Research Collection
Doctoral Thesis
The system of biotransformations multi-enzyme reaction engineering for one-pot synthesis of vicinal diols
Author(s): Schümperli, Michael
Publication Date: 2008
Permanent Link: https://doi.org/10.3929/ethz-a-005593490
Rights / License: In Copyright - Non-Commercial Use Permitted
This page was generated automatically upon download from the ETH Zurich Research Collection. For more information please consult the Terms of use.
ETH Library Diss. ETH NO. 17692
The System of Biotransformations: Multi-Enzyme Reaction Engineering for One-Pot Synthesis of Vicinal Diols
A dissertation submitted to ETH ZURICH
for the degree of Doctor of Sciences
presented by MICHAEL SCHÜMPERLI Dipl. Natw. ETH
15. 01. 1977
citizen of Chur (GR) and Wäldi (TG), Switzerland
accepted on the recommendation of Prof. Dr. Sven Panke (ETH Zurich, Switzerland), examiner Prof. Dr. Victor de Lorenzo (CNB Madrid, Spain), co-examiner Dr. Matthias Heinemann (ETH Zurich, Switzerland), co-examiner
2008
To Eva
Abstract i
Abstract
A System of Biotransformations (SBT) is an in vitro multi-enzyme reaction cascade basing on the metabolic reaction network of an organism. In this study, an SBT was developed for the production of dihydroxyacetone phosphate (DHAP), which was further extended for the production of a vicinal diol. As underlying reaction network, the metabolism of E. coli W3110 was used. DHAP is a metabolite of the glycolysis, an important pathway of the central carbon metabolism. The upper branch of this reaction cascade, starting from glucose, is used as DHAP production pathway, while the lower branch functions as cofactor recycling part. In a first step, the construction of the SBT pathway is accomplished with general tools for metabolic engineering. The pathway is subsequently produced by cultivation of the resulting mutant. In a second step, the SBT is used for the synthetic purpose as the cell-free extract (CFX) of the culture. To ensure the accumulation of DHAP, the deletion of the gene tpiA was necessary. TpiA acts as a triosephosphate isomerase, interconverting DHAP and glyceraldehyde 3-phosphate. On the basis of this single knock-out mutant, SBT production of DHAP was possible with a yield of 55% and an initial reaction velocity of 0.3 U mg-1 of total protein concentration. The glycolysis as part of the central carbon metabolism is highly interconnected with the cell’s reaction network. It is shown however, that in an in vitro system the number of presumably intermediate-diverting reactions can be as low as three reactions that cannot be eliminated without making growth of the strain impossible. Therefore, even the glycolysis can theoretically be almost completely insulated from the metabolism, the precondition for a high yield of DHAP on glucose. Consequently, the preliminary SBT was optimized by further insulation of the SBT pathway by the deletion of more genes. This measure led to a quadruple deletion mutant, E. coli W3110 tpiA cyaA ptsI zwf, whose CFX was used in optimized buffer for performing the SBT synthesis. The yield of DHAP production on glucose could be increased to 73% and the initial reaction velocity was 0.6 U mg-1 of total protein concentration. The DHAP-producing SBT is dependent on two expensive cofactors: ATP and NAD+. While the initial concentration of the latter can be reduced without any impact on the DHAP synthesis performance, the reduction of the ATP led to a drastically decreased initial reaction rate and to very low product concentration, irrespective of provided opportunities for cofactor recycling. According to model predictions, in the optimized SBT there are still side reactions active in the recycling pathway, thus reducing the recycling efficiency. Furthermore, it could be shown that the CFX contains Abstract ii
AXP-degrading activities leading to a rapid ATP depletion and thus to the early stop of the SBT. As main AXP degraders, the enzymes Amn, Apt, UshA and YgiF were identified. DHAP is a rather instable molecule, which is furthermore difficult to purify from the reaction mixture of the SBT. However, DHAP is just an intermediary – but important – product for the application of DHAP-dependent aldolases in the synthesis of vicinal diols. Therefore, the DHAP-producing SBT was extended by an additional fructose-1,6- bisphosphate aldolase (FBA) to catalyze D-threo (3S, 4R) 5-deoxy-5-ethyl-xylulose 1-phosphate (DEXP) from DHAP and the artificial acceptor aldehyde butanal. In fact, DEXP was efficiently produced from DHAP when FBA from rabbit muscle aldolase was supplied. Probably, FBA from E. coli has difficulties in handling non-natural acceptor aldehydes. To gain a deeper insight into the SBT’s reaction system, an MS-based tool was developed for real-time on-line analysis of the intermediary metabolites of the reaction pathway. Thereby, for every metabolism a data point is generated after only a few seconds, enabling high resolution and information content. However, the quantification of the collected data is hampered by ion suppression effects and so far only qualitative data can be collected. Nevertheless, a detailed insight into the SBT’s reaction cascade could be gained.
Zusammenfassung iii
Zusammenfassung
Ein “System of Biotransformations” (SBT) ist eine in vitro von einem Multienzymsystem katalysierte Reaktionskaskade, die auf dem metabolischen Reaktionsnetzwerk eines Organismus basiert. Diese Dissertation widmet sich der Entwicklung eines SBT für die Produktion von Dihydroxyaceton-Phosphat (DHAP) und dessen Umsetzung in vicinale Diole. Als grundlegendes Reaktionsnetzwerk wurde der Metabolismus von E. coli W3110 benutzt. DHAP ist ein Intermediat der Glycolyse, eines wichtigen Reaktionswegs des zentralen Kohlenstoffwechsels. Der obere Teil dieser Reaktionskaskade, mit Glucose als Substrat, wurde als DHAP-Produktionsweg genutzt. Der untere Teil diente der Cofaktor- Regeneration. In einem ersten Schritt wurden für die Konstruktion des SBT die Werkzeuge des “Metabolic Engineering“ benutzt. Der Reaktionsweg wurde anschliessend durch die Kultivierung der resultierenden Mutante vervielfältigt. In einem zweiten Schritt wurde das SBT für den Synthesezweck als zellfreier Extrakt der Kultur (CFX) eingesetzt. Um die Akkumulation von DHAP sicherzustellen, war die Beseitigung des Gens tpiA nötig. TpiA ist eine Triosephosphat-Isomerase und kann DHAP in Glyceraldehyd-3-Phosphat umwandeln und umgekehrt. Auf der Basis dieser knock-out-Mutante erreichte die SBT-Produktion von DHAP eine Ausbeute von 55% und eine Startreaktionsgeschwindigkeit von 0.3 U mg-1 Gesamtprotein. Die Glycolyse als Teil des zentralen Kohlenstoffwechsels ist sehr stark mit dem Metabolismus einer Zelle verknüpft. In einem in-vitro-System kann die Zahl der Reaktionen, die vermutlich Intermediate aus dem Syntheseweg abziehen, auf drei reduziert werden, die nicht eliminiert werden können, ohne die Wachstumsfähigkeit des Stammes zu verlieren. Theoretisch kann die Glycolyse also beinahe komplett vom Metabolismus isoliert werden, was die Voraussetzung für eine hohe DHAP-Ausbeute darstellt. Das erste SBT wurde daher durch die weitere Isolation, also durch das Deletieren weiterer Gene, optimiert. Diese Massnahme führte zu einer Mutante mit vier Gen-knock-outs, deren CFX für die DHAP-Produktion in optimiertem Reaktionspuffer benutzt wurde. Die Ausbeute stieg auf 73%, die Startreaktions- geschwindigkeit auf 0.6 U mg-1 Gesamtprotein. Das DHAP-produzierende SBT ist von zwei teuren Cofaktoren abhängig, die anfänglich nahezu stöchiometrisch eingesetzt wurden: von ATP und NAD+. Die Startkonzentration von NAD+ konnte dabei ohne negativen Einfluss auf die DHAP-Produktion reduziert werden. Die Reduktion von ATP dagegen führte trotz der Cofaktorregeneration zu einer drastisch verminderten Startreaktionsgeschwindigkeit und zu sehr tiefen Produkt- konzentrationen. Gemäss einer Modellanalyse sind im optimierten SBT immer noch Zusammenfassung iv
Nebenreaktionen im Recyclingweg aktiv, die dessen Effizienz verringern. Darüber hinaus enthält der CFX Enzyme, welche AXP, also ATP, ADP oder AMP, abbauen und sehr rasch zu einem Verschwinden des eingesetzten ATP und damit zum Erliegen der Reaktion führen. Als hauptsächlich für den AXP-Abbau verantwortliche Enzyme wurden Amn, Apt, UshA und YgiF identifiziert. DHAP ist ein instabiles Molekül, das nur schwer aus dem SBT isoliert werden kann. DHAP ist aber auch nur ein – zwar wichtiges – Zwischenprodukt für die Synthese von vicinalen Diolen mit DHAP-abhängigen Aldolasen. Das DHAP-produzierende SBT wurde durch eine zusätzliche Fructose-1,6-Bisphosphat-Aldolase (FBA) erweitert, um D-threo (3S, 4R)-5-Deoxy-5-Ethyl-Xylulose-1-Phosphat (DEXP) aus DHAP und dem nicht natürlichen Akzeptoraldehyd Butanal synthetisieren zu können. DHAP wurde effizient in DEXP umgesetzt, wenn die FBA aus Kaninchenmuskel (RAMA) hinzugefügt wurde. Vermutlich kann die E. coli-eigene FBA den nicht natürlichen Akzeptoraldehyd nicht oder kaum umsetzen. Mit RAMA wurde eine DEXP-Ausbeute von 80% bezüglich DHAP erreicht. Um tiefere Einblicke in das SBT-Reaktionssystem zu gewinnen, wurde das Reaktionskompartiment direkt mit einem Massenspektrometer gekoppelt. Damit konnten Echtzeitmessungen der intermediären Metaboliten des Reaktionswegs durchgeführt werden. Die Messzeiten der Analytik lagen im Sekundenbereich, womit hohe Auflösung und Informationsdichte erreicht wurden. Die Quantifizierung der Metabolitenkonzentrationen wurde jedoch durch Ionensuppressionseffekte erschwert, weshalb bisher nur qualitative Daten zur Verfügung stehen. Eine detaillierte Einsicht in das Reaktionsnetzwerk ist dennoch gelungen.
Table of Contents v
Table of Contents
Abstract i
Zusammenfassung iii
List of abbreviations 1
1. Introduction – the System of Biotransformations 5
2. Chemical and Enzymatic routes to dihydroxyacetone phosphate 11 2.1. Abstract 12 2.2. Introduction 12 2.3. DHAP in metabolism 15 2.4. Chemical routes to DHAP 18 2.5. Enzymatic routes to DHAP 20 2.6. Alternatives to DHAP 28 2.7. Conclusion 30 2.8. Acknowledgments 31 2.9. References 32
3. Development of a DHAP-producing System of Biotransformations 41 3.1. Abstract 42 3.2. Introduction 42 3.3. Materials and Methods 46 3.4. Results 51 3.5. Discussion 62 3.6. Acknowledgments 64 3.7. References 65
4. ATP sinks in cell-free extracts of E. coli 69 4.1. Abstract 70 4.2. Introduction 70 4.3. Materials and Methods 72 4.4. Results 74 Table of Contents vi
4.5. Discussion 82 4.6. Acknowledgments 83 4.7. References 84
5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 87 5.1. Abstract 88 5.2. Introduction 88 5.3. Materials and Methods 91 5.4. Results and Discussion 94 5.5. Conslusion and Outlook 100 5.6. Acknowledgments 100 5.7. References 101
6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS 103 6.1. Abstract 104 6.2. Introduction 104 6.3. Materials and Methods 108 6.4. Results 113 6.5. Discussion 122 6.6. Acknowledgments 123 6.7. References 124
7. Summary and Outlook 127
Acknowledgments 129
Curriculum Vitae 130
List of abbreviations 1
List of abbreviations
2PG 2-Phosphoglycerate 3PG 3-Phosphoglycerate 5PR1P 5-phosphoribosyl 1-pyrophosphate Ac Acetate AcK Acetyl kinase AcP Acetyl phosphate Ade Adenine ADP Adenosine diphosphate Ads Adenosine Ald Aldehyde ALDase Aldolase Amn AMP nucleosidase AMP Adenosine monophosphate AP Acid phosphatase AcP Acetyl phosphate Apt Adenine phosphoribosyltransferase ATP Adenosine triphosphate AXP ATP, ADP and AMP CAD Collision gas cAMP Cyclic adenosine monophosphate CAT Catalase CFX Cell-free extract CHAD Protein domain found in adenylate cyclases CIP Calf intestinal alkaline phosphatase CP Creatine phosphate CRP cAMP receptor protein CUR Curtain gas CyaA Adenylate cyclase CYTH Protein domain found in adenylate cyclases DAG Diacylglycerol
DDDP2 2.5-Diethoxy-p-dioxane-2,5-dimethanol-O-2-O-5 bisphosphate DEX D-threo (3S, 4R) 5-deoxy-5-ethyl-xylulose DEXP D-threo (3S, 4R) 5-deoxy-5-ethyl-xylulose 1-phosphate DHA Dihydroxyacetone List of abbreviations 2
DHAK Dihydoxyacetone kinase DHAP Dihydroxyacetone phosphate D-LAC D-Lactate DNA Deoxyribonucleic acid DOPS L-threo-(3.4-dihydroxyphenyl)serine DPG 1,3-bisphosphoglycerate EI Electron ionization EMR Enzyme membrane reactor ENO Enolase ESI Electrospray ionization F6P Fructose 6-phosphate FBA Flux balance analysis FBPA Fructosebisphosphate aldolase FBP Fructose 1,6-bisphosphate FPK Fructose 6-phosphate kinase FruA Fructose 1,6-bisphosphate aldolase FucA Fuculose 1-phosphate aldolase G3P Glycerol 3-phosphate G6P Glucose 6-phosphate G6P-DH Glucose 6-phosphate dehydrogenase GAP Glyceraldehyde 3-phosphate GAP-DH Glyceraldehyde 3-phosphate dehydrogenase GC Gas chromatography GDH Glycerol 3-phosphate dehydrogenase GI Glucose 6-phosphate isomerase GK Glycerol kinase Glc Glucose Gly Glycerol GlyP Glycerol 3-phosphate GPO L-Glycerolphosphate oxidase GpsA Glycerol 3-phosphate dehydrogenase GS1 Nebulizer gas GS2 Auxiliary gas HK Hexokinase HPLC High performance liquid chromatography IlvGIH Acetolactate synthase INV Invertase List of abbreviations 3
Ki Inhibition constant
KM Michaelis Menten constant LAC Lactate LC Liquid chromatography LDH Lactate dehydrogenase LGlyP L-Glycerol 3-phosphate M9-GYE M9 medium containing glucose and yeast extract MG Methylglyoxal MgsA Methylglyoxal synthase MK Myokinase MOPS 3-(N-Morpholino)propanesulfonic acid MRM Multi-reaction monitoring MS Mass spectrometry MT Mixing tee NAD+ Nicotinamid adenine dinucleotide NADH Nicotinamid adenine dinucleotide, reduced NADPH Nicotinamid adenine dinucleotide phosphate, reduced NMO N-methylmorpholine N-oxide NMP Nucleotide monophosphate PC Phosphatidylcholine PDHA Phosphatidyldihydroxyacetone PEP Phosphoenol pyruvate PFK Phosphofructokinase PGI Phosphoglucose isomerase PGK Phosphoglycerate kinase Pgm Phosphoglucomutase PGM Phosphoglycerate mutase PHY Phytase
Pi Inorganic phosphate PK Pyruvate kinase PK Pyruvate kinase PLC Phospholipase C PLD Phospholipase D
PPi Pyrophosphate PPP Pentose phosphate pathway
Prottot Total protein PTS Phosphotransferase system List of abbreviations 4
PtsI Phosphotransferase system enzyme I PYK Pyruvate kinase PYR Pyruvate R5P Ribulose 5-phosphate Rac-PGly Racemic glycerol phosphate RAMA Rabbit muscle aldolase RhuA Rhamnulose 1-phosphate aldolase rinit Initial reaction rate RNA Ribonucleic acid SAL Salicylic acid SBT System of Biotransformations SerA Phosphoglycerate dehydrogenase SM Standard mix ST Splitting tee TADEX D-threo (3S, 4R) 1,3,4-triacetyl-5-deoxy-5-ethyl-xylulose TagA Tagatose 1,6-bisphosphate aldolase TktB Transketolase B
TNATP ATP turnover number TPAP Tetrapropylammonium perruthenate TPI / TpiA Triosephosphate isomerase UshA 5’-nucleotidase WT Wild-type XI Xylose isomerase YgiF Putative adenylate cyclase Zwf Glucose 6-phosphate dehydrogenase
1. Introduction - The System of Biotransformations 5
1. Introduction – the System of Biotransformations
Enzyme-catalyzed reactions are widely known and appreciated as being specific, fast and executed in mild conditions, typically ambient pH and temperature and aqueous solvents. Specificity includes stereo-, regio- or chemoselectivity [1-3]. Many enzymes have broad substrate tolerance and catalyze reactions with non-natural substrates. For almost every type of reaction in organic chemistry, an enzyme-catalyzed pendant can be found [4]. Although naturally enzyme reactions are carried out in aqueous environments, many enzymes are also able to perform their task in organic solvents or ionic liquids [4-6]. Thus, biotransformations have become widely used in the production of fine chemicals to catalyze one or several steps in reaction sequences [1, 7]. However, many enzymatic reactions lead to a product very similar to the substrate, introducing only minor – nevertheless important – changes to the substrate [7]. A biotransformation in which the product hardly shares any similarities with the substrate therefore requires many enzymes in a reaction chain or reaction network [8]. In living organisms, enzymatic reaction networks are used to produce the vast variety of molecules found in nature. These reaction networks can be exploited in fermentations [9]. An organism can also be genetically modified to produce a specific target [10]. However, in this approach, both the substrate and the product have to fulfill several requirements. i) Neither substrate nor product should be toxic to the organism as otherwise sophisticated substrate feeding and product removal schemes have to be applied or low product concentrations have to be accepted [11]. ii) The organism has to be able to take up the substrate and secrete the product. This requirement hampers the utilization of a variety of starting materials and the production of compounds that cannot be transported out of the cell across the membrane. Toxicity and membrane transport are problems of much smaller or no importance in an in vitro multi-enzyme system. This approach has been implemented before, e.g. for the production of phosphorylated carbohydrates from sucrose, consisting of a reaction cascade of eight enzymes [12]. However, there are two issues which have to be considered. Firstly, the enzymes in an in vitro system have to be available. One possibility is the overproduction of every single enzyme in an organism with subsequent isolation and purification. However, this approach is time-consuming and laborious. Furthermore, using enzymes in solution, they might lack long-term operational stability and be difficult to recover and re-use. To overcome the latter problem, the enzymes can be immobilized [13]. Additionally, by the immobilization of several enzymes at the same 1. Introduction - The System of Biotransformations 6 time, the catalysts can spatially be brought close to one another, building up whole reaction cascades [14, 15]. As mentioned above, the expression and purification of enzymes is laborious and furthermore may result in a decreased enzymatic activity. It could be desirable to use cell-free extracts containing the required enzymes. The combination of the advantages of fermentation and using non-purified enzymes in an immobilized way led to a whole- cell approach, in which one or several organisms expressing desired enzymatic activities were permeabilized prior to their use as catalysts [16, 17]. The drawback of the utilization of non-purified enzymes is the potential presence of side activities possibly reducing the final yield of the reaction. Secondly, the dynamics of a multi-enzyme system are difficult to predict. The physical parameters of an in vivo reaction cascade cannot necessarily be derived from the known physical parameters measured in vitro of the single components of the reaction cascade [18]. It is likely that the same is true for in vitro reaction cascades, especially when the enzymes are immobilized. Furthermore, an enzyme might behave differently in different environments, depending on the composition of the matrix and the total protein concentration. Nevertheless, enzymatic in vitro reaction cascades are a promising tool to synthesize a valuable product from a cheap substrate in mild conditions, with high stereoselectivity and in only a few operational steps. The System of Biotransformations (SBT) is such a multi-enzymatic reaction cascade for in vitro synthesis of a desired compound. In a first phase, the basis of the reaction network topology of a living organism is used to design the production of a target compound. In a second phase, the dynamics of this reaction network are adapted for optimal target production. The SBT combines advantages of the fermentation and the in vitro reaction approach. The reaction cascade is produced in a first step in the cultivation of a – most likely genetically modified – organism. The synthetic task is performed in a second step as cell-free in vitro system. This thesis presents fundamental investigations towards the first SBT. As production target, dihydroxyacetone phosphate (DHAP) was chosen. DHAP is the key compound for the synthesis of vicinal diols using the DHAP-dependent aldolases, which have been demonstrated as very useful biocatalysts on many occasions [19-27]. Next, DHAP is phosphorylated and thus not able to cross the cellular membrane. No sustainable production routes have currently been implemented, and consequently the compound is difficult to obtain. In chapter 2, the role of DHAP in the enzyme-catalyzed aldol reactions is discussed together with its chemical and enzymatic production processes. Chapter 3 describes the development of the DHAP-producing SBT and its optimization. Chapter 4 is dedicated to the problem of cofactor recycling in a DHAP-producing SBT. In 1. Introduction - The System of Biotransformations 7 chapter 5, the SBT is extended for the production of an unnatural carbohydrate. In chapter 6, a comprehensive analytic tool basing on mass spectrometry is developed which can be used for the improved characterization of the SBT.
1. Introduction - The System of Biotransformations 8
References
1. Straathof, A.J.J., S. Panke, and A. Schmid, The production of fine chemicals by biotransformations. Curr Op Biotechnol, 2002. 13(6): p. 548-556. 2. Schmid, A., et al., The use of enzymes in the chemical industry in Europe. Curr Op Biotechnol, 2002. 13(4): p. 359-366. 3. Ishige, T., K. Honda, and S. Shimizu, Whole organism biocatalysis. Curr Op Chem Biol, 2005. 9(2): p. 174-180. 4. Faber, K., Biotransformations in Organic Chemistry. 5th ed. 2004, Berlin: Springer- Verlag. 5. Klibanov, A.M., Improving enzymes by using them in organic solvents. Nature, 2001. 409(6817): p. 241-246. 6. Kragl, U., M. Eckstein, and N. Kaftzik, Enzyme catalysis in ionic liquids. Current Opinion in Biotechnology, 2002. 13(6): p. 565-571. 7. Breuer, M., et al., Industrial Methods for the Production of Optically Active Intermediates. Angewandte Chemie International Edition, 2004. 43(7): p. 788- 824. 8. Panke, S., et al., Industrial multi-step biotransformations. Chimica Oggi, 2004. 9(22): p. 44-47. 9. Christensen, B. and J. Nielsen, Metabolic network analysis of Penicillium chrysogenum using C-13-labeled glucose. Biotechnol Bioeng, 2000. 68(6): p. 652- 659. 10. Jimenez, A., et al., Metabolic Engineering of the Purine Pathway for Riboflavin Production in Ashbya gossypii. Appl. Environ. Microbiol., 2005. 71(10): p. 5743-5751. 11. Booth, A.J., S.H. Ngiam, and G.J. Lye, Antibiotic purification from fermentation broths by counter-current chromatography: analysis of product purity and yield trade-offs. Bioproc Biosyst Eng, 2004. 27(1): p. 51-61. 12. Fessner, W.-D. and C. Walter, "Artificial metabolisms" for the Asymmetric One-Pot Synthesis of Branched-Chain Saccharides. Angew Chem Int Ed, 1992. 31(5): p. 614- 616. 13. Sheldon, R., A. , Enzyme Immobilization: The Quest for Optimum Performance. Advanced Synthesis & Catalysis, 2007. 349(8-9): p. 1289-1307. 14. Bruggink, A., R. Schoevaart, and T. Kieboom, Concepts of Nature in Organic Synthesis: Cascade Catalysis and Multistep Conversions in Concert. Organic Process Research & Development, 2003. 7(5): p. 622-640. 15. Nahalka, J., et al., Superbeads: Immobilization in "Sweet" Chemistry. Chemistry - A European Journal, 2003. 9(2): p. 372-377. 1. Introduction - The System of Biotransformations 9
16. Zhang, J., et al., Synthesis of Galactose-Containing Oligosaccharides through Superbeads and Superbug Approaches: Substrate Recognition along Different Biosynthetic Pathways, in Methods in Enzymology. 2003, Academic Press. p. 106- 124. 17. Endo, T., et al., Large-scale production of N-acetyllactosamine through bacterial coupling. Carbohydrate Res, 1999. 316: p. 179-183. 18. Teusink, B., et al., Can yeast glycolysis be understood in terms of in vitro kinetics of the constituent enzymes? Testing biochemistry. European Journal of Biochemistry, 2000. 267(17): p. 5313-5329. 19. Silvestri, M.G., et al., Asymmetric Aldol Reactions using Aldolases, in Topics in Stereochemistry, S.E. Denmark, Editor. 2003, John Wiley & Sons Inc. p. 267-341. 20. Bednarski, M.D., et al., Rabbit Muscle aldolase as a Catalyst in organic Synthesis. J Am Chem Soc, 1989. 111: p. 627-635. 21. Fessner, W.-D., Sinerius, G., Diastereoselective Enzymatic Aldol Additions: L- Rhamnulose and L-Fuculose 1-Phosphate Aldolases from E. coli. Angew Chem Int Ed, 1991. 30(5): p. 555-558. 22. Crestia, D., Guérard, C., Bolte, J., Demuynck, C., Rabbit muscle aldolase (RAMA) as a catalyst in a new approach for the synthesis of 3-deoxy-D-manno-2-octulosonic acid and analogues. J Mol Catal B: Enzymatic, 2001. 11: p. 207 - 217. 23. Zhu, W. and Z. Li, Synthesis of perfluoroalkylated sugars catalyzed by rabbit muscle aldolase (RAMA). J Chem Soc, Perkin Trans 1, 2000: p. 1105. 24. Schoevaart, R., Applications of aldolases in organic synthesis. 2000, Technische Universiteit Delft: Delft. p. 136. 25. Schultz, M., Waldmann, H., Vogt, W., Kunz, H., Stereospecific C-C-bond formation with rabbit muscle aldolase - a chemoenzymatic synthesis of (+)-exo- brevicomycin. Tetrahedron Lett, 1990. 31(6): p. 867 - 868. 26. Dean, S.M., W.A. Greenberg, and C.-H. Wong, Recent Advances in Aldolase- Catalyzed Asymmetric Synthesis. Advanced Synthesis & Catalysis, 2007. 349(8- 9): p. 1308-1320. 27. Espelt, L., et al., Stereoselective aldol additions catalyzed by dihydroxyacetone phosphate-dependent aldolases in emulsion systems: Preparation and structural characterization of linear and cyclic iminopolyols from aminoaldehydes. Chemistry - a European Journal, 2003. 9(20): p. 4887-4899.
2. Chemical and enzymatic routes to dihydroxyacetone phosphate 11
2. Chemical and enzymatic routes to dihydroxyacetone phosphate
Michael Schümperli, René Pellaux, Sven Panke
This review was published in Applied Microbiology and Biotechnology, 2007. 75(1): p. 33-45.
2. Chemical and enzymatic routes to dihydroxyacetone phosphate 12
2.1. Abstract
Stereoselective carbon-carbon bond formation with aldolases has become an indispensable tool in preparative synthetic chemistry. In particular the dihydroxyacetone phosphate (DHAP)-dependent aldolases are attractive because four different types are available that allow access to a complete set of diastereomers of vicinal diols from achiral aldehyde acceptors and the DHAP donor substrate. While the substrate specificity for the acceptor is rather relaxed, these enzymes show only very limited tolerance for substituting the donor. Therefore, access to DHAP is instrumental for the preparative exploitation of these enzymes, and several routes for its synthesis have become available. DHAP is unstable, so chemical synthetic routes have concentrated on producing a storable precursor that can easily be converted to DHAP immediately before its use. Enzymatic routes have concentrated on integrating the DHAP-formation with upstream or downstream catalytic steps, leading to multi- enzyme arrangements with up to seven enzymes operating simultaneously. While the various chemical routes suffer from either low yields, complicated work-up, or toxic reagents or catalysts, the enzymatic routes suffer from complex product mixtures and the need to assemble multiple enzymes into one reaction scheme. Both types of routes will require further improvement to serve as a basis for a scalable route to DHAP.
2.2. Introduction
Aldolases reversibly catalyze the enantioselective formation of carbon-carbon bonds, a key reaction in organic chemistry, and have become an indispensable tool in the preparative chemist’s toolbox. Next to their stereoselectivity, they allow circumventing or minimizing protective group chemistry and applying simultaneously several catalytic steps in one pot. Their mild reaction conditions make them particularly suitable for multifunctional complex molecules, and together these properties have led to a large number of applications of aldolases for preparative synthetic purposes, particularly in the production of carbohydrates and similar molecules [1, 2]. Recent industrial applications include the manufacturing of N-acetylneuraminic acid from pyruvate and N-acetylmannosamine for access to neuraminidase inhibitors [3], access to L-threo-(3,4- dihydroxyphenyl)serine (DOPS) in the treatment of parkinsonism [4], and novel routes to statin side chains in the manufacturing of cholesterole lowering drugs (summarized in [5]). 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 13
Generally speaking, aldolases catalyze the reversible addition of a ketone donor to an aldehyde acceptor. While the range of acceptors that different aldolases can utilize is frequently broad, aldolases typically tolerate only a very small number of donors. The most important natural donors are pyruvate, phosphoenol pyruvate, acetaldehyde, glycine, and dihydroxyacetone phosphate (DHAP), giving rise to the synthetically most useful classification of aldolases [6]. Glycine and acetaldehyde are easily accessible compounds and excellent routes to pyruvate have recently become available [7]. Correspondingly, the examples of industrial utilization mentioned above, stem from these classes. In contrast, access to phosphoenol pyruvate and in particular to DHAP remains difficult, which is reflected in prohibitively high prices for these compounds. This is in strong contradiction to the unique synthetic opportunities that have been elaborated over the last two decades based on DHAP-dependent aldolases. These opportunities rest on the fact that there are four types of enzymes which can provide access to a diastereomerically complete set of vicinal diols (Fig. 1) [8-12]. The corresponding diastereoselectivity is frequently excellent [13]. Consequently, the DHAP- dependent aldolases offer the opportunity to carry out in a controlled fashion different stereochemically complementary carbon-carbon bond aldol reactions. This has been widely exploited on laboratory scale over the past two decades [6, 14-16], in particular in the area of iminocyclitols, which can inhibit glycosidases and are therefore studied for antiviral, anticancer, antidiabetic, and pesticidal properties [17]. Representatives from the required four classes of enzymes have been throughly developed – they have been cloned, overexpressed, and purified [9, 18-23] and the 3D- structures are available, so rational mutagenesis or directed evolution can be applied to attempt adapting the substrate specificity of a given enzyme [24-27]. In summary, the DHAP-dependent aldolases are a particularly well-developed group of enzymes with enormous synthetic potential. However, in an industrial setting their applicability hinges on the availability of DHAP. Therefore, we summarize in this mini- review the routes that are available for the manufacturing of this compound. We will begin by summarizing the involvement of DHAP in metabolic pathways, as this has motivated the recruitment of the various enzymatic routes that will be discussed later. Subsequently, we will highlight synthetic chemical and enzymatic routes to DHAP, and finally discuss a number of possible alternatives to the use of DHAP. 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 14
OH O OH O
OH OH R R
OH OH 3S, 4R 3R, 4R FruA FucA
O
HO OPO3H2
+ R-CHO
RhuA TagA
OH O OH O
OH OH R R
OH OH 3R, 4S 3S, 4S Fig. 1: Four DHAP-dependent aldolases provide access to a diastereomerically complete set of vicinal diols. FruA: Fructose 1,6-bisphosphate aldolase; FucA: Fuculose 1-phosphate aldolase; RhuA: Rhamnulose 1-phosphate aldolase; TagA: Tagatose 1,6- bisphosphate aldolase.
2. Chemical and enzymatic routes to dihydroxyacetone phosphate 15
2.3. DHAP in metabolism
In metabolism, DHAP is involved in various pathways that have also inspired multi- enzymatic routes to its formation (see below). Firstly, it is an intermediate in the glycolysis and gluconeogenesis, formed in the reversible conversion of fructose 1,6- bisphosphate to glyceraldehyde 3-phosphate and DHAP. DHAP is then isomerized to glyceraldehyde 3-phosphate by triosephosphate isomerase. This enzyme proceeds by deprotonating DHAP to an enediolate intermediate, which is then re-protonated to give glyceraldehyde 3-phosphate (Fig. 2b).
Fig. 2: Most prominent conversions of DHAP. DHAP is deprotonated and degrades chemically along route a) to methylglyoxal. Alternatively, it is converted by triose phosphate isomerase along route b) to glyceraldehyde 3-phosphate.
Secondly, DHAP can be channeled into the lipid metabolism by NADH-dependent reduction to L-glycerol 3-phosphate by glycerol 3-phosphate dehydrogenase, or can be provided from glycerol by the reverse reaction or, more importantly in the present context, by the oxidation of L-glycerol 3-phosphate catalyzed by glycerol phosphate oxidase, which is a major route for glycerol assimilation in a number of microorganisms [28]. Alternatively, a glycerol phosphate oxidase route can be obtained from eukaryotes that use the mitochondrial glycerol phosphate shuttle to regenerate NAD+ instead of transferring reducing equivalents, such as in Trypanosomes [29]. 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 16
Thirdly, DHAP is an important intermediate in the anaerobic dissimilation of glycerol, during which glycerol is oxidized to dihydroxyacetone and channelled into the glycolysis via phosphorylation to DHAP by dihydroxyacetone kinases [30]. The latter class of enzymes is also important in the assimilation of formaldehyde in methylotrophic yeasts and in the detoxification of dihydroxyacetone in yeasts [31]. These dihydroxyacetone kinases do not accept glycerol (in contrast to glycerol kinases that accept dihydroxyacetone, see below) and exist in two structurally related classes: one that uses ATP as phosphate donor and has only one structural gene, and one that uses phosphoenol pyruvate and has typically three structural genes [32]. Only the ATP- dependent enzymes are relevant for the present discussion. Fourthly, DHAP has been found to be part of the metabolism of various sugars (in Escherichia coli, these are L-fucose, L-rhamnose, and galactitol) and the enzymes from the corresponding pathways – fuculose 1-phosphate aldolase [33], rhamnulose 1-phosphate aldolase [34], and tagatose 1,6-bisphosphate aldolase [35] - form together with fructose 1,6-bisphosphate aldolase the complete diastereomeric platform (Fig. 1). Next to these metabolic pathways, the knowledge of which has inspired the various enzymatic synthesis routes (see below), DHAP is an intermediate in the regeneration of ribulose 1,5-bisphosphate from C6 and C3 sugars in the dark reactions of photosynthesis, where the fructose 1,6-bisphosphate aldolase catalyzes the interconversion of erythrose 4-phosphate, DHAP, and sedoheptulose 1,7-bisphosphate.
One important property of DHAP is its rather low chemical stability (Fig 2): Degradation is initiated by the deprotonation of DHAP, just as it is the case for triosephosphate isomerase, but then the phosphate group is eliminated rather then the intermediate re-protonated to glyceraldehyde 3-phosphate. This elimination occurs under neutral and in particular under basic conditions and leads to the formation of methylglyoxal [36]. Typical chemical half-lives for DHAP under neutral to slightly basic conditions are between 3 h (37oC) and 30 h (25oC) [37, 38]. This degradation reaction is even a side reaction of triose phosphate isomerase catalysis [36], which makes methylglyoxal an inevitable side product of glycolysis [39]. As the latter compound is cytotoxic, it is then removed by conversion to D-lactate or pyruvate [40]. The chemical lability of DHAP has a profound influence on the strategies that have been followed in order to produce the molecule for preparative purposes. One strategy is to produce a stable precursor of DHAP rather than DHAP itself. This precursor would need to be converted to DHAP shortly before the reaction. Alternatively, DHAP can be produced in situ and immediately react further, so that the side reaction to methylglyoxal can be effectively suppressed. Both approaches would benefit if the step in which the DHAP is actually 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 17 produced is compatible – in terms of solvent and by-product profile - to the subsequent step, in which the DHAP is consumed again. This poses specific additional requirements on the activation and in situ production procedures. 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 18
2.4. Chemical routes to DHAP
Due to the multiple functionalities of DHAP, most of its published chemical syntheses require complicated multistep procedures including protection/deprotection steps. Essentially, synthetic routes towards three stable DHAP precursors have been described in the literature (Fig. 3): The cyclic dimeric DDDP2, one of a series of monomeric phosphorylated ketals, or dibenzyl-3-benzylhydroxyacetone phosphate. The first intermediate is available in several steps from the cheap dihydroxyacetone dimer (Fig. 3a). Several reaction sequences have been described. They are based on ketalization, phosphorylation by diphenylphosphorochloridate followed by hydrogenolysis [41], ketalization, phosphorylation with phosphorus oxychloride and isolation of the free phosphate as its barium salt [42], or ketalization, phosphorylation with dibenzyl N,N-diethylphosphoramidite followed by oxidation and hydrogenolysis [43]. In particular the improved route via dibenzyl N,N-diethylphosphoramidite produced good yields of the intermediate in the order of 73% [44]. From the stable precursors, DHAP can be obtained via hydrolysis by heating or acidification. These routes have the advantage that the stable intermediate can be produced efficiently, but the yield of the final step, the conversion to DHAP, has so far been only moderate, even though there might be some room for improvement. The monomeric phosphorylated ketals (Fig. 3b) are available by multi-step reactions from 3-chloro-1,2-propanediol [45], acetone [46], 1,3-dibromoacetone (available by acidic bromination of acetone) [47] or dihydroxyacetone dimer [48, 49]. Again, the step from the precursor to DHAP can be made by acidic treatment. In comparison to the routes summarized in Fig. 3a, the final step to DHAP gives usually excellent yields, but overall yields in the steps to the precursor are only moderate. In summary, even though it has been frequently shown that these routes lead to DHAP preparations that can be used directly in aldolase reactions, both sets of routes are multi-step, rather low yielding, include complicated purification procedures, and involve toxic and (in part) expensive chemicals. Recently, a convenient gram-scale synthesis of a benzylated DHAP precursor starting from cheap racemic benzylglycidol was reported (Fig. 3c) [50]. A Lewis acid-mediated regioselective epoxide ring opening with dibenzyl phosphate followed by catalytic oxidation of the secondary alcohol by TPAP/NMO (tetrapropylammonium perruthenate/ N-methylmorpholine N-oxide) led to dibenzyl-3-benzylhydroxyacetone phosphate as a stable stock material with an overall yield of 74%, which was then quantitatively hydrogenolyzed into DHAP, which in turn was directly suitable for enzymatic aldol reactions [51]. However, this method requires stoichiometric amounts 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 19 of the Lewis acid CuI and of NMO, which is inconvenient to prepare, and extensive use of the expensive catalyst TPAP. Furthermore, the starting material is not commercially available.
Fig. 3: Principal chemical routes DHAP. a) Routes to DDDP2. b) Route to monomeric phosphorylated ketals. c) Route to dibenzyl-3-benzylhydroxyacetone phosphate.
2. Chemical and enzymatic routes to dihydroxyacetone phosphate 20
2.5. Enzymatic routes to DHAP
Enzymatic routes to DHAP follow one of three general routes from cheap unphosphorylated precursors: from dihydroxyacetone to DHAP, from glycerol via glycerol phosphate to DHAP, or via multi-step routes that mimic glycolysis. In general, these routes require at some point the transfer of phosphate residues from donors with a high transfer potential to either ADP and then to unphosphorylated starting material or to the unphosphorylated starting material directly. The majority of the syntheses discussed below rely on ATP as the phosphorylating agent. As ATP is expensive, regeneration becomes a point of prime concern in such routes. As this is of central importance for the following and applies to most of the routes discussed below, we will start with a brief discussion of ATP regeneration.
Regeneration of phosphate donors An essential feature of the routes discussed below is the phosphorylation of a three carbon unit that requires spending a high energy phosphoester bond, unless reverse hydrolysis conditions are applied (see below). Recent advances in the regeneration of ATP as the most important phosphate donor have been discussed before [52]. In the set of syntheses discussed here, two regeneration systems have been of prime importance (Fig. 4a): ATP regeneration from acetyl phosphate by acetyl kinase or from phosphoenol pyruvate by pyruvate kinase. Even though these systems are well established and have been discussed frequently before [53], it might be useful for this minireview to summarize the most important features: acetyl phosphate is very easy to prepare but has the disadvantage that it hydrolyzes rather rapidly. In contrast, phosphoenol pyruvate is a rather stable compound, but it is more laborious and expensive to produce. Furthermore, the pyruvate that is formed during ATP regeneration acts as a competitive inhibitor of the kinase, for example with a Ki of only around 10 mM in the case of the muscle enzyme [54]. One additional system should be mentioned in this context: the regeneration of ATP from cheap polyphosphate by a polyphosphate kinase. This enzyme was overexpressed in E. coli, thereby reducing the problem of contaminating phosphatases that previously lead to phosphate accumulation [55]. It was successfully used in ATP regeneration with polyphosphate to kinate nucleoside bisphosphates. Unfortunately, the degree to which the nucleoside bisphosphates can be phosphorylated seems to be limited to around 50 % in the case of ADP for unclear reasons [55]. When required, polyphosphate can also be used to regenerate ATP from AMP by the combined activity of a polyphosphate:AMP phosphotransferase and an adenylate kinase [56]. Even though the 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 21 discussed enzymes appear to be synthetically very useful, they have so far never been used for DHAP production, most probably because of the requirements for purified enzymes to avoid unproductive dephosphorylation of the polyphosphate [57]. From dihydroxyacetone to DHAP The most straight forward enzymatic approach to DHAP starts at the dihydroxyacetone monomer (Fig. 4b), which forms from the cheap dimer in aqueous solution. This route has been pioneered by Wong and Whitesides [12], who exploited the relaxed substrate specificity of glycerol kinases to produce DHAP from dihydroxyacetone and ATP [58]. For example, glycerol kinase from Saccharomyces cerevisiae was used to produce DHAP on mole-scale from dihydroxyacetone and ATP [59]. Alternatively, dihydroxyacetone kinases have been employed as a substitute [60-62]. Although a number of these enzymes have been investigated (see [60] and references therein), it is difficult to assess the relative usefulness of the different species: The enzymes usually depend on divalent cations, measured specific activities are in the range of 20 U mg-1 of purified protein, and KM-values for dihydroxyacetone are generally below 100 M, which makes these enzymes in general useful for DHAP production. An alternative procedure for DHAP production that utilizes a cheap phosphate donor - and thus has the potential to circumvent the issue of cofactor regeneration completely - is based on the kinetically controlled transphosphorylation of dihydroxyacetone from cheap pyrophosphate by an acid phosphatase from Shigella flexneri [63]. In an optimized procedure, the enzyme allowed the production of up to 100 mM of DHAP in 6 h with yields of 20 % on dihydroxyacetone and pyrophosphate, before the yields started to fall again because of the hydrolytic activity of the phosphatase on DHAP. These are rather high values for transphosphorylation reactions when compared to similar protocols for the phosphorylation of glycerol (see below), but of course still far removed from preparative suitabilty. Furthermore, the enzyme has a large KM value of 3.6 M for dihydroxyacetone and its application leads to the accumulation of high concentrations of phosphate, which needs to be removed before application because it inhibits aldolases. A similar, though more complicated approach had been reported earlier, when an excess of dihydroxyacetone was used for transesterification of phosphatidylcholine to phosphatidyldihydroxyacetone in an organic phase by phospholipase D. The resulting compound was then isolated and hydrolyzed by phospholipase C to DHAP and 1,2-diacyclglycerol [64, 65]. Again, the yields on dihydroxyacetone are low, as it is typical for transesterifications.
2. Chemical and enzymatic routes to dihydroxyacetone phosphate 22
From glycerol to glycerol phosphate and DHAP An alternative route to DHAP is the phosphorylation of glycerol to L-glycerol 3-phosphate (sn-glycerol phosphate) and the subsequent oxidation to DHAP (Fig. 4c). Although this reaction sequence requires at least one enzymatic step more, it has the advantage that L-glycerol 3-phosphate is a stable compound that can be stored. The oxidation step is potentially very simple and requires only a glycerol phosphate oxidase and catalase (see below). Glycerol is somewhat more expensive than dihydroxyacetone, but this is poised to change at least to some extent with the increasing number of biodiesel installations that are coming on-stream and that produce glycerol as a side- product.
Fig. 4: Principal enzymatic routes to DHAP and the resulting sugar. Notation of enzymes and compounds as detailed for Tab. 1. Capital letters: alternative options. Numbers: Subsequent steps. a) Routes for ATP regeneration that have been used in the experiments depicted in b) to d). b): Routes starting from dihydroxyacetone. c) Routes passing through phosphorylated glycerol as intermediate. d) Multi-enzyme route starting from sucrose. e) Dephosphorylation of the sugar that resulted from the aldol reaction of DHAP and the aldehyde. For yields of the different routes, see Tab. 1. Capital letters indicate different enzymes that have been used for the same step. Numerals indicate that this step has been carried out in several subsequent steps. 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 23
The first step in this route is the production of a phosphorylated glycerol. Glycerol phosphorylated at the 3-position is chiral, and even though DHAP is not, this chirality is important because the L-glycerol phosphate oxidase that converts glycerol phosphate to DHAP is selective for L-glycerol 3-phosphate. Consequently, racemic mixtures can maximally deliver a 50% yield of DHAP on glycerol. However, as glycerol is cheap, this might be tolerable if the phosphorylation is cheap as well. One such procedure has been recently suggested to start from rac-glycidol, a common bulk chemical, the three-membered ring of which can be chemically opened with phosphate and gives then a mixture of the racemic terminal glycerol phosphate and the 2-glycerol phosphate [66]. These conditions can be optimized for yields of racemic glycerol phosphate up to 55 % or 28 % for the required L-glycerol 3-phosphate. Racemic glycerol phosphate without contaminating glycerol 2-phosphate is available enzymatically from glycerol and pyrophosphate by catalysis under reverse hydrolysis conditions with phytase (e.g. from Aspergillus ficuum) [67, 68]. Conditions could be optimized to 95 % glycerol and 150 mM pyrophosphate and resulted in quantitative conversion of pyrophosphate to racemic glycerol phosphate. Replacing phytase by calf intestinal alkaline phosphatase, it was possible to use a similar scheme and produce L-glycerol 3-phosphate from glycerol and pyrophosphate with a 9-fold excess over the D-enantiomer. The conditions were different with respect to pH and a reduced glycerol concentration (60 % in water, pH 7.9), and pyrophosphate was not consumed completely (55 %) [69]. Still, with enzymatic methods it is also easily possible to direct the production of glycerol phosphate completely to the desired L-enantiomer. For example the glycerol kinase from S. cerevisiae was exploited to produce exclusively L-glycerol 3-phosphate on a preparative scale [59] with ATP as the phosphate donor, and the enzyme from Bacillus stearothermophilus delivered L-glycerol 3-phosphate from ATP and glycerol up to concentrations of 300 mM (52 g L-1, limited by substrate inhibition) [37].
In order to obtain DHAP, the L-glycerol 3-phosphate has to be oxidized by a glycerol phosphate oxidase with consumption of oxygen and production of hydrogen peroxide. In this procedure, the inactivation of the glycerol phosphate oxidase by hydrogen peroxide was prevented by addition of catalase (Fig. 4c). Several L-glycerol phosphate oxidases have been investigated for their utility [70], and all show product inhibition from DHAP. Still, useful enzymes could be identified, in particular from Streptococcus sp., that still maintain 20 % of the activity at DHAP concentrations of 100 mM (17 g L-1), show hardly any substrate inhibition, tolerate relatively large concentrations of phosphate, are stable and inactivate relatively slowly at elevated levels of hydrogen 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 24 peroxide [70]. When investigated more closely, L-glycerol phosphate oxidase preparations from S. thermophilus were shown to have inhibition constants on the order of 60 to 80 mM for L-glycerol 3-phosphate and were essentially insensitive against accumulating phosphate. However, the enzymes were sensitive against air bubbles in particular and against oxygen in general [71]. Furthermore, oxidative conditions favored the production of polymers from DHAP. The authors evaluated several possibilities to provide mildly oxidative conditions and chose dosing of hydrogen peroxide in the presence of catalase to provide the required oxygen. Still the glycerol phosphate oxidase showed significant inactivation over the process (60 % remaining after 7 h), from the produced oxygen as well as from DHAP [71, 72]. Remarkably, the procedure appears to be rather insensitive to substantial amounts of the 2-glycerol phosphate isomer, as the L-glycerol 3-phosphate in the corresponding mix of three glycerol phosphate isomers from the ring opening of the rac-glycidol (see above) could be quantitatively converted to DHAP. However, rather large concentrations of enzymes were employed [66].
Integration of steps Due to the instability of DHAP, this molecule is usually not the desired product in any given reaction. Rather, it is a central intermediate that should be converted further with an appropriate aldehyde in the presence of an aldolase that provides one of the four possible stereochemistries. At the same time, for example the route starting from glycerol requires at least two enzymatic steps to DHAP plus cofactor regeneration and catalase – in short, it is highly desirable to integrate some or even all of these steps simultaneously into one pot, or if simultaneous reactions are impossible, than at least subsequently in one pot (Tab. 1). Enzyme-catalyzed reactions have here a unique advantage in that many of them have near-neutral pH, ambient temperature, and aqueous medium as their natural environment, so in principle integration of several enzymatic steps should be possible. On the other hand, some of the single reactions take place under conditions that are rather removed from such “average conditions” and consume/produce substrates/products that by experience might cause problems in integration. When starting from dihydroxyacetone, integration requirements are lower than for the other routes, as there is only one step from dihydroxyacetone to DHAP. In fact, this step, ATP regeneration, and aldolase reaction could be operated simultaneously in one pot at pH 7 [61]. The relatively broad pH tolerance of the rabbit muscle fructose 1,6-bisphosphate aldolase allowed even to integrate the operation of acid phosphatase with the aldolase operation: At pH 4.5, a compromise between the activities of the two 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 25 enzymes, the dihydroxyacetone was phosphorylated, the DHAP converted to the phosphorylated sugar, and the product was even dephosphorylated once the pyrophosphate was consumed - a particularly nice example of the potential benefits of the integration of several reaction steps [63]. Integration of routes passing through L-glycerol 3-phosphate is more challenging. The aldolase catalyzed coupling of DHAP to an aldehyde can be integrated with the oxidation of glycerol phosphate – which has the added benefit that this is an effective way to prevent inhibition of the glycerol phosphate oxidase by accumulating DHAP [70]. However, glycerol phosphate production with glycerol kinase cannot be integrated, because the kinase is rapidly inhibited by the oxygen required for the oxidation of the glycerol phosphate [70]. The substitution of ATP-regeneration by transphosphorylation catalyzed by phytase is also difficult to integrate. This step was optimized for a pH of 3.5 and a glycerol concentration of 95 %, and both conditions are not compatible with the subsequent requirement for oxidation of L-glycerol 3-phosphate by the glycerol phosphate oxidase and catalase. The aldolase step could most probably be integrated with the previous step. However the final step, the dephosphorylation of the DHAP-adduct, requires a second shift of pH back to acidic conditions [67]. Still, also this procedure is a one-pot procedure, and as a particular elegant feature the phytase that has been added for the first step was re-activated in the last step by the second pH shift. One exceptional example of integration of multiple enzymatic steps is the in vitro reconstitution of an eight enzyme pathway for simultaneous operation in one pot (Fig. 4d) [73]: In order to produce DHAP, an eight enzyme system was set up that started from sucrose to produce glucose and fructose, proceeded to fructose 6-phosphate, fructose 1,6-bisphosphate, and finally glyceraldehyde and DHAP. The required ATP was regenerated from phosphoenol pyruvate. Finally, DHAP was then converted by the aldolase to the phosphorylated ketose. This system was operated on a scale of 23 g L-1 of phosphorylated sugar. This multi-enzyme approach presented also access to the relatively expensive fructose 1,6-bisphosphate, which can also serve as a convenient but expensive starting material for DHAP [74].
The different multi-step enzymatic routes are compiled in Tab. 1. Even though nearly all of the routes mentioned there are one pot-routes and therefore rather convenient, please note that some of those routes cover only part of the overall sequence from cheap starting material to dephosphorylated sugar. Still, as entries 8 and 9 in Tab. 1 illustrate, the chance for smooth integration into one pot is high. On the other hand, it becomes also clear that none of these methods – to the best knowledge of the authors 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 26
- has ever been realized on more than lab scale. This suggests that many rather fundamental problems still remain to be discovered and solved, as indicated by the glycerol phosphate oxidase inactivation problem that appeared when the step from L-glycerol 3-phosphate was systematically investigated for scale up [37]. A lack in fundamental investigations regarding the scalability of the various procedures is also reflected in the rather poor yields, which in many cases have simply not been optimized. However, a major drawback remains the complex nature of the product mixtures. The desired product, be it DHAP, the phosphorylated, or even the dephospharoylated sugar, is never the most abundant species in the final mixture. This is particularly pronounced for the processes that utilize cheap phosphorylation protocols, which typically requires transesterification conditions and substantial excess of either pyrophosphate, dihydroxyacetone, or glycerol. Viewed from such a process perspective, the ATP-dependent phosphorylations of dihydroxyacetone or glycerol appear still to be the most promising methods. Even though the selected ATP regeneration process intrinsically leads to the accumulation of either acetate or pyruvate, this accumulation can most likely be kept to stoichiometric levels, which leaves the composition of the product mixture relatively simple, and, after dephosphorylation, the separation of the charged products from the sugar should be relatively easy to achieve.
2. Chemical and enzymatic routes to dihydroxyacetone phosphate 27
References Routeb) No. of No. overall # of max. # phosphate reported yieldb), g) composition of final mixtureb), stepsc) of enzymese) of donorb) scalef) h) potsd) enzymes simult. [75, 76] DHA to DHAP 1 1 2 (GK, AcK) 2 ATP, recycled 1.4 L or 66 g 98 % on DHA, 87 % on AcP 280 mM DHAP, 320 mM Ac, 40 mM Pi from AcP [65] DHA to DHAP via PDHA 2 2 2 (PLD, PLC) 1 PC 9 mLor 0.3 g 3 % on DHA, 52 % on PC aq: 42 mM DHAP; org: 42 mM DAG, 16 mM PC j) [61] DHA to P-sugar via DHAP 1 1 3 (DHAK, AcK, 3 ATP, recycled 30 mL or 0.1 g 84 % on DHA, 84 % on Ald, 30 mM P-sugar, 36 mM Pi, ALDase) from AcP 42 % on AcP 5 mM DHAP, 66 mM acetate [20] DHA to P-sugar via DHAP 1 1 4 (GK, PK, TPI, 4 ATP, recycled 200 mL or 5.1 g 40 % on DHA, 60 % on PEP 75 mM P-sugar 1, 12 mM P-sugar 2, ALDase) from PEP 125 mM DHA, 75 mM DHAP
[63] DHA to sugar via DHAP 1 1 2 (AP, ALDase) 2 PPi 10 mL or 0.2 g 95 % on ald (53 % isolated 95 mM sugar, 1.4 M Pi, 400 mM DHA and P-sugar yield), 13% on PPi, 19% on DHA [69] Gly to LGlyP 1 1 1 (CIP) 1 PPi 50 mL or 0.7 g 1 % on gly, 55 % on PPi (45 % 83 mM LGlyP, 8.1 M Gly, 217 mM Pi after isolation) [37] Gly to DHAP via LGlyP 2 1 4 (GK, PK, GPO, 2 ATP, recycled 350 mL or 6 g 84 % on Gly, 84 % on PEP 95 mM DHAP, 3 mM Pi, 103 mM Pyr, CAT) from PEP 3 mM glycerol, 5 mM Pi [75-78] Gly to P-Sugar via LGlyP 2 1 5 (GK, AcK, GPO, 3 ATP, recycled 10 mL or 0.2 g 90 % on Gly, 78 % on Ald, 93 mM P-sugar, 26 mM Ald, and DHAP CAT, ALDase) from AcP 41 % on AcP 7 mM LGlyP, 3 mM Gly, 220 mM Ac, 130 mM Pi [79] Sucrose to P-sugar via 2 1 8 (INV, XI, HK, 7 ATP, recycled 200 mL or 1 g 92% on sucrosesucrose, 8 % 18 mM P-sugar, 107 mM Pi, 107 mM Pyr, DHAP GI, FPK, ALDase, via PEP on PEP, 7 % on Ald 220 mM Ald TPI, PK) [80] FDP to P-sugar via DHAP 2 1 2 (ALDase, TPI) 2 FDP 3.75 L, 400 g 73 % on FDP, 49 % on Ald 50 mM FDP, 310 mM Ald, 290 mM P-sugar [81-83] Gly to sugar via rac-GlyP, 4 1 4 (PHY, GPO, 2 PPi 10 mL or 0.1 g 0.4 % on Gly, 39 % on PPi, 34 mM sugar, 7.5 M gly, 87 mM Pi, DHAP, and P-sugar CAT, ALDase) (29% isolated yield), 33.5 % 9 mM DHAP, 66 mM Ald on Ald [66] rac-Glycidol to sugar via 4 1 4 (GPO, CAT, 2 Pi 10 mL or 0.3 g 28 % on glycidol, (13% 140 mM sugar, 360 mM Gly, 500 mM rac-GlyP, DHAP, and P- ALDase, AP) isolated yield), 28 % on Pi, Pi, 60 mM Ald sugar 70 % on Ald Table 1: Enzymatic routes to DHAP b) Ac: Acetate; AcP: Acetyl phosphate; Ald: Aldehyde (together with DHAP substrate of the aldolase reaction); DAG: Diacylglycerol; DHA: Dihydroxyacetone; DHAP: Dihydroxyacetone phosphate; FDP: Fructose 1,6-bisphosphate; Gly: Glycerol ; GlyP: glycerol 3-phosphate; LGlyP: L-Glycerol 3-phosphate; Pi: Inorganic phosphate; PC: Phosphatidylcholine; PEP: Phosphoenol pyruvate; PDHA; Phosphatidyldihydroxyactone; PPi: Pyrophosphate; P-sugar: Monophosphorylated sugar (product of aldolase reaction of aldehyde and DHAP); Pyr: Pyruvate; sugar: Dephosphorylated P-sugar; rac-PGly: Racemic glycerol phosphate. c) Clearly distinct processing steps; addition of enzyme b after the reaction catalyzed by a is finished would be counted as two steps. Repeated addition of one agent (simulating a constant feed) was counted as one step. d) Indicates whether a purification step needs to be performed inbetween. e) AcK: Acetyl kinase; ALDase: Aldolase; AP: Acid phosphatase; CAT: Catalase; CIP: Calf intestinal alkaline phosphatase; DHAK: Dihydroxyacetone kinase; FPK: Fructose 6-phosphate kinase; GI: Glucose 6-phosphate isomerase; GK: Glycerol kinase; GPO: L-Glycerolphosphate oxidase; HK: Hexokinase; INV: Invertase; PHY: Phytase; PK: Pyruvate kinase; PLC: Phospholipase C; PLD: Phospholipase D; TPI: Triosephosphate isomerase; XI: Xylose isomerase. f) Volume of reaction and mass of produced product. When there was more than one step, the volume of the final step is mentioned. g) Where the yield was a function of time, the maximum yield was chosen. Isolated yields based on the substrate mentioned are reported when available. h) The compositions have been deduced from the yields and under the assumptions that [i] at the end of the reaction all phosphate donors such as AcP, PPi, PEP, have been consumed, and [ii] that DHAP does not degrade to methylglyoxal. Buffer components (on the assumption that in a real process pH would kept constant by addition of acid or base), enzymes, ATP, and salts deriving from pH switches are not considered. When there were several examples for one route, the example with the larger scale has been selected. j) Two-step and two phase reaction. The product was isolated after the first reaction. The phases of the second reaction are listed separately as aqueous (aq) and organic (org). 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 28
2.6. Alternatives to DHAP
The difficulties in DHAP production discussed above should have motivated the investigation of alternative approaches that would on the one hand allow exploiting the promising stereochemical possibilities of aldolases, but on the other hand avoid the difficulties connected to the production of DHAP. Two principal options can be considered: [i] using the available set of enzymes but using a donor-analogue that is easier to produce; or [ii] looking for similar enzymes that allow for a similarly attractive stereochemistry but do not require the phosphorylated donor. However, as already indicated in the beginning, the DHAP-dependent aldolases do not tolerate substantial modifications of the donor. Still, some modifications are possible, in particular at C1. This includes the use of phosphonates instead of a phosphate ester, which leads to a roughly ten-fold reduction in relative activity. Motivated by the potential of phosphonates to act as inhibitors for e.g. glycolysis in human parasites, the 4-hdydroxy-3-oxobutylphosphonate as the phosphonate analogue of DHAP was also investigated as a donor in other DHAP-dependent aldolases. In fact, two of the four available stereochemistries could be produced with this donor (with FruA and RhuA variants, see Fig. 1) [84]. But as this donor leads to sugars that cannot be dephosphorylated easily and the production does not seem to give advantage over production routes to DHAP, transfer of this approach into the area of preparative sugar synthesis is not helpful. Another example for tolerated modifications at C1 is the substitution of the phosphoester of dihydroxyacetone by an arsenate ester. The arsenate ester forms spontaneously, can then be converted by the aldolase, and also hydrolyzes again spontaneously. In fact, using the arsenate method, a number of unphosphorylated sugars have been prepared [85]. However, this method suffers from the required high concentrations of the highly toxic inorganic arsenate, which makes it highly unlikely that this method will be used preparatively. Modifications at C3 of DHAP are even less tolerated, and those modifications with which some residual activity has been detected typically lead to an up to 1000-fold reduction in relative activity [74]. Regarding the option of alternative enzymes, sources for those could be the environment or variants of the existing set of enzymes, obtained by directed evolution. However, the phosphate binding site for the donor is the most conserved element in the family of ( / )8-barrel proteins [86], to which the DHAP-aldolases belong, indicating that it might be a rather difficult undertaking to eliminate the donor-phosphate requirement by directed evolution. Furthermore, the phosphate group of DHAP has been suspected to play an important role in the catalytic mechanism of the aldolase 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 29 catalyzed reaction [87]. Correspondingly, no reports on the elimination of the phosphate requirement for the donor are available (to the best knowledge of the authors), although the preference for specific phosphorylated acceptors could be reduced for at least D-2-keto-3-deoxy-6-phosphogluconate aldolase [88]. With respect to novel enzymes, a fructose 6-phosphate aldolase (FSA) that accepts dihydroxyacetone instead of DHAP as the donor has recently been discovered in E. coli [89]. Next to its natural acceptor glyceraldehyde 3-phosphate (yielding fructose 6-phosphate instead of the bisphosphorylated fructose 1,6-bisphosphate), it also accepted non-phosphorylated acceptors and produced the same stereochemistry as FruA (3S,4R) [90]. Even though this is a very promising result, so far only this stereochemistry is available for dihydroxyacetone-dependent aldolases, which limits the usefulness of the approach. Finally, it has to be mentioned that a number of successful attempts were made to produce catalytic antibodies that do not rely on phosphorylated donors, can mimic aldolase enzymes, and are useful on preparative scale [91]. For the time being, these alternative routes do not possess the chemical versatility represented by the set of DHAP-dependent aldolases and it might still take quite some time before DHAP as a strategically important intermediate can be substituted.
2. Chemical and enzymatic routes to dihydroxyacetone phosphate 30
2.7. Conclusion
The synthetic potential of DHAP-dependent aldolases is so attractive that a variety of chemical and enzymatic routes have been devised to produce DHAP, either as the product itself, a precursor for it, or as an intermediate in a reaction sequence that continues, preferably until the unphosphorylated sugar. Still, it remains doubtful whether most of the routes above can indeed serve as the reliable basis of a scalable route to DHAP. Many of them have not advanced beyond small scale, suggesting that for example high enzyme concentrations might have masked effects that will not become obvious before influence of the various process parameters has been investigated in more detail. The chemical routes suffer, above all, from the requirement for toxic and in part expensive chemicals, which makes it doubtful that any of the routes discussed here will be further developed. Consequently attention focuses on the enzymatic routes, where no such problems exist. These routes are attractive as they are all one-pot routes, and some of them even one-step routes to the phosphorylated sugar product. In principle, this should be an excellent foundation to start a useful preparative route. Nevertheless, a considerable number of issues remain: With most of the enzymatic routes, the final concentration of product is relatively low, and the respective product is part of a rather involved mixture of compounds. Part of this might be due to the fact that only little work for optimization has been invested for most routes, so the true potential for scale up is difficult to assess. Where the degree of integration has remained limited and established ATP-regeneration systems have been applied, yields and composition of final mixtures were promising, but concentrations remained low [37, 59]. Another point to consider for the enzymatic routes is the number of enzymes in one route. Even though the different enzymes are available commercially to some extent or can be relatively easily overproduced in recombinant organisms, the multi-enzyme nature of the various routes remains a serious issue: providing between two and eight enzymes for the manufacturing of an intermediate of a synthesis simply might be too many for a sustainable synthesis. Of course, such a problem might be circumvented in the future by the construction of tailored strains. Finally, the utilization of phosphoenol pyruvate as phosphate donor on industrial scale is currently rather unlikely due to its price and limited availability. Acetyl phosphate can function as a substitute. Still, the integration of novel technologies such as polyphosphate-dependent ATP regeneration might facilitate further the syntheses, as this would introduce a cheap, directly available phosphate donor but avoid introducing 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 31 reverse hydrolysis conditions with the corresponding excess of some substrates. Of course, eliminating the requirement for ATP-regeneration altogether by recruiting the missing three stereochemistries for dihydroxyacetone-dependent enzymes might be the ultimate solution here. It remains unclear how the unavoidable requirement to go to higher product concentrations will influence the operation of the multi-enzyme systems. So far, the achieved product concentrations have exceeded the physiological range somewhat, but for a truly attractive synthesis, this needs to be improved. This will exceed the concentrations for which enzymes are typically investigated, and consequently there might be a number of so far undiscovered consequences. Finally, no adaptations aimed at process improvements has been reported with the enzymes involved in the different syntheses, for example by directed evolution or rational enzyme engineering approaches. Consequently, here might lie quite some potential for process improvement. In summary, a rather broad foundation of possible enzymatic routes to DHAP is available. Urgent issues to address are the design of engineered strains that eliminate problem of multi-enzyme system assembly, the increase of the product concentrations and the optimization of the composition of the final mixture for purification.
2.8. Acknowledgment
Michael Schümperli gratefully acknowledges funding from the EU-NEST project “EUROBIOSYN”.
2. Chemical and enzymatic routes to dihydroxyacetone phosphate 32
2.9. References
1. Gijsen, H.J.M., et al., Recent Advances in the Chemoenzymatic Synthesis of Carbohydrates and Carbohydrate Mimetics. Chem. Rev., 1996. 96(1): p. 443-474. 2. Wong, C.-H., et al., Enzymes in organic synthesis: application to the problems of carbohydrate recognition (Part 1). Angewandte Chemie International Edition in English, 1995. 34: p. 412-432. 3. Mahmoudian, M., et al., An efficient process for production of -acetylneuraminic acid using -acetylneuraminic acid aldolase. Enzyme and Microbial Technology, 1997. 20(5): p. 393-400. 4. Liu, J.Q., et al., Gene cloning and overproduction of low-specificity d-threonine aldolase from Alcaligenes xylosoxidans and its application for production of a key intermediate for parkinsonism drug. Applied Microbiology and Biotechnology, 2000. 54(1): p. 44-51. 5. Panke, S. and M. Wubbolts, Advances in biocatalytic synthesis of pharmaceutical intermediates. Current Opinion in Chemical Biology, 2005. 9(2): p. 188-194. 6. Fessner, W.D., et al., The Mechanism of Class II, Metal-Dependent Aldolases. Angew Chem Int Ed, 1996. 35(19): p. 2219-2221. 7. Zelic, B., et al., Process strategies to enhance pyruvate production with recombinant Escherichia coli: From repetitive fed-batch to in situ product recovery with fully integrated electrodialysis. Biotechnology and Bioengineering, 2004. 85(6): p. 638-646. 8. Brockamp, H.P. and M.R. Kula, Staphylococcus carnosus aldolase as catalyst for enzymatic aldol reactions. Tetrahedron Letters, 1990. 31(49): p. 7123-7126. 9. Fessner, W.-D., Sinerius, G., Diastereoselective Enzymatic Aldol Additions: L-Rhamnulose and L-Fuculose 1-Phosphate Aldolases from E. coli. Angew Chem Int Ed, 1991. 30(5): p. 555-558. 10. Fessner, W.-D. and O. Eyrisch, One-pot synthesis of tagatose 1,6-bisphosphate by diastereoselective enzymatic aldol addition. Angewandte Chemie-International Edition in English, 1992. 31: p. 56-58. 11. Ozaki, A., et al., Overproduction and substrate specificity of a bacterial fuculose-1- phosphate aldolase: a new enzymic catalyst for stereocontrolled aldol condensation. J. Am. Chem. Soc., 1990. 112(12): p. 4970-4971. 12. Wong, C.H. and G.M. Whitesides, Synthesis of sugars by aldolase-catalyzed condensation reactions. Journal of Organic Chemistry, 1983. 48(19): p. 3199-3205. 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 33
13. Schoevaart, R., F. van Rantwijk, and R.A. Sheldon, Stereochemistry of nonnatural aldol reactions catalyzed by DHAP aldolases. Biotechnology and Bioengineering, 2000. 70(3): p. 349-352. 14. Fessner, W.-D., Enzyme mediated C---C bond formation. Current Opinion in Chemical Biology, 1998. 2(1): p. 85-97. 15. Rowan, A.S. and H.C. J., Recent developments in preparative enzymatic syntheses of carbohydrates. Nat Prod Rep, 2006. 23: p. 412-443. 16. Silvestri, M.G., et al., Asymmetric Aldol Reactions using Aldolases, in Topics in Stereochemistry, S.E. Denmark, Editor. 2003, John Wiley & Sons Inc. p. 267-341. 17. Whalen, L.J. and C.H. Wong, Enzymes in organic synthesis: Aldolase-mediated synthesis of iminocyclitols and novel heterocycles. Aldrichimica Acta, 2006. 39(3): p. 63-71. 18. Ardao, I., et al., One step purification-immobilization of fuculose-1-phosphate aldolase, a class II DHAP dependent aldolase, by using metal-chelate supports. Enzyme and Microbial Technology, 2006. 39(1): p. 22-27. 19. Durany, O., C. de Mas, and J. Lopez-Santin, Fed-batch production of recombinant fuculose-1-phosphate aldolase in E. coli. Process Biochemistry, 2005. 40(2): p. 707- 716. 20. Eyrisch, O., G. Sinerius, and W.-D. Fessner, Facile enzymic de novo synthesis and NMR spectroscopic characterization of -tagatose 1,6-bisphosphate. Carbohydrate Research, 1993. 238: p. 287-306. 21. Garcia-Junceda, E., et al., A New Strategy for the Cloning, Overexpression and One Step Purification of Three DHAP-Dependent Aldolases: Rhamnulose-1-Phosphate Aldolase, Fuculose-1-Phosphate Aldolase and Tagatose-1,6-Diphosphate Aldolase. Bioorg Med Chem, 1995. 3(7): p. 945-953. 22. Vidal, L., et al., High-level production of recombinant His-tagged rhamnulose 1- phosphate aldolase in Escherichia coli. Journal of Chemical Technology & Biotechnology, 2003. 78(11): p. 1171-1179. 23. Von der Osten, C.H., et al., Use of a recombinant bacterial fructose-1,6- diphosphate aldolase in aldol reactions: preparative syntheses of 1-deoxynojirimycin, 1-deoxymannojirimycin, 1,4-dideoxy-1,4-imino-D-arabinitol, and fagomine. J. Am. Chem. Soc., 1989. 111(11): p. 3924-3927. 24. Cooper, S.J., et al., The crystal structure of a class II fructose-1,6-bisphosphate aldolase shows a novel binuclear metal-binding active site embedded in a familiar fold. Structure, 1996. 4(11): p. 1303-1315. 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 34
25. Dreyer, M.K. and G.E. Schulz, The Spatial Structure of the Class II -Fuculose-1- phosphate Aldolase from Escherichia coli. Journal of Molecular Biology, 1993. 231(3): p. 549-553. 26. Hall, D.R., et al., Structure of tagatose-1,6-bisphosphate aldolase - Insight into chiral discrimination, mechanism, and specificity of class II aldolases. Journal of Biological Chemistry, 2002. 277(24): p. 22018-22024. 27. Kroemer, M., I. Merkel, and G.E. Schulz, Structure and catalytic mechanism of L-rhamnulose-1-phosphate aldolase. Biochemistry, 2003. 42(36): p. 10560-10568. 28. Esders, T.W. and C.A. Michrina, Purification and properties of L-alpha- glycerophosphate oxidase from Streptococcus faecium ATCC 12755. J. Biol. Chem., 1979. 254(8): p. 2710-2715. 29. Operdoes, F.R., Compartmentation of carbohydrate metabolism in Trypanosomes. Annual Review of Microbiology, 1987. 41: p. 127-151. 30. Sun, J., et al., Comparative Genomic Analysis of dha Regulon and Related Genes for Anaerobic Glycerol Metabolism in Bacteria. Biotechnol. Prog., 2003. 19(2): p. 263-272. 31. Erni, B., et al., Small substrate, big surprise: fold, function and phylogeny of dihydroxyacetone kinases. Cellular and Molecular Life Sciences (CMLS), 2006. 63(7): p. 890-900. 32. Cheek, S., et al., A comprehensive update of the sequence and structure classification of kinases. BMC Structural Biology, 2005. 5(1): p. 6. 33. Chen, Y.M., Y. Zhu, and E.C.C. Lin, The organization of the fuc regulon specifying l-fucose dissimilation in Escherichia coli K12 as determined by gene cloning. Molecular and General Genetics MGG, 1987. 210(2): p. 331-337. 34. Sawada, H. and Y. Takagi, The metabolism of -rhamnose in Escherichia coli : III. - rhamnulose-phosphate aldolase. Biochimica et Biophysica Acta (BBA) - Specialized Section on Enzymological Subjects, 1964. 92(1): p. 26-32. 35. Nobelmann, B. and J.W. Lengeler, Molecular analysis of the gat genes from Escherichia coli and of their roles in galactitol transport and metabolism. J. Bacteriol., 1996. 178(23): p. 6790-6795. 36. Richard, J.P., Kinetic parameters for the elimination reaction catalyzed by triosephosphate isomerase and an estimation of the reaction's physiological significance. Biochemistry, 1991. 30(18): p. 4581-4585. 37. Hettwer, J., H. Oldenburg, and E. Flaschel, Enzymic routes to dihydroxyacetone phosphate or immediate precursors. J Mol Catal B: Enzymatic, 2002. 19-20: p. 215- 222. 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 35
38. Suau, T., et al., Influence of secondary reactions on the synthetic efficiency of DHAP-aldolases. Biotechnology and Bioengineering, 2006. 93(1): p. 48-55. 39. Phillips, S.A. and P.J. Thornalley, The formation of methylglyoxal from triose phosphates. Eur J Biochem, 1993. 212: p. 101-105. 40. Cooper, R.A., Metabolism of methylglyoxal in microorganisms. Ann Rev Microbiol, 1984. 38: p. 49-68. 41. Colbran, R.L., et al., A synthesis of dihydroxyacetone phosphate from dihydrxyacetaone. Carbohydrate Research, 1967. 4: p. 355-358. 42. Effenberger, F. and A. Straub, A Novel Convenient Preparation of Dihydroxyacetone Phosphate and its use in Enzymatic Aldol Reactions. Tetrahedron Lett, 1987. 25(15): p. 1641-1644. 43. Pederson, R.L., J. Esker, and C.H. Wong, An improved synthesis of dihydroxyacetone phosphate. Tetrahedron, 1991. 47(14-15): p. 2643-2648. 44. Jung, S.H., et al., An Efficient Multigram-Scale Preparation of Dihydroxyacetone Phosphate. Journal of Organic Chemistry, 1994. 59(23): p. 7182-7184. 45. Ballou, C.E. and H.O.L. Fischer, The Synthesis of Dihydroxyacetone Phosphate. Journal of the American Chemical Society, 1956. 78(8): p. 1659-1661. 46. Valentin, M.L. and J. Bolte, A convenient synthesis of DHAP from acetone. Bulletin de la Societe Chimique Francais, 1995. 132: p. 1167-1171. 47. Gefflaut, T., et al., A novel efficient synthesis of dihydroxyacetone phosphate and bromoacetol phosphate for use in enzymatic aldol syntheses. Journal of Organic Chemistry, 1997. 62(17): p. 5920-5922. 48. Charmantray, F., et al., Improved straightforward chemical synthesis of dihydroxyacetone phosphate through enzymatic desymmetrization of 2,2-dimethoxypropane-1,3-diol. Journal of Organic Chemistry, 2004. 69(26): p. 9310-9312. 49. Ferroni, E.L., et al., A Three-Step Preparation of Dihydroxyacetone Phosphate Dimethyl Acetal. J. Org. Chem., 1999. 64(13): p. 4943-4945. 50. Meyer, O., et al., Lewis acid mediated regioselective ring opening of benzylglycidol with dibenzyl phosphate: Short and attractive synthesis of dihydroxyacetone phosphate. Organic Letters, 2006. 8(19): p. 4347-4350. 51. Meyer, O., M. Rohmer, and C. Grosdemange-Billiard, Short and efficient synthesis of a stock material of dihydroxyacetone phosphate from glycidol. Tetrahedron Letters, 2004. 45(42): p. 7921-7923. 52. Zhao, H. and W.A. van der Donk, Regeneration of cofactors for use in biocatalysis. Current Opinion in Biotechnology, 2003. 14(6): p. 583-589. 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 36
53. Wong, C.H. and G.M. Whitesides, Enzymes in Synthetic Organic Chemistry. Tetrahedron Organic Chemistry Series, ed. J.E. Baldwin and P.D. Magnus. 1994, Tarrytown (N.Y.): Pergamon. 370. 54. Hirschbein, B.L., Mazenod, F. P., Whiesides, G. M., Synthesis of Phosphoenolyruvate and its use in Adenosine Triphosphate Cofactor Regenration. J Org Chem, 1982. 47: p. 3765-3766. 55. Noguchi, T. and T. Shiba, Use of Escherichia coli Polyphosphate Kinase for Oligosaccharde Synthesis. Biosci Biotechnol Biochem, 1998. 62(8): p. 1594-1596. 56. Resnick, S.M. and A.J.B. Zehnder, In vitro ATP regeneration from polyphosphate and AMP by polyphosphate:AMP phosphotransferase and adenylate kinase from Acinetobacter johnsonii 210A. Applied and Environmental Microbiology, 2000. 66(5): p. 2045-2051. 57. Shiba, T., et al., Inorganic Polyphosphate and Polyphosphate Kinase: Their Novel Biological Functions and Applications. Biochemistry (Moscow), 2000. 65(3): p. 315-323. 58. Crans, D.C. and G.M. Whitesides, Glycerol kinase: substrate specificity. Journal of the American Chemical Society, 1985. 107(24): p. 7008-7018. 59. Crans, D.C. and G.M. Whitesides, Glycerol kinase: synthesis of dihydroxyacetone phosphate, sn-glycerol-3-phosphate, and chiral analogues. Journal of the American Chemical Society, 1985. 107(24): p. 7019-7027. 60. Itoh, N., Y. Tujibata, and J.Q. Liu, Cloning and overexpression in Escherichia coli of the gene encoding dihydroxyacetone kinase isoenzyme I from Schizosaccharomyces pombe, and its application to dihydroxyacetone phosphate production. Applied Microbiology and Biotechnology, 1999. 51(2): p. 193-200. 61. Sanchez-Moreno, I., et al., Multienzyme system for dihydroxyacetone phosphate- dependent aldolase catalyzed C-C bond formation from dihydroxyacetone. Chem Comm, 2004: p. 1634-1635. 62. Yanase, H., et al., Enzymatic Preparation of [1,3-C-13]Dihydroxyacetone Phosphate from [C-13]Methanol and Hydroxypyruvate Using the Methanol-Assimilating System of Methylotrophic Yeasts. Applied Microbiology and Biotechnology, 1995. 43(2): p. 228-234. 63. Van Herk, T., et al., Simple enzymatic in situ generation of dihydroxyacetone phosphate and its use in a cascade reaction for the production of carbohydrates: Increased efficiency by phosphate cycling. Journal of Organic Chemistry, 2006. 71(16): p. 6244-6247. 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 37
64. Takami, M. and Y. Suzuki, Synthesis of novel phosphatidyldihydroxyacetone via transphosphatidylation reaction by phospholipase D. Biosci Biotechnol Biochem, 1994. 58: p. 2136 - 2139. 65. D'Arrigo, P., et al., Indirect Enzymatic Phosphorylatioin: Preparation of Dihydroxyacetone Phosphate. Chem Comm, 1995: p. 2505-2506. 66. Charmantray, F., et al., An efficient chemoenzymatic route to dihydroxy acetone phosphate from glycidol for the in situ aldolase-mediated synthesis of monosaccharides. Tetrahedron Letters, 2006. 47(19): p. 3261-3263. 67. Schoevaart, R., F. van Rantwijk, and R.A. Sheldon, A four-step enzymatic cascade for the one-pot synthesis of non-natural carbohydrates from glycerol. Journal of Organic Chemistry, 2000. 65(21): p. 6940-6943. 68. Schoevaart, R., F. van Rantwijk, and R.A. Sheldon, Carbohydrates from glycerol: an enzymatic four-step, one-pot synthesis. Chemical Communications, 1999(24): p. 2465-2466. 69. Pradines, A., et al., Enzymatic-Synthesis of Phosphoric Monoesters with Alkaline- Phosphatase in Reverse Hydrolysis Conditions. Tetrahedron, 1988. 44(20): p. 6373- 6386. 70. Fessner, W.D. and G. Sinerius, Synthesis of dihydroxyacetone phosphate (and isosteric analogues) by enzymatic oxidation - sugars from glycerol. Angewandte Chemie-International Edition in English, 1994. 33(2): p. 209-212. 71. Oldenburg, H., Kinteische Untersuchungen und Prozessoptimierung der Produktion von Dihydroxyacetonphosphat mit Hilfe der Glycerinphosphat- Oxidase. 1998, Universität Bielefeld: Bielefeld. 72. Hettwer, J., Kinetik aldolasekatalysierter Aldolkondensationen und Prozessoptimierung der Glycerin-3-Phoshat-Synthese. 1998, Universität Bielefeld: Bielefeld. 73. Fessner, W.-D. and C. Walter, "Artificial metabolisms" for the asymmetric one-pot synthesis of branched-chain saccharides. Angewandte Chemie-International Edition in English, 1992. 31(5): p. 614-616. 74. Bednarski, M.D., et al., Rabbit muscle aldolase as a catalyst in organic synthesis. Journal of the American Chemical Society, 1989. 111(2): p. 627-635. 75. Crans, D.C. and G.M. Whitesides, Glycerol Kinase - Synthesis of Dihydroxyacetone Phosphate, Sn-Glycerol-3-Phosphate, and Chiral Analogs. Journal of the American Chemical Society, 1985. 107(24): p. 7019-7027. 76. Crans, D.C. and G.M. Whitesides, Glycerol kinase: substrate specificity. J. Am. Chem. Soc., 1985. 107(24): p. 7008-7018. 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 38
77. Fessner, W.D. and G. Sinerius, Synthesis of Dihydroxyacetone Phosphate (and Isosteric Analogs) by Enzymatic Oxidation - Sugars from Glycerol. Angewandte Chemie-International Edition in English, 1994. 33(2): p. 209-212. 78. Wong, C.H. and G.M. Whitesides, Synthesis of sugars by aldolase-catalyzed condensation reactions. J. Org. Chem., 1983. 48(19): p. 3199-3205. 79. Fessner, W.-D. and C. Walter, "Artificial metabolisms" for the Asymmetric One-Pot Synthesis of Branched-Chain Saccharides. Angew Chem Int Ed, 1992. 31(5): p. 614- 616. 80. Bednarski, M.D., et al., Rabbit Muscle Aldolase as a Catalyst in Organic Synthesis. Journal of the American Chemical Society, 1989. 111(2): p. 627-635. 81. Schoevaart, R., van Rantwijk, F., Sheldon, R. A., Stereochemistry of Nonnatural Aldol Reactions Catalyzed by DHAP Aldolases. Biotechnol Bioeng, 2000. 70(3): p. 349-352. 82. Schoevaart, R., F. van Rantwijk, and R.A. Sheldon, Carbohydrates from glycerol: an enzymatic four-step, one-pot synthesis. Chem Comm, 1999. 1999: p. 2465-2466. 83. Schoevaart, R., F. van Rantwijk, and R.A. Sheldon, A Four-Step Enzymatic Cascade for the One-Pot Synthesis of Non-natural Carbohydrates form Glycerol. J. Org. Chem, 2000. 65: p. 6940-6943. 84. Arth, H.L. and W.D. Fessner, Practical synthesis of 4-hydroxy-3- oxobutylphosphonic acid and its evaluation as a bio-isosteric substrate of DHAP aldolase. Carbohydrate Research, 1997. 305(3-4): p. 313-321. 85. Drueckhammer, D.G., et al., Reversible and in situ formation of organic arsenates and vanadates as organic phosphate mimics in enzymatic reactions: mechanistic investigation of aldol reactions and synthetic applications. J. Org. Chem., 1989. 54(1): p. 70-77. 86. Nagano, N., C.A. Orengo, and J.M. Thornton, One Fold with Many Functions: The Evolutionary Relationships between TIM Barrel Families Based on their Sequences, Structures and Functions. Journal of Molecular Biology, 2002. 321(5): p. 741-765. 87. Periana, R.A., et al., Does substrate rather than protein provide the catalyst for .alpha.-proton abstraction in aldolase? J. Am. Chem. Soc., 1980. 102(11): p. 3923- 3927. 88. Fong, S., et al., Directed evolution of D-2-keto-3-deoxy-6-phosphogluconate aldolase to new variants for the efficient synthesis of D- and L-sugars. Chem Biol, 2000. 7: p. 873 - 883. 89. Schurmann, M. and G.A. Sprenger, Fructose-6-phosphate Aldolase Is a Novel Class I Aldolase from Escherichia coli and Is Related to a Novel Group of Bacterial Transaldolases. J. Biol. Chem., 2001. 276(14): p. 11055-11061. 2. Chemical and enzymatic routes to dihydroxyacetone phosphate 39
90. Schurmann, M., M. Schurmann, and G.A. Sprenger, Fructose 6-phosphate aldolase and 1-deoxy--xylulose 5-phosphate synthase from Escherichia coli as tools in enzymatic synthesis of 1-deoxysugars. Journal of Molecular Catalysis B: Enzymatic, 2002. 19-20: p. 247-252. 91. Schultz, P.G., J. Yin, and R.A. Lerner, The Chemistry of the Antibody Molecule. Angewandte Chemie International Edition, 2002. 41(23): p. 4427-4437.
3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 41
3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations
Michael Schümperli, Matthias Heinemann, Stephan Gomolka, Anne Kümmel, Sven Panke
MS determined the flux distributions, conducted the mutant construction, the fed- batch cultivations and corresponding homogenizations, SBT experiments including analysis and wrote the manuscript. MH was involved in the discussion of early SBT experiments. SG conducted a number of SBT experiments for a diploma thesis. AK implemented a genome-wide stoichiometric model in Matlab. SP supervised the research.
3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 42
3.1. Abstract
A System of Biotransformations (SBT) is an in vitro multi-enzyme reaction system to carry out complex chemical reaction sequences for the synthesis of valuable target compounds. In this chapter, an SBT largely based on naturally available enzymes of Escherichia coli is established and a protocol to insulate the SBT against competing reactions that would divert intermediates from the reaction chain of the SBT is implemented. As a model system, an SBT for the synthesis of dihydroxyacetone phosphate (DHAP) is investigated. The SBT was developed and applied in two steps: step I consisted of an in silico design part, the construction of the SBT-producing organism, the production of the synthetic system by fermentation and the preparation of the cell-free SBT. Step II consisted of the actual synthesis of the target molecule. DHAP was produced with an initial reaction rate of 0.3 U mg-1 of total protein and a yield of 55% after 6 hours with cofactor regeneration. Optimization of the SBT by further insulating the reaction chain from the metabolism and by changing the experimental conditions led to an SBT producing DHAP with an initial reaction rate of 0.6 U mg-1 and a yield of 72% after 6 hours. The reduction of the amount of ATP, but not of NAD+, from stoichiometric to catalytic concentration reduced the operative life time of the system significantly, indicating inefficient ATP recycling.
3.2. Introduction
A System of Biotransformations (SBT) is an in vitro multi-enzyme reaction system, whose implementation can be much facilitated if it is designed on the existing metabolism of an organism. In a first step, the organism’s metabolism is analyzed, a (part of a) reaction pathway is identified, and the organism is genetically engineered to enable product accumulation. The resulting strain is cultivated to amplify the enzymatic reaction system and the cells are ruptured to eliminate the cell membrane as an effective transport barrier as well as to produce the cell-free extract (CFX) containing the desired enzymatic activities. In a second step, the SBT is used for the synthesis of the target compound in cell-free in vitro system. In order to explore the scope of such an approach, an SBT for the synthesis of dihydroxyacetone phosphate (DHAP) was designed (fig. 1), which plays a central role in the diastereoselective synthesis of vicinal diols [chapter 2]. DHAP is an intermediate in the cell’s glycolysis and only 4 enzymatic steps away from a potential cheap starting 3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 43 material, glucose (Glc), which makes glycolysis an attractive multi-enzyme system for the production of DHAP. In the glycolysis, fructose-1,6-bisphosphate (FBP) is converted to glyceraldehyde 3-phosphate (GAP) and DHAP, which are the starting points of the lower branch of the glycolysis, converting DHAP to GAP through the triose phosphate isomerase (TpiA) and finally into pyruvate (PYR). In the course of this reaction chain, 2 moles of ATP are produced from each mole of GAP. Since 1 mole of Glc effectively produces 2 moles of GAP in regular glycolysis, a total of 4 moles of ATP are produced, of which 2 moles are re-invested for the conversion of glucose to GAP and 2 moles of ATP are available for the energy requirements of the cell. To allow accumulation of DHAP in a cell-free SBT, the TpiA must not be present to allow accumulation of DHAP. Thus, in the current absence of methods to selectively remove TpiA from the SBT after cell homogenization, the CFX used for the development of a DHAP producing SBT has to originate from a mutant strain lacking tpiA. The absence of TpiA will at the same time modify the stoichiometry of ATP production in the lower branch of the glycolysis to 2 moles of ATP per mole of Glc. Usually, E. coli removes an excess of DHAP by decomposition to the toxic methylglyoxal, catalyzed by methylglyoxal synthase or spontaneous reaction [1]. The production of this toxic compound appears to be a paradox. However, in E. coli a detoxification system for electrophiles exists, with which methylglyoxal can be converted to D-LAC. The production of methylglyoxal has been discussed to have a number of different physiological roles [1]. Furthermore, blocking of the glycolysis leads to mutants not able to grow on glucose as sole C-source [2]. The absence of TpiA thereby does not block the glycolysis, but halves the flux through the lower branch, undermining the net ATP production of the pathway. Therefore, elimination of tpiA from the genome can be expected to heavily reduce the cell’s viability [2-5]. Further, tpiA elimination will reduce the stoichiometry of ATP production to 2 moles of ATP per mole of Glc. In a somewhat limited point of view, the glycolysis from Glc to PYR without TpiA can be understood as an isolated DHAP production pathway, in which 2 moles of ATP have to be invested to produce 1 mole of DHAP from 1 mole of Glc, and 2 moles of ATP can be recycled in its lower branch when 1 mole phosphate is provided. In the course of this sequence, NAD+ is reduced in the step from GAP to 1,3-bisphosphoglycerate (DPG) introducing the need to add a second cofactor regeneration system to the SBT. This could be accomplished for example by the addition of lactate dehydrogenase (LDH), which catalyzes the reduction of PYR to lactate (LAC) thereby oxidizing NADH back to NAD+. Thus, in principle, a modified glycolysis can be used as an energetically and redox equivalent balanced DHAP-producing SBT. 3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 44
Of course, activities that cannot be expected to be working in a CFX or which might simply be lacking from the CFX of Glc-grown cells would need to be supplied externally or by genetic engineering. For example, Glc is taken up and phosphorylated via the membrane-located phosphotransferase system (PTS) in E. coli cells, which is expected to be inactive in CFX due to membrane removal and therefore needs to be replaced, for example by addition of a hexokinase (HK) [6, 7]. Further, the LDH of E. coli is normally not produced during aerobic growth on Glc, and thus needs to be supplied (fig. 1) [8]. In contrast to the reduced view of its role adopted above, the glycolysis serves in reality also as a reservoir for important anabolic intermediates. In other words, in any CFX produced from E. coli cells it is necessary to take the presence of a significant number of enzymes into account, whose role is to divert intermediates from the glycolysis into anabolic reactions. Therefore, to ensure a high yield of DHAP, the organism’s metabolism has to be adapted to its subsequent synthetic purpose in the SBT. For those enzymes that are essential for the growth of E. coli on Glc (including TpiA when E. coli is grown on minimal medium), this presents an obvious problem. In the following, a theoretical analysis of the essential enzymes in a given set of growth conditions and then steps towards the implementation of an effective SBT based on an adapted glycolytic pathway are presented.
3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 45
Fig. 1: SBT reaction path for DHAP production and cofactor recycling. Boxed: substrates and products of the reaction system. Underlined: enzymes added to the reaction system. Glc: glucose; G6P: glucose 6-phosphate; F6P: fructose 6-phosphate; FBP: fructose 1,6-bisphosphate; DHAP: dihydroxyacetone phosphate; GAP: glyceraldehyde 3-phosphate; 1,3-DPG: 1,3-diphosphoglycerate; 3PG: 3-phosphoglycerate; 2PG: 2-phosphoglycerate; PEP: phosphoenol pyruvate; PYR: pyruvate; LAC: lactate.
3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 46
3.3. Materials and Methods
Strains E. coli W3110 [9] and two derivatives, the knock-out strains W3110 tpiA::kan and W3110 tpiA cyaA ptsI zwf were used in this study. W3110 tpiA::kan was constructed by phage λ red recombination as described previously [10]. The gene tpiA was replaced by the kanamycin resistance gene of pKD13 using the primers P1 and P4 with the following 40-bp homology sequences up- and downstream of tpiA: 5’-GTTAAGGCGAAGA- GTTAAGGAAAGTAAGTGCCGGATATGA for P1 and 5’-CGTGGAGAATTAAAATGCG- ACATCCTTTAGTGATGGGTAA for P4. W3110 tpiA cyaA ptsI zwf was constructed by P1 phage transduction [11] using lytic phage P1 vir and selection against kanamycin (Calbiochem, San Diego, CA, USA). Between subsequent deletions, the kanamycin resistance gene was removed using the FLP helper plasmid pCP20 as described previously [10]. The single knock-out mutants originated from the Keio collection [12].
Media LB-medium was used for growing precultures and as medium on agar-plates. For fermentations M9-GYE (M9-medium supplemented with 4 g L-1 Glc and 5 g L-1 yeast extract (BD Biosciences, Basel, Switzerland) was used. LB and M9 media were prepared as described elsewhere [13].
SBT-buffer SBT-buffer contained either 100 mM 3-(N-Morpholino)propanesulfonic acid (MOPS) or sodium bicarbonate (NaHCO3) as indicated in the text, 0.84 mM KCl, 5 mM MgCl2 and
1 mM ZnSO4. MOPS-based SBT buffer was adjusted to pH = 7.0 with NaOH. The pH of
NaHCO3-based SBT buffer was left unadjusted at 7.6 – 7.7.
Step I: Preparation of the SBT A preculture of the investigated strain in 5 mL LB was used to inoculate a second preculture of 100 mL in M9-GYE medium, which in turn was used to inoculate a starting volume of 2.9 L of M9-GYE in a 5 L reactor vessel. The feed medium contained
-1 -1 100 g L yeast extract, 50 g L Glc and 22.2 mM MgSO4. The cultures were grown to an -1 OD600 of 18 to 20 corresponding to a cell dry weight (CDW) of 9 - 10 g L [14]. The feed was started upon entering of the culture into stationary phase, indicated by a rising DO value. The feed rate was approximately exponential. 3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 47
The biomass was harvested by centrifugation at 8800 g and 4°C for 10 min and resuspended in SBT buffer (the same used later on in the SBT experiments with the corresponding CFX) in a mass ratio of 2:1 (buffer/wet weight). The cells were ruptured in a high-pressure homogenizer (Haskel, Wesel, Germany) by recycling them over a pressure drop of 1000 bar over the orifice for 1.5 hours. The CFX was obtained by centrifugation at 50’000 g and 4°C for 30 min. The supernatant was divided in aliquots of 7 mL and stored at -80°C. Where required, prefiltered CFX (0.22 μm PVDF filter, Rotilabo syringe filters, Roth, Reinach, Switzerland) was washed 3 times by concentrating and refilling with fresh SBT buffer using 2 mL centrifugational ultrafiltration units with a polyethersulfone membrane with a cut-off at 20’000 Da (Sartorius, Dietikon, Switzerland). Centrifugation was carried out for 1 h at 3’220 g and 4°C.
Step II: Operation of the SBT SBT experiments were performed in jacketed beakers kept at a temperature of 37°C. The experiments were carried out in a volume of 10 mL containing either 10 mg mL-1 or
-1 1 mg mL of total protein concentration ([Prottot]) as determined by a commercial Bradford assay [15] (BioRad, Hercules, CA, USA) with bovine serum albumin (Sigma- Aldrich, Buchs, Switzerland) as the standard. The CFX was diluted to the required
[Prottot] with SBT buffer to a total volume of 9.5 mL. Subsequently, 100 μL of a 1.11 M
+ solution of Na2HPO4 in SBT buffer, 100 μL of a 0.6 M solution of NAD in SBT buffer, 0.6 or 0.06 U hexokinase (HK, amount depending on [Prottot], 1 U corresponding to the amount of enzyme converting 1 mmol of glucose in 1 min in presence of ATP at 25°C and a pH of 7.6.) and 33 or 3.3 U of L-lactate dehydrogenase (LDH) (Roche Diagnostics,
Rotkreuz, Switzerland, again according to [Prottot], 1 U corresponding to the amount of enzyme converting 1 mmol of pyruvate in 1 min at 25°C and a pH of 6.5), were added. The reactions were started by the addition to the equilibrated reactors of 200 μL of a 0.6 M solution of ATP in SBT buffer unless stated otherwise and 100 μL of a 1.11 M solution of Glc in SBT buffer. Samples of 200 μL were regularly removed, rapidly mixed with 200 μL of isopropanol and immediately put on ice. At the end of the experiment, all samples were centrifuged at 21’000 g and 4°C for 10 min. Finally, the supernatant was diluted two-fold in ddH2O and stored on ice or at -80°C for further analysis. Every reactor was operated in duplicate and every sample was analyzed twice. Thus, every data point consists of the arithmetic mean of 4 values.
3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 48
Analysis The Glc concentration was determined by applying a commercial enzymatic kit (R-Biopharm, Darmstadt, Germany). The DHAP concentration was determined as described previously [16] by conversion to glycerol 3-phosphate (G3P) with G3P- dehydrogenase (GDH) (Roche Diagnostics, Rotkreuz, Switzerland) and measurement of the corresponding decrease of NADH at 340 nm. Reaction rates were calculated from concentration vs. time plots and are given in U mg-1 of total protein.
The ATP turnover number (TNATP) was calculated by dividing the maximum DHAP concentration [DHAP]max times 2 by the initial ATP concentration [ATP]0, reflecting the fact that 2 moles of ATP needed to be converted to form 1 mole of DHAP from 1 mole of glucose (eq. 1):
2[DHAP]⋅ max TN ATP = 1 [ATP]0
Determination of the half-life time of DHAP Three jacketed beakers were filled with a 5 mM DHAP solution in SBT buffer to a volume of 10 mL and stirred at 37°C for 8 h. Samples were taken every hour and analyzed for DHAP concentration. The same procedure was used for the determination
-1 of the half-life time of DHAP in presence of CFX ([Prottot] = 10 mg mL ) and phosphate (11.1 mM).
Determination of glycogen, proteolytic activities and RNA in CFX The glycogen content of the CFX was determined using amyloglucosidase (Sigma- Aldrich, Buchs, Switzerland) to degrade glycogen to Glc and subsequent enzymatic analysis for Glc as described above. An amount of 14 U of amyloglucosidase (1 unit is supposed to liberate 1.0 mg (1.8 μmol) of Glc from starch in 3 min at pH = 4.5 and 55 °C)
-1 was added to 1 mL of undiluted CFX ([Prottot] = 46 mg mL ) and kept at 43°C for 4 hours. Controls were performed using 1 mL of SBT buffer instead of CFX, 1 mL of SBT buffer containing 10 mg of glycogen (Sigma-Aldrich, Buchs, Switzerland) and 1 mL of CFX containing 10 mg of glycogen. For the detection of proteolytic activities in the CFX, a colorimetric assay [17] was used.
Azocasein (Sigma-Aldrich, Buchs, Switzerland) was incubated with CFX ([Prottot] = 10 mg mL-1), SBT buffer, or SBT buffer containing 10 mg mL-1 trypsin (JRH Biosciences, Lenexa, KS, USA) at 37°C for 30 and 120 min. In the presence of proteolytic activities, azocasein should be digested, releasing an azo-dye which can be detected spectrophotometrically at 440 nm. 3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 49
To analyze the influence of the presence of RNA on the production of DHAP, 1.5 U of RNase A (RNA as substrate at 25°C, cleaving after pyrimidine nucleotides [18]) (Roche Diagnostics, Rotkreuz, Switzerland) were added to 7 mL of CFX and incubation at 30°C for 4 h. The RNase treated CFX was then washed to remove oligonucleotides and considered RNA free.
Stoichiometric model and flux balance analysis (FBA) Flux balance analysis (FBA) was used to hypothesize on the presence or absence of enzyme activities in the mutant used for SBT production during step I and consequently in the CFX used for the SBT in step II, based on the assumption that a gene which had a flux of zero in a given set of circumstances were not expressed. The flux values were regarded as highly speculative. Therefore, if FBA gave a flux of zero, the enzyme was considered as being inactive and its corresponding gene as not expressed. In case of flux values greater than 0, the enzyme was considered as being active. For this purpose, a genome-scale stoichiometric model for E. coli was used [19]. Optimizations were performed using LINDO API software (Lindo Systems, Chicago, IL, USA) via its Matlab (Mathworks, Natick, MA, USA) interface. To identify the maximal possible number of genes without which the cell still was able to grow, FBA was coupled to a mixed integer optimization:
2 max ∑ yi i 3 st.. maxμ st.. S×= r 0 4 min max 5 ()11−≤≤−∀∈yiir r i () y ii r i gene deletion targets min max 6 riii≤≤ r r ∀ i ∉ gene deletion targets
−∞≤rii ≤ max. uptake ∀ i ∈ exchange fluxes 7 y ∈ 0;1 i {} 8
The binary variable y defines the gene deletions and takes a value of 1 in case the gene is deleted. Eq. 5 restricts the corresponding reaction to have flux 0 and μ is the specific growth rate. If the resulting flux distribution would lead to growth rates below a defined minimal growth rate of 0.2 h-1, the resulting mutant was considered as unable to grow. The medium composition tested in silico for the production of a maximally isolated SBT-path contained water, Glc, metal ions, nitrogen, phosphate and sulfur sources and all amino acids. The medium components were allowed to enter and leave the cell 3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 50 unrestrained. Exchange fluxes for compounds present in the in silico medium were constrained to the influx direction (only positive values allowed). For metabolites that are capable of leaving the cell (i.e. acetate, LAC, carbon dioxide, etc.) the exchange fluxes were constrained to the efflux direction (only negative values allowed). 3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 51
3.4. Results
An SBT based on the enzymes of the glycolysis and the removal of TpiA from the set (fig. 1) should allow the production of maximally 1 mole of DHAP from 1 mole of Glc after investment of 2 moles of ATP. The regeneration of ATP can be achieved by converting the concomitantly produced GAP to PYR by the lower branch of the glycolysis. Finally, the 1 mole of NAD+ required to convert GAP to 1,3-bisphosphoglycerate (DPG) in the ATP regeneration pathway can be recycled by adding a LDH to the SBT. This results ideally in a 10-step synthesis system with the overall reaction equation
Glc + Pi = DHAP + LAC 9
Identification and qualification of theoretically present side reactions As part of the central carbon metabolism the glycolysis is highly interconnected with the cell’s metabolic network. Thus, the central task in designing a cell-free SBT for DHAP production based on the metabolism of E. coli and exploiting glycolytic enzymes is the insulation of the reaction pathway from other enzymatic activities that can be expected to be present in the CFX because they were required during the cultivation (step I). In order to specify these potentially flux-diverting activities, as a first step, all reactions containing one of the SBT’s metabolites as either substrate or product were extracted from a genome-scale metabolic model [19], leading to the identification of 90 reactions for the 12 intermediates of the SBT pathway (fig. 2). These 90 reactions represent the upper limit for the number of known enzymes involved in diverting intermediates from the SBT. The real number of interfering reactions will be smaller for several reasons: Firstly, not all of the gene products indicated in fig. 2 are actually required in step I, and those that are not required were assumed to be indeed absent in the further analysis. Secondly, many of those enzymes that are present will require cofactors or co-substrates that can be eliminated from the CFX by washing or rely on a membrane environment which is absent in the SBT step II. In order to obtain a hypothesis about the reactions required by E. coli during growth, FBA was carried out for the tpiA mutant strain in a hypothetical medium containing Glc as C-source and all proteinogenic amino acids, thus reflecting a medium containing a complex fraction. It has to be emphasized that in such circumstances, the flux values have to be considered as hypothetical because due to the complex fraction, the values obtained from the simulation cannot be confirmed by experiments with labeled 3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 52 substrates, as normally done [20-22]. Metabolic flux analysis is not possible in cultures grown on complex media. In this study, the interest was not focused on the flux values but rather on the binary statement if a flux could be expected for a particular reaction or not. If yes, the enzyme was considered to be active in the CFX used for the SBT, if not, the enzyme was considered to be absent. FBA indicated, according to the assumptions described in the Material and Methods section, that 73 of the 90 reactions extracted from the genome scale metabolic model could be ignored because they had no flux, leaving only 17 enzymatic activities to be considered as potentially disturbing (table 1). Next, of the 17 remaining reactions, 12 either relied on small molecule substrates or co- factors or were membrane bound (table 1). Thus, in CFX from E. coli W3110 tpiA::kan grown on a medium with Glc and all proteinogenic amino acids present, only five side reactions were expected to be active in step II: GpsA, IlvGIH, Pgm, SerA and TktB (table 1). This number is the result of a calculation based on a flux distribution for optimum growth on the medium defined in the Material and Methods section. However, the prediction of essential enzymatic activities in the same medium identified only GpsA, Pgm and TktB. Therefore, according to this theoretical analysis, it should be possible to delete the genes ilvGIH and serA without losing viability in step I, which in turn might improve the ATP recycling in step II. These theoretical considerations suggest that even for the glycolysis, which is part of the highly interconnected central carbon metabolism, there is a reasonable expectation that it can be insulated to a large extent from the remaining metabolic network in CFXs (fig 3). Therefore, we proceeded to implement some of the central modifications in the genome of E. coli W3110 to test these considerations.
3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 53
Fig. 2: SBT reaction chain, including all reactions involving SBT substrate, products and intermediates as extracted from a genome-scale metabolic model. Boxed metabolites are substrate and end products. : one reaction per compound; : multiple reactions per compound.
3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 54
Gene Reaction catalyzed by enzyme encoded by gene
aceE PYR + lipoamide => S-acetyldihydrolipoamide + CO2 alaB PYR + L-glutamate <=> 2-ketoglutarate + L-alanine
aroA PEP + shikimate 3-phosphate <=> 5-enolpyruvyl-shikimate 3-phosphate + Pi
PEP + D-erythrose 4-phosphate + H2O <=> 3-deoxy-D-arabino-heptulosonate aroFGH 7-phosphate + Pi dapA PYR + L-aspartate semialdehyde => 2 H2O + L-2,3-dihydrodipicolinate deoC GAP + acetaldehyde <=> deoxyribose 5-phosphate
eda PYR + CO2 <= oxaloacetate glmS F6P + L-glutamine => D-glucosamine 6-phosphate + L-glutamate gpsA DHAP + NADH <=> glycerol 3-phosphate + NAD+
ilvGIH 2 PYR => 2-acetolactate + CO2 metC PYR + ammonia + L-homocysteine <= cystathionine + H2O pgm G6P <=> Glc 1-phosphate serA 3-PG + NAD+ <=> 3-phospho-hydroxypyruvate + NADH talA GAP + D-sedoheptulose 7-phosphate <=> F6P + D-erythrose 4-phosphate tktB F6P + GAP <=> erythrose 4-phosphate + xylulose 5-phosphate zwf G6P + NADP+ <=> D-gluconolactone 6-phosphate + NADPH + H+
Glc + PEP => G6P + PYR pts requires intact membrane for functionality, thus expected to be non- functional in step II
Table 1: Side reactions of the SBT that were predicted to be operative in E. coli in step I. Substrates or cofactors that can presumably be removed by washing before step II are depicted in italics. Shaded reactions are expected to be active in step II.
Construction of the SBT As a central step towards an efficient DHAP-producing SBT, TpiA had to be eliminated from the CFX. Ideally, this would happen only after SBT step I, facilitating the growth of 3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 55 the strain before. The technologies to do this as part of a subproteomic post-growth switch are currently under development elsewhere (V. de Lorenzo, Centro Nacional de Biotecnología (CNB), Madrid, personal communication). Therefore, a tpiA knockout strain was produced. E. coli strains lacking tpiA were reported to be unable to grow on minimal medium containing Glc as sole C-source, based on both experimental observations [2, 3] and theoretical considerations [4, 5]. Indeed, also in our hands FBA predicted that a mutant lacking tpiA should be unable to grow when only glucose was added to the medium that served as the base of the calculations, assuming that a cell is unable to grow when its growth rate is calculated to be below 0.2. However, adding all proteinogenic amino acids to this in silico medium changed this. Consequently, cultivations of tpiA mutant strains on mineral medium supplemented by yeast extract (M9-GYE) were established and the results of the analysis could be verified. Furthermore, HK to allow Glc phosphorylation and LDH to regenerate NAD+ by PYR reduction were added to tpiA strain-derived CFXs. This was deemed necessary as the PTS system is membrane based and most likely not functional in a cell-free in vitro system and aerobically growing E. coli was not expected to synthesize LDH (see above).
DHAP stability DHAP has been reported to be instable in aqueous solutions with its degradation rate increasing with increasing pH [23]. At conditions of pH = 6.8 and T = 37°C (comparable to those employed in the SBT, see below) a half-live of 3 h has been found [23]. We reinvestigated the stability of DHAP for the composition and conditions (pH = 7.0; T = 37°C) used in SBT experiments either in SBT buffer or in SBT buffer in presence of
CFX and phosphate and found much increased half-lives of t1/2 = 10.5 h and t1/2 = 13.5 h respectively. Apparently, the stability of DHAP is strongly influenced by the composition of the experimental system. In the light of these stabilities, the measured DHAP concentration values were not corrected for degradation.
Realization of a DHAP producing SBT In order to test the indications from the theoretical analysis, E. coli W3110 tpiA was cultivated in M9-GYE medium to produce CFX for implementing an SBT for the production of DHAP. In partial agreement with the indications, DHAP was accumulated to a maximum concentration of 6.4 mM, while the Glc concentration dropped at the time from 11.1 to 0 mM. This corresponded to a yield on Glc of 59% (fig. 3a). No DHAP production was observed when CFX derived from the wild-type strain (E. coli W3110) was used (data not shown), confirming the importance of the tpiA deletion step. 3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 56
Next, the influence of washing the CFX was investigated. As was reasoned above, this step was expected to be required to eliminate a number of enzyme activities from the CFX. Remarkably, no significant difference between the DHAP accumulations in case of washed vs. non-washed CFX could be detected (fig. 3b). The reason for this is not quite clear. Even though the concentrations of metabolites in the cells are generally small [24] and can therefore be expected to be present only in small concentrations at the beginning of phase II, the pools could actually be refilled via Glc conversion. Cofactor concentrations can be higher [24] and thus could be high enough to maintain the activity of side reactions for some time, thus reducing the yield on Glc. On the other hand, it remains unclear which enzymes outside of the SBT have suffered during the preparation of it and which impact this has on the reaction network. This could be compounded by the fact that metabolites and cofactors are rather instable molecules that could have been degraded during CFX preparation and during SBT-operation. As a consequence, the laborious washing procedure therefore was omitted in subsequent experiments.
-1 Fig. 3: DHAP production with CFX derived from E. coli W3110 tpiA, [Prottot] = 10 mg mL , [ATP ] = 11.5 mM. a) Glc start concentration of 11.1 mM and control without glucose. b) Influence of washing on DHAP production: non-washed and washed CFX, Glc start concentration of 11.1 mM.
To verify that DHAP production was due to the added Glc, DHAP accumulation was investigated in the absence of Glc. Unexpectedly, DHAP production was also observed without any additions of glucose (fig. 3a). This was irrespective of whether the CFX had been washed or not (data not shown), suggesting that the unknown substrate for DHAP synthesis in the absence of Glc could not be a small molecule but had to be 3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 57 retained together with the protein fraction during the washing procedure. The roles of glycogen, proteins and RNA were investigated in this context. Glycogen is usually degraded via G6P and is fed directly into the glycolysis. Proteins can be degraded to amino acids, which can be converted to intermediates of the central carbon metabolism. They can be fed either directly or via the Entner-Doudoroff pathway into the pentose phosphate pathway (PPP), from which DHAP synthesis can be achieved via ribose 5-phosphate by transaldolase and transketolase reactions [25]. The PPP is also used in case of RNA degradation, when the RNA-derived nucleosides are cleaved to ribose 1-phosphate, which in turn can be converted to ribose 5-phosphate by phosphopentomutase [25]. For these pathways to play a role in the DHAP-producing SBT, the macromolecules would have to be broken down to the monomers first, which would have been likely to cause a delay in the DHAP production. Indeed, the DHAP production from the unknown substrate was roughly 5 times slower in the beginning than DHAP synthesis from Glc (fig. 3a). Consequently, CFX was incubated with enzymes to break down glycogen or RNA and washed, which should lead to the elimination of the respective monomers. However, neither glycogen nor RNA depletion had an effect on the DHAP production from an unknown substrate. Next, the protein degradation catalyzed by CFX was investigated, but no such degradation could be detected by the applied assay (data not shown). Irrespective of the fact that the source of the DHAP produced in the absence of added Glc remains unidentified for the time being, it must be a part of the CFX. This could be
-1 convincingly shown by the fact that a reduction of [Prottot] by a factor of 10 to 1 mg mL was still sufficient for DHAP production, but marginalized DHAP accumulation in the absence of glucose (fig 4). With Glc added, 6.1 mM DHAP were produced, which is in the same concentration range as the DHAP produced with the SBT with the higher total protein concentration. This corresponded to an uncorrected yield on Glc of 55%, while in the absence of Glc, only 0.7 mM DHAP accumulated.
3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 58
-1 Fig. 4: DHAP production with CFX derived from W3110 tpiA, [Prottot] = 1 mg mL , CFX not washed. Control reaction without addition of Glc.
Reduction of the initial cofactor concentration The experiments shown in figs. 3 and 4 were realized with half-stoichiometric ATP- and NAD+-concentrations (11.5 mM ATP and 5.75 mM NAD+), reflecting the concern that ATP- regeneration, and potentially also NAD+-regeneration, might not operate efficiently enough to allow substantial cofactor regeneration. Ultimately, such conditions are economically unsustainable and the effects of a reduction in cofactor concentrations were investigated. No significant effect of the NAD+ concentration on the DHAP accumulation patterns was observed (fig. 5a). This is in agreement with the consideration that the SBT required NAD+ rather than NADH: Even if the cofactor regeneration did not work, it can be expected that there are sufficient activities present in the CFX that help re-oxidizing the formed NADH. The reduction of the initial ATP concentration to 5.8 mM did not lead to decreased yield and reaction velocity either. However, when the initial concentration of ATP was reduced further to 1.2 mM, the rate of DHAP accumulation was significantly decreased
(fig. 5b). Concomitantly, the calculated minimum ATP turnover numbers (TNATP) increased from 1.1 to 2.1 and 2.5 for 11.5, 5.8 and 1.2 mM initial ATP concentrations respectively. The maximum possible TNATP of the system therefore appeared to be 2.5. Consequently, either the ATP regeneration pathway operated less efficient than anticipated due to removal of intermediates from the reaction path, or the reaction 3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 59 system that was taken into account was too narrow because it neglected undesired activities that lead to ATP degradation such as conversion of ATP to ADP due to unspecific enzyme activities or enzyme activities that are not yet known. These enzymatic ATP degrading activities do not necessarily have to be directly connected to the SBT pathway; nevertheless they might be important for the organism’s viability. Such activities, in particular if many different activities are considered in aggregate, might partly explain the rather low yields.
Fig. 5: DHAP accumulation in conditions of reduced cofactor concentrations. a) DHAP production with different initial NAD+ concentrations. b) DHAP production with
-1 different initial ATP concentrations. [Prottot] = 1 mg mL , CFX non-washed.
Optimization of the DHAP-producing SBT The performance of the SBT was strongly influenced by the composition of the system.
Changing the buffer from MOPS to a carbonate-based buffer, such as NaHCO3, led to an increase in the initial DHAP accumulation rate by more than 60% from 0.3 U mg-1 to
-1 0.5 U mg and in the yield on Glc from 55% to 66% (fig. 6a, b). NaHCO3 was chosen since there was a strong influence of MOPS in mass spectrometry based analytics [chapter 6]. It is unlikely that this improvement was due to the corresponding change in pH from 7.0 in the case of MOPS to 7.6 in the case of NaHCO3-based buffer, as control experiments with MOPS-based buffer had indicated that changing the pH in this range had no influence (data not shown). As we did not investigate the role of single enzymes in this context, it is difficult to pinpoint the mechanism behind this improvement in performance. Next, a couple of straightforward options were investigated that might contribute to a higher DHAP yield on glucose due to an increased insulation of the SBT from the 3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 60 remainder of the metabolic network. Specifically, the genes encoding for G6P dehydrogenase (zwf), PTS enzyme I (ptsI) and adenylate cyclase (cyaA) were deleted. Zwf had been classified as inactive in the SBT step II due to its dependence on NADP+ instead on NAD+ (table 1). However, the literature is in fact not quite conclusive in this respect [26]. It does not exclude that there might be a residual activity with NAD+. This was supported by qualitative MS-analysis of SBT samples that contained 6-phosphogluconate (data not shown). The PTS was considered as inactive in step II, which it most likely is with respect to Glc phosphorylation. However, not all enzymes of the PTS are membrane bound, as exemplified by the soluble PtsI which is involved in dephosphorylating PEP by transfer of Pi onto HPr, another soluble protein of the PTS [6]. Neither the substrate of the reaction catalyzed by PtsI nor the target of phosphorylation therefore can be considered as absent in step II of the SBT. However, if PtsI is functional, it can dephosphorylate PEP and substantially hamper the ATP recycling pathway. Adenylate cyclase (CyaA) catalyzes the conversion of ATP to cAMP, a global regulator. cAMP regulates various functions in E. coli, including carbohydrate uptake, the fructose regulon or glycogen synthesis, functions which have direct influence on the central carbon metabolism [27, 28]. Thereby, a high level of cAMP is found in cells growing on non-PTS C-sources or during starvation [29]. The deletion of the gene responsible for adenylate cyclase, cyaA, could lead to a mutant that cannot easily adapt the metabolism to changing environments by converting ATP into cAMP. In fact, the CFX of an E. coli W3110 tpiA cyaA ptsI zwf strain showed an improved initial reaction rate for DHAP production of 0.6 U mg-1 (fig. 6c), 20% higher than with CFX from W3110 tpiA. Which mutation in the end led to the improvement of the DHAP production is difficult to estimate, since neither the single knock-out mutants nor the various combinations of the gene deletions were tested. Interestingly, irrespective of the buffer system or the number of mutations that had been introduced into the strain from which the CFXs were produced, ATP was not completely degraded (fig. 6a-c). The concentrations even increased towards the end when Glc concentrations were low, clearly indicating that the ATP regeneration was functional. However, this stabilization of ATP concentrations occurred at rather low levels, between 2 and 4 mM. While it is difficult to estimate a reasonable level for the final ATP concentration without a comprehensive mass balance of glucose and the molecules derived from it, it appears likely that a substantial part of ATP has been degraded. This observation also suggests that operating the SBT at ATP concentrations much lower than stoichiometric might be difficult with the current system because ATP 3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 61 would be completely removed from the system. A result in agreement with this prediction was obtained when the initial ATP concentration was lowered to 1.2 mM, 10 times less than the initial Glc concentration (fig. 6d). Here, Glc consumption and DHAP formation ceased quickly after the experiment had been started. Nevertheless, the yield of DHAP with respect to consumed Glc in the experiment was 73% (1.1 mM
DHAP produced, 1.5 mM Glc consumed, TNATP = 1.9), corresponding to the yield with high initial ATP concentration present.
-1 Fig. 6: Optimization of the DHAP-producing SBT. a-c) [Prottot] = 1 mg mL , initial Glc concentration of 11.1 mM, initial ATP concentration of 11.5 mM. d) DHAP production with reduced initial ATP of 1.2 mM.
3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 62
3.5. Discussion
In vitro multi-enzyme reaction systems would be a convenient way to implement complex conversions of cheap starting materials to valuable compounds, in particular in the area of carbohydrate synthesis, where it is often doubtful whether hydrophilic starting materials can cross the cytoplasmic membrane to enter a cell or whether phosphorylated intermediates can exit it. However, such systems are laborious to assemble, so the implementation of such a multi-enzyme system based on the metabolism of E. coli was investigated. The theoretical analysis supported the notion that it should be possible, by a combination of knock-outs and biochemical treatments, to produce a sufficiently insulated reaction path even inside a CFX that in addition to the enzymes of interest contains hundreds of additional enzymes (figs. 2, 3). The experimental investigation of a tpiA deficient strain demonstrated that the theoretical analysis was to a significant extent correct. The 3 main issues of the system, DHAP accumulation using Glc as substrate and regeneration of NAD+ and ATP as cofactors, could be demonstrated. Theoretically, if there were no side reactions and the cofactor recycling pathway would work efficiently enough to account for every mole of dephosphorylated ATP, a yield of DHAP on Glc of 100% should be possible, assuming that the reaction catalyzed by pyruvate kinase (PYK) is irreversible. However, if the assumptions concerning the side reactions are correct, there are 5 reactions that can remove glycolytic intermediates in the upper branch, thus directly lowering the maximally possible yield of DHAP on glucose, and in the lower branch leading to decreased cofactor recycling efficiency and thus to a stop of DHAP production (fig. 7). One of the proposed side reactions, catalyzed by GpsA, utilizes DHAP as substrate to reduce it to G3P. The reaction is dependent on NADH, which is formed in the recycling pathway. Therefore, the reaction could be responsible for a substantial yield reduction. Furthermore, the conversion of DHAP to methylglyoxal (MG) was not taken into account. FBA predicted a flux of zero for the reaction and thus the methylglyoxal synthase was considered inactive. However, flux analysis of tpiA mutants grown on defined media revealed that MG production is substantial [30]. Furthermore, the bacteriocidal MG production was made partly responsible for the slow growth patterns of mutants bearing the tpiA gene deletion [1]. These earlier findings suggest that the possibility that MG synthase is active has to be taken into account. The Km values of both enzymes in E. coli, GpsA and MgsA, are 0.17 [31] and 0.2 mM [32] respectively, indicating that both enzymes could be operating at non-negligible conversion rates. 3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 63
Irrespective of these theoretical considerations, CFX from a tpiA gene deletion mutant produced DHAP with a yield of 55% on glucose, a value that could be increased by optimization to 72%. Therefore, the DHAP production in the SBT was more efficient than could be expected. Qualitative MS analysis of SBT samples revealed that G3P was produced (data not shown). MG could not be detected; however this could also have been due to the small and uncharged nature of the molecule, which makes it difficult to identify. The obvious way to overcome the consumption of DHAP would be the deletion of the genes encoding the two activities. However, FBA classified GpsA as essential and in fact, attempts to delete the corresponding gene, gpsA, failed. According to FBA and the essentiality analysis, there are two enzymatic activities remaining in the DHAP-producing pathway of the SBT performed with CFX from W3110 tpiA cyaA ptsI zwf, which could contribute to reduce the yield of DHAP on glucose: Pgm and TktB. Both reactions are considered as essential. The role these two reactions play in the SBT can best be estimated after the problem of potential DHAP consumption has been addressed. According to our estimation for the cofactor recycling pathway, there are still potentially active side reactions present as well, reducing its efficiency: SerA and IlvGIH. Both enzymatic activities are not considered as essential in cells growing on M9-GYE medium. Therefore, the corresponding genes could be deleted from the genome of the mutant used for SBT production to further enhance the cofactor regeneration efficiency. However, the sensitivity of the SBT to the reduction of the initial ATP concentration was unlikely to be the result of inefficient recycling only. For example, there might be many enzymes present in the CFX that can dephosphorylate ATP and its derivatives. Thus, if secondary ATP consumers are present, ATP is bound to be depleted in a short period of time (fig. 7). This problem is addressed in chapter 4 in more detail. The series of experiments including the development of a DHAP-producing SBT and its optimization clearly proved that the approach is feasible. Furthermore, the DHAP producing SBT still offers potential for improvement by further insulating the reaction pathway or optimization of the experimental conditions.
3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 64
Fig. 7: Factors detrimental for DHAP productivity (in red) in the optimized DHAP- producing SBT with CFX from E. coli W3110 tpiA cyaA ptsI zwf.
3.6. Acknowledgments
Dr. Johannes Bongaerts and Prof. Uwe Sauer are gratefully acknowledged for their help with the construction of the first gene deletion mutant. Financial support from the EU is gratefully acknowledged. The here described research was part of “EuroBioSyn – A Modular Platform for Biosynthesis of Complex Molecules”.
3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 65
3.7. References
1. Ferguson, G.P., et al., Methylglyoxal production in bacteria: suicide or survival? Arch Microbiol, 1998. 170: p. 209-219. 2. Irani, M.H. and P.K. Maitra, Properties of Escherichia coli Mutants Deficient in Enzymes of Glycolysis. J Bact, 1977. 132(2): p. 398-410. 3. Anderson, A. and R.A. Cooper, Gluconeogenesis in Escherichia Coli: The role of Triose Phosphate Isomerase. FEBS Lett, 1969. 4(1): p. 19-20. 4. Edwards, J.S. and B.O. Palsson, The Escherichia coli MG1655 in silico metabolic genotype: Its definition, characteristics, and capabilities. PNAS, 2000. 97(10): p. 5528-5533. 5. Segre, D., D. Vitkup, and G.M. Church, Analysis of optimality in natural and perturbed metabolic networks. PNAS, 2002. 99(23): p. 15112-15117. 6. Postma, P.W., J.W. Lengeler, and G.R. Jacobson, Phosphoenolpyruvate:Carbo- hydrate Phosphotransferase Systems, in Escherichia coli and Salmonella: cellular and molecular biology, A. Böck, et al., Editors. 1996, ASM Press. 7. Fraenkel, D.G., Glycolysis, in Escherichia coli and Salmonella: cellular and molecular biology, A. Böck, et al., Editors. 1996, ASM Press. 8. Böck, A. and G. Sawers, Fermentation, in Escherichia coli and Salmonella: cellular and molecular biology, A. Böck, et al., Editors. 1996, ASM Press. 9. Nieboer, M., J. Kingma, and B. Witholt, The alkane oxidation system of Pseudomonas oleovorans: induction of the alk genes in Escherichia coli W3110(pGEc47) affects membrane biogenesis and results in overexpression of alkane hydroxylase in a distinct cytoplasmic membrane subtraction. Molecular Microbiology, 1993. 8(6): p. 1039-1051. 10. Datsenko, K.A. and B.L. Wanner, One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products. PNAS, 2000. 97(2): p. 6640-6645. 11. Miller, J.H., A short course in bacterial genetics. 1992, Plainview, N. Y.: Cold Spring Harbor Laboratory Press. 12. Baba, T., et al., Construction of Escherichia coli K-12 in-frame, single-gene knockout mutants: the Keio collection. Mol Syst Biol, 2006. 13. Sambrook, J. and D.W. Russel, Molecular Cloning: A Laboratory Manual 3rd ed. 2001. 14. Sauer, U., et al., The Soluble and Membrane-bound Transhydrogenases UdhA and PntAB Have Divergent Functions in NADPH Metabolism of Escherichia coli. J. Biol. Chem., 2004. 279(8): p. 6613-6619.
3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 66
15. Bradford, M.M., A Rapid and Sensitive Method for the Quantitation of Microgram Quantities of Protein Utilizing the Principle of Protein-Dye Binding. Anal Biochem, 1976. 72: p. 248-254. 16. Bergmeyer, H.U., J. Bergmeyer, and M. Grassl, Methods of Enzymatic Analysis. 1984. 6: p. 342-350. 17. Tomarelli, R.M., J. Charney, and M.L. Harding, The use of azoalbumin as a substrate in the colorimetric determination of peptic and tryptic activity. J Lab Clin Med, 1949. 34: p. 428 - 433. 18. Lavallee, D.K. and R.B. Myers, Role of the pyrimidine base in ribonuclease A hydrolysis of RNA. Determination of the conformation of cyclic β-cytidine 2',3'- phosphate and cyclic β-uridine 2',3'-phosphate in solution. J. Am. Chem. Soc., 1978. 100(12): p. 3907-3912. 19. Reed, J.L., et al., An expanded genome-scale model of Escherichia coli K-12 (iJR904 GSM/GPR). Genome Biology, 2003. 4(9): p. R54.1-R54.12. 20. Fischer, E., N. Zamboni, and U. Sauer, High-throughput metabolic flux analysis based on gas chromatography-mass spectrometry derived 13C constraints. Anal Biochem, 2004. 325: p. 308-316. 21. Kleijn, R.J., et al., Metabolic flux analysis of a glycerol-overproducing Saccharomyces cerevisiae strain based on GC-MS, LC-MS and NMR-derived C-13- labelling data. Fems Yeast Research, 2007. 7(2): p. 216-231. 22. Iwatani, S., et al., Determination of metabolic flux changes during fed-batch cultivation from measurements of intracellular amino acids by LC-MS/MS. Journal of Biotechnology, 2007. 128(1): p. 93-111. 23. Hettwer, J., H. Oldenburg, and E. Flaschel, Enzymic routes to dihydroxyacetone phosphate or immediate precursors. J Mol Catal B: Enzymatic, 2002. 19-20: p. 215- 222. 24. Chassagnole, C., et al., Dynamic Modeling of the Central Carbon Metabolism of Escherichia coli. Biotechnol Bioeng, 2002. 79: p. 53-73. 25. Keseler, I.M., et al., EcoCyc: a comprehensive database resource for Escherichia coli. Nucl. Acids Res., 2005. 33(suppl 1): p. D334-D337. 26. Sanwal, B.D., Regulatory Mechanisms Involving Nicotinamide Adenine Nucleotides As Allosteric Effectors. III. Control of glucose 6-phosphate dehydrogenase. J. Biol. Chem., 1970. 245(7): p. 1626-1631. 27. Strozen, T.G., G.R. Langen, and S.P. Howard, Adenylate cyclase mutations rescue the degP temperature-sensitive phenotype and induce the sigma E and Cpx extracytoplasmic stress regulons in Escherichia coli. Journal of Bacteriology, 2005. 187(18): p. 6309-6316.
3. Development of a dihydroxyacetone phosphate-producing System of Biotransformations 67
28. Botsford, J.L. and J.G. Harman, Cyclic AMP in prokaryotes. Microbiol. Mol. Biol. Rev., 1992. 56(1): p. 100-122. 29. Peterkofsky, A., et al., Bacterial Adenylyl Cyclases. Prog Nucl Ac Res Mol Biol, 1993. 44: p. 31-65. 30. Fong, S.S., et al., Latent Pathway Activation and Increased Pathway Capacity Enable Escherichia coli Adaptation to Loss of Key Metabolic Enzymes. J. Biol. Chem., 2006. 281(12): p. 8024-8033. 31. Kito, M. and L.I. Pizer, Purification and Regulatory Properties of the Biosynthetic l- Glycerol 3-Phosphate Dehydrogenase from Escherichia coli. J. Biol. Chem., 1969. 244(12): p. 3316-3323. 32. Schomburg, I., et al., BRENDA, the enzyme database: updates and major new developments. Nucl. Acids Res., 2004. 32(suppl_1): p. D431-433.
4. ATP sinks in cell-free extracts of Escherichia coli 69
4. ATP sinks in cell-free extracts of Escherichia coli
Michael Schümperli, Stefan Makart, Sven Panke
MS conducted the database research, the mutant construction, the fed-batch cultivations and corresponding homogenizations, all ATP stability experiments including analysis and wrote the manuscript. SM performed the batch cultivations and corresponding homogenizations. SP supervised the research.
4. ATP sinks in cell-free extracts of Escherichia coli 70
4.1. Abstract
Systems of Biotransformations (SBTs) are multi-enzyme systems emulating cellular enzyme configurations in following complex chemical routes for converting readily available substrates into useful products. Necessarily, this entails enzyme sets that are involved in regeneration of energy and reducing equivalents, which requires careful identification of activities that could interfere with the pathway insulation and thus reduce either yield or efficiency of cofactor regeneration. When imposing a dihydroxyacetone phosphate (DHAP)-producing SBT on an Escherichia coli proteome, there were strong indications for enzyme-catalyzed ATP depletion in the cell-free extract (CFX). The comprehensive set of genes whose described function suggested that they might be involved in ATP-, ADP-, or AMP-degradation (AXP-degradation) as a main or side activity was identified and CFX was prepared from the corresponding knock-out strains. Knocking out the genes amn, apt, ushA or ygiF from an E. coli W3110 genome had a significant stabilizing effect on AXP concentrations. The most promising candidates to eliminate from a strain optimized for SBT operation appear to be amn and apt.
4.2. Introduction
In chapter 3, the concept of in vitro enzyme systems for the multi-step conversion of cheap starting materials into valuable compounds (System of Biotransformations, SBT) was introduced. Specifically, an SBT produced dihydroxyacetone phosphate (DHAP) from glucose (Glc) using the upper branch of the glycolysis for DHAP synthesis and recycled the spent ATP in the lower branch. 2 moles of ATP were consumed for the conversion of 1 mole of Glc to 1 mole of DHAP and 1 mole of glyceraldehyde 3-phosphate (GAP). Conversion of 1 mole of GAP to pyruvate (PYR) can theoretically regenerate 2 moles of ATP. Ignoring potential intermediate-diverting side activities, the net reaction should therefore be balanced in energy cofactor utilization. In presence of high ATP concentrations, DHAP was produced with a yield of 76% with respect to glucose in six hours. The reduction of the initial ATP concentration led to a decreased reaction velocity and conversion of glucose. The ATP turnover numbers
(TNATP) did not exceed 2.5, indicating that the ATP recycling might not have been working efficient enough to account for all ATP consumed. Apparently, ATP is lost either to chemical degradation or to degradation by dedicated or side activities of enzymes that are also present in the cell-free extract (CFX).
4. ATP sinks in cell-free extracts of Escherichia coli 71
Obviously, many enzymes are involved in AXP degradation, but most of them require another substrate for their normal activity, which cannot be expected to be present in the SBT. However, from the literature, it is clear that a number of enzymes are able to degrade AXPs without the need for a non-glycolytic metabolite or another cofactor. They can interfere with the SBT in several ways: i) by dephosphorylating ATP to ADP, which upsets the cofactor regeneration balance; ii) by dephosphorylating ADP to AMP, effectively removing the molecule from the recycling as there is no obvious possibility to re-convert AMP to ADP or even ATP in the CFX; iii) by converting ATP, ADP or AMP in another way than dephosphorylation, again effectively removing the cofactor from the possibility of reaction and regeneration. In cell-free protein synthesis systems, it was reported that phosphatase and ATPase activities in the cell-free extract dephosphorylate ATP and thus remove it from the reaction chain [1-6]. To prevent ATP depletion, efficient ATP regeneration systems were installed: Glucose 6-phosphate (G6P), acetyl phosphate (AcP), creatine phosphate (CP) or polyphosphate were used as phosphate donors for the ATP regeneration. In addition, a system using phosphoenolpyruvate (PEP) and containing amino acids, NAD+, oxalic acid, spermidine and putrescine, increased the ATP regeneration efficiency [7-9]. Alternatively, cell-free protein synthesis in wheat germ extract was prolonged when phosphatases were removed by immunodepletion: several phosphatase isozymes were removed by an antibody raised against the major isozyme, removing 20 – 40% of ATP hydrolyzing activity from the cell-free extract [5]. In this study, the effect of systematically removing enzymatic activities proven for or suspected of AXP degradation on AXP stability was examined.
4. ATP sinks in cell-free extracts of Escherichia coli 72
4.3. Materials and Methods
Database research The metabolic databases BRENDA [10], EcoCyc [11] and KEGG [12] were used for systematic identification of potential AXP degrading activities.
Strains All strains used in this study were derived from E. coli W3110 [13] and are listed in table 1. W3110 tpiA was constructed as described in chapter 3. W3110 tpiA cyaA and W3110 tpiA cyaA ygiF were constructed by P1 phage transduction as described in chapter 3.
Deletion in E. coli W3110 Eliminated activity Reference tpiA triose phosphate isomerase Chapter 3 tpiA cyaA triose phosphate isomerase, adenylate cyclase This study tpiA cyaA ygiF as above plus putative adenylate cyclase atpD H+-ATPase amn AMP nucleosidase aphA acid phosphatase apt adenine phosphoribosyltransferase Keio collec- hpt guanine phosphoribosyltransferase tion [14] nagD ribonucleotide monophosphatase phoA alkaline phosphatase sgcQ putative nucleoside triphosphatase ushA 5’-nucleotidase Table 1: Mutants used and created in this study.
Media LB-medium was used for growing precultures and as medium on agar-plates. For fermentations, M9-GYE (M9-medium supplemented with 4 g L-1 glucose and 5 g L-1 yeast extract (BD Biosciences, Basel, Switzerland) was used. LB and M9 media were prepared as described elsewhere [15].
4. ATP sinks in cell-free extracts of Escherichia coli 73
Fermentation and extraction Fed-batch fermentations with E. coli W3110 tpiA, tpiA cyaA, tpiA cyaA ygiF and atpD were performed as described in chapter 3. The other mutants were grown in batch fermentations on a scale of 3 L in a 5 L reactor vessel. A preculture of 5 mL grown in LB was used to inoculate a second preculture of 100 mL of M9-GYE which in turn was used to inoculate a starting volume of 2.9 L of M9-GYE. The cultures were grown to an OD600 of 8 – 10 corresponding to a cell dry weight (CDW) of 3 – 4 g L-1 [16]. The biomass was harvested as described in chapter 3. The growth rates of the mutants were identified in the exponential growth phase, in case of the fed-batch cultivations in the batch phase before the start of the feed.
ATP stability experiments Beaker reactors were filled with SBT buffer (100 mM 3-(N-morpholino)propanesulfonic acid (MOPS), 0.84 mM KCl, 5 mM MgCl2 and 1 mM ZnSO4, pH adjusted to 7.0 with -1 NaOH), CFX to a total protein concentration of 10 mg mL and Na2HPO4 solution (pH = 7.0) to a final concentration of 11.1 mM. The experiments were started by the addition of ATP to 11.5 mM. The total volume was 10 mL. Where required, CFX was washed as described in chapter 3.
Analysis Glc and DHAP concentrations were determined as described in chapter 3. ATP, ADP and AMP concentrations were determined as described previously [17] using glucose 6-phosphate dehydrogenase (G6P-DH), myokinase (MK), NADH (Sigma-Aldrich, Buchs, Switzerland), hexokinase (HK), L-lactate dehydrogenase (LDH), pyruvate kinase (PK) (Roche Diagnostics, Rotkreuz, Switzerland) and NADP+ (GERBU Biochemicals GmbH, Gaiberg, Germany). ATP was determined by the phosphorylation of glucose to G6P with HK and the subsequent oxidation to 6-phosphogluconolactone by G6P-DH with concomitant NADPH generation, which was followed spectrophotometrically at 340 nm. ADP was determined after phosphorylation to ATP by PK and phosphoenolpyruvate (PEP) and its subsequent reduction to lactate (LAC) by LDH with concomitant consumption of NADH, followed spectrophotometrically at 340 nm. AMP was determined likewise, after conversion to ADP by MK and concomitant ATP consumption. The ATP turnover number (TNATP) was calculated as described in chapter 3. AXP depletion patterns are given as percentages. The sum of the concentrations of ATP, ADP and AMP at the beginning of the experiment (time zero) was taken as 100%.
4. ATP sinks in cell-free extracts of Escherichia coli 74
4.4. Results
Fundamentally, enzyme-catalyzed ATP degradation in CFXs can be the result of two different phenomena: it can either be dominated by one or a few enzymes that are present in the CFX, or it can be the result of many specific or unspecific enzymatic activities. While the latter situation would be rather difficult to modify and might ultimately require the use of larger amounts of energy cofactors in the system or a switch to reaction systems consisting of purified constituents, the former situation could potentially be readily mitigated by eliminating the responsible enzymes, either by knocking-out the corresponding genes or by developing strategies to specifically remove the gene products from phase II of the SBT. Consequently, we started to systematically investigate CFXs of strains bearing deletions of genes encoding for enzymes that could be suspected of playing major roles in ATP degradation in CFXs. Such systematic schemes are much facilitated by the availability of a comprehensive collection of knock-out strains such as the Keio-collection [14], that was used in the experiments. For reference, ATP was incubated with and without washed and unwashed CFX derived from E. coli W3110 tpiA. While ATP was stable in SBT buffer in the absence of CFX over a period of 5 h at 37°C, it quickly degraded in the presence of CFX obtained from W3110 tpiA, with the initial concentration dropping by 8.2 mM in only one hour. The degradation rate was only slightly reduced by washing the CFX before the experiment (fig. 1). Consequently, ATP degradation is due to the presence of the CFX.
Fig. 1: ATP stability in SBT buffer, in the absence and presence of washed and unwashed CFX derived from E. coli W3110 tpiA.
4. ATP sinks in cell-free extracts of Escherichia coli 75
Next, the transformations that took place in the system were investigated. ATP degradation in unwashed CFX led quickly to the formation of ADP, indicating dephosphorylation of ATP as a strong contribution to ATP degradation. After that, the formation of AMP could be observed, either from the dephosphorylation of ADP or from the removal of a pyrophosphate from ATP. At the same time, the sum of the concentrations of AXPs declined substantially by 83%, indicating that AXPs were also degraded by other mechanisms (fig. 2, tpiA).
Gene Reaction Comment
Amn AMP + H2O = R5P + Ade Active in Pi limiting growth conditions [18]
Active independent of Pi-content of the AphA AXP + H2O = A(X-1)P + Pi medium [19]
Active in purine-rich growth environments Apt AMP + PPi = Ade + 5PR1P [20]
AtpD ATP + H2O = ADP + Pi Gene constitutively expressed [21] Active in absence of glucose in growth CyaA ATP = cAMP + PPi medium [22]
Gene expression cAMP/CRP regulated [11].
NagD AMP + H2O = Ads + Pi Active in absence of glucose from growth medium
PhoA AXP + H2O = A(X-1)P + Pi Active in Pi limiting growth conditions [23]
SgcQ ATP + H2O = ADP + Pi Conditions for expression/activity unclear Gene expression cAMP/CRP regulated [23].
UshA AMP + H2O = Ads + Pi Active in absence of glucose from growth medium
Gene co-transcribed with adenylyl transferase YgiF ATP = cAMP + PPi (glnE). Constitutively expressed [24] Table 2: Potential AXP degradation enzymes, their corresponding genes and catalyzed reactions. 5PR1P: 5-phosphoribosyl 1-pyrophosphate; Ade: adenine; Ads: adenosine; AXP:
ATP, ADP or AMP; cAMP: cyclic AMP; CRP: cAMP receptor protein; Pi: phosphate; PPi: pyrophosphate; R5P: ribulose 5-phosphate.
Consequently, enzymatic activities potentially leading to AXP degradation were identified from a genome wide reaction table [25] and a database research. Two search
4. ATP sinks in cell-free extracts of Escherichia coli 76 criteria were applied: i) the reaction should use an AXP as a substrate and ii) it must not rely on co-substrates and cofactors that could be expected to be absent after washing. The search yielded a set of 10 candidates whose activities are summarized in table 2. The influence of adenylate cyclase on the ATP depletion was investigated in the tpiA background, while the other functions were tested as single knock-out genes. The growth patterns of the tested mutants and the wild-type W3110 were explicitly different even though they were grown on mineral medium containing glucose and yeast extract (fig. 3). W3110 tpiA, tpiA cyaA and tpiA cyaA ygiF had very similar growth patterns: all three mutants had significantly reduced growth rates, which is due to the tpiA gene deletion [chapter 3]. In strains without the tpiA deletion, the growth rate of W3110 atpD was significantly lower than that of the wild-type, also reflecting the importance of the deleted function, the H+-ATPase. The other mutants showed growth rates in the range of the wild-type with differences not larger than 25%. Remarkably, ATP depletion in CFXs of the wild-type or the various mutant strains was slower than in CFX derived from mutants of the W3110 tpiA background or atpD. The latter mutants were grown in fed-batch cultures, while the others were cultivated in batch cultures. However, a calculation of the phosphate need of a growing E. coli culture based on data of the cellular composition [26] and the known composition of yeast extract [27] showed that there was no P-limitation to be expected in fed-batch or batch cultures. Conspicuously, the CFX having high ATP depleting activity originate from slow-growing cultures, while the others had growth rates in the range of the WT (fig. 3). This might suggest that the increased ATP degradation was caused by regulatory changes in cultures hampered in growth. CFX from W3110 tpiA cyaA was tested for ATP degradation. The ATP degradation was highly similar in the tpiA and tpiA cyaA double mutant (fig. 2), but ADP and AMP accumulated to somewhat higher values in case of the tpiA single knock-out. Eliminating ygiF in addition to tpiA and cyaA changed the AXP degradation pattern substantially (fig. 2). The AMP concentration increased steadily and ADP accumulated to somewhat higher levels before decreasing again. Due to the accumulating AMP, the total concentration of AXPs remained significantly higher in this case, relative to CFX derived from the tpiA mutant. YgiF has been annotated as a putative adenylate cyclase due to the presence of two domains also found in adenylate cyclases, CYTH and CHAD. The CYTH domain is presumably closely related to adenylate cyclase activity, while CHAD was predicted to functionally interact with CYTH [28]. Furthermore, ygiF was described as being cotranscribed with glnE, the gene encoding for glutamine synthetase adenylyltransferase, an enzyme covalently connecting AMP to subunits of glutamine
4. ATP sinks in cell-free extracts of Escherichia coli 77 synthetase by ATP consumption [11]. The necessity of cAMP formation as an intermediate in the course of this reaction is highly probable. It is difficult to pinpoint the role that YgiF played in this result: A higher concentration of AMP could be the result of an increase in ADP degradation – indicating that a previously prominent ADP degradation route is no longer active – or of a decrease in AMP degradation. Both effects could be direct or indirect consequences of the absence of YgiF depending on whether it catalyzes exactly these activities or interfered in a different manner with the degradation activity. However, the fact that the AMP concentration increased throughout the entire experiment indicates that elimination of AMP degradation activity seems to be the most likely explanation. ATP was approaching complete consumption after 120 min in most CFXs, but not in the CFXs derived from the amn, apt and ushA knock-out strains (fig. 2). Here, the ATP concentrations either remained high (apt), decreased but leveled off at a rather high (amn) or a lower value (ushA). Apparently, the absence of these gene products has significant effects on ATP degradation. Remarkably, the initial velocities of ATP degradation for the CFXs derived from W3110 tpiA and W3110 amn are rather similar (0.38 U mg-1 for tpiA and 0.32 U mg-1 for amn) and both were higher than that of the wild-type (0.16 U mg-1). This suggests that ATP dephosphorylation itself might not have been eliminated by deleting amn. When looking at the sum of the AXP concentrations, an overall decrease was observed in all CFXs but the one derived from the amn strain. However, for the apt and ushA strains, the decrease in AXP concentrations was significantly lower than for the wild- type or the tpiA mutant. While the majority of CFXs caused reduction of the AXP concentration by more than 65% in 120 min, AXP was stable in CFX derived from the amn mutant, decreased by only 30% in the CFX derived from the apt strain, or decreased by only 47% in the CFX of the ushA strain.
Apparently, the enzymes Amn, Apt, and UshA together with YgiF, play an important role in AXP degradation. The level of AXP in E. coli is reported to be changing as result of perturbations in the energy metabolism. The ratio of ATP to AMP is described to be relatively constant. The degradation of AMP is used to regulate the adenylate pool size [29, 30]. The enzyme catalyzing this reaction, Amn, is specific for AMP [30] and was reported to be allosterically activated by ATP and inhibited by phosphate [18], both present in the SBT. It is difficult to estimate if the activating or the inhibiting mechanism is stronger in SBT conditions. However, it was reported that ATP presence led to a Km value decreased by several orders of magnitude [18]. The removal of Amn from the CFX is therefore a plausible explanation for the increased levels of AMP (fig. 2).
4. ATP sinks in cell-free extracts of Escherichia coli 78
In E. coli with deleted amn gene, elevated ATP concentrations were reported [31]. In the CFX, an increased AMP and ATP content could be observed (fig. 2). This could be due to inhibitory effects of AMP on ATP and ADP dephosphorylating enzymes. Apt was reported to be located in the cytoplasma [32] and high Apt activity was found in cells growing in a purine-rich environment [20]. The transport of purines from the environment into the cell is mediated by a high-affinity transport system coupled to Apt to convert them to nucleotides [33]. The impact the deletion of apt had on the stability of AXP (fig. 2), lead to the conclusion that there possibly were purines in the growth medium, leading to high Apt activity. Physiologically, the production of AMP is favored [11]. The results presented here indicate that in the non-physiological SBT conditions, the reaction direction is reversed. UshA is a phosphatase and catalyzes the dephosphorylation of nucleotides (NMP) to the respective nucleosides. Its substrate range is quite broad and the periplasmic enzyme was reported to enable the uptake of extracellular nucleotides into the cell [34]. The expression of ushA could be an indication that there are nucleotides present in the medium the cell is taking up during growth. The source of these nucleotides could be the yeast extract used in the cultivations. The result indicated that all tested activities indeed contributed to some extent to AXP degradation. However, three of those activities, Amn, Apt and UshA, appear to play a major role. Apparently, most of the observed ATP degradation in the SBT is an effect of the combined actions of Amn, Apt, UshA and YgiF, making them prime targets for an improved SBT host strain. Thereby, in the CFX of a mutant lacking amn only dephosphorylation of ATP, ADP and AMP seems to take place, without removing any AXP from the system.
4. ATP sinks in cell-free extracts of Escherichia coli 79
Fig. 2: ATP stability in different CFXs from E. coli W3110 wild-type (WT) and various knock-out strains. Reference curves from the CFX derived from the tpiA knock-out strain in red.
4. ATP sinks in cell-free extracts of Escherichia coli 80
Fig. 2 continued.
4. ATP sinks in cell-free extracts of Escherichia coli 81
Fig. 3: Growth behavior of batch cultivated single knock-out mutants and wild-type W3110 as well as fed-batch cultivated atpD, tpiA and multi knock-out mutants in tpiA background.
4. ATP sinks in cell-free extracts of Escherichia coli 82
4.5. Discussion
Phosphorus is an important element in an E. coli cell. It is essential for several cellular components such as DNA, RNA or phospholipids. It is usually taken up as phosphate (Pi) from the environment. Pyrophosphate, polyphosphate, organophosphates and phosphonates are also reported as phosphate sources in E. coli, whose utilization requires the cleavage of the phosphate ester or phosphonate bond in the periplasm. In
Pi-limiting conditions, phosphatase synthesis in general is increased. Some of these phosphatases can make up to 6% of the total protein content in these conditions [23]. Consequently, the phosphate content of growth medium (step I of the SBT) can have a crucial influence. However, no P-limitation should have been occurred leading to the expression or activation of phosphatases. To avoid ATP depletion, several strategies can be followed. Firstly, the environment of the cells in the growth phase could be altered in a way that some of the identified functions leading to ATP depletion are not expressed and consequently are absent in the SBT. According to table 2, no glucose limitation should occur, and, hypothesizing that the complex fraction contained purines, adding yeast extract should be avoided. However, the growth of E. coli strains lacking tpiA, the most important prerequisite for DHAP production, is heavily impaired on mineral medium [chapter 3]. A subproteomic switch for tpiA, which would allow removing the gene product after the production of the CFX, is under construction in the laboratory of Prof. V. de Lorenzo (Centro Nacional de Biotecnología (CNB), Madrid). Secondly, all genes identified as mainly or partly responsible for ATP depletion can be deleted from the genome of the mutant designed for the SBT purpose. Possibly, there might be other enzymes able to take over the function of the missing ones. However, the fact that differences were seen in the AXP depletion patterns in CFXs from several single knock-out mutants suggests that the function of at least some dephosphorylating enzymes cannot or not fully be substituted. Removing the activities showing high AXP conversion by gene deletion therefore can be considered as one step towards a CFX with significantly higher ATP stability. Thirdly, another measure to counteract the AXP depletion could be the introduction of a second ATP recycling system modeled on the strategies mentioned in the introduction and including AMP in the regeneration circle. A combined strategy consisting of an environment that does not cause the expression and activation of AXP depleting enzymes in step I of the SBT, the utilization of a mutant lacking genes for amn and apt in the tpiA background and the addition of a further ATP
4. ATP sinks in cell-free extracts of Escherichia coli 83 recycling system could be a most promising way towards an SBT working with truly catalytic amounts of cofactors. In summary, ATP depletion appears to be a central problem in the current approaches to SBT production. However, the causes can be rationalized and promising strategies can be developed to overcome that obstacle.
4.6. Acknowledgments
Financial support from the EU is gratefully acknowledged. The here described research was part of “EuroBioSyn – A Modular Platform for Biosynthesis of Complex Molecules”.
4. ATP sinks in cell-free extracts of Escherichia coli 84
4.7. References
1. Kim, D.M. and J.R. Swartz, Prolonging Cell-Free Protein Synthesis by Selective Reagent Additions. Biotechnol. Prog., 2000. 16(3): p. 385-390. 2. Jewett, M.C. and J.R. Swartz, Substrate replenishment extends protein synthesis with an in vitro translation system designed to mimic the cytoplasm. Biotechnology and Bioengineering, 2004. 87(4): p. 465-471. 3. Matveev, S.V., et al., Effect of the ATP level on the overall protein biosynthesis rate in a wheat germ cell-free system. Biochimica et Biophysica Acta (BBA) - Protein Structure and Molecular Enzymology, 1996. 1293(2): p. 207-212. 4. Kitaoka, Y., N. Nishimura, and M. Niwano, Cooperativity of stabilized mRNA and enhanced translation activity in the cell-free system. Journal of Biotechnology, 1996. 48(1-2): p. 1-8. 5. Kawarasaki, Y., H. Nakano, and T. Yamane, Phosphatase-immunodepleted cell- free protein synthesis system. Journal of Biotechnology, 1998. 61(3): p. 199-208. 6. Kim, R.G. and C.Y. Choi, Expression-independent consumption of substrates in cell-free expression system from Escherichia coli. Journal of Biotechnology, 2000. 84(1): p. 27-32. 7. Kim, D.M. and J.R. Swartz, Regeneration of Adenosine Triphosphate from Glycolytic Intermediates for Cell-Free Protein Synthesis. Biotechnol. Bioeng., 2001. 74(4): p. 309 - 316. 8. Itoh, H., Y. Kawazoe, and T. Shiba, Enhancement of protein synthesis by an inorganic polyphosphate in an E. coli cell-free system. Journal of Microbiological Methods, 2006. 64(2): p. 241-249. 9. Calhoun, K.A. and J.R. Swartz, Energy Systems for ATP regeneration in Cell-Free Protein Synthesis Reactions, in In Vitro Transcription and Translation Protocols, G. G., Editor. 2007, Humana Press Inc.: Totowa, NJ. p. 3 - 17. 10. Schomburg, I., et al., BRENDA, the enzyme database: updates and major new developments. Nucl. Acids Res., 2004. 32(suppl_1): p. D431-433. 11. Keseler, I.M., et al., EcoCyc: a comprehensive database resource for Escherichia coli. Nucl. Acids Res., 2005. 33(suppl 1): p. D334-D337. 12. Kanehisa, M., et al., KEGG for linking genomes to life and the environment. Nucl. Acids Res., 2008. 36(suppl_1): p. D480-484. 13. Nieboer, M., J. Kingma, and B. Witholt, The alkane oxidation system of Pseudomonas oleovorans: induction of the alk genes in Escherichia coli W3110(pGEc47) affects membrane biogenesis and results in overexpression of 4. ATP sinks in cell-free extracts of Escherichia coli 85
alkane hydroxylase in a distinct cytoplasmic membrane subtraction. Molecular Microbiology, 1993. 8(6): p. 1039-1051. 14. Baba, T., et al., Construction of Escherichia coli K-12 in-frame, single-gene knockout mutants: the Keio collection. Mol Syst Biol, 2006. 15. Sambrook, J. and D.W. Russel, Molecular Cloning: A Laboratory Manual 3rd ed. 2001. 16. Sauer, U., et al., The Soluble and Membrane-bound Transhydrogenases UdhA and PntAB Have Divergent Functions in NADPH Metabolism of Escherichia coli. J. Biol. Chem., 2004. 279(8): p. 6613-6619. 17. Bergmeyer, H.U., J. Bergmeyer, and M. Grassl, Methods of Enzymatic Analysis. 1984. 6: p. 342-350. 18. Leung, H.B., et al., Structure and regulation of the AMP nucleosidase gene (amn) from Escherichia coli. Biochemistry, 1989. 28(22): p. 8726-8733. 19. Thaller, M.C., et al., Identification of the gene (aphA) encoding the class B acid phosphatase/phosphotransferase of Escherichia coli MG1655 and characterization of its product. FEMS Microbiology Letters, 1997. 146(2): p. 191-198. 20. Hochstadt, J., Adenine Phosphoribosyltransferase form Escherichia coli. Methods in Enzymology, 1978. 51: p. 558 - 567. 21. Senior, A.E., The proton-translocating ATPase of Escherichia coli. Annu Rev Biophys Biophys Chem, 1990. 19: p. 7-41. 22. Botsford, J.L. and J.G. Harman, Cyclic AMP in prokaryotes. Microbiol. Mol. Biol. Rev., 1992. 56(1): p. 100-122. 23. Wanner, B.L., Phosphorus Assimilation and Control of the Phosphate Regulon, in Escherichia coli and Salmonella: cellular and molecular biology, A. Böck, et al., Editors. 1996, ASM Press. 24. Heeswijk, W.C., et al., The genes of the glutamine synthetase adenylylation cascade are not regulated by nitrogen in Escherichia coli. Molecular Microbiology, 1993. 9(3): p. 443-457. 25. Reed, J.L., et al., An expanded genome-scale model of Escherichia coli K-12 (iJR904 GSM/GPR). Genome Biology, 2003. 4(9): p. R54.1-R54.12. 26. Nielsen, J. and J. Villadsen, Bioreaction Engineering Principles. 1994, New York: Plenum Press. 27. Difco Manual. 11 ed. 1998, Sparks, MD, USA: Becton Dickinson Co. 28. Iyer, L. and L. Aravind, The catalytic domains of thiamine triphosphatase and CyaB-like adenylyl cyclase define a novel superfamily of domains that bind organic phosphates. BMC Genomics, 2002. 3(1): p. 33. 4. ATP sinks in cell-free extracts of Escherichia coli 86
29. Leung, H.B. and V.L. Schramm, Adenylate degradation in Escherichia coli. The role of AMP nucleosidase and properties of the purified enzyme. J. Biol. Chem., 1980. 255(22): p. 10867-10874. 30. Zhang, Y., S.E. Cottet, and S.E. Ealick, Structure of Escherichia coli AMP Nucleosidase Reveals Similarity to Nucleoside Phosphorylases. Structure, 2004. 12(8): p. 1383-1394. 31. Morrison, B.A. and D.H. Shain, An AMP nucleosidase gene knockout in Escherichia coli elevates intracellular ATP levels and increases cold tolerance. Biology Letters, 2008. 4(1): p. 53-56. 32. Page, M.G. and K. Burton, The location of purine phosphoribosyltransferase activities in Escherichia coli. Biochem. J., 1978. 174(3): p. 717-725. 33. Zalkin, H. and P. Nygaard, Biosynthesis of Purine Nucleotides, in EcoSal— Escherichia coli and Salmonella: cellular and molecular biology. [Online.], A. Böck, et al., Editors. 2007, http://www.ecosal.org. ASM Press: Washington, D. C. 34. Schultz-Heienbrok, R., T. Maier, and N. Strater, A Large Hinge Bending Domain Rotation Is Necessary for the Catalytic Function of Escherichia colis 5'- Nucleotidase. Biochemistry, 2005. 44(7): p. 2244-2252.
5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 87
5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations
Michael Schümperli, Sven Panke
5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 88
5.1. Abstract
A dihydroxyacetone phosphate (DHAP) producing System of Biotransformations (SBT) was expanded by an aldolase reaction to produce D-threo (3S, 4R) 5-deoxy-5-ethyl- xylulose 1-phosphate (DEXP) from DHAP and butanal. The reaction was catalyzed either by Escherichia coli’s class II fructosebisphosphate aldolase (FBPA), which was also involved in DHAP formation, or by added class I rabbit muscle aldolase (RAMA). In a process separated into DHAP production and RAMA-catalyzed DHAP conversion to DEXP, a maximum of 80% of the formed DHAP was converted to DEXP, as judged by measurements of changes in DHAP concentrations. This is expected to be in broad agreement with the thermodynamics of the reaction. In a preparative experiment, 43 mg of D-threo (3S, 4R) 1,3,4-triacetyl-5-deoxy-5-ethyl-xylulose (TADEX), corresponding to approximately 35% of the originally produced amount of DEXP, were obtained from a 100 mL SBT reaction after dephosphorylation, extraction, acetylation and purification. Further experiments suggested that FBPA from E. coli was not efficiently accepting butanal as substrate, indicating that the SBT with the E. coli enzyme could be well suited to provide DHAP while other aldolases, including those with different diastereoselectivities, could be used for synthesis.
5.2. Introduction
The System of Biotransformations (SBT) is an enzyme-based system for the multi-step production of molecules from cheap precursors such as glucose (Glc). An SBT for the formation of dihydroxyacetone phosphate (DHAP) from Glc was described previously [chapter 3, 4]. It is based in large parts on the glycolysis as reaction pathway, completed by additional hexokinase (HK) and lactate dehydrogenase (LDH). The upper branch of the glycolysis from Glc to DHAP was used as the DHAP production pathway, whereas the lower branch from glyceraldehyde 3-phosphate (GAP) to pyruvate (PYR) and then lactate (LAC) was used for cofactor recycling. Though an efficient formation of DHAP is an indispensable element of the SBT, DHAP is not a suitable final product. Firstly, it has been reported to be unstable [1], although, given a set of appropriate reaction conditions, the degradation seems to be much lower than expected [chapter 3]. Secondly, the prime cause for interest in DHAP is its role as essential donor in the reactions of DHAP-dependent aldolases. These allow the diastereoselective formation of vicinal diols in all 4 possible diastereomeric configurations [chapter 2]. Therefore, rather than using the SBT for DHAP formation, it 5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 89 is attractive to add the DHAP-consuming reaction as well and to produce a vicinal diol of predetermined stereoconfiguration instead. As a DHAP-dependent aldolase, the fructosebisphosphate aldolase (FBPA) of Escherichia coli is already part of the SBT, the easiest option could be simply to add another acceptor aldehyde to the SBT next to GAP which is formed by the FBPA during the formation of DHAP (fig. 1). E. coli’s main FBPA belongs to the Zn2+-dependent aldolase class II and forms a dimer. The enzyme was not frequently used for synthetic purposes and is thus not very well characterized in that respect. It is known however that it catalyzes the aldol condensation of DHAP with an acceptor aldehyde to give a product of the D-threo (3S, 4R) configuration. The enzyme was used for the synthesis of 1-deoxynojirimycin and 1-deoxymannonojirimycin with azido aldehydes as acceptors [2] as well as for several non-natural fructose-like compounds using halogenated, hydroxylated or azido aldehydes [3]. E. coli possesses also a class I aldolase, which was detected only in cells growing on C3 carbon sources. Therefore, it is likely that the class I aldolase is involved in gluconeogenesis, while the class II enzyme catalyzes the glycolytic reaction [4]. The class I aldolase was considered to be absent due to growth of the culture on glucose. Alternatively, one could add a second DHAP-depending aldolase to the system. As the SBT has so far not been adapted to deal with multiple DHAP-dependent aldolases with different stereoselectivities, this second aldolase should preferably have the same stereoselectivity. Most frequently used for the formation of vicinal diols with the D-threo (3S, 4R)-configuration were the FBPAs from rabbit muscle (RAMA) and from Staphylococcus carnosus [5, 6], both belonging to the class I aldolases. Both aldolases have a remarkably large acceptor tolerance and the aldolase from S. carnosus was reported to be more stable. In this chapter, the feasibility of an integration of aldol formation into the SBT for DHAP formation was investigated. As a model reaction, the conversion of DHAP and butanal to D-threo (3S, 4R) 5-deoxy-5-ethyl-xylulose 1-phosphate (DEXP) was used. Butanal was described as feasible acceptor aldehyde for aldol condensations catalyzed by RAMA and FBPA from S. carnosus [5, 7]. The reaction rate was reported to be lower than with the natural acceptor GAP, but efficient conversion was achieved nevertheless. DEXP was produced with a yield on DHAP of 70 – 80%. The product is not commercially available, which has prompted the analysis to proceed via the conversion of DEXP to D-threo (3S, 4R) 1,3,4-triacetyl-5-deoxy-5-ethyl-xylulose (TADEX), which can be purified [5].
5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 90
Fig. 1: Expanded SBT containing the FBPA reaction to synthesize DEXP from DHAP and butanal. Glc: glucose; G6P: glucose 6-phosphate; F6P: fructose 6-phosphate; FBP: fructose 1,6-bisphosphate; DHAP: dihydroxyacetone phosphate; GAP: glyceraldehyde 3-phosphate; 1,3-DPG: 1,3-diphosphoglycerate; 3PG: 3-phosphoglycerate; 2PG: 2-phosphoglycerate; PEP: phosphoenol pyruvate; PYR: pyruvate; LAC: lactate, DEXP: D-threo (3S, 4R) 5-deoxy-5-ethyl-xylulose 1-phosphate; HK: hexokinase; PGI: phosphoglucose isomerase; PK: phosphofructokinase; FBPA: fructose-1,6-bisphosphate aldolase; RAMA: rabbit muscle aldolase; GAP-DH: glyceraldehyde 3-phosphate dehydrogenase; PGK: phosphoglycerate kinase; PGM: phosphoglycerate mutase; ENO: enolase; PYK: pyruvate kinase; LDH: lactate dehydrogenase; boxed: substrates and products of the SBT; underlined: external enzymes. 5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 91
5.3. Materials and Methods
Cell-free extracts Triose phosphate isomerase (TpiA)-free cell free extracts (CFX) of cultures of E. coli W3110 tpiA zwf::kanR and W3110 cya ptsI tpiA zwf [chapter 4] were produced as described previously [chapter 3].
Expanded SBTs SBT experiments were performed in jacketed, thermostated beakers at a temperature of 37°C. The experiments were carried out in a volume of 10 mL containing either 10 mg
-1 -1 mL or 1 mg mL total protein concentration ([Prottot]) as determined by a commercial Bradford assay [8] (BioRad, Hercules, CA, USA) and with bovine serum albumin (Sigma-
Aldrich, Buchs, Switzerland) as standard. The CFX was diluted to the required [Prottot] by SBT buffer (100 mM 3-(N-Morpholino)propanesulfonic acid (MOPS) or sodium bicarbonate (NaHCO3), 0.84 mM KCl, 5 mM MgCl2,1 mM ZnSO4 pH = 7.0 (MOPS) or 7.7
(NaHCO3)) to a total volume of 9.5 mL. Subsequently, 100 μL of a 1.11 M solution of
+ Na2HPO4 in SBT buffer, 100 μL of a 0.6 mM solution of NAD in SBT buffer, 0.6 or 0.06 U hexokinase (HK) and 33 or 3.3 U L-lactate dehydrogenase (LDH) (Roche Diagnostics,
Rotkreuz, Switzerland – for definition of units see chapter 3), depending on [Prottot], were added. The reactions were started by the addition of 200 μL of a 0.6 mM solution of ATP in SBT buffer unless stated otherwise and 100 μL of a 1.11 M solution of Glc in SBT buffer to the prewarmed reactors. The DEXP production was triggered by the addition of 1 mmol butanal (Sigma-Aldrich, Buchs, Switzerland) and, where necessary, of 9 U of RAMA (provided in aqueous solution, activity determined at 25°C with fructose 1,6-bisphosphate (FBP) as substrate, Roche Diagnostics, Rotkreuz, Switzerland). Initial reaction rates are given in U mg-1 of total protein calculated from the conversion of DHAP.
Derivatization of DEXP For a facilitated analysis, DEXP was converted to its dephosphorylated and acetylated derivative. For dephosphorylation of DEXP to 5-deoxy-5-ethyl-xylulose (DEX), the present proteins were precipitated by lowering the pH to 2.8 through addition of 4 M citric acid and then removed by centrifugation (10 min, 21’000 g, 4°C). The pH of the supernatant was adjusted to 4.8 with 2 M NaOH and approximately 100 U of potato acid phosphatase (activity defined for aqueous solution of pH = 4.8, at 37°C, with 5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 92 p-nitrophenyl phosphate as substrate, Sigma-Aldrich, Buchs, Switzerland) was added. After stirring over night at 37°C, the phosphatase was precipitated by the addition of 3 M HCl to lower the pH to 2.5 and the precipitated proteins were separated from the liquid phase by filtration through a wetted 0.22 μm polyvinylidene fluoride (PVDF) filter (Roth, Reinach, Switzerland). The resulting DEX-containing solution was extracted seven times with ethyl acetate and subsequently seven times with diethyl ether, each in a volume ratio of 1:1. The organic phases were dried over Na2SO4, which was later removed by filtration. The organic solvents were removed in a rotary evaporator. The residuals were united and dissolved in 67 mL of pyridine under a protective argon atmosphere. An aliquot of 33 mL of acetic anhydride was added for acetylation and the reaction was left stirring over night at room temperature under a protective atmosphere to complete the reaction to TADEX. Next, the remaining liquid phase was removed in a rotary evaporator at 40°C and a reduced pressure of 30 mbar and was co-evaporated three times with toluene. The remainders were further purified by column chromatography (silica gel, bed length = 20 cm, inner diameter = 2.5 cm) and applying a step gradient of 20%, 25%, 30% and 50% EtOAc in cyclohexane and 200 mL per step. According to the collected fractions, the product was eluted with 30% EtOAc in cyclohexane. The fractions containing the product were identified by thin layer chromatography on silica gel 60
F254 0.25 mm plates (Merck, Zug, Switzerland) developed in 50% EtOAc/cyclohexane.
TADEX had an Rf-value of 0.64 and was visualized by dipping the plate in a solution of
5 g Ce(SO4)2 and 25 g of (NH4)2MoO4 in 500 mL of 1:10 diluted H2SO4 and subsequent heating. The solvent from the united TADEX containing fractions was removed in a rotary evaporator, co-evaporated with dichloromethane and dried in a round bottom flask to obtain 43 mg (0.15 mmol) vacuum dried TADEX as yellow oil.
Analysis of TADEX 1H- and 13C-NMR spectra were recorded in a Varian VXR300 (300 MHz) spectrometer (Varian, Steinhausen, Switzerland). Chemical shifts δ are given in ppm relative to the resonance of the solvent (CDCl3). Electron ionization mass spectrometry (EI-MS) was performed by the MS service center at the Laboratory for Organic Chemistry, ETH Zürich. The NMR spectra corresponded well to those published [5].
5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 93
1 H-NMR (300 MHz, CDCl3): δ = 5.35 (m, H4), 5.29 (d, H3), 4.86 (d, H1b), 4.77 (d, H1a), 2.2 (s, 3H1), 2.14 (s, 3H3), 2.06 (s, 3H4), 1.6 (m, 2H5), 1.25 (m, 2H6), 0.95 (t, 3H7).
13 C-NMR (75 MHz, CDCl3) δ = 198.14 (C2), 170.04 (CO3), 169.90 (CO1), 169.66 (CO4). 76.60 (C3), 71.50 (C4), 66.79 (C1),
32.43 (C5), 20.88 (CH31), 20.55 (CH33), 20.49 (CH34), 18.61 (C6), 13.83 (C7).
+ + EI - MS: 288.1201 ([M] , C13H20O7; calc. 288.1209). 5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 94
5.4. Results and Discussion
Preparative DEXP synthesis and derivatization To expand the SBT towards the production of vicinal diols from cheap substrates, the addition of butanal was included, which was expected to react with the SBT-produced DHAP to give D-threo (3S, 4R)-configured DEXP (fig. 2). The reaction of butanal with DHAP catalyzed by FBPAs has been reported previously to give DEXP [5, 9]. As an unequivocal analysis of DEXP is rather laborious and has been performed previously [5], we confirmed first the chemical structure of the product that was formed in our SBT in order to use simpler indicators of reaction progression, such as DHAP consumption, for analysis later on.
Fig. 2: DHAP and DEXP synthesis after butanal addition in a two-phase SBT and subsequent analysis by derivatization of DEXP. a) DEXP production in 2-phase SBT with the addition of 10 mmol of butanal and 90 U of RAMA at the beginning of phase 2
-1 (arrow). CFX derived from E. coli W3110 tpiA cyaA ptsI zwf, [Prottot] = 1 mg mL , NaHCO3- based SBT buffer, pH = 7.7, V = 100 mL b) DEXP synthesis and analysis by dephosphorylation and acetylation.
To this end, an SBT with a volume of 100 mL was operated containing 11.1 mM (1.11 mmol) Glc for 180 min, in which time the Glc concentration dropped to 1.3 mM (0.13 mmol) and 6.6 mM (0.66 mmol) DHAP was produced. In a second phase, 10 mmol butanal (corresponding to 10-fold excess over DHAP) and 90 U of RAMA were added. As a result, the DHAP content decreased to 2 mM (0.2 mmol), while the remaining Glc was completely used up in the following 120 min. The resulting solution was treated as described in the Material and Methods section and it could be clearly proven by 1H-, 13C- NMR and EI-MS that TADEX resulted as product. We did not continue to verify the 5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 95
D-threo (3S, 4R)-configuration of the vicinal diol, which has been confirmed several times in the past [5, 10, 11] and which is not changed during work-up. Thus, we concluded that indeed D-threo (3S, 4R)-DEXP was produced in the expanded SBT, leading to D-threo (3S, 4R)-TADEX after work-up. As the decrease in DHAP concentration was to the largest extent dependent on the addition of RAMA (see also below), we further concluded that the aldol reaction was the main sink for DHAP in the system. Therefore, the consumption of DHAP is further used in the remainder of this chapter as the indicator of the progress of the aldol- reaction. Even though this cannot be more than a qualitative indicator as long as it is not possible to routinely quantify the reaction product, it was deemed to be sufficiently informative to judge on the feasibility of the various investigated scenarios. The preparative SBT was analyzed following this notion. A total of 43 mg of TADEX was isolated, corresponding to 0.15 mmol. The SBT produced 0.66 mmol of DHAP from
1.11 mmol Glc. Therefore, the yield of isolated TADEX on DHAP was 23% (YTADEX/DHAP) and
14% on Glc (YTADEX/Glc). These yields are rather low, but close to those obtained by others [5]. Several aspects have to be taken into account: First of all, the yield of the investigated reaction might be equilibrium-limited. The equilibrium constant for the reaction naturally catalyzed by RAMA (FBP to DHAP and GAP) is in the range of 104 M-1 [7], but the value for the reaction with butanal is not available. We therefore estimated it from the second phase of our experiment (fig. 2) under the assumption that the reaction had indeed reached equilibrium. Ideally, 0.46 mmol of DHAP were converted to 0.46 mmol of DEXP, corresponding to a conversion of 70%, which is in the range of what was observed before [12]. With these numbers, an equilibrium constant of 24 M-1 can be calculated. This suggests that the equilibrium lies much less on the synthetic side as it is the case with FBP. From the ideally 0.46 mmol of DEXP, 0.15 mmol TADEX were isolated. The yield of the derivatization of DEXP with subsequent isolation of TADEX therefore was in the order of 33%. Losses might have occurred during incomplete dephosphorylation and purification as well as during incomplete extractions with ethylacetate and diethylether. The acetylation step is generally considered as quantitative. Significant DHAP degradation was not considered likely (chapter 3), but the stability of DEXP and DEX are unknown and might have contributed to the rather low amount that has been recovered.
5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 96
Optimization of the expanded SBT Next, different possible operations to form DEXP in the SBT were investigated. The most important questions for optimization were when to add the butanal and whether the FBPA activity needed to be supplemented by additional enzyme. The first point might become important if the presence of higher concentrations of aldehyde interfered with the operation of the SBT that would be required to produce the DHAP, which would be a strong argument for establishing a multi-phase approach – producing DHAP first, then adding the aldehyde and consuming it in a second phase. The second point addresses the fact that it remains largely unclear how broad the substrate range for the E. coli enzyme actually is, so it might be well possible that E. coli’s FBPA might not be very active in converting the produced DHAP in the model reaction considered here. First, it was investigated whether butanal could be added already at the beginning (fig. 3). An SBT experiment was performed with 100 mM of butanal present and only the E. coli FBPA in the CFX without any extra additions. Compared to a parallel SBT without butanal, slightly less DHAP was produced (fig. 3a). This confirms the notion that the presence of the aldehyde does not significantly interfere with the operation of the E. coli FBPA. In addition, it is clear that at least the upper part of the glycolysis is not significantly impeded by the presence of the aldehyde. It is not possible to draw the same conclusion for the cofactor-regenerating part of the SBT, as ATP was added in half stoichiometric amounts. The slight difference between the DHAP formations could either indicate some effects of the aldehyde on the members of the SBT or signal that some DHAP was channeled into DEXP production. In any case, the DHAP concentration remained rather high, indicating that the active E. coli enzyme was not converting the butanal efficiently. For DHAP-depending aldolases, a lack of efficient conversion can have different reasons: butanal might not be a suitable substrate for the E. coli enzyme. It was found that phosphorylated aldehydes usually lead to more efficient conversion with DHAP- dependent aldolases than their non-phosphorylated analogues [7, 13]. Alternatively, it could be that the active site of the E. coli enzyme is simply formed in such a way that butanal does not fit well, in agreement with earlier comments on the enzyme’s narrow acceptor tolerance [9]. On the other hand, other authors did not report difficulties with non-phosphorylated acceptor aldehydes in E. coli derived FBPA catalyzed reactions [2, 3]. However, neither of them specifically tested butanal or other alkyl aldehydes. In order to confirm that the weak drop in DHAP concentration was mainly due to the inability of the E. coli enzyme to convert butanal efficiently, 9 U of RAMA were added to the system at reaction start (fig. 3b). Clearly, much less DHAP accumulated, indicating 5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 97 that more DHAP was converted to DEXP. In the best possible case – the complete difference in DHAP concentration between the experiments with and without butanal was due to DHAP conversion to DEXP – this would mean that approximately 60% of the formed DHAP had been converted to DEXP. The decreased DHAP accumulation in presence of RAMA indicates that butanal might indeed not be a good substrate for E. coli’s FBPA.
Fig 3. DHAP accumulation in the presence of 100 mM butanal. Control experiments were conducted without butanal. CFX derived from E. coli W3110 tpiA zwf::kan, [Prottot] = 10 mg mL-1, MOPS-based SBT-buffer. a) SBT without RAMA. b) SBT with 0.9 U mL-1 RAMA.
The second relevant question was whether the presence of butanal interfered with the SBT to such an extent that it might still be more efficient to use a two-phase approach. Butanal had been described to inactivate alcohol oxidase from Candida sp., but yeast transketolase was fairly stable in presence of 100 mM butanal [14, 15]. Therefore, the experiments were divided in two phases: in phase 1, DHAP was produced in an SBT as described in chapters 3 and 4, in absence of butanal. In phase 2, after addition of 1 mmol butanal and, where required, 9 U RAMA, DHAP was rapidly consumed and presumably DEXP was formed from DHAP (fig. 4). Should the aldehyde have no significant influence on the activity of the system then the final result should be determined by the position of the reaction equilibrium alone – in other words, the final concentrations of DHAP should be identical in the one- and two-phase approach. Alternatively, the two-phase approach might deliver better results.
5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 98
Fig 4. DHAP accumulation and consumption in a 2-phase approach. CFX derived from E.
-1 coli W3110 tpiA zwf::kan, [Prottot] = 10 mg mL , T = 37°C, MOPS-based SBT buffer. 1 mmol butanal (100 mM) was added at the beginning of the second phase. a) No additional aldolase added. b) 9 U RAMA (0.9 U mL-1) added at the beginning of the second phase.
First, the results described in fig. 3 were confirmed by carrying out a two-phase experiment with only the E. coli FBPA. When after 180 min phase 2 was started by the addition of butanal, the DHAP concentration decreased only slightly from 6.7 mM to 5.2 mM in 2 h (fig. 4a). Apparently, conversion of butanal by E. coli’s FBPA is indeed not efficient. In contrast, adding RAMA at the start of phase 2 led to a rapid decrease in DHAP concentration from 6.9 to 1.4 mM, representing a consumption of 80% of the available DHAP in only a few minutes (fig. 4b). Interestingly, the Glc concentration continued to decrease in phase 2 from 1.6 to 0.2 mM, and the DHAP concentration leveled off at approximately 2 mM, the same concentration as in the one-phase approach with RAMA (fig. 3) and the preparative experiment (fig. 2). This indicates that the final DHAP concentration is determined by the equilibrium position of the DEXP- butanal-DHAP system. Consequently, the system does not seem to be significantly impeded by the presence of 100 mM butanal. Furthermore, restarting glucose consumption after RAMA and butanal addition indicates that the accumulation of DHAP has contributed to the arresting of the SBT at the end of phase 1. DHAP is known to inhibit FBPA [16]. Thus, in presence of high DHAP concentrations, FBP starts to accumulate. The difference in free energy ΔGo’ of the PFK-catalyzed phosphorylation of F6P is -14.2 kJ mol-1 [17]. Therefore, the reaction equilibrium constant K0’ is 235. The ATP depletion helps to shift the reaction equilibrium to the substrate side. Thus, the reaction equilibrium could have been reached when FBPA could not transform FBP and 5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 99 most of the initially present ATP was consumed. The reaction equilibrium of the isomerization of G6P to F6P lies on the side of G6P, leading to an increasing G6P concentration after the reaction catalyzed by PFK came to a halt. With a ΔG0’ of -16.7 kJ mol-1 [17] for the phosphorylation of glucose to G6P, a similar situation is found as for the phosphorylation of F6P, and reaching the reaction equilibrium could have been possible. When DHAP was removed from the reaction system by DEXP synthesis, the metabolites and finally the remaining Glc could be converted again.
5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 100
5.5. Conclusion and Outlook
An aldolase reaction to produce a non-natural carbohydrate was successfully coupled to the DHAP-producing SBT, demonstrating the compatibility of the SBT to the production of non-natural carbohydrates from Glc. High productivity was achieved only by the addition of RAMA to the extended SBT and despite high concentrations of butanal present, no indications were found that the aldehyde inactivated the enzymes of the SBT. The results strongly suggested that E. coli’s FBPA did not efficiently catalyze the conversion of butanal and DHAP to DEXP because butanal was not a suitable substrate. This result could indicate that although the FBPA of E. coli is catalyzing the last step of the DHAP production in the upper branch of the SBT, it cannot well catalyze the condensation of DHAP with at least some non-natural acceptor aldehydes at any meaningful rate. The E. coli-based SBT might therefore be a suitable system to couple the formation of DHAP to its concomitant consumption in aldol reactions with additional aldolases of different stereoselectivities.
5.6. Acknowledgments
Prof. P. Seeberger and P. Bindschädler are gratefully acknowledged for their help with synthesis, purification and characterization of TADEX. Financial support from the EU is gratefully acknowledged. The here described research was part of “EuroBioSyn – A Modular Platform for Biosynthesis of Complex Molecules”.
5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 101
5.7. References
1. Hettwer, J., H. Oldenburg, and E. Flaschel, Enzymic routes to dihydroxyacetone phosphate or immediate precursors. J Mol Catal B: Enzymatic, 2002. 19-20: p. 215- 222. 2. Von der Osten, C.H., et al., Use of a recombinant bacterial fructose-1,6- diphosphate aldolase in aldol reactions: preparative syntheses of 1- deoxynojirimycin, 1-deoxymannojirimycin, 1,4-dideoxy-1,4-imino-D-arabinitol, and fagomine. J. Am. Chem. Soc., 1989. 111(11): p. 3924-3927. 3. Liu, K.K.-C., R.L. Pederson, and C.-H. Wong, Fructose 1,6-diphosphate aldolase- catalysed stereoselective C–C bond formation. J. Chem. Soc., Perkin Trans. 1, 1991: p. 2669 - 2673. 4. Keseler, I.M., et al., EcoCyc: a comprehensive database resource for Escherichia coli. Nucl. Acids Res., 2005. 33(suppl 1): p. D334-D337. 5. Schoevaart, R., F. van Rantwijk, and R.A. Sheldon, Class I fructose-1,6- bisphosphate aldolases as catalysts for asymmetric aldol reactions. Tetrahedron: Asymmetry, 1999. 10: p. 705-711. 6. Brockamp, H.P. and M.R. Kula, Purification and characterization of a class I fructose 1,6-bisphosphate aldolase from Staphylococcus carnosus. Applied Microbiology and Biotechnology, 1990. 34(3): p. 287-291. 7. Bednarski, M.D., et al., Rabbit Muscle aldolase as a Catalyst in organic Synthesis. J Am Chem Soc, 1989. 111: p. 627-635. 8. Bradford, M.M., A Rapid and Sensitive Method for the Quantitation of Microgram Quantities of Protein Utilizing the Principle of Protein-Dye Binding. Anal Biochem, 1976. 72: p. 248-254. 9. Fessner, W.D. and C. Walter, Enzymatic C-C bond formation in asymmetric synthesis, in Bioorganic Chemistry. 1997. p. 97-194. 10. Zhu, W. and Z. Li, Synthesis of perfluoroalkylated sugars catalyzed by rabbit muscle aldolase (RAMA). J Chem Soc, Perkin Trans 1, 2000: p. 1105. 11. Wong, C.-H. and G.M. Whitesides, C-C Bond Formation (Chapter 4), in Enzymes in Synthetic Organic Chemistry. 1994, Pergamon/Elsevier Science Ltd.: Oxford. p. 195-215. 12. Effenberger, F. and A. Straub, A Novel Convenient Preparation of Dihydroxyacetone Phosphate and its use in Enzymatic Aldol Reactions. Tetrahedron Lett, 1987. 25(15): p. 1641-1644. 5. Production of 5-deoxy-5-ethyl-xylulose 1-phosphate in a System of Biotransformations 102
13. Schoevaart, R., Applications of aldolases in organic synthesis. 2000, Technische Universiteit Delft: Delft. p. 136. 14. Kawakami, K. and T. Yoshida, Kinetic study of enzymatic reaction in aqueous- organic two-phase systems - an example of enhanced production of aldehydes by alcohol oxidase, in Biocatalysis in Non-Conventional Media, J. Tramper, Editor. 1992, Elsevier. p. 653 - 658. 15. Demuynck, C., et al., Enzyme-catalyzed synthesis of carbohydrates: synthetic potential of transketolase. Tetrahedron Letters, 1991. 32(38): p. 5085-5088. 16. Schomburg, I., et al., BRENDA, the enzyme database: updates and major new developments. Nucl. Acids Res., 2004. 32(suppl_1): p. D431-433. 17. Stryer, L., Biochemistry. 4 ed. 1988, New York: W. H. Freeman and Co.
6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 103
6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2
Michael Schümperli, René Pellaux, Matthias Bujara, Sven Panke
MS and RP developed the reactor coupling to the mass spectrometer and performed all experiments. MS conducted the enzymatic analysis and wrote the manuscript. RP performed the fine-tuning of the MRM method. MB tested the set-up for signal dynamics. SP supervised the research.
6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 104
6.1. Abstract
The comprehensive characterization of a dynamic multi-enzyme reaction system requires the fast and accurate quantification of the substrates, products and intermediates. Current measurement methods involve lengthy sample preparation, derivatization and/or chromatographic separation procedures, limiting typical sample throughput per analytical device to a few dozen samples per day, which makes the thorough characterization an extremely time-consuming activity. To drastically accelerate the determination of the concentrations of multiple intermediates in a perturbed system, an ESI-MS device was directly coupled to an enzyme-membrane reactor, eliminating all time-consuming steps and increasing the analysis frequency to a few seconds. However, submitting a complex mix of buffer and intermediates into the ESI-chamber limited in turn the accurate quantification of concentrations due to ion suppression, as could be shown by comparisons between MS and enzymatic measurements. Nevertheless, the continuous injection system allowed stable operation of the MS analytics, delivered reproducible results, could be used to determine an optimal initial ATP concentration for the enzymatic reaction system and delivered insight in the course of reactions from Glc to DHAP. Measures to moderate the impact of ion suppression and thus enable accurate quantification are underway.
6.2. Introduction
The behavior of a biological reaction system is difficult to predict by calculation based on the kinetic characteristics of the individual components available from the literature [1]. These data originate typically from measurements in which the enzyme is in a purified state and present at rather non-physiologically high concentration with respect to its own concentration, but at non-physiologically low concentration in terms of total protein. This artificial situation is substantially different from the situation of a multi-enzyme system, and in particular from a multi-enzyme system that operates as part of a cell-free extract (CFX). Here, the formation of complexes with other proteins might lead to channeling [2, 3], whereas in the purified form, the high overall protein concentration might provoke the formation of unexpected complexes or multimeric forms of one protein. In general, the different environmental conditions can have an impact on the value of the kinetic parameters such as the Km value or the reaction rate. Furthermore, a multi-step reaction system will be influenced by the presence of potentially all compounds that are formed during operation. Even for a system as well 6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 105 investigated as the glycolysis, it is unlikely that all relevant interactions have been identified, in particular when the system is forced to operate in non-physiological conditions, which is ultimately the aim of the Systems of Biotransformations (SBTs) described in previous chapters. Therefore, to enable accurate parametrization of a mechanistic model of a complex enzymatic reaction system, it is preferable to investigate the system in the state in which it will ultimately operate. This can in principle be done by adjusting specific suitable environmental conditions (e.g. high total protein concentration) and then investigating enzyme behavior one enzyme at a time. However, this approach is laborious, still does not capture the true system state and failed to produce an accurate model of the true system behavior [1]. Alternatively, the characterization is performed in real-time in the system. Such an approach has emerged for the characterization of in vivo metabolic networks in Escherichia coli and Saccharomyces cerevisiae. In principle, a culture is grown in defined medium, using 13C-labelled glucose either as pulse or as ingredient of the medium. Batch or continuous cultures are used. Samples are taken, derivatized where necessary and analyzed in HPLC, GC-MS or LC-MS [4-6]. If the response of the culture to a pulse is investigated, sampling and quenching need to be performed extremely rapidly [7]. A method to analyze the metabolites of the central carbon metabolism was developed recently. The mixture of chemically similar molecules required an elaborate LC- separation step before the MS-based quantification. The analysis time for one sample is in the range of 80 min, standard runs and column regeneration not included [8]. Next to the laborious nature of the method, one fundamental limitation of such an in vivo approach is the difficulty of system manipulation. Essentially, each manipulation of an in vivo system – such as the increase of an enzyme concentration or a change in kinetic properties – requires a genetic manipulation or at least the adaptation of the continuous cultivation system. In vitro systems have the unique advantage that they can be modified by the simple addition of more enzymes or enzyme variants, and pulses or steps can be carried out with, for example, phosphorylated intermediates that usually do not easily cross the membrane in vivo. Consequently, the depth to which the in vitro system can be analyzed is much larger. However, even though the analysis of an in vitro system might be easier, the data density that is required for a comprehensive analysis is still very high, and it becomes higher the more comprehensive the analysis is supposed to become. This is at odds with the current long analysis times discussed above, which motivated us to investigate fundamentally novel ways to address the analytic requirements. 6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 106
The crucial underlying reason for the elaborate protocols for dynamic sub-metabolomic measurements is to allow only single compounds to reach the detector and thus enable accurate identification and quantification. The first problem can in principle be directly addressed by MS2. The fragmentation and the association of fragments to parent ions allows to study complex ion patterns by selectively analyzing only characteristic ions. It is also the precondition for running an analysis in the multi- reaction monitoring (MRM) mode. In principle, the technique should allow the identification of several compounds that have been injected into the MS at the same time, typically via an electrospray ionization (ESI) chamber [9-13] which enables the removal of all the solvents to retain ionized analytes, which can then be injected into the spectrometer without running the risk of pressure increase inside the device [14]. However, concomitant submission of several compounds into an ESI-MS can lead to problems with quantification. A prime reason for this is ion suppression, which describes the change of signal intensity of a specific analyte at constant concentration in the presence of differing chemical conditions such as changing concentrations of a second metabolite [15-21]. Several possible causes for this effect have been described, including changing ion strength, ratio of organic and aqueous phase, presence of ion- pairing agents and the presence of less volatile compounds impairing droplet formation. Many mechanisms for ion suppression have been postulated, including ion pairing, ionization competition, change in conductivity, reduced solvent evaporation due to increased surface tension and hampered spray formation [15-21]. Generally, the effect appears to be poorly understood and inherently difficult to predict [21]. Specifically, it is unclear to which extent ion suppression would interfere with accurate quantification in our in vitro multi-enzyme systems. Consequently, given the attractive opportunity to drastically shorten analysis times for measurements, by eliminating chromatographic separation steps, a reactor system was implemented that allows direct injection of samples derived from a continuously operated reactor into an ESI-MS2 system. It was investigated whether the system had sufficient analytical power to enable the concomitant analysis of multiple analytes and whether it could be operated with enough accuracy to enable system characterization. As a model system, the formation of DHAP (dihydroxyacetone phosphate, all abbreviations of compounds are given in the legend of fig. 1) from Glc with regeneration of cofactors (chapter 3, fig. 1) was used.
6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 107
Fig. 1: SBT reaction pathway. Shaded sections indicate intermediates with potentially overlapping MS signals. Dark and medium grey: non-distinguishable isomers. Light grey: distinguishable despite identical mass signal. Glc: glucose; G6P: glucose 6-phosphate; F6P: fructose 6-phosphate; FBP: fructose 1,6-bisphosphate; DHAP: dihydroxyacetone phosphate; GAP: glyceraldehydes 3-phosphate; 1,3-DPG: 1,3-diphosphoglycerate; 3PG: 3-phosphoglycerate; 2PG: 2-phosphoglycerate; PEP: phosphoenol pyruvate; PYR: pyruvate; LAC: lactate; ATP: adenosine triphosphate; ADP: adenosine diphosphate; NAD+: nicotinamid adenine dinucleotid; NADH: nicotinamid adenine dinucleotid, reduced form. 6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 108
6.3. Materials and Methods
Instrumentation MS experiments were performed with an MDS Sciex 4000 QTrap triple quadrupole mass spectrometer (Applied Biosystems, Foster City, CA, USA). The HPLC pumps used in the reactor set-up were Hitachi-Merck LaChrom L-7100 pumps (Hitachi High- Technologies Europe GmbH, Krefeld, Germany). All capillaries, fittings, tees, unions and injectors were from Upchurch scientific (Oak Harbor, WA, USA). The enzyme-membrane reactor (EMR) was purchased from Jülich Chiral Solutions GmbH (Jülich, Germany).
Cell-free extract The cell-free extract (CFX) was prepared as described previously [chapter 3]. CFX with a total protein concentration of 4 mg mL-1, determined as described previously [chapter 3], was used.
Buffer:
The aqueous reaction buffer contained 20 mM NH4HCO3, 0.8 mM KCl, 2.5 mM + NH4NaHPO4, 1.25 mM NAD , 2.5 mM Glc, 0.4 mM 3-(N-Morpholino)propanesulfonic acid (MOPS) as internal standard and 25 mg mL-1 chloramphenicol to prevent microbial growth in the reactor. Where required, the appropriate amount of ATP was added. The pH of the buffer was 7.7 (unadjusted).
EMR – ESI set-up The set-up of the reactor/analysis system is depicted in fig. 2. The EMR (volume of 10 mL) was connected directly to the ESI and thus the MS. The flow through the EMR during reactions and analysis was 120 μL min-1. To reduce the concentrations of the compounds to a suitable range for analysis and to ensure compatibility of the solvent for ESI, the efflux from the reactor was passed through a series of flow reductions and dilutions. First, the leaving volume flow was split 1 to 9, then diluted 84-fold with a
3:1 (vol/vol) MeOH/H2O mixture containing 0.01 μM salicylic acid (SAL) as internal standard and 2 mM formic acid for acidification and the resulting flow was split again 33 to 1. This led to a final volume stream of 30 μL min-1 that continuously entered the ESI-chamber. The splitting ratio was defined by the back pressure of the capillaries and thus dependent on capillary length and inner diameter (fig. 2).
6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 109
Fig. 2: Schematic of the EMR-ESI set-up. P1, P2: HPLC-pumps; ST: Splitting tee; MT: mixing tee, dia: capillary diameter, l: capillary length. Important capillary lengths and diameters indicated, 1 in = 25.4 mm.
Characterization of the set-up To characterize the reactor set-up, it was connected to an HP 1050 UV photometer (Agilent, Basel, Switzerland) and phenylalanine (Sigma-Aldrich, Buchs, Switzerland) was used as a tracer (100 mM in water). For detection, a suboptimal wavelength of 270 nm (optimal 260 nm) was applied, since the concentration chosen for analysis would have given a nonlinear signal over the relevant range. 1 mL of the tracer was injected at t = 0.5 min (t = 0: start of recording data) into the EMR. The EMR was either directly or via the dilutive set-up connected to the detector.
General protocol for reactor experiments The EMR was charged with reaction buffer and CFX to a total protein concentration of 4 mg mL-1. The reactor was then flushed at 1 mL min-1 for 30 min. After that, the flux was reduced to 120 μL min---111 and the EMR connected to the MS. When the set-up was equilibrated (as indicated by constant signals of MOPS and SAL), enzymatic synthesis was triggered by pulses of either MgCl2 and glucose or, where required, ATP. Stepwise ATP concentration variations were performed from high to low concentrations (2.5 mM, 1.25 mM, 0.625 mM, 0.313 mM and 0.156 mM) with 30 min washing at 1 mL min-1 after each step.
MS analytics and quantification In the MRM operational mode of the triple-quadrupole MS, the first quadrupole (Q1) is used for the selection of the proper parent ion, the second quadrupole (Q2) works as collision cell to induce fragmentation of the selected parent ions and in the third quadrupole (Q3) the proper fragment is directed to the detector. Optimal ESI- and 6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 110
MS-parameters (declustering potential, collision energy, collision cell exit potential, fragment mass) for the fragmentation of every ion were determined previously.
Compound Company Order number
3PG, disodium salt Sigma-Aldrich, Buchs, Switzerland P8877
ADP, potassium salt Sigma-Aldrich 01899 dehydrate AMP, sodium salt Sigma-Aldrich A1752
ATP, disodium salt, Sigma-Aldrich 02055 hydrate DHAP, dilithium salt Sigma-Aldrich D7137
FBP, trisodium salt Sigma-Aldrich F6803
G3P, disodium salt, Sigma-Aldrich 50020 pentahydrate G6P, sodium salt Sigma-Aldrich 49282
GAP Sigma-Aldrich G5251
Glc Sigma-Aldrich 49139
LAC, sodium salt Sigma-Aldrich 71718
MOPS Sigma-Aldrich 69947
NAD+, hydrate GERBU Biotechnik GmbH, Gaiberg, Germany 1013
NADH, disodium salt GERBU 1051
PEP, potassium salt Sigma-Aldrich 79415
PYR, sodium salt Sigma-Aldrich 15990
SAL, sodium salt Sigma-Aldrich 71945 Table 1: Components used in the study. G3P: glycerol 3-phosphate; MOPS: 3-(N- morpholino)propanesulfonic acid; SAL: salicylate. Other abbreviations in fig. 1.
The MRM method was developed using the automated optimization routine for identification of the optimal parameters. A standard mix (SM1) containing all compounds given in table 2 dissolved in MeOH/H2O (3:1, vol/vol) was directly injected into the ESI chamber of the MS with a syringe pump (Harvard Apparatus, Holliston, MA, 6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 111
USA) at a flow rate of 30 μL min-1, corresponding to the later flow rate into the ESI chamber. The ionspray voltage was set to -4200 V, nebulizer gas (GS1), auxiliary gas (GS2), curtain gas (CUR) and collision gas (CAD) were set to 30, 40, 30 and 4 arbitrary units respectively. The temperature in the ESI chamber was set 200°C. Once the MRM method was defined, the linearity between the concentration of the compounds in the mixture and the signal intensity obtained from the MS- measurements were investigated. For this, an alternative standard mix, SM2, was used. SM2 contained all compounds given in table 2 except MOPS and SAL, which were used as internal standards of the aqueous and the organic phase respectively. The mix of appropriate dilution (into 20 mM NH4HCO3 with 4 mM MOPS) was injected to the set- up by a syringe pump replacing P1 and EMR (fig. 2) at a flow rate of 120 μL min-1. In this set-up, an analyte concentration of 10 mM in the syringe (the highest concentration used in the calibration) resulted in a concentration of 120 μM at time of injection into the ESI chamber.
Ion suppression experiment G6P and PEP were analyzed in the MS at concentrations of 0, 0.5, 1, 2 and 4 μM in presence of 0, 10 and 100 μM SM3. The concentration of SM3 comprises the sum of the concentration of all dissolved analytes. Thereby, the concentration of all analytes was identical. SM3 contained the compounds given in table 2 except G3P, Glc, MOPS and SAL, and except G6P or PEP respective to which compound was target of the ion suppression investigation. The SM containing G6P or PEP was diluted in water containing 250 μM MOPS and 2 mM formic acid. The solutions were directly injected into the ESI by a syringe pump.
Data handling The resulting signal intensities were normalized with the MOPS signal to account for variations in the flow stream. For quantification, the MRM signal intensities were converted to concentrations according to the previously measured standard slopes. To obtain the concentration of DHAP, the signal for mass/fragment 169/97 ((m/z)/(m/z)) was corrected for FBP. The signal intensities of 339/97 and 169/97 when measuring FBP were equal. Therefore, the real DHAP signal corresponded to the subtraction of the unique signal for FBP (339/97) from the observed signal intensity at 169/97.
6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 112
Enzyme assays Enzymatic quantification was performed for DHAP, Glc, G6P, Pyr, ATP, ADP and AMP. Glc was analyzed using the Roche enzymatic test kit (R-Biopharm, Darmstadt, Germany). All other metabolites were analyzed according to published protocols [22]. The required enzymes were purchased from Roche Diagnostics (Rotkreuz, Switzerland) or Sigma Aldrich (Buchs, Switzerland).
ATP turnover calculation
The ATP turnover number (TNATP) was calculated as described in chapter 3. 6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 113
6.4. Results
The comprehensive accurate real-time analysis of complex enzymatic reaction systems is essential for their characterization and modeling. In this study, direct injection of diluted effluxes from a reactor into an ESI-MS, bypassing all time-consuming chromatography- or derivatization-related steps, was investigated as a method for analysis for such a highly dynamic system.
MRM method MRM enables the time-resolved recording of signal intensities for specific fragment- parent ion pairs. This allows a rather accurate peak integration even in originally complex analyte mixtures [8]. The parameters of the MRM method have to be defined prior to analysis (table 2).
Collision Collision Parent ion Fragment Declustering cell exit Compound Time [ms] Energy M-H, [m/z] [m/z] potential [V] potential [eV] [V] PYR 87.016 43.1 500 -20 -12 -5 LAC 89.045 43.2 500 -25 -18 -5 SAL 136.944 93.2 500 -30 -20 -15 PEP 166.961 79.1 500 -20 -14 -11 DHAP 169.032 97 500 -25 -14 -15 GAP 169.01 151 500 -25 -12 -23 G3P 170.991 79.1 500 -35 -22 -11 Glc 179.049 89 500 -25 -12 -13 2PG 184.99 97 500 -25 -20 -15 MOPS 208.023 80 500 -65 -40 -11 G6P 258.978 97 500 -35 -22 -15 FBP 338.977 97 500 -30 -30 -15 AMP 346.048 134.2 500 -50 -42 -21 ADP 425.99 158.9 500 -50 -38 -25 ATP 506.022 408 500 -50 -32 -19 NAD+ 662.151 540.1 500 -30 -18 -29 NADH 664.143 346 500 -90 -48 -17 Table 2: MRM method. AMP: adenosine monophosphate;. G3P: glycerol 3-phosphate; MG: methylglyoxal; SAL: salicylate. Other abbreviations for metabolites in fig. 1.
6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 114
Distinguishing molecules of identical mass/charge ratios In MS, molecules with identical mass/charge ratio and fragmentation patterns cannot be distinguished. In case of the SBT, three groups of molecules – G6P and F6P, FBP, DHAP and GAP, 2PG and 3PG – can potentially have that problem (fig. 1), while all other metabolites could be resolved. DHAP and GAP have the same molecular mass and phosphate as the strongest fragment (m/z = 97). However, GAP showed an alternative, weak fragmentation (table 2). Therefore, GAP could be distinguished from DHAP. FBP can bear one or two negative charges. Only in the case of two negative charges is its m/z ratio identical to that of DHAP and GAP. Consequently, the MRM method was programmed to detect FBP by its singly charged molecular ion mass for detection (m/z = 339). Therefore, the concentration of DHAP could be calculated by subtraction of the contributions of FBP and GAP. However, during the experiments, GAP was never observed to accumulate, effectively facilitating the determination of the DHAP concentration. In case of the other two pairs with identical m/z ratios for the parent ions, no alternative fragmentation patterns could be detected, and consequently, they could not be distinguished.
Quantification Similar to the real measurement situation, the MS was calibrated with a mix of standard compounds (table 3). The mix contained DHAP, LAC and G3P in high concentrations, whereas all other analytes were present in lower concentrations. This reflected what was considered a likely scenario in the reactor: DHAP and LAC were the expected products of the reaction system and G3P had previously been detected in MS- analysis of samples of SBT experiments (chapter 3). The other compounds are intermediates in the reaction system and should ideally not accumulate to high concentrations. With this analyte mix, a first determination of the linear range of measurement was determined for each compound (table 3). Of course, using different dilutions of this mix fixed in relative compositon did not account for the changing analyte ratios in the course of an experiment and could therefore only give an indication about the range in which accurate measurements might be possible at all. Glucose was difficult to quantify with the here described MS method. Therefore, whenever glucose concentrations were required, they were determined by enzymatic assay.
6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 115
Linear Max conc Slope Correlation Compound range [mM] [signal intensity/mM] coefficient [mM] PYR 2 0.105 0.984 0 – 0.7 LAC 10 0.087 0.987 0 – 3.3 PEP 2 3.695 0.99 0 – 2.0
DHAP 10 2.27 0.996 0 – 10.0 G3P 10 2.033 0.996 0 – 10.0 2PG/3PG 2 2.389 0.994 0 – 2.0
G6P/F6P 2 4.414 0.992 0 – 2.0 FBP 2 2.127 0.999 0 – 2.0 AMP 2 0.381 0.983 0 – 2.0
ADP 2 1.071 0.994 0 – 2.0
ATP 2 0.565 0.997 0 – 2.0 Table 3: Calibration of the SBT metabolites: slopes, correlation coefficients and linear ranges. Reactor concentrations given. ESI chamber injection after dilution in the set-up by factor 84.
Ion suppression Ion suppression was investigated by measuring the signal intensities for different concentration of metabolites in presence of different concentrations of SM3. Ion suppression differed for different metabolites. The suppressing effects on G6P and PEP are shown to illustrate this. The signal for G6P was not suppressed in the tested conditions, whereas the PEP signal intensity was considerably decreased (fig. 3). G6P and PEP were measured at concentrations of 0, 0.5, 1, 2 and 4 μM. The experiment clearly shows that ion suppression can be significantly different for various analytes which makes it difficult to account for the effect in the experiment by calculation. Assuming that in a multi-enzyme system the composition of the system changes over time, the concentration space over which one would have to calibrate the system would be extremely large. 6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 116
Fig. 3: Ion suppression in presence of different amounts of SM3 (concentration given as the sum of all equally distributed analytes). a) no ion suppression of G6P. b) pronounced ion suppression of PEP.
Real-time measurement in a multi-enzyme reaction system For further investigation, the ESI-MS system was coupled to the efflux of a continuously operated EMR (fig. 2). In this reactor, an ultrafiltration membrane retained the enzymes. Consequently, the concentrations of the analytes in the efflux should no longer change after they passed the membrane. A direct injection of this efflux into the ESI-MS would then allow following the dynamics of the reactions system in the reactor by continuous MS-measurements. However, direct injection of the efflux into the MS is impossible, as the analyte concentrations are too high and the aqueous matrix is not suitable for ESI-MS. Therefore, the efflux was passed through a series of dilutions and splittings which introduced a time delay of about 7 min. However, this time delay should only change the time point at which the dynamics can be recorded, not the dynamics themselves. In order to make sure that the changes in intermediate concentrations over time are not additionally modulated by the tubing system between reactor and ESI-chamber, the residence time distributions were measured at the reactor outlet and immediately before the injection into the ESI-chamber (fig. 4). These distributions were broadly in agreement. Therefore, we think it is justified to call this mode of measurement real-time. The system was operated at a dilution rate of 0.12 mL min-1 (hydraulic residence time in the reactor 83 min). Typical times for an experiment were in the order of 2 h. Therefore, even though the reactor was operated with in- and efflux, the dynamics of the system resembled a batch reactor.
6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 117
Fig. 4: Influence of the post-reactor tubing system on the dynamics of the signal. Injection of 1 mL 100 mM phenylalanine solution, detection by UV/VIS.
Different ATP concentrations in the reactor Even though accurate quantification is not possible at the moment, qualitative statements about reactor performance can certainly be made with this particular set- up. A series of experiments was carried out to investigate the influence of the ATP concentration on the performance of the reaction system. The reactor containing the multi-enzyme system was equilibrated with a feed that contained all components and ATP in different concentrations. For each ATP concentration, the system was allowed to run into steady state. Efflux from the steady state was then used for a more accurate determination of analyte concentrations by enzyme assays [22]. This allowed evaluating the MS-determined concentration against a well established method of concentration measurement that has been used previously for the parametrization of a dynamic model for glycolysis [7]. Fig. 5 summarizes the results for 6 different ATP concentrations, and fig. 6 summarizes the results for all quantified metabolites and depicts the relative deviation of MS data to enzymatic data. Clearly, feeding high ATP concentrations into the reactor – concentrations close to the concentration of the available glucose of 2.5 mM – resulted in high DHAP and low glucose concentrations in steady state at the reactor outlet. With decreasing ATP concentrations, the amount of processed glucose decreased as well, which in turn led to lower DHAP concentrations.
6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 118
Fig. 5: DHAP concentration over time in response to a step-change in ATP concentration. Signal intensity normalized with MOPS for DHAP, ATP concentrations for the six different steps from left to right: 2.5, 1.25, 0.63, 0.32, 0.16 mM. a) sampling period for enzymatic analysis.
The two measurement methods – MS and enzymatic – showed good agreement from a qualitative but not from a quantitative point of view. The MS-based quantification led in many cases to higher concentrations than the enzymatic reference measurement. The values for DHAP, PYR and ATP can be directly compared (fig. 6c). For DHAP, the MS- based quantification returned on average 29% higher concentrations than the enzymatic reference measurement. In the cases of PYR and ATP the data was less homogenous. For the two lowest ATP concentrations in the feed, the enzymatic assays delivered higher concentrations. Clearly, the calibration that was derived under conditions of constant analyte ratios is not accurate enough when these ratios start to change, as it is the case in the dynamic measurements. The signal intensities of the analytes were divided by the one of the internal standard, MOPS, to account for intrinsic instabilities and to extend the linear range of the measurement.
6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 119
Fig. 6: MS-based and enzymatic quantifications of intermediates after ATP steps of various step size at steady state. a) Quantification according to MS-calibration. b) Quantification according to enzymatic assays. c) Deviations between measurement methods for DHAP, PYR and ATP and values for DHAP yield on consumed glucose for both measurement methods. The yield and the TNATP were calculated based on enzymatic data.
In the cases of ATP and PEP, the ratio of analyte/MOPS signal seemed to be inverted with decreasing ATP presence. Decreasing the ATP concentration in the first place signifies the decrease of ion strength of the matrix. Furthermore, the reaction system was less active in presence of low ATP concentrations, leading to less glucose transformation and less ion formation. Consequently, the ion suppression situation in presence of low ATP concentration has to be considered completely different compared to high ATP concentrations. 6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 120
On the other hand, in its current state the reaction/analysis system can be used to follow the development of analytes qualitatively. As an example of the scope, an experiment was performed in which an ATP pulse was fed into the reactor and the devolution of the signal over the reaction time was followed (fig. 7). The system is capable of following the progression of the different intermediates. This experiment impressively shows the potential of the analytic system, once quantification can be performed sufficiently accurately: The system allows following the conversions from G6P to LAC in detail. In particular, the successive accumulation of 2PG/3PG, PEP, PYR and LAC is in complete agreement with the anticipated reaction sequence of the system. Furthermore, the time course of the FBP concentration – which accumulates only after the DHAP concentration has increased significantly – suggests that accumulation is caused by the accumulating DHAP and either a product inhibition or the thermodynamic equilibrium. Furthermore, the fact that the F6P/G6P begins its accumulation before FBP starts to accumulate suggests that the phosphofructokinase plays an important role in the regulation of the system.
Fig. 7: Qualitative time course of intermediate concentrations in the DHAP-producing SBT after an ATP pulse.
Reproducibility The dilutive set-up described in this chapter can be used for real-time analysis of a reaction taking place in an EMR. However, fouling in the ESI-chamber, suboptimal 6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 121
spraying, the precipitation of e.g. MgCl2 or the evaporation of CO2 in the capillaries could lead to instable signals. Consequently, the inter-day reproducibility of the data was investigated. A series of three experiments with the addition of ATP pulses was performed on three different days, each experiment identical in set-up to the others. Each experiment comprised two ATP pulses and the time courses of the product and intermediate concentrations were followed. The data collected showed that the system has a good reproducibility (fig. 8). The concentration curves for DHAP and LAC, the two products in this system, can be reproduced rather accurately. The differences for the concentration curves of PYR are larger, as can be expected from the fact that the MS-signal for PYR is weak and it is an intermediate, so two sources of variations, production and consumption of PYR have to be considered where only one operates for DHAP and LAC.
Fig. 8: Inter-day reproducibility of ATP-pulse experiments in the EMR-ESI-MS set-up, followed by recording of the concentration of DHAP, LAC and PYR vs. time. Experiments were performed on three different days under identical conditions.
6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 122
6.5. Discussion
To characterize a complex system of enzymatic reactions, comprehensive analysis with high time resolution is necessary. The intermediates of the central carbon metabolism can be separated in an LC-step and analyzed by mass spectrometry, providing comprehensiveness and quantitative accuracy [8]. However, the LC-separation is time- consuming and therefore, only a limited number of samples can be analyzed. Sample quenching and storing becomes an issue. Omitting an LC-separation before injection enables the minimization of the measurement time. Like that, a continuous stream of analytes can be directed to the MS, providing the analysis of a specific analyte with a resolution of only seconds. The disadvantage is the increased vulnerability against ion suppression due to the continuously changing composition of the solution containing the analytes. Furthermore, it might be impossible to distinguish metabolites of identical parent ion mass, thus reducing the number of molecules that can be analyzed and increasing the insecurity of some values. In case of the relatively small number of glycolytic intermediates, this can be accepted and the advantage of high resolution by omitting the LC-separation is obvious. Nevertheless, particularly ion suppression leads to problems of data quality when the quantification of analytes is important. There are two aspects to the problem: on the one hand, different ions are affected differently in an otherwise identical solution (fig. 3), and on the other hand, the composition of the environment changes in time. An approach to solve this problem could be to increase the matrix-derived ion suppression to dampen the influence of the changing analyte concentrations. The results might then be comparable despite e.g. varying ATP concentrations, and the quantification should become more accurate. The drawback would be decreased sensitivity. Despite the insecure concentration values, the experiments have clearly shown that the initial ATP concentration can be reduced to 0.63 mM without hampering the DHAP production significantly (for a glucose starting concentration of 2.5 mM). TNATP thereby increased to 4, which seems to be the maximal turnover achievable in the current system. Both values are better than previous results (chapter 3), indicating that the conditions in the EMR were favorable to the ATP recycling in the DHAP-producing SBT. A possible reason could be the slightly less pronounced accumulation of intermediates thanks to the flux through the reactor. Furthermore, the MS-analysis has shown an increasing amount of lactate during the reaction, proving that the ATP regeneration pathway was functional. Finally, the set-up has clearly the power to provide detailed insights into the dynamics of the multi-enzyme system, provided the quantification problems can be overcome (fig. 7). 6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 123
In summary, the development of the MS-based analytic system allowing the real-time analysis of the course of the reaction taking place in a reactor coupled directly to the analytic system led to deep insights of the multi-enzyme reaction system, so far in a qualitative rather than in a quantitative way. The quantification of the metabolites was hampered by ion suppression. Possibilities to overcome this problem are in the focus of current investigations in our laboratory. When accurate quantification is possible, an extraordinary broad characterization of the reaction system will become possible.
6.6. Acknowledgments
Financial support from the EU is gratefully acknowledged. The here described research was part of “EuroBioSyn – A Modular Platform for Biosynthesis of Complex Molecules”.
6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 124
6.7. References
1. Teusink, B., et al., Can yeast glycolysis be understood in terms of in vitro kinetics of the constituent enzymes? Testing biochemistry. European Journal of Biochemistry, 2000. 267(17): p. 5313-5329. 2. Kholodenko, B.N., et al., Subtleties in control by metabolic channelling and enzyme organization. Molecular and Cellular Biochemistry, 1998. 184(1): p. 311- 320. 3. Rohwer, J.M., et al., Implications of macromolecular crowding for signal transduction and metabolite channeling. Proceedings of the National Academy of Sciences, 1998. 95(18): p. 10547-10552. 4. Fischer, E., N. Zamboni, and U. Sauer, High-throughput metabolic flux analysis based on gas chromatography-mass spectrometry derived 13C constraints. Anal Biochem, 2004. 325: p. 308-316. 5. Kleijn, R.J., et al., Metabolic flux analysis of a glycerol-overproducing Saccharomyces cerevisiae strain based on GC-MS, LC-MS and NMR-derived C-13- labelling data. Fems Yeast Research, 2007. 7(2): p. 216-231. 6. Iwatani, S., et al., Determination of metabolic flux changes during fed-batch cultivation from measurements of intracellular amino acids by LC-MS/MS. Journal of Biotechnology, 2007. 128(1): p. 93-111. 7. Theobald, U., et al., In Vivo Analysis of Glucose-Induced Fast Changes in Yeast Adenine Nucleotide Pool Applying a Rapid Sampling Technique. Anal Biochem, 1993. 214: p. 31-37. 8. Luo, B., et al., Simultaneous determination of multiple intracellular metabolites in glycolysis, pentose phosphate pathway and tricarboxylic acid cycle by liquid chromatography-mass spectrometry. Journal of Chromatography A, 2007. 1147(2): p. 153-164. 9. Zhao, X.N., et al., Determination of free fatty acids in bryophyte plants and soil by HPLC with fluorescence detection and identification by online MS. Chromatographia, 2007. 66(3-4): p. 197-206. 10. Gao, F., et al., Analysis of phospholipids by NACE with on-line ESI-MS. Electrophoresis, 2007. 28(9): p. 1418-1425. 11. Volpi, N., On-line HPLC/ESI-MS separation and characterization of hyaluronan oligosaccharides from 2-mers to 40-mers. Analytical Chemistry, 2007. 79(16): p. 6390-6397. 12. Tempels, F.W.A., et al., On-line coupling of SPE and CE-MS for peptide analysis. Electrophoresis, 2007. 28(9): p. 1319-1326. 6. Real-time quantification of an enzymatic reaction system by direct coupling to an ESI-MS2 125
13. Xing, J., C.F. Xie, and H.X. Lou, Recent applications of liquid chromatography-mass spectrometry in natural products bioanalysis. Journal of Pharmaceutical and Biomedical Analysis, 2007. 44(2): p. 368-378. 14. Herbert, C.G. and R.A.W. Johnstone, Electrospray Ionization (ESI), in Mass spectrometry basics. 2003, CRC Press. 15. Clifford, M.N., et al., A systematic study of carboxylic acids in negative ion mode electrospray ionisation mass spectrometry providing a structural model for ion suppression. Rapid Communications in Mass Spectrometry, 2007. 21(13): p. 2014- 2018. 16. de Rijke, E., et al., LC-MS Study To Reduce Ion Suppression and To Identify N- Lactoylguanosine 5'-Monophosphate in Bonito: A New Umami Molecule? J. Agric. Food Chem., 2007. 55(16): p. 6417-6423. 17. Annesley, T.M., Ion Suppression in Mass Spectrometry. Clin Chem, 2003. 49(7): p. 1041-1044. 18. Jessome, L.L. and D.A. Volmer, Ion Suppression: A Major Concern in Mass Spectrometry. The Application Notebook, 2007. Suppl: p. 83 - 89. 19. Antignac, J.-P., et al., The ion suppression phenomenon in liquid chromatography- mass spectrometry and its consequences in the field of residue analysis. Analytica Chimica Acta, 2005. 529(1-2): p. 129-136. 20. Francis Beaudry, P.V., Electrospray ionization suppression, a physical or a chemical phenomenon? Biomedical Chromatography, 2006. 20(2): p. 200-205. 21. Mallet, C.R., Z. Lu, and J.R. Mazzeo, A study of ion suppression effects in electrospray ionization from mobile phase additives and solid-phase extracts. Rapid Communications in Mass Spectrometry, 2004. 18(1): p. 49-58. 22. Bergmeyer, H.U., J. Bergmeyer, and M. Grassl, Methods of Enzymatic Analysis. 1984. 6: p. 342-350.
7. Summary and Outlook 127
7. Summary and Outlook
This thesis describes the development of the first System of Biotransformations (SBT), an in vitro multi-enzyme system for the synthesis of hardly accessible molecules. The basis of an SBT is the metabolic reaction network of an organism, which is genetically modified to exhibit the topology and dynamics for the production of a previously defined target compound. The steps towards the successful implementation of this concept are described for the example of DHAP production. The reaction cascade used for DHAP production and cofactor regeneration was the glycolysis. The organism used was E. coli, due to the numerous available tools for genetic engineering. The accumulation of DHAP could be demonstrated after a single – but important – genetic modification: the deletion of the gene encoding for the triosephosphate isomerase (TpiA). The yield of DHAP on glucose could be gradually increased from 55% to 73%, among other measures by further isolating the reaction pathway from the metabolism and thus channeling the metabolite flux towards DHAP. It could be demonstrated that the ATP and NAD+ regeneration was working and that in case of NAD+ the reduction of the initial cofactor concentration did not affect the DHAP synthesis. In contrast, the reduction of the initial ATP concentration by a factor of 10 led to slow reaction rates and very low product concentrations. The maximal ATP turnover was calculated to lie between 2.5 and 4. Two reasons are possible: i) the ATP regeneration pathway might suffer from side reactions thus reducing its efficiency and/or ii) the cell-free extract (CFX) used for the SBT contains significant activities to degrade AMP, ADP and ATP (AXPs). Indications for both reasons were found, and several enzymes heavily involved in AXP degradation could be identified. However, the ATP regeneration system was designed for the recycling of exactly the amount of ATP used for the DHAP production. Further ATP consumption could not be compensated in the current form. Improving the ATP regeneration in the SBT is essential. ATP is an expensive cofactor and cannot be used in nearly stoichiometric amounts in an economic synthesis. Three obvious measures can be taken for improvement. Firstly, the ATP regeneration pathway has to be further isolated. According to the in silico analyses performed, this should be possible without losing viability. Secondly, the identified ATP consuming enzyme activities have to be removed from the system. It is possible that this could influence the cell’s viability. Thirdly, an additional ATP regeneration system can be introduced to enhance the recycling efficiency. Another important point concerns the handling of the genetically modified E. coli. The deletion of tpiA from the genome led to an organism that is heavily hampered in 7. Summary and Outlook 128 growth, even when cultivated in a medium containing a complex fraction. This reflects the fundamental problem that some enzymes that are required for growth might interfere with the efficient functioning of the SBT. A tool to eliminate TpiA towards the end of the growth phase is under construction and will be implemented into the SBT- producing strain. Utilization of this tool is going to improve the growth pattern of the mutant considerably, restoring its growth-properties on mineral media and facilitating the introduction of further gene deletions. DHAP is a rather instable molecule whose efficient purification from the SBT’s reaction mix would be nearly impossible. However, DHAP is an intermediate necessary for the DHAP-dependent aldolases to perform the synthesis of vicinal diols. The SBT therefore was extended with the consecutive reaction to synthesize DEXP from DHAP and butanal with FBA. The aldolase from E. coli was an inefficient DEXP producer and the rabbit muscle enzyme had to be added to ensure fast DEXP synthesis. Pending further investigations, this might make the SBT a nearly ideal system for extension with any one of the different DHAP-dependent aldolases responsible for the four different stereoselectivities. For a better understanding of the system behavior during DHAP synthesis, an MS- based real-time analytic tool was developed, allowing the on-line record of the intermediates and products of the reaction cascade. Unlike other systems, a very fast measurement was possible, collecting data every few seconds for every metabolite. The quantification was hampered by ion suppression effects, due to which only qualitative measurements were possible. Currently, there are investigations underway to enable quantification. With feasible quantification, bottlenecks inside the reaction pathway can be uncovered and thus targets for further optimization of the SBT can be found.
The concept of the SBT has been proven to be feasible for the production of vicinal diols from DHAP. This opens the door to many other applications. The SBT has the potential to completely replace intricate and time-consuming multi-step synthesis. A complex oligosaccharide might be synthesized by the combination of several SBT modules, each one responsible for the synthesis of one specific subunit. A further module to combine the saccharides to an oligosaccharide could complete the reaction system. Several of these modules could be working in one pot, enabling the replacement of the individual cofactor recycling of each SBT by another module. The degree of freedom in combination of these SBT modules would be nearly unlimited, significantly facilitating the synthesis of oligosaccharides and thus eventually increasing the availability of molecules potentially comprising pharmaceutical activities.
Acknowledgments 129
Acknowledgments
In the first place, I would like to thank Prof. Sven Panke for supervising my research. I am grateful to him for his input in my work and helping me from the woodway to the highway.
Special thanks to Prof. Victor de Lorenzo and Dr. Matthias Heinemann for their co- referees of my thesis and for many fruitful discussions during my time as PhD student.
Dr. René Pellaux, Anne Kümmel, Matthias Bujara, Stephan Gomolka and Esther Tijchon contributed to this thesis and I thank them cordially for their efforts.
As partners in EuroBioSyn, I thank Prof. Sten Jorgensen, Dr. Jürgen Pleiss, Dr. Belén Calles, Florin Davidescu and Michael Knoll for their input in the interesting and agreeable meetings.
I thank all the former and present members of the BPL – Agnes, Alex, Andreas, Anne, Dushan, Esther, Frauke, Giorgia, Giovanni, Hafid, Laetitia, Maartije, Marcel, Martin, Matthias Be., Matthias Bu., René, Riri, Simone, Sonja, and Stefan – for their support and friendship.
I would also like to thank Eva, to whom this thesis is dedicated, for the support and love she gave me during all the years of my student’s life, and hopefully will give me for a great many years yet to come. I also thank my and her family for always being there for me.
Curriculum Vitae 130
Curriculum Vitae
Michael Schümperli
Born 15. 01. 1977 in Chur (GR)
Citizen of Chur (GR) and Wäldi (TG), Switzerland
Education:
05. 2003 – 03. 2008 PhD in Bioreaction Engineering in the group of Prof. Dr. Sven Panke, Institute of Process Engineering, ETH Zurich.
10. 1997 – 02. 2003 Studies in Biotechnology at ETH Zurich, Zürich, Switzerland. Graduation as Dipl. Natw.
08. 1992 – 06. 1997 Kantonsschule Chur, Graubünden, Switzerland. Matura Typus C (Scientific).
08. 1983 – 06. 1992 Primary and Secondary School, Chur, Graubünden, Switzerland.