Quantitative analysis of an engineered CO2-fixing Escherichia coli reveals great potential of heterotrophic CO2 fixation

Background Production of fuels from the abundant and wasteful CO2 is a promising approach to reduce carbon emission and consumption of fossil fuels. Autotrophic microbes naturally assimilate CO2 using energy from light, hydrogen, and/or sulfur. However, their slow growth rates call for investigation of the possibility of heterotrophic CO2 fixation. Although preliminary research has suggested that CO2 fixation in heterotrophic microbes is feasible after incorporation of a CO2-fixing bypass into the central carbon metabolic pathway, it remains unclear how much and how efficient that CO2 can be fixed by a heterotrophic microbe. Results A simple metabolic flux index was developed to indicate the relative strength of the CO2-fixation flux. When two sequential enzymes of the cyanobacterial Calvin cycle were incorporated into an E. coli strain, the flux of the CO2-fixing bypass pathway accounts for 13 % of that of the central carbon metabolic pathway. The value was increased to 17 % when the carbonic anhydrase involved in the cyanobacterial carbon concentrating mechanism was introduced, indicating that low intracellular CO2 concentration is one limiting factor for CO2 fixation in E. coli. The engineered CO2-fixing E. coli with carbonic anhydrase was able to fix CO2 at a rate of 19.6 mg CO2 L−1 h−1 or the specific rate of 22.5 mg CO2 g DCW−1 h−1. This CO2-fixation rate is comparable with the reported rates of 14 autotrophic cyanobacteria and algae (10.5–147.0 mg CO2 L−1 h−1 or the specific rates of 3.5–23.7 mg CO2 g DCW−1 h−1). Conclusions The ability of CO2 fixation was created and improved in E. coli by incorporating partial cyanobacterial Calvin cycle and carbon concentrating mechanism, respectively. Quantitative analysis revealed that the CO2-fixation rate of this strain is comparable with that of the autotrophic cyanobacteria and algae, demonstrating great potential of heterotrophic CO2 fixation. Electronic supplementary material The online version of this article (doi:10.1186/s13068-015-0268-1) contains supplementary material, which is available to authorized users.

Heterotrophic microbes usually do not assimilate CO 2 through the central metabolism. Recent studies indicated that incorporation of several steps of a natural carbon fixation pathway into a heterotrophic microbe may create a CO 2 -fixing bypass pathway which enables the host to assimilate CO 2 at the expense of carbohydrates. Examples include introduction of two enzymes of Calvin cycle into Escherichia coli and Saccharomyces cerevisiae, which resulted in enhanced CO 2 recycling in an air-tight fermentor [19] and an increased ethanol yield [20], respectively.
Although these preliminary data suggested that heterotrophic CO 2 -fixation is feasible, little is done to quantitatively analyze and evaluate the process. To date, simple approaches capable of evaluating the CO 2 flux in heterotrophic microbes are still lacking, since the metabolites of the CO 2 -fixing bypass pathway are indistinguishable from those of the central metabolic pathway. Due to lack of quantitative analysis, it remains unclear where the bottleneck for heterotrophic CO 2 -fixation is and whether the rate of heterotrophic CO 2 -fixation is higher, lower, or comparable with that of autotrophic CO 2 -fixation.
The aim of this study was to address the above issues through a quantitative and comprehensive analysis of the heterotrophic CO 2 -fixation process. To evaluate the strength of CO 2 flux, a metabolic flux index, MFI h-CO2 , was developed to indicate the metabolic flux ratio between the CO 2 -fixing bypass pathway and the central carbon metabolic pathway. The MFI h-CO2 was determined by addition of 13 C-labeled sodium bicarbonate into the culture medium, followed by quantification of the isotropic-labeled and unlabeled forms of one intracellular metabolite by liquid chromatography-mass spectrometry/ mass spectrometry (LC-MS/MS). Comparison of MFI h-CO2 values of several engineered CO 2 -fixing E. coli strains led to identification of the rate-limiting steps of heterotrophic CO 2 fixation. The strain with the highest MFI h-CO2 value was aerobically cultivated in minimal medium supplemented with xylose in a chamber filled with 5 % CO 2 . The mass of fixed CO 2 per liter culture of this strain per hour was calculated by the mass balance of carbon. The CO 2fixation rate in E. coli was then compared with those of several autotrophic microbes to evaluate the potential of heterotrophic CO 2 fixation.

Results
Development of a metabolic flux index, MFI h-CO2 , for relative quantification of heterotrophic CO 2 fixation It is costly and time-consuming to determine the absolute metabolic flux of CO 2 fixation by quantifying every isotropic-labeled metabolite upon the feed of 13 CO 2 during cultivation. As the metabolic flux of the central metabolism for a given strain is quite stable, the relative metabolic flux of the CO 2 -fixing bypass pathway over that of the central carbon metabolic pathway may give a quantitative understanding on the efficiency of CO 2 fixation. This relative value is then termed as the metabolic flux index of the heterotrophic CO 2 -fixation pathway, MFI h-CO2 . At the conjunction of the CO 2 -fixing bypass pathway and the central pathway, the metabolite generated by the two pathways can be differentiated by using 13 C-labeled CO 2 and unlabeled sugar. The amount of the labeled and unlabeled forms of the joint metabolite can be determined and used to calculate the metabolic flux ratio of the two pathways to obtain the MFI h-CO2 value.
Herein, we use a heterotrophic CO 2 -fixing E. coli strain as a model to elucidate how MFI h-CO2 is calculated. The strain was constructed by incorporating two sequential enzymes in the cyanobacterial Calvin cycle, phosphoribulokinase (PRK), and ribulose-1,5-bisphosphate carboxylase/ oxygenase (Rubisco) into the central metabolism of E. coli. The incorporated CO 2 -fixing bypass pathway starts at ribulose 5-phosphate (Ru5P) in the pentose phosphate pathway of the central metabolism and ends at 3phosphoglycerate (3PGA) in the glycolysis of the central metabolism ( Fig. 1). When the strain is cultured in medium supplemented with 13 C-labeled sodium bicarbonate, intracellular 13 CO 2 , either generated by diffusion of the extracellular dissolved 13 CO 2 or by the equilibrium of 13 C-labeled bicarbonate after its active transportation into cell, will be used as the substrate for Rubisco.
As shown in Fig. 1, we assume a mole of 3PGA is generated from the central pathway and b mole of 13 CO 2 is fixed by the Rubisco pathway in a given period of time. Then (a + b) mole of unlabeled 3PGA and b mole of 13 C-3PGA are generated. At the same period of time, we assume c mole of unlabeled 3PGA and d mole of 13 C-3PGA are channeled into the downstream metabolism. It was reported that a small fraction of 13 C isotope was coupled with all natural 12 C-containing compounds [21][22][23]. We then cultivated E. coli strains in medium free of any carbon isotope and determined the ratio of 13 C-3PGA to the unlabeled 3PGA as the basal isotopic level. The ratio was 3.45 % as shown in Additional file 1: Figure S1. We thus assume that 3.45 % of unlabeled 3PGA will convert to its isotopic form. Therefore, the actually detected molar amount of 13 C-3PGA (y) can be calculated by Eq. (1), while the actually detected unlabeled 3PGA (x) can be calculated by Eq. (2).
Under a metabolic steady-state, the relationship of d, c, x, and y is shown in Eq. (3).
In this case, only the concentration of 13 C-labeled and unlabeled 3PGA are required to be determined to calculate the MFI h-CO2 . Compared with quantification of all intracellular isotropic metabolites to calculate the absolute metabolic flux, we argue that the determination of MFI h-CO2 to evaluate the relative metabolic strength of the CO 2 -fixation pathway would be a simple and convenient alternative.

Construction of a heterotrophic CO 2 -fixing E. coli
The Rubisco-encoding genes rbcL-rbcX-rbcS from Synechococcus sp. PCC7002 and the PRK-encoding gene prk from Synechococcus elongatus PCC7942 were cloned into pET30a as described previously [24]. The resulted plasmid was designated as pET-RBC-PRK in this study. To verify the function of CO 2 -fixation pathway, Rubisco, and/or PRK were deactivated by introducing site-directed mutations to their conserved catalytic residues, yielding another three plasmids, pET-RBC197-PRK, pET-RBC-PRK2021, and pET-RBC197-PRK2021. Among them, RBC197 indicates a K197M mutation in the conserved catalytic site of the large subunit of Rubisco [25], and PRK2021 carries K20M and S21A mutations in the conserved nucleotidebinding sites of ATP-binding proteins [26].
The MFI h-CO2 values of strain BL21(DE3)/pET-RBC-PRK at different induction times were calculated to evaluate its relative CO 2 flux (Fig. 2c). For a period of 13 h induction, the MFI h-CO2 of the control strain BL21(DE3)/ pET30a was below 0.03. Whereas, the MFI h-CO2 values of strain BL21(DE3)/pET-RBC-PRK was increased from 0.07 at 3 h to 0.13 at 6 h and then slightly decreased to 0.12 at 13 h. The increase of MFI h-CO2 values from 3 to 6 h was associated with the increase of Rubisco expression level (Fig. 2d), suggesting that the increased Rubisco activity contributed to the increased metabolic flux of CO 2 fixation. When protein expression reached a high level from 6 h onwards, the MFI h-CO2 also reached its highest value.

Identification of the bottleneck of heterotrophic CO 2 fixation
Rubisco was generally considered as the rate-determining step in the Calvin cycle of autotrophic microbes due to its extremely low catalytic efficiency [28,29]. For the heterotrophic E. coli strain BL21(DE3)/pET-RBC-PRK harboring a partial Calvin cycle, accumulation of RuBP was observed even in the case of leaky-expression of PRK but overexpression of Rubisco. This result suggested that the Fig. 2 The intracellular 13 C-3PGA (a), cell growth (b), MFI h-CO2 values (c), and soluble protein expression (d) of BL21(DE3) strains harboring different plasmids. All strains were 1:100 inoculated into LB medium containing 100 mM NaH 13 CO 3 and shaken at 37°C. When the culture reached the mid-log phase (OD 600 = 0.4-0.6), 0.02 mM IPTG was added to induce Rubisco expression and the induction temperature was reduced to 22°C (zero point). The PRK-encoding gene under the control of a tryptophan-regulated promoter trpR-P trp was leakily expressed in LB medium. RbcL and RbcS are the large and small subunits of Rubisco, which are encoded by rbcL and rbcS genes, respectively. RbcX is the specific chaperon of Rubisco, which is encoded by the rbcX gene. Molecular weight standards from top to bottom are 80, 60, 40, 30, 20, and 12 kDa Rubisco-catalyzed reaction is one of the rate-limiting steps of the CO 2 -fixing bypass pathway in heterotrophic E. coli (Additional file 1: Figure S4A). Owing to the difficulty in improving the catalytic activity of Rubisco, we attempted to increase the substrate supply (RuBP or CO 2 ) for Rubisco to drive the reaction forward.
To increase the supply of RuBP, the weak promoter trpR-P trp for PRK expression was replaced by a strong promoter P T7 , yielding a plasmid pET-RBC-T7-PRK. A significant increase of PRK expression level and an 8.6fold increase of intracellular RuBP was observed after promoter replacement (Additional file 1: Figure S4). However, no significant difference in the MFI h-CO2 value (a P value of 0.36 using the Student T test) was observed after increasing the intracellular RuBP amount (Fig. 3), indicating that RuBP supply was not the rate-limiting factor.
To increase CO 2 supply, the unique cyanobacterial carbon concentrating mechanism (CCM) was introduced into E. coli. In cyanobacteria, bicarbonate is first transported to plasma membrane by bicarbonate transporter (BT), diffused into caboxysome, and then converted to CO 2 by carbonic anhydrase (CA) and finally catalyzed by Rubisco therein [30]. To mimic this CCM in E. coli, single BT-or CA-encoding gene from Synechococcus sp. PCC7002, and their combinations, were respectively introduced into E. coli. The bicA gene, which encodes a Na + -dependent BT with high flux rate [31], was fused with promoter trpR-P trp and then inserted into pET-RBC-PRK to generate pET-RBC-PRK-BT. The MFI h-CO2 value of strain BL21(DE3)/pET-RBC-PRK-BT exhibited a decrease of 34.1 % compared with that of strain BL21(DE3)/pET-RBC-PRK (Fig. 3). This can be speculated that the increase of intracellular bicarbonate might cause pH variance and possibly affect expression or function of Rubisco or PRK. Moreover, bicarbonate has to be converted to CO 2 so as to be catalyzed by Rubisco. The equilibrium of bicarbonate and CO 2 under intracellular condition (e.g., pH 7.5) give the ratio of [HCO 3 -]/[CO 2 ] to be 14 (the pK a of H 2 CO 3 is 6.35 [32]). The increment of intracellular CO 2 is thus only 7 % of that of bicarbonate. All these indicated that increasing the intracellular bicarbonate by BT expression was not an effective mean to improve heterotrophic CO 2 fixation.
The CA-encoding gene (ccaA) was fused with a mutated constitutive bacteriophage promoter P L -AA [33] and then inserted into pET-RBC-PRK and pET-RBC-PRK-BT. The resultant strains BL21(DE3)/pET-RBC-PRK-CA and BL21(DE3)/pET-RBC-PRK-BT-CA showed MFI h-CO2 values of 0.17 and 0.11, respectively, which were 39.8 and 40.7 % higher than those of their respective parent strains without CA insertion (Fig. 3). Overexpression of CA increased the metabolic flux of heterotrophic CO 2 -fixation, indicating that CO 2 supply is a limiting factor for CO 2 fixation in E. coli.

Determination of the CO 2 -fixation rate of the heterotrophic E. coli
It was reported that E. coli metabolized 99 % of the sugar carbon into biomass, CO 2 , and acetate under aerobic condition [34]. However, no obvious fermentation product was detected for the CO 2 -fixing and control E. coli strains after 24 h of aerobic cultivation (Additional file 1: Figure S5). The carbon balance calculation of the control strain BL21(DE3)/pET-RBC197-PRK2021 without the ability of CO 2 -fixation also confirmed that the biomass and released CO 2 accounted for 96 % of the consumed sugar carbon. According to the mass balance of carbon, the fixed CO 2 of the CO 2 -fixing E. coli strain can be calculated by Eq. (5), where all values are in the molar amount of carbon.
The specific CO 2 secretion rate of a given E. coli is a constant, which was 11.8 mmol g dry weight −1 h −1 reported in one literature [35] and 18.6 mmol g dry weight −1 h −1 in another [34]. Assuming the value is k, Eq. (5) can be transformed to Eq. (6).
Mass balance of carbon for the control strain BL21(DE3)/pET-RBC197-PRK2021, which harbored the two deactivated enzymes of the CO 2 -fixing pathway, can generate Eq. (7). Assuming the specific CO 2 secretion rate of the control strain is k', Eq. (7) will be transformed to Eq. (8).
Since CO 2 is mainly generated from the tricarboxylic acid cycle of E. coli under aerobic conditions, the incorporated CO 2 -fixing pathway, which is a bypass of the upstream glycolysis, would not affect the specific CO 2 secretion rate of the strain. Then, under the same cultivation condition, we can assume Eq. (9).
Solution to Eqs.
Two CO 2 -fixing E. coli strains and the control strain were aerobically cultivated in 200 mL of M9 minimal medium supplemented with 10 g L −1 xylose in an Erlenmeyer flask. The flask was placed in an air-tight container (10 L) prefilled with 5 % CO 2 and 95 % air and shaken at room temperature for 24 h. The pH variance, consumed xylose, and generated dry cell weight were determined (Table 1). All cultures maintained a stable pH, with a fluctuation of less than 0.2 unit. Calculation using Eq. (10) indicated that stains BL21(DE3)/pET-RBC-PRK and BL21(DE3)/pET-RBC-PRK-CA were able to fix 13.3 and 19.6 mg CO 2 L −1 h −1 , respectively. The 47.4 % of increment in the CO 2fixation rate after CA expression was similar to the 39.8 % of increment in the MFI h-CO2 value, which confirmed that the MFI h-CO2 was reliable for evaluating the CO 2 -fixation flux in the heterotrophic E. coli. The CO 2 -fixation rates of the heterotrophic E. coli strains constructed in this study were compared with those of the natural CO 2 -fixing autotrophic microbes (Table 2). Fourteen autotrophic microbes including microalgae, cyanobacteria, and nongreen algae fixed CO 2 at rates ranging from 10.5 to 147.0 mg CO 2 L −1 h −1 , with the median value of 21 mg CO 2 L −1 h −1 . The CO 2 -fixing E. coli strains were able to fix CO 2 at rates of 13.3-19.6 mg CO 2 L −1 h −1 , which were comparable to the capacity of the autotrophic microbes.

Discussion
Recycling CO 2 directly into fuels or chemicals is a potential approach to reduce carbon emission as well as to resolve energy crisis [6,7]. The past 5 years have witnessed great success in production of CO 2 -derived molecules that have potential to be used as fuels and chemicals by autotrophic microbes. Quantitative analysis in this study revealed that an engineered heterotrophic E. coli could assimilate CO 2 at a rate comparable to that of the autotrophic cyanobacteria and algae. It is noteworthy that the specific CO 2 -fixation rates of the E. coli strains were superior to most of the autotrophic microbes listed in Table 2. Since E. coli can easily grow to a high density in fermentors under well-controlled conditions, we believe that heterotrophic microbes might be an alternative candidate for CO 2 fixation with great potential.
The most striking advantage of using heterotrophic microbes for CO 2 fixation is their fast growth rates. The doubling times for E. coli and yeast are only 20 min [36] and 2 h [37], respectively, whereas those for common cyanobacteria and algae are in the range of 8-44 h [38,39]. Most autotrophic microbes use photosynthesis to provide energy for CO 2 assimilation and ultimately biomass accumulation. The theoretical maximum of solar energy conversion efficiency in photosynthesis is only 8-10 % [40], whereas the actual values for several species of cyanobacteria, microalgae, and plants do not exceed 3 % [41]. The low efficiency of photosynthesis can be ascribed to many inherent factors including insufficient absorption of all light wavelengths during light-dependent reactions and low carboxylation activity of Rubisco and existence of energy-consuming photorespiration during light-independent reactions [42]. Although many efforts have been made [43,44], dramatic increases in photosynthetic efficiency as well as growth rate are still big challenges for autotrophic microbes [44]. However, billions of years of evolution have enabled the heterotrophic microbes to efficiently assimilate the high-energy sugars to generate both carbon backbone and energy at the same time. Therefore, heterotrophic microbes might be a better   choice for CO 2 fixation, since the fixed CO 2 can be easily joined into the central metabolism and then be efficiently metabolized.
For the current version of the CO 2 -fixing E. coli strain constructed in this study, CO 2 was fixed at the expense of sugar consumption because all energy required for CO 2 fixation comes from sugar. However, it is not unbelievable that CO 2 fixation can occur without sugar consumption in heterotrophic microbes once energy can be supplied from other sources. The pioneer work by Liao's group has demonstrated that electricity can be used as the sole energy to convert CO 2 to higher alcohols in Ralstonia eutropha [8], opening the door of employing other energy forms for CO 2 fixation.
There is no doubt that improving the carboxylation activity of Rubisco is the ultimate way to increase the efficiency of CO 2 fixation in both autotrophic and heterotrophic microbes. However, decades of Rubisco engineering gained limited success [24,45]. In this work, the difficulty of Rubisco in access to CO 2 was found to be another limiting factor of heterotrophic CO 2 fixation. Expression of the CA from Synechococcus sp. PCC7002 under a weak constitutive promoter increased the E. coli CO 2 -fixation rate by 47.4 %. It is thus suggested that screening of the CA gene and optimization of its expression might be feasible ways to further improve the heterotrophic CO 2 -fixation rate. CA, which catalyzes the reversible interconversion of CO 2 and HCO 3 − , is widely existed in animals, plants, archaebacteria, and eubacteria, and plays an important role in many physiological functions [46]. Although some CAs prefer the direction of CO 2 hydration, the carboxysomal CAs in cyanobacteria and some chemoautotrophic bacteria favor the direction of HCO 3 − dehydration. To date, two forms of carboxysomal CAs (α and β), which are encoded by three types of genes with distinct sequences and structures (CsoSCA for α-CA and CcaA and CcmM for β-CA), were reported [47,48]. The selected CA-encoding gene from Synechococcus sp. PCC7002 in this study was the CcaA gene. Whether the other two types of CA-encoding genes can be expressed in E. coli and whether their expression can increase the heterotrophic CO 2 -fixation rate are now under investigation by our group. Moreover, a stronger inducible promoter might be employed to enhance the CA expression in a controllable way to further improve the CO 2 supply.
As a compensation for the low carboxylation activity of Rubisco, some autotrophic microbes have evolved some physical barriers (e.g., the semi-permeable caboxysome in cyanobacteria and the bundle sheath cells in C4 plants) to concentrate CO 2 around Rubisco. Inspired by these, we suppose that constraining CO 2 and the CO 2fixing enzyme in a microcompartment (e.g., reconstruction of the caboxysome in E. coli [49]) or recruiting the CO 2 -producing and CO 2 -fixing enzymes in a protein/ RNA scaffold in E. coli might be an alternative way to further improve its CO 2 -fixation rate.

Conclusions
In this study, quantitative analysis approaches have been developed for CO 2 fixation in heterotrophic microbes. The difficulty in access to CO 2 was found to be a limiting factor for heterotrophic CO 2 fixation. An E. coli strain capable of fixing CO 2 at a rate of 19.6 mg CO 2 L −1 h −1 or 22.5 mg CO 2 g DCW −1 h −1 was constructed by incorporation of partial cyanobacterial Calvin cycle and carbon concentrating mechanism. This work demonstrated that CO 2 fixation by the engineered heterotrophic E. coli can be as effective as the natural autotrophic cyanobacteria and algae, showing great potential of heterotrophic CO 2 fixation.

Plasmids construction
All plasmids were constructed based on pET30a (Additional file 1: Table S1) and transformed to E. coli BL21 (DE3) for protein expression. The primers used are listed in Additional file 1: Table S2.

Isotropic assay for CO 2 -fixation efficiency
A fresh single colony of the strain was inoculated into LB medium containing 50 ng μL −1 kanamycin and cultured overnight at 37°C. An aliquot of 100 μL of the overnight culture was inoculated into 40 mL fresh LB medium containing 50 ng μL −1 kanamycin, 100 mM hydroxyethylpiperazine ethanesulfonic acid (HEPES), and 100 mM NaH 13 CO 3 (Sigma). The culture was shaken at 37°C until its OD 600 reached 0.4-0.6. Then the temperature was reduced to 22°C for maximal protein expression. At intervals, 3 OD 600 of cells were harvested for SDS-PAGE and 8 mL of cells for intracellular metabolites extraction.
For intracellular metabolites extraction, all experiments were done on ice. At first, 10 mL of culture were rapidly centrifuged and washed in 10 mL cold (−20°C) aqueous methanol solution (60 %, v/v) to quench cell metabolism as soon as possible. The suspension was clarified at −20°C for 5 min at 20,000 g. The cell pellet was resuspended in 80 μL cold (−20°C) aqueous methanol solution (60 %, v/v). After addition of 100 μL of 0.3 M KOH (dissolved in 25 % ethanol), the mixture was stored at −80°C for more than 2 h to break the cell wall. The alkaline mixture was thawed on ice and neutralized by adding 2 μL of glacial acetic acid. Then the sample was centrifuged at −20°C for 10 min at 20,000 g. The supernatant was stored at −80°C before LC-MS/MS detection [50].

LC-MS/MS detection
Agilent 6460 series LC-MS/MS system equipped with a HPLC system and a triple-quadrupole Mass Spectrometer were used. All samples were separated by the reversed phase ion pair high performance liquid chromatography with Agilent XDC18 column (5uM, 150 mm × 4.6 mm). The negative ion and selected multiple reactions monitoring (MRM) mode were used for MS detection. Di-nbutylammonium acetate (DBAA) was used as the volatile ion pair reagent. DBAA and standard metabolites (3PGA and RuBP) were purchased from Sigma-Aldrich. Methanol was purchased from Fisher Scientific [51]. The mobile phase was the mixture of solution A (water with 5 mM DBAA) and solution B (methanol with 5 mM DBAA) prepared at the gradient shown in Additional file 1: Table S3. The flow rate was 0.6 mL min −1 . The injection volume was 50 μL and the column temperature was 40°C.
The settings for MS were as follows: gas temperature, 350°C; gas flow, 8 L min −1 ; nebulizer, 38 psi; sheath gas temperature, 350°C; sheath gas flow, 9 L min −1 ; capillary, −3500 V; nozzle voltage, 500 V. The dwell time was set at 200 ms. The MRM parameters were optimized by the standards, and the detailed values for Q1 (m/z of precursor ion), Q3 (m/z of product ion), fragmentor, and collision energy (CE) were listed in Additional file 1: Table S4. All metabolites were quantified by their standard curves.

HPLC detection
The concentrations of xylose in medium before and after cultivation were determined using an Agilent 1200 high performance liquid chromatography (Agilent Technologies, Santa Clara, CA, USA) with a refractive index (RI) detector. An Aminex HPX-87 H organic acid analysis column (7.8 × 300 mm) (Bio-Rad Laboratories, Inc, CA, USA) was maintained at 15°C with 0.05 mM sulfuric acid as mobile phase. The injection volume was 10 μL and the flow rate was 0.5 mL min −1 .