on MA production in M9 medium^a

	Value for culture with the following OD600 at induction:
	0.6		2.0		2.7
Time (h)	OD600	Concn of MA produced (mg/liter)	OD600	Concn of MA produced (mg/liter)	OD600	Concn of MA produced (mg/liter)
24	7.47	81.46	7.79	138.16	7.3	129.5
48	5.42	85.07	8.9	139.96	6.56	149.5
72	8.92	89.12	9.37	140	6.3	170

^aThe experiment was done in triplicates (biological replicates). Standard deviations were within 10% of the values shown.

FIG 3.

Growth, glucose consumption, and muconic acid production profiles for the FPTS strain with plasmids pMA3gEL and pMA5gO induced at an initial OD₆₀₀ of 2.7.

DISCUSSION

Production of value-added chemicals from biomass is the key factor for sustainability and reduction of emissions from the use of fossil resources. In this study, we have engineered an E. coli strain for the production of MA from glucose by functionally expressing a new nonnative pathway consisting of three genes, pobA, aroY, and catA. These nonnative pathway genes code for enzymes that catalyze the conversion of PHB to MA. PHB is an intermediate in the ubiquinone pathway, which branches out from chorismate in the aromatic amino acid pathway in E. coli (Fig. 1). The novelty of the proposed pathway is the use of PHB from the ubiquinone pathway as an intermediate for the production of MA. In the ubiquinone pathway, the second enzymatic step that converts PHB to 3-octaprenyl-4-hydoxybenzoate is the rate-limiting step (21). Hence, production of MA using the proposed pathway is benefitted by utilizing almost all the PHB. Also, the chorismate lyase enzyme that converts the chorismate to PHB produces pyruvate as a coproduct, which links the proposed MA production pathway to the growth of E. coli.

The aromatic amino acid pathway in E. coli is complex, and there are several control points, including the feedback inhibition of the first committing step in the aromatic amino acid pathway, i.e., condensation of E4P and PEP for the production of DAHP. The genes, aroF, aroH, and aroG, which code for the enzymes for this step, are feedback inhibited by the aromatic amino acids Tyr, Trp, and Phe, respectively (1, 22, 23). Deregulation of the feedback inhibition is essential to divert glucose to the aromatic amino acid pathway, and mutations were reported in the literature for the aroF and aroG genes that reduce or eliminate the feedback inhibition (12, 24). In this study, the single amino acid mutant aroF gene, renamed as aroF^FBR, was overexpressed. The enzyme chorismate lyase is also feedback inhibited by PHB (25). So the ubiC gene, encoding chorismate lyase, was also overexpressed in this study. Expression of the proposed nonnative pathway along with overexpression of E. coli native genes ubiC and aroF^FBR resulted in only a small amount of MA from the glucose. To further increase MA production, several native genes were knocked out and overexpressed in the engineered E. coli. Knockout of the pykF, ptsI, ptsH, and crr genes and overexpression of the tktA gene increased the MA production severalfold. All the genes knocked out from E. coli in this study are involved in the consumption of PEP, and overexpression of tktA could increase the production of E4P. Finally, additional copies of the genes aroE and aroL were introduced for enhanced production of chorismate from DAHP, which was channelized toward the nonnative pathway to increase MA. All these genetic modifications resulted in the production of 170 mg/liter of MA from glucose.

In contrast to the previous approaches, the proposed pathway does not block the aromatic amino acid or other downstream pathways but relies on the natural feedback inhibition of the chorismate utilizing enzymes and regulation of the ubiquinone pathway. Hence, the MA production was achieved solely from glucose and medium was not enriched with aromatic amino acids. As the proposed pathway converts PHB to MA, it is particularly interesting to use PHB from lignin depolymerization as an additional carbon source along with glucose. Hence, the present work is of wider importance in the context of genetic strategies to improve the microbial production of the industrially important chemical MA from simple glucose. Studies with different expression levels of genes with promoters of various strengths, optimization of medium components, and fed-batch fermentation experiments are required to further increase MA production from glucose.