High Quality 13C Metabolic Flux Analysis Using GC-MS“

10 APPLIED MICROBIOLOGY

Volume 10

To reduce the demand for fossil resources, biotechnological processes are increasingly applied. Therefore, the biological hosts used for manufacturing of products of interest have to be optimized in order to achieve the highest possible yield. For this optimization, Andreas Schmitz which is usually carried out by means of metabolic engineering strategies, the intracellular processes need to be investigated in advance by methods such as 13C-Metabolic 13 Flux Analysis (MFA). High Quality C metabolic flux The present thesis focused on improving analytical techniques for the measurement of analysis using GC-MS 13C-labeling patterns of metabolites. Parameters influencing the data quality of existing analytical techniques based on gas chromatography-mass spectrometry (GC-MS) for the classic 13C-MFA, were investigated. Focusing on the analysis of proteinogenic amino acids, a detailed protocol for sample preparation and GC-MS analysis was established to enable scientists to easily produce high quality labeling data for 13C-MFA.

Additionally, sample preparation and measurement methods for labeling determination of intracellular metabolites were implemented and further developed. The application of the developed methods facilitates an in-depth analysis of the glucose uptake in Pseu- domonas and the examination of cyclic Entner-Doudoroff-Pathway fluxes. Compared to classic 13C-MFA, the use of intracellular metabolites increased the information content and therefore additional fluxes could be resolved.

Furthermore, the potential of GC-MS/MS analyses was investigated to increase the positional 13C-label information of the MS data by introducing a second fragmentation

13 C metabolic flux analysis using GC-MS

step. This positional information on C-isotopes can be exploited to resolve additional 13 metabolic fluxes. Leucine and lysine were found to be promising amino acids in terms of acetyl-CoA labeling determination when analyzed in tandem MS. Labeling experiments with Saccharomyces cerevisiae revealed differences in the labeling of cytosolic and mitochondrial acetyl-CoA, indicating that there is no significant exchange of these two compartmental pools. High Quality

The methods presented in this study are highly sophisticated and are suited for determination of the labeling of various metabolites. The information gained by the application of these methods is suitable for use in 13C-MFA and thus enables a deeper insight into the metabolic activity inside the cell. Andreas Schmitz „High Quality 13C metabolic flux analysis using GC-MS“

Von der Fakultät für Mathematik, Informatik und Naturwissenschaften der RWTH Aachen University zur Erlangung des akademischen Grades eines Doktors der Naturwissenschaften genehmigte Dissertation

vorgelegt von

M.Sc. Angewandte Chemie Andreas Schmitz

aus

Mechernich, Nordrhein Westfalen, Deutschland

Berichter: Universitätsprofessor Dr.-Ing. Lars M. Blank Universitätsprofessor Dr. rer. nat. Marco Oldiges

Tag der mündlichen Prüfung: 16.07.2018

Diese Dissertation ist auf den Internetseiten der Universitätsbibliothek online verfügbar. Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über https://portal.dnb.de abrufbar.

Andreas Schmitz:

High Quality 13C metabolic flux analysis using GC-MS

1. Auflage, 2018

Gedruckt auf holz- und säurefreiem Papier, 100% chlorfrei gebleicht.

Apprimus Verlag, Aachen, 2018 Wissenschaftsverlag des Instituts für Industriekommunikation und Fachmedien an der RWTH Aachen Steinbachstr. 25, 52074 Aachen Internet: www.apprimus-verlag.de, E-Mail: [email protected]

Printed in Germany

ISBN 978-3-86359-660-6

D 82 (Diss. RWTH Aachen University, 2018) EIDESSTATTLICHE ERKLÄRUNG

Eidesstattliche Erklärung

“Hiermit erkläre ich, Andreas Schmitz, an Eides statt, dass ich die vorliegende Dissertation selbstständig verfasst und keine anderen als die angegebenen Quellen und Hilfsmittel benutzt habe.“

DANKSAGUNG

Danksagung An dieser Stelle möchte ich mich gerne bei einigen Personen bedanken, ohne welche diese Arbeit nicht entstanden wäre. Zunächst möchte ich mich bei meinem Doktorvater Prof. Dr. Lars M. Blank für die Möglichkeit bedanken, dieses interessante Thema im Rahmen meiner Doktorarbeit am Institut für Angewandte Mikrobiologie bearbeiten zu dürfen. Danke auch für die guten Gespräche und Diskussionen. Ein großer Dank geht ebenfalls an Dr. Birgitta Ebert, welche mir als Gruppenleiterin bei der Bearbeitung der Doktorarbeit stets zur Seite stand. Viele anregende Diskussionen führten letztlich zur erfolgreichen Planung der durchzuführenden Experimente als Grundlage für die vorliegende Arbeit. Weiterhin bedanke ich mich recht herzlich bei Prof. Dr. Marco Oldiges, für die bereitwillige Übernahme des Zweitgutachtens. Danken möchte ich auch meinen Studenten Vincent Wiebach, Michael Osthege, Tobias Alter, Isabella Albert und Sarah Maurer, welche im Rahmen von Bachelor- und Masterarbeiten sowie Forschungspraktika zu dieser Arbeit beigetragen haben. Meinen Kolleginnen und Kollegen am Institut für Angewandte Mikrobiologie, Ahmed, Andrea, Annette, Arnoldine, Benedikt, Bernd, Birthe, Carola, Christoph H., Christoph L., Christoph T., Dario, Eda, Eik, Elena, Elke, Gisela, Hamed, Henrik, Jan, Jannis, Kalle, Kerstin, Lars K., Maike, Manja, Mathias, Martin, Monika, Nick, Rabea, Salome, Sandra H., Sandra S., Sebastian K., Sebastian Z., Suresh, Theresa, Thiemo, Till, Ulf, Ulrike, Wie und Wing-Jin möchte ich ebenfalls an dieser Stelle danken. Ihr habt die Zeit am Institut zu einer für mich unvergesslichen gemacht! Es war schön mit euch nicht nur auf der Arbeit, sondern auch abseits davon viele gemeinsame und spaßige Augenblicke verbracht zu haben. Besonders großer Dank gilt an dieser Stelle meiner gesamten Familie. Ohne die großartige Unterstützung meiner Eltern, Helga und Helmut Schmitz, wäre dies alles nicht möglich gewesen. Ihr wart immer für mich da und habt mir immer unter die Arme gegriffen, wo Ihr nur konntet. Danke dafür! Der größte Dank gilt meiner Freundin Hannah! Danke, dass Du an meiner Seite bist und danke für die vielen aufmunternden Worte sowie die Unterstützung, besonders während des Zusammenschreibens! Es war nicht immer einfach, aber Du hast mich ermutigt, die Arbeit zu vollenden!

DANKE!

III

TABLE OF CONTENTS

EIDESSTATTLICHE ERKLÄRUNG ...... I DANKSAGUNG ...... III SUMMARY ...... VII ZUSAMMENFASSUNG ...... IX LIST OF ABBREVIATIONS ...... XI LIST OF FIGURES ...... XIII LIST OF TABLES ...... XVII 1. GENERAL INTRODUCTION ...... 1 1.1. General motivation of flux quantification in organisms ...... 2 1.2. 13C-metabolic flux analysis ...... 3 1.3. NMR and GC-MS techniques for stable isotope distribution quantification ...... 5 1.3.1. Amino acids determination ...... 7 1.3.2. Intracellular metabolite determination ...... 10 1.4. GC-MS/MS technique ...... 13 1.4.1. Amino acid fragmentation ...... 13 1.5. Aims of the Thesis ...... 14 2. RESULTS AND DISCUSSION ...... 17 2.1. Mass isotopomer distribution measurements for high quality 13C-metabolic flux analysis ...... 19 2.1.1. Summary ...... 20 2.1.2. Introduction ...... 20 2.1.3. Materials and Methods ...... 22 2.1.4. Results and Discussion ...... 23 2.1.5. Conclusion ...... 34 2.2. GC-MS based determination of mass isotopomer distributions for 13C-based metabolic flux analysis ...... 37 2.2.1. Summary ...... 38 2.2.2. Introduction ...... 38 2.2.3. Materials ...... 41 2.2.4. Methods ...... 42 2.2.5. Notes ...... 52 2.3. Investigation of glucose uptake in Pseudomonas putida ...... 57 2.3.1. Summary ...... 58 2.3.2. Introduction ...... 58 2.3.3. Materials and Methods ...... 60 2.3.4. Results and Discussion ...... 62

V TABLE OF CONTENTS

2.3.5. Conclusion ...... 79 2.4. High quality 13C-metabolic flux analysis for examination of cyclic ED activity in different Pseudomonas putida strains ...... 81 2.4.1. Summary ...... 82 2.4.2. Introduction ...... 82 2.4.3. Materials and Methods ...... 85 2.4.4. Results and Discussion ...... 89 2.4.5. Conclusion ...... 98 2.5. Tandem MS measurements for the determination of mitochondrial and cytosolic acetyl-CoA labeling in Saccharomyces cerevisiae ...... 99 2.5.1. Summary ...... 100 2.5.2. Introduction ...... 100 2.5.3. Materials and Methods ...... 103 2.5.4. Results and Discussion ...... 105 2.5.5. Conclusion ...... 120 3. GENERAL DISCUSSION ...... 121 3.1. Possible pitfalls of 13C-metabolic flux analysis ...... 122 3.2. Advantages and disadvantages of various labeling data for 13C-metabolic flux analysis ...... 123 3.3. New dimension of 13C-metabolic flux analysis by use of GC-tandem MS ...... 125 4. CONCLUSION ...... 127 5. REFERENCES ...... 129 6. CURRICULUM VITAE ...... 139

VI SUMMARY

Summary In order to reduce the demand of fossil resources such as crude oil or natural gas, biotechnological processes are increasingly applied for product manufacturing using biological hosts. To enhance the production efficiency, these biological hosts are generally optimized by means of metabolic engineering. Since good knowledge of intracellular processes in the selected biological hosts is mandatory prior metabolic engineering, these intracellular processes have to be investigated in advance by methods such as 13C-Metabolic Flux Analysis (MFA). The present thesis was focused on improving analytical techniques for the measurement of 13C-labeling pattern of metabolites. First, parameters influencing data quality of existing gas chromatography-mass spectrometry (GC-MS) based analytical techniques for the classic 13C- MFA, which is based on the analysis of proteinogenic amino acids, were analyzed. Scan mode, integration method and the amount of biomass analyzed emerged as the main parameters impacting the quality of the labeling data. On average, a 3.5-fold improvement of the data quality was achieved with measurements run in Single Ion Monitoring (SIM) mode. We set up a detailed protocol for sample preparation and GC-MS analysis to enable scientists new to this field to easily produce labeling data of high quality for 13C-MFA. 13C-MFA using labeling data of intracellular metabolites is becoming increasingly relevant in biotechnological research. It allows to perform 13C-tracer experiments at shorter time scales thereby reducing cost for the 13C-tracer but more importantly makes the analysis applicable for systems not maintaining stable metabolic steady states. Therefore, sample preparation and measurement methods for labeling determination of intracellular metabolites were implemented and further developed. Due to the high turnover rates of intracellular metabolites during growth, special attention was paid to the rapid termination of intracellular processes. Metabolite quenching with cold ethanol-saline solution was applied and showed best results for Pseudomonas putida strains. The metabolism was stopped immediately and intracellular metabolites were extracted subsequently using a cold methanol-chloroform-water solution. The developed methods were used for an in-depth analysis of the glucose uptake in Pseudomonas and for the examination of cyclic Entner-Doudoroff-Pathway fluxes. Compared to 13C-MFA with proteinogenic amino acids, the use of intracellular metabolites increased the information content and therefore additional fluxes could be resolved. Besides classical GC-MS approaches, the potential of GC-MS/MS analyses to increase the positional 13C-label information of the MS data by introducing a second fragmentation step was investigated as such positional information about 13C-isotopes can be exploited to resolve additional metabolic fluxes. For these GC-MS/MS measurements, cells were grown and harvested as for classical analysis of amino acids with GC-MS. Leucine and lysine were found to be promising amino acids in terms of acetyl-CoA labeling determination when being fragmented in tandem MS measurements. Labeling experiments with 1-13C-glucose and U- 13C-glucose revealed differences in the labeling of cytosolic and mitochondrial acetyl-CoA indicating that there is no high exchange of these two compartmental pools. The methods presented in this study are highly sophisticated and are suited for determining the labeling of various metabolites. The information gained by application of these methods is suitable for use in 13C-MFA and thus enables a deeper insight into the metabolic activity in the cell.

VII

ZUSAMMENFASSUNG

Zusammenfassung Zur Reduzierung des Bedarfs fossiler Rohstoffe wie Erdöl oder Erdgas werden immer häufiger biotechnologische Prozesse angewandt, um Produkte unter Verwendung biologischer Wirtsorganismen herzustellen. Zur Steigerung der Produktionseffizienz werden diese Organismen meist mittels Metabolic Engineering optimiert. Bevor die Organismen mittels Metabolic Engineering verändert werden ist es notwendig, die intrazellulären Vorgänge zu kennen. Diese Vorgänge müssen im Vorfeld mit Methoden wie der 13C-Stoffflussanalyse (MFA) untersucht werden. Der Fokus dieser Arbeit lag auf der Optimierung der Analytik zur Messung der 13C- Markierung von Metaboliten. Zunächst wurde der Einfluss, der für die 13C-MFA gängigen Gaschromatographie-Massenspektrometrie (GC-MS) Methoden auf die Datenqualität von proteinogenen Aminosäuren untersucht. Der Scanmodus, die Integrationsmethode sowie die Menge der analysierten Biomasse stellten sich als Haupteinflussfaktoren auf die Qualität der Markierungsdaten dar. Im Durchschnitt konnte die Datenqualität durch die Messung mittels Selected-Ion-Monitoring (SIM) um den Faktor 3.5 verbessert werden. Hierfür erstellten wir ein detailliertes Protokoll, das Wissenschaftlern ohne Erfahrungen auf diesem Fachgebiet die Probenvorbereitung sowie GC-MS Analyse beschreibt und somit eine einfache Messung der Markierungsdaten zur 13C-MFA ermöglicht. In der biotechnologischen Forschung wird es immer relevanter auch Markierungsdaten von intrazellulären Metaboliten für die 13C-MFA zu verwenden. Hierdurch benötigen die Experimente weniger Zeit, wodurch die Kosten für 13C-Substrate gesenkt werden können. Als weitaus wichtiger kann jedoch die Anwendung der Analyse für Experimente betrachtet werden, bei denen das Einstellen des metabolisch stationären Zustandes nicht möglich ist. Dazu wurden Probenvorbereitungs- und Messmethoden zur Markierungsbestimmung von intrazellulären Metaboliten implementiert und weiterentwickelt. Aufgrund der hohen Umsatzraten von intrazellulären Metaboliten während des Wachstums wurde besonderes Augenmerk auf die unverzügliche Unterbrechung der intrazellulären Prozesse gelegt. Metabolit-Quenching wurde mit kalter Ethanol-Kochsalz-Lösung durchgeführt und die intrazellulären Metabolite anschließend mittels kalter Methanol-Chloroform-Wasser-Lösung extrahiert. Die entwickelten Methoden wurden für eine eingehende Untersuchung der Glukoseaufnahme sowie zur Untersuchung von zyklischen Flüssen durch den Entner- Doudoroff Stoffwechselweg in Pseudomonaden verwendet. Im Vergleich zu 13C-MFA mit proteinogenen Aminosäuren konnte der Informationsgehalt durch die Verwendung intrazellulärer Metabolite erhöht und somit zusätzliche Stoffflüsse aufgelöst werden. Neben klassischen GC-MS-Ansätzen wurden GC-Tandem MS (MS/MS) Analysen durchgeführt. Hierdurch sollten Informationen über die Position von 13C-Isotopen in der Aminosäuren durch einen zweiten Fragmentierungsschritt gewonnen werden, um zusätzliche Stoffflüsse aufzulösen zu können. Für diese GC-MS/MS Messungen wurden die Zellen wie bei der klassischen Analyse von Aminosäuren mit GC-MS vorbereitet. Leucin und Lysin erwiesen sich als vielversprechende Aminosäuren im Hinblick auf die Acetyl-CoA- Markierung von S. cerevisiae. Experimente mit 1-13C-Glucose und U-13C-Glucose ergaben Unterschiede in der cytosolischen und mitochondrialen Acetyl-CoA Markierung, was auf einen geringen Austausch zwischen diesen beiden Kompartimenten hindeutet.

IX Die in dieser Arbeit vorgestellten Techniken sind hochentwickelte Methoden zur Bestimmung der Metabolitenmarkierung. Die durch die Anwendung dieser Methoden gewonnenen Informationen, sind für die Verwendung in 13C-MFA geeignet und ermöglichen somit einen tieferen Einblick in die Stoffwechselaktivität in der Zelle.

X LIST OF ABBREVIATIONS

List of Abbreviations

13C-MFA 13C-metabolic flux analysis 6pg 6-phosphogluconate 6Pgd 6-phosphogluconate dehydrogenase aCoA Acetyl-CoA ATP adenosine triphosphate autoSRM automatic selected reaction monitoring C. glutamicum Corynebacterium glutamicum CDW Cell dry weight ddH2O bidest water DNA deoxyribonucleic acid E. coli Escherichia coli ED Entner-Doudoroff Eda 2-keto-3-deoxy-6-phosphogluconate aldolase Edd phosphogluconate dehydratase EI electron impact EMP Emden-Meyerhoff-Parnas f6p fructose-6-phosphate fbp fructose-1,6-bisphosphate FTICR-MS fourier transform ion cyclotron resonance-mass spectrometry g3p glyceraldehyde-3-phosphate g6p glucose-6-phosphate GC gas chromatography Glk glucokinase GUI graphical user interface kdpg 2-keto-3-deoxy-6-phosphogluconate L. lactis Lactococcus lactis LC liquid chromatography m/z mass to charge MBDSTFA N-methyl-N-(tert-butyldimethylsilyl)-trifluoroacetamide MDV mass distribution vector MFA metabolic flux analysis MS mass spectrometry MS/MS tandem mass spectrometry MSTFA N-methyl-N-trimethylsilyl-trifluoroacetamide NADH nicotinamide adenine dinucleotide NADPH nicotinamide adenine dinucleotide phosphate NMR nuclear magnetic resonance NoD norm of difference oaa oxaloacetate OD600 optical density P. fluorescens Pseudomonas fluorescens P. putida Pseudomonas putida p5p pentose-5-phosphate Pdh pyruvate dehydrogenase Pfk-1 phosphofructokinase-1 Pgi phosphoglucose isomerase Pgl 6-phosphogluconolactonase pyr pyruvate RI refractive index RNA ribonucleic acid

XI LIST OF ABBREVIATIONS

RT retention time S. cerevisiae Saccharomyces cerevisiae SIM selected ion monitoring TBDMS tert-butyl-dimethylsilyl TCA tricarboxylic acid TMS trimethylsilyl UV ultraviolet light UV/Vis ultraviolet–visible spectroscopy Zwf glucose-6-phosphate dehydrogenase

XII LIST OF FIGURES

List of Figures Figure 1: Central carbon metabolism of Lactococcus lactis (a) and Escherichia coli (b) during growth on glucose adapted from [5]...... 3 Figure 2: Schematic overview of flux ratio analysis and global iterative fitting starting with a 13C-tracer experiment adapted from [15]...... 5 Figure 3: Schematic structure of a quadrupole mass spectrometer adapted from [22]...... 6 Figure 4: Amino acids (b) and their precursor metabolites (a) from CCM with their corresponding compartment in S. cerevisiae adapted from [25]...... 8 Figure 5: Schematic demonstration of the characteristic fragmentations observable during GC-MS measurement of MBDSTFA derivatized amino acids. Resulting fragments are M-15 (a), M-57 (b), M-159 (c), M-85 (d) and f302 (e). Adapted from [24]...... 9 Figure 6: Oximation of pyruvate using methoxyamine [48]...... 12 Figure 7: Silylation of methoxylated pyruvate using MSTFA [48]...... 12 Figure 8: Schematic structure of a triple quadrupole mass spectrometer adapted from [22]. .. 13 Figure 9: Presentation of transition from parent spectrum mass isotopomer (Mn) to daughter ion (mn) spectrum and the description in a so called transition matrix adapted from [56]...... 14 Figure 10: Workflow of 13C-MFA. Shown are the individual steps from 13C-labeling experiment to the final calculation of flux distributions. Steps marked in red are investigated in the following chapters...... 24 Figure 11: Comparison of NoDs of MDVs calculated from peak area vs. height...... 25 Figure 12: Impact of scan time on MDV quality. Shown are the NoDs for the serine fragments f302 (black), ser m-15 (red), ser m-57 (green), ser m-85 (blue) and ser m-159 (purple) computed from measurements with scan times between 0.001 and 0.01 s...... 26 Figure 13: Impact of scan width on MDV quality. Shown are the NoDs for the serine fragments f302 (black), ser m-15 (red), ser m-57 (green), ser m-85 (blue) and ser m-159 (purple) computed from measurements with scan width between 0. 1 and 1 amu...... 27 Figure 14: Comparison of NoDs computed from MDVs measured with FullScan (180-550 m/z; 0.25 s) and SIM (0.5 amu; 0.002 s) mode...... 28 Figure 15: Comparison of SIM and FullScan measurements. Shown is the dependence of NoD on the amount of CDW used for the analysis. MDVs were determined from measurement in SIM (green) and FullScan (red) mode, respectively. NoDs of serine m-159 from biomass were determined from quadruplicate measurements...... 29 Figure 16: Impact of cell dry weight used for the analysis on MDV quality. Shown are the NoDs of serine fragments f302 (black), ser m-15 (red), ser m-57 (green), ser m-85 (blue) and ser m-159 (purple) measured in quadruplicates...... 30 Figure 17: Impact of cell dry weight used for the analysis on MDV quality. Shown are the NoDs of leucine fragments leu m-15 (red), leu m-85 (blue) and leu m-158 (purple) measured in quadruplicates...... 31 Figure 18: Influence of hydrolysis time on MDV quality of measured amino acid fragments. Shown is the NoD of the serine fragments f302 (black), ser m-15

XIII LIST OF FIGURES

(red), ser m-57 (green), ser m-85 (blue) and ser m-159 (purple) measured in quadruplicates...... 32 Figure 19: Peak area of selected amino acids over hydrolysis time. Shown are average peak areas of histidine (red) and leucine (green) measured with SIM mode...... 33 Figure 20: Influence of hydrolysis time on MDV quality of measured amino acid fragments. Shown is the NoD of the histidine fragments f302 (black), his m- 15 (red), his m-57 (green), his m-85 (blue) and his m-159 (purple) measured in quadruplicates...... 34 Figure 21: Mass spectrum of the ion cluster of a three-carbon atom molecule, with ions M0 to M3 each representing a cluster of isotopomers (a). Single MS measurements do not allow differentiation of positional isotopomers but are limited to measurement of mass isotopomers (b). Normalizing the mass intensities to the sum of all masses yields the mass distribution vector (c)...... 40 Figure 22: Total ion current chromatogram of unlabeled biomass of a full scan measurement (a) and selected ion monitoring chromatogram of unlabeled biomass (b) ...... 43 Figure 23: Single mass peaks of an ion cluster of alanine m-85 containing two carbon atoms of the amino acid carbon backbone. The mass peaks M0 to MnumC+1 (shown in the first four chromatograms) are required for complete mass distribution determination...... 47 Figure 24: Peak identification settings in the Thermo Xcalibur data processing module ...... 47 Figure 25: Peak detection settings in the Thermo Xcalibur data processing module ...... 48 Figure 26: Dependence of MDV quality, expressed as the Norm of Difference (NoD), of two alanine fragments m-57 (green) and m-85 (blue) in dependence of the amount of biomass used. A fixed hydrolysis time of 6 h was used...... 49 Figure 27: Structure of the tab separated value (TSV) file with MS data to be processed with iMS2Flux ...... 50 Figure 28: Parameter setting via the graphical user interface of iMS2Flux ...... 51 Figure 29: Structure of the data output file with corrected mass distribution data calculated by iMS2Flux from raw MS data ...... 52 Figure 30: Assessment of MDV quality from full scan vs. SIM mode measurements. Shown is the Norm of difference of the serine fragment m-57 as function of the amount of biomass used calculated from data from full scan (red) and SIM mode measurements (green) ...... 54 Figure 31: Peak integration with an area scan window of 10 scans (top) and 40 scans (bottom) ...... 55 Figure 32: Norm of difference (NoD) of the MDV of the aspartate fragment m-57 calculated from either peak areas (red) or peak heights (green) ...... 56 Figure 33: Glucose uptake in P. putida. Glucose is incorporated and transformed to g6p via Glk. Alternatively, glucose is oxidized in the periplasm via glucose dehydrogenase to gluconate and further to 2-ketogluconate via gluconate dehydrogenase. Gluconate and 2-ketogluconate are transported into the cell and transformed to 6pg. g6p is oxidized to 6pg using Zwf. 6pg is the initial metabolite of the ED-Pathway resulting in g3p and pyr...... 59 Figure 34: Time course of substrate concentration (a), labeling (b) and intracellular g6p/f6p labeling (c) of P. putida KT2440 using glucose and gluconate as

XIV LIST OF FIGURES

initial carbon sources. Error bars on fractional labeling data represent the relative error of 10 %...... 67 Figure 35: Time course of substrate concentration (a), labeling (b) and intracellular g6p/f6p labeling (c) of P. putida S12 using glucose and gluconate as initial carbon sources. Error bars on fractional labeling data represent the relative error of 10 %...... 68 Figure 36: Time course of substrate concentration (a), labeling (b) and intracellular g6p/f6p labeling (c) of P. putida KT2440 Δglk using glucose and gluconate as initial carbon sources. Error bars on fractional labeling data represent the relative error of 10 %...... 69 Figure 37: Time course of substrate concentration (a), labeling (b) and intracellular g6p/f6p labeling (c) of P. putida KT2440 using glucose and 2-ketogluconate as initial carbon sources. Error bars on fractional labeling data represent the relative error of 10 %...... 72 Figure 38: Time course of substrate concentration (a), labeling (b) and intracellular g6p/f6p labeling (c) of P. putida S12 using glucose and 2-ketogluconate as initial carbon sources. Error bars on fractional labeling data represent the relative error of 10 %...... 73 Figure 39: Time course of substrate concentration (a), labeling (b) and intracellular g6p/f6p labeling (c) of P. putida KT2440 Δglk using glucose and 2- ketogluconate as initial carbon sources. Error bars on fractional labeling data represent the relative error of 10 %...... 74 Figure 40: Fractional labeling of intracellular central carbon metabolites in growth experiments using P. putida KT2440 (green), P. putida S12 (red) and P. putida KT2440 Δglk (blue). The cells were grown on U-13C-glucose and 12C-gluconate...... 76 Figure 41: Fractional labeling of intracellular central carbon metabolites in growth experiments using P. putida KT2440 (green), P. putida S12 (red) and P. putida KT2440 Δglk (blue). The cells were initially grown on U-13C- glucose and 12C-2-ketogluconate...... 78 Figure 42: Flux map of Pseudomonads containing oxidative glucose uptake, ED pathway, EMP pathway, Pentose Phosphate (PP) pathway, tricarboxylic acid (TCA) cycle and glyoxylate shunt...... 83 Figure 43: Schematic flux map of glucose uptake (oxidative glucose uptake not shown), ED pathway and partial EMP pathway in Pseudomonads. The reactions greyed out represent a potential cyclic system of ED and EMP pathway. Glucose is incorporated and transformed to 6pg. In the ED pathway, 6pg is transformed to pyr and g3p which can further be transformed to fbp and f6p in EMP pathway. Since f6p cannot be transformed to fbp via Pfk-1, f6p is potentially channeled to the ED pathway or PP pathway via g6p...... 84 Figure 44: Time course of substrate and product concentrations correlated with growth curve of P. putida KT2440 (a), P. putida S12 (b) and P. taiwanensis VLB120 (c) using a mixture of 20% U-13C-glucose and 80% 1-13C-glucose as initial carbon source...... 92 Figure 45: Flux distributions of the investigated Pseudomonas strains. The flux distributions of P. putida KT2440, P. putida S12 and P. taiwanensis VLB120 (top to bottom) were calculated from labeling data obtained in growth

XV LIST OF FIGURES

experiments on minimal media supplemented with 20:80 U-13C-glucose and 1-13C-glucose as substrate. All fluxes were normalized to a glucose uptake of 100 mmol/g/h. Numbers in squared brackets represent the corresponding confidence intervals...... 96 Figure 46: Central carbon metabolism of S. cerevisiae with the two compartments cytosol and mitochondrion. The main pathways Emden-Meyerhoff-Parnas pathway, Pentose Phosphate pathway and tricarboxylic acid cycle are highlighted...... 102 Figure 47: Chromatogram of amino acid standard mix derivatized with MBDSTFA. Intensity and retention time (RT) of a 6.7 mM mix scanned in mass to charge (m/z) ratios from 50-500 m/z...... 106 Figure 48: Mass spectral data of leucine (a), aspartate (b) and lysine (c) measured by GC- MS in full scan mode from 50-500 m/z [127]...... 108 Figure 49: Chromatogram of amino acid standard mix derivatized with MSTFA. Intensity and RT of a 6.7 mM mix scanned from 100-550 m/z...... 110 Figure 50: Comparison of the TBDMS lysine derivative (a; derivatized with MBDSTFA) and as single (b) or double (c) silylated TMS derivative (derivatized with MSTFA) [127]...... 111 Figure 51: Mass spectral data of MSTFA derivatized lysine measured by GC-MS in full scan mode from 100-550 m/z [127]...... 111 Figure 52: Mass spectral data of leucine measured by GC-MS in full scan mode from 50- 500 m/z. M0 to M+3 mass isotopomers are exemplarily shown for leucine fragment f302 [127]...... 113 Figure 53: Parent to daughter ion transition of M0 (a) and M+1 (b) of leucine fragment f302 and corresponding daughter ion spectra measured by GC-MS/MS in scan mode from 199-205 m/z [127]...... 114 Figure 54: Comparison of naturally labeled leucine (a) and 13C-labeled leucine (b) fragment f302 with the correlating mass spectra...... 115 Figure 55: Physiological data of S. cerevisiae grown on an 80/20 (n/n) mixture of 1-13C- glucose and U-13C-glucose. Samples used for tandem MS measurements were harvested in mid exponential phase [127]...... 117

XVI LIST OF TABLES

List of Tables Table 1: Amino acid fragments occurring in GC-MS measurements and their corresponding mass of the monoisotopic peak. The masses greyed out occur in two different fragments of the same amino acid...... 10 Table 2: Recommended GC settings ...... 43 Table 3: Recommended MS settings ...... 44 Table 4: m/z ratios analyzed in SIM mode ...... 44 Table 5: Growth-phase dependent conversion (-) and accumulation (+) rates of glucose, gluconate and 2-ketogluconate for P. putida KT2440 (a), P. putida S12 (b) and P. putida KT2440 Δglk (c) grown on U-13C-glucose and 12C-gluconate...... 63 Table 6: Time dependent uptake (-) and accumulation (+) rates of glucose, gluconate and 2-ketogluconate. P. putida KT2440 (a), P. putida S12 (b) and P. putida KT2440 Δglk (c) were grown on U-13C-glucose and 12C-2-ketogluconate...... 64 Table 7: (a) Enzymatic activity of selected enzymes simulated in Cobra Toolbox. 13C- glucose was used as sole carbon source for simulating the flux distributions. The position of the selected enzymes in CCM is shown in Figure 42...... 90 Table 8: Measured fragments suitable for 13C-MFA. The fragments measured by GC-MS after hydrolysis or quenching as well as fragments measured by LC-MS after quenching are shown. “X” represents the method, with which the fragment was measured. The backbone carbon atoms contained in the fragment are indicated with “1”. The backbone carbon atoms which were cleaved of from the fragment are indicated with “0”...... 94 Table 9: Masses of characteristic fragments of MBDSTFA derivatized leucine, aspartate and lysine...... 107 Table 10: Parent - daughter ion combinations identified by autoSRM analysis with the corresponding fragmentation energies, daughter ion intensity and postulated parent-to-daughter transition. Combinations which are greyed out are not fulfilling the requirements of high intensity and informative fragmentation and were therefore not used in further investigations [127]...... 109 Table 11: Combination of parent- and daughter ions identified by autoSRM analysis with the corresponding fragmentation energies, daughter ion intensity and parent- to-daughter transition [127]...... 112 Table 12: Determination of tandem MS transitions for parent ion (M+1) to daughter ion. .... 114 Table 13: Determination of tandem MS transition for 13C-labeled parent ion to daughter ion...... 116 Table 14: Measured tandem MS transitions of leucine 302 > 200 gained from a 13C- labeling experiment using an 80/20 (n/n) mixture of 1-13C-glucose and U-13C- glucose. Abundances were corrected for natural labeling and original biomass using iMS2Flux. Data taken from [127]...... 118 Table 15: Measured tandem MS transitions of lysine 434 > 243 gained from a 13C- labeling experiment using an 80/20 (n/n) mixture of 1-13C-glucose and U-13C- glucose. Abundances were corrected by natural labeling and original biomass using iMS2Flux [127]...... 119

XVII

Chapter 1

General Introduction

1 GENERAL INTRODUCTION

1. GENERAL INTRODUCTION 1.1. General motivation of flux quantification in organisms Within the last decades, the demand for fossil resources such as crude oil and natural gas increased significantly. The growing utilization as energy source, fuel or as substrate for chemical production plays an essential role in depletion of these finite resources. Therefore, a shift from fossil to renewable resources, or in other words from a crude oil-based to a bioeconomy, is pursued. The usage of biological resources displays one possibility. These biological resources can be either biological material itself or products manufactured by the use of biological hosts in biotechnological processes. A broad range of substances which are currently manufactured based on crude oil can be produced biotechnologically. This includes bio fuels, pharmaceutically valuable substrates or fine chemicals [1-3]. Changing the metabolic network of an organism in order to enhance the flux towards a product of choice is of great interest. The development is supported by metabolic engineering with targeted genetic modifications for manipulating microbial hosts. Modification of an organism can be carried out more targeted and effective if the intracellular processes of the cell are known in detail. Obtaining information about intracellular processes can be beneficial for both the general understanding of an organism, as well as verification of changes introduced by metabolic engineering. The phenotype of the manipulated organism can be compared to the phenotype of the unaffected organism and the efficiency of the engineered organism can be investigated. The information to be compared in the assessment of the intracellular processes describes the use of individual intracellular reaction rates (fluxes) within the cell. These fluxes are finally quantified in order to compare the unaffected and manipulated organism. In organisms with linear metabolic pathways, the intracellular carbon fluxes can simply be quantified by balancing substrate uptake and product formations rates. Simple metabolic networks like in Lactococcus lactis (Figure 1a), can therefore be investigated under the assumption of stationary conditions, meaning constant metabolite pool sizes within the investigated time span [4,5].

2 CHAPTER 1

Figure 1: Central carbon metabolism of Lactococcus lactis (a) and Escherichia coli (b) during growth on glucose adapted from [5].

In organisms with parallel, branched, and cyclic metabolic pathways not directly connected to fluxes that can be measured, like in Escherichia coli (Figure 1b), a stoichiometric approach for quantification of carbon fluxes is insufficient due to the high degree of freedom. Hence, these equation systems are composed of too few equations with too many unknowns wherefore the number of equations has to be enhanced. 13C-based metabolic flux analysis (13C-MFA) is used for enhancing the number of equations and allows the deciphering of those. This is achieved by balancing unresolvable fluxes integrating additional constraints from stable isotope data reflecting information about carbon utilization within the metabolic network [4-6].

1.2. 13C-metabolic flux analysis The method of 13C-MFA has been developed as an extension of stoichiometric flux analysis. Due to the investigation of carbon utilization, complex networks with parallel and cyclic

3 GENERAL INTRODUCTION pathways can be resolved. In 13C-MFA, the cells are fed with a substrate labeled with stable 13C-isotopes at predetermined positions. Typically used 13C-isotope labeled substrates are uniformly labeled U-13C-glucose or 1-13C-glucose [7,8]. Alternative 13C-isotope labeled substrates such as acetate, lactate, and methanol have already been reported in several studies [9,10]. Since the labeled substrates are taken up by the organism, they are incorporated into metabolites and biomass. Depending on the metabolic pathways used for the formation of metabolites, differences in the 13C-isotope distribution can be observed within the metabolites [11]. The cells are harvested and subsequently extracted or hydrolyzed, whereby the metabolites of interest become detectable. These metabolites are measured and their 13C- labeling is detected. In combination with growth data such as biomass formation and substrate uptake rates, these labeling data are used for computationally solving the equation system. These computations can be carried out with the support of various software programs such as OpenFLUX [12], 13CFLUX2 [13], Metran or FiatFlux [14]. These programs differ regarding the underlying algorithm [15]. In OpenFLUX, 13CFLUX2 and Metran global iterative fitting is used for metabolic flux calculation. 13C-labeling patterns of measured metabolites and extracellular rates such as uptake and production rates are added as constraints by the user. Based on a model of the cellular carbon metabolism, semi-random flux distributions are used for simulation of theoretical 13C-labeling patterns and extracellular rates. This procedure is repeated, starting from the previously calculated parameters until the deviation between the simulated and measured data is reaching a minimum. The whole metabolic network is used for flux calculation, directly resulting in absolute flux values for every flux within the model. The flux calculation approach used in FiatFlux is the so called flux ratio analysis [14]. In contrast to global iterative fitting, the flux ratio analysis only focusses on selected fluxes affected by different pathways and therefore local labeling changes are considered. In this case, the labeling data of selected metabolites are used for calculating the relative amount this metabolite is formed via given pathways. Extracellular rates are not needed for determination of flux ratios (Figure 2) [8].

4 CHAPTER 1

Figure 2: Schematic overview of flux ratio analysis and global iterative fitting starting with a 13C-tracer experiment adapted from [15].

Since flux ratio analysis only focusses on local fluxes, a number of different local flux ratios is necessary for constraining the entire network. For this purpose, a high number of comparable labeling patterns is required, which complicates the overall interpretation of the network. Since comparison of different experiments is facilitated with absolute values, a quantification of the respective fluxes within the network is necessary. Therefore, extracellular rates are needed, which are used in combination with the previously calculated flux ratios for the quantification of metabolic net fluxes [4]. Besides 13C-labeling patterns, physiological data is needed for quantification of metabolic fluxes. These physiological data imply various factors that describe the growth of the investigated organism. First of all the growth rate has to be named, which describes the amount of cells formed within a defined time and volume. Also the uptake and production rates of extracellular metabolites are required. A precise quantification of the metabolic fluxes can only be made if the C balance is closed with the help of the data mentioned before.

1.3. NMR and GC-MS techniques for stable isotope distribution quantification The isotope distribution of metabolites of interest, extracted from cells grown with 13C- labeled substrates can be determined by spectroscopic or spectrometric methods. Here, nuclear magnetic resonance (NMR) spectroscopy [16] as well as mass spectrometry (MS) or tandem mass spectrometry (MS/MS) coupled to gas- (GC) or liquid (LC) chromatography are used most for labeling determination [17,18].

5 GENERAL INTRODUCTION

The most commonly used technique for labeling determination is 2-dimensional [13C-1H]- COSY NMR spectroscopy. As this technique is based on measuring the nuclear spin of 13C, a direct interpretation of 13C-13C-coupling within the analyte and hence valuable positional 13C- information is gained [19,20]. This technique has extensively been reported for the measurement of proteinogenic amino acids, which were obtained by hydrolyzing the cellular proteins using hydrochloric acid at high temperatures. This powerful technique has the advantage of measuring various analytes at the same time without complex sample preparation prior analysis [21]. However, due to the low sensitivity of 13C-NMR high amounts of analytes within the sample as well as a high labeling fraction within these analytes are required for sufficient resolution [16]. In biological labeling experiments, often the analyte concentration is the limiting factor due to the experimental conditions and the high cost of labeled substrates. Although the information content of NMR spectroscopic measurements is very high concerning labeling positions in the metabolites, mass spectrometry has evolved to be the most used technique for 13C-MFA within the last 15 years. A reason for this is the higher sensitivity of MS in comparison to NMR spectroscopy. Furthermore, MS is a very versatile tool as it can be coupled to gas- or liquid chromatography, whereby the measurements concerning metabolic flux analysis will focus on GC-MS in the following chapters. This technique is used in many laboratories and enables separation of very complex sample mixtures within the chromatographic system prior to mass spectrometry (Figure 3). Analytes are entering the mass spectrometer, which works under high vacuum and reach the ion source where they are bombarded with an electron beam. Due to this bombardment, specific chemical bonds within the analytes are destroyed resulting in distinct fragmentation of the analytes. The charged fragments are focused via charged lenses and accelerated into the mass separator. The most common used mass separator technique is the quadrupole mass filter consisting of four cylindrical rods providing an altering electrical field. Depending on this electrical field, selected mass to charge (m/z) ratios are stabilized in flight and are able to reach the detector. The intensity of the m/z ratio passing the quadrupole is recorded by an electron multiplier. Fast changes in the quadrupoles electrical field lead to rapid scanning of different m/z ratios and therefore a mass spectrum representing the intensities of measured m/z ratios is recorded.

Figure 3: Schematic structure of a quadrupole mass spectrometer adapted from [22].

6 CHAPTER 1

One fragment however, shows not only one single m/z ratio within the mass spectrum, but several m/z ratios due to incorporated heavy isotopes. The lightest m/z ratio of a fragment (M0) represents the measured fragment with solely light isotopes. Incorporation of heavy isotopes of C, O, N, H, P, and Si lead to increasing fragment mass and therefore increasing m/z ratios. Thus, a single heavy isotope increases the mass M0 by 1 and therefore a signal for m/z M1 can be observed. Depending on the fragment size, the number of M0+n is varying. Incorporation of 13C-atoms due to a labeling experiment leads to rising intensity of high m/z ratios for the measured fragment. In contrast to NMR spectroscopic measurements, MS does not directly lead to information about labeling positions, as it cannot measure interferences between the atoms [16,23]. Here, only the relative amount of heavy isotopes within a fragment is measured. This relative amount represents the sum of all isotopomers with the same mass, also called mass isotopomers. In GC-MS measurements for 13C-MFA the abundance of all mass isotopomers is used for the determination of mass isotopomer distributions and are normalized to a mass distribution vector (MDV). If fragments containing different carbon backbone atoms are available, the labeling state of selected atoms of the original analyte can be determined by creating an equation system. This equation system combines the measured relative abundance of mass isotopomers as a function of the relative abundance of positional isotopomers. As soon as the labeling information is covered by the measured carbon backbone atoms, mathematical solutions of this equation system will lead to positional labeling information of the investigated metabolite. 1.3.1. Amino acids determination Typically used metabolites for 13C-MFA are intracellular metabolites as well as proteinogenic amino acids. In contrast to free intracellular metabolites with small pool sizes, proteinogenic amino acids represent roughly 50% of the biomass. These proteinogenic amino acids are formed from precursors placed in central carbon metabolism (CCM) and linked by peptide bonds followed by incorporation into biomass. Typical pool sizes of intracellular metabolites within CCM are orders of magnitude smaller than those of proteinogenic amino acids [8]. In case of substrate depletion, the pool sizes of free metabolites rapidly decrease (sub-seconds to low minutes), whereby the pool sizes of proteinogenic amino acids change rather slow (higher minutes to hours). Therefore, the possibility of sampling without introducing changes in the labeling patterns is simple, as no change in labeling is observed in minutes, explaining why proteinogenic amino acids are most often used for 13C-MFA. A straight forward sampling method has been published, in which the biomass is harvested by centrifugation, the pelleted biomass is hydrolyzed using hydrochloric acid at a temperature around 90° C and the hydrolysate dried afterwards [24]. The remaining amino acids within the dried residue have subsequently to be prepared for the subsequent measurement. Since the amino acid metabolism is well-known, the amino acid labeling can be used to infer the labeling patterns of the corresponding precursors, i.e., of central carbon metabolites. Here, several studies have been reported, declaring the correlation of amino acid carbon atoms and their origin. After Szyperski reported the relationship of amino acids and their precursors in E. coli in 1996, Maaheimo et al. reported this relationship in Saccharomyces cerevisiae in 2001 (Figure 4).

7 GENERAL INTRODUCTION

Figure 4: Amino acids (b) and their precursor metabolites (a) from CCM with their corresponding compartment in S. cerevisiae adapted from [25].

For GC-MS analysis of amino acids, the dried samples need to be dissolved within a water free solvent and derivatized prior measurement. The most commonly used derivatizing reagent for amino acid measurement with GC-MS is N-methyl-N-(tert-butyldimethylsilyl)- trifluoroacetamide (MBDSTFA). All hydroxyl and amino groups of the amino acids are chemically modified by addition of a tert-butyl-dimethylsilyl (TBDMS) group. During mass spectrometric analysis of TBDMS derivatized amino acids, the analytes are cleaved into characteristic fragments after electron impact ionization. As illustrated in Figure 5, various groups can be split off the analyte, resulting in fragments that differ in the number of carbon atoms in the molecule backbone. The five most abundant fragments found for most TBDMS derivatized amino acids are described in the following. Two fragments containing the complete carbon backbone are measurable in case of methyl- (M-15) or tert-butyl-group (M- 57) cleavage. Another two fragments containing all carbon backbone atoms except of C1, can be determined when a C(O)O-TBDMS ion (M-159) or a CO2 molecule in combination with a tert-butyl-group (M-85) is cleaved off. The last typical fragment (f302) includes C1 and C2 of the carbon backbone and occurs due to breaking of the C-C bond between C2 and C3 [15,24].

8 CHAPTER 1

Figure 5: Schematic demonstration of the characteristic fragmentations observable during GC-MS measurement of MBDSTFA derivatized amino acids. Resulting fragments are M-15 (a), M-57 (b), M-159 (c), M-85 (d) and f302 (e). Adapted from [24].

Proteinogenic amino acids analyses leads to an overall of 99 possible fragments, which can potentially be used for 13C-MFA. For glycine the fragment f302 cannot be found, because this amino acid only has two carbon backbone atoms. For glycine the largest fragment identified is M-15 with 288 m/z. During acidic hydrolysis, the amino acids asparagine and glutamine are oxidized to aspartate and glutamate, while the amino acids cysteine and tryptophane are completely degraded. Additionally, arginine concentration is below limit of detection in GC-MS measurements [26]. Therefore, these five amino acids cannot be used for 13C-MFA. In GC-MS measurement fragments with different structures but equal m/z ratios are determined and pooled in a single signal. These overlapping fragments can therefore not be distinguished and cannot be used for 13C-MFA. Exemplarily, this phenomenon can be observed for the fragment f302 in measurements of alanine (M-15), isoleucine (M-57), leucine (M-57), and threonine (M-159). Since the carbon backbone of these overlapping fragments differ, no unambiguous allocation of the measured mass isotopomer distributions and carbon atoms within the fragments can be made. Therefore, these overlapping fragments cannot be used for further metabolic flux analysis. Nevertheless, 66 possible fragments can be measured and their mass isotopomer distribution can be clearly assigned to the carbon backbone atoms (Table 1) and be used for 13C-MFA.

9 GENERAL INTRODUCTION

Table 1: Amino acid fragments occurring in GC-MS measurements and their corresponding mass of the monoisotopic peak. The masses greyed out occur in two different fragments of the same amino acid. amino acid (M-15) (M-57) (M-85) (M-159) f302 alanine 302 260 232 158 302 aspartate 460 418 390 316 302 glutamate 474 432 404 330 302 glycine 288 246 218 144 histidine 482 440 412 338 302 isoleucine 344 302 274 200 302 leucine 344 302 274 200 302 lysine 473 431 403 329 302 methionine 362 320 292 218 302 phenylalanine 378 336 308 234 302 proline 328 286 258 184 302 serine 432 390 362 288 302 threonine 446 404 376 302 302 tyrosine 508 466 438 364 302 valine 330 288 260 186 302

1.3.2. Intracellular metabolite determination As an alternative to measurements of proteinogenic amino acids for 13C-MFA, the labeling patterns of intracellular metabolites can be measured directly. These measurements provide more information as amino acids are only linked to a limited number of central carbon metabolites. Direct measurement of the labeling pattern of intracellular metabolites increases the information available to constraint the flux models. In contrast to proteinogenic amino acids, which contribute as mentioned above about 50% to the total biomass, free metabolites account depending on growth conditions to about 1 to 3 % of biomass. In metabolic steady state the intracellular metabolite pool size stays constant and equilibrium of metabolite formation and usage is assumed. Due to high turnover rates of the intracellular metabolite pools, changes in growth and hence in pathway usage rapidly affect the concentrations as well as labeling patterns of the metabolites [27]. Exemplarily, in case of glucose-6-phosphate (g6p) and adenosine triphosphate (ATP), these turnover rates were found to be in a range of 1 – 2 s, indicating a complete exchange of molecules within the metabolite pool in this time [28-30]. In contrast to this rapid turnover of intracellular metabolites, turnover rates of several minutes or longer can be found for proteinogenic amino acids as mentioned above [15]. These high turnover rates result in more difficult harvesting for measurement of these metabolites since changes in metabolite labeling can cause false results in 13C-MFA. These rapid adaptations of the organism’s metabolism to growth conditions have to be considered for later sample treatment. Therefore, the biggest challenge is sampling, since the metabolism has to be stopped rapidly, otherwise changes in pool size or metabolite labeling occurs during harvesting. Various techniques for fast sampling combined with interruption of metabolism have been reported. One can distinguish methods were the culture broth is directly quenched and

10 CHAPTER 1 methods with prior separation of cells and supernatant. In the first approach, the cultures are directly quenched, while the whole culture broth is treated either with cold or hot solvents. Depending on the investigated organism, various combinations of temperatures and solvents have been published. Eukaryotic organisms such as Saccharomyces cerevisiae are most often quenched using methanol (60 % v/v) solutions at temperatures below -40 °C [31-33]. The investigated organisms show cell leakage due to the combination of solvent and cold temperature. This cell leakage results in loss of intracellular metabolites with the result of even lower analyte concentrations causing poor analytical detectability and reproducibility [34]. Even though this cell leakage does not directly affect the quality of labeling determination, various combinations of applied solvents and temperatures for cell quenching were reported. These studies investigated the reduction of cell leakage for selected organisms in order to quantify metabolite pools. Strategies with boiling or cold ethanol, boiling sodium hydroxide, cold glycerol saline, and liquid nitrogen were tested for various organisms such as S. cerevisiae, Escherichia coli, Corynebacterium glutamicum, Lactococcus lactis, and Pseudomonas fluorescens. However, these different quenching strategies showed varying degrees of success with regard to reproducibility and analyte quantity. Reproducible results were observed for approaches using cold ethanol sodium chloride solution (40 % v/v) at -20 °C for gram negative (E. coli), gram positive (C. glutamicum), and eukaryotic (S. cerevisiae) organisms [32]. The second approach utilizes cell quenching and metabolite extraction after previous removal of the supernatant. As an example, fast filtration is a common way of separating the microorganism from the medium. Fast filtration often represents the only way of separating, e.g., filamentous growing organisms due to their poor ability of forming a cell pellet during centrifugation [35]. Most commonly, fast filtration is performed by the use of filter membranes in combination with vacuum or pressured air for enhancement of filtration speed. After separation of cells and supernatant, the filter is immediately soaked in liquid nitrogen for interruption of metabolism. Metabolite extraction from biomass is performed in a subsequent step [36-38]. Another way of beforehand separation of the organism and supernatant is silicone oil / perchloric acid quenching [39]. The culture broth is placed on top of a silicone oil layer, which in turn is located upon a perchloric acid layer. Centrifugation transfers the cells into the perchloric acid phase, while the supernatant remains on top of the silicon oil phase. When the cells reach the perchloric acid phase, they are quenched and extracted simultaneously [40,41]. However, fragile metabolites such as some phosphorylated compounds are not stable in acidic conditions leading to changes in metabolite composition during extraction [42]. When not treated with perchloric acid, the cells have to be extracted after harvesting and quenching. Various extraction methods can be used depending on the investigated organisms and metabolites of interest. The used method needs to be appropriate for both cell lysis and metabolite extraction. Extraction can either be performed at low or high temperatures. For cold extraction, pure methanol at – 40°C can be applied, while cell disruption may be enhanced by application of freeze-thaw cycles. Also, cold perchloric acid can be used for quenching at cold temperatures in combination with freeze-thaw cycles [43-45]. Hot extraction methods can be executed using boiling ethanol, boiling methanol or boiling α- aminobutyrate solution [37,43,44]. The solvents for metabolite extraction are most often mixed with buffers for stabilization of pH and therefore stabilization of the extracted metabolites. Afterwards, the extracts can be treated with chloroform, removing non-polar

11 GENERAL INTRODUCTION metabolites for reduction of analytical background. In addition to extraction with a single solvent followed by phase separation, a mixture of organic and aqueous solutions can be applied for direct extraction. Here, one-phase mixtures of methanol, chloroform, and water or buffer can be used for metabolite extraction at -20 °C or 37 °C [43,46,47]. Due to cell treatment with both polar and non-polar solvents, the extraction of analytes is enhanced. Depending on the analytes of interest, either the organic or aqueous fraction is used for analysis. However, mechanical treatments like vortexing or shaking are most often used for supporting cell disruption. The extracts are subsequently removed from cell debris by filtration or centrifugation and are finally dried for later analysis. Prior to GC-MS analysis the dried metabolites have to be solved and derivatized. A two-step derivatization is a common technique for the subsequent measurement of intracellular metabolites. Within the first step of derivatization, aldehydes and ketones are oximized using methoxyamine hydrochloride (Figure 6).

Figure 6: Oximation of pyruvate using methoxyamine [48].

This oximation reaction is used for thermostabilization of the analytes to withstand the temperature profile within the GC. Oximation of ketones and aldehydes results in reduced structure changes like in oxo-cyclo-tautomerism of glucose [49]. The number of possible glucose structures can therefore be reduced, resulting in fewer peaks in GC-MS analysis [50,51]. Within the second step of derivatization, hydroxyl- and amino-groups are treated with N- methyl-N-trimethylsilyl-trifluoroacetamide (MSTFA, Figure 7). The derivatization with MSTFA reduces the boiling point of the analytes and therefore increases volatility required for gas chromatography. Furthermore, silylated analytes are more thermostable, lowering thermal induced decay within the gas chromatographic system [22,52].

Figure 7: Silylation of methoxylated pyruvate using MSTFA [48].

12 CHAPTER 1

1.4. GC-MS/MS technique Mass spectrometry is the most widely used technique for the measurements of 13C-labeling patterns and methods for the measurement of proteinogenic amino acids as well as intracellular metabolites have been published [53,54]. Recently published studies showed that MS/MS promises a significant improvement of 13C- labeling measurements [55-57]. These measurements can be carried out with a triple quadrupole mass spectrometer (Figure 8). Here, comparable to a quadrupole MS, the analytes are fragmented within an ion chamber and subsequently separated by mass within a quadrupole. Selected masses passing the first quadrupole are fragmented a second time within the second quadrupole, acting as a collision cell. These new arising masses are transferred into the third quadrupole and separated again by mass prior detection [58].

Figure 8: Schematic structure of a triple quadrupole mass spectrometer adapted from [22].

A second fragmentation of fragments provides a new dimension of metabolite measurement and thus new possibilities for direct position determination of 13C-labeled atoms. GC-MS/MS measurements can therefore be used for combining the advantages of both NMR and classical GC-MS measurements. Hence, the positional labeling can be determined directly using a measurement technique with a high sensitivity such as GC-MS [55]. For sufficient determination of metabolic fluxes, the allocation of measured labeling information has to be clarified prior computational analysis. Therefore, the carbon atoms within the measured fragments in GC-MS/MS have to be clearly assigned to the atoms of the metabolite and thereby to their precursor metabolites. The origin of all measurable fragments has to be determined in advance to labeling measurements. Here, either measurements of labeled standards or computational fragment predictions can be used for position determination [57,59,60]. 1.4.1. Amino acid fragmentation Amino acids are promising metabolites for GC-MS/MS measurements due to their high concentration within the cells and straight forward sample preparation. The typical fragments occurring in the ion source of an MS are already known and therefore can be used for further fragmentation in tandem mass spectrometry. Within the collision cell of the tandem MS, a single mass isotopomer is used as parent ion (Mn) and fragmented by collision with argon. This parent ion fragmentation results in new ions called daughter ions (mn). The transition of

13 GENERAL INTRODUCTION every single mass isotopomer within the parent fragment to the new generated daughter ion spectrum is recorded and the results displayed in a transition matrix (Figure 9).

Figure 9: Presentation of transition from parent spectrum mass isotopomer (Mn) to daughter ion (mn) spectrum and the description in a so called transition matrix adapted from [56].

The results of tandem MS represented in the transition matrix can be used as mass distribution vectors demonstrating the relative abundances of every transition. The enhancement of data quality applying GC-MS/MS has previously been reported, since the use of tandem MS data led to 2- to 5-fold smaller standard deviations of determined fluxes within a gluconeogenesis model [56,57]. Comparison to GC-MS data showed better resolvability of fluxes when tandem MS data was used, as the number of measurable positional isotopomers is increased, enhancing the accuracy of labeling data used for 13C-MFA [55,56].

1.5. Aims of the Thesis The aim of this thesis was the establishment of high quality metabolic flux analysis using 13C- tracer for the deciphering of biochemical mysteries that could not be elucidated with existing techniques. 13C-data quality evaluation was first in focus to reevaluate previously reported parameters from analyte quantity to GC-MS settings. The resulting protocol allowed simple sample preparation and data acquisition at high quality especially for proteinogenic amino acids. To further standardize the protocol a semi-automation using an auto sampler was developed. Many of the biochemical mysteries can only be inferred from labeling patterns of intracellular metabolites, hence the protocol was extended by including optimized quenching of bacteria to stop rapidly enzymatic activities. A list of central carbon metabolites was established that

14 CHAPTER 1 covered metabolites suitable for gas chromatographic measurements resulting in high quality data for later 13C-MFA. The information content of molecules larger than 3 carbon atoms cannot be fully resolved using standard GC-MS. Hence, a GC-MS/MS protocol for selected proteinogenic amino acids was established to recover most information from the analytes. The battery of adapted and improved protocols was used to establish high quality carbon flux distributions in Pseudomonas strains. A first focus was on the quantification of a proposed cyclic operation of the Entner-Doudoroff (ED) pathway. Batch fermentations with mixtures of 1-13C- and U-13C-glucose were carried out and both amino acids and free metabolites were isolated and prepared for GC-MS analysis. The resulting labeling data of amino acids and intracellular metabolites were used for computational quantification of the cyclic ED pathway. Pseudomonas putida is able to convert glucose within the periplasmic membrane to gluconate and further to 2-ketogluconate. For all three compounds a transporter exist and hence all three compounds are direct carbon sources that can be metabolized to 6-phosphogluconate (6pg) representing the initial metabolite of the ED pathway. Since the three possibilities to synthesize 6pg from glucose differ in their cofactor balance, it is of fundamental and applied interest to quantify the fluxes through these alternative pathways. Again, the here established methods for sampling, derivatization, and GC-MS analysis were used with the aim to decipher the alternatives in glucose consumption quantitatively.

Chapter 2

Results and Discussion

Chapter 2.1

Mass isotopomer distribution measurements for high quality 13C metabolic flux analysis

19 MASS ISOTOPOMER DISTRIBUTION MEASUREMENTS

2. RESULTS AND DISCUSSION 2.1. Mass isotopomer distribution measurements for high quality 13C-metabolic flux analysis 2.1.1. Summary Metabolic flux analysis (MFA) is a powerful tool for intracellular flux quantification yielding insight into the cellular behavior. In 13C-MFA, cells are fed using a 13C-labeled substrate, followed by measurement of amino acids and mass distribution vectors (MDV) determination. Precise flux quantitation requires highly accurate physiological data as well as a well-defined metabolic model. Additionally, accurate determination of MDVs is a crucial requirement. In this work, the key steps from harvesting of biomass to measurement of mass abundances were investigated and their influence on MDV quality evaluated. Starting with investigation of the measurement mode in mass spectrometric determination, an average 3.5 fold increase in quality of the MDVs was achieved for analytes measured in selected ion monitoring in comparison to full scan mode. Furthermore, the impact of amount of sampled biomass and hydrolyzation time on MDV quality was studied. Finally, measurements of samples manually or automatically derivatized were compared. 2.1.2. Introduction Stable 13C-isotope based metabolic flux analysis (13C-MFA) is a key methodology for the quantification of reaction rates (fluxes) in biological systems. The knowledge of metabolic fluxes provides insight into the cellular physiology, which is of importance for example in biomedical research to elucidate the emergence of diseases or for verification of successful flux redirection in metabolic engineering [8]. In this approach, cells are fed with a specifically 13C-labeled substrate [8,15]. This substrate is taken up by the cells and the 13C-tracer incorporated into metabolites. Suitable substrates for performing labeling experiments can be both single isotopomers of a substrate with labeled C atoms at known positions and mixtures of differently labeled substrates. For example, 100 % 1-13C-glucose with labeling on the first C atom can be used or a mixture of U-13C-glucose, wherein each C atom is labeled, and naturally labeled glucose. A mixture of U-13C-glucose and 1-13C-glucose is another possible substrate composition. The distribution of the heavy isotopes in metabolites is a function of the metabolic network structure and the activity of alternative pathways, i.e. the metabolic flux distribution. Consequently, 13C-enrichment data of metabolites provides information from which the in vivo fluxes can be inferred. Different labeling strategies result in different metabolite labeling patterns, which vary in their power to resolve metabolic fluxes. Consequently, a careful a priori design of the 13C-tracer is mandatory and the optimal design depends on the specific biological question to answer, which governs the importance to precisely resolve particular fluxes or pathways in the metabolic network. [61]. Often, the 13C-label distribution of proteinogenic amino acids is measured instead of free metabolic intermediates, due to the high abundance, stability and easy analytics [62]. Isotopic labeling patterns of amino acids can be determined by spectroscopic methods, such as nuclear magnetic resonance (NMR) spectroscopy [63-65] and spectrometric methods, such as mass spectrometry (MS) coupled to gas- (GC) or liquid chromatography (LC) [23,66,67]. Although the information content of NMR measurements is higher as positional labeling data are obtained, mass spectrometry based methods are more often used owing to the high sensitivity.

20 CHAPTER 2.1

Mass spectrometry separates analytes by mass and gives information about the distribution of mass isotopomers, i.e., isotope isomers differing in the number of incorporated heavy isotopes. These mass isotopomer distributions determined by mass spectrometry are represented by mass distribution vectors (MDVs) and carry information about the activity of intracellular reactions, which can be used to compute metabolic fluxes. The measured labeling patterns of amino acids are used for metabolic flux analysis. By applying suitable software such as 13C-FLUX2, OpenFLUX (both based on global iterative fitting [6,12]) and FiatFlux (flux ratio analysis [14]), a flux distribution compliant with the measured extracellular rates and amino acid labeling can be calculated. Special attention needs to be paid on the quality of the metabolic models used. The correctness of the model must be ensured, since even an inadequate model can lead to a good fit of the experimental data while yielding an erroneous flux distribution [68]. Therefore, it is necessary that the stoichiometry of the model is complete and the carbon atom transitions correctly defined. Furthermore, for organisms with more than one compartment, the exact assignment of a reaction to the respective compartment has to be stated within the model [69]. For high quality 13C-MFA, various requirements have to be fulfilled. On the one hand, a high precision and accuracy of physiological data such as growth rate and extracellular uptake and secretion rates has to be ensured. On the other hand, the samples for 13C-MFA have to be harvested during a metabolic and isotopic steady state. In chemostat cultivations, this isotopic steady state is assumed to be achieved after 5 volume changes at constant conditions. In a batch cultivation under unrestricted growth conditions, the cells harvested in mid exponential growth phase are assumed to be in an isotopic steady state. As chemostat and batch cultivations result often in very different environments the physiology and hence the steady states might differ [15]. The third requirement for high quality 13C-MFA is a very precise measurement and interpretation of metabolite labeling finally represented in the MDVs. This high MDV quality can be achieved by fulfilling the following criteria: First of all, a sufficient mass separation of 1 amu is required, to discriminate single carbon isotopes. Furthermore, a sufficient analyte separation during gas chromatography of amino acids is mandatory to prevent overlapping peaks caused by amino acids or other cellular components. Such overlapping peaks of two substances cause erroneous MDVs making those fragments unsuitable for metabolic flux analysis. An additional objective is a high measurement accuracy, characterized by the best possible agreement between determined and theoretical MDVs when unlabeled biomass is measured. Various approaches for improving the accuracy of MDVs have already been reported [23,26,70]. Worth mentioning is the study published by Antoniewicz et al., who presented an accurate method for MDV measurements of tert-butyl-dimethylsilyl (TBDMS) -derivatized amino acids with errors smaller than 0.4 mol %. In this study, based on sensitivity studies, an error < 0.5 mol % was mentioned to be suitable for flux calculation [26]. In this study a method for MDV determination of proteinogenic amino acids was developed, thoroughly investigating the effect of single parameters of the sample preparation and measurement procedure on data quality. Specifically, the influence of the measurement technique, i.e., full scan vs selected ion monitoring mode, sample size and biomass hydrolysis times was investigated. Furthermore, the reproducibility of measurements of samples manually derivatized or automatically via an autosampler was compared.

21 MASS ISOTOPOMER DISTRIBUTION MEASUREMENTS

2.1.3. Materials and Methods 2.1.3.1. Strains and Growth Conditions Escherichia coli K12 MG1655 was used in all growth experiments. The precultures as well as the main cultures were cultivated at 37 °C and 200 rpm in mineral salt medium that contained (per liter) 3.88 g K2HPO4, 2.12 g NaH2PO4 ∙ 2 H2O, 2.00 g (NH4)2SO4 and 3.6 g Glucose complemented with 0.01 g EDTA, 0.10 g MgCl2 ∙ 6 H2O, 2.0 mg ZnSO4 ∙ 7 H2O, 1.0 mg CaCl2 ∙ 2 H2O, 5.0 mg FeSO4 ∙ 7 H2O, 0.2 mg Na2MoO4 ∙ 2 H2O, 0.2 mg CuSO4 ∙ 5 H2O, 0.4 mg CoCl2 ∙ 6 H2O and 1.0 mg MnCl2 ∙ 2 H2O [71]. 2.1.3.2. Hydrolysis of Harvested Biomass Samples with defined amounts of biomass were taken from the culture and centrifuged (16.873 x g, 5 min). The supernatant was discarded and the cells were resuspended in 0.9 % (w/v) NaCl solution and centrifuged again. After discarding the supernatant, the cell pellet was resuspended in 150 μL of 6 M HCl and vortexed for 15 s. The suspension was transferred into a conical 1.5 mL glass vial. The vial was tightly closed and incubated in a heating block at 105 °C for hydrolysis of the biomass. After hydrolyzation, HCl was evaporated at 85 °C under a hood. The dried sample was stored in a closed vial at room temperature. 2.1.3.3. Analytics Optical Density Measurement

22 CHAPTER 2.1

30 m; inner diameter, 0.25 mm; film thickness, 0.25 μm). Separation of amino acids was performed at a constant helium flow rate of 1 ml/min. A sample volume of 1 μL was injected into a split/splitless injector at 270 °C while a split ratio of 1:15 was used. The temperature of the GC oven was kept constant for 1 min at 140 ºC and afterwards increased with a gradient of 10 ºC/min to 310 ºC and again kept constant for 1 min. Mass spectrometry analysis was performed on a Thermo Scientific (Waltham, MA, USA) ISQ single quadrupole mass spectrometer. The temperatures of the transfer line and the ion source were both set to 280 ºC. Ionization was performed by electron impact (EI) ionization at 70 eV. GC-MS raw data were analyzed using Xcalibur. 2.1.3.4. Data Evaluation The measured data were evaluated by Thermo Xcalibur 2.2 SP1.48 (Thermo Scientific, Waltham, MA, USA) applying the Processing Setup. Here the ICIS peak identification method was used and the parameters “Baseline window,” “Area noise factor,” and “Peak noise factor” were adjusted for every single peak. 2.1.3.5. Norm of Difference The evaluated data were normalized for sufficient comparability. This normalization was done by calculation of the norm of difference (NoD, Equation 1) where xtheor,i represents the relative abundance of mass M+i. in the theoretical MDV and xmeas,i represents the relative abundance of the same mass in the experimentally determined MDV.

௡ ଶ ൌ ඩ෍൫ݔ௧௛௘௢௥ǡ௜ െݔ௠௘௔௦ǡ௜൯ Equation 1 ܦ݋ܰ ௜ୀଵ

2.1.4. Results and Discussion For the GC-MS analysis of proteinogenic amino acids, biomass was hydrolyzed and the released amino acids derivatized with MBDSTFA. 15 out of the 20 proteinogenic amino acids were detected; glutamine and asparagine were deaminated to glutamate and aspartate, while cysteine and tryptophan were destroyed during hydrolysis and arginine was below detection limit. The accuracy of the mass isotopomer measurements was assessed by calculating the mass distribution vector, i.e., the relative abundance of mass isotopomers and comparing the data to theoretical mass distributions using the NoD (Equation 1). Since the NoD is the square root of the summed square deviations, a small NoD represents a good congruency while high NoD represents a bad congruency of theoretical and measured MDVs. Measurement quality can also be assessed by comparison of mass isotopomers with identical carbon backbone composition, which should have equal MDVs, and by inspection of absolute intensities of the single masses. The latter are used to check if the measurements are within the linear range of the detector and fragments with mass abundances out of this range, are to be rejected from further analyses.

23 MASS ISOTOPOMER DISTRIBUTION MEASUREMENTS

In the following sections, the NoD is applied for investigation and optimization of various steps concerning the workflow of 13C-MFA (Figure 10). Starting with the GC-MS method optimization, the measuring method as well as the extraction of the mass spectral data are investigated and their influence on MDV quality is tested. With this optimized measurement method, the MDV qualities are then investigated as a function of the biomass concentration, hydrolysis time of the biomass and the derivatization technique.

Figure 10: Workflow of 13C-MFA. Shown are the individual steps from 13C-labeling experiment to the final calculation of flux distributions. Steps marked in red are investigated in the following chapters.

2.1.4.1. Impact of MS data extraction method The GC-MS measurements result in a chromatogram in which the intensity of each measured mass is plotted against time. For calculating the MDVs which are necessary for 13C-analysis, the mass peaks represented in the chromatograms have to be integrated as precise as possible. This peak integration can either be performed by evaluation of peak area or peak height. Mass isotopomer abundances can be determined either from the peak height or the peak area, which is gained by integrating the peak height over the full peak width. In general, the peak area is less susceptible to noise while peak height should be used for overlapping peaks. Full scan measurements of samples with different biomass concentrations were performed and MDVs were determined using either peak area or height. Comparison of NoDs determined from biological quadruplicates of corresponding MDVs (Figure 11) revealed a slight shift towards higher NoD values for MDVs calculated via peak height. This indicates more robust data quality using peak area for calculation of MDVs. This can be explained by the discrimination of mass isotopomers over the entire peak width [23] or variations in peak width, which are not taken into account when peak height is used. To obtain accurate mass isotopomer ratios, MDVs presented in the following sections were determined by integrating the signals over the whole analyte peak width.

24 CHAPTER 2.1

1.5

1.0 NoD [ Height ]

0.5

0.0 0.0 0.5 1.0 1.5 NoD Area [ ]

Figure 11: Comparison of NoDs of MDVs calculated from peak area vs. height.

2.1.4.2. MDV determination from selected ion monitoring measurements In general, two detection modes are possible for MS measurements. In full scan measurements, all masses in a defined scan range are analyzed, while in the selected ion monitoring (SIM) mode the quadrupole focusses only on selected ions with defined m/z ratios. These are channeled to the detector while all other ions are excluded from the analysis. The full scan mode is required for untargeted metabolomics aiming for the detection of all analytes in a sample. For targeted metabolite analysis, the SIM mode is the favored analysis method as it increases the sensitivity (higher signal to noise ratio) and mass resolution. Mass isotopomer measurements in SIM mode can be optimized by adjusting the scan time and the mass width. GC-MS based analysis of MBDSTFA derivatized proteinogenic amino acids results in 64 fragments that can potentially be used for 13C-MFA. The scan time represents the time a single mass is measured. Exemplarily, the NoDs of 5 serine fragments

25 MASS ISOTOPOMER DISTRIBUTION MEASUREMENTS obtained from measurements with varying scan time is plotted in Figure 12. No clear correlation of the scan time on the NoD was observed. The scan time for all further measurements was set to 0.002 s.

0.012 f302 -15 -57 -85 0.010 -159

0.008

0.006

Norm of0.004 Difference [ ]

0.002

0.000 0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.010 0.011 Scan time [s]

Figure 12: Impact of scan time on MDV quality. Shown are the NoDs for the serine fragments f302 (black), ser m-15 (red), ser m-57 (green), ser m-85 (blue) and ser m-159 (purple) computed from measurements with scan times between 0.001 and 0.01 s.

In addition to the scan time, the scan width also plays an important role when GC-MS measurements are performed. Since every fragment consists of several mass isotopomers with mass differences of ca. 1 amu, the detected masses must be assigned correctly so that correlation of the measured abundance to the corresponding mass isotopomer is precise. This is achieved by careful adjustment of the scan width. In order to analyze the influence of the scan width on the MDV quality, different scan widths between 0.1 and 1 amu were tested. As illustrated in Figure 13, the dependence of scan width on the NoD shows a clear trend. A high scan width of 1 amu resulted in large NoD values. This can be ascribed to the incorrect assignment of measured masses to mass isotopomers. Reduction of the scan window to 0.75 amu decreased the NoD, but further reduction did not lead to better improvement of MDV quality. Thus, a scan width of 0.75 amu or smaller is suitable for accurate measurements of MDVs. All further measurements were done with a scan width of 0.5 amu.

26 CHAPTER 2.1

0.05 f302 -15 -57 -85 -159 0.04

0.03

0.02 Norm [ of ] Difference

0.01

0.00 0.0 0.1 0 .2 0 .3 0 .4 0 .5 0 .6 0 .7 0 .8 0 .9 1 .0 1 .1 Scan width [amu]

Figure 13: Impact of scan width on MDV quality. Shown are the NoDs for the serine fragments f302 (black), ser m-15 (red), ser m-57 (green), ser m-85 (blue) and ser m-159 (purple) computed from measurements with scan width between 0. 1 and 1 amu.

A comparison of samples measured with FullScan and SIM mode is shown in Figure 14. The logarithmic presentation of this comparison shows a predominantly higher NoD for measurements carried out using the FullScan mode with an average of approx. 3.5 folds higher NoDs than NoDs resulting from SIM measurements. In contrast, some fragments showed significantly higher NoDs when measured in SIM mode. This can be explained by the higher sensitivity of the SIM mode that might lead to detector saturation for the most abundant masses. Accordingly, the abundance of these masses is underestimated leading to a skewed mass distribution vector, and thus an increased NoD for these fragments.

27 MASS ISOTOPOMER DISTRIBUTION MEASUREMENTS

1.0u10 1

1.0u10 0

1.0u10 -1 NoD [ FullScan ]

1.0u10 -2

1.0u10 -3 1.0u10 -3 1.0u10 -2 1.0u10 -1 1.0u10 0 1.0u10 1 NoD SIM [ ]

Figure 14: Comparison of NoDs computed from MDVs measured with FullScan (180-550 m/z; 0.25 s) and SIM (0.5 amu; 0.002 s) mode.

By using the SIM mode, a better MDV quality was achieved for 54 out of 64 measurable fragments when a biomass amount of 0.3 mgCDW was measured. This is exemplarily shown for serine m-159 in Figure 15, for which the NoD is 9.5 fold higher when measured in FullScan mode as compared to SIM mode. For the remaining ten fragments, no differences between the two detection modes became apparent.

28 CHAPTER 2.1

0.125 FullScan SIM

0.100

0.075

0.050

Norm0.025 of Difference [ ]

0.000 0.10.20.30.40.50.60.70.80.91.01.11.21.31.41.5 CDW [mg]

-0.025

Figure 15: Comparison of SIM and FullScan measurements. Shown is the dependence of NoD on the amount of CDW used for the analysis. MDVs were determined from measurement in SIM (green) and FullScan (red) mode, respectively. NoDs of serine m-159 from biomass were determined from quadruplicate measurements.

2.1.4.3. Influence of Biomass Amount on MDV Quality The mass distribution vector is calculated from the abundance of the single mass isotopomers. This requires that the correlation of signal and concentration of these masses is identical and that all mass intensities are within the linear range. The first requirement could not be proven as no standards of the isotopomers are available. Measurements within the linear range can be guaranteed by using an appropriate biomass sample size from which appropriate amounts of the amino acids are released during acidic hydrolysis. First, amino acid standards ranging from 0.001 to 10 mM were measured and peak areas of mass fragments were determined. These measurements revealed a common linear range from 0.025 to 2.5 mM for all 15 measurable amino acids.

Assuming an average amino acid concentration in E. coli biomass of 279 μmol/gCDW [72], amounts of hydrolyzed biomass between 0.01 and 1.5 mgCDW should yield amino acid concentrations in the linear detection range and therefore should be suitable for 13C-MFA. To investigate the influence of hydrolyzed biomass amount on MDV accuracy, measurements of ten different biomass amounts from 0.01 to 1.5 mgCDW were conducted and the NoD was determined (Figure 16). In these more complex biological samples, biomass amounts of less than 0.05 mgCDW resulted in high NoD values presumably due to an insufficient signal to noise ratio for higher mass isotopomers. These high mass isotopomers have a low abundance especially in the naturally labeled biomass used here. Measurements of biomass samples

29 MASS ISOTOPOMER DISTRIBUTION MEASUREMENTS

between 0.05 mgCDW to 1.5 mgCDW resulted in relatively invariant NoDs for most amino acids as exemplarily shown in Figure 16 for serine.

0.3 f302 -15 -57 -85 -159

0.2 Norm [ of ] Difference

0.1

0.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5

CDW [mg]

Figure 16: Impact of cell dry weight used for the analysis on MDV quality. Shown are the NoDs of serine fragments f302 (black), ser m-15 (red), ser m-57 (green), ser m-85 (blue) and ser m-159 (purple) measured in quadruplicates.

The relationship between CDW and NoD for leucine (Figure 17) shows an increase in NoD above 0.3 mg CDW. The NoD increase of fragment leu m-159 is explained by a high abundance of the mass 200 m/z. The intensity of this mass exceeded the upper detector limit above a biomass sample of 0.3 mg, leading to underestimation of its abundance and erroneous evaluation of the mass distribution vector and thus an increase in NoD.

30 CHAPTER 2.1

0.35 -15 -85 -159

0.30

0.25

0.20

0.15 Norm [ of ] Difference

0.10

0.05

0.00 0 .0 0 .1 0 .2 0 .3 0 .4 0 .5 0 .6 0 .7 0 .8 0 .9 1 .0 1 .1 1 .2 1 .3 1 .4 1 .5 CDW [mg]

Figure 17: Impact of cell dry weight used for the analysis on MDV quality. Shown are the NoDs of leucine fragments leu m-15 (red), leu m-85 (blue) and leu m-158 (purple) measured in quadruplicates.

The representation of the CDW / NoD correlation exemplarily shown for serine and leucine indicated that the various amino acid fragments differ in their quality in dependence of the amount of biomass used. With respect to the 64 measured fragments, 55 fragments were below limit of detection for biomass samples below 0.02 mg. Nevertheless, all fragments are usable at a CDW amount of 0.05 mg, except of all histidine fragments, isoleucine and leucine m-15, lysine m-85, methionine m-159 and threonine m-15. These 10 fragments showed an increase of NoD below 0.1 mgCDW. 12 fragments out of 64 showed a too high, i.e. exceeding the detector limit, intensity for samples above 0.4 mgCDW. For mass distribution measurements in E. coli K12 MG1655 sample sizes between 0.1 mg CDW and 0.4 mg CDW resulted in MDVs with lowest NoDs. As the amount of amino acid incorporated into the protein can be quite variable, it is necessary to adapt these investigations to the respective experiment. For example, different substrates, substrate concentrations, temperature as well as oxygen concentration in the medium can cause differences in the metabolism and result in distinct amino acid ratios of the biomass. In addition, it is useful to carry out these studies for the specific organisms used for 13C-MFA, since these can also vary in the proteinogenic amino acid composition, so that the MDV quality of individual fragments can be affected adversely.

31 MASS ISOTOPOMER DISTRIBUTION MEASUREMENTS

2.1.4.4. Effects of Different Hydrolysis Times on MDV Quality The proteinogenic amino acids are released from the biomass protein by hydrolysis with hydrochloric acid at elevated temperatures. Incubation times between 12 and 24 h have been reported [15,26,73]. This step can be regarded as crucial as it is assumed that the protein is not hydrolyzed uniformly and thus the amino acids are released at different times of hydrolysis. To investigate the efficiency and the effect of the hydrolysis time on the MDVs, 0.3 mg of CDW were resuspended in 150 μL 6 M HCl and incubated at 105 °C. Twelve hydrolysis times between 1 and 24 were tested. The amount of released amino acids increased with increasing hydrolysis time, e.g., for leucine up to 10 h (Figure 19). A reduction of peak area after a hydrolysis time of 10 h could be observed for various amino acids, e.g., histidine, indicating thermal degradation of the analytes. In contrast to influences on peak areas, for most amino acids the hydrolysis time did not significantly impact the MDV quality as exemplarily shown for serine (Figure 18).

0.015 f302 -15 -57 -85 -159

0.010

Norm [ of ] Difference 0.005

0.000 0 3 6 9 12 15 18 21 24 Time [h]

Figure 18: Influence of hydrolysis time on MDV quality of measured amino acid fragments. Shown is the NoD of the serine fragments f302 (black), ser m-15 (red), ser m-57 (green), ser m-85 (blue) and ser m-159 (purple) measured in quadruplicates.

32 CHAPTER 2.1

2.0u10 9 1.0u108

histidine leucine

8.0u107

1.5u10 9 ekra[-] Peakarea Histidine

6.0u107 Leucine Peakarea [-]

1.0u10 9

4.0u107

5.0u10 8 2.0u107 0 3 6 9 12 15 18 21 24 Time [h] Figure 19: Peak area of selected amino acids over hydrolysis time. Shown are average peak areas of histidine (red) and leucine (green) measured with SIM mode.

An exception is fragment m-85 of histidine, for which a time dependent change of NoD was observed (Figure 20). The MDV quality of this fragment was substantially improved when the hydrolysis time was prolonged. The higher NoDs at short hydrolysis times might be caused by co-elution of cellular compounds overlapping with histidine masses of 415 and 416 m/z. These compounds most likely are decomposed during extended exposure to HCl. Therefore, it can again be shown, that various cellular components can have negative influences on mass isotopomer measurements. It is of great importance to detect and prevent these effects prior to routine analysis to facilitate a high quality 13C-MFA. During this work samples of hydrolyzed E. coli K12 MG1655 cells grown in minimal media were used. Potentially, the observed phenomenon of histidine is specific for hydrolyzation of this strain and growth conditions. Other microorganisms or media might cause different analytical background, which in turn might lead to other interferences. Hence, the optimal hydrolysis time needs to be examined for the given organism and adapted for sufficient quality of MDVs.

33 MASS ISOTOPOMER DISTRIBUTION MEASUREMENTS

0.5 f302 -15 -57 -85 -159

0.4

0.3

0.2 Norm [ of ] Difference

0.1

0.0 0 3 6 9 12 15 18 21 24 Time [h]

Figure 20: Influence of hydrolysis time on MDV quality of measured amino acid fragments. Shown is the NoD of the histidine fragments f302 (black), his m-15 (red), his m-57 (green), his m-85 (blue) and his m-159 (purple) measured in quadruplicates.

2.1.4.5. MDV Quality of Automatized and Manually Derivatized Samples For measurement of amino acids by GC-MS, dried cell hydrolysates were dissolved in 30 μL acetonitrile and derivatized using 30 μL MBDSTFA and incubated for one hour at 85 °C. Various steps such as liquid transfer, incubation time and temperature can have an influence on the efficiency of this derivatization procedure. Here, the impact of manually and automized derivatization on the quality and reproducibility of the measured data was analyzed. The comparison of the two methods was performed by measurement of 10 manually and 10 automatically derivatized biological replicates. Evaluation of the peak areas of individual mass isotopomers revealed a 4 % smaller standard deviation in case of automated derivatization, suggesting a better reproducibility of the measurements. Despite the better reproducibility of peak areas, no significant difference of MDVs was observed. 2.1.5. Conclusion Accurate determination of labeling data is an essential step for high quality of 13C-MFA. In this study, MDVs of the 64 amino acid fragments usable for 13C-MFA were investigated and the influence of detection mode, biomass amount, hydrolysis time and derivatization procedure on the quality of these MDVs was determined. The MS detection mode had a high impact on MDV accuracy, with selected ion monitoring resulting in highly superior data with an average 3.5 fold better quality of MDVs in comparison to FullScan measurements. In this

34 CHAPTER 2.1

SIM mode, factors such as scan width and scan time play an essential role and had therefore to be optimized. For the given combination of measurement system and analytes these factors were found to be optimal at a scan time of 0.002 s/mass and a scan width of 0.5 amu. Due to sensitivity enhancement achieved with the SIM mode, individual mass fragments forfeited their good quality because of intensities exceeding the detector limit at standard biomass amount per sample. Hence, the optimal amount of biomass hydrolyzed for analysis was investigated as well. Biomass amounts between 0.05 and 0.3 mgCDW were found to be sufficient for MDV measurements in E. coli at a hydrolysis time of 6 h. Even though the various hydrolysis times did not show an influence on the MDV quality of the E. coli samples analyzed in this work, these investigations have to be repeated if other strains are used as amino acid abundance might be strain specific. Therefore, a highly concentrated amino acid in an organism might exceed the detector limit when the hydrolysis time is varied. Concerning the derivatization method, no significant changes in MDV quality could be found in case of manual and automated derivatization. Nevertheless, a reduction of standard deviations by 4 % for peak areas of individual mass isotopomers indicated a better reproducibility of these peak areas for automated compared to manual derivatization.

Chapter 2.2

GC-MS based determination of mass isotopomer distributions for 13C-based metabolic flux analysis

Published as: Schmitz A., Ebert B.E., Blank L.M. (2015), GC-MS-Based determination of mass isotopomer distributions for 13C-based metabolic flux analysis. In: McGenity T., Timmis K., Nogales B. (eds) Hydrocarbon and Lipid Microbiology Protocols. Springer Protocols Handbooks. Springer, Berlin, Heidelberg

37 GC-MS BASED DETERMINATION OF MASS ISOTOPOMER DISTRIBUTIONS

2.2. GC-MS based determination of mass isotopomer distributions for 13C-based metabolic flux analysis 2.2.1. Summary 13C-metabolic flux analysis (13C-MFA) is currently the most powerful tool to determine intracellular reaction rates in biological systems, valuable, e.g., for the identification of metabolic engineering targets or the elucidation of metabolic pathway activity and regulation. The method exploits that the carbon backbone of metabolites is often manipulated differently by alternative pathways. If the cells are fed with a specifically 13C-labeled carbon source, the activity of alternative pathways determine the incorporation of the stable isotopes into metabolites and biomass constituents resulting in pathway-specific labeling patterns. In conventional 13C-MFA, cells are fed with a 13C-labeled carbon source and harvested when the metabolic intermediates or biomass constituents have reached an isotopic labeling steady state. The method is today applied on all types of cells, from microbial to plant to mammalian cells or whole organs, and on diverse carbon sources. State-of-the-art 13C-MFA most often applies mass spectrometry (MS) for the determination of 13C-labeling patterns in intracellular metabolites. Integration of the 13C-enrichment data with biochemical reaction networks and metabolic modelling then allows the calculation of intracellular fluxes. In this chapter, we give a step-by-step protocol for the set-up and validation of gas chromatograph-mass spectrometry based analysis of 13C-mass isotopomer distributions of proteinogenic amino acids. In this protocol we demonstrate the general strategy for development of new 13C-MFA methods and point out possible pitfalls of this method. 2.2.2. Introduction Knowledge of intracellular reaction rates gives an in-depth insight into cellular metabolism important in research fields such as systems biology, analysis of human diseases, metabolic engineering, or bioremediation [8]. Intracellular reaction rates are most comprehensively determined by 13C- metabolic flux analysis (13C-MFA), which exploits the fact that different pathways producing the same final product often introduce asymmetries in the distribution of carbon atoms. If cells are fed with a carbon source labeled with stable 13C-isotopes at specific positions, the 13C-isotope abundance and position of the isotope in metabolic products depends on the activity of the metabolic pathways. Measurement of the abundance of isotopomers (contraction of isotope isomers, isomers differing only in the amount of incorporated 13C-isotopes and their position) allows the inference of pathway activities by fitting the data to a mathematical model relating the isotopomer distribution of intracellular metabolites with metabolic fluxes. First published 13C- MFA studies were conducted with 13C-labeled glucose, but in the in the last decades – with the maturation of 13C-MFA tools and its broader applications – an increasing number of studies using alternative sources, including acetate, glutamate, lactate or mixtures of glucose and methanol have been reported [7,9-11,74]. In theory, fluxes during growth on complex mixtures of carbon sources, e.g., biodegradation of oil, can be determined, provided suitable stable isotopic tracers are available Commercially available are, for example, 13C-isotopomers of linear and branched alkanes (e.g., hexadecane, dodecane, decane, isobutene, butane, propane, methane) aromatic hydrocarbons (e.g., naphthalene, fluorine) or fatty acids and

38 CHAPTER 2.2 alcohols (e.g., stearic acid, 1-dodecanol). However, such analyses require parallel experiments, which are run under identical conditions but in which single substrates are specifically labeled with 13C-isotopes, making these studies experimentally demanding. Recently, methods applicable for metabolic flux analysis in defined consortia have been reported [75-77] and advances in the sensitivity and accuracy of measurements of dilute intracellular metabolites and the development of new algorithm for the evaluation of non- stationary isotopic labeling data allow now the determination of fluxes in non-growing cells or (photo)-autotrophic organisms [78-80]. For instance, 13C-based MFA of oleaginous algae and oilseed rape embryos during heterotrophic and phototrophic growth, combined with metabolite profiling or stoichiometric metabolic modeling, gave insights into the global metabolism, laying the foundation for systematic metabolic engineering [79,81]. Also, 13C- MFA has proven valuable in identifying metabolic bottlenecks in the synthesis of fatty acids in E. coli [82]. 13C-labeling states can be determined by nuclear magnetic resonance (NMR) spectrometry [65] or mass spectrometry (MS) based methods [23]. While NMR distinguishes positional isotopomers that is isotopomers with the same number 13C-isotopes but incorporated at different positions in the carbon backbone of the analyte, mass spectrometry filters analytes by their mass to charge (m/z) ratio. Consequently, MS cannot distinguish positional isotopomers, but only determines the distribution of mass isotopomers. For small analytes, positional isotopomers can be determined if suitable fragments of the analyte are formed. Anyhow, the information content of mass isotopomer distributions is most often sufficient to determine all fluxes of the central carbon metabolism (CCM) of many organisms. High sensitivity and accuracy, the rather low cost for equipment and maintenance as well its robustness, have favored the use of MS-based methods for 13C-MFA. In classical 13C-MFA, the labeling pattern of abundant proteinogenic amino acids is measured as proxies for the label enrichment of their precursor metabolites of the CCM. Prior to the MS analysis, the sample analytes are separated by chromatographic methods. Both liquid- (LC) and gas chromatographic (GC) methods are applicable while the latter is currently the more frequently used separation method for the analysis of proteinogenic amino acids. Also, direct infusion of complex samples into mass spectrometers is possible with Fourier transform ion cyclotron resonance-mass spectrometry (FTICR-MS) having an extraordinary mass resolution power and the ability to discriminate analytes purely on mass [83]. However, such devices are only found in specialized analytical laboratories. 13C-MFA requires information about the labeling state of metabolites, which is given in form of a mass distribution vector (MDV) indicating the relative abundance of mass isotopomers within the particular ion cluster (Figure 21). While mass accuracy, the difference between measured and real mass is important for unambiguous identification of unknown compounds, for the determination of mass isotopomers differing only in the total amount of incorporated isotopes, here 13C, mass resolution power, the ability to separate two adjacent masses, is crucial. With a mass difference of 1 amu and a maximal mass of relevant silylated amino acid fragments of 508 amu (monoisotopic mass), MS instruments with a mass resolution power of 508 (m/∆m = 508 amu/1 amu) are sufficient for MDV determination. Hence, analytical requirements for MDV measurements for 13C-based MFA are not too demanding and can be achieved by regular MS instruments.

39 GC-MS BASED DETERMINATION OF MASS ISOTOPOMER DISTRIBUTIONS

Figure 21: Mass spectrum of the ion cluster of a three-carbon atom molecule, with ions M0 to M3 each representing a cluster of isotopomers (a). Single MS measurements do not allow differentiation of positional isotopomers but are limited to measurement of mass isotopomers (b). Normalizing the mass intensities to the sum of all masses yields the mass distribution vector (c).

In the following we describe a step-by-step protocol for the development and validation of a GC-MS method for the determination of the 13C-enrichment of proteinogenic amino acids emphasizing parameters important for high-quality data, which are briefly described in the following. x In contrast to metabolome profiling, not the complete total ion chromatogram, but only single metabolite fragments are of interest. This allows the set-up of Selected Ion Monitoring (SIM) methods, which excel full scan modes in signal-to-noise ratio and increased accuracy of low abundant masses. x The most abundant and less abundant masses of a molecule might differ by several orders of magnitude. For accurate determination of all masses, the MS instrument has to have a high sensitivity and a wide linear range allowing accurate measurements of all signals. Masses above or below the linear range will be over- or underestimated and will screw the complete MDV. The linear range of the detector can be determined by pure standard analyte measurements. x In real samples, the concentration of single amino acids will not only depend on the amount of biomass used and the amino acid composition of the strain but also on the hydrolysis time. Verifying the optimal amount of biomass and hydrolysis time requires combinatorial testing of these two factors. x Peak integration can highly impact the quality of the MDV [26]. Currently no automatic peak picking and integration methods exist that give optimal results. Hence, processing of mass spectral data requires special attention.

40 CHAPTER 2.2

2.2.3. Materials 2.2.3.1. Harvesting and hydrolysis of cells 1. Centrifuge suited for Eppendorf tubes, e.g., Eppendorf Centrifuge 5418R equipped with standard rotor FA-45-18-11 2. HCl: 6 M aqueous solution 3. Incubator at 105 °C e.g., Thermo Scientific™ Reacti-Therm III heating module (Part Number: TS-18824) 4. Conical glass vials, e.g., Phenomenex Verex Vial (Part Number: AR0-3940-13)

2.2.3.2. Derivatization 1. Acetonitrile: HPLC-MS grade 2. N-methyl-N-tert-butyldimethylsilyl-trifluoroacetamide (MBDSTFA) (see Note 1), e.g., CS Chromatographie Service, Langerwehe, Germany

2.2.3.3. GC-MS analysis 1. Quadrupole GC-MS system with electron impact (EI) ionization 2. Fused silica column with a 1,4-bis(dimethylsiloxy)phenylene dimethyl polysiloxane or similar phase, e.g., Thermo Scientific™ TraceGOLD TG-5SilMS (length: 30m; inner diameter: 0.25 mm; film thickness: 0.25 μm) or Restek Rxi-5Sil MS (length: 15m; inner diameter: 0.25 mm; film thickness: 0.25 μm) (see Note 2) 3. Software for data acquisition, e.g., Thermo Scientific™ Xcalibur Software 4. Split/splitless liner, e.g., Thermo Scientific™ liner (Part Number: 092141) 5. Helium 5.0 (purity ≥ 99.999 Vol.-%) 6. 1.5 mL conical autosampler vials 7. Acetonitrile: LC-MS grade 8. Isopropanol: LC-MS grade 9. Hexane: analytical grade

2.2.3.4. Measurement of amino acid standards Amino acid salts of high purity (> 98%): glycine, L-alanine, L-aspartic acid, L-glutamic acid monosodium salt, L-histidine, L-isoleucine, L-leucine, L-lysine monohydrochloride, L- methionine, L-phenylalanine, L-proline, L-serine, L-threonine, L-tyrosine, L-valine (e.g., Sigma Aldrich; Carl Roth GmbH & Co. KG) Mass spectral library for peak identification in GC-MS, e.g., NIST Mass Spectral Library 2.2.3.5. Data extraction and correction 1. Software for automated peak analysis and integration, e.g., Thermo Xcalibur Quan Browser in Thermo Scientific™ Xcalibur 2. Personal computer with the open source software iMS2Flux [84], available at http://ims2flux.sourceforge.net/.

41 GC-MS BASED DETERMINATION OF MASS ISOTOPOMER DISTRIBUTIONS

2.2.4. Methods In this chapter, we describe the development of GC-MS based analytical methods for the determination of the 13C-enrichment of proteinogenic amino acids for 13C-based MFA. The chapter does not cover all other steps of 13C-MFA that is cultivation of the microorganisms in 13C-tracer medium and computational assessment of intracellular reaction rates. For a detailed description of these steps the reader is referred to, e.g., [15,85,86]. 2.2.4.1. Biomass hydrolysis While 13C-MFA requires biomass samples taken during metabolic and isotopic (pseudo) steady state conditions, here, for the development of the analytical method, samples can be taken in any phase and from cultures without 13C-enriched substrates (see 2.2.4.7). To make the protein-bound amino acids accessible, the biomass polymers are first cleaved into the respective monomers by acidic hydrolysis at elevated temperatures as described below (see Note 3). 1. Harvest the cells by centrifugation at 16.873 x g for 5 min. 2. Discard the supernatant and wash cells with 0.9 % (w/v) NaCl. 3. Centrifuge the cell suspension at 16.873 x g for 5 min. 4. Discard supernatant. 5. Resuspend the cells in 150 μL 6 M HCl and transfer the suspension into a glass vial (see Note 4). 6. Incubate the closed vial for 6 h 105 °C for hydrolysis of biomass constituents. 7. Remove HCl using a heating block at 85 °C under a hood (see Note 5). 8. Close the vial when sample is completely dried and store at room temperature.

2.2.4.2. Derivatization For gas chromatographical separation of amino acids, the analytes have to be derivatized. The following steps are advised. 1. Resuspend the dried cell hydrolysate in 30 μL acetonitrile and 30 μL MBDSTFA (see Note 6-9). 2. Tightly cap the vial and vortex the sample for 15 s. 3. Incubate the closed vial at 85 °C for 1 h using a heating block. 4. Allow the samples to cool down to room temperature and analyze samples immediately (see Note 10).

2.2.4.3. Set–up of GC-MS analysis method Gas chromatographic separation of tert-butyl-dimethylsilyl (TBDMS) -derivatized amino acids is possible with fused silica capillary columns. We achieved good separation of 15 proteinogenic amino acids with columns of 15 m or 30 m length. 1. Set the injection parameters, basic GC settings and GC temperature profile to the values listed in Table 2.

42 CHAPTER 2.2

Table 2: Recommended GC settings

Injection parameters Three washing cycles with acetonitrile prior to injection One washing step with sample Seven plunger strokes for bubble elimination Injection volume: 1 μL Pre-injection dwell time: 3 sec Post-injection dwell time: 3 sec Five washing cycles with each acetonitrile and isopropanol GC settings Constant gas flow 1 mL/min Injection mode Split Split flow 15 mL/min (see Note 11) Split ratio 1/15 Injector temperature 270 °C Vacuum compensation On Oven program column length 15 m 30 m inner diameter 250 μm 250 μm film thickness 250 μm 250 μm initial temperature 140 °C held for 1 min 140 °C held for 1 min number of ramps 1 1 rate 10 °C/min 15 °C/min final temperature 310 °C held for 1 min 310 °C held for 3 min

Figure 22: Total ion current chromatogram of unlabeled biomass of a full scan measurement (a) and selected ion monitoring chromatogram of unlabeled biomass (b)

2. After chromatographic separation, analytes are ionized and fragmented in the MS and filtered by their mass-to-charge (m/z) ratio (Figure 22). The abundance of single masses can be measured in full scan mode (see Note 12) or in SIM mode (see Note 13). Set the MS settings for analysis in full scan and SIM mode as given in Table 3. 3. Run measurements in full scan mode for analyte identification (section 2.2.4.7) and to check for sample impurities and potential signal interference. We recommend running this analysis prior to the SIM mode for analysis of new or re-engineered

43 GC-MS BASED DETERMINATION OF MASS ISOTOPOMER DISTRIBUTIONS

strains and for any new analysis device (see Note 14). For accurate determination of 13C-mass isotopomers run analyses in SIM mode (see Note 15).

Table 3: Recommended MS settings full scan mode SIM mode MS transfer line temperature 280 °C 280 °C Ion source temperature 280 °C 280 °C Ionization mode EI, 70 eV EI, 70 eV Start mass 180 amu - End mass 550 amu - Scan time 0.25 sec - Dwell time per mass - 0.002 sec Scan width - 0.5 amu (see Note 16)

4. To cover all mass isotopomers of the ion cluster, masses with a m/z ratio of M0 (the lightest possible mass) to MnumC+1, with numC being the number of carbon atoms in the amino acid carbon backbone, have to be measured within the SIM mode. See Table 4 for a list of all relevant ions.

Table 4: m/z ratios analyzed in SIM mode Amino acid m/z ratio Alanine 232, 233, 234, 235, 260, 261, 262, 263, 264

Glycine 218, 219, 220, 246, 247, 248, 249, 288, 289, 290, 291

Valine 186, 187, 188, 189, 190, 191, 260, 261, 262, 263, 264, 265, 288, 289, 290, 291, 292, 293, 294, 302, 303, 304, 305, 330, 331, 332, 333, 334, 335, 336

Leucine 200, 201, 202, 203, 204, 205, 206, 274, 275, 276, 277, 278, 279, 280, 344, 345, 346, 347, 348, 349, 350, 351

Isoleucine 200, 201, 202, 203, 204, 205, 206, 274, 275, 276, 277, 278, 279, 280, 344, 345, 346, 347, 348, 349, 350, 351

Proline 184, 185, 186, 187, 188, 189, 258, 259, 260, 261, 262, 263, 286, 287, 288, 289, 290, 291, 292, 302, 303, 304, 305, 328, 329, 330, 331, 332, 333, 334

Methionine 218, 219, 220, 221, 222, 223, 292, 293, 294, 295, 296, 297, 302, 303, 304, 305, 320, 321, 322, 323, 324, 325, 326, 362, 363, 364, 365, 366, 367, 368

Serine 288, 289, 290, 291, 302, 303, 304, 305, 362, 363, 364, 365, 390, 391, 392, 393, 394, 432, 433, 434, 435, 436

44 CHAPTER 2.2

Threonine 376, 377, 378, 379, 380, 404, 405, 406, 407, 408, 409, 446, 447, 448, 449, 450, 451

Phenylalanine 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 302, 303, 304, 305, 308, 309, 310, 311, 312, 313, 314, 315, 316, 317, 336, 337, 338, 339, 340, 341, 342, 343, 344, 345, 346, 378, 379, 380, 381, 382, 383, 384, 385, 386, 387, 388

Aspartate 302, 303, 304, 305, 316, 317, 318, 319, 320, 390, 391, 392, 393, 394, 418, 419, 420, 421, 422, 423, 460, 461, 462, 463, 464, 465

Glutamate 302, 303, 304, 305, 330, 331, 332, 333, 334, 335, 404, 405, 406, 407, 408, 409, 432, 433, 434, 435, 436, 437, 438, 474, 475, 476, 477, 478, 479, 480

Lysine 302, 303, 304, 305, 329, 330, 331, 332, 333, 334, 335, 403, 404, 405, 406, 407, 408, 409, 431, 432, 433, 434, 435, 436, 437, 438, 473, 474, 475, 476, 477, 478, 479, 480

Histidine 302, 303, 304, 305, 338, 339, 340, 341, 342, 343, 344, 412, 413, 414, 415, 416, 417, 418, 440, 441, 442, 443, 444, 445, 446, 447, 482, 483, 484, 485, 486, 487, 488, 489

Tyrosine 302, 303, 304, 305, 364, 365, 366, 367, 368, 369, 370, 371, 372, 373, 438, 439, 440, 441, 442, 443, 444, 445, 446, 447, 466, 467, 468, 469, 470, 471, 472, 473, 474, 475, 476, 508, 509, 510, 511, 512, 513, 514, 515, 516, 517, 518

2.2.4.4. Analyte Identification For analyte identification, pure standard analyte measurements of all 15 amino acids detectable in the biomass hydrolysate (see Note 17) have to be performed.

1. Solve amino acid salts in bidest water (ddH2O) to a concentration of 10 mM. 2. Dilute the 10 mM stock solution to 1 mM and pipette 60 μL of the single standard solution into a glass vial. 3. Evaporate water by placing the standards in a heating block set to 85 °C. 4. Derivatize the dried standards as described in section 2.2.4.2 and analyze the standards using the full scan mode described in section 2.2.4.3. 5. Analytes can be identified by their specific retention time (RT) in single standard measurements and by comparison with mass spectral databases such as the NIST database.

2.2.4.5. Determination of the linear range For precise mass distribution measurements, all signals have to be within the linear range of the MS detector (see Note 18).

45 GC-MS BASED DETERMINATION OF MASS ISOTOPOMER DISTRIBUTIONS

1. Prepare a dilution series with concentrations between 0.001 mM to 10 mM from the concentrated 10 mM amino acid standard solution (section 2.2.4.7). Use ddH2O for dilution. At least five standards with different concentrations should be prepared. 2. Analyze the samples using the SIM mode. 3. For every analyte plot the peak area vs. concentration. Graphical inspection of the plot reveals the linear range and lower detection limit (see Note 19). 4. When measuring real samples, exclude all fragments with masses out of the linear range from further analysis.

2.2.4.6. MS data processing While it is sufficient for analyte quantification to analyze the total ion chromatogram or the intensity of one single mass, for determination of a mass distribution vector all masses in the ion cluster have to be analyzed individually. An example of this differentiation of mass peaks can be seen in Figure 23. Mass peak integration was shown to impact the accuracy of MDVs [23,26] and hence requires special attention. We describe here exemplarily the automated data processing, i.e., peak detection and integration for the software Thermo Xcalibur. 1. Open the MS Data file (.raw) of the amino acid measurement in the Thermo Xcalibur Processing Setup. 2. Name every mass, which has to be processed, with a unique identifier (Figure 24). 3. Choose the ICIS peak identification method (see Note 20). 4. Specify the expected RT of the current ion. From the ‘Trace’ dropdown menu choose ‘mass range’ and specify the m/z ratio of the current mass.

46 CHAPTER 2.2

Figure 23: Single mass peaks of an ion cluster of alanine m-85 containing two carbon atoms of the amino acid carbon backbone. The mass peaks M0 to MnumC+1 (shown in the first four chromatograms) are required for complete mass distribution determination.

Figure 24: Peak identification settings in the Thermo Xcalibur data processing module

5. Set the peak detection parameters as shown in Figure 25. The parameters “Baseline window”, “Area noise factor” and “Peak noise factor” are peak specific ad have to be set individually. The same holds for the “Area scan window” in the advanced settings (see Note 21).

47 GC-MS BASED DETERMINATION OF MASS ISOTOPOMER DISTRIBUTIONS

6. Repeat step 1 to 5 for all relevant masses. For every fragment, the masses M0 (lightest mass within the fragment) to MnumC+1 (M0 + number of carbon -atoms of the amino acid + 1) have to be considered. For example, the alanine fragment m-85 (see Note 10), with a monoisotopic mass (M0) of 232 amu contains two carbon atoms of the amino acid carbon backbone, the masses from m/z 232 (M0) to 235 (MnumC+1) are used. 7. After all amino acid fragments are included, reprocess the measured data using the processing method. Export the processed data (peak areas and/or peak height) to Excel for further data analysis by following the instructions specific for the used MS instrument (see Note 22).

Figure 25: Peak detection settings in the Thermo Xcalibur data processing module

2.2.4.7. Method optimization The amount of the individual amino acids in the hydrolyzed biomass sample depends on the biomass used and the hydrolysis time (see Note 23). A basic requirement for high-quality MDV determination is that the signals of all relevant amino acid fragments are within the linear range of the detector determined in section 2.2.4.5. Measurements within this range still can be flawed for example due to overlapping fragments. A quantitative measure of MDV quality is the norm of difference (NoD, Equation 2) between experimental and theoretical relative abundances of the masses within the fragment. To assess the optimal combination of amount of biomass and hydrolysis time, test combinations of these two variables and evaluate the quality of the resulting MDVs by comparison with theoretical values. 1. Calculate theoretical MDVs for unlabeled and derivatized amino acid fragments considering the natural abundance of heavy isotopes. This can be done by software tools such as IsoPro3, a mass spectral isotopic distribution simulator [87].

48 CHAPTER 2.2

2. Prepare samples with different amounts of unlabeled biomass (e.g., 0.05, 0.1, 0.2, 0.3, 0.5 and 1 mg) and hydrolyze for varying durations (e.g., 1, 2, 4, 6, 8, 10 and 20 h). 3. Derivatize the samples and perform GC-MS measurements in SIM mode. 4. Extract peak areas of all fragments using the automated integration method described in 2.2.4.6. 5. Calculate the mass distribution vector for all fragments by normalizing the abundance of the masses of the respective fragment (M0 to MnumC+1) to 1. 6. For evaluation of the quality of the experimental data calculate the NoD as defined in Equation 2.

௡ ଶ ൌ ඩ෍൫ݔ௧௛௘௢௥ǡ௜ െݔ௠௘௔௦ǡ௜൯ Equation 2 ܦ݋ܰ ௜ୀଵ

xtheor,i and xmeas,i are the relative abundance of mass M+i. in the theoretically and experimentally determined MDV (Figure 26).

0.20 -57

-85

0.15

0.10 Norm [ of ] Difference

0.05

0.00 0.00.10.20.30.40.50.60.70.80.91.01.11.21.31.41.5 CDW [mg]

Figure 26: Dependence of MDV quality, expressed as the Norm of Difference (NoD), of two alanine fragments m-57 (green) and m-85 (blue) in dependence of the amount of biomass used. A fixed hydrolysis time of 6 h was used.

49 GC-MS BASED DETERMINATION OF MASS ISOTOPOMER DISTRIBUTIONS

7. For MDVs with NoDs greater than 0.01, check if peak intensities are within the linear range and if peaks are properly integrated. Combinations of biomass and hydrolysis time resulting in MDVs of poor quality are not suited for 13C-MFA. 8. If necessary, adapt the MS data processing method for the respective ion cluster. Check quality of MDVs after re-integration of masses. 9. This quality check is iteratively done until the parameters biomass concentration, hydrolysis time, detection mode, and peak integration are at their best configuration.

2.2.4.8. Data Correction The extracted mass spectrometric data represent the peak intensities of single masses of an amino acid fragment. Natural abundance of stable isotopes in the amino acid or the derivatization groups, e. g., 15N, 2H, 29Si, contribute to mass isotopomers of higher masses while unlabeled biomass introduced by inoculation causes a dilution of the 13C-content. For flux quantitation using 13C-labeling data, the MS raw data have to be corrected for these effects. We here describe the correction procedure with the software iMS2FLUX [84]. 1. Arrange peak areas of the single masses in a *.txt file in the format shown in Figure 27. The first column specifies the compound name, the second column the mass. Each column after this lists MS data from one chromatogram or sample. Replicate measurements are consecutively arranged and labeled with increasing numbers. Save as *.txt file.

Figure 27: Structure of the tab separated value (TSV) file with MS data to be processed with iMS2Flux

2. Configure data related options, data checking options, correction options and print options using the graphical user interface (GUI) of iMS2Flux (Figure 28). 3. Load the generated data file. The path to the .txt file can be typed into the field ‘MS Data File’ or selected by browsing.

50 CHAPTER 2.2

4. Choose the type of MS Data File. 5. In the field “Data Related Options” specify the derivatization method, the type of measured analytes and the number of measured masses. With the described SIM mode the masses M0 to MnumC+1 are measured. 6. In “Data Checking Options” thresholds for detector overflow and minimum detection limit can be set. The software automatically excludes data exceeded or above or below set thresholds (see Note 24). Mark “Average Carbon Labeling” to write a result file listing the average carbon labeling of the analytes. Inspection of these data is useful for further data quality assessment (see Note 25). 7. Data correction options are specified in “Perform Corrections”. Correction for proton loss/gain, natural abundance, and original biomass can be selected. For the correction of amino acid fragments, natural abundance, and original biomass are of utmost importance. For every single data set within the data file the respective amount of original biomass has to be specified in an additional *.txt file.

Figure 28: Parameter setting via the graphical user interface of iMS2Flux

Different reports of corrected data can be generated: Processed data (Figure 29), average over replicates and standard deviations over replicates can be reported. Furthermore, the corrected mass isotopomer data can directly be formatted as required for analysis in 13CFLUX [6], 13CFLUX2 [13], and OpenFLUX [12].

51 GC-MS BASED DETERMINATION OF MASS ISOTOPOMER DISTRIBUTIONS

Figure 29: Structure of the data output file with corrected mass distribution data calculated by iMS2Flux from raw MS data

2.2.5. Notes 1. Derivatization for GC analysis can be done using various reagents. Here, N-methyl- N-tert-butyldimethylsilyl-trifluoroacetamide (MBDSTFA) is used because the silylated analytes form specific fragments within the ionization chamber during MS measurement. The most abundant fragments detected for all amino acids are m-15 (- CH3), m-57 (-C4H9), m-85 (-C5H9O), m-159 (-C7H15O2Si) and a fragment of m/z of 302, in which the amino acid side chain is cleaved off. This fragmentation leads to analytes with the carbon atoms C1-Cn (-15/-57), C2-Cn (-85/-159) and C1-C2 (f302). 2. Amino acid analysis in GC-MS has been developed by applying one of the most commonly used, non-polar stationary phases in gas chromatography. Gas chromatography of amino acids using a 5% diphenyl/95% dimethyl polysiloxane phase ensures a good separation of these compounds and reproducible elution of the amino acids in the order: Ala, Gly, Val, Leu, Ile, Pro, Met, Ser, Thr, Phe, Asp, Glu, Lys, His, Tyr. 3. The optimal method depends on laboratory equipment and the particular microorganism used for metabolic flux analysis. With our equipment, samples of 0.1 to 0.3 mg biomass are required for the analysis of proteinogenic amino acids in E. coli. 4. The dry residue heavily sticks to the vessel wall and is difficult to dissolve; often complete solution is only achieved during heating in the derivatization phase. In order to prevent sample loss, perform all steps from 2.2.4.1-6 till GC-MS analysis in the same glass vial. 5. The drying process can be accelerated through application of a constant inert gas stream (e.g., nitrogen); avoid prolonged drying periods as these can complicate later resuspension.

52 CHAPTER 2.2

6. Minimize the exposure of the silane to humid air by quickly transferring the reagent to the vials as the silane is readily destroyed in the presence of water. Ideally the silane is aliquoted and only few times exposed to air. A septum and syringe can also be used to reduce air exposure. 7. Because of differences in sample volume during hydrolysis (150 μL) and the end volume (less than 60 μL) a conical glass vial should be used, otherwise the liquid level might be too low for reliable sampling by the autosampler. 8. To diminish dilution of analytes, the volumes of acetonitrile and MBDSTFA can be reduced. Care has to be taken that the final sample volume is not below the minimal level the autosampler can handle. 9. Make sure that the sample is completely dry as the silane is readily destroyed in the presence of water. If the sample has been stored, incubate again for ca. 1 h to remove any moisture. 10. Samples can be stored at room temperature for several days if they are completely dry. Any moisture in the vial will destroy the silylated compound. 11. Sample injection is possible in split or splitless mode. During splitless injection the whole vaporized sample is transferred via an inert gas flow onto the column. During split injection only a defined fraction of the sample is transferred onto the column while the rest is purged out of the system without reaching the separation column. We worked with a split ratio of 1:15, the minimum of the system used under the set conditions (gas, flow through column, column dimension etc.). If the analyte concentrations are sufficient, use a split mode. The reduced solvent load will protect the column. 12. In full scan mode, all masses between 180 and 550 amu are scanned, covering all relevant amino acid fragments except alanine m-159 (158 amu) and glycine m-159 (144 amu). We decided to exclude these two fragments as the high signal to noise ratio below 180 amu did not allow reliable measurements. Besides amino acids other analytes present in the lyzed biomass can be detected if separated with the applied gas chromatographic program. 13. In SIM mode, only selected analyte specific ions are measured. Here, for every

amino acid fragment the masses M0 until MnumC+1 are analyzed while all other masses are masked. This procedure considerably increases the signal to noise ratio and decreases the detection limit. 14. We noticed differences in signal interference with different GC-MS instruments. For example, we regularly achieve good data for proline or valine (fragment f302), which in the literature have been reported to be unusable due to co-elution with other compounds or overlapping fragments. 15. The improvement in MDV quality determined from measurements in SIM mode vs. full scan mode is exemplarily shown in Figure 30. Besides reduced noise, SIM measurements result in a better mass resolution as this mode allows scanning narrower mass windows (here M0 ± 0.25 amu), which prevents measurement of overlapping masses.

53 GC-MS BASED DETERMINATION OF MASS ISOTOPOMER DISTRIBUTIONS

0.06

FullScan

SIM

0.05

0.04

0.03

Norm [ of ] Difference 0.02

0.01

0.00 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 CDW [mg]

Figure 30: Assessment of MDV quality from full scan vs. SIM mode measurements. Shown is the Norm of difference of the serine fragment m-57 as function of the amount of biomass used calculated from data from full scan (red) and SIM mode measurements (green)

16. With our set-up, a scan width of 0.5 amu yielded the best mass resolution. 17. From the 20 proteinogenic amino acids five acids (glutamine, asparagine, tryptophan, cysteine) are deaminated or degraded during the hydrolysis process. Arginine generally has a too low signal-to-noise ratio. 18. If the intensity of a single mass is above the detector limit, its abundance is underestimated, screwing the complete mass distribution vector. The same holds for masses below the detection limit, making these measurements unsuitable for flux analysis. 19. For our set-up, we specified a linear range for amino acids between 0.025 mM and 2.5 mM. 20. The Thermo Xcalibur processing set-up provides three integration methods, Avalon, Genesis and ICIS. The Thermo manual recommends the use of Avalon for ultraviolet–visible spectroscopy (UV/Vis) data, Genesis for full scan MS data and ICIS for MS measurements using SIM mode. 21. The settings for detecting a peak have to be specified for every mass of an ion cluster. We recommend using the same settings for all masses within an ion cluster instead of integrating every mass individually as this avoids different baseline windows and error prone MDVs. As peak width increases with analyte concentration, the peak detection parameters have to be adjusted to avoid incomplete integration (Figure 31).

54 CHAPTER 2.2

Figure 31: Peak integration with an area scan window of 10 scans (top) and 40 scans (bottom)

22. In experiments with unlabeled biomass we observed that mass distribution vectors calculated using the peak area generally coincided better with theoretical values as those calculated from peak height (Figure 32).

55 GC-MS BASED DETERMINATION OF MASS ISOTOPOMER DISTRIBUTIONS

0.05

Area

Height

0.04

0.03

0.02 Norm [ of ] Difference

0.01

0.00 0.0 0.1 0.20.30.40.50.60.70.80.91.01.11.21.31.41.5 CDW [mg]

Figure 32: Norm of difference (NoD) of the MDV of the aspartate fragment m-57 calculated from either peak areas (red) or peak heights (green)

23. We observed that alanine, valine and leucine signals increase with prolonged hydrolysis time (up to 16 h, decreasing again for hydrolysis times above 24 h), while, e.g., methionine showed the highest signal for a hydrolysis time of 4 h. 24. Use the upper bound of the linear range determined in section 2.2.4.5 as detector limit threshold, and as poor peak threshold the limit of detection of the amino acid. 25. The average carbon labeling should reflect the fraction of 13C-isotopes in the carbon source if a mixture of fully labeled and naturally labeled substrate has been used for the tracer experiment. The average carbon labeling can therefore be used to test data quality.

Chapter 2.3

Investigation of glucose uptake in Pseudomonas putida

Author Contributions: Sarah Maurer performed initial experiments for 13C measurements of intracellular metabolites. Andreas Schmitz planed and designed the project, performed the experiments and analyzed the results. The chapter was written with the help of Birgitta Ebert. Birgitta Ebert and Lars M. Blank supervised and conceived the study.

57 INVESTIGATION OF GLUCOSE UPTAKE IN PSEUDOMONAS PUTIDA

2.3. Investigation of glucose uptake in Pseudomonas putida 2.3.1. Summary The bacterium Pseudomonas putida is a promising microbial host for industrial processes such as the production of fine chemicals due to its high ability to adapt to external influences. This includes the high tolerance towards solvents and the flexible utilization of substrates for which it has been studied extensively. Even though its physiological behavior and biochemical capabilities are well known, there are still some open questions regarding this organism’s metabolism, the glucose uptake being one example. Besides the direct import into the cytosol and subsequent phosphorylation to glucose-6-phosphate (g6p), an oxidative pathway exists, in which glucose is oxidized in the periplasmic space to gluconate and further to 2-ketogluconate. Both gluconate and 2-ketogluconate can be transported into the cell and catabolized to 6-phosphogluconate (6pg), the initial metabolite of the Entner-Doudoroff (ED) pathway. In this study, stable isotope tracer experiments were applied to investigate the glucose uptake fluxes of P. putida strains differing in their physiological behavior, more specifically growth rate and substrate uptake rate, and comprising gene knockouts curtailing the glucose uptake pathways. Analysis of the labeling of the ED pathway and Pentose Phosphate (PP) pathway intermediates showed a pronounced activity of phosphoglucose isomerase (Pgi) for P. putida S12 compared to the two other strains, P. putida KT2440 and P. putida KT2440 Δglk. The metabolite labeling of metabolites participating in the ED and the PP pathway might indicate substrate channeling of 6pg originating from g6p into the ED-Pathway and 6pg originating from gluconate into the PP pathway. 2.3.2. Introduction Pseudomonas putida is a promising host for biotechnological processes. Due to the tolerance towards various chemicals such as solvents and toxic compounds as well as its ability of using a broad substrate range, P. putida can adapt to many external environments [88-91]. Hence, this highly versatile organism is increasingly evaluated for production of industrial relevant chemicals. Worth mentioning are alcohols like ethanol, aromatic acids such as ortho- aminobenzoate, and rhamnolipids, which are all already produced using P. putida [92-94]. In addition to physiological analyses of Pseudomonads, the genomes of different strains have been sequenced and annotated providing extended knowledge about the relationship between the genome and their biochemical and physiological capabilities. Noteworthy is the great knowledge concerning the solvent tolerance mechanism of P. putida S12 [95,96]. Likewise, the genome basis for the degradation of aromatic substances were extensively examined for example in P. putida KT2440 [97]. Nevertheless, even in these already well-studied organisms knowledge gaps about central metabolic processes, such as the glucose uptake, exist. Various studies [90,98-101] were published showing that Pseudomonads catabolize glucose via different pathways (Figure 33). One possibility is the direct import of glucose via an ABC-transporter and the subsequent phosphorylation to glucose-6-phosphate (g6p) by means of glucokinase (Glk) activity. The main fraction of g6p is subsequently converted to 6-phosphogluconate (6pg) via glucose-6-phosphate dehydrogenase (Zwf) and 6-phosphogluconolactonase (Pgl). An alternative uptake of glucose is the periplasmic

58 CHAPTER 2.3 oxidation of glucose to gluconate via glucose dehydrogenase. Gluconate can either be transported into the cell or further oxidized to 2-ketogluconate via gluconate dehydrogenase. 2-ketogluconate as well as gluconate can be transported into the cell and be converted to 6pg [99,100,102]. Conversion of gluconate to 6pg occurs by means of gluconokinase, while 2- ketogluconate is reduced to 6pg via 2-ketogluconate reductase [103]. 6pg represents the starting point of the Entner-Doudoroff (ED) pathway, the main glucose catabolic pathway present in Pseudomonads [104]. Phosphogluconate dehydratase (Edd) and 2-keto-3-deoxy gluconate aldolase are used for glyceraldehyde-3-phosphate (g3p) and pyruvate (pyr) formation, which are further degraded in the tricarboxylic acid (TCA) cycle (Figure 33) [105].

Figure 33: Glucose uptake in P. putida. Glucose is incorporated and transformed to g6p via Glk. Alternatively, glucose is oxidized in the periplasm via glucose dehydrogenase to gluconate and further to 2-ketogluconate via gluconate dehydrogenase. Gluconate and 2- ketogluconate are transported into the cell and transformed to 6pg. g6p is oxidized to 6pg using Zwf. 6pg is the initial metabolite of the ED-Pathway resulting in g3p and pyr.

During growth on glucose, P. putida exhibits different growth phases, which differ in biomass yield and growth rates. Within the first phase, glucose is consumed and gluconate as well as 2-ketogluconate accumulate. The second phase generally starts when glucose is depleted. During this phase, gluconate is consumed while the 2-ketogluconate concentration further increases. The third phase is characterized by the utilization of 2-ketogluconate. In well resolved growth curves, a reduction of the growth rate is apparent and occurs simultaneously with the shift between the growth phases [106]. Even though these growth phases can be observed, the uptake fluxes of glucose, gluconate and 2-ketogluconate cannot be determined directly from their extracellular concentration profiles.

59 INVESTIGATION OF GLUCOSE UPTAKE IN PSEUDOMONAS PUTIDA

Here, we set out to disclose the activity of the alternative carbon uptake pathways by applying 13C-tracer experiments. For these 13C-tracer experiments, cells are fed with specifically labeled carbon sources such as 13C-glucose and labeling of metabolized substances is measured via mass spectrometry (MS) or nuclear magnetic resonance (NMR) spectroscopy. For investigation of the alternative carbon uptake pathways, the labeling of intracellular metabolites needs to be analyzed [32,54]. These metabolites have to be extracted from cells, whose metabolism needs to be stopped rapidly. Subsequently, labeling determination of these metabolites can be performed using, e.g., gas chromatography (GC)-MS. 2.3.3. Materials and Methods 2.3.3.1. Strains and Growth Conditions The three Pseudomonas strains used in this study were P. putida KT2440, P. putida KT2440 Δglk, an isogenic mutant of P. putida KT2440, in which the glucokinase gene glk is deleted, and P. putida S12. The medium used was a mineral salt medium that contained (per liter) 3.88 g K2HPO4, 2.12 g NaH2PO4 ∙ 2 H2O, 2.00 g (NH4)2SO4, 0.01 g EDTA, 0.10 g MgCl2 ∙ 6 H2O, 2.0 mg ZnSO4 ∙ 7 H2O, 1.0 mg CaCl2 ∙ 2 H2O, 5.0 mg FeSO4 ∙ 7 H2O, 0.2 mg Na2MoO4 ∙ 2 H2O, 0.2 mg CuSO4 ∙ 5 H2O, 0.4 mg CoCl2 ∙ 6 H2O and 1.0 mg MnCl2 ∙ 2 H2O [71]. The medium was complemented with 10 mmol/L U-13C-glucose and 10 mmol/L gluconate or 2-ketogluconate, respectively. Precultures were performed in 500 mL shake flasks with 50 mL medium complemented with 20 mmol/L naturally labeled glucose at 30 °C and 250 rpm. The main cultures were cultivated in New Brunswick BioFlo 115 benchtop bioreactors (Eppendorf, Germany) at 30 °C. The batch fermentations were executed using a 1.3 L reactors with a working volume of 500 mL. Air supply was realized using headspace aeration with 2 vvm of compressed air. The culture was stirred with 750 rpm at the beginning and the stirring speed was increased when the dO2 signal dropped to 35 %. 2.3.3.2. Sample Preparation

For cell quenching, cell broth containing at a minimum 10 mgCDW was directly transferred into a cold 70 % ethanol / NaCl (0.9 % (w/v)) solution prechilled to -52 °C and shaken immediately. The solution was centrifuged (4800 x g) for 7 minutes at -10 °C and the supernatant was discarded. The cell pellet was stored at -80 °C until further analysis. Extraction of quenched cells was performed with a -50 °C cold methanol:chloroform:water (2.5:1:1) solution. The cell pellets were resuspended in 2 mL of the extraction solution and vortexed. The solution was kept at -20 °C while being shaken on an overhead shaker for at least 1 h. After extraction, the solution was centrifuged for 7 minutes at -10 °C, 1.3 mL of the supernatant were transferred into an Eppendorf tube and mixed with 300 μL H2O. The solution was centrifuged for 10 minutes at 4 °C and 1.2 mL of the water phase were filtered into glass vials. The samples were dried by first evaporating the methanol fraction in a centrifugal concentrator (ScanSpeed MaxiVac Beta, Labogene, Allerød, Denmark) at 20 °C for 60 min. The concentrated samples were frozen at -80 °C and subsequently placed in a lyophilizer (ScanSpeed MaxiVac Beta, Labogene, Allerød, Denmark) at – 110 °C over night for water elimination until complete dryness of the samples.

60 CHAPTER 2.3

2.3.3.3. Analytics Optical Density Measurement

Cell concentrations in liquid cultures were determined with optical density (OD600) measurements. These were carried out with a spectral photometer (Ultrospec 10 cell density meter, Amersham Biosicences, Glattbrugg, Schweiz) at a wavelength of 600 nm. Cell dry weight (CDW) concentrations were calculated with an OD600 / CDW conversion factor of 0.40 gCDW/L. High Performance Liquid Chromatography Measurement Extracellular concentrations of substrates and metabolites were quantified with high performance liquid chromatography (HPLC) measurements. The culture was centrifuged (16.873 x g, 5 min) and the supernatant was measured using a HPLC (Beckmann System Gold, Beckmann Coulter, Krefeld, Germany), with external column oven (Jetstream 2 Plus, Knauer, Berlin, Germany) equipped with an organic acid resin column (polystyrol- divinylbenzol copolymer, PS-DVB: 300 × 8.0 mm, CS-Chromatographie, Langerwehe, -1 Germany). 5 mM H2SO4 was used as eluent at a flow of 0.8 mL h min at 75 °C. The analytes were detected by an ultraviolet light (UV) detector (UV detector LC-166, Beckmann Coulter, Krefeld, Germany) at a wavelength of 210 nm and a refractive index (RI) detector (LCD 201, Gynkotek, Munich, Germany) system. Extracellular Rates Extracellular rates such as growth rate and carbon uptake / production rates of glucose, gluconate and 2-ketogluconate were determined with Microsoft Excel 2013. The growth rate was calculated as a linear function of the logarithm of OD600 over time. Uptake and production yields were calculated as the function of analyte and biomass concentration. Substrate uptake and product formation rates were calculated by multiplication of growth rate and corresponding yield. Derivatization Dried biomass samples were derivatized with 30 μL of pyridine / methoxyamine (20 mg/ml) at 67 °C for 1 h and 30 μL N-Methyl-N-trimethylsilyl-trifluoroacetamide (MSTFA, CS - Chromatographie Service GmbH, Langerwehe, Germany) for 1 h at 67 °C. The samples were cooled down to room temperature and analyzed immediately. Gas Chromatography–Mass Spectrometry GC separation was performed on a Thermo Scientific (Thermo Scientific, Waltham, MA, USA) Trace GC Ultra equipped with a Thermo Scientific AS 3000 autosampler. The column used was a Thermo Scientific (Thermo Scientific, Waltham, MA, USA) TraceGOLD TG- 5SilMS capillary column (length, 30 m; inner diameter, 0.25 mm; film thickness, 0.25 μm). Intracellular metabolites were separated at a constant flow rate of 1 mLhelium/min. 1 μL of the sample was injected into a programmed temperature vaporization injector at constant temperature of 270 °C while splitless injection was performed. The temperature of the GC oven was kept constant for 5 min at 70 ºC and afterwards increased with a gradient of 6 ºC/min to 320 ºC and again kept constant for 10 min. Metabolites in the supernatant were separated at a constant flow rate of 1 mLhelium/min. 1 μL of the sample was injected into a split/splitless injector at constant temperature of 270 °C while splitless injection was

61 INVESTIGATION OF GLUCOSE UPTAKE IN PSEUDOMONAS PUTIDA performed. The temperature of the GC oven was kept constant for 5 min at 70 ºC and afterwards increased with a gradient of 6 ºC/min to 250 ºC. The temperature was then increased with a gradient of 15 ºC/min to 320 ºC and kept constant for 1 min. Mass spectrometry analysis of intracellular metabolites was performed on a Thermo Scientific (Thermo Scientific, Waltham, MA, USA) TSQ triple quadrupole mass spectrometer. The temperature of both the transfer line and the ion source was set to 280 ºC. Ionization was performed by electron impact (EI) ionization at 70 eV and masses from 60 to 600 were detected with a scan time of 0.25 sec/scan. Mass spectrometry analysis of supernatant metabolites was performed on a Thermo Scientific (Thermo Scientific, Waltham, MA, USA) ISQ mass spectrometer using the same settings as described above. GC-MS raw data were analyzed using Xcalibur [107]. 2.3.3.4. Sampling plan Samples for determination of the required parameters, such as growth rate, extracellular rates and labeling of intracellular metabolites were harvested at fixed time points. Starting with the OD600 immediately after inoculation, the first sample was harvested after 1.5 hours. At intervals of 30 minutes additional samples were taken to determine a time course of the respective parameters. 2.3.3.5. Data evaluation iMS2Flux was used for correction of labeling data for natural abundance of heavy isotopes and label dilution by unlabeled biomass introduced with the inoculum [84]. 2.3.4. Results and Discussion The glucose uptake of the two wildtype strains Pseudomonas putida KT2440 and Pseudomonas putida S12 as well as the glk knockout mutant P. putida KT2440 Δglk was investigated. We have chosen these three strains as P. putida KT2440 and P. putida S12 show difference in the kinetics and amount of extracellular gluconate and 2-ketogluconate accumulation while P. putida KT2440 Δglk is not capable to directly take-up glucose due to its glk deficiency but has to oxidize it to gluconate or 2-ketogluconate in the periplasm. 13C- tracer experiments were performed with three different Pseudomonas strains cultivated in batch fermentation. Since no reconfiguration of the carbon backbone of the various metabolites involved in glucose uptake take place, the labeling of these metabolites will be equivalent to the labeling of glucose fed to the cells. For differentiation of the periplasmic pathways of glucose uptake metabolism, the cells were fed with a co-feed of 13C-glucose and 12C-gluconate or 13C-glucose and 12C-2-ketogluconate in equimolar amounts. Besides the physiological monitoring, the concentration and 13C-labeling of glucose, gluconate and 2- ketogluconate in the medium was measured as well as the labeling of intracellular metabolites. Combining this information, the activity of periplasmic oxidation reactions in contrast to uptake reactions should be distinguishable. The physiological as well as the labeling data of extracellular and intracellular metabolites are shown in the following sections. 2.3.4.1. Physiology of Pseudomonads growing on a cofeed of U-13C-glucose and 12C- gluconate Using equimolar amounts of U-13C-glucose and 12C-gluconate as initial carbon sources, only small differences in growth rates were observed for the three investigated strains. Average growth rates of 0.42 ± 0.06 1/h for P. putida KT2440, 0.44 ± 0.04 1/h for P. putida S12 and

62 CHAPTER 2.3

0.48 ± 0.05 1/h P. putida KT2440 Δglk were determined. For P. putida KT2440 and P. putida S12, the growth rates were calculated for growth between 1.5 to 3.5 h whereas for P. putida KT2440 Δglk a time interval between 1.5 to 4.5 h was chosen for calculation. Even though the growth rates were similar, differences in the physiology were observed in terms of substrate conversion, accumulation of gluconate and 2-ketogluconate as well as the length of lag phases. For growth of P. putida KT2440 Δglk a prolonged lag-phase was observed, explainable by the missing Glk enzyme responsible for phosphorylation of glucose to g6p. Without this Glk enzyme, P. putida cannot use glucose directly but has to convert it into gluconate first. However, gluconate was present in the growth medium right from the start and could have been taken up directly. Both P. putida KT2440 and P. putida KT2440 Δglk showed a temporal reduction in OD600 value increase when glucose and gluconate were depleted and the accumulated 2-ketogluconate was consumed. This diauxic growth indicates that 2-ketogluconate is not consumed while glucose and gluconate are present. In case of P. putida S12 no clear halting of the OD600 values was observed after glucose and gluconate depletion.

Table 5: Growth-phase dependent conversion (-) and accumulation (+) rates of glucose, gluconate and 2-ketogluconate for P. putida KT2440 (a), P. putida S12 (b) and P. putida KT2440 Δglk (c) grown on U-13C-glucose and 12C-gluconate. Strain time [h] glucose gluconate 2-ketogluconate formation rate formation rate formation rate [mmol/g/h] [mmol/g/h] [mmol/g/h] 0 – 1.5 - 16 ± 2.3 + 2 ± 0.3 + 8 ± 1.1 a 1.5 – 2.5 -13 ± 1.9 + 9 ± 1.3 3 – 4.5 -5 ± 0.7 0 – 1.5 - 15 ± 1.4 - 8 ± 0.7 + 18 ± 1.6 b 1.5 – 2 - 2 ± 0.2 - 15 ± 1.4 + 13 ± 1.2 2 – 4 -5 ± 0.5 0 – 3 - 8 ± 0.8 - 9 ± 0.9 + 13 ± 1.4 c 3 – 5 -5 ± 0.5

The substrate consumption rate and accumulation rate of extracellular oxidation products of the investigated strains can be divided in distinct phases, which are characterized by glucose consumption, gluconate consumption and 2-ketogluconate consumption, respectively (Table 5a - c). In case of P. putida KT2440 (Figure 44a), the three phases were clearly separated, whereby gluconate and 2-ketogluconate accumulated during the glucose conversion phase. This distinct separation was not observed for P. putida S12 (Figure 35a), as the initially fed gluconate was already consumed during the glucose conversion phase. Nevertheless, 2- ketogluconate accumulated until glucose as well as gluconate were completely depleted and was taken up within the last growth phase at a rate of 5 mmol/g/h for both strains. In contrast to P. putida KT2440, the glk knockout mutant P. putida KT2440 Δglk (Figure 36a) showed only two growth phases. Glucose and gluconate were consumed simultaneously at similar rates (Table 5c). During this first growth phase, 2-ketogluconate accumulated until both glucose and gluconate were completely depleted after 3 h. In the second growth phase, 2- ketogluconate was taken up with an uptake rate of 5 mmol/g/h as determined for the other

63 INVESTIGATION OF GLUCOSE UPTAKE IN PSEUDOMONAS PUTIDA investigated strains. As the Glk enzyme encoding gene is knocked out in this strain, glucose cannot directly be taken up and converted to g6p but has to be oxidized to gluconate. This deficiency might explain the reduced degradation rate of glucose in this strain, which was half of the degradation rate determined for P. putida KT2440. 2.3.4.2. Physiology of Pseudomonads growing on a cofeed of U-13C-glucose and 12C- 2-ketogluconate Physiological investigation of P. putida KT2440 (Figure 37a), P. putida S12 (Figure 38a) and P. putida KT2440 Δglk (Figure 39a) fed with equimolar amounts of U-13C-glucose and 12C-2- ketogluconate revealed equal growth rates for P. putida KT2440 and P. putida KT2440 Δglk, determined to be 0.69 ± 0.031/h and 0.67 ± 0.04 1/h respectively. The strain P. putida S12 showed a lower growth rate of 0.58 ± 0.05 1/h. For P. putida KT2440 the growth rate was calculated for growth between 1.0 to 2.5 h whereas for P. putida S12 and P. putida KT2440 Δglk a time interval between 1.0 to 3.0 h was chosen for calculation. Despite the lower growth rate, P. putida S12 reached its maximum OD600 faster than the other strains. As in the glucose-gluconate co-feed experiment, a lower increase of the OD600 values of P. putida KT2440 and P. putida KT2440 Δglk was observed when glucose and gluconate were completely consumed and 2-ketogluconate remained as single carbon source indicating that 2- ketogluconate is not taken up in the presence of glucose and gluconate. In contrast to these strains and in concordance with the previous cofeed experiment, such a pausing of OD600 value increase was not observed for P. putida S12, in which either 2-ketogluconate catabolism is not repressed by the other carbon sources or which is able to faster adapt to varying carbon sources.

Table 6: Time dependent uptake (-) and accumulation (+) rates of glucose, gluconate and 2- ketogluconate. P. putida KT2440 (a), P. putida S12 (b) and P. putida KT2440 Δglk (c) were grown on U-13C-glucose and 12C-2-ketogluconate. time [h] glucose rate gluconate rate 2-ketogluconate [mmol/g/h] [mmol/g/h] rate [mmol/g/h] 0 – 1 - 28 ± 1.2 + 22 ± 1.0 + 8 ± 0.3 a 1 – 2.5 - 6 ± 0.3 - 8 ± 0.3 + 5 ± 0.2 2.5 – 4.5 - 5 ± 0.2 0 – 1 - 28 ± 2.4 + 14 ± 1.2 + 11 ± 0.9 b 1 – 2 - 5 ± 0.4 - 10 ± 0.9 + 9 ± 0.8 2 – 4 -6 ± 0.5 0 – 1 - 21 ± 1.3 + 10 ± 0.6 + 4 ± 0.2 1 – 2 - 15 ± 0.9 - 3 ± 0.2 + 7 ± 0.4 c 2 – 2.5 - 7 ± 0.4 - 3 ± 0.2 - 3 ± 0.2 3 – 4 - 7 ± 0.4

All three investigated strains showed distinct growth phases when glucose and 2-ketogluconate were fed as carbon sources (Table 6). Within the first growth phase, glucose was degraded and gluconate as well as 2-ketogluconate accumulated. The second growth phase was represented by simultaneous degradation of glucose and gluconate while 2-ketogluconate further accumulated. 2-ketogluconate consumption characterized the third

64 CHAPTER 2.3 growth phase, in which comparable 2-ketogluconate uptake rates were determined for all strains. The glucose conversion rate was consistently higher during the first growth phase of the glucose-2-ketogluconate cofeed experiment compared to the rate observed when a mixture of glucose and gluconate was provided. This might be explained by an inhibition of glucose transport in presence of gluconate as has been reported for Pseudomonas aeruginosa [108,109]. Again, P. putida KT2440 Δglk degraded glucose slower than the other two strains. This is an indication for simultaneous conversion of glucose via Glk and glucose dehydrogenase in P. putida KT2440 and P. putida S12. 2.3.4.3. Fractional labeling of glucose, gluconate and 2-ketogluconate in cofeed experiments The investigated strains were fed with equimolar amounts of either U-13C-glucose and 12C- gluconate or U-13C-glucose and 12C-2-ketogluconate. During breakdown of the labeled carbon source, the 13C-isotopes are incorporated into the resulting compounds. This labeling can be found in extracellular gluconate and 2-ketogluconate as well as intracellular compounds. Here, for the elucidation of glucose uptake, the label incorporation into the compounds resulting directly from labeled glucose were of great interest. Therefore, we determined and compared the labeling of extracellular glucose, gluconate and 2-ketogluconate. First, we fed the three Pseudomonas cells with U-13C-glucose and 12C-gluconate (Figure 44b - Figure 36b). As expected, the fractional labeling, i.e., the percentage of 13C-isotopes, of the uniformly labeled glucose was determined to approx. 98 % in all experiments, validating the GC-MS measurements.

The labeling of gluconate showed an increase over time for all strains. The maximal fractional labeling was measured when glucose was completely consumed and reached the fractional labeling of glucose for P. putida S12 and P. putida KT2440 Δglk but only about 70 % for P. putida KT2440. This might be explained by either a higher direct uptake of glucose by P. putida KT2440 or a faster consumption of gluconate by P. putida S12 and P. putida KT2440 Δglk. The 13C-incorporation into 2-ketogluconate followed the same transient for all three strains. Interestingly, the final fractional labeling differed from that of the gluconate pool and reached only approx. 45 % for all strains. For experiments using U-13C-glucose and unlabeled 2-ketogluconate as carbon sources, the labeling changes in extracellular metabolites are illustrated in panel b of Figure 37 to Figure 39. Overall, the three strains showed a very similar behavior in the time-dependent label incorporation. Labeling of extracellular glucose remained constant at approx. 98 %. Equally, a constant gluconate labeling of approx. 97 % was found in all experiments. This was expected as gluconate was not present at the beginning of the growth experiment and should exclusively originate from the uniformly labeled glucose. The 2-ketogluoconate labeling increased over time and a conversion of glucose to 2-ketogluconate was observed comparable to experiments using U-13C-glucose and unlabeled gluconate. For P. putida KT2440 and P. putida S12, the 2-ketogluconate concentration increased by 37 and 47 % respectively. At this point in time the fractional labeling of 2-ketogluconate was 42 and 47 % respectively, indicating that only a small amount of 2-ketogluconate has been taken up by the cell, which would have resulted in higher fractional labeling. In P. putida KT2440 Δglk, a net 2-

65 INVESTIGATION OF GLUCOSE UPTAKE IN PSEUDOMONAS PUTIDA ketogluconate formation of 29 % and a fractional labeling of 36 % was observed. Apparently, the P. putida KT2440 mutant was only slowly consuming 2-ketogluconate in the presence of other carbon sources, as well.

66 CHAPTER 2.3

a 25 6

5 20

4 [-] density optical

15 OD glucose gluconate 3 2-ketogluconate total C6 10

2 concentration [m m ol/L]

5 1

0 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 tim e [h]

b 1.0

0.8

glucose 0.6 gluconate 2-ketogluconate

0.4 fractional labeling [-]

0.2

0.0 0.00.51.01.52.02.53.03.54.04.55.05.56.0 tim e [h]

c 1.0

0.8

g6p 0.6 f6p

0.4 fractional labeling [-]

0.2

0.0 0.00.51.01.52.02.53.03.54.04.55.05.56.0 tim e [h]

Figure 34: Time course of substrate concentration (a), labeling (b) and intracellular g6p/f6p labeling (c) of P. putida KT2440 using glucose and gluconate as initial carbon sources. Error bars on fractional labeling data represent the relative error of 10 %.

67 INVESTIGATION OF GLUCOSE UPTAKE IN PSEUDOMONAS PUTIDA

a 25 6

5 20

4 [-] density optical

15 OD glucose gluconate 3 2-ketogluconate total C6 10

2 concentration [m m ol/L]

5 1

0 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 tim e [h]

b 1.0

0.8

glucose 0.6 gluconate 2-ketogluconate

0.4 fractional labeling [-]

0.2

0.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 tim e [h]

c 1.0

0.8

g6p 0.6 f6p

0.4 fractional labeling [-]

0.2

0.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 tim e [h]

Figure 35: Time course of substrate concentration (a), labeling (b) and intracellular g6p/f6p labeling (c) of P. putida S12 using glucose and gluconate as initial carbon sources. Error bars on fractional labeling data represent the relative error of 10 %.

68 CHAPTER 2.3

a 25 6

5 20

4 [-] density optical

15 OD glucose gluconate 3 2-ketogluconate total C6 10

2 ocnrto [mmol/L] concentration

5 1

0 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 tim e [h]

b 1.0

0.8

glucose 0.6 gluconate 2-ketogluconate

0.4 fractional labeling [-]

0.2

0.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 tim e [h]

c 1.0

0.8

g6p 0.6 f6p

0.4 fractional labeling [-]

0.2

0.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 tim e [h]

Figure 36: Time course of substrate concentration (a), labeling (b) and intracellular g6p/f6p labeling (c) of P. putida KT2440 Δglk using glucose and gluconate as initial carbon sources. Error bars on fractional labeling data represent the relative error of 10 %.

69 INVESTIGATION OF GLUCOSE UPTAKE IN PSEUDOMONAS PUTIDA

2.3.4.4. Fractional labeling of glucose-6-phosphate and fructose-6-phosphate in cofeed experiments The quantitative description of the physiology of the three strains in the different media with the time profiles of the pool size and labeling of the extracellular carbon sources revealed different kinetics but could not unambiguously determine the actual activity of the alternative carbon uptake pathways. To gain a deeper insight, labeling changes in intracellular metabolites were determined. In particular, the labeling of g6p and f6p was measured. These metabolites were extracted from quenched cells, and the 13C-labeling pattern determined by GC-MS. The fractional labeling of fragments containing carbon atoms C5-C6 of f6p and g6p is illustrated in Figure 34c to Figure 36c. In case of P. putida KT2440 and P. putida S12, a high fractional labeling of both fragments was determined after the first 1.5 hours of the growth experiment, in which U-13C-glucose and unlabeled gluconate were supplied. The fractional labeling of g6p in P. putida KT2440 and P. putida S12 was 89 % and 86 %, respectively. As these values are in between the fractional labeling values of glucose and gluconate during this period, this observation indicates simultaneous glucose uptake via both the phosphorylative pathway and via gluconate during the first 1.5 hours. The labeling of f6p and g6p strongly correlated. But while in P. putida S12 the labeling was almost identical the fractional labeling of f6p in P. putida KT2440 was considerably lower than that of g6p. f6p is produced from g6p via Pgi and from the PP pathway, hence can originate from both glucose and gluconate (Figure 41). A partially cyclic ED pathway flux is also conceivable, in which the formed g3p is recycled back to f6p or g6p, enabling formation of f6p and g6p from gluconate, as well. On the one hand, the different labeling of g6p of f6p shows that the Pgi activity is not high enough to equilibrate the two metabolites and that the two metabolites originate partially, from different pathways and carbon sources, i.e. g6p synthesis from glucose and f6p synthesis from gluconate or 2- ketogluconate via the ED or the PP pathway. Interestingly, the fractional labeling of g6p in KT2440 remained above the fractional labeling of 2-ketogluconate until all carbon sources were depleted. This might be due to a large intracellular pool of g6p or the fueling of the g6p pool by an internal source, e.g., glycogen, with higher labeling but measurement artefacts cannot entirely be excluded. In glucose-gluconate cofeed experiments, the labeling of both g6p and f6p in P. putida KT2440 Δglk was constantly lower compared to those determined for P. putida KT2440 and P. putida S12. For both metabolites, the same time course for labeling over time was monitored, resulting in a labeling of approx. 51 % after 4 hours of the growth experiment. This equal labeling pattern was expected for this mutant as g6p can only originate from f6p, which is either derived from the PP pathway or a cyclic ED pathway activity. However, the labeling of both fragments was higher than that of gluconate or 2-ketogluconate in the first 2 h of the experiment. We don’t have a proper explanation for this, but it could be that the gluconate and 2-ketoglucoante pools in the periplasmic and extracellular space are not completely equilibrated with the measured extracellular space. Contrary to U-13C-glucose and 12C-gluconate cofeed experiments, all strains showed a high fractional labeling for g6p and f6p after 1 h when fed with U-13C-glucose and 12C-2- ketogluconate (panel c of Figure 37 to Figure 39) and both, g6p and f6p did not show significant differences in their labeling during the first 1.5 hours of the experiment, indicating uptake of mainly uniformly labeled glucose or gluconate within this time in all strains.

70 CHAPTER 2.3

In P. putida KT2440 the labeling of g6p and f6p drifted apart after complete degradation of glucose (~2 h). The faster decrease in f6p labeling indicates simultaneous consumption of gluconate and 2-ketogluconate during this growth phase and formation of f6p mainly from the less labeled 2-ketogluconate either from PP pathway or g3p originating from semi cyclic ED activity ending at f6p. The higher labeling of g6p shows that there is no extensive recycling of g6p via a cyclic operation of the ED pathway. As for the glucose-gluconate cofeed, the final fractional labeling of both fragments was higher than that of 2-ketogluconate, the sole carbon source in the last hour of measurements, explainable either by a nonstationary labeling state or by an additional internal, higher labeled source, e.g., glycogen. In P. putida S12, the labeling of g6p and f6p was very similar to each other over the whole experiment, comparable to experiments using U-13C-glucose and 12C-gluconate. The equilibration of these two pools can only be explained by a recycling of g6p from either PP or ED pathway intermediates, i.e., f6p or g3p. As stated for the P. putida KT2440 experiment, the reduction of labeling after glucose depletion is due to simultaneous usage of gluconate and 2-ketogluconate as carbon source. As g6p and f6p showed a high fractional labeling of 95 % and 92 %, respectively, after 1 h in P. putida KT2440 Δglk, both are apparently mainly formed from the uniformly labeled gluconate derived from U-13C-glucose. After 2 hours of the experiment, glucose and gluconate were almost depleted and the labeling of g6p and f6p was reduced to 81 % and 68 %, respectively. This reduction indicates a simultaneous uptake of uniformly labeled gluconate and partially labeled 2-ketogluconate when gluconate became limiting. The fractional labeling of g6p declined more slowly than that of f6p although g6p is theoretically synthesized from f6p via Pgi activity in this mutant. Again, the difference between g6p and f6p labeling is an indication for either large differences in the pool sizes of these two metabolites or a highly labeled internal carbon source, which might have had accumulated in the beginning of the growth experiment and which fueled the g6p pool after gluconate depletion.

71 INVESTIGATION OF GLUCOSE UPTAKE IN PSEUDOMONAS PUTIDA

a 25 6

5 20

4 [-] density optical

15 OD glucose gluconate 3 2-ketogluconate total C6 10

2 ocnrto [mmol/L] concentration

5 1

0 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 tim e [h]

b 1.0

0.8

glucose 0.6 gluconate 2-ketogluconate

0.4 fractional labeling [-]

0.2

0.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 tim e [h]

c 1.0

0.8

g6p 0.6 f6p

0.4 fractional labeling [-]

0.2

0.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 tim e [h]

Figure 37: Time course of substrate concentration (a), labeling (b) and intracellular g6p/f6p labeling (c) of P. putida KT2440 using glucose and 2-ketogluconate as initial carbon sources. Error bars on fractional labeling data represent the relative error of 10 %.

72 CHAPTER 2.3

a 25 6

5 20

4 [-] density optical

15 OD glucose gluconate 3 2-ketogluconate total C6 10

2 ocnrto [mmol/L] concentration

5 1

0 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 tim e [h]

b 1.0

0.8

glucose 0.6 gluconate 2-ketogluconate

0.4 fractional labeling [-]

0.2

0.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 tim e [h]

c 1.0

0.8

g6p 0.6 f6p

0.4 fractional labeling [-]

0.2

0.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 tim e [h]

Figure 38: Time course of substrate concentration (a), labeling (b) and intracellular g6p/f6p labeling (c) of P. putida S12 using glucose and 2-ketogluconate as initial carbon sources. Error bars on fractional labeling data represent the relative error of 10 %.

73 INVESTIGATION OF GLUCOSE UPTAKE IN PSEUDOMONAS PUTIDA

a 25 6

5 20

4 pia est [-] density optical

OD glucose 3 gluconate

10 2-ketogluconate total C6

concentration [m m ol/L] 2

5 1

0 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 tim e [h]

b 1.0

0.8

glucose 0.6 gluconate 2-ketogluconate

fractional labeling0.4 [-]

0.2

0.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 tim e [h]

c 1.0

0.8

g6p 0.6 f6p

fractional labeling0.4 [-]

0.2

0.0 0.00.51.01.52.02.53.03.54.04.55.05.56.0 tim e [h]

Figure 39: Time course of substrate concentration (a), labeling (b) and intracellular g6p/f6p labeling (c) of P. putida KT2440 Δglk using glucose and 2-ketogluconate as initial carbon sources. Error bars on fractional labeling data represent the relative error of 10 %.

74 CHAPTER 2.3

The equal labeling g6p and f6p in P. putida S12 revealed a pronounced Pgi activity with a net flux from f6p to g6p, indicating a potentially cyclic operating ED pathway. The same phenomenon was also observed when P. putida KT2440 Δglk was fed with U-13C-glucose and 12C-gluconate. This is expected as P. putida KT2440 Δglk is not able to directly produce g6p from glucose but has to be formed from f6p. Interestingly, however, the labeling of g6p and f6p during growth on a mixture of glucose and 2-ketogluconate deviated substantially and the final fractional labeling of g6p was much higher as expected. The same behavior was observed for P. putida KT2440. One hypothesis for this is the formation of g6p from a highly labeled storage molecule such as glycogen, which might have had accumulated intracellularly during growth on uniformly labeled glucose or gluconate. The equal fractional labeling of g6p and f6p in P. putida KT2440 Δglk fed with U-13C-glucose and 12C-gluconate is explained by full conversion of labeled glucose to 13C-gluconate and uptake this intermediate from a gluconate pool with an overall labeling significantly below 100 % most of the duration of the experiment. Potentially accumulating glycogen, hence, would also not highly labeled. 2.3.4.5. Fractional labeling of intracellular metabolites Since the evaluation of g6p and f6p labeling did only allow for few statements concerning the uptake of differently labeled substrates, the labeling of the remaining intracellular metabolites was inspected to gain more information. Metabolites from the ED pathway as well as the PP pathway were measured using GC-MS and the labeling over time was determined. In experiments with U-13C-glucose and 12C-gluconate it was observed, that except for g6p, f6p and pyruvate the three strains showed only small differences in time-dependent labeling (Figure 40). For P. putida S12, a rapid drop of the labeling was observed for the three metabolites. This drop can be an indication for a fast change of the substrate used for growth as well as a quick turnover rate of the metabolites. In P. putida KT2440 and P. putida KT2440 Δglk the reduction of metabolite labeling was smoother, signaling a longer adaptation phase for growth on 2-ketogluconate. Except for pyruvate, a very similar labeling kinetics was found for the metabolites in P. putida KT2440 and P. putida KT2440 Δglk. In case of P. putida KT2440 Δglk, the pyruvate labeling was consistently higher as for the other strains. A reason for this delayed and slowed-down label reduction might be the reduced carbon uptake rate and consequently reduced metabolic activity compared to the two wildtype strains.

75 INVESTIGATION OF GLUCOSE UPTAKE IN PSEUDOMONAS PUTIDA

Figure 40: Fractional labeling of intracellular central carbon metabolites in growth experiments using P. putida KT2440 (green), P. putida S12 (red) and P. putida KT2440 Δglk (blue). The cells were grown on U-13C-glucose and 12C-gluconate.

76 CHAPTER 2.3

When the labeling of the remaining metabolites was compared, it was observed that the labeling of PP pathway metabolites was lower than the labeling of the intermediates of the other pathways. As for the ED pathway, 6pg represents the initial metabolite for PP pathway. The labeling of metabolites originating from 6pg should therefore be equal without any dependence on their appearance. Substrate channeling, i.e., the direct transfer of intermediates from one enzyme to the downstream enzyme of a pathway could be an explanation for this phenomenon. The here observed labeling pattern might be explained by channeling of 6pg originating from either g6p (glucose) or gluconate to Edd, the first enzyme of the ED pathway and 6-phosphogluconate dehydrogenase (6Pgd), the first enzyme of the PP pathway, respectively. In contrast to experiments using U-13C-glucose and 12C-gluconate, a very similar labeling course was observed for the three strains as well as the different pathway intermediates when cells were fed using U-13C-glucose and 12C-2-ketogluconate. While labeling of ribose-5- phosphate (r5p) and ribulose-5-phosphate (ru5p) was lower compared to the remaining metabolites in experiments using gluconate as substrate, no significant differences between the metabolites was observed when 2-ketogluconate was used as substrate. Since glucose and the formed gluconate were both fully labeled in experiments with U-13C-glucose and 12C-2- ketogluconate, no differences in labeling within the two hypothesized 6pg pools was observed and the resulting metabolites showed comparable labeling courses.

77 INVESTIGATION OF GLUCOSE UPTAKE IN PSEUDOMONAS PUTIDA

Figure 41: Fractional labeling of intracellular central carbon metabolites in growth experiments using P. putida KT2440 (green), P. putida S12 (red) and P. putida KT2440 Δglk (blue). The cells were initially grown on U-13C-glucose and 12C-2-ketogluconate.

78 CHAPTER 2.3

2.3.5. Conclusion P. putida has the ability to oxidize glucose to gluconate and 2-ketogluconate within the periplasmic space in addition to direct uptake and phosphorylation to g6p. When grown on glucose in combination with either gluconate or 2-ketogluconate, P. putida KT2440, P. putida S12 as well as P. putida KT2440 Δglk showed differences in growth behavior as well as substrate uptake rates. P. putida S12 revealed a better ability of a simultaneous uptake of different substrates compared to the other strains. A high phoshoglucose isomerase activity could be shown for P. putida S12 by investigating the labeling of intracellular metabolites. High labeling of g6p after glucose depletion in P. putida KT2440 and P. putida KT2440 Δglk points to partial g6p formation from fully labeled glycogen that might have accumulated at the beginning of the growth experiment. Variances in PP pathway intermediate labeling observed for the different substrate strategies, are an indication for substrate channeling. In this process, the intermediate formed by a selected enzyme is directly handed over to another enzyme. In such a scenario, 6pg originating from gluconate and g6p might be independently used in the ED pathway and PP pathway. Higher labeling of ED pathway metabolites in comparison to PP pathway metabolites in experiments using glucose and gluconate indicated that 6pg used for PP pathway originates from gluconate. Concludingly, it could be shown, that 13C-labeling experiments followed by determination of intracellular metabolite labeling are crucial for evaluation of intracellular processes. Deeper insights into the cell’s metabolism have to be carried out by means of 13C-MFA, interpreting metabolite labeling data on basis of central carbon metabolism modeling.

Chapter 2.4

High quality 13C-metabolic flux analysis for examination of cyclic ED activity in different Pseudomonas putida strains

Author Contributions: Andreas Schmitz planed and designed the project, performed the experiments and analyzed the results. The chapter was written with the help of Birgitta Ebert. Birgitta Ebert and Lars M. Blank supervised and conceived the study. The LC-MS analysis was kindly performed by Prof. Marco Oldiges and Petra Geilenkirchen (Bioprocesses and Bioanalytics Group at Institute of Bio- and Geosciences: IBG-1 Biotechnology, Forschungszentrum Jülich, Germany)

81 HIGH QUALITY 13C-MFA FOR EXAMINATION OF CYCLIC ED ACTIVITY

2.4. High quality 13C-metabolic flux analysis for examination of cyclic ED activity in different Pseudomonas putida strains 2.4.1. Summary Among Gram-negative bacteria, the species of Pseudomonads is of high industrial interest due to its high tolerance against various growth conditions. For example, Pseudomonads are able to grow in the presence of solvents and toxic substances as well as on different substrates. Because of this, Pseudomonads are increasingly considered to be a promising host for biotechnological processes and are already used for these processes. Despite the widespread use of this organism and the examination of its metabolism, there are still many open questions regarding the metabolism, including the central carbon network operation. Pseudomonads are not able to metabolize glucose via the Emden-Meyerhoff-Parnas (EMP) pathway but mainly employ the Entner-Doudoroff (ED) pathway and to some extent the pentose phosphate pathway. It is assumed that the ED pathway is operated in a cyclic mode, in which glyceraldehyde-3-phosphate (g3p) and pyruvate are recycled to fructose-6-phosphate (f6p) and glucose-6-phosphate (g6p) to cover the cellular demand for nicotinamide adenine dinucleotide phosphate (NADPH). In this work, experimental design simulations employing stoichiometric and isotopomer modeling were carried out to support the examination of the assumed cyclic ED pathway and revealed a mixture of U-13C-glucose and 1-13C-glucose to be suitable for the identifiability of the desired fluxes. Growth experiments of the three Pseudomonas strains, P. putida KT2440, P. putida S12, and P. taiwanensis VLB120, were carried out using a 20:80 mixture of U-13C- glucose and 1-13C-glucose as initial substrate. The samples harvested during the growth experiments were analyzed by means of gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LC-MS) after appropriate sample preparation. These data were further investigated by means of 13C-metabolic flux analysis (13C-MFA). This 13C-MFA of the three strains, revealed a pronounced usage of the cyclic ED pathway under the used growth conditions indicating a high demand on NADPH during growth. 2.4.2. Introduction Because of their versatile metabolism, high metabolic activity and tolerance of harsh growth conditions, Pseudomonads are promising hosts for biotechnological production of miscellaneous industrially usable substances. Aromatic acids, alcohols and rhamnolipids only reflect a small selection of possible substances producible with Pseudomonads [92-94]. Glucose, as the currently most frequently used substrate for biotechnological processes is metabolized via the Entner-Doudoroff (ED) pathway in Pseudomonads. In contrast to most model microorganisms, like Escherichia coli, Pseudomonads do not possess a functional Emden-Meyerhoff-Parnas (EMP) pathway and instead employ the ED pathway as the main catabolic route for glucose degradation (Figure 42) [110,111]. This is explained by the missing enzyme phosphofructokinase-1 (Pfk-1) which catalyzes the reaction from fructose-6- phosphate (f6p) to fructose-1,6-bisphosphate (fbp). A further feature of Pseudomonads is that glucose can either be converted to gluconate and 2-ketogluconate in the periplasmic space or directly be transported into the cytosol where it is converted to glucose-6-phosphate (g6p). This intermediate further reacts to 6-phosphogluconate (6pg). After incorporation into the cytosol, gluconate and 2-ketogluconate are also converted to 6pg (Chapter 2.3) [99,112], the starting compound of the ED pathway, which is converted to 2-keto-3-deoxy-6-

82 CHAPTER 2.4 phosphogluconate (kdpg) by phosphogluconate dehydratase (Edd) activity and then cleaved to glyceraldehyde-3-phosphate (g3p) and pyruvate (pyr) by the use of 2-keto-3-deoxy-6- phosphogluconate aldolase (Eda) [105,113].

Figure 42: Flux map of Pseudomonads containing oxidative glucose uptake, ED pathway, EMP pathway, Pentose Phosphate (PP) pathway, tricarboxylic acid (TCA) cycle and glyoxylate shunt.

In addition to the ED pathway, the PP pathway and TCA cycle, and a partial EMP pathway from fbp via f6p to g6p exist in Pseudomonads. In P. putida KT2440, the enzymes responsible for these reactions could be shown to be active and a potential cyclic system of ED pathway and EMP pathway was reported (Figure 43) [114].

83 HIGH QUALITY 13C-MFA FOR EXAMINATION OF CYCLIC ED ACTIVITY

Figure 43: Schematic flux map of glucose uptake (oxidative glucose uptake not shown), ED pathway and partial EMP pathway in Pseudomonads. The reactions greyed out represent a potential cyclic system of ED and EMP pathway. Glucose is incorporated and transformed to 6pg. In the ED pathway, 6pg is transformed to pyr and g3p which can further be transformed to fbp and f6p in EMP pathway. Since f6p cannot be transformed to fbp via Pfk-1, f6p is potentially channeled to the ED pathway or PP pathway via g6p.

This cyclic system is of great interest for biotechnological approaches due to differences in cofactor regeneration among the alternative pathways. These differences can be seen in nicotinamide adenine dinucleotide phosphate (NADPH) and nicotinamide adenine dinucleotide (NADH) production by the cell, when different pathways are used. NADPH is

84 CHAPTER 2.4 formed within PP pathway and ED pathway whilst NADH is formed in the lower EMP pathway branch from g3p to pyr [110,115]. Thus, when the organism has a higher demand of NADPH than NADH, the cyclic use of ED pathway and EMP pathway is likely. These cofactor balances should be considered when designing effective metabolic engineering strategies. For this purpose the pathways need to be further investigated. A common way of investigating metabolic network operation is 13C-metabolic flux analysis (13C-MFA). Typically, this technique is carried out by measuring the labeling of metabolites such as proteinogenic amino acids after growth on labeled substrates [4,17,18]. The metabolite labeling is determined and used for computational flux analysis. Labeling of amino acids in this case is composed of the labeling of the precursor molecules in central carbon metabolism (CCM) [8,24]. Besides 13C-MFA using proteinogenic amino acids, approaches using intracellular central carbon metabolites are increasingly used. The challenge for measuring these intracellular central carbon metabolites in microorganisms is the high turnover rate of these metabolites. Therefore, the metabolism has to be stopped rapidly to prevent changes in metabolite labeling that would falsify the results of the 13C-MFA. One way of rapidly stopping the metabolism of an organism is enzyme quenching [32,116]. For enzyme quenching, the organism is transferred into an appropriate amount of cold quenching solution to stop intracellular reactions. The composition of quenching solution as well as their temperature are depending on the microorganism used and their ability to metabolite leakage [117]. Hence, eukaryotic organisms like Saccharomyces cerevisiae should be quenched with pure methanol at low temperature below 40°C whereas a cold 40% ethanol-sodium chloride (0.8%, w/v) solution at -20°C is suitable for prokaryotic organisms like Escherichia coli and Corynebacterium glutamicum [32,118]. Within this project, a method for clarifying cyclic ED pathway activities had to be found. Therefore, strategies of experimental design based on constraint-based modeling in combination with isotopomer modeling were performed for evaluation of intracellular metabolites and 13C-glucose isomers that allow to unambiguously identify cyclic ED pathway operation. The optimal labeling strategy was applied for flux determination of three different Pseudomonas strains in batch fermentation. 13C-incorporation in proteinogenic amino acids and central carbon metabolite was analyzed by gas chromatography-mass spectrometry (GC- MS) and liquid chromatography-mass spectrometry (LC-MS). 13C-MFA was performed combining the labeling information gained by the three different measuring types. 2.4.3. Materials and Methods 2.4.3.1. Strains and Growth Conditions The three Pseudomonas strains used in this study were P. putida KT2440, P. taiwanensis VLB120, and P. putida S12. The medium used was a mineral salt medium that contained (per liter) 3.88 g K2HPO4, 2.12 g NaH2PO4 ∙ 2 H2O, 2.00 g (NH4)2SO4, 0.01 g EDTA, 0.10 g MgCl2 ∙ 6 H2O, 2.0 mg ZnSO4 ∙ 7 H2O, 1.0 mg CaCl2 ∙ 2 H2O, 5.0 mg FeSO4 ∙ 7 H2O, 0.2 mg Na2MoO4 ∙ 2 H2O, 0.2 mg CuSO4 ∙ 5 H2O, 0.4 mg CoCl2 ∙ 6 H2O and 1.0 mg MnCl2 ∙ 2 H2O [71]. The medium was completed with 3.6 g/L glucose composed of 20% U-13C-glucose and 80% 1-13C-glucose. Precultures were performed in 500 mL shake flasks with 50 mL medium and 12C-glucose at 30 °C and 250 rpm. The main cultures were cultivated in New Brunswick BioFlo 115 benchtop bioreactors (Eppendorf, Germany) at 30 °C. The batch fermentations were executed using a 1.3 L reactor with a working volume of 500 mL. Air supply was realized using

85 HIGH QUALITY 13C-MFA FOR EXAMINATION OF CYCLIC ED ACTIVITY headspace aeration with 2 vvm of compressed air. The culture was stirred with 750 rpm at the beginning and stirring speed was increased when the dO2 signal dropped to 35 %. 2.4.3.2. Sample Preparation Hydrolysis of Biomass for Amino Acid Analysis

0.3 mgCDW biomass was taken from the culture and centrifuged (16.873 x g, 5 min). The supernatant was discarded and the cells were resuspended with 0.9% (w/v) NaCl solution and centrifuged again. After discarding the supernatant, the cell pellet was resuspended in 150 μL of 6 M HCl and vortexed for 15 s. The suspension was transferred into a conical 1.5 mL glass vial. The vial was tightly closed and incubated for 6 h in a heating block at 105°C for hydrolysis of the biomass. After hydrolyzation, HCl was evaporated at 85°C under a hood. The dried sample was stored in a closed vial at room temperature. Quenching and Extraction of Biomass for Intracellular Metabolite Analysis

For cell quenching, a minimal amount of 10 mgCDW was directly transferred into a cold 70% ethanol / NaCl (0.9% (w/v)) solution at -52 °C and shaken immediately. The solution was centrifuged (4800 x g) for 7 minutes at -10 °C and the supernatant was discarded afterwards. The cell pellet was stored at -80 °C. Extraction of quenched cells was performed with a -50 °C cold methanol:chloroform:water (2.5:1:1) solution. The cells were resuspended in 2 mL of the extraction solution and vortexed afterwards. The solution was kept at -20 °C while being shaken on an overhead shaker for at least 1 h. After extraction, the solution was centrifuged for 7 minutes at -10 °C, 1.3 mL of the supernatant were transferred into an Eppendorf tube and mixed with 300 μL H2O. The solution was centrifuged for 10 minutes at 4 °C and 1.2 mL of the water phase were filtered into glass vials. The samples were dried by first evaporating the methanol fraction in a centrifugal concentrator (ScanSpeed MaxiVac Beta, Labogene, Allerød, Denmark) at 20 °C for 60 min. The concentrated samples were frozen at -80 °C and subsequently placed in a lyophilizer (ScanSpeed MaxiVac Beta, Labogene, Allerød, Denmark) at – 110 °C over night for water elimination until complete dryness of the samples. 2.4.3.3. Analytics Optical Density Measurement

86 CHAPTER 2.4 divinylbenzol copolymer, PS-DVB: 300 × 8.0 mm, CS-Chromatographie, Langerwehe, -1 Germany). 5 mM H2SO4 was used as eluent at a flow of 0.8 mL h min at 75 °C. The analytes were detected by an ultraviolet light (UV) detector (UV detector LC-166, Beckmann Coulter, Krefeld, Germany) at a wavelength of 210 nm and a refractive index (RI) detector (LCD 201, Gynkotek, Munich, Germany) system. Extracellular Rates Extracellular rates such as growth rate and carbon uptake / production rates of glucose, gluconate and 2-ketogluconate were determined with Microsoft Excel 2013. The growth rate was calculated as a linear function of the logarithm of OD600 over time. Uptake and production yields were calculated as the function of analyte and biomass concentration. Substrate uptake and product formation rates were calculated by multiplication of growth rate and corresponding yield. Derivatization Hydrolyzed biomass samples were derivatized with N-methyl-N-(tert-butyldimethylsilyl)- trifluoroacetamide (MBDSTFA, CS - Chromatographie Service GmbH, Langerwehe, Germany). The dried cell hydrolysate was resuspended in 30 μL acetonitrile and 30 μL MBDSTFA. The glass vial was tightly closed, vortexed for 15 s and incubated for 1 h in a heating block at 85°C. Quenched biomass samples were derivatized with 30 μL of pyridine / methoxyamine (20 mg/ml) at 67 °C for 1 h. After incubation 30 μL N-Methyl-N-trimethylsilyl- trifluoroacetamide (MSTFA, CS - Chromatographie Service GmbH, Langerwehe, Germany) was added and samples were incubated again for 1 h at 67 °C. The samples were cooled down to room temperature and analyzed immediately. Gas Chromatography–Mass Spectrometry GC separation was performed on a Thermo Scientific (Thermo Scientific, Waltham, MA, USA) Trace GC Ultra equipped with a Thermo Scientific AS 3000 autosampler. The column used was a Thermo Scientific (Thermo Scientific, Waltham, MA, USA) TraceGOLD TG- 5SilMS capillary column (length, 30 m; inner diameter, 0.25 mm; film thickness, 0.25 μm). Separation of amino acids was performed at a constant flow rate of 1 ml helium min-1. A sample volume of 1 μL was injected into a split/splitless injector at 270 °C while a split ratio of 1:15 was used. The temperature of the GC oven was kept constant for 1 min at 140 ºC and afterwards increased with a gradient of 10 ºC min-1 to 310 ºC and again kept constant for 1 min. Intracellular metabolites were separated at a constant flow rate of 1 mL helium min-1. 1 μL sample was injected into a programmed temperature vaporization injector at constant temperature of 270 °C while splitless injection was performed. The temperature of the GC oven was kept constant for 5 min at 70 ºC and afterwards increased with a gradient of 6 ºC min-1 to 320 ºC and again kept constant for 10 min. Mass spectrometry analysis of amino acids was performed on a Thermo Scientific (Thermo Scientific, Waltham, MA, USA) ISQ single quadrupole mass spectrometer. The temperatures of the transfer line and the ion source were both set to 280 ºC. Ionization was performed by

87 HIGH QUALITY 13C-MFA FOR EXAMINATION OF CYCLIC ED ACTIVITY electron impact (EI) ionization at 70 eV. Amino acid fragments were determined with a selected ion monitoring mode. Mass spectrometry analysis of intracellular metabolites was performed on a Thermo Scientific (Thermo Scientific, Waltham, MA, USA) TSQ triple quadrupole mass spectrometer. The temperatures of the transfer line and the ion source were both set to 280 ºC. Ionization was performed by EI ionization at 70 eV while masses from 60 to 600 were detected with a scan time of 0.25 sec scan-1. GC-MS raw data were analyzed using Xcalibur [107]. Liquid Chromatography–Tandem Mass Spectrometry LC separation was performed on a Jasco (Jasco, Tokyo, Japan) X-LC 3000 Series HPLC system. The HPLC system was equipped with a Phenomenex (Phenomenex, Aschaffenburg, Germany) Synergi Hydro-RP (C18) column (length, 100 mm; inner diameter, 3.1 mm; particle size, 2.5 μm; pore size, 80 Å). For separation of intracellular metabolites, a gradient system of eluent A (10mM tributylamine aqueous solution adjusted pH to 4.95 with 15mM acetic acid) and eluent B (methanol) was applied at a temperature of 60 °C. The time course of the gradient system was as follows: 2 min (100% A),5 min (80% A), 8 min (80% A), 10 min (65%), 14 min (0% A), 15 min (0% A), 15.5 min (100% A) and 17 min (100% A). A flow rate of 0.45 ml min-1 and an injection volume of 10 μl were used for LC measurement. The HPLC system was coupled to an Applied Biosystems/MDS Sciex (AB/MDS Sciex, Concord, Canada) API4000 triple quadrupol mass spectrometer equipped with a TurboIon spray source. Negative ion mode at – 4500 V and multiple reaction monitoring mode were used for operating the MS measurement. The measured data were acquired and evaluated via Applied Biosystems/MDS Sciex (AB/MDS Sciex, Concord, Canada) Analyst software (version 1.5) [119]. 2.4.3.4. Experimental Design For targeted application of 13C-labeled substrates in growth experiments and subsequent measurement of metabolite fragments, the experiments were simulated in silico before being applied in vitro. The flux balance analysis was conducted using COBRA Toolbox v2.0, a MATLAB package for constraint-based simulations of metabolic network capabilities [120]. The metabolic network for P. putida was implemented and parameters such as glucose uptake fluxes and active as well as inactive PP pathway were preset. Flux distributions were subsequently simulated using a maximum pyruvate formation as objective function with respect to these constraints. The computed flux distributions were used for forward simulation of metabolite mass distribution vectors (MDVs) in OpenFLUX [12]. These simulations were run for various 13C-labeled glucose isotopomer mixtures and flux distributions with distinct intracellular fluxes. The resulting simulated labeling data were investigated searching for metabolite fragments showing distinct MDVs for different flux distributions and therefore the most informative labeling strategy. 2.4.3.5. 13C-Metabolic Flux Analysis The measured data were evaluated by Thermo Xcalibur 2.2 SP1.48 (Thermo Scientific, Waltham, MA, USA) applying the Processing Setup. Here the ICIS peak identification

88 CHAPTER 2.4 method was used and the parameters “Baseline window,” “Area noise factor,” and “Peak noise factor” were adjusted peak specific for determination of mass distribution data. iMS2Flux was used for correction of mass distribution data for natural labeling and original biomass [84]. Corrected labeling data were used in combination with physiological data for 13C-metabolic flux analysis. OpenFLUX was applied for calculation of intracellular fluxes in CCM of P. putida [12]. 2.4.4. Results and Discussion Growth experiments with different Pseudomonads were exemplarily carried out with the three wild-type strains P. putida KT2440, P. putida S12, and P. taiwanensis VLB120. The strains examined in this work differ in their growth properties and in the presence of glucose uptake pathways. While all strains can directly take up glucose to intracellularly phosphorylate it to glucose-6-phosphate, the periplasmic oxidation of glucose differs. In P. taiwanensis VLB120 glucose is oxidized periplasmically to gluconate, which in contrast to the other Pseudomonads strains tested here cannot be further converted to 2-ketogluconate in this compartment. This might have an effect on the energy metabolism of this strain and therefore on the entire CCM. The experimental design studies using the COBRA toolbox were performed to determine the optimal labeling strategy for the 13C-MFA. The main goal of the tracer experiment was to investigate and quantify a cyclic ED pathway activity. The results of the experimental design, the resulting growth experiments with corresponding physiological data as well as the final 13C-MFA using OpenFLUX are presented in the following sections. 2.4.4.1. Experimental Design of the 13C-Tracer Experiment Prior to the laboratory experiment, in silico simulations of the experiment can be helpful in order to minimize the experimental effort which can be noted in cost savings in the case of growth experiments with 13C-labeled substrates. In this work, simulations were carried out with the help of the COBRA Toolbox and OpenFLUX before the growth experiments for simulation of the best suitable 13C-tracer strategy concerning ED pathway quantification. More specifically, we computed flux distributions of the CCM network of Pseudomonas with different fluxes through a linear or cyclic ED pathway using the Flux Balance Analysis algorithm provided with the COBRA toolbox [120]. These flux distributions were characterized by either a strongly pronounced, weakly pronounced or absent cyclic ED pathway. In addition, variations of the fluxes involved in the glucose uptake as well as two differently pronounced fluxes into PP pathway were used for the simulations. In total 30 different flux distributions were calculated and used as input for an isotopomer model implemented in OpenFLUX [12], with which the incorporation of the 13C-isotopes into central carbon intermediates was calculated for different single 13C-glucose isotopomers or mixtures of it. In addition to simulations with glucose as the sole carbon source, in silico cofeed experiments with 13C-labeled glucose and unlabeled gluconate or 2-ketogluconate were computed (data not shown). Finally, three labeling strategies with glucose as sole carbon source (50% U-13C-glucose with 50% 1-13C-glucose / 20% U-13C-glucose with 80% 1-13C- glucose / 100% 1-13C-glucose) were examined in detail by simulation of the corresponding MDVs. The resulting mass distribution data were inspected to determine, which applied 13C- tracer resulted in flux distribution dependent 13C-label incorporation of metabolites (Table 7).

89 HIGH QUALITY 13C-MFA FOR EXAMINATION OF CYCLIC ED ACTIVITY

Table 7: (a) Enzymatic activity of selected enzymes simulated in Cobra Toolbox. 13C-glucose was used as sole carbon source for simulating the flux distributions. The position of the selected enzymes in CCM is shown in Figure 42. (b) For each of the six flux distributions calculated by COBRA Toolbox, MDVs for selected metabolite fragments were simulated in OpenFLUX. For these simulations the three labeling strategies 50% U-13C-glucose with 50% 1-13C-glucose (I), 20% U-13C-glucose with 80% 1- 13C-glucose (II) and 100% 1-13C-glucose (III) were used. MDVs of the fragments marked in dark grey were insensitive to changes in the flux distribution for the labeling strategy simulated. MDVs of the fragments marked in light grey varied in dependence on the flux distribution. a enzymatic activity [%] 1 2 3 4 5 6 glucose dehydrogenase 100 75 50 100 100 100 gluconate dehydrogenase 100 75 50 100 60 17 glucokinase 0 25 50 0 0 0 gluconokinase 0 0 0 0 40 83 2-ketogluconate kinase 100 75 50 100 60 17 glucose-6-phosphate isomerase -100 -50 0 -100 -50 -6.7 6-phosphogluconate dehydrogenase 0 0 0 10 10 10 b labeling strategy I II III I II III flux distribution 1 - 3 4 - 6 pyr1 pyr1-3 f6p5-6 f6p4-6 f6p1-6 g6p5-6 g6p4-5 g6p1-6 6pg5-6 6pg1-4 6pg1-6

As can be seen in Table 7, the MDVs of the fragments f6p5-6, f6p4-6, g6p5-6, g6p4-6 as well as 6pg5-6 did not show any correlation with flux variations regardless of the labeling strategy used for the simulation. The labeling of pyr1, pyr1-3 and f6p1-6 was only affected for flux distributions with an active PP pathway. The fragments which are unrestrictedly usable for the investigation of the ED pathway fluxes were g6p1-6, 6pg1-4 and 6pg1-6. The comparison of the different labeling strategies showed that 100% 1-13C-glucose is not fully suitable for the investigation of the ED pathway fluxes because no differences in the MDVs of f6p1-6 and g6p1-6 were detectable when the PP pathway was active. The isotopomer mixtures of either 50% U-13C-glucose and 50% 1-13C-glucose or 20% U-13C-glucose plus 80% 1-13C-glucose showed comparable results in the isotopomer simulation, wherefore the following growth

90 CHAPTER 2.4 experiments were performed using 20% U-13C-glucose and 80% 1-13C-glucose as carbon source. This decision was based on the price differences of U-13C-glucose and 1-13C-glucose, which makes the experiments more cost-effective with a 20:80 mixture. 2.4.4.2. Physiology of Pseudomonads Grown on U-13C-Glucose and 1-13C-Glucose When the three strains were grown on glucose, examination of these experiments revealed differences in the growth behavior of the strains. The strain P. taiwanensis VLB120 grew slightly slower at a growth rate of 0.63 ± 0.05 1/h than the strains P. putida KT2440 with 0.66 ± 0.02 1/h and P. putida S12 with 0.70 ± 0.04 1/h. Even though only slight differences were observed concerning the growth rates of the respective strains, significant differences were found in the extracellular concentrations of gluconate and 2-ketogluconate of the investigated strains.

91 HIGH QUALITY 13C-MFA FOR EXAMINATION OF CYCLIC ED ACTIVITY

25 6 a OD glucose gluconate

2-ketogluconate 5 20 total C6

4 [-] density optical

2 concentration [mmol/L]

5 1

0 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 tim e [h]

25 6 OD b glucose gluconate

2-ketogluconate 5 20 total C6

4 [-] density optical

2 concentration [mmol/L]

5 1

0 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 11.0 11.5 tim e [h]

25 6 c OD glucose gluconate

2-ketogluconate 5 20 total C6

4 [-] density optical

2 concentration [mmol/L]

5 1

0 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 tim e [h]

Figure 44: Time course of substrate and product concentrations correlated with growth curve of P. putida KT2440 (a), P. putida S12 (b) and P. taiwanensis VLB120 (c) using a mixture of 20% U-13C-glucose and 80% 1-13C-glucose as initial carbon source.

92 CHAPTER 2.4

As observed in earlier studies (see Chapter 2.3), the strain P. putida KT2440 (Figure 44 a) did not show a pronounced accumulation of gluconate and 2-ketogluconate. During the growth of P. putida KT2440, a change in the growth behavior was observable after approximately 6 h, which is the time the concentration of gluconate in the medium began to decrease. This change in growth behavior might be attributed to adaptation to the use of gluconate as main carbon source, wherefore the organism grew with a lower growth rate during this time whilst the carbon source was changed. Such a behavior was also observed in the growth of P. taiwanensis VLB120 (Figure 44 c). Compared to the growth of P. putida KT2440, a temporary reduction of the growth rate was also observed after approx. 6 h of growth. This change in growth behavior can also be associated with the decrease in gluconate concentration in the medium. In contrast to the growth of P. putida KT2440, no further glucose was present in the medium after this change in the growth behavior of P. taiwanensis VLB120, so that the strain used gluconate as sole carbon source. Gluconate accumulation during growth on glucose was significantly higher than for P. putida KT2440, whereas no 2-ketogluconate was found in the medium due to the fact that this strain has no gluconate dehydrogenase. This also explains the increased accumulation of gluconate of up to 11.75 mmol/L, since it can only be converted via gluconokinase. The lack of gluconate dehydrogenase in P. taiwanensis VLB120 can also be a justification for the slightly lower growth rate compared to the two P. putida strains, which makes this strain less efficient in energy metabolism since the reaction of gluconate to 2-ketogluconate provides electrons which directly are channeled into the electron transport chain. Even though the strain P. putida S12 showed the highest growth rate of the three strains compared, a slight lag phase was observed at the beginning of the growth experiment (Figure 44 b). Comparable to growth of P. putida KT2440, accumulation of gluconate and 2- ketogluconate could be observed. However, in P. putida S12 the detected concentrations of gluconate with 1.56 mmol/L and 2-ketogluconate with 2.71 mmol/L were both higher than for P. putida KT2440. In contrast to the other two strains, no change in growth behavior could be observed in the exponential phase of growth. An explanation for this might be that the metabolic fluxes in the periplasm are already intensively used during the growth on glucose in P. putida S12. Therefore, the strain already utilizes gluconate and 2-ketogluconate for growth and adaptation of the metabolism to these substrates is not necessary. 2.4.4.3. Flux Analysis of Pseudomonads Grown on U-13C-Glucose and 1-13C-Glucose The above described growth experiments were executed using a mixture of 20:80 of U-13C- glucose and 1-13C-glucose for evaluation of the cyclic ED pathway fluxes. Intracellular fluxes were determined by 13C-MFA based on the labeling of intracellular metabolites measured by GC-MS. Therefore, during the growth experiments of the three tested strains, samples were taken for the labeling analysis of different metabolites. Both samples for the measurement of intracellular metabolites, which were immediately quenched, and samples the labeling measurement of proteinogenic amino acids were taken. In addition to the samples used for amino acid analysis, the quenched samples were analyzed using GC-MS and LC-MS to measure different metabolites as well as fragments to increase the information content for 13C- MFA.

93 HIGH QUALITY 13C-MFA FOR EXAMINATION OF CYCLIC ED ACTIVITY

Table 8: Measured fragments suitable for 13C-MFA. The fragments measured by GC-MS after hydrolysis or quenching as well as fragments measured by LC-MS after quenching are shown. “X” represents the method, with which the fragment was measured. The backbone carbon atoms contained in the fragment are indicated with “1”. The backbone carbon atoms which were cleaved of from the fragment are indicated with “0”.

backbone carbon atoms measuring technique

ALA 011 ASP 1100 ASP 1111 HIS 011111 HIS 111111 ILE 11111 LEU 11111 LYS 11111 hydrolysis / GC-MS PRO 01111 PRO 11111 SER 011 SER 110 SER 111 THR 0111 TYR 110000000 pg2 111 pg3 111 cit 110111 dhap 111 e4p 0111 f6p 000111 f6p 000011 g6p 110000 g6p 000011 quenching / GC-MS g6p 001111 gap 111 pep 111 r5p 00011 ru5p 00111 ru5p 00011 s7p 0000011 s7p 0001111 suc 1111 g6p 111111 quenching / LC-MS g6p 111111

The measurements resulted in 35 fragments of different amino acids and metabolites, with sufficient quality for the subsequent flux analysis (Table 8). GC-MS measurements of the hydrolyzed biomass resulted in 15 fragments from a total of 10 amino acids which could be

94 CHAPTER 2.4 used for the subsequent flux analysis. The quenched samples yielded 18 fragments from 13 intracellular metabolites in the GC-MS analysis and 2 fragments from 2 intracellular metabolites in the LC-MS analysis. The 35 fragments used were corrected with iMS2Flux and further utilized in OpenFLUX to calculate the flux distributions of the respective strains. The physiological data obtained from the growth experiments (Chapter 2.4.4.2) were applied for the 13C-MFA. Due to the selected labeling strategy, accurate determination of glucose uptake fluxes in the periplasm was not possible. Therefore, the glucose uptake rate as well as gluconate and 2-ketogluconate accumulation rates were aggregated to a total C6 uptake rate. This was in the case of P. putida KT2440 9.87 ± 0.48 mmol/g/h in P. putida S12 10.14 ± 0.71 mmol/g/h and in P. taiwanensis VLB120 9.18 ± 0.67 mmol/g/h. These total C6 uptake rates were combined with the determined growth rates to calculate the fluxes of precursors into biomass synthesis during the respective growth experiments which was further used as a basis for the 13C-MFA. In this 13C-MFA, no significant differences in the flux distribution of the respective strains were observed for the growth conditions studied (Figure 45).

95 HIGH QUALITY 13C-MFA FOR EXAMINATION OF CYCLIC ED ACTIVITY

Figure 45: Flux distributions of the investigated Pseudomonas strains. The flux distributions of P. putida KT2440, P. putida S12 and P. taiwanensis VLB120 (top to bottom) were calculated from labeling data obtained in growth experiments on minimal media supplemented with 20:80 U-13C-glucose and 1-13C-glucose as substrate. All fluxes were normalized to a glucose uptake of 100 mmol/g/h. Numbers in squared brackets represent the corresponding confidence intervals.

For a total C6 uptake rate normalized to 100%, a PP pathway flux of approx. 23% was present. Compared to previously published flux distributions of P. putida KT2440, the value

96 CHAPTER 2.4 calculated here significantly exceeds the published values of approx. 10% [90,114]. The PP pathway fluxes determined are higher compared to the data Blank et al. stated in 2008. As a result of differences in the determined PP pathway fluxes, further key fluxes such as TCA cycle fluxes differ as well. With regard to the two strains P. putida KT2440 and P. putida S12, Blank et al. published differences between these strains as well. In our study, investigation of P. putida KT2440, P. taiwanensis VLB120 and P. putida S12 did not reveal significant differences concerning the flux distributions. Thus, a flux of approximately 55 % from pyr to oxaloacetate (oaa) was computed for the three investigated strains, which is comparable to the value for P. putida KT2440 determined by Blank et al. The published flux from pyr to oaa in P. putida S12 was 82 ± 5 %, which is higher than the flux calculated in this study. The differences between the published flux distributions and the distributions determined here can be explained by the growth behavior of the strains. Thus, an overall lower C6 uptake rate as well as growth rate was observed for the strains grown in shake flasks. As the strains in this study were grown in bioreactors, higher growth rates and carbon uptake rates could be observed. These different physiological data might explain the higher PP pathway and therefore lower TCA cycle fluxes. Another factor for the different flux distributions might be the chosen labeling strategy for determination of the cyclic ED pathway. A flux of f6p to g6p of about 25.5 % was estimated for each of the strains tested. This very pronounced flux is an indication that the organism, e.g., for the regulation of the cofactor balance strongly utilizes the upper part of the CCM. NADPH is formed during oxidation of g6p to 6pg and in the decarboxylation of 6pg to pentose-5-phosphate (p5p), which serves as redox cofactor in the cell. An explanation for this increased need for NADPH is the conversion of glucose to gluconate and 2-ketogluconate in the periplasm. In these reactions, electrons are released which pass directly into the electron transport chain. Furthermore, the transport of protons through the membrane is thereby promoted, whereby the formation of adenosine triphosphate (ATP) required for the metabolism is driven. To ensure that these reactions take place, an excess of NADPH is required to convert the formed 2-ketogluconate into 6pg, which is subsequently incorporated into the ED pathway [90,106,121]. Due to the labeling strategy used, the metabolic reactions in the periplasm could not be resolved. Therefore, no correlation between the determined cyclic ED pathway operation and the flux of 2-ketogluconate to 6pg could be made. Likewise, the fluxes from glucose to g6p as well as from g6p to 6pg are subject to a low resolution and therefore low confidence. This low resolution of the glucose uptake fluxes and therefore missing information can also be an explanation for the very equal fluxes of P. taiwanensis VLB120 compared to P. putida KT2440 and S12. Despite the missing gluconate dehydrogenase activity and the resulting differences in carbon uptake metabolism, no significant differences were found in the determined flux distribution for the investigated growth conditions. The use of labeling data evaluated from labeling experiments with various labeled substrates in combination with additional metabolite fragments and various growth conditions might lead to a better resolution of fluxes and should be further extended.

97 HIGH QUALITY 13C-MFA FOR EXAMINATION OF CYCLIC ED ACTIVITY

2.4.5. Conclusion For target-oriented planning of scientific experiments, in silico simulations performed prior to the experiments are useful. These simulations are used in order to minimize the experimental effort and therefore result in a more efficient and cost-effective workflow. The simulations carried out in this work using COBRA Toolbox and OpenFLUX revealed a suitable labeling strategy for the growth experiments. Therefore, a mixture of U-13C-glucose and 1-13C-glucose indicated a considerably better solubility of the cyclic ED pathway fluxes compared to mixtures of U-13C-glucose and 12C-glucose or 100% 1-13C-glucose. In order to minimize experimental cost, a 20:80 mixture of U-13C-glucose and 1-13C-glucose was used to perform growth experiments of the three strains P. putida KT2440, P. putida S12 and P. taiwanensis VLB120. The samples taken during these growth experiments were analyzed by GC-MS and LC-MS and yielded 35 different fragments which were sufficient for 13C-MFA. The 13C-MFA was carried out using the experimentally determined growth and labeling data. However, the flux analysis did not reveal different flux distributions among the examined strains despite their slightly different growth characteristics. The determined flux distributions showed the significant use of a cyclically operating ED pathway, but were different to previously published flux distributions of P. putida KT2440 and P. putida S12. The chosen growth conditions as well as the chosen labeling strategy might be the reason why some fluxes are subject to high errors. The labeling strategy which was chosen to focus on the evaluation of the cyclic ED pathway resulted in a good resolution of these cyclic ED pathway fluxes on one hand. On the other hand, this labeling strategy might explain the inadequate determination of further individual fluxes. Unfortunately, the labeling data of 6pg, a very central metabolite of the ED pathway, were not used here due to low quality, which means that important information was not available. Future work should therefore focus on the sufficient measurability of 6pg. Furthermore, it is useful to perform experiments with different 13C- tracers in parallel and to combine the labeling data of the respective experiments in one 13C- MFA. This strategy was shown to results in increased precision of the entire flux distribution in the CCM.

Chapter 2.5

Tandem MS measurements for the determination of mitochondrial and cytosolic acetyl-CoA labeling in Saccharomyces cerevisiae

Partially submitted as: Albert, I. (2014), Entwicklung von Triple-Quadrupol-Massenspektrometrie-Methoden für die 13C-Stoffflussanalyse, Master Thesis, RWTH Aachen University, Institute of Applied Microbiology

Author Contributions: Vincent Wiebach performed initial physiology experiments and simulations. Andreas Schmitz planed and designed the project. Isabella Albert performed the experiments and analyzed the results. The chapter was written by Andreas Schmitz with the help of Birgitta Ebert. Birgitta Ebert and Lars M. Blank supervised and conceived the study.

99 TANDEM MS MEASUREMENTS FOR THE DETERMINATION OF ACETYL-COA LABELING

2.5. Tandem MS measurements for the determination of mitochondrial and cytosolic acetyl- CoA labeling in Saccharomyces cerevisiae 2.5.1. Summary The technique of 13C-metabolic flux analysis (13C-MFA) is increasingly used in biotechnology to gain insight into the intracellular reaction rates of an organism. To this end, the organism is fed with 13C-labeled substrate and the labeling patterns in synthesized metabolites is determined. Classically, determination of 13C-labeling patterns of metabolites for 13C-MFA, is carried out by mass spectrometry (MS), and most often coupled to an upstream gas chromatographic (GC) separation. Nevertheless, fragments obtained from GC-MS analysis often are insufficient to explicitly determine all positional isotopomers (or: isotopologues), inherently limiting the resolvability of intracellular fluxes. As an alternative, GC coupled to tandem mass spectrometry (MS/MS) can be used, which introduces a second fragmentation of the analyte fragments generated in the initial ionization step and separated in the first quadrupole. Mass spectral analyses of these daughter ions yields detailed information about positional 13C-labeling. In this chapter, the potential of GC-MS/MS methods for 13C-MFA is exemplarily shown for elucidation of the acetyl-CoA metabolism in Saccharomyces cerevisiae. The second fragmentation of proteinogenic amino acids allowed an in-depth insight into the positional carbon labeling of these amino acids, which was further used to infer the labeling of cytosolic and mitochondrial acetyl-CoA. Specifically, the fragmentation of lysine (LYS434 > LYS243) and leucine (LEU302 > LEU200) was shown to be suitable for comparison of cytosolic and mitochondrial acetyl-CoA and revealed a higher C1-labeling of lysine. Thus no acetyl-CoA transport is active in this yeast, but rather acetyl-CoA synthesis via the pyruvate dehydrogenase is active as revealed in experiments with 13C-labeled glucose. 2.5.2. Introduction Quantification of intracellular carbon fluxes in microorganisms is increasingly required as it is a beneficial tool in systems biology and therefore finds application in biotechnological laboratories. Flux quantification provides basic knowledge about an organism’s intracellular behavior under certain conditions and can support and verify targeted genetic interventions in metabolic engineering [8]. A commonly used technique for carbon flux quantification is stable 13C-isotope based metabolic flux analysis (13C-MFA). In this approach, cells are fed with a substrate such as glucose labeled with 13C-isotopes at specific positions, which are incorporated into metabolites upon degradation of the carbon source. The resulting labeling patterns depend on the relative activity of the metabolic pathways. Integration of experimentally determined 13C- labeling data of proteinogenic amino acids or central carbon metabolites (Figure 4) into a metabolic model, together with extracellular uptake and secretion and the growth rate allows quantification of intracellular carbon fluxes. For these calculations, software, e.g., 13C- FLUX2 or OpenFLUX [6,12,14] is available, which derive the flux distribution by mathematically fitting the labeling data and combining these results with extracellular rates with a central carbon metabolism (CCM) model. The measurements for labeling pattern determination are most commonly performed with mass-spectrometry (MS) coupled to chromatographic techniques. Here, depending on the characteristics of the metabolite of interest, gas- (GC) or liquid chromatography (LC) is used for analyte separation [23,66,67]. Besides mass spectrometric methods, nuclear magnetic resonance (NMR) spectroscopy is used for determination of isotopic labeling patterns [63-65].

100 CHAPTER 2.5

In contrast to MS methods, NMR distinguishes the 13C-positions within the measured metabolite and therefore is a more suitable method for labeling pattern determination. Nevertheless, MS methods are more often used for 13C-MFA. Reasons for this are lower costs for technical equipment on the one hand and high versatility in terms of various analytical questions as well as robustness of MS on the other hand. The disadvantage of not directly receiving positional information in GC-MS measurements is compensated by measurement of analyte fragments containing different carbon atoms of the metabolite and usage of the labeling patterns of these fragments is usually sufficient for 13C-MFA. Yet, the lack of positional labeling information in MS based techniques can limit resolvability of metabolic fluxes and complicates statements about labeling of metabolites and their precursors. A promising technique combining the advantages of MS and positional labeling information is gas chromatography coupled to tandem mass spectrometry (GC-MS/MS) as shown by Jeffrey et al. in 2002 [55]. In classical GC-MS devices, the analytes are separated within the GC and ionized in the ion chamber of the MS. During this electron impact (EI) ionization the analytes are fragmented and separated within the quadrupole MS. The advantage of tandem mass spectrometry is the combination of three quadrupoles (Figure 8) of which the second functions as collision chamber. In the second quadrupole, the fragments generated by EI ionization (parent ions) are fragmented a second time by impact with a collision gas and new fragments (daughter ions) occur. The labeling of appropriate daughter ion yields a better information about the positional labeling of the parent ion. To exploit the information content, every mass isotopomer (isotopomer with the same number of labeled atoms) of the parent ion is fragmented and the mass isotopomer distribution of the daughter ion is measured [56]. After the mass isotopomer distributions for all required parent to daughter ion transitions are measured, the data can be processed into a mass distribution vector (MDV) for the respective daughter ion. This MDV can afterwards be used for either flux distribution calculation or direct comparison of fragments containing the same precursor atoms. Tandem MS-based 13C- MFA might have potential to better resolve the CCM fluxes in S. cerevisiae. This eukaryotic organism has different organelles, such as mitochondria, vacuoles or peroxisomes. Some metabolites occur in several compartments (Figure 46). One example is acetyl-CoA which can be synthesized either in the mitochondrion by pyruvate dehydrogenase (Pdh) from mitochondrial pyruvate or via the cytosolic acetyl CoA synthetase from cytosolic acetate which itself is part of the glucose fermentation pathway. Published 13C-MFA studies used a model that included a direct exchange of acetyl-CoA between these two compartments enabled by a carnitine-acetyltranferase-system (carnitine shuttle), which was assumed to always be active [63,122]. However, it has been shown that S. cerevisiae is unable to synthesize carnitine making this shuttle nonfunctional when the strain is in minimal media without carnitine supplementation. [123,124]. 13C-MFA might shed light on the actual intercompartmental transport of acetyl-CoA or other metabolites.

101 TANDEM MS MEASUREMENTS FOR THE DETERMINATION OF ACETYL-COA LABELING

Figure 46: Central carbon metabolism of S. cerevisiae with the two compartments cytosol and mitochondrion. The main pathways Emden-Meyerhoff-Parnas pathway, Pentose Phosphate pathway and tricarboxylic acid cycle are highlighted.

The aim of this study was to develop a GC–MS/MS method for the determination of the labeling of cytosolic and mitochondrial acetyl-CoA. Even though acetyl-CoA labeling can be measured directly with LC-MS, cytosolic and mitochondrial acetyl-CoA cannot be discriminated and hence the labeling of the different acetyl-CoA pools cannot be determined. In this work, the labeling of these two acetyl-CoA pools shall be investigated by fragmentation of different amino acids containing carbon atoms originating from either cytosolic or mitochondrial acetyl-CoA in their carbon backbone with tandem MS

102 CHAPTER 2.5 measurements. Therefore, various amino acid fragments had to be found for which the labeling of single backbone atoms is correlated to acetyl-CoA atoms. Promising amino acids such as leucine and lysine, containing acetyl-CoA atoms from the different compartments had to be compared for indirect comparison of acetyl-CoA labeling within the compartments. 2.5.3. Materials and Methods 2.5.3.1. Strains and Growth Conditions S. cerevisiae CEN.PK113-7D was used in all growth experiments and cultivated at 30 °C and 200 rpm in mineral salt medium [125] published by Verduyn et al. that contained (per liter) 20.42 g potassium hydrogen phthalate, 3.00 g KH2PO4, 5.00 g (NH4)2SO4, 0.5 g MgSO4 ∙ 7 H2O, and 3.6 g Glucose complemented with 0.15 g Na2EDTA, 0.045 g ZnSO4 · 7 H2O, 0.01 g MnCl2 · 2 H2O, 0.003 g CoCl2 · 6 H2O, 0.003 g CuSO4 · 5 H2O, 0.004 g Na2MoO4 · 2 H2O, 0.045 g CaCl2 ·2 H2O, 0.03 g FeSO4 · 7 H2O, 0.01 g H3BO3, 0.001 g KI and 0.0005 g biotin, 0.01 g calcium pantothenate, 0.01 g nicotonic acid, 0.25 g inositol, 0.01 g thiamine · HCl, 0.01 g pyridoxine · HCl, 0.002 g p-aminobenzoic acid [125,126]. 2.5.3.2. Amino Acid Standards Amino acid standards (Carl Roth GmbH + Co. KG, Karlsruhe, Germany) were prepared using 400 μL of an amino acid mix containing 10 mM of L-alanine, L-valine, L-leucine, L- isoleucine, L-threonine, L-aspartate, L-glutamate and L-lysine in water. The final amino acid concentration within the measurement solution was 6.7 mM, comparable to 0.3 mg dried biomass [127]. 2.5.3.3. Hydrolysis of Harvested Biomass Samples with defined amounts of biomass were taken from the culture and centrifuged (16.873 x g, 5 min). The supernatant was discarded and the cells were resuspended with 0.9 % (w/v) NaCl solution and centrifuged again. After discarding the supernatant, the cell pellet was resuspended in 150 μL of 6 M HCl and vortexed for 15 s. The suspension was transferred into a conical 1.5 mL glass vial. The vial was tightly closed and incubated in a heating block at 105 °C for hydrolysis of the biomass. After hydrolyzation, HCl was evaporated at 85 °C under a hood. The dried sample was stored in a closed vial at room temperature. 2.5.3.4. Analytics Optical Density Measurement

103 TANDEM MS MEASUREMENTS FOR THE DETERMINATION OF ACETYL-COA LABELING

Knauer, Berlin, Germany) equipped with an organic acid resin column (polystyrol- divinylbenzol copolymer, PS-DVB: 300 × 8.0 mm, CS-Chromatographie, Langerwehe, -1 Germany). 5 mM H2SO4 was used as eluent at a flow of 0.7 mL h at 50 °C. The analytes were detected by an ultraviolet light (UV) detector (UV detector LC-166, Beckmann Coulter, Krefeld, Germany) at a wavelength of 210 nm and a refractive index (RI) detector (LCD 201, Gynkotek, Munich, Germany) system. Extracellular Rates Extracellular rates such as growth rate and carbon uptake / production rates of glucose, acetate, ethanol and glycerol were determined with Microsoft Excel 2013. The growth rate was calculated as a linear function of the logarithm of OD600 over time. Uptake and production yields were calculated as the function of analyte and biomass concentration. Substrate uptake and product formation rates were calculated by multiplication of growth rate and corresponding yield. Derivatization Amino acid standards and hydrolyzed biomass samples were dried and derivatized prior measurement. The dried samples were resuspended in 30 μL acetonitrile and 30 μL derivatizing reagent. Derivatization was performed using either N-methyl-N-(tert- butyldimethylsilyl)-trifluoroacetamide (MBDSTFA, CS - Chromatographie Service GmbH, Langerwehe, Germany) or N-methyl-N-trimethylsilyl-trifluoroacetamide (MSTFA, CS - Chromatographie Service GmbH, Langerwehe, Germany). The glass vial was tightly closed, vortexed for 15 s and incubated for 1 h using a heating block at 85 °C. The samples were cooled down to room temperature and analyzed immediately. Gas Chromatography–Mass Spectrometry Gas chromatography separation was performed on a Thermo Scientific (Thermo Scientific, Waltham, MA, USA) Trace GC Ultra equipped with a Thermo Scientific AS 3000 autosampler. The column used was a Restek (Restek GmbH, Bad Homburg, Germany) Rxi- 5Sil MS capillary column (length, 15 m; inner diameter, 0.25 mm; film thickness, 0.25 μm). Separation of amino acids was performed at a constant flow rate of 1 mLhelium/min. A sample volume of 1 μL was injected into a split/splitless injector at 270 °C with a split ratio of 1:10. For MBDSTFA derivatized samples, the temperature of the GC oven was kept constant for 1 min at 140 ºC and was afterwards increased with a gradient of 10 ºC/min to 270 ºC and again kept constant for 1 min. The GC oven for MSTFA derivatized samples was kept constant for 1 min at 70 °C and was increased to 170 °C with a gradient of 10 °C/min. The temperature was then increased with a gradient of 20 °C/min to 270 °C and kept constant for 1 min. Mass spectrometry analysis was performed with a Thermo Scientific (Thermo Scientific, Waltham, MA, USA) TSQ triple quadrupole mass spectrometer for both, GC-MS and GC-MS/MS measurements. The temperatures of the transfer line and the ion source were both set to 280 ºC. Ionization was performed by electron impact (EI) ionization at 70 eV. For GC-MS/MS measurements, the collision cell was operated with argon gas. Using automatic selected reaction monitoring (autoSRM), collision energies within a range of 4 to 25 eV were found to be optimal depending on the parent – daughter fragment pair. GC-MS

104 CHAPTER 2.5 measurements were performed without operation of the collision cell and the three quadrupoles were used as one big mass filter. Data generated with GC-MS were recorded with full scan mode and GC-MS/MS data were recorded in selected ion monitoring (SIM) mode. 2.5.3.5. Data Evaluation The measured data were evaluated with Thermo Xcalibur 2.2 SP1.48 (Thermo Scientific, Waltham, MA, USA) applying the Processing Setup. Here the “ICIS” peak identification method was used and the parameters “Baseline window,” “Area noise factor,” and “Peak noise factor” were adjusted to peak specific optimal values. iMS2Flux was used for correction of labeled data for natural labeling and original biomass [84]. 2.5.4. Results and Discussion Metabolites involved in CCM of S. cerevisiae can be found in different compartments within the cell, with the cytoplasm and the mitochondrion being the compartments, in which the CCM reactions during growth on glucose predominantly take place. For 13C-based MFA it is necessary to specify the localization of metabolites, i.e., to define two distinct pools for metabolites in the two compartments. Since intercompartmental transport of these metabolites can play an essential role for metabolic engineering, a good knowledge of transport reactions is mandatory. Exemplarily, the knowledge of intercompartmental transport reactions of acetyl-CoA is essential for targeted manipulations in order to reduce acetate formation during growth of S. cerevisiae. The method development for GC-MS/MS analysis of selected amino acids involved in acetyl-CoA metabolism is shown in the following sections. By the use of this novel method, the reactions involving cytosolic and mitochondrial acetyl-CoA is examined. 2.5.4.1. Amino Acid Analysis The GC-MS method for the measurement of the 13C-labeling distribution of proteinogenic amino acids described in Chapter 2.2 was used for full scan measurements of standard mixtures of 8 amino acids derivatized with MBDSTFA (Figure 47).

105 TANDEM MS MEASUREMENTS FOR THE DETERMINATION OF ACETYL-COA LABELING

Figure 47: Chromatogram of amino acid standard mix derivatized with MBDSTFA. Intensity and retention time (RT) of a 6.7 mM mix scanned in mass to charge (m/z) ratios from 50-500 m/z. Alanine, RT 2.60 min; valine, RT 3.52 min; leucine, RT 3.87 min; isoleucine, RT 4.15 min; threonine, RT 6.74 min; aspartate, RT 7.92 min; glutamate, RT 8.91 min; lysine, RT 9.78 min. Figure taken from [127].

The amino acids leucine, lysine and aspartate contain carbon atoms originating from cytosolic and mitochondrial acetyl-CoA (Figure 4). These three amino acids are well separated from each other and the residual amino acids contained in the standard mix (Figure 47). This clear separation is mandatory, since overlapping fragments of different amino acids causes false results in 13C-MFA. The mass spectra of these amino acids (Figure 48) were extracted and further examined. The common mass peaks of MBDSTFA derivatized compounds were identified to be the masses 73 and 147 m/z. As these masses do not contain carbon atoms originating from the amino acids, they were neglected in the further work. The remaining masses with high intensity were further investigated. Using MBDSTFA for derivatization, the typical fragments formed are f302, M-15, M-57, M-85 and M-159. Here, M represents the mass of the molecular ion and f302 the fragment with a m/z of 302, which contains the carbon atoms C1 and C2 of the amino acid [24]. The masses of the fragments of leucine, lysine and aspartate are listed in Table 9.

106 CHAPTER 2.5

Table 9: Masses of characteristic fragments of MBDSTFA derivatized leucine, aspartate and lysine.

M-15 M-57 M-85 M-159 leucine 344 302 274 200 aspartate 460 418 390 316 lysine 473 431 403 329

Leucine fragment M-57 (302 m/z) cannot be used for 13C-MFA using GC-MS, because it has the same mass as fragment f302. Additionally, lysine fragments M-15 (473 m/z) and M-85 (403 m/z) are not usable due to low intensity and bad signal to noise ratio [26].

107 TANDEM MS MEASUREMENTS FOR THE DETERMINATION OF ACETYL-COA LABELING

c Figure 48: Mass spectral data of leucine (a), aspartate (b) and lysine (c) measured by GC-MS in full scan mode from 50-500 m/z [127].

For the development of a tandem mass spectrometry method, an autoSRM [128] was applied, in which mass fragments from full scan measurements were selected and fragmented using different fragmentation energies in the collision chamber. For identification of acetyl-CoA labeling, daughter ions containing a single backbone carbon atom originating from acetyl- CoA can be used. Likewise, fragments are sufficient, if one of the C atoms originating from acetyl-CoA was removed during the second fragmentation step. Feasible backbone atoms are C1 or C2 in case of leucine and lysine, and C1, C2 or C3 in case of aspartate. Application of autoSRM yielded 6 leucine, 6 aspartate, and 3 lysine ion pairs (Table 10).

108 CHAPTER 2.5

Table 10: Parent - daughter ion combinations identified by autoSRM analysis with the corresponding fragmentation energies, daughter ion intensity and postulated parent-to- daughter transition. Combinations which are greyed out are not fulfilling the requirements of high intensity and informative fragmentation and were therefore not used in further investigations [127]. parent daughter fragmentation intensity [ ] Parent-to- ion [m/z] ion [m/z] energy [eV] daughter transition 5 leucine 344 145 10 2∙10 ± 7 % [C1-C6] > C1 5 344 200 15 4.5∙10 ± 6.5 % [C1-C6] > [C2-C6] 7 302 200 15 1.1∙10 ± 3.5 % [C1-C6] > [C2-C6] 5 274 232 10 3.5∙10 ± 8.5 % [C2-C6] > [C2-C3] 7 274 218 10 4.2∙10 ± 4 % [C2-C6] > C2 6 200 158 10 1.2∙10 ± 1.5 % [C2-C6] > [C2-C3]

5 aspartate 460 286 10 3.6∙10 ± 12 % [C1-C4] > [C1-C4] 7 418 117 12 1.2∙10 ± 3 % [C1-C4] > [C3-C4] 6 418 103 20 2.2∙10 ± 11 % [C1-C4] > C1 6 390 346 12 7.1∙10 ± 5 % [C2-C4] > [C2-C3] 6 390 216 15 8.2∙10 ± 6 % [C2-C4] > C2 6 316 142 10 3∙10 ± 11 % [C1-C6] > C2

3 lysine 431 273 5, 10, 15, 20 < 10 [C1-C6] > [C2-C6] 4 431 245 5, 10, 15, 20 < 10 [C1-C6] > [C2-C4] 3 329 142 5, 10, 15, 20 < 10 [C2-C6] > C2

A threshold intensity of the daughter ion of 105 was set and ions below this threshold were excluded due to their low signal to noise ratio. This requirement was met for all leucine and aspartate daughter ions. Only the intensity of the lysine daughter ions was too low for the detected daughter ion and could therefore not be used for further application. In addition to sufficient intensity, the backbone carbon atoms of measured fragments need to carry information about acetyl-CoA labeling. This requirement was fulfilled for four leucine as well as four aspartate fragments (Table 10). However, no suitable lysine fragments were found with autoSRM focusing on standard fragments for this amino acid. Therefore, additional parent ions were searched within the mass spectral data from full scan measurements. Five promising fragments (170, 198, 272, 300, and 488 m/z) with adequate intensity were found and further tested. No unambiguous structure could be assigned to the fragments 198 m/z and 272 m/z, since various fragments were found carrying the same mass. Even though these fragments carry different backbone carbon atoms their mass is equal. Therefore, these fragments were not used for further investigation. The remaining three fragments were tested in tandem MS applying various fragmentation energies but no daughter ions with sufficient intensity as well as informative fragmentation were found (data not shown).

109 TANDEM MS MEASUREMENTS FOR THE DETERMINATION OF ACETYL-COA LABELING

2.5.4.1.1. MSTFA derivatization of Lysine In a second approach lysine was derivatized with the alternative derivatization agent MSTFA. In a first analysis, an amino acid standard mix was derivatized with MSTFA and measured using GC-MS in full scan mode (Figure 49).

Figure 49: Chromatogram of amino acid standard mix derivatized with MSTFA. Intensity and RT of a 6.7 mM mix scanned from 100-550 m/z. valine + TMS, RT 3.29 min; alanine + 2*TMS, RT 3.51 min; leucine + TMS, RT 4.08 min; isoleucine + TMS, RT 4.33 min; valine + 2*TMS, RT 4.88 min; leucine + 2*TMS, RT 5.61 min; isoleucine + 2*TMS, RT 5.88 min; threonine + 3*TMS, RT 7.12 min; aspartate + 2*TMS, RT 7.4 min; aspartate + 3*TMS, RT 8.73 min; glutamate + 3*TMS, RT 9.84 min; lysine + 3*TMS, RT 10.77 min; lysine + 4*TMS, RT 12.52 min [127].

In contrast to the amino acid derivatization with MBDSTFA, two differently silylated products were observed for amino acids derivatized with MSTFA for 1 h at 85 °C. Figure 50 exemplarily shows the two trimethylsilyl (TMS) derived lysine derivatives. The amino group can be derivatized with either one or two TMS groups, which is not observed when derivatization is executed with MBDSTFA. This can be explained by steric hindrance when lysine is derivatized with MBDSTFA, because the resulting tert-butyl-dimethylsilyl (TBDMS) group is bigger than TMS preventing the linkage of a second TBDMS group.

110 CHAPTER 2.5

Figure 50: Comparison of the TBDMS lysine derivative (a; derivatized with MBDSTFA) and as single (b) or double (c) silylated TMS derivative (derivatized with MSTFA) [127].

Using the MSTFA derivatized lysine with four TMS (lysine + 4 ∙ TMS) groups led to less symmetric isobaric fragments. Inspection of the mass spectral data of lysine + 4 ∙ TMS led to identification of two fragments (434 and 317 m/z; Figure 51) with high intensity, which were promising candidates as parent ions for tandem MS measurements. Application of autoSRM yielded three parent - daughter ion pairs which fulfilled the requirements of sufficient intensity above 105 and useful fragmentation (Table 11).

Figure 51: Mass spectral data of MSTFA derivatized lysine measured by GC-MS in full scan mode from 100-550 m/z [127].

111 TANDEM MS MEASUREMENTS FOR THE DETERMINATION OF ACETYL-COA LABELING

Table 11: Combination of parent- and daughter ions identified by autoSRM analysis with the corresponding fragmentation energies, daughter ion intensity and parent-to-daughter transition [127]. parent daughter fragmentation intensity [-] Parent-to-daughter ion [m/z] ion [m/z] energy [eV] transition 5 lysine 434 243 12 5∙10 ± 9 % [C1-C6] > [C2-C6] 6 434 230 12 1.5∙10 ± 9 % [C1-C6] > [C2-C6] 6 317 102 10 8.3∙10 ± 4 % [C2-C6] > C2

2.5.4.2. Validation of Amino Acid Fragmentation The postulated structure of fragments identified in the GC-MS/MS analyses (2.5.4.1) had to be validated for later use in 13C-MFA. This validation was performed by comparing theoretical fragmentation patterns with measured patterns of 13C-labeled analytes on the one hand, and naturally labeled analytes on the other hand. Labeled and unlabeled analytes were fragmented and compared to expectations. Only fragments passing both strategies can be used for 13C-MFA.

2.5.4.2.1. M+1 analysis In a mass spectrometric analysis, analytes are fragmented in the ion chamber. The various fragments generated in the ion chamber are subsequently separated in the quadrupole according to their mass. Each of these so-called mass isotopomers has the same structure but differs in the number of incorporated isotopes. Thus, the lightest mass isotopomer (M0) of a fragment results from the presence of light isotopes alone. The occurrence of a single heavy isotope leads to a mass increase of 1 and is denoted as M+1. A molecule carrying two heavy isotopes has a mass shift of +2 and is indicated as M+2. Depending on the number (n) of heavy isotopes incorporated in a fragment, a single fragment can comprise up to M+n+1 mass isotopomers. As an example, for fragment f302 of naturally labeled leucine, M0 has a mass of 302 m/z. Further mass isotopomers up to a mass of 305 m/z (M+3) can be identified with high intensities (Figure 52). Every of these mass isotopomers, except of M0, can be composed of several positional isotopomers, which have the same chemical structure and integer mass but differ in the position of the heavy isotope and the element (13C, 2H, 17O, 15N, 29Si, etc.).

112 CHAPTER 2.5

M+1

M+2

M+3

Figure 52: Mass spectral data of leucine measured by GC-MS in full scan mode from 50-500 m/z. M0 to M+3 mass isotopomers are exemplarily shown for leucine fragment f302 [127].

In the first validation test, the fragmentation of the mass isotopomer M+1 of a parent ion was investigated. As explained above, this mass isotopomer M+1 includes only a single heavy isotope which can be located at any position in the molecule. In case of M+1 fragmentation in the collision cell, mass isotopomers measured for daughter ions should be restricted to m0 and m+1 (Figure 53). The intensities of m0 and m+1 depend on the relative abundance the heavy isotope is located within the daughter ion.

113 TANDEM MS MEASUREMENTS FOR THE DETERMINATION OF ACETYL-COA LABELING

Figure 53: Parent to daughter ion transition of M0 (a) and M+1 (b) of leucine fragment f302 and corresponding daughter ion spectra measured by GC-MS/MS in scan mode from 199-205 m/z [127].

Nevertheless, when the daughter ion spectra were scanned, occasionally additional masses were detected. This can be explained by false postulated parent to daughter ion transitions on the one hand and by masses with slightly different masses overlapping in the spectral data on the other hand. Therefore, only parent to daughter ion transitions, where solely m0 and m+1 mass isotopomers were found for daughter ions of M+1 were used in the further experiments. The results of measurements of daughter ions originating from M+1 parent ions for the three analyzed amino acids are shown in Table 12.

Table 12: Determination of tandem MS transitions for parent ion (M+1) to daughter ion. Measured transitions that are greyed out do not fit the expected transitions. Data taken from [127].

M+1 M0 ion pair expected transition measured transition leucine 344 > 145 345 > 146 345 > 147 344 > 200 345 > 201 345 > 201 302 >200 303 > 201 303 > 201 274 > 218 275 > 219 275 > 219 aspartate 418 > 117 419 > 118 419 > 117 418 > 103 419 > 104 419 > 103 390 > 216 391 > 217 391 > 217 316 > 142 317 > 143 317 > 142 lysine 434 > 243 435 > 244 435 > 244 434 > 230 435 > 231 435 > 231 317 > 102 318 > 103 318 > 102

114 CHAPTER 2.5

6 out of 11 transitions were verified, while one leucine, three aspartate and one lysine transition did not pass the M+1 test. Leucine transition 345 > 146 showed a 1 u higher daughter ion isotopomer than expected. This 1 u higher daughter ion isotopomer represents an m+2 isotopomer with two heavy isotopes, which cannot occur if a parent ion M+1 is fragmented. An explanation for this can be an overlay with the mass 147 m/z which is typically present in mass spectra of MBDSTFA derivatized samples. This ion-pair was excluded from further analysis. The remaining ion-pairs of aspartate and lysine showed a 1 u lower daughter ion isotopomer than expected. 2.5.4.2.2. Analysis of 13C-labeled biomass Validation of tandem MS fragmentations was further pursued by analyzing uniformly 13C- labeled biomass. S. cerevisiae was fed with U-13C-glucose and hydrolyzed afterwards. By feeding cells with labeled glucose, the resulting amino acids become also fully labeled since 13C-atoms from glucose are incorporated in these metabolites. As a result, the masses of amino acid fragments will change according to the number of backbone carbons. Figure 54 shows the mass spectral data of fragment f302 of naturally labelled leucine and shifts in mass spectra of measurements of the same fragment of uniformly 13C-labeled leucine.

Figure 54: Comparison of naturally labeled leucine (a) and 13C-labeled leucine (b) fragment f302 with the correlating mass spectra.

115 TANDEM MS MEASUREMENTS FOR THE DETERMINATION OF ACETYL-COA LABELING

This shift in the mass spectrum can be used to verify the structure of daughter ions, whose masses should equally shift by the amount of backbone carbon atoms. Evaluation of tandem MS measurements of amino acids from 13C-labeled biomass was applied for validating the tandem MS method. Here, two out of ten ion pairs did not fit the expectations, whereby aspartate 390 > 216 daughter ion was found to be 1 u heavier and lysine 434 > 230 daughter ion was 1 u lighter than expected (Table 13). These faulty predicted tandem MS transitions were excluded from further experiments.

Table 13: Determination of tandem MS transition for 13C-labeled parent ion to daughter ion. Measured transitions colored in grey do not fit the expected transitions. Data taken from [127]. 13C-ion pair M0 ion pair expected transition measured transition leucine 344 > 200 350 > 205 350 > 205 302 >200 308 > 205 308 > 205 274 > 218 279 > 219 279 > 219 aspartate 418 > 117 422 > 119 422 > 119 418 > 103 422 > 104 422 > 104 390 > 216 393 > 217 393 > 218 316 > 142 319 > 143 319 > 143 lysine 434 > 243 440 > 248 440 > 248 434 > 230 440 > 235 440 > 234 317 > 102 322 > 103 322 > 103

2.5.4.3. 13C-Labeling Experiment The method developed in Chapters 2.5.4.1 and 2.5.4.2 was used for the determination of acetyl-CoA labeling in S. cerevisiae. S. cerevisiae CEN.PK113-7D was grown with 20 mmol/L of an 80/20 (n/n) mixture of 1-13C-glucose and U-13C-glucose in a 500-mL shake flask. The growth curve and physiological data are shown in Figure 55. Glucose was completely consumed after 10 hours and acetate, ethanol (not shown) as well as glycerol (not shown) were produced. Ethanol and glycerol were taken up after glucose depletion whereas acetate was produced constantly within the recorded time range.

116 CHAPTER 2.5

25 2.5 OD glucose acetate

20 2.0 pia est [-] density optical

15 1.5

10 1.0 concentration [m m ol/L]

5 0.5

0 0.0 024681012 tim e [h]

Figure 55: Physiological data of S. cerevisiae grown on an 80/20 (n/n) mixture of 1-13C- glucose and U-13C-glucose. Samples used for tandem MS measurements were harvested in mid exponential phase [127].

For comparison of cytosolic and mitochondrial acetyl-CoA labeling, the tandem MS transitions of fragments had to be used which were carrying the information of equal carbon backbone atoms of acetyl-CoA. Therefore, leucine 302 > 200 (Table 14) and lysine 434 > 243 (Table 15) were chosen which both show the transition from [C1-C6] of the parent ion to [C2- C6] of the daughter ion. Using this data, the labeling of C1 can be determined, which represents the C1 of mitochondrial acetyl-CoA in case of leucine and cytosolic acetyl-CoA in case of lysine.

117 TANDEM MS MEASUREMENTS FOR THE DETERMINATION OF ACETYL-COA LABELING

Table 14: Measured tandem MS transitions of leucine 302 > 200 gained from a 13C-labeling experiment using an 80/20 (n/n) mixture of 1-13C-glucose and U-13C-glucose. Abundances were corrected for natural labeling and original biomass using iMS2Flux. Data taken from [127]. M > m relative abundance M0 302 > 200 1.000 302 > 201 0.000 302 > 202 0.000 302 > 203 0.000 302 > 204 0.000 302 > 205 0.000 M1 303 > 200 0.050 303 > 201 0.950 303 > 202 0.000 303 > 203 0.000 303 > 204 0.000 303 > 205 0.000 M2 304 > 200 0.000 304 > 201 0.165 304 > 202 0.835 304 > 203 0.000 304 > 204 0.000 304 > 205 0.000 M3 305 > 200 0.000 305 > 201 0.000 305 > 202 0.316 305 > 203 0.684 305 > 204 0.000 305 > 205 0.000

118 CHAPTER 2.5

Table 15: Measured tandem MS transitions of lysine 434 > 243 gained from a 13C-labeling experiment using an 80/20 (n/n) mixture of 1-13C-glucose and U-13C-glucose. Abundances were corrected by natural labeling and original biomass using iMS2Flux [127]. M > m abundance M0 434 > 243 1.000 434 > 244 0.000 434 > 245 0.000 434 > 246 0.000 434 > 247 0.000 434 > 248 0.000 M1 435 > 243 0.078 435 > 244 0.922 435 > 245 0.000 435 > 246 0.000 435 > 247 0.000 435 > 248 0.000 M2 436 > 243 0.000 436 > 244 0.262 436 > 245 0.738 436 > 246 0.000 436 > 247 0.000 436 > 248 0.000 M3 437 > 243 0.000 437 > 244 0.000 437 > 245 0.393 437 > 246 0.607 437 > 247 0.000 437 > 248 0.000

The data, corrected for natural abundance and original biomass, showed a difference in C1 labeling in the compared fragments. A slight difference was observed for the labeling pattern for the fragment derived from the parent ion M+1 of leucine and lysine. 5 % and 7.9 % of the 13 C1 of the M+1 mass isotopomer of leucine and lysine respectively, were C-isotopes. Larger 13 differences were observed for the fragment of the parent ion M2. In this case, the C-fraction of the C1 atom of this mass isotopomer was 16.5 % for leucine and 26.2 % for lysine. The C1 labeling in the M3 mass isotopomer was 31.6 % for leucine and 39.3 % for lysine, respectively.

For the given labeling strategy and growth conditions, the labeling of lysine C1 was constantly higher. This carbon atom in lysine originates from the C1 of cytosolic acetyl-CoA while the C1 of leucine stems from the C1 of mitochondrial acetyl-CoA. These differences in labeling of lysine and leucine can only be explained by differences in labeling of the two acetyl-CoA and therefore the pyruvate pools. However, cytosolic pyruvate is a direct precursor of mitochondrial pyruvate. Differences between the two pyruvate pools can therefore only occur if malic acid enzyme activity results in the conversion of malate to pyruvate. This effect of different pyruvate labeling was verified with computational simulations (data not shown). In these simulations using the COBRA toolbox and OpenFLUX, a significant influence of malic acid enzyme on the labeling of mitochondrial pyruvate could be shown [120,129]. Besides

119 TANDEM MS MEASUREMENTS FOR THE DETERMINATION OF ACETYL-COA LABELING changes in pyruvate labeling due to malic acid enzyme activity, differences in acetyl-CoA labeling in the two metabolite pools indicate that a direct exchange of acetyl-CoA is not active or of a low activity that does not allow equilibration of the two pools [123]. Hence, there is no evidence for the presence of an active carnitine shuttle for acetyl-CoA transfer between cytosol and mitochondria. 2.5.5. Conclusion As any eukaryotic organism, S. cerevisiae consists of different compartments such as nucleus, membranes, cytoplasm and mitochondria. CCM of S. cerevisiae is taking place in cytoplasm and mitochondria and several metabolites are present in both compartments. One example is acetyl-CoA, a key metabolite of the CCM, participating in reactions present in the cytosol and almost all organelles including the mitochondria. Applying tandem MS measurements of selected amino acids of S. cerevisiae, we were able to partially resolve the labeling of cytosolic and mitochondrial acetyl-CoA. These measurements disclosed differences in carbon labeling of the C1 atom of acetyl-CoA. An equilibration of the cytosolic and mitochondrial acetyl-CoA pools in minimal media could be refuted, supporting the hypothesis that carnitine cannot be synthesized by this yeast. Additionally, it could be shown, that mitochondrial malic enzyme activity takes place. This malic enzyme activity is the reason for different labeling of cytosolic and mitochondrial acetyl-CoA since mitochondrial pyruvate is regenerated and further decarboxylated to mitochondrial acetyl-CoA.

120

Chapter 3

General Discussion

121 GENERAL DISCUSSION

3. GENERAL DISCUSSION 3.1. Possible pitfalls of 13C-metabolic flux analysis Essential for the successful application of 13C-metabolic flux analysis (13C-MFA) is a sufficient quality of measurement data. Various critical steps influencing the quality of labeling data can be observed when performing 13C-labeling experiments (Chapter 2.1). Since the substrates used for labeling experiments have a direct impact on the information content of labeling data, the substrate composition has to be well known for sufficient 13C-MFA quality. This can be achieved by the use of minimal media supplemented with trace elements and vitamins and the addition of a defined carbon source. Besides the substrate used, the quality of the labeling data used for 13C-MFA may also be influenced by the determination of 13C-labeling itself [26]. These metabolites are extracted and subsequently measured for labeling determination. Methods for metabolite measurement have already been extensively investigated and quality of labeling data was improved [15,24,26]. Exemplarily, the errors of measured amino acid mass distribution vectors (MDVs) for 13C-MFA have been improved to less than 0.4 mol % [26]. Nevertheless, 13C-MFA is further on error prone, depending on the various kinds of data used for flux analysis. Besides labeling data, physiological data are used for calculation of metabolic fluxes as well. These physiological data, i.e., specific growth, substrate uptake and production rates, are determined from time profiles of cell, substrate and secreted metabolite concentrations in the medium. Labeling data are typically associated with small measurement errors compared to physiological data used for 13C-MFA. Therefore, erroneous fluxes calculated in 13C-MFA may be explained by the used physiological data. One source of inaccuracy during determination of physiological data is the change of concentration of extracellular metabolites and biomass due to evaporation of medium or loss of volatile metabolites. This effect leads to over- and underestimation of metabolite concentrations. Faulty concentrations of biomass and analytes influence the calculated rates used as input for the computational analysis and consequently affect metabolic flux estimation. The focus in literature is usually placed on minimizing the error of measured MDVs, assuming that determination of physiological data is trivial [17,26]. Even though the quality of labeling data is fundamental for precise flux quantification, the quality of physiological data needs to be sufficient as well. Therefore, future studies should additionally focus on precise and reproducible determination of cell and analyte concentrations. This can be achieved by parallel experiments while measuring concentrations over time and determining the influence of medium vaporization on concentrations of analytes and cells. Results in computational 13C-MFA can still vary due to the used metabolic model even though measured rates and extracellular concentrations are very precise. The metabolic models mostly rest on central carbon metabolism (CCM) models combined with information on biomass composition of the analyzed organism. Both, the metabolic model and biomass composition are based on biochemical knowledge, with sometimes additional evidence [68]. Due to the experimental background, the biomass composition of the investigated organism is prone to statistical deviations. This can be explained since these data just represent the results for the conditions given by biochemical experiments. The growth conditions can vary within the different metabolic flux experiments, wherefore metabolic fluxes as well biomass composition fluxes can also vary depending on the experimental settings. Differences in growth media such as the use of complex or minimal media affect the metabolic fluxes as

122 CHAPTER 3 well as the present pH value, oxygen availibility or carbon source concentrations. The established metabolic model for a selected organism must therefore be suitable for the used conditions 13C-MFA and cover all reactions of CCM which take place during the growth conditions. An insufficient metabolic model or biomass composition may affect the calculation of intracellular fluxes if changes in the experimental setup for MFA take place and the model does not cover these conditions [130]. To avoid this in future studies, the biomass composition should be determined for the experimental setup prior to 13C-MFA and combined with the used metabolic model.

3.2. Advantages and disadvantages of various labeling data for 13C-metabolic flux analysis A metabolic model representing the metabolism of the investigated organism, physiological data, and labeling data are utilized for flux calculation by 13C-MFA. Besides high data quality, the type of labeling data used for computational analysis is of high impact. Considering the metabolites used for 13C-MFA it can be distinguished between two groups of metabolites. Proteinogenic amino acids are the first group representing analytes originating from precursors out of CCM (Chapter 1.3.1). The amino acids are incorporated into proteins and accumulate in conjunction with cell growth over time. Consequently, proteinogenic amino acids are easy to extract in an amount sufficient for labeling determination. Even though proteinogenic amino acids are often used for 13C-MFA due to their straightforward sample preparation, they only provide limited information for flux quantification. Reasons for the restricted information content of amino acids lie within the origin of their carbon backbone atoms. The carbon atoms utilized for amino acid biosynthesis originate from central carbon metabolites, whereby carbon atoms of different metabolites can be used for formation of a single amino acid. Measuring amino acids with gas chromatography-mass spectrometry (GC- MS), leads to fragmentation of the molecules within the ion chamber. In classical mass spectrometric analysis of amino acids, positional isotopomers cannot be distinguished wherefore labeling information of selected carbon atoms have to be determined by comparison of detected mass isotopomers [17,18,62]. However, only a limited number of fragments suitable for MFA are gained within the ion source for a given derivatization strategy [23,24]. On one hand, this can be explained by a high measurement background at a mass range of fragments with low molecular mass. On the other hand, various fragments occurring during ionization of amino acids carry the same molecular mass and cannot be assigned unambiguously. Because of the finite number of precursor carbon atoms in combination with the loss of labeling information during mass spectrometric analysis, the obtainable information sufficient for 13C-MFA cannot be enhanced by GC-MS analysis of amino acids. For overcoming the problems concerning 13C-MFA using proteinogenic amino acids, various cellular components were investigated with respect to their applicability in 13C-MFA. In this context intracellular metabolites from CCM as well as components from deoxyribonucleic acid (DNA), ribonucleic acid (RNA) or polysaccharides are worth mentioning. Labeling of nucleosides isolated from DNA or RNA have successfully been measured and applied to 13C- MFA by Miranda-Santos et al. and Long et al. [131,132]. The glucose moiety of polysaccharides like glycogen could also be measured by GC-MS and applied to 13C-MFA by Christensen et al., Guzman et al. and Long et al. [66,131,133]. The group of analytes investigated in this work is represented by intracellular metabolites from CCM (Chapter

123 GENERAL DISCUSSION

1.3.2) [28,30]. Prior to the analysis of intracellular metabolites, the metabolism of the organism has to be stopped rapidly due to the high turnover rates of these metabolites. A sufficient sampling velocity with subsequent inactivation of the metabolism can be achieved by fast sampling methods combined with fast filtration or quenching. Due to fast changes in metabolic flux behavior and therefore changes in metabolite labeling, these methods are very error prone. A broad variation of methods have extensively been investigated in recent years and were applied to a wide variety of organisms [27,32,134,135]. However, a simple method transfer between different organisms is most often problematical, due to the fact that the cell membranes of the various organisms can react differently to the stress during sample preparation by quenching or filtration, whereby metabolites can be transported out of the cells [37,118]. This must be avoided, since important labeled metabolites might be lost before the metabolites can be extracted from the cells. For highly concentrated intracellular metabolites the phenomenon of metabolite leakage might not cause any problems in labeling determination. For low concentrated metabolites as well as for the isotopically non-stationary 13C-MFA, a high reliability on sample preparation without loss of any metabolites must be ensured [136]. If not, the comparability and reproducibility of the experiments is influenced as the measurement accuracy decreases at low concentrations. Exemplarily, labeling information of individual metabolites is erroneous or not obtained at all due to insufficient intensity, moreover the background has a strong impact on the fluxes calculated by 13C-MFA. In case of the isotopically non-stationary 13C-MFA, the pool size of the metabolites is also considered for solving the time dependent resolution of the fluxes [137-139]. Fluctuating metabolite concentrations can thus have a negative influence on the evaluation of these data, even if the concentrations are sufficient. In addition to concentration changes due to cell leakage, the extraction process itself can have an effect on the metabolites as well as the solvents used for quenching, and sample preparation methods for GC-MS measurement. Phosphorylated metabolites present in the pentose phosphate pathway, glycolysis and Entner- Doudoroff (ED) pathway have to be named in this context as these substances are usually unstable which can lead to a loss of the phosphate group. Exemplarily, experiments with standard substances revealed the loss of the phosphate group from 6-phosphogluconate (6pg) and resulted in formation of gluconate (data not shown). For labeling experiments with subsequent measurement of both intracellular gluconate and 6pg, this phosphate group cleavage prior to the GC analysis can cause blending of the gluconate and 6pg metabolite pools. If the two pools of gluconate and 6pg are labeled differently, the mixture of both pools after cleavage of the phosphate group results in faulty information supplied to 13C-MFA. Selection of solvents for the respective sample preparation must therefore be considered with care as the solvent has to be suitable for the investigated organism as well as for the metabolites to be measured. For quenching of different Pseudomonas strains, an adapted version of a cold ethanol saline solution published by Spura et al. in 2009 was applied in this work (cf. Chapter 2.3 and 2.4) [32]. For sufficient labeling determination in intracellular metabolites a high sample volume is required. Therefore, the sample amount was increased at low biomass concentrations compared to the volume of quenching solution. To avoid a too high temperature in the final mixture, a quenching solution with a lower temperature was needed compared to the -20 °C Spura et al. published. Since the quenching solution was stored at – 52 °C, the amount of ethanol had to be adapted, in order to prevent freezing of the quenching solution at this temperature. It could be demonstrated, that methods for individual

124 CHAPTER 3 organisms need to be optimized for application on further organisms. Nevertheless, the content of information using the labeling information of additional intracellular metabolites can be considerable enhanced, if the challenges in sample preparation and analysis are avoided.

3.3. New dimension of 13C-metabolic flux analysis by use of GC-tandem MS The solvability of individual metabolic fluxes in 13C-MFA depends to a large extent on the information obtained during the measurement of metabolites. This information content can base on the measured metabolites as well as on the technique used for the measurement. The frequently used measurement of proteinogenic amino acids by means of GC-MS generally provides sufficient information for investigating the CCM of microorganisms [15,18,23,62]. Thus, it is usually possible to quantify individual metabolic pathways within the CCM during growth on glucose [111,140,141]. For this reason, the GC-MS technology has been intensively used in the past and can still be named as a standard tool in biotechnology. However, the technique is often limited if it is necessary to examine specific questions concerning the CCM [58,142]. Some fluxes that are part of the tricarboxylic acid (TCA) cycle can exemplarily not be fully quantified with the information obtained by the GC-MS analysis of amino acids. Due to the insufficiency of information available for calculation of these fluxes, the results are error prone. The reason for this insufficiency of information is the selection of tested metabolites on one hand, and the computational data analysis of the measured labeling data on the other hand. The analysis of the intracellular metabolites can provide more information than the measurement of proteinogenic amino acids since only selected atoms of the precursor molecules are found in the amino acids. During amino acid synthesis, important labeling information is lost which cannot be obtained during the subsequent analysis of the amino acids. Furthermore, in case of measurements by means of GC-MS, the results do not provide information on the labeling of individual atoms but only on the labeling of a complete fragment. In this case, labeling of individual atoms can only be determined when the labeling of various GC-MS fragments is used for solving an established equation system. This equation system is used for mathematically comparing the different labeling information. Nevertheless, this method cannot always be used for all atoms present in amino acids, wherefore important information may not be generated for the later flux analysis. In the worst case, individual fluxes cannot be determined adequately. The GC coupled to tandem mass spectrometry (MS/MS) technique used in this work offers the advantage of combining the high sensitivity of mass spectrometric analysis with a further dimension of fragmentation, which can lead to higher information content (Chapter 1.4 & 2.5). For this purpose, the measurement of the individual fragment transitions from mother to daughter ions is used. A fragment containing a known number of labeled atoms is fragmented a second time by MS/MS and the labeling of the resulting ions is determined. The labeling of measured daughter ions can be used for determination of the atom labeling abundance within the carbon backbone of the mother ions (Chapter 1.4.1).

125 GENERAL DISCUSSION

First, this technique offers the possibility of separating distinct fragments of the same mass by a second fragmentation [56]. Specific fragmentations in the collision cell are used to generate a daughter ion with the desired carbon backbones. As a result, a unique separation of the fragments can be ensured. The fragmentation of every single parent ion isotopomer, and subsequent summation of the respective daughter ions leads to the MDV of the fragment of interest. By doing so, the information content of individual selected fragments can be determined which is not possible by single quadrupole MS measurements since this technique does not allow a sufficient labeling determination. The additional labeling data of selected fragments can help in these cases to solve the equation system and can therefore provide a complete evaluation of the positional labeling. Moreover, a possibility of using GC-MS/MS data for 13C-MFA was shown by Choi et al. in 2011, based on the work of Jeffrey et al. in 2002. The direct use of the above-mentioned fragment transitions for the 13C-MFA was shown, which gave a 2- to 5-fold higher accuracy for the calculation of the flows [55,56]. In addition, this method was used to calculate fluxes, which could not be calculated by the use of single quadrupole MS data. The higher accuracy compared to single quadrupole MS measurements can be explained by the fact that only a small number of data is used for the calculation when using data from these single quadrupole MS measurements. Small inaccuracies in the data applied for the calculation lead in this case to an inaccuracy in the determined fluxes. Using fragment transitions from tandem MS measurements for 13C-MFA results in a higher amount of data for the calculation. Therefore, determination of the position-specific labeling is much less error-prone compared to full-scan data due to the high number of information. This lower susceptibility to errors also leads to a higher accuracy in the calculated fluxes in 13C-MFA. Nevertheless, this type of analysis can also cause problems. It should be mentioned that the computer programs used for the 13C-MFA must be able to process this kind of measurement data for calculating metabolic fluxes [143]. By default, this is not provided for the commonly used software programs, wherefore this possibility must be created through programming beforehand. In addition to the computer-based problems concerning the processing of measured data, further difficulties can arise due to the analysis itself, depending on the molecules analyzed. For complete tandem MS analysis and subsequent use for flux analysis, various fragments have to be analyzed, wherefore many scans are required in a very short time, depending on the molecule. Since an analyte separated in gas chromatography will reach the detector only for a few seconds, this limitation in detection time can lead to an incomplete analysis of the molecule. This problem is especially observed in big molecules in which several fragments generated with varying fragmentation energies are analyzed. Therefore, the sample must usually be analyzed several times by means of GC-MS/MS in order to obtain a complete analysis of the analyte. Depending on the measuring method used, these additional measurements can present a great disadvantage in terms of time and cost. Compared to the nuclear magnetic resonance (NMR) analysis used for 13C-MFA, the use of tandem MS data seems nevertheless to be a promising alternative. The higher sensitivity of mass spectrometric analysis can provide a significant advantage in the further GC-MS/MS analysis of intracellular metabolites. As already described, these are predominantly present in low concentrations, which makes an NMR analysis more difficult. Such problems can be circumvented by the use of tandem MS measurements, especially as previous investigations of the two techniques showed largely comparable results in the analysis of amino acids and the subsequent MFA [55].

126

Chapter 4

Conclusion

127 CONCLUSION

4. CONCLUSION The resolution of 13C-MFA depends on the data available to constrain the models, besides the metabolic network structure. In this thesis, methods to evaluate and improve the quality of 13C-MFA were investigated. The standard 13C-MFA using the information from proteinogenic amino acids was revisited regarding data quality, by evaluating the individual steps from sample preparation to GC-MS measurement. By optimizing the different steps, from the amount of biomass to the parameters during GC-MS analysis measurement, their contribution to high quality labeling determination was evaluated. The developed protocol offers the possibility to optimize the individual steps of sample preparation and labeling measurement. Optimization might be required when novel organisms are investigated, in which the amino acid composition differs significantly or interfering compounds exist that interfere with the standard protocol. Circular pathways are difficult to resolve and might often require information from a particular analyte or even analyte fragment. Here, the different alternative pathways for glucose utilization in Pseudomonas putida, the ED and PP pathway, were investigated to prove or disprove a potentially cyclic operation. To gather the additionally required information a method for the determination of labeling patterns of intracellular metabolites was investigated. In biomass, rapidly quenched with an ethanol:saline solution at -50°C, intracellular metabolites could be detected with abundances that allowed 13C-labeling determination. With the optimized sampling protocol in hand the glucose metabolism of P. putida was investigated revealing that glucose or gluconate might be specifically channeled into the ED or the PP pathway. Combined with simulations for determining the most suitable labeling strategy to resolve the cyclic ED pathway, growth experiments with different Pseudomonas strains were carried out. The simulations suggested to extend the analyses with further intracellular metabolites, for which, in addition to GC-MS, LC-MS analyses were applied. The 13C-MFA performed with this extended dataset indicated that the cyclic ED pathway is functioning in a cyclic manner in order to produce e.g., NADPH, which is subsequently used in biomass formation and mechanisms against reactive oxygen species; possibly partially explaining the high stress tolerance reported in the literature. In contrast to extension of the analyte dataset used for 13C-MFA with intracellular metabolites, integration of positional 13C-labeling information in the proteinogenic amino acids was used here to improve flux identifiability. To this end, the applicability of GC- MS/MS analysis was evaluated. Compared to standard GC-MS analysis of proteinogenic amino acids, a higher information content could be obtained. The additional information available was used to investigate the acetyl-CoA metabolism in S. cerevisiae. The specific analysis of the C1 carbon atom of leucine and lysine revealed differences in the labeling of the cytosolic and mitochondrial acetyl-CoA pool, verifying the hypothesis that acetyl-CoA is not transported between the two compartments in the used CEN.PK yeast strain. Having shown the potential of GC-MS/MS analyses for 13C-MFA, this technique should be further developed and fully integrated into 13C-MFA software. The estimation of the quality of data, combined with an extended analyte range and analytical possibilities, will improve the scope and quality of 13C-MFA and when combined in workflows that are easily acquired by interested users, this analytical technique will find broader use and contribute to the understanding of the operation of metabolic networks.

128 CHAPTER 5

5. REFERENCES 1. Yadav, V.G., De Mey, M., Giaw Lim, C., Kumaran Ajikumar, P., Stephanopoulos, G. (2012), The future of metabolic engineering and synthetic biology: towards a systematic practice. Metabolic Engineering, 14, 233-241. 2. Lee, J.W., Na, D., Park, J.M., Lee, J., Choi, S., Lee, S.Y. (2012), Systems metabolic engineering of microorganisms for natural and non-natural chemicals. Nature Chemical Biology, 8, 536-546. 3. Park, J.H., Lee, S.Y. (2008), Towards systems metabolic engineering of microorganisms for amino acid production. Current Opinion in Biotechnology, 19, 454-460. 4. Wiechert, W. (2001), 13C metabolic flux analysis. Metabolic Engineering, 3, 195-206. 5. Blank, L.M., Kuepfer, L. (2010), Metabolic flux distributions: genetic information, computational predictions, and experimental validation. Applied Microbiology and Biotechnology, 86, 1243-1255. 6. Wiechert, W., Mollney, M., Petersen, S., de Graaf, A.A. (2001), A universal framework for 13C metabolic flux analysis. Metabolic Engineering, 3, 265-283. 7. Becker, J., Klopprogge, C., Zelder, O., Heinzle, E., Wittmann, C. (2005), Amplified expression of fructose 1,6-bisphosphatase in Corynebacterium glutamicum increases in vivo flux through the pentose phosphate pathway and lysine production on different carbon sources. Applied and Environmental Microbiology, 71, 8587-8596. 8. Sauer, U. (2006), Metabolic networks in motion: 13C-based flux analysis. Molecular Systems Biology, 2, 62. 9. Jorda, J., Jouhten, P., Camara, E., Maaheimo, H., Albiol, J., Ferrer, P. (2012), Metabolic flux profiling of recombinant protein secreting Pichia pastoris growing on glucose:methanol mixtures. Microbial Cell Factories, 11, 57. 10. Hua, Q., Joyce, A.R., Palsson, B.O., Fong, S.S. (2007), Metabolic characterization of Escherichia coli strains adapted to growth on lactate. Applied and Environmental Microbiology, 73, 4639-4647. 11. Metallo, C.M., Walther, J.L., Stephanopoulos, G. (2009), Evaluation of 13C isotopic tracers for metabolic flux analysis in mammalian cells. Journal of Biotechnology, 144, 167-174. 12. Quek, L.-E., Wittmann, C., Nielsen, L.K., Krömer, J.O. (2009), OpenFLUX: efficient modelling software for 13C-based metabolic flux analysis. Microbial Cell Factories, 8, 25. 13. Weitzel, M., Noh, K., Dalman, T., Niedenfuhr, S., Stute, B., Wiechert, W. (2013), 13CFLUX2 - high-performance software suite for 13C-metabolic flux analysis. Bioinformatics, 29, 143-145. 14. Zamboni, N., Fischer, E., Sauer, U. (2005), FiatFlux -a software for metabolic flux analysis from 13C-glucose experiments. BMC Bioinformatics, 6, 209. 15. Zamboni, N., Fendt, S.M., Ruhl, M., Sauer, U. (2009), 13C-based metabolic flux analysis. Nature Protocols, 4, 878-892. 16. Szyperski, T. (1998), 13C-NMR, MS and metabolic flux balancing in biotechnology research. Quarterly Reviews of Biophysics, 31, 41-106. 17. Christensen, B., Nielsen, J. (1999), Isotopomer analysis using GC-MS. Metabolic Engineering, 1, 282-290. 18. Wittmann, C. (2002), Metabolic flux analysis using mass spectrometry. Advances in Biochemical Engineering/Biotechnology, 74, 39-64. 19. Emmerling, M., Dauner, M., Ponti, A., Fiaux, J., Hochuli, M., Szyperski, T., Wuthrich, K., Bailey, J.E., Sauer, U. (2002), Metabolic flux responses to pyruvate kinase knockout in Escherichia coli. Journal of Bacteriology, 184, 152-164.

129 REFERENCES

20. Sauer, U., Lasko, D.R., Fiaux, J., Hochuli, M., Glaser, R., Szyperski, T., Wuthrich, K., Bailey, J.E. (1999), Metabolic flux ratio analysis of genetic and environmental modulations of Escherichia coli central carbon metabolism. Journal of Bacteriology, 181, 6679-6688. 21. Szyperski, T. (1995), Biosynthetically directed fractional 13C-labeling of proteinogenic amino acids. An efficient analytical tool to investigate intermediary metabolism. European Journal of Biochemistry, 232, 433-448. 22. Halket, J.M., Waterman, D., Przyborowska, A.M., Patel, R.K., Fraser, P.D., Bramley, P.M. (2005), Chemical derivatization and mass spectral libraries in metabolic profiling by GC-MS and LC-MS/MS. Journal of Experimental Botany, 56, 219-243. 23. Dauner, M., Sauer, U. (2000), GC-MS analysis of amino acids rapidly provides rich information for isotopomer balancing. Biotechnology Progress, 16, 642-649. 24. Nanchen, A., Fuhrer, T., Sauer, U. (2007), Determination of metabolic flux ratios from 13C-experiments and gas chromatography-mass spectrometry data: protocol and principles. In Methods in Molecular Biology, 2006/10/13 ed.; Weckwerth, W., Ed. Vol. 358, pp 177-197, 25. Maaheimo, H., Fiaux, J., Cakar, Z.P., Bailey, J.E., Sauer, U., Szyperski, T. (2001), Central carbon metabolism of Saccharomyces cerevisiae explored by biosynthetic fractional 13C labeling of common amino acids. European Journal of Biochemistry, 268, 2464-2479. 26. Antoniewicz, M.R., Kelleher, J.K., Stephanopoulos, G. (2007), Accurate assessment of amino acid mass isotopomer distributions for metabolic flux analysis. Analytical Chemistry, 79, 7554-7559. 27. Wellerdiek, M., Winterhoff, D., Reule, W., Brandner, J., Oldiges, M. (2009), Metabolic quenching of Corynebacterium glutamicum: efficiency of methods and impact of cold shock. Bioprocess and Biosystems Engineering, 32, 581-592. 28. Mashego, M.R., Rumbold, K., De Mey, M., Vandamme, E., Soetaert, W., Heijnen, J.J. (2007), Microbial metabolomics: past, present and future methodologies. Biotechnology Letters, 29, 1-16. 29. Weibel, K.E., Mor, J.R., Fiechter, A. (1974), Rapid sampling of yeast cells and automated assays of adenylate, citrate, pyruvate and glucose-6-phosphate pools. Analytical Biochemistry, 58, 208-216. 30. Schaub, J., Schiesling, C., Reuss, M., Dauner, M. (2006), Integrated sampling procedure for metabolome analysis. Biotechnology Progress, 22, 1434-1442. 31. Cronwright, G.R., Rohwer, J.M., Prior, B.A. (2002), Metabolic control analysis of glycerol synthesis in Saccharomyces cerevisiae. Applied and Environmental Microbiology, 68, 4448-4456. 32. Spura, J., Reimer, L.C., Wieloch, P., Schreiber, K., Buchinger, S., Schomburg, D. (2009), A method for enzyme quenching in microbial metabolome analysis successfully applied to gram-positive and gram-negative bacteria and yeast. Analytical Biochemistry, 394, 192-201. 33. Luo, B., Groenke, K., Takors, R., Wandrey, C., Oldiges, M. (2007), Simultaneous determination of multiple intracellular metabolites in glycolysis, pentose phosphate pathway and tricarboxylic acid cycle by liquid chromatography-mass spectrometry. Journal of Chromatography A, 1147, 153-164. 34. Wittmann, C., Kromer, J.O., Kiefer, P., Binz, T., Heinzle, E. (2004), Impact of the cold shock phenomenon on quantification of intracellular metabolites in bacteria. Analytical Biochemistry, 327, 135-139. 35. Smart, K.F., Aggio, R.B., Van Houtte, J.R., Villas-Boas, S.G. (2010), Analytical platform for metabolome analysis of microbial cells using methyl chloroformate

130 CHAPTER 5

derivatization followed by gas chromatography-mass spectrometry. Nature Protocols, 5, 1709-1729. 36. da Luz, J.A., Hans, E., Zeng, A.-P. (2014), Automated fast filtration and on-filter quenching improve the intracellular metabolite analysis of microorganisms. Engineering in Life Sciences, 14, 135-142. 37. Bolten, C.J., Kiefer, P., Letisse, F., Portais, J.C., Wittmann, C. (2007), Sampling for metabolome analysis of microorganisms. Analytical Chemistry, 79, 3843-3849. 38. Bolten, C.J., Wittmann, C. (2008), Appropriate sampling for intracellular amino acid analysis in five phylogenetically different yeasts. Biotechnology Letters, 30, 1993- 2000. 39. Shimazu, M., Vetcher, L., Galazzo, J.L., Licari, P., Santi, D.V. (2004), A sensitive and robust method for quantification of intracellular short-chain coenzyme A esters. Analytical Biochemistry, 328, 51-59. 40. Pflug, S. (2011), Globale Metabolomanalyse von Corynebacterium glutamicum, Dissertation, Universität Stuttgart, Institut für Bioverfahrenstechnik 41. Theobald, U., Mailinger, W., Reuss, M., Rizzi, M. (1993), In vivo analysis of glucose- induced fast changes in yeast adenine nucleotide pool applying a rapid sampling technique. Analytical Biochemistry, 214, 31-37. 42. Chassagnole, C., Noisommit-Rizzi, N., Schmid, J.W., Mauch, K., Reuss, M. (2002), Dynamic modeling of the central carbon metabolism of Escherichia coli. Biotechnology and Bioengineering, 79, 53-73. 43. Maharjan, R.P., Ferenci, T. (2003), Global metabolite analysis: the influence of extraction methodology on metabolome profiles of Escherichia coli. Analytical Biochemistry, 313, 145-154. 44. Villas-Boas, S.G., Hojer-Pedersen, J., Akesson, M., Smedsgaard, J., Nielsen, J. (2005), Global metabolite analysis of yeast: evaluation of sample preparation methods. Yeast, 22, 1155-1169. 45. Hajjaj, H., Blanc, P.J., Goma, G., Francois, J. (1998), Sampling techniques and comparative extraction procedures for quantitative determination of intra- and extracellular metabolites in filamentous fungi. FEMS Microbiology Letters, 164, 195- 200. 46. Meinert, S., Rapp, S., Schmitz, K., Noack, S., Kornfeld, G., Hardiman, T. (2013), Quantitative quenching evaluation and direct intracellular metabolite analysis in Penicillium chrysogenum. Analytical Biochemistry, 438, 47-52. 47. Yoshida, R., Tamura, T., Takaoka, C., Harada, K., Kobayashi, A., Mukai, Y., Fukusaki, E. (2010), Metabolomics-based systematic prediction of yeast lifespan and its application for semi-rational screening of ageing-related mutants. Aging Cell, 9, 616-625. 48. Dettmer, K., Aronov, P.A., Hammock, B.D. (2007), Mass spectrometry-based metabolomics. Mass spectrometry reviews, 26, 51-78. 49. Gullberg, J., Jonsson, P., Nordstrom, A., Sjostrom, M., Moritz, T. (2004), Design of experiments: an efficient strategy to identify factors influencing extraction and derivatization of Arabidopsis thaliana samples in metabolomic studies with gas chromatography-mass spectrometry. Analytical Biochemistry, 331, 283-295. 50. Fiehn, O., Kopka, J., Trethewey, R.N., Willmitzer, L. (2000), Identification of uncommon plant metabolites based on calculation of elemental compositions using gas chromatography and quadrupole mass spectrometry. Analytical Chemistry, 72, 3573-3580. 51. Tam, Y.Y., Normanly, J. (1998), Determination of indole-3-pyruvic acid levels in Arabidopsis thaliana by gas chromatography-selected ion monitoring-mass spectrometry. Journal of Chromatography A, 800, 101-108.

131 REFERENCES

52. Cerdan-Calero, M., Sendra, J.M., Sentandreu, E. (2012), Gas chromatography coupled to mass spectrometry analysis of volatiles, sugars, organic acids and aminoacids in Valencia late orange juice and reliability of the automated mass spectral deconvolution and identification system for their automatic identification and quantification. Journal of Chromatography A, 1241, 84-95. 53. Calheiros Gomes, L., Simoes, M. (2012), 13C metabolic flux analysis: from the principle to recent applications. Current Bioinformatics, 7, 77-86. 54. Jung, J.Y., Oh, M.K. (2015), Isotope labeling pattern study of central carbon metabolites using GC-MS. Journal of Chromatography B, 974, 101-108. 55. Jeffrey, F.M., Roach, J.S., Storey, C.J., Sherry, A.D., Malloy, C.R. (2002), 13C isotopomer analysis of glutamate by tandem mass spectrometry. Analytical Biochemistry, 300, 192-205. 56. Choi, J., Antoniewicz, M.R. (2011), Tandem mass spectrometry: a novel approach for metabolic flux analysis. Metabolic Engineering, 13, 225-233. 57. Choi, J., Grossbach, M.T., Antoniewicz, M.R. (2012), Measuring complete isotopomer distribution of aspartate using gas chromatography-tandem mass spectrometry. Analytical Chemistry, 84, 4628-4632. 58. Antoniewicz, M.R. (2013), Tandem mass spectrometry for measuring stable-isotope labeling. Current Opinion in Biotechnology, 24, 48-53. 59. Tepper, N., Shlomi, T. (2013), An integrated computational approach for metabolic flux analysis coupled with inference of tandem-MS collisional fragments. Bioinformatics, 29, 3045-3052. 60. Heinonen, M., Rantanen, A., Mielikainen, T., Kokkonen, J., Kiuru, J., Ketola, R.A., Rousu, J. (2008), FiD: a software for ab initio structural identification of product ions from tandem mass spectrometric data. Rapid Communications in Mass Spectrometry, 22, 3043-3052. 61. Antoniewicz, M.R. (2013), 13C metabolic flux analysis: optimal design of isotopic labeling experiments. Current Opinion in Biotechnology. 62. Wittmann, C. (2007), Fluxome analysis using GC-MS. Microbial Cell Factories, 6. 63. Fiaux, J., Cakar, Z.P., Sonderegger, M., Wuthrich, K., Szyperski, T., Sauer, U. (2003), Metabolic-flux profiling of the yeasts Saccharomyces cerevisiae and Pichia stipitis. Eukaryotic Cell, 2, 170-180. 64. Frey, A.D., Fiaux, J., Szyperski, T., Wuthrich, K., Bailey, J.E., Kallio, P.T. (2001), Dissection of central carbon metabolism of hemoglobin-expressing Escherichia coli by 13C nuclear magnetic resonance flux distribution analysis in microaerobic bioprocesses. Applied and Environmental Microbiology, 67, 680-687. 65. de Graaf, A.A., Mahle, M., Mollney, M., Wiechert, W., Stahmann, P., Sahm, H. (2000), Determination of full 13C isotopomer distributions for metabolic flux analysis using heteronuclear spin echo difference NMR spectroscopy. Journal of Biotechnology, 77, 25-35. 66. Christensen, B., Karoly Gombert, A., Nielsen, J. (2002), Analysis of flux estimates based on 13C-labelling experiments. European Journal of Biochemistry, 269, 2795- 2800. 67. Kleijn, R.J., Geertman, J.M., Nfor, B.K., Ras, C., Schipper, D., Pronk, J.T., Heijnen, J.J., van Maris, A.J., van Winden, W.A. (2007), Metabolic flux analysis of a glycerol- overproducing Saccharomyces cerevisiae strain based on GC-MS, LC-MS and NMR- derived C-labelling data. FEMS Yeast Research, 7, 216-231. 68. Lehnen, M., Ebert, B.E., Blank, L.M. (2017), A comprehensive evaluation of constraining amino acid biosynthesis in compartmented models for metabolic flux analysis. Metabolic Engineering Communications, 5, 34-44.

132 CHAPTER 5

69. van Winden, W., Verheijen, P., Heijnen, S. (2001), Possible pitfalls of flux calculations based on 13C-labeling. Metabolic Engineering, 3, 151-162. 70. Klapa, M.I., Aon, J.-C., Stephanopoulos, G. (2003), Systematic quantification of complex metabolic flux networks using stable isotopes and mass spectrometry. European Journal of Biochemistry, 270, 3525-3542. 71. Hartmans, S., Smits, J.P., van der Werf, M.J., Volkering, F., de Bont, J.A. (1989), Metabolism of styrene oxide and 2-phenylethanol in the styrene-degrading Xanthobacter strain 124X. Applied and Environmental Microbiology, 55, 2850-2855. 72. Pramanik, J., Keasling, J.D. (1997), Stoichiometric model of Escherichia coli metabolism: incorporation of growth-rate dependent biomass composition and mechanistic energy requirements. Biotechnology and Bioengineering, 56, 398-421. 73. Ebert, B.E., Kurth, F., Grund, M., Blank, L.M., Schmid, A. (2011), Response of Pseudomonas putida KT2440 to increased NADH and ATP demand. Applied and Environmental Microbiology, 77, 6597-6605. 74. Zhao, J., Shimizu, K. (2003), Metabolic flux analysis of Escherichia coli K12 grown on 13C-labeled acetate and glucose using GC-MS and powerful flux calculation method. Journal of Biotechnology, 101, 101-117. 75. Shaikh, A.S., Tang, Y.J., Mukhopadhyay, A., Keasling, J.D. (2008), Isotopomer distributions in amino acids from a highly expressed protein as a proxy for those from total protein. Analytical Chemistry, 80, 886-890. 76. Ghosh, A., Nilmeier, J., Weaver, D., Adams, P.D., Keasling, J.D., Mukhopadhyay, A., Petzold, C.J., Martin, H.G. (2014), A peptide-based method for 13C metabolic flux analysis in microbial communities. PLoS Computational Biology, 10, e1003827. 77. Ruhl, M., Hardt, W.D., Sauer, U. (2011), Subpopulation-specific metabolic pathway usage in mixed cultures as revealed by reporter protein-based 13C analysis. Applied and Environmental Microbiology, 77, 1816-1821. 78. Shaikh, A.S., Tang, Y.J., Mukhopadhyay, A., Martin, H.G., Gin, J., Benke, P.I., Keasling, J.D. (2010), Study of stationary phase metabolism via isotopomer analysis of amino acids from an isolated protein. Biotechnology Progress, 26, 52-56. 79. Wu, C., Xiong, W., Dai, J., Wu, Q. (2015), Genome-based metabolic mapping and 13C flux analysis reveal systematic properties of an oleaginous microalga Chlorella protothecoides. Plant Physiology, 167, 586-599. 80. Ma, F., Jazmin, L.J., Young, J.D., Allen, D.K. (2014), Isotopically nonstationary 13C flux analysis of changes in Arabidopsis thaliana leaf metabolism due to high light acclimation. Proceedings of the National Academy of Sciences of the United States of America, 111, 16967-16972. 81. Hay, J., Schwender, J. (2011), Computational analysis of storage synthesis in developing Brassica napus L. (oilseed rape) embryos: flux variability analysis in relation to 13C metabolic flux analysis. Plant Journal, 67, 513-525. 82. He, L., Xiao, Y., Gebreselassie, N., Zhang, F., Antoniewiez, M.R., Tang, Y.J., Peng, L. (2014), Central metabolic responses to the overproduction of fatty acids in Escherichia coli based on 13C-metabolic flux analysis. Biotechnology and Bioengineering, 111, 575-585. 83. Blank, L.M., Desphande, R.R., Schmid, A., Hayen, H. (2012), Analysis of carbon and nitrogen co-metabolism in yeast by ultrahigh-resolution mass spectrometry applying 13C- and 15N-labeled substrates simultaneously. Analytical and Bioanalytical Chemistry, 403, 2291-2305. 84. Poskar, C.H., Huege, J., Krach, C., Franke, M., Shachar-Hill, Y., Junker, B.H. (2012), iMS2Flux - a high-throughput processing tool for stable isotope labeled mass spectrometric data used for metabolic flux analysis. BMC Bioinformatics, 13, 295.

133 REFERENCES

85. Ebert, B.E., Blank, L.M. (2014), Successful downsizing for high-throughput 13C-MFA applications. Methods in Molecular Biology, 1191, 127-142. 86. Quek, L.E., Nielsen, L.K. (2014), Steady-state 13C fluxomics using OpenFLUX. Methods in Molecular Biology, 1191, 209-224. 87. IsoPro 3.1 (2014), https://sites.google.com/site/isoproms/, Accessed 20 Dec 2014. 88. Nikel, P.I., Kim, J., de Lorenzo, V. (2014), Metabolic and regulatory rearrangements underlying glycerol metabolism in Pseudomonas putida KT2440. Environmental Microbiology, 16, 239-254. 89. Wierckx, N.J., Ballerstedt, H., de Bont, J.A., Wery, J. (2005), Engineering of solvent- tolerant Pseudomonas putida S12 for bioproduction of phenol from glucose. Applied and Environmental Microbiology, 71, 8221-8227. 90. Blank, L.M., Ionidis, G., Ebert, B.E., Buhler, B., Schmid, A. (2008), Metabolic response of Pseudomonas putida during redox biocatalysis in the presence of a second octanol phase. FEBS Journal, 275, 5173-5190. 91. Ruhl, J., Schmid, A., Blank, L.M. (2009), Selected Pseudomonas putida strains able to grow in the presence of high butanol concentrations. Applied and Environmental Microbiology, 75, 4653-4656. 92. Nikel, P.I., de Lorenzo, V. (2014), Robustness of Pseudomonas putida KT2440 as a host for ethanol biosynthesis. New Biotechnology, 31, 562-571. 93. Kuepper, J., Dickler, J., Biggel, M., Behnken, S., Jager, G., Wierckx, N., Blank, L.M. (2015), Metabolic engineering of Pseudomonas putida KT2440 to produce anthranilate from glucose. Frontiers in Microbiology, 6, 1310. 94. Wittgens, A., Tiso, T., Arndt, T.T., Wenk, P., Hemmerich, J., Muller, C., Wichmann, R., Kupper, B., Zwick, M., Wilhelm, S., Hausmann, R., Syldatk, C., Rosenau, F., Blank, L.M. (2011), Growth independent rhamnolipid production from glucose using the non-pathogenic Pseudomonas putida KT2440. Microbial Cell Factories, 10, 80. 95. Kuepper, J., Ruijssenaars, H.J., Blank, L.M., de Winde, J.H., Wierckx, N. (2015), Complete genome sequence of solvent-tolerant Pseudomonas putida S12 including megaplasmid pTTS12. Journal of Biotechnology, 200, 17-18. 96. Tao, F., Shen, Y., Fan, Z., Tang, H., Xu, P. (2012), Genome sequence of Pseudomonas putida S12, a potential platform strain for industrial production of valuable chemicals. Journal of Bacteriology, 194, 5985-5986. 97. Jimenez, J.I., Minambres, B., Garcia, J.L., Diaz, E. (2002), Genomic analysis of the aromatic catabolic pathways from Pseudomonas putida KT2440. Environmental Microbiology, 4, 824-841. 98. Chavarria, M., Kleijn, R.J., Sauer, U., Pfluger-Grau, K., de Lorenzo, V. (2012), Regulatory tasks of the phosphoenolpyruvate-phosphotransferase system of Pseudomonas putida in central carbon metabolism. mBio, 3. 99. del Castillo, T., Ramos, J.L., Rodriguez-Herva, J.J., Fuhrer, T., Sauer, U., Duque, E. (2007), Convergent peripheral pathways catalyze initial glucose catabolism in Pseudomonas putida: genomic and flux analysis. Journal of Bacteriology, 189, 5142- 5152. 100. Hunt, J.C., Phibbs, P.V. (1983), Regulation of alternate peripheral pathways of glucose catabolism during aerobic and anaerobic growth of Pseudomonas aeruginosa. Journal of Bacteriology, 154, 793-802. 101. Puchalka, J., Oberhardt, M.A., Godinho, M., Bielecka, A., Regenhardt, D., Timmis, K.N., Papin, J.A., dos Santos, V.A.P.M. (2008), Genome-Scale reconstruction and analysis of the Pseudomonas putida KT2440 metabolic network facilitates applications in biotechnology. Plos Computational Biology, 4. 102. Lessie, T.G., Phibbs, P.V. (1984), Alternative pathways of carbohydrate utilization in pseudomonads. Annual Review of Microbiology, 38, 359-387.

134 CHAPTER 5

103. Vicente, M., Pedro, M.A., Torrontegui, G., Canovas, J.L. (1975), The uptake of glucose and gluconate by Pseudomonas putida. Molecular and Cellular Biochemistry, 7, 59-64. 104. Vicente, M., Canovas, J.L. (1973), Glucolysis in Pseudomonas putida: physiological role of alternative routes from the analysis of defective mutants. Journal of Bacteriology, 116, 908-914. 105. Daddaoua, A., Krell, T., Alfonso, C., Morel, B., Ramos, J.L. (2010), Compartmentalized glucose metabolism in Pseudomonas putida is controlled by the PtxS repressor. Journal of Bacteriology, 192, 4357-4366. 106. Latrach Tlemcani, L., Corroler, D., Barillier, D., Mosrati, R. (2008), Physiological states and energetic adaptation during growth of Pseudomonas putida mt-2 on glucose. Archives of Microbiology, 190, 141-150. 107. Schmitz, A., Ebert, B.E., Blank, L.M. (2015), GC-MS-based determination of mass isotopomer distributions for 13C-based metabolic flux analysis. Humana Press: pp 1- 21, 10.1007/8623_2015_78 108. Whiting, P.H., Midgley, M., Dawes, E.A. (1976), The role of glucose limitation in the regulation of the transport of glucose, gluconate and 2-oxogluconate, and of glucose metabolism in Pseudomonas aeruginosa. Journal of General Microbiology, 92, 304- 310. 109. Whiting, P.H., Midgley, M., Dawes, E.A. (1976), The regulation of transport of glucose, gluconate and 2-oxogluconate and of glucose catabolism in Pseudomonas aeruginosa. Biochemichal Journal, 154, 659-668. 110. Fuhrer, T., Sauer, U. (2009), Different biochemical mechanisms ensure network-wide balancing of reducing equivalents in microbial metabolism. Journal of Bacteriology, 191, 2112-2121. 111. Furch, T., Preusse, M., Tomasch, J., Zech, H., Wagner-Dobler, I., Rabus, R., Wittmann, C. (2009), Metabolic fluxes in the central carbon metabolism of Dinoroseobacter shibae and Phaeobacter gallaeciensis, two members of the marine Roseobacter clade. BMC Microbiology, 9, 209. 112. del Castillo, T., Duque, E., Ramos, J.L. (2008), A set of activators and repressors control peripheral glucose pathways in Pseudomonas putida to yield a common central intermediate. Journal of Bacteriology, 190, 2331-2339. 113. Conway, T. (1992), The Entner-Doudoroff pathway - history, physiology and molecular-biology. FEMS Microbiology Reviews, 103, 1-28. 114. Nikel, P.I., Chavarria, M., Fuhrer, T., Sauer, U., de Lorenzo, V. (2015), Pseudomonas putida KT2440 strain metabolizes glucose through a cycle formed by enzymes of the Entner-Doudoroff, Embden-Meyerhof-Parnas, and Pentose Phosphate pathways. Journal of Biological Chemistry, 290, 25920-25932. 115. Nogales, J., Palsson, B.O., Thiele, I. (2008), A genome-scale metabolic reconstruction of Pseudomonas putida KT2440: iJN746 as a cell factory. BMC Systems Biology, 2. 116. Pieterse, B., Jellema, R.H., van der Werf, M.J. (2006), Quenching of microbial samples for increased reliability of microarray data. Journal of Microbiological Methods, 64, 207-216. 117. Winder, C.L., Dunn, W.B., Schuler, S., Broadhurst, D., Jarvis, R., Stephens, G.M., Goodacre, R. (2008), Global metabolic profiling of Escherichia coli cultures: an evaluation of methods for quenching and extraction of intracellular metabolites. Analytical Chemistry, 80, 2939-2948. 118. Canelas, A.B., Ras, C., Pierick, A., Dam, J.C., Heijnen, J.J., Gulik, W.M. (2008), Leakage-free rapid quenching technique for yeast metabolomics. Metabolomics, 4, 226-239.

135 REFERENCES

119. Paczia, N., Nilgen, A., Lehmann, T., Gatgens, J., Wiechert, W., Noack, S. (2012), Extensive exometabolome analysis reveals extended overflow metabolism in various microorganisms. Microbial Cell Factories, 11. 120. Schellenberger, J., Que, R., Fleming, R.M., Thiele, I., Orth, J.D., Feist, A.M., Zielinski, D.C., Bordbar, A., Lewis, N.E., Rahmanian, S., Kang, J., Hyduke, D.R., Palsson, B.O. (2011), Quantitative prediction of cellular metabolism with constraint- based models: the COBRA Toolbox v2.0. Nature Protocols, 6, 1290-1307. 121. Matsushita, K., Yamada, M., Shinagawa, E., Adachi, O., Ameyama, M. (1980), Function of ubiquinone in the electron transport system of Pseudomonas aeruginosa grown aerobically. Journal of Biochemistry, 88, 757-764. 122. Pronk, J.T., Wenzel, T.J., Luttik, M.A., Klaassen, C.C., Scheffers, W.A., Steensma, H.Y., van Dijken, J.P. (1994), Energetic aspects of glucose metabolism in a pyruvate- dehydrogenase-negative mutant of Saccharomyces cerevisiae. Microbiology, 140 ( Pt 3), 601-610. 123. Chen, Y., Siewers, V., Nielsen, J. (2012), Profiling of cytosolic and peroxisomal acetyl-CoA metabolism in Saccharomyces cerevisiae. Plos One, 7, e42475. 124. van Roermund, C.W., Hettema, E.H., van den Berg, M., Tabak, H.F., Wanders, R.J. (1999), Molecular characterization of carnitine-dependent transport of acetyl-CoA from peroxisomes to mitochondria in Saccharomyces cerevisiae and identification of a plasma membrane carnitine transporter, Agp2p. EMBO J, 18, 5843-5852. 125. Verduyn, C., Postma, E., Scheffers, W.A., Vandijken, J.P. (1990), Physiology of Saccharomyces cerevisiae in anaerobic glucose-limited chemostat cultures. Journal of General Microbiology, 136, 395-403. 126. Bruinenberg, P.M., van Dijken, J.P., Scheffers, W.A. (1983), An enzymic analysis of NADPH production and consumption in Candida utilis. Journal of General Microbiology, 129, 965-971. 127. Albert, I. (2014), Entwicklung von Triple-Quadrupol-Massenspektrometrie-Methoden für die 13C-Stoffflussanalyse, Master Thesis, RWTH Aachen University, Institute of Applied Microbiology 128. Inc., T.F.S., TSQ 8000 AutoSRM User Guide. 2012. 129. Wiebach, V. (2014), Aufklärung des Acetyl-CoA Metabolismus in Saccharomyces cerevisiae mittels 13C-Flussanalyse, Research internship, RWTH Aachen University, Institute of Applied Microbiology 130. Wittmann, C., Heinzle, E. (1999), Mass spectrometry for metabolic flux analysis. Biotechnology and Bioengineering, 62, 739-750. 131. Long, C.P., Au, J., Gonzalez, J.E., Antoniewicz, M.R. (2016), 13C metabolic flux analysis of microbial and mammalian systems is enhanced with GC-MS measurements of glycogen and RNA labeling. Metabolic Engineering, 38, 65-72. 132. Miranda-Santos, I., Gramacho, S., Pineiro, M., Martinez-Gomez, K., Fritz, M., Hollemeyer, K., Salvador, A., Heinzle, E. (2015), Mass isotopomer analysis of nucleosides isolated from RNA and DNA using GC-MS. Analytical Chemistry, 87, 617-623. 133. Guzman, S., Marin, S., Miranda, A., Selivanov, V.A., Centelles, J.J., Harmancey, R., Smih, F., Turkieh, A., Durocher, Y., Zorzano, A., Rouet, P., Cascante, M. (2014), 13C metabolic flux analysis shows that resistin impairs the metabolic response to insulin in L6E9 myotubes. BMC Systems Biology, 8, 109. 134. Faijes, M., Mars, A.E., Smid, E.J. (2007), Comparison of quenching and extraction methodologies for metabolome analysis of Lactobacillus plantarum. Microbial Cell Factories, 6, 27.

136 CHAPTER 5

135. Villas-Boas, S.G., Bruheim, P. (2007), Cold glycerol-saline: the promising quenching solution for accurate intracellular metabolite analysis of microbial cells. Analytical Biochemistry, 370, 87-97. 136. Noh, K., Wiechert, W. (2006), Experimental design principles for isotopically instationary 13C labeling experiments. Biotechnology and Bioengineering, 94, 234- 251. 137. Noh, K., Gronke, K., Luo, B., Takors, R., Oldiges, M., Wiechert, W. (2007), Metabolic flux analysis at ultra short time scale: isotopically non-stationary 13C labeling experiments. Journal of Biotechnology, 129, 249-267. 138. Nicolae, A., Wahrheit, J., Bahnemann, J., Zeng, A.P., Heinzle, E. (2014), Non- stationary 13C metabolic flux analysis of Chinese hamster ovary cells in batch culture using extracellular labeling highlights metabolic reversibility and compartmentation. BMC Systems Biology, 8, 50. 139. Noack, S., Noh, K., Moch, M., Oldiges, M., Wiechert, W. (2011), Stationary versus non-stationary 13C-MFA: a comparison using a consistent dataset. Journal of Biotechnology, 154, 179-190. 140. Blank, L.M., Lehmbeck, F., Sauer, U. (2005), Metabolic-flux and network analysis in fourteen hemiascomycetous yeasts. FEMS Yeast Research, 5, 545-558. 141. Fischer, E., Sauer, U. (2003), Metabolic flux profiling of Escherichia coli mutants in central carbon metabolism using GC-MS. European Journal of Biochemistry, 270, 880-891. 142. Ruhl, M., Rupp, B., Noh, K., Wiechert, W., Sauer, U., Zamboni, N. (2012), Collisional fragmentation of central carbon metabolites in LC-MS/MS increases precision of 13C metabolic flux analysis. Biotechnology and Bioengineering, 109, 763- 771. 143. Tepper, N., Shlomi, T. (2015), Efficient modeling of MS/MS data for metabolic flux analysis. PLoS One, 10, e0130213.

137

CHAPTER 6

6. CURRICULUM VITAE

Personal Data Name: Andreas Schmitz Born: March 07, 1986 in Mechernich, North Rhine Westphalia, Germany Nationality: German

Work Experience and Education Since 08/2016 Head of quality control at the Peter Greven GmbH, Bad Münstereifel, Germany 11/2011 – 07/2015 Scientific assistant (PhD Student) at the RWTH Aachen, Institute of Applied Microbiology, Aachen, Germany Supervisor: Univ. Prof. Dr. Lars M. Blank 09/2009 – 10/2011 Student of Applied Chemistry at the Hochschule Niederrhein, University of Applied Sciences, Krefeld, Germany Degree: Master of Science 07/2006 – 09/2009 Student of Chemistry and Material Sciences at the Hochschule Bonn- Rhein-Sieg, University of Applied Science, Rheinbach, Germany Degree: Bachelor of Science 07/1996 – 06/2005 Attendance at the Gymnasium am Turmhof, Mechernich, Germany Degree: Abitur

Publications Schmitz, A., Ebert, B. E., Blank, L. M. (2015), GC-MS-Based determination of mass isotopomer distributions for 13C-based metabolic flux analysis - In: McGenity T., Timmis K., Nogales B. (eds) Hydrocarbon and Lipid Microbiology Protocols. Springer Protocols Handbooks. Springer, Berlin, Heidelberg Oral Presentations Schmitz, A., Ebert, B. E., Blank, L. M. (2015): Methodenentwicklung für die 13C- Stoffflussanalyse mittels Triple Quad GC-MS – Eingeladener Vortrag beim 10. Chromatographie Anwendertreffen ThermoScientific, March 03rd 2015, Düsseldorf, Germany

139 CURRICULUM VITAE

Poster Presentations Ebert, B. E., Schmitz, A., Blank, L. M. (2014): High-quality flux distributions of Saccharomyces cerevisiae - COBRA 2014 3rd Conference on constraint based reconstruction and analysis, May 20th-23rd 2014, Charlottesville, VA, USA

Schmitz, A., Albert, I., Ebert, B. E., Blank, L. M. (2014): 13C Metabolic Flux Analysis of S. cerevisiae Using Gas Chromatography-Tandem Mass Spectrometry - Microbiology and Infection 2014 - 4th Joint conference, October 05th- 08th 2014, Dresden, Germany

Lehnen, M., Schmitz, A., Ebert, B. E., Blank, L. M. (2016): Broadening the application range of compartmented eukaryotic metabolic models for cases of unknown cellular localization of amino acid synthesis pathways – VAAM Annual Conference 2016, March 13rd-16th 2016, Jena, Germany

Lehnen, M., Schmitz, A., Ebert, B. E., Blank, L. M. (2016): Broadening the application range of compartmented eukaryotic metabolic models for cases of unknown cellular localization of amino acid synthesis pathways – BioProcessing Days 2016, March 22nd-23rd 2016, Recklinghausen, Germany

140