Evolution of pyruvate kinase-deficient Escherichia coli mutants enables glycerol-based cell growth and succinate production

Authors


Abstract

Aims

The aim of this study was to engineer Escherichia coli strains that efficiently produce succinate from glycerol under anaerobic conditions after an aerobic growth phase.

Methods and Results

We constructed E. coli strain ss195 with deletions of pykA and pykF, which resulted in slow growth on glycerol as sole carbon source. This growth defect was overcome by the selection of fast-growing mutants. Whole-genome resequencing of the evolved mutant ss251 identified the mutation A595S in PEP carboxylase (Ppc). Reverse metabolic engineering by introducing the wild-type allele revealed that this mutation is crucial for the described phenotype. Strain ss251 and derivatives thereof produced succinate with high yields above 80% mol mol−1 from glycerol under nongrowth conditions.

Conclusions

The results show that during the aerobic growth of ss251, the formation of pyruvate proceeds via the proposed POMP pathway, starting with the carboxylation of PEP by Ppc. The resulting oxaloacetate is reduced by malate dehydrogenase (Mdh) to malate, which is then decarboxylated back to pyruvate by a malic enzyme (MaeA or MaeB). Mutation of ppc is crucial for fast growth of pykAF mutants on glycerol.

Significance and Impact of Study

An E. coli mutant that is capable of achieving high yields of succinate (a top valued-added chemical) from glycerol (an abundant carbon source) was constructed. The identified ppc mutation could be applied to other production strains that require strong PEP carboxylation fluxes.

Introduction

Succinate has become a valued building block for deriving commodity and speciality chemicals from biomass (Werpy and Petersen 2004). Traditionally produced from petrol at relatively high costs, in these times of dwindling fossil resources, industry is increasingly moving towards the bio-based production of the chemical. Low-cost succinic acid has major potential as a platform chemical from which numerous products can be created (Bozell and Petersen 2010; Chen 2010). So far, succinate has been produced using Anaerobiospirillum succiniciproducens (Lee et al. 2010), Actinobacillus succinogenes (Borges and Pereira 2010), Mannheimia succiniciproducens (Lee et al. 2008), Corynebacterium glutamicum (Inui et al. 2004) and Escherichia coli. Most publications on succinate production deal with E. coli as producer strain and glucose as substrate (Stols and Donnelly 1997; Chatterjee et al. 2001; Lin et al. 2005a,c; Sanchez et al. 2005; Andersson et al. 2007; Jantama et al. 2008). To increase the commercial attractiveness of bio-based succinate, manufacturing costs need to fall below the costs generated by the petrol-based succinate production. To minimize bioreactor investment and operation costs, anaerobic processes need to replace aerobic ones and inexpensive carbon sources need to be found. Glycerol is one such abundant carbon source, especially as it is a by-product of biodiesel production, which is expected to reach a worldwide volume of 41 billion litres by 2021 (Thompson and He 2006; OECD 2010). The engineering of strains that efficiently produce succinate on this substrate therefore has high economic potential. There are different routes for producing succinate from glycerol, aerobically via the glyoxylate cycle and anaerobically via phosphoenolpyruvate carboxykinase (Pck) or phosphoenolpyruvate carboxylase (Ppc; Table 1 and Fig. 1). A competing route is the production of pyruvate from PEP by the pyruvate kinases PykA and PykF. Escherichia coli bacteria in which the pyruvate kinase genes pykA and pykF are deleted should not produce pyruvate and pyruvate-derived products such as lactate or ethanol from glycerol. Such mutants were therefore expected to achieve maximal succinate production. In theory, one mol succinate per mol glycerol can be produced anaerobically, provided that one mol CO2 is also fixed (Dharmadi et al. 2006; Chen et al. 2010). This conversion is redox-neutral and thus possible under anaerobic conditions (Table 1). Ideally, a strain would produce biomass when grown aerobically on glycerol and after a short adaption phase to anaerobic conditions would switch to succinate production. Unfortunately, pyruvate kinase-deficient mutants do not grow on glycerol or, worse still, are not viable at all (Pertierra and Cooper 1977; Cunningham et al. 2009).

Table 1. Succinate yields, redox and energy balances resulting from the use of different metabolic routes starting from glycerol
Pathway of succinate formationYield (mol succinate mol glycerol−1)CO2 fixation (mol CO2 mol glycerol−1)Redox balanceATP gain (mol ATP mol glycerol−1)
SCElectron transport chaina
  1. SC, substrate chain phosphorylation.

  2. a

    involves ATP synthase activity. A consensus stoichiometry of 3 H+ translocated per ATP synthesized has been used for the calculations.

  3. b

    The deletion of sdhABCD is required or at least very beneficial for succinic acid production through this pathway under aerobic or microaerobic conditions (Lin et al. 2005b,c).

  4. c

    [H] = NADH or UQH2.

  5. d

    Glp = route through GlpACB–FrdABCD complex (Miki and Lin 1973, 1975).

Glyoxylate cyclebvia PDH complex0·5−1+3·5 [H]c, balancing by transfer to oxygen1·02·3–8
Carboxylation of PEP by Ppc1·0+1Closed under anaerobiosis00·1–0·67 (Glp)d
Carboxylation of PEP by Pck1·0+1Closed under anaerobiosis1·00·1–0·67 (Glp)d
Figure 1.

Glycerol metabolism of Escherichia coli. The figure was constructed with the assistance of the EcoCyc database (Keseler et al. 2011). Black arrows and black letters indicate the desired metabolic flow and intermediates from glycerol to succinate for maximum yield. Dotted arrows indicate the postulated POMP (PEP-oxaloacetate-malate-pyruvate) pathway. The following abbreviations were used: PEP, phosphoenolpyruvate; GldA, glycerol dehydrogenase; GlpK, glycerol kinase; MQ, menaquinone; DHA(P), dihydroxyacetone (phosphate); GlpACB, glycerol phosphate dehydrogenase (anaerobic); DhaKLM, dihydroxyacetone kinase complex; PykA(F), pyruvate kinase A(F); Pps, PEP synthetase; Pck, PEP carboxykinase; Ppc, PEP carboxylase; MaeA(B), malic enzyme A(B); Mdh, malate dehydrogenase; FumABC, fumarase A, B or C; FRDC, fumarate reductase complex.

A possible way to grow on glycerol would be what we called the POMP pathway (P from PEP, O from oxaloacetate, M from malate and P from pyruvate) for production of the essential metabolite pyruvate. PEP is the end product of glycolysis in the pykAF deficient strain as it cannot be dephosphorylated to pyruvate. Instead, it is carboxylated to oxaloacetate by the PEP carboxylase (Ppc); oxaloacetate is then reduced by malate dehydrogenase (Mdh) to malate, which in turn is decarboxylated to pyruvate by a malic enzyme (MaeA or MaeB).

Alternatively, the genes pps, gldA and mgsA may also be responsible for pyruvate formation on routes other than the POMP pathway. The gene pps encodes the enzyme phosphoenolpyruvate synthase that produces PEP from pyruvate in an ATP-dependent reaction (Imanaka et al. 2006). Pyruvate might be produced in the reverse reaction. The gene gldA encodes a glycerol dehydrogenase that produces dihydroxyacetone from glycerol. In combination with the enzyme dihydroxyacetone kinase (dhaKLM), dihydroxyacetone phosphate and pyruvate are produced in a PEP-dependent reaction (Bachler et al. 2005; Subedi et al. 2008). Finally, the mgsA gene product produces methylglyoxalate from dihydroxyacetone phosphate, which could be converted to pyruvate via lactate (Mazumdar et al. 2010).

The present study was therefore aimed at the identification of pyruvate kinase-deficient strains that would remain vital or even grow on glycerol as well as produce high quantities of succinate from glycerol.

Materials and methods

Strains, media and growth conditions

The strains and plasmids used are listed in Table 2 and Table S1. Tables S2, S3, S4 and S5 list the oligonucleotides used. Cultures were grown at 37°C in LB. Mutant selection involved the antibiotics ampicillin (100 μg ml−1), chloramphenicol (25 μg ml−1), kanamycin (50 μg ml−1 when resistance genes were on plasmids and 25 μg ml−1 for genes on chromosomes), apramycin (50 μg ml−1) and nourseothricin (10–40 μg ml−1, obtained from Jena Bioscience GmbH, Jena, Germany). M9 minimal medium (Sambrook et al. 1989) was used with 0·4% (w/v) either glucose or glycerol. Modified M9hP (M9 high or 100 mmol l−1 phosphate) was used for biotransformation: 500 ml 1·15x M9hP salts, 0·5 ml 0·1 M CaCl2, 0·5 ml 1 M MgSO4, 0·2 ml 10 mg ml−1 thiamine. The 1·15x M9hP salts stock solution consisted of 20·47 g Na2HPO4·2 H2O, 15·65 g KH2PO4, 1·15 g NaCl, 2·30 g NH4Cl and 2 l H2O.

Table 2. Escherichia coli strains and plasmids used
Strain or plasmidRelevant genes/propertiesaReference or construction
  1. a

    (AmpR), (AprR), (CmlR), (KanR) or (NouR) = resistance against ampicillin, apramycin, chloramphenicol, kanamycin or nourseothricin; aac(3)IV = aminoglycoside-3-N-acetyltransferase gene, bla = β-lactamase gene, cat = chloramphenicol acetyltransferase gene, nat1 = nourseothricin acetyltransferase gene.

  2. b

    IIG = Institute of Industrial Genetics, Stuttgart University;

  3. c

    P1 (A) × (B) = P1-transduction of genes from strain A to strain B;

  4. d

    (pJOE5555·1) Removal of the resistance marker by site-specific recombination between two mrpS sites using the plasmid-encoded recombinase MrpA.

Strains
BW25113lacIq rrnBT14 ΔlacZWJ16 hsdR514 ΔaraBADAH33 ΔrhaBADLD78Datsenko and Wanner (2000)
WA66xBW25113 Lac+ ΔxthA::FRTWegerer, A., IIGb
ss173WA66x ΔldhA::FRT-aac(3)IV-FRT ΔpoxB::FRT ΔpflB::FRTP1 (MW4) x (ss171)c
ss195WA66x ΔpykF::FRT ΔpykA::FRT-kan-FRT ΔldhA::FRT ΔpoxB::FRT ΔpflB::FRT ΔtdcE::FRT-aac(3)IV-FRTP1 (ss145) x (ss182)c
ss251ss195 (selected for faster growth on glycerol, transfer #16)Serial passages
ss271ss251 ΔgldA::mrpS-nat1-mrpSP1 (ss262) x (ss251)c
ss274ss251 Δpps::mrpS-nat1-mrpSP1 (ss266) x (ss251)c
ss279ss251 ΔgldA::mrpSss271 (pJOE5555·1)d
ss281ss251 ΔmgsA::mrpS-nat1-mrpS ΔgldA::mrpSP1 (ss231) x (ss279)c
ss310ss251 ΔmaeA::mrpS-nat1-mrpS ΔgldA::mrpSP1 (ss296) x (ss279)c
ss374ss251 Δppc::mrpS-nat1-mrpS ΔgldA::mrpSP1 (ss225) x (ss279)c
ss431ss251 ΔmaeB::mrpS-nat1-mrpS ΔmaeA::mrpS ΔgldA::mrpSP1 (ss415) x (ss430)
Plasmids
pCP20flp bla (AmpR) cat (CmlR) repA101Cherepanov and Wackernagel (1995)
pIJ773FRT-aac(3)IV-FRT (AprR) bla (AmpR)Gust et al. (2003)
pIJ790cat (CmlR) repA101 araC λ-RED (gam, bet, exo)Gust et al. (2003)
pJOE5555·1Prha-mrpA cat (CmlR) repA101Altenbuchner, J., IIGb
pJOE6038·1Additional recA on plasmid pIJ790Altenbuchner, J., IIGb
pWA38·4bla (AmpR) mrpS-kan-mrpS (KanR)Wegerer, A., IIGb
pSS55·11bla (AmpR) mrpS-nat1-mrpS (NouR)Exchange of resistance gene in pWA38·4
pSS57·2bla (AmpR) mrpS-aac(3)IV-mrpS (AprR)Exchange of resistance gene in pSS55·11

Construction of gene deletions with the λ-Red recombination system

Chromosomal gene deletions were constructed using the λ-Red recombination system (Baba et al. 2006) and the E. coli strain WA66x pJOE6038·1. WA66x is a Lac+ derivative of BW25113 obtained by P1 transduction (unpublished). The plasmid pJOE6038·1 is a derivative of the recombination plasmid pIJ790, which has the E. coli recA gene integrated in addition to the λ-Red exo, bet and gam genes. The resistance gene cassettes flanked by site-specific recombination sites for the recombinases, FLP and MrpA, were amplified from the following plasmids: pWA38·4 (mrpS-KmR-mrpS), pIJ773 (FRT-aac(3)IV-FRT), pSS55·11 (mrpS-nat1-mrpS) and pSS57·2 (mrpS-aac(3)IV-mrpS). The oligonucleotides used for gene deletion are summarized in Table S3. Electroporation of WA66x pJOE6038·1 was carried out as described by Gust et al. (2003). The chromosomal regions with the antibiotic resistance cassettes and corresponding gene deletions engineered in E. coli WA66x pJOE6083·1 were transduced using the bacteriophage P1kc as described by Lennox (1955). For verification, a bacterial colony picked from an LB agar plate was resuspended in 100 μl water and centrifuged, and the cells were resuspended in 100 μl water and stored at −20°C until further use. 5 μl of this suspension was amplified by PCR (50 μl reaction volume) according to the manufacturer's instructions using Taq DNA polymerase. Analysis of the deletions and gene replacements of pykA, pykF and tdcE by PCR in strains ss195, ss251, ss279 and, as a control, in ss173 is shown in Fig. S1.

Excision of antibiotic resistance gene cassettes

The plasmid pCP20, which contains a temperature-sensitive replicon that enables the thermal induction of FLP synthesis, was used to excise antibiotic resistance gene cassettes flanked by FRT sites as described by Cherepanov and Wackernagel (1995). The same procedure was applied to the plasmid pJOE5555·1 that carries the recombinase gene mrpA and mrpS sites flanking antibiotic resistance gene cassettes (Warth et al. 2011). The expression of mrpA was induced by supplementing the plates with 0·2% (w/v) l-rhamnose. More details about this and other methods used can be found at the study by Söllner (2012).

Determination of growth rate

An overnight culture in M9 medium was used to inoculate cultures in 100-ml Erlenmeyer flasks filled with 15 ml growth medium containing either glucose or glycerol as carbon sources. No antibiotics were added. The flasks were incubated at 37°C [Gio Gyrotory® Shaker; NBS (Edison, NJ, USA) at c. 200 rev min−1]; the initial cell density was between 0·01 and 0·1 OD600. The resulting parameter μ (h−1) indicates the maximum growth rate per hour to the base e.

Aerobic selection for faster growth on glycerol

As a selection of faster growth, the pykAF mutant ss195 was repeatedly grown in M9 medium with glycerol. Starting from a preculture in M9 with 0·4% glucose, 100 ml M9 medium with 0·4% glycerol in a 500-ml shake flask was inoculated at an OD600 of 0·01 (transfer #1) and incubated aerobically at 37°C. When an OD600 of about 1 was reached, the cells were transferred again to new medium. The volume of the inoculum for the next culture was adjusted in relation to the growth rate of the cultures so that subsequent cultures were able to reach a cell density of 1 OD600 within 24 h.

Whole-genome resequencing and data analysis

Strain E. coli ss251 was sequenced by GATC Biotech AG (Konstanz, Germany) on the Illumina/Solexa SBS platform. The individual reads (both forward and reverse directions) were 51 bp long; the sequence data were analysed with the Geneious Pro trial 5.6.5 (Biomatters Ltd., Newark, NJ, USA) software package. The reads were mapped to the reference genome of E. coli K12 MG1655 (accession numbers: NC_000913.2/GI:49175990), and then, a search for nucleotide exchanges and deletions was performed.

Small-scale biotransformation assay in 1·5-ml plastic cups

A 5-ml overnight culture in M9 with 0·4% glucose and the relevant antibiotic was used to inoculate a 15-ml preculture in a 100-ml flask containing M9 supplemented with 0·4% glycerol as carbon source (in case of strains ss310 and ss431, the preculture was additionally supplemented with 10 mmol l−1 pyruvate). The culture was incubated for 4 h at 37°C until a cell density of c. 0·6 OD600 was reached. The cells were pelleted and resuspended in 5 ml 1·15x M9hP without any carbon source and stored on ice. A biotransformation solution was prepared by mixing the cells with 1·15x M9hP solution, 150 μl of 20% (w/v) glycerol and 600 μl of 1 M NaHCO3 to a final volume of 6 ml (1x M9hP). The cell density of a standard biotransformation assay was 0·5 OD600. Aliquots of 1·2 ml were transferred into 1·79-ml plastic cups, and the lids were closed tightly (residual air space = 0·59 ml) to create anaerobic or microaerobic conditions. The plastic cups were incubated at 37°C (rotator, Neolab, c. 10 rev min−1, radius 10 cm) and only opened once to determine cell density and analyse the extracellular metabolites using HPLC. The cells were pelleted, and the supernatant was stored at −20°C for subsequent analysis.

Analytics

The concentrations of glycerol, lactate, acetate, formate, succinate and ethanol were determined using an Agilent 1200 series HPLC system equipped with a REZEX ROA column (300 × 7·8 mm, Phenomenex, Aschaffenburg, Germany) that enabled the detection of organic acids; 5 mmol l−1 H2SO4 was used as mobile phase. The samples were prepared as follows: 1 ml diluted sample was mixed with 45 μl 4 M NH3 (pH 10·2) and 100 μl 1·2 M MgSO4 and incubated for 5 min at room temperature to precipitate phosphate as magnesium ammonium phosphate hexahydrate. After 5 min centrifugation, 500 μl of the supernatant was mixed with 500 μl 0·1 M H2SO4. After a 15 min incubation period and 15 min centrifugation, the supernatant was used for HPLC analysis.

13C analysis

To measure carbon dioxide fixation, the above-described ‘small-scale’ (1·5 ml) biotransformations were performed using fully labelled NaH13CO3 (98 atom% 13C, Isotec™, Miamisburg, OH, USA) instead of its naturally labelled (12C) hydrogen carbonate isotope. The resulting metabolite labelling patterns were analysed using a TurboMass GC–MS system (PerkinElmer LAS, Rodgau, Germany) as described by Vielhauer et al. (2011). The resulting MS measurement data set had to be further revised to separate the naturally occurring 13C labels from the ones artificially accomplished by applying the software-based mass isotope correction method (MATLAB-based software tool), described by Wahl et al. (2004).

Results

Construction of pyruvate kinase-negative mutants and their growth in minimal medium supplemented with glycerol

To engineer E. coli for succinate production, the strain BW25113 was chosen because of the enhanced levels of recombination achieved using the λ-Red method. The strain ss173 was obtained from a Lac+ derivative of BW25113 by deleting pyruvate consuming genes encoding lactate dehydrogenase (ldhA), pyruvate oxidase (poxB) and pyruvate formate-lyase (pflB). Succinate production with this strain was very low (data not shown). Therefore, E. coli ss173 was used to introduce the deletions of pykA and pykF. In addition, the PFL-like keto acid formate-lyase gene (tdcE) was also deleted (see Fig. S1). The resulting strain E. coli ss195 (ss173 ΔpykA ΔpykF ΔtdcE) is therefore unable to express pyruvate kinases. Strain ss195 grew very slowly (μ = 0·09 ± 0·01 h−1) when glycerol was used as the sole source of carbon. The growth rate on glucose-containing medium (μ = 0·52 ± 0·01 h−1) was very similar to that of PB25 (μ = 0·51 h−1), a JM101 derivative that also lacks the genes pykA and pykF (Ponce 1999). Strain ss173, which is the predecessor of ss195 that contains both wild-type pyruvate kinase genes, grew at a rate of 0·33 ± 0·03 h−1 on glycerol-containing medium.

Selection for faster growth on glycerol through serial dilutions in liquid culture

As the strain ss195 grew ineffectively on glycerol, spontaneous mutants with higher growth rates were selected in a series of dilutions in glycerol-supplemented M9 medium. The relative growth rate increased steadily during the transfer experiments (data not shown) and was around three times higher at the end of the experiment. After 16 transfers, strain ss251 (= ss195*T16, T for transfer number), which grew relatively well on glucose and glycerol (Table 3), was chosen for further analysis. The requirement of pyruvate for growth on various PTS and non-PTS sugars of ss251 and predecessors ss173 and ss195 are shown in Table S6.

Table 3. Growth rates of pyruvate kinase-deficient strains before and after the selection for faster growth. Growth rates were measured in M9 minimal medium with glucose or glycerol as carbon source at 37°C
Strain (relevant genotype, short description)Growth rate μ (h−1)
Glucose 0·4 %Glycerol 0·4 %
  1. a

    Selected strain.

  2. b

    20 mmol l−1 succinate in addition to glucose restored growth.

ss195 (ΔpykAF, ancestor)0·52 ± 0·010·09 ± 0·01
ss251 (ss195a Transfer 16, selected mutant)0·56 ± 0·010·28 ± 0·02
ss271 (ss251 ΔgldA)0·57 ± 0·020·26 ± 0·01
ss274 (ss251 Δpps)0·54 ± 0·020·28 ± 0·02
ss279 (ss271 without resistance marker)0·49 ± 0·010·28 ± 0·01
ss281 (ss279 ΔmgsA)0·53 ± 0·020·27 ± 0·02
ss310 (ss279 ΔmaeA)0·50 ± 0·010·09 ± 0·01
ss431 (ss279 ΔmaeA ΔmaeB)0·37 ± 0·01No growth
ss374 (ss279 Δppc)No growthbNo growth

Characterization of the growth of mutant ss251 and analysis of flux through the newly defined POMP pathway

The usage of the POMP pathway by the mutant ss251 was assessed by deleting several genes in the ss251 strain, including those which were expected to play a key role in the POMP pathway (ppc, maeA, maeB) or any other route that leads to pyruvate (pps, gldA and mgsA). The growth rates of these mutants on glycerol are shown in Table 3. The cultivation of the strains on glucose served as a control for their ability to grow in minimal medium. The deletion of pps in ss251 resulted in strain ss274, and the deletion of gldA led to strain ss271. The nourseothricin resistance cassette used for inactivation of gldA was removed from ss271, and the resulting strain ss279 was used to produce the gldA mgsA double mutant ss281. Compared with ss251, no significant change in growth rate was observed in these three mutants. This was in contrast to the outcome of deleting ppc in ss279. The Δppc mutant ss374 did not grow at all on glycerol. The deletion of maeA in strain ss279 led to strain ss310, which displayed a growth rate of μ = 0·09 ± 0·01 h−1, which was similar to that of ss195 prior to the adaptive evolution process. The deletion of the malic enzyme genes maeA and maeB (ss431) led to a complete inability of the strains to grow on glycerol.

The growth rates of the mutants indicated that only the deletion of the genes whose products are involved in the POMP pathway led to a reduction in the strains' ability to grow on glycerol.

Whole-genome resequencing along with reverse genetic engineering identified a mutation of ppc that underlies the fast-growing phenotype of ss251

Assembly and alignment of the whole-genome sequence reads to the reference genome of E. coli K12 MG1655 verified the identity of ss251 as a BW25113 derivative (e.g. hsdR514 ΔaraBADAH33 ΔrhaBADLD78) and, even more importantly, confirmed all gene deletions that were engineered during the construction of ss251 (data not shown). 15 mutations that lead to amino acid changes were detected (Table S7). Most mutations affect proteins that are most likely unrelated to the general metabolism on M9 medium with glycerol (e.g. prophage genes, nitrate reductase NapA or the cobalamin transporter BtuB).

A transversion in the ppc gene that causes an amino acid exchange (A595S; codon GCG -> TCG) in the α-helix 26 of Ppc was detected. Because Ppc catalyses the first and committing step of the proposed POMP pathway, further tests were carried out to elucidate whether the mutation of ppc was the reason for the bacteria's faster growth on glycerol. Ppc was deleted in strain ss251 to make room for either the wild-type or mutant allele in the resulting strain ss374. The reintegration was performed by PCR products (primers s6325/s6326) that contained the full-length ppc ORF and some flanking sequences from strain ss195 (before undergoing serial passage on glycerol) or from strain ss251 (evolved mutant). The successful integration of functional Ppc was tested on M9 agar plates with glucose. The colonies of ss374 with the PCR product of wild-type ppc showed a growth rate of 0·06 ± 0·01 h−1 (average of three colonies tested in M9 glycerol), and colonies with ppc from ss251 showed growth rates of 0·27 ± 0·01 h−1.

Small-scale biotransformation assay

A small-scale assay was developed to monitor succinate production from glycerol by the pyruvate kinase-deficient mutants. Microcentrifuge tubes were filled with 1·2 ml mineral salt medium containing glycerol, NaHCO3 and the cells. The remaining air volume allowed the cells to adapt slowly to anaerobic conditions. For this adaption process, cell concentration was critical. A cell density of 0·5 OD600 (which represents a dry cell weight of 0·2 g l−1) was optimal for producing high quantities of succinate. Smaller cell numbers produced mainly biomass at the expense of succinate. Higher (10-fold) initial densities prevented the cells from adapting to anaerobic conditions and even led to cell lysis (data not shown).

The pyruvate kinase-deficient mutants were tested using this assay. For detailed investigations, strain ss251 was replaced by ss279 to avoid any glycerol being converted into pyruvate via glycerol dehydrogenase GldA. On the first day of the experiment, the density of ss279 cells increased from 0·5 to about 0·8 OD600, but did not increase further over the remaining course (8 days) of the experiment (Fig. 2). During the first 6 days of the 9-day experiment (d0–d6), ss279 consumed 57·8 ± 2·8 mmol l−1 glycerol, achieving succinate yields (mol mol−1) of 81·5 ± 2·0%. Higher yields of succinate were calculated for the period from day 2 to day 6, during which cell density was constant. Hereby, the succinate yield of ss279 increased to 89·5 ± 3·7% (Table 4).

Table 4. Succinate yields of biotransformations with the strains ss279, ss310 and ss431 using glycerol as substrate. Experiments were carried out with a cell density of 0·5 OD600 at 37°C. The minimal medium was M9hP with 60 mmol l−1 glycerol and 100 mmol l−1 NaH13CO3. Two independent biotransformations were performed with each strain. The average extracellular concentrations of glycerol or succinate are shown for different time intervals
StrainTime frame (days)Glycerol consumed (mmol l−1)Succinate formed (mmol l−1)Yield (mol succinate mol glycerol−1)
  1. a

    i.a. ΔpykA ΔpykF ΔgldA.

  2. b

    i.a. ΔpykA ΔpykF ΔgldA ΔmaeA.

  3. c

    i.a. ΔpykA ΔpykF ΔgldA ΔmaeA ΔmaeB.

  4. d

    complete consumption of the carbon source glycerol.

ss279ad0–d657·8 ± 2·847·1 ± 1·10·81 ± 0·02
ss310b57·3 ± 0·745·7 ± 0·70·80 ± 0·01 
ss431c46·4 ± 0·937·1 ± 0·30·80 ± 0·01 
ss279ad0–d963·7 ± 2·2d48·7 ± 0·60·77 ± 0·04
ss310b62·3 ± 0·1d50·4 ± 1·60·81 ± 0·03
ss431c59·4 ± 0·5d47·4 ± 0·30·80 ± 0·01
ss279ad1–d640·0 ± 0·635·8 ± 1·00·90 ± 0·04
ss310b41·7 ± 0·735·7 ± 0·20·86 ± 0·01
ss431c36·1 ± 0·630·8 ± 0·40·85 ± 0·03
Figure 2.

Biotransformations of glycerol to succinic acid using Escherichia coli strains ss279 and ss310, respectively. Time course of concentrations of glycerol (cglycerol; initial concentration of 60 mmol l−1) and succinic acid (csuccinate, with a final value of about 45 mmol l−1 after 9 days of incubation) and the measured optical density (OD600) of the clones are assayed in closed 1·5-ml plastic tubes. Depicted data are from two independent biotransformations (with standard errors below 10%). Cell density of 0·5 OD600 equals to 0·2 g cell dry weight l−1. Hence, starting with a specific production rate qp,max of 1·25 mmol succinate g−1 cell dry weight h−1, qp declined after 6 days of incubation to 0·9 mmol g−1 h−1 and further lapsed to near zero after at least 9 days of resting cell incubation. (image) ss279 and (image)ss310.

The influence of maeA and maeB on succinate production was analysed by comparing strain ss279, with its successor strains ss310 (ΔmaeA) and ss413 (ΔmaeA, ΔmaeB). The cell density of ss310 was similar to that of ss279 (Fig. 2). The strain also consumed about the same amount of glycerol (57·3 ± 0·7 mmol l−1) as ss279 within the first 6 days and had similar succinate yields (79·8 ± 0·3% mol mol−1; Table 4). Only a marginal increase in cell density was seen with ss431 (not shown).

Verification of CO2 fixation by determining 13C-enriched succinate

The question as to whether the bacteria produce oxaloacetate and hence succinate through the addition of hydrogen carbonate to phosphoenolpyruvate (PEP) was further analysed using stable isotope labelling experiments. Strains ss279 and ss310 were cultivated for 144 h in plastic cups with 13C-labelled hydrogen carbonate. In both strains, more than 90% of the succinate that had accumulated outside of the cell was labelled (mass shift to m+1 and m+2; Fig. 3). ss279 produced small quantities of pyruvate, alanine and valine, all of which had one 13C-labelled carbon (data not shown). The side products were not quantified in this experiment.

Figure 3.

Labelling pattern of extracellularly accumulated succinate formed during cultivation of ss279 and ss310 in 1·5-ml plastic tubes filled with M9hP 0·4% (w/v) glycerol and 100 mmol l−1 NaH13CO3. Succinate with the mass m0 consists of 12C atoms. The masses m+1 and m+2 indicate one and two 13C atoms in the succinate, respectively. (image) m+2; (image) m+1 and (image) m0.

Discussion

Anaerobic growth of E. coli on sugars normally favours the formation of metabolites that originate from pyruvate such as lactate, acetate and ethanol. So, the goal of this project was to use pyruvate kinase-deficient strains to avoid formation of pyruvate during succinate production from glycerol under anaerobic, nongrowth conditions. Strains devoid of pyruvate kinases are known to grow ineffectively on glycerol. Iterative transfer of ss195 on medium with glycerol as single carbon source allowed the isolation of the much faster growing mutant ss251.

We assume that the mutant strain ss251 uses the postulated POMP detour for pyruvate formation during aerobic growth. This assumption is based on the observation of detrimental effects arising from the deletion of the genes ΔmaeA and Δppc, whereas the deletion of other possible pyruvate-forming pathways (ΔgldA, Δpps, ΔmgsA) had no significant effect on growth. An exhaustive description of those alternative pyruvate-forming pathways was recently published (Feuer et al. 2012). The additional deletion of maeB in strain ss310 (ΔpykAF ΔmaeA) led to a strain that did not grow on glycerol at all, whereas its ability to grow on glucose was barely affected. Whole-genome resequencing of ss251 identified a mutation in the gene ppc amongst others. Ppc catalyses the carboxylation of PEP, which is the first and committing step of the proposed POMP pathway. Therefore, ss251 derivatives containing wild-type and mutant ppc alleles were constructed by reverse metabolic engineering. The resulting growth phenotypes were reassessed on glycerol-containing M9 medium. The growth rate of the strain with wild-type ppc was 0·06 ± 0·01 h−1, which is similar to that of strain ss195 (μ = 0·09 ± 0·01 h−1). Colonies with ss251 ppc showed the same growth rate as ss251 (μ = 0·27 ± 0·01 h−1), which was three times higher than that of ss195. This confirmed that Ppc (A595S) is responsible for the bacteria's ability to grow faster. The 3D structure of the tetrameric E. coli Ppc is known (Kai et al. 1999). The enzyme is inhibited by aspartate or malate or their analogues (Yoshinaga 1977), and allosteric activation by acetyl-CoA is crucial for its activity (Terada et al. 1991). In order for the POMP detour to be effective and provide all of the cell's pyruvate, we assume that the mutation of ppc causes a loss of inhibition by the citric acid cycle intermediates l-aspartate and l-malate or that the resulting protein variant is independent of acetyl-CoA activation. The replacement of Lys620 to serine has already shown that E. coli is no longer sensitive to the feedback inhibitors l-Asp and l-malate (Yano and Izui 1997). Overexpression of heterologous Ppc or increasing the acetyl-CoA pool has already been applied to raise the succinate yields of E. coli when grown on glucose (Lin et al. 2004, 2005d).

It was expected that strains channelling metabolic flux through the POMP detour might also produce succinate efficiently under microaerobic or anaerobic conditions.

Derivatives of ss251 were checked for their ability to produce succinate using small-scale biotransformation assays. The strains ss279 (=ss251 ΔgldA), ss310 (=ss279 ΔmaeA) and ss431 (=ss279 ΔmaeA ΔmaeB) efficiently converted glycerol to succinate with yields of about 85% mol mol−1 and concentrations of about 50 mmol l−1 during a zero-growth phase. This equates to 85% of the theoretical maximum. It was not possible to find the remaining 15% in the form of metabolites such as malate, fumarate, formate, pyruvate or ethanol. The HPLC detection limit of all these compounds was between 3 (fumarate) and 6 (ethanol) mmol l−1, which is why these compounds could reach substantial quantities without being detected. A somewhat higher product concentration of 120 mmol l−1 succinate was reported when using 221 mmol l−1 glycerol, but this corresponded to a yield of only 54% with an E. coli strain (ΔadhE Δpta ΔldhA Δppc ΔpoxB) that overexpressed the ppc from Lactococcus lactis (Blankschien et al. 2010). A succinate yield of 80 ± 1% from glycerol was achieved with E. coli CpflB ΔptsI) that had a promoter-up mutation in the pck gene (Zhang et al. 2010). However, the authors reported that only 128 mmol l−1 of the 543 mmol l−1 glycerol in the cultures was used, whereas in our biotransformation, glycerol was completely consumed. It seems that a 100% biotransformation of glycerol to succinate has not yet been achieved.

Because strains ss310 and ss279 produced more than 75% (mol mol−1) succinate (C4 molecule) from glycerol (C3 molecule) in a zero-growth phase (Table 4 and Fig. 2), it can be concluded that a net fixation of CO2 occurred. This and the high incorporation of 13C hydrogen carbonate into succinate are in line with the proposed pathway of succinate formation resulting from intermediate PEP carboxylation. One-atom labelling patterns in pyruvate, alanine and valine (data not shown) further support the hypothesis of the POMP pathway where 13C atoms originating from hydrogen carbonate eventually lead to the labelling of these intermediates. The twofold labelling of succinate is significantly higher than one would expect from naturally occurring 13C in glycerol and must be the result of repeated addition of labelled inline image to the C3 body of PEP or pyruvate via PEP carboxylase or malic enzyme, respectively. Notably, the backbone C-skeleton may circulate between oxaloacetate, malate and fumarate. Hereby, oxaloacetate or malate may be decarboxylated again at the unlabelled end to PEP or pyruvate followed by renewed synthesis of oxaloacetate or malate, which results in the double labelling as indicated.

The succinate production process with ΔpykAF strains will be optimized in future to make it economically feasible. This will involve scale-up and transfer of the biotransformation process from plastic cups to large fermenters allowing the increase in cell densities and succinate concentrations.

Acknowledgements

We would like to thank Dr. Oliver Vielhauer and Mira Lenfers-Lücker for assistance with HPLC and GC analytics. The authors further thank the Stiftung Baden-Württemberg for financial support.

Conflict of interest

The authors declare that there are no conflict of interests.

Ancillary