Genome engineering Escherichia coli for L-DOPA overproduction from glucose

Genome engineering has become a powerful tool for creating useful strains in research and industry. In this study, we applied singleplex and multiplex genome engineering approaches to construct an E. coli strain for the production of L-DOPA from glucose. We first used the singleplex genome engineering approach to create an L-DOPA-producing strain, E. coli DOPA-1, by deleting transcriptional regulators (tyrosine repressor tyrR and carbon storage regulator A csrA), altering glucose transport from the phosphotransferase system (PTS) to ATP-dependent uptake and the phosphorylation system overexpressing galactose permease gene (galP) and glucokinase gene (glk), knocking out glucose-6-phosphate dehydrogenase gene (zwf) and prephenate dehydratase and its leader peptide genes (pheLA) and integrating the fusion protein chimera of the downstream pathway of chorismate. Then, multiplex automated genome engineering (MAGE) based on 23 targets was used to further improve L-DOPA production. The resulting strain, E. coli DOPA-30N, produced 8.67 g/L of L-DOPA in 60 h in a 5 L fed-batch fermentation. This titer is the highest achieved in metabolically engineered E. coli having PHAH activity from glucose.


Removal of transcriptional regulators. Tyrosine repressor (TyrR) is a transcriptional dual regulator that
represses the transcription of several genes encoding enzymes involved in aromatic acid biosynthesis 23 . Carbon storage regulator A (CsrA) is a regulator of carbohydrate metabolism. CsrA regulates the levels of three enzymes that participate directly in phosphoenolpyruvate (PEP) metabolism. It activates pyruvate kinase (PykF) and represses PEP carboxykinase (PckA) and PEP synthase (PpsA) in E. coli 24 . It has been reported that the inactivation of tyrR and csrA improves aromatic compound production [25][26][27][28] . Thus, we first deleted tyrR and csrA to obtain E. coli AROM-1 (Fig. 1), resulting in a slight increase in L-DOPA production from 138.7 ± 4.9 mg/L to 148.3 ± 11.7 (Table 1). Munoz et al. also reported that knocking out tyrR enhanced L-DOPA production in E. coli 16 . Increasing the availabilities of the precursor PEP by altering glucose transport. Increasing PEP availability is a common strategy for engineering E. coli strains for the overproduction of aromatic compounds. In E. coli, glucose is mainly transported and phosphorylated by the phosphotransferase system (PTS). Under standard growth conditions, 50% of the glycolytic intermediate PEP resulting from the catabolism of glucose is used as the phosphate donor for phosphorylation and translocation by the PTS. The properties of the PTS limit the production of compounds that have PEP as a precursor. Carmona et al. suggested that inactivation of the PTS is the primary strategy for engineering E. coli to overproduce aromatic metabolites 29 . Thus, we deleted the PTS (ptsHIcrr) to further improve L-DOPA production. The inactivation of the PTS increased the L-DOPA titer to 176 ± 3.6 mg/L (Table 1). Non-PEP-mediated glucose transport and phosphorylation systems have successfully been used for the replacement of the PTS to increase PEP availability [30][31][32] . Thus, we integrated the galP and glk  under the control of the P37 promoter into the E. coli knockout strain AROM-2 to obtain E. coli AROM-3. The titer of L-DOPA and growth of E. coli AROM-3 harboring pQE30-hpaBC showed no significant difference compared to E. coli AROM-2 (p < 0.05, Table 1).
Knockout of Glucose-6-phosphate dehydrogenase gene. Glucose-6-phosphate dehydrogenase (encoded by zwf) catalyzes the oxidization of glucose-6-phosphate to gluconate-6-phosphate. It has been reported that knocking out zwf drives more carbon flux into the Embden-Meyerhof-Parnas (EMP) pathway and tricarboxylic acid (TCA) cycle 33 . They also found that the zwf mutant is able to synthesize pentose phosphate (PP) pathway-derived compounds independently from the oxidative part of the PP pathway by directing its carbon flow from the EMP pathway directly into the non-oxidative part of the PP pathway. Thus, we disrupted zwf in E. coli AROM-3 to obtain E. coli AROM-4. E. coli AROM-4 (pQE30-hpaBC) produced L-DOPA at 205.3 ± 2.5 mg/L, which was greater than E. coli AROM-3 (pQE30-hpaBC) ( Table 1). The stoichiometric analysis demonstrated that the yield of the aromatic compound DAHP approaches the theoretical maximum when E4P is provided by the nonoxidative part of the PP pathway and pyruvate is recycled to PEP by PpsA 34 . The improvement of L-DOPA titer after zwf deletion was experimentally demonstrated for the first time.
Removal of competing pathway. Prephenate can be converted into either tyrosine or phenylalanine. To eliminate the loss of prephenate to the competing reaction (phenylalanine biosynthesis), we deleted prephenate dehydratase and its leader peptide genes (pheLA) in E. coli AROM-4 to obtain E. coli TYR-1. The pheLA deletion slightly increased the L-DOPA titer to 209.2 ± 0.9 mg/L ( Table 1). Some other groups have previously reported that the pheLA deletion increases L-tyrosine production 35,36 . Coordinating expression of the downstream pathway of chorismate. The bifunctional enzyme Chorismate (CHA) mutase/prephenate dehydrogenase, TyrA, catalyzes the first and second step of L-tyrosine biosynthesis ( Fig. 1). TyrA catalyzes both reactions in separate domains of the protein, and the CHA mutase/prephenate hydrogenase is feedback-inhibited by L-tyrosine (up to 95% inhibition of the prephenate dehydrogenase and 45% of the CHA mutase activity 28 . Feedback-resistant mutants of the TyrA E. coli enzyme have been used for L-tyrosine overproduction 35,36 . Thus, TyrA fbr [M53I/A354V] was used to deregulate the feedback inhibition by tyrosine. Substrate channeling is a powerful tool for balancing the expression of genes. It can increase the catalytic efficiency of the sequential reactions in a biosynthetic pathway 37,38 . To increase the rate of CHA conversion to L-DOPA, we first fused the tyrA fbr , tyrB and hpaBC genes with a (G 4 S) 3 linker, then integrated the fusion protein chimera under the control of the 7P37 promoter into the chromosome of E. coli TYR-1 to obtain E. coli DOPA-1. E. coli DOPA-1 produced 307.4 ± 3.7 mg/L of L-DOPA.
Multiplex automated genome engineering. MAGE is an efficient and rapid tool for the genome engineering of bacterial strains. We selected aroF, aroG, aroB, aroD, ydiB, aroE, ppsA, tktA, nadK, aroL, aroK, aroA, tyrA, tyrB and tyrA fbr (M53I/A354V) as target sites to tune translation by ribosome binding site (RBS) replacement (Fig. 1). The RBS sequences were designed to be DDRRRRRDDDD (D = A, G, T; R = A, G) with a total pool complexity of 3.5 × 10 5 (3 6 × 2 5 × 15). Six genes (aroF P148L , aroG D146N , tyrA M53I , tyrA A354V , rpoD D521E and rpoA V257R ) were targeted for amino acid mutations in their open reading frames (ORF). The introduction mutations in aroF, aroG and tyrA were used to remove product feedback inhibition 23,[26][27][28]35,36 . The rpoD and rpoA mutants have been successfully used to increase tyrosine production 39 . Two genes (trpD and trpE) were targeted for inactivation by introducing a revertible premature stop codon into each ORF. To increase the MAGE allelic replacement frequency, the methyl-directed mismatch repair protein gene (mutS) of E. coli DOPA-1 was first deleted to obtain E. coli DOPA-2. E. coli DOPA-2 (pSIM6) was used as the starting strain for MAGE. After 30 cycles of MAGE, 1.3 × 10 10 genetic variants (4.3 × 10 8 bp variations per cycle for 30 MAGE cycles 19 ) were generated. According to an allelic replacement efficiency calculation 22 , 30 MAGE cycles generate 2.3% of genomes with at least 3 out of 23 targeted loci and 6.1 × 10 −12 of genomes with all 23 targeted loci. One hundred clones from the 5th, 10th, 15th, 20th and 25th cycle and 1000 clones from the 30th cycle were screened in deep-well microplate culture. L-DOPA can be easily oxidized to dopachrome and then polymerized nonenzymatically to form the black pigment  Table 1. L-DOPA production in different E. coli strains*. * Experiments were conducted in triplicates, and measurements are presented with their means and s.d.
melanin 40 . Thus, we selected strains that produced darker cultures for further analysis. Darker cultures in the 48-well microplates were selected for HPLC analysis to determine L-DOPA concentration. Six MAGE strains from the 30 th cycle showed higher L-DOPA concentrations in the deep-well microplate analysis, and these were further analyzed in shake flasks. Of the six strains, strain 30-30 produced the highest level of L-DOPA, which was 34% higher than that of the starting strain E. coli DOPA-2 ( Table 2). Table 2 also shows that all MAGE strains produced more tyrosine and total tyrosine plus L-DOPA than the starting strain. The reason may be because the above modification strategies were used to increase the availability of the precursor, tyrosine.  Table 2, not all of the tyrosine was converted to L-DOPA in E. coli DOPA-30. In order to convert all L-tyrosine into L-DOPA, we added a single additional copy of the hpaBC into pQE30-hpaBC to obtain pQE30-2hpaBC and transformed the plasmid into E. coli DOPA-30. As shown in Table 3, overexpression of hpaBC in E. coli DOPA-30 indeed increased L-DOPA production, but this strain cannot also convert all the L-tyrosine into L-DOPA. However, the engineered E. coli with the hpaBC reported by Munoz et al. produced few L-tyrosine 16 . Comparing the sequence of the hpaBC in pQE30-2hpaBC with that reported by Munoz et al. 16 , the 5′ -UTR sequence of the hpaC has been changed. The change may lead to the imbalanced expression between the hpaB and hpaC. Is this change resulted in the accumulation of L-tyrosine in the engineered strain? We re-amplified the hpaBC operon with the native 5-UTR sequence of the hpaC to obtain pQE30-hpaBC N . As shown in Table 3, E. coli DOPA-30 harboring pQE30-hpaBC N cannot produce L-tyrosine. Thus, the hpaBC in E. coli DOPA-30 was replaced with the hpaBC N to obtain E. coli DOPA-30N. As shown in Table 3, E. coli DOPA-30N cannot also produce L-tyrosine and produced 614.3 mg/L of L-DOPA.
Fed-batch fermentation. Fed-batch fermentation of E. coli DOPA-30N was performed in a 5 L bioreactor.
As shown in Fig. 2, the strain produced 8.67 g/L of L-DOPA at 60 h. The OD 600 of the culture reached 110. The L-DOPA productivity was 144.5 mg/L/h. The L-DOPA yield from glucose was 62.7 mg/g. The titer and yield were 5.7-and 1.2-fold higher than that reported by Muñoz et al. 15 , respectively. In addition, it was found that all the L-tyrosine was converted to L-DOPA after 40 h. The similar phenomenon was also observed by Muñoz et al. 15 . It indicates that the rate of hydroxylation of L-tyrosine by the HpaBC is slower than the rate of L-tyrosine synthesis. Therefore, the catalytic efficiency of the PHAH encoded by hpaBC should be improved.
Comparison with other microorganisms. L-DOPA production by microorganisms is summarized in Table 4. The L-DOPA titer obtained in this study is higher by a factor of 5.7 than the highest level previously reported using metabolically engineered E. coli strain that have PHAH activity from glucose 16 . The value is also higher than that obtained in microorganisms that have tyrosinase activity from tyrosine [2][3][4][5][6][7][8] . However, the value in this study is lower than that obtained in some microorganisms with Tpl activity from catechol and pyruvate 9,10,12 .
It indicates that further works should be carried out for improving L-DOPA production.   Although the L-DOPA titer of our engineered E. coli is considerably higher than that previously reported, all of the tyrosine was converted to L-DOPA only after 40 h (Fig. 2). It indicates that PHAH is the rate-limited step for L-DOPA biosynthesis in this strain. The catalytic efficiency of the PHAH encoded by hpaBC should be improved. Directed evolution may be used to increase its catalytic efficiency. Because only three targets were found in the MAGE strain (Supplementary Table 1), we can apply other strategies to further enhance the availability of tyrosine, such as upregulating tktA, increasing NADPH availability and upregulating hpaBC.
In conclusion, we first constructed an L-DOPA-producing E. coli strain, DOPA-1, using a singleplex genome engineering approach based on knockouts of genes and integration of the tyrA fbr , tyrB and hpaBC fusion protein chimera. MAGE based on 23 targets was then used to further improve L-DOPA production, which yielded the strain E. coli DOPA-30N. E. coli DOPA-30N produced 8.67 g/L of L-DOPA in 60 h in a 5L fed-batch fermentation. This titer is the highest reported in metabolically engineered E. coli that has PHAH activity from glucose. This strain, E. coli DOPA-30N, can serve as a base strain for developing more efficient strains capable of producing L-DOPA or other aromatic compounds. The rapid and efficient markerless deletion approach using the IPTG-inducible ccdB as a counter-selectable marker will be generally useful for gene knockout of E. coli.

Methods
Strains, plasmids and primers. The strains and plasmids used in this study are listed in Table 5 Genetic methods. The genes hpaB and hpaC were amplified from E. coli W using the primers hpaB-F/ hpaB-R and hpaC-F/hpaC-R, respectively. The hpaB fragment was cloned into the SacI/KpnI sites of pQE30 to obtain pQE30-hpaB. The hpaC fragment was cloned into the KpnI/SalI sites of pQE30-hpaB to obtain pQE30-hpaBC. The hpaBC genes were also amplified from pQE30-hpaBC using the primers hpaBC-F/hpaBC-R and then cloned into the SalI/HindIII sites of pQE30-hpaBC to obtain pQE30-2hpaBC. The hpaBC operon was amplied from E. coli W using the primers hpaB-F/hpaC-R and then cloned into the SacI/SalI to obtain pQE30-hpaBC N .   The knockouts of the csrA, tyrR and mutS genes were carried out according to the one-step inactivity method 42 with the help of the pSIM6 plasmid 43 expressing the lambda red recombination system. Gene knockouts were verified by colony PCR using appropriate primers (Supplementary Table 2).
The knockouts of other genes were carried out by a two-step recombination method using lambda red recombination and I-SceI cleavage as described as in Supplementary Fig. 1. The method was first reported by Yu et al. 44 . They used sacB as the counter-selectable marker. However, the efficiency of the first recombination is very low (24%) because sacB generally results in a certain number of false-positive colonies in the screening process due to mutation of sacB 45 . Thus, we used the IPTG-inducible ccdB gene as the counter-selectable marker. The ccdB gene was amplified from pOSIP-CH 46 using the primers ccdBF/ccdBR, then cloned into the HindIII/XbaI sites of pXMJ19 47 to obtain pXMJ-ccdB. The plasmid pXMJ-ccdB was digested by HindIII, blunted and self-ligated to obtain pEC-ccdB* . The IPTG-inducible ccdB gene was amplified from pXMJ-ccdB* using the primers ccdB* F/ccdB* R, then cloned into pMD18 to obtain pMD-lacI-P tac ccdB. A kan resistance gene (encoding aminoglycoside 3′ -phosphotransferase) containing I-SceI recognition sites was amplified from pK-JL 48 using the primers kanF/kanR and then cloned into the XhoI/SpeI sites of pMD-lacI-P trc ccdB to obtain pMD-ccdBKanS. The I-Scel endonuclease gene was synthesized by Suzhou GENEWIZ, Inc. (Suzhou, China) and ligated into pUC57 to obtain pUC57-I-SceI. The I-Scel was cut from pUC57-I-SceI by EcoRI/KpnI and cloned into pBAD30 49 to obtain pBAD30-I-SceI. The arabinose-inducible I-Scel was amplified from pBAD30-I-SceI using the primers  IsceIF/IsceIR and cloned into the NdeI site of pSIM6 50 to obtain pSIMIS. The efficiency of the first recombination of the method reached 80.3%, which was much higher than that based on the sacB (24%, Supplementary Table 3).
The replacement of 5′ -UTR of the hpaC in E. coli DOPA-30 was carried out by the CRIPR-Cas method as described by Jang et al. 53 . The sgRNA fragment was amplified from pTargetF using the primers hpaCN20F/ hpaCN20R and then cloned into the SpeI/XhoI sites of pTargetF to obtain the sgRNA plasmid pTargetF-hpaC. The target fragment was amplied from pQE30-hpaBC N using the primers hpaB/hpaBC.

MAGE and Screening of MAGE strains.
Oligos were mixed in equimolar amounts to reach a final total oligo concentration of 1 μ M. MAGE cycling was performed as previously described [19][20][21] . In brief, E. coli DOPA-3 harboring pSIM6 was grown in a 20-mL conical tube containing 5 mL of LB medium supplemented with 100 μ g/mL ampicillin at 30 °C with 200 rpm agitation until the OD 600 reached 0.5 to 0.7. Then, the cultures were heat-shocked in a shaking water bath at 42 °C for 15 min to induce the expression of λ Red recombination genes (gam, bet and exo). The cells were then chilled to 4 °C and centrifuged at 11,000 rpm for 30 s at 4 °C. The cultures were washed three times with ice-cold sterile 10% glycerol to remove salts. The cells were resuspended in 50 μ L oligo mixture. Electroporation was carried out at 1.8 kV in 1-mm gap cuvettes on a Bio-Rad MicroPulser, BTX ECM-830. Cells were incubated in fresh LB low salt medium at 30 °C until their OD 600 reached 0.4 to 0.6. The processes were repeated 30 times (30 MAGE cycles). After 5, 10, 15, 20, 25 and 30 cycles, the cells were grown overnight in 50 mL LB low salt medium and stored at − 80 °C in a 15% (v/v) glycerol solution.
Cells from the 5th, 10th, 15th, 20th, 25th and 30th cycles were diluted, plated onto LB-agar plates with ampicillin and cultured overnight. Individual colonies were inoculated in individual wells of a 48-well deep-well microplate (4.6 mL) containing 600 mL of the fermentation medium without ascorbic acid and incubated at 30 °C with 200 rpm agitation for 48 h on a Microtron shaker (Infors). Because L-DOPA can be easily oxidized to dopachrome and then polymerized nonenzymatically to form melanin 40 , darker cultures were selected for HPLC analysis to determine L-DOPA concentration. Cultures with higher L-DOPA concentrations in the deep-well microplate analysis were selected for shake flask analysis. In the screening process, the culture temperature was set to 30 °C because the cells harbored pSIM6. L-DOPA production in shake flasks. For L-DOPA production, a single colony was inoculated into 5 mL of LB medium in a 20-mL conical tube which was cultured overnight at 37 °C in a rotary shaker at 200 rpm. The overnight seed culture was then inoculated into 50 mL of fermentation medium with a starting OD 600 of 0.1. The fermentation medium (pH 7.0) contains (g/L): peptone 10, yeast extract 5, NaCl 10, glucose 14, ascorbic acid 0. 45  Fed-batch culture for L-DOPA production. The seed culture produced in 5 mL of LB medium was subcultured in 6 × 50 mL LB medium for 10-12 h with shaking at 200 rpm at 37 °C. The seed culture (~300 mL) was inoculated into a 5 L fermenter (Biostat B5, B. Braun, Germany) containing 3 L of fermentation medium with an initial OD 600 of approximately 0.4. The fermentation medium (pH 7.0) contains (g/L): peptone 10, yeast extract 5, NaCl 10, glucose 25, (NH 4 ) 2 SO 4 4 7.5, ascorbic acid 18. The feed was introduced continuously into the fermenter by using the pH-stat feeding strategy. Once the glucose is exhausted, the pH rises rapidly. When the pH was higher than 7.0 by 0.1 U, the feed was automatically added to the fermenter. A total of 680 mL feed solution was added.
Samples were periodically withdrawn, and the following parameters were measured: OD 600 , residual glucose concentration, tyrosine concentration and L-DOPA concentration. Fermentation experiments were carried out in triplicate.
Analytical methods. Growth was monitored by measuring the optical density at 600 nm. Tyrosine and L-DOPA in the supernatants were analyzed using a Shimadzu HPLC system (LC-20 A, Shimadzu, Japan) equipped with an Inertsil ODS-SP column (5 μ m, 4.6 × 150 mm, GL Sciences Inc., Tokyo, Japan). The mobile phase was 0.2% TFA in 40% methanol, with a flow rate of 0.5 mL/min, at 30 °C. A photodiode array detector (SPD-M20A) operating at 280 nm was used, and a standard curve was constructed from serial dilutions of a standard stock solution. Glucose concentration was determined by using glucose oxidase and a glucose assay kit (Shanghai Rongsheng Biotech Corporation, Shanghai, China).