Fed-batch production of l -phenylalanine from glycerol and ammonia with recombinant Escherichia coli

Biochemical Engineering Journal 83 (2014) 62–69

Contents lists available at ScienceDirect

Biochemical Engineering Journal journal homepage: www.elsevier.com/locate/bej

Regular Article

Fed-batch production of l-phenylalanine from glycerol and ammonia with recombinant Escherichia coli Michael Weiner a , Christoph Albermann b , Katrin Gottlieb b , Georg A. Sprenger b , Dirk Weuster-Botz a,∗ a b

Lehrstuhl für Bioverfahrenstechnik, Technische Universität München, Garching, Germany Institut für Mikrobiologie, Universität Stuttgart, Stuttgart, Germany

a r t i c l e

i n f o

Article history: Received 21 August 2013 Received in revised form 30 October 2013 Accepted 1 December 2013 Available online 9 December 2013 Keywords: l-Phenylalanine Glycerol Escherichia coli Fed-batch Plasmid stability

a b s t r a c t Glycerol was used as carbon source for l-phenylalanine production with recombinant Escherichia coli. In contrast to glucose, no consumption of the precursor phosphoenolpyruvate (PEP) is necessary for glycerol uptake. Additional lactic acid feeding was necessary for growth because the genes encoding the PEP consuming pyruvate kinase isoenzymes have been deleted. Thus a fed-batch process was developed with feeding of lactic acid and glycerol for biomass formation followed by feeding of glycerol and ammonia for l-phenylalanine production. Unfortunately, plasmid instability was observed in the first process. Plasmid stability could be successfully assured by replacing an ampicillin resistance gene by a kanamycin resistance gene cassette. The resulting maximum l-phenylalanine concentration of 13.4 g L−1 was improved by −1 26% and biomass specific productivity (22 mgL-phe g−1 h ) was raised by 69%. The final l-phenylalanine CDW −1 concentration of 13.4 g L was thus improved by a factor of 2.4 compared to earlier reports. © 2013 Elsevier B.V. All rights reserved.

1. Introduction l-Phenylalanine is one of the commercially most important aromatic amino acids [1–4]. The main relevance for the industrial synthesis of l-phenylalanine is based on its role as building block for the artificial sweetener aspartame [2]. Furthermore, lphenylalanine is used as additive in the food and feed industry [2] as well as for pharmaceutically active compounds like HIV protease inhibitor, anti-inflammatory drugs [4], and catecholamines [5]. High l-phenylalanine concentrations of 38 g L−1 [6], 46 g L−1 [7] and 50 g L−1 [8] can be achieved with metabolically designed Escherichia coli (E. coli) strains with glucose as carbon source. A rational approach for metabolic engineering of E. coli for enhanced l-phenylalanine production was applied [6; reviewed in 2]. First, an E. coli strain was used (F4, a derivative of wild type strain W3110) which had been engineered to carry a precise chromosomal deletion of the gene cluster pheA-aroF-tyrA. Deletion of tyrA was necessary to avoid the formation of l-tyrosine. Then, the gene pheA* encoding a feedback resistant enzyme PheA (chorismatemutase/prephenatedehydratase) and the wild type genes, aroF (tyrosine-sensitive DAHP synthase), aroB (dehydroquinate synthase), and aroL (shikimate kinase) were combined to an artificial operon under the control of a lacI/Ptac expression system

∗ Corresponding author. Tel.: +49 089 28915712. E-mail address: [email protected] (D. Weuster-Botz). 1369-703X/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.bej.2013.12.001

(pF81) which was equipped with an ampicillin resistance gene (bla) (in the following specified as pF81amp ). The combination of these genes led to the minimization of intermediates of the aromatic biosynthesis pathway in E. coli (e.g. 3-deoxy-d-arabinoheptulosonate-7-phosphate, shikimate and 3-dehydroshikimate). l-Phenylalanine concentrations of up to 38 g L−1 were thus achieved in a fed-batch process using glucose as carbon source [6]. In recent years, alternative carbon sources like glycerol have been studied for microbial l-phenylalanine production [1,5,9–11]. Glycerol is an inevitable by-product of biodiesel production [12]. For each kilogram of biodiesel approximately 75–100 g of crude glycerol are produced [12,13]. Through the exponential growth of biodiesel production from vegetable oils and animal fats in the last decade [12,13] the market price of crude glycerol decreased considerably [12]. With a price of US $ 0.13–0.24 per kilogram [14], glycerol has become an attractive alternative carbon source compared to sugars like glucose. Glycerol as carbon source also provides application-oriented advantages. For glycerol the degree of reduction per carbon () is significantly higher (C3 H8 O3 :  = 4.67) than that of glucose (C6 H12 O6 :  = 4) [15]. Glycerol uptake in E. coli occurs either by passive diffusion or through protein-assisted facilitated diffusion [16,17]. Compared to the uptake of glucose by the phosphotransferase system (PTS), no phosphoenolpyruvate (PEP) is necessary for glycerol uptake. Therefore pure stoichiometric carbon balancing without consideration of energy metabolism reveals a carbon recovery of 90% for glycerol, whereas PTS using glucose conversion

M. Weiner et al. / Biochemical Engineering Journal 83 (2014) 62–69

can only achieve 45% [11]. For energetic balancing one negative aspect of glycerol metabolism is its first phosphorylation step to glycerol-3-phosphate and as a consequence the necessary ATP regeneration. This point is partly compensated by ubiquinol generation during DHAP synthesis. Comparable to ATP loss in glycerol metabolism, ATP loss is obtained on glucose during fructose1,6-bisphosphate synthesis, which is produced heading for PEP generation. As two molecules of PEP are necessary for the aromatic biosynthesis pathways, PEP-saving could be an interesting feature and glycerol could be an interesting carbon source for microbial PEP-consuming reactions [18]. Reports on microbial l-phenylalanine production from glycerol as carbon source have mainly highlighted single enzyme variations like alteration of phenylalanine dehydrogenase in combination with amino acid exporter and glycerol transport facilitator [10]. Thereby results focus on comparing different recombinant E. coli. The product concentrations are rather low if simple batch processes are applied: 0.4 g L−1 within 240 h [10] up to 5.6 g L−1 within 70 h [5]. This work deals with the design of a controlled fed-batch process for l-phenylalanine-production with glycerol as carbon source making use of genetically engineered E. coli [2]. An E. coli strain with enhanced expression of the shikimate pathway key enzymes and deletion of the pyruvate kinase genes was used to provide more phosphoenolpyruvate for the aromatic biosynthesis route. This attempt is similar to earlier approaches improving PEP availability if glucose is the carbon source for aromatics [18,21–23]. As glycerol catabolism via the lower glycolytic trunk is reduced by the loss of pyruvate kinase activity, these strains require a source of pyruvate to resume good growth in minimal media with glycerol (growth on glucose is less affected as the PTS provides pyruvate from PEP; [18,21–23]). 2. Material and methods 2.1. Microorganisms and plasmids The bacterial strains used in this study were E. coli DH5␣ and E. coli FUS4.11kan (E. coli W3110 pheA-tyrA-aroF, lacIZYA::Ptac aroFBL, pykA::FRT, pykF::FRT-Kan-FRT [11]). E. coli FUS4.11kan was constructed from strain E. coli F4 [6,19] by the introduction of an artificial chromosomal Ptac -aroF-aroB-aroL operon, resulting in strain E. coli FUS4 [11]. Chromosomal deletion of the genes encoding the two pyruvate kinase isoenzymes by a recombineering method [20], yielding in a strain with pykA gene deleted and pykF disrupted by a kanamycin resistance cassette (strain FUS4.11kan ) [11]. The kanamycin-resistance cassette of strain E. coli FUS4.11kan was eliminated using the plasmid pCP20 as previously described [24] to yield strain FUS4.11. The exchange of the ampicillin resistance gene (bla) of pF81amp [6] by a kanamycin resistance gene (kan) was conducted using the ␭-Red recombineering technique as described by Sharan et al. [25]. In brief, the kan gene was PCR amplified from plasmid pJF-crtZFRT-kan-FRT [26] using the following primers: 5 TTCTAAATACATTCAAATATGTATCCGCTCATGAGACAATAACCCTGATAGCTTGCAGTGGGCTTACATGG 3 and 5 TGAAGTTTTAAATCAATCTAAAGTATATATGAGTAAACTTGGTCTGACAGGAACTCCAGCATGAGATCCCCG 3 . The PCR product was electroporated into ␭-Red expressing E. coli DH5␣ cells carrying plasmid pF81amp . The transformants were selected on LB-kanamycin-agar plates. Plasmid DNA of these kanamycin-resistant clones was then isolated and retransformed

63

into strain DH5␣ with selection for kanamycin resistance. The resulting plasmid, pF81kan , was verified by restriction analysis and DNA sequencing. Maps of plasmids pF81amp and pF81kan are shown in the supplementary material (Fig. S1). Fig. 1 represents a schematic pathway illustration of metabolically engineered E. coli FUS4.11kan pF81amp and E. coli FUS4.11 pF81kan . To underline differing genetic features during growth and production phase, metabolic pathways are illustrated before (Fig. 1A) and after induction of protein expression (Fig. 1B). Transformation of E. coli strains by plasmids pF81amp or pF81kan was performed by heat-shock transformation of chemical competent cells. The resulting recombinant strains E. coli FUS4.11kan pF81amp and E. coli FUS4.11 pF81kan were stored on defined medium agar plates at 4 ◦ C before inoculation of the preculture. 2.2. Growth medium and preculture A defined minimal medium [27] was modified and used for all cultivations, which contained (g L−1 ): 3.00 KH2 PO4 , 12.00 K2 HPO4 , 5.00 (NH4 )2 SO4 , 0.10 NaCl, 0.30 MgSO4 ·7H2 O, 0.015 CaCl2 ·2H2 O, 0.1125 FeSO4 ·7H2 O, 1.50 sodium citrate, 0.0075 thiamin, 0.075 lphenylalanine, 0.075 l-tyrosine and antibiotics (0.10 ampicillin or 0.05 kanamycin). For precultures the pH was adjusted to pH 7.0 before sterilization of 1.04-times concentrated salt components (121 ◦ C, 20 min) and addition of filtrated (0.2 ␮m) stock solutions of the temperature sensitive components (MgSO4 ·7H2 O: 300.00 g L−1 , CaCl2 ·2H2 O: 15.00 g L−1 , FeSO4 ·7H2 O: 22.50 g L−1 , sodium citrate: 300.00 g L−1 , thiamin: 7.50 g L−1 , l-phenylalanine: 10.00 g L−1 , ltyrosine: 15.00 g L−1 (titrated with 10 M potassium hydroxide for solubility) and ampicillin or kanamycin: 50.00 g L−1 ). A trace-element solution adapted from literature [28] was used (1 mL L−1 ) in the fed-batch processes containing (g L−1 ): 11.20 MnSO4 ·H2 O, 10.00 AlCl3 ·6H2 O, 7.33 CoCl2 ·6H2 O, 2.00 ZnSO4 ·7H2 O, 2.00 Na2 MoO4 ·2H2 O, 1.00 CuCl2 ·2H2 O and 0.50 H3 BO3 . Precultures were obtained in two steps: First two single E. coli colonies were picked from the agar plate to inoculate 20 mL of defined minimal medium with 4.5 g L−1 glycerol and 2.5 g L−1 lactic acid in two 100 mL shake flask. After incubation in an orbital shaker for 24 h at 37 ◦ C and 100 rpm the whole culture was used to inoculate ten 500 mL shake flasks. Each shake-flask contains 100 mL of defined medium with 4.5 g L−1 glycerol and 2.5 g L−1 lactic acid. Flasks were incubated for 20 h at 37 ◦ C and 200 rpm. The complete preculture was used to inoculate the stirred tank reactor used for fed-batch fermentations. 2.3. Bioreactor and feeding solutions Fed-batch cultivations were performed in a 42 L stainless steel stirred tank reactor with four equidistant baffles and three six-bladed Rushton impellers (Techfors, Infors HT, Bottmingen, Switzerland). The vessel was filled with a concentrated solution of the salts of the medium before in situ sterilization for 20 min at 121 ◦ C. After cooling down (37 ◦ C), microfiltrated stock solutions of the temperature sensitive medium components and separately sterilized glycerol (1000 g L−1 ) and lactic acid (600 g L−1 ) were pumped aseptically into the reactor to achieve initial substrate concentrations of 2.4 g L−1 glycerol and 1.6 g L−1 lactic acid at an initial volume of 15 L. Oxygen supply was realized by air-gassing (up to 1.75 vvm). Dissolved oxygen concentration (probe: 322756800, Mettler Toledo GmbH, Giessen, Germany), was kept above 40% air saturation by controlling stirrer speed and gassing rate. The pH (probe: 405-DPAS-SC-K8S/120, Mettler Toledo GmbH, Giessen, Germany) was controlled to pH 7.0 by addition of either H3 PO4 (42%) or NH4 OH (25%). Four different feed solutions were used successively for fedbatch operation. Feed solutions contained carbon sources (glycerol

64

M. Weiner et al. / Biochemical Engineering Journal 83 (2014) 62–69

Fig. 1. Schematic pathway illustration of metabolic engineered Escherichia coli FUS4.11kan pF81amp and Escherichia coli FUS4.11 pF81kan before (A) and after (B) addition of IPTG. Black lines indicate single enzymatic reaction with encoding genes (gray boxes and lines). Black dashed lines indicate multiple enzymatic reactions. Dotted black lines indicate feedback regulations in the aromatic amino acid pathway. Black crosses describe knock-outs of enzymatic reactions or feedback inhibition. Fat lines indicate overexpression of enzymatic reaction. Abbreviations: Glucose-6-phosphate (G6P), fructose-6-phosphate (F6P), dihydroxyacetone phosphate (DHAP), glyceraldehyde-3-phosphate (GAP), phosphoenolpyruvate (PEP), pyruvate (PYR), tricarboxylic acid cycle (TCA), erythrose-4-phosphate (E4P), 3-deoxy-d-arabino-heptulosonate-7-phosphate (DAHP), 3-dehydroquinate (3DHQ), 3-dehydroshikimate (3DHS), shikimate (SHK), shikimate-3-phosphate (SHK3P), 5-enolpyruvolylshikimate-3-phosphate (5EPS3P), chorismate (CHOR), prephenate (PRE), phenylpyruvate (PHPYR), 4-hydroxyphenylpyruvate (HPP), l-phenylalanine (L-Phe), L-tyrosine (l-Tyr), l-tryptophan (L-Trp). The descriptions pheAFBR encodes a feedback resistant version of PheA.

and/or lactic acid), nitrogen-sources and antibiotics (ampicillin or kanamycin). Amino acids for the complementation of auxotrophies were added to feed solutions only during growth phase. Fed batch media composition is shown in Table 1. An in-situ sampling device built in-house was used for sampling [29]. Reaction temperature was kept constant at 37 ◦ C during the whole fed-batch process. CO2 production was monitored in the offgas (Easy Line, ABB Automation, Zurich, Switzerland).

was determined in parallelized preliminary experiments (data not shown). By variation of the initial carbon source ratio optimal growth conditions were determined for this carbon source ratio. After the initial batch-phase of about 8 h a controlled exponential substrate limited supply of fed-batch media 1 (see Table 1) was started automatically triggered by the increase of dissolved-oxygen concentration. Media supply was adjusted to a fixed feeding profile

2.4. Process design The l-phenylalanine fed-batch production process was divided into two process phases: First biomass formation and second lphenylalanine production. A schematic overview on carbon source supply during the two process phases is shown in Fig. 2. Within the first fed-batch phase biomass formation was realized by feeding both carbon sources (glycerol and lactic acid) as well as the supplementing amino acids l-phenylalanine and l-tyrosine. Feeding of lactic acid as source of pyruvate was necessary to allow biomass formation due to the lack of both pyruvate kinase genes (pykA, pykF) [11]. An initial concentration of 2.4 g L−1 glycerol and 1.6 g L−1 lactic acid was used in the batch-phase. The carbon source ratio

Fig. 2. Schematic illustration of carbon supply for substrate limited fed-batch production of l-phenylalanine with recombinant E. coli in stirred-tank bioreactors on a 15 L-scale using the two carbon sources glycerol and lactic acid.

M. Weiner et al. / Biochemical Engineering Journal 83 (2014) 62–69

65

Table 1 Composition of media applied for fed-batch production of l-phenylalanine with recombinant E. coli. Component

Fed-batch media 1, g L−1

Fed-batch media 2a, g L−1

Fed-batch media 2b, g L−1

Fed-batch media 3, g L−1

Glycerol Lactic acid l-Phenylalanine l-Tyrosine (NH4 )2 SO4 (NH4 )2 HPO4 Ampicillin or kanamycin

100.00 60.00 2.50 3.60 80.00 – 0.40 or 0.10

– 200.00 1.00 4.20 40.00 – 0.27 or 0.10

400.00 – 0.50 3.75 – – 0.40 or 0.10

800.00 – – – 37.50 37.50 0.40 or 0.10

FFeed (t) (see Eq. (1)) adapted from Jenzsch et al. [30] with a desired growth rate Set of 0.1 h−1 , an initial volume V0 of 15 L, a yield coefficient YX/S of 0.4 g g−1 estimated for lactic acid as growth limiting carbon source and an initial biomass concentration cX0 as well as the substrate concentration in fed-batch media cS . For this purpose lactic acid was assumed to be the growth limiting substrate during cultivations with pyruvate kinase negative strains [30]. FFeed (t) =

Set · cX0 · V0 exp(Set ·t) YX/S · cS

(1)

Both carbon sources were added simultaneously in the first fedbatch phase. After about 13 h of feeding, the two carbon sources glycerol and lactic acid were supplied with two separate fed-batch media (2a and 2b). A controlled exponential lactic acid supply (fedbatch media 2a) was applied to circumvent growth inhibition by lactic acid, which was shown for low concentrations of 2 g L−1 [31]. Glycerol supply (fed-batch media 2b) was kept constant. A continuous expression of enzymes from glycerol metabolism is assured by keeping a constant glycerol supply. This should reduce adaption time in the following production phase on glycerol. During biomass formation the concentration of l-tyrosine and l-phenylalanine was kept above 10 mg L−1 to avoid growth limitation. This was assured by supplementing both amino acids to the fed-batch media (Table 1). This approach resulted in low concentrations of these components (l-phenylalanine < 2 mM and l-tyrosine < 3 mM) in the reactor at the beginning of the production phase. This was important, because l-tyrosine feedback inhibits the aroF encoded isoenzyme of the first enzymatic step (3-deoxyd-arabino-heptulosonate-7-phosphate synthase) of the aromatic amino acid synthesis [2,4,19,32]. After a process time of about 42 h, biomass formation was expected to be completed. Expression of enzymes for l-phenylalanine production was induced by adding isopropyl ␤d-1-thiogalactopyranoside (IPTG, 0.3 mM final concentration) to the reactor. Afterwards, glycerol was fed as sole carbon source together with two nitrogen sources ((NH4 )2 SO4 and (NH4 )2 HPO4 ) for l-phenylalanine production (see Table 1 fed-batch media 3). The h−1 and the volumetric feed rate was adjusted to 0.18 gglycerol g−1 CDW flux was kept constant. This feed rate was chosen, because previous studies (data not shown) resulted in the highest l-phenylalanine yield. 2.5. Analytical methods The concentrations of glycerol, acetic acid and lactic acid in the culture supernatant were determined by using high-performance liquid chromatography (Agilent 1100 Series, Agilent, Santa Clara, USA) equipped with a RI detector (Agilent 1200 Series, Agilent, Santa Clara, USA). For analytical metabolite separation an Aminex HPX-87H column (BioRad, München, Germany) was used, working at 60 ◦ C (Oven: Mistral, Spark Holland, AJ Emmen, The Netherlands) and a constant flow rate (0.7 mL min−1 ) of sulfuric acid (5 mM). The concentrations of l-phenylalanine and l-tyrosine in the culture supernatant were determined by using high-performance liquid chromatography (Smartline HPLC, Knauer, Berlin, Germany)

with a fluorescence detector (RF20A, Shimadzu, Kyoto, Japan). The applied derivatisation and chromatographic method was adapted from HPLC application 15992 (Phenomenex, Torrance, USA): An automated sample-derivatisation with o-phthaldialdehyde (OPA), mercaptopropionic acid (MCA) and iodo acetic acid (IAA) was performed for labeling of the amino acids. A 40 mM bicine buffer (pH 10.2) was used for dilution and stabilization of derivatisation agents. 10 ␮L of sample were mixed with 658 ␮L of a MCA solution (0.3 mM in bicine buffer) for derivatisation. Then 20 ␮L of a 3.5 mM IAA solution (in bicine buffer) was added. After mixing 70 ␮L of a 11 mM OPA (in 92.8% bicine:7.1% methanol:0.1% MCA (v/v)) solution was added. The resulting solution was mixed and 20 ␮L were injected onto the HPLC-column. Chromatographic analysis was performed using an analytical scale (150 mm × 4.6 mm, 5 ␮m) Gemini C18 column (Phenomenex, Torrance, USA) operating at 40 ◦ C. The total flow rate was kept constant at 1.0 mL min−1 using a gradient of two mobile phases: eluent A 40 mM sodium dihydrogen phosphate (pH 7.75) and eluent B (45% methanol:45% acetonitrile:10% water (v/v)). Starting condition for the applied gradient was 100% solvent A for 3 min. From minutes 3.0–8.5, a constant decrease to 75% solvent A and 25% solvent B was applied. From minutes 8.5–28.5, the ratio decreased linear to 60% solvent A and 40% solvent B. After a constant phase of 2 min, gradient was switched to 100% B for 2 min. From minutes 32.0–34.0 a constant gradient of 20% solvent A and 80% solvent B was applied. A linear increase to 100% solvent A was performed for 4 min. Conditions were kept constant till end of analysis (43 min). Detection by fluorescence was performed using the wavelengths 340 nm (excitation) and 450 nm (emission). Ammonium concentration in the supernatant was measured using an enzymatic assay (Kit No.: 11 112 732 035, Boehringer Mannheim/R-Biopharm, Darmstadt, Germany). Prior to all analytics samples of culture supernatant were microfiltrated (pore size: 0.2 ␮m) and stored cold (4 ◦ C) after sampling from the fed-batch process. Analytical errors of glycerol, lactic acid, acetic acid, amino acids and ammonium were estimated by sample duplets. 3. Results and discussion To allow comparability of process data, a clearly structured and rational designed multi-phase process was designed (see Section 2). Within this standardized process a direct evaluation of recombinant strains for l-phenylalanine production was possible. The process performance of E. coli FUS4.11kan pF81amp is shown in Fig. 3. After the initial batch phase the substrate-limited fedbatch-process was started automatically at a process time of 7 h. A constant growth rate of approximately 0.1 h−1 was achieved due to the preset exponential feeding. A biomass-concentration of 20.4 g L−1 was measured at a process time of 42.3 h (see Fig. 3A). Supply of the auxotrophic amino acids was sufficient. 0.41 g L−1 l-tyrosine (Fig. 3C) and 0.24 g L−1 l-phenylalanine (Fig. 3A) were measured after a process time of 42.3 h. Protein expression for l-phenylalanine production was induced with 0.3 mM IPTG at a process-time of 42.3 h. l-Phenylalanine was produced with glycerol as the sole carbon source. Biomass

66

M. Weiner et al. / Biochemical Engineering Journal 83 (2014) 62–69

Fig. 3. l-Phenylalanine production process with Escherichia coli FUS4.11kan pF81amp in a stirred-tank reactor on a 15 L-scale with glycerol and ammonia. First vertical dashed line (7.0 h) indicates the end of the batch phase. Second vertical dashed line (42.3 h) indicates time of the induction of protein expression. (A) Concentrations of cell dry weight (CDW) and l-phenylalanine (L-phe), (B) concentrations of glycerol and ammonia (NH4 + ), (C) concentrations of lactic acid, l-tyrosine (L-tyr) and acetic acid, and (D) ratio of colony forming units on LB agar plates without (CFULB ) and with ampicillin (CFUamp+ ).

concentration further increased from 20.4 g L−1 to 33.3 g L−1 parallel to l-phenylalanine production. The continuing growth of the pyruvate kinase negative strain on glycerol was unexpected. Additionally performed batch experiments (data not shown) with E. coli FUS4.11kan pF81amp and glycerol as carbon source confirmed the ability to grow on glycerol. The maximum of the biomass specific l-phenylalanine production rate was estimated to be 13 mgL-phe g−1 h−1 . The maximum product specific yield CDW

was 0.12 gL-phe g−1 . After 70.9 h, a maximal l-phenylalanine glycerol

concentration of 10.6 g L−1 was measured. Initiation of acetic acid formation was observed at the same process time. Acetic acid accumulated up to 5.1 g L−1 until the end of cultivation (Fig. 3C). Due to the simultaneous start of acetic acid formation and the end of l-phenylalanine production inhibitory effects of high acetic acid concentration on l-phenylalanine production can be negated. Furthermore formation of lactic acid was detected at 68.5 h. Lactic acid accumulated to a maximum of 0.5 g L−1 after 75.7 h. The obtained by product formation of acetic acid and lactic acid supports the presented ability of E. coli FUS4.11kan pF81amp to grow on glycerol. These results indicate an alternative active pathway for pyruvate synthesis. Feasible pathways may be a combination of PEP carboxylase and malic enzymes, methylglyoxal pathway or isochorismate degradation. In the last few hours of the process glycerol accumulation was observed as well up to a concentration of 23.4 g L−1 after 86.2 h (Fig. 3B). A complete inhibition of carbon flux into the aromatic amino acid pathway by feedback regulation of l-tyrosine can be excluded at the process time l-phenylalanine production stopped (70.9 h). l-Tyrosine concentration decreased during the whole production phase. At a process time of 70.9 h l-tyrosine concentration was below 20 mg L−1 . A limiting ammonia supply during production phase can be ruled out. Fig. 3B demonstrates a slowly increasing ammonia concentration in the medium up to 6.9 g L−1 (86.2 h). In preliminary batch cultivations (data not shown) inhibitory effects of varying

ammonium (1.4–8.2 g L−1 ) concentrations were investigated and no influence on the l-phenylalanine production rate was obtained. Therefore the increased final concentration of ammonium in the 15 L-scale process was also excluded for production break-off reasoning. The relative number of plasmid carrying cells was measured as function of process time to investigate plasmid stability. The number of colony forming units (CFU) was measured by plating out cells (fourfold, after appropriate dilution) on LB plates with 100 mg L−1 ampicillin (CFUamp+ ) or without ampicillin (CFU) after a process time of 20.3 h, 42.3 h, 60.7 h, and 86.2 h, respectively. The ratio between CFU and CFUamp+ is shown in Fig. 3D. It is evident that the percentage of plasmid carrying cells decreased during the whole fed-batch process although ampicillin was used in batch and fed-batch-media as selection marker. Plasmid instability in E. coli may be caused by degradation of ampicillin through the action of ␤-lactamase. It has been reported that extracellular activity of ␤-lactamase can increase up to 90% within 24 h [33], followed by a rapid inactivation of ampicillin. Friehs [34] described that extracellular ␤-lactamase activity of a 500 ␮L culture-broth is sufficient to completely inactivate ampicillin of a 40 L cultivation within 1 min. This allows growth of plasmid-free cells as they are no longer inhibited by ampicillin. As plasmid-bearing cells then have a growth disadvantage, plasmidfree cells could take over. The plasmid loss (shown in Fig. 3D) clearly argues against the suitability of ampicillin as selection-marker for longtime cultivations with E. coli. To reduce plasmid loss during cultivation, we wanted to exchange the antibiotic selection marker from ampicillin to kanamycin. However, strain FUS4.11kan carried a chromosomal kanamycin resistance cassette. Therefore, this cassette was first removed (see Section 2.1) to yield strain FUS4.11. Then, a kanamycin resistance cassette was recombineered onto plasmid pF81amp to yield pF81kan . A similar fed-batch process on a 15 L-scale was performed (Fig. 4) to compare the strain E. coli FUS4.11 pF81kan to the

M. Weiner et al. / Biochemical Engineering Journal 83 (2014) 62–69

67

Fig. 4. l-Phenylalanine production process with E. coli FUS4.11 pF81kan in a stirred-tank reactor on a 15 L-scale with glycerol and ammonia. First vertical dashed line (8.9 h) indicates the end of the batch phase. Second vertical dashed line (43.5 h) indicates time of the induction of protein expression. (A) Concentrations of cell dry weight (CDW) and l-phenylalanine (L-phe), (B) concentrations of glycerol and ammonia (NH4 + ), (C) concentrations of lactic acid, l-tyrosine (L-tyr) and acetic acid, and (D) ratio of colony forming units on LB agar plates without (CFULB ) and with kanamycin (CFUkan+ ).

previously described results. After the initial batch phase the substrate-limited fed-batch-process was initiated at a process time of 8.9 h achieving a constant growth rate of approximately 0.1 h−1 as before. A biomass concentration of 22.1 g L−1 was achieved after at a process time of 43.5 h (Fig. 4A). l-Phenylalanine and l-tyrosine supply was sufficient with final concentrations of 0.46 g L−1 ltyrosine (Fig. 4C) and 0.21 g L−1 l-phenylalanine at 43.5 h (Fig. 4A). After induction of protein expression with 0.3 mM IPTG at a process-time of 43.5 h l-phenylalanine production was initiated with glycerol as sole carbon source, which was added with a supply rate of 0.18 gglycerol g−1 h−1 . Compared to the process perCDW formance of E. coli FUS4.11kan pF81amp a significantly reduced biomass formation was observed in the l-phenylalanine production phase, as biomass concentration increased only from 22.1 g L−1 to 27.0 g L−1 after a process time of 68.5 h. E. coli FUS4.11 pF81kan showed an increased maximal biomass specific production rate (22 mgL-phe g−1 h−1 ) and an increased substrate specific prodCDW uct yield (0.15 gL-phe g−1 ) compared to the ampicillin resistant glycerol

strain. A maximum l-phenylalanine concentration of 13.4 g L−1 was measured after a process time of 73.9 h (30.4 h after induction, see Fig. 4A). A notable decrease in l-phenylalanine production rate was observed after 65.9 h (22.4 h after induction). Accumulation of acetic acid and lactic acid started at the same process time. Also during this cultivation no increased concentration of acetic acid (0.2 g L−1 ) was detected at the time of the l-phenylalanine production break-off. Final concentrations of 6.6 g L−1 acetic acid and 0.8 g L−1 lactic acid were measured at the end of the process (87.8 h) (Fig. 4C). First, glycerol accumulation was observed after 73.9 h and resulted in a final concentration of 23.1 g L−1 . Again, ammonia supply was sufficient at any time of the fed-batch process resulting in a final concentration of 8.1 g L−1 at 87.8 h (Fig. 4B). Plasmid stability was investigated as well throughout the fedbatch process with E. coli FUS4.11 pF81kan . The number of colony forming units (CFU) was determined with (CFUkan+ ) and without

kanamycin (50 mg L−1 ) after a process time of 20.9 h, 43.5 h, 68.5 h, and 87.8 h, respectively. The ratio between CFU and CFUkan+ is shown in Fig. 4D. The CFU-ratio remains constant close to 1.0 indicating no plasmid loss throughout the whole fed-batch process. The enhanced plasmid stability of E. coli FUS4.11 pF81kan compared to the ampicillin-resistant E. coli FUS4.11kan pF81amp clearly improved the fed-batch production of l-phenylalanine with glycerol as carbon source. An improvement of the maximum l-phenylalanine concentration (13.4 g L−1 ) by 26% was observed and biomass specific productivity was raised by 69% to h−1 . The product yield on glycerol was increased 22 mgL-phe g−1 CDW

by 25% to 0.15 gL-phe g−1 as well. The increase in product glycerol yield correlates well with a reduced biomass formation in the lphenylalanine production phase. The measured product yield of 0.15 gL-phe g−1 represents 27% of the theoretical maximum [11]. glycerol

The theoretical maximal yield of 0.55 gL-phe g−1 was estimated glycerol for non-growing cells neglecting energy metabolism. Comparing the calculated yield coefficient to estimated yield coefficients of other groups (< 0.05 gL-phe g−1 ) [10], nevertheless demonglycerol strates an improved glycerol conversion to l-phenylalanine. For further analysis of the standardized process performance and to point out possible reasons for the reduced product yield, carbon balances were calculated for both production processes (Fig. 5A and B). Complete analytical process coverage is shown by high recovery rates of both carbon balances (E. coli FUS4.11kan pF81amp : >93%; E. coli FUS4.11 pF81kan : >97%). Comparing carbon balance of E. coli FUS4.11kan pF81amp to E. coli FUS4.11 pF81kan demonstrates a reduced carbon flux into biomass formation during l-phenylalanine production with E. coli FUS4.11 pF81kan . During this cultivation more glycerol was provided for l-phenylalanine production. Both carbon balances further indicate a high glycerol conversion in carbon dioxide by respiration. These high respiratory rates may indicate a high demand of maintenance and energy

68

M. Weiner et al. / Biochemical Engineering Journal 83 (2014) 62–69

Fig. 5. Carbon balances for the l-phenylalanine fed-batch production process with E. coli FUS4.11kan pF81amp (A) and E. coli FUS4.11 pF81kan (B). Balances are standardized to 100% at the time of IPTG addition. Substances considered: Biomass (CDW), CO2 , glycerol, l-phenylalanine (L-phe) and other (lactic acid, l-tyrosine, carbon eliminated by sampling).

metabolism and therefore can explain the variation between the experimentally determined yield coefficient and the theoretically estimated. The reasons for the relatively fast break down of l-phenylalanine production a few hours after induction of l-phenylalanine production, followed by accumulation of acetic acid, lactic acid and glycerol in the reactor are presently under study. Proteome analyses, metabolome analyses and flux analyses combined with rapid media transition studies [35] will be applied at varying process times in the l-phenylalanine production phase. 4. Conclusions A substrate-limited fed-batch process was successfully established with recombinant E. coli in a stirred-tank reactor on a 15 L-scale for l-phenylalanine production with glycerol as carbon source and ammonia as nitrogen source. Assuring plasmid stability by replacing the selectivity marker ampicillin with kanamycin was necessary to achieve l-phenylalanine concentrations of up to 13.4 g L−1 in the fed-batch process. This is an improvement by a factor of 2.4 compared to the so far published maximum lphenylalanine concentration of 5.6 g L−1 [5] achieved with glycerol as carbon source. Acknowledgements This work was funded by the German Research Foundation (Grant Nos. WE 2715/11-1 and SP 503/7-1, DFG, Germany). Initial strain construction of Katrin Gottlieb was supported by a grant of DBU (German Environment Foundation AZ 20006/881). The mentoring of Michael Weiner by the TUM Graduate School is gratefully acknowledged. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.bej.2013.12.001. References ˜ [1] K. Martínez-Gómez, N. Flores, H.M. Castaneda, G. Martínez-Batallar, O.T. Ramírez, G. Gosset, S. Encarnación, F. Bolivar, New insights into Escherichia coli metabolism: carbon scavenging, acetate metabolism and carbon recycling responses during growth on glycerol, Microb. Cell Fact. 11 (2012) 46. [2] G.A. Sprenger, From scratch to value: engineering Escherichia coli wild type cells to the production of l-phenylalanine and other fine chemicals derived from chorismate, Appl. Microbiol. Biotechnol. 75 (2007) 739–749. [3] M. Ikeda, Towards bacterial strains overproducing l-tryptophan and other aromatics by metabolic engineering, Appl. Microbiol. Biotechnol. 69 (2006) 615–626.

[4] J. Bongaerts, M. Krämer, U. Müller, L. Raeven, M. Wubbolts, Metabolic engineering for microbial production of aromatic amino acids and derived compounds, Metab. Eng. 3 (4) (2001) 289–300. [5] M. Khamduang, K. Packdibamrung, J. Chutmanop, Y. Chisti, P. Srinophakun, Production of l-phenylalanine from glycerol by a recombinant Escherichia coli, J. Ind. Microbiol. Biotechnol. 36 (10) (2009) 1267–1274. [6] N. Rüffer, U. Heidersdorf, I. Kretzers, G.A. Sprenger, L. Raeven, R. Takors, Fully integrated l-phenylalanine separation and concentration using reactiveextraction with liquid–liquid centrifuges in a fed-batch process with E. coli, Bioprocess Biosyst. Eng. 26 (2004) 239–248. [7] K.B. Konstantinov, N. Nishio, T. Seki, T. Yoshida, Physiologically motivated strategies for control of the fed-batch cultivation of recombinant Escherichia coli for phenylalanine production, J. Ferment. Bioeng. 71 (1991) 350–355. [8] K. Backman, M.J. O’Connor, A. Maruya, E. Rudd, D. McKay, R. Balakrishnan, M. Radjai, V. DiPasquantonio, D. Shoda, R. Hatch, K. Venkatasubramanian, Genetic engineering of metabolic pathways applied to the production of phenylalanine, Ann. N. Y. Acad. Sci. 589 (1990) 16–24. [9] K. Martínez, R. de Anda, G. Hernández, A. Escalante, G. Gosset, O.T. Ramírez, F.G. Bolívar, Coutilization of glucose and glycerol enhances the production of aromatic compounds in an Escherichia coli strain lacking the phosphoenolpyruvate: carbohydrate phosphotransferase system, Microb. Cell Fact. 7 (2008) 1. [10] M. Thongchuang, Y. Chisti, P. Pongsawasdi, K. Packdibamrung, Design of a recombinant Escherichia coli for producing l-phenylalanine from glycerol, World J. Microbiol. Biotechnol. 28 (2012) 2937–2943. [11] K. Gottlieb, Nutzung von Glycerin und Rohglycerin als C-Quelle fürdie Produktion von L-Phenylalanin mit rekombinanten Escherichia coli-Stämmen (Ph.D. dissertation), University of Stuttgart, 2011. [12] S.S. Yazdani, R. Gonzalez, Anaerobic fermentation of glycerol: a path to economic viability for the biofuels industry, Curr. Opin. Biotechnol. 18 (2007) 213–219. [13] J.C. Thompson, B.B. He, Characterization of crude glycerol from biodiesel production from multiple feedstocks, Appl. Eng. Agric. 22 (2) (2006) 261–265. [14] R.W. Nicol, K. Marchand, W.D. Lubitz, Bioconversion of crude glycerol by fungi, Appl. Microbiol. Biotechnol. 93 (2012) 1865–1875. [15] S.S. Yazdani, R. Gonzalez, Engineering Escherichia coli for the efficient conversion of glycerol to ethanol and co-products, Metab. Eng. 10 (6) (2008) 340–351. [16] D. Lu, P. Grayson, K. Schulten, Glycerol conductance and physical asymmetry of the Escherichia coli glycerol facilitator GlpF, Biophys. J. 85 (5) (2003) 2977–2987. [17] K.B. Heller, E.C.C. Lin, T.H. Wilson, Substrate specificity and transport properties of the glycerol facilitator of Escherichia coli, J. Bacteriol. 144 (1980) 274–278. [18] E. Meza, J. Becker, F. Bolivar, G. Gosset, C. Wittmann, Consequences of phosphoenolpyruvate:sugarphosphotranferase system and pyruvate kinase isozymes inactivation in central carbon metabolism flux distribution in Escherichia coli, Microb. Cell Fact. 11 (2012) 127. [19] M. Gerigk, R. Bujnicki, E. Ganpo-Nkwenkwa, J. Bongaerts, G. Sprenger, R. Takors, Process control for enhanced l-phenylalanine production using different recombinant Escherichia coli strains, Biotechnol. Bioeng. 80 (7) (2002) 746–754. [20] K.A. Datsenko, B.L. Wanner, One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products, Proc. Natl. Acad. Sci. U. S. A. 97 (2000) 6640–6645. [21] E. Ponce, N. Flores, A. Martinez, F. Valle, F. Bolívar, Cloning of the two pyruvate kinase isoenzyme structural genes from Escherichia coli: the relative roles of these enzymes in pyruvate biosynthesis, J. Bacteriol. 177 (19) (1995) 5719–5722. [22] E. Ponce, A. Martinez, F. Bolivar, F. Valle, Stimulation of glucose catabolism through the pentose pathway by the absence of the two pyruvate kinase isoenzymes in Escherichia coli, Biotechnol. Bioeng. 58 (2–3) (1998) 292–295. [23] G. Sprenger, et al., Microbial preparation of substances from aromatic metabolism/I, WO98/18936 (1998). [24] P.P. Cherepanov, W. Wackernagel, Gene disruption in Escherichia coli: TcR and KmR cassettes with the option of Flp-catalyzed excision of the antibioticresistance determinant, Gene 158 (1995) 9–14.

M. Weiner et al. / Biochemical Engineering Journal 83 (2014) 62–69 [25] S.K. Sharan, L.C. Thomason, S.G. Kuznetsov, D.L. Court, Recombineering: a homologous recombination-based method of genetic engineering, Nat. Protoc. 4 (2009) 206–223. [26] C. Albermann, N. Trachtmann, G.A. Sprenger, A simple and reliable method to conduct and monitor expression cassette integration into the Escherichia coli chromosome, Biotechnol. J. 5 (1) (2010) 32–38. [27] C. Albermann, S. Ghanegaonkar, K. Lemuth, T. Vallon, M. Reuss, W. Armbruster, G.A. Sprenger, Biosynthesis of the vitamin E compound ␦-tocotrienol in recombinant Escherichia coli cells, Chembiochem 9 (2008) 2524–2533. [28] J.G. Pan, J.S. Rhee, J.M. Lebeault, Physiological constraints in increasing biomass concentration of Escherichia coli B in fed-batch culture, Biotechnol. Lett. 9 (2) (1987) 89–94. [29] J. Hiller, E. Franco-Lara, V. Papaioannou, D. Weuster-Botz, Fast sampling and quenching procedures for microbial metabolic profiling, Biotechnol. Lett. 29 (2007) 1161–1167. [30] M. Jenzsch, S. Gnoth, M. Beck, M. Kleinschmidt, R. Simutis, A. Lübbert, Open-loop control of the biomass concentration within the growth phase

[31]

[32]

[33]

[34] [35]

69

of recombinant protein production processes, J. Biotechnol. 127 (2006) 84–94. S.A. Ibrahim, H. Yang, C.W. Seo, Antimicrobial activity of lactic acid and copper on growth of Salmonella and Escherichia coli O157:H7 in laboratory medium and carrot juice, Feed Chem. 109 (2007) 137–143. R. Jossek, J. Bongaerts, G.A. Sprenger, Characterization of a new feedbackresistant 3-deoxy-d-arabino-heptulosonate 7-phosphate synthase of Escherichia coli, FEMS Microbiol. Lett. 202 (2001) 145–148. G. Georgiou, M.L. Shuler, D.B. Wilson, Release of periplasmic enzymes and other physiological effects of ␤-lactamase overproduction in Escherichia coli, Biotechnol. Bioeng. 32 (1988) 741–748. K. Friehs, Plasmid copy number and plasmid stability, Adv. Biochem. Eng. Biotechnol. 86 (2004) 47–82. H. Link, B. Anselment, D. Weuster-Botz, Rapid media transition: an experimental approach for steady state analysis of metabolic pathways, Biotechnol. Prog. 26 (1) (2010) 1–10.