An in Silico Characterization of Microbial Electrosynthesis for Metabolic Engineering of Biochemicals

An in silico Characterization of Microbial Electrosynthesis for Metabolic Engineering of Biochemicals

Aditya Pandit

A thesis submitted in conformity with the requirement for the degree of Masters of Applied Science

Graduate Department of Chemical Engineering and Applied Chemistry University of Toronto

An in silico Characterization of Microbial Electrosynthesis for Metabolic Engineering of Biochemicals

Aditya Vikram Pandit

Masters of Applied Science

Graduate Department of Chemical Engineering and Applied Chemistry University of Toronto

2012

ABSTRACT

A critical concern in metabolic engineering is the need to balance the demand and supply of redox intermediates. Bioelectrochemical techniques offer a promising method to alleviate redox imbalances during the synthesis of biochemicals. Broadly, these techniques reduce intracellular NAD+ to NADH and therefore manipulate the cell‘s redox balance. The cellular response to such redox changes and the additional reducing can be harnessed to produce desired metabolites. In the context of microbial fermentation, these bioelectrochemical techniques can improve product yields and/or productivity.

We have developed a method to characterize the role of bioelectrosynthesis in chemical production using the genome-scale metabolic model of E. coli. The results elucidate the role of bioelectrosynthesis and its impact on biomass growth, cellular ATP yields and biochemical production.

The results also suggest that strain design strategies can change for fermentation

processes that employ microbial electrosynthesis and suggest that dynamic operating strategies lead to maximizing productivity.

iii

ACKNOWLEDGMENTS

I would like to give my thanks to my supervisor, Prof. Mahadevan for giving me the opportunity to do my masters degree in his lab. It has been a great experience and I have learned a lot under his guidance. As an undergraduate, I never thought that I would ever work in anything related to biology. However, Prof. Mahadevan has passed on his passion and enthusiasm for metabolic engineering and microbiology on to me, and I am very glad to have done graduate work in metabolic engineering and microbiology.

I would like to extend my appreciation to my parents for their continuing support, guidance and encouragement. Most importantly, they have instilled in me a passion for learning – and for that I am grateful.

Finally, I would like to extend my thanks to all my colleagues in Biozone for their support, great conversations and levity during my studies. In particular, thanks to Nicholas Bourdakos for all the insightful conversations.

TABLE OF CONTENTS Abstract ...... ii Acknowledgments ...... iv Table of Contents...... v List of Tables...... vi List of Figures ...... vii List of Appendices ...... viii Glossary...... ix Chapter 1 ...... 1 Introduction and Background ...... 1 1.1 Importance of Developing Biochemicals ...... 1 1.2 Cofactor Manipulation as a Means to Drive Product Synthesis ..... 2 1.3 Bioelectrochemcial Techniques to Drive Product Synthesis ...... 3 1.4 Bioelectrochemcial Systems ...... 4 1.5 Extracellular Electron Transfer ...... 6 1.6 Computational Strategies for Strain Design ...... 7 Chapter 2 ...... 8 Knowledge Gaps and Statement of Objectives ...... 8 Chapter 3 ...... 10 Methods and Materials ...... 10 3.1 Modelling Bioelectrosynthesis for Chemical Production ...... 10 3.2 Flux Balance Analysis ...... 11 3.3 Modelling Electrode Interactions ...... 12 3.4 Augmenting the Model to Incorporate Heterologous Pathways ..14 3.5 Selection of Substrates and Products for Analysis ...... 14 Chapter 4 ...... 17 Results and Discussion ...... 17 4.1 Impact on ATP Yield and Biomass ...... 17

4.2 Impact on CO2 Fixing Pathways ...... 23 4.3 Impact of Product and Substrate Degree of Reduction on Bioelectrosynthesis ...... 25 4.4 Growth Coupled Electrical Enhancement ...... 31 4.5 Bioelectrosynthesis on Substrate Mixtures ...... 39 4.6 Limitations of Modelling Results ...... 41 Chapter 5: ...... 44 Conclusions and Recommendations ...... 44 Bibliography ...... 47 Appendix A ...... 54 Appendix B...... 57 Appendix C ...... 62 Appendix D ...... 66

LIST OF TABLES

Number Page

1 Milestones of Achievements in Bioelectrosynthesis ...... 4 2 NADH Produced or Consumed per Substrate or Product ...... 15 3 SPEEQ Values for Substrate Product Couplings ...... 15

4 Effect of CO2 Fixation on Growth Rate ...... 24 5 SPEEQ Values for Substrate Product Couplings ...... 27 6 Predicted Fluxes Through Selected Reactions ...... 29 7 Predicted Changes in Fluxes Through Selected Pathways ...... 37 8 Summary of Increases in Theoretical Product Yield Coupled to Biomass Growth ...... 40

LIST OF FIGURES

Number Page Figure 1 Milestones of Achievements in Bioelectrosynthesis ...... 4

Figure 2 Typical Bioelectrochemical System (BES) ...... 5

Figure 3 Central Metabolism and NADH Regeneration Maps ...... 13

Figure 4 Percentage Improvements in ATP Yield and Biomass Yield as a Result of Electrical Enhancement ...... 18

Figure 5 Map of Metabolic Flux Distributions for the Wild Type Metabolism .....22

Figure 6 Theoretical Increases in Product Yield Resulting from Electrical Enhancement ...... 27

Figure 7 Theoretical Increases in Product Yield of Growth Coupled Resulting from Electrical Enhancement...... 33

Figure 8 Production Envelopes for Growth Coupled and Electrically Enhanced 35

Figure 9 Production Envelopes for Three Growth Coupled Strategies for Ethanol...... 38

Figure 10 Changes in Product Yield, Biomass Yield and Substrate Specific Productivity as a Function of Electron Exchange Rate ...... 39

vii

LIST OF APPENDICES

Page Appendix A Model Reactions/Deletion Strategies ...... 54

Appendix B Modelling Results - Maximizing Biomass ...... 57

Appendix C Changes in Metabolic Flux Distribution ...... 62

Appendix D Maximization of Product ...... 66

viii

GLOSSARY

3HB 3-Hydroxybutyrate ACK Acetate kinase ACTP Acetyl-P ADH Alcohol Dehydrogenase AKG 2-ketoglutarate ADP Adenosine diphosphate ATP Adenosine triphosphate ATPs4rpp ATP syntahse BDOL 1,4-Butanediol BES Bioelectrochemcial System BTOL Butanol CIT Citrate CS Citrate synthase DHAP Dihydroxyacetone phosphate E4P Erythrose-4-phosphate EE Electrical enhancement ETOH Ethanol F6P Fructose-6-Phospate FBA Flux Balance Analysis FUM Fumarate FRD2 Fumarate reductase G3P Glucose-3-Phosphate G6P Glucose-6-Phosphate GAPD Glyceraldehyde phosphate dehydrogenase GLX Glyoxylate ICIT Isocitrate ICDHyr Isocitrate dehydrogenase Lactate Lactate MAL Malate NADH Nicotinamide adenine dinucleotide OAA Oxaloacetate ORP Oxidoreduction Potential PEP Phosphoenolpyruvate PDH Pyruvate dehydrogenase PGI Phosphoglucose isomerase PGK Phosphoglucomutase PFL Pyruvate formate lyase

PPC Phosphoenolpyruvate carboxylase PYK Pyruvate kinase RPE Ribulose-5-phosphate-3-epimerase RPI Ribose-P isomerase Ru5P Rubisco-5-phosphate S7P Seudoheptulose-7-phosphate SPEEQ Substrate Product Electron Equivalence Quotient Xu5P Xylulose-5-phosphate

CHAPTER 1 INTRODUCTION AND BACKGROUND

1.1 Importance of Developing Biochemicals In response to economic and environmental considerations, there has been increased interest in the development of commercial bioprocesses that produce biofuels and specialty chemicals. Conventional petrochemical routes that produce fuels and chemicals are environmentally unfriendly. Furthermore, as the cost of petrochemical feedstocks (crude oil) becomes increasingly expensive, alternative processes or feedstocks are eventually expected to replace or at least supplement them. The view held by proponents of biochemicals and biofuels is that advancements in metabolic engineering will eventually enable bioprocesses to replace or offset production by conventional petrochemical processes. Bioprocesses have the capacity to manufacture bulk and fine chemicals and fuels at significant scale but offer the advantage of being cost competitive and sustainable. The successful commercialization of these bioprocesses requires that the bioproducts which are converted from biomass based feedstocks are produced at sufficiently high yields and productivity so that the process is economically viable (Keasling, 2010). Hence, engineering the metabolism of organisms that drive these fermentative processes, such as Escherichia coli and Saccharomyces cerevisiae, is required to achieve desired process objectives (product yield, productivity, and titer) (Stephanopoulos 2007; Jarboe et al. 2010; Li, Cann, and Liao 2010). Metabolic engineering attempts to optimize the cellular metabolism of an organism to satisfy these desired process objectives. The metabolic engineering of an organism to overproduce a biochemical can be a long and iterative process involving many different strategies. Many of these strategies can be lumped together into like categories. The overproduction of a desired metabolite can be achieved by, 1) introducing exogenous metabolic pathways or enzymes. For example Sanchez and coworkers showed that by expressing a heterologous pyruvate carboxylase from Lactococcus lactis a greater flux could be directed to oxaloacetate to increase succinate yields (Sánchez, Bennett, San 2005); 2) Manipulating native and/or competing metabolic

1 pathways has also been shown to improve product yields. For example, Millard and colleagues showed that the overexpression of a native phosphoenolpyruvate carboxylase in E. coli resulted in a 3.5 fold increase in succinate production (Millard, Chao, J C Liao, & Donnelly, 1996). Finally, 3) manipulating cellular redox and energy reactions in order to overproduce desired metabolites as was shown by Singh et al for improved succinate production (Singh et al. 2011). Perturbations that are made to the cellular metabolism during bioelectrosynthesis fall most closely into this third category and electrical reducing power has the ability to affect the internal cellular redox conditions of the cell.

1.2 Cofactor Manipulation as a Means to Drive Product Synthesis Redox cofactors such as NADH or NADPH play an important role in cellular metabolism, and altering the cellular redox balance has been regarded as an essential initial step in metabolic engineering for achieving bioprocess objectives (Berríos-Rivera, Bennett, and San 2002). A fundamental study performed by Bennett and co-workers showed the ability to modulate the fermentation secretion pattern of E. coli in response to redox disturbances by altering the substrate (San et al. 2002). While altering the substrate is not a practical method to drive increases in product yield, the same study highlighted how the genetic manipulation of various enzymatic pathways offer a means to increase the NAD(P)H available to the cell (San et al. 2002). Increases in cellular NAD(P)H levels has been demonstrated as an effective way to increase the synthesis of desired products (Chemler, Fowler, McHugh, & Koffas, 2010; Chin, Khankal, Monroe, Maranas, & Cirino, 2009; Fasan et al., 2010). In essence, these strategies provide additional reducing power to the cell that can then be utilized to drive the synthesis of NADH dependent product reactions. An alternative approach for manipulating the redox metabolism is the use of bioelectrochemical techniques such as those that supply reducing power by generating NADH within the cell through interactions with an electrode. These techniques have been shown to be effective at increasing synthesis of several products including ethanol, n- butanol, and succinate in a variety of hosts including S. cerevisiae, Clostridium acetobutylicum,

Actinobacillus succinogenes (Shin, Zeikus, and Jain 2002; Kim and Kim 1988; Park and Zeikus 1999) Other microbes have also been successfully utilized for biotransformations or product synthesis (Peguin and Soucaille 1996; Park et al. 2005).

1.3 Bioelectrochemcial Techniques to Drive Product Synthesis These types of bioelectrochemical techniques, which refer to electricity driven product synthesis, are generally known as microbial electrosynthesis or bioelectrosynthesis (Rabaey & Rozendal, 2010). In 1979, Hongo and co-workers were among the first to use these techniques to increase product synthesis by showing that it was possible to improve L- glutamine yields (Hongo M et al. 1979). Considerable progress has been made since 1979 in the area of bioelectrosynthesis. Reduced carriers such as neutral red and methyl viologen have been shown to increase the yields of products such as ethanol, butanol and succinate and direct electron transfer has also been demonstrated in mediator free bioelectrochemical systems where electron transfer has occurred between the cell membrane and the cathode (Rabaey & Rozendal, 2010). In 2004, it was demonstrated that an electrode could serve as a sole energy source for Geobacter sulfurreducens (Bond, and Lovley 2004). Recently, Lovley et al. demonstrated a proof of concept for this approach; Sporomusa ovata was used to generate acetate and small amounts of organic compounds by reducing carbon dioxide with an electrode powered by solar energy (Nevin, Woodard, Franks, Summers, & D. R. Lovley, 2010). However, practical industrial implementation would require further genetic perturbations to the central metabolic pathways of S. ovata to develop strains that could produce valuable chemicals instead of acetate. Figure 1 shows a timeline of some of the major accomplishments since 1979.

Figure 1 Milestones of Achievements in Bioelectrosynthesis

Table 1 Milestones of Achievements in Bioelectrosynthesis Year Milestone 1979 One of the first demonstrations improving metabolite yield in the presence of a current supply 1988 Clostridium acetobutylicum ferments butanol at greater yields in the presence of mediated current supply 1999 Neutral red serves as the sole electro donor for Actinobacillus succinogenes and is linked to ATP production in the cell 2001 Trichosporon capitatum is ―stimulated‖ to produce 6-bromo-2-tetralol in the presence of a mediated current supply 2004 Geobacter sulfurreducens uses cathodes as the sole electron source 2008 Methane production by cathodic biofilms 2010 Sporomusa ovata is used to produce acetate and oxo-butyrate using electrical

current as the electron source and CO2 as the carbon source. 2010 Acetate is converted to ethanol by a microbial population in the presence of mediated current supply

1.4 Bioelectrochemcial Systems A bioelectrochemical system (BES) is a compartmentalized reactor system composed of an anodic and cathodic compartment separated by a proton exchange membrane. The anodic compartment contains an electrode where a substrate serves as the electron donor to the anode as it becomes oxidized. In the cathodic compartment, electrons are taken up by an electron acceptor causing the chemical species to become reduced. Theoretically, either reaction at the anode or the cathode can be catalyzed chemically (abiotically) or microbially (biotically). Microbial fuel cells are examples where the anodic reaction is microbially

4 catalyzed, during which a chemical substrate such as acetate is metabolized by a microbe such as G. sulfurreducens and the anode is in turn used as the terminal electron acceptor for the microbe (Rabaey et al., 2007). At the cathode, the electrons react with protons and oxygen in the presence of a platinum catalyst to form water. BESs that are used for bioelectrosynthesis operate reversely to a fuel cell. The cathode compartment is catalysed by a microbe (Rabaey et al., 2007). Electrons are accepted by the microbe and incorporated into its metabolism. The microbe uses the electrons and the chemical substrate present in the cathodic compartment to produce biochemicals and biomass. The electrons are supplied by the anodic compartment where a chemical species oxidized. Water is typically oxidized at the anode into protons, electrons and oxygen. Wastewater containing organic substrates such as acetate can also serve as electron donor at the anode (Rabaey & Rozendal, 2010)

Figure 2 Typical Bioelectrochemical System (BES) A typical BES consists of two compartments separated by an ion exchange membrane that separates the oxidation reaction from the reduction reaction.

1.5 Extracellular Electron Transfer Bioelectrosynthesis requires a conduit for electrons to travel from the inner membrane to the outer membrane. In electricigens, this conduit is made up of a series of proteins present on the outer membrane and others spanning the periplasmic space. These proteins are c-type cytochromes containing reducible heme groups (Londer et al, 2007; Logan, 2009; Shelobolina et al, 2007). Conductive nanowires extending from the surface of the microbe have also been identified and are implicated in the transfer of extracellular electrons (Reguera et al, 2005). The mechanistic information on how microbes transfer electrons from the inner membrane to a final electron acceptor such as an electrode is generally well understood (Rabaey & Rozendal, 2010). While the specific proteins responsible for electron transfer are different for different species, the general mechanism by which electron transfer occurs is analogous. For example Geobacter and Shewanella employ similar strategies of using cytochromes and conductive pili for extracellular electron transfer. Bioelectrosynthesis requires electron transfer to proceed in reverse, and electron transport in Geobacter has been shown to be bi-directional. Interestingly, a number of organisms that are not natural electricigens have been shown to accept electrons via a mediator driven process. However as of yet, very little has been elucidated on the mechanism of this reverse electron flow. In Acidithiobacillus ferrooxidans reverse electron transfer from iron(II) to the cell is thought to occur via enzymes and electron transfer proteins that are present on the membranes and that span the periplasmic space (Valdes et al, 2008). A reversible NADH dehydrogenase then accepts the electrons from the periplasmic electron carrier in order to drive the synthesis of NADH. It could be postulated that Geobacter or other species capable of reverse electron transfer may employ similar strategies or have homologous proteins present in their electron transfer chain. However, much remains yet to be learned about the mechanism by which this process occurs. Electron flow in mediator driven electron transfer that has been observed in non- electricigens such as E. coli, A. succinogenes and Clostridium acetobutylicum, but mechanistic information has not yet been elucidated. Park et al. provided a hypothetical model in

6 which they suggested the role of mediators such as neutral red play may be as electronophores or electron channels. In their model, neutral red, a hydrophobic compound, is capable of binding to the cell membrane. Neutral red becomes reduced by the electrode by direct contact and can then directly reduce NAD+ to NADH.

1.6 Computational Strategies for Strain Design Computational algorithms that can be used to predict cellular response aid in the design of strains for the over-production of biochemicals by an organism. The metabolism of an organism such as E. coli can be represented in silico by a genome scale model. Computational algorithms such as Opt-Knock, OptForce and EMILiO can then be used to identify knockout, overexpression and/or inhibition strategies ( Burgard, Pharkya, and Maranas 2003; Ranganathan, Suthers, and Maranas 2010; Yang, Cluett and Mahadevan 2010) in order to overproduce the target metabolite. However, the benefits of analysis using genome scale models extend beyond strain design and can be diverse. In addition to the above mentioned application for genome- scale models for metabolic engineering, studies that incorporate genome scale models can have other useful applications: 1) they can provide valuable insight in understanding metabolic capabilities of the underlying network, 2) they can provide a model-based explaination to help interpret the causes of microbial physiology and 3) they can predict and analyze bacterial growth phenotype (Durot, Bourguignon, & Schachter, 2009). In short, metabolic reconstructions of microbial metabolism can be an important tool for the microbiologist in understanding, predicting and then ultimately explaining bacterial phenotype, particularly as it relates to its genotype. Currently, however, these computational tools have been limited to perturbations that affect the metabolism through genetic manipulation techniques and not through electrochemical techniques. The lack of similar computational tools that could aid in the understanding of electrochemical perturbations to the cellular metabolism and the growing interest in microbial electrosynthesis motivates the development of a computational framework that can be used in the rational design of strains but also explain observed growth behaviour.

CHAPTER 2 KNOWLEDGE GAPS AND STATEMENT OF OBJECTIVES

At the most fundamental level, bioelectrosynthesis affects the cellular metabolism by altering the cell‘s internal redox state as NADH is electrically generated. In the context of using bioelectrochemical techniques for industrial synthesis of biofuels or biochemicals, two overarching questions need to be addressed. The first being what specific changes to the cell‘s metabolism are to be expected as NADH is generated electrically within the cell? And secondly, can these changes be harnessed through metabolic engineering to improve the synthesis of these desired products?

Earlier studies have shown that bioelectrosynthesis does indeed impact fermentation products and biomass growth rates (Park and Zeikus 1999; Peguin and Soucaille 1996; Park et al. 2005). Yet, while it is generally accepted that bioelectrosynthesis leads to enhanced yields of products, there exists a lack of understanding of how these techniques improve yield in a systematic framework that takes into account redox manipulation. In other words, it has been demonstrated, as an example, that succinate and ethanol yields can be enhanced using these techniques. But are the improvements in the expected yield to be the same for the different products? Or, as an example, would yields vary with other factors such as with substrate – fermentation on glucose versus glycerol.

Therefore, the overarching hypothesis of this project is that a computational framework makes it possible to evaluate the role of microbial electrosynthesis for the production of biochemicals and identify the limitations of using electricity to drive product synthesis as well. The insight gained by this using this framework helps to explain observed bacterial behaviour and provides guidance for where additional study research needs to be.

Statement of Objectives The lack of such a framework means that there is also a lack of systematic understanding of the instances where bioelectrosynthesis can lead to improved product yield (electrical enhancement). Hence, in order to characterize the role of electrical enhancement on chemical production, in this study we:

1. Develop a method to analyze the impact of microbial electrosynthesis on biochemical production using the genome-scale metabolism of E coli;

2. Examine the role of microbial electrosynthesis on biomass growth;

3. Examine the impact on biochemical production for a suite of top value-added chemicals;

4. Identify which conditions microbial electrosynthesis is best suited as a tool to improve yield; and

5. Comprehensively evaluate how microbial electrosynthesis may impact strain design and process productivity.

CHAPTER 3 METHODS AND MATERIALS

3.1 Modelling Bioelectrosynthesis for Chemical Production The iAF1260 metabolic reconstruction of Escherichia coli was used as the basis for all in silico evaluations (Feist et al., 2007). The iAF1260 model was selected because it is well-curated, studied and experimentally validated. Moreover, genetic tools for E. coli are well established and it is used widely in industry as platform for biochemical production. Even though E. coli is not a natural electricigen in the way that Geobacter or Shewanella are, E. coli can be electroactive and interact with an electrode in the presence of mediators (Xie, Li, and Tang 2010a; Xie, Li, and Tang 2010b)

Recently E. coli was shown to reduce solid α-Fe2O3 in the absence of mediators after portions of the extracellular electron transfer chain of Shewanella oneidensis was expressed in E. coli (Jensen et al., 2010). This ability to reduce metals is a trait that mimics natural electricigens such as Geobacter sulfurreducens, suggesting that direct electron transfer between electrodes may be a possibility. In short, E. coli has shown to exhibit electroactive characteristics. But because it is a model organism it can be used to understand general biological phenomenon and then apply the insight gained to the particular metabolic workings of other organisms. Therefore, extrapolating the findings made using genome scale models can help to rationalize observations made in other organisms such as those bioelectrosynthesis examples in the literature that used A. succinogenes or S. cerevisiae, but for which detailed or reliable metabolic models do not exist. E. coli has the additional benefit of being widely used to produce biofuels and biochemicals whereas electricigens like Geobacter and Shewanella are generally not considered as suitable platforms for producing biochemicals. Hence, the E. coli metabolic reconstruction is suitable for analyzing the various aspects of electrical generation of NADH on the metabolic capacity for producing value-added bioproducts.

3.2 Flux Balance Analysis Flux balance analysis (FBA) provides a mathematical framework for modelling and predicting the metabolism of microorganisms (Becker et al., 2007). The FBA solution solves the intercellular and exchange reaction fluxes by solving a metabolic model. The model can be thought of as a set of the reactions present in the cell‘s metabolism. These reactions can be put into the form of a stoichiometric matrix (m x n) where the rows represent all of the different metabolites present in the system, and the columns represent all of the gene protein reactions. Each reaction can be constrained by thermodynamic data that constrains the directionality of the reaction to prevent flux through thermodynamically infeasible pathways. Reactions are further constrained by known and/or measured limits on reaction fluxes such as nutrient uptake rates or enzymatic capacity. The model also contains a biomass reaction that represents the synthesis of the cell from its precursor metabolites. In general, this stoichiometric matrix, S, contains more fluxes than there are metabolites. Therefore, the general system is underdetermined and there exists more than one feasible solution for the flux distribution. Therefore, in order to arrive at a unique solution, linear programming can be employed where the optimal flux distribution is obtained when the matrix is solved using an objective function that corresponds to a defined cellular goal. The most common objective function in FBA is the maximization of the biomass reaction based on the ‗hypothesis that cellular metabolism is programmed through evolution for optimal resource utilization and growth‘ (Mahadevan, 2009). The LP problem can be formulated in the following way: Maximize (CT v) s.t. S●v = 0

vlb ≤ v ≤ vub i = 1, 2, 3, ... , n

where S is the stoichiometric matrix and vi is the flux thorugh reaction i.

3.3 Modelling Electrode Interactions To account for the interactions with an electrode, two reactions were added to the model reconstruction. These reactions are described in the Appendix A and represent the net reaction that occurs between the electrode and free NAD+ in the cytoplasm through the quinone pool. They are based on pathways used by bacteria such as Acidithiobacillus ferrooxidans, whose outer membrane has cytochromes that are responsible for oxidizing iron ore (Gaël Brasseur et al. 2002; Valdés et al. 2008; Elbehti, Brasseur, and Lemesle- Meunier 2000). Figure 3A shows this process schematically. While these proteins are not native, E. coli is known to interact with an electrode when neutral red serves as a mediator, and this interaction has been demonstrated to affect its cellular metabolism (D H Park & J G Zeikus, 2000). Furthermore, E. coli has shown dissimilatory iron reduction under cymA expression from Shewanella oneidensis (Gescher, Cordova, & Spormann, 2008). Figure 3A shows this process schematically. The lower value for the electron uptake rate was set based on known and measurable reverse electron flows that occur in electricigens. Bond and Lovley measured the electron donation rate to an electrode for G. sulfurreducens at 1.2 μmol/mgProtein-min or 1.07A/gDW (Bond and Lovley 2003). The same study reported a corresponding current production at 1.143A/m2. Xie et al showed that E. coli is capable of becoming redox active in the presence of an electrode modulated by a computerized potentiostat (Xie, Li, and Tang 2010b). Based on their data, an electron donation rate of ca. 600 mA/gDW was calculated. The electron uptake rate used in this study (ca. 800mA/gDW) is consistent with these reported values. The calculation of growth rates and product flux was computationally determined using a Flux Balance Analysis (FBA) framework. The FBA framework has been previously used to analyze growth of E. coli (Varma and Palsson 1994; Varma and Palsson 1994). Computations were performed in MATLAB (The Math-Works Inc., Natick, MA, USA) and the COBRA Toolbox (Becker et al., 2007).

Figure 3 Central Metabolism and NADH Regeneration Maps (A) A schematic of the bioelectrosynthesis process. (i) shows proteins that would likely be necessary for reverse electron flow to drive bioelectrosynthesis. (ii) shows bioelectrosynthesis by a mediator driven process. (B) The central metabolism of E. coli shown with the products that were analysed in this study. The key abbreviations are provided in the Glossary.

3.4 Augmenting the Model to Incorporate Heterologous Pathways In addition to the electron transfer reactions described earlier, a number of other modifications were made to the model to account for growth-coupled product flux and the production of biochemicals via pathways not native to E. coli. These heterologous pathways had been previously identified in literature and were incorporated into the model (Feist et al. 2010; da Silva, Mack, and Contiero 2009; Nakamura and Whited 2003; Atsumi et al. 2008; Burgard, Van Dien, and Burk 2009). A summary of these reactions and the corresponding growth coupled knockouts is provided in the Appendix A. All simulations were performed under anaerobic conditions. A substrate uptake rate of 10 mmol/gDW-hr for glucose was used. For ease of comparison to other substrates on an equal carbon basis, the following values were used for other substrates: 10 mmol/gDW-hr for gluconate and sorbitol. 12 mmol/gDW-hr for xylose; 20 mmol/gDW-hr for glycerol and 5 mmol/gDW-hr for maltose. The electron uptake rate was limited to a maximum of 30 mmol/gDW-hr. The model was allowed to choose any uptake rate that maximized the product flux (or biomass) between 0 and 30 mmol/gDW- hr.

3.5 Selection of Substrates and Products for Analysis E. coli is capable of growing on a number of different carbon sources. The carbon source utilized by the organism can impact the distribution of the fermentation products during anaerobic growth. The change in the by-product secretion patterns, and the corresponding yield of a product, is often associated with the degree of reduction (defined below) of the substrate and the external redox state. Electrical enhancement introduces another perturbation to the cell‘s redox state. Therefore, to characterize the general impact of electrical enhancement on product yield, we considered various product-substrate pairs, and modelled growth under a variety of different carbon substrates. Many of these product compounds have been identified as commercially valuable by the US Department of Energy (Werpy et al, 2004). Table 2 shows the list of carbon sources that were selected as substrates and products.

Table 2 NADH Produced or Consumed per Substrate or Product Products Substrates NADH NADH NADH NADH Consumed # of C Produced # of C Succinate 2 0.5 Glucose 2 0.33 1,3-Propanediol 3 1.0 Xylose 1.7 0.33 1,4-Butanediol 6 1.5 Glycerol 2 0.67 n-Butanol 4 1.0 Maltose 4 0.33 Ethanol 2 1.0 Sorbitol 3 0.50 Gluconate 1 0.17

Table 3 SPEEQ Values for Substrate Product Couplings Succinate 1,3-Propanediol 1,4-Butanediol n-Butanol Ethanol Gluconate 0.17 0.33 0.17 0.11 0.17 Glucose 0.33 0.67 0.33 0.22 0.33 Xylose 0.33 0.67 0.33 0.22 0.33 Sorbitol 0.50 1.00 0.50 0.33 0.50 Glycerol 0.67 1.33 0.67 0.44 0.67 Maltose 0.33 0.67 0.33 0.22 0.33

To establish a systematic method to analyze the relationship between the substrate-product coupling, we compared the yield improvements as a function of the degree of reduction of the substrate and product. We defined the degree of reduction as the number of reducing equivalents produced by a substrate during its oxidation to pyruvate or the number of reducing equivalents consumed to produce a desired metabolite from pyruvate, divided by the carbon length of that substrate or product. These values are shown in Table 2. We then divided the degree of reduction of the substrate by the degree of reduction of the product to create single parameter which we defined as the Substrate Product Electron Equivalence Quotient (SPEEQ). We related this parameter to product yield improvement. SPEEQ provides information on the relative degree of reduction of substrate to the product, and if SPEEQ is greater than one, it implies that the substrate is more reduced than the product and vice-versa. The products specified in Table 2 were carefully selected so that we could test the effect of the products‘ degree of reduction on the yield improvement. For example,

15 products such as ethanol and n-butanol allowed us to test the effect of keeping the degree of reduction constant while varying the carbon length and pathway dependency, while products such as succinate, 1,3-propanediol and 1,4-butanediol allowed us to evaluate how different degrees of reduction impact yield. Figure 3B shows the pathways present in the central metabolism of E. coli that consume or produce these substrates and products.

CHAPTER 4 RESULTS AND DISCUSSION

4.1 Impact on ATP Yield and Biomass Characterizing changes to ATP yield and biomass growth rate are fundamental to understanding how electrical enhancement impacts cellular metabolism. To determine the extent to which electrical enhancement can influence ATP production, we calculated the maximal ATP yield that could be generated in the presence of electrical enhancement. While the maximization of ATP (without biomass) is not perhaps a physiologically valid objective function, it serves two purposes in understanding how bioelectrosynthesis influences the energetics of the underlying metabolic network. Firstly, it helps to isolate and identify the energy producing pathways that could make the largest contribution to additional ATP for the cell. Secondly, comparison of relative increases in maximal ATP production against increases in biomass yield can help distinguish whether the biomass production is limited by ATP, redox, or carbon availability. By analyzing the differences in ATP and biomass yields, it is possible identify the constraints on the metabolic network for increasing ATP and biomass yields by bioelectrosynthesis. The results of the simulations are shown in Figure 4A. Figure 4A shows electrical enhancement is capable of increasing the ATP yield in range of 5-13% of the base case (without electrical enhancement). More importantly, the predicted absolute increase in the ATP produced was identical, 1.9mmol/gDW-hr in all cases except for sorbitol and glycerol utilizing conditions. These two exceptions can be explained by their already high degree of reduction, and therefore the extra electrons supplied during bioelectrosynthesis have a lower or no benefit on ATP production. The results suggest two important points. The increase in ATP produced by the cell during electrical enhancement is directly proportional to the current supplied to the cell. Secondly, the underlying mechanism by which ATP is produced is linked to specific metabolic pathways that are described in detail in the following section.

Figure 4 Percentage Improvements in ATP Yield and Biomass Yield as a Result of Electrical Enhancement (A) The improvement in the theoretically maximum synthesis of ATP by electrical enhancement. (B) The improvement in biomass growth rate and ATP yield under electrically enhanced conditions.

Since there are a limited number of reactions in the central metabolism that are capable of generating ATP under anaerobic conditions, (acetate kinase (ACKr), pyruvate kinase (PYK), phosphoglycerate kinase (PGK), ATP synthase (ATPs4rpp)), analysis of these four pathways can suggest a mechanism by which electrical enhancement improves ATP yields.

Physiologically, E. coli operates ATP synthase as a low flux ATP driven proton pump under anaerobic conditions (Kasimoglu, S. J. Park, Malek, Tseng, & Gunsalus, 1996). This pump discharges protons during anaerobic growth to generate an electrochemical proton gradient usually required for solute transport and flagellar rotation. During bioelectrosynthesis, electrons that are transferred into the cell drive the production of NADH. This production of NADH results in consumption of protons in the cytoplasm, leading to a lower concentration of protons in the cytoplasm. For a large enough current exchange rate, it may lead to a greater concentration of proton in the periplasm than in the cytoplasm and hence a proton gradient. The creation of this proton gradient, it is hypothesized, may allow ATP synthase to generate ATP and in order to replace the consumed protons. For every four moles of protons transported into the cell, ATP synthase also produces one mole of ATP. Hence, the addition of electrons can directly enhance ATP generation through the reversal of ATP synthase from ATP consumption (proton efflux) to ATP production (proton influx). Appendix B provides a sample of the flux distributions. In their pivitol study in 1999, Park and Zeikus found that electrical enhancement with neutral red as an electron mediator for succinate respiring A. succinogenes was able to drive proton translocation and increase ATP synthesis (Park and Zeikus 1999). Their experimental results showed that the strain undergoing bioelectrosynthesis consumed significantly more glucose and had greater biomass and succinate production at the end of the batch. In a patent filed by the same authors, they extended their hypothetical model to E. coli (Zeikus and Park 2001). Our in silico analysis helps to explain mechanistically how bioelectrosynthesis increases ATP synthesis even when flux through the ATP producing acetate kinase reaction decreases and oxidative phosphorylation is not available to the cell. This analysis seems to indicate that energy yields within the cell could be manipulated by controlling the current of the electrode, since an increased current to the microbe could facilitate the generation of a proton gradient by driving proton consumption. The ability to increase ATP yields within a cell offers new perspectives on how strains could be designed to maximize growth rates, particularly under fermentative

19 conditions when cells are usually energy starved. We examined the ATP production when the objective function was assumed to be maximization of growth rates in silico to further test this idea. The change in the net ATP flux produced by the four pathways (kinases and the ATP synthase) available to E. coli under anaerobic conditions (biomass objective) was calculated and plotted as a percentage of the base case (no electrical enhancement). Figure 4B shows that increases in biomass yield from electrical enhancement do occur and that the enhancements, with the exception of pyruvate and glycerol utilizing conditions, fall in a limited range. Figures 4A and 4B together help elucidate the role that generation of NADH by an electrochemically controlled bioelectrochemical system (BES) has on biomass yields through the manipulation of both cellular energetics (ATP) and cellular redox conditions (NADH). Consider that for glucose and xylose utilizing conditions, there is an expected increase of 6% and 8% in ATP yield, and 10% and 12% in biomass yield, respectively. When gluconate is used as the substrate, the increase in ATP and biomass yield is 5% and 15%, respectively. In contrast, a significant disparity between increases in ATP yield and biomass yield can be observed for pyruvate (12% and 31% increase in ATP & biomass respectively). These results suggest that the improvements in yield are not the sole result of additional ATP generated by a proton motive force (described above). For example, if increases in biomass yield arose only from ATP, then the increase in biomass yield on pyruvate predicted by the simulation should be less marked. Rather, it is the combination of reducing power available to the cell (dependent on current supply at the electrode and the substrate‘s degree of reduction), and ATP that leads to higher growth yields. This result suggests that relative increases in biomass yields are greater for oxidized substrates than for more reduced ones. The above conclusion has the following implications for industrial bioelectrosynthesis. In the presence of an NADH drain on the system, such as biosynthetic pathways, or pathways that might be required for the production of highly reduced metabolites, bioelectrosynthesis offers the possibility of boosting biomass yields by supplying reducing power as well as producing a proton motive force that could

20 generate ATP. This concept is particularly important for strain designs that result in poor growth rates. These strains are usually not industrially viable because of their poor growth rates, but could be made viable by electrically boosting growth rates while maintaining specific knockout strategies geared towards metabolite production (See n-butanol case in Section 4.4). Analysis of the flux distributions of the central metabolism provide insight into how these improvements can be achieved. The wild-type metabolic network of E. coli shows increases in flux through parts of the pentose phosphate pathway, mid glycolysis and the branched TCA cycle. Changes to the flux of the reactions that belong to these pathways are generally similar and are approximately 10% of the base case. Significant changes in flux appear in those reactions that are capable of consuming NADH. These changes include: 1) the pyruvate metabolism for which ethanol producing pathways have higher fluxes (90% increase); 2) the acetate producing pathways which have a lower flux value (90% decrease);. and 3) ATP synthase shows a change in directionality (for reasons previously described). A metabolic map describing the changes to fluxes in the central metabolism is in Figure 5. .

Figure 5 Map of Metabolic Flux Distributions for the Wild Type Metabolism A metabolic map of the central metabolism of E. coli shows the expected changes in the flux distribution. As more reducing power is available to the cell, there are increase in the pentose phosphate pathway and branches of the TCA cycle which produce biomass precursors.

It is interesting to note from Figure 5, that the there is a global response in the increase in flux in biosynthetic pathways. While it might be initially expected that the redox stress placed on the cellular metabolism by an electrode may actually drive carbon away from biomass precursors to reduced metabolites like ethanol and result in a lower expected growth rate, the modelling results suggest that an opposite behaviour can be expected.

Carbon flux is diverted from metabolites like acetate, and CO2 is utilized as more efficiently as carbon source – the combination of which lead to greater biomass yield. Those reactions present in the lower glycolysis show negligible increases in carbon flux.

Significance of Findings 1. Modelling results help explain the mechanism that leads to higher ATP yields observed in strains growing under electrochemically reduced conditions.

2. Impacts of electrochemically reduced conditions are eventually limited by the degree of reduction of the substrate.

3. Analysis of the flux distributions shows increased fluxes through pentose phosphate pathway and reductive branch of TCA cycle. This suggests that electrochemically reduced conditions have a broad impact on the cellular metabolism and are not limited to a few localized redox reactions, a finding that is important for metabolic engineering strategies that target various areas of the cellular metabolism (ex. those that may target amino acid metabolism).

4.2 Impact on CO2 Fixing Pathways Bioelectrosynthesis provides reducing power and if used in conjunction with highly oxidized carbon sources such CO2, it may be possible to substantially improve yields (Travick, Burk, and Burgard 2010). We explored this concept further by incorporating the Wood-Ljungdahl pathway into the model (Ragsdale & Pierce, 2008) (See Appendix A for pathway).

The results shown in Table 3 suggest that simultaneous CO2 fixation and electrical generation of NADH can lead to a substantial improvement in biomass yield and growth rate. The improvement in growth rate is almost two fold relative to the wild type under electrically enhanced and carbon fixing conditions. The Wood-Ljungdahl pathway produces acetyl-CoA as the final product. A fraction of this can be converted to acetate to generate ATP while part of the acetyl-CoA can be transformed to biomass precursors provided that NADH cofactor requirements for these reactions are met. Electrical enhancement provides some of the reducing power that is necessary to meet these

23 requirements. Without enhancement, glucose is the sole supply of reducing power leading to lower relative yields.

Table 4 Effect of CO2 Fixation on Growth Rate Glucose & Glucose & Wood- Glucose with Wood- Enhancement Condition Glucose Ljungdahl Enhancement Ljungdahl on only CO Pathway with 2 Pathway Enhancement Growth 0.19 0.21 0.31 0.38* 0.008 Rate (hr-1) Growth 3.1 3.4 5.2 5.8 -- Yield (x102)** *Carbon from CO2 represented roughly 10% of total carbon. Carbon from CO2 as a faction of total carbon (glucose + CO2) was negligible for other conditions. ** Defined as Growth Rate/(6*Total Incoming Glucose + Total Incoming Carbon Dioxide)

The growth yield shows the total biomass produced relative to total incoming carbon is larger with electrical enhancement than without. This increase occurs due to a combination of additional CO2 fixed and less carbon secreted as metabolites such as formate. Therefore, bioelectrosynthesis improves the specificity of the biomass reaction by incorporating more carbon into biomass precursors. While carbon fixation can be used to increase biomass and improve product yields in the presence of glucose (or some other substrate), we wanted to evaluate the potential for bioelectrosynthesis when CO2 is the sole carbon source. The maximum theoretical succinate flux was used as the criteria for evaluating this potential. The maximum theoretical succinate production was calculated to be 0.38 mmol/gDW-hr under anaerobic conditions at a maximum electron uptake rate of 30 mmol/gDW-hr. The calculated succinate flux is very low and this suggests that much larger electron uptake rates are required to achieve product fluxes that are comparable to succinate production from glucose. The results seem to suggest that the best strategy, in the short term, for synthesizing chemicals or fuels would be the co-utilization of CO2 with other hexose and pentose sugars. Our data suggests the microbes capable of reducing CO2 to a desired metabolite such as succinate would require an electron uptake flux an order of magnitude larger than what we used in the model. The equivalent electron flux for substituting glucose would be 2.4 102 mmol/gDW-hr for 10 mmol/gDW-hr of glucose. This flux

24 corresponds to current density of 6.8 A/m2 assuming biofilm cell densities reported in the literature (Bond and Lovley 2003). In some cases it may be possible to supply a current sufficient for bioelectrosynthesis to microbes, because lower rates may be required for synthesizing less reduced metabolites. However, using current microbial fuel cell technology as a basis for drawing some conclusions, it becomes apparent that achieving sufficiently high current exchange rates with an electrode can be difficult in non- electricigens since the one of the highest reported electron transfer rates for electricigens in microbial fuel cells (even though this transfer is in the direction of current generation) is 7.6 A/m2 (Logan, 2009). This potential limitation in supply of electrons also suggests avenues for future research to improve the electron transfer rate through a better understanding of the microbe-electrode interactions particularly when the electrode is used as the donor (Inoue et al., 2010). The results are significant because they help put into perspective some of the challenges that will need to be overcome in order for microbial electrosynthesis to use CO2 as the sole carbon source.

Significance of Findings 1. Significant improvements in extracellular exchange rates are required before carbon dioxide can be used as the sole carbon source to produce desired biochemicals using bioelectrosynthesis

2. Electrochemically reduced conditions not only have the ability to increase product yield, but also increase specificity of desired product.

4.3 Impact of Product and Substrate Degree of Reduction on Bioelectrosynthesis Product yield improvements due to electrical enhancement are dependent on the biochemical compound produced and the substrate utilized. We computed the maximum theoretical product flux for each biochemical compound of interest from each of the substrates. The maximum product yield occurs under conditions of no biomass synthesis (where biomass formation is not the objective function) and can be computationally determined by solving the stoichiometric model and maximizing the product exchange

25 flux rather than the biomass growth reaction. Maximizing for product synthesis redirects carbon flux and therefore electron transfer to product formation rather than biomass synthesis and establishes the upper limit on the product yield. Changes to the theoretical maximum product flux under electrical enhancement provides insight into the relationship between the substrate and the product, and helps to identify the substrate product electron equivalence quotient (SPEEQ) values for which electrical enhancement is most useful. There are two discernible conclusions from Figure 6. The first is the expected trend that shows the achievable product yield improvement is generally dependent on the SPEEQ leading to the conclusion that when the reducing power of the substrate is high relative to the product‘s low degree of reduction, electrical enhancement is of little significance. The converse of this also appears true; electrical enhancement is much more significant in improving yields when the reducing power of the substrate is relatively small compared to the product‘s degree of reduction. The trend between the SPEEQ value and the increase in the predicted yield was expected to be linear – and therefore a linear trendline is shown in Figure 6. And while to other types of trend fit the data (ex. polynomial, exponential, logarithimic, etc), and the data does seem to follow a somewhat linear patter, the strength of this correlation is weak as the data points are scattered. Since the maximum product formation is directly dependent on the number of reducing equivalents and the number of carbons available, the trend is expected to be much tighter. The departure from these expected results can largely be explained by three factors that genome scale models consider. The first is the constraints on the underlying network such as ATP maintenance requirements that require certain portion of carbon to be directed towards ATP production. The second is the pathway constraints. The production of desired products from substrates is not direction, and can require the co-factors, additional metabolites as part of the reaction stoichiometry or can produce other metabolite. Additionally, where the substrate enters the metabolic can be of significance since it can lead to the presence of different cofactors. As an example, the production of NADH from glycolysis under glucose utilization or the production of NADPH under xylose utilization. The third, carbon fixation pathways, is

26 explained below. All these factors can cause non-linearity in the relationship between SPEEQ and the predicted increase in product yield.

Figure 6 Theoretical Increases in Product Yield Resulting from Electrical Enhancement Theoretical improvement in product yield resulting from electrical enhancement using the wild type metabolic network. Substrate Product Electron Equivalence Quotient Represents (SPEEQ) the various product-substrate combinations. SPEEQ has units of moles of NADH produced/moles of NADH consumed. Six substrates are modeled: gluconate, glucose, xylose, sorbitol, glycerol and maltose. SPEEQ is explained in greater detail in Section 3.5

Table 5 SPEEQ Values for Substrate Product Couplings Succinate 1,3-Propanediol 1,4-Butanediol n-Butanol Ethanol Gluconate 0.17 0.33 0.17 0.11 0.17 Glucose 0.33 0.67 0.33 0.22 0.33 Xylose 0.33 0.67 0.33 0.22 0.33 Sorbitol 0.50 1.00 0.50 0.33 0.50 Glycerol 0.67 1.33 0.67 0.44 0.67 Maltose 0.33 0.67 0.33 0.22 0.33

While these results are along expected lines, there are a few data points that cluster above the trending line (circled), which suggest large improvements in yield are possible even when the substrate product electron equivalence quotient (SPEEQ) is relatively high. Analysis shows that these points are associated with the production of succinate and 1,4- butanediol, and can be attributed to the CO2 reducing pathways present in the E. coli metabolism. Phosphoenolpyruvate carboxylase (ppc) is the enzyme responsible for this function. In the model, ppc, fixes carbon using gaseous CO2 to produce OAA from PEP. Consequently, the additional NADH generated by reactions at the electrode induces the cell to shift its metabolism towards pathways that balance the redox allowing the incorporation of carbon dioxide. This results in a substantial improvement in yield. Hence, this provides an explanation for the counter intuitive observation that large improvements are possible even when SPEEQ is high. It should be noted however, recently, yield improvements may be possible for high SPEEQ conditions when an electrode serves as an electrode sink rather than an electron donor. Flynn et al. were able to show that Shewanella oneidensis could be used to ferment glycerol to ethanol in the presence of an electrode based electron acceptor (Flynn, Ross, Hunt, D. R. Bond, & Gralnick, 2010). A closer look at the output from the simulations reveals the importance that ppc activity has on yield improvement. For succinate production, the flux through this reaction can be increased by as much as 14-44% for optimum production under electrically enhanced conditions. The relative increase in 1,4-butanediol production is even greater when the ppc activity is present. The predicted flux of n-butanol is also increased under electrically enhanced conditions, however this is attributable largely to the additional reducing power available to drive the reactions forward. The flux through ppc is relatively low for n-butanol synthesis compared to other products since acetyl-CoA is the precursor metabolite and not oxaloacetate. Additionally, the sole point that shows an increase in yield is associated with 1,3-propanediol production from glycerol. This occurs because glycerol is a substrate that is far more reduced than 1,3-propanediol. Fluxes for a few key reactions with and without electrical enhancement (EE) are shown in Table 4.

Table 6 Predicted Fluxes Through Selected Reactions Reaction Succinate 1,4-Butanediol Ethanol n-Butanol Without With Without With Without With Without With EE EE EE EE EE EE EE EE ATPS4rpp -1.8 5.3 2.7 11.7 -11.6 13.1 -11.6 -1.6 ppc 13.7 18.4 8.6 10 0 13.6 0 2.5 ackr -3.9 -1.4 2.1 6.8 0 0 0 0 Data is shown for ATP Synthase (ATPS4rpp), phosphoenolpyruvate carboxylase (ppc), and acetate kinase (ackr) with and without electrical enhancement (EE) *Units: mmol/gDW-hr

While studies have shown that overexpression of ppc has resulted in higher succinate production under anaerobic conditions (Millard et al., 1996) this data suggests that electrical enhancement could serve as a means to regulate the flux through ppc as a response to NADH generation. Increasing the applied current to the cells would result in increased NADH which would subsequently change the cell‘s NADH/NAD+ ratio. Those pathways that are capable of regenerating NAD+ would in turn see higher fluxes.

When these pathways incorporate CO2, the product yields are relatively higher. This result suggests that as other pathways of carbon fixation that regenerate NAD+ are incorporated into the metabolism, their fluxes may be controlled by directly influencing the NADH generation rate through an electrode. An important factor to consider in the implementation of this strategy will be the in vivo rates at which these enzymes function. FBA places no limit on the flux through the ppc reaction and does not take into account CO2 concentrations required in the media to drive its activity. For example, flux values for the metabolic simulations performed in this work have been as high as 17mmol/gDW-hr for succinate runs. Higher NADH regeneration or greater substrate flux would potentially drive the simulated flux higher. Experimental results conducted on phosphoenolpyruvate carboxylases and carboxykinases in the literature reveal that the pathway that converts phosphoenolpyruvate to succinate via ppc has a flux of about 2mmol/gDW-hr during ppc overexpression in non-optimized cells at a glucose uptake rate of 11.2mmol/gDW-hr (Gokarn, Eiteman, & Altman, 2000). However, as many studies have shown, activity of ppc (as measured in cell extracts) does not guarantee high product flux. The cellular metabolism needs to optimized for product flux. In the wild-type MG1665 strain, it was shown that the flux though the PPC reaction 29 was about 1mmol/gDW-hr (Zhu, Shalel-Levanon, G. Bennett, & K.-Y. San, 2006). Yet, other studies have shown increased succinate production at high concentrations of bicarbonate in the medium (10g/L)(P. Kim, Laivenieks, Vieille, J. Gregory, & J Gregory Zeikus, 2004; Kwon, Lee, & P. Kim, 2008). in silico flux distributions calculated in this study have been much higher. As an example, flux through PPC during growth coupled succinate production was approximately 18 mmol/gDW-hr. The results suggest the role of having not only having strong promoters, but ensuring that the enzyme is active in vivo and the CO2 pressures are high enough to drive product formation. The notion of subjecting microbes to extracellular oxidioreduction potentials (ORP) as a means to modify metabolic fluxes is not an entirely new one. Riondet and coworkers demonstrated the effect of extracellular oxidoreduction potential on metabolic fluxes. One of their findings showed that moving between low and high reducing conditions using H2 resulted in an increase in total CO2 used and higher phosphoenolcarboxylase activity (Riondet, Cachon, Waché, Alcaraz, & Diviès, 2000). Park et al used electrochemical techniques to control the extracellular ORP, observing more ethanol and lactate at high reducing conditions, suggesting the metabolism of W. kimchii can be modulated by redox potentials outside the cell (Park et al. 2005). While the results of the simulations are in agreement with these experiments, they go further in suggesting that extracellular ORP is not the sole consideration for driving product synthesis. The substrate product coupling and as well, if and perhaps how many, carboxylation steps are present in the product pathway, will influence the required electron exchange rate at the electrode. This suggests that electrical energy in these systems needs to be considered from a process optimization perspective that examines product pathway as well as the SPEEQ, and not simply as an on/off switch that alters metabolic fluxes. Theoretical improvements in yield are notably similar for products that have the same degree of reduction but are metabolically many reaction steps apart. Ethanol and n- butanol are examples of these and they exhibit similar enhancement. Additionally substrates with the same degree of reduction (glucose, maltose, xylose) show almost identical levels of enhancement once again suggesting the importance that the SPEEQ value has on product yield improvement.

Significance of Findings 1. Importance of substrate product coupling will become more significant as desired products are more reduced. Therefore, bioelectrosynthesis will become much more significant for biochemicals such as higher order alcohols and -diols that are of particular importance as fuel substitutes or as constituents in biopolymers.

2. Controlling fluxes in response to electrochemically reduced conditions offers new perspectives dynamic control over cellular metabolism.

3. Bioelectrosynthesis naturally boosts the ability for the cellular metabolism to fix carbon dioxide as fluxes through carboxylation reactions increase. The rate of carbon fixation in vivo will be dependent on both carbon dioxide concentration and enzyme kinetic limitations. Flux balance analysis does not take these to parameters into account.

4.4 Growth Coupled Electrical Enhancement Under growth coupled product formation, electrical enhancement can impact the metabolism in a number of ways. We evaluated two scenarios where electrical enhancement could significantly impact the cell, and therefore, the growth coupled product flux. We examined the impact of electrical enhancement on (1) growth coupled strategies for various substrate-product couplings to show that electrical enhancement can be used under growth coupled scenarios (see Appendix A), and (2) strain design strategies and changes to their corresponding substrate specific productivity under electrically enhanced conditions.

4.4.1 Growth Coupled Strategies Figure 7 characterizes the effect that electrical enhancement has on product flux in these strains coupled to biomass production. These results show that in most cases, electrical enhancement is compatible with growth coupled strain designs, and that the improvement in product flux is a function of SPEEQ. As in Section 4.3, there are a few data points that deviate from the general trend. These are associated with carbon fixation. Interestingly,

31 two points for butanol production show a decrease in product yield because bioelectrosynthesis results in increased biomass yield instead as electron availability enables carbon to be rerouted to form biomass. This result suggests that it is important to recalculate the growth coupled strategies to obtain effective coupling during bioelectrosynthesis. The n-butanol case is important because it shows that while the theoretically achievable increase is similar to ethanol (same SPEEQ condition), under growth coupled strategies, the product pathway influences whether those electrons end up in the product or in the biomass. The pathway that allows n-butanol formation has acetyl- CoA as the precursor metabolite, and is preferentially favoured for the formation of biomass precursors rather than n-butanol during bioelectrosynthesis. The yield calculations in Figure 7 are based on substrate uptake rates. We further explored product yield with respect to all incoming carbon (Figure D1 in Appendix D).

These data show that once total improvement in yield is normalized for assimilated CO2 the points that deviate from the general trend show better correlation with the SPEEQ condition. However, this occurs only for the conditions when CO2 was assimilated, a small subset of all data. Additional analysis showed that where CO2 was fixed, the product yield relative to total incoming carbon was slightly lower (<10%) than yield with respect to substrate uptake rate in most cases. An exception was succinate production under electrically enhanced conditions, because the carbon fixing reaction, ppc, is part of the succinate producing pathway. The succinate yield relative to total incoming carbon is 21% and 16% for glucose and xylose, respectively. While this is lower than the yield relative to only glucose or xylose, bioelectrosynthesis still improves the overall efficiency of product synthesis since a greater proportion of total incoming carbon ends up in the product.

Figure 7 Theoretical Increases in Product Yield of Growth Coupled Resulting from Electrical Enhancement Theoretical improvement in product yield resulting from electrical enhancement for strains with reactions knocked out to produce the desired product

During biomass growth, NAD+ must be regenerated by reducing partially oxidized metabolic intermediates such as pyruvate to lactate or ethanol (etc.) that are then excreted from the cell. The amount of each intermediate can be modulated by the cell to balance the reducing equivalents consumed and produced during fermentation so that it can grow on a variety of different substrates. Product formation is often coupled to growth via knock-outs that perturb these NAD+ regenerating pathways, forcing carbon flux to be directed towards desired products. Electrical enhancement introduces another redox disturbance to the cell, to direct additional carbon through these pathways. This perturbation can often come at the expense of growth rate.

To understand this better, we examined the production envelopes for individual biochemicals under electrically and non-electrically enhanced conditions with glucose as the substrate. All four graphs in Figure 8 show that electrical enhancement enlarges the production envelope. A larger production envelope means that the possible solutions to coupling product formation to biomass growth may require different or additional genetic perturbations to achieve the theoretical maximum product flux at any given growth rate. Specifically, the flux distribution of the metabolic network change as fluxes through some reactions are now higher while others are now lower. Reactions that showed to flux may now be active during bioelectrosynthesis. Consequently, these changes in these flux distributions can lead to increases in the original growth rate and/or predicted product flux. Figure 8C clearly shows this for ethanol production. Not only is the new optimum at a higher growth rate but the ethanol flux is much greater. The benefit of improvement in both yield and biomass growth rate is the resulting improvement in overall substrate- specific productivity of these fermentation processes. These results support the conclusions that were obtained in Section 4.2. However, when comparing these different production envelopes. We find that the electrons (or NADH) supplied (generated) by an electrode can, but do not necessarily appear in the desired product. For example, in the case of n-butanol production (Figure 6B), electrical enhancement growth rate increases substantially, as product flux decreases slightly. Therefore, despite the additional NADH that is available for product formation, the additional reducing equivalents do not appear as more product. In this particular case, the result is slightly lower product yield, but greater substrate specific productivity – which is still beneficial to process economics. By comparison, electrical enhancement increases the succinate yield at the expense of growth rate. The maximum growth rate for succinate occurs when no NADH is generated by the electrode. In the scenario for ethanol production, growth rate and product flux increase simultaneously with electron uptake. This scenario results in the greatest possible increase in substrate specific productivity. This result shows the inherent trade-off that can exist between the improved yield and the growth rate of the organism, but that it appears to be conditional on the active metabolic pathways. Consider further that the optimum electron uptake rate predicted

34 computationally for ethanol production is 20.6 mmol/gDW-hr with a predicted growth rate of 0.075 hr-1. By comparison, forcing electrons at 30 mmol/gDW-hr reduces growth rate to 0.058 hr-1 and increases product flux to 21.6 mmol/gDW-hr. This result suggests that the best strategy for electrically enhancing a bioprocess would be a dynamic one. Having no electrical enhancement during the beginning of a batch would maximize growth rate, and once a threshold cell density is reached, growth rate can be sacrificed for improved product yield with electrical enhancement. However, additional optimization will be required to identify the optimal time of switching from a growth phase to a production phase (Anesiadis, Cluett, & Mahadevan, 2008). The implementation of such a dynamic strategy can have a significant impact on substrate specific productivity and the metabolic knockout strategies.

Figure 8 Production Envelopes for Growth Coupled and Electrically Enhanced Metabolite Production on Glucose The production envelopes show changes to growth rate and maximum product flux profiles resulting from electrical enhancement. The trade-off between these suggests that an optima exists both for strain design and for a dynamic strategy. 35

Figure 7 shows a relatively large increase in product yield for many products (dependent on the substrate). The large increase in product yield with a relatively small input of electricity to the system can result in significant improvement process economics. For example, the cost of succinic acid is approximately $0.66/kg. An increase from 13 to 18 mmol/gDW-hr as predicted by the model, can result in an increase of $3.68×10-4 on a gDW-hr basis. By comparison, only $5.9×10-5 on a gDW-hr basis of electricity costs are required to achieve this at an electricity price of $0.06/kWh (Rabaey, Girguis, & Nielsen, 2011). Therefore, with only a small increase in material cost (additional electricity), these bioprocesses can be significantly optimized.

4.4.2 Changes in Flux Distributions for Growth Coupled Products Under growth coupled product synthesis, fewer reactions are available for the cell to respond to electrically induced redox changes. Inspite of the changes in the metabolism, the network has to balance the additional reducing equivalents supplied from the electrode. Therefore, the specific pathways or the magnitude of that response may be different from the wild type. We analysed the changes in the flux distribution for the four scenarios in Section 4.4 and a summary of the results are provided in Table 5. Detailed changes to the flux distributions are shown as metabolic maps in Appendix C. For example, flux through the non-oxidative branch of the TCA cycle is significantly increased at the expense of other pathways such as the pentose phosphate pathway (PPP) and oxidative TCA cycle. Glycolysis is not significantly affected. ATP synthase activity is significantly upregulated to produce ATP as described earlier. While these changes in metabolism are essential to balance the additional reducing equivalent load, the changes do not appear to disrupt growth coupled product formation in most cases except butanol.

Table 7 Predicted Changes in Fluxes Through Selected Pathways WildType 1,4-Butanediol Ethanol n-Butanol Succinate Glycolysis Upper 10% 34% -18% 32% -1% (Reaction) (PFK) (PFK) Lower 0% ~0% 0% -1% -1% Branched TCA Oxidative 10% 4% -18% 32% -91% Reductive 10% 40% -18% 32% 40% Pentose Phosphate Pathway Oxidative 10% 4% -50% 14% -91% Non-Oxidative 0% 4% >300% 15-20% -91% ATP Synthase -216% 246% 146% -357% -638%

4.4.3 Large Knock-out Strategy and Productivity The trade-off between product flux and growth rate is capable of significantly affecting the overall productivity of a system. We examined three specific knock-out strategies for ethanol production to characterize the effect of electrical enhancement on strain design and its productivity. The production envelopes for three different strategies are shown in Figure 8 and are based on results obtained by Feist et al. (Feist et al., 2010). The figure shows that growth coupled product flux is highest for the 10 knockout strategy, followed by the 5 knockout strategy, and finally the 3 knockout strategy. Intuitively, we expect that increasing the number of knockouts results in product fluxes that are closer to the theoretically maximum flux resulting in improved yields. We show, however, that under electrically enhanced conditions, this strategy does not necessarily provide the best product yield. Instead, the 5 knockout strategy and 3 knockout strategy are far superior on a basis of product yield and growth rate. This strongly suggests that while existing strains can be used for bioelectrosynthesis, Opt-Knock or other strain design algorithms such as EMILiO or OptForce that consider electrode reactions will generally lead to different strain designs that are capable of improving product flux. In addition, since electrical enhancement is essentially current supplied at an electrode, by varying the current, it is possible to operate along a particular section of the

37 production envelope. A number of these operating points represented by blue dots are shown for the 3 knockout strategy. Depending on the magnitude of this current, it is possible to operate along the maximum growth rate, or at some other point that improves the product flux while lowering growth rate. This flexibility has significant consequences for implementing a dynamic strategy that maximizes productivity. Such a strategy would use no current (or low current) early during a batch to maximize biomass and then increase the current to maximize product flux. Figure 10 shows how yield, growth rate and substrate specific productivity vary with incremental increases in electron exchange flux.

Figure 9 Production Envelopes for Three Growth Coupled Strategies for Ethanol The production envelopes for various growth coupled strategies for ethanol production show the impact that electrical enhancement has on the strategy`s maximum product flux and growth rate. These parameters greatly influence overall bioprocess productivity.

Figure 10 Changes in Product Yield, Biomass Yield and Substrate Specific Productivity as a Function of Electron Exchange Rate Increases in ethanol yield and decreases in biomass yield suggest the inherent optimum that exists in BESs and the need to operate many of them in a dynamic mode to maximize productivity.

Significance of Findings 1. Growth coupling strategies for product formation are generally compatible with strains undergoing bioelectrosynthesis. However, based on the specific product pathway, these strategies may need to be recalculated when designing for bioelectrosynthesis.

2. While bioelectrosynthesis is generally thought of as a static process, the above analysis provides evidence that bioelectrosynthesis can be thought of as a dynamic process that varies the current at different times during the batch in order to maximize the overall productivity of the process.

4.5 Bioelectrosynthesis on Substrate Mixtures The fermentable substrates that are used by microorganisms are generally produced by the enzymatic hydrolysis of cellulosic biomass. The hydrolyzed biomass results in mixtures of sugars such as glucose and xylose, and the proportion of each sugar in the mixture is dependent on the type of biomass hydrolyzed (ex. softwood, hardwood, corn husks, etc).

Therefore, it is worthwhile examining the affect of bioelectrosynthesis on mixtures of fermentable substrates. To that end, we modelled three different scenarios: 1) 60% glucose, 40% xylose; 2) 40% glucose, 60% xylose; 3) 50% glucose, 50% glycerol, on a carbon basis. Substrate mixtures can have SPEEQ values that are different than pure substrates.

Table 8 Summary of Increases in Theoretical Product Yield Coupled to Biomass Growth Glucose 60 - Glucose 40 - Glucose 50 - Xylose Glucose Glycerol Xylose 40 Xylose 60 Glycerol 50 WT 10% 11% 6% 12% 10% 0% Biomass Ethanol 14% 15% -- 15% 14% -- Succinate 40% 39% 17% 29% 40% -- n-Butanol -4% -4% 7% -4% -4% 18% Butanediol 39% 39% 31% 39% 39% 25%

Results from the modelled simulations show that for many cases, the increase in the theoretical yield on the mixture can be approximated by a weighted average of the yields of the two components that make up the mixture. For example, the increase in the wild type biomass is 11% on the 60% xylose mixture and 12% on pure xylose. The 50%- glycerol mixture for butanediol production had an increase in yield of 31%, which is close to the average of 39% and 25% of each pure substrate. However, discrepancies from this general trend did exist and one notable exception occurred for succinate production on 40% glucose and 60% xylose. The predicted increase in the theoretical product yield for the 40%-glucose-60%xylose mixture was greater than the average of the two component substrates. An explanation for this is that by utilizing both substrates, the model was capable of accepting a higher current from the electrode which resulted in a larger yield. The significance of this is that the dual substrate utilization lessens the burden of the redox stress on the metabolism. Put simply, under sole glucose utilization, there exists a limit on the electron exchange rate, after which the metabolism is incapable of oxidizing NADH to NAD+. When glucose and xylose are co-utilized, the predicted limit on the maximum electron exchange rate increases. This occurs because the presence of another substrate makes it possible to synthesize metabolites by way of alternate pathways. For example, biomass pre-cursors in the pentose phosphate pathway (PPP) can be synthesized from 40 xylose without using the oxidative branch of the PPP, whereas this branch is required when glucose is the sole carbon source. This inherent flexibility and recognizing that NADH is a global co-factor means that the cell can manage greater redox stresses and use it to drive product synthesis. As explained, co-utilization appears to be helpful when the maximum electron exchange rate has been reached (as it was for sole xylose utilization). Physiologically, this can occur if the strain design strategy has a large number of genetic perturbations that make bioelectrosynthesis unfavourable because the metabolic network is too rigid to accommodate an electrochemical redox stress. Conclusions from the above example show that co-utilization may increase flexibility in the metabolism, and improve the favourability of bioelectrosynthesis. It is worthwhile noting that the co-utilization of substrates can be difficult to accomplish in vivo. Suppression of the cell‘s natural regulatory network that prevents co- utilization is a field of study that is still ongoing. Manipulation of the cell‘s regulatory and/or signalling network, or its metabolic network may be a difficult task in the presence of redox stress induced by an electrode.

4.6 Limitations of Modelling Results This study used an FBA framework to evaluate the implications of electrochemical generation of reducing power to drive the synthesis of biochemicals. Several important results were elucidated on the role of bioelectrosynthesis and its impact on biomass growth, cellular ATP yields and biochemical productions. However, it is important to understand these results in light of the limitations of using an FBA model. There are a number of drawbacks to using an FBA model. Firstly, an FBA approach to understanding cellular metabolism involves prediction of cellular growth rates based solely on stoichiometric constraints, without any kinetic information. The solution predicts cell growth and internal cellular fluxes at an optimal level assuming pseudo steady- state cell growth. This optimum flux distribution is best represented by log phase growth and FBA is unable to accurately predict the dynamics of growth and product secretion. This would have implications on the conclusions made in Section 4.4.3 that address the

41 use of bioelectrosynthesis in a dynamic fashion. Specifically, FBA would need additional kinetic parameters for optimizing dynamic strategies that vary the electrode exchange rates to maximize product secretion. Another limitation of the modelling results is that it does not take into account enzymatic constraints and allosteric regulations. Therefore, while the model predicts high flux for many reactions, such as those that oxidize NADH or carboxylases like ppc, there may in fact exist a constraint on the enzymatic capacity to allow for the predicted high flux. At the enzymatic level, allosteric and feedback regulation that come into play during the accumulation of metabolite pools may lead lower than expected yields Finally, the model is based on altering the NADH/NAD+ ratio of the cellular metabolism to drive product synthesis. One of the limitations of using an FBA model is that it does not take regulatory effects into account. The model predicts that, by placing a redox burden on the cell, there are a large number of reactions available to the cell for it to mitigate that disturbance. Cellular regulation at various levels such as gene expression regulation of enzymes may in fact restrict the number of reactions available to the cell. Consequently, the cell may not be able to cope with the burden of a high redox stress or may exhibit poor growth rate even when the computational model predicts growth. . However, despite many of these limitations to using FBA, it is in fact a valuable tool in metabolic engineering. Computational algorithms have been applied to engineer strains for the over-production of metabolites and provided insight into engineering strategies. As an example, Fong an co-workers validated lactate formation coupled to growth rate (Fong et al., 2005). Therefore, while computational analysis have their limitations, they are nonetheless very valuable in understanding how changes in the underlying metabolic network of an organism can be harnessed to over-produce desired biochemicals. In the context to microbial electrosynthesis, they provide a number of valuable insights into the potential for using electricity to drive product synthesis. Sometimes these computational strategies will require additional experimental validation as was the case for Yim and co-workers who used computational techniques to engineer E. coli to produce 1,4-butanediol, and used 13C metabolic flux analysis to fine tune the strains to maximize 1,4-butanediol production (Yim et al., 2011). Nonetheless, this framework

42 provides a systematic understanding of the benefits and limitations of these bioelectrochemical techniques and suggests avenues for further research.

CHAPTER 5 CONCLUSIONS AND RECOMMENDATIONS

Traditional strategies that attempt to manipulate NADH availability, namely, the utilization of carbon source with different redox states and the genetic manipulation of the host cell are generally well understood. Bioelectrosynthesis is a third method that has been used; however a systematic understanding of the potential of this technique hasn‘t been fully evaluated. We have provided a systematic understanding of how electrical enhancement can impact cellular metabolism. Our evaluation of electrical enhancement on a complete genome scale network of the model organism, E. coli, shows that electrical generation of reducing equivalents is capable of influencing the three most important aspects of bioprocesses: biochemical yield (g product/g substrate), cellular ATP yield (mmol ATP/mmol substrate) and cellular growth rate (1/hr). Understanding these processes has provided insight into the long term strategies for electrical enhancement techniques and we have highlighted the significance of these findings both for the application of electrical enhancement at an industrial scale as well as for rational strain design. Our results showed that manipulating cellular redox conditions can force the metabolism to increase the utilization of pathways that fix carbon dioxide. For E. coli we were able to increase the activity of the phosphoenolpyruvate carboxylase pathway because the increased demand to regenerate NAD+ required the use of this pathway. Since there is significant interest to use bioelectrosynthesis on carbon dioxide only, we extended this concept to compare the difference between sole CO2 utilization and co-utilization of CO2 and hexose. Our results suggest that while bioelectrosynthesis is possible on CO2 alone, its very low yield probably make it unsuitable at an industrial scale unless significant advances are made in understanding and optimizing electron transfer rates from an electrode. A limit on the current uptake rate feasible by microbes means that in the short term at least, electrical enhancement strategies that co-utilize CO2 and sugar substrates are more feasible. The extent to which electrical enhancement is capable of influencing ATP and biomass yields provided a basis for understanding the implications that bioelectrosynthesis

44 has on strain design. Our results suggest that electrical enhancement is generally compatible with existing metabolic engineering strategies, although sometimes it may be necessary to recalculate specific strategies under electrically enhanced conditions. The potential trade-off between the biomass growth rate and product yields suggests that there is room for process optimization during bioelectrosynthesis where the organism may grow under non-electrically conditions initially and then electrically enhanced conditions once certain growth conditions have been met.

Recommendations Finally additional work needs to be done to understand the microbiology and physiology of electricigens and their ability to accept electrons. There are a number of possible limitations that can arise, and in particular incorporation of components of the electron transport chain from organisms such as Shewanella, Geobacter or Acidithiobacillus into E. coli to improve extracellular electron exchange rates would be the next logical step in addressing the limitations of extracellular electrons transfer in E. coli. Our results suggest that influencing the NADH/NAD+ ratio is possible with an electrode – and in particular that this may form the basis for controlling fluxes through some reactions, notably those that oxidize NADH and alleviate redox constraints. Further work needs to examine the network level adaptations of an organism in response to these redox perturbations using metabolic flux analysis with 13C isotope labelled substrates and the scale-up of these systems, in much the same way that fuel cells are being considered for scale up. While initial work suggests that increasing current exchange fluxes offer a means to manipulate metabolism, enzymatic limitations may apply, constraining the extent to which the fermentation products can be practically modulated. It is worthwhile examining electrical enhancement in the context of enzymatic capacity to understand the limitations of bioelectrosynthesis. Additional suggested work includes the concept of evolving strains on a bioelectrochemical system to induce higher expression levels of desired pathways. Consider that bioelectrosynthesis places a redox stress on the cellular metabolism, requiring the cell to respond by upregulating pathways that could compensate for stress.

Once the stress has been ―switched off‖, the cells should still exhibit higher activity in the desired pathways, although after several generations, the upregulation should decrease. Batch fermentation requires cells to be stable for relatively short periods of times during which the desired metabolite can be produced. Therefore, it may be possible therefore to evolve strains to a desired level using an electrode and then proceed using normal fermentation using these strains with upregulated product pathways. This would have the benefit of improving product yields without the need for large surface areas that bioelectrochemical systems usually require. Furthermore, analyzing changes to DNA and RNA after strain evolution on electrodes would also provide a lot of insight in understanding necessary mutations to upregulate certain pathways. Transcriptome and proteome analysis of these strains would be valuable to the larger practise of metabolic engineering, in particular, designing conventional strains that have upregulated product pathways. These may be of particular importance to carboxylases and dehydrogenases. This computational study lays the framework for understanding where electrical enhancement is most useful and evaluating potential benefits as well as limitations. The concept of electrical enhancement is promising for products that are highly reduced. These concepts could be beneficial in developing strategies for chemicals or fuels (e.g., jet fuels, biodiesel) that require large amounts of reducing power via NADH. Yields of large molecules such as those that would be required to make fuels could be improved by electrical enhancement. Broadly, the experimental validation of the principles described herein is critical to further understand and optimize microbial bioelectrosynthesis of key biochemical products and develop economically viable and environmentally friendly commercial bioprocesses.

BIBLIOGRAPHY

Anesiadis, N., Cluett, W. R., & Mahadevan, R. (2008). Dynamic metabolic engineering for increasing bioprocess productivity. Metabolic engineering, 10(5), 255-66. doi:10.1016/j.ymben.2008.06.004

Atsumi, S., Cann, A. F., Connor, M. R., Shen, C. R., Smith, K. M., Brynildsen, M. P., Chou, K. J. Y., et al. (2008). Metabolic engineering of Escherichia coli for 1-butanol production. Metabolic engineering, 10(6), 305-11. doi:10.1016/j.ymben.2007.08.003

Becker, S. a, Feist, A. M., Mo, M. L., Hannum, G., Palsson, B. Ø., & Herrgard, M. J. (2007). Quantitative prediction of cellular metabolism with constraint-based models: the COBRA Toolbox. Nature protocols, 2(3), 727-38. doi:10.1038/nprot.2007.99

Berríos-Rivera, S. J., Bennett, George N, & San, K.-Y. (2002). Metabolic Engineering of Escherichia coli: Increase of NADH Availability by Overexpressing an NAD+- Dependent Formate Dehydrogenase. Metabolic Engineering, 4(3), 217-229. doi:10.1006/mben.2002.0227

Bond, D. (2003). Electricity production by Geobacter sulfurreducens attached to electrodes. Applied and environmental microbiology, 69(3), 1548-1555. doi:10.1128/AEM.69.3.1548

Brasseur, Gaël, Bruscella, P., Bonnefoy, V., & Lemesle-Meunier, Danielle. (2002). The bc(1) complex of the iron-grown acidophilic chemolithotrophic bacterium Acidithiobacillus ferrooxidans functions in the reverse but not in the forward direction. Is there a second bc(1) complex? Biochimica et biophysica acta, 1555(1-3), 37-43.

Burgard, A. P., Pharkya, P., & Maranas, C. D. (2003). Optknock: a bilevel programming framework for identifying gene knockout strategies for microbial strain optimization. Biotechnology and bioengineering, 84(6), 647-57. doi:10.1002/bit.10803

Burgard, A, Dien, S. V., & Burk, M. (2009). Methods and organisms for the growth- coupled production of 1, 4-butanediol. WO Patent WO/2009/023,493.

Chemler, J. a, Fowler, Z. L., McHugh, K. P., & Koffas, M. a G. (2010). Improving NADPH availability for natural product biosynthesis in Escherichia coli by metabolic engineering. Metabolic engineering, 12(2), 96-104. Elsevier. doi:10.1016/j.ymben.2009.07.003

Chin, J. W., Khankal, R., Monroe, C. a, Maranas, C. D., & Cirino, P. C. (2009). Analysis of NADPH supply during xylitol production by engineered Escherichia coli. Biotechnology and bioengineering, 102(1), 209-20. doi:10.1002/bit.22060

Durot, M., Bourguignon, P.-Y., & Schachter, V. (2009). Genome-scale models of bacterial metabolism: reconstruction and applications. FEMS microbiology reviews, 33(1), 164-90. doi:10.1111/j.1574-6976.2008.00146.x

Elbehti, A., Brasseur, G, & Lemesle-Meunier, D. (2000). First evidence for existence of an uphill electron transfer through the bc(1) and NADH-Q oxidoreductase complexes of the acidophilic obligate chemolithotrophic ferrous ion-oxidizing bacterium Thiobacillus ferrooxidans. Journal of bacteriology, 182(12), 3602-6.

Energy, U. D. O. (n.d.). Top Value Added Chemicals from Biomass Volume I — Results of Screening for Potential Candidates from Sugars and Synthesis Gas Top Value Added Chemicals From Biomass Volume I : Results of Screening for Potential Candidates. Program.

Fasan, R., Crook, N. C., Peters, M. W., Meinhold, P., Buelter, T., Landwehr, M., Cirino, P. C., et al. (2010). Improved product-per-glucose yields in P450-dependent propane biotransformations using engineered Escherichia coli. Biotechnology and bioengineering, 108(3), 500-510. doi:10.1002/bit.22984

Feist, A. M., Henry, C. S., Reed, J. L., Krummenacker, M., Joyce, A. R., Karp, P. D., Broadbelt, L. J., et al. (2007). A genome-scale metabolic reconstruction for Escherichia coli K-12 MG1655 that accounts for 1260 ORFs and thermodynamic information. Molecular systems biology, 3(121), 121. doi:10.1038/msb4100155

Feist, A. M., Zielinski, D. C., Orth, J. D., Schellenberger, J., Herrgard, M. J., & Palsson, B. Ø. (2010). Model-driven evaluation of the production potential for growth- coupled products of Escherichia coli. Metabolic engineering, 12(3), 173-86. doi:10.1016/j.ymben.2009.10.003

Flynn, J. M., Ross, D. E., Hunt, K. A., Bond, D. R., & Gralnick, J. A. (2010). Enabling unbalanced fermentations by using engineered electrode-interfaced bacteria. mBio, 1(5), e00190–10. Am Soc Microbiol. doi:10.1128/mBio.00190-10

Fong, S. S., Burgard, A. P., Herring, C. D., Knight, E. M., Blattner, F. R., Maranas, C. D., & Palsson, Bernhard O. (2005). In silico design and adaptive evolution of Escherichia coli for production of lactic acid. Biotechnology and bioengineering, 91(5), 643-8. doi:10.1002/bit.20542

Gescher, J. S., Cordova, C. D., & Spormann, A. M. (2008). Dissimilatory iron reduction in Escherichia coli: identification of CymA of Shewanella oneidensis and NapC of E. coli

as ferric reductases. Molecular microbiology, 68(3), 706-19. doi:10.1111/j.1365- 2958.2008.06183.x

Gokarn, R., Eiteman, M., & Altman, E. (2000). Metabolic analysis of Escherichia coli in the presence and absence of the carboxylating enzymes phosphoenolpyruvate carboxylase and pyruvate carboxylase. Applied and environmental microbiology, 66(5), 1844. Am Soc Microbiol.

Gregory, K. B., Bond, D. R., & Lovley, Derek R. (2004). Graphite electrodes as electron donors for anaerobic respiration. Environmental microbiology, 6(6), 596-604. doi:10.1111/j.1462-2920.2004.00593.x

Hongo M, I. (1979). Application of electro-energizing method to L-glutamic acid fermentation. Agric Biol Chem, 43, 2075–2081.

Inoue, K., Leang, C., Franks, A. E., Woodard, T. L., Nevin, K. P., & Lovley, Derek R. (2010). Specific localization of the c-type cytochrome OmcZ at the anode surface in current-producing biofilms of Geobacter sulfurreducens. Environmental Microbiology Reports, no-no. doi:10.1111/j.1758-2229.2010.00210.x

Jarboe, L. R., Zhang, X., Wang, X., Moore, J. C., Shanmugam, K. T., & Ingram, L. O. (2010). Metabolic engineering for production of biorenewable fuels and chemicals: contributions of synthetic biology. Journal of biomedicine & biotechnology, 2010, 761042. doi:10.1155/2010/761042

Jensen, H. M., Albers, a E., Malley, K. R., Londer, Y. Y., Cohen, B. E., Helms, B. a, Weigele, P., et al. (2010). Engineering of a synthetic electron conduit in living cells. Proceedings of the National Academy of Sciences. doi:10.1073/pnas.1009645107

Kasimoglu, E., Park, S. J., Malek, J., Tseng, C. P., & Gunsalus, R. P. (1996). Transcriptional regulation of the proton-translocating ATPase (atpIBEFHAGDC) operon of Escherichia coli: control by cell growth rate. Journal of bacteriology, 178(19), 5563-7.

Keasling, J. D. (2010). Manufacturing Molecules Through Metabolic Engineering. Science, 330(6009), 1355-1358. doi:10.1126/science.1193990

Kim, P., Laivenieks, M., Vieille, C., Gregory, J., & Zeikus, J Gregory. (2004). Effect of Overexpression of Actinobacillus succinogenes Phosphoenolpyruvate Carboxykinase on Succinate Production in Escherichia coli Society. doi:10.1128/AEM.70.2.1238

Kim, T., & Kim, B. (1988). Electron flow shift in Clostridium acetobutylicum fermentation by electrochemically introduced reducing equivalent. Biotechnology Letters, i(2), 123- 128.

Kwon, Y.-D., Lee, S. Y., & Kim, P. (2008). A Physiology Study of Escherichia coli Overexpressing Phosphoenolpyruvate Carboxykinase. Bioscience, Biotechnology, and Biochemistry, 72(4), 1138-1141. doi:10.1271/bbb.70831

Li, H., Cann, A. F., & Liao, James C. (2010). Biofuels: Biomolecular Engineering Fundamentals and Advances. Annual Review of Chemical and Biomolecular Engineering, 1(1), 19-36. doi:10.1146/annurev-chembioeng-073009-100938

Logan, B. E. (2009). Exoelectrogenic bacteria that power microbial fuel cells. Nature reviews. Microbiology, 7(5), 375-81. doi:10.1038/nrmicro2113

Mahadevan, R. (2009). Constraint-Based Genome Scale Models of Cellular Metabolism. The Metabolic Pathway Engineering Handbook: Fundamentals (pp. 18-1).

Millard, C. S., Chao, Y. P., Liao, J C, & Donnelly, M. I. (1996). Enhanced production of succinic acid by overexpression of phosphoenolpyruvate carboxylase in Escherichia coli. Applied and environmental microbiology, 62(5), 1808-10.

Nakamura, C. E., & Whited, G. M. (2003). Metabolic engineering for the microbial production of 1, 3-propanediol. Current opinion in biotechnology. Elsevier. doi:10.1016/j.copbio.2003.08.005

Nevin, K. P., Woodard, T. L., Franks, A. E., Summers, Z. M., & Lovley, D. R. (2010). Microbial Electrosynthesis: Feeding Microbes Electricity To Convert Carbon Dioxide and Water to Multicarbon Extracellular Organic Compounds. mBio, 1(2), e00103-10-e00103-10. doi:10.1128/mBio.00103-10

Park, D H, & Zeikus, J G. (2000). Electricity generation in microbial fuel cells using neutral red as an electronophore. Applied and environmental microbiology, 66(4), 1292-7.

Park, D H, & Zeikus, J G. (1999). Utilization of electrically reduced neutral red by Actinobacillus succinogenes: physiological function of neutral red in membrane-driven fumarate reduction and energy conservation. Journal of bacteriology, 181(8), 2403-10.

Park, S. M., Sang, B. I., Park, D. W., & Park, Doo Hyun. (2005). Electrochemical reduction of xylose to xylitol by whole cells or crude enzyme of Candida peltata. Journal of microbiology (Seoul, Korea), 43(5), 451-5.

Peguin, S., & Soucaille, P. (1996). Modulation of Metabolism of Clostridium acetobutylicum Grown in Chemostat Culture in a Three-Electrode Potentiostatic System with Methyl Viologen as Electron Carrier. Biotechnology and Bioengineering,, 51, 342-348.

Rabaey, K., & Rozendal, R. a. (2010). Microbial electrosynthesis - revisiting the electrical route for microbial production. Nature reviews. Microbiology, 8(10), 706-16. Nature Publishing Group. doi:10.1038/nrmicro2422

Rabaey, K., Girguis, P., & Nielsen, L. K. (2011). Metabolic and practical considerations on microbial electrosynthesis. Current opinion in biotechnology, 22(3), 371-7. Elsevier Ltd. doi:10.1016/j.copbio.2011.01.010

Rabaey, K., Rodríguez, J., Blackall, L. L., Keller, J., Gross, P., Batstone, D., Verstraete, W., et al. (2007). Microbial ecology meets electrochemistry: electricity-driven and driving communities. The ISME journal, 1(1), 9-18. doi:10.1038/ismej.2007.4

Ragsdale, S. W., & Pierce, E. (2008). Acetogenesis and the Wood-Ljungdahl pathway of

CO2 fixation. Biochimica et biophysica acta, 1784(12), 1873-98. Elsevier. doi:10.1016/j.bbapap.2008.08.012

Ranganathan, S., Suthers, P. F., & Maranas, C. D. (2010). OptForce: An Optimization Procedure for Identifying All Genetic Manipulations Leading to Targeted Overproductions. (N. D. Price, Ed.)PLoS Computational Biology, 6(4), e1000744. doi:10.1371/journal.pcbi.1000744

Riondet, C., Cachon, R., Waché, Y., Alcaraz, G., & Diviès, C. (2000). Extracellular oxidoreduction potential modifies carbon and electron flow in Escherichia coli. Journal of bacteriology, 182(3), 620-6.

San, K.-Y., Bennett, George N, Berríos-Rivera, S. J., Vadali, R. V., Yang, Y.-T., Horton, E., Rudolph, F. B., et al. (2002). Metabolic engineering through cofactor manipulation and its effects on metabolic flux redistribution in Escherichia coli. Metabolic engineering, 4(2), 182-92. doi:10.1006/mben.2001.0220

Shin, H. S., Zeikus, J G, & Jain, M. K. (2002). Electrically enhanced ethanol fermentation by Clostridium thermocellum and Saccharomyces cerevisiae. Applied microbiology and biotechnology, 58(4), 476-81. doi:10.1007/s00253-001-0923-2 da Silva, G. P., Mack, M., & Contiero, J. (2009). Glycerol: a promising and abundant carbon source for industrial microbiology. Biotechnology Advances, 27(1), 30–39. Elsevier. doi:10.1016/j.biotechadv.2008.07.006

Singh, A., Cher Soh, K., Hatzimanikatis, V., & Gill, R. T. (2011). Manipulating redox and ATP balancing for improved production of succinate in E. coli. Metabolic engineering, 13(1), 76-81. Elsevier. doi:10.1016/j.ymben.2010.10.006

Stephanopoulos, G. (2007). Challenges in engineering microbes for biofuels production. Science (New York, N.Y.), 315(5813), 801-4. doi:10.1126/science.1139612

Sun Mi, P., Hye Sun, K., Dae Won, P., & Doo Hyun, P. (2005). Electrochemical control of metabolic flux of Weissella kimchii sk10: Neutral red immobilized in cytoplasmic membrane as electron channel. Journal of microbiology and biotechnology, 15(1), 80-85. Korean Society for Applied Microbiology.

Sánchez, A. M., Bennett, G.N., & San, K. Y. (2005). Efficient succinic acid production from glucose through overexpression of pyruvate carboxylase in an Escherichia coli alcohol dehydrogenase and lactate dehydrogenase mutant. Biotechnology progress, 21(2), 358–365. [New York, NY: American Institute of Chemical Engineers, c1985- .

Travick, J. D., Burk, M., & Burgard, A. P. (2010). Microogranisms and Methods for Conversion of Syngas and Other Carbon Sources to Useful Products. doi:20100304453A1

Valdés, J., Pedroso, I., Quatrini, R., Dodson, R. J., Tettelin, H., Blake, R., Eisen, J. A., et al. (2008). Acidithiobacillus ferrooxidans metabolism: from genome sequence to industrial applications. BMC genomics, 9, 597. doi:10.1186/1471-2164-9-597

Varma, A., & Palsson, B. (1994). Metabolic flux balancing: basic concepts, scientific and practical use. Nature Biotechnology.

Varma, A., & Palsson, B O. (1994). Stoichiometric flux balance models quantitatively predict growth and metabolic by-product secretion in wild-type Escherichia coli W3110. Applied and environmental microbiology, 60(10), 3724-31.

Xie, X. H., Li, E. L., & Tang, Z. K. (2010a). Sudden Emergence of Redox Active Escherichia coli Phenotype: Cyclic Voltammetric Evidence of the Overlapping Pathways. Int. J. Electrochem. Sci, 5, 1070–1081.

Xie, X. H., Li, E. L., & Tang, Z. K. (2010b). Redox modulation and non-invasive evaluation of phenotypic adaptation of Escherichia coli Biofilm. Int. J. Electrochem. Sci, 5, 1379–1389.

Yang, L., & Cluett, W. R. (2010). Rapid design of system-wide metabolic network modifications using iterative linear programming. Dynamics and Control, (Dycops), 377-382.

Yim, H., Haselbeck, R., Niu, W., Pujol-Baxley, C., Burgard, Anthony, Boldt, J., Khandurina, J., et al. (2011). Metabolic engineering of Escherichia coli for direct production of 1,4-butanediol. Nature chemical biology, 7(7), 445-444. Nature Publishing Group. doi:10.1038/nchembio.580

Zeikus, J G, & Park, D H. (2001). Electrochemical methods for generation of a biological proton motive force and pyridine nucleotide cofactor regeneration.

Zhu, J., Shalel-Levanon, S., Bennett, G., & San, K.-Y. (2006). Effect of the global redox sensing/regulation networks on Escherichia coli and metabolic flux distribution based on C-13 labeling experiments. Metabolic engineering, 8(6), 619-27. doi:10.1016/j.ymben.2006.07.002

APPENDIX A - MODEL REACTIONS/DELETION STRATEGIES

Table A1 - Model Reactions/Deletion Strategies

Reaction Added Lower Upper E.C. # Model Rxn Description/Notes Bounds Bounds Name Interface with electrode EE -> -30 0 EX_EE Exchange reaction for electrons between electrode and cell NAD[c] + 2 EE + h[c] -> NADH 0 1000 QE_rxn Simplified reaction represents production of NADH through menaquinone pool with two electrode electrons

n-Butanol Pathway 2 accoa[c] -> aacoa[c] + coa[c] -1000 1000 2.3.1.9 ACACT1r' Acetoacetyl Coenzyme A Thiolase aacoa[c] + nadh[c] + h[c] -> 3hbcoa[c] + nad[c] -1000 1000 .1.1.35 HACD1i Hydroxybutryl-CoA Dehydrogenase 3hbcoa[c] -> b2coa[c] + h2o[c] -1000 1000 4.2.1.17 ECOAH1' Enoyl-CoA hydratase b2coa[c] + nadh[c] + h[c] -> btcoa[c] + nad[c] -1000 1000 1.3.99.2 BDH' Butyryl-CoA dehydrogenase; Crotonoyl-CoA + NADH + H+ <==> Butanoyl-CoA + NAD; reaction in model exists as 'ACOAD1f' with FAD. This reaction, was deleted btcoa[c] + nadh[c] + h[c] -> btal + coa[c] + nad[c] 0 1000 1.2.1.10 ADH Acetaldehyde dehydrogenase (acetylating) btal + nadh[c] + h[c] -> btol[c] + nad[c] 0 1000 1.1.1.- ADH1' Butanoate metabolism

btol[c] -> btol[e] -1000 1000 BTOLtex' btol[e] -> 0 1000 EX_btol' Butanol exchange reaction

Wood-Ljungdahl Pathway nadh[c] + co2[c] + h[c] -> nad[c] + for[c] 0 1000 1.2.1.43 'WL1' Formate dehydrogenase. This reaction normally uses NADPH as the cofactor Knock out FDH4pp and FDH5pp which do rev. rxn 'for[c] + atp[c] + thf[c] -> 10fthf[c] + adp[c] + pi[c] 0 1000 6.3.4.3 WL2' 10-Formyl-H4 folate synthetase; Knock out FTHFD with does rev. rxn '10fthf[c] + h[c] -> methf[c] + h2o[c] 3.5.4.9 WL3 Exists already as MTHFC; 5,10- methenyl-H4 folate cyclohydrolase nadph[c] + methf[c] -> mlthf[c] + nadp[c] 1.5.1.15 WL4 Exists already as MTHFD using NAPDH instead; 5,10-methylene-H4 folate dehydrogenase nadh[c] + mlthf[c] + 2 h[c] -> 5mthf[c] + nad[c] 1.1.99.15 WL5 5,10-methylene-H4folate reductase; Exists as MTHFR2 but with NADH cofactor nadh[c] + co2[c] + h[c] -> nad[c] + co[c] + h2o[c] 0 1000 WL6' CO dehydrogenase/acetyl-CoA synthase 'coa[c] + 5mthf[c] + co[c] -> thf[c] + accoa[c] 0 1000 WL7'

1,4 Butanediol Production 'succoa[c] + 2 nadh[c] + 2 h[c] -> 4hb[c] + 2 nad[c] 0 1000 1.2.1.16 '4HDHa' 4-hydroxybutyrate dehydrogenase + coa[c] '4hb[c] + accoa[c] -> 4hbcoa[c] + ac[c] 0 1000 '4HDHac' 4-hydroxybutyrate CoA transferase '4hbcoa[c] + nadh[c] + h[c] -> 4hbal[c] + coa[c] + 0 1000 'BDH' Butanal dehydrogenase nad[c] '4hbal[c] + nadh[c] + h[c] -> bdol[c] + nad[c] -1000 1000 1.1.1.202 '4hbal' 1,3-propanediol dehydrogenase - Works with butanediol; Gene AAK09379.1

bdol[c] -> bdol[e] -1000 10000 BDOLtex 'bdol[e] -> 0 1000 'EX_bdol' Butanol exchange reaction

1,3-Propanediol glyc[c] -> 3hp[c] + h2o[c] 0 1000 Gylcerol --> 3-HP + H2O '3hp[c] + nadh[c] + h[c] -> 13pdo[c] + nad[c] 0 1000 3-HPA + NADH + H --> 1,3PDO '13pdo[c] -> 13pdo[e] -1000 1000 13pdo[e] -> 0 1000 Propanediol exchange

Table A2 - Knock-out Strategies for Growth Coupled Product Formation

Succinate LDH_D PFL ALCD2x 'THD2pp' n-Butanol 'LDH_D' 'ACALD' 'ALCD2x' 'PTAr' 'ACKr' Ethanol 'ACKr' 'TPI' 'PFL' 1,4-Butanediol LDH_D PFL ALCD2x 'THD2pp'

Table A3 - Knock -out Strategies for Growth Coupled Ethanol Formation

3KO ACKr TPI PFL 5KO ACALD ATPS4rpp LDH_D PFK PFL 10KO ABTA ACALD ACKr ATPS4rpp F6PA GLUDy LDH_D MGSA PFL TPI

APPENDIX B MODELLING RESULTS - MAXIMIZING BIOMASS

Table B1 - Modelling Results for Maximizing Biomass (No Enhancement)

Glucose Xylose Pryuvate Sorbitol Gluconate

Exchange Flux Exchange Flux Exchange Flux Exchange Flux Exchange Flux DM_4HBA 4.20E-05 DM_4HBA 3.54E-05 DM_4HBA 1.31E-05 DM_4HBA 3.37E-05 DM_4HBA 2.95E-05 DM_5DRIB 8.39E-05 DM_5DRIB 7.07E-05 DM_5DRIB 2.62E-05 DM_5DRIB 6.74E-05 DM_5DRIB 5.89E-05 DM_HMFURN 8.39E-05 DM_HMFURN7.07E-05 DM_HMFURN2.62E-05 DM_HMFURN6.74E-05 DM_HMFURN5.89E-05 EX_ac(e) 8.61685 EX_ac(e) 8.83399 EX_ac(e) 19.0989 EX_ac(e) 3.88827 EX_ac(e) 11.7695 EX_ca2(e) -0.00089 EX_ca2(e) -0.00075 EX_ca2(e) -0.00028 EX_ca2(e) -0.00072 EX_ca2(e) -0.00063 EX_cl(e) -0.00089 EX_cl(e) -0.00075 EX_cl(e) -0.00028 EX_cl(e) -0.00072 EX_cl(e) -0.00063 EX_co2(e) -0.09117 EX_co2(e) -0.07686 EX_co2(e) 0.910533 EX_co2(e) -0.07328 EX_co2(e) 1.87214 EX_cobalt2(e) -0.00059 EX_cobalt2(e)-0.0005 EX_cobalt2(e)-0.00019 EX_cobalt2(e)-0.00048 EX_cobalt2(e)-0.00042 EX_cu2(e) -0.00059 EX_cu2(e) -0.0005 EX_cu2(e) -0.00019 EX_cu2(e) -0.00048 EX_cu2(e) -0.00042 EX_etoh(e) 8.49536 EX_etoh(e) 8.73157 EX_fe2(e) -0.00044 EX_etoh(e) 13.7906 EX_etoh(e) 5.55648 EX_fe2(e) -0.00142 EX_fe2(e) -0.0012 EX_fe3(e) -0.00042 EX_fe2(e) -0.00114 EX_fe2(e) -0.001 EX_fe3(e) -0.00134 EX_fe3(e) -0.00113 EX_for(e) 18.409 EX_fe3(e) -0.00107 EX_fe3(e) -0.00094 EX_for(e) 17.9104 EX_for(e) 18.2384 EX_glyclt(e)3.93E-05 EX_for(e) 18.3204 EX_for(e) 17.8867 EX_glc(e) -10 EX_glyclt(e)0.000106 EX_h2o(e) -17.5235 EX_glyclt(e)0.000101 EX_glcn(e) -10 EX_glyclt(e) 0.000126 EX_h2o(e) -4.57757 EX_h(e) 18.0863 EX_h2o(e) 0.169992 EX_glyclt(e)8.84E-05 EX_h2o(e) -3.5678 EX_h(e) 28.6351 EX_k(e) -0.01043 EX_h(e) 23.6987 EX_h2o(e) -0.1587 EX_h(e) 28.3809 EX_k(e) -0.02817 EX_mg2(e) -0.00046 EX_k(e) -0.02686 EX_h(e) 20.9583 EX_k(e) -0.03341 EX_mg2(e) -0.00125 EX_mn2(e) -0.00019 EX_mg2(e) -0.00119 EX_k(e) -0.02347 EX_mg2(e) -0.00149 EX_mn2(e) -0.0005 EX_mobd(e)-0.00019 EX_mn2(e) -0.00048 EX_mg2(e) -0.00104 EX_mn2(e) -0.00059 EX_mobd(e) -0.0005 EX_nh4(e) -0.63309 EX_mobd(e)-0.00048 EX_mn2(e) -0.00042 EX_mobd(e) -0.00059 EX_nh4(e) -1.71044 EX_pi(e) -0.05644 EX_nh4(e) -1.63082 EX_mobd(e)-0.00042 EX_nh4(e) -2.02896 EX_pi(e) -0.15247 EX_pyr(e) -20 EX_pi(e) -0.14538 EX_nh4(e) -1.42534 EX_pi(e) -0.18087 EX_so4(e) -0.03969 EX_so4(e) -0.01469 EX_sbt_D(e) -10 EX_pi(e) -0.12706 EX_so4(e) -0.04708 EX_succ(e) 0.052934 EX_succ(e) 0.019593 EX_so4(e) -0.03784 EX_so4(e) -0.03308 EX_succ(e) 0.062791 EX_xyl_D(e) -12 EX_zn2(e) -0.00019 EX_succ(e) 0.050469 EX_succ(e) 0.044111 EX_zn2(e) -0.00059 EX_zn2(e) -0.0005 Ec_biomass_iAF1260_core_59p81M0.058707 EX_zn2(e) -0.00048 EX_zn2(e) -0.00042 Ec_biomass_iAF1260_core_59p81M0.188145 Ec_biomass_iAF1260_core_59p81M0.158609 Ec_biomass_iAF1260_core_59p81M0.151225 Ec_biomass_iAF1260_core_59p81M0.132171

Biomass Yield on Substrate0.00314 0.00264 0.00098 0.00252 0.00220 Biomass Yield (total0.00313 carbon) 0.00264 0.00098 0.00252 0.00220

Table B2 - Modelling Results for Maximizing Biomass (Enhancement)

Glucose Xylose Pyruvate Sorbitol Gluconate

Exchange Flux Exchange Flux Exchange Flux Exchange Flux Exchange Flux DM_4HBA 4.61E-05 DM_4HBA 3.95E-05 DM_4HBA 1.71E-05 DM_4HBA 3.58E-05 DM_4HBA 3.38E-05 DM_5DRIB 9.21E-05 DM_5DRIB 7.90E-05 DM_5DRIB 3.42E-05 DM_5DRIB 7.16E-05 DM_5DRIB 6.75E-05 DM_HMFURN 9.21E-05 DM_HMFURN7.90E-05 DM_HMFURN3.42E-05 DM_HMFURN7.16E-05 DM_HMFURN6.75E-05 EX_ac(e) 0.981142 EX_ac(e) 1.19828 EX_ac(e) 11.936 EX_ca2(e) -0.00076 EX_ac(e) 3.79845 EX_ca2(e) -0.00098 EX_ca2(e) -0.00084 EX_ca2(e) -0.00036 EX_cl(e) -0.00076 EX_ca2(e) -0.00072 EX_cl(e) -0.00098 EX_cl(e) -0.00084 EX_cl(e) -0.00036 EX_co2(e) -0.07783 EX_cl(e) -0.00072 EX_co2(e) -0.10011 EX_co2(e) -0.0858 EX_co2(e) -0.03718 EX_cobalt2(e)-0.00051 EX_co2(e) 2.1451 EX_cobalt2(e) -0.00065 EX_cobalt2(e)-0.00056 EX_cobalt2(e)-0.00024 EX_cu2(e) -0.00051 EX_cobalt2(e)-0.00048 EX_cu2(e) -0.00065 EX_cu2(e) -0.00056 EX_cu2(e) -0.00024 EX_etoh(e) 17.5346 EX_cu2(e) -0.00048 EX_etoh(e) 15.8477 EX_etoh(e) 16.0839 EX_etoh(e) 6.88644 EX_fe2(e) -0.00121 EX_etoh(e) 13.1376 EX_fe2(e) -0.00156 EX_fe2(e) -0.00134 EX_fe2(e) -0.00058 EX_fe3(e) -0.00114 EX_fe2(e) -0.00114 EX_fe3(e) -0.00147 EX_fe3(e) -0.00126 EX_fe3(e) -0.00055 EX_for(e) 18.216 EX_fe3(e) -0.00108 EX_for(e) 17.7054 EX_for(e) 18.0334 EX_for(e) 19.1479 EX_glyclt(e)0.000107 EX_for(e) 17.5785 EX_glc(e) -10 EX_glyclt(e)0.000118 EX_glyclt(e)5.13E-05 EX_h2o(e) 4.31053 EX_glcn(e) -10 EX_glyclt(e) 0.000138 EX_h2o(e) 3.55353 EX_h2o(e) -9.87708 EX_h(e) 4.52194 EX_glyclt(e)0.000101 EX_h2o(e) 4.5633 EX_h(e) -9.02375 EX_h(e) -18.1602 EX_k(e) -0.02853 EX_h2o(e) 8.04716 EX_h(e) -9.27792 EX_k(e) -0.03145 EX_k(e) -0.01363 EX_mg2(e) -0.00127 EX_h(e) -17.1309 EX_k(e) -0.03669 EX_mg2(e) -0.0014 EX_mg2(e) -0.00061 EX_mn2(e) -0.00051 EX_k(e) -0.0269 EX_mg2(e) -0.00163 EX_mn2(e) -0.00056 EX_mn2(e) -0.00024 EX_mobd(e)-0.00051 EX_mg2(e) -0.0012 EX_mn2(e) -0.00065 EX_mobd(e)-0.00056 EX_mobd(e)-0.00024 EX_nh4(e) -1.73219 EX_mn2(e) -0.00048 EX_mobd(e) -0.00065 EX_nh4(e) -1.90952 EX_nh4(e) -0.82737 EX_pi(e) -0.15441 EX_mobd(e)-0.00048 EX_nh4(e) -2.22804 EX_pi(e) -0.17022 EX_pi(e) -0.07375 EX_sbt_D(e) -10 EX_nh4(e) -1.63316 EX_pi(e) -0.19861 EX_so4(e) -0.04431 EX_pyr(e) -20 EX_so4(e) -0.0402 EX_pi(e) -0.14558 EX_so4(e) -0.0517 EX_succ(e) 0.059095 EX_so4(e) -0.0192 EX_succ(e) 0.053607 EX_so4(e) -0.0379 EX_succ(e) 0.068952 EX_xyl_D(e) -12 EX_succ(e) 0.025605 EX_zn2(e) -0.00051 EX_succ(e) 0.050542 EX_zn2(e) -0.00065 EX_zn2(e) -0.00056 EX_zn2(e) -0.00024 Ec_biomass_iAF1260_core_59p81M0.160625 EX_zn2(e) -0.00048 Ec_biomass_iAF1260_core_59p81M0.206605 Ec_biomass_iAF1260_core_59p81M0.177069 Ec_biomass_iAF1260_core_59p81M0.076722 EX_EE -15.2767 Ec_biomass_iAF1260_core_59p81M0.151442 EX_EE -30 EX_EE -30 EX_EE -30 EX_EE -30

Biomass Yield on Substrate0.003443 0.002951 0.001279 0.002677 0.002524 Biomass Yield (total carbon)0.003438 0.002947 0.001278 0.002674 0.002524

Table B3 - Modelling Results – Effect of CO2 Fixation on Growth Rate

WT WT + EE WT + WL WT + WL + EE WT + WL + EE - Glc Exchange Flux Exchange Flux Exchange Flux Exchange Flux Exchange Flux

DM_4HBA 4.20E-05 DM_4HBA 4.61E-05 DM_4HBA 6.89E-05 DM_4HBA 8.58E-05 DM_4HBA 1.78E-06 DM_5DRIB 8.39E-05 DM_5DRIB 9.21E-05 DM_5DRIB 0.000138 DM_5DRIB 0.000172 DM_5DRIB 3.57E-06 DM_HMFURN 8.39E-05 DM_HMFURN9.21E-05 DM_HMFURN 0.000138 DM_HMFURN0.000172 DM_HMFURN3.57E-06 EX_ac(e) 8.61685 EX_ac(e) 0.981142 EX_ac(e) 23.162 EX_ac(e) 25.2335 EX_ac(e) 3.57307 EX_ca2(e) -0.00089 EX_ca2(e) -0.00098 EX_ca2(e) -0.00146 EX_ca2(e) -0.001823 EX_ca2(e) -3.79E-05 EX_cl(e) -0.00089 EX_cl(e) -0.00098 EX_cl(e) -0.00146 EX_cl(e) -0.001823 EX_cl(e) -3.79E-05 EX_co2(e) -0.09117 EX_co2(e) -0.10011 EX_co2(e) 0.605629 EX_co2(e) -6.74571 EX_co2(e) -7.48433 EX_cobalt2(e) -0.00059 EX_cobalt2(e)-0.00065 EX_cobalt2(e) -0.00098 EX_cobalt2(e)-0.001216 EX_cobalt2(e)-2.53E-05 EX_cu2(e) -0.00059 EX_cu2(e) -0.00065 EX_cu2(e) -0.00098 EX_cu2(e) -0.001216 EX_cu2(e) -2.53E-05 EX_etoh(e) 8.49536 EX_etoh(e) 15.8477 EX_fe2(e) -0.00233 EX_fe2(e) -0.002907 EX_fe2(e) -6.04E-05 EX_fe2(e) -0.00142 EX_fe2(e) -0.00156 EX_fe3(e) -0.0022 EX_fe3(e) -0.002735 EX_fe3(e) -5.68E-05 EX_fe3(e) -0.00134 EX_fe3(e) -0.00147 EX_glc(e) -10 EX_glc(e) -10 EX_glyclt(e)5.35E-06 EX_for(e) 17.9104 EX_for(e) 17.7054 EX_glyclt(e) 0.000207 EX_glyclt(e)0.000258 EX_h2o(e) 7.72899 EX_glc(e) -10 EX_glc(e) -10 EX_h2o(e) 8.85017 EX_h2o(e) 18.5226 EX_h(e) -33.7406 EX_glyclt(e) 0.000126 EX_glyclt(e)0.000138 EX_h(e) 5.38849 EX_h(e) -30.0116 EX_k(e) -0.00142 EX_h2o(e) -3.5678 EX_h2o(e) 4.5633 EX_k(e) -0.05489 EX_k(e) -0.068366 EX_mg2(e)-6.31E-05 EX_h(e) 28.3809 EX_h(e) -9.27792 EX_mg2(e) -0.00244 EX_mg2(e)-0.003039 EX_mn2(e)-2.53E-05 EX_k(e) -0.03341 EX_k(e) -0.03669 EX_mn2(e) -0.00098 EX_mn2(e)-0.001216 EX_mobd(e)-2.53E-05 EX_mg2(e) -0.00149 EX_mg2(e) -0.00163 EX_mobd(e) -0.00098 EX_mobd(e)-0.001216 EX_nh4(e) -0.08624 EX_mn2(e) -0.00059 EX_mn2(e) -0.00065 EX_nh4(e) -3.33309 EX_nh4(e) -4.15124 EX_pi(e) -0.00769 EX_mobd(e) -0.00059 EX_mobd(e)-0.00065 EX_pi(e) -0.29712 EX_pi(e) -0.370052 EX_so4(e) -0.002 EX_nh4(e) -2.02896 EX_nh4(e) -2.22804 EX_so4(e) -0.07735 EX_so4(e) -0.096332 EX_succ(e) 0.002669 EX_pi(e) -0.18087 EX_pi(e) -0.19861 EX_succ(e) 0.10315 EX_succ(e) 0.12847 EX_zn2(e) 0.000 EX_so4(e) -0.04708 EX_so4(e) -0.0517 EX_zn2(e) -0.00098 EX_zn2(e) -0.001216 Ec_biomass_iAF1260_core_59p81M0.007997 EX_succ(e) 0.062791 EX_succ(e) 0.068952 Ec_biomass_iAF1260_core_59p81M0.309076 Ec_biomass_iAF1260_core_59p81M0.384943 EX_EE -30 EX_zn2(e) -0.00059 EX_zn2(e) -0.00065 EX_EE -30 Ec_biomass_iAF1260_core_59p81M0.188145 Ec_biomass_iAF1260_core_59p81M0.206605 EX_EE -30 Biomass Yield on Substrate0.00314 0.00344 0.00515 0.00642 Biomass Yield (total carbon)0.00313 0.00344 0.00515 0.00577

Table B4 - Modelling Results for Maximizing ATP

Glucose Xylose Exchange Flux Exchange Flux ACALD -10.00 ACALD -10.000 ACKr -10.00 ACKr -10.00 ACt2rpp -10.00 ACt2rpp -10 ACtex -10.00 ACtex -10 - ADK3 1000.00 ADK3 -1000 - ALATA_L 1000.00 ADNt2pp 1000 ALCD2x -10.00 ADNt2rpp -1000 ATPM 27.50 ALATA_L -1000 ATPS4rpp -2.5 ALCD2x -10 CYTDt2pp 1000 ATPM 24.5 CYTDt2rpp -1000 ATPS4rpp -5.5 DHAPT 10 DHAPT 8 ENO 20 ENO 20 ETOHt2rpp -10 ETOHt2rpp -10 ETOHtex -10 ETOHtex -10 EX_ac(e) 10 EX_ac(e) 10 EX_etoh(e) 10 EX_etoh(e) 10 EX_for(e) 20 EX_for(e) 20 EX_glc(e) -10 EX_h2o(e) -10 EX_h2o(e) -10 EX_h(e) 30 EX_h(e) 30 EX_xyl_D(e) -12 F6PA 10 F6PA 8 FORtex -20 FORtex -20 FORtppi 20 FORtppi 20 GAPD 20 GAPD 20 GLCptspp 10 GLCtex -1000 GLCtex -990 GLCtexi 1000 GLCtexi 1000 H2Otex 10 H2Otex 10 H2Otpp 10 H2Otpp 10 Htex -30 Htex -30 ICHORS -1000

ICHORS -1000 ICHORSi 1000 ICHORSi 1000 INSt2pp 1000 NDPK1 -1000 INSt2rpp -1000 PFL 20 NAt3pp 1000 PGI 10 NDPK1 -1000 PGK -20 PFL 20 PGM -20 PGK -20 PPM -1000 PGM -20 PRPPS -1000 PPM -1000 PTAr 10 PRPPS -1000 R15BPK 1000 PTAr 10 R1PK 1000 PYK 12 TPI 10 R15BPK 1000 VALTA 1000 R1PK 1000 VPAMT 1000 RPE -4 RPI -4 SERt2rpp -1000 SERt4pp 1000 TALA 4 THMDt2pp 1000 THMDt2rpp -1000 TKT1 4 TKT2 4 TPI 8 VALTA 1000 VPAMT 1000 XYLI1 12 XYLK 12 XYLt2pp 12 XYLtex 12

APPENDIX C: CHANGES IN METABOLIC FLUX DISTRIBUTION

Changes in Metabolic Flux Distribution (Flux Map)

Fig C1 – Map of Metabolic Flux Distributions for Ethanol

Fig C2 – Map of Metabolic Flux Distributions for Succinate

Fig C3 – Map of Metabolic Flux Distributions for Butanol

Fig C4 – Map of Metabolic Flux Distributions for Butanediol

APPENDIX D: MAXIMIZATION OF PRODUCT

Table D1: Sample of Modelling Results – Exchange Fluxes Maximization of Product on Glucose

Exchange Flux Exchange Flux (Enhanced) Succinate EX_ac(e) 3.92667 EX_ac(e) 1.42667 EX_co2(e) -7.44952 EX_co2(e) -16.7352 EX_glc(e) -10 EX_glc(e) -10 EX_h2o(e) 7.44952 EX_h2o(e) 16.7352 EX_h(e) 33.7248 EX_h(e) 8.36762 EX_succ(e) 14.899 EX_succ(e) 18.4705 EX_EE -30 Butanediol EX_ac(e) 6.42685 EX_ac(e) 3.9611 EX_co2(e) 12.8581 EX_co2(e) 8.74849 EX_glc(e) -10 EX_glc(e) -10 EX_h2o(e) 4.28603 EX_h2o(e) 12.9162 EX_h(e) 6.42685 EX_h(e) -26.0389 EX_bdol 8.57205 EX_EE -30 EX_bdol 10.8323 Butanol EX_co2(e) 20 EX_co2(e) 15 EX_glc(e) -10 EX_glc(e) -10 EX_h2o(e) 10 EX_h2o(e) 18.75 EX_btol 10 EX_h(e) -30 EX_EE -30 EX_btol 11.25 Ethanol EX_co2(e) 20 EX_co2(e) 15 EX_etoh(e) 20 EX_etoh(e) 22.5 EX_glc(e) -10 EX_glc(e) -10 EX_h2o(e) 7.5 EX_h(e) -30 EX_EE -30

Figure D1 - Increase in Product Yield Normalized to Total Incoming

Carbon from CO2

0.7

0.6 0.5 0.4 0.3 0.2 0.1

Increase in Product Yield Product in Increase 0 -0.1 0 0.2 0.4 0.6 0.8 Substrate Product Electron Equivalence Quotient