Introduction

In Escherichia coli, phosphoenolpyruvate (PEP), a C3 metabolite of the glycolysis, is carboxylated into oxaloacetate (OAA), a C4 metabolite of the TCA cycle, via phosphoenolpyruvate carboxylase (PPC, EC 4.1.1.31) under glycolytic conditions [4, 7]. The OAA is decarboxylated into PEP under gluconeogenic conditions via phosphoenolpyruvate carboxykinase (PCK, EC 4.1.1.49) in vivo [4, 16, 36]. However, examination of the enzymatic characteristics in vitro revealed that PCK could also mediate the reversible reaction under conditions of high CO2 concentrations, and that the carboxylation of PEP by PCK could result in the production of ATP, as opposed to PPC, which produces inorganic phosphate [4, 5, 21]. The overexpression of PCK enabled E. coli to harbor higher ATP concentration [13, 19, 20], which in turn allowed the E. coli to achieve enhanced expression and secretion of recombinant protein [12]. It was suggested that the enhanced protein production facilitated by the artificial increase of ATP was due to the up-regulation of genes in biosynthetic pathways, increase of ribosome concentration, and activation of the amino acid-charging reactions on tRNAs. Although the harboring of higher ATP concentration via PCK-overexpression was proven to be beneficial for the production of recombinant protein, the limitation of retarded host’s growth rate was also observed. The PCK-overexpressing E. coli was reported to have a 20 % decrease in the growth rate [20]. A similar phenomenon was also observed in the E. coli strains harboring greater reducing equivalents [14, 15, 18], either through the overexpression of NADH-dependent malic enzyme (MaeA), or in phosphoglucose isomerase (pgi)-disrupted E. coli. The resulting increases of both [NADH] and [NADPH] resulted in decreased growth rates [14].

The degree of artificial ATP increase by plasmid-derived PCK-overexpression was thought to be altered by modified enzymatic activity [21]. The directed evolution of enzymes has been a powerful technique for the improvement of enzymatic characteristics, such as providing faster turnover rate [11] and modification of the optimal conditions [30]. It has not yet been determined whether the growth retardation of the PCK-overexpressing E. coli could be altered by the modification of PCK enzyme activity.

Bacteriorhodopsin is a proton-pumping membrane-integrated protein present in many cyanobacteria [17, 27]. Photon absorption on retinal, a prosthetic molecule in bacteriorhodopsin, triggers isomerization and release of a proton outside the membrane, resulting in the generation of proton motive force. Lee and Jung reported that the expression of a bacteriorhodopsin from Gloeobacter violaceus PCC7421, a primitive cyanobacteria, caused regeneration of light-driven ATP in a membrane inverted system of E. coli [6, 23, 27]. Expression of the Gloeobacter rhodopsin (GR) in E. coli was predicted as another way of artificially increasing the ATP concentration without perturbation of the carbon flux.

In this light, we report that the overexpression of a directly evolved PCK, as well as Gloeobacter rhodopsin with illumination of the culture, can help to gain an understanding of the relationship between retarded growth rate and the artificial increase of ATP in E. coli. Steady-state cells overexpressing PCKWT from chemostat cultures with different dilution rates were also used to verify the relationship. A reason for the retarded growth observed in the E. coli with artificial ATP increase is suggested based on the results.

Materials and methods

Strains and plasmids

All strains and plasmids used in this study are listed in Table 1. E. coli W3110 (Korean Collection of Type Culture, KCTC 2223) and its derivatives were used for studies on the relationship between artificial ATP increase and growth rate. E. coli DH5α (Invitrogen, CA, USA) was utilized for plasmid construction and propagation. E. coli ER2566 (NEB, MA, USA) was used for the expression host in the directed evolution of PCK. Standard molecular cloning procedures were utilized herein [34].

Table 1 Strains and plasmids used in this study
Full size table

The plasmids carrying the genes encoding E. coli PCK (pEcPck) and Gloeobacter rhodopsin (pKJ606-GR) were described in previous studies [19, 23]. To verify the effect of membrane protein expression as control, the gene of dicarboxylate transporter (dctA) from E. coli were PCR-amplified based on the pTrc(Plac hemA +-maeB-dctA) template [24] and the oligonucleotides of CATATGAAAACCTCTCTGTTTAAAAGCC (Nde I site underlined) and GCGGCCGCTTAAGAGGATAATTCGT (Not I site underlined). The DNA fragments (1.3 kb) were sub-cloned into a T-vector (T-blunt, Solgent Co., Daejeon, Korea). After sequence verification, the T-vectors containing PCR-fragments were double-digested with Nde I-Not I and the DNA fragment (1.3 kb) were ligated with the 9.0 kb fragment of the same digested pKA001 (the same with the parental vector of pKJ606-GR except a Nde I site removal), resulting in pKJ606-dctA.

Escherichia coli PEP carboxykinase (pckA) was directly evolved by introducing random mutations, followed by screening based on the PEP carboxylation activity, according to the method in a previous study [11]. A random mutation was introduced into the pckA gene by an error-prone PCR mutagenesis kit (ClonTech Laboratories, CA, USA) with a mutation rate of 2.7 mutations/kb using the pEcPck plasmid DNA as a PCR template and the following oligonucleotides: 5′-CATATGCGCGTTAACAATGGTT (Nde I site underlined) and GAATTCTTACAGTTTCGGACCAGC (EcoR I site underlined). The mutated PCR-fragments (1.5 kb) were double-digested with Nde I and EcoR I, and then ligated with the same digested pET28a+ vector (Merck Millipore, Darmstadt, Germany). The vectors were transformed into the ER2566 strain and used for screening, enzyme purification, and characterization.

The selected PCK derivative (PCK67) was PCR-amplified from the template pET-Pck67, using the oligonucleotides GAATTCCGCGTTAACAATGGTT (EcoR I site underlined) and CTGCAGTTACAGTTTCGGACCAGC (Pst I site underlined). The DNA fragment was ligated with the same digested pTrc99a vector (GenBank M22744.1) after purification, resulting in pEcPck67, and further transformed into the strain W3110. Oligonucleotide synthesis and DNA sequencing were carried out in the facility of Macrogen Co. (Seoul, Korea).

Culture conditions

All E. coli strains were grown in Luria–Bertani (LB) medium [34]. The W3110 derivatives expressing PCK were grown in LB glucose-based medium [13, 19]. A single colony was inoculated in 3 mL of LB medium containing ampicillin (50 μg/mL) and incubated at 37 °C and 220 rpm. Next, 500 μL of overnight grown culture was then transferred into a 250 mL Erlenmyer flask containing 50 mL of medium (25 g/L LB broth, 50 μg/mL ampicillin, 0.1 mM isopropyl β-d-thiogalactopyranoside (IPTG), 9 g/L glucose, 10 g/L sodium bicarbonate). A glutathione stock was dissolved in 0.1 mM dithiothreitol (DTT) solution to prevent oxidation, and added to the media at concentrations of 0–1 mM when necessary. Cells were incubated at 37 °C and 220 rpm for 8 h, and samples were taken every hour.

For the anaerobic cultures of the W3110 derivatives expressing PCK, a 150 mL anaerobic glass serum vial (Bellco Glass Inc., Vineland, NJ, USA) containing 100 mL of the LB glucose-based medium was tightly sealed with a rubber bung with an aluminum cap and further flushed repeatedly with nitrogen gas to remove oxygen prior to inoculation. One mL of overnight grown E. coli strains were inoculated to the sterilized anaerobic vials using a 1 mL syringe and the inoculated vial was maintained at 37 °C and 100 rpm. Culture broth samples were taken using a 1 mL syringe to avoid oxygen contamination.

Batch cultures of the W3110 expressing membrane-integrated proteins were performed in LB glucose-based medium [34] under illumination or dark conditions [6, 22]. A single colony was cultured in a 15 mL tube containing 3 mL of LB medium with ampicillin (50 μg/mL) at 37 °C and 220 rpm overnight, and then an aliquot (500 μL) was transferred to a 250 mL Erlenmyer flask containing 50 mL of minimal media (3 g/L glucose, 0.1 mM IPTG, 50 μg/mL ampicillin, 0.8 g/L NH4Cl, 0.5 g/L NaCl, 7.5 g/L Na2HPO4∙2H2O, 3 g/L KH4PO4, 0.2 g/L MgSO4∙7H2O, 0.1 g/L CaCl2, 1 mg/L thiamine∙HCl) supplemented with all-trans-retinal (Sigma) (5 μM, dissolved in ethanol). For illumination conditions, the E. coli expressing GR were then cultured in a rotary shaking incubator (37 °C and 220 rpm) equipped with light bulb illumination (two 13 W bulbs at a 20-cm distance). For dark conditions, 250 mL Erlenmyer flasks were wrapped with aluminum foil and turned the light bulb off during the cultivation. Batch cultures of the W3110 expressing GR or DctA were performed in the dark conditions. Samples were taken every hour. All batch experiments were repeated at least 3 times.

Chemostat cultures of W3110 expressing PCKWT were performed in minimal glucose-based medium to determine the intracellular ATP and ROS concentrations at the steady states. One milliliter of overnight culture of a single colony, grown in LB medium containing ampicillin (50 μg/mL), was transferred to a 250 mL fermenter jar (Mini-Chemostat Fermenter; Hanil Inc., Bucheon, Korea) containing 100 mL of M9 minimal glucose medium (3 g/L glucose, 0.1 mM IPTG, 50 μg/mL ampicillin, 0.8 g/L NH4Cl, 0.5 g/L NaCl, 7.5 g/L Na2HPO4∙2H2O, 3 g/L KH4PO4, 0.2 g/L MgSO4∙7H2O, 0.1 g/L CaCl2, 1 mg/L thiamine∙HCl). The fermenters were operated at 37 °C and 200 rpm with aeration (150 mL/min). The fresh feed medium (2 L) had the same composition as the initial medium. The inlet and outlet flow rates were set at 10, 20, and 30 mL/h (dilution rates of 0.1, 0.2, and 0.3 h−1, respectively). Steady-state samples were taken after 10 turnovers of the culture volume.

Screening and characterization

The ER2566 strain carrying the pET28a+ vector with pckA derivatives was grown on LB agar plates containing kanamycin (50 µg/mL), and then the colonies were transferred to a 96-well plate containing 400 µL of LB-kanamycin medium. The cells were incubated overnight at 37 °C, 500 rpm, and then 40 μL was used to inoculate another 96-well plate containing 360 µL of fresh medium supplemented with IPTG (0.1 mM). The cells were allowed to express enzymes for 6 h at 37 °C, 500 rpm. After incubation, 200 µL of culture was mixed with 20 µL of lysis buffer (EasyLyse bacterial protein extraction buffer, Epicentre, Madison, WI, USA) and 4 µL of 1 M MgCl2 to disrupt the cells for 10 min at room temperature. The lysed cell extract (50 µL) was then reacted with 150 µL of reaction buffer [2 mM PEP, 50 mM NaHCO3, 5 mM MgCl2, and 2 mM ADP in 100 mM HEPES (2-[4-(2-hydroxyethyl)piperazin-1-yl] ethanesulfonic acid) buffer, pH 7.0, 2.5 U of malate dehydrogenase, 0.14 mM NADH]. After a 10 min reaction at 37 °C, the OAA production was estimated by measuring the absorbance at 340 nm, with an extinction coefficient of 6.22 mM−1 cm−1. Among the tested colonies, 3 displaying increased enzyme activities were selected and subjected to purification using Ni+-affinity chromatography. Enzyme kinetic parameters were determined as previously described [21].

Analysis

Cell growth was measured by determining the optical density at 600 nm (OD600 nm) and then converted into dry cell weight (DCW) using the coefficient of 0.31 g/L.

Intracellular ATP concentration was immediately determined after sampling using an ATP bioluminescent assay kit (Sigma-Aldrich, MO, USA) and a luminometer (Promega, Madison, WI, USA), according to the manufacturer’s instructions. Measurement was performed after sampling and disrupting the cells by sonication (UP200S ultrasonic processor, Hielscher Ultrasonics Co., Teltow, Germany) at 140 W for 30 s on ice [12].

Relative ROSs of the cells were determined using CM-H2DCFDA (5-[and -6]-chloromethyl-2′,7′ -dichlorodihydrofluorescein diacetate), a free radical indicator [37]. A 1 mL aliquot of steady-state cells was harvested from the chemostat (Biotron Inc., Bucheon, Korea), washed with 50 mM phosphate buffer (0.2 mL, pH 7.0), and mixed with 20 μM CM-H2DCFDA (dissolved in dimethyl sulfoxide). After incubation for 30 min at 37 °C, the supernatant was discarded and the pellet was re-suspended in 0.2 mL of the same buffer, then subjected to disruption by sonication (140 W, 30 s) on ice. Cell extracts were transferred into a 96-well black-walled plate with a flat bottom, and the fluorescence intensity was measured using a fluorometer reader (Synergy MX, BioTek, VT, USA) with excitation/emission at 492/530 nm, respectively. Relative ROS was displayed as the ratio of ROSsample/ROScontrol. The ATP and ROS concentrations were analyzed at least 3 times.

Results

Directed evolution of PCK

To acquire a PEP carboxykinase (PCK) with greater enzymatic activity, random mutations were introduced into the wild-type PCK gene (pckA). Among 600 clones expressing mutated enzymes, three were selected based on the greater enzyme activity displayed than the original cell extracts. The enzyme activities of the selected clones were characterized after purification (Table 2). The clone PCK67 demonstrated the greatest increase of PEP carboxylation activity, with 62 % greater catalytic efficiency for the PEP substrate than that of PCKWT, with the same catalytic efficiency for the bicarbonate substrate. The DNA sequencing results of PCK67 indicated that 748 T was changed to G, resulting in change of the serine residue at 250 to alanine.

Table 2 Characteristics of PEP carboxykinase derivatives
Full size table

Effect of PCK-overexpression-mediated increase of ATP on growth rate

To investigate the relationship between artificial ATP increase and growth rate, the activities of ATP-generating enzymes were determined by examining the growth rates of cells expressing the PCKWT or the evolved PCK (PCK67) in different host backgrounds (Table 3). Wild-type E. coli carrying empty vector, which should have the lowest intracellular PCK activity among the tested strains, showed a maximum growth rate of 1.03 h−1 and the intracellular ATP concentration of 0.28 μmol/gcell at log phase under the given conditions. The W3110 P − ppc pckA strain expressed the pckA gene under the control of the glycolytic ppc promoter from its chromosome, and previous study had shown ~80 % of the PEP carboxylation reaction was a result of PPC while ~20 % PCK [21]. This promoter engineered strain was expected to have higher PCK activity than the wild-type strain, and showed a slower growth rate with more accumulation of ATP (0.86 h−1 and 0.54 μmol/gcell, respectively). The E. coli overexpressing PCK from the plasmid, expected to display the highest PCKWT activity, showed an even lower growth rate of 0.84 h−1, with the ATP concentration of 1.00 μmol/gcell. Upon overexpression of PCK67, which possessed 60 % greater activity than PCKWT, the highest overall ATP concentration of 1.45 μmol/gcell and lowest growth rate of 0.66 h−1 were observed.

Table 3 Effects of PCK-derived artificial ATP increase on growth rate under aerobic conditions
Full size table

PPC (PEP carboxylase) shares substrates (PEP and bicarbonate) with PCK (PEP carboxykinase). To eliminate competition for substrate, the wild-type strain was replaced with a PPC-disrupted strain (W3110 ppc −). The W3110 ppc − derivatives could hardly grow in glucose-containing media (both a minimal glucose and a LB glucose medium), though the strains grew well in complex medium (LB) without glucose presence, and the intracellular ATP concentration could not be determined.

Relationship between PCK-mediated artificial ATP increase and ROS

Because toxic reactive oxygen species (ROS) are generated within the electron transfer chain in energy metabolism, which is related to membrane potential, it was assumed that the artificial increase of ATP might also have accelerated the diversion of electrons to oxygen molecules, resulting in an increase of the ROS. To verify this assumption, chemostat cultures were performed, and the intracellular ATP and ROS concentrations were estimated using steady-state cells (Fig. 1). The intracellular ATP concentration itself was not directly related to the growth rate; however, the difference in ATP between the control (W3110 [empty vector]) and the strain overexpressing PCKWT (W3110 [pEcPck]) increased as the dilution rate increased (∆ATP = 0.3 μmol/gcell, ∆ATP = 0.45 μmol/gcell, ∆ATP = 1.05 μmol/gcell at 0.1, 0.2, and 0.3 h−1, respectively). At the same time, while great variation of the ROS concentrations did not occur in the control, the PCKWT-overexpressing W3110 showed drastic increase of the ROS as the ∆ATP increased. The ROS differences between the control and the PCKWT-overexpressing strain were as follows: ∆ROS = 0.5 AU (arbitrary unit)/gcell at D = 0.1 h−1, ∆ROS = 1.2 AU/gcell at D = 0.2 h−1, and ∆ROS = 7.2 AU/gcell at D = 0.3 h−1. The W3110 (pEcPck) never reached the steady state at D = 0.4 h−1 or higher dilution because it was eventually washed out, though the control strain could grow at D = 0.4 h−1 or higher dilution rate. Therefore, the artificial ATP increase was correlated to the ROS concentration in steady-state cells.

Fig. 1
figure 1

Intracellular ATP and relative ROS contents of PCK-overexpressing E. coli at steady state. Bars represent for ATP and symbols for ROS Control, W3110 strain carrying empty vector (white circle, white bar); PCKWT, W3110 strain carrying pEcPck (black circle, black bar). Data are represented as mean ± SD from at least three independent experiments

Full size image

To verify this chemostat results, batch cultures of the few PCK-derived artificial ATP-increased strains from the previous results were repeated and their ROS concentrations were analyzed (Table 3, repeated experimental set). Indeed, the strains with the artificial ATP increase by pckA promoter exchange and Pck67-overexpression (W3110 P −ppc pckA [pTrc99a] and W3110 [pEcPck67] in Table 3) contained 34 and 60 % greater intracellular ROS than the control, respectively. Therefore, it was confirmed that the PCK-derived artificial ATP-increased strains contained greater ROS concentration in chemostat and batch cultures.

Effect of ROS scavenger on growth rate in PCK-overexpressing E. coli

To confirm that the observed increase in ROS caused by the artificial ATP increase was related to the growth retardation, cells were grown in medium containing glutathione as an antioxidant, and their growth rates in the log phase were estimated (Table 4). The growth rate of the control strain (W3110 harboring an empty vector) did not vary depending on the glutathione concentration, which was added in the range of 0–1 mM. However, the strain with artificial ATP increase (W3110 [pEcPck]) demonstrated recovery of the growth rate as the glutathione concentration increased. The growth retardation of W3110 (pEcPck) was fully recovered in the 0.1 mM glutathione-containing medium.

Table 4 Effects of glutathione addition on growth retardation caused by PCK-derived artificial ATP increase
Full size table

Effect of ATP increase via expression of bacteriorhodopsin on growth rate

Because the overexpression of PCK or knockout of PPC might have caused perturbations of metabolic flux (between PEP and OAA node), the PCK-derived artificial ATP increase might have not been directly related to the growth retardation. To clear the issue by eliminating the perturbation of carbon flux, Gloeobacter rhodopsin (GR), a light-induced proton pump [6, 23], was introduced into E. coli, and the cells were cultured under illumination (Fig. 2a). Escherichia coli W3110 (empty vector), the negative control, displayed the maximum growth rate of 0.35 h−1, with the intracellular ATP concentration of 0.72 μmol/gcell and ROS of 905 AU/gcell at the log phase. In contrast, the W3110 (pKJ606-GR) strain displayed the lower growth rate of 0.19 h−1, the higher intracellular ATP of 1.14 μmol/gcell and ROS of 1646 AU/gcell which corresponded to a 47 % lower growth rate with 81 % greater intracellular ATP and 82 % greater relative ROS than the control. The artificial ATP increase without carbon flux variation (GR-expression+illumination) also reduced growth rate with increase of ROS.

Fig. 2
figure 2

Growth profile, intracellular ATP, and ROS of membrane-integrated protein expressing E. coli strains. a GR-expression under illumination, b GR-expression under dark conditions, c DctA-expression. Circle symbols for growth, open bars for ATP, and shaded bars for ROS Control, W3110 strain carrying empty vector (white circle, white bars); membrane protein expressions (black circle, gray bars). ATP and ROS were measured at 6 h. Data are represented as mean ± SD from at least three independent experiments

Full size image

In the dark conditions (no artificial ATP increase conditions), growth profiles of E. coli strains expressing GR and dicarboxylate transporter (dctA), a control membrane-integrated protein with no energy generation, were also examined (Fig. 2b, c). The GR-expression under dark conditions even reduced the growth rate (26 % of control) with elaborated ROS (6.5-fold of control). The ATP concentration was 0.34 μmol/gcell (vs. 0.76 μmol/gcell of control). The DctA-expressing E. coli also showed similar trend with the GR-expressing E. coli under dark conditions, reduced growth rate (61 % of control), elaborated ROS (11.1-fold of control), with lower ATP level (0.14 μmol/gcell vs. 0.76 μmol/gcell of control).

Growth rate comparison of the PCK-mediated artificial ATP increases under anaerobic conditions

All the evidences of the above agreed that the ROS increase came from the artificial ATP increase was the reason of the growth retardation. To complement the theory, E. coli strains with various artificial ATP increase levels were cultivated under strict anaerobic conditions, which no ROS should be provoked, and compared with the growth under aerobic conditions (Fig. 3). Under aerobic conditions, the most ATP-harboring strain (W3110 [pEcPck67]) had the most ROS and the slowest growth rate (Fig. 3a, came from the data set of Table 3). Unlike the aerobically grown E. coli strains that showed the reciprocal relationship between the growth rate and the artificial ATP increase, no oxygen conditions enabled the E. coli strains to grow faster as the more artificial ATP increased in order of W3110(pEcPck67) > W3110(pEcPck) > W3110 P − ppc pckA(empty vector) > W3110(empty vector) (Fig. 3b). Relative ROS was varied in a limited range of ~1.2-fold compared with those from the aerobic conditions (relative ROS ~1.8-fold).

Fig. 3
figure 3

Growth profiles, intracellular ATP, and ROS of PCK-derived artificial ATP-increased E. coli strains. a Aerobic conditions, b Anaerobic conditions. Circle symbols for growth, open bars for ATP, and shaded bars for ROS Control, W3110 carrying empty vector (white); W3110 P − ppc pckA (light gray); W3110 overexpressing PCKWT (dark gray); W3110 overexpressing PCK67, (black). ATP and ROS were measured at mid-log phase (4 h for aerobic conditions). Data are represented as mean ± SD from at least three independent experiments

Full size image

Discussion and conclusions

Artificial ATP increases of cells could have biotechnological benefits, such as allowing enhancement of the translation and secretion of recombinant protein [12]. Active transport of fermentation substrate influx and product efflux could be another potential benefit of using cells harboring extra ATP considering that the energy state in E. coli (logarithm function of [ATP]/[ADP]) was linearly correlated to the activity of phosphotransferase uptake system [32], and one might expect that artificial ATP increase would enhance the glucose-derived metabolism. However, observed retardation of the growth rate caused by the artificial increase of ATP was noted as a limitation for bio-process applications [21]. The results of this study indicated that the artificial increase in ATP indeed caused the cells to grow more slowly, in a dose-dependent manner (Tables 3; Fig. 1). The artificial ATP increase was achieved herein by the “extra ATP-generating” activities of PCK, and the growth rate was observed to be inversely related to the amount of ATP (Table 2; Fig. 3a). The strain overexpressing the PCK mutant with the highest activity (PCK67) without a substrate-competing PEP carboxylase (PPC) host (W3110 ppc −), expected to harbor the greatest ATP concentration, did not grow in the presence of glucose as other reports (no growth under aerobic conditions of ppc-disrupted E. coli on M9 medium with glucose [31], on M9 medium with glycerol [10], and MOPS medium with glucose [1]). The results from expression of Gloeobacter rhodopsin (GR) also agreed with the theory that growth was inhibited by the artificial ATP increase (Fig. 2a). The artificial ATP increase by GR-expression contributed to the increase of ATP under illumination conditions, resulting in reduction of the growth rate. Because the ATP increases by GR-expression did not provoke perturbations of the carbon flux, unlike those by metabolic PCK expression, this result confirmed that the artificial ATP increase itself was responsible for the observed growth retardation.

Superoxide radical, hydroxyl radical, and hydrogen peroxide are reactive oxygen species (ROS), which are mainly formed during respiration by the incomplete reduction of oxygen [8]. Because oxidative stress caused by the high oxidizing potential of ROS leads to cellular damage, such as mutations and metabolic pathway disruption, increases of ROS could inhibit cell growth [22]. ROS damage from the oxygenic bio-process of aerobic organisms is unavoidable, so cells defend themselves by expressing ROS-neutralizing enzymes, or by producing ROS-scavenging metabolites [28]. We suspected that ATP synthase and the electron transfer reactions might somehow be influenced by the artificial increase of ATP, and that oxygen, the final electron acceptor, might consequently be incompletely reduced. Indeed, the E. coli with artificial ATP increase showed a dramatic increase of ROS at the steady state (Fig. 1) and the repeated experimental sets of batch cultures (Table 3). Conversely, the addition of the antioxidant glutathione (>0.1 mM) into the media allowed restoration of the retarded growth of the E. coli (Table 4) (The intracellular glutathione concentration in glucose-fed, exponentially growing E. coli was known to be 2.4 mM [2]), which supports the theory that retardation of the growth rate by artificial ATP increase was mediated by increased ROS. In the no ROS-provoking (strict anaerobic) conditions, the growth of artificial ATP-increased strains even grew faster than control strain (Fig. 3b). In the case of eukaryotic mitochondrial oxidation state, it was reported that the correlations between ATP and ROS can result in programmed cell death and axonal degeneration [29, 33]. Similar studies in other species were found to support the hypothesis that the growth retardation by ROS can be recovered by the use of antioxidants as free radical scavengers in other living organisms [25, 35].

Mechanistic reason of the ROS increase by the artificial ATP increase requires more studies. A plausible explanation is that the artificial-driven ATP increase might have led more protons accumulation across membrane and the strengthened membrane potential might further have influenced on many membrane-bound dehydrogenases and mobile electron carriers to mis-transfer a pair of electrons to oxygen molecule, which would increase ROS. Although an E. coli expressing periplasmic recombinant protein (e,g., human interleukin-2) has been reported to provoke ROS increase [26], the growth defect therefrom was not as severe as the expression of membrane-integrated proteins in this study (Fig. 2b, c). The overexpression of membrane-integrated proteins, i.e., GR under dark conditions and dicarboxylate transporter, should have interfered the movement of electron carriers, ed the drastic ROS increase (6.5-fold Fig. 2b and 11.1-fold in Fig. 2c). These results would strengthen the plausible explanation. Measuring the redox state of membrane components, genetic deletions of the proteins in oxidative phosphorylation [3] would be also helpful to understand the mechanistic reason.

Although E. coli possesses many genes to defend against ROS, it might have incorrectly responded to the additional ROS driven by the artificial increase in ATP, an unnatural physiological condition. Table 5 presents the variation of ROS defense genes in the PCK-overexpressing E. coli W3110 at D = 0.1 h−1 steady state in complex glucose medium [20]. The previous DNA microarray results indicated that the higher ATP-harboring cell by the PCK-overexpression up-regulated the transcriptions of genes in biosynthetic pathway (i.e., amino acids and nucleotides) and flagella structural genes, while down-regulated the transcription of genes in carbon metabolism (Gene Expression Omnibus access code: GSM211801). Superoxide dismutases, peroxidases, and catalases, however, were slightly down-regulated or showed no difference in expression levels even though the ATP was artificially increased, which must have increased the ROS levels as well. If the ROS defense genes have been expressed in the strain with artificial ATP increase, one might expect recovery of the growth rate without the addition of antioxidants to the medium. The artificial ATP increase was correlated with the faster growth rate and yield under the anaerobic conditions (Fig. 3), which support the above expectation.

Table 5 Variation of transcriptome of high ATP-charged cells, focused on defense against ROS
Full size table

In conclusion, the reduction in growth rate occurred as a result of the artificial ATP increase, coupled with increase in intracellular ROS, regardless of carbon perturbation. The growth rate could be even faster as the more artificial ATP increase under anaerobic conditions. Therefore, an ROS removal strategy is required for metabolically engineered cells harboring greater ATP for practical applications.