Developing a pyruvate-driven metabolic scenario for growth-coupled microbial production

Abstract

Microbial-based chemical synthesis serves as a promising approach for sustainable production of industrially important products. However, limited production performance caused by metabolic burden or genetic variations poses one of the major challenges in achieving an economically viable biomanufacturing process. To address this issue, one superior strategy is to couple the product synthesis with cellular growth, which renders production obligatory for cell survival. Here we create a pyruvate-driven metabolic scenario in engineered Escherichia coli for growth-coupled bioproduction, with which we demonstrate its application in boosting production of anthranilate and its derivatives. Deletion of a minimal set of endogenous pyruvate-releasing pathways engenders anthranilate synthesis as the salvage route for pyruvate generation to support cell growth, concomitant with simultaneous anthranilate production. Further introduction of native and non-native downstream pathways affords production enhancement of two anthranilate-derived high-value products including L-tryptophan and cis, cis-muconic acid from different carbon sources. The work reported here presents a new growth-coupled strategy with demonstrated feasibility for promoting microbial production.

1. Introduction

As a green alternative to petrol-based chemical synthesis or natural extraction, engineering microbial workhorses to produce natural products or industrially important commodity chemicals from cheap and renewable biomass is more economically and environmentally appealing (Alper and Stephanopoulos, 2009; Bailey, 1991; Ehrenworth and Peralta-Yahya, 2017; Jullesson et al., 2015; Sun et al., 2015). Tailoring microbial hosts with natural or synthetic pathways, together with genetic optimization of endogenous metabolic networks, are conventional strategies that can readily confer microbes with novel or improved production capabilities (Luo et al., 2015; Lynch, 2016; Smanski et al., 2016; Trantas et al., 2015). However, the severe competition or conflict between cell growth and bioproduction may either lead to compromised cell fitness or loss-of-function phenotypes, ascribed to the adverse metabolic burden from intermediate or product toxicities and/or genetic variations (McNerney et al., 2015; Tsoi et al., 2018; Wu et al., 2016).

To address these issues, several advanced strategies have been devised, including the sensor-regulator based dynamic regulation systems (Xu et al., 2014; Yang et al., 2018; Zhang et al., 2012), the synthetic intermediate or product addiction systems (Rugbjerg et al., 2018; Xiao et al., 2016), NADH or ATP driven systems (Lan and Liao, 2012; Shen et al., 2011), and quorum-sensing circuit based metabolic control systems (Gupta et al., 2017; Williams et al., 2015). The underlying principle of different strategies is to create a metabolic scenario whereby bioproduction process is adapted to cellular growth, achieving the balance between biomass formation and product synthesis.

To render a mandatory drive for bioproduction, one potentially superior strategy is to achieve growth-coupled bioproduction (Klamt and Mahadevan, 2015; von Kamp and Klamt, 2017). Generally, deletion of essential genes in microbial host forces rewiring of carbon flux and necessitates alternative salvage pathways (Long et al., 2018; Pontrelli et al., 2018). When the target synthetic pathway serves as the sole compensation route, product synthesis will be obligatorily coupled to cellular growth and thus an integral part of the microbial metabolism. To enable growth-coupled production, one typical approach is to create auxotroph that can be rescued by synthetic pathways. For instance, introducing the citramalate pathway into an isoleucine auxotrophic Escherichia coli enabled growth-based evolution of citramalate synthase and thus increased production of 1-propanol and 1-butanol (Atsumi and Liao, 2008). An alternative coupling approach is to hijack enzymatic reactions in central carbon metabolism by enforcing the synthetic pathways for complementation; in this respect, the tricarboxylic acid (TCA) cycle was most frequently harnessed to drive bioproduction. For example, disruption of 2-ketoglutarate (2-KG) dehydrogenase gene (sucA) forced to the carbon flux to go through the deacetoxycephalosporin C synthase catalyzed reaction for 2-KG to succinate conversion, concomitant with obligatory conversion of penicillin G to 7-aminodeacetoxycephalosporanic acid (Lin et al., 2015). Similarly, coupling proline 4-hydroxylase mediated conversion of 2-KG to succinate with an incomplete TCA cycle in Corynebacterium glutamicum, led to a 60% increase in 4-hydroxy-L-proline production with no obvious impact on the cellular growth rate (Zhang et al., 2019). Interruption of the succinyl-CoA synthetase (sucCD) in C. glutamicum could enforce the lysine pathway as a compensation route for conversion of succinyl-CoA to succinate, via which the lysine yield was increased by 60% (Kind et al., 2013). With growth coupling, product synthesis is indispensably coined into the host metabolism, whose loss of function will always lead to growth retardation or disadvantage. Additionally, it is amenable to boost the production capabilities of a growth-coupled strain through adaptive laboratory evolution and via growth-based selection (Dinh et al., 2018; von Kamp and Klamt, 2017). Despite its exceptional advantage and feasibility, growth-coupled strategy has been less exploited in practical metabolic engineering applications. This is probably due to the paucity of connecting reactions that can couple synthetic pathways with host cell metabolisms, or the existence of multiple known or unknown native metabolic repair pathways that may bypass or compromise the synthetic pathways (Long et al., 2018; Pontrelli et al., 2018).

In this work, we sought to establish a pyruvate-based driving force for microbial production. Pyruvate is a crucial central metabolite linking glycolysis and TCA cycle. Carbon dissimilation via microbial central carbon metabolism always leads to pyruvate formation, whose further oxidation via TCA cycle releases energy and reducing power for microbial growth. Due to its essential role for cell survival in broad genetic contexts, we aim to create a pyruvate-driven metabolic scenario for growth-coupled bioproduction purposes. Rewriting the carbon metabolism and deleting endogenous pyruvate-releasing pathways in E. coli could enforce pyruvate-forming synthetic pathways to regenerate pyruvate for cell growth, achieving growth-coupled production of target chemicals. As proof-of-concept demonstration of our design, we chose anthranilate (AA) as our target as its biosynthesis involves pyruvate release; and more importantly, AA serves as a multi-functional platform chemical for either biosynthesis of complex value-added natural products or degradation to important commodity chemicals (Balderas-Hernandez et al., 2009; Eudes et al., 2013; Ikeda, 2006; Pavlikova et al., 2018; Sun et al., 2013). With anthranilate overproduction, further introduction of downstream biosynthetic pathways diverted anthranilate to two derivatives including L-tryptophan (L-Trp) and cis, cis-muconic acid (MA). Our work presented a platform for the rational rewriting of carbon metabolism to enable growth-coupled chemical production with superior capability.

2. Materials and methods

2.1. Bacterial strains, plasmids and chemicals

Bacterial strains and plasmids used in this study were listed in Table 1. E. coli XL1-Blue (Stratagene) was used for plasmid construction. E. coli ATCC 31884 (Tribe, 1987) and QH4 (Huang et al., 2013) strains were starting strains for development of production strains. E. coli BW25113 (F′) was used for bioconversion experiments. Plasmids pZE12-luc (Lutz and Bujard, 1997), pCS27 (Shen and Liao, 2008), pSA74 (Huo et al., 2011) were used for gene expression, and pCP20 (Cherepanov and Wackernagel, 1995) was used for elimination of kanamycin resistance marker during gene disruption. Phusion DNA polymerase, restriction endonucleases and quick ligase were purchased from NEB (New England Biolabs). Standard chemicals including anthranilate, L-Trp, catechol, and cis, cis-muconic acid were purchased from Sigma-Aldrich unless otherwise specified.

Table 1.

Plasmids and bacterial strains used in this study.

Plasmid	Description	Source
pZE12-luc	PLlacO1, colE ori, Ampr	Lutz and Bujard (1997)
pCS27	PLlacO1, P15A ori, KanR	Shen and Liao (2008)
pSA74	PLlacO1, pSC101 ori, ClR	Huo et al. (2011)
pCP20	Flippase, AmpR, and temperature-sensitive replicon	Cherepanov and Wackernagel (1995)
pZE-PAPC	pZE12-luc containing antABC from Pseudomonas aeruginosa (paantABC) and catA from P. putida KT2440 (ppcatA)	Sun et al. (2013)
pCS-trpEfbrG	pCS27 containing trpEfbrG (encoding feedback-inhibition resistant TrpE with an S40R mutation and the N-terminal domain of TrpD (designated as TrpG)) from E. coli	Sun et al. (2013)
pCS-trpEfbrG-APTA	pCS-trpEfbrG containing aroL, ppsA, tktA, aroGfbr (encoding feedback-inhibition resistant AroG with a D146N mutation) from E. coli	This study
pCS-serA*	pCS27 containing serA* (encoding feedback-inhibition resistant SerA with H344A/N364A mutation)	This study
pCS-serA*-prsAwt	pCS27 containing PLlacO1-serA*-prsA	This study
pCS-serA*-prsA-A95T	pCS27 containing PLlacO1-serA*-prsA-A95T	This study
pCS-serA*-prsA-G226V	pCS27 containing PLlacO1-serA*-prsA-G226V	This study
pCS-trpEfbrG-serA*-prsA-A95T	pCS27 containing two operons: PLlacO1-trpEfbrG and PLlacO1-serA*-prsA-A95T	This study
pSA-trpDBCA	pSA74 containing trpDBCA from E. coli	Sun et al. (2014b)
Strain	Description	Source
E. coli XL1-Blue	recA1 endA1 gyrA96 thi-1 hsdR17 supE44 relA1 lac [F′ proAB lacIqZΔM15Tn10 (TeiR)]	Stratagene
E. coli BW25113 (F′)	rrnBT14 ΔlacZWJ16 hsdR514 ΔaraBADAH33 ΔrhaBADLD78 F′ [traD36 proAB laclqZΔM15 Tn10(Tetr)]	Atsumi et al. (2008)
E. coli ATCC 31884	A phenylalanine over-producing derivative of E. coli K-12 with aroH367, tyrR366, tna-2, lacY5, aroF394fbr, malT384, pheA101fbr, pheO352, aroG397fbr	ATCC
E. coli QH4	E. coli ATCC31884 ΔpheLA ΔtyrA	Huang et al. (2013)
E. coli JW1	E. coli QH4 ΔpykA ΔpykF	This study
E. coli JW2	E. coli QH4 ΔpykA ΔpykF ΔgldA	This study
E. coli JW3	E. coli QH4 ΔpykA ΔpykF ΔgldA ΔmaeB	This study
E. coli JW4	E. coli QH4 ΔtrpE	This study
E. coli JW5	E. coli QH4 ΔpykA ΔpykF ΔgldA ΔtrpE	This study
E. coli JW6	E. coli QH4 ΔpykA ΔpykF ΔgldA ΔmaeB ΔtrpE	This study
E. coli JW7	E. coli QH4 ΔpykA ΔpykF ΔgldA ΔtnaA ΔtrpR with F′ plasmid [traD36 proAB laclqZΔM15] transduced from E. coli XL-1 Blue	This study
E. coli JW8	E. coli QH4 with ΔpykA ΔpykF ΔgldA ΔmaeB ΔtnaA ΔtrpR with F′ plasmid [traD36 proAB ladqZΔM15] transduced from E. coli XL-1 Blue	This study
E. coli AA1	E. coli QH4 with pCS-trpEfbrG	This study
E. coli AA2	E. coli JW1 with pCS-trpEfbrG	This study
E. coli AA3	E. coli JW2 with pCS-trpEfbrG	This study
E. coli AA4	E. coli JW3 with pCS-trpEfbrG	This study
E. coli TP1	E. coli QH4 with pCS-trpEfbrG-serA*-prsA-A95T and pSA-trpDBCA	This study
E. coli TP2	E. coli JW7 with pCS-trpEfbrG-serA*-prsA-A95T and pSA-trpDBCA	This study
E. coli TP3	E. coli JW8 with pCS-trpEfbrG-serA*-prsA-A95T and pSA-trpDBCA	This study
E. coli MA1	E. coli QH4 with pCS-trpEfbrG and pZE-PAPC	This study
E. coli MA2	E. coli JW2 with pCS-trpEfbrG and pZE-PAPC	This study
E. coli MA3	E. coli JW3 with pCS-trpEfbrG and pZE-PAPC	This study
E. coli MA4	E. coli JW2 with pCS-trpEfbrG-APTA and pZE-PAPC	This study
E. coli MA5	E. coli JW3 with pCS-trpEfbrG-APTA and pZE-PAPC	This study

2.2. Plasmid construction

All manipulations of DNA were conducted referring the standard molecular cloning protocols (Sambrook et al., 1989). The feedback-inhibition resistant mutant of serA from E. coli BW25113 (F′) with H344A/N364A mutation (serA^*) was created by overlapping PCR using primers containing mutations, and was cloned into the medium-copy-number plasmid pCS27 between Acc65I and SalI sites, yielding pCS-serA^*. The wildtype prsA from E. coli BW25113 (F′) and its mutants obtained via overlapping PCR, were digested with SalI and BamHI, and constructed into pCS-serA^* to generate pCS27-serA^*-prsAwt, pCS-serA^*-prsA-A95T and pCS-serA^*-prsA-G226V. The pCS-trpE^fbrG was obtained from our previous work (Sun et al., 2013). The expression cassette of P_LlacO1-trpE^fbrG was then digested and cloned into pCS-serA^*-prsA-A95T between SpeI and SacI sites, yielding pCS-trpE^fbrG-serA^*-prsA-A95T. The trpDBCA operon was placed on the low-copy-number plasmid pSA74, and the resulting plasmid pSA-trpDBCA was obtained from our previous work (Sun et al., 2014b). pCS-trpE^fbrG-APTA was obtained by amplifying and inserting the APTA module containing aroL, ppsA, tktA, aroG^fbr (encoding feedback-inhibition resistant AroG with a D146N mutation) from previously constructed pCS-APTA (Lin et al., 2013) into pCS-trpE^fbrG between SpeI and SacI sites. pZE-PAPC was obtained from previous work (Sun et al., 2013).

2.3. Gene disruption

The disruption of chromosomal genes in E. coli was conducted by P1 phage-based transduction (Thomason et al., 2007). Specifically, P1 lysates for target knockouts were prepared from the Keio collection strains (CGSC) (Baba et al., 2006). The transduced cells were screened on LB-agar plates with appropriate antibiotic and 100 mM sodium citrate. Removal of antibiotic resistance was conducted by electroporation of pCP20 into target strains (Datsenko and Wanner, 2000). Knockout strains were confirmed by colony PCR.

2.4. Culture media and conditions

Luria-Bertani (LB) medium containing 10 g/L tryptone, 5 g/L yeast extract and 10 g/L sodium chloride was used for cell propagation. The M9 minimal medium (per liter) containing 6 g Na₂HPO₄, 0.5 g NaCl, 3 g KH₂PO₄, 1 g NH₄Cl, 246.5 mg MgSO₄·7H₂O and 14.7 mg CaCl₂·2H₂O was used as the basis medium for growth-related tests and shake flask cultivation. Antibiotics were added when necessary at the concentration of 100, 50 and 34 μg/mL for ampicillin (Amp), kanamycin (Kan) and chloramphenicol (Cl), respectively. For growth-related tests, M9 medium was supplemented with 20 g/L glycerol, 100 mg/L L-Phe, L-Tyr and L-Trp of each, and 0.5 or 5 g/L yeast extract when needed. For shake flask cultivation, M9 minimal medium was supplemented with 20 g/L glycerol or xylose and 5 g/L yeast extract (M9Y). Overnight cell culture was inoculated with 2% into 20 mL M9Y medium in 125 mL flasks and incubated for 3h at 37 °C and 280 rpm. Isopropyl β-D-1-thiogalactopyranoside (IPTG) was added at a final concentration of 0.5 mM and cell cultures were switched to 30 °C for induction. For anthranilate to L-Trp bioconversion experiments, anthranilate was added along with IPTG to cell cultures at a final concentration of 2 g/L. Cell cultures were sampled every 12 h. The optical density at 600 nm (OD₆₀₀) of cell culture was measured in 96-well plates in Biotek Synergy HT Microplate Reader (Biotek). Products in the culture were quantified by high-performance liquid chromatography (HPLC) analysis after centrifugation at 13,523×g for 10 min and filtration with 0.45 μm filter.

2.5. Growth test conditions and measurement

To investigate the pyruvate-dependent growth, seed cultures were inoculated in 3 mL LB culture in test tube at 37 °C for overnight. 100 μL of seed cultures were collected by centrifugation at 3381×g for 2 min, and the cell pellets were resuspended and inoculated into M9 minimal medium containing 20 g/L glycerol, 100 mg/L L-Phe, L-Tyr, and L-Trp of each. To test the effect of pyruvate supplementation, pyruvate was added with final concentrations ranging from 0 to 2 g/L. Cell cultures were sampled every 6 or 12 h for growth measurement (OD₆₀₀) in 96-well plates. The specific growth rate, μ, was calculated from the OD₆₀₀ growth curve at exponential phase, as described previously (μ = ΔlnOD₆₀₀/Δt, where t is time) (Berney et al., 2006; Lindqvist and Barmark, 2014). Similar approach was used for investigating the trpE-dependent growth of the engineered cells, with additional supplementation of 0.5 g/L yeast extract to support the initial cell growth.

2.6. HPLC analysis

Both standard chemicals and products in cell cultures including anthranilate, L-Trp, catechol, and cis, cis-muconic acid were quantified by a reverse phase HPLC system 1260 infinity II (Agilent technologies) equipped with a ZORBAX SB-C18 column and a 1260 infinity II Diode Array Detector WR. The mobile phase was consisted of 0.1% trifluoroacetic acid (solvent A) and methanol (solvent B). The flow rate was set at 1 mL/min and temperature at 28 °C. The analyzing method was set as: 5% solvent B for 2 min, from 5 to 50% solvent B for 6 min, from 50 to 80% solvent B at 2 min, from 80 to 5% solvent B for 5 min, and 5% solvent B for additional 2 min. The concentrations of anthranilate, L-Trp, catechol, and cis, cis-muconic acid were determined by the peak areas at absorbances of 320 nm, 276 nm, 276 nm and 260 nm, respectively. Glycerol and xylose concentrations were quantified by the Dionex HPLC system with Coregel-64H column (Transgenomic). 4 mN H₂SO₄ was used as the mobile phase and the flow rate was set at 0.4 mL/min. The oven temperature was set at 45 °C.

3. Results

3.1. Metabolic design for pyruvate-driven growth-coupled bioproduction

To enable pyruvate-driven growth-coupled bioproduction, the centerpiece of our metabolic design is to rewire carbon metabolism and create a metabolic scenario of pyruvate-based coupling of cell growth and target synthetic pathways. Pyruvate is a key metabolite from central carbon metabolism, and can support cell growth alone via TCA cycle mediated oxidation (Antonovsky et al., 2016). Besides central carbon metabolism, pyruvate release is also a commonplace in synthetic pathways, especially those involving aromatic biosynthesis or catabolism (Johnson and Beckham, 2015; Noda et al., 2016, 2017). We reason that it is possible to tightly couple cell growth with target synthetic pathways by: (1) deleting endogenous pyruvate-releasing pathways to create a pyruvate auxotrophic E. coli, (2) rendering the target biosynthetic pathway as the primary or even the sole pathway to regenerate pyruvate. We began our design by focusing on rewiring glycerol metabolism for growth-coupled AA production in E. coli (Fig. 1). Glycerol is chosen as the carbon source due to its increased availability, low price, high degree of reduction, and most importantly, a higher maximal theoretical yield for aromatic compounds (Chen and Liu, 2016; Durnin et al., 2009; Trondle et al., 2018). AA can be natively produced by E. coli via the shikimate pathway with 1 mol of pyruvate released per mole of AA produced, making it an ideal pyruvate-driven target pathway.

Fig. 1.

Engineering pyruvate-driven growth-coupled bioproduction in E. coli. The scheme represents key metabolic pathways, metabolites and genes involved in carbon metabolism, and biosynthetic pathways for anthranilate, L-tryptophan and cis, cis-muconic acid. Deleted genes are indicated in red and over-expressed genes are indicated in blue. Metabolite abbreviations: G6P, glucose-6-phosphate; F6P, fructose-6-phosphate; F1,6BP, fructose 1,6-bisphosphate; 6PGNL, 6-phosphogluconolactone; 6PGNT, 6-phosphogluconate; RU5P, ribulose-5-phosphate; R5P, ribose-5-phosphate; X5P, xylulose-5-phosphate; S7P, seudoheptulose-7-phosphate; E4P, erythrose-4-phosphate; G3P, glycerol 3-phosphate; DHA, dihydroxyacetone; DHAP, dihydroxyacetone phosphate; GA3P, glyceraldehyde- 3-phosphate; 3 PG, 3-phosphoglycerate; PEP, phosphoenolpyruvate; PYR, pyruvate; OAA, oxaloacetate; MAL, malate; TCA cycle, tricarboxylic acid cycle; DAHP, 3-deoxy-D-arabino-heptulosonate-7-phosphate; PRPP, 5-phospho-α-D-ribose 1-diphosphate; L-Ser, L-serine; L-Gln, L-glutamine; L-Glu, L-glutamate; L-Phe, L-phenylalanine; L-Tyr, L-tyrosine; L-Trp, L-tryptophan; MA, cis, cis-muconic acid. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

To render AA pathway as the main source for pyruvate generation, native pyruvate-releasing pathways in glycerol dissimilation and central carbon metabolism need to be blocked (Fig. 1). Glycerol dissimilation into the central carbon metabolism in E. coli involves two pathways, the major GlpK-GlpD/GlpABC mediated respiratory pathway and the minor GldA-DhaKLM mediated fermentative pathway (Durnin et al., 2009). To preclude pyruvate leaking via phosphoenolpyruvate (PEP)-dependent dihydroxyacetone kinase (DhaKLM), gldA, encoding glycerol dehydrogenase that catalyzes the first committed step of the fermentative pathway, will be deleted. In central carbon metabolism, two major pyruvate-releasing reactions are involved: the pyruvate kinases (PykA and PykF) mediated conversion of PEP to pyruvate, and malate dehydrogenase (MaeB) mediated decarboxylation of malate to pyruvate. In AA biosynthesis pathway, anthranilate synthase (TrpED) is responsible for conversion of chorismite to AA, concomitant with pyruvate release. Deletion of all corresponding genes would possibly make E. coli auxotrophic for pyruvate, or retaining trpED would make E. coli growth dependent on AA pathway. Moreover, deletion of gldA and pykAF would reserve PEP for shikimate pathway.

3.2. Engineering pyruvate-driven growth-coupled production of anthranilate

To create and validate an AA pathway dependent E. coli, we started knocking out endogenous pyruvate-releasing genes within strain E. coli QH4 (Fig. 2A), a derivative of L-phenylalanine (L-Phe) overproducer E. coli ATCC31884 with disrupted L-Phe and L-tyrosine (L-Tyr) branches (Huang et al., 2013). We first deleted the pyruvate kinase genes pykA and pykF in E. coli QH4, yielding strain JW1 (Table 1). Pyruvate kinases catalyze the conversion of PEP to pyruvate, and knocking out of pykA and pykF alleviates pyruvate formation but does not impair cell viability (Emmerling et al., 2002; Xu et al., 2012; Yang et al., 2018). Subsequent stepwise deletion of gldA and maeB yielded two more strains, JW2 and JW3. When cultivated in minimal medium containing glycerol (20 g/L) and limited yeast extract (0.5 g/L), JW2 and JW3 showed 2.4- and 3.2-fold decrease in specific growth rate (μ = 0.08 and 0.06 h⁻¹, respectively) compared with their parental strain QH4 (μ = 0.19 h⁻¹) (Fig. 2B). When further knocking out anthranilate synthase gene trpE in JW2 or JW3, cell growth of the resultant strain JW5 or JW6 was arrested within 36 h. Noteworthy, the cell growth of maeB deleting strains, JW3 and JW6 (μ = 0.04 h⁻¹), were more severely impaired. In contrast, QH4 with only trpE knockout (JW4) exhibited no growth defect (μ = 0.19 h⁻¹) (Fig. 2B), and could grow even with no yeast extract (Fig. 2C). Taken together, these results established that disruption of the major pyruvate-releasing genes including pykA, pykF, gldA and maeB significantly compromised cell growth, possibly caused by limited pyruvate supply. Further knockout of trpE almost deprived cell of growth potential, as JW5 and JW6 could barely grow in glycerol minimal medium with limited yeast extract, and did not show detectable growth when further removing yeast extract from the medium (Fig. 2D and E). This revealed that the cell growth of JW2 or JW3 was in part reliant on pyruvate released from AA pathway in glycerol minimal medium with limited yeast extract. When cultivated in glycerol minimal medium with surplus of yeast extract (5 g/L), all mutant strains showed improved growth while JW5 and JW6 still showed compromised growth rate and potential (Fig. S1).

Fig. 2.

Engineering pyruvate-driven E. coli strains for growth-coupled AA production. (A) Deletion of endogenous pyruvate-releasing genes to enable pyruvate formation mainly via AA biosynthetic pathway. (B) Investigating the growth potential of E. coli mutants derived from QH4 in glycerol minimal medium with 0.5 g/L yeast extract. Growth curve of JW4 (C), JW5 (D) and JW6 (E) in glycerol minimal medium (no yeast extract) with gradient concentrations of pyruvate, from 0 (circles) to 0.5 (triangles), 1 (squares) and 2 g/L (diamonds). (F) Residual carbon source in the medium after 30 h cultivation of JW4, JW5 and JW6 with glycerol minimal medium (no yeast extract) containing 2 g/L pyruvate. (G) AA production profiles and growth curves of strains AA1, AA2, AA3 and AA4 in 48 h. Error bars represent s.d. (n = 3).

To investigate if exogenous pyruvate could restore cell growth of pyruvate auxotrophic E. coli, JW5 and JW6 were cultivated in glycerol minimal medium (containing100 mg/L L-Phe, L-Tyr and L-Trp but with no yeast extract) supplemented with different concentrations of pyruvate (0–2 g/L). As expected, the control strain JW4 showed no growth difference in different pyruvate concentrations, while the growth of JW5 and JW6 were restored by pyruvate and their growth potentials were dependent on the amount of pyruvate supplied (Fig. 2C, D, E). This demonstrated that knocking out of pykA, pykF, gldA with or without maeB rendered E. coli growth pyruvate-dependent. This also implied that these genetic manipulations enabled high correlation of pyruvate supply and cell viability that the high-performers with pyruvate-forming synthetic pathways will be potentially more advantageous. Of note, exogenous pyruvate addition facilitated glycerol utilization in JW5 and JW6 but not in JW4 (Fig. 2F), as reflected by the difference in their growth potentials with 2 g/L pyruvate (Fig. 2C, D, E).

Given the maintained but reduced cell growth attributed to single chromosomal copy of trpE, we reason that plasmid-based over-expression of anthranilate synthase in our engineered E. coli strains will promote not only cell growth but also AA production. To verify that, the feedback-inhibition resistant anthranilate synthase (TrpE^fbrG), composed of TrpE with an S40R mutation and the N-terminal domain of TrpD (designated TrpG) (Sun et al., 2013), was expressed under control of P_LlacO1 promoter on the medium-copy-number plasmid. The resultant plasmid pCS-trpE^fbrG was transferred into E. coli QH4, JW1, JW2 and JW3, resulting in AA producer strains AA1 to AA4 (Table 1). All transformed strains were subjected to shake flask cultivation using M9 minimal medium containing 20 g/L glycerol with 5 g/L yeast extract. After 48 h of aerobic cultivation, all recombinant strains (AA2 to AA4) showed slightly decreased growth but increased AA production than the control strain AA1 (Fig. 2G). Especially, AA4 with the most knockouts afforded the highest titer of 1.78 g/L AA, which is 2.1-fold to that produced by AA1 (0.85 g/L). In particular, AA4 achieved an AA yield of 0.143 g/g glycerol, which is 1.86-fold to that of strain AA3 (0.077 g/g glycerol) and 2.65-fold to that of control strain AA1 (0.054 g/g glycerol) (Table 2). This demonstrated that, with strong growth coupling, the engineered pyruvate-driven mechanism could significantly improve AA production, even without enhancing the upstream shikimate pathway flux.

Table 2.

Summary of production performance by the engineered pyruvate-driven E. coli strains.

Host strain	Glycerol						Xylose
	AA		L-Trp		MA		AA		L-Trp		MA
	Titer a	Yield b	Titer	Yield	Titer	Yield	Titer	Yield	Titer	Yield	Titer	Yield
E. coli QH4	0.84	0.054	0.73	0.037	0.89	0.071	0.56	0.046	0.82	0.05	0.68	0.059
E. coli JW2	1.00	0.077	-c	-	1.82	0.122	1.35	0.131	-	-	0.46	0.028
E. coli JW3	1.78	0.143	-	-	1.60	0.125	1.90	0.178	-	-	0.47	0.029
E. coli JW7	-	-	1.73	0.132	-	-	-	-	1.31	0.09	-	-
E. coli JW8	-	-	1.32	0.168	-	-	-	-	0.97	0.08	-	-

^aPeak titers were used for comparison with titer unit, g/L.

^byields corresponding to peak titers were used for comparison with yield unit, g/g.

^c, not tested.

3.3. Enhancing L-Trp production via the pyruvate-driven system

With establishment of the pyruvate-driven system for AA production, we next evaluated whether the engineered AA platform would translate into AA-derived bioproduction pathways. Natively, AA serves as the direct precursor for L-Trp, an essential aromatic-group amino acid for humans and a fundamental precursor to a variety of high-value compounds like neurotransmitters 5-hydroxytryptophan (5-HTP) and serotonin, and drugs like indolylglucosinolate and monoterpene indole alkaloids (MIAs) (Ehrenworth and Peralta-Yahya, 2017; Lin et al., 2014a; Wang et al., 2016). L-Trp biosynthesis is subjected to multi-level regulation, including TrpR-mediated transcriptional repression and TnaA mediated L-Trp degradation (Chen et al., 2018; Panichkin et al., 2016). Therefore, to alleviate L-Trp degradation and also to deregulate TrpR mediated repression, both tnaA and trpR were knocked out in JW2 and JW3, with subsequent transduction of the F′ plasmid (harboring latcl^q) from E. coli XL-Blue to avoid leaky expression and to facilitate transformation. The resultant strains were designated JW7 and JW8 (Table 1).

To fully exploit the AA platform, we set to address rate-limiting steps during AA conversion to L-Trp. AA undergoes five enzymatic steps to L-Trp, during which supply of two co-substrates is of utmost importance: L-serine (L-Ser) and 5-phospho-α-D-ribose 1-diphosphate (PRPP) (Chen and Zeng, 2017). To assess the contribution of L-Ser supply, 2 g/L AA was fed to E. coli BW25113 (F′) expressing the feedback-inhibition resistant D-3-phosphoglycerate dehydrogenase SerA^* (SerA H344A/N364A) and trpDBCA operon (pSA-trpDBCA). SerA^* overexpression led to production of 1.27 g/L L-Trp in 48 h, a 6.2-fold increase over the control strain with no SerA^* over-expression (Fig. 3a, Fig. S2A). Considering the residual of 0.93 g/L AA, SerA^* overexpression resulted in an AA-to-L-Trp conversion ratio to 42.6% of the theoretical maximum (Fig. S2B). To further avoid AA build up, we sought to increase the PRPP pool by over-expressing the ribose-phosphate diphosphokinase (PrsA). However, co-expression of wildtype prsA (prsAwt) with serA^* and trpDBCA decreased L-Trp to 0.58 g/L, suggesting an adverse effect by PRPP surplus (Fig. 3A). We next turned to two PrsA mutants, A95T and G226V, which were chosen because of their evolutionally beneficial role in producing a hemi-autotrophic E. coli that can grow on CO₂ and pyruvate (Antonovsky et al., 2016; Herz et al., 2017). Surprisingly, although with reduced ribose-phosphate diphosphokinase activities (Antonovsky et al., 2016), L-Trp production was increased to 1.36 g/L by PrsA G26V (45.7% of theoretical maximum), while increased to 2.02 g/L by PrsA A95T (68.0% of theoretical maximum) (Fig. 3A, Fig. S2A). Meanwhile, with PrsA A95T expressing, the residual AA was significantly decreased to 0.45 g/L at 48 h. These results suggested that, by modulating L-Ser and PRPP pool, AA can be efficiently converted to L-Trp.

Fig. 3.

Enhancing L-Trp production from glycerol. (A) Improving bioconversion of anthranilate (orange) to L-Trp (blue) by increasing L-Ser supply (over-expressing the feedback-inhibition resistant mutant SerA^*) and PRPP supply (over-expressing PrsA or its variants) in E. coli BW25113 (F′). NC, negative control with neither SerA^* nor PrsA over-expression. (B) De novo production of L-Trp from glycerol by TP1 (gold), TP2 (green) and TP3 (purple). Cell density and L-Trp titers were measured every 12 h for 60 h. Error bars represent s.d. (n = 3). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

With engineered hosts and optimized L-Trp pathway, we then tested the ability of pyruvate-driven system in boosting L-Trp production from glycerol. The complete L-Trp pathway was reassembled on two plasmids, pCS-trpE^fbrG-serA^*-prsA-A95T and pSA-trpDBCA. They were respectively transferred into QH4, JW7 and JW8, with the resultant strains designated TP1 to TP3 (Table 1). After 60 h of aerobic cultivation in glycerol minimal medium, all recombinant strains reached peak L-Trp titers in 60 h with no detectable accumulation of AA. The parental strain TP1 could produce 0.73 g/L L-Trp, while the engineered TP2 produced the highest titer of L-Trp of 1.73 g/L, a 2.37-fold to that produced by TP1 (Fig. 3B, Fig. S3). The L-Trp yield by TP1 reached 0.037 g/g glycerol, whereas TP2 achieved a 3.57-fold increased L-Trp yield (0.132 g/g) (Table 2). Interestingly, although TP3 produced less L-Trp than TP2, it also consumed less glycerol, resulting in the highest L-Trp yield to 0.168 g/g glycerol. This demonstrated that our pyruvate-driven system could significantly improve both L-Trp titers and yields.

3.4. Enhancing cis, cis-muconic acid production via the pyruvate-driven system

To test the applicability of our AA platform for non-native bioproduction pathways, we implemented it into an AA derived non-native product MA. MA is microbially produced from renewable biomass as the synthetic precursor for adipic acid and terephthalic acid, which have been widely used as important platform chemicals for synthesis of nylon-6,6, polyurethane and polyethylene terephthalate (PET) (Barton et al., 2018; Kruyer and Peralta-Yahya, 2017; Vardon et al., 2016). Our lab previously established three independent MA synthetic pathways starting from chorismate: salicylic acid (SA) pathway, 2,3-dihydrox- ybenzoic acid (DHBA) pathway and AA pathway (Lin et al., 2014b; Sun et al., 2013, 2014a). The AA-based MA pathway requires three enzymatic steps downstream of chorismate: conversion of chorismate to AA by TrpEG, conversion of AA to catechol by anthranilate 1,2-dioxygenase (ADO) and benzene ring cleavage of catechol to MA via catechol 1,2-dioxygenase (CDO). Although the AA-based MA pathway is shorter and less toxic, its production potential has been significantly limited by the AA supply (Lin et al., 2014b; Sun et al., 2013).

To harness our AA platform for MA production, the synthetic pathway comprising ADO (antABC from Pseudomonas aeruginosa) and CDO (catA from P. putida KT2440), was introduced into AA overproducer strains on a high-copy-number plasmid (pZE-PAPC). Co-transformation of pZE-PAPC and pCS-trpE^fbrG into QH4, JW2 and JW3 yielded MA producer strains, MA1 to MA3. After cultivation in glycerol minimal medium for 72 h, MA2 and MA3 respectively afforded production of 1.82 and 1.60 g/L MA, along with 0.14 and 0.43 g/L residual catechol, while QH4 resulted in only 0.89 g/L MA with almost no accumulation of catechol (Fig. 4, Fig. S4). Interestingly, AA is barely accumulated in all production strains, suggesting that the downstream CDO is rate-limiting in MA production. To further increase carbon flux to shikimate pathway, we incorporated the APTA module (harboring aroL, ppsA, tktA, and aroG^fbr) on the pCS-trpE^fbrG as an independent operon (pCS-trpE^fbrG-APTA). However, when transformed with pZE-PAPC and pCS-trpE^fbrG-APTA into JW2 and JW3, the resultant strains MA4 and MA5 showed substantially decreased MA production (0.86 and 0.71 g/L, respectively) (Fig. 4, Fig. S5). This may be explained by increased metabolic burden, possibly caused by excessive protein expression. Taking into account that the engineered pyruvate-driven strains are efficient in driving carbon flux to AA, minimal expression of downstream genes would allow optimal production. Compared with the previous report on AA-based MA production (Sun et al., 2013), our engineered strain increased MA titer by 4.7-fold with less over-expressed genes, demonstrating the robustness of our pyruvate-driven system for microbial production of AA-derived non-native products.

Fig. 4.

MA production from glycerol using the engineered platform strains. Comparison of MA production by MAI, MA2, MA3, MA4 and MA5 in glycerol minimal medium. Product profiles include: MA (dark grey) and catechol (light grey). All data points are reported as mean from three independent experiments and the peak titers were applied for comparison. Error bars represent s.d. (n = 3).

3.5. Extending the pyruvate-driven system to xylose metabolism

After demonstrating the application of the pyruvate-driven system in glycerol-based medium, we hypothesize that it can be potentially applied to utilizing other commonly used carbon sources like lignocellulosic sugar glucose and xylose. To extend the flexibility of carbon sources, the key point is to eliminate pyruvate release during carbon assimilation. Since glucose uptake primarily relies on PEP-dependent phosphotransferase system (PTS), to simplify the genetic manipulations, we therefore switch to xylose as proof-of-concept demonstration. Xylose is the second most abundant sugar in lignocellulosic biomass and has been widely used as a potential feedstock for microbial production of chemicals in recent decades (Li et al., 2018; Tai et al., 2016; Wang et al., 2017). In E. coli, xylose enters the pentose phosphate pathway via an ATP-dependent isomerase pathway with the combined action of xylose isomerase (XylA) and xylulokinase (XylB), leaving no additional pyruvate releasing reactions (Fig. 1).

We first tested the applicability of our final pyruvate-driven strains for AA biosynthesis using xylose. Shake flask experiments were performed with strains AA1, AA3 and AA4 in minimal medium containing 20 g/L xylose. After 72 h of cultivation, the control strain AA1 only produced 0.56 g/L AA, while the engineered strain AA3 and AA4 produced 1.38 g/L and 1.90 g/L AA, respectively (Fig. 5, Fig. S6). Particularly, AA4 achieved the highest yield of 0.178 g/g xylose, which is 1.36-fold to that of AA3 (0.131 g/g) and 3.87-fold to that of AA1 (0.046 g/g) (Table 2). The production trend was in parallel with that observed with glycerol, indicating the applicability of the pyruvate- driven system for xylose-based chemical production.

Fig. 5.

Extending pyruvate-driven system for xylose-based bioproduction. Production performance of AA producers (AA1, AA3 and AA4), L-Trp producers (TP1, TP2 and TP3) and MA producers (MA1, MA2 and MA3) in xylose minimal medium. Product profiles include: AA (orange), L-Trp (blue), MA (dark grey) and catechol (light grey). All data points are reported as mean from three independent experiments and the peak titers were applied for comparison. Error bars represent s.d. (n = 3). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Subsequently, the pyruvate-driven system was deployed for L-Trp and MA biosynthesis. When cultivating TP1, TP2 and TP3 in xylose minimal medium, TP2 afforded the highest L-Trp production with a peak titer of 1.31 g/L at 36 h, which is 1.35-fold and 1.6-fold increase to that produced by TP3 (0.97 g/L) and TP1 (0.82 g/L), respectively (Fig. 5 and Table 2). The residual of AA (> 0.5 g/L) in TP2 and TP3 cultures indicated that the downstream AA-to-L-Trp conversion in xylose-based condition is less efficient than that in glycerol-based condition (Fig. S7). Considering the difference of substrate reluctance and metabolic infrastructure, the choice of carbon source might have different effects on downstream synthetic pathways. This effect is more notable in MA production, which did not show an increase trend in xylose medium as observed in glycerol medium. With the engineered pyruvate-driven system, MA2 and MA3 only afforded 0.46 and 0.47 g/L MA, which are less than that produced by the control strain MA1 (0.68 g/L) (Fig. 5 and Table 2). The substantial accumulation of intermediates including AA (0.37–0.97 g/L) and catechol (0.43–0.80 g/L) in MA2 and MA3 cultures indicated that the downstream MA pathway is inefficient in conversion of AA to MA in xylose-based condition (Fig. S8). Although the MA concentrations from xylose were lower than from glycerol, the high accumulation of metabolites including AA and catechol still demonstrated the functionality of the pyruvate-driven system in driving carbon flux from xylose to MA pathway.

4. Discussion

Improving the microbial production performance is a long-term pursue in metabolic engineering fields. Efforts to repurpose microbes for chemical production have been largely focused on maximizing heterologous synthetic pathways and minimizing endogenous competing pathways (Lee et al., 2012; Zhang et al., 2011). The heavy metabolic burden, extensive genetic manipulations, metabolite toxicities or sophisticated metabolic regulations might compromise cell viability and production performance. To facilitate bioproduction, engineering inherent driving forces for metabolic engineering applications have been devised and employed mainly based upon cofactor-driven systems, like NADH-driven system in E. coli and ATP-driven system in cyanobacteria (Lan and Liao, 2012; Shen et al., 2011). However, these cofactor-driven systems are conditionally functioning, like NADH- driven system only working in microaerobic or anaerobic conditions and ATP-driven system limited to photosynthetic organisms (Wang et al., 2013). Here we report the development of a pyruvate-driven system to couple product synthesis with cellular growth, which enables the bioproduction coined as an integral part of host metabolism. Via growth coupling, product synthesis will be positively correlated with cell fitness, promoting high-performance variants to dominate the population and provoking adaptive evolution of cell robustness to lessen metabolic burden. Noteworthy, due to the crucial role of pyruvate in microbial carbon metabolism, the pyruvate-driven system can be readily adapted to any production microorganisms with no constraint of cultivation conditions. Yet, to fully deploy the pyruvate-driven system for microbial production, two key prerequisites need to be fulfilled, removing major endogenous pyruvate-releasing reactions in carbon metabolism and anchoring pyruvate-forming synthetic pathways.

In E. coli, although there exist more than thirty genes that are involved in pyruvate formation, knocking out several major pyruvate-releasing genes can render E. coli pyruvate auxotrophic or dependent on specific pyruvate-forming pathways. When using glycerol as the carbon source, knocking out of a minimal set of only four genes including pykA, pykF, gldA and maeB rendered cellular growth dependent on anthrani-late biosynthetic pathway, since further anthranilate synthase gene (trpE) knockout would endow E. coli pyruvate auxotrophic. Over-expression of feedback-inhibition resistant anthranilate synthase (TrpE^fbrG) in the engineered platform strain JW3 (QH4 ΔgldA ΔpykA ΔpykF ΔmaeB) showed slightly decreased growth but 2.1-fold increase of AA production compared to the parental strain QH4, indicating the coupled effects of growth and bioproduction. The pyruvate-driven system can be theoretically applied to all commonly used carbon sources that are catabolized to pyruvate including lignocellulosic biomass derived glucose and xylose. To extend the feedstock flexibility of the pyruvate-driven system, the key principle of the metabolic design is to eliminate pyruvate formation during sugar catabolism. For glucose, replacing the PEP-dependent PTS transport system with galactose permease (GalP) and ATP-dependent glucokinase (Glk) would preclude pyruvate formation as has been demonstrated previously (Noda et al., 2016, 2017); while for xylose, the native ATP-dependent pathway would result in no additional pyruvate release. Thus, as no additional genetic modifications needed, we directly shifted the carbon source to xylose with our engineered platform strain. The reproducible trend of titer and yield improvement from xylose demonstrated the applicability of pyruvate-driven system in different carbon sources.

The pyruvate-forming pathways are quite common in nature. Among those, the shikimate pathway is of remarkable industrial importance because of its involvement in production of a variety of high- value aromatic compounds. The branches of shikimate pathway that start from chorismate or its derivatives involve pyruvate release; for instances, one pyruvate is released during conversion of chorismate to AA or 4-hydroxybenzoate, isochorismate to 2,3-dihydroxy-2,3-dihydrobenzoate or salicylate, and 4-amino-4-deoxychorismate to 4-aminobenzoate (Noda et al., 2016). These products could serve as precursors for an expanded spectrum of new products of interest. In the present study, we focused on establishing and demonstrating the pyruvate-driven system for AA and its derivative overproduction. We demonstrated that, with no bottlenecks in AA-derived synthetic pathways, the benefit of pyruvate-driven system was extended to AA-derived products including L-Trp and MA.

One key strength of the pyruvate-driven system is its capacity to improve the practical yield, a feature that can be seen in growth-coupled chemical production. With the advanced pyruvate-driven system, the carbon utilization, as well as product synthesis, were internally coupled to cellular growth. As we observed in the current study, installing the pyruvate-driven system increased the yield of AA, L-Trp and MA from different carbon source. From a practical perspective, the pyruvate-driven system could potentially afford high-yield and feed-stock-efficient production by controlling carbon expenditure and flux redirection.

In summary, our research underlines the great potential of setting up pyruvate-driven growth-coupled metabolic design for microbial chemical production. With proof-of-concept demonstration with AA and its derivative pathways using different carbon sources, we conclude that the pyruvate-driven system could be readily achievable in different metabolic or genetic contexts. Current limitation of the pyruvate-driven system may be the low carbon flux toward the synthetic pathways, which might restrict pyruvate supply for cellular growth or provoke the emergence of alternative metabolic routes for complementation (Pontrelli et al., 2018). By increasing the carbon flux through synthetic pathways and adaptive evolution of rationally engineered growth- coupled strains, we expect further increase of cell fitness and improvement of production performance in the future.